1 /*- 2 * Copyright (c) 2001 The NetBSD Foundation, Inc. 3 * All rights reserved. 4 * 5 * This code is derived from software contributed to The NetBSD Foundation 6 * by Matt Thomas <matt@3am-software.com> of Allegro Networks, Inc. 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 NetBSD 19 * Foundation, Inc. and its contributors. 20 * 4. Neither the name of The NetBSD Foundation nor the names of its 21 * contributors may be used to endorse or promote products derived 22 * from this software without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 25 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 26 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 27 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 28 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 34 * POSSIBILITY OF SUCH DAMAGE. 35 */ 36 /*- 37 * Copyright (C) 1995, 1996 Wolfgang Solfrank. 38 * Copyright (C) 1995, 1996 TooLs GmbH. 39 * All rights reserved. 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. All advertising materials mentioning features or use of this software 50 * must display the following acknowledgement: 51 * This product includes software developed by TooLs GmbH. 52 * 4. The name of TooLs GmbH may not be used to endorse or promote products 53 * derived from this software without specific prior written permission. 54 * 55 * THIS SOFTWARE IS PROVIDED BY TOOLS GMBH ``AS IS'' AND ANY EXPRESS OR 56 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 57 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 58 * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 59 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 60 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 61 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 62 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 63 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 64 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 65 * 66 * $NetBSD: pmap.c,v 1.28 2000/03/26 20:42:36 kleink Exp $ 67 */ 68 /*- 69 * Copyright (C) 2001 Benno Rice. 70 * All rights reserved. 71 * 72 * Redistribution and use in source and binary forms, with or without 73 * modification, are permitted provided that the following conditions 74 * are met: 75 * 1. Redistributions of source code must retain the above copyright 76 * notice, this list of conditions and the following disclaimer. 77 * 2. Redistributions in binary form must reproduce the above copyright 78 * notice, this list of conditions and the following disclaimer in the 79 * documentation and/or other materials provided with the distribution. 80 * 81 * THIS SOFTWARE IS PROVIDED BY Benno Rice ``AS IS'' AND ANY EXPRESS OR 82 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 83 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 84 * IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 85 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 86 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 87 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 88 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 89 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 90 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 91 */ 92 93 #include <sys/cdefs.h> 94 __FBSDID("$FreeBSD$"); 95 96 /* 97 * Manages physical address maps. 98 * 99 * In addition to hardware address maps, this module is called upon to 100 * provide software-use-only maps which may or may not be stored in the 101 * same form as hardware maps. These pseudo-maps are used to store 102 * intermediate results from copy operations to and from address spaces. 103 * 104 * Since the information managed by this module is also stored by the 105 * logical address mapping module, this module may throw away valid virtual 106 * to physical mappings at almost any time. However, invalidations of 107 * mappings must be done as requested. 108 * 109 * In order to cope with hardware architectures which make virtual to 110 * physical map invalidates expensive, this module may delay invalidate 111 * reduced protection operations until such time as they are actually 112 * necessary. This module is given full information as to which processors 113 * are currently using which maps, and to when physical maps must be made 114 * correct. 115 */ 116 117 #include "opt_kstack_pages.h" 118 119 #include <sys/param.h> 120 #include <sys/kernel.h> 121 #include <sys/ktr.h> 122 #include <sys/lock.h> 123 #include <sys/msgbuf.h> 124 #include <sys/mutex.h> 125 #include <sys/proc.h> 126 #include <sys/sysctl.h> 127 #include <sys/systm.h> 128 #include <sys/vmmeter.h> 129 130 #include <sys/kdb.h> 131 132 #include <dev/ofw/openfirm.h> 133 134 #include <vm/vm.h> 135 #include <vm/vm_param.h> 136 #include <vm/vm_kern.h> 137 #include <vm/vm_page.h> 138 #include <vm/vm_map.h> 139 #include <vm/vm_object.h> 140 #include <vm/vm_extern.h> 141 #include <vm/vm_pageout.h> 142 #include <vm/vm_pager.h> 143 #include <vm/uma.h> 144 145 #include <machine/_inttypes.h> 146 #include <machine/cpu.h> 147 #include <machine/platform.h> 148 #include <machine/frame.h> 149 #include <machine/md_var.h> 150 #include <machine/psl.h> 151 #include <machine/bat.h> 152 #include <machine/hid.h> 153 #include <machine/pte.h> 154 #include <machine/sr.h> 155 #include <machine/trap.h> 156 #include <machine/mmuvar.h> 157 158 #include "mmu_if.h" 159 160 #define MOEA_DEBUG 161 162 #define TODO panic("%s: not implemented", __func__); 163 void moea64_release_vsid(uint64_t vsid); 164 uintptr_t moea64_get_unique_vsid(void); 165 166 static __inline register_t 167 cntlzd(volatile register_t a) { 168 register_t b; 169 __asm ("cntlzd %0, %1" : "=r"(b) : "r"(a)); 170 return b; 171 } 172 173 #define PTESYNC() __asm __volatile("ptesync"); 174 #define TLBSYNC() __asm __volatile("tlbsync; ptesync"); 175 #define SYNC() __asm __volatile("sync"); 176 #define EIEIO() __asm __volatile("eieio"); 177 178 /* 179 * The tlbie instruction must be executed in 64-bit mode 180 * so we have to twiddle MSR[SF] around every invocation. 181 * Just to add to the fun, exceptions must be off as well 182 * so that we can't trap in 64-bit mode. What a pain. 183 */ 184 struct mtx tlbie_mutex; 185 186 static __inline void 187 TLBIE(uint64_t vpn) { 188 #ifndef __powerpc64__ 189 register_t vpn_hi, vpn_lo; 190 register_t msr; 191 register_t scratch; 192 #endif 193 194 vpn <<= ADDR_PIDX_SHFT; 195 vpn &= ~(0xffffULL << 48); 196 197 mtx_lock_spin(&tlbie_mutex); 198 #ifdef __powerpc64__ 199 __asm __volatile("\ 200 ptesync; \ 201 tlbie %0; \ 202 eieio; \ 203 tlbsync; \ 204 ptesync;" 205 :: "r"(vpn) : "memory"); 206 #else 207 vpn_hi = (uint32_t)(vpn >> 32); 208 vpn_lo = (uint32_t)vpn; 209 210 __asm __volatile("\ 211 mfmsr %0; \ 212 mr %1, %0; \ 213 insrdi %1,%5,1,0; \ 214 mtmsrd %1; isync; \ 215 ptesync; \ 216 \ 217 sld %1,%2,%4; \ 218 or %1,%1,%3; \ 219 tlbie %1; \ 220 \ 221 mtmsrd %0; isync; \ 222 eieio; \ 223 tlbsync; \ 224 ptesync;" 225 : "=r"(msr), "=r"(scratch) : "r"(vpn_hi), "r"(vpn_lo), "r"(32), "r"(1) 226 : "memory"); 227 #endif 228 mtx_unlock_spin(&tlbie_mutex); 229 } 230 231 #define DISABLE_TRANS(msr) msr = mfmsr(); mtmsr(msr & ~PSL_DR); isync() 232 #define ENABLE_TRANS(msr) mtmsr(msr); isync() 233 234 #define VSID_MAKE(sr, hash) ((sr) | (((hash) & 0xfffff) << 4)) 235 #define VSID_TO_HASH(vsid) (((vsid) >> 4) & 0xfffff) 236 #define VSID_HASH_MASK 0x0000007fffffffffULL 237 238 #define PVO_PTEGIDX_MASK 0x007UL /* which PTEG slot */ 239 #define PVO_PTEGIDX_VALID 0x008UL /* slot is valid */ 240 #define PVO_WIRED 0x010UL /* PVO entry is wired */ 241 #define PVO_MANAGED 0x020UL /* PVO entry is managed */ 242 #define PVO_BOOTSTRAP 0x080UL /* PVO entry allocated during 243 bootstrap */ 244 #define PVO_FAKE 0x100UL /* fictitious phys page */ 245 #define PVO_LARGE 0x200UL /* large page */ 246 #define PVO_VADDR(pvo) ((pvo)->pvo_vaddr & ~ADDR_POFF) 247 #define PVO_ISFAKE(pvo) ((pvo)->pvo_vaddr & PVO_FAKE) 248 #define PVO_PTEGIDX_GET(pvo) ((pvo)->pvo_vaddr & PVO_PTEGIDX_MASK) 249 #define PVO_PTEGIDX_ISSET(pvo) ((pvo)->pvo_vaddr & PVO_PTEGIDX_VALID) 250 #define PVO_PTEGIDX_CLR(pvo) \ 251 ((void)((pvo)->pvo_vaddr &= ~(PVO_PTEGIDX_VALID|PVO_PTEGIDX_MASK))) 252 #define PVO_PTEGIDX_SET(pvo, i) \ 253 ((void)((pvo)->pvo_vaddr |= (i)|PVO_PTEGIDX_VALID)) 254 #define PVO_VSID(pvo) ((pvo)->pvo_vpn >> 16) 255 256 #define MOEA_PVO_CHECK(pvo) 257 258 #define LOCK_TABLE() mtx_lock(&moea64_table_mutex) 259 #define UNLOCK_TABLE() mtx_unlock(&moea64_table_mutex); 260 #define ASSERT_TABLE_LOCK() mtx_assert(&moea64_table_mutex, MA_OWNED) 261 262 struct ofw_map { 263 cell_t om_va; 264 cell_t om_len; 265 cell_t om_pa_hi; 266 cell_t om_pa_lo; 267 cell_t om_mode; 268 }; 269 270 /* 271 * Map of physical memory regions. 272 */ 273 static struct mem_region *regions; 274 static struct mem_region *pregions; 275 static u_int phys_avail_count; 276 static int regions_sz, pregions_sz; 277 278 extern struct pmap ofw_pmap; 279 280 extern void bs_remap_earlyboot(void); 281 282 283 /* 284 * Lock for the pteg and pvo tables. 285 */ 286 struct mtx moea64_table_mutex; 287 struct mtx moea64_slb_mutex; 288 289 /* 290 * PTEG data. 291 */ 292 static struct lpteg *moea64_pteg_table; 293 u_int moea64_pteg_count; 294 u_int moea64_pteg_mask; 295 296 /* 297 * PVO data. 298 */ 299 struct pvo_head *moea64_pvo_table; /* pvo entries by pteg index */ 300 struct pvo_head moea64_pvo_kunmanaged = /* list of unmanaged pages */ 301 LIST_HEAD_INITIALIZER(moea64_pvo_kunmanaged); 302 303 uma_zone_t moea64_upvo_zone; /* zone for pvo entries for unmanaged pages */ 304 uma_zone_t moea64_mpvo_zone; /* zone for pvo entries for managed pages */ 305 306 #define BPVO_POOL_SIZE 327680 307 static struct pvo_entry *moea64_bpvo_pool; 308 static int moea64_bpvo_pool_index = 0; 309 310 #define VSID_NBPW (sizeof(u_int32_t) * 8) 311 #ifdef __powerpc64__ 312 #define NVSIDS (NPMAPS * 16) 313 #define VSID_HASHMASK 0xffffffffUL 314 #else 315 #define NVSIDS NPMAPS 316 #define VSID_HASHMASK 0xfffffUL 317 #endif 318 static u_int moea64_vsid_bitmap[NVSIDS / VSID_NBPW]; 319 320 static boolean_t moea64_initialized = FALSE; 321 322 /* 323 * Statistics. 324 */ 325 u_int moea64_pte_valid = 0; 326 u_int moea64_pte_overflow = 0; 327 u_int moea64_pvo_entries = 0; 328 u_int moea64_pvo_enter_calls = 0; 329 u_int moea64_pvo_remove_calls = 0; 330 SYSCTL_INT(_machdep, OID_AUTO, moea64_pte_valid, CTLFLAG_RD, 331 &moea64_pte_valid, 0, ""); 332 SYSCTL_INT(_machdep, OID_AUTO, moea64_pte_overflow, CTLFLAG_RD, 333 &moea64_pte_overflow, 0, ""); 334 SYSCTL_INT(_machdep, OID_AUTO, moea64_pvo_entries, CTLFLAG_RD, 335 &moea64_pvo_entries, 0, ""); 336 SYSCTL_INT(_machdep, OID_AUTO, moea64_pvo_enter_calls, CTLFLAG_RD, 337 &moea64_pvo_enter_calls, 0, ""); 338 SYSCTL_INT(_machdep, OID_AUTO, moea64_pvo_remove_calls, CTLFLAG_RD, 339 &moea64_pvo_remove_calls, 0, ""); 340 341 vm_offset_t moea64_scratchpage_va[2]; 342 uint64_t moea64_scratchpage_vpn[2]; 343 struct lpte *moea64_scratchpage_pte[2]; 344 struct mtx moea64_scratchpage_mtx; 345 346 uint64_t moea64_large_page_mask = 0; 347 int moea64_large_page_size = 0; 348 int moea64_large_page_shift = 0; 349 350 /* 351 * Allocate physical memory for use in moea64_bootstrap. 