1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */ 27 /* All Rights Reserved */ 28 29 /* 30 * University Copyright- Copyright (c) 1982, 1986, 1988 31 * The Regents of the University of California 32 * All Rights Reserved 33 * 34 * University Acknowledgment- Portions of this document are derived from 35 * software developed by the University of California, Berkeley, and its 36 * contributors. 37 */ 38 39 #ifndef _VM_HAT_H 40 #define _VM_HAT_H 41 42 #pragma ident "%Z%%M% %I% %E% SMI" 43 44 #include <sys/types.h> 45 #include <sys/t_lock.h> 46 #include <vm/faultcode.h> 47 #include <sys/kstat.h> 48 #include <sys/siginfo.h> 49 50 #ifdef __cplusplus 51 extern "C" { 52 #endif 53 54 /* 55 * VM - Hardware Address Translation management. 56 * 57 * This file describes the machine independent interfaces to 58 * the hardware address translation management routines. Other 59 * machine specific interfaces and structures are defined 60 * in <vm/hat_xxx.h>. The hat layer manages the address 61 * translation hardware as a cache driven by calls from the 62 * higher levels of the VM system. 63 */ 64 65 struct hat; 66 struct kpme; 67 struct memseg; 68 69 #include <vm/page.h> 70 71 /* 72 * a callback used with hat_unload_callback() 73 * start and end mark are set to a range of unloaded addresses 74 * and the function is invoked with a pointer to this data structure 75 */ 76 typedef struct hat_callback { 77 caddr_t hcb_start_addr; 78 caddr_t hcb_end_addr; 79 void (*hcb_function)(struct hat_callback *); 80 void *hcb_data; 81 } hat_callback_t; 82 83 #ifdef _KERNEL 84 85 /* 86 * One time hat initialization 87 */ 88 void hat_init(void); 89 90 /* 91 * Notify hat of a system dump 92 */ 93 void hat_dump(void); 94 95 /* 96 * Operations on an address space: 97 * 98 * struct hat *hat_alloc(as) 99 * allocated a hat structure for as. 100 * 101 * void hat_free_start(hat) 102 * informs hat layer process has finished executing but as has not 103 * been cleaned up yet. 104 * 105 * void hat_free_end(hat) 106 * informs hat layer as is being destroyed. hat layer cannot use as 107 * pointer after this call. 108 * 109 * void hat_swapin(hat) 110 * allocate any hat resources required for process being swapped in. 111 * 112 * void hat_swapout(hat) 113 * deallocate hat resources for process being swapped out. 114 * 115 * size_t hat_get_mapped_size(hat) 116 * returns number of bytes that have valid mappings in hat. 117 * 118 * void hat_stats_enable(hat) 119 * void hat_stats_disable(hat) 120 * enables/disables collection of stats for hat. 121 * 122 * int hat_dup(parenthat, childhat, addr, len, flags) 123 * Duplicate address translations of the parent to the child. Supports 124 * the entire address range or a range depending on flag, 125 * zero returned on success, non-zero on error 126 * 127 * void hat_thread_exit(thread) 128 * Notifies the HAT that a thread is exiting, called after it has been 129 * reassigned to the kernel AS. 130 */ 131 132 struct hat *hat_alloc(struct as *); 133 void hat_free_start(struct hat *); 134 void hat_free_end(struct hat *); 135 int hat_dup(struct hat *, struct hat *, caddr_t, size_t, uint_t); 136 void hat_swapin(struct hat *); 137 void hat_swapout(struct hat *); 138 size_t hat_get_mapped_size(struct hat *); 