Blame view
mm/huge_memory.c
79.1 KB
71e3aac07 thp: transparent ... |
1 2 3 4 5 6 |
/* * 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. */ |
ae3a8c1c2 mm/huge_memory.c:... |
7 |
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
71e3aac07 thp: transparent ... |
8 9 |
#include <linux/mm.h> #include <linux/sched.h> |
f7ccbae45 sched/headers: Pr... |
10 |
#include <linux/sched/coredump.h> |
6a3827d75 sched/headers: Pr... |
11 |
#include <linux/sched/numa_balancing.h> |
71e3aac07 thp: transparent ... |
12 13 14 15 16 |
#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... |
17 |
#include <linux/shrinker.h> |
ba76149f4 thp: khugepaged |
18 |
#include <linux/mm_inline.h> |
e9b61f198 thp: reintroduce ... |
19 |
#include <linux/swapops.h> |
4897c7655 thp: prepare for ... |
20 |
#include <linux/dax.h> |
ba76149f4 thp: khugepaged |
21 |
#include <linux/khugepaged.h> |
878aee7d6 thp: freeze khuge... |
22 |
#include <linux/freezer.h> |
f25748e3c mm, dax: convert ... |
23 |
#include <linux/pfn_t.h> |
a664b2d85 thp: madvise(MADV... |
24 |
#include <linux/mman.h> |
3565fce3a mm, x86: get_user... |
25 |
#include <linux/memremap.h> |
325adeb55 mm: huge_memory: ... |
26 |
#include <linux/pagemap.h> |
49071d436 thp: add debugfs ... |
27 |
#include <linux/debugfs.h> |
4daae3b4b mm: mempolicy: Us... |
28 |
#include <linux/migrate.h> |
43b5fbbd2 mm/huge_memory.c:... |
29 |
#include <linux/hashtable.h> |
6b251fc96 userfaultfd: call... |
30 |
#include <linux/userfaultfd_k.h> |
33c3fc71c mm: introduce idl... |
31 |
#include <linux/page_idle.h> |
baa355fd3 thp: file pages s... |
32 |
#include <linux/shmem_fs.h> |
6b31d5955 mm, oom: fix pote... |
33 |
#include <linux/oom.h> |
97ae17497 thp: implement re... |
34 |
|
71e3aac07 thp: transparent ... |
35 36 37 |
#include <asm/tlb.h> #include <asm/pgalloc.h> #include "internal.h" |
ba76149f4 thp: khugepaged |
38 |
/* |
8bfa3f9a0 mm/huge_memory.c:... |
39 40 41 42 43 44 |
* 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 |
45 |
*/ |
71e3aac07 thp: transparent ... |
46 |
unsigned long transparent_hugepage_flags __read_mostly = |
13ece886d thp: transparent ... |
47 |
#ifdef CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS |
ba76149f4 thp: khugepaged |
48 |
(1<<TRANSPARENT_HUGEPAGE_FLAG)| |
13ece886d thp: transparent ... |
49 50 51 52 |
#endif #ifdef CONFIG_TRANSPARENT_HUGEPAGE_MADVISE (1<<TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG)| #endif |
444eb2a44 mm: thp: set THP ... |
53 |
(1<<TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG)| |
79da5407e thp: introduce sy... |
54 55 |
(1<<TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG)| (1<<TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG); |
ba76149f4 thp: khugepaged |
56 |
|
9a982250f thp: introduce de... |
57 |
static struct shrinker deferred_split_shrinker; |
f000565ad thp: set recommen... |
58 |
|
97ae17497 thp: implement re... |
59 |
static atomic_t huge_zero_refcount; |
56873f43a mm:add KPF_ZERO_P... |
60 |
struct page *huge_zero_page __read_mostly; |
4a6c12972 thp: huge zero pa... |
61 |
|
6fcb52a56 thp: reduce usage... |
62 |
static struct page *get_huge_zero_page(void) |
97ae17497 thp: implement re... |
63 64 65 66 |
{ struct page *zero_page; retry: if (likely(atomic_inc_not_zero(&huge_zero_refcount))) |
4db0c3c29 mm: remove rest o... |
67 |
return READ_ONCE(huge_zero_page); |
97ae17497 thp: implement re... |
68 69 |
zero_page = alloc_pages((GFP_TRANSHUGE | __GFP_ZERO) & ~__GFP_MOVABLE, |
4a6c12972 thp: huge zero pa... |
70 |
HPAGE_PMD_ORDER); |
d8a8e1f0d thp, vmstat: impl... |
71 72 |
if (!zero_page) { count_vm_event(THP_ZERO_PAGE_ALLOC_FAILED); |
5918d10a4 thp: fix huge zer... |
73 |
return NULL; |
d8a8e1f0d thp, vmstat: impl... |
74 75 |
} count_vm_event(THP_ZERO_PAGE_ALLOC); |
97ae17497 thp: implement re... |
76 |
preempt_disable(); |
5918d10a4 thp: fix huge zer... |
77 |
if (cmpxchg(&huge_zero_page, NULL, zero_page)) { |
97ae17497 thp: implement re... |
78 |
preempt_enable(); |
5ddacbe92 mm: free compound... |
79 |
__free_pages(zero_page, compound_order(zero_page)); |
97ae17497 thp: implement re... |
80 81 82 83 84 85 |
goto retry; } /* We take additional reference here. It will be put back by shrinker */ atomic_set(&huge_zero_refcount, 2); preempt_enable(); |
4db0c3c29 mm: remove rest o... |
86 |
return READ_ONCE(huge_zero_page); |
4a6c12972 thp: huge zero pa... |
87 |
} |
6fcb52a56 thp: reduce usage... |
88 |
static void put_huge_zero_page(void) |
4a6c12972 thp: huge zero pa... |
89 |
{ |
97ae17497 thp: implement re... |
90 91 92 93 94 |
/* * 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... |
95 |
} |
6fcb52a56 thp: reduce usage... |
96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 |
struct page *mm_get_huge_zero_page(struct mm_struct *mm) { if (test_bit(MMF_HUGE_ZERO_PAGE, &mm->flags)) return READ_ONCE(huge_zero_page); if (!get_huge_zero_page()) return NULL; if (test_and_set_bit(MMF_HUGE_ZERO_PAGE, &mm->flags)) put_huge_zero_page(); return READ_ONCE(huge_zero_page); } void mm_put_huge_zero_page(struct mm_struct *mm) { if (test_bit(MMF_HUGE_ZERO_PAGE, &mm->flags)) put_huge_zero_page(); } |
488964666 hugepage: convert... |
115 116 |
static unsigned long shrink_huge_zero_page_count(struct shrinker *shrink, struct shrink_control *sc) |
4a6c12972 thp: huge zero pa... |
117 |
{ |
488964666 hugepage: convert... |
118 119 120 |
/* 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... |
121 |
|
488964666 hugepage: convert... |
122 123 124 |
static unsigned long shrink_huge_zero_page_scan(struct shrinker *shrink, struct shrink_control *sc) { |
97ae17497 thp: implement re... |
125 |
if (atomic_cmpxchg(&huge_zero_refcount, 1, 0) == 1) { |
5918d10a4 thp: fix huge zer... |
126 127 |
struct page *zero_page = xchg(&huge_zero_page, NULL); BUG_ON(zero_page == NULL); |
5ddacbe92 mm: free compound... |
128 |
__free_pages(zero_page, compound_order(zero_page)); |
488964666 hugepage: convert... |
129 |
return HPAGE_PMD_NR; |
97ae17497 thp: implement re... |
130 131 132 |
} return 0; |
4a6c12972 thp: huge zero pa... |
133 |
} |
97ae17497 thp: implement re... |
134 |
static struct shrinker huge_zero_page_shrinker = { |
488964666 hugepage: convert... |
135 136 |
.count_objects = shrink_huge_zero_page_count, .scan_objects = shrink_huge_zero_page_scan, |
97ae17497 thp: implement re... |
137 138 |
.seeks = DEFAULT_SEEKS, }; |
71e3aac07 thp: transparent ... |
139 |
#ifdef CONFIG_SYSFS |
71e3aac07 thp: transparent ... |
140 141 142 |
static ssize_t enabled_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { |
444eb2a44 mm: thp: set THP ... |
143 144 145 146 147 148 149 150 151 |
if (test_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags)) return sprintf(buf, "[always] madvise never "); else if (test_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags)) return sprintf(buf, "always [madvise] never "); else return sprintf(buf, "always madvise [never] "); |
71e3aac07 thp: transparent ... |
152 |
} |
444eb2a44 mm: thp: set THP ... |
153 |
|
71e3aac07 thp: transparent ... |
154 155 156 157 |
static ssize_t enabled_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { |
21440d7eb mm, thp: add new ... |
158 |
ssize_t ret = count; |
ba76149f4 thp: khugepaged |
159 |
|
21440d7eb mm, thp: add new ... |
160 161 162 163 164 165 166 167 168 169 170 171 172 173 |
if (!memcmp("always", buf, min(sizeof("always")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); } else if (!memcmp("madvise", buf, min(sizeof("madvise")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); } else if (!memcmp("never", buf, min(sizeof("never")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG, &transparent_hugepage_flags); } else ret = -EINVAL; |
ba76149f4 thp: khugepaged |
174 175 |
if (ret > 0) { |
b46e756f5 thp: extract khug... |
176 |
int err = start_stop_khugepaged(); |
ba76149f4 thp: khugepaged |
177 178 179 |
if (err) ret = err; } |
ba76149f4 thp: khugepaged |
180 |
return ret; |
71e3aac07 thp: transparent ... |
181 182 183 |
} static struct kobj_attribute enabled_attr = __ATTR(enabled, 0644, enabled_show, enabled_store); |
b46e756f5 thp: extract khug... |
184 |
ssize_t single_hugepage_flag_show(struct kobject *kobj, |
71e3aac07 thp: transparent ... |
185 186 187 |
struct kobj_attribute *attr, char *buf, enum transparent_hugepage_flag flag) { |
e27e6151b mm/thp: use conve... |
188 189 190 |