352 */ 353 static vm_offset_t moea64_bootstrap_alloc(vm_size_t, u_int); 354 355 /* 356 * PTE calls. 357 */ 358 static int moea64_pte_insert(u_int, struct lpte *); 359 360 /* 361 * PVO calls. 362 */ 363 static int moea64_pvo_enter(pmap_t, uma_zone_t, struct pvo_head *, 364 vm_offset_t, vm_offset_t, uint64_t, int); 365 static void moea64_pvo_remove(struct pvo_entry *); 366 static struct pvo_entry *moea64_pvo_find_va(pmap_t, vm_offset_t); 367 static struct lpte *moea64_pvo_to_pte(const struct pvo_entry *); 368 369 /* 370 * Utility routines. 371 */ 372 static void moea64_bootstrap(mmu_t mmup, 373 vm_offset_t kernelstart, vm_offset_t kernelend); 374 static void moea64_cpu_bootstrap(mmu_t, int ap); 375 static void moea64_enter_locked(pmap_t, vm_offset_t, vm_page_t, 376 vm_prot_t, boolean_t); 377 static boolean_t moea64_query_bit(vm_page_t, u_int64_t); 378 static u_int moea64_clear_bit(vm_page_t, u_int64_t); 379 static void moea64_kremove(mmu_t, vm_offset_t); 380 static void moea64_syncicache(pmap_t pmap, vm_offset_t va, 381 vm_offset_t pa, vm_size_t sz); 382 static void tlbia(void); 383 #ifdef __powerpc64__ 384 static void slbia(void); 385 #endif 386 387 /* 388 * Kernel MMU interface 389 */ 390 void moea64_change_wiring(mmu_t, pmap_t, vm_offset_t, boolean_t); 391 void moea64_clear_modify(mmu_t, vm_page_t); 392 void moea64_clear_reference(mmu_t, vm_page_t); 393 void moea64_copy_page(mmu_t, vm_page_t, vm_page_t); 394 void moea64_enter(mmu_t, pmap_t, vm_offset_t, vm_page_t, vm_prot_t, boolean_t); 395 void moea64_enter_object(mmu_t, pmap_t, vm_offset_t, vm_offset_t, vm_page_t, 396 vm_prot_t); 397 void moea64_enter_quick(mmu_t, pmap_t, vm_offset_t, vm_page_t, vm_prot_t); 398 vm_paddr_t moea64_extract(mmu_t, pmap_t, vm_offset_t); 399 vm_page_t moea64_extract_and_hold(mmu_t, pmap_t, vm_offset_t, vm_prot_t); 400 void moea64_init(mmu_t); 401 boolean_t moea64_is_modified(mmu_t, vm_page_t); 402 boolean_t moea64_is_prefaultable(mmu_t, pmap_t, vm_offset_t); 403 boolean_t moea64_is_referenced(mmu_t, vm_page_t); 404 boolean_t moea64_ts_referenced(mmu_t, vm_page_t); 405 vm_offset_t moea64_map(mmu_t, vm_offset_t *, vm_offset_t, vm_offset_t, int); 406 boolean_t moea64_page_exists_quick(mmu_t, pmap_t, vm_page_t); 407 int moea64_page_wired_mappings(mmu_t, vm_page_t); 408 void moea64_pinit(mmu_t, pmap_t); 409 void moea64_pinit0(mmu_t, pmap_t); 410 void moea64_protect(mmu_t, pmap_t, vm_offset_t, vm_offset_t, vm_prot_t); 411 void moea64_qenter(mmu_t, vm_offset_t, vm_page_t *, int); 412 void moea64_qremove(mmu_t, vm_offset_t, int); 413 void moea64_release(mmu_t, pmap_t); 414 void moea64_remove(mmu_t, pmap_t, vm_offset_t, vm_offset_t); 415 void moea64_remove_all(mmu_t, vm_page_t); 416 void moea64_remove_write(mmu_t, vm_page_t); 417 void moea64_zero_page(mmu_t, vm_page_t); 418 void moea64_zero_page_area(mmu_t, vm_page_t, int, int); 419 void moea64_zero_page_idle(mmu_t, vm_page_t); 420 void moea64_activate(mmu_t, struct thread *); 421 void moea64_deactivate(mmu_t, struct thread *); 422 void *moea64_mapdev(mmu_t, vm_offset_t, vm_size_t); 423 void *moea64_mapdev_attr(mmu_t, vm_offset_t, vm_size_t, vm_memattr_t); 424 void moea64_unmapdev(mmu_t, vm_offset_t, vm_size_t); 425 vm_offset_t moea64_kextract(mmu_t, vm_offset_t); 426 void moea64_page_set_memattr(mmu_t, vm_page_t m, vm_memattr_t ma); 427 void moea64_kenter_attr(mmu_t, vm_offset_t, vm_offset_t, vm_memattr_t ma); 428 void moea64_kenter(mmu_t, vm_offset_t, vm_offset_t); 429 boolean_t moea64_dev_direct_mapped(mmu_t, vm_offset_t, vm_size_t); 430 static void moea64_sync_icache(mmu_t, pmap_t, vm_offset_t, vm_size_t); 431 432 static mmu_method_t moea64_methods[] = { 433 MMUMETHOD(mmu_change_wiring, moea64_change_wiring), 434 MMUMETHOD(mmu_clear_modify, moea64_clear_modify), 435 MMUMETHOD(mmu_clear_reference, moea64_clear_reference), 436 MMUMETHOD(mmu_copy_page, moea64_copy_page), 437 MMUMETHOD(mmu_enter, moea64_enter), 438 MMUMETHOD(mmu_enter_object, moea64_enter_object), 439 MMUMETHOD(mmu_enter_quick, moea64_enter_quick), 440 MMUMETHOD(mmu_extract, moea64_extract), 441 MMUMETHOD(mmu_extract_and_hold, moea64_extract_and_hold), 442 MMUMETHOD(mmu_init, moea64_init), 443 MMUMETHOD(mmu_is_modified, moea64_is_modified), 444 MMUMETHOD(mmu_is_prefaultable, moea64_is_prefaultable), 445 MMUMETHOD(mmu_is_referenced, moea64_is_referenced), 446 MMUMETHOD(mmu_ts_referenced, moea64_ts_referenced), 447 MMUMETHOD(mmu_map, moea64_map), 448 MMUMETHOD(mmu_page_exists_quick,moea64_page_exists_quick), 449 MMUMETHOD(mmu_page_wired_mappings,moea64_page_wired_mappings), 450 MMUMETHOD(mmu_pinit, moea64_pinit), 451 MMUMETHOD(mmu_pinit0, moea64_pinit0), 452 MMUMETHOD(mmu_protect, moea64_protect), 453 MMUMETHOD(mmu_qenter, moea64_qenter), 454 MMUMETHOD(mmu_qremove, moea64_qremove), 455 MMUMETHOD(mmu_release, moea64_release), 456 MMUMETHOD(mmu_remove, moea64_remove), 457 MMUMETHOD(mmu_remove_all, moea64_remove_all), 458 MMUMETHOD(mmu_remove_write, moea64_remove_write), 459 MMUMETHOD(mmu_sync_icache, moea64_sync_icache), 460 MMUMETHOD(mmu_zero_page, moea64_zero_page), 461 MMUMETHOD(mmu_zero_page_area, moea64_zero_page_area), 462 MMUMETHOD(mmu_zero_page_idle, moea64_zero_page_idle), 463 MMUMETHOD(mmu_activate, moea64_activate), 464 MMUMETHOD(mmu_deactivate, moea64_deactivate), 465 MMUMETHOD(mmu_page_set_memattr, moea64_page_set_memattr), 466 467 /* Internal interfaces */ 468 MMUMETHOD(mmu_bootstrap, moea64_bootstrap), 469 MMUMETHOD(mmu_cpu_bootstrap, moea64_cpu_bootstrap), 470 MMUMETHOD(mmu_mapdev, moea64_mapdev), 471 MMUMETHOD(mmu_mapdev_attr, moea64_mapdev_attr), 472 MMUMETHOD(mmu_unmapdev, moea64_unmapdev), 473 MMUMETHOD(mmu_kextract, moea64_kextract), 474 MMUMETHOD(mmu_kenter, moea64_kenter), 475 MMUMETHOD(mmu_kenter_attr, moea64_kenter_attr), 476 MMUMETHOD(mmu_dev_direct_mapped,moea64_dev_direct_mapped), 477 478 { 0, 0 } 479 }; 480 481 MMU_DEF(oea64_mmu, MMU_TYPE_G5, moea64_methods, 0); 482 483 static __inline u_int 484 va_to_pteg(uint64_t vsid, vm_offset_t addr, int large) 485 { 486 uint64_t hash; 487 int shift; 488 489 shift = large ? moea64_large_page_shift : ADDR_PIDX_SHFT; 490 hash = (vsid & VSID_HASH_MASK) ^ (((uint64_t)addr & ADDR_PIDX) >> 491 shift); 492 return (hash & moea64_pteg_mask); 493 } 494 495 static __inline struct pvo_head * 496 vm_page_to_pvoh(vm_page_t m) 497 { 498 499 return (&m->md.mdpg_pvoh); 500 } 501 502 static __inline void 503 moea64_attr_clear(vm_page_t m, u_int64_t ptebit) 504 { 505 506 mtx_assert(&vm_page_queue_mtx, MA_OWNED); 507 m->md.mdpg_attrs &= ~ptebit; 508 } 509 510 static __inline u_int64_t 511 moea64_attr_fetch(vm_page_t m) 512 { 513 514 return (m->md.mdpg_attrs); 515 } 516 517 static __inline void 518 moea64_attr_save(vm_page_t m, u_int64_t ptebit) 519 { 520 521 mtx_assert(&vm_page_queue_mtx, MA_OWNED); 522 m->md.mdpg_attrs |= ptebit; 523 } 524 525 static __inline void 526 moea64_pte_create(struct lpte *pt, uint64_t vsid, vm_offset_t va, 527 uint64_t pte_lo, int flags) 528 { 529 530 ASSERT_TABLE_LOCK(); 531 532 /* 533 * Construct a PTE. Default to IMB initially. Valid bit only gets 534 * set when the real pte is set in memory. 535 * 536 * Note: Don't set the valid bit for correct operation of tlb update. 537 */ 538 pt->pte_hi = (vsid << LPTE_VSID_SHIFT) | 539 (((uint64_t)(va & ADDR_PIDX) >> ADDR_API_SHFT64) & LPTE_API); 540 541 if (flags & PVO_LARGE) 542 pt->pte_hi |= LPTE_BIG; 543 544 pt->pte_lo = pte_lo; 545 } 546 547 static __inline void 548 moea64_pte_synch(struct lpte *pt, struct lpte *pvo_pt) 549 { 550 551 ASSERT_TABLE_LOCK(); 552 553 pvo_pt->pte_lo |= pt->pte_lo & (LPTE_REF | LPTE_CHG); 554 } 555 556 static __inline void 557 moea64_pte_clear(struct lpte *pt, uint64_t vpn, u_int64_t ptebit) 558 { 559 ASSERT_TABLE_LOCK(); 560 561 /* 562 * As shown in Section 7.6.3.2.3 563 */ 564 pt->pte_lo &= ~ptebit; 565 TLBIE(vpn); 566 } 567 568 static __inline void 569 moea64_pte_set(struct lpte *pt, struct lpte *pvo_pt) 570 { 571 572 ASSERT_TABLE_LOCK(); 573 pvo_pt->pte_hi |= LPTE_VALID; 574 575 /* 576 * Update the PTE as defined in section 7.6.3.1. 577 * Note that the REF/CHG bits are from pvo_pt and thus should have 578 * been saved so this routine can restore them (if desired). 579 */ 580 pt->pte_lo = pvo_pt->pte_lo; 581 EIEIO(); 582 pt->pte_hi = pvo_pt->pte_hi; 583 PTESYNC(); 584 moea64_pte_valid++; 585 } 586 587 static __inline void 588 moea64_pte_unset(struct lpte *pt, struct lpte *pvo_pt, uint64_t vpn) 589 { 590 ASSERT_TABLE_LOCK(); 591 pvo_pt->pte_hi &= ~LPTE_VALID; 592 593 /* 594 * Force the reg & chg bits back into the PTEs. 595 */ 596 SYNC(); 597 598 /* 599 * Invalidate the pte. 600 */ 601 pt->pte_hi &= ~LPTE_VALID; 602 TLBIE(vpn); 603 604 /* 605 * Save the reg & chg bits. 606 */ 607 moea64_pte_synch(pt, pvo_pt); 608 moea64_pte_valid--; 609 } 610 611 static __inline void 612 moea64_pte_change(struct lpte *pt, struct lpte *pvo_pt, uint64_t vpn) 613 { 614 615 /* 616 * Invalidate the PTE 617 */ 618 moea64_pte_unset(pt, pvo_pt, vpn); 619 moea64_pte_set(pt, pvo_pt); 620 } 621 622 static __inline uint64_t 623 moea64_calc_wimg(vm_offset_t pa, vm_memattr_t ma) 624 { 625 uint64_t pte_lo; 626 int i; 627 628 if (ma != VM_MEMATTR_DEFAULT) { 629 switch (ma) { 630 case VM_MEMATTR_UNCACHEABLE: 631 return (LPTE_I | LPTE_G); 632 case VM_MEMATTR_WRITE_COMBINING: 633 case VM_MEMATTR_WRITE_BACK: 634 case VM_MEMATTR_PREFETCHABLE: 635 return (LPTE_I); 636 case VM_MEMATTR_WRITE_THROUGH: 637 return (LPTE_W | LPTE_M); 638 } 639 } 640 641 /* 642 * Assume the page is cache inhibited and access is guarded unless 643 * it's in our available memory array. 644 */ 645 pte_lo = LPTE_I | LPTE_G; 646 for (i = 0; i < pregions_sz; i++) { 647 if ((pa >= pregions[i].mr_start) && 648 (pa < (pregions[i].mr_start + pregions[i].mr_size))) { 649 pte_lo &= ~(LPTE_I | LPTE_G); 650 pte_lo |= LPTE_M; 651 break; 652 } 653 } 654 655 return pte_lo; 656 } 657 658 /* 659 * Quick sort callout for comparing memory regions. 660 */ 661 static int mr_cmp(const void *a, const void *b); 662 static int om_cmp(const void *a, const void *b); 663 664 static int 665 mr_cmp(const void *a, const void *b) 666 { 667 const struct mem_region *regiona; 668 const struct mem_region *regionb; 669 670 regiona = a; 671 regionb = b; 672 if (regiona->mr_start < regionb->mr_start) 673 return (-1); 674 else if (regiona->mr_start > regionb->mr_start) 675 return (1); 676 else 677 return (0); 678 } 679 680 static int 681 om_cmp(const void *a, const void *b) 682 { 683 const struct ofw_map *mapa; 684 const struct ofw_map *mapb; 685 686 mapa = a; 687 mapb = b; 688 if (mapa->om_pa_hi < mapb->om_pa_hi) 689 return (-1); 690 else if (mapa->om_pa_hi > mapb->om_pa_hi) 691 return (1); 692 else if (mapa->om_pa_lo < mapb->om_pa_lo) 693 return (-1); 694 else if (mapa->om_pa_lo > mapb->om_pa_lo) 695 return (1); 696 else 697 return (0); 698 } 699 700 static void 701 moea64_cpu_bootstrap(mmu_t mmup, int ap) 702 { 703 int i = 0; 704 #ifdef __powerpc64__ 705 struct slb *slb = PCPU_GET(slb); 706 #endif 707 708 /* 709 * Initialize segment registers and MMU 710 */ 711 712 mtmsr(mfmsr() & ~PSL_DR & ~PSL_IR); isync(); 713 714 /* 715 * Install kernel SLB entries 716 */ 717 718 #ifdef __powerpc64__ 719 slbia(); 720 721 for (i = 0; i < 64; i++) { 722 if (!