1 /* $OpenBSD: uvm_page.h,v 1.25 2007/04/18 18:51:37 art Exp $ */
2 /* $NetBSD: uvm_page.h,v 1.19 2000/12/28 08:24:55 chs Exp $ */
3
4 /*
5 * Copyright (c) 1997 Charles D. Cranor and Washington University.
6 * Copyright (c) 1991, 1993, The Regents of the University of California.
7 *
8 * All rights reserved.
9 *
10 * This code is derived from software contributed to Berkeley by
11 * The Mach Operating System project at Carnegie-Mellon University.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. All advertising materials mentioning features or use of this software
22 * must display the following acknowledgement:
23 * This product includes software developed by Charles D. Cranor,
24 * Washington University, the University of California, Berkeley and
25 * its contributors.
26 * 4. Neither the name of the University nor the names of its contributors
27 * may be used to endorse or promote products derived from this software
28 * without specific prior written permission.
29 *
30 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
31 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
32 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
33 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
34 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
35 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
36 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
37 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
38 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
39 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
40 * SUCH DAMAGE.
41 *
42 * @(#)vm_page.h 7.3 (Berkeley) 4/21/91
43 * from: Id: uvm_page.h,v 1.1.2.6 1998/02/04 02:31:42 chuck Exp
44 *
45 *
46 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
47 * All rights reserved.
48 *
49 * Permission to use, copy, modify and distribute this software and
50 * its documentation is hereby granted, provided that both the copyright
51 * notice and this permission notice appear in all copies of the
52 * software, derivative works or modified versions, and any portions
53 * thereof, and that both notices appear in supporting documentation.
54 *
55 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
56 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
57 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
58 *
59 * Carnegie Mellon requests users of this software to return to
60 *
61 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
62 * School of Computer Science
63 * Carnegie Mellon University
64 * Pittsburgh PA 15213-3890
65 *
66 * any improvements or extensions that they make and grant Carnegie the
67 * rights to redistribute these changes.
68 */
69
70 #ifndef _UVM_UVM_PAGE_H_
71 #define _UVM_UVM_PAGE_H_
72
73 /*
74 * uvm_page.h
75 */
76
77 /*
78 * Resident memory system definitions.
79 */
80
81 /*
82 * Management of resident (logical) pages.
83 *
84 * A small structure is kept for each resident
85 * page, indexed by page number. Each structure
86 * is an element of several lists:
87 *
88 * A hash table bucket used to quickly
89 * perform object/offset lookups
90 *
91 * A list of all pages for a given object,
92 * so they can be quickly deactivated at
93 * time of deallocation.
94 *
95 * An ordered list of pages due for pageout.
96 *
97 * In addition, the structure contains the object
98 * and offset to which this page belongs (for pageout),
99 * and sundry status bits.
100 *
101 * Fields in this structure are locked either by the lock on the
102 * object that the page belongs to (O) or by the lock on the page
103 * queues (P) [or both].
104 */
105
106 #include <uvm/uvm_extern.h>
107 #include <uvm/uvm_pglist.h>
108
109 struct vm_page {
110 TAILQ_ENTRY(vm_page) pageq; /* queue info for FIFO
111 * queue or free list (P) */
112 TAILQ_ENTRY(vm_page) hashq; /* hash table links (O)*/
113 TAILQ_ENTRY(vm_page) listq; /* pages in same object (O)*/
114
115 struct vm_anon *uanon; /* anon (O,P) */
116 struct uvm_object *uobject; /* object (O,P) */
117 voff_t offset; /* offset into object (O,P) */
118
119 u_int pg_flags; /* object flags [O or P] */
120
121 u_int pg_version; /* version count [O] */
122 u_int wire_count; /* wired down map refs [P] */
123
124 u_int loan_count; /* number of active loans
125 * to read: [O or P]
126 * to modify: [O _and_ P] */
127 paddr_t phys_addr; /* physical address of page */
128
129 #ifdef __HAVE_VM_PAGE_MD
130 struct vm_page_md mdpage; /* pmap-specific data */
131 #endif
132 #if defined(UVM_PAGE_TRKOWN)
133 /* debugging fields to track page ownership */
134 pid_t owner; /* proc that set PG_BUSY */
135 char *owner_tag; /* why it was set busy */
136 #endif
137 };
138
139 /*
140 * These are the flags defined for vm_page.
141 *
142 * Note: PG_FILLED and PG_DIRTY are added for the filesystems.
143 */
144
145 /*
146 * locking rules:
147 * PG_ ==> locked by object lock
148 * PQ_ ==> lock by page queue lock
149 * PQ_FREE is locked by free queue lock and is mutex with all other PQs
150 * pg_flags may only be changed using the atomic operations.
151 *
152 * PG_ZERO is used to indicate that a page has been pre-zero'd. This flag
153 * is only set when the page is on no queues, and is cleared when the page
154 * is placed on the free list.
