| 1 | /* $NetBSD: uvm_aobj.c,v 1.124 2016/07/28 07:52:06 martin Exp $ */ |
| 2 | |
| 3 | /* |
| 4 | * Copyright (c) 1998 Chuck Silvers, Charles D. Cranor and |
| 5 | * Washington University. |
| 6 | * All rights reserved. |
| 7 | * |
| 8 | * Redistribution and use in source and binary forms, with or without |
| 9 | * modification, are permitted provided that the following conditions |
| 10 | * are met: |
| 11 | * 1. Redistributions of source code must retain the above copyright |
| 12 | * notice, this list of conditions and the following disclaimer. |
| 13 | * 2. Redistributions in binary form must reproduce the above copyright |
| 14 | * notice, this list of conditions and the following disclaimer in the |
| 15 | * documentation and/or other materials provided with the distribution. |
| 16 | * |
| 17 | * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
| 18 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
| 19 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
| 20 | * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 21 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
| 22 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 23 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 24 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 25 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
| 26 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 27 | * |
| 28 | * from: Id: uvm_aobj.c,v 1.1.2.5 1998/02/06 05:14:38 chs Exp |
| 29 | */ |
| 30 | |
| 31 | /* |
| 32 | * uvm_aobj.c: anonymous memory uvm_object pager |
| 33 | * |
| 34 | * author: Chuck Silvers <chuq@chuq.com> |
| 35 | * started: Jan-1998 |
| 36 | * |
| 37 | * - design mostly from Chuck Cranor |
| 38 | */ |
| 39 | |
| 40 | #include <sys/cdefs.h> |
| 41 | __KERNEL_RCSID(0, "$NetBSD: uvm_aobj.c,v 1.124 2016/07/28 07:52:06 martin Exp $" ); |
| 42 | |
| 43 | #ifdef _KERNEL_OPT |
| 44 | #include "opt_uvmhist.h" |
| 45 | #endif |
| 46 | |
| 47 | #include <sys/param.h> |
| 48 | #include <sys/systm.h> |
| 49 | #include <sys/kernel.h> |
| 50 | #include <sys/kmem.h> |
| 51 | #include <sys/pool.h> |
| 52 | #include <sys/atomic.h> |
| 53 | |
| 54 | #include <uvm/uvm.h> |
| 55 | |
| 56 | /* |
| 57 | * An anonymous UVM object (aobj) manages anonymous-memory. In addition to |
| 58 | * keeping the list of resident pages, it may also keep a list of allocated |
| 59 | * swap blocks. Depending on the size of the object, this list is either |
| 60 | * stored in an array (small objects) or in a hash table (large objects). |
| 61 | * |
| 62 | * Lock order |
| 63 | * |
| 64 | * uao_list_lock -> |
| 65 | * uvm_object::vmobjlock |
| 66 | */ |
| 67 | |
| 68 | /* |
| 69 | * Note: for hash tables, we break the address space of the aobj into blocks |
| 70 | * of UAO_SWHASH_CLUSTER_SIZE pages, which shall be a power of two. |
| 71 | */ |
| 72 | |
| 73 | #define UAO_SWHASH_CLUSTER_SHIFT 4 |
| 74 | #define UAO_SWHASH_CLUSTER_SIZE (1 << UAO_SWHASH_CLUSTER_SHIFT) |
| 75 | |
| 76 | /* Get the "tag" for this page index. */ |
| 77 | #define UAO_SWHASH_ELT_TAG(idx) ((idx) >> UAO_SWHASH_CLUSTER_SHIFT) |
| 78 | #define UAO_SWHASH_ELT_PAGESLOT_IDX(idx) \ |
| 79 | ((idx) & (UAO_SWHASH_CLUSTER_SIZE - 1)) |
| 80 | |
| 81 | /* Given an ELT and a page index, find the swap slot. */ |
| 82 | #define UAO_SWHASH_ELT_PAGESLOT(elt, idx) \ |
| 83 | ((elt)->slots[UAO_SWHASH_ELT_PAGESLOT_IDX(idx)]) |
| 84 | |
| 85 | /* Given an ELT, return its pageidx base. */ |
| 86 | #define UAO_SWHASH_ELT_PAGEIDX_BASE(ELT) \ |
| 87 | ((elt)->tag << UAO_SWHASH_CLUSTER_SHIFT) |
| 88 | |
| 89 | /* The hash function. */ |
| 90 | #define UAO_SWHASH_HASH(aobj, idx) \ |
| 91 | (&(aobj)->u_swhash[(((idx) >> UAO_SWHASH_CLUSTER_SHIFT) \ |
| 92 | & (aobj)->u_swhashmask)]) |
| 93 | |
| 94 | /* |
| 95 | * The threshold which determines whether we will use an array or a |
| 96 | * hash table to store the list of allocated swap blocks. |
| 97 | */ |
| 98 | #define UAO_SWHASH_THRESHOLD (UAO_SWHASH_CLUSTER_SIZE * 4) |
| 99 | #define UAO_USES_SWHASH(aobj) \ |
| 100 | ((aobj)->u_pages > UAO_SWHASH_THRESHOLD) |
| 101 | |
| 102 | /* The number of buckets in a hash, with an upper bound. */ |
| 103 | #define UAO_SWHASH_MAXBUCKETS 256 |
| 104 | #define UAO_SWHASH_BUCKETS(aobj) \ |
| 105 | (MIN((aobj)->u_pages >> UAO_SWHASH_CLUSTER_SHIFT, UAO_SWHASH_MAXBUCKETS)) |
| 106 | |
| 107 | /* |
| 108 | * uao_swhash_elt: when a hash table is being used, this structure defines |
| 109 | * the format of an entry in the bucket list. |
| 110 | */ |
| 111 | |
| 112 | struct uao_swhash_elt { |
| 113 | LIST_ENTRY(uao_swhash_elt) list; /* the hash list */ |
| 114 | voff_t tag; /* our 'tag' */ |
| 115 | int count; /* our number of active slots */ |
| 116 | int slots[UAO_SWHASH_CLUSTER_SIZE]; /* the slots */ |
| 117 | }; |
| 118 | |
| 119 | /* |
| 120 | * uao_swhash: the swap hash table structure |
| 121 | */ |
| 122 | |
| 123 | LIST_HEAD(uao_swhash, uao_swhash_elt); |
| 124 | |
| 125 | /* |
| 126 | * uao_swhash_elt_pool: pool of uao_swhash_elt structures. |
| 127 | * Note: pages for this pool must not come from a pageable kernel map. |
| 128 | */ |
| 129 | static struct pool uao_swhash_elt_pool __cacheline_aligned; |
| 130 | |
| 131 | /* |
| 132 | * uvm_aobj: the actual anon-backed uvm_object |
| 133 | * |
| 134 | * => the uvm_object is at the top of the structure, this allows |
| 135 | * (struct uvm_aobj *) == (struct uvm_object *) |
| 136 | * => only one of u_swslots and u_swhash is used in any given aobj |
| 137 | */ |
| 138 | |
| 139 | struct uvm_aobj { |
| 140 | struct uvm_object u_obj; /* has: lock, pgops, memq, #pages, #refs */ |
| 141 | pgoff_t u_pages; /* number of pages in entire object */ |
| 142 | int u_flags; /* the flags (see uvm_aobj.h) */ |
| 143 | int *u_swslots; /* array of offset->swapslot mappings */ |
| 144 | /* |
| 145 | * hashtable of offset->swapslot mappings |
| 146 | * (u_swhash is an array of bucket heads) |
| 147 | */ |
| 148 | struct uao_swhash *u_swhash; |
| 149 | u_long u_swhashmask; /* mask for hashtable */ |
