Blame view
mm/huge_memory.c
76.1 KB
71e3aac07 thp: transparent ... |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
/* * Copyright (C) 2009 Red Hat, Inc. * * This work is licensed under the terms of the GNU GPL, version 2. See * the COPYING file in the top-level directory. */ #include <linux/mm.h> #include <linux/sched.h> #include <linux/highmem.h> #include <linux/hugetlb.h> #include <linux/mmu_notifier.h> #include <linux/rmap.h> #include <linux/swap.h> |
97ae17497 thp: implement re... |
15 |
#include <linux/shrinker.h> |
ba76149f4 thp: khugepaged |
16 17 18 |
#include <linux/mm_inline.h> #include <linux/kthread.h> #include <linux/khugepaged.h> |
878aee7d6 thp: freeze khuge... |
19 |
#include <linux/freezer.h> |
a664b2d85 thp: madvise(MADV... |
20 |
#include <linux/mman.h> |
325adeb55 mm: huge_memory: ... |
21 |
#include <linux/pagemap.h> |
4daae3b4b mm: mempolicy: Us... |
22 |
#include <linux/migrate.h> |
43b5fbbd2 mm/huge_memory.c:... |
23 |
#include <linux/hashtable.h> |
97ae17497 thp: implement re... |
24 |
|
71e3aac07 thp: transparent ... |
25 26 27 |
#include <asm/tlb.h> #include <asm/pgalloc.h> #include "internal.h" |
ba76149f4 thp: khugepaged |
28 |
/* |
8bfa3f9a0 mm/huge_memory.c:... |
29 30 31 32 33 34 |
* By default transparent hugepage support is disabled in order that avoid * to risk increase the memory footprint of applications without a guaranteed * benefit. When transparent hugepage support is enabled, is for all mappings, * and khugepaged scans all mappings. * Defrag is invoked by khugepaged hugepage allocations and by page faults * for all hugepage allocations. |
ba76149f4 thp: khugepaged |
35 |
*/ |
71e3aac07 thp: transparent ... |
36 |
unsigned long transparent_hugepage_flags __read_mostly = |
13ece886d thp: transparent ... |
37 |
#ifdef CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS |
ba76149f4 thp: khugepaged |
38 |
(1<<TRANSPARENT_HUGEPAGE_FLAG)| |
13ece886d thp: transparent ... |
39 40 41 42 |
#endif #ifdef CONFIG_TRANSPARENT_HUGEPAGE_MADVISE (1<<TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG)| #endif |
d39d33c33 thp: enable direc... |
43 |
(1<<TRANSPARENT_HUGEPAGE_DEFRAG_FLAG)| |
79da5407e thp: introduce sy... |
44 45 |
(1<<TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG)| (1<<TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG); |
ba76149f4 thp: khugepaged |
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
/* default scan 8*512 pte (or vmas) every 30 second */ static unsigned int khugepaged_pages_to_scan __read_mostly = HPAGE_PMD_NR*8; static unsigned int khugepaged_pages_collapsed; static unsigned int khugepaged_full_scans; static unsigned int khugepaged_scan_sleep_millisecs __read_mostly = 10000; /* during fragmentation poll the hugepage allocator once every minute */ static unsigned int khugepaged_alloc_sleep_millisecs __read_mostly = 60000; static struct task_struct *khugepaged_thread __read_mostly; static DEFINE_MUTEX(khugepaged_mutex); static DEFINE_SPINLOCK(khugepaged_mm_lock); static DECLARE_WAIT_QUEUE_HEAD(khugepaged_wait); /* * default collapse hugepages if there is at least one pte mapped like * it would have happened if the vma was large enough during page * fault. */ static unsigned int khugepaged_max_ptes_none __read_mostly = HPAGE_PMD_NR-1; static int khugepaged(void *none); |
ba76149f4 thp: khugepaged |
66 |
static int khugepaged_slab_init(void); |
ba76149f4 thp: khugepaged |
67 |
|
43b5fbbd2 mm/huge_memory.c:... |
68 69 |
#define MM_SLOTS_HASH_BITS 10 static __read_mostly DEFINE_HASHTABLE(mm_slots_hash, MM_SLOTS_HASH_BITS); |
ba76149f4 thp: khugepaged |
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 |
static struct kmem_cache *mm_slot_cache __read_mostly; /** * struct mm_slot - hash lookup from mm to mm_slot * @hash: hash collision list * @mm_node: khugepaged scan list headed in khugepaged_scan.mm_head * @mm: the mm that this information is valid for */ struct mm_slot { struct hlist_node hash; struct list_head mm_node; struct mm_struct *mm; }; /** * struct khugepaged_scan - cursor for scanning * @mm_head: the head of the mm list to scan * @mm_slot: the current mm_slot we are scanning * @address: the next address inside that to be scanned * * There is only the one khugepaged_scan instance of this cursor structure. */ struct khugepaged_scan { struct list_head mm_head; struct mm_slot *mm_slot; unsigned long address; |
2f1da6421 mm/huge_memory.c:... |
96 97 |
}; static struct khugepaged_scan khugepaged_scan = { |
ba76149f4 thp: khugepaged |
98 99 |
.mm_head = LIST_HEAD_INIT(khugepaged_scan.mm_head), }; |
f000565ad thp: set recommen... |
100 101 102 103 104 105 |
static int set_recommended_min_free_kbytes(void) { struct zone *zone; int nr_zones = 0; unsigned long recommended_min; |
f000565ad thp: set recommen... |
106 |
|
17c230afa thp: use khugepag... |
107 |
if (!khugepaged_enabled()) |
f000565ad thp: set recommen... |
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 |
return 0; for_each_populated_zone(zone) nr_zones++; /* Make sure at least 2 hugepages are free for MIGRATE_RESERVE */ recommended_min = pageblock_nr_pages * nr_zones * 2; /* * Make sure that on average at least two pageblocks are almost free * of another type, one for a migratetype to fall back to and a * second to avoid subsequent fallbacks of other types There are 3 * MIGRATE_TYPES we care about. */ recommended_min += pageblock_nr_pages * nr_zones * MIGRATE_PCPTYPES * MIGRATE_PCPTYPES; /* don't ever allow to reserve more than 5% of the lowmem */ recommended_min = min(recommended_min, (unsigned long) nr_free_buffer_pages() / 20); recommended_min <<= (PAGE_SHIFT-10); |
42aa83cb6 mm: show message ... |
129 130 131 132 133 134 |
if (recommended_min > min_free_kbytes) { if (user_min_free_kbytes >= 0) pr_info("raising min_free_kbytes from %d to %lu " "to help transparent hugepage allocations ", min_free_kbytes, recommended_min); |
f000565ad thp: set recommen... |
135 |
min_free_kbytes = recommended_min; |
42aa83cb6 mm: show message ... |
136 |
} |
f000565ad thp: set recommen... |
137 138 139 140 |
setup_per_zone_wmarks(); return 0; } late_initcall(set_recommended_min_free_kbytes); |
ba76149f4 thp: khugepaged |
141 142 143 144 |
static int start_khugepaged(void) { int err = 0; if (khugepaged_enabled()) { |
ba76149f4 thp: khugepaged |
145 146 147 148 149 150 151 152 153 154 |
if (!khugepaged_thread) khugepaged_thread = kthread_run(khugepaged, NULL, "khugepaged"); if (unlikely(IS_ERR(khugepaged_thread))) { printk(KERN_ERR "khugepaged: kthread_run(khugepaged) failed "); err = PTR_ERR(khugepaged_thread); khugepaged_thread = NULL; } |
911891afe thp: move khugepa... |
155 156 |
if (!list_empty(&khugepaged_scan.mm_head)) |
ba76149f4 thp: khugepaged |
157 |
wake_up_interruptible(&khugepaged_wait); |
f000565ad thp: set recommen... |
158 159 |
set_recommended_min_free_kbytes(); |
911891afe thp: move khugepa... |
160 |
} else if (khugepaged_thread) { |
911891afe thp: move khugepa... |
161 162 163 |
kthread_stop(khugepaged_thread); khugepaged_thread = NULL; } |
637e3a27e thp: remove unnec... |
164 |
|
ba76149f4 thp: khugepaged |
165 166 |
return err; } |
71e3aac07 thp: transparent ... |
167 |
|
97ae17497 thp: implement re... |
168 |
static atomic_t huge_zero_refcount; |
5918d10a4 thp: fix huge zer... |
169 |
static struct page *huge_zero_page __read_mostly; |
97ae17497 thp: implement re... |
170 |
|
5918d10a4 thp: fix huge zer... |
171 |
static inline bool is_huge_zero_page(struct page *page) |
4a6c12972 thp: huge zero pa... |
172 |
{ |
5918d10a4 thp: fix huge zer... |
173 |
return ACCESS_ONCE(huge_zero_page) == page; |
97ae17497 thp: implement re... |
174 |
} |
4a6c12972 thp: huge zero pa... |
175 |
|
97ae17497 thp: implement re... |
176 177 |
static inline bool is_huge_zero_pmd(pmd_t pmd) { |
5918d10a4 thp: fix huge zer... |
178 |
return is_huge_zero_page(pmd_page(pmd)); |
97ae17497 thp: implement re... |
179 |
} |
5918d10a4 thp: fix huge zer... |
180 |
static struct page *get_huge_zero_page(void) |
97ae17497 thp: implement re... |
181 182 183 184 |
{ struct page *zero_page; retry: if (likely(atomic_inc_not_zero(&huge_zero_refcount))) |
5918d10a4 thp: fix huge zer... |
185 |
return ACCESS_ONCE(huge_zero_page); |
97ae17497 thp: implement re... |
186 187 |
zero_page = alloc_pages((GFP_TRANSHUGE | __GFP_ZERO) & ~__GFP_MOVABLE, |
4a6c12972 thp: huge zero pa... |
188 |
HPAGE_PMD_ORDER); |
d8a8e1f0d thp, vmstat: impl... |
189 190 |
if (!zero_page) { count_vm_event(THP_ZERO_PAGE_ALLOC_FAILED); |
5918d10a4 thp: fix huge zer... |
191 |
return NULL; |
d8a8e1f0d thp, vmstat: impl... |
192 193 |
} count_vm_event(THP_ZERO_PAGE_ALLOC); |
97ae17497 thp: implement re... |
194 |
preempt_disable(); |
5918d10a4 thp: fix huge zer... |
195 |
if (cmpxchg(&huge_zero_page, NULL, zero_page)) { |
97ae17497 thp: implement re... |
196 197 198 199 200 201 202 203 |
preempt_enable(); __free_page(zero_page); goto retry; } /* We take additional reference here. It will be put back by shrinker */ atomic_set(&huge_zero_refcount, 2); preempt_enable(); |
5918d10a4 thp: fix huge zer... |
204 |
return ACCESS_ONCE(huge_zero_page); |
4a6c12972 thp: huge zero pa... |
205 |
} |
97ae17497 thp: implement re... |
206 |
static void put_huge_zero_page(void) |
4a6c12972 thp: huge zero pa... |
207 |
{ |
97ae17497 thp: implement re... |
208 209 210 211 212 |
/* * Counter should never go to zero here. Only shrinker can put * last reference. */ BUG_ON(atomic_dec_and_test(&huge_zero_refcount)); |
4a6c12972 thp: huge zero pa... |
213 |
} |
488964666 hugepage: convert... |
214 215 |
static unsigned long shrink_huge_zero_page_count(struct shrinker *shrink, struct shrink_control *sc) |
4a6c12972 thp: huge zero pa... |
216 |
{ |
488964666 hugepage: convert... |
217 218 219 |
/* we can free zero page only if last reference remains */ return atomic_read(&huge_zero_refcount) == 1 ? HPAGE_PMD_NR : 0; } |
97ae17497 thp: implement re... |
220 |
|
488964666 hugepage: convert... |
221 222 223 |
static unsigned long shrink_huge_zero_page_scan(struct shrinker *shrink, struct shrink_control *sc) { |
97ae17497 thp: implement re... |
224 |
if (atomic_cmpxchg(&huge_zero_refcount, 1, 0) == 1) { |
5918d10a4 thp: fix huge zer... |
225 226 227 |
struct page *zero_page = xchg(&huge_zero_page, NULL); BUG_ON(zero_page == NULL); __free_page(zero_page); |
488964666 hugepage: convert... |
228 |
return HPAGE_PMD_NR; |
97ae17497 thp: implement re... |
229 230 231 |
} return 0; |
4a6c12972 thp: huge zero pa... |
232 |
} |
97ae17497 thp: implement re... |
233 |
static struct shrinker huge_zero_page_shrinker = { |
488964666 hugepage: convert... |
234 235 |
.count_objects = shrink_huge_zero_page_count, .scan_objects = shrink_huge_zero_page_scan, |
97ae17497 thp: implement re... |
236 237 |
.seeks = DEFAULT_SEEKS, }; |
71e3aac07 thp: transparent ... |
238 |
#ifdef CONFIG_SYSFS |
ba76149f4 thp: khugepaged |
239 |
|
71e3aac07 thp: transparent ... |
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 |
static ssize_t double_flag_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf, enum transparent_hugepage_flag enabled, enum transparent_hugepage_flag req_madv) { if (test_bit(enabled, &transparent_hugepage_flags)) { VM_BUG_ON(test_bit(req_madv, &transparent_hugepage_flags)); return sprintf(buf, "[always] madvise never "); } else if (test_bit(req_madv, &transparent_hugepage_flags)) return sprintf(buf, "always [madvise] never "); else return sprintf(buf, "always madvise [never] "); } static ssize_t double_flag_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count, enum transparent_hugepage_flag enabled, enum transparent_hugepage_flag req_madv) { if (!memcmp("always", buf, min(sizeof("always")-1, count))) { set_bit(enabled, &transparent_hugepage_flags); clear_bit(req_madv, &transparent_hugepage_flags); } else if (!memcmp("madvise", buf, min(sizeof("madvise")-1, count))) { clear_bit(enabled, &transparent_hugepage_flags); set_bit(req_madv, &transparent_hugepage_flags); } else if (!memcmp("never", buf, min(sizeof("never")-1, count))) { clear_bit(enabled, &transparent_hugepage_flags); clear_bit(req_madv, &transparent_hugepage_flags); } else return -EINVAL; return count; } static ssize_t enabled_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return double_flag_show(kobj, attr, buf, TRANSPARENT_HUGEPAGE_FLAG, TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG); } static ssize_t enabled_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { |
ba76149f4 thp: khugepaged |
291 292 293 294 295 296 297 |
ssize_t ret; ret = double_flag_store(kobj, attr, buf, count, TRANSPARENT_HUGEPAGE_FLAG, TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG); if (ret > 0) { |
911891afe thp: move khugepa... |
298 299 300 301 302 |
int err; mutex_lock(&khugepaged_mutex); err = start_khugepaged(); mutex_unlock(&khugepaged_mutex); |
ba76149f4 thp: khugepaged |
303 304 305 306 307 |
if (err) ret = err; } return ret; |
71e3aac07 thp: transparent ... |
308 309 310 311 312 313 314 315 |
} static struct kobj_attribute enabled_attr = __ATTR(enabled, 0644, enabled_show, enabled_store); static ssize_t single_flag_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf, enum transparent_hugepage_flag flag) { |
e27e6151b mm/thp: use conve... |
316 317 318 |
return sprintf(buf, "%d ", !!test_bit(flag, &transparent_hugepage_flags)); |
71e3aac07 thp: transparent ... |
319 |
} |
e27e6151b mm/thp: use conve... |
320 |
|
71e3aac07 thp: transparent ... |
321 322 323 324 325 |
static ssize_t single_flag_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count, enum transparent_hugepage_flag flag) { |
e27e6151b mm/thp: use conve... |
326 327 328 329 330 331 332 333 334 335 |
unsigned long value; int ret; ret = kstrtoul(buf, 10, &value); if (ret < 0) return ret; if (value > 1) return -EINVAL; if (value) |
71e3aac07 thp: transparent ... |
336 |
set_bit(flag, &transparent_hugepage_flags); |
e27e6151b mm/thp: use conve... |
337 |
else |
71e3aac07 thp: transparent ... |
338 |
clear_bit(flag, &transparent_hugepage_flags); |
71e3aac07 thp: transparent ... |
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 |
return count; } /* * Currently defrag only disables __GFP_NOWAIT for allocation. A blind * __GFP_REPEAT is too aggressive, it's never worth swapping tons of * memory just to allocate one more hugepage. */ static ssize_t defrag_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return double_flag_show(kobj, attr, buf, TRANSPARENT_HUGEPAGE_DEFRAG_FLAG, TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG); } static ssize_t defrag_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { return double_flag_store(kobj, attr, buf, count, TRANSPARENT_HUGEPAGE_DEFRAG_FLAG, TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG); } static struct kobj_attribute defrag_attr = __ATTR(defrag, 0644, defrag_show, defrag_store); |
