mem.c revision 104908
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: head/sys/i386/i386/mem.c 104908 2002-10-11 14:58:34Z mike $
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/lock.h>
54#include <sys/malloc.h>
55#include <sys/memrange.h>
56#include <sys/mutex.h>
57#include <sys/proc.h>
58#include <sys/signalvar.h>
59#include <sys/systm.h>
60#include <sys/uio.h>
61
62#include <machine/db_machdep.h>
63#include <machine/frame.h>
64#include <machine/psl.h>
65#include <machine/specialreg.h>
66
67#include <vm/vm.h>
68#include <vm/pmap.h>
69#include <vm/vm_extern.h>
70
71static dev_t memdev, kmemdev, iodev;
72
73static	d_open_t	mmopen;
74static	d_close_t	mmclose;
75static	d_read_t	mmrw;
76static	d_ioctl_t	mmioctl;
77static	d_mmap_t	memmmap;
78
79#define CDEV_MAJOR 2
80static struct cdevsw mem_cdevsw = {
81	/* open */	mmopen,
82	/* close */	mmclose,
83	/* read */	mmrw,
84	/* write */	mmrw,
85	/* ioctl */	mmioctl,
86	/* poll */	(d_poll_t *)seltrue,
87	/* mmap */	memmmap,
88	/* strategy */	nostrategy,
89	/* name */	"mem",
90	/* maj */	CDEV_MAJOR,
91	/* dump */	nodump,
92	/* psize */	nopsize,
93	/* flags */	D_MEM,
94};
95
96MALLOC_DEFINE(M_MEMDESC, "memdesc", "memory range descriptors");
97
98struct mem_range_softc mem_range_softc;
99
100static int
101mmclose(dev_t dev, int flags, int fmt, struct thread *td)
102{
103	switch (minor(dev)) {
104	case 14:
105		td->td_frame->tf_eflags &= ~PSL_IOPL;
106	}
107	return (0);
108}
109
110static int
111mmopen(dev_t dev, int flags, int fmt, struct thread *td)
112{
113	int error;
114
115	switch (minor(dev)) {
116	case 0:
117	case 1:
118		if (flags & FWRITE) {
119			error = securelevel_gt(td->td_ucred, 0);
120			if (error != 0)
121				return (error);
122		}
123		break;
124	case 14:
125		error = suser(td);
126		if (error != 0)
127			return (error);
128		error = securelevel_gt(td->td_ucred, 0);
129		if (error != 0)
130			return (error);
131		td->td_frame->tf_eflags |= PSL_IOPL;
132		break;
133	}
134	return (0);
135}
136
137/*ARGSUSED*/
138static int
139mmrw(dev_t dev, struct uio *uio, int flags)
140{
141	int o;
142	u_int c = 0, v;
143	struct iovec *iov;
144	int error = 0;
145	vm_offset_t addr, eaddr;
146
147	GIANT_REQUIRED;
148
149	while (uio->uio_resid > 0 && error == 0) {
150		iov = uio->uio_iov;
151		if (iov->iov_len == 0) {
152			uio->uio_iov++;
153			uio->uio_iovcnt--;
154			if (uio->uio_iovcnt < 0)
155				panic("mmrw");
156			continue;
157		}
158		switch (minor(dev)) {
159
160/* minor device 0 is physical memory */
161		case 0:
162			v = uio->uio_offset;
163			v &= ~PAGE_MASK;
164			pmap_kenter((vm_offset_t)ptvmmap, v);
165			o = (int)uio->uio_offset & PAGE_MASK;
166			c = (u_int)(PAGE_SIZE - ((int)iov->iov_base & PAGE_MASK));
167			c = min(c, (u_int)(PAGE_SIZE - o));
168			c = min(c, (u_int)iov->iov_len);
169			error = uiomove((caddr_t)&ptvmmap[o], (int)c, uio);
170			pmap_qremove((vm_offset_t)ptvmmap, 1);
171			continue;
172
173/* minor device 1 is kernel memory */
174		case 1:
175			c = iov->iov_len;
176
177			/*
178			 * Make sure that all of the pages are currently resident so
179			 * that we don't create any zero-fill pages.
180			 */
181			addr = trunc_page(uio->uio_offset);
182			eaddr = round_page(uio->uio_offset + c);
183
184			if (addr < (vm_offset_t)VADDR(PTDPTDI, 0))
185				return (EFAULT);
186			if (eaddr >= (vm_offset_t)VADDR(APTDPTDI, 0))
187				return (EFAULT);
188			for (; addr < eaddr; addr += PAGE_SIZE)
189				if (pmap_extract(kernel_pmap, addr) == 0)
190					return (EFAULT);
191
192			if (!kernacc((caddr_t)(int)uio->uio_offset, c,
193			    uio->uio_rw == UIO_READ ?
