device_pager.c revision 171779
1/*-
2 * Copyright (c) 1990 University of Utah.
3 * Copyright (c) 1991, 1993
4 *	The Regents of the University of California.  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.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 4. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 *	@(#)device_pager.c	8.1 (Berkeley) 6/11/93
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD: head/sys/vm/device_pager.c 171779 2007-08-07 15:36:25Z kib $");
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/conf.h>
43#include <sys/lock.h>
44#include <sys/proc.h>
45#include <sys/mutex.h>
46#include <sys/mman.h>
47#include <sys/sx.h>
48
49#include <vm/vm.h>
50#include <vm/vm_object.h>
51#include <vm/vm_page.h>
52#include <vm/vm_pager.h>
53#include <vm/uma.h>
54
55static void dev_pager_init(void);
56static vm_object_t dev_pager_alloc(void *, vm_ooffset_t, vm_prot_t,
57		vm_ooffset_t);
58static void dev_pager_dealloc(vm_object_t);
59static int dev_pager_getpages(vm_object_t, vm_page_t *, int, int);
60static void dev_pager_putpages(vm_object_t, vm_page_t *, int,
61		boolean_t, int *);
62static boolean_t dev_pager_haspage(vm_object_t, vm_pindex_t, int *,
63		int *);
64
65/* list of device pager objects */
66static struct pagerlst dev_pager_object_list;
67/* protect list manipulation */
68static struct mtx dev_pager_mtx;
69
70
71static uma_zone_t fakepg_zone;
72
73static vm_page_t dev_pager_getfake(vm_paddr_t);
74static void dev_pager_putfake(vm_page_t);
75static void dev_pager_updatefake(vm_page_t, vm_paddr_t);
76
77struct pagerops devicepagerops = {
78	.pgo_init =	dev_pager_init,
79	.pgo_alloc =	dev_pager_alloc,
80	.pgo_dealloc =	dev_pager_dealloc,
81	.pgo_getpages =	dev_pager_getpages,
82	.pgo_putpages =	dev_pager_putpages,
83	.pgo_haspage =	dev_pager_haspage,
84};
85
86static void
87dev_pager_init()
88{
89	TAILQ_INIT(&dev_pager_object_list);
90	mtx_init(&dev_pager_mtx, "dev_pager list", NULL, MTX_DEF);
91	fakepg_zone = uma_zcreate("DP fakepg", sizeof(struct vm_page),
92	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR,
93	    UMA_ZONE_NOFREE|UMA_ZONE_VM);
94}
95
96/*
97 * MPSAFE
98 */
99static vm_object_t
100dev_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot, vm_ooffset_t foff)
101{
102	struct cdev *dev;
103	vm_object_t object, object1;
104	vm_pindex_t pindex;
105	unsigned int npages;
106	vm_paddr_t paddr;
107	vm_offset_t off;
108	struct cdevsw *csw;
109
110	/*
111	 * Offset should be page aligned.
112	 */
113	if (foff & PAGE_MASK)
114		return (NULL);
115
116	size = round_page(size);
117	pindex = OFF_TO_IDX(foff + size);
118
119	/*
120	 * Make sure this device can be mapped.
121	 */
122	dev = handle;
123	csw = dev_refthread(dev);
124	if (csw == NULL)
125		return (NULL);
126
127	/*
128	 * Check that the specified range of the device allows the desired
129	 * protection.
130	 *
131	 * XXX assumes VM_PROT_* == PROT_*
132	 */
133	npages = OFF_TO_IDX(size);
134	for (off = foff; npages--; off += PAGE_SIZE)
135		if ((*csw->d_mmap)(dev, off, &paddr, (int)prot) != 0) {
136			dev_relthread(dev);
137			return (NULL);
138		}
139
140	/*
141	 * Lock to prevent object creation race condition.
142	 */
143	mtx_lock(&dev_pager_mtx);
144
145	/*
146	 * Look up pager, creating as necessary.
147	 */
148	object1 = NULL;
149	object = vm_pager_object_lookup(&dev_pager_object_list, handle);
150	if (object == NULL) {
151		/*
152		 * Allocate object and associate it with the pager.
153		 */
154		mtx_unlock(&dev_pager_mtx);
155		object1 = vm_object_allocate(OBJT_DEVICE, pindex);
156		mtx_lock(&dev_pager_mtx);
157		object = vm_pager_object_lookup(&dev_pager_object_list, handle);
158		if (object != NULL) {
159			/*
160			 * We raced with other thread while allocating object.
