pmap.c revision 1.15
1 /* $NetBSD: pmap.c,v 1.15 2016/07/11 16:06:09 matt Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 2001 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center and by Chris G. Demetriou. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1992, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * This code is derived from software contributed to Berkeley by 38 * the Systems Programming Group of the University of Utah Computer 39 * Science Department and Ralph Campbell. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)pmap.c 8.4 (Berkeley) 1/26/94 66 */ 67 68 #include <sys/cdefs.h> 69 70 __KERNEL_RCSID(0, "$NetBSD: pmap.c,v 1.15 2016/07/11 16:06:09 matt Exp $"); 71 72 /* 73 * Manages physical address maps. 74 * 75 * In addition to hardware address maps, this 76 * module is called upon to provide software-use-only 77 * maps which may or may not be stored in the same 78 * form as hardware maps. These pseudo-maps are 79 * used to store intermediate results from copy 80 * operations to and from address spaces. 81 * 82 * Since the information managed by this module is 83 * also stored by the logical address mapping module, 84 * this module may throw away valid virtual-to-physical 85 * mappings at almost any time. However, invalidations 86 * of virtual-to-physical mappings must be done as 87 * requested. 88 * 89 * In order to cope with hardware architectures which 90 * make virtual-to-physical map invalidates expensive, 91 * this module may delay invalidate or reduced protection 92 * operations until such time as they are actually 93 * necessary. This module is given full information as 94 * to which processors are currently using which maps, 95 * and to when physical maps must be made correct. 96 */ 97 98 #include "opt_modular.h" 99 #include "opt_multiprocessor.h" 100 #include "opt_sysv.h" 101 102 #define __PMAP_PRIVATE 103 104 #include <sys/param.h> 105 #include <sys/atomic.h> 106 #include <sys/buf.h> 107 #include <sys/cpu.h> 108 #include <sys/mutex.h> 109 #include <sys/pool.h> 110 #include <sys/atomic.h> 111 #include <sys/mutex.h> 112 #include <sys/atomic.h> 113 114 #include <uvm/uvm.h> 115 116 #if defined(MULTIPROCESSOR) && defined(PMAP_VIRTUAL_CACHE_ALIASES) \ 117 && !defined(PMAP_NO_PV_UNCACHED) 118 #error PMAP_VIRTUAL_CACHE_ALIASES with MULTIPROCESSOR requires \ 119 PMAP_NO_PV_UNCACHED to be defined 120 #endif 121 122 PMAP_COUNTER(remove_kernel_calls, "remove kernel calls"); 123 PMAP_COUNTER(remove_kernel_pages, "kernel pages unmapped"); 124 PMAP_COUNTER(remove_user_calls, "remove user calls"); 125 PMAP_COUNTER(remove_user_pages, "user pages unmapped"); 126 PMAP_COUNTER(remove_flushes, "remove cache flushes"); 127 PMAP_COUNTER(remove_tlb_ops, "remove tlb ops"); 128 PMAP_COUNTER(remove_pvfirst, "remove pv first"); 129 PMAP_COUNTER(remove_pvsearch, "remove pv search"); 130 131 PMAP_COUNTER(prefer_requests, "prefer requests"); 132 PMAP_COUNTER(prefer_adjustments, "prefer adjustments"); 133 134 PMAP_COUNTER(idlezeroed_pages, "pages idle zeroed"); 135 136 PMAP_COUNTER(kenter_pa, "kernel fast mapped pages"); 137 PMAP_COUNTER(kenter_pa_bad, "kernel fast mapped pages (bad color)"); 138 PMAP_COUNTER(kenter_pa_unmanaged, "kernel fast mapped unmanaged pages"); 139 PMAP_COUNTER(kremove_pages, "kernel fast unmapped pages"); 140 141 PMAP_COUNTER(page_cache_evictions, "pages changed to uncacheable"); 142 PMAP_COUNTER(page_cache_restorations, "pages changed to cacheable"); 143 144 PMAP_COUNTER(kernel_mappings_bad, "kernel pages mapped (bad color)"); 145 PMAP_COUNTER(user_mappings_bad, "user pages mapped (bad color)"); 146 PMAP_COUNTER(kernel_mappings, "kernel pages mapped"); 147 PMAP_COUNTER(user_mappings, "user pages mapped"); 148 PMAP_COUNTER(user_mappings_changed, "user mapping changed"); 149 PMAP_COUNTER(kernel_mappings_changed, "kernel mapping changed"); 150 PMAP_COUNTER(uncached_mappings, "uncached pages mapped"); 151 PMAP_COUNTER(unmanaged_mappings, "unmanaged pages mapped"); 152 PMAP_COUNTER(managed_mappings, "managed pages mapped"); 153 PMAP_COUNTER(mappings, "pages mapped"); 154 PMAP_COUNTER(remappings, "pages remapped"); 155 PMAP_COUNTER(unmappings, "pages unmapped"); 156 PMAP_COUNTER(primary_mappings, "page initial mappings"); 157 PMAP_COUNTER(primary_unmappings, "page final unmappings"); 158 PMAP_COUNTER(tlb_hit, "page mapping"); 159 160 PMAP_COUNTER(exec_mappings, "exec pages mapped"); 161 PMAP_COUNTER(exec_synced_mappings, "exec pages synced"); 162 PMAP_COUNTER(exec_synced_remove, "exec pages synced (PR)"); 163 PMAP_COUNTER(exec_synced_clear_modify, "exec pages synced (CM)"); 164 PMAP_COUNTER(exec_synced_page_protect, "exec pages synced (PP)"); 165 PMAP_COUNTER(exec_synced_protect, "exec pages synced (P)"); 166 PMAP_COUNTER(exec_uncached_page_protect, "exec pages uncached (PP)"); 167 PMAP_COUNTER(exec_uncached_clear_modify, "exec pages uncached (CM)"); 168 PMAP_COUNTER(exec_uncached_zero_page, "exec pages uncached (ZP)"); 169 PMAP_COUNTER(exec_uncached_copy_page, "exec pages uncached (CP)"); 170 PMAP_COUNTER(exec_uncached_remove, "exec pages uncached (PR)"); 171 172 PMAP_COUNTER(create, "creates"); 173 PMAP_COUNTER(reference, "references"); 174 PMAP_COUNTER(dereference, "dereferences"); 175 PMAP_COUNTER(destroy, "destroyed"); 176 PMAP_COUNTER(activate, "activations"); 177 PMAP_COUNTER(deactivate, "deactivations"); 178 PMAP_COUNTER(update, "updates"); 179 #ifdef MULTIPROCESSOR 180 PMAP_COUNTER(shootdown_ipis, "shootdown IPIs"); 181 #endif 182 PMAP_COUNTER(unwire, "unwires"); 183 PMAP_COUNTER(copy, "copies"); 184 PMAP_COUNTER(clear_modify, "clear_modifies"); 185 PMAP_COUNTER(protect, "protects"); 186 PMAP_COUNTER(page_protect, "page_protects"); 187 188 #define PMAP_ASID_RESERVED 0 189 CTASSERT(PMAP_ASID_RESERVED == 0); 190 191 #ifndef PMAP_SEGTAB_ALIGN 192 #define PMAP_SEGTAB_ALIGN /* nothing */ 193 #endif 194 #ifdef _LP64 195 pmap_segtab_t pmap_kstart_segtab PMAP_SEGTAB_ALIGN; /* first mid-level segtab for kernel */ 196 #endif 197 pmap_segtab_t pmap_kern_segtab PMAP_SEGTAB_ALIGN = { /* top level segtab for kernel */ 198 #ifdef _LP64 199 .seg_seg[(VM_MIN_KERNEL_ADDRESS & XSEGOFSET) >> SEGSHIFT] = &pmap_kstart_segtab, 200 #endif 201 }; 202 203 struct pmap_kernel kernel_pmap_store = { 204 .kernel_pmap = { 205 .pm_count = 1, 206 .pm_segtab = &pmap_kern_segtab, 207 .pm_minaddr = VM_MIN_KERNEL_ADDRESS, 208 .pm_maxaddr = VM_MAX_KERNEL_ADDRESS, 209 }, 210 }; 211 212 struct pmap * const kernel_pmap_ptr = &kernel_pmap_store.kernel_pmap; 213 214 struct pmap_limits pmap_limits = { /* VA and PA limits */ 215 .virtual_start = VM_MIN_KERNEL_ADDRESS, 216 }; 217 218 #ifdef UVMHIST 219 static struct kern_history_ent pmapexechistbuf[10000]; 220 static struct kern_history_ent pmaphistbuf[10000]; 221 UVMHIST_DEFINE(pmapexechist); 222 UVMHIST_DEFINE(pmaphist); 223 #endif 224 225 /* 226 * The pools from which pmap structures and sub-structures are allocated. 227 */ 228 struct pool pmap_pmap_pool; 229 struct pool pmap_pv_pool; 230 231 #ifndef PMAP_PV_LOWAT 232 #define PMAP_PV_LOWAT 16 233 #endif 234 int pmap_pv_lowat = PMAP_PV_LOWAT; 235 236 bool pmap_initialized = false; 237 #define PMAP_PAGE_COLOROK_P(a, b) \ 238 ((((int)(a) ^ (int)(b)) & pmap_page_colormask) == 0) 239 u_int pmap_page_colormask; 240 241 #define PAGE_IS_MANAGED(pa) (pmap_initialized && uvm_pageismanaged(pa)) 242 243 #define PMAP_IS_ACTIVE(pm) \ 244 ((pm) == pmap_kernel() || \ 245 (pm) == curlwp->l_proc->p_vmspace->vm_map.pmap) 246 247 /* Forward function declarations */ 248 void pmap_page_remove(struct vm_page *); 249 static void pmap_pvlist_check(struct vm_page_md *); 250 void pmap_remove_pv(pmap_t, vaddr_t, struct vm_page *, bool); 251 void pmap_enter_pv(pmap_t, vaddr_t, struct vm_page *, pt_entry_t *, u_int); 252 253 /* 254 * PV table management functions. 