Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Newer
Older
100644 476 lines (412 sloc) 11.612 kB
d5e5fd2 @dalingrin Initial Palm source
dalingrin authored
1 /*
2 * High memory handling common code and variables.
3 *
4 * (C) 1999 Andrea Arcangeli, SuSE GmbH, andrea@suse.de
5 * Gerhard Wichert, Siemens AG, Gerhard.Wichert@pdb.siemens.de
6 *
7 *
8 * Redesigned the x86 32-bit VM architecture to deal with
9 * 64-bit physical space. With current x86 CPUs this
10 * means up to 64 Gigabytes physical RAM.
11 *
12 * Rewrote high memory support to move the page cache into
13 * high memory. Implemented permanent (schedulable) kmaps
14 * based on Linus' idea.
15 *
16 * Copyright (C) 1999 Ingo Molnar <mingo@redhat.com>
17 */
18
19 #include <linux/mm.h>
20 #include <linux/module.h>
21 #include <linux/swap.h>
22 #include <linux/bio.h>
23 #include <linux/pagemap.h>
24 #include <linux/mempool.h>
25 #include <linux/blkdev.h>
26 #include <linux/init.h>
27 #include <linux/hash.h>
28 #include <linux/highmem.h>
29 #include <asm/tlbflush.h>
30
31 /*
32 * Virtual_count is not a pure "count".
33 * 0 means that it is not mapped, and has not been mapped
34 * since a TLB flush - it is usable.
35 * 1 means that there are no users, but it has been mapped
36 * since the last TLB flush - so we can't use it.
37 * n means that there are (n-1) current users of it.
38 */
39 #ifdef CONFIG_HIGHMEM
40
41 unsigned long totalhigh_pages __read_mostly;
42 EXPORT_SYMBOL(totalhigh_pages);
43
44 unsigned int nr_free_highpages (void)
45 {
46 pg_data_t *pgdat;
47 unsigned int pages = 0;
48
49 for_each_online_pgdat(pgdat) {
50 pages += zone_page_state(&pgdat->node_zones[ZONE_HIGHMEM],
51 NR_FREE_PAGES);
52 if (zone_movable_is_highmem())
53 pages += zone_page_state(
54 &pgdat->node_zones[ZONE_MOVABLE],
55 NR_FREE_PAGES);
56 }
57
58 return pages;
59 }
60
61 static int pkmap_count[LAST_PKMAP];
62 static unsigned int last_pkmap_nr;
63 static __cacheline_aligned_in_smp DEFINE_SPINLOCK(kmap_lock);
64
65 pte_t * pkmap_page_table;
66
67 static DECLARE_WAIT_QUEUE_HEAD(pkmap_map_wait);
68
69 /*
70 * Most architectures have no use for kmap_high_get(), so let's abstract
71 * the disabling of IRQ out of the locking in that case to save on a
72 * potential useless overhead.
73 */
74 #ifdef ARCH_NEEDS_KMAP_HIGH_GET
75 #define lock_kmap() spin_lock_irq(&kmap_lock)
76 #define unlock_kmap() spin_unlock_irq(&kmap_lock)
77 #define lock_kmap_any(flags) spin_lock_irqsave(&kmap_lock, flags)
78 #define unlock_kmap_any(flags) spin_unlock_irqrestore(&kmap_lock, flags)
79 #else
80 #define lock_kmap() spin_lock(&kmap_lock)
81 #define unlock_kmap() spin_unlock(&kmap_lock)
82 #define lock_kmap_any(flags) \
83 do { spin_lock(&kmap_lock); (void)(flags); } while (0)
84 #define unlock_kmap_any(flags) \
85 do { spin_unlock(&kmap_lock); (void)(flags); } while (0)
86 #endif
87
88 static void flush_all_zero_pkmaps(void)
89 {
90 int i;
91 int need_flush = 0;
92
93 flush_cache_kmaps();
94
95 for (i = 0; i < LAST_PKMAP; i++) {
96 struct page *page;
97
98 /*
99 * zero means we don't have anything to do,
100 * >1 means that it is still in use. Only
101 * a count of 1 means that it is free but
102 * needs to be unmapped
103 */
104 if (pkmap_count[i] != 1)
105 continue;
106 pkmap_count[i] = 0;
107
108 /* sanity check */
109 BUG_ON(pte_none(pkmap_page_table[i]));
110
111 /*
112 * Don't need an atomic fetch-and-clear op here;
113 * no-one has the page mapped, and cannot get at
114 * its virtual address (and hence PTE) without first
115 * getting the kmap_lock (which is held here).
