device_pager.c revision 183383
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 183383 2008-09-26 14:50:49Z 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	mtx_lock(&dev_pager_mtx);
141
142	/*
143	 * Look up pager, creating as necessary.
144	 */
145	object1 = NULL;
146	object = vm_pager_object_lookup(&dev_pager_object_list, handle);
147	if (object == NULL) {
148		/*
149		 * Allocate object and associate it with the pager.  Initialize
150		 * the object's pg_color based upon the physical address of the
151		 * device's memory.
152		 */
153		mtx_unlock(&dev_pager_mtx);
154		object1 = vm_object_allocate(OBJT_DEVICE, pindex);
155		object1->flags |= OBJ_COLORED;
156		object1->pg_color = atop(paddr) - OFF_TO_IDX(off - PAGE_SIZE);
157		mtx_lock(&dev_pager_mtx);
158		object = vm_pager_object_lookup(&dev_pager_object_list, handle);
159		if (object != NULL) {
160			/*
161			 * We raced with other thread while allocating object.
162			 */
163			if (pindex > object->size)
164				object->size = pindex;
165		} else {
166			object = object1;
167			object1 = NULL;
168			object->handle = handle;
169			TAILQ_INIT(&object->un_pager.devp.devp_pglist);
170			TAILQ_INSERT_TAIL(&dev_pager_object_list, object,
171			    pager_object_list);
172		}
173	} else {
174		if (pindex > object->size)
175			object->size = pindex;
176	}
177	mtx_unlock(&dev_pager_mtx);
178	dev_relthread(dev);
179	vm_object_deallocate(object1);
180	return (object);
181}
182
183static void
184dev_pager_dealloc(object)
185	vm_object_t object;
186{
187	vm_page_t m;
188
189	VM_OBJECT_UNLOCK(object);
190	mtx_lock(&dev_pager_mtx);
191	TAILQ_REMOVE(&dev_pager_object_list, object, pager_object_list);
192	mtx_unlock(&dev_pager_mtx);
193	VM_OBJECT_LOCK(object);
194	/*
195	 * Free up our fake pages.
196	 */
197	while ((m = TAILQ_FIRST(&object->un_pager.devp.devp_pglist)) != 0) {
198		TAILQ_REMOVE(&object->un_pager.devp.devp_pglist, m, pageq);
199		dev_pager_putfake(m);
200	}
201}
202
203static int
204dev_pager_getpages(object, m, count, reqpage)
205	vm_object_t object;
206	vm_page_t *m;
207	int count;
208	int reqpage;
209{
210	vm_pindex_t offset;
211	vm_paddr_t paddr;
212	vm_page_t page;
213	struct cdev *dev;
214	int i, ret;
215	int prot;
216	struct cdevsw *csw;
217	struct thread *td;
218	struct file *fpop;
219
220	VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
221	dev = object->handle;
222	offset = m[reqpage]->pindex;
223	VM_OBJECT_UNLOCK(object);
224	csw = dev_refthread(dev);
225	if (csw == NULL)
226		panic("dev_pager_getpage: no cdevsw");
227	prot = PROT_READ;	/* XXX should pass in? */
228
229	td = curthread;
230	fpop = td->td_fpop;
231	td->td_fpop = NULL;
232	ret = (*csw->d_mmap)(dev, (vm_offset_t)offset << PAGE_SHIFT, &paddr, prot);
233	KASSERT(ret == 0, ("dev_pager_getpage: map function returns error"));
234	td->td_fpop = fpop;
235	dev_relthread(dev);
236
237	if ((m[reqpage]->flags & PG_FICTITIOUS) != 0) {
238		/*
239		 * If the passed in reqpage page is a fake page, update it with
240		 * the new physical address.
241		 */
242		VM_OBJECT_LOCK(object);
243		dev_pager_updatefake(m[reqpage], paddr);
244		if (count > 1) {
245			vm_page_lock_queues();
246			for (i = 0; i < count; i++) {
247				if (i != reqpage)
248					vm_page_free(m[i]);
249			}
250			vm_page_unlock_queues();
251		}
252	} else {
253		/*
254		 * Replace the passed in reqpage page with our own fake page and
255		 * free up the all of the original pages.
256		 */
257		page = dev_pager_getfake(paddr);
258		VM_OBJECT_LOCK(object);
259		TAILQ_INSERT_TAIL(&object->un_pager.devp.devp_pglist, page, pageq);
260		vm_page_lock_queues();
261		for (i = 0; i < count; i++)
262			vm_page_free(m[i]);
263		vm_page_unlock_queues();
264		vm_page_insert(page, object, offset);
265		m[reqpage] = page;
266	}
267
268	return (VM_PAGER_OK);
269}
270
271static void
272dev_pager_putpages(object, m, count, sync, rtvals)
273	vm_object_t object;
274	vm_page_t *m;
275	int count;
276	boolean_t sync;
277	int *rtvals;
278{
279	panic("dev_pager_putpage called");
280}
281
282static boolean_t
283dev_pager_haspage(object, pindex, before, after)
284	vm_object_t object;
285	vm_pindex_t pindex;
286	int *before;
287	int *after;
288{
289	if (before != NULL)
290		*before = 0;
291	if (after != NULL)
292		*after = 0;
293	return (TRUE);
294}
295
296/*
297 * Instantiate a fictitious page.  Unlike physical memory pages, only
298 * the machine-independent fields must be initialized.
299 */
300static vm_page_t
301dev_pager_getfake(paddr)
302	vm_paddr_t paddr;
303{
304	vm_page_t m;
305
306	m = uma_zalloc(fakepg_zone, M_WAITOK);
307
308	m->flags = PG_FICTITIOUS;
309	m->oflags = VPO_BUSY;
310	m->valid = VM_PAGE_BITS_ALL;
311	m->dirty = 0;
312	m->busy = 0;
313	m->queue = PQ_NONE;
314	m->object = NULL;
315
316	m->wire_count = 1;
317	m->hold_count = 0;
318	m->phys_addr = paddr;
319
320	return (m);
321}
322
323static void
324dev_pager_putfake(m)
325	vm_page_t m;
326{
327	if (!(m->flags & PG_FICTITIOUS))
328		panic("dev_pager_putfake: bad page");
329	uma_zfree(fakepg_zone, m);
330}
331
332static void
333dev_pager_updatefake(m, paddr)
334	vm_page_t m;
335	vm_paddr_t paddr;
336{
337	if (!(m->flags & PG_FICTITIOUS))
338		panic("dev_pager_updatefake: bad page");
339	m->phys_addr = paddr;
340	m->valid = VM_PAGE_BITS_ALL;
341}
342