79da5407e thp: introduce sy... |
365 366 367 368 369 370 371 372 373 374 375 376 377 378 |
static ssize_t use_zero_page_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return single_flag_show(kobj, attr, buf, TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG); } static ssize_t use_zero_page_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { return single_flag_store(kobj, attr, buf, count, TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG); } static struct kobj_attribute use_zero_page_attr = __ATTR(use_zero_page, 0644, use_zero_page_show, use_zero_page_store); |
71e3aac07 thp: transparent ... |
379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 |
#ifdef CONFIG_DEBUG_VM static ssize_t debug_cow_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return single_flag_show(kobj, attr, buf, TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG); } static ssize_t debug_cow_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { return single_flag_store(kobj, attr, buf, count, TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG); } static struct kobj_attribute debug_cow_attr = __ATTR(debug_cow, 0644, debug_cow_show, debug_cow_store); #endif /* CONFIG_DEBUG_VM */ static struct attribute *hugepage_attr[] = { &enabled_attr.attr, &defrag_attr.attr, |
79da5407e thp: introduce sy... |
400 |
&use_zero_page_attr.attr, |
71e3aac07 thp: transparent ... |
401 402 403 404 405 406 407 408 |
#ifdef CONFIG_DEBUG_VM &debug_cow_attr.attr, #endif NULL, }; static struct attribute_group hugepage_attr_group = { .attrs = hugepage_attr, |
ba76149f4 thp: khugepaged |
409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 |
}; static ssize_t scan_sleep_millisecs_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_scan_sleep_millisecs); } static ssize_t scan_sleep_millisecs_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { unsigned long msecs; int err; |
3dbb95f78 mm: replace stric... |
425 |
err = kstrtoul(buf, 10, &msecs); |
ba76149f4 thp: khugepaged |
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 |
if (err || msecs > UINT_MAX) return -EINVAL; khugepaged_scan_sleep_millisecs = msecs; wake_up_interruptible(&khugepaged_wait); return count; } static struct kobj_attribute scan_sleep_millisecs_attr = __ATTR(scan_sleep_millisecs, 0644, scan_sleep_millisecs_show, scan_sleep_millisecs_store); static ssize_t alloc_sleep_millisecs_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_alloc_sleep_millisecs); } static ssize_t alloc_sleep_millisecs_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { unsigned long msecs; int err; |
3dbb95f78 mm: replace stric... |
452 |
err = kstrtoul(buf, 10, &msecs); |
ba76149f4 thp: khugepaged |
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 |
if (err || msecs > UINT_MAX) return -EINVAL; khugepaged_alloc_sleep_millisecs = msecs; wake_up_interruptible(&khugepaged_wait); return count; } static struct kobj_attribute alloc_sleep_millisecs_attr = __ATTR(alloc_sleep_millisecs, 0644, alloc_sleep_millisecs_show, alloc_sleep_millisecs_store); static ssize_t pages_to_scan_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_pages_to_scan); } static ssize_t pages_to_scan_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { int err; unsigned long pages; |
3dbb95f78 mm: replace stric... |
478 |
err = kstrtoul(buf, 10, &pages); |
ba76149f4 thp: khugepaged |
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 |
if (err || !pages || pages > UINT_MAX) return -EINVAL; khugepaged_pages_to_scan = pages; return count; } static struct kobj_attribute pages_to_scan_attr = __ATTR(pages_to_scan, 0644, pages_to_scan_show, pages_to_scan_store); static ssize_t pages_collapsed_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_pages_collapsed); } static struct kobj_attribute pages_collapsed_attr = __ATTR_RO(pages_collapsed); static ssize_t full_scans_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_full_scans); } static struct kobj_attribute full_scans_attr = __ATTR_RO(full_scans); static ssize_t khugepaged_defrag_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return single_flag_show(kobj, attr, buf, TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG); } static ssize_t khugepaged_defrag_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { return single_flag_store(kobj, attr, buf, count, TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG); } static struct kobj_attribute khugepaged_defrag_attr = __ATTR(defrag, 0644, khugepaged_defrag_show, khugepaged_defrag_store); /* * max_ptes_none controls if khugepaged should collapse hugepages over * any unmapped ptes in turn potentially increasing the memory * footprint of the vmas. When max_ptes_none is 0 khugepaged will not * reduce the available free memory in the system as it * runs. Increasing max_ptes_none will instead potentially reduce the * free memory in the system during the khugepaged scan. */ static ssize_t khugepaged_max_ptes_none_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%u ", khugepaged_max_ptes_none); } static ssize_t khugepaged_max_ptes_none_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { int err; unsigned long max_ptes_none; |
3dbb95f78 mm: replace stric... |
548 |
err = kstrtoul(buf, 10, &max_ptes_none); |
ba76149f4 thp: khugepaged |
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 |
if (err || max_ptes_none > HPAGE_PMD_NR-1) return -EINVAL; khugepaged_max_ptes_none = max_ptes_none; return count; } static struct kobj_attribute khugepaged_max_ptes_none_attr = __ATTR(max_ptes_none, 0644, khugepaged_max_ptes_none_show, khugepaged_max_ptes_none_store); static struct attribute *khugepaged_attr[] = { &khugepaged_defrag_attr.attr, &khugepaged_max_ptes_none_attr.attr, &pages_to_scan_attr.attr, &pages_collapsed_attr.attr, &full_scans_attr.attr, &scan_sleep_millisecs_attr.attr, &alloc_sleep_millisecs_attr.attr, NULL, }; static struct attribute_group khugepaged_attr_group = { .attrs = khugepaged_attr, .name = "khugepaged", |
71e3aac07 thp: transparent ... |
574 |
}; |
71e3aac07 thp: transparent ... |
575 |
|
569e55900 thp: improve the ... |
576 |
static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj) |
71e3aac07 thp: transparent ... |
577 |
{ |
71e3aac07 thp: transparent ... |
578 |
int err; |
569e55900 thp: improve the ... |
579 580 |
*hugepage_kobj = kobject_create_and_add("transparent_hugepage", mm_kobj); if (unlikely(!*hugepage_kobj)) { |
2c79737af mm: clean up tran... |
581 582 |
printk(KERN_ERR "hugepage: failed to create transparent hugepage kobject "); |
569e55900 thp: improve the ... |
583 |
return -ENOMEM; |
ba76149f4 thp: khugepaged |
584 |
} |
569e55900 thp: improve the ... |
585 |
err = sysfs_create_group(*hugepage_kobj, &hugepage_attr_group); |
ba76149f4 thp: khugepaged |
586 |
if (err) { |
2c79737af mm: clean up tran... |
587 588 |
printk(KERN_ERR "hugepage: failed to register transparent hugepage group "); |
569e55900 thp: improve the ... |
589 |
goto delete_obj; |
ba76149f4 thp: khugepaged |
590 |
} |
569e55900 thp: improve the ... |
591 |
err = sysfs_create_group(*hugepage_kobj, &khugepaged_attr_group); |
ba76149f4 thp: khugepaged |
592 |
if (err) { |
2c79737af mm: clean up tran... |
593 594 |
printk(KERN_ERR "hugepage: failed to register transparent hugepage group "); |
569e55900 thp: improve the ... |
595 |
goto remove_hp_group; |
ba76149f4 thp: khugepaged |
596 |
} |
569e55900 thp: improve the ... |
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 |
return 0; remove_hp_group: sysfs_remove_group(*hugepage_kobj, &hugepage_attr_group); delete_obj: kobject_put(*hugepage_kobj); return err; } static void __init hugepage_exit_sysfs(struct kobject *hugepage_kobj) { sysfs_remove_group(hugepage_kobj, &khugepaged_attr_group); sysfs_remove_group(hugepage_kobj, &hugepage_attr_group); kobject_put(hugepage_kobj); } #else static inline int hugepage_init_sysfs(struct kobject **hugepage_kobj) { return 0; } static inline void hugepage_exit_sysfs(struct kobject *hugepage_kobj) { } #endif /* CONFIG_SYSFS */ static int __init hugepage_init(void) { int err; struct kobject *hugepage_kobj; if (!has_transparent_hugepage()) { transparent_hugepage_flags = 0; return -EINVAL; } err = hugepage_init_sysfs(&hugepage_kobj); if (err) return err; |
ba76149f4 thp: khugepaged |
637 638 639 640 |
err = khugepaged_slab_init(); if (err) goto out; |
97ae17497 thp: implement re... |
641 |
register_shrinker(&huge_zero_page_shrinker); |
97562cd24 thp: disable tran... |
642 643 644 645 646 647 648 |
/* * By default disable transparent hugepages on smaller systems, * where the extra memory used could hurt more than TLB overhead * is likely to save. The admin can still enable it through /sys. */ if (totalram_pages < (512 << (20 - PAGE_SHIFT))) transparent_hugepage_flags = 0; |
ba76149f4 thp: khugepaged |
649 |
start_khugepaged(); |
569e55900 thp: improve the ... |
650 |
return 0; |
ba76149f4 thp: khugepaged |
651 |
out: |
569e55900 thp: improve the ... |
652 |
hugepage_exit_sysfs(hugepage_kobj); |
ba76149f4 thp: khugepaged |
653 |
return err; |
71e3aac07 thp: transparent ... |
654 |
} |
a64fb3cd6 mm: audit/fix non... |
655 |
subsys_initcall(hugepage_init); |
71e3aac07 thp: transparent ... |
656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 |
static int __init setup_transparent_hugepage(char *str) { int ret = 0; if (!str) goto out; if (!strcmp(str, "always")) { set_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); ret = 1; } else if (!strcmp(str, "madvise")) { clear_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); ret = 1; } else if (!strcmp(str, "never")) { clear_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); ret = 1; } out: if (!ret) printk(KERN_WARNING "transparent_hugepage= cannot parse, ignored "); return ret; } __setup("transparent_hugepage=", setup_transparent_hugepage); |
b32967ff1 mm: numa: Add THP... |
689 |
pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma) |
71e3aac07 thp: transparent ... |
690 691 692 693 694 |
{ if (likely(vma->vm_flags & VM_WRITE)) pmd = pmd_mkwrite(pmd); return pmd; } |
3122359a6 thp: move maybe_p... |
695 |
static inline pmd_t mk_huge_pmd(struct page *page, pgprot_t prot) |
b3092b3b7 thp: cleanup: int... |
696 697 |
{ pmd_t entry; |
3122359a6 thp: move maybe_p... |
698 |
entry = mk_pmd(page, prot); |
b3092b3b7 thp: cleanup: int... |
699 700 701 |
entry = pmd_mkhuge(entry); return entry; } |
71e3aac07 thp: transparent ... |
702 703 704 705 706 |
static int __do_huge_pmd_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long haddr, pmd_t *pmd, struct page *page) { |
71e3aac07 thp: transparent ... |
707 |
pgtable_t pgtable; |
c4088ebdc mm: convert the r... |
708 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
709 |
|
309381fea mm: dump page whe... |
710 |
VM_BUG_ON_PAGE(!PageCompound(page), page); |
71e3aac07 thp: transparent ... |
711 |
pgtable = pte_alloc_one(mm, haddr); |
edad9d2c3 mm, thp: allow fa... |
712 |
if (unlikely(!pgtable)) |
71e3aac07 thp: transparent ... |
713 |
return VM_FAULT_OOM; |
71e3aac07 thp: transparent ... |
714 715 |
clear_huge_page(page, haddr, HPAGE_PMD_NR); |
52f37629f THP: fix comment ... |
716 717 718 719 720 |
/* * The memory barrier inside __SetPageUptodate makes sure that * clear_huge_page writes become visible before the set_pmd_at() * write. */ |
71e3aac07 thp: transparent ... |
721 |
__SetPageUptodate(page); |
c4088ebdc mm: convert the r... |
722 |
ptl = pmd_lock(mm, pmd); |
71e3aac07 thp: transparent ... |
723 |
if (unlikely(!pmd_none(*pmd))) { |
c4088ebdc mm: convert the r... |
724 |
spin_unlock(ptl); |
b9bbfbe30 thp: memcg huge m... |
725 |
mem_cgroup_uncharge_page(page); |
71e3aac07 thp: transparent ... |
726 727 728 729 |
put_page(page); pte_free(mm, pgtable); } else { pmd_t entry; |
3122359a6 thp: move maybe_p... |
730 731 |
entry = mk_huge_pmd(page, vma->vm_page_prot); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); |
71e3aac07 thp: transparent ... |
732 |
page_add_new_anon_rmap(page, vma, haddr); |
6b0b50b06 mm/THP: add pmd a... |
733 |
pgtable_trans_huge_deposit(mm, pmd, pgtable); |
71e3aac07 thp: transparent ... |
734 |
set_pmd_at(mm, haddr, pmd, entry); |
71e3aac07 thp: transparent ... |
735 |
add_mm_counter(mm, MM_ANONPAGES, HPAGE_PMD_NR); |
e1f56c89b mm: convert mm->n... |
736 |
atomic_long_inc(&mm->nr_ptes); |
c4088ebdc mm: convert the r... |
737 |
spin_unlock(ptl); |
71e3aac07 thp: transparent ... |
738 |
} |
aa2e878ef mm, thp: remove u... |
739 |
return 0; |
71e3aac07 thp: transparent ... |
740 |
} |
cc5d462f7 mm: use __GFP_OTH... |
741 |
static inline gfp_t alloc_hugepage_gfpmask(int defrag, gfp_t extra_gfp) |
0bbbc0b33 thp: add numa awa... |
742 |
{ |
cc5d462f7 mm: use __GFP_OTH... |
743 |
return (GFP_TRANSHUGE & ~(defrag ? 0 : __GFP_WAIT)) | extra_gfp; |
0bbbc0b33 thp: add numa awa... |
744 745 746 747 |
} static inline struct page *alloc_hugepage_vma(int defrag, struct vm_area_struct *vma, |
cc5d462f7 mm: use __GFP_OTH... |
748 749 |
unsigned long haddr, int nd, gfp_t extra_gfp) |
0bbbc0b33 thp: add numa awa... |
750 |
{ |
cc5d462f7 mm: use __GFP_OTH... |
751 |
return alloc_pages_vma(alloc_hugepage_gfpmask(defrag, extra_gfp), |
5c4b4be3b mm: use correct n... |
752 |
HPAGE_PMD_ORDER, vma, haddr, nd); |
0bbbc0b33 thp: add numa awa... |
753 |
} |
c4088ebdc mm: convert the r... |
754 |
/* Caller must hold page table lock. */ |
3ea41e621 thp: avoid race o... |
755 |
static bool set_huge_zero_page(pgtable_t pgtable, struct mm_struct *mm, |
97ae17497 thp: implement re... |
756 |
struct vm_area_struct *vma, unsigned long haddr, pmd_t *pmd, |
5918d10a4 thp: fix huge zer... |
757 |
struct page *zero_page) |
fc9fe822f thp: copy_huge_pm... |
758 759 |
{ pmd_t entry; |
3ea41e621 thp: avoid race o... |
760 761 |
if (!pmd_none(*pmd)) return false; |
5918d10a4 thp: fix huge zer... |
762 |
entry = mk_pmd(zero_page, vma->vm_page_prot); |
fc9fe822f thp: copy_huge_pm... |
763 764 |
entry = pmd_wrprotect(entry); entry = pmd_mkhuge(entry); |
6b0b50b06 mm/THP: add pmd a... |
765 |
pgtable_trans_huge_deposit(mm, pmd, pgtable); |
fc9fe822f thp: copy_huge_pm... |
766 |
set_pmd_at(mm, haddr, pmd, entry); |
e1f56c89b mm: convert mm->n... |
767 |
atomic_long_inc(&mm->nr_ptes); |
3ea41e621 thp: avoid race o... |