139 int hat_stats_enable(struct hat *); 140 void hat_stats_disable(struct hat *); 141 void hat_thread_exit(kthread_t *); 142 143 /* 144 * Operations on a named address within a segment: 145 * 146 * void hat_memload(hat, addr, pp, attr, flags) 147 * load/lock the given page struct 148 * 149 * void hat_memload_array(hat, addr, len, ppa, attr, flags) 150 * load/lock the given array of page structs 151 * 152 * void hat_devload(hat, addr, len, pf, attr, flags) 153 * load/lock the given page frame number 154 * 155 * void hat_unlock(hat, addr, len) 156 * unlock a given range of addresses 157 * 158 * void hat_unload(hat, addr, len, flags) 159 * void hat_unload_callback(hat, addr, len, flags, callback) 160 * unload a given range of addresses (has optional callback) 161 * 162 * void hat_sync(hat, addr, len, flags) 163 * synchronize mapping with software data structures 164 * 165 * void hat_map(hat, addr, len, flags) 166 * 167 * void hat_setattr(hat, addr, len, attr) 168 * void hat_clrattr(hat, addr, len, attr) 169 * void hat_chgattr(hat, addr, len, attr) 170 * modify attributes for a range of addresses. skips any invalid mappings 171 * 172 * uint_t hat_getattr(hat, addr, *attr) 173 * returns attr for <hat,addr> in *attr. returns 0 if there was a 174 * mapping and *attr is valid, nonzero if there was no mapping and 175 * *attr is not valid. 176 * 177 * size_t hat_getpagesize(hat, addr) 178 * returns pagesize in bytes for <hat, addr>. returns -1 if there is 179 * no mapping. This is an advisory call. 180 * 181 * pfn_t hat_getpfnum(hat, addr) 182 * returns pfn for <hat, addr> or PFN_INVALID if mapping is invalid. 183 * 184 * pfn_t hat_getkpfnum(addr) 185 * returns pfn for non-memory mapped addr in kernel address space 186 * or PFN_INVALID if mapping is invalid or is kernel memory. 187 * 188 * int hat_probe(hat, addr) 189 * return 0 if no valid mapping is present. Faster version 190 * of hat_getattr in certain architectures. 191 * 192 * int hat_share(dhat, daddr, shat, saddr, len, szc) 193 * 194 * void hat_unshare(hat, addr, len, szc) 195 * 196 * void hat_chgprot(hat, addr, len, vprot) 197 * This is a deprecated call. New segment drivers should store 198 * all attributes and use hat_*attr calls. 199 * Change the protections in the virtual address range 200 * given to the specified virtual protection. If vprot is ~PROT_WRITE, 201 * then remove write permission, leaving the other permissions 202 * unchanged. If vprot is ~PROT_USER, remove user permissions. 203 */ 204 205 void hat_memload(struct hat *, caddr_t, struct page *, uint_t, uint_t); 206 void hat_memload_array(struct hat *, caddr_t, size_t, struct page **, 207 uint_t, uint_t); 208 209 void hat_devload(struct hat *, caddr_t, size_t, pfn_t, uint_t, int); 210 void hat_unlock(struct hat *, caddr_t, size_t); 211 void hat_unload(struct hat *, caddr_t, size_t, uint_t); 212 void hat_unload_callback(struct hat *, caddr_t, size_t, uint_t, 213 hat_callback_t *); 214 void hat_sync(struct hat *, caddr_t, size_t, uint_t); 215 void hat_map(struct hat *, caddr_t, size_t, uint_t); 216 void hat_setattr(struct hat *, caddr_t, size_t, uint_t); 217 void hat_clrattr(struct hat *, caddr_t, size_t, uint_t); 218 void hat_chgattr(struct hat *, caddr_t, size_t, uint_t); 219 