116 * So no dangers, even with speculative execution.
117 */
118 page = pte_page(pkmap_page_table[i]);
119 pte_clear(&init_mm, (unsigned long)page_address(page),
120 &pkmap_page_table[i]);
121
122 set_page_address(page, NULL);
123 need_flush = 1;
124 }
125 if (need_flush)
126 flush_tlb_kernel_range(PKMAP_ADDR(0), PKMAP_ADDR(LAST_PKMAP));
127 }
128
129 /**
130 * kmap_flush_unused - flush all unused kmap mappings in order to remove stray mappings
131 */
132 void kmap_flush_unused(void)
133 {
134 lock_kmap();
135 flush_all_zero_pkmaps();
136 unlock_kmap();
137 }
138
139 static inline unsigned long map_new_virtual(struct page *page)
140 {
141 unsigned long vaddr;
142 int count;
143
144 start:
145 count = LAST_PKMAP;
146 /* Find an empty entry */
147 for (;;) {
148 last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
149 if (!last_pkmap_nr) {
150 flush_all_zero_pkmaps();
151 count = LAST_PKMAP;
152 }
153 if (!pkmap_count[last_pkmap_nr])
154 break; /* Found a usable entry */
155 if (--count)
156 continue;
157
158 /*
159 * Sleep for somebody else to unmap their entries
160 */
161 {
162 DECLARE_WAITQUEUE(wait, current);
163
164 __set_current_state(TASK_UNINTERRUPTIBLE);
165 add_wait_queue(&pkmap_map_wait, &wait);
166 unlock_kmap();
167 schedule();
168 remove_wait_queue(&pkmap_map_wait, &wait);
169 lock_kmap();
170
171 /* Somebody else might have mapped it while we slept */
172 if (page_address(page))
173 return (unsigned long)page_address(page);
174
175 /* Re-start */
176 goto start;
177 }
178 }
179 vaddr = PKMAP_ADDR(last_pkmap_nr);
180 set_pte_at(&init_mm, vaddr,
181 &(pkmap_page_table[last_pkmap_nr]), mk_pte(page, kmap_prot));
182
183 pkmap_count[last_pkmap_nr] = 1;
184 set_page_address(page, (void *)vaddr);
185
186 return vaddr;
187 }
188
189 /**
190 * kmap_high - map a highmem page into memory
191 * @page: &struct page to map
192 *
193 * Returns the page's virtual memory address.
194 *
195 * We cannot call this from interrupts, as it may block.
196 */
197 void *kmap_high(struct page *page)
198 {
199 unsigned long vaddr;
200
201 /*
202 * For highmem pages, we can't trust "virtual" until
203 * after we have the lock.
204 */
205 lock_kmap();
206 vaddr = (unsigned long)page_address(page);
207 if (!vaddr)
208 vaddr = map_new_virtual(page);
209 pkmap_count[PKMAP_NR(vaddr)]++;
210 BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 2);
211 unlock_kmap();
212 return (void*) vaddr;
213 }
214
215 EXPORT_SYMBOL(kmap_high);
216
217 #ifdef ARCH_NEEDS_KMAP_HIGH_GET
218 /**
219 * kmap_high_get - pin a highmem page into memory
220 * @page: &struct page to pin
221 *
222 * Returns the page's current virtual memory address, or NULL if no mapping
223 * exists. If and only if a non null address is returned then a
224 * matching call to kunmap_high() is necessary.
225 *
226 * This can be called from any context.