(slb[i].slbe & SLBE_VALID)) 723 continue; 724 725 __asm __volatile ("slbmte %0, %1" :: 726 "r"(slb[i].slbv), "r"(slb[i].slbe)); 727 } 728 #else 729 for (i = 0; i < 16; i++) 730 mtsrin(i << ADDR_SR_SHFT, kernel_pmap->pm_sr[i]); 731 #endif 732 733 /* 734 * Install page table 735 */ 736 737 __asm __volatile ("ptesync; mtsdr1 %0; isync" 738 :: "r"((uintptr_t)moea64_pteg_table 739 | (64 - cntlzd(moea64_pteg_mask >> 11)))); 740 tlbia(); 741 } 742 743 static void 744 moea64_add_ofw_mappings(mmu_t mmup, phandle_t mmu, size_t sz) 745 { 746 struct ofw_map translations[sz/sizeof(struct ofw_map)]; 747 register_t msr; 748 vm_offset_t off; 749 vm_paddr_t pa_base; 750 int i, ofw_mappings; 751 752 bzero(translations, sz); 753 if (OF_getprop(mmu, "translations", translations, sz) == -1) 754 panic("moea64_bootstrap: can't get ofw translations"); 755 756 CTR0(KTR_PMAP, "moea64_add_ofw_mappings: translations"); 757 sz /= sizeof(*translations); 758 qsort(translations, sz, sizeof (*translations), om_cmp); 759 760 for (i = 0, ofw_mappings = 0; i < sz; i++) { 761 CTR3(KTR_PMAP, "translation: pa=%#x va=%#x len=%#x", 762 (uint32_t)(translations[i].om_pa_lo), translations[i].om_va, 763 translations[i].om_len); 764 765 if (translations[i].om_pa_lo % PAGE_SIZE) 766 panic("OFW translation not page-aligned!"); 767 768 pa_base = translations[i].om_pa_lo; 769 770 #ifdef __powerpc64__ 771 pa_base += (vm_offset_t)translations[i].om_pa_hi << 32; 772 #else 773 if (translations[i].om_pa_hi) 774 panic("OFW translations above 32-bit boundary!"); 775 #endif 776 777 /* Now enter the pages for this mapping */ 778 779 DISABLE_TRANS(msr); 780 for (off = 0; off < translations[i].om_len; off += PAGE_SIZE) { 781 if (moea64_pvo_find_va(kernel_pmap, 782 translations[i].om_va + off) != NULL) 783 continue; 784 785 moea64_kenter(mmup, translations[i].om_va + off, 786 pa_base + off); 787 788 ofw_mappings++; 789 } 790 ENABLE_TRANS(msr); 791 } 792 } 793 794 #ifdef __powerpc64__ 795 static void 796 moea64_probe_large_page(void) 797 { 798 uint16_t pvr = mfpvr() >> 16; 799 800 switch (pvr) { 801 case IBM970: 802 case IBM970FX: 803 case IBM970MP: 804 powerpc_sync(); isync(); 805 mtspr(SPR_HID4, mfspr(SPR_HID4) & ~HID4_970_DISABLE_LG_PG); 806 powerpc_sync(); isync(); 807 808 /* FALLTHROUGH */ 809 case IBMCELLBE: 810 moea64_large_page_size = 0x1000000; /* 16 MB */ 811 moea64_large_page_shift = 24; 812 break; 813 default: 814 moea64_large_page_size = 0; 815 } 816 817 moea64_large_page_mask = moea64_large_page_size - 1; 818 } 819 820 static void 821 moea64_bootstrap_slb_prefault(vm_offset_t va, int large) 822 { 823 struct slb *cache; 824 struct slb entry; 825 uint64_t esid, slbe; 826 uint64_t i; 827 828 cache = PCPU_GET(slb); 829 esid = va >> ADDR_SR_SHFT; 830 slbe = (esid << SLBE_ESID_SHIFT) | SLBE_VALID; 831 832 for (i = 0; i < 64; i++) { 833 if (cache[i].slbe == (slbe | i)) 834 return; 835 } 836 837 entry.slbe = slbe; 838 entry.slbv = KERNEL_VSID(esid) << SLBV_VSID_SHIFT; 839 if (large) 840 entry.slbv |= SLBV_L; 841 842 slb_insert_kernel(entry.slbe, entry.slbv); 843 } 844 #endif 845 846 static void 847 moea64_setup_direct_map(mmu_t mmup, vm_offset_t kernelstart, 848 vm_offset_t kernelend) 849 { 850 register_t msr; 851 vm_paddr_t pa; 852 vm_offset_t size, off; 853 uint64_t pte_lo; 854 int i; 855 856 if (moea64_large_page_size == 0) 857 hw_direct_map = 0; 858 859 DISABLE_TRANS(msr); 860 if (hw_direct_map) { 861 PMAP_LOCK(kernel_pmap); 862 for (i = 0; i < pregions_sz; i++) { 863 for (pa = pregions[i].mr_start; pa < pregions[i].mr_start + 864 pregions[i].mr_size; pa += moea64_large_page_size) { 865 pte_lo = LPTE_M; 866 867 /* 868 * Set memory access as guarded if prefetch within 869 * the page could exit the available physmem area. 870 */ 871 if (pa & moea64_large_page_mask) { 872 pa &= moea64_large_page_mask; 873 pte_lo |= LPTE_G; 874 } 875 if (pa + moea64_large_page_size > 876 pregions[i].mr_start + pregions[i].mr_size) 877 pte_lo |= LPTE_G; 878 879 moea64_pvo_enter(kernel_pmap, moea64_upvo_zone, 880 &moea64_pvo_kunmanaged, pa, pa, 881 pte_lo, PVO_WIRED | PVO_LARGE | 882 VM_PROT_EXECUTE); 883 } 884 } 885 PMAP_UNLOCK(kernel_pmap); 886 } else { 887 size = moea64_pteg_count * sizeof(struct lpteg); 888 off = (vm_offset_t)(moea64_pteg_table); 889 for (pa = off; pa < off + size; pa += PAGE_SIZE) 890 moea64_kenter(mmup, pa, pa); 891 size = sizeof(struct pvo_head) * moea64_pteg_count; 892 off = (vm_offset_t)(moea64_pvo_table); 893 for (pa = off; pa < off + size; pa += PAGE_SIZE) 894 moea64_kenter(mmup, pa, pa); 895 size = BPVO_POOL_SIZE*sizeof(struct pvo_entry); 896 off = (vm_offset_t)(moea64_bpvo_pool); 897 for (pa = off; pa < off + size; pa += PAGE_SIZE) 898 moea64_kenter(mmup, pa, pa); 899 900 /* 901 * Map certain important things, like ourselves. 902 * 903 * NOTE: We do not map the exception vector space. That code is 904 * used only in real mode, and leaving it unmapped allows us to 905 * catch NULL pointer deferences, instead of making NULL a valid 906 * address. 907 */ 908 909 for (pa = kernelstart & ~PAGE_MASK; pa < kernelend; 910 pa += PAGE_SIZE) 911 moea64_kenter(mmup, pa, pa); 912 } 913 ENABLE_TRANS(msr); 914 } 915 916 static void 917 moea64_bootstrap(mmu_t mmup, vm_offset_t kernelstart, vm_offset_t kernelend) 918 { 919 ihandle_t mmui; 920 phandle_t chosen; 921 phandle_t mmu; 922 size_t sz; 923 int i, j; 924 vm_size_t size, physsz, hwphyssz; 925 vm_offset_t pa, va; 926 register_t msr; 927 void *dpcpu; 928 929 #ifndef __powerpc64__ 930 /* We don't have a direct map since there is no BAT */ 931 hw_direct_map = 0; 932 933 /* Make sure battable is zero, since we have no BAT */ 934 for (i = 0; i < 16; i++) { 935 battable[i].batu = 0; 936 battable[i].batl = 0; 937 } 938 #else 939 moea64_probe_large_page(); 940 941 /* Use a direct map if we have large page support */ 942 if (moea64_large_page_size > 0) 943 hw_direct_map = 1; 944 else 945 hw_direct_map = 0; 946 #endif 947 948 /* Get physical memory regions from firmware */ 949 mem_regions(&pregions, &pregions_sz, ®ions, ®ions_sz); 950 CTR0(KTR_PMAP, "moea64_bootstrap: physical memory"); 951 952 qsort(pregions, pregions_sz, sizeof(*pregions), mr_cmp); 953 if (sizeof(phys_avail)/sizeof(phys_avail[0]) < regions_sz) 954 panic("moea64_bootstrap: phys_avail too small"); 955 qsort(regions, regions_sz, sizeof(*regions), mr_cmp); 956 phys_avail_count = 0; 957 physsz = 0; 958 hwphyssz = 0; 959 TUNABLE_ULONG_FETCH("hw.physmem", (u_long *) &hwphyssz); 960 for (i = 0, j = 0; i < regions_sz; i++, j += 2) { 961 CTR3(KTR_PMAP, "region: %#x - %#x (%#x)", regions[i].mr_start, 962 regions[i].mr_start + regions[i].mr_size, 963 regions[i].mr_size); 964 if (hwphyssz != 0 && 965 (physsz + regions[i].mr_size) >= hwphyssz) { 966 if (physsz < hwphyssz) { 967 phys_avail[j] = regions[i].mr_start; 968 phys_avail[j + 1] = regions[i].mr_start + 969 hwphyssz - physsz; 970 physsz = hwphyssz; 971 phys_avail_count++; 972 } 973 break; 974 } 975 phys_avail[j] = regions[i].mr_start; 976 phys_avail[j + 1] = regions[i].mr_start + regions[i].mr_size; 977 phys_avail_count++; 978 physsz += regions[i].mr_size; 979 } 980 981 /* Check for overlap with the kernel and exception vectors */ 982 for (j = 0; j < 2*phys_avail_count; j+=2) { 983 if (phys_avail[j] < EXC_LAST) 984 phys_avail[j] += EXC_LAST; 985 986 if (kernelstart >= phys_avail[j] && 987 kernelstart < phys_avail[j+1]) { 988 if (kernelend < phys_avail[j+1]) { 989 phys_avail[2*phys_avail_count] = 990 (kernelend & ~PAGE_MASK) + PAGE_SIZE; 991 phys_avail[2*phys_avail_count + 1] = 992 phys_avail[j+1]; 993 phys_avail_count++; 994 } 995 996 phys_avail[j+1] = kernelstart & ~PAGE_MASK; 997 } 998 999 if (kernelend >= phys_avail[j] && 1000 kernelend < phys_avail[j+1]) { 1001 if (kernelstart > phys_avail[j]) { 1002 phys_avail[2*phys_avail_count] = phys_avail[j]; 1003 phys_avail[2*phys_avail_count + 1] = 1004 kernelstart & ~PAGE_MASK; 1005 phys_avail_count++; 1006 } 1007 1008 phys_avail[j] = (kernelend & ~PAGE_MASK) + PAGE_SIZE; 1009 } 1010 } 1011 1012 physmem = btoc(physsz); 1013 1014 /* 1015 * Allocate PTEG table. 1016 */ 1017 #ifdef PTEGCOUNT 1018 moea64_pteg_count = PTEGCOUNT; 1019 #else 1020 moea64_pteg_count = 0x1000; 1021 1022 while (moea64_pteg_count < physmem) 1023 moea64_pteg_count <<= 1; 1024 1025 moea64_pteg_count >>= 1; 1026 #endif /* PTEGCOUNT */ 1027 1028 size = moea64_pteg_count * sizeof(struct lpteg); 1029 CTR2(KTR_PMAP, "moea64_bootstrap: %d PTEGs, %d bytes", 1030 moea64_pteg_count, size); 1031 1032 /* 1033 * We now need to allocate memory. This memory, to be allocated, 1034 * has to reside in a page table. The page table we are about to 1035 * allocate. We don't have BAT. So drop to data real mode for a minute 1036 * as a measure of last resort. We do this a couple times. 1037 */ 1038 1039 moea64_pteg_table = (struct lpteg *)moea64_bootstrap_alloc(size, size); 1040 DISABLE_TRANS(msr); 1041 bzero((void *)moea64_pteg_table, moea64_pteg_count * sizeof(struct lpteg)); 1042 ENABLE_TRANS(msr); 1043 1044 moea64_pteg_mask = moea64_pteg_count - 1; 1045 1046 CTR1(KTR_PMAP, "moea64_bootstrap: PTEG table at %p", moea64_pteg_table); 1047 1048 /* 1049 * Allocate pv/overflow lists. 1050 */ 1051 size = sizeof(struct pvo_head) * moea64_pteg_count; 1052 1053 moea64_pvo_table = (struct pvo_head *)moea64_bootstrap_alloc(size, 1054 PAGE_SIZE); 1055 CTR1(KTR_PMAP, "moea64_bootstrap: PVO table at %p", moea64_pvo_table); 1056 1057 DISABLE_TRANS(msr); 1058 for (i = 0; i < moea64_pteg_count; i++) 1059 LIST_INIT(&moea64_pvo_table[i]); 1060 ENABLE_TRANS(msr); 1061 1062 /* 1063 * Initialize the lock that synchronizes access to the pteg and pvo 1064 * tables. 1065 */ 1066 mtx_init(&moea64_table_mutex, "pmap table", NULL, MTX_DEF | 1067 MTX_RECURSE); 1068 mtx_init(&moea64_slb_mutex, "SLB table", NULL, MTX_DEF); 1069 1070 /* 1071 * Initialize the TLBIE lock. TLBIE can only be executed by one CPU. 1072 */ 1073 mtx_init(&tlbie_mutex, "tlbie mutex", NULL, MTX_SPIN); 1074 1075 /* 1076 * Initialise the unmanaged pvo pool. 1077 */ 1078 moea64_bpvo_pool = (struct pvo_entry *)moea64_bootstrap_alloc( 1079 BPVO_POOL_SIZE*sizeof(struct pvo_entry), 0); 1080 moea64_bpvo_pool_index = 0; 1081 1082 /* 1083 * Make sure kernel vsid is allocated as well as VSID 0. 1084 */ 1085 #ifndef __powerpc64__ 1086 moea64_vsid_bitmap[(KERNEL_VSIDBITS & (NVSIDS - 1)) / VSID_NBPW] 1087 |= 1 << (KERNEL_VSIDBITS % VSID_NBPW); 1088 moea64_vsid_bitmap[0] |= 1; 1089 #endif 1090 1091 /* 1092 * Initialize the kernel pmap (which is statically allocated). 