255 */ 256 void *pmap_pv_page_alloc(struct pool *, int); 257 void pmap_pv_page_free(struct pool *, void *); 258 259 struct pool_allocator pmap_pv_page_allocator = { 260 pmap_pv_page_alloc, pmap_pv_page_free, 0, 261 }; 262 263 #define pmap_pv_alloc() pool_get(&pmap_pv_pool, PR_NOWAIT) 264 #define pmap_pv_free(pv) pool_put(&pmap_pv_pool, (pv)) 265 266 #if !defined(MULTIPROCESSOR) || !defined(PMAP_MD_NEED_TLB_MISS_LOCK) 267 #define pmap_md_tlb_miss_lock_enter() do { } while(/*CONSTCOND*/0) 268 #define pmap_md_tlb_miss_lock_exit() do { } while(/*CONSTCOND*/0) 269 #endif /* !MULTIPROCESSOR || !PMAP_MD_NEED_TLB_MISS_LOCK */ 270 271 #ifndef MULTIPROCESSOR 272 kmutex_t pmap_pvlist_mutex __cacheline_aligned; 273 #endif 274 275 /* 276 * Debug functions. 277 */ 278 279 static inline void 280 pmap_asid_check(pmap_t pm, const char *func) 281 { 282 #ifdef DEBUG 283 if (!PMAP_IS_ACTIVE(pm)) 284 return; 285 286 struct pmap_asid_info * const pai = PMAP_PAI(pm, cpu_tlb_info(curcpu())); 287 tlb_asid_t asid = tlb_get_asid(); 288 if (asid != pai->pai_asid) 289 panic("%s: inconsistency for active TLB update: %u <-> %u", 290 func, asid, pai->pai_asid); 291 #endif 292 } 293 294 static void 295 pmap_addr_range_check(pmap_t pmap, vaddr_t sva, vaddr_t eva, const char *func) 296 { 297 #ifdef DEBUG 298 if (pmap == pmap_kernel()) { 299 if (sva < VM_MIN_KERNEL_ADDRESS) 300 panic("%s: kva %#"PRIxVADDR" not in range", 301 func, sva); 302 if (eva >= pmap_limits.virtual_end) 303 panic("%s: kva %#"PRIxVADDR" not in range", 304 func, eva); 305 } else { 306 if (eva > VM_MAXUSER_ADDRESS) 307 panic("%s: uva %#"PRIxVADDR" not in range", 308 func, eva); 309 pmap_asid_check(pmap, func); 310 } 311 #endif 312 } 313 314 /* 315 * Misc. functions. 316 */ 317 318 bool 319 pmap_page_clear_attributes(struct vm_page_md *mdpg, u_int clear_attributes) 320 { 321 volatile unsigned long * const attrp = &mdpg->mdpg_attrs; 322 #ifdef MULTIPROCESSOR 323 for (;;) { 324 u_int old_attr = *attrp; 325 if ((old_attr & clear_attributes) == 0) 326 return false; 327 u_int new_attr = old_attr & ~clear_attributes; 328 if (old_attr == atomic_cas_ulong(attrp, old_attr, new_attr)) 329 return true; 330 } 331 #else 332 unsigned long old_attr = *attrp; 333 if ((old_attr & clear_attributes) == 0) 334 return false; 335 *attrp &= ~clear_attributes; 336 return true; 337 #endif 338 } 339 340 void 341 pmap_page_set_attributes(struct vm_page_md *mdpg, u_int set_attributes) 342 { 343 #ifdef MULTIPROCESSOR 344 atomic_or_ulong(&mdpg->mdpg_attrs, set_attributes); 345 #else 346 mdpg->mdpg_attrs |= set_attributes; 347 #endif 348 } 349 350 static void 351 pmap_page_syncicache(struct vm_page *pg) 352 { 353 #ifndef MULTIPROCESSOR 354 struct pmap * const curpmap = curlwp->l_proc->p_vmspace->vm_map.pmap; 355 #endif 356 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 357 pv_entry_t pv = &mdpg->mdpg_first; 358 kcpuset_t *onproc; 359 #ifdef MULTIPROCESSOR 360 kcpuset_create(&onproc, true); 361 KASSERT(onproc != NULL); 362 #else 363 onproc = NULL; 364 #endif 365 VM_PAGEMD_PVLIST_READLOCK(mdpg); 366 pmap_pvlist_check(mdpg); 367 368 if (pv->pv_pmap != NULL) { 369 for (; pv != NULL; pv = pv->pv_next) { 370 #ifdef MULTIPROCESSOR 371 kcpuset_merge(onproc, pv->pv_pmap->pm_onproc); 372 if (kcpuset_match(onproc, kcpuset_running)) { 373 break; 374 } 375 #else 376 if (pv->pv_pmap == curpmap) { 377 onproc = curcpu()->ci_data.cpu_kcpuset; 378 break; 379 } 380 #endif 381 } 382 } 383 pmap_pvlist_check(mdpg); 384 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 385 kpreempt_disable(); 386 pmap_md_page_syncicache(pg, onproc); 387 kpreempt_enable(); 388 #ifdef MULTIPROCESSOR 389 kcpuset_destroy(onproc); 390 #endif 391 } 392 393 /* 394 * Define the initial bounds of the kernel virtual address space. 395 */ 396 void 397 pmap_virtual_space(vaddr_t *vstartp, vaddr_t *vendp) 398 { 399 400 *vstartp = pmap_limits.virtual_start; 401 *vendp = pmap_limits.virtual_end; 402 } 403 404 vaddr_t 405 pmap_growkernel(vaddr_t maxkvaddr) 406 { 407 vaddr_t virtual_end = pmap_limits.virtual_end; 408 maxkvaddr = pmap_round_seg(maxkvaddr) - 1; 409 410 /* 411 * Reserve PTEs for the new KVA space. 412 */ 413 for (; virtual_end < maxkvaddr; virtual_end += NBSEG) { 414 pmap_pte_reserve(pmap_kernel(), virtual_end, 0); 415 } 416 417 /* 418 * Don't exceed VM_MAX_KERNEL_ADDRESS! 419 */ 420 if (virtual_end == 0 || virtual_end > VM_MAX_KERNEL_ADDRESS) 421 virtual_end = VM_MAX_KERNEL_ADDRESS; 422 423 /* 424 * Update new end. 425 */ 426 pmap_limits.virtual_end = virtual_end; 427 return virtual_end; 428 } 429 430 /* 431 * Bootstrap memory allocator (alternative to vm_bootstrap_steal_memory()). 432 * This function allows for early dynamic memory allocation until the virtual 433 * memory system has been bootstrapped. After that point, either kmem_alloc 434 * or malloc should be used. This function works by stealing pages from the 435 * (to be) managed page pool, then implicitly mapping the pages (by using 436 * their k0seg addresses) and zeroing them. 437 * 438 * It may be used once the physical memory segments have been pre-loaded 439 * into the vm_physmem[] array. Early memory allocation MUST use this 440 * interface! This cannot be used after vm_page_startup(), and will 441 * generate a panic if tried. 442 * 443 * Note that this memory will never be freed, and in essence it is wired 444 * down. 445 * 446 * We must adjust *vstartp and/or *vendp iff we use address space 447 * from the kernel virtual address range defined by pmap_virtual_space(). 448 */ 449 vaddr_t 450 pmap_steal_memory(vsize_t size, vaddr_t *vstartp, vaddr_t *vendp) 451 { 452 size_t npgs; 453 paddr_t pa; 454 vaddr_t va; 455 struct vm_physseg *maybe_seg = NULL; 456 u_int maybe_bank = vm_nphysseg; 457 458 size = round_page(size); 459 npgs = atop(size); 460 461 aprint_debug("%s: need %zu pages\n", __func__, npgs); 462 463 for (u_int bank = 0; bank < vm_nphysseg; bank++) { 464 struct vm_physseg * const seg = VM_PHYSMEM_PTR(bank); 465 if (uvm.page_init_done == true) 466 panic("pmap_steal_memory: called _after_ bootstrap"); 467 468 aprint_debug("%s: seg %u: %#"PRIxPADDR" %#"PRIxPADDR" %#"PRIxPADDR" %#"PRIxPADDR"\n", 469 __func__, bank, 470 seg->avail_start, seg->start, 471 seg->avail_end, seg->end); 472 473 if (seg->avail_start != seg->start 474 || seg->avail_start >= seg->avail_end) { 475 aprint_debug("%s: seg %u: bad start\n", __func__, bank); 476 continue; 477 } 478 479 if (seg->avail_end - seg->avail_start < npgs) { 480 aprint_debug("%s: seg %u: too small for %zu pages\n", 481 __func__, bank, npgs); 482 continue; 483 } 484 485 if (!pmap_md_ok_to_steal_p(seg, npgs)) { 486 continue; 487 } 488 489 /* 490 * Always try to allocate from the segment with the least 491 * amount of space left. 492 */ 493 #define VM_PHYSMEM_SPACE(s) ((s)->avail_end - (s)->avail_start) 494 if (maybe_seg == NULL 495 || VM_PHYSMEM_SPACE(seg) < VM_PHYSMEM_SPACE(maybe_seg)) { 496 maybe_seg = seg; 497 maybe_bank = bank; 498 } 499 } 500 501 if (maybe_seg) { 502 struct vm_physseg * const seg = maybe_seg; 503 u_int bank = maybe_bank; 504 505 /* 506 * There are enough pages here; steal them! 507 */ 508 pa = ptoa(seg->avail_start); 509 seg->avail_start += npgs; 510 seg->start += npgs; 511 512 /* 513 * Have we used up this segment? 514 */ 515 if (seg->avail_start == seg->end) { 516 if (vm_nphysseg == 1) 517 panic("pmap_steal_memory: out of memory!"); 518 519 aprint_debug("%s: seg %u: %zu pages stolen (removed)\n", 520 __func__, bank, npgs); 521 /* Remove this segment from the list. */ 522 vm_nphysseg--; 523 for (u_int x = bank; x < vm_nphysseg; x++) { 524 /* structure copy */ 525 VM_PHYSMEM_PTR_SWAP(x, x + 1); 526 } 527 } else { 528 aprint_debug("%s: seg %u: %zu pages stolen (%#"PRIxPADDR" left)\n", 529 __func__, bank, npgs, VM_PHYSMEM_SPACE(seg)); 530 } 531 532 va = pmap_md_map_poolpage(pa, size); 533 memset((void *)va, 0, size); 534 return va; 535 } 536 537 /* 538 * If we got here, there was no memory left. 539 */ 540 panic("pmap_steal_memory: no memory to steal %zu pages", npgs); 541 } 542 543 /* 544 * Initialize the pmap module. 545 * Called by vm_init, to initialize any structures that the pmap 546 * system needs to map virtual memory. 