227 */
228 void *kmap_high_get(struct page *page)
229 {
230 unsigned long vaddr, flags;
231
232 lock_kmap_any(flags);
233 vaddr = (unsigned long)page_address(page);
234 if (vaddr) {
235 BUG_ON(pkmap_count[PKMAP_NR(vaddr)] < 1);
236 pkmap_count[PKMAP_NR(vaddr)]++;
237 }
238 unlock_kmap_any(flags);
239 return (void*) vaddr;
240 }
241 #endif
242
243 /**
244 * kunmap_high - map a highmem page into memory
245 * @page: &struct page to unmap
246 *
247 * If ARCH_NEEDS_KMAP_HIGH_GET is not defined then this may be called
248 * only from user context.
249 */
250 void kunmap_high(struct page *page)
251 {
252 unsigned long vaddr;
253 unsigned long nr;
254 unsigned long flags;
255 int need_wakeup;
256
257 lock_kmap_any(flags);
258 vaddr = (unsigned long)page_address(page);
259 BUG_ON(!vaddr);
260 nr = PKMAP_NR(vaddr);
261
262 /*
263 * A count must never go down to zero
264 * without a TLB flush!
265 */
266 need_wakeup = 0;
267 switch (--pkmap_count[nr]) {
268 case 0:
269 BUG();
270 case 1:
271 /*
272 * Avoid an unnecessary wake_up() function call.
273 * The common case is pkmap_count[] == 1, but
274 * no waiters.
275 * The tasks queued in the wait-queue are guarded
276 * by both the lock in the wait-queue-head and by
277 * the kmap_lock. As the kmap_lock is held here,
278 * no need for the wait-queue-head's lock. Simply
279 * test if the queue is empty.
280 */
281 need_wakeup = waitqueue_active(&pkmap_map_wait);
282 }
283 unlock_kmap_any(flags);
284
285 /* do wake-up, if needed, race-free outside of the spin lock */
286 if (need_wakeup)
287 wake_up(&pkmap_map_wait);
288 }
289
290 EXPORT_SYMBOL(kunmap_high);
291 #endif
292
293 #if defined(HASHED_PAGE_VIRTUAL)
294
295 #define PA_HASH_ORDER 7
296
297 /*
298 * Describes one page->virtual association
299 */
300 struct page_address_map {
301 struct page *page;
302 void *virtual;
303 struct list_head list;
304 };
305
306 /*
307 * page_address_map freelist, allocated from page_address_maps.
308 */
309 static struct list_head page_address_pool; /* freelist */
310 static spinlock_t pool_lock; /* protects page_address_pool */
311
312 /*
313 * Hash table bucket
314 */
315 static struct page_address_slot {
316 struct list_head lh; /* List of page_address_maps */
317 spinlock_t lock; /* Protect this bucket's list */
318 } ____cacheline_aligned_in_smp page_address_htable[1<<PA_HASH_ORDER];
319
320 static struct page_address_slot *page_slot(struct page *page)
321 {
322 return &page_address_htable[hash_ptr(page, PA_HASH_ORDER)];
323 }
324
325 /**
326 * page_address - get the mapped virtual address of a page
327 * @page: &struct page to get the virtual address of
328 *
329 * Returns the page's virtual address.