uint_t hat_getattr(struct hat *, caddr_t, uint_t *); 220 ssize_t hat_getpagesize(struct hat *, caddr_t); 221 pfn_t hat_getpfnum(struct hat *, caddr_t); 222 int hat_probe(struct hat *, caddr_t); 223 int hat_share(struct hat *, caddr_t, struct hat *, caddr_t, size_t, uint_t); 224 void hat_unshare(struct hat *, caddr_t, size_t, uint_t); 225 void hat_chgprot(struct hat *, caddr_t, size_t, uint_t); 226 void hat_reserve(struct as *, caddr_t, size_t); 227 pfn_t va_to_pfn(void *); 228 uint64_t va_to_pa(void *); 229 230 /* 231 * hat_getkpfnum() is never supported on amd64 and will be 232 * removed from other platforms in future release 233 */ 234 #if !defined(__amd64) 235 pfn_t hat_getkpfnum(caddr_t); 236 #endif 237 238 239 /* 240 * Kernel Physical Mapping (segkpm) hat interface routines. 241 */ 242 caddr_t hat_kpm_mapin(struct page *, struct kpme *); 243 void hat_kpm_mapout(struct page *, struct kpme *, caddr_t); 244 caddr_t hat_kpm_page2va(struct page *, int); 245 struct page *hat_kpm_vaddr2page(caddr_t); 246 int hat_kpm_fault(struct hat *, caddr_t); 247 void hat_kpm_mseghash_clear(int); 248 void hat_kpm_mseghash_update(pgcnt_t, struct memseg *); 249 void hat_kpm_addmem_mseg_update(struct memseg *, pgcnt_t, offset_t); 250 void hat_kpm_addmem_mseg_insert(struct memseg *); 251 void hat_kpm_addmem_memsegs_update(struct memseg *); 252 caddr_t hat_kpm_mseg_reuse(struct memseg *); 253 void hat_kpm_delmem_mseg_update(struct memseg *, struct memseg **); 254 void hat_kpm_split_mseg_update(struct memseg *, struct memseg **, 255 struct memseg *, struct memseg *, struct memseg *); 256 void hat_kpm_walk(void (*)(void *, void *, size_t), void *); 257 258 /* 259 * Operations on all translations for a given page(s) 260 * 261 * void hat_page_setattr(pp, flag) 262 * void hat_page_clrattr(pp, flag) 263 * used to set/clr red/mod bits. 264 * 265 * uint hat_page_getattr(pp, flag) 266 * If flag is specified, returns 0 if attribute is disabled 267 * and non zero if enabled. If flag specifes multiple attributs 268 * then returns 0 if ALL atriibutes are disabled. This is an advisory 269 * call. 270 * 271 * int hat_pageunload(pp, forceflag) 272 * unload all translations attached to pp. 273 * 274 * uint_t hat_pagesync(pp, flags) 275 * get hw stats from hardware into page struct and reset hw stats 276 * returns attributes of page 277 * 278 * ulong_t hat_page_getshare(pp) 279 * returns approx number of mappings to this pp. A return of 0 implies 280 * there are no mappings to the page. 281 * 282 * faultcode_t hat_softlock(hat, addr, lenp, ppp, flags); 283 * called to softlock pages for zero copy tcp 284 * 285 * void hat_page_demote(pp); 286 * unload all large mappings to pp and decrease p_szc of all 287 * constituent pages according to the remaining mappings. 