1 /* 2 * Copyright (c) 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * The Mach Operating System project at Carnegie-Mellon University. 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 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * from: @(#)vm_kern.c 8.3 (Berkeley) 1/12/94 37 * 38 * 39 * Copyright (c) 1987, 1990 Carnegie-Mellon University. 40 * All rights reserved. 41 * 42 * Authors: Avadis Tevanian, Jr., Michael Wayne Young 43 * 44 * Permission to use, copy, modify and distribute this software and 45 * its documentation is hereby granted, provided that both the copyright 46 * notice and this permission notice appear in all copies of the 47 * software, derivative works or modified versions, and any portions 48 * thereof, and that both notices appear in supporting documentation. 49 * 50 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" 51 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND 52 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. 53 * 54 * Carnegie Mellon requests users of this software to return to 55 * 56 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU 57 * School of Computer Science 58 * Carnegie Mellon University 59 * Pittsburgh PA 15213-3890 60 * 61 * any improvements or extensions that they make and grant Carnegie the 62 * rights to redistribute these changes. 63 * 64 * $FreeBSD: src/sys/vm/vm_kern.c,v 1.61.2.2 2002/03/12 18:25:26 tegge Exp $ 65 * $DragonFly: src/sys/vm/vm_kern.c,v 1.28 2007/01/02 04:52:31 dillon Exp $ 66 */ 67 68 /* 69 * Kernel memory management. 70 */ 71 72 #include <sys/param.h> 73 #include <sys/systm.h> 74 #include <sys/proc.h> 75 #include <sys/malloc.h> 76 #include <sys/kernel.h> 77 #include <sys/sysctl.h> 78 79 #include <vm/vm.h> 80 #include <vm/vm_param.h> 81 #include <sys/lock.h> 82 #include <vm/pmap.h> 83 #include <vm/vm_map.h> 84 #include <vm/vm_object.h> 85 #include <vm/vm_page.h> 86 #include <vm/vm_pageout.h> 87 #include <vm/vm_kern.h> 88 #include <vm/vm_extern.h> 89 90 struct vm_map kernel_map; 91 struct vm_map exec_map; 92 struct vm_map clean_map; 93 struct vm_map buffer_map; 94 95 /* 96 * kmem_alloc_pageable: 97 * 98 * Allocate pageable memory to the kernel's address map. 99 * "map" must be kernel_map or a submap of kernel_map. 100 */ 101 vm_offset_t 102 kmem_alloc_pageable(vm_map_t map, vm_size_t size) 103 { 104 vm_offset_t addr; 105 int result; 106 107 size = round_page(size); 108 addr = vm_map_min(map); 109 result = vm_map_find(map, NULL, (vm_offset_t) 0, 110 &addr, size, 111 TRUE, 112 VM_MAPTYPE_NORMAL, 113 VM_PROT_ALL, VM_PROT_ALL, 114 0); 115 if (result != KERN_SUCCESS) { 116 return (0); 117 } 118 return (addr); 119 } 120 121 /* 122 * kmem_alloc_nofault: 123 * 124 * Same as kmem_alloc_pageable, except that it create a nofault entry. 125 */ 126 vm_offset_t 127 kmem_alloc_nofault(vm_map_t map, vm_size_t size) 128 { 129 vm_offset_t addr; 130 int result; 131 132 size = round_page(size); 133 addr = vm_map_min(map); 134 result = vm_map_find(map, NULL, (vm_offset_t) 0, 135 &addr, size, 136 TRUE, 137 VM_MAPTYPE_NORMAL, 138 VM_PROT_ALL, VM_PROT_ALL, 139 MAP_NOFAULT); 140 if (result != KERN_SUCCESS) { 141 return (0); 142 } 143 return (addr); 144 } 145 146 /* 147 * Allocate wired-down memory in the kernel's address map 148 * or a submap. 