768 |
return true; |
fc9fe822f thp: copy_huge_pm... |
769 |
} |
71e3aac07 thp: transparent ... |
770 771 772 773 774 775 |
int do_huge_pmd_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pmd_t *pmd, unsigned int flags) { struct page *page; unsigned long haddr = address & HPAGE_PMD_MASK; |
71e3aac07 thp: transparent ... |
776 |
|
128ec037b thp: do_huge_pmd_... |
777 |
if (haddr < vma->vm_start || haddr + HPAGE_PMD_SIZE > vma->vm_end) |
c02925540 thp: consolidate ... |
778 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
779 780 781 782 783 784 |
if (unlikely(anon_vma_prepare(vma))) return VM_FAULT_OOM; if (unlikely(khugepaged_enter(vma))) return VM_FAULT_OOM; if (!(flags & FAULT_FLAG_WRITE) && transparent_hugepage_use_zero_page()) { |
c4088ebdc mm: convert the r... |
785 |
spinlock_t *ptl; |
128ec037b thp: do_huge_pmd_... |
786 787 788 789 790 |
pgtable_t pgtable; struct page *zero_page; bool set; pgtable = pte_alloc_one(mm, haddr); if (unlikely(!pgtable)) |
ba76149f4 thp: khugepaged |
791 |
return VM_FAULT_OOM; |
128ec037b thp: do_huge_pmd_... |
792 793 794 |
zero_page = get_huge_zero_page(); if (unlikely(!zero_page)) { pte_free(mm, pgtable); |
81ab4201f mm: add VM counte... |
795 |
count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
796 |
return VM_FAULT_FALLBACK; |
b9bbfbe30 thp: memcg huge m... |
797 |
} |
c4088ebdc mm: convert the r... |
798 |
ptl = pmd_lock(mm, pmd); |
128ec037b thp: do_huge_pmd_... |
799 800 |
set = set_huge_zero_page(pgtable, mm, vma, haddr, pmd, zero_page); |
c4088ebdc mm: convert the r... |
801 |
spin_unlock(ptl); |
128ec037b thp: do_huge_pmd_... |
802 803 804 |
if (!set) { pte_free(mm, pgtable); put_huge_zero_page(); |
edad9d2c3 mm, thp: allow fa... |
805 |
} |
edad9d2c3 mm, thp: allow fa... |
806 |
return 0; |
71e3aac07 thp: transparent ... |
807 |
} |
128ec037b thp: do_huge_pmd_... |
808 809 810 811 |
page = alloc_hugepage_vma(transparent_hugepage_defrag(vma), vma, haddr, numa_node_id(), 0); if (unlikely(!page)) { count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
812 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
813 |
} |
d715ae08f memcg: rename hig... |
814 |
if (unlikely(mem_cgroup_charge_anon(page, mm, GFP_KERNEL))) { |
128ec037b thp: do_huge_pmd_... |
815 |
put_page(page); |
17766dde3 mm, thp: count th... |
816 |
count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
817 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
818 819 820 821 |
} if (unlikely(__do_huge_pmd_anonymous_page(mm, vma, haddr, pmd, page))) { mem_cgroup_uncharge_page(page); put_page(page); |
17766dde3 mm, thp: count th... |
822 |
count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
823 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
824 |
} |
17766dde3 mm, thp: count th... |
825 |
count_vm_event(THP_FAULT_ALLOC); |
128ec037b thp: do_huge_pmd_... |
826 |
return 0; |
71e3aac07 thp: transparent ... |
827 828 829 830 831 832 |
} int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm, pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr, struct vm_area_struct *vma) { |
c4088ebdc mm: convert the r... |
833 |
spinlock_t *dst_ptl, *src_ptl; |
71e3aac07 thp: transparent ... |
834 835 836 837 838 839 840 841 842 |
struct page *src_page; pmd_t pmd; pgtable_t pgtable; int ret; ret = -ENOMEM; pgtable = pte_alloc_one(dst_mm, addr); if (unlikely(!pgtable)) goto out; |
c4088ebdc mm: convert the r... |
843 844 845 |
dst_ptl = pmd_lock(dst_mm, dst_pmd); src_ptl = pmd_lockptr(src_mm, src_pmd); spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING); |
71e3aac07 thp: transparent ... |
846 847 848 849 850 851 852 |
ret = -EAGAIN; pmd = *src_pmd; if (unlikely(!pmd_trans_huge(pmd))) { pte_free(dst_mm, pgtable); goto out_unlock; } |
fc9fe822f thp: copy_huge_pm... |
853 |
/* |
c4088ebdc mm: convert the r... |
854 |
* When page table lock is held, the huge zero pmd should not be |
fc9fe822f thp: copy_huge_pm... |
855 856 857 858 |
* under splitting since we don't split the page itself, only pmd to * a page table. */ if (is_huge_zero_pmd(pmd)) { |
5918d10a4 thp: fix huge zer... |
859 |
struct page *zero_page; |
3ea41e621 thp: avoid race o... |
860 |
bool set; |
97ae17497 thp: implement re... |
861 862 863 864 865 |
/* * get_huge_zero_page() will never allocate a new page here, * since we already have a zero page to copy. It just takes a * reference. */ |
5918d10a4 thp: fix huge zer... |
866 |
zero_page = get_huge_zero_page(); |
3ea41e621 thp: avoid race o... |
867 |
set = set_huge_zero_page(pgtable, dst_mm, vma, addr, dst_pmd, |
5918d10a4 thp: fix huge zer... |
868 |
zero_page); |
3ea41e621 thp: avoid race o... |
869 |
BUG_ON(!set); /* unexpected !pmd_none(dst_pmd) */ |
fc9fe822f thp: copy_huge_pm... |
870 871 872 |
ret = 0; goto out_unlock; } |
de466bd62 mm: numa: avoid u... |
873 |
|
71e3aac07 thp: transparent ... |
874 875 |
if (unlikely(pmd_trans_splitting(pmd))) { /* split huge page running from under us */ |
c4088ebdc mm: convert the r... |
876 877 |
spin_unlock(src_ptl); spin_unlock(dst_ptl); |
71e3aac07 thp: transparent ... |
878 879 880 881 882 883 |
pte_free(dst_mm, pgtable); wait_split_huge_page(vma->anon_vma, src_pmd); /* src_vma */ goto out; } src_page = pmd_page(pmd); |
309381fea mm: dump page whe... |
884 |
VM_BUG_ON_PAGE(!PageHead(src_page), src_page); |
71e3aac07 thp: transparent ... |
885 886 887 888 889 890 |
get_page(src_page); page_dup_rmap(src_page); add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR); pmdp_set_wrprotect(src_mm, addr, src_pmd); pmd = pmd_mkold(pmd_wrprotect(pmd)); |
6b0b50b06 mm/THP: add pmd a... |
891 |
pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable); |
71e3aac07 thp: transparent ... |
892 |
set_pmd_at(dst_mm, addr, dst_pmd, pmd); |
e1f56c89b mm: convert mm->n... |
893 |
atomic_long_inc(&dst_mm->nr_ptes); |
71e3aac07 thp: transparent ... |
894 895 896 |
ret = 0; out_unlock: |
c4088ebdc mm: convert the r... |
897 898 |
spin_unlock(src_ptl); spin_unlock(dst_ptl); |
71e3aac07 thp: transparent ... |
899 900 901 |
out: return ret; } |
a1dd450bc mm: thp: set the ... |
902 903 904 905 906 907 |
void huge_pmd_set_accessed(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pmd_t *pmd, pmd_t orig_pmd, int dirty) { |
c4088ebdc mm: convert the r... |
908 |
spinlock_t *ptl; |
a1dd450bc mm: thp: set the ... |
909 910 |
pmd_t entry; unsigned long haddr; |
c4088ebdc mm: convert the r... |
911 |
ptl = pmd_lock(mm, pmd); |
a1dd450bc mm: thp: set the ... |
912 913 914 915 916 917 918 919 920 |
if (unlikely(!pmd_same(*pmd, orig_pmd))) goto unlock; entry = pmd_mkyoung(orig_pmd); haddr = address & HPAGE_PMD_MASK; if (pmdp_set_access_flags(vma, haddr, pmd, entry, dirty)) update_mmu_cache_pmd(vma, address, pmd); unlock: |
c4088ebdc mm: convert the r... |
921 |
spin_unlock(ptl); |
a1dd450bc mm: thp: set the ... |
922 |
} |
71e3aac07 thp: transparent ... |
923 924 925 926 927 928 929 |
static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pmd_t *pmd, pmd_t orig_pmd, struct page *page, unsigned long haddr) { |
c4088ebdc mm: convert the r... |
930 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
931 932 933 934 |
pgtable_t pgtable; pmd_t _pmd; int ret = 0, i; struct page **pages; |
2ec74c3ef mm: move all mmu ... |
935 936 |
unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
71e3aac07 thp: transparent ... |
937 938 939 940 941 942 943 944 945 |
pages = kmalloc(sizeof(struct page *) * HPAGE_PMD_NR, GFP_KERNEL); if (unlikely(!pages)) { ret |= VM_FAULT_OOM; goto out; } for (i = 0; i < HPAGE_PMD_NR; i++) { |
cc5d462f7 mm: use __GFP_OTH... |
946 947 |
pages[i] = alloc_page_vma_node(GFP_HIGHUSER_MOVABLE | __GFP_OTHER_NODE, |
19ee151e1 mm: preserve orig... |
948 |
vma, address, page_to_nid(page)); |
b9bbfbe30 thp: memcg huge m... |
949 |
if (unlikely(!pages[i] || |
d715ae08f memcg: rename hig... |
950 |
mem_cgroup_charge_anon(pages[i], mm, |
b9bbfbe30 thp: memcg huge m... |
951 952 |
GFP_KERNEL))) { if (pages[i]) |
71e3aac07 thp: transparent ... |
953 |
put_page(pages[i]); |
b9bbfbe30 thp: memcg huge m... |
954 955 956 957 958 959 |
mem_cgroup_uncharge_start(); while (--i >= 0) { mem_cgroup_uncharge_page(pages[i]); put_page(pages[i]); } mem_cgroup_uncharge_end(); |
71e3aac07 thp: transparent ... |
960 961 962 963 964 965 966 967 |
kfree(pages); ret |= VM_FAULT_OOM; goto out; } } for (i = 0; i < HPAGE_PMD_NR; i++) { copy_user_highpage(pages[i], page + i, |
0089e4853 mm/huge_memory: f... |
968 |
haddr + PAGE_SIZE * i, vma); |
71e3aac07 thp: transparent ... |
969 970 971 |
__SetPageUptodate(pages[i]); cond_resched(); } |
2ec74c3ef mm: move all mmu ... |
972 973 974 |
mmun_start = haddr; mmun_end = haddr + HPAGE_PMD_SIZE; mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); |
c4088ebdc mm: convert the r... |
975 |
ptl = pmd_lock(mm, pmd); |
71e3aac07 thp: transparent ... |
976 977 |
if (unlikely(!pmd_same(*pmd, orig_pmd))) goto out_free_pages; |
309381fea mm: dump page whe... |
978 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
71e3aac07 thp: transparent ... |
979 |
|
2ec74c3ef mm: move all mmu ... |
980 |
pmdp_clear_flush(vma, haddr, pmd); |
71e3aac07 thp: transparent ... |
981 |
/* leave pmd empty until pte is filled */ |
6b0b50b06 mm/THP: add pmd a... |
982 |
pgtable = pgtable_trans_huge_withdraw(mm, pmd); |
71e3aac07 thp: transparent ... |
983 984 985 986 987 988 989 990 991 992 993 994 995 |
pmd_populate(mm, &_pmd, pgtable); for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) { pte_t *pte, entry; entry = mk_pte(pages[i], vma->vm_page_prot); entry = maybe_mkwrite(pte_mkdirty(entry), vma); page_add_new_anon_rmap(pages[i], vma, haddr); pte = pte_offset_map(&_pmd, haddr); VM_BUG_ON(!pte_none(*pte)); set_pte_at(mm, haddr, pte, entry); pte_unmap(pte); } kfree(pages); |
71e3aac07 thp: transparent ... |
996 997 998 |
smp_wmb(); /* make pte visible before pmd */ pmd_populate(mm, pmd, pgtable); page_remove_rmap(page); |
c4088ebdc mm: convert the r... |
999 |
spin_unlock(ptl); |
71e3aac07 thp: transparent ... |
1000 |
|
2ec74c3ef mm: move all mmu ... |
1001 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
1002 1003 1004 1005 1006 1007 1008 |
ret |= VM_FAULT_WRITE; put_page(page); out: return ret; out_free_pages: |
c4088ebdc mm: convert the r... |
1009 |
spin_unlock(ptl); |
2ec74c3ef mm: move all mmu ... |
1010 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
b9bbfbe30 thp: memcg huge m... |
1011 1012 1013 |
mem_cgroup_uncharge_start(); for (i = 0; i < HPAGE_PMD_NR; i++) { mem_cgroup_uncharge_page(pages[i]); |
71e3aac07 thp: transparent ... |
1014 |
put_page(pages[i]); |
b9bbfbe30 thp: memcg huge m... |
1015 1016 |
} mem_cgroup_uncharge_end(); |
71e3aac07 thp: transparent ... |
1017 1018 1019 1020 1021 1022 1023 |
kfree(pages); goto out; } int do_huge_pmd_wp_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, pmd_t *pmd, pmd_t orig_pmd) { |
c4088ebdc mm: convert the r... |
1024 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
1025 |
int ret = 0; |
93b4796de thp: do_huge_pmd_... |
1026 |
struct page *page = NULL, *new_page; |
71e3aac07 thp: transparent ... |
1027 |
unsigned long haddr; |
2ec74c3ef mm: move all mmu ... |
1028 1029 |
unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
71e3aac07 thp: transparent ... |
1030 |
|
c4088ebdc mm: convert the r... |
1031 |
ptl = pmd_lockptr(mm, pmd); |
71e3aac07 thp: transparent ... |
1032 |
VM_BUG_ON(!vma->anon_vma); |
93b4796de thp: do_huge_pmd_... |
1033 1034 1035 |
haddr = address & HPAGE_PMD_MASK; if (is_huge_zero_pmd(orig_pmd)) goto alloc; |
c4088ebdc mm: convert the r... |
1036 |
spin_lock(ptl); |
71e3aac07 thp: transparent ... |
1037 1038 1039 1040 |
if (unlikely(!pmd_same(*pmd, orig_pmd))) goto out_unlock; page = pmd_page(orig_pmd); |
309381fea mm: dump page whe... |
1041 |
VM_BUG_ON_PAGE(!PageCompound(page) || !PageHead(page), page); |
71e3aac07 thp: transparent ... |
1042 1043 1044 1045 1046 |
if (page_mapcount(page) == 1) { pmd_t entry; entry = pmd_mkyoung(orig_pmd); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); if (pmdp_set_access_flags(vma, haddr, pmd, entry, 1)) |
b113da657 mm: Add and use u... |
1047 |
update_mmu_cache_pmd(vma, address, pmd); |
71e3aac07 thp: transparent ... |
1048 1049 1050 1051 |
ret |= VM_FAULT_WRITE; goto out_unlock; } get_page(page); |
c4088ebdc mm: convert the r... |
1052 |
spin_unlock(ptl); |
93b4796de thp: do_huge_pmd_... |
1053 |
alloc: |
71e3aac07 thp: transparent ... |
1054 1055 |
if (transparent_hugepage_enabled(vma) && !transparent_hugepage_debug_cow()) |
0bbbc0b33 thp: add numa awa... |
1056 |
new_page = alloc_hugepage_vma(transparent_hugepage_defrag(vma), |
cc5d462f7 mm: use __GFP_OTH... |
1057 |
vma, haddr, numa_node_id(), 0); |
71e3aac07 thp: transparent ... |
1058 1059 1060 1061 |
else new_page = NULL; if (unlikely(!new_page)) { |
eecc1e426 thp: fix copy_pag... |
1062 |
if (!page) { |
e9b71ca91 mm, thp: drop do_... |
1063 1064 |
split_huge_page_pmd(vma, address, pmd); ret |= VM_FAULT_FALLBACK; |
93b4796de thp: do_huge_pmd_... |
1065 1066 1067 |
} else { ret = do_huge_pmd_wp_page_fallback(mm, vma, address, pmd, orig_pmd, page, haddr); |
9845cbbd1 mm, thp: fix infi... |
1068 |
if (ret & VM_FAULT_OOM) { |
93b4796de thp: do_huge_pmd_... |
1069 |
split_huge_page(page); |
9845cbbd1 mm, thp: fix infi... |
1070 1071 |
ret |= VM_FAULT_FALLBACK; } |
93b4796de thp: do_huge_pmd_... |
1072 1073 |
put_page(page); } |
17766dde3 mm, thp: count th... |
1074 |
count_vm_event(THP_FAULT_FALLBACK); |
71e3aac07 thp: transparent ... |
1075 1076 |
goto out; } |
d715ae08f memcg: rename hig... |
1077 |
if (unlikely(mem_cgroup_charge_anon(new_page, mm, GFP_KERNEL))) { |
b9bbfbe30 thp: memcg huge m... |
1078 |
put_page(new_page); |
93b4796de thp: do_huge_pmd_... |
1079 1080 1081 |
if (page) { split_huge_page(page); put_page(page); |
9845cbbd1 mm, thp: fix infi... |
1082 1083 1084 |
} else split_huge_page_pmd(vma, address, pmd); ret |= VM_FAULT_FALLBACK; |
17766dde3 mm, thp: count th... |
1085 |
count_vm_event(THP_FAULT_FALLBACK); |
b9bbfbe30 thp: memcg huge m... |
1086 1087 |
goto out; } |
17766dde3 mm, thp: count th... |
1088 |
count_vm_event(THP_FAULT_ALLOC); |
eecc1e426 thp: fix copy_pag... |
1089 |
if (!page) |
93b4796de thp: do_huge_pmd_... |
1090 1091 1092 |
clear_huge_page(new_page, haddr, HPAGE_PMD_NR); else copy_user_huge_page(new_page, page, haddr, vma, HPAGE_PMD_NR); |
71e3aac07 thp: transparent ... |
1093 |
__SetPageUptodate(new_page); |
2ec74c3ef mm: move all mmu ... |
1094 1095 1096 |
mmun_start = haddr; mmun_end = haddr + HPAGE_PMD_SIZE; mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); |