| 150 | LIST_ENTRY(uvm_aobj) u_list; /* global list of aobjs */ |
| 151 | int u_freelist; /* freelist to allocate pages from */ |
| 152 | }; |
| 153 | |
| 154 | static void uao_free(struct uvm_aobj *); |
| 155 | static int uao_get(struct uvm_object *, voff_t, struct vm_page **, |
| 156 | int *, int, vm_prot_t, int, int); |
| 157 | static int uao_put(struct uvm_object *, voff_t, voff_t, int); |
| 158 | |
| 159 | #if defined(VMSWAP) |
| 160 | static struct uao_swhash_elt *uao_find_swhash_elt |
| 161 | (struct uvm_aobj *, int, bool); |
| 162 | |
| 163 | static bool uao_pagein(struct uvm_aobj *, int, int); |
| 164 | static bool uao_pagein_page(struct uvm_aobj *, int); |
| 165 | #endif /* defined(VMSWAP) */ |
| 166 | |
| 167 | static struct vm_page *uao_pagealloc(struct uvm_object *, voff_t, int); |
| 168 | |
| 169 | /* |
| 170 | * aobj_pager |
| 171 | * |
| 172 | * note that some functions (e.g. put) are handled elsewhere |
| 173 | */ |
| 174 | |
| 175 | const struct uvm_pagerops = { |
| 176 | .pgo_reference = uao_reference, |
| 177 | .pgo_detach = uao_detach, |
| 178 | .pgo_get = uao_get, |
| 179 | .pgo_put = uao_put, |
| 180 | }; |
| 181 | |
| 182 | /* |
| 183 | * uao_list: global list of active aobjs, locked by uao_list_lock |
| 184 | */ |
| 185 | |
| 186 | static LIST_HEAD(aobjlist, uvm_aobj) uao_list __cacheline_aligned; |
| 187 | static kmutex_t uao_list_lock __cacheline_aligned; |
| 188 | |
| 189 | /* |
| 190 | * hash table/array related functions |
| 191 | */ |
| 192 | |
| 193 | #if defined(VMSWAP) |
| 194 | |
| 195 | /* |
| 196 | * uao_find_swhash_elt: find (or create) a hash table entry for a page |
| 197 | * offset. |
| 198 | * |
| 199 | * => the object should be locked by the caller |
| 200 | */ |
| 201 | |
| 202 | static struct uao_swhash_elt * |
| 203 | uao_find_swhash_elt(struct uvm_aobj *aobj, int pageidx, bool create) |
| 204 | { |
| 205 | struct uao_swhash *swhash; |
| 206 | struct uao_swhash_elt *elt; |
| 207 | voff_t page_tag; |
| 208 | |
| 209 | swhash = UAO_SWHASH_HASH(aobj, pageidx); |
| 210 | page_tag = UAO_SWHASH_ELT_TAG(pageidx); |
| 211 | |
| 212 | /* |
| 213 | * now search the bucket for the requested tag |
| 214 | */ |
| 215 | |
| 216 | LIST_FOREACH(elt, swhash, list) { |
| 217 | if (elt->tag == page_tag) { |
| 218 | return elt; |
| 219 | } |
| 220 | } |
| 221 | if (!create) { |
| 222 | return NULL; |
| 223 | } |
| 224 | |
| 225 | /* |
| 226 | * allocate a new entry for the bucket and init/insert it in |
| 227 | */ |
| 228 | |
| 229 | elt = pool_get(&uao_swhash_elt_pool, PR_NOWAIT); |
| 230 | if (elt == NULL) { |
| 231 | return NULL; |
| 232 | } |
| 233 | LIST_INSERT_HEAD(swhash, elt, list); |
| 234 | elt->tag = page_tag; |
| 235 | elt->count = 0; |
| 236 | memset(elt->slots, 0, sizeof(elt->slots)); |
| 237 | return elt; |
| 238 | } |
| 239 | |
| 240 | /* |
| 241 | * uao_find_swslot: find the swap slot number for an aobj/pageidx |
| 242 | * |
| 243 | * => object must be locked by caller |
| 244 | */ |
| 245 | |
| 246 | int |
| 247 | uao_find_swslot(struct uvm_object *uobj, int pageidx) |
| 248 | { |
| 249 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 250 | struct uao_swhash_elt *elt; |
| 251 | |
| 252 | /* |
| 253 | * if noswap flag is set, then we never return a slot |
| 254 | */ |
| 255 | |
| 256 | if (aobj->u_flags & UAO_FLAG_NOSWAP) |
| 257 | return 0; |
| 258 | |
| 259 | /* |
| 260 | * if hashing, look in hash table. |
| 261 | */ |
| 262 | |
| 263 | if (UAO_USES_SWHASH(aobj)) { |
| 264 | elt = uao_find_swhash_elt(aobj, pageidx, false); |
| 265 | return elt ? UAO_SWHASH_ELT_PAGESLOT(elt, pageidx) : 0; |
| 266 | } |
| 267 | |
| 268 | /* |
| 269 | * otherwise, look in the array |
| 270 | */ |
| 271 | |
| 272 | return aobj->u_swslots[pageidx]; |
| 273 | } |
| 274 | |
| 275 | /* |
| 276 | * uao_set_swslot: set the swap slot for a page in an aobj. |
| 277 | * |
| 278 | * => setting a slot to zero frees the slot |
| 279 | * => object must be locked by caller |
| 280 | * => we return the old slot number, or -1 if we failed to allocate |
| 281 | * memory to record the new slot number |
| 282 | */ |
| 283 | |
| 284 | int |
| 285 | uao_set_swslot(struct uvm_object *uobj, int pageidx, int slot) |
| 286 | { |
| 287 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 288 | struct uao_swhash_elt *elt; |
| 289 | int oldslot; |
| 290 | UVMHIST_FUNC("uao_set_swslot" ); UVMHIST_CALLED(pdhist); |
| 291 | UVMHIST_LOG(pdhist, "aobj %p pageidx %d slot %d" , |
| 292 | aobj, pageidx, slot, 0); |
| 293 | |
| 294 | KASSERT(mutex_owned(uobj->vmobjlock) || uobj->uo_refs == 0); |
| 295 | |
| 296 | /* |
| 297 | * if noswap flag is set, then we can't set a non-zero slot. |
| 298 | */ |
| 299 | |
| 300 | if (aobj->u_flags & UAO_FLAG_NOSWAP) { |
| 301 | KASSERTMSG(slot == 0, "uao_set_swslot: no swap object" ); |
| 302 | return 0; |
| 303 | } |
| 304 | |
| 305 | /* |
| 306 | * are we using a hash table? if so, add it in the hash. |
| 307 | */ |
| 308 | |
| 309 | if (UAO_USES_SWHASH(aobj)) { |
| 310 | |
| 311 | /* |
| 312 | * Avoid allocating an entry just to free it again if |
| 313 | * the page had not swap slot in the first place, and |
| 314 | * we are freeing. |
| 315 | */ |
| 316 | |
| 317 | elt = uao_find_swhash_elt(aobj, pageidx, slot != 0); |
| 318 | if (elt == NULL) { |
| 319 | return slot ? -1 : 0; |
| 320 | } |
| 321 | |
| 322 | oldslot = UAO_SWHASH_ELT_PAGESLOT(elt, pageidx); |
| 323 | UAO_SWHASH_ELT_PAGESLOT(elt, pageidx) = slot; |
| 324 | |
| 325 | /* |
| 326 | * now adjust the elt's reference counter and free it if we've |
| 327 | * dropped it to zero. |
| 328 | */ |
| 329 | |
| 330 | if (slot) { |
| 331 | if (oldslot == 0) |
| 332 | elt->count++; |
| 333 | } else { |
| 334 | if (oldslot) |
| 335 | elt->count--; |
| 336 | |
| 337 | if (elt->count == 0) { |
| 338 | LIST_REMOVE(elt, list); |
| 339 | pool_put(&uao_swhash_elt_pool, elt); |
| 340 | } |
| 341 | } |
| 342 | } else { |
| 343 | /* we are using an array */ |
| 344 | oldslot = aobj->u_swslots[pageidx]; |
| 345 | aobj->u_swslots[pageidx] = slot; |
| 346 | } |
| 347 | return oldslot; |
| 348 | } |
