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/queue.h>
57 #include <sys/random.h>
58 #include <sys/signalvar.h>
60 #include <sys/vnode.h>
61 #include <sys/sysctl.h>
63 #include <sys/signal2.h>
64 #include <sys/spinlock2.h>
68 #include <vm/vm_map.h>
69 #include <vm/vm_extern.h>
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;
78 static d_mmap_t memmmap;
80 static d_kqfilter_t mmkqfilter;
81 static int memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake);
84 static struct dev_ops mem_ops = {
85 { "mem", 0, D_MPSAFE | D_QUICK },
91 .d_kqfilter = mmkqfilter,
98 static struct dev_ops mem_ops_mem = {
99 { "mem", 0, D_MEM | D_MPSAFE | D_QUICK },
105 .d_kqfilter = mmkqfilter,
109 .d_uksmap = memuksmap
112 static struct dev_ops mem_ops_noq = {
113 { "mem", 0, D_MPSAFE },
119 .d_kqfilter = mmkqfilter,
123 .d_uksmap = memuksmap
126 static int rand_bolt;
128 static cdev_t zerodev = NULL;
129 static struct lock mem_lock = LOCK_INITIALIZER("memlk", 0, 0);
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 *);
135 struct mem_range_softc mem_range_softc;
137 static int seedenable;
138 SYSCTL_INT(_kern, OID_AUTO, seedenable, CTLFLAG_RW, &seedenable, 0, "");
141 mmopen(struct dev_open_args *ap)
143 cdev_t dev = ap->a_head.a_dev;
146 switch (minor(dev)) {
150 * /dev/mem and /dev/kmem
152 if (ap->a_oflags & FWRITE) {
153 if (securelevel > 0 || kernel_mem_readonly)
160 * /dev/kpmap can only be opened for reading.
162 if (ap->a_oflags & FWRITE)
167 error = priv_check_cred(ap->a_cred, PRIV_ROOT, 0);
170 if (securelevel > 0 || kernel_mem_readonly) {
174 error = cpu_set_iopl();
184 mmclose(struct dev_close_args *ap)
186 cdev_t dev = ap->a_head.a_dev;
189 switch (minor(dev)) {
191 error = cpu_clr_iopl();
202 mmrw(cdev_t dev, struct uio *uio, int flags)
212 while (uio->uio_resid > 0 && error == 0) {
214 if (iov->iov_len == 0) {
217 if (uio->uio_iovcnt < 0)
221 switch (minor(dev)) {
224 * minor device 0 is physical memory, /dev/mem
227 v &= ~(long)PAGE_MASK;
228 pmap_kenter((vm_offset_t)ptvmmap, v);
229 o = (int)uio->uio_offset & PAGE_MASK;
230 c = (u_int)(PAGE_SIZE - ((uintptr_t)iov->iov_base & PAGE_MASK));
231 c = min(c, (u_int)(PAGE_SIZE - o));
232 c = min(c, (u_int)iov->iov_len);
233 error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
234 pmap_kremove((vm_offset_t)ptvmmap);
239 * minor device 1 is kernel memory, /dev/kmem
241 vm_offset_t saddr, eaddr;
247 * Make sure that all of the pages are currently
248 * resident so that we don't create any zero-fill
251 saddr = trunc_page(uio->uio_offset);
252 eaddr = round_page(uio->uio_offset + c);
257 * Make sure the kernel addresses are mapped.
258 * platform_direct_mapped() can be used to bypass
259 * default mapping via the page table (virtual kernels
260 * contain a lot of out-of-band data).
