| 1 | /* |
| 2 | * Copyright (c) 1990 University of Utah. |
| 3 | * Copyright (c) 1991 The Regents of the University of California. |
| 4 | * 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 | * 3. All advertising materials mentioning features or use of this software |
| 19 | * must display the following acknowledgement: |
| 20 | * This product includes software developed by the University of |
| 21 | * California, Berkeley and its contributors. |
| 22 | * 4. Neither the name of the University nor the names of its contributors |
| 23 | * may be used to endorse or promote products derived from this software |
| 24 | * without specific prior written permission. |
| 25 | * |
| 26 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 27 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 28 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 29 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 30 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 31 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 32 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 33 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 34 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 35 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 36 | * SUCH DAMAGE. |
| 37 | * |
| 38 | * @(#)vnode_pager.c 7.5 (Berkeley) 4/20/91 |
| 39 | */ |
| 40 | |
| 41 | /* |
| 42 | * Page to/from files (vnodes). |
| 43 | * |
| 44 | * TODO: |
| 45 | * pageouts |
| 46 | * fix credential use (uses current process credentials now) |
| 47 | */ |
| 48 | #include "vnodepager.h" |
| 49 | #if NVNODEPAGER > 0 |
| 50 | |
| 51 | #include "param.h" |
| 52 | #include "proc.h" |
| 53 | #include "malloc.h" |
| 54 | #include "vnode.h" |
| 55 | #include "uio.h" |
| 56 | #include "mount.h" |
| 57 | |
| 58 | #include "vm_param.h" |
| 59 | #include "lock.h" |
| 60 | #include "queue.h" |
| 61 | #include "vm_prot.h" |
| 62 | #include "vm_object.h" |
| 63 | #include "vm_page.h" |
| 64 | #include "vnode_pager.h" |
| 65 | |
| 66 | queue_head_t vnode_pager_list; /* list of managed vnodes */ |
| 67 | |
| 68 | #ifdef DEBUG |
| 69 | int vpagerdebug = 0x00; |
| 70 | #define VDB_FOLLOW 0x01 |
| 71 | #define VDB_INIT 0x02 |
| 72 | #define VDB_IO 0x04 |
| 73 | #define VDB_FAIL 0x08 |
| 74 | #define VDB_ALLOC 0x10 |
| 75 | #define VDB_SIZE 0x20 |
| 76 | #endif |
| 77 | |
| 78 | void |
| 79 | vnode_pager_init() |
| 80 | { |
| 81 | #ifdef DEBUG |
| 82 | if (vpagerdebug & VDB_FOLLOW) |
| 83 | printf("vnode_pager_init()\n"); |
| 84 | #endif |
| 85 | queue_init(&vnode_pager_list); |
| 86 | } |
| 87 | |
| 88 | /* |
| 89 | * Allocate (or lookup) pager for a vnode. |
| 90 | * Handle is a vnode pointer. |
| 91 | */ |
| 92 | vm_pager_t |
| 93 | vnode_pager_alloc(handle, size, prot) |
| 94 | caddr_t handle; |
| 95 | vm_size_t size; |
| 96 | vm_prot_t prot; |
| 97 | { |
| 98 | register vm_pager_t pager; |
| 99 | register vn_pager_t vnp; |
| 100 | vm_object_t object; |
| 101 | struct vattr vattr; |
| 102 | struct vnode *vp; |
| 103 | struct proc *p = curproc; /* XXX */ |
| 104 | |
| 105 | #ifdef DEBUG |