547 */ 548 void 549 pmap_init(void) 550 { 551 UVMHIST_INIT_STATIC(pmapexechist, pmapexechistbuf); 552 UVMHIST_INIT_STATIC(pmaphist, pmaphistbuf); 553 554 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 555 556 /* 557 * Initialize the segtab lock. 558 */ 559 mutex_init(&pmap_segtab_lock, MUTEX_DEFAULT, IPL_HIGH); 560 561 /* 562 * Set a low water mark on the pv_entry pool, so that we are 563 * more likely to have these around even in extreme memory 564 * starvation. 565 */ 566 pool_setlowat(&pmap_pv_pool, pmap_pv_lowat); 567 568 /* 569 * Set the page colormask but allow pmap_md_init to override it. 570 */ 571 pmap_page_colormask = ptoa(uvmexp.colormask); 572 573 pmap_md_init(); 574 575 /* 576 * Now it is safe to enable pv entry recording. 577 */ 578 pmap_initialized = true; 579 } 580 581 /* 582 * Create and return a physical map. 583 * 584 * If the size specified for the map 585 * is zero, the map is an actual physical 586 * map, and may be referenced by the 587 * hardware. 588 * 589 * If the size specified is non-zero, 590 * the map will be used in software only, and 591 * is bounded by that size. 592 */ 593 pmap_t 594 pmap_create(void) 595 { 596 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 597 PMAP_COUNT(create); 598 599 pmap_t pmap = pool_get(&pmap_pmap_pool, PR_WAITOK); 600 memset(pmap, 0, PMAP_SIZE); 601 602 KASSERT(pmap->pm_pai[0].pai_link.le_prev == NULL); 603 604 pmap->pm_count = 1; 605 pmap->pm_minaddr = VM_MIN_ADDRESS; 606 pmap->pm_maxaddr = VM_MAXUSER_ADDRESS; 607 608 pmap_segtab_init(pmap); 609 610 #ifdef MULTIPROCESSOR 611 kcpuset_create(&pmap->pm_active, true); 612 kcpuset_create(&pmap->pm_onproc, true); 613 KASSERT(pmap->pm_active != NULL); 614 KASSERT(pmap->pm_onproc != NULL); 615 #endif 616 617 UVMHIST_LOG(pmaphist, " <-- done (pmap=%p)", pmap, 0, 0, 0); 618 619 return pmap; 620 } 621 622 /* 623 * Retire the given physical map from service. 624 * Should only be called if the map contains 625 * no valid mappings. 626 */ 627 void 628 pmap_destroy(pmap_t pmap) 629 { 630 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 631 UVMHIST_LOG(pmaphist, "(pmap=%p)", pmap, 0, 0, 0); 632 633 if (atomic_dec_uint_nv(&pmap->pm_count) > 0) { 634 PMAP_COUNT(dereference); 635 UVMHIST_LOG(pmaphist, " <-- done (deref)", 0, 0, 0, 0); 636 return; 637 } 638 639 PMAP_COUNT(destroy); 640 KASSERT(pmap->pm_count == 0); 641 kpreempt_disable(); 642 pmap_md_tlb_miss_lock_enter(); 643 pmap_tlb_asid_release_all(pmap); 644 pmap_segtab_destroy(pmap, NULL, 0); 645 pmap_md_tlb_miss_lock_exit(); 646 647 #ifdef MULTIPROCESSOR 648 kcpuset_destroy(pmap->pm_active); 649 kcpuset_destroy(pmap->pm_onproc); 650 pmap->pm_active = NULL; 651 pmap->pm_onproc = NULL; 652 #endif 653 654 pool_put(&pmap_pmap_pool, pmap); 655 kpreempt_enable(); 656 657 UVMHIST_LOG(pmaphist, " <-- done (freed)", 0, 0, 0, 0); 658 } 659 660 /* 661 * Add a reference to the specified pmap. 662 */ 663 void 664 pmap_reference(pmap_t pmap) 665 { 666 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 667 UVMHIST_LOG(pmaphist, "(pmap=%p)", pmap, 0, 0, 0); 668 PMAP_COUNT(reference); 669 670 if (pmap != NULL) { 671 atomic_inc_uint(&pmap->pm_count); 672 } 673 674 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 675 } 676 677 /* 678 * Make a new pmap (vmspace) active for the given process. 679 */ 680 void 681 pmap_activate(struct lwp *l) 682 { 683 pmap_t pmap = l->l_proc->p_vmspace->vm_map.pmap; 684 685 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 686 UVMHIST_LOG(pmaphist, "(l=%p (pmap=%p))", l, pmap, 0, 0); 687 PMAP_COUNT(activate); 688 689 kpreempt_disable(); 690 pmap_md_tlb_miss_lock_enter(); 691 pmap_tlb_asid_acquire(pmap, l); 692 if (l == curlwp) { 693 pmap_segtab_activate(pmap, l); 694 } 695 pmap_md_tlb_miss_lock_exit(); 696 kpreempt_enable(); 697 698 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 699 } 700 701 /* 702 * Remove this page from all physical maps in which it resides. 703 * Reflects back modify bits to the pager. 704 */ 705 void 706 pmap_page_remove(struct vm_page *pg) 707 { 708 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 709 710 kpreempt_disable(); 711 VM_PAGEMD_PVLIST_LOCK(mdpg); 712 pmap_pvlist_check(mdpg); 713 714 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 715 716 pv_entry_t pv = &mdpg->mdpg_first; 717 if (pv->pv_pmap == NULL) { 718 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 719 kpreempt_enable(); 720 UVMHIST_LOG(pmaphist, " <-- done (empty)", 0, 0, 0, 0); 721 return; 722 } 723 724 pv_entry_t npv; 725 pv_entry_t pvp = NULL; 726 727 for (; pv != NULL; pv = npv) { 728 npv = pv->pv_next; 729 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 730 if (pv->pv_va & PV_KENTER) { 731 UVMHIST_LOG(pmaphist, " pv %p pmap %p va %" 732 PRIxVADDR" skip", pv, pv->pv_pmap, pv->pv_va, 0); 733 734 KASSERT(pv->pv_pmap == pmap_kernel()); 735 736 /* Assume no more - it'll get fixed if there are */ 737 pv->pv_next = NULL; 738 739 /* 740 * pvp is non-null when we already have a PV_KENTER 741 * pv in pvh_first; otherwise we haven't seen a 742 * PV_KENTER pv and we need to copy this one to 743 * pvh_first 744 */ 745 if (pvp) { 746 /* 747 * The previous PV_KENTER pv needs to point to 748 * this PV_KENTER pv 749 */ 750 pvp->pv_next = pv; 751 } else { 752 pv_entry_t fpv = &mdpg->mdpg_first; 753 *fpv = *pv; 754 KASSERT(fpv->pv_pmap == pmap_kernel()); 755 } 756 pvp = pv; 757 continue; 758 } 759 #endif 760 const pmap_t pmap = pv->pv_pmap; 761 vaddr_t va = trunc_page(pv->pv_va); 762 pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 763 KASSERTMSG(ptep != NULL, "%#"PRIxVADDR " %#"PRIxVADDR, va, 764 pmap_limits.virtual_end); 765 pt_entry_t pte = *ptep; 766 UVMHIST_LOG(pmaphist, " pv %p pmap %p va %"PRIxVADDR 767 " pte %#"PRIxPTE, pv, pmap, va, pte_value(pte)); 768 if (!pte_valid_p(pte)) 769 continue; 770 const bool is_kernel_pmap_p = (pmap == pmap_kernel()); 771 if (is_kernel_pmap_p) { 772 PMAP_COUNT(remove_kernel_pages); 773 } else { 774 PMAP_COUNT(remove_user_pages); 775 } 776 if (pte_wired_p(pte)) 777 pmap->pm_stats.wired_count--; 778 pmap->pm_stats.resident_count--; 779 780 pmap_md_tlb_miss_lock_enter(); 781 const pt_entry_t npte = pte_nv_entry(is_kernel_pmap_p); 782 *ptep = npte; 783 /* 784 * Flush the TLB for the given address. 785 */ 786 pmap_tlb_invalidate_addr(pmap, va); 787 pmap_md_tlb_miss_lock_exit(); 788 789 /* 790 * non-null means this is a non-pvh_first pv, so we should 791 * free it. 792 */ 793 if (pvp) { 794 KASSERT(pvp->pv_pmap == pmap_kernel()); 795 KASSERT(pvp->pv_next == NULL); 796 pmap_pv_free(pv); 797 } else { 798 pv->pv_pmap = NULL; 799 pv->pv_next = NULL; 800 } 801 } 802 803 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 804 pmap_page_clear_attributes(mdpg, VM_PAGEMD_UNCACHED); 805 #endif 806 pmap_pvlist_check(mdpg); 807 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 808 kpreempt_enable(); 809 810 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 811 } 812 813 814 /* 815 * Make a previously active pmap (vmspace) inactive. 816 */ 817 void 818 pmap_deactivate(struct lwp *l) 819 { 820 pmap_t pmap = l->l_proc->p_vmspace->vm_map.pmap; 821 822 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 823 UVMHIST_LOG(pmaphist, "(l=%p (pmap=%p))", l, pmap, 0, 0); 824 PMAP_COUNT(deactivate); 825 826 kpreempt_disable(); 827 KASSERT(l == curlwp || l->l_cpu == curlwp->l_cpu); 828 pmap_md_tlb_miss_lock_enter(); 829 curcpu()->ci_pmap_user_segtab = PMAP_INVALID_SEGTAB_ADDRESS; 830 #ifdef _LP64 831 curcpu()->ci_pmap_user_seg0tab = NULL; 832 #endif 833 pmap_tlb_asid_deactivate(pmap); 834 pmap_md_tlb_miss_lock_exit(); 835 kpreempt_enable(); 836 837 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 838 } 839 840 void 841 pmap_update(struct pmap *pmap) 842 { 843 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 844 UVMHIST_LOG(pmaphist, "(pmap=%p)", pmap, 0, 0, 0); 845 PMAP_COUNT(update); 846 847 kpreempt_disable(); 848 #if defined(MULTIPROCESSOR) && defined(PMAP_NEED_TLB_SHOOTDOWN) 849 u_int pending = atomic_swap_uint(&pmap->pm_shootdown_pending, 0); 850 if (pending && pmap_tlb_shootdown_bystanders(pmap)) 851 PMAP_COUNT(shootdown_ipis); 852 #endif 853 pmap_md_tlb_miss_lock_enter(); 854 #if defined(DEBUG) && !defined(MULTIPROCESSOR) 855 pmap_tlb_check(pmap, pmap_md_tlb_check_entry); 856 #endif /* DEBUG */ 857 858 /* 859 * If pmap_remove_all was called, we deactivated ourselves and nuked 860 * our ASID. Now we have to reactivate ourselves. 