xref: /netbsd-src/sys/rump/librump/rumpvfs/vm_vfs.c (revision a5847cc334d9a7029f6352b847e9e8d71a0f9e0c)
1 /*	$NetBSD: vm_vfs.c,v 1.32 2011/06/19 18:28:24 hannken Exp $	*/
2 
3 /*
4  * Copyright (c) 2008-2011 Antti Kantee.  All Rights Reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  *
15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS
16  * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
17  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
18  * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
21  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25  * SUCH DAMAGE.
26  */
27 
28 #include <sys/cdefs.h>
29 __KERNEL_RCSID(0, "$NetBSD: vm_vfs.c,v 1.32 2011/06/19 18:28:24 hannken Exp $");
30 
31 #include <sys/param.h>
32 
33 #include <sys/buf.h>
34 #include <sys/vnode.h>
35 
36 #include <uvm/uvm.h>
37 #include <uvm/uvm_readahead.h>
38 
39 /*
40  * release resources held during async io.  this is almost the
41  * same as uvm_aio_aiodone() from uvm_pager.c and only lacks the
42  * call to uvm_aio_aiodone_pages(): unbusies pages directly here.
43  */
44 void
45 uvm_aio_aiodone(struct buf *bp)
46 {
47 	struct uvm_object *uobj;
48 	int i, npages = bp->b_bufsize >> PAGE_SHIFT;
49 	struct vm_page **pgs;
50 	vaddr_t va;
51 	int pageout = 0;
52 
53 	KASSERT(npages > 0);
54 	pgs = kmem_alloc(npages * sizeof(*pgs), KM_SLEEP);
55 	for (i = 0; i < npages; i++) {
56 		va = (vaddr_t)bp->b_data + (i << PAGE_SHIFT);
57 		pgs[i] = uvm_pageratop(va);
58 		if (pgs[i]->flags & PG_PAGEOUT) {
59 			KASSERT((pgs[i]->flags & PG_FAKE) == 0);
60 			pageout++;
61 			pgs[i]->flags &= ~PG_PAGEOUT;
62 			pgs[i]->flags |= PG_RELEASED;
63 		}
64 	}
65 
66 	uvm_pagermapout((vaddr_t)bp->b_data, npages);
67 
68 	/* get uobj because we need it after pages might be recycled */
69 	uobj = pgs[0]->uobject;
70 	KASSERT(uobj);
71 
72 	mutex_enter(uobj->vmobjlock);
73 	mutex_enter(&uvm_pageqlock);
74 	uvm_page_unbusy(pgs, npages);
75 	mutex_exit(&uvm_pageqlock);
76 	mutex_exit(uobj->vmobjlock);
77 
78 	uvm_pageout_done(pageout);
79 
80 	if (BUF_ISWRITE(bp) && (bp->b_cflags & BC_AGE) != 0) {
81 		mutex_enter(bp->b_objlock);
82 		vwakeup(bp);
83 		mutex_exit(bp->b_objlock);
84 	}
85 
86 	putiobuf(bp);
87 
88 	kmem_free(pgs, npages * sizeof(*pgs));
89 }
90 
91 void
92 uvm_aio_biodone(struct buf *bp)
93 {
94 
95 	uvm_aio_aiodone(bp);
96 }
97 
98 /*
99  * UBC
100  */
101 
102 #define PAGERFLAGS (PGO_SYNCIO | PGO_NOBLOCKALLOC | PGO_NOTIMESTAMP)
103 
104 void
105 ubc_zerorange(struct uvm_object *uobj, off_t off, size_t len, int flags)
106 {
107 	struct vm_page **pgs;
108 	struct uvm_object *pguobj;
109 	int maxpages = MIN(32, round_page(len) >> PAGE_SHIFT);
110 	int rv, npages, i;
111 
112 	if (maxpages == 0)
113 		return;
114 
115 	pgs = kmem_alloc(maxpages * sizeof(pgs), KM_SLEEP);
116 	mutex_enter(uobj->vmobjlock);
117 	while (len) {
118 		npages = MIN(maxpages, round_page(len) >> PAGE_SHIFT);
119 		memset(pgs, 0, npages * sizeof(struct vm_page *));
120 		rv = uobj->pgops->pgo_get(uobj, trunc_page(off),
121 		    pgs, &npages, 0, VM_PROT_READ | VM_PROT_WRITE,