| 106 | if (vpagerdebug & (VDB_FOLLOW|VDB_ALLOC)) |
| 107 | printf("vnode_pager_alloc(%x, %x, %x)\n", handle, size, prot); |
| 108 | #endif |
| 109 | /* |
| 110 | * Pageout to vnode, no can do yet. |
| 111 | */ |
| 112 | if (handle == NULL) |
| 113 | return(NULL); |
| 114 | |
| 115 | /* |
| 116 | * Vnodes keep a pointer to any associated pager so no need to |
| 117 | * lookup with vm_pager_lookup. |
| 118 | */ |
| 119 | vp = (struct vnode *)handle; |
| 120 | pager = (vm_pager_t)vp->v_vmdata; |
| 121 | if (pager == NULL) { |
| 122 | /* |
| 123 | * Allocate pager structures |
| 124 | */ |
| 125 | pager = (vm_pager_t)malloc(sizeof *pager, M_VMPAGER, M_WAITOK); |
| 126 | if (pager == NULL) |
| 127 | return(NULL); |
| 128 | vnp = (vn_pager_t)malloc(sizeof *vnp, M_VMPGDATA, M_WAITOK); |
| 129 | if (vnp == NULL) { |
| 130 | free((caddr_t)pager, M_VMPAGER); |
| 131 | return(NULL); |
| 132 | } |
| 133 | /* |
| 134 | * And an object of the appropriate size |
| 135 | */ |
| 136 | if (VOP_GETATTR(vp, &vattr, p->p_ucred, p) == 0) { |
| 137 | object = vm_object_allocate(round_page(vattr.va_size)); |
| 138 | vm_object_enter(object, pager); |
| 139 | vm_object_setpager(object, pager, 0, TRUE); |
| 140 | } else { |
| 141 | free((caddr_t)vnp, M_VMPGDATA); |
| 142 | free((caddr_t)pager, M_VMPAGER); |
| 143 | return(NULL); |
| 144 | } |
| 145 | /* |
| 146 | * Hold a reference to the vnode and initialize pager data. |
| 147 | */ |
| 148 | VREF(vp); |
| 149 | vnp->vnp_flags = 0; |
| 150 | vnp->vnp_vp = vp; |
| 151 | vnp->vnp_size = vattr.va_size; |
| 152 | queue_enter(&vnode_pager_list, pager, vm_pager_t, pg_list); |
| 153 | pager->pg_handle = handle; |
| 154 | pager->pg_type = PG_VNODE; |
| 155 | pager->pg_ops = &vnodepagerops; |
| 156 | pager->pg_data = (caddr_t)vnp; |
| 157 | vp->v_vmdata = (caddr_t)pager; |
| 158 | } else { |
| 159 | /* |
| 160 | * vm_object_lookup() will remove the object from the |
| 161 | * cache if found and also gain a reference to the object. |
| 162 | */ |
| 163 | object = vm_object_lookup(pager); |
| 164 | #ifdef DEBUG |
| 165 | vnp = (vn_pager_t)pager->pg_data; |
| 166 | #endif |
| 167 | } |
| 168 | #ifdef DEBUG |
| 169 | if (vpagerdebug & VDB_ALLOC) |
| 170 | printf("vnode_pager_setup: vp %x sz %x pager %x object %x\n", |
| 171 | vp, vnp->vnp_size, pager, object); |
| 172 | #endif |
| 173 | return(pager); |
| 174 | } |
| 175 | |
| 176 | void |
| 177 | vnode_pager_dealloc(pager) |
| 178 | vm_pager_t pager; |
| 179 | { |
| 180 | register vn_pager_t vnp = (vn_pager_t)pager->pg_data; |
| 181 | register struct vnode *vp; |
| 182 | struct proc *p = curproc; /* XXX */ |
| 183 | |
| 184 | #ifdef DEBUG |
| 185 | if (vpagerdebug & VDB_FOLLOW) |
| 186 | printf("vnode_pager_dealloc(%x)\n", pager); |
| 187 | #endif |
| 188 | if (vp = vnp->vnp_vp) { |
| 189 | vp->v_vmdata = NULL; |
| 190 | vp->v_flag &= ~VTEXT; |
| 191 | #if 0 |
| 192 | /* can hang if done at reboot on NFS FS */ |
| 193 | (void) VOP_FSYNC(vp, p->p_ucred, p); |
| 194 | #endif |
| 195 | vrele(vp); |
| 196 | } |
| 197 | queue_remove(&vnode_pager_list, pager, vm_pager_t, pg_list); |
| 198 | free((caddr_t)vnp, M_VMPGDATA); |