288 */ 289 290 void hat_page_setattr(struct page *, uint_t); 291 void hat_page_clrattr(struct page *, uint_t); 292 uint_t hat_page_getattr(struct page *, uint_t); 293 int hat_pageunload(struct page *, uint_t); 294 uint_t hat_pagesync(struct page *, uint_t); 295 ulong_t hat_page_getshare(struct page *); 296 faultcode_t hat_softlock(struct hat *, caddr_t, size_t *, 297 struct page **, uint_t); 298 void hat_page_demote(struct page *); 299 300 /* 301 * Rountine to expose supported HAT features to PIM. 302 */ 303 enum hat_features { 304 HAT_SHARED_PT, /* Shared page tables */ 305 HAT_DYNAMIC_ISM_UNMAP, /* hat_pageunload() handles ISM pages */ 306 HAT_VMODSORT /* support for VMODSORT flag of vnode */ 307 }; 308 309 int hat_supported(enum hat_features, void *); 310 311 /* 312 * Services provided to the hat: 313 * 314 * void as_signal_proc(as, siginfo) 315 * deliver signal to all processes that have this as. 316 * 317 * int hat_setstat(as, addr, len, rmbits) 318 * informs hatstat layer that ref/mod bits need to be updated for 319 * address range. Returns 0 on success, 1 for failure. 320 */ 321 void as_signal_proc(struct as *, k_siginfo_t *siginfo); 322 void hat_setstat(struct as *, caddr_t, size_t, uint_t); 323 324 /* 325 * Flags to pass to hat routines. 326 * 327 * Certain flags only apply to some interfaces: 328 * 329 * HAT_LOAD Default flags to load a translation to the page. 330 * HAT_LOAD_LOCK Lock down mapping resources; hat_map(), hat_memload(), 331 * and hat_devload(). 332 * HAT_LOAD_ADV Advisory load - Load translation if and only if 333 * sufficient MMU resources exist (i.e., do not steal). 334 * HAT_LOAD_SHARE A flag to hat_memload() to indicate h/w page tables 335 * that map some user pages (not kas) is shared by more 336 * than one process (eg. ISM). 337 * HAT_LOAD_CONTIG Pages are contigous 338 * HAT_LOAD_NOCONSIST Do not add mapping to mapping list. 339 * HAT_LOAD_REMAP Reload a valid pte with a different page frame. 340 * HAT_RELOAD_SHARE Reload a shared page table entry. Some platforms 341 * may require different actions than on the first 342 * load of a shared mapping. 343 * HAT_NO_KALLOC Do not kmem_alloc while creating the mapping; at this 344 * point, it's setting up mapping to allocate internal 345 * hat layer data structures. This flag forces hat layer 346 * to tap its reserves in order to prevent infinite 347 * recursion. 348 * HAT_LOAD_AUTOLPG Get MMU specific disable_auto_large_pages 349 */ 350 351 /* 352 * Flags for hat_memload/hat_devload 353 */ 354 #define HAT_FLAGS_RESV 0xFF000000 /* resv for hat impl */ 355 #define HAT_LOAD 0x00 356 #define HAT_LOAD_LOCK 0x01 357 #define HAT_LOAD_ADV 0x04 358 #define HAT_LOAD_CONTIG 0x10 359 #define HAT_LOAD_NOCONSIST 0x20 360 #define HAT_LOAD_SHARE 0x40 361 #define HAT_LOAD_REMAP 0x80 362 #define HAT_RELOAD_SHARE 0x100 363 #define HAT_NO_KALLOC 0x200 364 #define HAT_LOAD_TEXT 0x400 365 #define HAT_LOAD_AUTOLPG 0x800 366 367 /* 368 * Attributes for hat_memload/hat_devload/hat_*attr 369 * are a superset of prot flags defined in mman.h. 370 */ 371 #define HAT_PLAT_ATTR_MASK 0xF00000 372 #define HAT_PROT_MASK 0x0F 373 374 #define HAT_NOFAULT 0x10 375 #define HAT_NOSYNC 0x20 376 377 /* 378 * Advisory ordering attributes. Apply only to device mappings. 379 * 380 * HAT_STRICTORDER: the CPU must issue the references in order, as the 381 * programmer specified. This is the default. 382 * HAT_UNORDERED_OK: the CPU may reorder the references (this is all kinds 383 * of reordering; store or load with store or load). 384 * HAT_MERGING_OK: merging and batching: the CPU may merge individual stores 385 * to consecutive locations (for example, turn two consecutive byte 386 * stores into one halfword store), and it may batch individual loads 387 * (for example, turn two consecutive byte loads into one halfword load). 