| 349 | |
| 350 | #endif /* defined(VMSWAP) */ |
| 351 | |
| 352 | /* |
| 353 | * end of hash/array functions |
| 354 | */ |
| 355 | |
| 356 | /* |
| 357 | * uao_free: free all resources held by an aobj, and then free the aobj |
| 358 | * |
| 359 | * => the aobj should be dead |
| 360 | */ |
| 361 | |
| 362 | static void |
| 363 | uao_free(struct uvm_aobj *aobj) |
| 364 | { |
| 365 | struct uvm_object *uobj = &aobj->u_obj; |
| 366 | |
| 367 | KASSERT(mutex_owned(uobj->vmobjlock)); |
| 368 | uao_dropswap_range(uobj, 0, 0); |
| 369 | mutex_exit(uobj->vmobjlock); |
| 370 | |
| 371 | #if defined(VMSWAP) |
| 372 | if (UAO_USES_SWHASH(aobj)) { |
| 373 | |
| 374 | /* |
| 375 | * free the hash table itself. |
| 376 | */ |
| 377 | |
| 378 | hashdone(aobj->u_swhash, HASH_LIST, aobj->u_swhashmask); |
| 379 | } else { |
| 380 | |
| 381 | /* |
| 382 | * free the array itsself. |
| 383 | */ |
| 384 | |
| 385 | kmem_free(aobj->u_swslots, aobj->u_pages * sizeof(int)); |
| 386 | } |
| 387 | #endif /* defined(VMSWAP) */ |
| 388 | |
| 389 | /* |
| 390 | * finally free the aobj itself |
| 391 | */ |
| 392 | |
| 393 | uvm_obj_destroy(uobj, true); |
| 394 | kmem_free(aobj, sizeof(struct uvm_aobj)); |
| 395 | } |
| 396 | |
| 397 | /* |
| 398 | * pager functions |
| 399 | */ |
| 400 | |
| 401 | /* |
| 402 | * uao_create: create an aobj of the given size and return its uvm_object. |
| 403 | * |
| 404 | * => for normal use, flags are always zero |
| 405 | * => for the kernel object, the flags are: |
| 406 | * UAO_FLAG_KERNOBJ - allocate the kernel object (can only happen once) |
| 407 | * UAO_FLAG_KERNSWAP - enable swapping of kernel object (" ") |
| 408 | */ |
| 409 | |
| 410 | struct uvm_object * |
| 411 | uao_create(vsize_t size, int flags) |
| 412 | { |
| 413 | static struct uvm_aobj kernel_object_store; |
| 414 | static kmutex_t kernel_object_lock; |
| 415 | static int kobj_alloced __diagused = 0; |
| 416 | pgoff_t pages = round_page(size) >> PAGE_SHIFT; |
| 417 | struct uvm_aobj *aobj; |
| 418 | int refs; |
| 419 | |
| 420 | /* |
| 421 | * Allocate a new aobj, unless kernel object is requested. |
| 422 | */ |
| 423 | |
| 424 | if (flags & UAO_FLAG_KERNOBJ) { |
| 425 | KASSERT(!kobj_alloced); |
| 426 | aobj = &kernel_object_store; |
| 427 | aobj->u_pages = pages; |
| 428 | aobj->u_flags = UAO_FLAG_NOSWAP; |
| 429 | refs = UVM_OBJ_KERN; |
| 430 | kobj_alloced = UAO_FLAG_KERNOBJ; |
| 431 | } else if (flags & UAO_FLAG_KERNSWAP) { |
| 432 | KASSERT(kobj_alloced == UAO_FLAG_KERNOBJ); |
| 433 | aobj = &kernel_object_store; |
| 434 | kobj_alloced = UAO_FLAG_KERNSWAP; |
| 435 | refs = 0xdeadbeaf; /* XXX: gcc */ |
| 436 | } else { |
| 437 | aobj = kmem_alloc(sizeof(struct uvm_aobj), KM_SLEEP); |
| 438 | aobj->u_pages = pages; |
| 439 | aobj->u_flags = 0; |
| 440 | refs = 1; |
| 441 | } |
| 442 | |
| 443 | /* |
| 444 | * no freelist by default |
| 445 | */ |
| 446 | |
| 447 | aobj->u_freelist = VM_NFREELIST; |
| 448 | |
| 449 | /* |
| 450 | * allocate hash/array if necessary |
| 451 | * |
| 452 | * note: in the KERNSWAP case no need to worry about locking since |
| 453 | * we are still booting we should be the only thread around. |
| 454 | */ |
| 455 | |
| 456 | if (flags == 0 || (flags & UAO_FLAG_KERNSWAP) != 0) { |
| 457 | #if defined(VMSWAP) |
| 458 | const int kernswap = (flags & UAO_FLAG_KERNSWAP) != 0; |
| 459 | |
| 460 | /* allocate hash table or array depending on object size */ |
| 461 | if (UAO_USES_SWHASH(aobj)) { |
| 462 | aobj->u_swhash = hashinit(UAO_SWHASH_BUCKETS(aobj), |
| 463 | HASH_LIST, kernswap ? false : true, |
| 464 | &aobj->u_swhashmask); |
| 465 | if (aobj->u_swhash == NULL) |
| 466 | panic("uao_create: hashinit swhash failed" ); |
| 467 | } else { |
| 468 | aobj->u_swslots = kmem_zalloc(pages * sizeof(int), |
| 469 | kernswap ? KM_NOSLEEP : KM_SLEEP); |
| 470 | if (aobj->u_swslots == NULL) |
| 471 | panic("uao_create: swslots allocation failed" ); |
| 472 | } |
| 473 | #endif /* defined(VMSWAP) */ |
| 474 | |
| 475 | if (flags) { |
| 476 | aobj->u_flags &= ~UAO_FLAG_NOSWAP; /* clear noswap */ |
| 477 | return &aobj->u_obj; |
| 478 | } |
| 479 | } |
| 480 | |
| 481 | /* |
| 482 | * Initialise UVM object. |
| 483 | */ |
| 484 | |
| 485 | const bool kernobj = (flags & UAO_FLAG_KERNOBJ) != 0; |
| 486 | uvm_obj_init(&aobj->u_obj, &aobj_pager, !kernobj, refs); |
| 487 | if (__predict_false(kernobj)) { |
| 488 | /* Initialisation only once, for UAO_FLAG_KERNOBJ. */ |
| 489 | mutex_init(&kernel_object_lock, MUTEX_DEFAULT, IPL_NONE); |
| 490 | uvm_obj_setlock(&aobj->u_obj, &kernel_object_lock); |
| 491 | } |
| 492 | |
| 493 | /* |
| 494 | * now that aobj is ready, add it to the global list |
| 495 | */ |
| 496 | |
| 497 | mutex_enter(&uao_list_lock); |
| 498 | LIST_INSERT_HEAD(&uao_list, aobj, u_list); |
| 499 | mutex_exit(&uao_list_lock); |
| 500 | return(&aobj->u_obj); |
| 501 | } |
| 502 | |
| 503 | /* |
| 504 | * uao_set_pgfl: allocate pages only from the specified freelist. |
| 505 | * |
| 506 | * => must be called before any pages are allocated for the object. |
| 507 | * => reset by setting it to VM_NFREELIST, meaning any freelist. |
| 508 | */ |
| 509 | |
| 510 | void |
| 511 | uao_set_pgfl(struct uvm_object *uobj, int freelist) |
| 512 | { |
| 513 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 514 | |
| 515 | KASSERTMSG((0 <= freelist), "invalid freelist %d" , freelist); |
| 516 | KASSERTMSG((freelist <= VM_NFREELIST), "invalid freelist %d" , |
| 517 | freelist); |
| 518 | |
| 519 | aobj->u_freelist = freelist; |
| 520 | } |
| 521 | |
| 522 | /* |
| 523 | * uao_pagealloc: allocate a page for aobj. |
| 524 | */ |
| 525 | |
| 526 | static inline struct vm_page * |
| 527 | uao_pagealloc(struct uvm_object *uobj, voff_t offset, int flags) |
| 528 | { |
| 529 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 530 | |
| 531 | if (__predict_true(aobj->u_freelist == VM_NFREELIST)) |
| 532 | return uvm_pagealloc(uobj, offset, NULL, flags); |
| 533 | else |
| 534 | return uvm_pagealloc_strat(uobj, offset, NULL, flags, |
| 535 | UVM_PGA_STRAT_ONLY, aobj->u_freelist); |
