Add an argument to vfs_add_vnodeops() to specify VVF_* flags for the vop_ops
[dragonfly.git] / sys / vm / vm_swap.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)vm_swap.c 8.5 (Berkeley) 2/17/94
34 * $FreeBSD: src/sys/vm/vm_swap.c,v 1.96.2.2 2001/10/14 18:46:47 iedowse Exp $
dc1be39c 35 * $DragonFly: src/sys/vm/vm_swap.c,v 1.19 2005/09/17 07:43:12 dillon Exp $
984263bc
MD
36 */
37
38#include "opt_swap.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/sysproto.h>
43#include <sys/buf.h>
44#include <sys/proc.h>
fad57d0e 45#include <sys/nlookup.h>
984263bc
MD
46#include <sys/dmap.h> /* XXX */
47#include <sys/vnode.h>
48#include <sys/fcntl.h>
49#include <sys/blist.h>
50#include <sys/kernel.h>
51#include <sys/lock.h>
52#include <sys/conf.h>
53#include <sys/stat.h>
cdd46d2e 54#include <sys/thread2.h>
984263bc
MD
55#include <vm/vm.h>
56#include <vm/vm_extern.h>
57#include <vm/swap_pager.h>
58#include <vm/vm_zone.h>
59
60/*
61 * Indirect driver for multi-controller paging.
62 */
63
64#ifndef NSWAPDEV
65#define NSWAPDEV 4
66#endif
67static struct swdevt should_be_malloced[NSWAPDEV];
68static struct swdevt *swdevt = should_be_malloced;
69static int nswap; /* first block after the interleaved devs */
70static int nswdev = NSWAPDEV;
71int vm_swap_size;
72
1388df65 73static int swapdev_strategy (struct vop_strategy_args *ap);
984263bc
MD
74struct vnode *swapdev_vp;
75
76/*
77 * swapdev_strategy:
78 *
79 * VOP_STRATEGY() for swapdev_vp.
80 * Perform swap strategy interleave device selection.
81 *
82 * The bp is expected to be locked and *not* B_DONE on call.
83 */
84
85static int
57e43348
MD
86swapdev_strategy(struct vop_strategy_args /* {
87 struct vnode *a_vp;
88 struct buf *a_bp;
89 } */ *ap)
984263bc 90{
cdd46d2e 91 int sz, off, seg, index;
5f910b2f 92 struct swdevt *sp;
984263bc
MD
93 struct vnode *vp;
94 struct buf *bp;
95
96 bp = ap->a_bp;
97 sz = howmany(bp->b_bcount, PAGE_SIZE);
98
99 /*
100 * Convert interleaved swap into per-device swap. Note that
101 * the block size is left in PAGE_SIZE'd chunks (for the newswap)
102 * here.
103 */
104 if (nswdev > 1) {
105 off = bp->b_blkno % dmmax;
106 if (off + sz > dmmax) {
107 bp->b_error = EINVAL;
108 bp->b_flags |= B_ERROR;
109 biodone(bp);
110 return 0;
111 }
112 seg = bp->b_blkno / dmmax;
113 index = seg % nswdev;
114 seg /= nswdev;
115 bp->b_blkno = seg * dmmax + off;
116 } else {
117 index = 0;
118 }
119 sp = &swdevt[index];
120 if (bp->b_blkno + sz > sp->sw_nblks) {
121 bp->b_error = EINVAL;
122 bp->b_flags |= B_ERROR;
123 biodone(bp);
124 return 0;
125 }
126 bp->b_dev = sp->sw_device;
127 if (sp->sw_vp == NULL) {
128 bp->b_error = ENODEV;
129 bp->b_flags |= B_ERROR;
130 biodone(bp);
131 return 0;
132 }
133
134 /*
135 * Convert from PAGE_SIZE'd to DEV_BSIZE'd chunks for the actual I/O
136 */
137 bp->b_blkno = ctodb(bp->b_blkno);
138
139 vhold(sp->sw_vp);
cdd46d2e 140 crit_enter();
984263bc
MD
141 if ((bp->b_flags & B_READ) == 0) {
142 vp = bp->b_vp;
143 if (vp) {
144 vp->v_numoutput--;
145 if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) {
146 vp->v_flag &= ~VBWAIT;
147 wakeup(&vp->v_numoutput);
148 }
149 }
150 sp->sw_vp->v_numoutput++;
151 }
152 pbreassignbuf(bp, sp->sw_vp);
cdd46d2e 153 crit_exit();
984263bc
MD
154 VOP_STRATEGY(bp->b_vp, bp);
155 return 0;
156}
157
158/*
159 * Create a special vnode op vector for swapdev_vp - we only use
160 * VOP_STRATEGY(), everything else returns an error.
