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 $ 35 * $DragonFly: src/sys/vm/vm_swap.c,v 1.21 2006/02/17 19:18:08 dillon Exp $ 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> 45 #include <sys/nlookup.h> 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> 54 #include <sys/thread2.h> 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 67 static struct swdevt should_be_malloced[NSWAPDEV]; 68 static struct swdevt *swdevt = should_be_malloced; 69 static int nswap; /* first block after the interleaved devs */ 70 static int nswdev = NSWAPDEV; 71 int vm_swap_size; 72 73 static int swapdev_strategy (struct vop_strategy_args *ap); 74 struct vnode *swapdev_vp; 75 76 /* 77 * swapdev_strategy: 78 * 79 * vn_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 * (struct vnode *a_vp, struct bio *b_bio) 85 */ 86 87 static int 88 swapdev_strategy(struct vop_strategy_args *ap) 89 { 90 struct bio *bio = ap->a_bio; 91 struct bio *nbio; 92 struct buf *bp = bio->bio_buf; 93 int sz, off, seg, index; 94 struct swdevt *sp; 95 struct vnode *vp; 96 97 vp = ap->a_vp; 98 sz = howmany(bp->b_bcount, PAGE_SIZE); 99 100 /* 101 * Convert interleaved swap into per-device swap. Note that 102 * the block size is left in PAGE_SIZE'd chunks (for the newswap) 103 * here. 104 */ 105 nbio = push_bio(bio); 106 if (nswdev > 1) { 107 off = bio->bio_blkno % dmmax; 108 if (off + sz > dmmax) { 109 bp->b_error = EINVAL; 110 bp->b_flags |= B_ERROR; 111 biodone(bio); 112 return 0; 113 } 114 seg = bio->bio_blkno / dmmax; 115 index = seg % nswdev; 116 seg /= nswdev; 117 nbio->bio_blkno = seg * dmmax + off; 118 } else { 119 index = 0; 120 nbio->bio_blkno = bio->bio_blkno; 121 } 122 sp = &swdevt[index]; 123 if (nbio->bio_blkno + sz > sp->sw_nblks) { 124 bp->b_error = EINVAL; 125 bp->b_flags |= B_ERROR; 126 /* I/O was never started on nbio, must biodone(bio) */ 127 biodone(bio); 128 return 0; 129 } 130 if (sp->sw_vp == NULL) { 131 bp->b_error = ENODEV; 132 bp->b_flags |= B_ERROR; 133 /* I/O was never started on nbio, must biodone(bio) */ 134 biodone(bio); 135 return 0; 136 } 137 138 /* 139 * Convert from PAGE_SIZE'd to DEV_BSIZE'd chunks for the actual I/O. 140 * Issue a strategy call on the appropriate swap vnode. Note that 141 * bp->b_vp is not modified. Strategy code is always supposed to 142 * use the passed vp. 143 * 144 * XXX do a dev_dstrategy() call on sp->sw_device instead of on 145 * sp->sw_vp ? 146 */ 147 nbio->bio_blkno = ctodb(nbio->bio_blkno); 148 vn_strategy(sp->sw_vp, nbio); 149 return 0; 150 } 151 152 /* 153 * Create a special vnode op vector for swapdev_vp - we only use 154 * vn_strategy(), everything else returns an error. 155 */ 156 struct vop_ops *swapdev_vnode_vops; 157 static struct vnodeopv_entry_desc swapdev_vnodeop_entries[] = { 158 { &vop_default_desc, vop_defaultop }, 159 { &vop_strategy_desc, (void *) swapdev_strategy }, 160 { NULL, NULL } 161 }; 162 static struct vnodeopv_desc swapdev_vnodeop_opv_desc = 163 { &swapdev_vnode_vops, swapdev_vnodeop_entries, 0 }; 164 165 VNODEOP_SET(swapdev_vnodeop_opv_desc); 166 167 /* 168 * swapon_args(char *name) 169 * 170 * System call swapon(name) enables swapping on device name, 171 * which must be in the swdevsw. Return EBUSY 172 * if already swapping on this device. 173 */ 174 /* ARGSUSED */ 175 int 176 swapon(struct swapon_args *uap) 177 { 178 struct thread *td = curthread; 179 struct vattr attr; 180 struct vnode *vp; 181 struct nlookupdata nd; 182 int error; 183 struct ucred *cred; 184 185 KKASSERT(td->td_proc); 186 cred = td->td_proc->p_ucred; 187 188 error = suser(td); 189 if (error) 190 return (error); 191 192 vp = NULL; 193 error = nlookup_init(&nd, uap->name, UIO_USERSPACE, NLC_FOLLOW); 194 if (error == 0) 195 error = nlookup(&nd); 196 if (error == 0) 197 error = cache_vref(nd.nl_ncp, nd.nl_cred, &vp); 198 nlookup_done(&nd); 199 if (error) 200 return (error); 201 202 if (vn_isdisk(vp, &error)) 203 error = swaponvp(td, vp, 0); 204 else if (vp->v_type == VREG && vp->v_tag == VT_NFS && 205 (error = VOP_GETATTR(vp, &attr, td)) == 0) { 206 /* 207 * Allow direct swapping to NFS regular files in the same 208 * way that nfs_mountroot() sets up diskless swapping. 