c4088ebdc mm: convert the r... |
1097 |
spin_lock(ptl); |
93b4796de thp: do_huge_pmd_... |
1098 1099 |
if (page) put_page(page); |
b9bbfbe30 thp: memcg huge m... |
1100 |
if (unlikely(!pmd_same(*pmd, orig_pmd))) { |
c4088ebdc mm: convert the r... |
1101 |
spin_unlock(ptl); |
b9bbfbe30 thp: memcg huge m... |
1102 |
mem_cgroup_uncharge_page(new_page); |
71e3aac07 thp: transparent ... |
1103 |
put_page(new_page); |
2ec74c3ef mm: move all mmu ... |
1104 |
goto out_mn; |
b9bbfbe30 thp: memcg huge m... |
1105 |
} else { |
71e3aac07 thp: transparent ... |
1106 |
pmd_t entry; |
3122359a6 thp: move maybe_p... |
1107 1108 |
entry = mk_huge_pmd(new_page, vma->vm_page_prot); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); |
2ec74c3ef mm: move all mmu ... |
1109 |
pmdp_clear_flush(vma, haddr, pmd); |
71e3aac07 thp: transparent ... |
1110 1111 |
page_add_new_anon_rmap(new_page, vma, haddr); set_pmd_at(mm, haddr, pmd, entry); |
b113da657 mm: Add and use u... |
1112 |
update_mmu_cache_pmd(vma, address, pmd); |
eecc1e426 thp: fix copy_pag... |
1113 |
if (!page) { |
93b4796de thp: do_huge_pmd_... |
1114 |
add_mm_counter(mm, MM_ANONPAGES, HPAGE_PMD_NR); |
97ae17497 thp: implement re... |
1115 1116 |
put_huge_zero_page(); } else { |
309381fea mm: dump page whe... |
1117 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
93b4796de thp: do_huge_pmd_... |
1118 1119 1120 |
page_remove_rmap(page); put_page(page); } |
71e3aac07 thp: transparent ... |
1121 1122 |
ret |= VM_FAULT_WRITE; } |
c4088ebdc mm: convert the r... |
1123 |
spin_unlock(ptl); |
2ec74c3ef mm: move all mmu ... |
1124 1125 |
out_mn: mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
1126 1127 |
out: return ret; |
2ec74c3ef mm: move all mmu ... |
1128 |
out_unlock: |
c4088ebdc mm: convert the r... |
1129 |
spin_unlock(ptl); |
2ec74c3ef mm: move all mmu ... |
1130 |
return ret; |
71e3aac07 thp: transparent ... |
1131 |
} |
b676b293f mm, thp: fix mapp... |
1132 |
struct page *follow_trans_huge_pmd(struct vm_area_struct *vma, |
71e3aac07 thp: transparent ... |
1133 1134 1135 1136 |
unsigned long addr, pmd_t *pmd, unsigned int flags) { |
b676b293f mm, thp: fix mapp... |
1137 |
struct mm_struct *mm = vma->vm_mm; |
71e3aac07 thp: transparent ... |
1138 |
struct page *page = NULL; |
c4088ebdc mm: convert the r... |
1139 |
assert_spin_locked(pmd_lockptr(mm, pmd)); |
71e3aac07 thp: transparent ... |
1140 1141 1142 |
if (flags & FOLL_WRITE && !pmd_write(*pmd)) goto out; |
85facf257 thp: avoid dumpin... |
1143 1144 1145 |
/* Avoid dumping huge zero page */ if ((flags & FOLL_DUMP) && is_huge_zero_pmd(*pmd)) return ERR_PTR(-EFAULT); |
2b4847e73 mm: numa: seriali... |
1146 1147 1148 |
/* Full NUMA hinting faults to serialise migration in fault paths */ if ((flags & FOLL_NUMA) && pmd_numa(*pmd)) goto out; |
71e3aac07 thp: transparent ... |
1149 |
page = pmd_page(*pmd); |
309381fea mm: dump page whe... |
1150 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
71e3aac07 thp: transparent ... |
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 |
if (flags & FOLL_TOUCH) { pmd_t _pmd; /* * We should set the dirty bit only for FOLL_WRITE but * for now the dirty bit in the pmd is meaningless. * And if the dirty bit will become meaningful and * we'll only set it with FOLL_WRITE, an atomic * set_bit will be required on the pmd to set the * young bit, instead of the current set_pmd_at. */ _pmd = pmd_mkyoung(pmd_mkdirty(*pmd)); |
8663890a9 mm/thp: use the c... |
1162 1163 1164 |
if (pmdp_set_access_flags(vma, addr & HPAGE_PMD_MASK, pmd, _pmd, 1)) update_mmu_cache_pmd(vma, addr, pmd); |
71e3aac07 thp: transparent ... |
1165 |
} |
b676b293f mm, thp: fix mapp... |
1166 1167 1168 1169 1170 1171 1172 1173 |
if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) { if (page->mapping && trylock_page(page)) { lru_add_drain(); if (page->mapping) mlock_vma_page(page); unlock_page(page); } } |
71e3aac07 thp: transparent ... |
1174 |
page += (addr & ~HPAGE_PMD_MASK) >> PAGE_SHIFT; |
309381fea mm: dump page whe... |
1175 |
VM_BUG_ON_PAGE(!PageCompound(page), page); |
71e3aac07 thp: transparent ... |
1176 |
if (flags & FOLL_GET) |
70b50f94f mm: thp: tail pag... |
1177 |
get_page_foll(page); |
71e3aac07 thp: transparent ... |
1178 1179 1180 1181 |
out: return page; } |
d10e63f29 mm: numa: Create ... |
1182 |
/* NUMA hinting page fault entry point for trans huge pmds */ |
4daae3b4b mm: mempolicy: Us... |
1183 1184 |
int do_huge_pmd_numa_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, pmd_t pmd, pmd_t *pmdp) |
d10e63f29 mm: numa: Create ... |
1185 |
{ |
c4088ebdc mm: convert the r... |
1186 |
spinlock_t *ptl; |
b8916634b mm: Prevent paral... |
1187 |
struct anon_vma *anon_vma = NULL; |
b32967ff1 mm: numa: Add THP... |
1188 |
struct page *page; |
d10e63f29 mm: numa: Create ... |
1189 |
unsigned long haddr = addr & HPAGE_PMD_MASK; |
8191acbd3 mm: numa: Sanitiz... |
1190 |
int page_nid = -1, this_nid = numa_node_id(); |
90572890d mm: numa: Change ... |
1191 |
int target_nid, last_cpupid = -1; |
8191acbd3 mm: numa: Sanitiz... |
1192 1193 |
bool page_locked; bool migrated = false; |
6688cc054 mm: numa: Do not ... |
1194 |
int flags = 0; |
d10e63f29 mm: numa: Create ... |
1195 |
|
c4088ebdc mm: convert the r... |
1196 |
ptl = pmd_lock(mm, pmdp); |
d10e63f29 mm: numa: Create ... |
1197 1198 |
if (unlikely(!pmd_same(pmd, *pmdp))) goto out_unlock; |
de466bd62 mm: numa: avoid u... |
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 |
/* * If there are potential migrations, wait for completion and retry * without disrupting NUMA hinting information. Do not relock and * check_same as the page may no longer be mapped. */ if (unlikely(pmd_trans_migrating(*pmdp))) { spin_unlock(ptl); wait_migrate_huge_page(vma->anon_vma, pmdp); goto out; } |
d10e63f29 mm: numa: Create ... |
1209 |
page = pmd_page(pmd); |
a1a46184e mm: numa: Do not ... |
1210 |
BUG_ON(is_huge_zero_page(page)); |
8191acbd3 mm: numa: Sanitiz... |
1211 |
page_nid = page_to_nid(page); |
90572890d mm: numa: Change ... |
1212 |
last_cpupid = page_cpupid_last(page); |
03c5a6e16 mm: numa: Add pte... |
1213 |
count_vm_numa_event(NUMA_HINT_FAULTS); |
04bb2f947 sched/numa: Adjus... |
1214 |
if (page_nid == this_nid) { |
03c5a6e16 mm: numa: Add pte... |
1215 |
count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL); |
04bb2f947 sched/numa: Adjus... |
1216 1217 |
flags |= TNF_FAULT_LOCAL; } |
4daae3b4b mm: mempolicy: Us... |
1218 |
|
ff9042b11 mm: Wait for THP ... |
1219 |
/* |
6688cc054 mm: numa: Do not ... |
1220 1221 1222 1223 1224 1225 1226 1227 |
* Avoid grouping on DSO/COW pages in specific and RO pages * in general, RO pages shouldn't hurt as much anyway since * they can be in shared cache state. */ if (!pmd_write(pmd)) flags |= TNF_NO_GROUP; /* |
ff9042b11 mm: Wait for THP ... |
1228 1229 1230 |
* Acquire the page lock to serialise THP migrations but avoid dropping * page_table_lock if at all possible */ |
b8916634b mm: Prevent paral... |
1231 1232 1233 1234 |
page_locked = trylock_page(page); target_nid = mpol_misplaced(page, vma, haddr); if (target_nid == -1) { /* If the page was locked, there are no parallel migrations */ |
a54a407fb mm: Close races b... |
1235 |
if (page_locked) |
b8916634b mm: Prevent paral... |
1236 |
goto clear_pmdnuma; |
2b4847e73 mm: numa: seriali... |
1237 |
} |
4daae3b4b mm: mempolicy: Us... |
1238 |
|
de466bd62 mm: numa: avoid u... |
1239 |
/* Migration could have started since the pmd_trans_migrating check */ |
2b4847e73 mm: numa: seriali... |
1240 |
if (!page_locked) { |
c4088ebdc mm: convert the r... |
1241 |
spin_unlock(ptl); |
b8916634b mm: Prevent paral... |
1242 |
wait_on_page_locked(page); |
a54a407fb mm: Close races b... |
1243 |
page_nid = -1; |
b8916634b mm: Prevent paral... |
1244 1245 |
goto out; } |
2b4847e73 mm: numa: seriali... |
1246 1247 1248 1249 |
/* * Page is misplaced. Page lock serialises migrations. Acquire anon_vma * to serialises splits */ |
b8916634b mm: Prevent paral... |
1250 |
get_page(page); |
c4088ebdc mm: convert the r... |
1251 |
spin_unlock(ptl); |
b8916634b mm: Prevent paral... |
1252 |
anon_vma = page_lock_anon_vma_read(page); |
4daae3b4b mm: mempolicy: Us... |
1253 |
|
c69307d53 sched/numa: Fix c... |
1254 |
/* Confirm the PMD did not change while page_table_lock was released */ |
c4088ebdc mm: convert the r... |
1255 |
spin_lock(ptl); |
b32967ff1 mm: numa: Add THP... |
1256 1257 1258 |
if (unlikely(!pmd_same(pmd, *pmdp))) { unlock_page(page); put_page(page); |
a54a407fb mm: Close races b... |
1259 |
page_nid = -1; |
4daae3b4b mm: mempolicy: Us... |
1260 |
goto out_unlock; |
b32967ff1 mm: numa: Add THP... |
1261 |
} |
ff9042b11 mm: Wait for THP ... |
1262 |
|
c3a489cac mm: numa: ensure ... |
1263 1264 1265 1266 1267 1268 |
/* Bail if we fail to protect against THP splits for any reason */ if (unlikely(!anon_vma)) { put_page(page); page_nid = -1; goto clear_pmdnuma; } |
a54a407fb mm: Close races b... |
1269 1270 1271 1272 |
/* * Migrate the THP to the requested node, returns with page unlocked * and pmd_numa cleared. */ |
c4088ebdc mm: convert the r... |
1273 |
spin_unlock(ptl); |
b32967ff1 mm: numa: Add THP... |
1274 |
migrated = migrate_misplaced_transhuge_page(mm, vma, |
340ef3902 mm: numa: cleanup... |
1275 |
pmdp, pmd, addr, page, target_nid); |
6688cc054 mm: numa: Do not ... |
1276 1277 |
if (migrated) { flags |= TNF_MIGRATED; |
8191acbd3 mm: numa: Sanitiz... |
1278 |
page_nid = target_nid; |
6688cc054 mm: numa: Do not ... |
1279 |
} |
b32967ff1 mm: numa: Add THP... |
1280 |
|
8191acbd3 mm: numa: Sanitiz... |
1281 |
goto out; |
b32967ff1 mm: numa: Add THP... |
1282 |
clear_pmdnuma: |
a54a407fb mm: Close races b... |
1283 |
BUG_ON(!PageLocked(page)); |
d10e63f29 mm: numa: Create ... |
1284 1285 1286 1287 |
pmd = pmd_mknonnuma(pmd); set_pmd_at(mm, haddr, pmdp, pmd); VM_BUG_ON(pmd_numa(*pmdp)); update_mmu_cache_pmd(vma, addr, pmdp); |
a54a407fb mm: Close races b... |
1288 |
unlock_page(page); |
d10e63f29 mm: numa: Create ... |
1289 |
out_unlock: |
c4088ebdc mm: convert the r... |
1290 |
spin_unlock(ptl); |
b8916634b mm: Prevent paral... |
1291 1292 1293 1294 |
out: if (anon_vma) page_unlock_anon_vma_read(anon_vma); |
8191acbd3 mm: numa: Sanitiz... |
1295 |
if (page_nid != -1) |
6688cc054 mm: numa: Do not ... |
1296 |
task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR, flags); |
8191acbd3 mm: numa: Sanitiz... |
1297 |
|
d10e63f29 mm: numa: Create ... |
1298 1299 |
return 0; } |
71e3aac07 thp: transparent ... |
1300 |
int zap_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma, |
f21760b15 thp: add tlb_remo... |
1301 |
pmd_t *pmd, unsigned long addr) |
71e3aac07 thp: transparent ... |
1302 |
{ |
bf929152e mm, thp: change p... |
1303 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
1304 |
int ret = 0; |
bf929152e mm, thp: change p... |
1305 |
if (__pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
025c5b245 thp: optimize awa... |
1306 1307 |
struct page *page; pgtable_t pgtable; |
f5c8ad472 mm: thp: Use more... |
1308 |
pmd_t orig_pmd; |
a6bf2bb03 mm/THP: withdraw ... |
1309 1310 1311 1312 1313 1314 |
/* * For architectures like ppc64 we look at deposited pgtable * when calling pmdp_get_and_clear. So do the * pgtable_trans_huge_withdraw after finishing pmdp related * operations. */ |
f5c8ad472 mm: thp: Use more... |
1315 |
orig_pmd = pmdp_get_and_clear(tlb->mm, addr, pmd); |
025c5b245 thp: optimize awa... |
1316 |
tlb_remove_pmd_tlb_entry(tlb, pmd, addr); |
a6bf2bb03 mm/THP: withdraw ... |
1317 |
pgtable = pgtable_trans_huge_withdraw(tlb->mm, pmd); |
479f0abbf thp: zap_huge_pmd... |
1318 |
if (is_huge_zero_pmd(orig_pmd)) { |
e1f56c89b mm: convert mm->n... |
1319 |
atomic_long_dec(&tlb->mm->nr_ptes); |
bf929152e mm, thp: change p... |
1320 |
spin_unlock(ptl); |
97ae17497 thp: implement re... |
1321 |
put_huge_zero_page(); |
479f0abbf thp: zap_huge_pmd... |
1322 1323 1324 |
} else { page = pmd_page(orig_pmd); page_remove_rmap(page); |
309381fea mm: dump page whe... |
1325 |
VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); |
479f0abbf thp: zap_huge_pmd... |
1326 |
add_mm_counter(tlb->mm, MM_ANONPAGES, -HPAGE_PMD_NR); |
309381fea mm: dump page whe... |
1327 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
e1f56c89b mm: convert mm->n... |
1328 |
atomic_long_dec(&tlb->mm->nr_ptes); |
bf929152e mm, thp: change p... |
1329 |
spin_unlock(ptl); |
479f0abbf thp: zap_huge_pmd... |
1330 1331 |
tlb_remove_page(tlb, page); } |
025c5b245 thp: optimize awa... |
1332 1333 1334 |
pte_free(tlb->mm, pgtable); ret = 1; } |
71e3aac07 thp: transparent ... |
1335 1336 |
return ret; } |
0ca1634d4 thp: mincore tran... |
1337 1338 1339 1340 |
int mincore_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end, unsigned char *vec) { |
bf929152e mm, thp: change p... |
1341 |
spinlock_t *ptl; |
0ca1634d4 thp: mincore tran... |
1342 |
int ret = 0; |
bf929152e mm, thp: change p... |
1343 |
if (__pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
025c5b245 thp: optimize awa... |
1344 1345 1346 1347 |
/* * All logical pages in the range are present * if backed by a huge page. */ |
bf929152e mm, thp: change p... |
1348 |
spin_unlock(ptl); |
025c5b245 thp: optimize awa... |
1349 1350 1351 |
memset(vec, 1, (end - addr) >> PAGE_SHIFT); ret = 1; } |
0ca1634d4 thp: mincore tran... |
1352 1353 1354 |
return ret; } |
37a1c49a9 thp: mremap suppo... |
1355 1356 1357 1358 1359 |
int move_huge_pmd(struct vm_area_struct *vma, struct vm_area_struct *new_vma, unsigned long old_addr, unsigned long new_addr, unsigned long old_end, pmd_t *old_pmd, pmd_t *new_pmd) { |
bf929152e mm, thp: change p... |
1360 |
spinlock_t *old_ptl, *new_ptl; |
37a1c49a9 thp: mremap suppo... |
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 |
int ret = 0; pmd_t pmd; struct mm_struct *mm = vma->vm_mm; if ((old_addr & ~HPAGE_PMD_MASK) || (new_addr & ~HPAGE_PMD_MASK) || old_end - old_addr < HPAGE_PMD_SIZE || (new_vma->vm_flags & VM_NOHUGEPAGE)) goto out; /* * The destination pmd shouldn't be established, free_pgtables() * should have release it. */ if (WARN_ON(!pmd_none(*new_pmd))) { VM_BUG_ON(pmd_trans_huge(*new_pmd)); goto out; } |
bf929152e mm, thp: change p... |
1380 1381 1382 1383 1384 |
/* * We don't have to worry about the ordering of src and dst * ptlocks because exclusive mmap_sem prevents deadlock. */ ret = __pmd_trans_huge_lock(old_pmd, vma, &old_ptl); |
025c5b245 thp: optimize awa... |
1385 |
if (ret == 1) { |
bf929152e mm, thp: change p... |
1386 1387 1388 |