return sprintf(buf, "%d ", !!test_bit(flag, &transparent_hugepage_flags)); |
71e3aac07 thp: transparent ... |
191 |
} |
e27e6151b mm/thp: use conve... |
192 |
|
b46e756f5 thp: extract khug... |
193 |
ssize_t single_hugepage_flag_store(struct kobject *kobj, |
71e3aac07 thp: transparent ... |
194 195 196 197 |
struct kobj_attribute *attr, const char *buf, size_t count, enum transparent_hugepage_flag flag) { |
e27e6151b mm/thp: use conve... |
198 199 200 201 202 203 204 205 206 207 |
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 ... |
208 |
set_bit(flag, &transparent_hugepage_flags); |
e27e6151b mm/thp: use conve... |
209 |
else |
71e3aac07 thp: transparent ... |
210 |
clear_bit(flag, &transparent_hugepage_flags); |
71e3aac07 thp: transparent ... |
211 212 213 |
return count; } |
71e3aac07 thp: transparent ... |
214 215 216 |
static ssize_t defrag_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { |
444eb2a44 mm: thp: set THP ... |
217 |
if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags)) |
21440d7eb mm, thp: add new ... |
218 219 |
return sprintf(buf, "[always] defer defer+madvise madvise never "); |
444eb2a44 mm: thp: set THP ... |
220 |
if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags)) |
21440d7eb mm, thp: add new ... |
221 222 223 224 225 226 227 228 229 230 |
return sprintf(buf, "always [defer] defer+madvise madvise never "); if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags)) return sprintf(buf, "always defer [defer+madvise] madvise never "); if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags)) return sprintf(buf, "always defer defer+madvise [madvise] never "); return sprintf(buf, "always defer defer+madvise madvise [never] "); |
71e3aac07 thp: transparent ... |
231 |
} |
21440d7eb mm, thp: add new ... |
232 |
|
71e3aac07 thp: transparent ... |
233 234 235 236 |
static ssize_t defrag_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { |
21440d7eb mm, thp: add new ... |
237 238 239 240 241 242 |
if (!memcmp("always", buf, min(sizeof("always")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags); |
21440d7eb mm, thp: add new ... |
243 244 245 246 247 248 |
} else if (!memcmp("defer+madvise", buf, min(sizeof("defer+madvise")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags); |
4fad7fb6b mm, thp: fix sett... |
249 250 251 252 253 254 |
} else if (!memcmp("defer", buf, min(sizeof("defer")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags); |
21440d7eb mm, thp: add new ... |
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 |
} else if (!memcmp("madvise", buf, min(sizeof("madvise")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags); set_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags); } else if (!memcmp("never", buf, min(sizeof("never")-1, count))) { clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags); clear_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags); } else return -EINVAL; return count; |
71e3aac07 thp: transparent ... |
271 272 273 |
} static struct kobj_attribute defrag_attr = __ATTR(defrag, 0644, defrag_show, defrag_store); |
79da5407e thp: introduce sy... |
274 275 276 |
static ssize_t use_zero_page_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { |
b46e756f5 thp: extract khug... |
277 |
return single_hugepage_flag_show(kobj, attr, buf, |
79da5407e thp: introduce sy... |
278 279 280 281 282 |
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) { |
b46e756f5 thp: extract khug... |
283 |
return single_hugepage_flag_store(kobj, attr, buf, count, |
79da5407e thp: introduce sy... |
284 285 286 287 |
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); |
49920d287 mm: make transpar... |
288 289 290 291 292 293 294 295 296 |
static ssize_t hpage_pmd_size_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { return sprintf(buf, "%lu ", HPAGE_PMD_SIZE); } static struct kobj_attribute hpage_pmd_size_attr = __ATTR_RO(hpage_pmd_size); |
71e3aac07 thp: transparent ... |
297 298 299 300 |
#ifdef CONFIG_DEBUG_VM static ssize_t debug_cow_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf) { |
b46e756f5 thp: extract khug... |
301 |
return single_hugepage_flag_show(kobj, attr, buf, |
71e3aac07 thp: transparent ... |
302 303 304 305 306 307 |
TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG); } static ssize_t debug_cow_store(struct kobject *kobj, struct kobj_attribute *attr, const char *buf, size_t count) { |
b46e756f5 thp: extract khug... |
308 |
return single_hugepage_flag_store(kobj, attr, buf, count, |
71e3aac07 thp: transparent ... |
309 310 311 312 313 314 315 316 317 |
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... |
318 |
&use_zero_page_attr.attr, |
49920d287 mm: make transpar... |
319 |
&hpage_pmd_size_attr.attr, |
e496cf3d7 thp: introduce CO... |
320 |
#if defined(CONFIG_SHMEM) && defined(CONFIG_TRANSPARENT_HUGE_PAGECACHE) |
5a6e75f81 shmem: prepare hu... |
321 322 |
&shmem_enabled_attr.attr, #endif |
71e3aac07 thp: transparent ... |
323 324 325 326 327 |
#ifdef CONFIG_DEBUG_VM &debug_cow_attr.attr, #endif NULL, }; |
8aa95a21b mm/huge_memory.c:... |
328 |
static const struct attribute_group hugepage_attr_group = { |
71e3aac07 thp: transparent ... |
329 |
.attrs = hugepage_attr, |
ba76149f4 thp: khugepaged |
330 |
}; |
569e55900 thp: improve the ... |
331 |
static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj) |
71e3aac07 thp: transparent ... |
332 |
{ |
71e3aac07 thp: transparent ... |
333 |
int err; |
569e55900 thp: improve the ... |
334 335 |
*hugepage_kobj = kobject_create_and_add("transparent_hugepage", mm_kobj); if (unlikely(!*hugepage_kobj)) { |
ae3a8c1c2 mm/huge_memory.c:... |
336 337 |
pr_err("failed to create transparent hugepage kobject "); |
569e55900 thp: improve the ... |
338 |
return -ENOMEM; |
ba76149f4 thp: khugepaged |
339 |
} |
569e55900 thp: improve the ... |
340 |
err = sysfs_create_group(*hugepage_kobj, &hugepage_attr_group); |
ba76149f4 thp: khugepaged |
341 |
if (err) { |
ae3a8c1c2 mm/huge_memory.c:... |
342 343 |
pr_err("failed to register transparent hugepage group "); |
569e55900 thp: improve the ... |
344 |
goto delete_obj; |
ba76149f4 thp: khugepaged |
345 |
} |
569e55900 thp: improve the ... |
346 |
err = sysfs_create_group(*hugepage_kobj, &khugepaged_attr_group); |
ba76149f4 thp: khugepaged |
347 |
if (err) { |
ae3a8c1c2 mm/huge_memory.c:... |
348 349 |
pr_err("failed to register transparent hugepage group "); |
569e55900 thp: improve the ... |
350 |
goto remove_hp_group; |
ba76149f4 thp: khugepaged |
351 |
} |
569e55900 thp: improve the ... |
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 |
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; } |
ff20c2e0a mm: Some arch may... |
388 389 390 391 392 393 394 395 396 |
/* * hugepages can't be allocated by the buddy allocator */ MAYBE_BUILD_BUG_ON(HPAGE_PMD_ORDER >= MAX_ORDER); /* * we use page->mapping and page->index in second tail page * as list_head: assuming THP order >= 2 */ MAYBE_BUILD_BUG_ON(HPAGE_PMD_ORDER < 2); |
569e55900 thp: improve the ... |
397 398 |
err = hugepage_init_sysfs(&hugepage_kobj); if (err) |
65ebb64f4 thp: handle error... |
399 |
goto err_sysfs; |
ba76149f4 thp: khugepaged |
400 |
|
b46e756f5 thp: extract khug... |
401 |
err = khugepaged_init(); |
ba76149f4 thp: khugepaged |
402 |
if (err) |
65ebb64f4 thp: handle error... |
403 |
goto err_slab; |
ba76149f4 thp: khugepaged |
404 |
|
65ebb64f4 thp: handle error... |
405 406 407 |
err = register_shrinker(&huge_zero_page_shrinker); if (err) goto err_hzp_shrinker; |
9a982250f thp: introduce de... |
408 409 410 |
err = register_shrinker(&deferred_split_shrinker); if (err) goto err_split_shrinker; |
97ae17497 thp: implement re... |
411 |
|
97562cd24 thp: disable tran... |
412 413 414 415 416 |
/* * 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. */ |
79553da29 thp: cleanup khug... |
417 |
if (totalram_pages < (512 << (20 - PAGE_SHIFT))) { |
97562cd24 thp: disable tran... |
418 |
transparent_hugepage_flags = 0; |
79553da29 thp: cleanup khug... |
419 420 |
return 0; } |
97562cd24 thp: disable tran... |
421 |
|
79553da29 thp: cleanup khug... |
422 |
err = start_stop_khugepaged(); |
65ebb64f4 thp: handle error... |
423 424 |
if (err) goto err_khugepaged; |
ba76149f4 thp: khugepaged |
425 |
|
569e55900 thp: improve the ... |
426 |
return 0; |
65ebb64f4 thp: handle error... |
427 |
err_khugepaged: |
9a982250f thp: introduce de... |
428 429 |