330 */
331 void *page_address(struct page *page)
332 {
333 unsigned long flags;
334 void *ret;
335 struct page_address_slot *pas;
336
337 if (!PageHighMem(page))
338 return lowmem_page_address(page);
339
340 pas = page_slot(page);
341 ret = NULL;
342 spin_lock_irqsave(&pas->lock, flags);
343 if (!list_empty(&pas->lh)) {
344 struct page_address_map *pam;
345
346 list_for_each_entry(pam, &pas->lh, list) {
347 if (pam->page == page) {
348 ret = pam->virtual;
349 goto done;
350 }
351 }
352 }
353 done:
354 spin_unlock_irqrestore(&pas->lock, flags);
355 return ret;
356 }
357
358 EXPORT_SYMBOL(page_address);
359
360 /**
361 * set_page_address - set a page's virtual address
362 * @page: &struct page to set
363 * @virtual: virtual address to use
364 */
365 void set_page_address(struct page *page, void *virtual)
366 {
367 unsigned long flags;
368 struct page_address_slot *pas;
369 struct page_address_map *pam;
370
371 BUG_ON(!PageHighMem(page));
372
373 pas = page_slot(page);
374 if (virtual) { /* Add */
375 BUG_ON(list_empty(&page_address_pool));
376
377 spin_lock_irqsave(&pool_lock, flags);
378 pam = list_entry(page_address_pool.next,
379 struct page_address_map, list);
380 list_del(&pam->list);
381 spin_unlock_irqrestore(&pool_lock, flags);
382
383 pam->page = page;
384 pam->virtual = virtual;
385
386 spin_lock_irqsave(&pas->lock, flags);
387 list_add_tail(&pam->list, &pas->lh);
388 spin_unlock_irqrestore(&pas->lock, flags);
389 } else { /* Remove */
390 spin_lock_irqsave(&pas->lock, flags);
391 list_for_each_entry(pam, &pas->lh, list) {
392 if (pam->page == page) {
393 list_del(&pam->list);
394 spin_unlock_irqrestore(&pas->lock, flags);
395 spin_lock_irqsave(&pool_lock, flags);
396 list_add_tail(&pam->list, &page_address_pool);
397 spin_unlock_irqrestore(&pool_lock, flags);
398 goto done;
399 }
400 }
401 spin_unlock_irqrestore(&pas->lock, flags);
402 }
403 done:
404 return;
405 }
406
407 static struct page_address_map page_address_maps[LAST_PKMAP];
408
409 void __init page_address_init(void)
410 {
411 int i;
412
413 INIT_LIST_HEAD(&page_address_pool);
414 for (i = 0; i < ARRAY_SIZE(page_address_maps); i++)
415 list_add(&page_address_maps[i].list, &page_address_pool);
416 for (i = 0; i < ARRAY_SIZE(page_address_htable); i++) {
417 INIT_LIST_HEAD(&page_address_htable[i].lh);
418 spin_lock_init(&page_address_htable[i].lock);
419 }
420 spin_lock_init(&pool_lock);
421 }
422
423 #endif /* defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) */
424
425 #ifdef CONFIG_DEBUG_HIGHMEM
426
427 void debug_kmap_atomic(enum km_type type)
428 {
429 static int warn_count = 10;
430
431 if (unlikely(warn_count < 0))
432 return;
433
434 if (unlikely(in_interrupt())) {
435 if (in_nmi()) {
436 if (type != KM_NMI && type != KM_NMI_PTE) {
437 WARN_ON(1);
438 warn_count--;
439 }
440 } else if (in_irq()) {
441 if (type != KM_IRQ0 && type != KM_IRQ1 &&
442 type != KM_BIO_SRC_IRQ && type != KM_BIO_DST_IRQ &&
443 type != KM_BOUNCE_READ && type != KM_IRQ_PTE) {
444 WARN_ON(1);
445 warn_count--;
446 }
447 } else if (!irqs_disabled()) { /* softirq */
448 if (type != KM_IRQ0 && type != KM_IRQ1 &&
449 type != KM_SOFTIRQ0 && type != KM_SOFTIRQ1 &&
450 type != KM_SKB_SUNRPC_DATA &&
451 type != KM_SKB_DATA_SOFTIRQ &&
452 type != KM_BOUNCE_READ) {
453 WARN_ON(1);
454 warn_count--;
455 }
456 }
457 }
458
459 if (type == KM_IRQ0 || type == KM_IRQ1 || type == KM_BOUNCE_READ ||
460 type == KM_BIO_SRC_IRQ || type == KM_BIO_DST_IRQ ||
461 type == KM_IRQ_PTE || type == KM_NMI ||
462 type == KM_NMI_PTE ) {
463 if (!irqs_disabled()) {
464 WARN_ON(1);
465 warn_count--;
466 }
467 } else if (type == KM_SOFTIRQ0 || type == KM_SOFTIRQ1) {
468 if (irq_count() == 0 && !irqs_disabled()) {
469 WARN_ON(1);
470 warn_count--;
471 }
472 }
473 }
474
475 #endif
Something went wrong with that request. Please try again.