388 * This also implies re-ordering. 389 * HAT_LOADCACHING_OK: the CPU may cache the data it fetches and reuse it 390 * until another store occurs. The default is to fetch new data 391 * on every load. This also implies merging. 392 * HAT_STORECACHING_OK: the CPU may keep the data in the cache and push it to 393 * the device (perhaps with other data) at a later time. The default is 394 * to push the data right away. This also implies load caching. 395 */ 396 #define HAT_STRICTORDER 0x0000 397 #define HAT_UNORDERED_OK 0x0100 398 #define HAT_MERGING_OK 0x0200 399 #define HAT_LOADCACHING_OK 0x0300 400 #define HAT_STORECACHING_OK 0x0400 401 #define HAT_ORDER_MASK 0x0700 402 403 /* endian attributes */ 404 #define HAT_NEVERSWAP 0x0000 405 #define HAT_STRUCTURE_BE 0x1000 406 #define HAT_STRUCTURE_LE 0x2000 407 #define HAT_ENDIAN_MASK 0x3000 408 409 /* flags for hat_softlock */ 410 #define HAT_COW 0x0001 411 412 /* 413 * Flags for hat_unload 414 */ 415 #define HAT_UNLOAD 0x00 416 #define HAT_UNLOAD_NOSYNC 0x02 417 #define HAT_UNLOAD_UNLOCK 0x04 418 #define HAT_UNLOAD_OTHER 0x08 419 #define HAT_UNLOAD_UNMAP 0x10 420 421 /* 422 * Flags for hat_pagesync, hat_getstat, hat_sync 423 */ 424 #define HAT_SYNC_DONTZERO 0x00 425 #define HAT_SYNC_ZERORM 0x01 426 /* Additional flags for hat_pagesync */ 427 #define HAT_SYNC_STOPON_REF 0x02 428 #define HAT_SYNC_STOPON_MOD 0x04 429 #define HAT_SYNC_STOPON_RM (HAT_SYNC_STOPON_REF | HAT_SYNC_STOPON_MOD) 430 #define HAT_SYNC_STOPON_SHARED 0x08 431 432 /* 433 * Flags for hat_dup 434 * 435 * HAT_DUP_ALL dup entire address space 436 * HAT_DUP_COW dup plus hat_clrattr(..PROT_WRITE) on newas 437 */ 438 #define HAT_DUP_ALL 1 439 #define HAT_DUP_COW 2 440 441 442 /* 443 * Flags for hat_map 444 */ 445 #define HAT_MAP 0x00 446 447 /* 448 * Flag for hat_pageunload 449 */ 450 #define HAT_ADV_PGUNLOAD 0x00 451 #define HAT_FORCE_PGUNLOAD 0x01 452 453 /* 454 * Attributes for hat_page_*attr, hat_setstats and 455 * returned by hat_pagesync. 456 */ 457 #define P_MOD 0x1 /* the modified bit */ 458 #define P_REF 0x2 /* the referenced bit */ 459 #define P_RO 0x4 /* Read only page */ 460 461 #define hat_ismod(pp) (hat_page_getattr(pp, P_MOD)) 462 #define hat_isref(pp) (hat_page_getattr(pp, P_REF)) 463 #define hat_isro(pp) (hat_page_getattr(pp, P_RO)) 464 465 #define hat_setmod(pp) (hat_page_setattr(pp, P_MOD)) 466 #define hat_setref(pp) (hat_page_setattr(pp, P_REF)) 467 #define hat_setrefmod(pp) (hat_page_setattr(pp, P_REF|P_MOD)) 468 469 #define hat_clrmod(pp) (hat_page_clrattr(pp, P_MOD)) 470 #define hat_clrref(pp) (hat_page_clrattr(pp, P_REF)) 471 #define hat_clrrefmod(pp) (hat_page_clrattr(pp, P_REF|P_MOD)) 472 473 #define hat_page_is_mapped(pp) (hat_page_getshare(pp)) 474 475 /* 476 * hat_setup is being used in sparc/os/sundep.c 477 */ 478 void hat_setup(struct hat *, int); 479 480 /* 481 * Flags for hat_setup 482 */ 483 #define HAT_DONTALLOC 0 484 #define HAT_ALLOC 1 485 #define HAT_INIT 2 486 487 /* 488 * Other routines, for statistics 489 */ 490 int hat_startstat(struct as *); 491 void hat_getstat(struct as *, caddr_t, size_t, uint_t, char *, int); 492 void hat_freestat(struct as *, int); 493 void hat_resvstat(size_t, struct as *, caddr_t); 494 495 /* 496 * Transitionary routine while we still allow hat_getkpfnum(caddr_t) 497 * to return a pfn for kernel memory, but want to warn the user that 498 * it isn't supported. 