unregister_shrinker(&deferred_split_shrinker); err_split_shrinker: |
65ebb64f4 thp: handle error... |
430 431 |
unregister_shrinker(&huge_zero_page_shrinker); err_hzp_shrinker: |
b46e756f5 thp: extract khug... |
432 |
khugepaged_destroy(); |
65ebb64f4 thp: handle error... |
433 |
err_slab: |
569e55900 thp: improve the ... |
434 |
hugepage_exit_sysfs(hugepage_kobj); |
65ebb64f4 thp: handle error... |
435 |
err_sysfs: |
ba76149f4 thp: khugepaged |
436 |
return err; |
71e3aac07 thp: transparent ... |
437 |
} |
a64fb3cd6 mm: audit/fix non... |
438 |
subsys_initcall(hugepage_init); |
71e3aac07 thp: transparent ... |
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 |
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) |
ae3a8c1c2 mm/huge_memory.c:... |
466 467 |
pr_warn("transparent_hugepage= cannot parse, ignored "); |
71e3aac07 thp: transparent ... |
468 469 470 |
return ret; } __setup("transparent_hugepage=", setup_transparent_hugepage); |
b32967ff1 mm: numa: Add THP... |
471 |
pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma) |
71e3aac07 thp: transparent ... |
472 473 474 475 476 |
{ if (likely(vma->vm_flags & VM_WRITE)) pmd = pmd_mkwrite(pmd); return pmd; } |
9a982250f thp: introduce de... |
477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 |
static inline struct list_head *page_deferred_list(struct page *page) { /* * ->lru in the tail pages is occupied by compound_head. * Let's use ->mapping + ->index in the second tail page as list_head. */ return (struct list_head *)&page[2].mapping; } void prep_transhuge_page(struct page *page) { /* * we use page->mapping and page->indexlru in second tail page * as list_head: assuming THP order >= 2 */ |
9a982250f thp: introduce de... |
492 493 494 495 |
INIT_LIST_HEAD(page_deferred_list(page)); set_compound_page_dtor(page, TRANSHUGE_PAGE_DTOR); } |
74d2fad13 thp, dax: add thp... |
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 |
unsigned long __thp_get_unmapped_area(struct file *filp, unsigned long len, loff_t off, unsigned long flags, unsigned long size) { unsigned long addr; loff_t off_end = off + len; loff_t off_align = round_up(off, size); unsigned long len_pad; if (off_end <= off_align || (off_end - off_align) < size) return 0; len_pad = len + size; if (len_pad < len || (off + len_pad) < off) return 0; addr = current->mm->get_unmapped_area(filp, 0, len_pad, off >> PAGE_SHIFT, flags); if (IS_ERR_VALUE(addr)) return 0; addr += (off - addr) & (size - 1); return addr; } unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr, unsigned long len, unsigned long pgoff, unsigned long flags) { loff_t off = (loff_t)pgoff << PAGE_SHIFT; if (addr) goto out; if (!IS_DAX(filp->f_mapping->host) || !IS_ENABLED(CONFIG_FS_DAX_PMD)) goto out; addr = __thp_get_unmapped_area(filp, len, off, flags, PMD_SIZE); if (addr) return addr; out: return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags); } EXPORT_SYMBOL_GPL(thp_get_unmapped_area); |
82b0f8c39 mm: join struct f... |
538 |
static int __do_huge_pmd_anonymous_page(struct vm_fault *vmf, struct page *page, |
bae473a42 mm: introduce fau... |
539 |
gfp_t gfp) |
71e3aac07 thp: transparent ... |
540 |
{ |
82b0f8c39 mm: join struct f... |
541 |
struct vm_area_struct *vma = vmf->vma; |
00501b531 mm: memcontrol: r... |
542 |
struct mem_cgroup *memcg; |
71e3aac07 thp: transparent ... |
543 |
pgtable_t pgtable; |
82b0f8c39 mm: join struct f... |
544 |
unsigned long haddr = vmf->address & HPAGE_PMD_MASK; |
6b31d5955 mm, oom: fix pote... |
545 |
int ret = 0; |
71e3aac07 thp: transparent ... |
546 |
|
309381fea mm: dump page whe... |
547 |
VM_BUG_ON_PAGE(!PageCompound(page), page); |
00501b531 mm: memcontrol: r... |
548 |
|
49f4a8c52 mm, thp: do not c... |
549 550 |
if (mem_cgroup_try_charge(page, vma->vm_mm, gfp | __GFP_NORETRY, &memcg, true)) { |
6b251fc96 userfaultfd: call... |
551 552 553 554 |
put_page(page); count_vm_event(THP_FAULT_FALLBACK); return VM_FAULT_FALLBACK; } |
00501b531 mm: memcontrol: r... |
555 |
|
bae473a42 mm: introduce fau... |
556 |
pgtable = pte_alloc_one(vma->vm_mm, haddr); |
00501b531 mm: memcontrol: r... |
557 |
if (unlikely(!pgtable)) { |
6b31d5955 mm, oom: fix pote... |
558 559 |
ret = VM_FAULT_OOM; goto release; |
00501b531 mm: memcontrol: r... |
560 |
} |
71e3aac07 thp: transparent ... |
561 |
|
c79b57e46 mm: hugetlb: clea... |
562 |
clear_huge_page(page, vmf->address, HPAGE_PMD_NR); |
52f37629f THP: fix comment ... |
563 564 565 566 567 |
/* * The memory barrier inside __SetPageUptodate makes sure that * clear_huge_page writes become visible before the set_pmd_at() * write. */ |
71e3aac07 thp: transparent ... |
568 |
__SetPageUptodate(page); |
82b0f8c39 mm: join struct f... |
569 570 |
vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd); if (unlikely(!pmd_none(*vmf->pmd))) { |
6b31d5955 mm, oom: fix pote... |
571 |
goto unlock_release; |
71e3aac07 thp: transparent ... |
572 573 |
} else { pmd_t entry; |
6b251fc96 userfaultfd: call... |
574 |
|
6b31d5955 mm, oom: fix pote... |
575 576 577 |
ret = check_stable_address_space(vma->vm_mm); if (ret) goto unlock_release; |
6b251fc96 userfaultfd: call... |
578 579 580 |
/* Deliver the page fault to userland */ if (userfaultfd_missing(vma)) { int ret; |
82b0f8c39 mm: join struct f... |
581 |
spin_unlock(vmf->ptl); |
f627c2f53 memcg: adjust to ... |
582 |
mem_cgroup_cancel_charge(page, memcg, true); |
6b251fc96 userfaultfd: call... |
583 |
put_page(page); |
bae473a42 mm: introduce fau... |
584 |
pte_free(vma->vm_mm, pgtable); |
82b0f8c39 mm: join struct f... |
585 |
ret = handle_userfault(vmf, VM_UFFD_MISSING); |
6b251fc96 userfaultfd: call... |
586 587 588 |
VM_BUG_ON(ret & VM_FAULT_FALLBACK); return ret; } |
3122359a6 thp: move maybe_p... |
589 590 |
entry = mk_huge_pmd(page, vma->vm_page_prot); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); |
d281ee614 rmap: add argumen... |
591 |
page_add_new_anon_rmap(page, vma, haddr, true); |
f627c2f53 memcg: adjust to ... |
592 |
mem_cgroup_commit_charge(page, memcg, false, true); |
00501b531 mm: memcontrol: r... |
593 |
lru_cache_add_active_or_unevictable(page, vma); |
82b0f8c39 mm: join struct f... |
594 595 |
pgtable_trans_huge_deposit(vma->vm_mm, vmf->pmd, pgtable); set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry); |
bae473a42 mm: introduce fau... |
596 597 |
add_mm_counter(vma->vm_mm, MM_ANONPAGES, HPAGE_PMD_NR); atomic_long_inc(&vma->vm_mm->nr_ptes); |
82b0f8c39 mm: join struct f... |
598 |
spin_unlock(vmf->ptl); |
6b251fc96 userfaultfd: call... |
599 |
count_vm_event(THP_FAULT_ALLOC); |
71e3aac07 thp: transparent ... |
600 |
} |
aa2e878ef mm, thp: remove u... |
601 |
return 0; |
6b31d5955 mm, oom: fix pote... |
602 603 604 605 606 607 608 609 |
unlock_release: spin_unlock(vmf->ptl); release: if (pgtable) pte_free(vma->vm_mm, pgtable); mem_cgroup_cancel_charge(page, memcg, true); put_page(page); return ret; |
71e3aac07 thp: transparent ... |
610 |
} |
444eb2a44 mm: thp: set THP ... |
611 |
/* |
21440d7eb mm, thp: add new ... |
612 613 614 615 616 617 618 |
* always: directly stall for all thp allocations * defer: wake kswapd and fail if not immediately available * defer+madvise: wake kswapd and directly stall for MADV_HUGEPAGE, otherwise * fail if not immediately available * madvise: directly stall for MADV_HUGEPAGE, otherwise fail if not immediately * available * never: never stall for any thp allocation |
444eb2a44 mm: thp: set THP ... |
619 620 621 |
*/ static inline gfp_t alloc_hugepage_direct_gfpmask(struct vm_area_struct *vma) { |
21440d7eb mm, thp: add new ... |
622 |
const bool vma_madvised = !!(vma->vm_flags & VM_HUGEPAGE); |
251603549 mm, thp: remove _... |
623 |
|
21440d7eb mm, thp: add new ... |
624 |
if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_DIRECT_FLAG, &transparent_hugepage_flags)) |
251603549 mm, thp: remove _... |
625 |
return GFP_TRANSHUGE | (vma_madvised ? 0 : __GFP_NORETRY); |
21440d7eb mm, thp: add new ... |
626 627 628 629 630 631 632 633 |
if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_FLAG, &transparent_hugepage_flags)) return GFP_TRANSHUGE_LIGHT | __GFP_KSWAPD_RECLAIM; if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_KSWAPD_OR_MADV_FLAG, &transparent_hugepage_flags)) return GFP_TRANSHUGE_LIGHT | (vma_madvised ? __GFP_DIRECT_RECLAIM : __GFP_KSWAPD_RECLAIM); if (test_bit(TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG, &transparent_hugepage_flags)) return GFP_TRANSHUGE_LIGHT | (vma_madvised ? __GFP_DIRECT_RECLAIM : 0); |