| 536 | } |
| 537 | |
| 538 | /* |
| 539 | * uao_init: set up aobj pager subsystem |
| 540 | * |
| 541 | * => called at boot time from uvm_pager_init() |
| 542 | */ |
| 543 | |
| 544 | void |
| 545 | uao_init(void) |
| 546 | { |
| 547 | static int uao_initialized; |
| 548 | |
| 549 | if (uao_initialized) |
| 550 | return; |
| 551 | uao_initialized = true; |
| 552 | LIST_INIT(&uao_list); |
| 553 | mutex_init(&uao_list_lock, MUTEX_DEFAULT, IPL_NONE); |
| 554 | pool_init(&uao_swhash_elt_pool, sizeof(struct uao_swhash_elt), |
| 555 | 0, 0, 0, "uaoeltpl" , NULL, IPL_VM); |
| 556 | } |
| 557 | |
| 558 | /* |
| 559 | * uao_reference: hold a reference to an anonymous UVM object. |
| 560 | */ |
| 561 | void |
| 562 | uao_reference(struct uvm_object *uobj) |
| 563 | { |
| 564 | /* Kernel object is persistent. */ |
| 565 | if (UVM_OBJ_IS_KERN_OBJECT(uobj)) { |
| 566 | return; |
| 567 | } |
| 568 | atomic_inc_uint(&uobj->uo_refs); |
| 569 | } |
| 570 | |
| 571 | /* |
| 572 | * uao_detach: drop a reference to an anonymous UVM object. |
| 573 | */ |
| 574 | void |
| 575 | uao_detach(struct uvm_object *uobj) |
| 576 | { |
| 577 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 578 | struct vm_page *pg; |
| 579 | |
| 580 | UVMHIST_FUNC("uao_detach" ); UVMHIST_CALLED(maphist); |
| 581 | |
| 582 | /* |
| 583 | * Detaching from kernel object is a NOP. |
| 584 | */ |
| 585 | |
| 586 | if (UVM_OBJ_IS_KERN_OBJECT(uobj)) |
| 587 | return; |
| 588 | |
| 589 | /* |
| 590 | * Drop the reference. If it was the last one, destroy the object. |
| 591 | */ |
| 592 | |
| 593 | UVMHIST_LOG(maphist," (uobj=0x%x) ref=%d" , uobj,uobj->uo_refs,0,0); |
| 594 | if (atomic_dec_uint_nv(&uobj->uo_refs) > 0) { |
| 595 | UVMHIST_LOG(maphist, "<- done (rc>0)" , 0,0,0,0); |
| 596 | return; |
| 597 | } |
| 598 | |
| 599 | /* |
| 600 | * Remove the aobj from the global list. |
| 601 | */ |
| 602 | |
| 603 | mutex_enter(&uao_list_lock); |
| 604 | LIST_REMOVE(aobj, u_list); |
| 605 | mutex_exit(&uao_list_lock); |
| 606 | |
| 607 | /* |
| 608 | * Free all the pages left in the aobj. For each page, when the |
| 609 | * page is no longer busy (and thus after any disk I/O that it is |
| 610 | * involved in is complete), release any swap resources and free |
| 611 | * the page itself. |
| 612 | */ |
| 613 | |
| 614 | mutex_enter(uobj->vmobjlock); |
| 615 | mutex_enter(&uvm_pageqlock); |
| 616 | while ((pg = TAILQ_FIRST(&uobj->memq)) != NULL) { |
| 617 | pmap_page_protect(pg, VM_PROT_NONE); |
| 618 | if (pg->flags & PG_BUSY) { |
| 619 | pg->flags |= PG_WANTED; |
| 620 | mutex_exit(&uvm_pageqlock); |
| 621 | UVM_UNLOCK_AND_WAIT(pg, uobj->vmobjlock, false, |
| 622 | "uao_det" , 0); |
| 623 | mutex_enter(uobj->vmobjlock); |
| 624 | mutex_enter(&uvm_pageqlock); |
| 625 | continue; |
| 626 | } |
| 627 | uao_dropswap(&aobj->u_obj, pg->offset >> PAGE_SHIFT); |
| 628 | uvm_pagefree(pg); |
| 629 | } |
| 630 | mutex_exit(&uvm_pageqlock); |
| 631 | |
| 632 | /* |
| 633 | * Finally, free the anonymous UVM object itself. |
| 634 | */ |
| 635 | |
| 636 | uao_free(aobj); |
| 637 | } |
| 638 | |
| 639 | /* |
| 640 | * uao_put: flush pages out of a uvm object |
| 641 | * |
| 642 | * => object should be locked by caller. we may _unlock_ the object |
| 643 | * if (and only if) we need to clean a page (PGO_CLEANIT). |
| 644 | * XXXJRT Currently, however, we don't. In the case of cleaning |
| 645 | * XXXJRT a page, we simply just deactivate it. Should probably |
| 646 | * XXXJRT handle this better, in the future (although "flushing" |
| 647 | * XXXJRT anonymous memory isn't terribly important). |
| 648 | * => if PGO_CLEANIT is not set, then we will neither unlock the object |
| 649 | * or block. |
| 650 | * => if PGO_ALLPAGE is set, then all pages in the object are valid targets |
| 651 | * for flushing. |
| 652 | * => NOTE: we rely on the fact that the object's memq is a TAILQ and |
| 653 | * that new pages are inserted on the tail end of the list. thus, |
| 654 | * we can make a complete pass through the object in one go by starting |
| 655 | * at the head and working towards the tail (new pages are put in |
| 656 | * front of us). |
| 657 | * => NOTE: we are allowed to lock the page queues, so the caller |
| 658 | * must not be holding the lock on them [e.g. pagedaemon had |
| 659 | * better not call us with the queues locked] |
| 660 | * => we return 0 unless we encountered some sort of I/O error |
| 661 | * XXXJRT currently never happens, as we never directly initiate |
| 662 | * XXXJRT I/O |
| 663 | * |
| 664 | * note on page traversal: |
| 665 | * we can traverse the pages in an object either by going down the |
| 666 | * linked list in "uobj->memq", or we can go over the address range |
| 667 | * by page doing hash table lookups for each address. depending |
| 668 | * on how many pages are in the object it may be cheaper to do one |
| 669 | * or the other. we set "by_list" to true if we are using memq. |
| 670 | * if the cost of a hash lookup was equal to the cost of the list |
| 671 | * traversal we could compare the number of pages in the start->stop |
| 672 | * range to the total number of pages in the object. however, it |
| 673 | * seems that a hash table lookup is more expensive than the linked |
| 674 | * list traversal, so we multiply the number of pages in the |
| 675 | * start->stop range by a penalty which we define below. |
| 676 | */ |
| 677 | |
| 678 | static int |
| 679 | uao_put(struct uvm_object *uobj, voff_t start, voff_t stop, int flags) |
| 680 | { |
| 681 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 682 | struct vm_page *pg, *nextpg, curmp, endmp; |
| 683 | bool by_list; |
| 684 | voff_t curoff; |
| 685 | UVMHIST_FUNC("uao_put" ); UVMHIST_CALLED(maphist); |
| 686 | |
| 687 | KASSERT(mutex_owned(uobj->vmobjlock)); |
| 688 | |
| 689 | curoff = 0; |
| 690 | if (flags & PGO_ALLPAGES) { |
| 691 | start = 0; |
| 692 | stop = aobj->u_pages << PAGE_SHIFT; |
| 693 | by_list = true; /* always go by the list */ |
| 694 | } else { |
| 695 | start = trunc_page(start); |