1093 */ 1094 #ifdef __powerpc64__ 1095 for (i = 0; i < 64; i++) { 1096 pcpup->pc_slb[i].slbv = 0; 1097 pcpup->pc_slb[i].slbe = 0; 1098 } 1099 #else 1100 for (i = 0; i < 16; i++) 1101 kernel_pmap->pm_sr[i] = EMPTY_SEGMENT + i; 1102 #endif 1103 1104 kernel_pmap->pmap_phys = kernel_pmap; 1105 kernel_pmap->pm_active = ~0; 1106 1107 PMAP_LOCK_INIT(kernel_pmap); 1108 1109 /* 1110 * Now map in all the other buffers we allocated earlier 1111 */ 1112 1113 moea64_setup_direct_map(mmup, kernelstart, kernelend); 1114 1115 /* 1116 * Set up the Open Firmware pmap and add its mappings if not in real 1117 * mode. 1118 */ 1119 1120 chosen = OF_finddevice("/chosen"); 1121 if (chosen != -1 && OF_getprop(chosen, "mmu", &mmui, 4) != -1) { 1122 #ifndef __powerpc64__ 1123 moea64_pinit(mmup, &ofw_pmap); 1124 1125 for (i = 0; i < 16; i++) 1126 ofw_pmap.pm_sr[i] = kernel_pmap->pm_sr[i]; 1127 #endif 1128 1129 if ((mmu = OF_instance_to_package(mmui)) == -1) 1130 panic("moea64_bootstrap: can't get mmu package"); 1131 if ((sz = OF_getproplen(mmu, "translations")) == -1) 1132 panic("moea64_bootstrap: can't get ofw translation count"); 1133 if (sz > 6144 /* tmpstksz - 2 KB headroom */) 1134 panic("moea64_bootstrap: too many ofw translations"); 1135 1136 moea64_add_ofw_mappings(mmup, mmu, sz); 1137 } 1138 1139 #ifdef SMP 1140 TLBSYNC(); 1141 #endif 1142 1143 /* 1144 * Calculate the last available physical address. 1145 */ 1146 for (i = 0; phys_avail[i + 2] != 0; i += 2) 1147 ; 1148 Maxmem = powerpc_btop(phys_avail[i + 1]); 1149 1150 /* 1151 * Initialize MMU and remap early physical mappings 1152 */ 1153 moea64_cpu_bootstrap(mmup,0); 1154 mtmsr(mfmsr() | PSL_DR | PSL_IR); isync(); 1155 pmap_bootstrapped++; 1156 bs_remap_earlyboot(); 1157 1158 /* 1159 * Set the start and end of kva. 1160 */ 1161 virtual_avail = VM_MIN_KERNEL_ADDRESS; 1162 virtual_end = VM_MAX_SAFE_KERNEL_ADDRESS; 1163 1164 /* 1165 * Map the entire KVA range into the SLB. We must not fault there. 1166 */ 1167 #ifdef __powerpc64__ 1168 for (va = virtual_avail; va < virtual_end; va += SEGMENT_LENGTH) 1169 moea64_bootstrap_slb_prefault(va, 0); 1170 #endif 1171 1172 /* 1173 * Figure out how far we can extend virtual_end into segment 16 1174 * without running into existing mappings. Segment 16 is guaranteed 1175 * to contain neither RAM nor devices (at least on Apple hardware), 1176 * but will generally contain some OFW mappings we should not 1177 * step on. 1178 */ 1179 1180 #ifndef __powerpc64__ /* KVA is in high memory on PPC64 */ 1181 PMAP_LOCK(kernel_pmap); 1182 while (virtual_end < VM_MAX_KERNEL_ADDRESS && 1183 moea64_pvo_find_va(kernel_pmap, virtual_end+1) == NULL) 1184 virtual_end += PAGE_SIZE; 1185 PMAP_UNLOCK(kernel_pmap); 1186 #endif 1187 1188 /* 1189 * Allocate some things for page zeroing. We put this directly 1190 * in the page table, marked with LPTE_LOCKED, to avoid any 1191 * of the PVO book-keeping or other parts of the VM system 1192 * from even knowing that this hack exists. 1193 */ 1194 1195 if (!hw_direct_map) { 1196 mtx_init(&moea64_scratchpage_mtx, "pvo zero page", NULL, 1197 MTX_DEF); 1198 for (i = 0; i < 2; i++) { 1199 struct lpte pt; 1200 uint64_t vsid; 1201 int pteidx, ptegidx; 1202 1203 moea64_scratchpage_va[i] = (virtual_end+1) - PAGE_SIZE; 1204 virtual_end -= PAGE_SIZE; 1205 1206 LOCK_TABLE(); 1207 1208 vsid = va_to_vsid(kernel_pmap, 1209 moea64_scratchpage_va[i]); 1210 moea64_pte_create(&pt, vsid, moea64_scratchpage_va[i], 1211 LPTE_NOEXEC, 0); 1212 pt.pte_hi |= LPTE_LOCKED; 1213 1214 moea64_scratchpage_vpn[i] = (vsid << 16) | 1215 ((moea64_scratchpage_va[i] & ADDR_PIDX) >> 1216 ADDR_PIDX_SHFT); 1217 ptegidx = va_to_pteg(vsid, moea64_scratchpage_va[i], 0); 1218 pteidx = moea64_pte_insert(ptegidx, &pt); 1219 if (pt.pte_hi & LPTE_HID) 1220 ptegidx ^= moea64_pteg_mask; 1221 1222 moea64_scratchpage_pte[i] = 1223 &moea64_pteg_table[ptegidx].pt[pteidx]; 1224 1225 UNLOCK_TABLE(); 1226 } 1227 } 1228 1229 /* 1230 * Allocate a kernel stack with a guard page for thread0 and map it 1231 * into the kernel page map. 1232 */ 1233 pa = moea64_bootstrap_alloc(KSTACK_PAGES * PAGE_SIZE, PAGE_SIZE); 1234 va = virtual_avail + KSTACK_GUARD_PAGES * PAGE_SIZE; 1235 virtual_avail = va + KSTACK_PAGES * PAGE_SIZE; 1236 CTR2(KTR_PMAP, "moea_bootstrap: kstack0 at %#x (%#x)", pa, va); 1237 thread0.td_kstack = va; 1238 thread0.td_kstack_pages = KSTACK_PAGES; 1239 for (i = 0; i < KSTACK_PAGES; i++) { 1240 moea64_kenter(mmup, va, pa); 1241 pa += PAGE_SIZE; 1242 va += PAGE_SIZE; 1243 } 1244 1245 /* 1246 * Allocate virtual address space for the message buffer. 1247 */ 1248 pa = msgbuf_phys = moea64_bootstrap_alloc(MSGBUF_SIZE, PAGE_SIZE); 1249 msgbufp = (struct msgbuf *)virtual_avail; 1250 va = virtual_avail; 1251 virtual_avail += round_page(MSGBUF_SIZE); 1252 while (va < virtual_avail) { 1253 moea64_kenter(mmup, va, pa); 1254 pa += PAGE_SIZE; 1255 va += PAGE_SIZE; 1256 } 1257 1258 /* 1259 * Allocate virtual address space for the dynamic percpu area. 1260 */ 1261 pa = moea64_bootstrap_alloc(DPCPU_SIZE, PAGE_SIZE); 1262 dpcpu = (void *)virtual_avail; 1263 va = virtual_avail; 1264 virtual_avail += DPCPU_SIZE; 1265 while (va < virtual_avail) { 1266 moea64_kenter(mmup, va, pa); 1267 pa += PAGE_SIZE; 1268 va += PAGE_SIZE; 1269 } 1270 dpcpu_init(dpcpu, 0); 1271 } 1272 1273 /* 1274 * Activate a user pmap. The pmap must be activated before its address 1275 * space can be accessed in any way. 1276 */ 1277 void 1278 moea64_activate(mmu_t mmu, struct thread *td) 1279 { 1280 pmap_t pm; 1281 1282 pm = &td->td_proc->p_vmspace->vm_pmap; 1283 pm->pm_active |= PCPU_GET(cpumask); 1284 1285 #ifdef __powerpc64__ 1286 PCPU_SET(userslb, pm->pm_slb); 1287 #else 1288 PCPU_SET(curpmap, pm->pmap_phys); 1289 #endif 1290 } 1291 1292 void 1293 moea64_deactivate(mmu_t mmu, struct thread *td) 1294 { 1295 pmap_t pm; 1296 1297 pm = &td->td_proc->p_vmspace->vm_pmap; 1298 pm->pm_active &= ~(PCPU_GET(cpumask)); 1299 #ifdef __powerpc64__ 1300 PCPU_SET(userslb, NULL); 1301 #else 1302 PCPU_SET(curpmap, NULL); 1303 #endif 1304 } 1305 1306 void 1307 moea64_change_wiring(mmu_t mmu, pmap_t pm, vm_offset_t va, boolean_t wired) 1308 { 1309 struct pvo_entry *pvo; 1310 struct lpte *pt; 1311 uint64_t vsid; 1312 int i, ptegidx; 1313 1314 PMAP_LOCK(pm); 1315 pvo = moea64_pvo_find_va(pm, va & ~ADDR_POFF); 1316 1317 if (pvo != NULL) { 1318 LOCK_TABLE(); 1319 pt = moea64_pvo_to_pte(pvo); 1320 1321 if (wired) { 1322 if ((pvo->pvo_vaddr & PVO_WIRED) == 0) 1323 pm->pm_stats.wired_count++; 1324 pvo->pvo_vaddr |= PVO_WIRED; 1325 pvo->pvo_pte.lpte.pte_hi |= LPTE_WIRED; 1326 } else { 1327 if ((pvo->pvo_vaddr & PVO_WIRED) != 0) 1328 pm->pm_stats.wired_count--; 1329 pvo->pvo_vaddr &= ~PVO_WIRED; 1330 pvo->pvo_pte.lpte.pte_hi &= ~LPTE_WIRED; 1331 } 1332 1333 if (pt != NULL) { 1334 /* Update wiring flag in page table. */ 1335 moea64_pte_change(pt, &pvo->pvo_pte.lpte, 1336 pvo->pvo_vpn); 1337 } else if (wired) { 1338 /* 1339 * If we are wiring the page, and it wasn't in the 1340 * page table before, add it. 1341 */ 1342 vsid = PVO_VSID(pvo); 1343 ptegidx = va_to_pteg(vsid, PVO_VADDR(pvo), 1344 pvo->pvo_vaddr & PVO_LARGE); 1345 1346 i = moea64_pte_insert(ptegidx, &pvo->pvo_pte.lpte); 1347 if (i >= 0) { 1348 PVO_PTEGIDX_CLR(pvo); 1349 PVO_PTEGIDX_SET(pvo, i); 1350 } 1351 } 1352 1353 UNLOCK_TABLE(); 1354 } 1355 PMAP_UNLOCK(pm); 1356 } 1357 1358 /* 1359 * This goes through and sets the physical address of our 1360 * special scratch PTE to the PA we want to zero or copy. Because 1361 * of locking issues (this can get called in pvo_enter() by 1362 * the UMA allocator), we can't use most other utility functions here 1363 */ 1364 1365 static __inline 1366 void moea64_set_scratchpage_pa(int which, vm_offset_t pa) { 1367 1368 KASSERT(!hw_direct_map, ("Using OEA64 scratchpage with a direct map!")); 1369 mtx_assert(&moea64_scratchpage_mtx, MA_OWNED); 1370 1371 moea64_scratchpage_pte[which]->pte_hi &= ~LPTE_VALID; 1372 TLBIE(moea64_scratchpage_vpn[which]); 1373 1374 moea64_scratchpage_pte[which]->pte_lo &= 1375 ~(LPTE_WIMG | LPTE_RPGN); 1376 moea64_scratchpage_pte[which]->pte_lo |= 1377 moea64_calc_wimg(pa, VM_MEMATTR_DEFAULT) | (uint64_t)pa; 1378 EIEIO(); 1379 1380 moea64_scratchpage_pte[which]->pte_hi |= LPTE_VALID; 1381 PTESYNC(); isync(); 1382 } 1383 1384 void 1385 moea64_copy_page(mmu_t mmu, vm_page_t msrc, vm_page_t mdst) 1386 { 1387 vm_offset_t dst; 1388 vm_offset_t src; 1389 1390 dst = VM_PAGE_TO_PHYS(mdst); 1391 src = VM_PAGE_TO_PHYS(msrc); 1392 1393 if (hw_direct_map) { 1394 kcopy((void *)src, (void *)dst, PAGE_SIZE); 1395 } else { 1396 mtx_lock(&moea64_scratchpage_mtx); 1397 1398 moea64_set_scratchpage_pa(0,src); 1399 moea64_set_scratchpage_pa(1,dst); 1400 1401 kcopy((void *)moea64_scratchpage_va[0], 1402 (void *)moea64_scratchpage_va[1], PAGE_SIZE); 1403 1404 mtx_unlock(&moea64_scratchpage_mtx); 1405 } 1406 } 1407 1408 void 1409 moea64_zero_page_area(mmu_t mmu, vm_page_t m, int off, int size) 1410 { 1411 vm_offset_t pa = VM_PAGE_TO_PHYS(m); 1412 1413 if (!moea64_initialized) 1414 panic("moea64_zero_page: can't zero pa %#" PRIxPTR, pa); 1415 if (size + off > PAGE_SIZE) 1416 panic("moea64_zero_page: size + off > PAGE_SIZE"); 1417 1418 if (hw_direct_map) { 1419 bzero((caddr_t)pa + off, size); 1420 } else { 1421 mtx_lock(&moea64_scratchpage_mtx); 1422 moea64_set_scratchpage_pa(0,pa); 1423 bzero((caddr_t)moea64_scratchpage_va[0] + off, size); 1424 mtx_unlock(&moea64_scratchpage_mtx); 1425 } 1426 } 1427 1428 /* 1429 * Zero a page of physical memory by temporarily mapping it 1430 */ 1431 void 1432 moea64_zero_page(mmu_t mmu, vm_page_t m) 1433 { 1434 vm_offset_t pa = VM_PAGE_TO_PHYS(m); 1435 vm_offset_t va, off; 1436 1437 if (!moea64_initialized) 1438 panic("moea64_zero_page: can't zero pa %#zx", pa); 1439 1440 if (!hw_direct_map) { 1441 mtx_lock(&moea64_scratchpage_mtx); 1442 1443 moea64_set_scratchpage_pa(0,pa); 1444 va = moea64_scratchpage_va[0]; 1445 } else { 1446 va = pa; 1447 } 1448 1449 for (off = 0; off < PAGE_SIZE; off += cacheline_size) 1450 __asm __volatile("dcbz 0,%0" :: "r"(va + off)); 1451 1452 if (!hw_direct_map) 1453 mtx_unlock(&moea64_scratchpage_mtx); 1454 } 1455 1456 void 1457 moea64_zero_page_idle(mmu_t mmu, vm_page_t m) 1458 { 1459 1460 moea64_zero_page(mmu, m); 1461 } 1462 1463 /* 1464 * Map the given physical page at the specified virtual address in the 1465 * target pmap with the protection requested. If specified the page 1466 * will be wired down. 1467 */ 1468 void 1469 moea64_enter(mmu_t mmu, pmap_t pmap, vm_offset_t va, vm_page_t m, 1470 vm_prot_t prot, boolean_t wired) 1471 { 1472 1473 vm_page_lock_queues(); 1474 PMAP_LOCK(pmap); 1475 moea64_enter_locked(pmap, va, m, prot, wired); 1476 vm_page_unlock_queues(); 1477 PMAP_UNLOCK(pmap); 1478 } 1479 1480 /* 1481 * Map the given physical page at the specified virtual address in the 1482 * target pmap with the protection requested. If specified the page 1483 * will be wired down. 1484 * 1485 * The page queues and pmap must be locked. 