251603549 mm, thp: remove _... |
634 |
return GFP_TRANSHUGE_LIGHT; |
444eb2a44 mm: thp: set THP ... |
635 |
} |
c4088ebdc mm: convert the r... |
636 |
/* Caller must hold page table lock. */ |
d295e3415 dax: don't use se... |
637 |
static bool set_huge_zero_page(pgtable_t pgtable, struct mm_struct *mm, |
97ae17497 thp: implement re... |
638 |
struct vm_area_struct *vma, unsigned long haddr, pmd_t *pmd, |
5918d10a4 thp: fix huge zer... |
639 |
struct page *zero_page) |
fc9fe822f thp: copy_huge_pm... |
640 641 |
{ pmd_t entry; |
7c4141645 dax: revert userf... |
642 643 |
if (!pmd_none(*pmd)) return false; |
5918d10a4 thp: fix huge zer... |
644 |
entry = mk_pmd(zero_page, vma->vm_page_prot); |
fc9fe822f thp: copy_huge_pm... |
645 |
entry = pmd_mkhuge(entry); |
12c9d70bd mm: fix memory le... |
646 647 |
if (pgtable) pgtable_trans_huge_deposit(mm, pmd, pgtable); |
fc9fe822f thp: copy_huge_pm... |
648 |
set_pmd_at(mm, haddr, pmd, entry); |
e1f56c89b mm: convert mm->n... |
649 |
atomic_long_inc(&mm->nr_ptes); |
7c4141645 dax: revert userf... |
650 |
return true; |
fc9fe822f thp: copy_huge_pm... |
651 |
} |
82b0f8c39 mm: join struct f... |
652 |
int do_huge_pmd_anonymous_page(struct vm_fault *vmf) |
71e3aac07 thp: transparent ... |
653 |
{ |
82b0f8c39 mm: join struct f... |
654 |
struct vm_area_struct *vma = vmf->vma; |
077fcf116 mm/thp: allocate ... |
655 |
gfp_t gfp; |
71e3aac07 thp: transparent ... |
656 |
struct page *page; |
82b0f8c39 mm: join struct f... |
657 |
unsigned long haddr = vmf->address & HPAGE_PMD_MASK; |
71e3aac07 thp: transparent ... |
658 |
|
128ec037b thp: do_huge_pmd_... |
659 |
if (haddr < vma->vm_start || haddr + HPAGE_PMD_SIZE > vma->vm_end) |
c02925540 thp: consolidate ... |
660 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
661 662 |
if (unlikely(anon_vma_prepare(vma))) return VM_FAULT_OOM; |
6d50e60cd mm, thp: fix coll... |
663 |
if (unlikely(khugepaged_enter(vma, vma->vm_flags))) |
128ec037b thp: do_huge_pmd_... |
664 |
return VM_FAULT_OOM; |
82b0f8c39 mm: join struct f... |
665 |
if (!(vmf->flags & FAULT_FLAG_WRITE) && |
bae473a42 mm: introduce fau... |
666 |
!mm_forbids_zeropage(vma->vm_mm) && |
128ec037b thp: do_huge_pmd_... |
667 668 669 670 |
transparent_hugepage_use_zero_page()) { pgtable_t pgtable; struct page *zero_page; bool set; |
6b251fc96 userfaultfd: call... |
671 |
int ret; |
bae473a42 mm: introduce fau... |
672 |
pgtable = pte_alloc_one(vma->vm_mm, haddr); |
128ec037b thp: do_huge_pmd_... |
673 |
if (unlikely(!pgtable)) |
ba76149f4 thp: khugepaged |
674 |
return VM_FAULT_OOM; |
6fcb52a56 thp: reduce usage... |
675 |
zero_page = mm_get_huge_zero_page(vma->vm_mm); |
128ec037b thp: do_huge_pmd_... |
676 |
if (unlikely(!zero_page)) { |
bae473a42 mm: introduce fau... |
677 |
pte_free(vma->vm_mm, pgtable); |
81ab4201f mm: add VM counte... |
678 |
count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
679 |
return VM_FAULT_FALLBACK; |
b9bbfbe30 thp: memcg huge m... |
680 |
} |
82b0f8c39 mm: join struct f... |
681 |
vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd); |
6b251fc96 userfaultfd: call... |
682 683 |
ret = 0; set = false; |
82b0f8c39 mm: join struct f... |
684 |
if (pmd_none(*vmf->pmd)) { |
6b31d5955 mm, oom: fix pote... |
685 686 687 688 |
ret = check_stable_address_space(vma->vm_mm); if (ret) { spin_unlock(vmf->ptl); } else if (userfaultfd_missing(vma)) { |
82b0f8c39 mm: join struct f... |
689 690 |
spin_unlock(vmf->ptl); ret = handle_userfault(vmf, VM_UFFD_MISSING); |
6b251fc96 userfaultfd: call... |
691 692 |
VM_BUG_ON(ret & VM_FAULT_FALLBACK); } else { |
bae473a42 mm: introduce fau... |
693 |
set_huge_zero_page(pgtable, vma->vm_mm, vma, |
82b0f8c39 mm: join struct f... |
694 695 |
haddr, vmf->pmd, zero_page); spin_unlock(vmf->ptl); |
6b251fc96 userfaultfd: call... |
696 697 698 |
set = true; } } else |
82b0f8c39 mm: join struct f... |
699 |
spin_unlock(vmf->ptl); |
6fcb52a56 thp: reduce usage... |
700 |
if (!set) |
bae473a42 mm: introduce fau... |
701 |
pte_free(vma->vm_mm, pgtable); |
6b251fc96 userfaultfd: call... |
702 |
return ret; |
71e3aac07 thp: transparent ... |
703 |
} |
444eb2a44 mm: thp: set THP ... |
704 |
gfp = alloc_hugepage_direct_gfpmask(vma); |
077fcf116 mm/thp: allocate ... |
705 |
page = alloc_hugepage_vma(gfp, vma, haddr, HPAGE_PMD_ORDER); |
128ec037b thp: do_huge_pmd_... |
706 707 |
if (unlikely(!page)) { count_vm_event(THP_FAULT_FALLBACK); |
c02925540 thp: consolidate ... |
708 |
return VM_FAULT_FALLBACK; |
128ec037b thp: do_huge_pmd_... |
709 |
} |
9a982250f thp: introduce de... |
710 |
prep_transhuge_page(page); |
82b0f8c39 mm: join struct f... |
711 |
return __do_huge_pmd_anonymous_page(vmf, page, gfp); |
71e3aac07 thp: transparent ... |
712 |
} |
ae18d6dcf thp: change inser... |
713 |
static void insert_pfn_pmd(struct vm_area_struct *vma, unsigned long addr, |
3b6521f53 mm/huge_memory.c:... |
714 715 |
pmd_t *pmd, pfn_t pfn, pgprot_t prot, bool write, pgtable_t pgtable) |
5cad465d7 mm: add vmf_inser... |
716 717 718 719 720 721 |
{ struct mm_struct *mm = vma->vm_mm; pmd_t entry; spinlock_t *ptl; ptl = pmd_lock(mm, pmd); |
f25748e3c mm, dax: convert ... |
722 723 724 |
entry = pmd_mkhuge(pfn_t_pmd(pfn, prot)); if (pfn_t_devmap(pfn)) entry = pmd_mkdevmap(entry); |
01871e59a mm, dax: fix live... |
725 726 727 |
if (write) { entry = pmd_mkyoung(pmd_mkdirty(entry)); entry = maybe_pmd_mkwrite(entry, vma); |
5cad465d7 mm: add vmf_inser... |
728 |
} |
3b6521f53 mm/huge_memory.c:... |
729 730 731 732 733 |
if (pgtable) { pgtable_trans_huge_deposit(mm, pmd, pgtable); atomic_long_inc(&mm->nr_ptes); } |
01871e59a mm, dax: fix live... |
734 735 |
set_pmd_at(mm, addr, pmd, entry); update_mmu_cache_pmd(vma, addr, pmd); |
5cad465d7 mm: add vmf_inser... |
736 |
spin_unlock(ptl); |
5cad465d7 mm: add vmf_inser... |
737 738 739 |
} int vmf_insert_pfn_pmd(struct vm_area_struct *vma, unsigned long addr, |
f25748e3c mm, dax: convert ... |
740 |
pmd_t *pmd, pfn_t pfn, bool write) |
5cad465d7 mm: add vmf_inser... |
741 742 |
{ pgprot_t pgprot = vma->vm_page_prot; |
3b6521f53 mm/huge_memory.c:... |
743 |
pgtable_t pgtable = NULL; |
5cad465d7 mm: add vmf_inser... |
744 745 746 747 748 749 750 751 752 |
/* * If we had pmd_special, we could avoid all these restrictions, * but we need to be consistent with PTEs and architectures that * can't support a 'special' bit. */ BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))); BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) == (VM_PFNMAP|VM_MIXEDMAP)); BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags)); |
f25748e3c mm, dax: convert ... |
753 |
BUG_ON(!pfn_t_devmap(pfn)); |
5cad465d7 mm: add vmf_inser... |
754 755 756 |
if (addr < vma->vm_start || addr >= vma->vm_end) return VM_FAULT_SIGBUS; |
308a047c3 x86/pat, mm: Make... |
757 |
|
3b6521f53 mm/huge_memory.c:... |
758 759 760 761 762 |
if (arch_needs_pgtable_deposit()) { pgtable = pte_alloc_one(vma->vm_mm, addr); if (!pgtable) return VM_FAULT_OOM; } |
308a047c3 x86/pat, mm: Make... |
763 |
track_pfn_insert(vma, &pgprot, pfn); |
3b6521f53 mm/huge_memory.c:... |
764 |
insert_pfn_pmd(vma, addr, pmd, pfn, pgprot, write, pgtable); |
ae18d6dcf thp: change inser... |
765 |
return VM_FAULT_NOPAGE; |
5cad465d7 mm: add vmf_inser... |
766 |
} |
dee410792 /dev/dax, core: f... |
767 |
EXPORT_SYMBOL_GPL(vmf_insert_pfn_pmd); |
5cad465d7 mm: add vmf_inser... |
768 |
|
a00cc7d9d mm, x86: add supp... |
769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 |
#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD static pud_t maybe_pud_mkwrite(pud_t pud, struct vm_area_struct *vma) { if (likely(vma->vm_flags & VM_WRITE)) pud = pud_mkwrite(pud); return pud; } static void insert_pfn_pud(struct vm_area_struct *vma, unsigned long addr, pud_t *pud, pfn_t pfn, pgprot_t prot, bool write) { struct mm_struct *mm = vma->vm_mm; pud_t entry; spinlock_t *ptl; ptl = pud_lock(mm, pud); entry = pud_mkhuge(pfn_t_pud(pfn, prot)); if (pfn_t_devmap(pfn)) entry = pud_mkdevmap(entry); if (write) { entry = pud_mkyoung(pud_mkdirty(entry)); entry = maybe_pud_mkwrite(entry, vma); } set_pud_at(mm, addr, pud, entry); update_mmu_cache_pud(vma, addr, pud); spin_unlock(ptl); } int vmf_insert_pfn_pud(struct vm_area_struct *vma, unsigned long addr, pud_t *pud, pfn_t pfn, bool write) { pgprot_t pgprot = vma->vm_page_prot; /* * If we had pud_special, we could avoid all these restrictions, * but we need to be consistent with PTEs and architectures that * can't support a 'special' bit. */ BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP))); BUG_ON((vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)) == (VM_PFNMAP|VM_MIXEDMAP)); BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags)); BUG_ON(!pfn_t_devmap(pfn)); if (addr < vma->vm_start || addr >= vma->vm_end) return VM_FAULT_SIGBUS; track_pfn_insert(vma, &pgprot, pfn); insert_pfn_pud(vma, addr, pud, pfn, pgprot, write); return VM_FAULT_NOPAGE; } EXPORT_SYMBOL_GPL(vmf_insert_pfn_pud); #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */ |
3565fce3a mm, x86: get_user... |
822 |
static void touch_pmd(struct vm_area_struct *vma, unsigned long addr, |
01ca97274 mm, thp: Do not m... |
823 |
pmd_t *pmd, int flags) |
3565fce3a mm, x86: get_user... |
824 825 |
{ pmd_t _pmd; |
01ca97274 mm, thp: Do not m... |
826 827 828 |
_pmd = pmd_mkyoung(*pmd); if (flags & FOLL_WRITE) _pmd = pmd_mkdirty(_pmd); |
3565fce3a mm, x86: get_user... |
829 |
if (pmdp_set_access_flags(vma, addr & HPAGE_PMD_MASK, |
01ca97274 mm, thp: Do not m... |
830 |
pmd, _pmd, flags & FOLL_WRITE)) |
3565fce3a mm, x86: get_user... |
831 832 833 834 835 836 837 838 839 840 841 842 |
update_mmu_cache_pmd(vma, addr, pmd); } struct page *follow_devmap_pmd(struct vm_area_struct *vma, unsigned long addr, pmd_t *pmd, int flags) { unsigned long pfn = pmd_pfn(*pmd); struct mm_struct *mm = vma->vm_mm; struct dev_pagemap *pgmap; struct page *page; assert_spin_locked(pmd_lockptr(mm, pmd)); |
8310d48b1 mm/huge_memory.c:... |
843 844 845 846 847 |
/* * When we COW a devmap PMD entry, we split it into PTEs, so we should * not be in this function with `flags & FOLL_COW` set. */ WARN_ONCE(flags & FOLL_COW, "mm: In follow_devmap_pmd with FOLL_COW set"); |
3565fce3a mm, x86: get_user... |
848 849 850 851 852 853 854 855 856 |
if (flags & FOLL_WRITE && !pmd_write(*pmd)) return NULL; if (pmd_present(*pmd) && pmd_devmap(*pmd)) /* pass */; else return NULL; if (flags & FOLL_TOUCH) |
01ca97274 mm, thp: Do not m... |
857 |
touch_pmd(vma, addr, pmd, flags); |
3565fce3a mm, x86: get_user... |
858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 |
/* * device mapped pages can only be returned if the * caller will manage the page reference count. */ if (!(flags & FOLL_GET)) return ERR_PTR(-EEXIST); pfn += (addr & ~PMD_MASK) >> PAGE_SHIFT; pgmap = get_dev_pagemap(pfn, NULL); if (!pgmap) return ERR_PTR(-EFAULT); page = pfn_to_page(pfn); get_page(page); put_dev_pagemap(pgmap); return page; } |
71e3aac07 thp: transparent ... |
876 877 878 879 |
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... |
880 |
spinlock_t *dst_ptl, *src_ptl; |
71e3aac07 thp: transparent ... |
881 882 |
struct page *src_page; pmd_t pmd; |
12c9d70bd mm: fix memory le... |
883 |
pgtable_t pgtable = NULL; |
628d47ce9 thp: skip file hu... |
884 |
int ret = -ENOMEM; |
71e3aac07 thp: transparent ... |
885 |
|
628d47ce9 thp: skip file hu... |
886 887 888 889 890 891 892 |
/* Skip if can be re-fill on fault */ if (!vma_is_anonymous(vma)) return 0; pgtable = pte_alloc_one(dst_mm, addr); if (unlikely(!pgtable)) goto out; |
71e3aac07 thp: transparent ... |
893 |
|
c4088ebdc mm: convert the r... |
894 895 896 |
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 ... |
897 898 899 |
ret = -EAGAIN; pmd = *src_pmd; |
84c3fc4e9 mm: thp: check pm... |
900 901 902 903 904 905 906 907 908 |
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION if (unlikely(is_swap_pmd(pmd))) { swp_entry_t entry = pmd_to_swp_entry(pmd); VM_BUG_ON(!is_pmd_migration_entry(pmd)); if (is_write_migration_entry(entry)) { make_migration_entry_read(&entry); pmd = swp_entry_to_pmd(entry); |
ab6e3d093 mm: soft-dirty: k... |
909 910 |
if (pmd_swp_soft_dirty(*src_pmd)) pmd = pmd_swp_mksoft_dirty(pmd); |
84c3fc4e9 mm: thp: check pm... |
911 912 |
set_pmd_at(src_mm, addr, src_pmd, pmd); } |
dd8a67f9a mm/huge_memory.c:... |
913 914 915 |
add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR); atomic_long_inc(&dst_mm->nr_ptes); pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable); |
84c3fc4e9 mm: thp: check pm... |
916 917 918 919 920 |
set_pmd_at(dst_mm, addr, dst_pmd, pmd); ret = 0; goto out_unlock; } #endif |
628d47ce9 thp: skip file hu... |
921 |
if (unlikely(!pmd_trans_huge(pmd))) { |
71e3aac07 thp: transparent ... |
922 923 924 |
pte_free(dst_mm, pgtable); goto out_unlock; } |
fc9fe822f thp: copy_huge_pm... |
925 |
/* |
c4088ebdc mm: convert the r... |
926 |
* When page table lock is held, the huge zero pmd should not be |
fc9fe822f thp: copy_huge_pm... |
927 928 929 930 |
* 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... |
931 |
struct page *zero_page; |
97ae17497 thp: implement re... |
932 933 934 935 936 |
/* * 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. */ |
6fcb52a56 thp: reduce usage... |
937 |
zero_page = mm_get_huge_zero_page(dst_mm); |
6b251fc96 userfaultfd: call... |
938 |
set_huge_zero_page(pgtable, dst_mm, vma, addr, dst_pmd, |
5918d10a4 thp: fix huge zer... |
939 |
zero_page); |
fc9fe822f thp: copy_huge_pm... |
940 941 942 |
ret = 0; goto out_unlock; } |
de466bd62 mm: numa: avoid u... |
943 |
|
628d47ce9 thp: skip file hu... |
944 945 946 947 948 949 950 |
src_page = pmd_page(pmd); VM_BUG_ON_PAGE(!PageHead(src_page), src_page); get_page(src_page); page_dup_rmap(src_page, true); add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR); atomic_long_inc(&dst_mm->nr_ptes); pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable); |
71e3aac07 thp: transparent ... |
951 952 953 954 |
pmdp_set_wrprotect(src_mm, addr, src_pmd); pmd = pmd_mkold(pmd_wrprotect(pmd)); set_pmd_at(dst_mm, addr, dst_pmd, pmd); |
71e3aac07 thp: transparent ... |
955 956 957 |
ret = 0; out_unlock: |
c4088ebdc mm: convert the r... |
958 959 |
spin_unlock(src_ptl); spin_unlock(dst_ptl); |
71e3aac07 thp: transparent ... |
960 961 962 |
out: return ret; } |
a00cc7d9d mm, x86: add supp... |
963 964 |
#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD static void touch_pud(struct vm_area_struct *vma, unsigned long addr, |
01ca97274 mm, thp: Do not m... |
965 |
pud_t *pud, int flags) |
a00cc7d9d mm, x86: add supp... |
966 967 |
{ pud_t _pud; |
01ca97274 mm, thp: Do not m... |
968 969 970 |
_pud = pud_mkyoung(*pud); if (flags & FOLL_WRITE) _pud = pud_mkdirty(_pud); |
a00cc7d9d mm, x86: add supp... |
971 |
if (pudp_set_access_flags(vma, addr & HPAGE_PUD_MASK, |
01ca97274 mm, thp: Do not m... |
972 |
pud, _pud, flags & FOLL_WRITE)) |
a00cc7d9d mm, x86: add supp... |
973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 |
update_mmu_cache_pud(vma, addr, pud); } struct page *follow_devmap_pud(struct vm_area_struct *vma, unsigned long addr, pud_t *pud, int flags) { unsigned long pfn = pud_pfn(*pud); struct mm_struct *mm = vma->vm_mm; struct dev_pagemap *pgmap; struct page *page; assert_spin_locked(pud_lockptr(mm, pud)); if (flags & FOLL_WRITE && !pud_write(*pud)) return NULL; if (pud_present(*pud) && pud_devmap(*pud)) /* pass */; else return NULL; if (flags & FOLL_TOUCH) |
01ca97274 mm, thp: Do not m... |
995 |
touch_pud(vma, addr, pud, flags); |
a00cc7d9d mm, x86: add supp... |
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 |
/* * device mapped pages can only be returned if the * caller will manage the page reference count. */ if (!(flags & FOLL_GET)) return ERR_PTR(-EEXIST); pfn += (addr & ~PUD_MASK) >> PAGE_SHIFT; pgmap = get_dev_pagemap(pfn, NULL); if (!pgmap) return ERR_PTR(-EFAULT); page = pfn_to_page(pfn); get_page(page); put_dev_pagemap(pgmap); return page; } int copy_huge_pud(struct mm_struct *dst_mm, struct mm_struct *src_mm, pud_t *dst_pud, pud_t *src_pud, unsigned long addr, struct vm_area_struct *vma) { spinlock_t *dst_ptl, *src_ptl; pud_t pud; int ret; dst_ptl = pud_lock(dst_mm, dst_pud); src_ptl = pud_lockptr(src_mm, src_pud); spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING); ret = -EAGAIN; pud = *src_pud; if (unlikely(!pud_trans_huge(pud) && !pud_devmap(pud))) goto out_unlock; /* * When page table lock is held, the huge zero pud should not be * under splitting since we don't split the page itself, only pud to * a page table. */ if (is_huge_zero_pud(pud)) { /* No huge zero pud yet */ } pudp_set_wrprotect(src_mm, addr, src_pud); pud = pud_mkold(pud_wrprotect(pud)); set_pud_at(dst_mm, addr, dst_pud, pud); ret = 0; out_unlock: spin_unlock(src_ptl); spin_unlock(dst_ptl); return ret; } void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud) { pud_t entry; unsigned long haddr; bool write = vmf->flags & FAULT_FLAG_WRITE; vmf->ptl = pud_lock(vmf->vma->vm_mm, vmf->pud); if (unlikely(!pud_same(*vmf->pud, orig_pud))) goto unlock; entry = pud_mkyoung(orig_pud); if (write) entry = pud_mkdirty(entry); haddr = vmf->address & HPAGE_PUD_MASK; if (pudp_set_access_flags(vmf->vma, haddr, vmf->pud, entry, write)) update_mmu_cache_pud(vmf->vma, vmf->address, vmf->pud); unlock: spin_unlock(vmf->ptl); } #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */ |
82b0f8c39 mm: join struct f... |
1073 |
void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd) |
a1dd450bc mm: thp: set the ... |
1074 1075 1076 |
{ pmd_t entry; unsigned long haddr; |
20f664aab mm: pmd dirty emu... |
1077 |
bool write = vmf->flags & FAULT_FLAG_WRITE; |
a1dd450bc mm: thp: set the ... |
1078 |
|
82b0f8c39 mm: join struct f... |
1079 1080 |
vmf->ptl = pmd_lock(vmf->vma->vm_mm, vmf->pmd); if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) |
a1dd450bc mm: thp: set the ... |
1081 1082 1083 |
goto unlock; entry = pmd_mkyoung(orig_pmd); |
20f664aab mm: pmd dirty emu... |
1084 1085 |
if (write) entry = pmd_mkdirty(entry); |
82b0f8c39 mm: join struct f... |
1086 |
haddr = vmf->address & HPAGE_PMD_MASK; |
20f664aab mm: pmd dirty emu... |
1087 |
if (pmdp_set_access_flags(vmf->vma, haddr, vmf->pmd, entry, write)) |
82b0f8c39 mm: join struct f... |
1088 |
update_mmu_cache_pmd(vmf->vma, vmf->address, vmf->pmd); |
a1dd450bc mm: thp: set the ... |
1089 1090 |
unlock: |
82b0f8c39 mm: join struct f... |
1091 |
spin_unlock(vmf->ptl); |
a1dd450bc mm: thp: set the ... |
1092 |
} |
82b0f8c39 mm: join struct f... |
1093 |
static int do_huge_pmd_wp_page_fallback(struct vm_fault *vmf, pmd_t orig_pmd, |
bae473a42 mm: introduce fau... |
1094 |
struct page *page) |
71e3aac07 thp: transparent ... |
1095 |
{ |
82b0f8c39 mm: join struct f... |
1096 1097 |
struct vm_area_struct *vma = vmf->vma; unsigned long haddr = vmf->address & HPAGE_PMD_MASK; |
00501b531 mm: memcontrol: r... |
1098 |
struct mem_cgroup *memcg; |
71e3aac07 thp: transparent ... |
1099 1100 1101 1102 |
pgtable_t pgtable; pmd_t _pmd; int ret = 0, i; struct page **pages; |
2ec74c3ef mm: move all mmu ... |
1103 1104 |
unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
71e3aac07 thp: transparent ... |
1105 1106 1107 1108 1109 1110 1111 1112 1113 |
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++) { |
41b6167e8 mm: get rid of __... |
1114 |
pages[i] = alloc_page_vma_node(GFP_HIGHUSER_MOVABLE, vma, |
82b0f8c39 mm: join struct f... |
1115 |
vmf->address, page_to_nid(page)); |
b9bbfbe30 thp: memcg huge m... |
1116 |
if (unlikely(!pages[i] || |
bae473a42 mm: introduce fau... |
1117 1118 |
mem_cgroup_try_charge(pages[i], vma->vm_mm, GFP_KERNEL, &memcg, false))) { |
b9bbfbe30 thp: memcg huge m... |
1119 |
if (pages[i]) |
71e3aac07 thp: transparent ... |
1120 |
put_page(pages[i]); |
b9bbfbe30 thp: memcg huge m... |
1121 |
while (--i >= 0) { |
00501b531 mm: memcontrol: r... |
1122 1123 |
memcg = (void *)page_private(pages[i]); set_page_private(pages[i], 0); |
f627c2f53 memcg: adjust to ... |
1124 1125 |
mem_cgroup_cancel_charge(pages[i], memcg, false); |
b9bbfbe30 thp: memcg huge m... |
1126 1127 |
put_page(pages[i]); } |
71e3aac07 thp: transparent ... |
1128 1129 1130 1131 |
kfree(pages); ret |= VM_FAULT_OOM; goto out; } |
00501b531 mm: memcontrol: r... |
1132 |
set_page_private(pages[i], (unsigned long)memcg); |
71e3aac07 thp: transparent ... |
1133 1134 1135 1136 |
} for (i = 0; i < HPAGE_PMD_NR; i++) { copy_user_highpage(pages[i], page + i, |
0089e4853 mm/huge_memory: f... |
1137 |
haddr + PAGE_SIZE * i, vma); |
71e3aac07 thp: transparent ... |
1138 1139 1140 |
__SetPageUptodate(pages[i]); cond_resched(); } |
2ec74c3ef mm: move all mmu ... |
1141 1142 |
mmun_start = haddr; mmun_end = haddr + HPAGE_PMD_SIZE; |
bae473a42 mm: introduce fau... |
1143 |
mmu_notifier_invalidate_range_start(vma->vm_mm, mmun_start, mmun_end); |
2ec74c3ef mm: move all mmu ... |
1144 |
|
82b0f8c39 mm: join struct f... |
1145 1146 |
vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd); if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) |
71e3aac07 thp: transparent ... |
1147 |
goto out_free_pages; |
309381fea mm: dump page whe... |
1148 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
71e3aac07 thp: transparent ... |
1149 |
|
82b0f8c39 mm: join struct f... |
1150 |
pmdp_huge_clear_flush_notify(vma, haddr, vmf->pmd); |
71e3aac07 thp: transparent ... |
1151 |
/* leave pmd empty until pte is filled */ |
82b0f8c39 mm: join struct f... |
1152 |
pgtable = pgtable_trans_huge_withdraw(vma->vm_mm, vmf->pmd); |
bae473a42 mm: introduce fau... |
1153 |
pmd_populate(vma->vm_mm, &_pmd, pgtable); |
71e3aac07 thp: transparent ... |
1154 1155 |
for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) { |
bae473a42 mm: introduce fau... |
1156 |
pte_t entry; |
71e3aac07 thp: transparent ... |
1157 1158 |
entry = mk_pte(pages[i], vma->vm_page_prot); entry = maybe_mkwrite(pte_mkdirty(entry), vma); |
00501b531 mm: memcontrol: r... |
1159 1160 |
memcg = (void *)page_private(pages[i]); set_page_private(pages[i], 0); |
82b0f8c39 mm: join struct f... |
1161 |
page_add_new_anon_rmap(pages[i], vmf->vma, haddr, false); |
f627c2f53 memcg: adjust to ... |
1162 |
mem_cgroup_commit_charge(pages[i], memcg, false, false); |
00501b531 mm: memcontrol: r... |
1163 |
lru_cache_add_active_or_unevictable(pages[i], vma); |
82b0f8c39 mm: join struct f... |
1164 1165 1166 1167 |
vmf->pte = pte_offset_map(&_pmd, haddr); VM_BUG_ON(!pte_none(*vmf->pte)); set_pte_at(vma->vm_mm, haddr, vmf->pte, entry); pte_unmap(vmf->pte); |
71e3aac07 thp: transparent ... |
1168 1169 |
} kfree(pages); |
71e3aac07 thp: transparent ... |
1170 |
smp_wmb(); /* make pte visible before pmd */ |
82b0f8c39 mm: join struct f... |
1171 |
pmd_populate(vma->vm_mm, vmf->pmd, pgtable); |
d281ee614 rmap: add argumen... |
1172 |
page_remove_rmap(page, true); |
82b0f8c39 mm: join struct f... |
1173 |
spin_unlock(vmf->ptl); |
71e3aac07 thp: transparent ... |
1174 |
|
bae473a42 mm: introduce fau... |
1175 |
mmu_notifier_invalidate_range_end(vma->vm_mm, mmun_start, mmun_end); |
2ec74c3ef mm: move all mmu ... |
1176 |
|
71e3aac07 thp: transparent ... |
1177 1178 1179 1180 1181 1182 1183 |
ret |= VM_FAULT_WRITE; put_page(page); out: return ret; out_free_pages: |
82b0f8c39 mm: join struct f... |
1184 |
spin_unlock(vmf->ptl); |
bae473a42 mm: introduce fau... |
1185 |
mmu_notifier_invalidate_range_end(vma->vm_mm, mmun_start, mmun_end); |
b9bbfbe30 thp: memcg huge m... |
1186 |
for (i = 0; i < HPAGE_PMD_NR; i++) { |
00501b531 mm: memcontrol: r... |
1187 1188 |
memcg = (void *)page_private(pages[i]); set_page_private(pages[i], 0); |
f627c2f53 memcg: adjust to ... |
1189 |
mem_cgroup_cancel_charge(pages[i], memcg, false); |
71e3aac07 thp: transparent ... |
1190 |
put_page(pages[i]); |
b9bbfbe30 thp: memcg huge m... |
1191 |
} |
71e3aac07 thp: transparent ... |
1192 1193 1194 |
kfree(pages); goto out; } |
82b0f8c39 mm: join struct f... |
1195 |
int do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd) |
71e3aac07 thp: transparent ... |
1196 |
{ |
82b0f8c39 mm: join struct f... |
1197 |
struct vm_area_struct *vma = vmf->vma; |
93b4796de thp: do_huge_pmd_... |
1198 |
struct page *page = NULL, *new_page; |
00501b531 mm: memcontrol: r... |
1199 |
struct mem_cgroup *memcg; |
82b0f8c39 mm: join struct f... |
1200 |
unsigned long haddr = vmf->address & HPAGE_PMD_MASK; |
2ec74c3ef mm: move all mmu ... |
1201 1202 |
unsigned long mmun_start; /* For mmu_notifiers */ unsigned long mmun_end; /* For mmu_notifiers */ |
3b3636924 mm, memcg: sync a... |
1203 |
gfp_t huge_gfp; /* for allocation and charge */ |
bae473a42 mm: introduce fau... |
1204 |
int ret = 0; |
71e3aac07 thp: transparent ... |
1205 |
|
82b0f8c39 mm: join struct f... |
1206 |
vmf->ptl = pmd_lockptr(vma->vm_mm, vmf->pmd); |
81d1b09c6 mm: convert a few... |
1207 |
VM_BUG_ON_VMA(!vma->anon_vma, vma); |
93b4796de thp: do_huge_pmd_... |
1208 1209 |
if (is_huge_zero_pmd(orig_pmd)) goto alloc; |
82b0f8c39 mm: join struct f... |
1210 1211 |
spin_lock(vmf->ptl); if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) |
71e3aac07 thp: transparent ... |