155 */
156
157 #define PG_BUSY 0x00000001 /* page is locked */
158 #define PG_WANTED 0x00000002 /* someone is waiting for page */
159 #define PG_TABLED 0x00000004 /* page is in VP table */
160 #define PG_CLEAN 0x00000008 /* page has not been modified */
161 #define PG_CLEANCHK 0x00000010 /* clean bit has been checked */
162 #define PG_RELEASED 0x00000020 /* page released while paging */
163 #define PG_FAKE 0x00000040 /* page is not yet initialized */
164 #define PG_RDONLY 0x00000080 /* page must be mapped read-only */
165 #define PG_ZERO 0x00000100 /* page is pre-zero'd */
166
167 #define PG_PAGER1 0x00001000 /* pager-specific flag */
168 #define PG_MASK 0x0000ffff
169
170 #define PQ_FREE 0x00010000 /* page is on free list */
171 #define PQ_INACTIVE 0x00020000 /* page is in inactive list */
172 #define PQ_ACTIVE 0x00040000 /* page is in active list */
173 #define PQ_ANON 0x00100000 /* page is part of an anon, rather
174 than an uvm_object */
175 #define PQ_AOBJ 0x00200000 /* page is part of an anonymous
176 uvm_object */
177 #define PQ_SWAPBACKED (PQ_ANON|PQ_AOBJ)
178 #define PQ_ENCRYPT 0x00400000 /* page needs {en,de}cryption */
179 #define PQ_MASK 0x00ff0000
180
181 #define PG_PMAP0 0x01000000 /* Used by some pmaps. */
182 #define PG_PMAP1 0x02000000 /* Used by some pmaps. */
183 #define PG_PMAP2 0x04000000 /* Used by some pmaps. */
184 #define PG_PMAP3 0x08000000 /* Used by some pmaps. */
185
186 /*
187 * physical memory layout structure
188 *
189 * MD vmparam.h must #define:
190 * VM_PHYSEG_MAX = max number of physical memory segments we support
191 * (if this is "1" then we revert to a "contig" case)
192 * VM_PHYSSEG_STRAT: memory sort/search options (for VM_PHYSEG_MAX > 1)
193 * - VM_PSTRAT_RANDOM: linear search (random order)
194 * - VM_PSTRAT_BSEARCH: binary search (sorted by address)
195 * - VM_PSTRAT_BIGFIRST: linear search (sorted by largest segment first)
196 * - others?
197 * XXXCDC: eventually we should purge all left-over global variables...
198 */
199 #define VM_PSTRAT_RANDOM 1
200 #define VM_PSTRAT_BSEARCH 2
201 #define VM_PSTRAT_BIGFIRST 3
202
203 /*
204 * vm_physmemseg: describes one segment of physical memory
205 */
206 struct vm_physseg {
207 paddr_t start; /* PF# of first page in segment */
208 paddr_t end; /* (PF# of last page in segment) + 1 */
209 paddr_t avail_start; /* PF# of first free page in segment */
210 paddr_t avail_end; /* (PF# of last free page in segment) +1 */
211 int free_list; /* which free list they belong on */
212 struct vm_page *pgs; /* vm_page structures (from start) */
213 struct vm_page *lastpg; /* vm_page structure for end */
214 #ifdef __HAVE_PMAP_PHYSSEG
215 struct pmap_physseg pmseg; /* pmap specific (MD) data */
216 #endif
217 };
218
219 #ifdef _KERNEL
220
221 /*
222 * globals
223 */
224
225 extern boolean_t vm_page_zero_enable;
226
227 /*
228 * Each pageable resident page falls into one of three lists:
229 *
230 * free
231 * Available for allocation now.
232 * inactive
233 * Not referenced in any map, but still has an
234 * object/offset-page mapping, and may be dirty.
235 * This is the list of pages that should be
236 * paged out next.
237 * active
238 * A list of pages which have been placed in
239 * at least one physical map. This list is
240 * ordered, in LRU-like fashion.
241 */
242
243 extern struct pglist vm_page_queue_free; /* memory free queue */
244 extern struct pglist vm_page_queue_active; /* active memory queue */
245 extern struct pglist vm_page_queue_inactive; /* inactive memory queue */
246
247 /*
248 * physical memory config is stored in vm_physmem.