861 */ 862 if (__predict_false(pmap->pm_flags & PMAP_DEFERRED_ACTIVATE)) { 863 pmap->pm_flags ^= PMAP_DEFERRED_ACTIVATE; 864 pmap_tlb_asid_acquire(pmap, curlwp); 865 pmap_segtab_activate(pmap, curlwp); 866 } 867 pmap_md_tlb_miss_lock_exit(); 868 kpreempt_enable(); 869 870 UVMHIST_LOG(pmaphist, " <-- done%s", 871 (pmap == pmap_kernel()) ? " (kernel)" : "", 0, 0, 0); 872 } 873 874 /* 875 * Remove the given range of addresses from the specified map. 876 * 877 * It is assumed that the start and end are properly 878 * rounded to the page size. 879 */ 880 881 static bool 882 pmap_pte_remove(pmap_t pmap, vaddr_t sva, vaddr_t eva, pt_entry_t *ptep, 883 uintptr_t flags) 884 { 885 const pt_entry_t npte = flags; 886 const bool is_kernel_pmap_p = (pmap == pmap_kernel()); 887 888 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 889 UVMHIST_LOG(pmaphist, "(pmap=%p %sva=%#"PRIxVADDR"..%#"PRIxVADDR, 890 pmap, (is_kernel_pmap_p ? "(kernel) " : ""), sva, eva); 891 UVMHIST_LOG(pmaphist, "ptep=%p, flags(npte)=%#"PRIxPTR")", 892 ptep, flags, 0, 0); 893 894 KASSERT(kpreempt_disabled()); 895 896 for (; sva < eva; sva += NBPG, ptep++) { 897 const pt_entry_t pte = *ptep; 898 if (!pte_valid_p(pte)) 899 continue; 900 if (is_kernel_pmap_p) { 901 PMAP_COUNT(remove_kernel_pages); 902 } else { 903 PMAP_COUNT(remove_user_pages); 904 } 905 if (pte_wired_p(pte)) 906 pmap->pm_stats.wired_count--; 907 pmap->pm_stats.resident_count--; 908 struct vm_page * const pg = PHYS_TO_VM_PAGE(pte_to_paddr(pte)); 909 if (__predict_true(pg != NULL)) { 910 pmap_remove_pv(pmap, sva, pg, pte_modified_p(pte)); 911 } 912 pmap_md_tlb_miss_lock_enter(); 913 *ptep = npte; 914 /* 915 * Flush the TLB for the given address. 916 */ 917 pmap_tlb_invalidate_addr(pmap, sva); 918 pmap_md_tlb_miss_lock_exit(); 919 } 920 921 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 922 923 return false; 924 } 925 926 void 927 pmap_remove(pmap_t pmap, vaddr_t sva, vaddr_t eva) 928 { 929 const bool is_kernel_pmap_p = (pmap == pmap_kernel()); 930 const pt_entry_t npte = pte_nv_entry(is_kernel_pmap_p); 931 932 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 933 UVMHIST_LOG(pmaphist, "(pmap=%p, va=%#"PRIxVADDR"..%#"PRIxVADDR")", 934 pmap, sva, eva, 0); 935 936 if (is_kernel_pmap_p) { 937 PMAP_COUNT(remove_kernel_calls); 938 } else { 939 PMAP_COUNT(remove_user_calls); 940 } 941 #ifdef PMAP_FAULTINFO 942 curpcb->pcb_faultinfo.pfi_faultaddr = 0; 943 curpcb->pcb_faultinfo.pfi_repeats = 0; 944 curpcb->pcb_faultinfo.pfi_faultpte = NULL; 945 #endif 946 kpreempt_disable(); 947 pmap_addr_range_check(pmap, sva, eva, __func__); 948 pmap_pte_process(pmap, sva, eva, pmap_pte_remove, npte); 949 kpreempt_enable(); 950 951 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 952 } 953 954 /* 955 * pmap_page_protect: 956 * 957 * Lower the permission for all mappings to a given page. 958 */ 959 void 960 pmap_page_protect(struct vm_page *pg, vm_prot_t prot) 961 { 962 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 963 pv_entry_t pv; 964 vaddr_t va; 965 966 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 967 UVMHIST_LOG(pmaphist, "(pg=%p (pa %#"PRIxPADDR") prot=%#x)", 968 pg, VM_PAGE_TO_PHYS(pg), prot, 0); 969 PMAP_COUNT(page_protect); 970 971 switch (prot) { 972 case VM_PROT_READ|VM_PROT_WRITE: 973 case VM_PROT_ALL: 974 break; 975 976 /* copy_on_write */ 977 case VM_PROT_READ: 978 case VM_PROT_READ|VM_PROT_EXECUTE: 979 pv = &mdpg->mdpg_first; 980 kpreempt_disable(); 981 VM_PAGEMD_PVLIST_READLOCK(mdpg); 982 pmap_pvlist_check(mdpg); 983 /* 984 * Loop over all current mappings setting/clearing as apropos. 985 */ 986 if (pv->pv_pmap != NULL) { 987 while (pv != NULL) { 988 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 989 if (pv->pv_va & PV_KENTER) { 990 pv = pv->pv_next; 991 continue; 992 } 993 #endif 994 const pmap_t pmap = pv->pv_pmap; 995 va = trunc_page(pv->pv_va); 996 const uintptr_t gen = 997 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 998 pmap_protect(pmap, va, va + PAGE_SIZE, prot); 999 KASSERT(pv->pv_pmap == pmap); 1000 pmap_update(pmap); 1001 if (gen != VM_PAGEMD_PVLIST_READLOCK(mdpg)) { 1002 pv = &mdpg->mdpg_first; 1003 } else { 1004 pv = pv->pv_next; 1005 } 1006 pmap_pvlist_check(mdpg); 1007 } 1008 } 1009 pmap_pvlist_check(mdpg); 1010 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1011 kpreempt_enable(); 1012 break; 1013 1014 /* remove_all */ 1015 default: 1016 pmap_page_remove(pg); 1017 } 1018 1019 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1020 } 1021 1022 static bool 1023 pmap_pte_protect(pmap_t pmap, vaddr_t sva, vaddr_t eva, pt_entry_t *ptep, 1024 uintptr_t flags) 1025 { 1026 const vm_prot_t prot = (flags & VM_PROT_ALL); 1027 1028 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1029 UVMHIST_LOG(pmaphist, "(pmap=%p %sva=%#"PRIxVADDR"..%#"PRIxVADDR, 1030 pmap, (pmap == pmap_kernel() ? "(kernel) " : ""), sva, eva); 1031 UVMHIST_LOG(pmaphist, "ptep=%p, flags(npte)=%#"PRIxPTR")", 1032 ptep, flags, 0, 0); 1033 1034 KASSERT(kpreempt_disabled()); 1035 /* 1036 * Change protection on every valid mapping within this segment. 1037 */ 1038 for (; sva < eva; sva += NBPG, ptep++) { 1039 pt_entry_t pte = *ptep; 1040 if (!pte_valid_p(pte)) 1041 continue; 1042 struct vm_page * const pg = PHYS_TO_VM_PAGE(pte_to_paddr(pte)); 1043 if (pg != NULL && pte_modified_p(pte)) { 1044 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1045 if (VM_PAGEMD_EXECPAGE_P(mdpg)) { 1046 KASSERT(mdpg->mdpg_first.pv_pmap != NULL); 1047 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1048 if (VM_PAGEMD_CACHED_P(mdpg)) { 1049 #endif 1050 UVMHIST_LOG(pmapexechist, 1051 "pg %p (pa %#"PRIxPADDR"): %s", 1052 pg, VM_PAGE_TO_PHYS(pg), 1053 "syncicached performed", 0); 1054 pmap_page_syncicache(pg); 1055 PMAP_COUNT(exec_synced_protect); 1056 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1057 } 1058 #endif 1059 } 1060 } 1061 pte = pte_prot_downgrade(pte, prot); 1062 if (*ptep != pte) { 1063 pmap_md_tlb_miss_lock_enter(); 1064 *ptep = pte; 1065 /* 1066 * Update the TLB if needed. 1067 */ 1068 pmap_tlb_update_addr(pmap, sva, pte, PMAP_TLB_NEED_IPI); 1069 pmap_md_tlb_miss_lock_exit(); 1070 } 1071 } 1072 1073 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1074 1075 return false; 1076 } 1077 1078 /* 1079 * Set the physical protection on the 1080 * specified range of this map as requested. 1081 */ 1082 void 1083 pmap_protect(pmap_t pmap, vaddr_t sva, vaddr_t eva, vm_prot_t prot) 1084 { 1085 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1086 UVMHIST_LOG(pmaphist, 1087 "(pmap=%p, va=%#"PRIxVADDR"..%#"PRIxVADDR", prot=%u)", 1088 pmap, sva, eva, prot); 1089 PMAP_COUNT(protect); 1090 1091 if ((prot & VM_PROT_READ) == VM_PROT_NONE) { 1092 pmap_remove(pmap, sva, eva); 1093 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1094 return; 1095 } 1096 1097 /* 1098 * Change protection on every valid mapping within this segment. 1099 */ 1100 kpreempt_disable(); 1101 pmap_addr_range_check(pmap, sva, eva, __func__); 1102 pmap_pte_process(pmap, sva, eva, pmap_pte_protect, prot); 1103 kpreempt_enable(); 1104 1105 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1106 } 1107 1108 #if defined(PMAP_VIRTUAL_CACHE_ALIASES) && !defined(PMAP_NO_PV_UNCACHED) 1109 /* 1110 * pmap_page_cache: 1111 * 1112 * Change all mappings of a managed page to cached/uncached. 