263 if (uio->uio_rw != UIO_READ)
264 prot |= VM_PROT_WRITE;
265 error = kvm_access_check(saddr, eaddr, prot);
268 error = uiomove((caddr_t)(vm_offset_t)uio->uio_offset,
274 * minor device 2 (/dev/null) is EOF/RATHOLE
276 if (uio->uio_rw == UIO_READ)
282 * minor device 3 (/dev/random) is source of filth
283 * on read, seeder on write
286 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
287 c = min(iov->iov_len, PAGE_SIZE);
288 if (uio->uio_rw == UIO_WRITE) {
289 error = uiomove(buf, (int)c, uio);
293 error = add_buffer_randomness_src(buf, c, RAND_SRC_SEEDING);
294 } else if (error == 0) {
298 poolsize = read_random(buf, c);
302 if ((flags & IO_NDELAY) != 0)
303 return (EWOULDBLOCK);
306 c = min(c, poolsize);
307 error = uiomove(buf, (int)c, uio);
312 * minor device 4 (/dev/urandom) is source of muck
313 * on read, writes are disallowed.
315 c = min(iov->iov_len, PAGE_SIZE);
316 if (uio->uio_rw == UIO_WRITE) {
320 if (CURSIG(curthread->td_lwp) != 0) {
322 * Use tsleep() to get the error code right.
323 * It should return immediately.
325 error = tsleep(&rand_bolt, PCATCH, "urand", 1);
326 if (error != 0 && error != EWOULDBLOCK)
330 buf = kmalloc(PAGE_SIZE, M_TEMP, M_WAITOK);
331 poolsize = read_random_unlimited(buf, c);
332 c = min(c, poolsize);
333 error = uiomove(buf, (int)c, uio);
335 /* case 5: read/write not supported, mmap only */
336 /* case 6: read/write not supported, mmap only */
339 * minor device 12 (/dev/zero) is source of nulls
340 * on read, write are disallowed.
342 if (uio->uio_rw == UIO_WRITE) {
347 zbuf = (caddr_t)kmalloc(PAGE_SIZE, M_TEMP,
350 c = min(iov->iov_len, PAGE_SIZE);
351 error = uiomove(zbuf, (int)c, uio);
358 iov->iov_base = (char *)iov->iov_base + c;
360 uio->uio_offset += c;
369 mmread(struct dev_read_args *ap)
371 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
375 mmwrite(struct dev_write_args *ap)
377 return(mmrw(ap->a_head.a_dev, ap->a_uio, ap->a_ioflag));
380 /*******************************************************\
381 * allow user processes to MMAP some memory sections *
382 * instead of going through read/write *
383 \*******************************************************/
385 static int user_kernel_mapping(vm_map_backing_t ba, int num,
386 vm_ooffset_t offset, vm_ooffset_t *resultp);
389 memuksmap(vm_map_backing_t ba, int op, cdev_t dev, vm_page_t fake)
400 * We only need to track mappings for /dev/lpmap, all process
401 * mappings will be deleted when the process exits and we
402 * do not need to track kernel mappings.
404 if (minor(dev) == 7) {
406 spin_lock(&lp->lwp_spin);
407 TAILQ_INSERT_TAIL(&lp->lwp_lpmap_backing_list,
409 spin_unlock(&lp->lwp_spin);
414 * We only need to track mappings for /dev/lpmap, all process
415 * mappings will be deleted when the process exits and we
416 * do not need to track kernel mappings.
418 if (minor(dev) == 7) {
420 spin_lock(&lp->lwp_spin);
421 TAILQ_REMOVE(&lp->lwp_lpmap_backing_list, ba, entry);
422 spin_unlock(&lp->lwp_spin);
426 switch (minor(dev)) {
429 * minor device 0 is physical memory
431 fake->phys_addr = ptoa(fake->pindex);
435 * minor device 1 is kernel memory
437 fake->phys_addr = vtophys(ptoa(fake->pindex));
443 * minor device 5 is /dev/upmap (see sys/upmap.h)
444 * minor device 6 is /dev/kpmap (see sys/upmap.h)
445 * minor device 7 is /dev/lpmap (see sys/upmap.h)
448 error = user_kernel_mapping(ba,
452 fake->phys_addr = result;
467 mmioctl(struct dev_ioctl_args *ap)
469 cdev_t dev = ap->a_head.a_dev;
472 lockmgr(&mem_lock, LK_EXCLUSIVE);
474 switch (minor(dev)) {
476 error = mem_ioctl(dev, ap->a_cmd, ap->a_data,
477 ap->a_fflag, ap->a_cred);
481 error = random_ioctl(dev, ap->a_cmd, ap->a_data,
482 ap->a_fflag, ap->a_cred);
489 lockmgr(&mem_lock, LK_RELEASE);
495 * Operations for changing memory attributes.