149 */ 150 vm_offset_t 151 kmem_alloc3(vm_map_t map, vm_size_t size, int kmflags) 152 { 153 vm_offset_t addr; 154 vm_offset_t i; 155 int count; 156 157 size = round_page(size); 158 159 if (kmflags & KM_KRESERVE) 160 count = vm_map_entry_kreserve(MAP_RESERVE_COUNT); 161 else 162 count = vm_map_entry_reserve(MAP_RESERVE_COUNT); 163 164 /* 165 * Use the kernel object for wired-down kernel pages. Assume that no 166 * region of the kernel object is referenced more than once. 167 * 168 * Locate sufficient space in the map. This will give us the final 169 * virtual address for the new memory, and thus will tell us the 170 * offset within the kernel map. 171 */ 172 vm_map_lock(map); 173 if (vm_map_findspace(map, vm_map_min(map), size, 1, &addr)) { 174 vm_map_unlock(map); 175 if (kmflags & KM_KRESERVE) 176 vm_map_entry_krelease(count); 177 else 178 vm_map_entry_release(count); 179 return (0); 180 } 181 vm_object_reference(&kernel_object); 182 vm_map_insert(map, &count, 183 &kernel_object, addr, addr, addr + size, 184 VM_MAPTYPE_NORMAL, 185 VM_PROT_ALL, VM_PROT_ALL, 186 0); 187 vm_map_unlock(map); 188 if (kmflags & KM_KRESERVE) 189 vm_map_entry_krelease(count); 190 else 191 vm_map_entry_release(count); 192 193 /* 194 * Guarantee that there are pages already in this object before 195 * calling vm_map_wire. This is to prevent the following 196 * scenario: 197 * 198 * 1) Threads have swapped out, so that there is a pager for the 199 * kernel_object. 2) The kmsg zone is empty, and so we are 200 * kmem_allocing a new page for it. 3) vm_map_wire calls vm_fault; 201 * there is no page, but there is a pager, so we call 202 * pager_data_request. But the kmsg zone is empty, so we must 203 * kmem_alloc. 4) goto 1 5) Even if the kmsg zone is not empty: when 204 * we get the data back from the pager, it will be (very stale) 205 * non-zero data. kmem_alloc is defined to return zero-filled memory. 206 * 207 * We're intentionally not activating the pages we allocate to prevent a 208 * race with page-out. vm_map_wire will wire the pages. 209 */ 210 211 for (i = 0; i < size; i += PAGE_SIZE) { 212 vm_page_t mem; 213 214 mem = vm_page_grab(&kernel_object, OFF_TO_IDX(addr + i), 215 VM_ALLOC_ZERO | VM_ALLOC_NORMAL | VM_ALLOC_RETRY); 216 if ((mem->flags & PG_ZERO) == 0) 217 vm_page_zero_fill(mem); 218 mem->valid = VM_PAGE_BITS_ALL; 219 vm_page_flag_clear(mem, PG_ZERO); 220 vm_page_wakeup(mem); 221 } 222 223 /* 224 * And finally, mark the data as non-pageable. 225 */ 226 227 vm_map_wire(map, (vm_offset_t) addr, addr + size, kmflags); 228 229 return (addr); 230 } 231 232 /* 233 * kmem_free: 234 * 235 * Release a region of kernel virtual memory allocated 236 * with kmem_alloc, and return the physical pages 237 * associated with that region. 238 * 239 * This routine may not block on kernel maps. 240 */ 241 void 242 kmem_free(vm_map_t map, vm_offset_t addr, vm_size_t size) 243 { 244 vm_map_remove(map, trunc_page(addr), round_page(addr + size)); 245 } 246 247 /* 248 * kmem_suballoc: 249 * 250 * Used to break a system map into smaller maps, usually to reduce 251 * contention and to provide large KVA spaces for subsystems like the 252 * buffer cache. 253 * 254 * parent Map to take range from 255 * result 256 * size Size of range to find 257 * min, max Returned endpoints of map 258 * pageable Can the region be paged 259 */ 260 void 261 kmem_suballoc(vm_map_t parent, vm_map_t result, 262 vm_offset_t *min, vm_offset_t *max, vm_size_t size) 263 { 264 int ret; 265 266 size = round_page(size); 267 268 *min = (vm_offset_t) vm_map_min(parent); 269 ret = vm_map_find(parent, NULL, (vm_offset_t) 0, 270 min, size, 271 TRUE, 272 VM_MAPTYPE_UNSPECIFIED, 273 VM_PROT_ALL, VM_PROT_ALL, 274 0); 275 if (ret != KERN_SUCCESS) { 276 kprintf("kmem_suballoc: bad status return of %d.