1486 */ 1487 1488 static void 1489 moea64_enter_locked(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot, 1490 boolean_t wired) 1491 { 1492 struct pvo_head *pvo_head; 1493 uma_zone_t zone; 1494 vm_page_t pg; 1495 uint64_t pte_lo; 1496 u_int pvo_flags; 1497 int error; 1498 1499 if (!moea64_initialized) { 1500 pvo_head = &moea64_pvo_kunmanaged; 1501 pg = NULL; 1502 zone = moea64_upvo_zone; 1503 pvo_flags = 0; 1504 } else { 1505 pvo_head = vm_page_to_pvoh(m); 1506 pg = m; 1507 zone = moea64_mpvo_zone; 1508 pvo_flags = PVO_MANAGED; 1509 } 1510 1511 if (pmap_bootstrapped) 1512 mtx_assert(&vm_page_queue_mtx, MA_OWNED); 1513 PMAP_LOCK_ASSERT(pmap, MA_OWNED); 1514 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) != 0 || 1515 (m->oflags & VPO_BUSY) != 0 || VM_OBJECT_LOCKED(m->object), 1516 ("moea64_enter_locked: page %p is not busy", m)); 1517 1518 /* XXX change the pvo head for fake pages */ 1519 if ((m->flags & PG_FICTITIOUS) == PG_FICTITIOUS) { 1520 pvo_flags &= ~PVO_MANAGED; 1521 pvo_head = &moea64_pvo_kunmanaged; 1522 zone = moea64_upvo_zone; 1523 } 1524 1525 pte_lo = moea64_calc_wimg(VM_PAGE_TO_PHYS(m), pmap_page_get_memattr(m)); 1526 1527 if (prot & VM_PROT_WRITE) { 1528 pte_lo |= LPTE_BW; 1529 if (pmap_bootstrapped && 1530 (m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0) 1531 vm_page_flag_set(m, PG_WRITEABLE); 1532 } else 1533 pte_lo |= LPTE_BR; 1534 1535 if (prot & VM_PROT_EXECUTE) 1536 pvo_flags |= VM_PROT_EXECUTE; 1537 1538 if (wired) 1539 pvo_flags |= PVO_WIRED; 1540 1541 if ((m->flags & PG_FICTITIOUS) != 0) 1542 pvo_flags |= PVO_FAKE; 1543 1544 error = moea64_pvo_enter(pmap, zone, pvo_head, va, VM_PAGE_TO_PHYS(m), 1545 pte_lo, pvo_flags); 1546 1547 /* 1548 * Flush the page from the instruction cache if this page is 1549 * mapped executable and cacheable. 1550 */ 1551 if ((pte_lo & (LPTE_I | LPTE_G | LPTE_NOEXEC)) == 0) { 1552 moea64_syncicache(pmap, va, VM_PAGE_TO_PHYS(m), PAGE_SIZE); 1553 } 1554 } 1555 1556 static void 1557 moea64_syncicache(pmap_t pmap, vm_offset_t va, vm_offset_t pa, vm_size_t sz) 1558 { 1559 1560 /* 1561 * This is much trickier than on older systems because 1562 * we can't sync the icache on physical addresses directly 1563 * without a direct map. Instead we check a couple of cases 1564 * where the memory is already mapped in and, failing that, 1565 * use the same trick we use for page zeroing to create 1566 * a temporary mapping for this physical address. 1567 */ 1568 1569 if (!pmap_bootstrapped) { 1570 /* 1571 * If PMAP is not bootstrapped, we are likely to be 1572 * in real mode. 1573 */ 1574 __syncicache((void *)pa, sz); 1575 } else if (pmap == kernel_pmap) { 1576 __syncicache((void *)va, sz); 1577 } else if (hw_direct_map) { 1578 __syncicache((void *)pa, sz); 1579 } else { 1580 /* Use the scratch page to set up a temp mapping */ 1581 1582 mtx_lock(&moea64_scratchpage_mtx); 1583 1584 moea64_set_scratchpage_pa(1,pa & ~ADDR_POFF); 1585 __syncicache((void *)(moea64_scratchpage_va[1] + 1586 (va & ADDR_POFF)), sz); 1587 1588 mtx_unlock(&moea64_scratchpage_mtx); 1589 } 1590 } 1591 1592 /* 1593 * Maps a sequence of resident pages belonging to the same object. 1594 * The sequence begins with the given page m_start. This page is 1595 * mapped at the given virtual address start. Each subsequent page is 1596 * mapped at a virtual address that is offset from start by the same 1597 * amount as the page is offset from m_start within the object. The 1598 * last page in the sequence is the page with the largest offset from 1599 * m_start that can be mapped at a virtual address less than the given 1600 * virtual address end. Not every virtual page between start and end 1601 * is mapped; only those for which a resident page exists with the 1602 * corresponding offset from m_start are mapped. 1603 */ 1604 void 1605 moea64_enter_object(mmu_t mmu, pmap_t pm, vm_offset_t start, vm_offset_t end, 1606 vm_page_t m_start, vm_prot_t prot) 1607 { 1608 vm_page_t m; 1609 vm_pindex_t diff, psize; 1610 1611 psize = atop(end - start); 1612 m = m_start; 1613 vm_page_lock_queues(); 1614 PMAP_LOCK(pm); 1615 while (m != NULL && (diff = m->pindex - m_start->pindex) < psize) { 1616 moea64_enter_locked(pm, start + ptoa(diff), m, prot & 1617 (VM_PROT_READ | VM_PROT_EXECUTE), FALSE); 1618 m = TAILQ_NEXT(m, listq); 1619 } 1620 vm_page_unlock_queues(); 1621 PMAP_UNLOCK(pm); 1622 } 1623 1624 void 1625 moea64_enter_quick(mmu_t mmu, pmap_t pm, vm_offset_t va, vm_page_t m, 1626 vm_prot_t prot) 1627 { 1628 1629 vm_page_lock_queues(); 1630 PMAP_LOCK(pm); 1631 moea64_enter_locked(pm, va, m, prot & (VM_PROT_READ | VM_PROT_EXECUTE), 1632 FALSE); 1633 vm_page_unlock_queues(); 1634 PMAP_UNLOCK(pm); 1635 } 1636 1637 vm_paddr_t 1638 moea64_extract(mmu_t mmu, pmap_t pm, vm_offset_t va) 1639 { 1640 struct pvo_entry *pvo; 1641 vm_paddr_t pa; 1642 1643 PMAP_LOCK(pm); 1644 pvo = moea64_pvo_find_va(pm, va); 1645 if (pvo == NULL) 1646 pa = 0; 1647 else 1648 pa = (pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN) | 1649 (va - PVO_VADDR(pvo)); 1650 PMAP_UNLOCK(pm); 1651 return (pa); 1652 } 1653 1654 /* 1655 * Atomically extract and hold the physical page with the given 1656 * pmap and virtual address pair if that mapping permits the given 1657 * protection. 1658 */ 1659 vm_page_t 1660 moea64_extract_and_hold(mmu_t mmu, pmap_t pmap, vm_offset_t va, vm_prot_t prot) 1661 { 1662 struct pvo_entry *pvo; 1663 vm_page_t m; 1664 vm_paddr_t pa; 1665 1666 m = NULL; 1667 pa = 0; 1668 PMAP_LOCK(pmap); 1669 retry: 1670 pvo = moea64_pvo_find_va(pmap, va & ~ADDR_POFF); 1671 if (pvo != NULL && (pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) && 1672 ((pvo->pvo_pte.lpte.pte_lo & LPTE_PP) == LPTE_RW || 1673 (prot & VM_PROT_WRITE) == 0)) { 1674 if (vm_page_pa_tryrelock(pmap, 1675 pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN, &pa)) 1676 goto retry; 1677 m = PHYS_TO_VM_PAGE(pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN); 1678 vm_page_hold(m); 1679 } 1680 PA_UNLOCK_COND(pa); 1681 PMAP_UNLOCK(pmap); 1682 return (m); 1683 } 1684 1685 static void * 1686 moea64_uma_page_alloc(uma_zone_t zone, int bytes, u_int8_t *flags, int wait) 1687 { 1688 /* 1689 * This entire routine is a horrible hack to avoid bothering kmem 1690 * for new KVA addresses. Because this can get called from inside 1691 * kmem allocation routines, calling kmem for a new address here 1692 * can lead to multiply locking non-recursive mutexes. 1693 */ 1694 static vm_pindex_t color; 1695 vm_offset_t va; 1696 1697 vm_page_t m; 1698 int pflags, needed_lock; 1699 1700 *flags = UMA_SLAB_PRIV; 1701 needed_lock = !PMAP_LOCKED(kernel_pmap); 1702 1703 if (needed_lock) 1704 PMAP_LOCK(kernel_pmap); 1705 1706 if ((wait & (M_NOWAIT|M_USE_RESERVE)) == M_NOWAIT) 1707 pflags = VM_ALLOC_INTERRUPT | VM_ALLOC_WIRED; 1708 else 1709 pflags = VM_ALLOC_SYSTEM | VM_ALLOC_WIRED; 1710 if (wait & M_ZERO) 1711 pflags |= VM_ALLOC_ZERO; 1712 1713 for (;;) { 1714 m = vm_page_alloc(NULL, color++, pflags | VM_ALLOC_NOOBJ); 1715 if (m == NULL) { 1716 if (wait & M_NOWAIT) 1717 return (NULL); 1718 VM_WAIT; 1719 } else 1720 break; 1721 } 1722 1723 va = VM_PAGE_TO_PHYS(m); 1724 1725 moea64_pvo_enter(kernel_pmap, moea64_upvo_zone, 1726 &moea64_pvo_kunmanaged, va, VM_PAGE_TO_PHYS(m), LPTE_M, 1727 PVO_WIRED | PVO_BOOTSTRAP); 1728 1729 if (needed_lock) 1730 PMAP_UNLOCK(kernel_pmap); 1731 1732 if ((wait & M_ZERO) && (m->flags & PG_ZERO) == 0) 1733 bzero((void *)va, PAGE_SIZE); 1734 1735 return (void *)va; 1736 } 1737 1738 void 1739 moea64_init(mmu_t mmu) 1740 { 1741 1742 CTR0(KTR_PMAP, "moea64_init"); 1743 1744 moea64_upvo_zone = uma_zcreate("UPVO entry", sizeof (struct pvo_entry), 1745 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 1746 UMA_ZONE_VM | UMA_ZONE_NOFREE); 1747 moea64_mpvo_zone = uma_zcreate("MPVO entry", sizeof(struct pvo_entry), 1748 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 1749 UMA_ZONE_VM | UMA_ZONE_NOFREE); 1750 1751 if (!hw_direct_map) { 1752 uma_zone_set_allocf(moea64_upvo_zone,moea64_uma_page_alloc); 1753 uma_zone_set_allocf(moea64_mpvo_zone,moea64_uma_page_alloc); 1754 } 1755 1756 moea64_initialized = TRUE; 1757 } 1758 1759 boolean_t 1760 moea64_is_referenced(mmu_t mmu, vm_page_t m) 1761 { 1762 1763 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1764 ("moea64_is_referenced: page %p is not managed", m)); 1765 return (moea64_query_bit(m, PTE_REF)); 1766 } 1767 1768 boolean_t 1769 moea64_is_modified(mmu_t mmu, vm_page_t m) 1770 { 1771 1772 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1773 ("moea64_is_modified: page %p is not managed", m)); 1774 1775 /* 1776 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be 1777 * concurrently set while the object is locked. Thus, if PG_WRITEABLE 1778 * is clear, no PTEs can have LPTE_CHG set. 1779 */ 1780 VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED); 1781 if ((m->oflags & VPO_BUSY) == 0 && 1782 (m->flags & PG_WRITEABLE) == 0) 1783 return (FALSE); 1784 return (moea64_query_bit(m, LPTE_CHG)); 1785 } 1786 1787 boolean_t 1788 moea64_is_prefaultable(mmu_t mmu, pmap_t pmap, vm_offset_t va) 1789 { 1790 struct pvo_entry *pvo; 1791 boolean_t rv; 1792 1793 PMAP_LOCK(pmap); 1794 pvo = moea64_pvo_find_va(pmap, va & ~ADDR_POFF); 1795 rv = pvo == NULL || (pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) == 0; 1796 PMAP_UNLOCK(pmap); 1797 return (rv); 1798 } 1799 1800 void 1801 moea64_clear_reference(mmu_t mmu, vm_page_t m) 1802 { 1803 1804 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1805 ("moea64_clear_reference: page %p is not managed", m)); 1806 moea64_clear_bit(m, LPTE_REF); 1807 } 1808 1809 void 1810 moea64_clear_modify(mmu_t mmu, vm_page_t m) 1811 { 1812 1813 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1814 ("moea64_clear_modify: page %p is not managed", m)); 1815 VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED); 1816 KASSERT((m->oflags & VPO_BUSY) == 0, 1817 ("moea64_clear_modify: page %p is busy", m)); 1818 1819 /* 1820 * If the page is not PG_WRITEABLE, then no PTEs can have LPTE_CHG 1821 * set. If the object containing the page is locked and the page is 1822 * not VPO_BUSY, then PG_WRITEABLE cannot be concurrently set. 1823 */ 1824 if ((m->flags & PG_WRITEABLE) == 0) 1825 return; 1826 moea64_clear_bit(m, LPTE_CHG); 1827 } 1828 1829 /* 1830 * Clear the write and modified bits in each of the given page's mappings. 1831 */ 1832 void 1833 moea64_remove_write(mmu_t mmu, vm_page_t m) 1834 { 1835 struct pvo_entry *pvo; 1836 struct lpte *pt; 1837 pmap_t pmap; 1838 uint64_t lo; 1839 1840 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1841 ("moea64_remove_write: page %p is not managed", m)); 1842 1843 /* 1844 * If the page is not VPO_BUSY, then PG_WRITEABLE cannot be set by 1845 * another thread while the object is locked. Thus, if PG_WRITEABLE 1846 * is clear, no page table entries need updating. 