161			 */
162			if (pindex > object->size)
163				object->size = pindex;
164		} else {
165			object = object1;
166			object1 = NULL;
167			object->handle = handle;
168			TAILQ_INIT(&object->un_pager.devp.devp_pglist);
169			TAILQ_INSERT_TAIL(&dev_pager_object_list, object,
170			    pager_object_list);
171		}
172	} else {
173		if (pindex > object->size)
174			object->size = pindex;
175	}
176	mtx_unlock(&dev_pager_mtx);
177	dev_relthread(dev);
178	vm_object_deallocate(object1);
179	return (object);
180}
181
182static void
183dev_pager_dealloc(object)
184	vm_object_t object;
185{
186	vm_page_t m;
187
188	VM_OBJECT_UNLOCK(object);
189	mtx_lock(&dev_pager_mtx);
190	TAILQ_REMOVE(&dev_pager_object_list, object, pager_object_list);
191	mtx_unlock(&dev_pager_mtx);
192	VM_OBJECT_LOCK(object);
193	/*
194	 * Free up our fake pages.
195	 */
196	while ((m = TAILQ_FIRST(&object->un_pager.devp.devp_pglist)) != 0) {
197		TAILQ_REMOVE(&object->un_pager.devp.devp_pglist, m, pageq);
198		dev_pager_putfake(m);
199	}
200}
201
202static int
203dev_pager_getpages(object, m, count, reqpage)
204	vm_object_t object;
205	vm_page_t *m;
206	int count;
207	int reqpage;
208{
209	vm_pindex_t offset;
210	vm_paddr_t paddr;
211	vm_page_t page;
212	struct cdev *dev;
213	int i, ret;
214	int prot;
215	struct cdevsw *csw;
216
217	VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
218	dev = object->handle;
219	offset = m[reqpage]->pindex;
220	VM_OBJECT_UNLOCK(object);
221	csw = dev_refthread(dev);
222	if (csw == NULL)
223		panic("dev_pager_getpage: no cdevsw");
224	prot = PROT_READ;	/* XXX should pass in? */
225
226	ret = (*csw->d_mmap)(dev, (vm_offset_t)offset << PAGE_SHIFT, &paddr, prot);
227	KASSERT(ret == 0, ("dev_pager_getpage: map function returns error"));
228	dev_relthread(dev);
229
230	if ((m[reqpage]->flags & PG_FICTITIOUS) != 0) {
231		/*
232		 * If the passed in reqpage page is a fake page, update it with
233		 * the new physical address.
234		 */
235		VM_OBJECT_LOCK(object);
236		dev_pager_updatefake(m[reqpage], paddr);
237		if (count > 1) {
238			vm_page_lock_queues();
239			for (i = 0; i < count; i++) {
240				if (i != reqpage)
241					vm_page_free(m[i]);
242			}
243			vm_page_unlock_queues();
244		}
245	} else {
246		/*
247		 * Replace the passed in reqpage page with our own fake page and
248		 * free up the all of the original pages.
249		 */
250		page = dev_pager_getfake(paddr);
251		VM_OBJECT_LOCK(object);
252		TAILQ_INSERT_TAIL(&object->un_pager.devp.devp_pglist, page, pageq);
253		vm_page_lock_queues();
254		for (i = 0; i < count; i++)
255			vm_page_free(m[i]);
256		vm_page_unlock_queues();
257		vm_page_insert(page, object, offset);
258		m[reqpage] = page;
259	}
260
261	return (VM_PAGER_OK);
262}
263
264static void
265dev_pager_putpages(object, m, count, sync, rtvals)
266	vm_object_t object;
267	vm_page_t *m;
268	int count;
269	boolean_t sync;
270	int *rtvals;
271{
272	panic("dev_pager_putpage called");
273}
274
275static boolean_t
276dev_pager_haspage(object, pindex, before, after)
277	vm_object_t object;
278	vm_pindex_t pindex;
279	int *before;
280	int *after;
281{
282	if (before != NULL)
283		*before = 0;
284	if (after != NULL)
285		*after = 0;
286	return (TRUE);
287}
288
289/*
290 * Instantiate a fictitious page.  Unlike physical memory pages, only
291 * the machine-independent fields must be initialized.
292 */
293static vm_page_t
294dev_pager_getfake(paddr)
295	vm_paddr_t paddr;
296{
297	vm_page_t m;
298
299	m = uma_zalloc(fakepg_zone, M_WAITOK);
300
301	m->flags = PG_FICTITIOUS;
302	m->oflags = VPO_BUSY;
303	m->valid = VM_PAGE_BITS_ALL;
304	m->dirty = 0;
305	m->busy = 0;
306	m->queue = PQ_NONE;
307	m->object = NULL;
308
309	m->wire_count = 1;
310	m->hold_count = 0;
311	m->phys_addr = paddr;
312
313	return (m);
314}
315
316static void
317dev_pager_putfake(m)
318	vm_page_t m;
319{
320	if (!(m->flags & PG_FICTITIOUS))
321		panic("dev_pager_putfake: bad page");
322	uma_zfree(fakepg_zone, m);
323}
324
325static void
326dev_pager_updatefake(m, paddr)
327	vm_page_t m;
328	vm_paddr_t paddr;
329{
330	if (!(m->flags & PG_FICTITIOUS))
331		panic("dev_pager_updatefake: bad page");
332	m->phys_addr = paddr;
333	m->valid = VM_PAGE_BITS_ALL;
334}
335