\n", ret); 277 panic("kmem_suballoc"); 278 } 279 *max = *min + size; 280 pmap_reference(vm_map_pmap(parent)); 281 vm_map_init(result, *min, *max, vm_map_pmap(parent)); 282 if ((ret = vm_map_submap(parent, *min, *max, result)) != KERN_SUCCESS) 283 panic("kmem_suballoc: unable to change range to submap"); 284 } 285 286 /* 287 * kmem_alloc_wait: 288 * 289 * Allocates pageable memory from a sub-map of the kernel. If the submap 290 * has no room, the caller sleeps waiting for more memory in the submap. 291 * 292 * This routine may block. 293 */ 294 295 vm_offset_t 296 kmem_alloc_wait(vm_map_t map, vm_size_t size) 297 { 298 vm_offset_t addr; 299 int count; 300 301 size = round_page(size); 302 303 count = vm_map_entry_reserve(MAP_RESERVE_COUNT); 304 305 for (;;) { 306 /* 307 * To make this work for more than one map, use the map's lock 308 * to lock out sleepers/wakers. 309 */ 310 vm_map_lock(map); 311 if (vm_map_findspace(map, vm_map_min(map), size, 1, &addr) == 0) 312 break; 313 /* no space now; see if we can ever get space */ 314 if (vm_map_max(map) - vm_map_min(map) < size) { 315 vm_map_entry_release(count); 316 vm_map_unlock(map); 317 return (0); 318 } 319 vm_map_unlock(map); 320 tsleep(map, 0, "kmaw", 0); 321 } 322 vm_map_insert(map, &count, 323 NULL, (vm_offset_t) 0, 324 addr, addr + size, 325 VM_MAPTYPE_NORMAL, 326 VM_PROT_ALL, VM_PROT_ALL, 327 0); 328 vm_map_unlock(map); 329 vm_map_entry_release(count); 330 return (addr); 331 } 332 333 /* 334 * kmem_free_wakeup: 335 * 336 * Returns memory to a submap of the kernel, and wakes up any processes 337 * waiting for memory in that map. 338 */ 339 void 340 kmem_free_wakeup(vm_map_t map, vm_offset_t addr, vm_size_t size) 341 { 342 int count; 343 344 count = vm_map_entry_reserve(MAP_RESERVE_COUNT); 345 vm_map_lock(map); 346 vm_map_delete(map, trunc_page(addr), round_page(addr + size), &count); 347 wakeup(map); 348 vm_map_unlock(map); 349 vm_map_entry_release(count); 350 } 351 352 /* 353 * kmem_init: 354 * 355 * Create the kernel_map and insert mappings to cover areas already 356 * allocated or reserved thus far. That is, the area (KvaStart,start) 357 * and (end,KvaEnd) must be marked as allocated. 358 * 359 * We could use a min_offset of 0 instead of KvaStart, but since the 360 * min_offset is not used for any calculations other then a bounds check 361 * it does not effect readability. KvaStart is more appropriate. 362 * 363 * Depend on the zalloc bootstrap cache to get our vm_map_entry_t. 364 */ 365 void 366 kmem_init(vm_offset_t start, vm_offset_t end) 367 { 368 vm_map_t m; 369 int count; 370 371 m = vm_map_create(&kernel_map, &kernel_pmap, KvaStart, KvaEnd); 372 vm_map_lock(m); 373 /* N.B.: cannot use kgdb to debug, starting with this assignment ... */ 374 m->system_map = 1; 375 count = vm_map_entry_reserve(MAP_RESERVE_COUNT); 376 if (KvaStart != start) { 377 vm_map_insert(m, &count, NULL, (vm_offset_t) 0, 378 KvaStart, start, 379 VM_MAPTYPE_NORMAL, 380 VM_PROT_ALL, VM_PROT_ALL, 381 0); 382 } 383 if (KvaEnd != end) { 384 vm_map_insert(m, &count, NULL, (vm_offset_t) 0, 385 end, KvaEnd, 386 VM_MAPTYPE_NORMAL, 387 VM_PROT_ALL, VM_PROT_ALL, 388 0); 389 } 390 /* ... and ending with the completion of the above `insert' */ 391 vm_map_unlock(m); 392 vm_map_entry_release(count); 393 } 394 395 static int 396 kvm_size(SYSCTL_HANDLER_ARGS) 397 { 398 unsigned long ksize = KvaSize; 399 400 return sysctl_handle_long(oidp, &ksize, 0, req); 401 } 402 SYSCTL_PROC(_vm, OID_AUTO, kvm_size, CTLTYPE_LONG|CTLFLAG_RD, 403 0, 0, kvm_size, "IU", "Size of KVM"); 404 405 static int 406 kvm_free(SYSCTL_HANDLER_ARGS) 407 { 408 unsigned long kfree = virtual_end - kernel_vm_end; 409 410 return sysctl_handle_long(oidp, &kfree, 0, req); 411 } 412 SYSCTL_PROC(_vm, OID_AUTO, kvm_free, CTLTYPE_LONG|CTLFLAG_RD, 413 0, 0, kvm_free, "IU", "Amount of KVM free"); 414 415