249 */
250
251 extern struct vm_physseg vm_physmem[VM_PHYSSEG_MAX];
252 extern int vm_nphysseg;
253
254 /*
255 * handle inline options
256 */
257
258 #ifdef UVM_PAGE_INLINE
259 #define PAGE_INLINE static __inline
260 #else
261 #define PAGE_INLINE /* nothing */
262 #endif /* UVM_PAGE_INLINE */
263
264 /*
265 * prototypes: the following prototypes define the interface to pages
266 */
267
268 void uvm_page_init(vaddr_t *, vaddr_t *);
269 #if defined(UVM_PAGE_TRKOWN)
270 void uvm_page_own(struct vm_page *, char *);
271 #endif
272 #if !defined(PMAP_STEAL_MEMORY)
273 boolean_t uvm_page_physget(paddr_t *);
274 #endif
275 void uvm_page_rehash(void);
276 void uvm_pageidlezero(void);
277
278 PAGE_INLINE int uvm_lock_fpageq(void);
279 PAGE_INLINE void uvm_unlock_fpageq(int);
280
281 PAGE_INLINE void uvm_pageactivate(struct vm_page *);
282 vaddr_t uvm_pageboot_alloc(vsize_t);
283 PAGE_INLINE void uvm_pagecopy(struct vm_page *, struct vm_page *);
284 PAGE_INLINE void uvm_pagedeactivate(struct vm_page *);
285 void uvm_pagefree(struct vm_page *);
286 void uvm_page_unbusy(struct vm_page **, int);
287 PAGE_INLINE struct vm_page *uvm_pagelookup(struct uvm_object *, voff_t);
288 PAGE_INLINE void uvm_pageunwire(struct vm_page *);
289 PAGE_INLINE void uvm_pagewait(struct vm_page *, int);
290 PAGE_INLINE void uvm_pagewake(struct vm_page *);
291 PAGE_INLINE void uvm_pagewire(struct vm_page *);
292 PAGE_INLINE void uvm_pagezero(struct vm_page *);
293
294 PAGE_INLINE int uvm_page_lookup_freelist(struct vm_page *);
295
296 static struct vm_page *PHYS_TO_VM_PAGE(paddr_t);
297 static int vm_physseg_find(paddr_t, int *);
298
299 /*
300 * macros
301 */
302
303 #define uvm_lock_pageq() simple_lock(&uvm.pageqlock)
304 #define uvm_unlock_pageq() simple_unlock(&uvm.pageqlock)
305
306 #define uvm_pagehash(obj,off) \
307 (((unsigned long)obj+(unsigned long)atop(off)) & uvm.page_hashmask)
308
309 #define UVM_PAGEZERO_TARGET (uvmexp.free)
310
311 #define VM_PAGE_TO_PHYS(entry) ((entry)->phys_addr)
312
313 /*
314 * when VM_PHYSSEG_MAX is 1, we can simplify these functions
315 */
316
317 /*
318 * vm_physseg_find: find vm_physseg structure that belongs to a PA
319 */
320 static __inline int
321 vm_physseg_find(pframe, offp)
322 paddr_t pframe;
323 int *offp;
324 {
325 #if VM_PHYSSEG_MAX == 1
326
327 /* 'contig' case */
328 if (pframe >= vm_physmem[0].start && pframe < vm_physmem[0].end) {
329 if (offp)
330 *offp = pframe - vm_physmem[0].start;
331 return(0);
332 }
333 return(-1);
334
335 #elif (VM_PHYSSEG_STRAT == VM_PSTRAT_BSEARCH)
336 /* binary search for it */
337 int start, len, try;
338
339 /*
340 * if try is too large (thus target is less than than try) we reduce
341 * the length to trunc(len/2) [i.e. everything smaller than "try"]
342 *
343 * if the try is too small (thus target is greater than try) then
344 * we set the new start to be (try + 1). this means we need to
345 * reduce the length to (round(len/2) - 1).
346 *
347 * note "adjust" below which takes advantage of the fact that
348 * (round(len/2) - 1) == trunc((len - 1) / 2)
349 * for any value of len we may have
350 */
351
352 for (start = 0, len = vm_nphysseg ; len != 0 ; len = len / 2) {
353 try = start + (len / 2); /* try in the middle */
354
355 /* start past our try? */
356 if (pframe >= vm_physmem[try].start) {
357 /* was try correct? */
358 if (pframe < vm_physmem[try].end) {
359 if (offp)
360 *offp = pframe - vm_physmem[try].start;
361 return(try); /* got it */
362 }
363 start = try + 1; /* next time, start here */
364 len--; /* "adjust" */
365 } else {
366 /*
367 * pframe before try, just reduce length of
368 * region, done in "for" loop
369 */
370 }
371 }
372 return(-1);
373
374 #else
375 /* linear search for it */
376 int lcv;
377
378 for (lcv = 0; lcv < vm_nphysseg; lcv++) {
379 if (pframe >= vm_physmem[lcv].start &&
380 pframe < vm_physmem[lcv].end) {
381 if (offp)
382 *offp = pframe - vm_physmem[lcv].start;
383 return(lcv); /* got it */
384 }
385 }
386 return(-1);
387
388 #endif
389 }
390
391 /*
392 * PHYS_TO_VM_PAGE: find vm_page for a PA. used by MI code to get vm_pages
393 * back from an I/O mapping (ugh!). used in some MD code as well.
394 */
395 static __inline struct vm_page *
396 PHYS_TO_VM_PAGE(pa)
397 paddr_t pa;
398 {
399 paddr_t pf = atop(pa);
400 int off;
401 int psi;
402
403 psi = vm_physseg_find(pf, &off);
404 if (psi != -1)
405 return(&vm_physmem[psi].pgs[off]);
406 return(NULL);
407 }
408
409 #define VM_PAGE_IS_FREE(entry) ((entry)->pg_flags & PQ_FREE)
410
411 #endif /* _KERNEL */
412
413 #endif /* _UVM_UVM_PAGE_H_ */