161 */
2d3e977e 162struct vop_ops *swapdev_vnode_vops;
984263bc 163static struct vnodeopv_entry_desc swapdev_vnodeop_entries[] = {
2d3e977e
MD
164 { &vop_default_desc, vop_defaultop },
165 { &vop_strategy_desc, (void *) swapdev_strategy },
984263bc
MD
166 { NULL, NULL }
167};
168static struct vnodeopv_desc swapdev_vnodeop_opv_desc =
dc1be39c 169 { &swapdev_vnode_vops, swapdev_vnodeop_entries, 0 };
984263bc
MD
170
171VNODEOP_SET(swapdev_vnodeop_opv_desc);
172
173/*
41c20dac
MD
174 * swapon_args(char *name)
175 *
984263bc
MD
176 * System call swapon(name) enables swapping on device name,
177 * which must be in the swdevsw. Return EBUSY
178 * if already swapping on this device.
179 */
984263bc
MD
180/* ARGSUSED */
181int
41c20dac 182swapon(struct swapon_args *uap)
984263bc 183{
dadab5e9 184 struct thread *td = curthread;
984263bc 185 struct vattr attr;
5f910b2f 186 struct vnode *vp;
fad57d0e 187 struct nlookupdata nd;
984263bc 188 int error;
dadab5e9 189 struct ucred *cred;
984263bc 190
dadab5e9
MD
191 KKASSERT(td->td_proc);
192 cred = td->td_proc->p_ucred;
193
194 error = suser(td);
984263bc
MD
195 if (error)
196 return (error);
197
fad57d0e
MD
198 vp = NULL;
199 error = nlookup_init(&nd, uap->name, UIO_USERSPACE, NLC_FOLLOW);
200 if (error == 0)
201 error = nlookup(&nd);
202 if (error == 0)
203 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp);
204 nlookup_done(&nd);
984263bc
MD
205 if (error)
206 return (error);
207
984263bc 208 if (vn_isdisk(vp, &error))
e4c9c0c8 209 error = swaponvp(td, vp, 0);
984263bc 210 else if (vp->v_type == VREG && vp->v_tag == VT_NFS &&
3b568787 211 (error = VOP_GETATTR(vp, &attr, td)) == 0) {
984263bc
MD
212 /*
213 * Allow direct swapping to NFS regular files in the same
214 * way that nfs_mountroot() sets up diskless swapping.
215 */
e4c9c0c8 216 error = swaponvp(td, vp, attr.va_size / DEV_BSIZE);
984263bc
MD
217 }
218
219 if (error)
220 vrele(vp);
221
222 return (error);
223}
224
225/*
226 * Swfree(index) frees the index'th portion of the swap map.
227 * Each of the nswdev devices provides 1/nswdev'th of the swap
228 * space, which is laid out with blocks of dmmax pages circularly
229 * among the devices.
230 *
231 * The new swap code uses page-sized blocks. The old swap code used
232 * DEV_BSIZE'd chunks.