| 696 | if (stop == 0) { |
| 697 | stop = aobj->u_pages << PAGE_SHIFT; |
| 698 | } else { |
| 699 | stop = round_page(stop); |
| 700 | } |
| 701 | if (stop > (aobj->u_pages << PAGE_SHIFT)) { |
| 702 | printf("uao_flush: strange, got an out of range " |
| 703 | "flush (fixed)\n" ); |
| 704 | stop = aobj->u_pages << PAGE_SHIFT; |
| 705 | } |
| 706 | by_list = (uobj->uo_npages <= |
| 707 | ((stop - start) >> PAGE_SHIFT) * UVM_PAGE_TREE_PENALTY); |
| 708 | } |
| 709 | UVMHIST_LOG(maphist, |
| 710 | " flush start=0x%lx, stop=0x%x, by_list=%d, flags=0x%x" , |
| 711 | start, stop, by_list, flags); |
| 712 | |
| 713 | /* |
| 714 | * Don't need to do any work here if we're not freeing |
| 715 | * or deactivating pages. |
| 716 | */ |
| 717 | |
| 718 | if ((flags & (PGO_DEACTIVATE|PGO_FREE)) == 0) { |
| 719 | mutex_exit(uobj->vmobjlock); |
| 720 | return 0; |
| 721 | } |
| 722 | |
| 723 | /* |
| 724 | * Initialize the marker pages. See the comment in |
| 725 | * genfs_putpages() also. |
| 726 | */ |
| 727 | |
| 728 | curmp.flags = PG_MARKER; |
| 729 | endmp.flags = PG_MARKER; |
| 730 | |
| 731 | /* |
| 732 | * now do it. note: we must update nextpg in the body of loop or we |
| 733 | * will get stuck. we need to use nextpg if we'll traverse the list |
| 734 | * because we may free "pg" before doing the next loop. |
| 735 | */ |
| 736 | |
| 737 | if (by_list) { |
| 738 | TAILQ_INSERT_TAIL(&uobj->memq, &endmp, listq.queue); |
| 739 | nextpg = TAILQ_FIRST(&uobj->memq); |
| 740 | } else { |
| 741 | curoff = start; |
| 742 | nextpg = NULL; /* Quell compiler warning */ |
| 743 | } |
| 744 | |
| 745 | /* locked: uobj */ |
| 746 | for (;;) { |
| 747 | if (by_list) { |
| 748 | pg = nextpg; |
| 749 | if (pg == &endmp) |
| 750 | break; |
| 751 | nextpg = TAILQ_NEXT(pg, listq.queue); |
| 752 | if (pg->flags & PG_MARKER) |
| 753 | continue; |
| 754 | if (pg->offset < start || pg->offset >= stop) |
| 755 | continue; |
| 756 | } else { |
| 757 | if (curoff < stop) { |
| 758 | pg = uvm_pagelookup(uobj, curoff); |
| 759 | curoff += PAGE_SIZE; |
| 760 | } else |
| 761 | break; |
| 762 | if (pg == NULL) |
| 763 | continue; |
| 764 | } |
| 765 | |
| 766 | /* |
| 767 | * wait and try again if the page is busy. |
| 768 | */ |
| 769 | |
| 770 | if (pg->flags & PG_BUSY) { |
| 771 | if (by_list) { |
| 772 | TAILQ_INSERT_BEFORE(pg, &curmp, listq.queue); |
| 773 | } |
| 774 | pg->flags |= PG_WANTED; |
| 775 | UVM_UNLOCK_AND_WAIT(pg, uobj->vmobjlock, 0, |
| 776 | "uao_put" , 0); |
| 777 | mutex_enter(uobj->vmobjlock); |
| 778 | if (by_list) { |
| 779 | nextpg = TAILQ_NEXT(&curmp, listq.queue); |
| 780 | TAILQ_REMOVE(&uobj->memq, &curmp, |
| 781 | listq.queue); |
| 782 | } else |
| 783 | curoff -= PAGE_SIZE; |
| 784 | continue; |
| 785 | } |
| 786 | |
| 787 | switch (flags & (PGO_CLEANIT|PGO_FREE|PGO_DEACTIVATE)) { |
| 788 | |
| 789 | /* |
| 790 | * XXX In these first 3 cases, we always just |
| 791 | * XXX deactivate the page. We may want to |
| 792 | * XXX handle the different cases more specifically |
| 793 | * XXX in the future. |
| 794 | */ |
| 795 | |
| 796 | case PGO_CLEANIT|PGO_FREE: |
| 797 | case PGO_CLEANIT|PGO_DEACTIVATE: |
| 798 | case PGO_DEACTIVATE: |
| 799 | deactivate_it: |
| 800 | mutex_enter(&uvm_pageqlock); |
| 801 | /* skip the page if it's wired */ |
| 802 | if (pg->wire_count == 0) { |
| 803 | uvm_pagedeactivate(pg); |
| 804 | } |
| 805 | mutex_exit(&uvm_pageqlock); |
| 806 | break; |
| 807 | |
| 808 | case PGO_FREE: |
| 809 | /* |
| 810 | * If there are multiple references to |
| 811 | * the object, just deactivate the page. |
| 812 | */ |
| 813 | |
| 814 | if (uobj->uo_refs > 1) |
| 815 | goto deactivate_it; |
| 816 | |
| 817 | /* |
| 818 | * free the swap slot and the page. |
| 819 | */ |
| 820 | |
| 821 | pmap_page_protect(pg, VM_PROT_NONE); |
| 822 | |
| 823 | /* |
| 824 | * freeing swapslot here is not strictly necessary. |
| 825 | * however, leaving it here doesn't save much |
| 826 | * because we need to update swap accounting anyway. |
| 827 | */ |
| 828 | |
| 829 | uao_dropswap(uobj, pg->offset >> PAGE_SHIFT); |
| 830 | mutex_enter(&uvm_pageqlock); |
| 831 | uvm_pagefree(pg); |
| 832 | mutex_exit(&uvm_pageqlock); |
| 833 | break; |
| 834 | |
| 835 | default: |
| 836 | panic("%s: impossible" , __func__); |
| 837 | } |
| 838 | } |
| 839 | if (by_list) { |
| 840 | TAILQ_REMOVE(&uobj->memq, &endmp, listq.queue); |
| 841 | } |
| 842 | mutex_exit(uobj->vmobjlock); |
| 843 | return 0; |
| 844 | } |
| 845 | |
| 846 | /* |
| 847 | * uao_get: fetch me a page |
| 848 | * |
| 849 | * we have three cases: |
| 850 | * 1: page is resident -> just return the page. |
| 851 | * 2: page is zero-fill -> allocate a new page and zero it. |
| 852 | * 3: page is swapped out -> fetch the page from swap. |
| 853 | * |
| 854 | * cases 1 and 2 can be handled with PGO_LOCKED, case 3 cannot. |
| 855 | * so, if the "center" page hits case 3 (or any page, with PGO_ALLPAGES), |
| 856 | * then we will need to return EBUSY. |
| 857 | * |
| 858 | * => prefer map unlocked (not required) |
| 859 | * => object must be locked! we will _unlock_ it before starting any I/O. |
| 860 | * => flags: PGO_ALLPAGES: get all of the pages |
| 861 | * PGO_LOCKED: fault data structures are locked |
| 862 | * => NOTE: offset is the offset of pps[0], _NOT_ pps[centeridx] |
| 863 | * => NOTE: caller must check for released pages!! |
| 864 | */ |
| 865 | |
| 866 | static int |
| 867 | uao_get(struct uvm_object *uobj, voff_t offset, struct vm_page **pps, |
| 868 | int *npagesp, int centeridx, vm_prot_t access_type, int advice, int flags) |
| 869 | { |
| 870 | voff_t current_offset; |
| 871 | struct vm_page *ptmp = NULL; /* Quell compiler warning */ |
| 872 | int lcv, gotpages, maxpages, swslot, pageidx; |
| 873 | bool done; |
| 874 | UVMHIST_FUNC("uao_get" ); UVMHIST_CALLED(pdhist); |
| 875 | |
| 876 | UVMHIST_LOG(pdhist, "aobj=%p offset=%d, flags=%d" , |
| 877 | (struct uvm_aobj *)uobj, offset, flags,0); |
| 878 | |
| 879 | /* |
| 880 | * get number of pages |
| 881 | */ |
| 882 | |
| 883 | maxpages = *npagesp; |