1113 */ 1114 void 1115 pmap_page_cache(struct vm_page *pg, bool cached) 1116 { 1117 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1118 1119 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1120 UVMHIST_LOG(pmaphist, "(pg=%p (pa %#"PRIxPADDR") cached=%s)", 1121 pg, VM_PAGE_TO_PHYS(pg), cached ? "true" : "false", 0); 1122 1123 KASSERT(kpreempt_disabled()); 1124 KASSERT(VM_PAGEMD_PVLIST_LOCKED_P(mdpg)); 1125 1126 if (cached) { 1127 pmap_page_clear_attributes(mdpg, VM_PAGEMD_UNCACHED); 1128 PMAP_COUNT(page_cache_restorations); 1129 } else { 1130 pmap_page_set_attributes(mdpg, VM_PAGEMD_UNCACHED); 1131 PMAP_COUNT(page_cache_evictions); 1132 } 1133 1134 for (pv_entry_t pv = &mdpg->mdpg_first; pv != NULL; pv = pv->pv_next) { 1135 pmap_t pmap = pv->pv_pmap; 1136 vaddr_t va = trunc_page(pv->pv_va); 1137 1138 KASSERT(pmap != NULL); 1139 KASSERT(pmap != pmap_kernel() || !pmap_md_direct_mapped_vaddr_p(va)); 1140 pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 1141 if (ptep == NULL) 1142 continue; 1143 pt_entry_t pte = *ptep; 1144 if (pte_valid_p(pte)) { 1145 pte = pte_cached_change(pte, cached); 1146 pmap_md_tlb_miss_lock_enter(); 1147 *ptep = pte; 1148 pmap_tlb_update_addr(pmap, va, pte, PMAP_TLB_NEED_IPI); 1149 pmap_md_tlb_miss_lock_exit(); 1150 } 1151 } 1152 1153 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1154 } 1155 #endif /* PMAP_VIRTUAL_CACHE_ALIASES && !PMAP_NO_PV_UNCACHED */ 1156 1157 /* 1158 * Insert the given physical page (p) at 1159 * the specified virtual address (v) in the 1160 * target physical map with the protection requested. 1161 * 1162 * If specified, the page will be wired down, meaning 1163 * that the related pte can not be reclaimed. 1164 * 1165 * NB: This is the only routine which MAY NOT lazy-evaluate 1166 * or lose information. That is, this routine must actually 1167 * insert this page into the given map NOW. 1168 */ 1169 int 1170 pmap_enter(pmap_t pmap, vaddr_t va, paddr_t pa, vm_prot_t prot, u_int flags) 1171 { 1172 const bool wired = (flags & PMAP_WIRED) != 0; 1173 const bool is_kernel_pmap_p = (pmap == pmap_kernel()); 1174 u_int update_flags = (flags & VM_PROT_ALL) != 0 ? PMAP_TLB_INSERT : 0; 1175 #ifdef UVMHIST 1176 struct kern_history * const histp = 1177 ((prot & VM_PROT_EXECUTE) ? &pmapexechist : &pmaphist); 1178 #endif 1179 1180 UVMHIST_FUNC(__func__); UVMHIST_CALLED(*histp); 1181 #define VM_PROT_STRING(prot) \ 1182 &"\0 " \ 1183 "(R)\0 " \ 1184 "(W)\0 " \ 1185 "(RW)\0 " \ 1186 "(X)\0 " \ 1187 "(RX)\0 " \ 1188 "(WX)\0 " \ 1189 "(RWX)\0"[UVM_PROTECTION(prot)*6] 1190 UVMHIST_LOG(*histp, "(pmap=%p, va=%#"PRIxVADDR", pa=%#"PRIxPADDR, 1191 UVMHIST_LOG(*histp, "(pmap=%p, va=%#"PRIxVADDR", pa=%#"PRIxPADDR, 1192 pmap, va, pa, 0); 1193 UVMHIST_LOG(*histp, "prot=%#x%s flags=%#x%s)", 1194 prot, VM_PROT_STRING(prot), flags, VM_PROT_STRING(flags)); 1195 1196 const bool good_color = PMAP_PAGE_COLOROK_P(pa, va); 1197 if (is_kernel_pmap_p) { 1198 PMAP_COUNT(kernel_mappings); 1199 if (!good_color) 1200 PMAP_COUNT(kernel_mappings_bad); 1201 } else { 1202 PMAP_COUNT(user_mappings); 1203 if (!good_color) 1204 PMAP_COUNT(user_mappings_bad); 1205 } 1206 pmap_addr_range_check(pmap, va, va, __func__); 1207 1208 KASSERTMSG(prot & VM_PROT_READ, "no READ (%#x) in prot %#x", 1209 VM_PROT_READ, prot); 1210 1211 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 1212 struct vm_page_md * const mdpg = (pg ? VM_PAGE_TO_MD(pg) : NULL); 1213 1214 if (pg) { 1215 /* Set page referenced/modified status based on flags */ 1216 if (flags & VM_PROT_WRITE) { 1217 pmap_page_set_attributes(mdpg, VM_PAGEMD_MODIFIED|VM_PAGEMD_REFERENCED); 1218 } else if (flags & VM_PROT_ALL) { 1219 pmap_page_set_attributes(mdpg, VM_PAGEMD_REFERENCED); 1220 } 1221 1222 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1223 if (!VM_PAGEMD_CACHED_P(mdpg)) { 1224 flags |= PMAP_NOCACHE; 1225 PMAP_COUNT(uncached_mappings); 1226 } 1227 #endif 1228 1229 PMAP_COUNT(managed_mappings); 1230 } else { 1231 /* 1232 * Assumption: if it is not part of our managed memory 1233 * then it must be device memory which may be volatile. 1234 */ 1235 if ((flags & PMAP_CACHE_MASK) == 0) 1236 flags |= PMAP_NOCACHE; 1237 PMAP_COUNT(unmanaged_mappings); 1238 } 1239 1240 pt_entry_t npte = pte_make_enter(pa, mdpg, prot, flags, 1241 is_kernel_pmap_p); 1242 1243 kpreempt_disable(); 1244 1245 pt_entry_t * const ptep = pmap_pte_reserve(pmap, va, flags); 1246 if (__predict_false(ptep == NULL)) { 1247 kpreempt_enable(); 1248 UVMHIST_LOG(*histp, " <-- ENOMEM", 0, 0, 0, 0); 1249 return ENOMEM; 1250 } 1251 const pt_entry_t opte = *ptep; 1252 1253 /* Done after case that may sleep/return. */ 1254 if (pg) 1255 pmap_enter_pv(pmap, va, pg, &npte, 0); 1256 1257 /* 1258 * Now validate mapping with desired protection/wiring. 1259 * Assume uniform modified and referenced status for all 1260 * MIPS pages in a MACH page. 1261 */ 1262 if (wired) { 1263 pmap->pm_stats.wired_count++; 1264 npte = pte_wire_entry(npte); 1265 } 1266 1267 UVMHIST_LOG(*histp, "new pte %#"PRIxPTE" (pa %#"PRIxPADDR")", 1268 pte_value(npte), pa, 0, 0); 1269 1270 if (pte_valid_p(opte) && pte_to_paddr(opte) != pa) { 1271 pmap_remove(pmap, va, va + NBPG); 1272 PMAP_COUNT(user_mappings_changed); 1273 } 1274 1275 KASSERT(pte_valid_p(npte)); 1276 const bool resident = pte_valid_p(opte); 1277 if (resident) { 1278 update_flags |= PMAP_TLB_NEED_IPI; 1279 } else { 1280 pmap->pm_stats.resident_count++; 1281 } 1282 1283 pmap_md_tlb_miss_lock_enter(); 1284 *ptep = npte; 1285 pmap_tlb_update_addr(pmap, va, npte, update_flags); 1286 pmap_md_tlb_miss_lock_exit(); 1287 kpreempt_enable(); 1288 1289 if (pg != NULL && (prot == (VM_PROT_READ | VM_PROT_EXECUTE))) { 1290 KASSERT(mdpg != NULL); 1291 PMAP_COUNT(exec_mappings); 1292 if (!VM_PAGEMD_EXECPAGE_P(mdpg) && pte_cached_p(npte)) { 1293 if (!pte_deferred_exec_p(npte)) { 1294 UVMHIST_LOG(*histp, 1295 "va=%#"PRIxVADDR" pg %p: %s syncicache%s", 1296 va, pg, "immediate", ""); 1297 pmap_page_syncicache(pg); 1298 pmap_page_set_attributes(mdpg, 1299 VM_PAGEMD_EXECPAGE); 1300 PMAP_COUNT(exec_synced_mappings); 1301 } else { 1302 UVMHIST_LOG(*histp, "va=%#"PRIxVADDR 1303 " pg %p: %s syncicache: pte %#x", 1304 va, pg, "defer", npte); 1305 } 1306 } else { 1307 UVMHIST_LOG(*histp, 1308 "va=%#"PRIxVADDR" pg %p: %s syncicache%s", 1309 va, pg, "no", 1310 (pte_cached_p(npte) 1311 ? " (already exec)" 1312 : " (uncached)")); 1313 } 1314 } else if (pg != NULL && (prot & VM_PROT_EXECUTE)) { 1315 KASSERT(mdpg != NULL); 1316 KASSERT(prot & VM_PROT_WRITE); 1317 PMAP_COUNT(exec_mappings); 1318 pmap_page_syncicache(pg); 1319 pmap_page_clear_attributes(mdpg, VM_PAGEMD_EXECPAGE); 1320 UVMHIST_LOG(*histp, 1321 "va=%#"PRIxVADDR" pg %p: %s syncicache%s", 1322 va, pg, "immediate", " (writeable)"); 1323 } 1324 1325 UVMHIST_LOG(*histp, " <-- 0 (OK)", 0, 0, 0, 0); 1326 return 0; 1327 } 1328 1329 void 1330 pmap_kenter_pa(vaddr_t va, paddr_t pa, vm_prot_t prot, u_int flags) 1331 { 1332 pmap_t pmap = pmap_kernel(); 1333 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 1334 struct vm_page_md * const mdpg = (pg ? VM_PAGE_TO_MD(pg) : NULL); 1335 1336 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1337 UVMHIST_LOG(pmaphist, 1338 "(va=%#"PRIxVADDR", pa=%#"PRIxPADDR", prot=%u, flags=%#x)", 1339 va, pa, prot, flags); 1340 PMAP_COUNT(kenter_pa); 1341 1342 if (mdpg == NULL) { 1343 PMAP_COUNT(kenter_pa_unmanaged); 1344 if ((flags & PMAP_CACHE_MASK) == 0) 1345 flags |= PMAP_NOCACHE; 1346 } else { 1347 if ((flags & PMAP_NOCACHE) == 0 && !PMAP_PAGE_COLOROK_P(pa, va)) 1348 PMAP_COUNT(kenter_pa_bad); 1349 } 1350 1351 pt_entry_t npte = pte_make_kenter_pa(pa, mdpg, prot, flags); 1352 kpreempt_disable(); 1353 pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 1354 KASSERTMSG(ptep != NULL, "%#"PRIxVADDR " %#"PRIxVADDR, va, 1355 pmap_limits.virtual_end); 1356 KASSERT(!pte_valid_p(*ptep)); 1357 1358 /* 1359 * No need to track non-managed pages or PMAP_KMPAGEs pages for aliases 1360 */ 1361 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1362 if (pg != NULL && (flags & PMAP_KMPAGE) == 0) { 1363 