194			    VM_PROT_READ : VM_PROT_WRITE))
195				return (EFAULT);
196			error = uiomove((caddr_t)(int)uio->uio_offset, (int)c, uio);
197			continue;
198		}
199
200		if (error)
201			break;
202		iov->iov_base = (char *)iov->iov_base + c;
203		iov->iov_len -= c;
204		uio->uio_offset += c;
205		uio->uio_resid -= c;
206	}
207	return (error);
208}
209
210/*******************************************************\
211* allow user processes to MMAP some memory sections	*
212* instead of going through read/write			*
213\*******************************************************/
214static int
215memmmap(dev_t dev, vm_offset_t offset, int prot)
216{
217	switch (minor(dev))
218	{
219
220	/* minor device 0 is physical memory */
221	case 0:
222        	return (i386_btop(offset));
223
224	/* minor device 1 is kernel memory */
225	case 1:
226        	return (i386_btop(vtophys(offset)));
227
228	default:
229		return (-1);
230	}
231}
232
233/*
234 * Operations for changing memory attributes.
235 *
236 * This is basically just an ioctl shim for mem_range_attr_get
237 * and mem_range_attr_set.
238 */
239static int
240mmioctl(dev_t dev, u_long cmd, caddr_t data, int flags, struct thread *td)
241{
242	int nd, error = 0;
243	struct mem_range_op *mo = (struct mem_range_op *)data;
244	struct mem_range_desc *md;
245
246	/* is this for us? */
247	if ((cmd != MEMRANGE_GET) &&
248	    (cmd != MEMRANGE_SET))
249		return (ENOTTY);
250
251	/* any chance we can handle this? */
252	if (mem_range_softc.mr_op == NULL)
253		return (EOPNOTSUPP);
254
255	/* do we have any descriptors? */
256	if (mem_range_softc.mr_ndesc == 0)
257		return (ENXIO);
258
259	switch (cmd) {
260	case MEMRANGE_GET:
261		nd = imin(mo->mo_arg[0], mem_range_softc.mr_ndesc);
262		if (nd > 0) {
263			md = (struct mem_range_desc *)
264				malloc(nd * sizeof(struct mem_range_desc),
265				       M_MEMDESC, M_WAITOK);
266			error = mem_range_attr_get(md, &nd);
267			if (!error)
268				error = copyout(md, mo->mo_desc,
269					nd * sizeof(struct mem_range_desc));
270			free(md, M_MEMDESC);
271		}
272		else
273			nd = mem_range_softc.mr_ndesc;
274		mo->mo_arg[0] = nd;
275		break;
276
277	case MEMRANGE_SET:
278		md = (struct mem_range_desc *)malloc(sizeof(struct mem_range_desc),
279						    M_MEMDESC, M_WAITOK);
280		error = copyin(mo->mo_desc, md, sizeof(struct mem_range_desc));
281		/* clamp description string */
282		md->mr_owner[sizeof(md->mr_owner) - 1] = 0;
283		if (error == 0)
284			error = mem_range_attr_set(md, &mo->mo_arg[0]);
285		free(md, M_MEMDESC);
286		break;
287	}
288	return (error);
289}
290
291/*
292 * Implementation-neutral, kernel-callable functions for manipulating
293 * memory range attributes.
294 */
295int
296mem_range_attr_get(struct mem_range_desc *mrd, int *arg)
297{
298	/* can we handle this? */
299	if (mem_range_softc.mr_op == NULL)
300		return (EOPNOTSUPP);
301
302	if (*arg == 0)
303		*arg = mem_range_softc.mr_ndesc;
304	else
305		bcopy(mem_range_softc.mr_desc, mrd,
306			(*arg) * sizeof(struct mem_range_desc));
307	return (0);
308}
309
310int
311mem_range_attr_set(struct mem_range_desc *mrd, int *arg)
312{
313	/* can we handle this? */
314	if (mem_range_softc.mr_op == NULL)
315		return (EOPNOTSUPP);
316
317	return (mem_range_softc.mr_op->set(&mem_range_softc, mrd, arg));
318}
319
320#ifdef SMP
321void
322mem_range_AP_init(void)
323{
324	if (mem_range_softc.mr_op && mem_range_softc.mr_op->initAP)
325		(mem_range_softc.mr_op->initAP(&mem_range_softc));
326}
327#endif
328
329static int
330mem_modevent(module_t mod, int type, void *data)
331{
332	switch(type) {
333	case MOD_LOAD:
334		if (bootverbose)
335			printf("mem: <memory & I/O>\n");
336		/* Initialise memory range handling */
337		if (mem_range_softc.mr_op != NULL)
338			mem_range_softc.mr_op->init(&mem_range_softc);
339
340		memdev = make_dev(&mem_cdevsw, 0, UID_ROOT, GID_KMEM,
341			0640, "mem");
342		kmemdev = make_dev(&mem_cdevsw, 1, UID_ROOT, GID_KMEM,
343			0640, "kmem");
344		iodev = make_dev(&mem_cdevsw, 14, UID_ROOT, GID_WHEEL,
345			0600, "io");
346		return (0);
347
348	case MOD_UNLOAD:
349		destroy_dev(memdev);
350		destroy_dev(kmemdev);
351		destroy_dev(iodev);
352		return (0);
353
354	case MOD_SHUTDOWN:
355		return (0);
356
357	default:
358		return (EOPNOTSUPP);
359	}
360}
361
362DEV_MODULE(mem, mem_modevent, NULL);
363