497 * This is basically just an ioctl shim for mem_range_attr_get
498 * and mem_range_attr_set.
501 mem_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
504 struct mem_range_op *mo = (struct mem_range_op *)data;
505 struct mem_range_desc *md;
507 /* is this for us? */
508 if ((cmd != MEMRANGE_GET) &&
509 (cmd != MEMRANGE_SET))
512 /* any chance we can handle this? */
513 if (mem_range_softc.mr_op == NULL)
516 /* do we have any descriptors? */
517 if (mem_range_softc.mr_ndesc == 0)
522 nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
524 md = (struct mem_range_desc *)
525 kmalloc(nd * sizeof(struct mem_range_desc),
526 M_MEMDESC, M_WAITOK);
527 error = mem_range_attr_get(md, &nd);
529 error = copyout(md, mo->mo_desc,
530 nd * sizeof(struct mem_range_desc));
531 kfree(md, M_MEMDESC);
533 nd = mem_range_softc.mr_ndesc;
539 md = (struct mem_range_desc *)kmalloc(sizeof(struct mem_range_desc),
540 M_MEMDESC, M_WAITOK);
541 error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
542 /* clamp description string */
543 md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
545 error = mem_range_attr_set(md, &mo->mo_arg[0]);
546 kfree(md, M_MEMDESC);
553 * Implementation-neutral, kernel-callable functions for manipulating
554 * memory range attributes.
557 mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
559 /* can we handle this? */
560 if (mem_range_softc.mr_op == NULL)
564 *arg = mem_range_softc.mr_ndesc;
566 bcopy(mem_range_softc.mr_desc, mrd, (*arg) * sizeof(struct mem_range_desc));
572 mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
574 /* can we handle this? */
575 if (mem_range_softc.mr_op == NULL)
578 return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
582 mem_range_AP_init(void)
584 if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
585 mem_range_softc.mr_op->initAP(&mem_range_softc);
589 random_ioctl(cdev_t dev, u_long cmd, caddr_t data, int flags, struct ucred *cred)
595 * Even inspecting the state is privileged, since it gives a hint
596 * about how easily the randomness might be guessed.
601 /* Really handled in upper layer */
605 intr = *(int16_t *)data;
606 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
608 if (intr < 0 || intr >= MAX_INTS)
610 register_randintr(intr);
613 intr = *(int16_t *)data;
614 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
616 if (intr < 0 || intr >= MAX_INTS)
618 unregister_randintr(intr);
624 intr = *(int16_t *)data;
625 if ((error = priv_check_cred(cred, PRIV_ROOT, 0)) != 0)
627 if (intr < 0 || intr >= MAX_INTS)
629 intr = next_registered_randintr(intr);
630 if (intr == MAX_INTS)
632 *(u_int16_t *)data = intr;
642 mm_filter_read(struct knote *kn, long hint)
648 mm_filter_write(struct knote *kn, long hint)
654 dummy_filter_detach(struct knote *kn) {}
656 /* Implemented in kern_nrandom.c */
657 static struct filterops random_read_filtops =
658 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, random_filter_read };
660 static struct filterops mm_read_filtops =
661 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_read };
663 static struct filterops mm_write_filtops =
664 { FILTEROP_ISFD|FILTEROP_MPSAFE, NULL, dummy_filter_detach, mm_filter_write };
667 mmkqfilter(struct dev_kqfilter_args *ap)
669 struct knote *kn = ap->a_kn;
670 cdev_t dev = ap->a_head.a_dev;
673 switch (kn->kn_filter) {
675 switch (minor(dev)) {
677 kn->kn_fop = &random_read_filtops;
680 kn->kn_fop = &mm_read_filtops;
685 kn->kn_fop = &mm_write_filtops;
688 ap->a_result = EOPNOTSUPP;
696 iszerodev(cdev_t dev)
698 return (zerodev == dev);
702 * /dev/lpmap, /dev/upmap, /dev/kpmap.