| 199 | free((caddr_t)pager, M_VMPAGER); |
| 200 | } |
| 201 | |
| 202 | vnode_pager_getpage(pager, m, sync) |
| 203 | vm_pager_t pager; |
| 204 | vm_page_t m; |
| 205 | boolean_t sync; |
| 206 | { |
| 207 | |
| 208 | #ifdef DEBUG |
| 209 | if (vpagerdebug & VDB_FOLLOW) |
| 210 | printf("vnode_pager_getpage(%x, %x)\n", pager, m); |
| 211 | #endif |
| 212 | return(vnode_pager_io((vn_pager_t)pager->pg_data, m, UIO_READ)); |
| 213 | } |
| 214 | |
| 215 | boolean_t |
| 216 | vnode_pager_putpage(pager, m, sync) |
| 217 | vm_pager_t pager; |
| 218 | vm_page_t m; |
| 219 | boolean_t sync; |
| 220 | { |
| 221 | int err; |
| 222 | |
| 223 | #ifdef DEBUG |
| 224 | if (vpagerdebug & VDB_FOLLOW) |
| 225 | printf("vnode_pager_putpage(%x, %x)\n", pager, m); |
| 226 | #endif |
| 227 | if (pager == NULL) |
| 228 | return; |
| 229 | err = vnode_pager_io((vn_pager_t)pager->pg_data, m, UIO_WRITE); |
| 230 | if (err == VM_PAGER_OK) { |
| 231 | m->clean = TRUE; /* XXX - wrong place */ |
| 232 | pmap_clear_modify(VM_PAGE_TO_PHYS(m)); /* XXX - wrong place */ |
| 233 | } |
| 234 | return(err); |
| 235 | } |
| 236 | |
| 237 | boolean_t |
| 238 | vnode_pager_haspage(pager, offset) |
| 239 | vm_pager_t pager; |
| 240 | vm_offset_t offset; |
| 241 | { |
| 242 | register vn_pager_t vnp = (vn_pager_t)pager->pg_data; |
| 243 | daddr_t bn; |
| 244 | int err; |
| 245 | |
| 246 | #ifdef DEBUG |
| 247 | if (vpagerdebug & VDB_FOLLOW) |
| 248 | printf("vnode_pager_haspage(%x, %x)\n", pager, offset); |
| 249 | #endif |
| 250 | |
| 251 | /* |
| 252 | * Offset beyond end of file, do not have the page |
| 253 | */ |
| 254 | if (offset >= vnp->vnp_size) { |
| 255 | #ifdef DEBUG |
| 256 | if (vpagerdebug & (VDB_FAIL|VDB_SIZE)) |
| 257 | printf("vnode_pager_haspage: pg %x, off %x, size %x\n", |
| 258 | pager, offset, vnp->vnp_size); |
| 259 | #endif |
| 260 | return(FALSE); |
| 261 | } |
| 262 | |
| 263 | /* |
| 264 | * Read the index to find the disk block to read |
| 265 | * from. If there is no block, report that we don't |
| 266 | * have this data. |
| 267 | * |
| 268 | * Assumes that the vnode has whole page or nothing. |
| 269 | */ |
| 270 | err = VOP_BMAP(vnp->vnp_vp, |
| 271 | offset / vnp->vnp_vp->v_mount->mnt_stat.f_bsize, |
| 272 | (struct vnode **)0, &bn); |
| 273 | if (err) { |
| 274 | #ifdef DEBUG |
| 275 | if (vpagerdebug & VDB_FAIL) |
| 276 | printf("vnode_pager_haspage: BMAP err %d, pg %x, off %x\n", |
| 277 | err, pager, offset); |
| 278 | #endif |
| 279 | return(TRUE); |
| 280 | } |
| 281 | return((long)bn < 0 ? FALSE : TRUE); |
| 282 | } |
| 283 | |
| 284 | /* |
| 285 | * (XXX) |
| 286 | * Lets the VM system know about a change in size for a file. |
| 287 | * If this vnode is mapped into some address space (i.e. we have a pager |
| 288 | * for it) we adjust our own internal size and flush any cached pages in |
| 289 | * the associated object that are affected by the size change. |
| 290 | * |
| 291 | * Note: this routine may be invoked as a result of a pager put |
| 292 | * operation (possibly at object termination time), so we must be careful. |
| 293 | */ |