new_ptl = pmd_lockptr(mm, new_pmd); if (new_ptl != old_ptl) spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING); |
025c5b245 thp: optimize awa... |
1389 1390 |
pmd = pmdp_get_and_clear(mm, old_addr, old_pmd); VM_BUG_ON(!pmd_none(*new_pmd)); |
3592806cf thp: move preallo... |
1391 |
|
b3084f4db powerpc/thp: Fix ... |
1392 1393 |
if (pmd_move_must_withdraw(new_ptl, old_ptl)) { pgtable_t pgtable; |
3592806cf thp: move preallo... |
1394 1395 |
pgtable = pgtable_trans_huge_withdraw(mm, old_pmd); pgtable_trans_huge_deposit(mm, new_pmd, pgtable); |
3592806cf thp: move preallo... |
1396 |
} |
b3084f4db powerpc/thp: Fix ... |
1397 1398 1399 |
set_pmd_at(mm, new_addr, new_pmd, pmd_mksoft_dirty(pmd)); if (new_ptl != old_ptl) spin_unlock(new_ptl); |
bf929152e mm, thp: change p... |
1400 |
spin_unlock(old_ptl); |
37a1c49a9 thp: mremap suppo... |
1401 1402 1403 1404 |
} out: return ret; } |
f123d74ab mm: Only flush TL... |
1405 1406 1407 1408 1409 1410 |
/* * Returns * - 0 if PMD could not be locked * - 1 if PMD was locked but protections unchange and TLB flush unnecessary * - HPAGE_PMD_NR is protections changed and TLB flush necessary */ |
cd7548ab3 thp: mprotect: tr... |
1411 |
int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, |
4b10e7d56 mm: mempolicy: Im... |
1412 |
unsigned long addr, pgprot_t newprot, int prot_numa) |
cd7548ab3 thp: mprotect: tr... |
1413 1414 |
{ struct mm_struct *mm = vma->vm_mm; |
bf929152e mm, thp: change p... |
1415 |
spinlock_t *ptl; |
cd7548ab3 thp: mprotect: tr... |
1416 |
int ret = 0; |
bf929152e mm, thp: change p... |
1417 |
if (__pmd_trans_huge_lock(pmd, vma, &ptl) == 1) { |
025c5b245 thp: optimize awa... |
1418 |
pmd_t entry; |
f123d74ab mm: Only flush TL... |
1419 |
ret = 1; |
a4f1de176 mm: fix kernel BU... |
1420 |
if (!prot_numa) { |
f123d74ab mm: Only flush TL... |
1421 |
entry = pmdp_get_and_clear(mm, addr, pmd); |
1667918b6 mm: numa: clear n... |
1422 1423 |
if (pmd_numa(entry)) entry = pmd_mknonnuma(entry); |
4b10e7d56 mm: mempolicy: Im... |
1424 |
entry = pmd_modify(entry, newprot); |
f123d74ab mm: Only flush TL... |
1425 |
ret = HPAGE_PMD_NR; |
56eecdb91 mm: Use ptep/pmdp... |
1426 |
set_pmd_at(mm, addr, pmd, entry); |
a4f1de176 mm: fix kernel BU... |
1427 1428 |
BUG_ON(pmd_write(entry)); } else { |
4b10e7d56 mm: mempolicy: Im... |
1429 |
struct page *page = pmd_page(*pmd); |
a1a46184e mm: numa: Do not ... |
1430 |
/* |
1bc115d87 mm: numa: Scan pa... |
1431 1432 1433 1434 |
* Do not trap faults against the zero page. The * read-only data is likely to be read-cached on the * local CPU cache and it is less useful to know about * local vs remote hits on the zero page. |
a1a46184e mm: numa: Do not ... |
1435 |
*/ |
1bc115d87 mm: numa: Scan pa... |
1436 |
if (!is_huge_zero_page(page) && |
4b10e7d56 mm: mempolicy: Im... |
1437 |
!pmd_numa(*pmd)) { |
56eecdb91 mm: Use ptep/pmdp... |
1438 |
pmdp_set_numa(mm, addr, pmd); |
f123d74ab mm: Only flush TL... |
1439 |
ret = HPAGE_PMD_NR; |
4b10e7d56 mm: mempolicy: Im... |
1440 1441 |
} } |
bf929152e mm, thp: change p... |
1442 |
spin_unlock(ptl); |
025c5b245 thp: optimize awa... |
1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 |
} return ret; } /* * Returns 1 if a given pmd maps a stable (not under splitting) thp. * Returns -1 if it maps a thp under splitting. Returns 0 otherwise. * * Note that if it returns 1, this routine returns without unlocking page * table locks. So callers must unlock them. */ |
bf929152e mm, thp: change p... |
1455 1456 |
int __pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma, spinlock_t **ptl) |
025c5b245 thp: optimize awa... |
1457 |
{ |
bf929152e mm, thp: change p... |
1458 |
*ptl = pmd_lock(vma->vm_mm, pmd); |
cd7548ab3 thp: mprotect: tr... |
1459 1460 |
if (likely(pmd_trans_huge(*pmd))) { if (unlikely(pmd_trans_splitting(*pmd))) { |
bf929152e mm, thp: change p... |
1461 |
spin_unlock(*ptl); |
cd7548ab3 thp: mprotect: tr... |
1462 |
wait_split_huge_page(vma->anon_vma, pmd); |
025c5b245 thp: optimize awa... |
1463 |
return -1; |
cd7548ab3 thp: mprotect: tr... |
1464 |
} else { |
025c5b245 thp: optimize awa... |
1465 1466 1467 |
/* Thp mapped by 'pmd' is stable, so we can * handle it as it is. */ return 1; |
cd7548ab3 thp: mprotect: tr... |
1468 |
} |
025c5b245 thp: optimize awa... |
1469 |
} |
bf929152e mm, thp: change p... |
1470 |
spin_unlock(*ptl); |
025c5b245 thp: optimize awa... |
1471 |
return 0; |
cd7548ab3 thp: mprotect: tr... |
1472 |
} |
117b0791a mm, thp: move ptl... |
1473 1474 1475 1476 1477 1478 1479 1480 |
/* * This function returns whether a given @page is mapped onto the @address * in the virtual space of @mm. * * When it's true, this function returns *pmd with holding the page table lock * and passing it back to the caller via @ptl. * If it's false, returns NULL without holding the page table lock. */ |
71e3aac07 thp: transparent ... |
1481 1482 1483 |
pmd_t *page_check_address_pmd(struct page *page, struct mm_struct *mm, unsigned long address, |
117b0791a mm, thp: move ptl... |
1484 1485 |
enum page_check_address_pmd_flag flag, spinlock_t **ptl) |
71e3aac07 thp: transparent ... |
1486 |
{ |
b5a8cad37 thp: close race b... |
1487 1488 |
pgd_t *pgd; pud_t *pud; |
117b0791a mm, thp: move ptl... |
1489 |
pmd_t *pmd; |
71e3aac07 thp: transparent ... |
1490 1491 |
if (address & ~HPAGE_PMD_MASK) |
117b0791a mm, thp: move ptl... |
1492 |
return NULL; |
71e3aac07 thp: transparent ... |
1493 |
|
b5a8cad37 thp: close race b... |
1494 1495 |
pgd = pgd_offset(mm, address); if (!pgd_present(*pgd)) |
117b0791a mm, thp: move ptl... |
1496 |
return NULL; |
b5a8cad37 thp: close race b... |
1497 1498 1499 1500 |
pud = pud_offset(pgd, address); if (!pud_present(*pud)) return NULL; pmd = pmd_offset(pud, address); |
117b0791a mm, thp: move ptl... |
1501 |
*ptl = pmd_lock(mm, pmd); |
b5a8cad37 thp: close race b... |
1502 |
if (!pmd_present(*pmd)) |
117b0791a mm, thp: move ptl... |
1503 |
goto unlock; |
71e3aac07 thp: transparent ... |
1504 |
if (pmd_page(*pmd) != page) |
117b0791a mm, thp: move ptl... |
1505 |
goto unlock; |
94fcc585f thp: avoid breaki... |
1506 1507 1508 1509 1510 1511 1512 1513 1514 |
/* * split_vma() may create temporary aliased mappings. There is * no risk as long as all huge pmd are found and have their * splitting bit set before __split_huge_page_refcount * runs. Finding the same huge pmd more than once during the * same rmap walk is not a problem. */ if (flag == PAGE_CHECK_ADDRESS_PMD_NOTSPLITTING_FLAG && pmd_trans_splitting(*pmd)) |
117b0791a mm, thp: move ptl... |
1515 |
goto unlock; |
71e3aac07 thp: transparent ... |
1516 1517 1518 |
if (pmd_trans_huge(*pmd)) { VM_BUG_ON(flag == PAGE_CHECK_ADDRESS_PMD_SPLITTING_FLAG && !pmd_trans_splitting(*pmd)); |
117b0791a mm, thp: move ptl... |
1519 |
return pmd; |
71e3aac07 thp: transparent ... |
1520 |
} |
117b0791a mm, thp: move ptl... |
1521 1522 1523 |
unlock: spin_unlock(*ptl); return NULL; |
71e3aac07 thp: transparent ... |
1524 1525 1526 1527 1528 1529 1530 |
} static int __split_huge_page_splitting(struct page *page, struct vm_area_struct *vma, unsigned long address) { struct mm_struct *mm = vma->vm_mm; |
117b0791a mm, thp: move ptl... |
1531 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
1532 1533 |
pmd_t *pmd; int ret = 0; |
2ec74c3ef mm: move all mmu ... |
1534 1535 1536 |
/* For mmu_notifiers */ const unsigned long mmun_start = address; const unsigned long mmun_end = address + HPAGE_PMD_SIZE; |
71e3aac07 thp: transparent ... |
1537 |
|
2ec74c3ef mm: move all mmu ... |
1538 |
mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
1539 |
pmd = page_check_address_pmd(page, mm, address, |
117b0791a mm, thp: move ptl... |
1540 |
PAGE_CHECK_ADDRESS_PMD_NOTSPLITTING_FLAG, &ptl); |
71e3aac07 thp: transparent ... |
1541 1542 1543 1544 1545 |
if (pmd) { /* * We can't temporarily set the pmd to null in order * to split it, the pmd must remain marked huge at all * times or the VM won't take the pmd_trans_huge paths |
5a505085f mm/rmap: Convert ... |
1546 |
* and it won't wait on the anon_vma->root->rwsem to |
71e3aac07 thp: transparent ... |
1547 1548 |
* serialize against split_huge_page*. */ |
2ec74c3ef mm: move all mmu ... |
1549 |
pmdp_splitting_flush(vma, address, pmd); |
71e3aac07 thp: transparent ... |
1550 |
ret = 1; |
117b0791a mm, thp: move ptl... |
1551 |
spin_unlock(ptl); |
71e3aac07 thp: transparent ... |
1552 |
} |
2ec74c3ef mm: move all mmu ... |
1553 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
1554 1555 1556 |
return ret; } |
5bc7b8aca mm: thp: add spli... |
1557 1558 |
static void __split_huge_page_refcount(struct page *page, struct list_head *list) |
71e3aac07 thp: transparent ... |
1559 1560 |
{ int i; |
71e3aac07 thp: transparent ... |
1561 |
struct zone *zone = page_zone(page); |
fa9add641 mm/memcg: apply a... |
1562 |
struct lruvec *lruvec; |
70b50f94f mm: thp: tail pag... |
1563 |
int tail_count = 0; |
71e3aac07 thp: transparent ... |
1564 1565 1566 |
/* prevent PageLRU to go away from under us, and freeze lru stats */ spin_lock_irq(&zone->lru_lock); |
fa9add641 mm/memcg: apply a... |
1567 |
lruvec = mem_cgroup_page_lruvec(page, zone); |
71e3aac07 thp: transparent ... |
1568 |
compound_lock(page); |
e94c8a9cb memcg: make mem_c... |
1569 1570 |
/* complete memcg works before add pages to LRU */ mem_cgroup_split_huge_fixup(page); |
71e3aac07 thp: transparent ... |
1571 |
|
45676885b thp: improve orde... |
1572 |
for (i = HPAGE_PMD_NR - 1; i >= 1; i--) { |
71e3aac07 thp: transparent ... |
1573 |
struct page *page_tail = page + i; |
70b50f94f mm: thp: tail pag... |
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 |
/* tail_page->_mapcount cannot change */ BUG_ON(page_mapcount(page_tail) < 0); tail_count += page_mapcount(page_tail); /* check for overflow */ BUG_ON(tail_count < 0); BUG_ON(atomic_read(&page_tail->_count) != 0); /* * tail_page->_count is zero and not changing from * under us. But get_page_unless_zero() may be running * from under us on the tail_page. If we used * atomic_set() below instead of atomic_add(), we * would then run atomic_set() concurrently with * get_page_unless_zero(), and atomic_set() is * implemented in C not using locked ops. spin_unlock * on x86 sometime uses locked ops because of PPro * errata 66, 92, so unless somebody can guarantee * atomic_set() here would be safe on all archs (and * not only on x86), it's safer to use atomic_add(). */ atomic_add(page_mapcount(page) + page_mapcount(page_tail) + 1, &page_tail->_count); |
71e3aac07 thp: transparent ... |
1595 1596 1597 |
/* after clearing PageTail the gup refcount can be released */ smp_mb(); |
a6d30ddda thp: fix the wron... |
1598 1599 1600 1601 1602 1603 |
/* * retain hwpoison flag of the poisoned tail page: * fix for the unsuitable process killed on Guest Machine(KVM) * by the memory-failure. */ page_tail->flags &= ~PAGE_FLAGS_CHECK_AT_PREP | __PG_HWPOISON; |
71e3aac07 thp: transparent ... |
1604 1605 1606 1607 |
page_tail->flags |= (page->flags & ((1L << PG_referenced) | (1L << PG_swapbacked) | (1L << PG_mlocked) | |
e180cf806 thp, mm: avoid Pa... |
1608 1609 1610 |
(1L << PG_uptodate) | (1L << PG_active) | (1L << PG_unevictable))); |
71e3aac07 thp: transparent ... |
1611 |
page_tail->flags |= (1L << PG_dirty); |
70b50f94f mm: thp: tail pag... |
1612 |
/* clear PageTail before overwriting first_page */ |
71e3aac07 thp: transparent ... |
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 |
smp_wmb(); /* * __split_huge_page_splitting() already set the * splitting bit in all pmd that could map this * hugepage, that will ensure no CPU can alter the * mapcount on the head page. The mapcount is only * accounted in the head page and it has to be * transferred to all tail pages in the below code. So * for this code to be safe, the split the mapcount * can't change. But that doesn't mean userland can't * keep changing and reading the page contents while * we transfer the mapcount, so the pmd splitting * status is achieved setting a reserved bit in the * pmd, not by clearing the present bit. */ |
71e3aac07 thp: transparent ... |
1629 1630 1631 1632 |
page_tail->_mapcount = page->_mapcount; BUG_ON(page_tail->mapping); page_tail->mapping = page->mapping; |
45676885b thp: improve orde... |
1633 |
page_tail->index = page->index + i; |
90572890d mm: numa: Change ... |
1634 |
page_cpupid_xchg_last(page_tail, page_cpupid_last(page)); |
71e3aac07 thp: transparent ... |
1635 1636 1637 1638 1639 |
BUG_ON(!PageAnon(page_tail)); BUG_ON(!PageUptodate(page_tail)); BUG_ON(!PageDirty(page_tail)); BUG_ON(!PageSwapBacked(page_tail)); |
5bc7b8aca mm: thp: add spli... |
1640 |
lru_add_page_tail(page, page_tail, lruvec, list); |
71e3aac07 thp: transparent ... |
1641 |
} |
70b50f94f mm: thp: tail pag... |
1642 1643 |
atomic_sub(tail_count, &page->_count); BUG_ON(atomic_read(&page->_count) <= 0); |
71e3aac07 thp: transparent ... |
1644 |
|
fa9add641 mm/memcg: apply a... |
1645 |
__mod_zone_page_state(zone, NR_ANON_TRANSPARENT_HUGEPAGES, -1); |
79134171d thp: transparent ... |
1646 |
|
71e3aac07 thp: transparent ... |
1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 |
ClearPageCompound(page); compound_unlock(page); spin_unlock_irq(&zone->lru_lock); for (i = 1; i < HPAGE_PMD_NR; i++) { struct page *page_tail = page + i; BUG_ON(page_count(page_tail) <= 0); /* * Tail pages may be freed if there wasn't any mapping * like if add_to_swap() is running on a lru page that * had its mapping zapped. And freeing these pages * requires taking the lru_lock so we do the put_page * of the tail pages after the split is complete. */ put_page(page_tail); } /* * Only the head page (now become a regular page) is required * to be pinned by the caller. */ BUG_ON(page_count(page) <= 0); } static int __split_huge_page_map(struct page *page, struct vm_area_struct *vma, unsigned long address) { struct mm_struct *mm = vma->vm_mm; |
117b0791a mm, thp: move ptl... |
1676 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
1677 1678 1679 1680 |
pmd_t *pmd, _pmd; int ret = 0, i; pgtable_t pgtable; unsigned long haddr; |
71e3aac07 thp: transparent ... |
1681 |
pmd = page_check_address_pmd(page, mm, address, |
117b0791a mm, thp: move ptl... |
1682 |
PAGE_CHECK_ADDRESS_PMD_SPLITTING_FLAG, &ptl); |
71e3aac07 thp: transparent ... |
1683 |
if (pmd) { |
6b0b50b06 mm/THP: add pmd a... |
1684 |
pgtable = pgtable_trans_huge_withdraw(mm, pmd); |
71e3aac07 thp: transparent ... |
1685 |
pmd_populate(mm, &_pmd, pgtable); |
e3ebcf643 thp: remove assum... |
1686 1687 |