1847 */ 1848 VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED); 1849 if ((m->oflags & VPO_BUSY) == 0 && 1850 (m->flags & PG_WRITEABLE) == 0) 1851 return; 1852 vm_page_lock_queues(); 1853 lo = moea64_attr_fetch(m); 1854 SYNC(); 1855 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) { 1856 pmap = pvo->pvo_pmap; 1857 PMAP_LOCK(pmap); 1858 LOCK_TABLE(); 1859 if ((pvo->pvo_pte.lpte.pte_lo & LPTE_PP) != LPTE_BR) { 1860 pt = moea64_pvo_to_pte(pvo); 1861 pvo->pvo_pte.lpte.pte_lo &= ~LPTE_PP; 1862 pvo->pvo_pte.lpte.pte_lo |= LPTE_BR; 1863 if (pt != NULL) { 1864 moea64_pte_synch(pt, &pvo->pvo_pte.lpte); 1865 lo |= pvo->pvo_pte.lpte.pte_lo; 1866 pvo->pvo_pte.lpte.pte_lo &= ~LPTE_CHG; 1867 moea64_pte_change(pt, &pvo->pvo_pte.lpte, 1868 pvo->pvo_vpn); 1869 if (pvo->pvo_pmap == kernel_pmap) 1870 isync(); 1871 } 1872 } 1873 UNLOCK_TABLE(); 1874 PMAP_UNLOCK(pmap); 1875 } 1876 if ((lo & LPTE_CHG) != 0) { 1877 moea64_attr_clear(m, LPTE_CHG); 1878 vm_page_dirty(m); 1879 } 1880 vm_page_flag_clear(m, PG_WRITEABLE); 1881 vm_page_unlock_queues(); 1882 } 1883 1884 /* 1885 * moea64_ts_referenced: 1886 * 1887 * Return a count of reference bits for a page, clearing those bits. 1888 * It is not necessary for every reference bit to be cleared, but it 1889 * is necessary that 0 only be returned when there are truly no 1890 * reference bits set. 1891 * 1892 * XXX: The exact number of bits to check and clear is a matter that 1893 * should be tested and standardized at some point in the future for 1894 * optimal aging of shared pages. 1895 */ 1896 boolean_t 1897 moea64_ts_referenced(mmu_t mmu, vm_page_t m) 1898 { 1899 1900 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 1901 ("moea64_ts_referenced: page %p is not managed", m)); 1902 return (moea64_clear_bit(m, LPTE_REF)); 1903 } 1904 1905 /* 1906 * Modify the WIMG settings of all mappings for a page. 1907 */ 1908 void 1909 moea64_page_set_memattr(mmu_t mmu, vm_page_t m, vm_memattr_t ma) 1910 { 1911 struct pvo_entry *pvo; 1912 struct pvo_head *pvo_head; 1913 struct lpte *pt; 1914 pmap_t pmap; 1915 uint64_t lo; 1916 1917 if (m->flags & PG_FICTITIOUS) { 1918 m->md.mdpg_cache_attrs = ma; 1919 return; 1920 } 1921 1922 vm_page_lock_queues(); 1923 pvo_head = vm_page_to_pvoh(m); 1924 lo = moea64_calc_wimg(VM_PAGE_TO_PHYS(m), ma); 1925 LIST_FOREACH(pvo, pvo_head, pvo_vlink) { 1926 pmap = pvo->pvo_pmap; 1927 PMAP_LOCK(pmap); 1928 LOCK_TABLE(); 1929 pt = moea64_pvo_to_pte(pvo); 1930 pvo->pvo_pte.lpte.pte_lo &= ~LPTE_WIMG; 1931 pvo->pvo_pte.lpte.pte_lo |= lo; 1932 if (pt != NULL) { 1933 moea64_pte_change(pt, &pvo->pvo_pte.lpte, 1934 pvo->pvo_vpn); 1935 if (pvo->pvo_pmap == kernel_pmap) 1936 isync(); 1937 } 1938 UNLOCK_TABLE(); 1939 PMAP_UNLOCK(pmap); 1940 } 1941 m->md.mdpg_cache_attrs = ma; 1942 vm_page_unlock_queues(); 1943 } 1944 1945 /* 1946 * Map a wired page into kernel virtual address space. 1947 */ 1948 void 1949 moea64_kenter_attr(mmu_t mmu, vm_offset_t va, vm_offset_t pa, vm_memattr_t ma) 1950 { 1951 uint64_t pte_lo; 1952 int error; 1953 1954 pte_lo = moea64_calc_wimg(pa, ma); 1955 1956 PMAP_LOCK(kernel_pmap); 1957 error = moea64_pvo_enter(kernel_pmap, moea64_upvo_zone, 1958 &moea64_pvo_kunmanaged, va, pa, pte_lo, 1959 PVO_WIRED | VM_PROT_EXECUTE); 1960 1961 if (error != 0 && error != ENOENT) 1962 panic("moea64_kenter: failed to enter va %#zx pa %#zx: %d", va, 1963 pa, error); 1964 1965 /* 1966 * Flush the memory from the instruction cache. 1967 */ 1968 if ((pte_lo & (LPTE_I | LPTE_G)) == 0) { 1969 __syncicache((void *)va, PAGE_SIZE); 1970 } 1971 PMAP_UNLOCK(kernel_pmap); 1972 } 1973 1974 void 1975 moea64_kenter(mmu_t mmu, vm_offset_t va, vm_offset_t pa) 1976 { 1977 1978 moea64_kenter_attr(mmu, va, pa, VM_MEMATTR_DEFAULT); 1979 } 1980 1981 /* 1982 * Extract the physical page address associated with the given kernel virtual 1983 * address. 1984 */ 1985 vm_offset_t 1986 moea64_kextract(mmu_t mmu, vm_offset_t va) 1987 { 1988 struct pvo_entry *pvo; 1989 vm_paddr_t pa; 1990 1991 /* 1992 * Shortcut the direct-mapped case when applicable. We never put 1993 * anything but 1:1 mappings below VM_MIN_KERNEL_ADDRESS. 1994 */ 1995 if (va < VM_MIN_KERNEL_ADDRESS) 1996 return (va); 1997 1998 PMAP_LOCK(kernel_pmap); 1999 pvo = moea64_pvo_find_va(kernel_pmap, va); 2000 KASSERT(pvo != NULL, ("moea64_kextract: no addr found for %#" PRIxPTR, 2001 va)); 2002 pa = (pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN) + (va - PVO_VADDR(pvo)); 2003 PMAP_UNLOCK(kernel_pmap); 2004 return (pa); 2005 } 2006 2007 /* 2008 * Remove a wired page from kernel virtual address space. 2009 */ 2010 void 2011 moea64_kremove(mmu_t mmu, vm_offset_t va) 2012 { 2013 moea64_remove(mmu, kernel_pmap, va, va + PAGE_SIZE); 2014 } 2015 2016 /* 2017 * Map a range of physical addresses into kernel virtual address space. 2018 * 2019 * The value passed in *virt is a suggested virtual address for the mapping. 2020 * Architectures which can support a direct-mapped physical to virtual region 2021 * can return the appropriate address within that region, leaving '*virt' 2022 * unchanged. We cannot and therefore do not; *virt is updated with the 2023 * first usable address after the mapped region. 2024 */ 2025 vm_offset_t 2026 moea64_map(mmu_t mmu, vm_offset_t *virt, vm_offset_t pa_start, 2027 vm_offset_t pa_end, int prot) 2028 { 2029 vm_offset_t sva, va; 2030 2031 sva = *virt; 2032 va = sva; 2033 for (; pa_start < pa_end; pa_start += PAGE_SIZE, va += PAGE_SIZE) 2034 moea64_kenter(mmu, va, pa_start); 2035 *virt = va; 2036 2037 return (sva); 2038 } 2039 2040 /* 2041 * Returns true if the pmap's pv is one of the first 2042 * 16 pvs linked to from this page. This count may 2043 * be changed upwards or downwards in the future; it 2044 * is only necessary that true be returned for a small 2045 * subset of pmaps for proper page aging. 2046 */ 2047 boolean_t 2048 moea64_page_exists_quick(mmu_t mmu, pmap_t pmap, vm_page_t m) 2049 { 2050 int loops; 2051 struct pvo_entry *pvo; 2052 boolean_t rv; 2053 2054 KASSERT((m->flags & (PG_FICTITIOUS | PG_UNMANAGED)) == 0, 2055 ("moea64_page_exists_quick: page %p is not managed", m)); 2056 loops = 0; 2057 rv = FALSE; 2058 vm_page_lock_queues(); 2059 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) { 2060 if (pvo->pvo_pmap == pmap) { 2061 rv = TRUE; 2062 break; 2063 } 2064 if (++loops >= 16) 2065 break; 2066 } 2067 vm_page_unlock_queues(); 2068 return (rv); 2069 } 2070 2071 /* 2072 * Return the number of managed mappings to the given physical page 2073 * that are wired. 2074 */ 2075 int 2076 moea64_page_wired_mappings(mmu_t mmu, vm_page_t m) 2077 { 2078 struct pvo_entry *pvo; 2079 int count; 2080 2081 count = 0; 2082 if ((m->flags & PG_FICTITIOUS) != 0) 2083 return (count); 2084 vm_page_lock_queues(); 2085 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) 2086 if ((pvo->pvo_vaddr & PVO_WIRED) != 0) 2087 count++; 2088 vm_page_unlock_queues(); 2089 return (count); 2090 } 2091 2092 static uintptr_t moea64_vsidcontext; 2093 2094 uintptr_t 2095 moea64_get_unique_vsid(void) { 2096 u_int entropy; 2097 register_t hash; 2098 uint32_t mask; 2099 int i; 2100 2101 entropy = 0; 2102 __asm __volatile("mftb %0" : "=r"(entropy)); 2103 2104 mtx_lock(&moea64_slb_mutex); 2105 for (i = 0; i < NVSIDS; i += VSID_NBPW) { 2106 u_int n; 2107 2108 /* 2109 * Create a new value by mutiplying by a prime and adding in 2110 * entropy from the timebase register. This is to make the 2111 * VSID more random so that the PT hash function collides 2112 * less often. (Note that the prime casues gcc to do shifts 2113 * instead of a multiply.) 2114 */ 2115 moea64_vsidcontext = (moea64_vsidcontext * 0x1105) + entropy; 2116 hash = moea64_vsidcontext & (NVSIDS - 1); 2117 if (hash == 0) /* 0 is special, avoid it */ 2118 continue; 2119 n = hash >> 5; 2120 mask = 1 << (hash & (VSID_NBPW - 1)); 2121 hash = (moea64_vsidcontext & VSID_HASHMASK); 2122 if (moea64_vsid_bitmap[n] & mask) { /* collision? */ 2123 /* anything free in this bucket? */ 2124 if (moea64_vsid_bitmap[n] == 0xffffffff) { 2125 entropy = (moea64_vsidcontext >> 20); 2126 continue; 2127 } 2128 i = ffs(~moea64_vsid_bitmap[n]) - 1; 2129 mask = 1 << i; 2130 hash &= VSID_HASHMASK & ~(VSID_NBPW - 1); 2131 hash |= i; 2132 } 2133 KASSERT(!(moea64_vsid_bitmap[n] & mask), 2134 ("Allocating in-use VSID %#zx\n", hash)); 2135 moea64_vsid_bitmap[n] |= mask; 2136 mtx_unlock(&moea64_slb_mutex); 2137 return (hash); 2138 } 2139 2140 mtx_unlock(&moea64_slb_mutex); 2141 panic("%s: out of segments",__func__); 2142 } 2143 2144 #ifdef __powerpc64__ 2145 void 2146 moea64_pinit(mmu_t mmu, pmap_t pmap) 2147 { 2148 PMAP_LOCK_INIT(pmap); 2149 2150 pmap->pm_slb_tree_root = slb_alloc_tree(); 2151 pmap->pm_slb = slb_alloc_user_cache(); 2152 pmap->pm_slb_len = 0; 2153 } 2154 #else 2155 void 2156 moea64_pinit(mmu_t mmu, pmap_t pmap) 2157 { 2158 int i; 2159 uint32_t hash; 2160 2161 PMAP_LOCK_INIT(pmap); 2162 2163 if (pmap_bootstrapped) 2164 pmap->pmap_phys = (pmap_t)moea64_kextract(mmu, 2165 (vm_offset_t)pmap); 2166 else 2167 pmap->pmap_phys = pmap; 2168 2169 /* 2170 * Allocate some segment registers for this pmap. 2171 */ 2172 hash = moea64_get_unique_vsid(); 2173 2174 for (i = 0; i < 16; i++) 2175 pmap->pm_sr[i] = VSID_MAKE(i, hash); 2176 2177 KASSERT(pmap->pm_sr[0] != 0, ("moea64_pinit: pm_sr[0] = 0")); 2178 } 2179 #endif 2180 2181 /* 2182 * Initialize the pmap associated with process 0. 2183 */ 2184 void 2185 moea64_pinit0(mmu_t mmu, pmap_t pm) 2186 { 2187 moea64_pinit(mmu, pm); 2188 bzero(&pm->pm_stats, sizeof(pm->pm_stats)); 2189 } 2190 2191 /* 2192 * Set the physical protection on the specified range of this map as requested. 2193 */ 2194 void 2195 moea64_protect(mmu_t mmu, pmap_t pm, vm_offset_t sva, vm_offset_t eva, 2196 vm_prot_t prot) 2197 { 2198 struct pvo_entry *pvo; 2199 struct lpte *pt; 2200 2201 CTR4(KTR_PMAP, "moea64_protect: pm=%p sva=%#x eva=%#x prot=%#x", pm, sva, 2202 eva, prot); 2203 2204 2205 KASSERT(pm == &curproc->p_vmspace->vm_pmap || pm == kernel_pmap, 2206 ("moea64_protect: non current pmap")); 2207 2208 if ((prot & VM_PROT_READ) == VM_PROT_NONE) { 2209 moea64_remove(mmu, pm, sva, eva); 2210 return; 2211 } 2212 2213 vm_page_lock_queues(); 2214 PMAP_LOCK(pm); 2215 for (; sva < eva; sva += PAGE_SIZE) { 2216 pvo = moea64_pvo_find_va(pm, sva); 2217 if (pvo == NULL) 2218 continue; 2219 2220 /* 2221 * Grab the PTE pointer before we diddle with the cached PTE 2222 * copy. 2223 */ 2224 LOCK_TABLE(); 2225 pt = moea64_pvo_to_pte(pvo); 2226 2227 /* 2228 * Change the protection of the page. 2229 */ 2230 pvo->pvo_pte.lpte.pte_lo &= ~LPTE_PP; 2231 pvo->pvo_pte.lpte.pte_lo |= LPTE_BR; 2232 pvo->pvo_pte.lpte.pte_lo &= ~LPTE_NOEXEC; 2233 if ((prot & VM_PROT_EXECUTE) == 0) 2234 pvo->pvo_pte.lpte.pte_lo |= LPTE_NOEXEC; 2235 2236 /* 2237 * If the PVO is in the page table, update that pte as well. 2238 */ 2239 if (pt != NULL) { 2240 moea64_pte_change(pt, &pvo->pvo_pte.lpte, pvo->pvo_vpn); 2241 if ((pvo->pvo_pte.lpte.pte_lo & 2242 (LPTE_I | LPTE_G | LPTE_NOEXEC)) == 0) { 2243 moea64_syncicache(pm, sva, 2244 pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN, 2245 PAGE_SIZE); 2246 } 2247 } 2248 UNLOCK_TABLE(); 2249 } 2250 vm_page_unlock_queues(); 2251 PMAP_UNLOCK(pm); 2252 } 2253 2254 /* 2255 * Map a list of wired pages into kernel virtual address space. This is 2256 * intended for temporary mappings which do not need page modification or 2257 * references recorded. Existing mappings in the region are overwritten. 