122 		    0, PAGERFLAGS | PGO_PASTEOF);
123 		KASSERT(npages > 0);
124 
125 		for (i = 0, pguobj = NULL; i < npages; i++) {
126 			struct vm_page *pg;
127 			uint8_t *start;
128 			size_t chunkoff, chunklen;
129 
130 			pg = pgs[i];
131 			if (pg == NULL)
132 				break;
133 			if (pguobj == NULL)
134 				pguobj = pg->uobject;
135 			KASSERT(pguobj == pg->uobject);
136 
137 			chunkoff = off & PAGE_MASK;
138 			chunklen = MIN(PAGE_SIZE - chunkoff, len);
139 			start = (uint8_t *)pg->uanon + chunkoff;
140 
141 			memset(start, 0, chunklen);
142 			pg->flags &= ~PG_CLEAN;
143 
144 			off += chunklen;
145 			len -= chunklen;
146 		}
147 		mutex_enter(pguobj->vmobjlock);
148 		uvm_page_unbusy(pgs, npages);
149 		if (pguobj != uobj) {
150 			mutex_exit(pguobj->vmobjlock);
151 			mutex_enter(uobj->vmobjlock);
152 		}
153 	}
154 	mutex_exit(uobj->vmobjlock);
155 	kmem_free(pgs, maxpages * sizeof(pgs));
156 
157 	return;
158 }
159 
160 #define len2npages(off, len)						\
161     ((round_page(off+len) - trunc_page(off)) >> PAGE_SHIFT)
162 
163 int
164 ubc_uiomove(struct uvm_object *uobj, struct uio *uio, vsize_t todo,
165 	int advice, int flags)
166 {
167 	struct vm_page **pgs;
168 	struct uvm_object *pguobj;
169 	int npages = len2npages(uio->uio_offset, todo);
170 	size_t pgalloc;
171 	int i, rv, pagerflags;
172 	vm_prot_t prot;
173 
174 	pgalloc = npages * sizeof(pgs);
175 	pgs = kmem_alloc(pgalloc, KM_SLEEP);
176 
177 	pagerflags = PAGERFLAGS;
178 	if (flags & UBC_WRITE)
179 		pagerflags |= PGO_PASTEOF;
180 	if (flags & UBC_FAULTBUSY)
181 		pagerflags |= PGO_OVERWRITE;
182 
183 	prot = VM_PROT_READ;
184 	if (flags & UBC_WRITE)
185 		prot |= VM_PROT_WRITE;
186 
187 	mutex_enter(uobj->vmobjlock);
188 	do {
189 		npages = len2npages(uio->uio_offset, todo);
190 		memset(pgs, 0, pgalloc);
191 		rv = uobj->pgops->pgo_get(uobj, trunc_page(uio->uio_offset),
192 		    pgs, &npages, 0, prot, 0, pagerflags);
193 		if (rv)
194 			goto out;
195 
196 		for (i = 0, pguobj = NULL; i < npages; i++) {
197 			struct vm_page *pg;
198 			size_t xfersize;
199 			off_t pageoff;
200 
201 			pg = pgs[i];
202 			if (pg == NULL)
203 				break;
204 			if (pguobj == NULL)
205 				pguobj = pg->uobject;
206 			KASSERT(pguobj == pg->uobject);
207 
208 			pageoff = uio->uio_offset & PAGE_MASK;
209 			xfersize = MIN(MIN(todo, PAGE_SIZE), PAGE_SIZE-pageoff);
210 			KASSERT(xfersize > 0);
211 			rv = uiomove((uint8_t *)pg->uanon + pageoff,
212 			    xfersize, uio);
213 			if (rv) {
214 				mutex_enter(pguobj->vmobjlock);
215 				uvm_page_unbusy(pgs, npages);
216 				mutex_exit(pguobj->vmobjlock);
217 				goto out;
218 			}
219 			if (uio->uio_rw == UIO_WRITE)
220 				pg->flags &= ~(PG_CLEAN | PG_FAKE);
221 			todo -= xfersize;
222 		}
223 		mutex_enter(pguobj->vmobjlock);
224 		uvm_page_unbusy(pgs, npages);
225 		if (pguobj != uobj) {
226 			mutex_exit(pguobj->vmobjlock);
227 			mutex_enter(uobj->vmobjlock);
228 		}
229 	} while (todo);
230 	mutex_exit(uobj->vmobjlock);
231 
232  out:
233 	kmem_free(pgs, pgalloc);
234 	return rv;
235 }
236