haddr = address; for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) { |
71e3aac07 thp: transparent ... |
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 |
pte_t *pte, entry; BUG_ON(PageCompound(page+i)); entry = mk_pte(page + i, vma->vm_page_prot); entry = maybe_mkwrite(pte_mkdirty(entry), vma); if (!pmd_write(*pmd)) entry = pte_wrprotect(entry); else BUG_ON(page_mapcount(page) != 1); if (!pmd_young(*pmd)) entry = pte_mkold(entry); |
1ba6e0b50 mm: numa: split_h... |
1698 1699 |
if (pmd_numa(*pmd)) entry = pte_mknuma(entry); |
71e3aac07 thp: transparent ... |
1700 1701 1702 1703 1704 |
pte = pte_offset_map(&_pmd, haddr); BUG_ON(!pte_none(*pte)); set_pte_at(mm, haddr, pte, entry); pte_unmap(pte); } |
71e3aac07 thp: transparent ... |
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 |
smp_wmb(); /* make pte visible before pmd */ /* * Up to this point the pmd is present and huge and * userland has the whole access to the hugepage * during the split (which happens in place). If we * overwrite the pmd with the not-huge version * pointing to the pte here (which of course we could * if all CPUs were bug free), userland could trigger * a small page size TLB miss on the small sized TLB * while the hugepage TLB entry is still established * in the huge TLB. Some CPU doesn't like that. See * http://support.amd.com/us/Processor_TechDocs/41322.pdf, * Erratum 383 on page 93. Intel should be safe but is * also warns that it's only safe if the permission * and cache attributes of the two entries loaded in * the two TLB is identical (which should be the case * here). But it is generally safer to never allow * small and huge TLB entries for the same virtual * address to be loaded simultaneously. So instead of * doing "pmd_populate(); flush_tlb_range();" we first * mark the current pmd notpresent (atomically because * here the pmd_trans_huge and pmd_trans_splitting * must remain set at all times on the pmd until the * split is complete for this pmd), then we flush the * SMP TLB and finally we write the non-huge version * of the pmd entry with pmd_populate. */ |
46dcde735 thp: introduce pm... |
1732 |
pmdp_invalidate(vma, address, pmd); |
71e3aac07 thp: transparent ... |
1733 1734 |
pmd_populate(mm, pmd, pgtable); ret = 1; |
117b0791a mm, thp: move ptl... |
1735 |
spin_unlock(ptl); |
71e3aac07 thp: transparent ... |
1736 |
} |
71e3aac07 thp: transparent ... |
1737 1738 1739 |
return ret; } |
5a505085f mm/rmap: Convert ... |
1740 |
/* must be called with anon_vma->root->rwsem held */ |
71e3aac07 thp: transparent ... |
1741 |
static void __split_huge_page(struct page *page, |
5bc7b8aca mm: thp: add spli... |
1742 1743 |
struct anon_vma *anon_vma, struct list_head *list) |
71e3aac07 thp: transparent ... |
1744 1745 |
{ int mapcount, mapcount2; |
bf181b9f9 mm anon rmap: rep... |
1746 |
pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT); |
71e3aac07 thp: transparent ... |
1747 1748 1749 1750 1751 1752 |
struct anon_vma_chain *avc; BUG_ON(!PageHead(page)); BUG_ON(PageTail(page)); mapcount = 0; |
bf181b9f9 mm anon rmap: rep... |
1753 |
anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) { |
71e3aac07 thp: transparent ... |
1754 1755 1756 |
struct vm_area_struct *vma = avc->vma; unsigned long addr = vma_address(page, vma); BUG_ON(is_vma_temporary_stack(vma)); |
71e3aac07 thp: transparent ... |
1757 1758 |
mapcount += __split_huge_page_splitting(page, vma, addr); } |
05759d380 thp: split_huge_p... |
1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 |
/* * It is critical that new vmas are added to the tail of the * anon_vma list. This guarantes that if copy_huge_pmd() runs * and establishes a child pmd before * __split_huge_page_splitting() freezes the parent pmd (so if * we fail to prevent copy_huge_pmd() from running until the * whole __split_huge_page() is complete), we will still see * the newly established pmd of the child later during the * walk, to be able to set it as pmd_trans_splitting too. */ if (mapcount != page_mapcount(page)) printk(KERN_ERR "mapcount %d page_mapcount %d ", mapcount, page_mapcount(page)); |
71e3aac07 thp: transparent ... |
1773 |
BUG_ON(mapcount != page_mapcount(page)); |
5bc7b8aca mm: thp: add spli... |
1774 |
__split_huge_page_refcount(page, list); |
71e3aac07 thp: transparent ... |
1775 1776 |
mapcount2 = 0; |
bf181b9f9 mm anon rmap: rep... |
1777 |
anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) { |
71e3aac07 thp: transparent ... |
1778 1779 1780 |
struct vm_area_struct *vma = avc->vma; unsigned long addr = vma_address(page, vma); BUG_ON(is_vma_temporary_stack(vma)); |
71e3aac07 thp: transparent ... |
1781 1782 |
mapcount2 += __split_huge_page_map(page, vma, addr); } |
05759d380 thp: split_huge_p... |
1783 1784 1785 1786 |
if (mapcount != mapcount2) printk(KERN_ERR "mapcount %d mapcount2 %d page_mapcount %d ", mapcount, mapcount2, page_mapcount(page)); |
71e3aac07 thp: transparent ... |
1787 1788 |
BUG_ON(mapcount != mapcount2); } |
5bc7b8aca mm: thp: add spli... |
1789 1790 1791 1792 1793 1794 1795 1796 |
/* * Split a hugepage into normal pages. This doesn't change the position of head * page. If @list is null, tail pages will be added to LRU list, otherwise, to * @list. Both head page and tail pages will inherit mapping, flags, and so on * from the hugepage. * Return 0 if the hugepage is split successfully otherwise return 1. */ int split_huge_page_to_list(struct page *page, struct list_head *list) |
71e3aac07 thp: transparent ... |
1797 1798 1799 |
{ struct anon_vma *anon_vma; int ret = 1; |
5918d10a4 thp: fix huge zer... |
1800 |
BUG_ON(is_huge_zero_page(page)); |
71e3aac07 thp: transparent ... |
1801 |
BUG_ON(!PageAnon(page)); |
062f1af21 mm: thp: acquire ... |
1802 1803 1804 1805 1806 1807 1808 1809 1810 |
/* * The caller does not necessarily hold an mmap_sem that would prevent * the anon_vma disappearing so we first we take a reference to it * and then lock the anon_vma for write. This is similar to * page_lock_anon_vma_read except the write lock is taken to serialise * against parallel split or collapse operations. */ anon_vma = page_get_anon_vma(page); |
71e3aac07 thp: transparent ... |
1811 1812 |
if (!anon_vma) goto out; |
062f1af21 mm: thp: acquire ... |
1813 |
anon_vma_lock_write(anon_vma); |
71e3aac07 thp: transparent ... |
1814 1815 1816 1817 1818 |
ret = 0; if (!PageCompound(page)) goto out_unlock; BUG_ON(!PageSwapBacked(page)); |
5bc7b8aca mm: thp: add spli... |
1819 |
__split_huge_page(page, anon_vma, list); |
81ab4201f mm: add VM counte... |
1820 |
count_vm_event(THP_SPLIT); |
71e3aac07 thp: transparent ... |
1821 1822 1823 |
BUG_ON(PageCompound(page)); out_unlock: |
08b52706d mm/rmap: rename a... |
1824 |
anon_vma_unlock_write(anon_vma); |
062f1af21 mm: thp: acquire ... |
1825 |
put_anon_vma(anon_vma); |
71e3aac07 thp: transparent ... |
1826 1827 1828 |
out: return ret; } |
9050d7eba mm: include VM_MI... |
1829 |
#define VM_NO_THP (VM_SPECIAL | VM_HUGETLB | VM_SHARED | VM_MAYSHARE) |
78f11a255 mm: thp: fix /dev... |
1830 |
|
60ab3244e thp: khugepaged: ... |
1831 1832 |
int hugepage_madvise(struct vm_area_struct *vma, unsigned long *vm_flags, int advice) |
0af4e98b6 thp: madvise(MADV... |
1833 |
{ |
a664b2d85 thp: madvise(MADV... |
1834 1835 |
switch (advice) { case MADV_HUGEPAGE: |
1e1836e84 mm: revert "thp: ... |
1836 1837 1838 1839 1840 1841 1842 1843 1844 |
#ifdef CONFIG_S390 /* * qemu blindly sets MADV_HUGEPAGE on all allocations, but s390 * can't handle this properly after s390_enable_sie, so we simply * ignore the madvise to prevent qemu from causing a SIGSEGV. */ if (mm_has_pgste(vma->vm_mm)) return 0; #endif |
a664b2d85 thp: madvise(MADV... |
1845 1846 1847 |
/* * Be somewhat over-protective like KSM for now! */ |
78f11a255 mm: thp: fix /dev... |
1848 |
if (*vm_flags & (VM_HUGEPAGE | VM_NO_THP)) |
a664b2d85 thp: madvise(MADV... |
1849 1850 1851 |
return -EINVAL; *vm_flags &= ~VM_NOHUGEPAGE; *vm_flags |= VM_HUGEPAGE; |
60ab3244e thp: khugepaged: ... |
1852 1853 1854 1855 1856 1857 1858 |
/* * If the vma become good for khugepaged to scan, * register it here without waiting a page fault that * may not happen any time soon. */ if (unlikely(khugepaged_enter_vma_merge(vma))) return -ENOMEM; |
a664b2d85 thp: madvise(MADV... |
1859 1860 1861 1862 1863 |
break; case MADV_NOHUGEPAGE: /* * Be somewhat over-protective like KSM for now! */ |
78f11a255 mm: thp: fix /dev... |
1864 |
if (*vm_flags & (VM_NOHUGEPAGE | VM_NO_THP)) |
a664b2d85 thp: madvise(MADV... |
1865 1866 1867 |
return -EINVAL; *vm_flags &= ~VM_HUGEPAGE; *vm_flags |= VM_NOHUGEPAGE; |
60ab3244e thp: khugepaged: ... |
1868 1869 1870 1871 1872 |
/* * Setting VM_NOHUGEPAGE will prevent khugepaged from scanning * this vma even if we leave the mm registered in khugepaged if * it got registered before VM_NOHUGEPAGE was set. */ |
a664b2d85 thp: madvise(MADV... |
1873 1874 |
break; } |
0af4e98b6 thp: madvise(MADV... |
1875 1876 1877 |
return 0; } |
ba76149f4 thp: khugepaged |
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 |
static int __init khugepaged_slab_init(void) { mm_slot_cache = kmem_cache_create("khugepaged_mm_slot", sizeof(struct mm_slot), __alignof__(struct mm_slot), 0, NULL); if (!mm_slot_cache) return -ENOMEM; return 0; } |
ba76149f4 thp: khugepaged |
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 |
static inline struct mm_slot *alloc_mm_slot(void) { if (!mm_slot_cache) /* initialization failed */ return NULL; return kmem_cache_zalloc(mm_slot_cache, GFP_KERNEL); } static inline void free_mm_slot(struct mm_slot *mm_slot) { kmem_cache_free(mm_slot_cache, mm_slot); } |
ba76149f4 thp: khugepaged |
1899 1900 1901 |
static struct mm_slot *get_mm_slot(struct mm_struct *mm) { struct mm_slot *mm_slot; |
ba76149f4 thp: khugepaged |
1902 |
|
b67bfe0d4 hlist: drop the n... |
1903 |
hash_for_each_possible(mm_slots_hash, mm_slot, hash, (unsigned long)mm) |
ba76149f4 thp: khugepaged |
1904 1905 |
if (mm == mm_slot->mm) return mm_slot; |
43b5fbbd2 mm/huge_memory.c:... |
1906 |
|
ba76149f4 thp: khugepaged |
1907 1908 1909 1910 1911 1912 |
return NULL; } static void insert_to_mm_slots_hash(struct mm_struct *mm, struct mm_slot *mm_slot) { |
ba76149f4 thp: khugepaged |
1913 |
mm_slot->mm = mm; |
43b5fbbd2 mm/huge_memory.c:... |
1914 |
hash_add(mm_slots_hash, &mm_slot->hash, (long)mm); |
ba76149f4 thp: khugepaged |
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 |
} static inline int khugepaged_test_exit(struct mm_struct *mm) { return atomic_read(&mm->mm_users) == 0; } int __khugepaged_enter(struct mm_struct *mm) { struct mm_slot *mm_slot; int wakeup; mm_slot = alloc_mm_slot(); if (!mm_slot) return -ENOMEM; /* __khugepaged_exit() must not run from under us */ VM_BUG_ON(khugepaged_test_exit(mm)); if (unlikely(test_and_set_bit(MMF_VM_HUGEPAGE, &mm->flags))) { free_mm_slot(mm_slot); return 0; } spin_lock(&khugepaged_mm_lock); insert_to_mm_slots_hash(mm, mm_slot); /* * Insert just behind the scanning cursor, to let the area settle * down a little. */ wakeup = list_empty(&khugepaged_scan.mm_head); list_add_tail(&mm_slot->mm_node, &khugepaged_scan.mm_head); spin_unlock(&khugepaged_mm_lock); atomic_inc(&mm->mm_count); if (wakeup) wake_up_interruptible(&khugepaged_wait); return 0; } int khugepaged_enter_vma_merge(struct vm_area_struct *vma) { unsigned long hstart, hend; if (!vma->anon_vma) /* * Not yet faulted in so we will register later in the * page fault if needed. */ return 0; |
78f11a255 mm: thp: fix /dev... |
1964 |
if (vma->vm_ops) |
ba76149f4 thp: khugepaged |
1965 1966 |
/* khugepaged not yet working on file or special mappings */ return 0; |
b3b9c2932 mm, x86, pat: rew... |
1967 |
VM_BUG_ON(vma->vm_flags & VM_NO_THP); |
ba76149f4 thp: khugepaged |
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 |
hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK; hend = vma->vm_end & HPAGE_PMD_MASK; if (hstart < hend) return khugepaged_enter(vma); return 0; } void __khugepaged_exit(struct mm_struct *mm) { struct mm_slot *mm_slot; int free = 0; spin_lock(&khugepaged_mm_lock); mm_slot = get_mm_slot(mm); if (mm_slot && khugepaged_scan.mm_slot != mm_slot) { |
43b5fbbd2 mm/huge_memory.c:... |
1983 |
hash_del(&mm_slot->hash); |
ba76149f4 thp: khugepaged |
1984 1985 1986 |
list_del(&mm_slot->mm_node); free = 1; } |
d788e80a8 mm/huge_memory.c:... |
1987 |
spin_unlock(&khugepaged_mm_lock); |
ba76149f4 thp: khugepaged |
1988 1989 |
if (free) { |
ba76149f4 thp: khugepaged |
1990 1991 1992 1993 |
clear_bit(MMF_VM_HUGEPAGE, &mm->flags); free_mm_slot(mm_slot); mmdrop(mm); } else if (mm_slot) { |
ba76149f4 thp: khugepaged |
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 |
/* * This is required to serialize against * khugepaged_test_exit() (which is guaranteed to run * under mmap sem read mode). Stop here (after we * return all pagetables will be destroyed) until * khugepaged has finished working on the pagetables * under the mmap_sem. */ down_write(&mm->mmap_sem); up_write(&mm->mmap_sem); |
d788e80a8 mm/huge_memory.c:... |
2004 |
} |
ba76149f4 thp: khugepaged |
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 |
} static void release_pte_page(struct page *page) { /* 0 stands for page_is_file_cache(page) == false */ dec_zone_page_state(page, NR_ISOLATED_ANON + 0); unlock_page(page); putback_lru_page(page); } static void release_pte_pages(pte_t *pte, pte_t *_pte) { while (--_pte >= pte) { pte_t pteval = *_pte; if (!pte_none(pteval)) release_pte_page(pte_page(pteval)); } } |
ba76149f4 thp: khugepaged |
2023 2024 2025 2026 2027 2028 |
static int __collapse_huge_page_isolate(struct vm_area_struct *vma, unsigned long address, pte_t *pte) { struct page *page; pte_t *_pte; |
344aa35c2 thp: clean up __c... |
2029 |
int referenced = 0, none = 0; |
ba76149f4 thp: khugepaged |
2030 2031 2032 2033 2034 2035 |
for (_pte = pte; _pte < pte+HPAGE_PMD_NR; _pte++, address += PAGE_SIZE) { pte_t pteval = *_pte; if (pte_none(pteval)) { if (++none <= khugepaged_max_ptes_none) continue; |
344aa35c2 thp: clean up __c... |
2036 |
else |
ba76149f4 thp: khugepaged |
2037 |
goto out; |
ba76149f4 thp: khugepaged |
2038 |
} |
344aa35c2 thp: clean up __c... |
2039 |
if (!pte_present(pteval) || !pte_write(pteval)) |
ba76149f4 thp: khugepaged |
2040 |
goto out; |
ba76149f4 thp: khugepaged |
2041 |
page = vm_normal_page(vma, address, pteval); |
344aa35c2 thp: clean up __c... |
2042 |
if (unlikely(!page)) |
ba76149f4 thp: khugepaged |
2043 |
goto out; |
344aa35c2 thp: clean up __c... |
2044 |
|
309381fea mm: dump page whe... |
2045 2046 2047 |