pmap_enter_pv(pmap, va, pg, &npte, PV_KENTER); 1364 } 1365 #endif 1366 1367 /* 1368 * We have the option to force this mapping into the TLB but we 1369 * don't. Instead let the next reference to the page do it. 1370 */ 1371 pmap_md_tlb_miss_lock_enter(); 1372 *ptep = npte; 1373 pmap_tlb_update_addr(pmap_kernel(), va, npte, 0); 1374 pmap_md_tlb_miss_lock_exit(); 1375 kpreempt_enable(); 1376 #if DEBUG > 1 1377 for (u_int i = 0; i < PAGE_SIZE / sizeof(long); i++) { 1378 if (((long *)va)[i] != ((long *)pa)[i]) 1379 panic("%s: contents (%lx) of va %#"PRIxVADDR 1380 " != contents (%lx) of pa %#"PRIxPADDR, __func__, 1381 ((long *)va)[i], va, ((long *)pa)[i], pa); 1382 } 1383 #endif 1384 1385 UVMHIST_LOG(pmaphist, " <-- done (ptep=%p)", ptep, 0, 0, 0); 1386 } 1387 1388 /* 1389 * Remove the given range of addresses from the kernel map. 1390 * 1391 * It is assumed that the start and end are properly 1392 * rounded to the page size. 1393 */ 1394 1395 static bool 1396 pmap_pte_kremove(pmap_t pmap, vaddr_t sva, vaddr_t eva, pt_entry_t *ptep, 1397 uintptr_t flags) 1398 { 1399 const pt_entry_t new_pte = pte_nv_entry(true); 1400 1401 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1402 UVMHIST_LOG(pmaphist, 1403 "(pmap=%p, sva=%#"PRIxVADDR", eva=%#"PRIxVADDR", ptep=%p)", 1404 pmap, sva, eva, ptep); 1405 1406 KASSERT(kpreempt_disabled()); 1407 1408 for (; sva < eva; sva += NBPG, ptep++) { 1409 pt_entry_t pte = *ptep; 1410 if (!pte_valid_p(pte)) 1411 continue; 1412 1413 PMAP_COUNT(kremove_pages); 1414 struct vm_page * const pg = PHYS_TO_VM_PAGE(pte_to_paddr(pte)); 1415 if (pg != NULL) { 1416 pmap_remove_pv(pmap, sva, pg, !pte_readonly_p(pte)); 1417 } 1418 1419 pmap_md_tlb_miss_lock_enter(); 1420 *ptep = new_pte; 1421 pmap_tlb_invalidate_addr(pmap, sva); 1422 pmap_md_tlb_miss_lock_exit(); 1423 } 1424 1425 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1426 1427 return false; 1428 } 1429 1430 void 1431 pmap_kremove(vaddr_t va, vsize_t len) 1432 { 1433 const vaddr_t sva = trunc_page(va); 1434 const vaddr_t eva = round_page(va + len); 1435 1436 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1437 UVMHIST_LOG(pmaphist, "(va=%#"PRIxVADDR", len=%#"PRIxVSIZE")", 1438 va, len, 0, 0); 1439 1440 kpreempt_disable(); 1441 pmap_pte_process(pmap_kernel(), sva, eva, pmap_pte_kremove, 0); 1442 kpreempt_enable(); 1443 1444 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1445 } 1446 1447 void 1448 pmap_remove_all(struct pmap *pmap) 1449 { 1450 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1451 UVMHIST_LOG(pmaphist, "(pm=%p)", pmap, 0, 0, 0); 1452 1453 KASSERT(pmap != pmap_kernel()); 1454 1455 kpreempt_disable(); 1456 /* 1457 * Free all of our ASIDs which means we can skip doing all the 1458 * tlb_invalidate_addrs(). 1459 */ 1460 pmap_md_tlb_miss_lock_enter(); 1461 #ifdef MULTIPROCESSOR 1462 // This should be the last CPU with this pmap onproc 1463 KASSERT(!kcpuset_isotherset(pmap->pm_onproc, cpu_index(curcpu()))); 1464 if (kcpuset_isset(pmap->pm_onproc, cpu_index(curcpu()))) 1465 #endif 1466 pmap_tlb_asid_deactivate(pmap); 1467 #ifdef MULTIPROCESSOR 1468 KASSERT(kcpuset_iszero(pmap->pm_onproc)); 1469 #endif 1470 pmap_tlb_asid_release_all(pmap); 1471 pmap_md_tlb_miss_lock_exit(); 1472 pmap->pm_flags |= PMAP_DEFERRED_ACTIVATE; 1473 1474 #ifdef PMAP_FAULTINFO 1475 curpcb->pcb_faultinfo.pfi_faultaddr = 0; 1476 curpcb->pcb_faultinfo.pfi_repeats = 0; 1477 curpcb->pcb_faultinfo.pfi_faultpte = NULL; 1478 #endif 1479 kpreempt_enable(); 1480 1481 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1482 } 1483 1484 /* 1485 * Routine: pmap_unwire 1486 * Function: Clear the wired attribute for a map/virtual-address 1487 * pair. 1488 * In/out conditions: 1489 * The mapping must already exist in the pmap. 1490 */ 1491 void 1492 pmap_unwire(pmap_t pmap, vaddr_t va) 1493 { 1494 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1495 UVMHIST_LOG(pmaphist, "(pmap=%p, va=%#"PRIxVADDR")", pmap, va, 0, 0); 1496 PMAP_COUNT(unwire); 1497 1498 /* 1499 * Don't need to flush the TLB since PG_WIRED is only in software. 1500 */ 1501 kpreempt_disable(); 1502 pmap_addr_range_check(pmap, va, va, __func__); 1503 pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 1504 KASSERTMSG(ptep != NULL, "pmap %p va %#"PRIxVADDR" invalid STE", 1505 pmap, va); 1506 pt_entry_t pte = *ptep; 1507 KASSERTMSG(pte_valid_p(pte), 1508 "pmap %p va %#"PRIxVADDR" invalid PTE %#"PRIxPTE" @ %p", 1509 pmap, va, pte_value(pte), ptep); 1510 1511 if (pte_wired_p(pte)) { 1512 pmap_md_tlb_miss_lock_enter(); 1513 *ptep = pte_unwire_entry(pte); 1514 pmap_md_tlb_miss_lock_exit(); 1515 pmap->pm_stats.wired_count--; 1516 } 1517 #ifdef DIAGNOSTIC 1518 else { 1519 printf("%s: wiring for pmap %p va %#"PRIxVADDR" unchanged!\n", 1520 __func__, pmap, va); 1521 } 1522 #endif 1523 kpreempt_enable(); 1524 1525 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1526 } 1527 1528 /* 1529 * Routine: pmap_extract 1530 * Function: 1531 * Extract the physical page address associated 1532 * with the given map/virtual_address pair. 1533 */ 1534 bool 1535 pmap_extract(pmap_t pmap, vaddr_t va, paddr_t *pap) 1536 { 1537 paddr_t pa; 1538 1539 if (pmap == pmap_kernel()) { 1540 if (pmap_md_direct_mapped_vaddr_p(va)) { 1541 pa = pmap_md_direct_mapped_vaddr_to_paddr(va); 1542 goto done; 1543 } 1544 if (pmap_md_io_vaddr_p(va)) 1545 panic("pmap_extract: io address %#"PRIxVADDR"", va); 1546 1547 if (va >= pmap_limits.virtual_end) 1548 panic("%s: illegal kernel mapped address %#"PRIxVADDR, 1549 __func__, va); 1550 } 1551 kpreempt_disable(); 1552 const pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 1553 if (ptep == NULL || !pte_valid_p(*ptep)) { 1554 kpreempt_enable(); 1555 return false; 1556 } 1557 pa = pte_to_paddr(*ptep) | (va & PGOFSET); 1558 kpreempt_enable(); 1559 done: 1560 if (pap != NULL) { 1561 *pap = pa; 1562 } 1563 return true; 1564 } 1565 1566 /* 1567 * Copy the range specified by src_addr/len 1568 * from the source map to the range dst_addr/len 1569 * in the destination map. 1570 * 1571 * This routine is only advisory and need not do anything. 1572 */ 1573 void 1574 pmap_copy(pmap_t dst_pmap, pmap_t src_pmap, vaddr_t dst_addr, vsize_t len, 1575 vaddr_t src_addr) 1576 { 1577 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1578 PMAP_COUNT(copy); 1579 } 1580 1581 /* 1582 * pmap_clear_reference: 1583 * 1584 * Clear the reference bit on the specified physical page. 1585 */ 1586 bool 1587 pmap_clear_reference(struct vm_page *pg) 1588 { 1589 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1590 1591 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1592 UVMHIST_LOG(pmaphist, "(pg=%p (pa %#"PRIxPADDR"))", 1593 pg, VM_PAGE_TO_PHYS(pg), 0,0); 1594 1595 bool rv = pmap_page_clear_attributes(mdpg, VM_PAGEMD_REFERENCED); 1596 1597 UVMHIST_LOG(pmaphist, " <-- %s", rv ? "true" : "false", 0, 0, 0); 1598 1599 return rv; 1600 } 1601 1602 /* 1603 * pmap_is_referenced: 1604 * 1605 * Return whether or not the specified physical page is referenced 1606 * by any physical maps. 1607 */ 1608 bool 1609 pmap_is_referenced(struct vm_page *pg) 1610 { 1611 return VM_PAGEMD_REFERENCED_P(VM_PAGE_TO_MD(pg)); 1612 } 1613 1614 /* 1615 * Clear the modify bits on the specified physical page. 1616 */ 1617 bool 1618 pmap_clear_modify(struct vm_page *pg) 1619 { 1620 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1621 pv_entry_t pv = &mdpg->mdpg_first; 1622 pv_entry_t pv_next; 1623 1624 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1625 UVMHIST_LOG(pmaphist, "(pg=%p (%#"PRIxPADDR"))", 1626 pg, VM_PAGE_TO_PHYS(pg), 0,0); 1627 PMAP_COUNT(clear_modify); 1628 1629 if (VM_PAGEMD_EXECPAGE_P(mdpg)) { 1630 if (pv->pv_pmap == NULL) { 1631 UVMHIST_LOG(pmapexechist, 1632 "pg %p (pa %#"PRIxPADDR"): %s", 1633 pg, VM_PAGE_TO_PHYS(pg), "execpage cleared", 0); 1634 pmap_page_clear_attributes(mdpg, VM_PAGEMD_EXECPAGE); 1635 PMAP_COUNT(exec_uncached_clear_modify); 1636 } else { 1637 UVMHIST_LOG(pmapexechist, 1638 "pg %p (pa %#"PRIxPADDR"): %s", 1639 pg, VM_PAGE_TO_PHYS(pg), "syncicache performed", 0); 1640 pmap_page_syncicache(pg); 1641 PMAP_COUNT(exec_synced_clear_modify); 1642 } 1643 } 1644 if (!pmap_page_clear_attributes(mdpg, VM_PAGEMD_MODIFIED)) { 1645 UVMHIST_LOG(pmaphist, " <-- false", 0, 0, 0, 0); 1646 return false; 1647 } 1648 if (pv->pv_pmap == NULL) { 1649 UVMHIST_LOG(pmaphist, " <-- true (no mappings)", 0, 0, 0, 0); 1650 return true; 1651 } 1652 1653 /* 1654 * remove write access from any pages that are dirty 1655 * so we can tell if they are written to again later. 