| 884 | |
| 885 | /* |
| 886 | * step 1: handled the case where fault data structures are locked. |
| 887 | */ |
| 888 | |
| 889 | if (flags & PGO_LOCKED) { |
| 890 | |
| 891 | /* |
| 892 | * step 1a: get pages that are already resident. only do |
| 893 | * this if the data structures are locked (i.e. the first |
| 894 | * time through). |
| 895 | */ |
| 896 | |
| 897 | done = true; /* be optimistic */ |
| 898 | gotpages = 0; /* # of pages we got so far */ |
| 899 | for (lcv = 0, current_offset = offset ; lcv < maxpages ; |
| 900 | lcv++, current_offset += PAGE_SIZE) { |
| 901 | /* do we care about this page? if not, skip it */ |
| 902 | if (pps[lcv] == PGO_DONTCARE) |
| 903 | continue; |
| 904 | ptmp = uvm_pagelookup(uobj, current_offset); |
| 905 | |
| 906 | /* |
| 907 | * if page is new, attempt to allocate the page, |
| 908 | * zero-fill'd. |
| 909 | */ |
| 910 | |
| 911 | if (ptmp == NULL && uao_find_swslot(uobj, |
| 912 | current_offset >> PAGE_SHIFT) == 0) { |
| 913 | ptmp = uao_pagealloc(uobj, current_offset, |
| 914 | UVM_FLAG_COLORMATCH|UVM_PGA_ZERO); |
| 915 | if (ptmp) { |
| 916 | /* new page */ |
| 917 | ptmp->flags &= ~(PG_FAKE); |
| 918 | ptmp->pqflags |= PQ_AOBJ; |
| 919 | goto gotpage; |
| 920 | } |
| 921 | } |
| 922 | |
| 923 | /* |
| 924 | * to be useful must get a non-busy page |
| 925 | */ |
| 926 | |
| 927 | if (ptmp == NULL || (ptmp->flags & PG_BUSY) != 0) { |
| 928 | if (lcv == centeridx || |
| 929 | (flags & PGO_ALLPAGES) != 0) |
| 930 | /* need to do a wait or I/O! */ |
| 931 | done = false; |
| 932 | continue; |
| 933 | } |
| 934 | |
| 935 | /* |
| 936 | * useful page: busy/lock it and plug it in our |
| 937 | * result array |
| 938 | */ |
| 939 | |
| 940 | /* caller must un-busy this page */ |
| 941 | ptmp->flags |= PG_BUSY; |
| 942 | UVM_PAGE_OWN(ptmp, "uao_get1" ); |
| 943 | gotpage: |
| 944 | pps[lcv] = ptmp; |
| 945 | gotpages++; |
| 946 | } |
| 947 | |
| 948 | /* |
| 949 | * step 1b: now we've either done everything needed or we |
| 950 | * to unlock and do some waiting or I/O. |
| 951 | */ |
| 952 | |
| 953 | UVMHIST_LOG(pdhist, "<- done (done=%d)" , done, 0,0,0); |
| 954 | *npagesp = gotpages; |
| 955 | if (done) |
| 956 | return 0; |
| 957 | else |
| 958 | return EBUSY; |
| 959 | } |
| 960 | |
| 961 | /* |
| 962 | * step 2: get non-resident or busy pages. |
| 963 | * object is locked. data structures are unlocked. |
| 964 | */ |
| 965 | |
| 966 | if ((flags & PGO_SYNCIO) == 0) { |
| 967 | goto done; |
| 968 | } |
| 969 | |
| 970 | for (lcv = 0, current_offset = offset ; lcv < maxpages ; |
| 971 | lcv++, current_offset += PAGE_SIZE) { |
| 972 | |
| 973 | /* |
| 974 | * - skip over pages we've already gotten or don't want |
| 975 | * - skip over pages we don't _have_ to get |
| 976 | */ |
| 977 | |
| 978 | if (pps[lcv] != NULL || |
| 979 | (lcv != centeridx && (flags & PGO_ALLPAGES) == 0)) |
| 980 | continue; |
| 981 | |
| 982 | pageidx = current_offset >> PAGE_SHIFT; |
| 983 | |
| 984 | /* |
| 985 | * we have yet to locate the current page (pps[lcv]). we |
| 986 | * first look for a page that is already at the current offset. |
| 987 | * if we find a page, we check to see if it is busy or |
| 988 | * released. if that is the case, then we sleep on the page |
| 989 | * until it is no longer busy or released and repeat the lookup. |
| 990 | * if the page we found is neither busy nor released, then we |
| 991 | * busy it (so we own it) and plug it into pps[lcv]. this |
| 992 | * 'break's the following while loop and indicates we are |
| 993 | * ready to move on to the next page in the "lcv" loop above. |
| 994 | * |
| 995 | * if we exit the while loop with pps[lcv] still set to NULL, |
| 996 | * then it means that we allocated a new busy/fake/clean page |
| 997 | * ptmp in the object and we need to do I/O to fill in the data. |
| 998 | */ |
| 999 | |
| 1000 | /* top of "pps" while loop */ |
| 1001 | while (pps[lcv] == NULL) { |
| 1002 | /* look for a resident page */ |
| 1003 | ptmp = uvm_pagelookup(uobj, current_offset); |
| 1004 | |
| 1005 | /* not resident? allocate one now (if we can) */ |
| 1006 | if (ptmp == NULL) { |
| 1007 | |
| 1008 | ptmp = uao_pagealloc(uobj, current_offset, 0); |
| 1009 | |
| 1010 | /* out of RAM? */ |
| 1011 | if (ptmp == NULL) { |
| 1012 | mutex_exit(uobj->vmobjlock); |
| 1013 | UVMHIST_LOG(pdhist, |
| 1014 | "sleeping, ptmp == NULL\n" ,0,0,0,0); |
| 1015 | uvm_wait("uao_getpage" ); |
| 1016 | mutex_enter(uobj->vmobjlock); |
| 1017 | continue; |
| 1018 | } |
| 1019 | |
| 1020 | /* |
| 1021 | * safe with PQ's unlocked: because we just |
| 1022 | * alloc'd the page |
| 1023 | */ |
| 1024 | |
| 1025 | ptmp->pqflags |= PQ_AOBJ; |
| 1026 | |
| 1027 | /* |
| 1028 | * got new page ready for I/O. break pps while |
| 1029 | * loop. pps[lcv] is still NULL. |
| 1030 | */ |
| 1031 | |
| 1032 | break; |
| 1033 | } |
| 1034 | |
| 1035 | /* page is there, see if we need to wait on it */ |
| 1036 | if ((ptmp->flags & PG_BUSY) != 0) { |
| 1037 | ptmp->flags |= PG_WANTED; |
| 1038 | UVMHIST_LOG(pdhist, |
| 1039 | "sleeping, ptmp->flags 0x%x\n" , |
| 1040 | ptmp->flags,0,0,0); |
| 1041 | UVM_UNLOCK_AND_WAIT(ptmp, uobj->vmobjlock, |
| 1042 | false, "uao_get" , 0); |
| 1043 | mutex_enter(uobj->vmobjlock); |
| 1044 | continue; |
| 1045 | } |
| 1046 | |
| 1047 | /* |
| 1048 | * if we get here then the page has become resident and |
| 1049 | * unbusy between steps 1 and 2. we busy it now (so we |
| 1050 | * own it) and set pps[lcv] (so that we exit the while |
| 1051 | * loop). |
| 1052 | */ |
| 1053 | |
| 1054 | /* we own it, caller must un-busy */ |
| 1055 | ptmp->flags |= PG_BUSY; |
| 1056 | UVM_PAGE_OWN(ptmp, "uao_get2" ); |
| 1057 | pps[lcv] = ptmp; |
| 1058 | } |
| 1059 | |
| 1060 | /* |
| 1061 | * if we own the valid page at the correct offset, pps[lcv] will |
| 1062 | * point to it. nothing more to do except go to the next page. |
| 1063 | */ |
| 1064 | |
| 1065 | if (pps[lcv]) |
| 1066 | continue; /* next lcv */ |
| 1067 | |
| 1068 | /* |
| 1069 | * we have a "fake/busy/clean" page that we just allocated. |