705 user_kernel_mapping(vm_map_backing_t ba, int num, vm_ooffset_t offset,
706 vm_ooffset_t *resultp)
721 * /dev/upmap - maps RW per-process shared user-kernel area.
725 * If this is a child currently in vfork the pmap is shared
726 * with the parent! We need to actually set-up the parent's
727 * p_upmap, not the child's, and we need to set the invfork
728 * flag. Userland will probably adjust its static state so
729 * it must be consistent with the parent or userland will be
730 * really badly confused.
732 * (this situation can happen when user code in vfork() calls
733 * libc's getpid() or some other function which then decides
734 * it wants the upmap).
739 if (p->p_flags & P_PPWAIT) {
749 * Create the kernel structure as required, set the invfork
750 * flag if we are faulting in on a vfork().
752 if (p->p_upmap == NULL)
753 proc_usermap(p, invfork);
754 if (p->p_upmap && invfork)
755 p->p_upmap->invfork = invfork;
758 * Extract address for pmap
761 offset < roundup2(sizeof(*p->p_upmap), PAGE_SIZE)) {
762 /* only good for current process */
763 *resultp = pmap_kextract((vm_offset_t)p->p_upmap +
770 * /dev/kpmap - maps RO shared kernel global page
772 * Extract address for pmap
775 offset < roundup2(sizeof(*kpmap), PAGE_SIZE)) {
776 *resultp = pmap_kextract((vm_offset_t)kpmap + offset);
782 * /dev/lpmap - maps RW per-thread shared user-kernel area.
789 * Create the kernel structure as required
791 if (lp->lwp_lpmap == NULL)
792 lwp_usermap(lp, -1); /* second arg not yet XXX */
795 * Extract address for pmap
798 offset < roundup2(sizeof(*lp->lwp_lpmap), PAGE_SIZE)) {
799 /* only good for current process */
800 *resultp = pmap_kextract((vm_offset_t)lp->lwp_lpmap +
812 mem_drvinit(void *unused)
815 /* Initialise memory range handling */
816 if (mem_range_softc.mr_op != NULL)
817 mem_range_softc.mr_op->init(&mem_range_softc);
819 make_dev(&mem_ops_mem, 0, UID_ROOT, GID_KMEM, 0640, "mem");
820 make_dev(&mem_ops_mem, 1, UID_ROOT, GID_KMEM, 0640, "kmem");
821 make_dev(&mem_ops, 2, UID_ROOT, GID_WHEEL, 0666, "null");
822 make_dev(&mem_ops, 3, UID_ROOT, GID_WHEEL, 0644, "random");
823 make_dev(&mem_ops, 4, UID_ROOT, GID_WHEEL, 0644, "urandom");
824 make_dev(&mem_ops, 5, UID_ROOT, GID_WHEEL, 0666, "upmap");
825 make_dev(&mem_ops, 6, UID_ROOT, GID_WHEEL, 0444, "kpmap");
826 make_dev(&mem_ops, 7, UID_ROOT, GID_WHEEL, 0666, "lpmap");
827 zerodev = make_dev(&mem_ops, 12, UID_ROOT, GID_WHEEL, 0666, "zero");
828 make_dev(&mem_ops_noq, 14, UID_ROOT, GID_WHEEL, 0600, "io");
831 SYSINIT(memdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE + CDEV_MAJOR, mem_drvinit,