2258 */ 2259 void 2260 moea64_qenter(mmu_t mmu, vm_offset_t va, vm_page_t *m, int count) 2261 { 2262 while (count-- > 0) { 2263 moea64_kenter(mmu, va, VM_PAGE_TO_PHYS(*m)); 2264 va += PAGE_SIZE; 2265 m++; 2266 } 2267 } 2268 2269 /* 2270 * Remove page mappings from kernel virtual address space. Intended for 2271 * temporary mappings entered by moea64_qenter. 2272 */ 2273 void 2274 moea64_qremove(mmu_t mmu, vm_offset_t va, int count) 2275 { 2276 while (count-- > 0) { 2277 moea64_kremove(mmu, va); 2278 va += PAGE_SIZE; 2279 } 2280 } 2281 2282 void 2283 moea64_release_vsid(uint64_t vsid) 2284 { 2285 int idx, mask; 2286 2287 mtx_lock(&moea64_slb_mutex); 2288 idx = vsid & (NVSIDS-1); 2289 mask = 1 << (idx % VSID_NBPW); 2290 idx /= VSID_NBPW; 2291 KASSERT(moea64_vsid_bitmap[idx] & mask, 2292 ("Freeing unallocated VSID %#jx", vsid)); 2293 moea64_vsid_bitmap[idx] &= ~mask; 2294 mtx_unlock(&moea64_slb_mutex); 2295 } 2296 2297 2298 void 2299 moea64_release(mmu_t mmu, pmap_t pmap) 2300 { 2301 2302 /* 2303 * Free segment registers' VSIDs 2304 */ 2305 #ifdef __powerpc64__ 2306 slb_free_tree(pmap); 2307 slb_free_user_cache(pmap->pm_slb); 2308 #else 2309 KASSERT(pmap->pm_sr[0] != 0, ("moea64_release: pm_sr[0] = 0")); 2310 2311 moea64_release_vsid(VSID_TO_HASH(pmap->pm_sr[0])); 2312 #endif 2313 2314 PMAP_LOCK_DESTROY(pmap); 2315 } 2316 2317 /* 2318 * Remove the given range of addresses from the specified map. 2319 */ 2320 void 2321 moea64_remove(mmu_t mmu, pmap_t pm, vm_offset_t sva, vm_offset_t eva) 2322 { 2323 struct pvo_entry *pvo; 2324 2325 vm_page_lock_queues(); 2326 PMAP_LOCK(pm); 2327 for (; sva < eva; sva += PAGE_SIZE) { 2328 pvo = moea64_pvo_find_va(pm, sva); 2329 if (pvo != NULL) 2330 moea64_pvo_remove(pvo); 2331 } 2332 vm_page_unlock_queues(); 2333 PMAP_UNLOCK(pm); 2334 } 2335 2336 /* 2337 * Remove physical page from all pmaps in which it resides. moea64_pvo_remove() 2338 * will reflect changes in pte's back to the vm_page. 2339 */ 2340 void 2341 moea64_remove_all(mmu_t mmu, vm_page_t m) 2342 { 2343 struct pvo_head *pvo_head; 2344 struct pvo_entry *pvo, *next_pvo; 2345 pmap_t pmap; 2346 2347 vm_page_lock_queues(); 2348 pvo_head = vm_page_to_pvoh(m); 2349 for (pvo = LIST_FIRST(pvo_head); pvo != NULL; pvo = next_pvo) { 2350 next_pvo = LIST_NEXT(pvo, pvo_vlink); 2351 2352 MOEA_PVO_CHECK(pvo); /* sanity check */ 2353 pmap = pvo->pvo_pmap; 2354 PMAP_LOCK(pmap); 2355 moea64_pvo_remove(pvo); 2356 PMAP_UNLOCK(pmap); 2357 } 2358 if ((m->flags & PG_WRITEABLE) && moea64_is_modified(mmu, m)) { 2359 moea64_attr_clear(m, LPTE_CHG); 2360 vm_page_dirty(m); 2361 } 2362 vm_page_flag_clear(m, PG_WRITEABLE); 2363 vm_page_unlock_queues(); 2364 } 2365 2366 /* 2367 * Allocate a physical page of memory directly from the phys_avail map. 2368 * Can only be called from moea64_bootstrap before avail start and end are 2369 * calculated. 2370 */ 2371 static vm_offset_t 2372 moea64_bootstrap_alloc(vm_size_t size, u_int align) 2373 { 2374 vm_offset_t s, e; 2375 int i, j; 2376 2377 size = round_page(size); 2378 for (i = 0; phys_avail[i + 1] != 0; i += 2) { 2379 if (align != 0) 2380 s = (phys_avail[i] + align - 1) & ~(align - 1); 2381 else 2382 s = phys_avail[i]; 2383 e = s + size; 2384 2385 if (s < phys_avail[i] || e > phys_avail[i + 1]) 2386 continue; 2387 2388 if (s == phys_avail[i]) { 2389 phys_avail[i] += size; 2390 } else if (e == phys_avail[i + 1]) { 2391 phys_avail[i + 1] -= size; 2392 } else { 2393 for (j = phys_avail_count * 2; j > i; j -= 2) { 2394 phys_avail[j] = phys_avail[j - 2]; 2395 phys_avail[j + 1] = phys_avail[j - 1]; 2396 } 2397 2398 phys_avail[i + 3] = phys_avail[i + 1]; 2399 phys_avail[i + 1] = s; 2400 phys_avail[i + 2] = e; 2401 phys_avail_count++; 2402 } 2403 2404 return (s); 2405 } 2406 panic("moea64_bootstrap_alloc: could not allocate memory"); 2407 } 2408 2409 static void 2410 tlbia(void) 2411 { 2412 vm_offset_t i; 2413 #ifndef __powerpc64__ 2414 register_t msr, scratch; 2415 #endif 2416 2417 TLBSYNC(); 2418 2419 for (i = 0; i < 0xFF000; i += 0x00001000) { 2420 #ifdef __powerpc64__ 2421 __asm __volatile("tlbiel %0" :: "r"(i)); 2422 #else 2423 __asm __volatile("\ 2424 mfmsr %0; \ 2425 mr %1, %0; \ 2426 insrdi %1,%3,1,0; \ 2427 mtmsrd %1; \ 2428 isync; \ 2429 \ 2430 tlbiel %2; \ 2431 \ 2432 mtmsrd %0; \ 2433 isync;" 2434 : "=r"(msr), "=r"(scratch) : "r"(i), "r"(1)); 2435 #endif 2436 } 2437 2438 EIEIO(); 2439 TLBSYNC(); 2440 } 2441 2442 #ifdef __powerpc64__ 2443 static void 2444 slbia(void) 2445 { 2446 register_t seg0; 2447 2448 __asm __volatile ("slbia"); 2449 __asm __volatile ("slbmfee %0,%1; slbie %0;" : "=r"(seg0) : "r"(0)); 2450 } 2451 #endif 2452 2453 static int 2454 moea64_pvo_enter(pmap_t pm, uma_zone_t zone, struct pvo_head *pvo_head, 2455 vm_offset_t va, vm_offset_t pa, uint64_t pte_lo, int flags) 2456 { 2457 struct pvo_entry *pvo; 2458 uint64_t vsid; 2459 int first; 2460 u_int ptegidx; 2461 int i; 2462 int bootstrap; 2463 2464 /* 2465 * One nasty thing that can happen here is that the UMA calls to 2466 * allocate new PVOs need to map more memory, which calls pvo_enter(), 2467 * which calls UMA... 2468 * 2469 * We break the loop by detecting recursion and allocating out of 2470 * the bootstrap pool. 2471 */ 2472 2473 first = 0; 2474 bootstrap = (flags & PVO_BOOTSTRAP); 2475 2476 if (!moea64_initialized) 2477 bootstrap = 1; 2478 2479 /* 2480 * Compute the PTE Group index. 2481 */ 2482 va &= ~ADDR_POFF; 2483 vsid = va_to_vsid(pm, va); 2484 ptegidx = va_to_pteg(vsid, va, flags & PVO_LARGE); 2485 2486 /* 2487 * Remove any existing mapping for this page. Reuse the pvo entry if 2488 * there is a mapping. 2489 */ 2490 LOCK_TABLE(); 2491 2492 moea64_pvo_enter_calls++; 2493 2494 LIST_FOREACH(pvo, &moea64_pvo_table[ptegidx], pvo_olink) { 2495 if (pvo->pvo_pmap == pm && PVO_VADDR(pvo) == va) { 2496 if ((pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN) == pa && 2497 (pvo->pvo_pte.lpte.pte_lo & LPTE_PP) == 2498 (pte_lo & LPTE_PP)) { 2499 if (!(pvo->pvo_pte.lpte.pte_hi & LPTE_VALID)) { 2500 /* Re-insert if spilled */ 2501 i = moea64_pte_insert(ptegidx, 2502 &pvo->pvo_pte.lpte); 2503 if (i >= 0) 2504 PVO_PTEGIDX_SET(pvo, i); 2505 moea64_pte_overflow--; 2506 } 2507 UNLOCK_TABLE(); 2508 return (0); 2509 } 2510 moea64_pvo_remove(pvo); 2511 break; 2512 } 2513 } 2514 2515 /* 2516 * If we aren't overwriting a mapping, try to allocate. 2517 */ 2518 if (bootstrap) { 2519 if (moea64_bpvo_pool_index >= BPVO_POOL_SIZE) { 2520 panic("moea64_enter: bpvo pool exhausted, %d, %d, %zd", 2521 moea64_bpvo_pool_index, BPVO_POOL_SIZE, 2522 BPVO_POOL_SIZE * sizeof(struct pvo_entry)); 2523 } 2524 pvo = &moea64_bpvo_pool[moea64_bpvo_pool_index]; 2525 moea64_bpvo_pool_index++; 2526 bootstrap = 1; 2527 } else { 2528 /* 2529 * Note: drop the table lock around the UMA allocation in 2530 * case the UMA allocator needs to manipulate the page 2531 * table. The mapping we are working with is already 2532 * protected by the PMAP lock. 2533 */ 2534 UNLOCK_TABLE(); 2535 pvo = uma_zalloc(zone, M_NOWAIT); 2536 LOCK_TABLE(); 2537 } 2538 2539 if (pvo == NULL) { 2540 UNLOCK_TABLE(); 2541 return (ENOMEM); 2542 } 2543 2544 moea64_pvo_entries++; 2545 pvo->pvo_vaddr = va; 2546 pvo->pvo_vpn = (uint64_t)((va & ADDR_PIDX) >> ADDR_PIDX_SHFT) 2547 | (vsid << 16); 2548 pvo->pvo_pmap = pm; 2549 LIST_INSERT_HEAD(&moea64_pvo_table[ptegidx], pvo, pvo_olink); 2550 pvo->pvo_vaddr &= ~ADDR_POFF; 2551 2552 if (!(flags & VM_PROT_EXECUTE)) 2553 pte_lo |= LPTE_NOEXEC; 2554 if (flags & PVO_WIRED) 2555 pvo->pvo_vaddr |= PVO_WIRED; 2556 if (pvo_head != &moea64_pvo_kunmanaged) 2557 pvo->pvo_vaddr |= PVO_MANAGED; 2558 if (bootstrap) 2559 pvo->pvo_vaddr |= PVO_BOOTSTRAP; 2560 if (flags & PVO_FAKE) 2561 pvo->pvo_vaddr |= PVO_FAKE; 2562 if (flags & PVO_LARGE) 2563 pvo->pvo_vaddr |= PVO_LARGE; 2564 2565 moea64_pte_create(&pvo->pvo_pte.lpte, vsid, va, 2566 (uint64_t)(pa) | pte_lo, flags); 2567 2568 /* 2569 * Remember if the list was empty and therefore will be the first 2570 * item. 2571 */ 2572 if (LIST_FIRST(pvo_head) == NULL) 2573 first = 1; 2574 LIST_INSERT_HEAD(pvo_head, pvo, pvo_vlink); 2575 2576 if (pvo->pvo_vaddr & PVO_WIRED) { 2577 pvo->pvo_pte.lpte.pte_hi |= LPTE_WIRED; 2578 pm->pm_stats.wired_count++; 2579 } 2580 pm->pm_stats.resident_count++; 2581 2582 /* 2583 * We hope this succeeds but it isn't required. 2584 */ 2585 i = moea64_pte_insert(ptegidx, &pvo->pvo_pte.lpte); 2586 if (i >= 0) { 2587 PVO_PTEGIDX_SET(pvo, i); 2588 } else { 2589 panic("moea64_pvo_enter: overflow"); 2590 moea64_pte_overflow++; 2591 } 2592 2593 if (pm == kernel_pmap) 2594 isync(); 2595 2596 UNLOCK_TABLE(); 2597 2598 #ifdef __powerpc64__ 2599 /* 2600 * Make sure all our bootstrap mappings are in the SLB as soon 2601 * as virtual memory is switched on. 2602 */ 2603 if (!pmap_bootstrapped) 2604 moea64_bootstrap_slb_prefault(va, flags & PVO_LARGE); 2605 #endif 2606 2607 return (first ? ENOENT : 0); 2608 } 2609 2610 static void 2611 moea64_pvo_remove(struct pvo_entry *pvo) 2612 { 2613 struct lpte *pt; 2614 2615 /* 2616 * If there is an active pte entry, we need to deactivate it (and 2617 * save the ref & cfg bits). 2618 */ 2619 LOCK_TABLE(); 2620 pt = moea64_pvo_to_pte(pvo); 2621 if (pt != NULL) { 2622 moea64_pte_unset(pt, &pvo->pvo_pte.lpte, pvo->pvo_vpn); 2623 PVO_PTEGIDX_CLR(pvo); 2624 } else { 2625 moea64_pte_overflow--; 2626 } 2627 2628 /* 2629 * Update our statistics. 2630 */ 2631 pvo->pvo_pmap->pm_stats.resident_count--; 2632 if (pvo->pvo_vaddr & PVO_WIRED) 2633 pvo->pvo_pmap->pm_stats.wired_count--; 2634 2635 /* 2636 * Save the REF/CHG bits into their cache if the page is managed. 2637 */ 2638 if ((pvo->pvo_vaddr & (PVO_MANAGED|PVO_FAKE)) == PVO_MANAGED) { 2639 struct vm_page *pg; 2640 2641 pg = PHYS_TO_VM_PAGE(pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN); 2642 if (pg != NULL) { 2643 moea64_attr_save(pg, pvo->pvo_pte.lpte.pte_lo & 2644 (LPTE_REF | LPTE_CHG)); 2645 } 2646 } 2647 2648 /* 2649 * Remove this PVO from the PV list. 2650 */ 2651 LIST_REMOVE(pvo, pvo_vlink); 2652 2653 /* 2654 * Remove this from the overflow list and return it to the pool 2655 * if we aren't going to reuse it. 2656 */ 2657 LIST_REMOVE(pvo, pvo_olink); 2658 2659 moea64_pvo_entries--; 2660 moea64_pvo_remove_calls++; 2661 2662 UNLOCK_TABLE(); 2663 2664 if (!