| 1070 | * do the needed "i/o", either reading from swap or zeroing. |
| 1071 | */ |
| 1072 | |
| 1073 | swslot = uao_find_swslot(uobj, pageidx); |
| 1074 | |
| 1075 | /* |
| 1076 | * just zero the page if there's nothing in swap. |
| 1077 | */ |
| 1078 | |
| 1079 | if (swslot == 0) { |
| 1080 | |
| 1081 | /* |
| 1082 | * page hasn't existed before, just zero it. |
| 1083 | */ |
| 1084 | |
| 1085 | uvm_pagezero(ptmp); |
| 1086 | } else { |
| 1087 | #if defined(VMSWAP) |
| 1088 | int error; |
| 1089 | |
| 1090 | UVMHIST_LOG(pdhist, "pagein from swslot %d" , |
| 1091 | swslot, 0,0,0); |
| 1092 | |
| 1093 | /* |
| 1094 | * page in the swapped-out page. |
| 1095 | * unlock object for i/o, relock when done. |
| 1096 | */ |
| 1097 | |
| 1098 | mutex_exit(uobj->vmobjlock); |
| 1099 | error = uvm_swap_get(ptmp, swslot, PGO_SYNCIO); |
| 1100 | mutex_enter(uobj->vmobjlock); |
| 1101 | |
| 1102 | /* |
| 1103 | * I/O done. check for errors. |
| 1104 | */ |
| 1105 | |
| 1106 | if (error != 0) { |
| 1107 | UVMHIST_LOG(pdhist, "<- done (error=%d)" , |
| 1108 | error,0,0,0); |
| 1109 | if (ptmp->flags & PG_WANTED) |
| 1110 | wakeup(ptmp); |
| 1111 | |
| 1112 | /* |
| 1113 | * remove the swap slot from the aobj |
| 1114 | * and mark the aobj as having no real slot. |
| 1115 | * don't free the swap slot, thus preventing |
| 1116 | * it from being used again. |
| 1117 | */ |
| 1118 | |
| 1119 | swslot = uao_set_swslot(uobj, pageidx, |
| 1120 | SWSLOT_BAD); |
| 1121 | if (swslot > 0) { |
| 1122 | uvm_swap_markbad(swslot, 1); |
| 1123 | } |
| 1124 | |
| 1125 | mutex_enter(&uvm_pageqlock); |
| 1126 | uvm_pagefree(ptmp); |
| 1127 | mutex_exit(&uvm_pageqlock); |
| 1128 | mutex_exit(uobj->vmobjlock); |
| 1129 | return error; |
| 1130 | } |
| 1131 | #else /* defined(VMSWAP) */ |
| 1132 | panic("%s: pagein" , __func__); |
| 1133 | #endif /* defined(VMSWAP) */ |
| 1134 | } |
| 1135 | |
| 1136 | if ((access_type & VM_PROT_WRITE) == 0) { |
| 1137 | ptmp->flags |= PG_CLEAN; |
| 1138 | pmap_clear_modify(ptmp); |
| 1139 | } |
| 1140 | |
| 1141 | /* |
| 1142 | * we got the page! clear the fake flag (indicates valid |
| 1143 | * data now in page) and plug into our result array. note |
| 1144 | * that page is still busy. |
| 1145 | * |
| 1146 | * it is the callers job to: |
| 1147 | * => check if the page is released |
| 1148 | * => unbusy the page |
| 1149 | * => activate the page |
| 1150 | */ |
| 1151 | |
| 1152 | ptmp->flags &= ~PG_FAKE; |
| 1153 | pps[lcv] = ptmp; |
| 1154 | } |
| 1155 | |
| 1156 | /* |
| 1157 | * finally, unlock object and return. |
| 1158 | */ |
| 1159 | |
| 1160 | done: |
| 1161 | mutex_exit(uobj->vmobjlock); |
| 1162 | UVMHIST_LOG(pdhist, "<- done (OK)" ,0,0,0,0); |
| 1163 | return 0; |
| 1164 | } |
| 1165 | |
| 1166 | #if defined(VMSWAP) |
| 1167 | |
| 1168 | /* |
| 1169 | * uao_dropswap: release any swap resources from this aobj page. |
| 1170 | * |
| 1171 | * => aobj must be locked or have a reference count of 0. |
| 1172 | */ |
| 1173 | |
| 1174 | void |
| 1175 | uao_dropswap(struct uvm_object *uobj, int pageidx) |
| 1176 | { |
| 1177 | int slot; |
| 1178 | |
| 1179 | slot = uao_set_swslot(uobj, pageidx, 0); |
| 1180 | if (slot) { |
| 1181 | uvm_swap_free(slot, 1); |
| 1182 | } |
| 1183 | } |
| 1184 | |
| 1185 | /* |
| 1186 | * page in every page in every aobj that is paged-out to a range of swslots. |
| 1187 | * |
| 1188 | * => nothing should be locked. |
| 1189 | * => returns true if pagein was aborted due to lack of memory. |
| 1190 | */ |
| 1191 | |
| 1192 | bool |
| 1193 | uao_swap_off(int startslot, int endslot) |
| 1194 | { |
| 1195 | struct uvm_aobj *aobj; |
| 1196 | |
| 1197 | /* |
| 1198 | * Walk the list of all anonymous UVM objects. Grab the first. |
| 1199 | */ |
| 1200 | mutex_enter(&uao_list_lock); |
| 1201 | if ((aobj = LIST_FIRST(&uao_list)) == NULL) { |
| 1202 | mutex_exit(&uao_list_lock); |
| 1203 | return false; |
| 1204 | } |
| 1205 | uao_reference(&aobj->u_obj); |
| 1206 | |
| 1207 | do { |
| 1208 | struct uvm_aobj *nextaobj; |
| 1209 | bool rv; |
| 1210 | |
| 1211 | /* |
| 1212 | * Prefetch the next object and immediately hold a reference |
| 1213 | * on it, so neither the current nor the next entry could |
| 1214 | * disappear while we are iterating. |
| 1215 | */ |
| 1216 | if ((nextaobj = LIST_NEXT(aobj, u_list)) != NULL) { |
| 1217 | uao_reference(&nextaobj->u_obj); |
| 1218 | } |
| 1219 | mutex_exit(&uao_list_lock); |
| 1220 | |
| 1221 | /* |
| 1222 | * Page in all pages in the swap slot range. |
| 1223 | */ |
| 1224 | mutex_enter(aobj->u_obj.vmobjlock); |
| 1225 | rv = uao_pagein(aobj, startslot, endslot); |
| 1226 | mutex_exit(aobj->u_obj.vmobjlock); |
| 1227 | |
| 1228 | /* Drop the reference of the current object. */ |
| 1229 | uao_detach(&aobj->u_obj); |
| 1230 | if (rv) { |
| 1231 | if (nextaobj) { |
| 1232 | uao_detach(&nextaobj->u_obj); |
| 1233 | } |
| 1234 | return rv; |
| 1235 | } |
| 1236 | |
| 1237 | aobj = nextaobj; |
| 1238 | mutex_enter(&uao_list_lock); |
| 1239 | } while (aobj); |
| 1240 | |
| 1241 | mutex_exit(&uao_list_lock); |
| 1242 | return false; |
| 1243 | } |
| 1244 | |
| 1245 | /* |
| 1246 | * page in any pages from aobj in the given range. |
| 1247 | * |
| 1248 | * => aobj must be locked and is returned locked. |
| 1249 | * => returns true if pagein was aborted due to lack of memory. |
| 1250 | */ |
| 1251 | static bool |
| 1252 | uao_pagein(struct uvm_aobj *aobj, int startslot, int endslot) |
| 1253 | { |
| 1254 | bool rv; |
| 1255 | |
| 1256 | if (UAO_USES_SWHASH(aobj)) { |
| 1257 | struct uao_swhash_elt *elt; |
| 1258 | int buck; |
| 1259 | |
| 1260 | restart: |
| 1261 | for (buck = aobj->u_swhashmask; buck >= 0; buck--) { |
| 1262 | for (elt = LIST_FIRST(&aobj->u_swhash[buck]); |
| 1263 | elt != NULL; |
| 1264 | elt = LIST_NEXT(elt, list)) { |
| 1265 | int i; |
| 1266 | |
| 1267 | for (i = 0; i < UAO_SWHASH_CLUSTER_SIZE; i++) { |
| 1268 | int slot = elt->slots[i]; |
| 1269 | |
| 1270 | /* |
| 1271 | * if the slot isn't in range, skip it. |
| 1272 | */ |
| 1273 | |