1656 * flush the VAC first if there is one. 1657 */ 1658 kpreempt_disable(); 1659 KASSERT(!VM_PAGEMD_PVLIST_LOCKED_P(mdpg)); 1660 VM_PAGEMD_PVLIST_READLOCK(mdpg); 1661 pmap_pvlist_check(mdpg); 1662 for (; pv != NULL; pv = pv_next) { 1663 pmap_t pmap = pv->pv_pmap; 1664 vaddr_t va = trunc_page(pv->pv_va); 1665 1666 pv_next = pv->pv_next; 1667 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1668 if (pv->pv_va & PV_KENTER) 1669 continue; 1670 #endif 1671 pt_entry_t * const ptep = pmap_pte_lookup(pmap, va); 1672 KASSERT(ptep); 1673 pt_entry_t pte = pte_prot_nowrite(*ptep); 1674 if (*ptep == pte) { 1675 continue; 1676 } 1677 KASSERT(pte_valid_p(pte)); 1678 const uintptr_t gen = VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1679 pmap_md_tlb_miss_lock_enter(); 1680 *ptep = pte; 1681 pmap_tlb_invalidate_addr(pmap, va); 1682 pmap_md_tlb_miss_lock_exit(); 1683 pmap_update(pmap); 1684 if (__predict_false(gen != VM_PAGEMD_PVLIST_READLOCK(mdpg))) { 1685 /* 1686 * The list changed! So restart from the beginning. 1687 */ 1688 pv_next = &mdpg->mdpg_first; 1689 pmap_pvlist_check(mdpg); 1690 } 1691 } 1692 pmap_pvlist_check(mdpg); 1693 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1694 kpreempt_enable(); 1695 1696 UVMHIST_LOG(pmaphist, " <-- true (mappings changed)", 0, 0, 0, 0); 1697 return true; 1698 } 1699 1700 /* 1701 * pmap_is_modified: 1702 * 1703 * Return whether or not the specified physical page is modified 1704 * by any physical maps. 1705 */ 1706 bool 1707 pmap_is_modified(struct vm_page *pg) 1708 { 1709 return VM_PAGEMD_MODIFIED_P(VM_PAGE_TO_MD(pg)); 1710 } 1711 1712 /* 1713 * pmap_set_modified: 1714 * 1715 * Sets the page modified reference bit for the specified page. 1716 */ 1717 void 1718 pmap_set_modified(paddr_t pa) 1719 { 1720 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 1721 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1722 pmap_page_set_attributes(mdpg, VM_PAGEMD_MODIFIED|VM_PAGEMD_REFERENCED); 1723 } 1724 1725 /******************** pv_entry management ********************/ 1726 1727 static void 1728 pmap_pvlist_check(struct vm_page_md *mdpg) 1729 { 1730 #ifdef DEBUG 1731 pv_entry_t pv = &mdpg->mdpg_first; 1732 if (pv->pv_pmap != NULL) { 1733 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1734 const u_int colormask = uvmexp.colormask; 1735 u_int colors = 0; 1736 #endif 1737 for (; pv != NULL; pv = pv->pv_next) { 1738 KASSERT(pv->pv_pmap != pmap_kernel() || !pmap_md_direct_mapped_vaddr_p(pv->pv_va)); 1739 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1740 colors |= __BIT(atop(pv->pv_va) & colormask); 1741 #endif 1742 } 1743 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1744 // Assert there if there more than 1 color mapped, that they 1745 // are uncached. 1746 KASSERTMSG(!pmap_md_virtual_cache_aliasing_p() 1747 || colors == 0 || (colors & (colors-1)) == 0 1748 || VM_PAGEMD_UNCACHED_P(mdpg), "colors=%#x uncached=%u", 1749 colors, VM_PAGEMD_UNCACHED_P(mdpg)); 1750 #endif 1751 } 1752 #endif /* DEBUG */ 1753 } 1754 1755 /* 1756 * Enter the pmap and virtual address into the 1757 * physical to virtual map table. 1758 */ 1759 void 1760 pmap_enter_pv(pmap_t pmap, vaddr_t va, struct vm_page *pg, pt_entry_t *nptep, 1761 u_int flags) 1762 { 1763 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1764 pv_entry_t pv, npv, apv; 1765 #ifdef UVMHIST 1766 bool first = false; 1767 #endif 1768 1769 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1770 UVMHIST_LOG(pmaphist, 1771 "(pmap=%p va=%#"PRIxVADDR" pg=%p (%#"PRIxPADDR")", 1772 pmap, va, pg, VM_PAGE_TO_PHYS(pg)); 1773 UVMHIST_LOG(pmaphist, "nptep=%p (%#"PRIxPTE"))", 1774 nptep, pte_value(*nptep), 0, 0); 1775 1776 KASSERT(kpreempt_disabled()); 1777 KASSERT(pmap != pmap_kernel() || !pmap_md_direct_mapped_vaddr_p(va)); 1778 KASSERTMSG(pmap != pmap_kernel() || !pmap_md_io_vaddr_p(va), 1779 "va %#"PRIxVADDR, va); 1780 1781 apv = NULL; 1782 VM_PAGEMD_PVLIST_LOCK(mdpg); 1783 again: 1784 pv = &mdpg->mdpg_first; 1785 pmap_pvlist_check(mdpg); 1786 if (pv->pv_pmap == NULL) { 1787 KASSERT(pv->pv_next == NULL); 1788 /* 1789 * No entries yet, use header as the first entry 1790 */ 1791 PMAP_COUNT(primary_mappings); 1792 PMAP_COUNT(mappings); 1793 #ifdef UVMHIST 1794 first = true; 1795 #endif 1796 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1797 KASSERT(VM_PAGEMD_CACHED_P(mdpg)); 1798 // If the new mapping has an incompatible color the last 1799 // mapping of this page, clean the page before using it. 1800 if (!PMAP_PAGE_COLOROK_P(va, pv->pv_va)) { 1801 pmap_md_vca_clean(pg, PMAP_WBINV); 1802 } 1803 #endif 1804 pv->pv_pmap = pmap; 1805 pv->pv_va = va | flags; 1806 } else { 1807 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1808 if (pmap_md_vca_add(pg, va, nptep)) { 1809 goto again; 1810 } 1811 #endif 1812 1813 /* 1814 * There is at least one other VA mapping this page. 1815 * Place this entry after the header. 1816 * 1817 * Note: the entry may already be in the table if 1818 * we are only changing the protection bits. 1819 */ 1820 1821 #ifdef PARANOIADIAG 1822 const paddr_t pa = VM_PAGE_TO_PHYS(pg); 1823 #endif 1824 for (npv = pv; npv; npv = npv->pv_next) { 1825 if (pmap == npv->pv_pmap 1826 && va == trunc_page(npv->pv_va)) { 1827 #ifdef PARANOIADIAG 1828 pt_entry_t *ptep = pmap_pte_lookup(pmap, va); 1829 pt_entry_t pte = (ptep != NULL) ? *ptep : 0; 1830 if (!pte_valid_p(pte) || pte_to_paddr(pte) != pa) 1831 printf("%s: found va %#"PRIxVADDR 1832 " pa %#"PRIxPADDR 1833 " in pv_table but != %#"PRIxPTE"\n", 1834 __func__, va, pa, pte_value(pte)); 1835 #endif 1836 PMAP_COUNT(remappings); 1837 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1838 if (__predict_false(apv != NULL)) 1839 pmap_pv_free(apv); 1840 1841 UVMHIST_LOG(pmaphist, " <-- done pv=%p%s", 1842 pv, " (reused)", 0, 0); 1843 return; 1844 } 1845 } 1846 if (__predict_true(apv == NULL)) { 1847 /* 1848 * To allocate a PV, we have to release the PVLIST lock 1849 * so get the page generation. We allocate the PV, and 1850 * then reacquire the lock. 1851 */ 1852 pmap_pvlist_check(mdpg); 1853 const uintptr_t gen = VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1854 1855 apv = (pv_entry_t)pmap_pv_alloc(); 1856 if (apv == NULL) 1857 panic("pmap_enter_pv: pmap_pv_alloc() failed"); 1858 1859 /* 1860 * If the generation has changed, then someone else 1861 * tinkered with this page so we should start over. 1862 */ 1863 if (gen != VM_PAGEMD_PVLIST_LOCK(mdpg)) 1864 goto again; 1865 } 1866 npv = apv; 1867 apv = NULL; 1868 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1869 /* 1870 * If need to deal with virtual cache aliases, keep mappings 1871 * in the kernel pmap at the head of the list. This allows 1872 * the VCA code to easily use them for cache operations if 1873 * present. 1874 */ 1875 pmap_t kpmap = pmap_kernel(); 1876 if (pmap != kpmap) { 1877 while (pv->pv_pmap == kpmap && pv->pv_next != NULL) { 1878 pv = pv->pv_next; 1879 } 1880 } 1881 #endif 1882 npv->pv_va = va | flags; 1883 npv->pv_pmap = pmap; 1884 npv->pv_next = pv->pv_next; 1885 pv->pv_next = npv; 1886 PMAP_COUNT(mappings); 1887 } 1888 pmap_pvlist_check(mdpg); 1889 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1890 if (__predict_false(apv != NULL)) 1891 pmap_pv_free(apv); 1892 1893 UVMHIST_LOG(pmaphist, " <-- done pv=%p%s", 1894 pv, first ? " (first pv)" : "",0,0); 1895 } 1896 1897 /* 1898 * Remove a physical to virtual address translation. 