(pvo->pvo_vaddr & PVO_BOOTSTRAP)) 2665 uma_zfree((pvo->pvo_vaddr & PVO_MANAGED) ? moea64_mpvo_zone : 2666 moea64_upvo_zone, pvo); 2667 } 2668 2669 static struct pvo_entry * 2670 moea64_pvo_find_va(pmap_t pm, vm_offset_t va) 2671 { 2672 struct pvo_entry *pvo; 2673 int ptegidx; 2674 uint64_t vsid; 2675 #ifdef __powerpc64__ 2676 uint64_t slbv; 2677 2678 if (pm == kernel_pmap) { 2679 slbv = kernel_va_to_slbv(va); 2680 } else { 2681 struct slb *slb; 2682 slb = user_va_to_slb_entry(pm, va); 2683 /* The page is not mapped if the segment isn't */ 2684 if (slb == NULL) 2685 return NULL; 2686 slbv = slb->slbv; 2687 } 2688 2689 vsid = (slbv & SLBV_VSID_MASK) >> SLBV_VSID_SHIFT; 2690 if (slbv & SLBV_L) 2691 va &= ~moea64_large_page_mask; 2692 else 2693 va &= ~ADDR_POFF; 2694 ptegidx = va_to_pteg(vsid, va, slbv & SLBV_L); 2695 #else 2696 va &= ~ADDR_POFF; 2697 vsid = va_to_vsid(pm, va); 2698 ptegidx = va_to_pteg(vsid, va, 0); 2699 #endif 2700 2701 LOCK_TABLE(); 2702 LIST_FOREACH(pvo, &moea64_pvo_table[ptegidx], pvo_olink) { 2703 if (pvo->pvo_pmap == pm && PVO_VADDR(pvo) == va) 2704 break; 2705 } 2706 UNLOCK_TABLE(); 2707 2708 return (pvo); 2709 } 2710 2711 static struct lpte * 2712 moea64_pvo_to_pte(const struct pvo_entry *pvo) 2713 { 2714 struct lpte *pt; 2715 int pteidx, ptegidx; 2716 uint64_t vsid; 2717 2718 ASSERT_TABLE_LOCK(); 2719 2720 /* If the PTEG index is not set, then there is no page table entry */ 2721 if (!PVO_PTEGIDX_ISSET(pvo)) 2722 return (NULL); 2723 2724 /* 2725 * Calculate the ptegidx 2726 */ 2727 vsid = PVO_VSID(pvo); 2728 ptegidx = va_to_pteg(vsid, PVO_VADDR(pvo), 2729 pvo->pvo_vaddr & PVO_LARGE); 2730 2731 /* 2732 * We can find the actual pte entry without searching by grabbing 2733 * the PTEG index from 3 unused bits in pvo_vaddr and by 2734 * noticing the HID bit. 2735 */ 2736 if (pvo->pvo_pte.lpte.pte_hi & LPTE_HID) 2737 ptegidx ^= moea64_pteg_mask; 2738 2739 pteidx = (ptegidx << 3) | PVO_PTEGIDX_GET(pvo); 2740 2741 if ((pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) && 2742 !PVO_PTEGIDX_ISSET(pvo)) { 2743 panic("moea64_pvo_to_pte: pvo %p has valid pte in pvo but no " 2744 "valid pte index", pvo); 2745 } 2746 2747 if ((pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) == 0 && 2748 PVO_PTEGIDX_ISSET(pvo)) { 2749 panic("moea64_pvo_to_pte: pvo %p has valid pte index in pvo " 2750 "pvo but no valid pte", pvo); 2751 } 2752 2753 pt = &moea64_pteg_table[pteidx >> 3].pt[pteidx & 7]; 2754 if ((pt->pte_hi ^ (pvo->pvo_pte.lpte.pte_hi & ~LPTE_VALID)) == 2755 LPTE_VALID) { 2756 if ((pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) == 0) { 2757 panic("moea64_pvo_to_pte: pvo %p has valid pte in " 2758 "moea64_pteg_table %p but invalid in pvo", pvo, pt); 2759 } 2760 2761 if (((pt->pte_lo ^ pvo->pvo_pte.lpte.pte_lo) & 2762 ~(LPTE_M|LPTE_CHG|LPTE_REF)) != 0) { 2763 panic("moea64_pvo_to_pte: pvo %p pte does not match " 2764 "pte %p in moea64_pteg_table difference is %#x", 2765 pvo, pt, 2766 (uint32_t)(pt->pte_lo ^ pvo->pvo_pte.lpte.pte_lo)); 2767 } 2768 2769 return (pt); 2770 } 2771 2772 if (pvo->pvo_pte.lpte.pte_hi & LPTE_VALID) { 2773 panic("moea64_pvo_to_pte: pvo %p has invalid pte %p in " 2774 "moea64_pteg_table but valid in pvo", pvo, pt); 2775 } 2776 2777 return (NULL); 2778 } 2779 2780 static __inline int 2781 moea64_pte_spillable_ident(u_int ptegidx) 2782 { 2783 struct lpte *pt; 2784 int i, j, k; 2785 2786 /* Start at a random slot */ 2787 i = mftb() % 8; 2788 k = -1; 2789 for (j = 0; j < 8; j++) { 2790 pt = &moea64_pteg_table[ptegidx].pt[(i + j) % 8]; 2791 if (pt->pte_hi & (LPTE_LOCKED | LPTE_WIRED)) 2792 continue; 2793 2794 /* This is a candidate, so remember it */ 2795 k = (i + j) % 8; 2796 2797 /* Try to get a page that has not been used lately */ 2798 if (!(pt->pte_lo & LPTE_REF)) 2799 return (k); 2800 } 2801 2802 return (k); 2803 } 2804 2805 static int 2806 moea64_pte_insert(u_int ptegidx, struct lpte *pvo_pt) 2807 { 2808 struct lpte *pt; 2809 struct pvo_entry *pvo; 2810 u_int pteg_bktidx; 2811 int i; 2812 2813 ASSERT_TABLE_LOCK(); 2814 2815 /* 2816 * First try primary hash. 2817 */ 2818 pteg_bktidx = ptegidx; 2819 for (pt = moea64_pteg_table[pteg_bktidx].pt, i = 0; i < 8; i++, pt++) { 2820 if ((pt->pte_hi & (LPTE_VALID | LPTE_LOCKED)) == 0) { 2821 pvo_pt->pte_hi &= ~LPTE_HID; 2822 moea64_pte_set(pt, pvo_pt); 2823 return (i); 2824 } 2825 } 2826 2827 /* 2828 * Now try secondary hash. 2829 */ 2830 pteg_bktidx ^= moea64_pteg_mask; 2831 for (pt = moea64_pteg_table[pteg_bktidx].pt, i = 0; i < 8; i++, pt++) { 2832 if ((pt->pte_hi & (LPTE_VALID | LPTE_LOCKED)) == 0) { 2833 pvo_pt->pte_hi |= LPTE_HID; 2834 moea64_pte_set(pt, pvo_pt); 2835 return (i); 2836 } 2837 } 2838 2839 /* 2840 * Out of luck. Find a PTE to sacrifice. 2841 */ 2842 pteg_bktidx = ptegidx; 2843 i = moea64_pte_spillable_ident(pteg_bktidx); 2844 if (i < 0) { 2845 pteg_bktidx ^= moea64_pteg_mask; 2846 i = moea64_pte_spillable_ident(pteg_bktidx); 2847 } 2848 2849 if (i < 0) { 2850 /* No freeable slots in either PTEG? We're hosed. */ 2851 panic("moea64_pte_insert: overflow"); 2852 return (-1); 2853 } 2854 2855 if (pteg_bktidx == ptegidx) 2856 pvo_pt->pte_hi &= ~LPTE_HID; 2857 else 2858 pvo_pt->pte_hi |= LPTE_HID; 2859 2860 /* 2861 * Synchronize the sacrifice PTE with its PVO, then mark both 2862 * invalid. The PVO will be reused when/if the VM system comes 2863 * here after a fault. 2864 */ 2865 pt = &moea64_pteg_table[pteg_bktidx].pt[i]; 2866 2867 if (pt->pte_hi & LPTE_HID) 2868 pteg_bktidx ^= moea64_pteg_mask; /* PTEs indexed by primary */ 2869 2870 LIST_FOREACH(pvo, &moea64_pvo_table[pteg_bktidx], pvo_olink) { 2871 if (pvo->pvo_pte.lpte.pte_hi == pt->pte_hi) { 2872 KASSERT(pvo->pvo_pte.lpte.pte_hi & LPTE_VALID, 2873 ("Invalid PVO for valid PTE!")); 2874 moea64_pte_unset(pt, &pvo->pvo_pte.lpte, pvo->pvo_vpn); 2875 PVO_PTEGIDX_CLR(pvo); 2876 moea64_pte_overflow++; 2877 break; 2878 } 2879 } 2880 2881 KASSERT(pvo->pvo_pte.lpte.pte_hi == pt->pte_hi, 2882 ("Unable to find PVO for spilled PTE")); 2883 2884 /* 2885 * Set the new PTE. 2886 */ 2887 moea64_pte_set(pt, pvo_pt); 2888 2889 return (i); 2890 } 2891 2892 static boolean_t 2893 moea64_query_bit(vm_page_t m, u_int64_t ptebit) 2894 { 2895 struct pvo_entry *pvo; 2896 struct lpte *pt; 2897 2898 if (moea64_attr_fetch(m) & ptebit) 2899 return (TRUE); 2900 2901 vm_page_lock_queues(); 2902 2903 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) { 2904 MOEA_PVO_CHECK(pvo); /* sanity check */ 2905 2906 /* 2907 * See if we saved the bit off. If so, cache it and return 2908 * success. 2909 */ 2910 if (pvo->pvo_pte.lpte.pte_lo & ptebit) { 2911 moea64_attr_save(m, ptebit); 2912 MOEA_PVO_CHECK(pvo); /* sanity check */ 2913 vm_page_unlock_queues(); 2914 return (TRUE); 2915 } 2916 } 2917 2918 /* 2919 * No luck, now go through the hard part of looking at the PTEs 2920 * themselves. Sync so that any pending REF/CHG bits are flushed to 2921 * the PTEs. 2922 */ 2923 SYNC(); 2924 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) { 2925 MOEA_PVO_CHECK(pvo); /* sanity check */ 2926 2927 /* 2928 * See if this pvo has a valid PTE. if so, fetch the 2929 * REF/CHG bits from the valid PTE. If the appropriate 2930 * ptebit is set, cache it and return success. 2931 */ 2932 LOCK_TABLE(); 2933 pt = moea64_pvo_to_pte(pvo); 2934 if (pt != NULL) { 2935 moea64_pte_synch(pt, &pvo->pvo_pte.lpte); 2936 if (pvo->pvo_pte.lpte.pte_lo & ptebit) { 2937 UNLOCK_TABLE(); 2938 2939 moea64_attr_save(m, ptebit); 2940 MOEA_PVO_CHECK(pvo); /* sanity check */ 2941 vm_page_unlock_queues(); 2942 return (TRUE); 2943 } 2944 } 2945 UNLOCK_TABLE(); 2946 } 2947 2948 vm_page_unlock_queues(); 2949 return (FALSE); 2950 } 2951 2952 static u_int 2953 moea64_clear_bit(vm_page_t m, u_int64_t ptebit) 2954 { 2955 u_int count; 2956 struct pvo_entry *pvo; 2957 struct lpte *pt; 2958 2959 vm_page_lock_queues(); 2960 2961 /* 2962 * Clear the cached value. 2963 */ 2964 moea64_attr_clear(m, ptebit); 2965 2966 /* 2967 * Sync so that any pending REF/CHG bits are flushed to the PTEs (so 2968 * we can reset the right ones). note that since the pvo entries and 2969 * list heads are accessed via BAT0 and are never placed in the page 2970 * table, we don't have to worry about further accesses setting the 2971 * REF/CHG bits. 2972 */ 2973 SYNC(); 2974 2975 /* 2976 * For each pvo entry, clear the pvo's ptebit. If this pvo has a 2977 * valid pte clear the ptebit from the valid pte. 2978 */ 2979 count = 0; 2980 LIST_FOREACH(pvo, vm_page_to_pvoh(m), pvo_vlink) { 2981 MOEA_PVO_CHECK(pvo); /* sanity check */ 2982 2983 LOCK_TABLE(); 2984 pt = moea64_pvo_to_pte(pvo); 2985 if (pt != NULL) { 2986 moea64_pte_synch(pt, &pvo->pvo_pte.lpte); 2987 if (pvo->pvo_pte.lpte.pte_lo & ptebit) { 2988 count++; 2989 moea64_pte_clear(pt, pvo->pvo_vpn, ptebit); 2990 } 2991 } 2992 pvo->pvo_pte.lpte.pte_lo &= ~ptebit; 2993 MOEA_PVO_CHECK(pvo); /* sanity check */ 2994 UNLOCK_TABLE(); 2995 } 2996 2997 vm_page_unlock_queues(); 2998 return (count); 2999 } 3000 3001 boolean_t 3002 moea64_dev_direct_mapped(mmu_t mmu, vm_offset_t pa, vm_size_t size) 3003 { 3004 struct pvo_entry *pvo; 3005 vm_offset_t ppa; 3006 int error = 0; 3007 3008 PMAP_LOCK(kernel_pmap); 3009 for (ppa = pa & ~ADDR_POFF; ppa < pa + size; ppa += PAGE_SIZE) { 3010 pvo = moea64_pvo_find_va(kernel_pmap, ppa); 3011 if (pvo == NULL || 3012 (pvo->pvo_pte.lpte.pte_lo & LPTE_RPGN) != ppa) { 3013 error = EFAULT; 3014 break; 3015 } 3016 } 3017 PMAP_UNLOCK(kernel_pmap); 3018 3019 return (error); 3020 } 3021 3022 /* 3023 * Map a set of physical memory pages into the kernel virtual 3024 * address space. Return a pointer to where it is mapped. This 3025 * routine is intended to be used for mapping device memory, 3026 * NOT real memory. 3027 */ 3028 void * 3029 moea64_mapdev_attr(mmu_t mmu, vm_offset_t pa, vm_size_t size, vm_memattr_t ma) 3030 { 3031 vm_offset_t va, tmpva, ppa, offset; 3032 3033 ppa = trunc_page(pa); 3034 offset = pa & PAGE_MASK; 3035 size = roundup(offset + size, PAGE_SIZE); 3036 3037 va = kmem_alloc_nofault(kernel_map, size); 3038 3039 if (!va) 3040 panic("moea64_mapdev: Couldn't alloc kernel virtual memory"); 3041 3042 for (tmpva = va; size > 0;) { 3043 moea64_kenter_attr(mmu, tmpva, ppa, ma); 3044 size -= PAGE_SIZE; 3045 tmpva += PAGE_SIZE; 3046 ppa += PAGE_SIZE; 3047 } 3048 3049 return ((void *)(va + offset)); 3050 } 3051 3052 void * 3053 moea64_mapdev(mmu_t mmu, vm_offset_t pa, vm_size_t size) 3054 { 3055 3056 return moea64_mapdev_attr(mmu, pa, size, VM_MEMATTR_DEFAULT); 3057 } 3058 3059 void 3060 moea64_unmapdev(mmu_t mmu, vm_offset_t va, vm_size_t size) 3061 { 3062 vm_offset_t base, offset; 3063 3064 base = trunc_page(va); 3065 offset = va & PAGE_MASK; 3066 size = roundup(offset + size, PAGE_SIZE); 3067 3068 kmem_free(kernel_map, base, size); 3069 } 3070 3071 static void 3072 moea64_sync_icache(mmu_t mmu, pmap_t pm, vm_offset_t va, vm_size_t sz) 3073 { 3074 struct pvo_entry *pvo; 3075 vm_offset_t lim; 3076 vm_paddr_t pa; 3077 vm_size_t len; 3078 3079 PMAP_LOCK(pm); 3080 while (sz > 0) { 3081 lim = round_page(va); 3082 len = MIN(lim - va, sz); 3083 pvo = moea64_pvo_find_va(pm, va & ~ADDR_POFF); 3084 if (pvo != NULL) { 3085 pa = (pvo->pvo_pte.pte.pte_lo & LPTE_RPGN) | 3086 (va & ADDR_POFF); 3087 moea64_syncicache(pm, va, pa, len); 3088 } 3089 va += len; 3090 sz -= len; 3091 } 3092 PMAP_UNLOCK(pm); 3093 } 3094