499 */ 500 void hat_getkpfnum_badcall(void *caller); 501 502 /* 503 * Relocation callback routines. Currently only sfmmu HAT supports 504 * these. 505 */ 506 extern int hat_add_callback(id_t, caddr_t, uint_t, uint_t, void *, 507 pfn_t *, void **); 508 extern id_t hat_register_callback(int, 509 int (*prehandler)(caddr_t, uint_t, uint_t, void *), 510 int (*posthandler)(caddr_t, uint_t, uint_t, void *, pfn_t), 511 int (*errhandler)(caddr_t, uint_t, uint_t, void *), int); 512 extern void hat_delete_callback(caddr_t, uint_t, void *, uint_t, void *); 513 514 /* 515 * hat_add_callback()/hat_delete_callback() flags. 516 */ 517 #define HAC_NOSLEEP 0x0 518 #define HAC_SLEEP 0x1 519 #define HAC_PAGELOCK 0x2 520 521 /* 522 * Suspend/unsuspend handler callback arguments. 523 */ 524 #define HAT_SUSPEND 0x0010 525 #define HAT_UNSUSPEND 0x0010 526 #define HAT_PRESUSPEND 0x0020 527 #define HAT_POSTUNSUSPEND 0x0020 528 529 /* 530 * Error handler callback arguments. See the block comments 531 * before the implementation of hat_add_callback() for an 532 * explanation of what these mean. 533 */ 534 #define HAT_CB_ERR_LEAKED 0x1 535 536 #endif /* _KERNEL */ 537 538 /* 539 * The size of the bit array for ref and mod bit storage must be a power of 2. 540 * 2 bits are collected for each page. Below the power used is 4, 541 * which is 16 8-bit characters = 128 bits, ref and mod bit information 542 * for 64 pages. 543 */ 544 #define HRM_SHIFT 4 545 #define HRM_BYTES (1 << HRM_SHIFT) 546 #define HRM_PAGES ((HRM_BYTES * NBBY) / 2) 547 #define HRM_PGPERBYTE (NBBY/2) 548 #define HRM_PGBYTEMASK (HRM_PGPERBYTE-1) 549 550 #define HRM_PGOFFMASK ((HRM_PGPERBYTE-1) << MMU_PAGESHIFT) 551 #define HRM_BASEOFFSET (((MMU_PAGESIZE * HRM_PAGES) - 1)) 552 #define HRM_BASEMASK (~(HRM_BASEOFFSET)) 553 554 #define HRM_BASESHIFT (MMU_PAGESHIFT + (HRM_SHIFT + 2)) 555 #define HRM_PAGEMASK (MMU_PAGEMASK ^ HRM_BASEMASK) 556 557 #define HRM_HASHSIZE 0x200 558 #define HRM_HASHMASK (HRM_HASHSIZE - 1) 559 560 #define HRM_BLIST_INCR 0x200 561 562 /* 563 * The structure for maintaining referenced and modified information 564 */ 565 struct hrmstat { 566 struct as *hrm_as; /* stat block belongs to this as */ 567 uintptr_t hrm_base; /* base of block */ 568 ushort_t hrm_id; /* opaque identifier, one of a_vbits */ 569 struct hrmstat *hrm_anext; /* as statistics block list */ 570 struct hrmstat *hrm_hnext; /* list for hashed blocks */ 571 uchar_t hrm_bits[HRM_BYTES]; /* the ref and mod bits */ 572 }; 573 574 /* 575 * For global monitoring of the reference and modified bits 576 * of all address spaces we reserve one id bit. 577 */ 578 #define HRM_SWSMONID 1 579 580 581 #ifdef _KERNEL 582 583 /* 584 * Hat locking functions 585 * XXX - these two functions are currently being used by hatstats 586 * they can be removed by using a per-as mutex for hatstats. 587 */ 588 void hat_enter(struct hat *); 589 void hat_exit(struct hat *); 590 591 #endif /* _KERNEL */ 592 593 #ifdef __cplusplus 594 } 595 #endif 596 597 #endif /* _VM_HAT_H */ 598