1212 1213 1214 |
goto out_unlock; page = pmd_page(orig_pmd); |
309381fea mm: dump page whe... |
1215 |
VM_BUG_ON_PAGE(!PageCompound(page) || !PageHead(page), page); |
1f25fe20a mm, thp: adjust c... |
1216 1217 |
/* * We can only reuse the page if nobody else maps the huge page or it's |
6d0a07edd mm: thp: calculat... |
1218 |
* part. |
1f25fe20a mm, thp: adjust c... |
1219 |
*/ |
ba3c4ce6d mm, THP, swap: ma... |
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 |
if (!trylock_page(page)) { get_page(page); spin_unlock(vmf->ptl); lock_page(page); spin_lock(vmf->ptl); if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) { unlock_page(page); put_page(page); goto out_unlock; } put_page(page); } if (reuse_swap_page(page, NULL)) { |
71e3aac07 thp: transparent ... |
1233 1234 1235 |
pmd_t entry; entry = pmd_mkyoung(orig_pmd); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); |
82b0f8c39 mm: join struct f... |
1236 1237 |
if (pmdp_set_access_flags(vma, haddr, vmf->pmd, entry, 1)) update_mmu_cache_pmd(vma, vmf->address, vmf->pmd); |
71e3aac07 thp: transparent ... |
1238 |
ret |= VM_FAULT_WRITE; |
ba3c4ce6d mm, THP, swap: ma... |
1239 |
unlock_page(page); |
71e3aac07 thp: transparent ... |
1240 1241 |
goto out_unlock; } |
ba3c4ce6d mm, THP, swap: ma... |
1242 |
unlock_page(page); |
ddc58f27f mm: drop tail pag... |
1243 |
get_page(page); |
82b0f8c39 mm: join struct f... |
1244 |
spin_unlock(vmf->ptl); |
93b4796de thp: do_huge_pmd_... |
1245 |
alloc: |
71e3aac07 thp: transparent ... |
1246 |
if (transparent_hugepage_enabled(vma) && |
077fcf116 mm/thp: allocate ... |
1247 |
!transparent_hugepage_debug_cow()) { |
444eb2a44 mm: thp: set THP ... |
1248 |
huge_gfp = alloc_hugepage_direct_gfpmask(vma); |
3b3636924 mm, memcg: sync a... |
1249 |
new_page = alloc_hugepage_vma(huge_gfp, vma, haddr, HPAGE_PMD_ORDER); |
077fcf116 mm/thp: allocate ... |
1250 |
} else |
71e3aac07 thp: transparent ... |
1251 |
new_page = NULL; |
9a982250f thp: introduce de... |
1252 1253 1254 |
if (likely(new_page)) { prep_transhuge_page(new_page); } else { |
eecc1e426 thp: fix copy_pag... |
1255 |
if (!page) { |
82b0f8c39 mm: join struct f... |
1256 |
split_huge_pmd(vma, vmf->pmd, vmf->address); |
e9b71ca91 mm, thp: drop do_... |
1257 |
ret |= VM_FAULT_FALLBACK; |
93b4796de thp: do_huge_pmd_... |
1258 |
} else { |
82b0f8c39 mm: join struct f... |
1259 |
ret = do_huge_pmd_wp_page_fallback(vmf, orig_pmd, page); |
9845cbbd1 mm, thp: fix infi... |
1260 |
if (ret & VM_FAULT_OOM) { |
82b0f8c39 mm: join struct f... |
1261 |
split_huge_pmd(vma, vmf->pmd, vmf->address); |
9845cbbd1 mm, thp: fix infi... |
1262 1263 |
ret |= VM_FAULT_FALLBACK; } |
ddc58f27f mm: drop tail pag... |
1264 |
put_page(page); |
93b4796de thp: do_huge_pmd_... |
1265 |
} |
17766dde3 mm, thp: count th... |
1266 |
count_vm_event(THP_FAULT_FALLBACK); |
71e3aac07 thp: transparent ... |
1267 1268 |
goto out; } |
bae473a42 mm: introduce fau... |
1269 |
if (unlikely(mem_cgroup_try_charge(new_page, vma->vm_mm, |
49f4a8c52 mm, thp: do not c... |
1270 |
huge_gfp | __GFP_NORETRY, &memcg, true))) { |
b9bbfbe30 thp: memcg huge m... |
1271 |
put_page(new_page); |
82b0f8c39 mm: join struct f... |
1272 |
split_huge_pmd(vma, vmf->pmd, vmf->address); |
bae473a42 mm: introduce fau... |
1273 |
if (page) |
ddc58f27f mm: drop tail pag... |
1274 |
put_page(page); |
9845cbbd1 mm, thp: fix infi... |
1275 |
ret |= VM_FAULT_FALLBACK; |
17766dde3 mm, thp: count th... |
1276 |
count_vm_event(THP_FAULT_FALLBACK); |
b9bbfbe30 thp: memcg huge m... |
1277 1278 |
goto out; } |
17766dde3 mm, thp: count th... |
1279 |
count_vm_event(THP_FAULT_ALLOC); |
eecc1e426 thp: fix copy_pag... |
1280 |
if (!page) |
c79b57e46 mm: hugetlb: clea... |
1281 |
clear_huge_page(new_page, vmf->address, HPAGE_PMD_NR); |
93b4796de thp: do_huge_pmd_... |
1282 1283 |
else copy_user_huge_page(new_page, page, haddr, vma, HPAGE_PMD_NR); |
71e3aac07 thp: transparent ... |
1284 |
__SetPageUptodate(new_page); |
2ec74c3ef mm: move all mmu ... |
1285 1286 |
mmun_start = haddr; mmun_end = haddr + HPAGE_PMD_SIZE; |
bae473a42 mm: introduce fau... |
1287 |
mmu_notifier_invalidate_range_start(vma->vm_mm, mmun_start, mmun_end); |
2ec74c3ef mm: move all mmu ... |
1288 |
|
82b0f8c39 mm: join struct f... |
1289 |
spin_lock(vmf->ptl); |
93b4796de thp: do_huge_pmd_... |
1290 |
if (page) |
ddc58f27f mm: drop tail pag... |
1291 |
put_page(page); |
82b0f8c39 mm: join struct f... |
1292 1293 |
if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) { spin_unlock(vmf->ptl); |
f627c2f53 memcg: adjust to ... |
1294 |
mem_cgroup_cancel_charge(new_page, memcg, true); |
71e3aac07 thp: transparent ... |
1295 |
put_page(new_page); |
2ec74c3ef mm: move all mmu ... |
1296 |
goto out_mn; |
b9bbfbe30 thp: memcg huge m... |
1297 |
} else { |
71e3aac07 thp: transparent ... |
1298 |
pmd_t entry; |
3122359a6 thp: move maybe_p... |
1299 1300 |
entry = mk_huge_pmd(new_page, vma->vm_page_prot); entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); |
82b0f8c39 mm: join struct f... |
1301 |
pmdp_huge_clear_flush_notify(vma, haddr, vmf->pmd); |
d281ee614 rmap: add argumen... |
1302 |
page_add_new_anon_rmap(new_page, vma, haddr, true); |
f627c2f53 memcg: adjust to ... |
1303 |
mem_cgroup_commit_charge(new_page, memcg, false, true); |
00501b531 mm: memcontrol: r... |
1304 |
lru_cache_add_active_or_unevictable(new_page, vma); |
82b0f8c39 mm: join struct f... |
1305 1306 |
set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry); update_mmu_cache_pmd(vma, vmf->address, vmf->pmd); |
eecc1e426 thp: fix copy_pag... |
1307 |
if (!page) { |
bae473a42 mm: introduce fau... |
1308 |
add_mm_counter(vma->vm_mm, MM_ANONPAGES, HPAGE_PMD_NR); |
97ae17497 thp: implement re... |
1309 |
} else { |
309381fea mm: dump page whe... |
1310 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
d281ee614 rmap: add argumen... |
1311 |
page_remove_rmap(page, true); |
93b4796de thp: do_huge_pmd_... |
1312 1313 |
put_page(page); } |
71e3aac07 thp: transparent ... |
1314 1315 |
ret |= VM_FAULT_WRITE; } |
82b0f8c39 mm: join struct f... |
1316 |
spin_unlock(vmf->ptl); |
2ec74c3ef mm: move all mmu ... |
1317 |
out_mn: |
bae473a42 mm: introduce fau... |
1318 |
mmu_notifier_invalidate_range_end(vma->vm_mm, mmun_start, mmun_end); |
71e3aac07 thp: transparent ... |
1319 1320 |
out: return ret; |
2ec74c3ef mm: move all mmu ... |
1321 |
out_unlock: |
82b0f8c39 mm: join struct f... |
1322 |
spin_unlock(vmf->ptl); |
2ec74c3ef mm: move all mmu ... |
1323 |
return ret; |
71e3aac07 thp: transparent ... |
1324 |
} |
8310d48b1 mm/huge_memory.c:... |
1325 1326 1327 1328 1329 1330 1331 1332 1333 |
/* * FOLL_FORCE can write to even unwritable pmd's, but only * after we've gone through a COW cycle and they are dirty. */ static inline bool can_follow_write_pmd(pmd_t pmd, unsigned int flags) { return pmd_write(pmd) || ((flags & FOLL_FORCE) && (flags & FOLL_COW) && pmd_dirty(pmd)); } |
b676b293f mm, thp: fix mapp... |
1334 |
struct page *follow_trans_huge_pmd(struct vm_area_struct *vma, |
71e3aac07 thp: transparent ... |
1335 1336 1337 1338 |
unsigned long addr, pmd_t *pmd, unsigned int flags) { |
b676b293f mm, thp: fix mapp... |
1339 |
struct mm_struct *mm = vma->vm_mm; |
71e3aac07 thp: transparent ... |
1340 |
struct page *page = NULL; |
c4088ebdc mm: convert the r... |
1341 |
assert_spin_locked(pmd_lockptr(mm, pmd)); |
71e3aac07 thp: transparent ... |
1342 |
|
8310d48b1 mm/huge_memory.c:... |
1343 |
if (flags & FOLL_WRITE && !can_follow_write_pmd(*pmd, flags)) |
71e3aac07 thp: transparent ... |
1344 |
goto out; |
85facf257 thp: avoid dumpin... |
1345 1346 1347 |
/* Avoid dumping huge zero page */ if ((flags & FOLL_DUMP) && is_huge_zero_pmd(*pmd)) return ERR_PTR(-EFAULT); |
2b4847e73 mm: numa: seriali... |
1348 |
/* Full NUMA hinting faults to serialise migration in fault paths */ |
8a0516ed8 mm: convert p[te|... |
1349 |
if ((flags & FOLL_NUMA) && pmd_protnone(*pmd)) |
2b4847e73 mm: numa: seriali... |
1350 |
goto out; |
71e3aac07 thp: transparent ... |
1351 |
page = pmd_page(*pmd); |
ca120cf68 mm: fix show_smap... |
1352 |
VM_BUG_ON_PAGE(!PageHead(page) && !is_zone_device_page(page), page); |
3565fce3a mm, x86: get_user... |
1353 |
if (flags & FOLL_TOUCH) |
01ca97274 mm, thp: Do not m... |
1354 |
touch_pmd(vma, addr, pmd, flags); |
de60f5f10 mm: introduce VM_... |
1355 |
if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) { |