VM_BUG_ON_PAGE(PageCompound(page), page); VM_BUG_ON_PAGE(!PageAnon(page), page); VM_BUG_ON_PAGE(!PageSwapBacked(page), page); |
ba76149f4 thp: khugepaged |
2048 2049 |
/* cannot use mapcount: can't collapse if there's a gup pin */ |
344aa35c2 thp: clean up __c... |
2050 |
if (page_count(page) != 1) |
ba76149f4 thp: khugepaged |
2051 |
goto out; |
ba76149f4 thp: khugepaged |
2052 2053 2054 2055 2056 2057 |
/* * We can do it before isolate_lru_page because the * page can't be freed from under us. NOTE: PG_lock * is needed to serialize against split_huge_page * when invoked from the VM. */ |
344aa35c2 thp: clean up __c... |
2058 |
if (!trylock_page(page)) |
ba76149f4 thp: khugepaged |
2059 |
goto out; |
ba76149f4 thp: khugepaged |
2060 2061 2062 2063 2064 2065 |
/* * Isolate the page to avoid collapsing an hugepage * currently in use by the VM. */ if (isolate_lru_page(page)) { unlock_page(page); |
ba76149f4 thp: khugepaged |
2066 2067 2068 2069 |
goto out; } /* 0 stands for page_is_file_cache(page) == false */ inc_zone_page_state(page, NR_ISOLATED_ANON + 0); |
309381fea mm: dump page whe... |
2070 2071 |
VM_BUG_ON_PAGE(!PageLocked(page), page); VM_BUG_ON_PAGE(PageLRU(page), page); |
ba76149f4 thp: khugepaged |
2072 2073 |
/* If there is no mapped pte young don't collapse the page */ |
8ee53820e thp: mmu_notifier... |
2074 2075 |
if (pte_young(pteval) || PageReferenced(page) || mmu_notifier_test_young(vma->vm_mm, address)) |
ba76149f4 thp: khugepaged |
2076 2077 |
referenced = 1; } |
344aa35c2 thp: clean up __c... |
2078 2079 |
if (likely(referenced)) return 1; |
ba76149f4 thp: khugepaged |
2080 |
out: |
344aa35c2 thp: clean up __c... |
2081 2082 |
release_pte_pages(pte, _pte); return 0; |
ba76149f4 thp: khugepaged |
2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 |
} static void __collapse_huge_page_copy(pte_t *pte, struct page *page, struct vm_area_struct *vma, unsigned long address, spinlock_t *ptl) { pte_t *_pte; for (_pte = pte; _pte < pte+HPAGE_PMD_NR; _pte++) { pte_t pteval = *_pte; struct page *src_page; if (pte_none(pteval)) { clear_user_highpage(page, address); add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1); } else { src_page = pte_page(pteval); copy_user_highpage(page, src_page, address, vma); |
309381fea mm: dump page whe... |
2101 |
VM_BUG_ON_PAGE(page_mapcount(src_page) != 1, src_page); |
ba76149f4 thp: khugepaged |
2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 |
release_pte_page(src_page); /* * ptl mostly unnecessary, but preempt has to * be disabled to update the per-cpu stats * inside page_remove_rmap(). */ spin_lock(ptl); /* * paravirt calls inside pte_clear here are * superfluous. */ pte_clear(vma->vm_mm, address, _pte); page_remove_rmap(src_page); spin_unlock(ptl); free_page_and_swap_cache(src_page); } address += PAGE_SIZE; page++; } } |
26234f36e thp: introduce kh... |
2123 |
static void khugepaged_alloc_sleep(void) |
ba76149f4 thp: khugepaged |
2124 |
{ |
26234f36e thp: introduce kh... |
2125 2126 2127 |
wait_event_freezable_timeout(khugepaged_wait, false, msecs_to_jiffies(khugepaged_alloc_sleep_millisecs)); } |
ba76149f4 thp: khugepaged |
2128 |
|
9f1b868a1 mm: thp: khugepag... |
2129 |
static int khugepaged_node_load[MAX_NUMNODES]; |
26234f36e thp: introduce kh... |
2130 |
#ifdef CONFIG_NUMA |
9f1b868a1 mm: thp: khugepag... |
2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 |
static int khugepaged_find_target_node(void) { static int last_khugepaged_target_node = NUMA_NO_NODE; int nid, target_node = 0, max_value = 0; /* find first node with max normal pages hit */ for (nid = 0; nid < MAX_NUMNODES; nid++) if (khugepaged_node_load[nid] > max_value) { max_value = khugepaged_node_load[nid]; target_node = nid; } /* do some balance if several nodes have the same hit record */ if (target_node <= last_khugepaged_target_node) for (nid = last_khugepaged_target_node + 1; nid < MAX_NUMNODES; nid++) if (max_value == khugepaged_node_load[nid]) { target_node = nid; break; } last_khugepaged_target_node = target_node; return target_node; } |
26234f36e thp: introduce kh... |
2155 2156 2157 2158 2159 2160 2161 |
static bool khugepaged_prealloc_page(struct page **hpage, bool *wait) { if (IS_ERR(*hpage)) { if (!*wait) return false; *wait = false; |
e3b4126c5 thp: khugepaged_p... |
2162 |
*hpage = NULL; |
26234f36e thp: introduce kh... |
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 |
khugepaged_alloc_sleep(); } else if (*hpage) { put_page(*hpage); *hpage = NULL; } return true; } static struct page *khugepaged_alloc_page(struct page **hpage, struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, int node) { |
309381fea mm: dump page whe... |
2177 |
VM_BUG_ON_PAGE(*hpage, *hpage); |
ce83d2174 thp: allocate mem... |
2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 |
/* * Allocate the page while the vma is still valid and under * the mmap_sem read mode so there is no memory allocation * later when we take the mmap_sem in write mode. This is more * friendly behavior (OTOH it may actually hide bugs) to * filesystems in userland with daemons allocating memory in * the userland I/O paths. Allocating memory with the * mmap_sem in read mode is good idea also to allow greater * scalability. */ |
9f1b868a1 mm: thp: khugepag... |
2188 2189 |
*hpage = alloc_pages_exact_node(node, alloc_hugepage_gfpmask( khugepaged_defrag(), __GFP_OTHER_NODE), HPAGE_PMD_ORDER); |
692e0b354 mm: thp: optimize... |
2190 2191 2192 2193 2194 |
/* * After allocating the hugepage, release the mmap_sem read lock in * preparation for taking it in write mode. */ up_read(&mm->mmap_sem); |
26234f36e thp: introduce kh... |
2195 |
if (unlikely(!*hpage)) { |
81ab4201f mm: add VM counte... |
2196 |
count_vm_event(THP_COLLAPSE_ALLOC_FAILED); |
ce83d2174 thp: allocate mem... |
2197 |
*hpage = ERR_PTR(-ENOMEM); |
26234f36e thp: introduce kh... |
2198 |
return NULL; |
ce83d2174 thp: allocate mem... |
2199 |
} |
26234f36e thp: introduce kh... |
2200 |
|
65b3c07b4 thp: fix the coun... |
2201 |
count_vm_event(THP_COLLAPSE_ALLOC); |
26234f36e thp: introduce kh... |
2202 2203 2204 |
return *hpage; } #else |
9f1b868a1 mm: thp: khugepag... |
2205 2206 2207 2208 |
static int khugepaged_find_target_node(void) { return 0; } |
10dc4155c mm: thp: cleanup:... |
2209 2210 2211 2212 2213 |
static inline struct page *alloc_hugepage(int defrag) { return alloc_pages(alloc_hugepage_gfpmask(defrag, 0), HPAGE_PMD_ORDER); } |
26234f36e thp: introduce kh... |
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 |
static struct page *khugepaged_alloc_hugepage(bool *wait) { struct page *hpage; do { hpage = alloc_hugepage(khugepaged_defrag()); if (!hpage) { count_vm_event(THP_COLLAPSE_ALLOC_FAILED); if (!*wait) return NULL; *wait = false; khugepaged_alloc_sleep(); } else count_vm_event(THP_COLLAPSE_ALLOC); } while (unlikely(!hpage) && likely(khugepaged_enabled())); return hpage; } static bool khugepaged_prealloc_page(struct page **hpage, bool *wait) { if (!*hpage) *hpage = khugepaged_alloc_hugepage(wait); if (unlikely(!*hpage)) return false; return true; } static struct page *khugepaged_alloc_page(struct page **hpage, struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, int node) { up_read(&mm->mmap_sem); VM_BUG_ON(!*hpage); return *hpage; } |
692e0b354 mm: thp: optimize... |
2254 |
#endif |
fa475e517 thp: introduce hu... |
2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 |
static bool hugepage_vma_check(struct vm_area_struct *vma) { if ((!(vma->vm_flags & VM_HUGEPAGE) && !khugepaged_always()) || (vma->vm_flags & VM_NOHUGEPAGE)) return false; if (!vma->anon_vma || vma->vm_ops) return false; if (is_vma_temporary_stack(vma)) return false; VM_BUG_ON(vma->vm_flags & VM_NO_THP); return true; } |
26234f36e thp: introduce kh... |
2268 2269 2270 2271 2272 2273 |
static void collapse_huge_page(struct mm_struct *mm, unsigned long address, struct page **hpage, struct vm_area_struct *vma, int node) { |
26234f36e thp: introduce kh... |
2274 2275 2276 2277 |
pmd_t *pmd, _pmd; pte_t *pte; pgtable_t pgtable; struct page *new_page; |
c4088ebdc mm: convert the r... |
2278 |
spinlock_t *pmd_ptl, *pte_ptl; |
26234f36e thp: introduce kh... |
2279 2280 |
int isolated; unsigned long hstart, hend; |
2ec74c3ef mm: move all mmu ... |
2281 2282 |
unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
26234f36e thp: introduce kh... |
2283 2284 2285 2286 2287 2288 2289 |
VM_BUG_ON(address & ~HPAGE_PMD_MASK); /* release the mmap_sem read lock. */ new_page = khugepaged_alloc_page(hpage, mm, vma, address, node); if (!new_page) return; |
d715ae08f memcg: rename hig... |
2290 |
if (unlikely(mem_cgroup_charge_anon(new_page, mm, GFP_KERNEL))) |
ce83d2174 thp: allocate mem... |
2291 |
return; |
ba76149f4 thp: khugepaged |
2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 |
/* * Prevent all access to pagetables with the exception of * gup_fast later hanlded by the ptep_clear_flush and the VM * handled by the anon_vma lock + PG_lock. */ down_write(&mm->mmap_sem); if (unlikely(khugepaged_test_exit(mm))) goto out; vma = find_vma(mm, address); |
a8f531ebc mm/huge_memory.c:... |
2303 2304 |
if (!vma) goto out; |
ba76149f4 thp: khugepaged |
2305 2306 2307 2308 |
hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK; hend = vma->vm_end & HPAGE_PMD_MASK; if (address < hstart || address + HPAGE_PMD_SIZE > hend) goto out; |
fa475e517 thp: introduce hu... |
2309 |
if (!hugepage_vma_check(vma)) |
a7d6e4ecd thp: prevent huge... |
2310 |
goto out; |
6219049ae mm: introduce mm_... |
2311 2312 |
pmd = mm_find_pmd(mm, address); if (!pmd) |
ba76149f4 thp: khugepaged |
2313 |
goto out; |
6219049ae mm: introduce mm_... |
2314 |
if (pmd_trans_huge(*pmd)) |
ba76149f4 thp: khugepaged |
2315 |
goto out; |
4fc3f1d66 mm/rmap, migratio... |
2316 |
anon_vma_lock_write(vma->anon_vma); |
ba76149f4 thp: khugepaged |
2317 2318 |
pte = pte_offset_map(pmd, address); |
c4088ebdc mm: convert the r... |
2319 |
pte_ptl = pte_lockptr(mm, pmd); |
ba76149f4 thp: khugepaged |
2320 |
|
2ec74c3ef mm: move all mmu ... |
2321 2322 2323 |
mmun_start = address; mmun_end = address + HPAGE_PMD_SIZE; mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); |
c4088ebdc mm: convert the r... |
2324 |
pmd_ptl = pmd_lock(mm, pmd); /* probably unnecessary */ |
ba76149f4 thp: khugepaged |
2325 2326 2327 2328 2329 2330 |
/* * After this gup_fast can't run anymore. This also removes * any huge TLB entry from the CPU so we won't allow * huge and small TLB entries for the same virtual address * to avoid the risk of CPU bugs in that area. */ |
2ec74c3ef mm: move all mmu ... |
2331 |
_pmd = pmdp_clear_flush(vma, address, pmd); |
c4088ebdc mm: convert the r... |
2332 |
spin_unlock(pmd_ptl); |
2ec74c3ef mm: move all mmu ... |
2333 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
ba76149f4 thp: khugepaged |
2334 |
|
c4088ebdc mm: convert the r... |
2335 |
spin_lock(pte_ptl); |
ba76149f4 thp: khugepaged |
2336 |
isolated = __collapse_huge_page_isolate(vma, address, pte); |
c4088ebdc mm: convert the r... |
2337 |
spin_unlock(pte_ptl); |
ba76149f4 thp: khugepaged |
2338 2339 |
if (unlikely(!isolated)) { |
453c71926 thp: keep highpte... |
2340 |
pte_unmap(pte); |
c4088ebdc mm: convert the r... |
2341 |
spin_lock(pmd_ptl); |
ba76149f4 thp: khugepaged |
2342 |
BUG_ON(!pmd_none(*pmd)); |
7c3425123 mm/THP: use pmd_p... |
2343 2344 2345 2346 2347 2348 |
/* * We can only use set_pmd_at when establishing * hugepmds and never for establishing regular pmds that * points to regular pagetables. Use pmd_populate for that */ pmd_populate(mm, pmd, pmd_pgtable(_pmd)); |
c4088ebdc mm: convert the r... |
2349 |
spin_unlock(pmd_ptl); |
08b52706d mm/rmap: rename a... |
2350 |
anon_vma_unlock_write(vma->anon_vma); |
ce83d2174 thp: allocate mem... |
2351 |
goto out; |
ba76149f4 thp: khugepaged |
2352 2353 2354 2355 2356 2357 |
} /* * All pages are isolated and locked so anon_vma rmap * can't run anymore. */ |
08b52706d mm/rmap: rename a... |
2358 |
anon_vma_unlock_write(vma->anon_vma); |
ba76149f4 thp: khugepaged |
2359 |
|
c4088ebdc mm: convert the r... |
2360 |
__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl); |
453c71926 thp: keep highpte... |
2361 |
pte_unmap(pte); |
ba76149f4 thp: khugepaged |
2362 2363 |
__SetPageUptodate(new_page); pgtable = pmd_pgtable(_pmd); |
ba76149f4 thp: khugepaged |
2364 |
|
3122359a6 thp: move maybe_p... |
2365 2366 |
_pmd = mk_huge_pmd(new_page, vma->vm_page_prot); _pmd = maybe_pmd_mkwrite(pmd_mkdirty(_pmd), vma); |
ba76149f4 thp: khugepaged |
2367 2368 2369 2370 2371 2372 2373 |
/* * spin_lock() below is not the equivalent of smp_wmb(), so * this is needed to avoid the copy_huge_page writes to become * visible after the set_pmd_at() write. */ smp_wmb(); |
c4088ebdc mm: convert the r... |
2374 |
spin_lock(pmd_ptl); |
ba76149f4 thp: khugepaged |
2375 2376 |
BUG_ON(!pmd_none(*pmd)); page_add_new_anon_rmap(new_page, vma, address); |
fce144b47 mm/THP: deposit t... |
2377 |
pgtable_trans_huge_deposit(mm, pmd, pgtable); |
ba76149f4 thp: khugepaged |
2378 |
set_pmd_at(mm, address, pmd, _pmd); |
b113da657 mm: Add and use u... |
2379 |
update_mmu_cache_pmd(vma, address, pmd); |
c4088ebdc mm: convert the r... |
2380 |
spin_unlock(pmd_ptl); |
ba76149f4 thp: khugepaged |
2381 2382 |
*hpage = NULL; |
420256ef0 thp: release page... |
2383 |
|
ba76149f4 thp: khugepaged |
2384 |
khugepaged_pages_collapsed++; |
ce83d2174 thp: allocate mem... |
2385 |
out_up_write: |
ba76149f4 thp: khugepaged |
2386 |
up_write(&mm->mmap_sem); |
0bbbc0b33 thp: add numa awa... |
2387 |
return; |
ce83d2174 thp: allocate mem... |
2388 |
out: |
678ff896a memcg: fix leak o... |
2389 |
mem_cgroup_uncharge_page(new_page); |
ce83d2174 thp: allocate mem... |
2390 |
goto out_up_write; |
ba76149f4 thp: khugepaged |
2391 2392 2393 2394 2395 2396 2397 |
} static int khugepaged_scan_pmd(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long address, struct page **hpage) { |
ba76149f4 thp: khugepaged |
2398 2399 2400 2401 2402 2403 |
pmd_t *pmd; pte_t *pte, *_pte; int ret = 0, referenced = 0, none = 0; struct page *page; unsigned long _address; spinlock_t *ptl; |
00ef2d2f8 mm: use NUMA_NO_NODE |
2404 |
int node = NUMA_NO_NODE; |
ba76149f4 thp: khugepaged |
2405 2406 |
VM_BUG_ON(address & ~HPAGE_PMD_MASK); |
6219049ae mm: introduce mm_... |
2407 2408 |
pmd = mm_find_pmd(mm, address); if (!pmd) |
ba76149f4 thp: khugepaged |
2409 |
goto out; |
6219049ae mm: introduce mm_... |
2410 |
if (pmd_trans_huge(*pmd)) |
ba76149f4 thp: khugepaged |
2411 |
goto out; |
9f1b868a1 mm: thp: khugepag... |
2412 |