| 294 | vnode_pager_setsize(vp, nsize) |
| 295 | struct vnode *vp; |
| 296 | u_long nsize; |
| 297 | { |
| 298 | register vn_pager_t vnp; |
| 299 | register vm_object_t object; |
| 300 | vm_pager_t pager; |
| 301 | |
| 302 | /* |
| 303 | * Not a mapped vnode |
| 304 | */ |
| 305 | if (vp == NULL || vp->v_type != VREG || vp->v_vmdata == NULL) |
| 306 | return; |
| 307 | /* |
| 308 | * Hasn't changed size |
| 309 | */ |
| 310 | pager = (vm_pager_t)vp->v_vmdata; |
| 311 | vnp = (vn_pager_t)pager->pg_data; |
| 312 | if (nsize == vnp->vnp_size) |
| 313 | return; |
| 314 | /* |
| 315 | * No object. |
| 316 | * This can happen during object termination since |
| 317 | * vm_object_page_clean is called after the object |
| 318 | * has been removed from the hash table, and clean |
| 319 | * may cause vnode write operations which can wind |
| 320 | * up back here. |
| 321 | */ |
| 322 | object = vm_object_lookup(pager); |
| 323 | if (object == NULL) |
| 324 | return; |
| 325 | |
| 326 | #ifdef DEBUG |
| 327 | if (vpagerdebug & (VDB_FOLLOW|VDB_SIZE)) |
| 328 | printf("vnode_pager_setsize: vp %x obj %x osz %d nsz %d\n", |
| 329 | vp, object, vnp->vnp_size, nsize); |
| 330 | #endif |
| 331 | /* |
| 332 | * File has shrunk. |
| 333 | * Toss any cached pages beyond the new EOF. |
| 334 | */ |
| 335 | if (nsize < vnp->vnp_size) { |
| 336 | vm_object_lock(object); |
| 337 | vm_object_page_remove(object, |
| 338 | (vm_offset_t)nsize, vnp->vnp_size); |
| 339 | vm_object_unlock(object); |
| 340 | } |
| 341 | vnp->vnp_size = (vm_offset_t)nsize; |
| 342 | vm_object_deallocate(object); |
| 343 | } |
| 344 | |
| 345 | vnode_pager_umount(mp) |
| 346 | register struct mount *mp; |
| 347 | { |
| 348 | register vm_pager_t pager, npager; |
| 349 | struct vnode *vp; |
| 350 | |
| 351 | pager = (vm_pager_t) queue_first(&vnode_pager_list); |
| 352 | while (!queue_end(&vnode_pager_list, (queue_entry_t)pager)) { |
| 353 | /* |
| 354 | * Save the next pointer now since uncaching may |
| 355 | * terminate the object and render pager invalid |
| 356 | */ |
| 357 | vp = ((vn_pager_t)pager->pg_data)->vnp_vp; |
| 358 | npager = (vm_pager_t) queue_next(&pager->pg_list); |
| 359 | if (mp == (struct mount *)0 || vp->v_mount == mp) |
| 360 | (void) vnode_pager_uncache(vp); |
| 361 | pager = npager; |
| 362 | } |
| 363 | } |
| 364 | |
| 365 | /* |
| 366 | * Remove vnode associated object from the object cache. |
| 367 | * |
| 368 | * Note: this routine may be invoked as a result of a pager put |
| 369 | * operation (possibly at object termination time), so we must be careful. |
| 370 | */ |
| 371 | boolean_t |
| 372 | vnode_pager_uncache(vp) |
| 373 | register struct vnode *vp; |
| 374 | { |
| 375 | register vm_object_t object; |
| 376 | boolean_t uncached, locked; |
| 377 | vm_pager_t pager; |
| 378 | |
| 379 | /* |
| 380 | * Not a mapped vnode |
| 381 | */ |
| 382 | pager = (vm_pager_t)vp->v_vmdata; |
| 383 | if (pager == NULL) |
| 384 | return (TRUE); |
| 385 | /* |
| 386 | * Unlock the vnode if it is currently locked. |
| 387 | * We do this since uncaching the object may result |