1899 * If cache was inhibited on this page, and there are no more cache 1900 * conflicts, restore caching. 1901 * Flush the cache if the last page is removed (should always be cached 1902 * at this point). 1903 */ 1904 void 1905 pmap_remove_pv(pmap_t pmap, vaddr_t va, struct vm_page *pg, bool dirty) 1906 { 1907 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 1908 pv_entry_t pv, npv; 1909 bool last; 1910 1911 UVMHIST_FUNC(__func__); UVMHIST_CALLED(pmaphist); 1912 UVMHIST_LOG(pmaphist, 1913 "(pmap=%p, va=%#"PRIxVADDR", pg=%p (pa %#"PRIxPADDR")", 1914 pmap, va, pg, VM_PAGE_TO_PHYS(pg)); 1915 UVMHIST_LOG(pmaphist, "dirty=%s)", dirty ? "true" : "false", 0, 0, 0); 1916 1917 KASSERT(kpreempt_disabled()); 1918 KASSERT((va & PAGE_MASK) == 0); 1919 pv = &mdpg->mdpg_first; 1920 1921 VM_PAGEMD_PVLIST_LOCK(mdpg); 1922 pmap_pvlist_check(mdpg); 1923 1924 /* 1925 * If it is the first entry on the list, it is actually 1926 * in the header and we must copy the following entry up 1927 * to the header. Otherwise we must search the list for 1928 * the entry. In either case we free the now unused entry. 1929 */ 1930 1931 last = false; 1932 if (pmap == pv->pv_pmap && va == trunc_page(pv->pv_va)) { 1933 npv = pv->pv_next; 1934 if (npv) { 1935 *pv = *npv; 1936 KASSERT(pv->pv_pmap != NULL); 1937 } else { 1938 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1939 pmap_page_clear_attributes(mdpg, VM_PAGEMD_UNCACHED); 1940 #endif 1941 pv->pv_pmap = NULL; 1942 last = true; /* Last mapping removed */ 1943 } 1944 PMAP_COUNT(remove_pvfirst); 1945 } else { 1946 for (npv = pv->pv_next; npv; pv = npv, npv = npv->pv_next) { 1947 PMAP_COUNT(remove_pvsearch); 1948 if (pmap == npv->pv_pmap && va == trunc_page(npv->pv_va)) 1949 break; 1950 } 1951 if (npv) { 1952 pv->pv_next = npv->pv_next; 1953 } 1954 } 1955 1956 pmap_pvlist_check(mdpg); 1957 VM_PAGEMD_PVLIST_UNLOCK(mdpg); 1958 1959 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 1960 pmap_md_vca_remove(pg, va, dirty, last); 1961 #endif 1962 1963 /* 1964 * Free the pv_entry if needed. 1965 */ 1966 if (npv) 1967 pmap_pv_free(npv); 1968 if (VM_PAGEMD_EXECPAGE_P(mdpg) && dirty) { 1969 if (last) { 1970 /* 1971 * If this was the page's last mapping, we no longer 1972 * care about its execness. 1973 */ 1974 UVMHIST_LOG(pmapexechist, 1975 "pg %p (pa %#"PRIxPADDR")%s: %s", 1976 pg, VM_PAGE_TO_PHYS(pg), 1977 last ? " [last mapping]" : "", 1978 "execpage cleared"); 1979 pmap_page_clear_attributes(mdpg, VM_PAGEMD_EXECPAGE); 1980 PMAP_COUNT(exec_uncached_remove); 1981 } else { 1982 /* 1983 * Someone still has it mapped as an executable page 1984 * so we must sync it. 1985 */ 1986 UVMHIST_LOG(pmapexechist, 1987 "pg %p (pa %#"PRIxPADDR")%s: %s", 1988 pg, VM_PAGE_TO_PHYS(pg), 1989 last ? " [last mapping]" : "", 1990 "performed syncicache"); 1991 pmap_page_syncicache(pg); 1992 PMAP_COUNT(exec_synced_remove); 1993 } 1994 } 1995 1996 UVMHIST_LOG(pmaphist, " <-- done", 0, 0, 0, 0); 1997 } 1998 1999 #if defined(MULTIPROCESSOR) 2000 struct pmap_pvlist_info { 2001 kmutex_t *pli_locks[PAGE_SIZE / 32]; 2002 volatile u_int pli_lock_refs[PAGE_SIZE / 32]; 2003 volatile u_int pli_lock_index; 2004 u_int pli_lock_mask; 2005 } pmap_pvlist_info; 2006 2007 void 2008 pmap_pvlist_lock_init(size_t cache_line_size) 2009 { 2010 struct pmap_pvlist_info * const pli = &pmap_pvlist_info; 2011 const vaddr_t lock_page = uvm_pageboot_alloc(PAGE_SIZE); 2012 vaddr_t lock_va = lock_page; 2013 if (sizeof(kmutex_t) > cache_line_size) { 2014 cache_line_size = roundup2(sizeof(kmutex_t), cache_line_size); 2015 } 2016 const size_t nlocks = PAGE_SIZE / cache_line_size; 2017 KASSERT((nlocks & (nlocks - 1)) == 0); 2018 /* 2019 * Now divide the page into a number of mutexes, one per cacheline. 2020 */ 2021 for (size_t i = 0; i < nlocks; lock_va += cache_line_size, i++) { 2022 kmutex_t * const lock = (kmutex_t *)lock_va; 2023 mutex_init(lock, MUTEX_DEFAULT, IPL_HIGH); 2024 pli->pli_locks[i] = lock; 2025 } 2026 pli->pli_lock_mask = nlocks - 1; 2027 } 2028 2029 kmutex_t * 2030 pmap_pvlist_lock_addr(struct vm_page_md *mdpg) 2031 { 2032 struct pmap_pvlist_info * const pli = &pmap_pvlist_info; 2033 kmutex_t *lock = mdpg->mdpg_lock; 2034 2035 /* 2036 * Allocate a lock on an as-needed basis. This will hopefully give us 2037 * semi-random distribution not based on page color. 2038 */ 2039 if (__predict_false(lock == NULL)) { 2040 size_t locknum = atomic_add_int_nv(&pli->pli_lock_index, 37); 2041 size_t lockid = locknum & pli->pli_lock_mask; 2042 kmutex_t * const new_lock = pli->pli_locks[lockid]; 2043 /* 2044 * Set the lock. If some other thread already did, just use 2045 * the one they assigned. 2046 */ 2047 lock = atomic_cas_ptr(&mdpg->mdpg_lock, NULL, new_lock); 2048 if (lock == NULL) { 2049 lock = new_lock; 2050 atomic_inc_uint(&pli->pli_lock_refs[lockid]); 2051 } 2052 } 2053 2054 /* 2055 * Now finally provide the lock. 2056 */ 2057 return lock; 2058 } 2059 #else /* !MULTIPROCESSOR */ 2060 void 2061 pmap_pvlist_lock_init(size_t cache_line_size) 2062 { 2063 mutex_init(&pmap_pvlist_mutex, MUTEX_DEFAULT, IPL_HIGH); 2064 } 2065 2066 #ifdef MODULAR 2067 kmutex_t * 2068 pmap_pvlist_lock_addr(struct vm_page_md *mdpg) 2069 { 2070 /* 2071 * We just use a global lock. 2072 */ 2073 if (__predict_false(mdpg->mdpg_lock == NULL)) { 2074 mdpg->mdpg_lock = &pmap_pvlist_mutex; 2075 } 2076 2077 /* 2078 * Now finally provide the lock. 2079 */ 2080 return mdpg->mdpg_lock; 2081 } 2082 #endif /* MODULAR */ 2083 #endif /* !MULTIPROCESSOR */ 2084 2085 /* 2086 * pmap_pv_page_alloc: 2087 * 2088 * Allocate a page for the pv_entry pool. 2089 */ 2090 void * 2091 pmap_pv_page_alloc(struct pool *pp, int flags) 2092 { 2093 struct vm_page * const pg = PMAP_ALLOC_POOLPAGE(UVM_PGA_USERESERVE); 2094 if (pg == NULL) 2095 return NULL; 2096 2097 return (void *)pmap_map_poolpage(VM_PAGE_TO_PHYS(pg)); 2098 } 2099 2100 /* 2101 * pmap_pv_page_free: 2102 * 2103 * Free a pv_entry pool page. 2104 */ 2105 void 2106 pmap_pv_page_free(struct pool *pp, void *v) 2107 { 2108 vaddr_t va = (vaddr_t)v; 2109 2110 KASSERT(pmap_md_direct_mapped_vaddr_p(va)); 2111 const paddr_t pa = pmap_md_direct_mapped_vaddr_to_paddr(va); 2112 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 2113 KASSERT(pg != NULL); 2114 #ifdef PMAP_VIRTUAL_CACHE_ALIASES 2115 kpreempt_disable(); 2116 pmap_md_vca_remove(pg, va, true, true); 2117 kpreempt_enable(); 2118 #endif 2119 pmap_page_clear_attributes(VM_PAGE_TO_MD(pg), VM_PAGEMD_POOLPAGE); 2120 uvm_pagefree(pg); 2121 } 2122 2123 #ifdef PMAP_PREFER 2124 /* 2125 * Find first virtual address >= *vap that doesn't cause 2126 * a cache alias conflict. 2127 */ 2128 void 2129 pmap_prefer(vaddr_t foff, vaddr_t *vap, vsize_t sz, int td) 2130 { 2131 vsize_t prefer_mask = ptoa(uvmexp.colormask); 2132 2133 PMAP_COUNT(prefer_requests); 2134 2135 prefer_mask |= pmap_md_cache_prefer_mask(); 2136 2137 if (prefer_mask) { 2138 vaddr_t va = *vap; 2139 vsize_t d = (foff - va) & prefer_mask; 2140 if (d) { 2141 if (td) 2142 *vap = trunc_page(va - ((-d) & prefer_mask)); 2143 else 2144 *vap = round_page(va + d); 2145 PMAP_COUNT(prefer_adjustments); 2146 } 2147 } 2148 } 2149 #endif /* PMAP_PREFER */ 2150 2151 #ifdef PMAP_MAP_POOLPAGE 2152 vaddr_t 2153 pmap_map_poolpage(paddr_t pa) 2154 { 2155 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 2156 KASSERT(pg); 2157 struct vm_page_md * const mdpg = VM_PAGE_TO_MD(pg); 2158 pmap_page_set_attributes(mdpg, VM_PAGEMD_POOLPAGE); 2159 2160 return pmap_md_map_poolpage(pa, NBPG); 2161 } 2162 2163 paddr_t 2164 pmap_unmap_poolpage(vaddr_t va) 2165 { 2166 KASSERT(pmap_md_direct_mapped_vaddr_p(va)); 2167 paddr_t pa = pmap_md_direct_mapped_vaddr_to_paddr(va); 2168 2169 struct vm_page * const pg = PHYS_TO_VM_PAGE(pa); 2170 KASSERT(pg != NULL); 2171 pmap_page_clear_attributes(VM_PAGE_TO_MD(pg), VM_PAGEMD_POOLPAGE); 2172 pmap_md_unmap_poolpage(va, NBPG); 2173 2174 return pa; 2175 } 2176 #endif /* PMAP_MAP_POOLPAGE */ 2177
Indexes created Fri Sep 26 20:09:58 GMT 2025