233 *
234 * XXX locking when multiple swapon's run in parallel
235 */
236int
e4c9c0c8 237swaponvp(struct thread *td, struct vnode *vp, u_long nblks)
984263bc 238{
e4c9c0c8
MD
239 u_long aligned_nblks;
240 struct ucred *cred;
5f910b2f
RG
241 struct swdevt *sp;
242 swblk_t vsbase;
984263bc 243 swblk_t dvbase;
e4c9c0c8
MD
244 dev_t dev;
245 int index;
984263bc 246 int error;
e4c9c0c8 247 long blk;
dadab5e9
MD
248
249 KKASSERT(td->td_proc);
250 cred = td->td_proc->p_ucred;
984263bc
MD
251
252 if (!swapdev_vp) {
6ddb7618 253 error = getspecialvnode(VT_NON, NULL, &swapdev_vnode_vops,
3446c007 254 &swapdev_vp, 0, 0);
984263bc
MD
255 if (error)
256 panic("Cannot get vnode for swapdev");
257 swapdev_vp->v_type = VNON; /* Untyped */
5fd012e0 258 vx_unlock(swapdev_vp);
984263bc
MD
259 }
260
261 ASSERT_VOP_UNLOCKED(vp, "swaponvp");
262 for (sp = swdevt, index = 0 ; index < nswdev; index++, sp++) {
263 if (sp->sw_vp == vp)
264 return EBUSY;
265 if (!sp->sw_vp)
266 goto found;
267
268 }
269 return EINVAL;
270 found:
5fd012e0 271 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
fad57d0e 272 error = VOP_OPEN(vp, FREAD | FWRITE, cred, NULL, td);
5fd012e0 273 VOP_UNLOCK(vp, 0, td);
984263bc
MD
274 if (error)
275 return (error);
276
e4c9c0c8
MD
277 /*
278 * v_rdev is not valid until after the VOP_OPEN() call. dev_psize()
279 * must be supported if a character device has been specified.
280 */
281 if (vp->v_type == VCHR)
282 dev = vp->v_rdev;
283 else
284 dev = NODEV;
285
335dda38 286 if (nblks == 0 && dev != NODEV && ((nblks = dev_dpsize(dev)) == -1)) {
3b568787 287 (void) VOP_CLOSE(vp, FREAD | FWRITE, td);
984263bc
MD
288 return (ENXIO);
289 }
290 if (nblks == 0) {
3b568787 291 (void) VOP_CLOSE(vp, FREAD | FWRITE, td);
984263bc
MD
292 return (ENXIO);
293 }
294
295 /*
296 * If we go beyond this, we get overflows in the radix
297 * tree bitmap code.
298 */
299 if (nblks > 0x40000000 / BLIST_META_RADIX / nswdev) {
300 printf("exceeded maximum of %d blocks per swap unit\n",
301 0x40000000 / BLIST_META_RADIX / nswdev);
3b568787 302 (void) VOP_CLOSE(vp, FREAD | FWRITE, td);
984263bc
MD
303 return (ENXIO);
304 }
305 /*
306 * nblks is in DEV_BSIZE'd chunks, convert to PAGE_SIZE'd chunks.
307 * First chop nblks off to page-align it, then convert.
308 *
309 * sw->sw_nblks is in page-sized chunks now too.
310 */
311 nblks &= ~(ctodb(1) - 1);
312 nblks = dbtoc(nblks);
313
314 sp->sw_vp = vp;
315 sp->sw_dev = dev2udev(dev);
316 sp->sw_device = dev;
317 sp->sw_flags |= SW_FREED;
318 sp->sw_nblks = nblks;
319
320 /*
321 * nblks, nswap, and dmmax are PAGE_SIZE'd parameters now, not
322 * DEV_BSIZE'd. aligned_nblks is used to calculate the
323 * size of the swap bitmap, taking into account the stripe size.
324 */
325 aligned_nblks = (nblks + (dmmax - 1)) & ~(u_long)(dmmax - 1);
326
327 if (aligned_nblks * nswdev > nswap)
328 nswap = aligned_nblks * nswdev;
329
330 if (swapblist == NULL)
331 swapblist = blist_create(nswap);
332 else
333 blist_resize(&swapblist, nswap, 0);
334
335 for (dvbase = dmmax; dvbase < nblks; dvbase += dmmax) {
336 blk = min(nblks - dvbase, dmmax);
337 vsbase = index * dmmax + dvbase * nswdev;
338 blist_free(swapblist, vsbase, blk);
339 vm_swap_size += blk;
340 }
341
342 return (0);
343}