209 */ 210 error = swaponvp(td, vp, attr.va_size / DEV_BSIZE); 211 } 212 213 if (error) 214 vrele(vp); 215 216 return (error); 217 } 218 219 /* 220 * Swfree(index) frees the index'th portion of the swap map. 221 * Each of the nswdev devices provides 1/nswdev'th of the swap 222 * space, which is laid out with blocks of dmmax pages circularly 223 * among the devices. 224 * 225 * The new swap code uses page-sized blocks. The old swap code used 226 * DEV_BSIZE'd chunks. 227 * 228 * XXX locking when multiple swapon's run in parallel 229 */ 230 int 231 swaponvp(struct thread *td, struct vnode *vp, u_long nblks) 232 { 233 u_long aligned_nblks; 234 struct ucred *cred; 235 struct swdevt *sp; 236 swblk_t vsbase; 237 swblk_t dvbase; 238 dev_t dev; 239 int index; 240 int error; 241 long blk; 242 243 KKASSERT(td->td_proc); 244 cred = td->td_proc->p_ucred; 245 246 if (!swapdev_vp) { 247 error = getspecialvnode(VT_NON, NULL, &swapdev_vnode_vops, 248 &swapdev_vp, 0, 0); 249 if (error) 250 panic("Cannot get vnode for swapdev"); 251 swapdev_vp->v_type = VNON; /* Untyped */ 252 vx_unlock(swapdev_vp); 253 } 254 255 ASSERT_VOP_UNLOCKED(vp, "swaponvp"); 256 for (sp = swdevt, index = 0 ; index < nswdev; index++, sp++) { 257 if (sp->sw_vp == vp) 258 return EBUSY; 259 if (!sp->sw_vp) 260 goto found; 261 262 } 263 return EINVAL; 264 found: 265 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 266 error = VOP_OPEN(vp, FREAD | FWRITE, cred, NULL, td); 267 VOP_UNLOCK(vp, 0, td); 268 if (error) 269 return (error); 270 271 /* 272 * v_rdev is not valid until after the VOP_OPEN() call. dev_psize() 273 * must be supported if a character device has been specified. 274 */ 275 if (vp->v_type == VCHR) 276 dev = vp->v_rdev; 277 else 278 dev = NODEV; 279 280 if (nblks == 0 && dev != NODEV && ((nblks = dev_dpsize(dev)) == -1)) { 281 VOP_CLOSE(vp, FREAD | FWRITE, td); 282 return (ENXIO); 283 } 284 if (nblks == 0) { 285 VOP_CLOSE(vp, FREAD | FWRITE, td); 286 return (ENXIO); 287 } 288 289 /* 290 * If we go beyond this, we get overflows in the radix 291 * tree bitmap code. 292 */ 293 if (nblks > 0x40000000 / BLIST_META_RADIX / nswdev) { 294 printf("exceeded maximum of %d blocks per swap unit\n", 295 0x40000000 / BLIST_META_RADIX / nswdev); 296 VOP_CLOSE(vp, FREAD | FWRITE, td); 297 return (ENXIO); 298 } 299 /* 300 * nblks is in DEV_BSIZE'd chunks, convert to PAGE_SIZE'd chunks. 301 * First chop nblks off to page-align it, then convert. 302 * 303 * sw->sw_nblks is in page-sized chunks now too. 304 */ 305 nblks &= ~(ctodb(1) - 1); 306 nblks = dbtoc(nblks); 307 308 sp->sw_vp = vp; 309 sp->sw_dev = dev2udev(dev); 310 sp->sw_device = dev; 311 sp->sw_flags |= SW_FREED; 312 sp->sw_nblks = nblks; 313 314 /* 315 * nblks, nswap, and dmmax are PAGE_SIZE'd parameters now, not 316 * DEV_BSIZE'd. aligned_nblks is used to calculate the 317 * size of the swap bitmap, taking into account the stripe size. 318 */ 319 aligned_nblks = (nblks + (dmmax - 1)) & ~(u_long)(dmmax - 1); 320 321 if (aligned_nblks * nswdev > nswap) 322 nswap = aligned_nblks * nswdev; 323 324 if (swapblist == NULL) 325 swapblist = blist_create(nswap); 326 else 327 blist_resize(&swapblist, nswap, 0); 328 329 for (dvbase = dmmax; dvbase < nblks; dvbase += dmmax) { 330 blk = min(nblks - dvbase, dmmax); 331 vsbase = index * dmmax + dvbase * nswdev; 332 blist_free(swapblist, vsbase, blk); 333 vm_swap_size += blk; 334 } 335 336 return (0); 337 } 338