memset(khugepaged_node_load, 0, sizeof(khugepaged_node_load)); |
ba76149f4 thp: khugepaged |
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 |
pte = pte_offset_map_lock(mm, pmd, address, &ptl); for (_address = address, _pte = pte; _pte < pte+HPAGE_PMD_NR; _pte++, _address += PAGE_SIZE) { pte_t pteval = *_pte; if (pte_none(pteval)) { if (++none <= khugepaged_max_ptes_none) continue; else goto out_unmap; } if (!pte_present(pteval) || !pte_write(pteval)) goto out_unmap; page = vm_normal_page(vma, _address, pteval); if (unlikely(!page)) goto out_unmap; |
5c4b4be3b mm: use correct n... |
2428 |
/* |
9f1b868a1 mm: thp: khugepag... |
2429 2430 2431 2432 |
* Record which node the original page is from and save this * information to khugepaged_node_load[]. * Khupaged will allocate hugepage from the node has the max * hit record. |
5c4b4be3b mm: use correct n... |
2433 |
*/ |
9f1b868a1 mm: thp: khugepag... |
2434 2435 |
node = page_to_nid(page); khugepaged_node_load[node]++; |
309381fea mm: dump page whe... |
2436 |
VM_BUG_ON_PAGE(PageCompound(page), page); |
ba76149f4 thp: khugepaged |
2437 2438 2439 2440 2441 |
if (!PageLRU(page) || PageLocked(page) || !PageAnon(page)) goto out_unmap; /* cannot use mapcount: can't collapse if there's a gup pin */ if (page_count(page) != 1) goto out_unmap; |
8ee53820e thp: mmu_notifier... |
2442 2443 |
if (pte_young(pteval) || PageReferenced(page) || mmu_notifier_test_young(vma->vm_mm, address)) |
ba76149f4 thp: khugepaged |
2444 2445 2446 2447 2448 2449 |
referenced = 1; } if (referenced) ret = 1; out_unmap: pte_unmap_unlock(pte, ptl); |
9f1b868a1 mm: thp: khugepag... |
2450 2451 |
if (ret) { node = khugepaged_find_target_node(); |
ce83d2174 thp: allocate mem... |
2452 |
/* collapse_huge_page will return with the mmap_sem released */ |
5c4b4be3b mm: use correct n... |
2453 |
collapse_huge_page(mm, address, hpage, vma, node); |
9f1b868a1 mm: thp: khugepag... |
2454 |
} |
ba76149f4 thp: khugepaged |
2455 2456 2457 2458 2459 2460 2461 |
out: return ret; } static void collect_mm_slot(struct mm_slot *mm_slot) { struct mm_struct *mm = mm_slot->mm; |
b9980cdcf mm: fix UP THP sp... |
2462 |
VM_BUG_ON(NR_CPUS != 1 && !spin_is_locked(&khugepaged_mm_lock)); |
ba76149f4 thp: khugepaged |
2463 2464 2465 |
if (khugepaged_test_exit(mm)) { /* free mm_slot */ |
43b5fbbd2 mm/huge_memory.c:... |
2466 |
hash_del(&mm_slot->hash); |
ba76149f4 thp: khugepaged |
2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 |
list_del(&mm_slot->mm_node); /* * Not strictly needed because the mm exited already. * * clear_bit(MMF_VM_HUGEPAGE, &mm->flags); */ /* khugepaged_mm_lock actually not necessary for the below */ free_mm_slot(mm_slot); mmdrop(mm); } } static unsigned int khugepaged_scan_mm_slot(unsigned int pages, struct page **hpage) |
2f1da6421 mm/huge_memory.c:... |
2483 2484 |
__releases(&khugepaged_mm_lock) __acquires(&khugepaged_mm_lock) |
ba76149f4 thp: khugepaged |
2485 2486 2487 2488 2489 2490 2491 |
{ struct mm_slot *mm_slot; struct mm_struct *mm; struct vm_area_struct *vma; int progress = 0; VM_BUG_ON(!pages); |
b9980cdcf mm: fix UP THP sp... |
2492 |
VM_BUG_ON(NR_CPUS != 1 && !spin_is_locked(&khugepaged_mm_lock)); |
ba76149f4 thp: khugepaged |
2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 |
if (khugepaged_scan.mm_slot) mm_slot = khugepaged_scan.mm_slot; else { mm_slot = list_entry(khugepaged_scan.mm_head.next, struct mm_slot, mm_node); khugepaged_scan.address = 0; khugepaged_scan.mm_slot = mm_slot; } spin_unlock(&khugepaged_mm_lock); mm = mm_slot->mm; down_read(&mm->mmap_sem); if (unlikely(khugepaged_test_exit(mm))) vma = NULL; else vma = find_vma(mm, khugepaged_scan.address); progress++; for (; vma; vma = vma->vm_next) { unsigned long hstart, hend; cond_resched(); if (unlikely(khugepaged_test_exit(mm))) { progress++; break; } |
fa475e517 thp: introduce hu... |
2520 2521 |
if (!hugepage_vma_check(vma)) { skip: |
ba76149f4 thp: khugepaged |
2522 2523 2524 |
progress++; continue; } |
ba76149f4 thp: khugepaged |
2525 2526 |
hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK; hend = vma->vm_end & HPAGE_PMD_MASK; |
a7d6e4ecd thp: prevent huge... |
2527 2528 2529 2530 |
if (hstart >= hend) goto skip; if (khugepaged_scan.address > hend) goto skip; |
ba76149f4 thp: khugepaged |
2531 2532 |
if (khugepaged_scan.address < hstart) khugepaged_scan.address = hstart; |
a7d6e4ecd thp: prevent huge... |
2533 |
VM_BUG_ON(khugepaged_scan.address & ~HPAGE_PMD_MASK); |
ba76149f4 thp: khugepaged |
2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 |
while (khugepaged_scan.address < hend) { int ret; cond_resched(); if (unlikely(khugepaged_test_exit(mm))) goto breakouterloop; VM_BUG_ON(khugepaged_scan.address < hstart || khugepaged_scan.address + HPAGE_PMD_SIZE > hend); ret = khugepaged_scan_pmd(mm, vma, khugepaged_scan.address, hpage); /* move to next address */ khugepaged_scan.address += HPAGE_PMD_SIZE; progress += HPAGE_PMD_NR; if (ret) /* we released mmap_sem so break loop */ goto breakouterloop_mmap_sem; if (progress >= pages) goto breakouterloop; } } breakouterloop: up_read(&mm->mmap_sem); /* exit_mmap will destroy ptes after this */ breakouterloop_mmap_sem: spin_lock(&khugepaged_mm_lock); |
a7d6e4ecd thp: prevent huge... |
2562 |
VM_BUG_ON(khugepaged_scan.mm_slot != mm_slot); |
ba76149f4 thp: khugepaged |
2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 |
/* * Release the current mm_slot if this mm is about to die, or * if we scanned all vmas of this mm. */ if (khugepaged_test_exit(mm) || !vma) { /* * Make sure that if mm_users is reaching zero while * khugepaged runs here, khugepaged_exit will find * mm_slot not pointing to the exiting mm. */ if (mm_slot->mm_node.next != &khugepaged_scan.mm_head) { khugepaged_scan.mm_slot = list_entry( mm_slot->mm_node.next, struct mm_slot, mm_node); khugepaged_scan.address = 0; } else { khugepaged_scan.mm_slot = NULL; khugepaged_full_scans++; } collect_mm_slot(mm_slot); } return progress; } static int khugepaged_has_work(void) { return !list_empty(&khugepaged_scan.mm_head) && khugepaged_enabled(); } static int khugepaged_wait_event(void) { return !list_empty(&khugepaged_scan.mm_head) || |
2017c0bff thp: remove wake_... |
2598 |
kthread_should_stop(); |
ba76149f4 thp: khugepaged |
2599 |
} |
d516904bd thp: merge page p... |
2600 |
static void khugepaged_do_scan(void) |
ba76149f4 thp: khugepaged |
2601 |
{ |
d516904bd thp: merge page p... |
2602 |
struct page *hpage = NULL; |
ba76149f4 thp: khugepaged |
2603 2604 |
unsigned int progress = 0, pass_through_head = 0; unsigned int pages = khugepaged_pages_to_scan; |
d516904bd thp: merge page p... |
2605 |
bool wait = true; |
ba76149f4 thp: khugepaged |
2606 2607 2608 2609 |
barrier(); /* write khugepaged_pages_to_scan to local stack */ while (progress < pages) { |
26234f36e thp: introduce kh... |
2610 |
if (!khugepaged_prealloc_page(&hpage, &wait)) |
d516904bd thp: merge page p... |
2611 |
break; |
26234f36e thp: introduce kh... |
2612 |
|
420256ef0 thp: release page... |
2613 |
cond_resched(); |
ba76149f4 thp: khugepaged |
2614 |
|
878aee7d6 thp: freeze khuge... |
2615 2616 |
if (unlikely(kthread_should_stop() || freezing(current))) break; |
ba76149f4 thp: khugepaged |
2617 2618 2619 2620 2621 2622 |
spin_lock(&khugepaged_mm_lock); if (!khugepaged_scan.mm_slot) pass_through_head++; if (khugepaged_has_work() && pass_through_head < 2) progress += khugepaged_scan_mm_slot(pages - progress, |
d516904bd thp: merge page p... |
2623 |
&hpage); |
ba76149f4 thp: khugepaged |
2624 2625 2626 2627 |
else progress = pages; spin_unlock(&khugepaged_mm_lock); } |
ba76149f4 thp: khugepaged |
2628 |
|
d516904bd thp: merge page p... |
2629 2630 |
if (!IS_ERR_OR_NULL(hpage)) put_page(hpage); |
0bbbc0b33 thp: add numa awa... |
2631 |
} |
2017c0bff thp: remove wake_... |
2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 |
static void khugepaged_wait_work(void) { try_to_freeze(); if (khugepaged_has_work()) { if (!khugepaged_scan_sleep_millisecs) return; wait_event_freezable_timeout(khugepaged_wait, kthread_should_stop(), msecs_to_jiffies(khugepaged_scan_sleep_millisecs)); return; } if (khugepaged_enabled()) wait_event_freezable(khugepaged_wait, khugepaged_wait_event()); } |
ba76149f4 thp: khugepaged |
2649 2650 2651 |
static int khugepaged(void *none) { struct mm_slot *mm_slot; |
878aee7d6 thp: freeze khuge... |
2652 |
set_freezable(); |
ba76149f4 thp: khugepaged |
2653 |
set_user_nice(current, 19); |
b7231789b thp: remove khuge... |
2654 2655 2656 2657 |
while (!kthread_should_stop()) { khugepaged_do_scan(); khugepaged_wait_work(); } |
ba76149f4 thp: khugepaged |
2658 2659 2660 2661 2662 2663 2664 |
spin_lock(&khugepaged_mm_lock); mm_slot = khugepaged_scan.mm_slot; khugepaged_scan.mm_slot = NULL; if (mm_slot) collect_mm_slot(mm_slot); spin_unlock(&khugepaged_mm_lock); |
ba76149f4 thp: khugepaged |
2665 2666 |
return 0; } |
c5a647d09 thp: implement sp... |
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 |
static void __split_huge_zero_page_pmd(struct vm_area_struct *vma, unsigned long haddr, pmd_t *pmd) { struct mm_struct *mm = vma->vm_mm; pgtable_t pgtable; pmd_t _pmd; int i; pmdp_clear_flush(vma, haddr, pmd); /* leave pmd empty until pte is filled */ |
6b0b50b06 mm/THP: add pmd a... |
2677 |
pgtable = pgtable_trans_huge_withdraw(mm, pmd); |
c5a647d09 thp: implement sp... |
2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 |
pmd_populate(mm, &_pmd, pgtable); for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) { pte_t *pte, entry; entry = pfn_pte(my_zero_pfn(haddr), vma->vm_page_prot); entry = pte_mkspecial(entry); pte = pte_offset_map(&_pmd, haddr); VM_BUG_ON(!pte_none(*pte)); set_pte_at(mm, haddr, pte, entry); pte_unmap(pte); } smp_wmb(); /* make pte visible before pmd */ pmd_populate(mm, pmd, pgtable); |
97ae17497 thp: implement re... |
2691 |
put_huge_zero_page(); |
c5a647d09 thp: implement sp... |
2692 |
} |
e180377f1 thp: change split... |
2693 2694 |
void __split_huge_page_pmd(struct vm_area_struct *vma, unsigned long address, pmd_t *pmd) |
71e3aac07 thp: transparent ... |
2695 |
{ |
c4088ebdc mm: convert the r... |
2696 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
2697 |
struct page *page; |
e180377f1 thp: change split... |
2698 |
struct mm_struct *mm = vma->vm_mm; |
c5a647d09 thp: implement sp... |
2699 2700 2701 |
unsigned long haddr = address & HPAGE_PMD_MASK; unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
e180377f1 thp: change split... |
2702 2703 |
BUG_ON(vma->vm_start > haddr || vma->vm_end < haddr + HPAGE_PMD_SIZE); |
71e3aac07 thp: transparent ... |
2704 |
|
c5a647d09 thp: implement sp... |
2705 2706 |
mmun_start = haddr; mmun_end = haddr + HPAGE_PMD_SIZE; |
750e8165f mm: fix BUG in __... |
2707 |
again: |
c5a647d09 thp: implement sp... |
2708 |
mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end); |
c4088ebdc mm: convert the r... |
2709 |
ptl = pmd_lock(mm, pmd); |
71e3aac07 thp: transparent ... |
2710 |
if (unlikely(!pmd_trans_huge(*pmd))) { |
c4088ebdc mm: convert the r... |
2711 |
spin_unlock(ptl); |
c5a647d09 thp: implement sp... |
2712 2713 2714 2715 2716 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); return; } if (is_huge_zero_pmd(*pmd)) { __split_huge_zero_page_pmd(vma, haddr, pmd); |
c4088ebdc mm: convert the r... |
2717 |
spin_unlock(ptl); |
c5a647d09 thp: implement sp... |
2718 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
2719 2720 2721 |
return; } page = pmd_page(*pmd); |
309381fea mm: dump page whe... |
2722 |
VM_BUG_ON_PAGE(!page_count(page), page); |
71e3aac07 thp: transparent ... |
2723 |
get_page(page); |
c4088ebdc mm: convert the r... |
2724 |
spin_unlock(ptl); |
c5a647d09 thp: implement sp... |
2725 |
mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
2726 2727 2728 2729 |
split_huge_page(page); put_page(page); |
750e8165f mm: fix BUG in __... |
2730 2731 2732 2733 2734 2735 2736 2737 |
/* * We don't always have down_write of mmap_sem here: a racing * do_huge_pmd_wp_page() might have copied-on-write to another * huge page before our split_huge_page() got the anon_vma lock. */ if (unlikely(pmd_trans_huge(*pmd))) goto again; |
71e3aac07 thp: transparent ... |
2738 |
} |
94fcc585f thp: avoid breaki... |
2739 |
|
e180377f1 thp: change split... |
2740 2741 2742 2743 2744 2745 2746 2747 2748 |
void split_huge_page_pmd_mm(struct mm_struct *mm, unsigned long address, pmd_t *pmd) { struct vm_area_struct *vma; vma = find_vma(mm, address); BUG_ON(vma == NULL); split_huge_page_pmd(vma, address, pmd); } |
94fcc585f thp: avoid breaki... |
2749 2750 2751 |
static void split_huge_page_address(struct mm_struct *mm, unsigned long address) { |
94fcc585f thp: avoid breaki... |
2752 2753 2754 |
pmd_t *pmd; VM_BUG_ON(!(address & ~HPAGE_PMD_MASK)); |
6219049ae mm: introduce mm_... |
2755 2756 |
pmd = mm_find_pmd(mm, address); if (!pmd) |
94fcc585f thp: avoid breaki... |
2757 2758 2759 2760 2761 |
return; /* * Caller holds the mmap_sem write mode, so a huge pmd cannot * materialize from under us. */ |
e180377f1 thp: change split... |
2762 |
split_huge_page_pmd_mm(mm, address, pmd); |
94fcc585f thp: avoid breaki... |
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 |
} void __vma_adjust_trans_huge(struct vm_area_struct *vma, unsigned long start, unsigned long end, long adjust_next) { /* * If the new start address isn't hpage aligned and it could * previously contain an hugepage: check if we need to split * an huge pmd. */ if (start & ~HPAGE_PMD_MASK && (start & HPAGE_PMD_MASK) >= vma->vm_start && (start & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= vma->vm_end) split_huge_page_address(vma->vm_mm, start); /* * If the new end address isn't hpage aligned and it could * previously contain an hugepage: check if we need to split * an huge pmd. */ if (end & ~HPAGE_PMD_MASK && (end & HPAGE_PMD_MASK) >= vma->vm_start && (end & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= vma->vm_end) split_huge_page_address(vma->vm_mm, end); /* * If we're also updating the vma->vm_next->vm_start, if the new * vm_next->vm_start isn't page aligned and it could previously * contain an hugepage: check if we need to split an huge pmd. */ if (adjust_next > 0) { struct vm_area_struct *next = vma->vm_next; unsigned long nstart = next->vm_start; nstart += adjust_next << PAGE_SHIFT; if (nstart & ~HPAGE_PMD_MASK && (nstart & HPAGE_PMD_MASK) >= next->vm_start && (nstart & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= next->vm_end) split_huge_page_address(next->vm_mm, nstart); } } |