| 388 | * in its destruction which may initiate paging |
| 389 | * activity which may necessitate locking the vnode. |
| 390 | */ |
| 391 | locked = VOP_ISLOCKED(vp); |
| 392 | if (locked) |
| 393 | VOP_UNLOCK(vp); |
| 394 | /* |
| 395 | * Must use vm_object_lookup() as it actually removes |
| 396 | * the object from the cache list. |
| 397 | */ |
| 398 | object = vm_object_lookup(pager); |
| 399 | if (object) { |
| 400 | uncached = (object->ref_count <= 1); |
| 401 | pager_cache(object, FALSE); |
| 402 | } else |
| 403 | uncached = TRUE; |
| 404 | if (locked) |
| 405 | VOP_LOCK(vp); |
| 406 | return(uncached); |
| 407 | } |
| 408 | |
| 409 | vnode_pager_io(vnp, m, rw) |
| 410 | register vn_pager_t vnp; |
| 411 | vm_page_t m; |
| 412 | enum uio_rw rw; |
| 413 | { |
| 414 | struct uio auio; |
| 415 | struct iovec aiov; |
| 416 | vm_offset_t kva, foff; |
| 417 | int error, size; |
| 418 | struct proc *p = curproc; /* XXX */ |
| 419 | |
| 420 | #ifdef DEBUG |
| 421 | if (vpagerdebug & VDB_FOLLOW) |
| 422 | printf("vnode_pager_io(%x, %x, %c): vnode %x\n", |
| 423 | vnp, m, rw == UIO_READ ? 'R' : 'W', vnp->vnp_vp); |
| 424 | #endif |
| 425 | foff = m->offset + m->object->paging_offset; |
| 426 | /* |
| 427 | * Return failure if beyond current EOF |
| 428 | */ |
| 429 | if (foff >= vnp->vnp_size) { |
| 430 | #ifdef DEBUG |
| 431 | if (vpagerdebug & VDB_SIZE) |
| 432 | printf("vnode_pager_io: vp %x, off %d size %d\n", |
| 433 | vnp->vnp_vp, foff, vnp->vnp_size); |
| 434 | #endif |
| 435 | return(VM_PAGER_BAD); |
| 436 | } |
| 437 | if (foff + PAGE_SIZE > vnp->vnp_size) |
| 438 | size = vnp->vnp_size - foff; |
| 439 | else |
| 440 | size = PAGE_SIZE; |
| 441 | /* |
| 442 | * Allocate a kernel virtual address and initialize so that |
| 443 | * we can use VOP_READ/WRITE routines. |
| 444 | */ |
| 445 | kva = vm_pager_map_page(m); |
| 446 | aiov.iov_base = (caddr_t)kva; |
| 447 | aiov.iov_len = size; |
| 448 | auio.uio_iov = &aiov; |
| 449 | auio.uio_iovcnt = 1; |
| 450 | auio.uio_offset = foff; |
| 451 | auio.uio_segflg = UIO_SYSSPACE; |
| 452 | auio.uio_rw = rw; |
| 453 | auio.uio_resid = size; |
| 454 | auio.uio_procp = (struct proc *)0; |
| 455 | #ifdef DEBUG |
| 456 | if (vpagerdebug & VDB_IO) |
| 457 | printf("vnode_pager_io: vp %x kva %x foff %x size %x", |
| 458 | vnp->vnp_vp, kva, foff, size); |
| 459 | #endif |
| 460 | if (rw == UIO_READ) |
| 461 | error = VOP_READ(vnp->vnp_vp, &auio, 0, p->p_ucred); |
| 462 | else |
| 463 | error = VOP_WRITE(vnp->vnp_vp, &auio, 0, p->p_ucred); |
| 464 | #ifdef DEBUG |
| 465 | if (vpagerdebug & VDB_IO) { |
| 466 | if (error || auio.uio_resid) |
| 467 | printf(" returns error %x, resid %x", |
| 468 | error, auio.uio_resid); |
| 469 | printf("\n"); |
| 470 | } |
| 471 | #endif |
| 472 | if (!error) { |
| 473 | register int count = size - auio.uio_resid; |
| 474 | |
| 475 | if (count == 0) |
| 476 | error = EINVAL; |
| 477 | else if (count != PAGE_SIZE && rw == UIO_READ) |
| 478 | bzero(kva + count, PAGE_SIZE - count); |
| 479 | } |
| 480 | vm_pager_unmap_page(kva); |
| 481 | return (error ? VM_PAGER_FAIL : VM_PAGER_OK); |
| 482 | } |
| 483 | #endif |