| 1274 | if (slot < startslot || |
| 1275 | slot >= endslot) { |
| 1276 | continue; |
| 1277 | } |
| 1278 | |
| 1279 | /* |
| 1280 | * process the page, |
| 1281 | * the start over on this object |
| 1282 | * since the swhash elt |
| 1283 | * may have been freed. |
| 1284 | */ |
| 1285 | |
| 1286 | rv = uao_pagein_page(aobj, |
| 1287 | UAO_SWHASH_ELT_PAGEIDX_BASE(elt) + i); |
| 1288 | if (rv) { |
| 1289 | return rv; |
| 1290 | } |
| 1291 | goto restart; |
| 1292 | } |
| 1293 | } |
| 1294 | } |
| 1295 | } else { |
| 1296 | int i; |
| 1297 | |
| 1298 | for (i = 0; i < aobj->u_pages; i++) { |
| 1299 | int slot = aobj->u_swslots[i]; |
| 1300 | |
| 1301 | /* |
| 1302 | * if the slot isn't in range, skip it |
| 1303 | */ |
| 1304 | |
| 1305 | if (slot < startslot || slot >= endslot) { |
| 1306 | continue; |
| 1307 | } |
| 1308 | |
| 1309 | /* |
| 1310 | * process the page. |
| 1311 | */ |
| 1312 | |
| 1313 | rv = uao_pagein_page(aobj, i); |
| 1314 | if (rv) { |
| 1315 | return rv; |
| 1316 | } |
| 1317 | } |
| 1318 | } |
| 1319 | |
| 1320 | return false; |
| 1321 | } |
| 1322 | |
| 1323 | /* |
| 1324 | * uao_pagein_page: page in a single page from an anonymous UVM object. |
| 1325 | * |
| 1326 | * => Returns true if pagein was aborted due to lack of memory. |
| 1327 | * => Object must be locked and is returned locked. |
| 1328 | */ |
| 1329 | |
| 1330 | static bool |
| 1331 | uao_pagein_page(struct uvm_aobj *aobj, int pageidx) |
| 1332 | { |
| 1333 | struct uvm_object *uobj = &aobj->u_obj; |
| 1334 | struct vm_page *pg; |
| 1335 | int rv, npages; |
| 1336 | |
| 1337 | pg = NULL; |
| 1338 | npages = 1; |
| 1339 | |
| 1340 | KASSERT(mutex_owned(uobj->vmobjlock)); |
| 1341 | rv = uao_get(uobj, pageidx << PAGE_SHIFT, &pg, &npages, |
| 1342 | 0, VM_PROT_READ | VM_PROT_WRITE, 0, PGO_SYNCIO); |
| 1343 | |
| 1344 | /* |
| 1345 | * relock and finish up. |
| 1346 | */ |
| 1347 | |
| 1348 | mutex_enter(uobj->vmobjlock); |
| 1349 | switch (rv) { |
| 1350 | case 0: |
| 1351 | break; |
| 1352 | |
| 1353 | case EIO: |
| 1354 | case ERESTART: |
| 1355 | |
| 1356 | /* |
| 1357 | * nothing more to do on errors. |
| 1358 | * ERESTART can only mean that the anon was freed, |
| 1359 | * so again there's nothing to do. |
| 1360 | */ |
| 1361 | |
| 1362 | return false; |
| 1363 | |
| 1364 | default: |
| 1365 | return true; |
| 1366 | } |
| 1367 | |
| 1368 | /* |
| 1369 | * ok, we've got the page now. |
| 1370 | * mark it as dirty, clear its swslot and un-busy it. |
| 1371 | */ |
| 1372 | uao_dropswap(&aobj->u_obj, pageidx); |
| 1373 | |
| 1374 | /* |
| 1375 | * make sure it's on a page queue. |
| 1376 | */ |
| 1377 | mutex_enter(&uvm_pageqlock); |
| 1378 | if (pg->wire_count == 0) |
| 1379 | uvm_pageenqueue(pg); |
| 1380 | mutex_exit(&uvm_pageqlock); |
| 1381 | |
| 1382 | if (pg->flags & PG_WANTED) { |
| 1383 | wakeup(pg); |
| 1384 | } |
| 1385 | pg->flags &= ~(PG_WANTED|PG_BUSY|PG_CLEAN|PG_FAKE); |
| 1386 | UVM_PAGE_OWN(pg, NULL); |
| 1387 | |
| 1388 | return false; |
| 1389 | } |
| 1390 | |
| 1391 | /* |
| 1392 | * uao_dropswap_range: drop swapslots in the range. |
| 1393 | * |
| 1394 | * => aobj must be locked and is returned locked. |
| 1395 | * => start is inclusive. end is exclusive. |
| 1396 | */ |
| 1397 | |
| 1398 | void |
| 1399 | uao_dropswap_range(struct uvm_object *uobj, voff_t start, voff_t end) |
| 1400 | { |
| 1401 | struct uvm_aobj *aobj = (struct uvm_aobj *)uobj; |
| 1402 | int swpgonlydelta = 0; |
| 1403 | |
| 1404 | KASSERT(mutex_owned(uobj->vmobjlock)); |
| 1405 | |
| 1406 | if (end == 0) { |
| 1407 | end = INT64_MAX; |
| 1408 | } |
| 1409 | |
| 1410 | if (UAO_USES_SWHASH(aobj)) { |
| 1411 | int i, hashbuckets = aobj->u_swhashmask + 1; |
| 1412 | voff_t taghi; |
| 1413 | voff_t taglo; |
| 1414 | |
| 1415 | taglo = UAO_SWHASH_ELT_TAG(start); |
| 1416 | taghi = UAO_SWHASH_ELT_TAG(end); |
| 1417 | |
| 1418 | for (i = 0; i < hashbuckets; i++) { |
| 1419 | struct uao_swhash_elt *elt, *next; |
| 1420 | |
| 1421 | for (elt = LIST_FIRST(&aobj->u_swhash[i]); |
| 1422 | elt != NULL; |
| 1423 | elt = next) { |
| 1424 | int startidx, endidx; |
| 1425 | int j; |
| 1426 | |
| 1427 | next = LIST_NEXT(elt, list); |
| 1428 | |
| 1429 | if (elt->tag < taglo || taghi < elt->tag) { |
| 1430 | continue; |
| 1431 | } |
| 1432 | |
| 1433 | if (elt->tag == taglo) { |
| 1434 | startidx = |
| 1435 | UAO_SWHASH_ELT_PAGESLOT_IDX(start); |
| 1436 | } else { |
| 1437 | startidx = 0; |
| 1438 | } |
| 1439 | |
| 1440 | if (elt->tag == taghi) { |
| 1441 | endidx = |
| 1442 | UAO_SWHASH_ELT_PAGESLOT_IDX(end); |
| 1443 | } else { |
| 1444 | endidx = UAO_SWHASH_CLUSTER_SIZE; |
| 1445 | } |
| 1446 | |
| 1447 | for (j = startidx; j < endidx; j++) { |
| 1448 | int slot = elt->slots[j]; |
| 1449 | |
| 1450 | KASSERT(uvm_pagelookup(&aobj->u_obj, |
| 1451 | (UAO_SWHASH_ELT_PAGEIDX_BASE(elt) |
| 1452 | + j) << PAGE_SHIFT) == NULL); |
| 1453 | if (slot > 0) { |
| 1454 | uvm_swap_free(slot, 1); |
| 1455 | swpgonlydelta++; |
| 1456 | KASSERT(elt->count > 0); |
| 1457 | elt->slots[j] = 0; |
| 1458 | elt->count--; |
| 1459 | } |
| 1460 | } |
| 1461 | |
| 1462 | if (elt->count == 0) { |
| 1463 | LIST_REMOVE(elt, list); |
| 1464 | pool_put(&uao_swhash_elt_pool, elt); |
| 1465 | } |
| 1466 | } |
| 1467 | } |
| 1468 | } else { |
| 1469 | int i; |
| 1470 | |
| 1471 | if (aobj->u_pages < end) { |
| 1472 | end = aobj->u_pages; |
| 1473 | } |
| 1474 | for (i = start; i < end; i++) { |
| 1475 | int slot = aobj->u_swslots[i]; |
| 1476 | |
| 1477 | if (slot > 0) { |
| 1478 | uvm_swap_free(slot, 1); |
| 1479 | swpgonlydelta++; |
| 1480 | } |
| 1481 | } |
| 1482 | } |
| 1483 | |
| 1484 | /* |
| 1485 | * adjust the counter of pages only in swap for all |
| 1486 | * the swap slots we've freed. |
| 1487 | */ |
| 1488 | |
| 1489 | if (swpgonlydelta > 0) { |
| 1490 | mutex_enter(&uvm_swap_data_lock); |
| 1491 | KASSERT(uvmexp.swpgonly >= swpgonlydelta); |
| 1492 | uvmexp.swpgonly -= swpgonlydelta; |
| 1493 | mutex_exit(&uvm_swap_data_lock); |
| 1494 | } |
| 1495 | } |
| 1496 | |
| 1497 | #endif /* defined(VMSWAP) */ |
| 1498 | |