e90309c9f thp: allow mlocke... |
1356 1357 1358 1359 |
/* * We don't mlock() pte-mapped THPs. This way we can avoid * leaking mlocked pages into non-VM_LOCKED VMAs. * |
9a73f61bd thp, mlock: do no... |
1360 1361 |
* For anon THP: * |
e90309c9f thp: allow mlocke... |
1362 1363 1364 1365 1366 1367 1368 |
* In most cases the pmd is the only mapping of the page as we * break COW for the mlock() -- see gup_flags |= FOLL_WRITE for * writable private mappings in populate_vma_page_range(). * * The only scenario when we have the page shared here is if we * mlocking read-only mapping shared over fork(). We skip * mlocking such pages. |
9a73f61bd thp, mlock: do no... |
1369 1370 1371 1372 1373 1374 |
* * For file THP: * * We can expect PageDoubleMap() to be stable under page lock: * for file pages we set it in page_add_file_rmap(), which * requires page to be locked. |
e90309c9f thp: allow mlocke... |
1375 |
*/ |
9a73f61bd thp, mlock: do no... |
1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 |
if (PageAnon(page) && compound_mapcount(page) != 1) goto skip_mlock; if (PageDoubleMap(page) || !page->mapping) goto skip_mlock; if (!trylock_page(page)) goto skip_mlock; lru_add_drain(); if (page->mapping && !PageDoubleMap(page)) mlock_vma_page(page); unlock_page(page); |
b676b293f mm, thp: fix mapp... |
1387 |
} |
9a73f61bd thp, mlock: do no... |
1388 |
skip_mlock: |
71e3aac07 thp: transparent ... |
1389 |
page += (addr & ~HPAGE_PMD_MASK) >> PAGE_SHIFT; |
ca120cf68 mm: fix show_smap... |
1390 |
VM_BUG_ON_PAGE(!PageCompound(page) && !is_zone_device_page(page), page); |
71e3aac07 thp: transparent ... |
1391 |
if (flags & FOLL_GET) |
ddc58f27f mm: drop tail pag... |
1392 |
get_page(page); |
71e3aac07 thp: transparent ... |
1393 1394 1395 1396 |
out: return page; } |
d10e63f29 mm: numa: Create ... |
1397 |
/* NUMA hinting page fault entry point for trans huge pmds */ |
82b0f8c39 mm: join struct f... |
1398 |
int do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd) |
d10e63f29 mm: numa: Create ... |
1399 |
{ |
82b0f8c39 mm: join struct f... |
1400 |
struct vm_area_struct *vma = vmf->vma; |
b8916634b mm: Prevent paral... |
1401 |
struct anon_vma *anon_vma = NULL; |
b32967ff1 mm: numa: Add THP... |
1402 |
struct page *page; |
82b0f8c39 mm: join struct f... |
1403 |
unsigned long haddr = vmf->address & HPAGE_PMD_MASK; |
8191acbd3 mm: numa: Sanitiz... |
1404 |
int page_nid = -1, this_nid = numa_node_id(); |
90572890d mm: numa: Change ... |
1405 |
int target_nid, last_cpupid = -1; |
8191acbd3 mm: numa: Sanitiz... |
1406 1407 |
bool page_locked; bool migrated = false; |
b191f9b10 mm: numa: preserv... |
1408 |
bool was_writable; |
6688cc054 mm: numa: Do not ... |
1409 |
int flags = 0; |
d10e63f29 mm: numa: Create ... |
1410 |
|
82b0f8c39 mm: join struct f... |
1411 1412 |
vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd); if (unlikely(!pmd_same(pmd, *vmf->pmd))) |
d10e63f29 mm: numa: Create ... |
1413 |
goto out_unlock; |
de466bd62 mm: numa: avoid u... |
1414 1415 1416 1417 1418 |
/* * 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. */ |
82b0f8c39 mm: join struct f... |
1419 1420 |
if (unlikely(pmd_trans_migrating(*vmf->pmd))) { page = pmd_page(*vmf->pmd); |
3c226c637 mm: numa: avoid w... |
1421 1422 |
if (!get_page_unless_zero(page)) goto out_unlock; |
82b0f8c39 mm: join struct f... |
1423 |
spin_unlock(vmf->ptl); |
5d8330621 mm: numa: do not ... |
1424 |
wait_on_page_locked(page); |
3c226c637 mm: numa: avoid w... |
1425 |
put_page(page); |
de466bd62 mm: numa: avoid u... |
1426 1427 |
goto out; } |
d10e63f29 mm: numa: Create ... |
1428 |
page = pmd_page(pmd); |
a1a46184e mm: numa: Do not ... |
1429 |
BUG_ON(is_huge_zero_page(page)); |
8191acbd3 mm: numa: Sanitiz... |
1430 |
page_nid = page_to_nid(page); |
90572890d mm: numa: Change ... |
1431 |
last_cpupid = page_cpupid_last(page); |
03c5a6e16 mm: numa: Add pte... |
1432 |
count_vm_numa_event(NUMA_HINT_FAULTS); |
04bb2f947 sched/numa: Adjus... |
1433 |
if (page_nid == this_nid) { |
03c5a6e16 mm: numa: Add pte... |
1434 |
count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL); |
04bb2f947 sched/numa: Adjus... |
1435 1436 |
flags |= TNF_FAULT_LOCAL; } |
4daae3b4b mm: mempolicy: Us... |
1437 |
|
bea66fbd1 mm: numa: group r... |
1438 |
/* See similar comment in do_numa_page for explanation */ |
288bc5494 mm/autonuma: let ... |
1439 |
if (!pmd_savedwrite(pmd)) |
6688cc054 mm: numa: Do not ... |
1440 1441 1442 |
flags |= TNF_NO_GROUP; /* |
ff9042b11 mm: Wait for THP ... |
1443 1444 1445 |
* Acquire the page lock to serialise THP migrations but avoid dropping * page_table_lock if at all possible */ |
b8916634b mm: Prevent paral... |
1446 1447 1448 1449 |
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... |
1450 |
if (page_locked) |
b8916634b mm: Prevent paral... |
1451 |
goto clear_pmdnuma; |
2b4847e73 mm: numa: seriali... |
1452 |
} |
4daae3b4b mm: mempolicy: Us... |
1453 |
|
de466bd62 mm: numa: avoid u... |
1454 |
/* Migration could have started since the pmd_trans_migrating check */ |
2b4847e73 mm: numa: seriali... |
1455 |
if (!page_locked) { |
3c226c637 mm: numa: avoid w... |
1456 1457 1458 |
page_nid = -1; if (!get_page_unless_zero(page)) goto out_unlock; |
82b0f8c39 mm: join struct f... |
1459 |
spin_unlock(vmf->ptl); |
b8916634b mm: Prevent paral... |
1460 |
wait_on_page_locked(page); |
3c226c637 mm: numa: avoid w... |
1461 |
put_page(page); |
b8916634b mm: Prevent paral... |
1462 1463 |
goto out; } |
2b4847e73 mm: numa: seriali... |
1464 1465 1466 1467 |
/* * Page is misplaced. Page lock serialises migrations. Acquire anon_vma * to serialises splits */ |
b8916634b mm: Prevent paral... |
1468 |
get_page(page); |
82b0f8c39 mm: join struct f... |
1469 |
spin_unlock(vmf->ptl); |
b8916634b mm: Prevent paral... |
1470 |
anon_vma = page_lock_anon_vma_read(page); |
4daae3b4b mm: mempolicy: Us... |
1471 |
|
c69307d53 sched/numa: Fix c... |
1472 |
/* Confirm the PMD did not change while page_table_lock was released */ |
82b0f8c39 mm: join struct f... |
1473 1474 |
spin_lock(vmf->ptl); if (unlikely(!pmd_same(pmd, *vmf->pmd))) { |
b32967ff1 mm: numa: Add THP... |
1475 1476 |
unlock_page(page); put_page(page); |
a54a407fb mm: Close races b... |
1477 |
page_nid = -1; |
4daae3b4b mm: mempolicy: Us... |
1478 |
goto out_unlock; |
b32967ff1 mm: numa: Add THP... |
1479 |
} |
ff9042b11 mm: Wait for THP ... |
1480 |
|
c3a489cac mm: numa: ensure ... |
1481 1482 1483 1484 1485 1486 |
/* 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... |
1487 |
/* |
8b1b436dd mm, locking: Rewo... |
1488 1489 1490 1491 1492 |
* Since we took the NUMA fault, we must have observed the !accessible * bit. Make sure all other CPUs agree with that, to avoid them * modifying the page we're about to migrate. * * Must be done under PTL such that we'll observe the relevant |
ccde85ba0 mm, locking: Fix ... |
1493 1494 1495 1496 |
* inc_tlb_flush_pending(). * * We are not sure a pending tlb flush here is for a huge page * mapping or not. Hence use the tlb range variant |
8b1b436dd mm, locking: Rewo... |
1497 1498 |
*/ if (mm_tlb_flush_pending(vma->vm_mm)) |
ccde85ba0 mm, locking: Fix ... |
1499 |
flush_tlb_range(vma, haddr, haddr + HPAGE_PMD_SIZE); |
8b1b436dd mm, locking: Rewo... |
1500 1501 |
/* |
a54a407fb mm: Close races b... |
1502 |
* Migrate the THP to the requested node, returns with page unlocked |
8a0516ed8 mm: convert p[te|... |
1503 |
* and access rights restored. |
a54a407fb mm: Close races b... |
1504 |
*/ |
82b0f8c39 mm: join struct f... |
1505 |
spin_unlock(vmf->ptl); |
8b1b436dd mm, locking: Rewo... |
1506 |
|
bae473a42 mm: introduce fau... |
1507 |
migrated = migrate_misplaced_transhuge_page(vma->vm_mm, vma, |
82b0f8c39 mm: join struct f... |
1508 |
vmf->pmd, pmd, vmf->address, page, target_nid); |
6688cc054 mm: numa: Do not ... |
1509 1510 |
if (migrated) { flags |= TNF_MIGRATED; |
8191acbd3 mm: numa: Sanitiz... |
1511 |
page_nid = target_nid; |
074c23817 mm: numa: slow PT... |
1512 1513 |
} else flags |= TNF_MIGRATE_FAIL; |
b32967ff1 mm: numa: Add THP... |
1514 |
|
8191acbd3 mm: numa: Sanitiz... |
1515 |
goto out; |
b32967ff1 mm: numa: Add THP... |
1516 |
clear_pmdnuma: |
a54a407fb mm: Close races b... |
1517 |
BUG_ON(!PageLocked(page)); |
288bc5494 mm/autonuma: let ... |
1518 |
was_writable = pmd_savedwrite(pmd); |
4d9424669 mm: convert p[te|... |
1519 |
pmd = pmd_modify(pmd, vma->vm_page_prot); |
b7b04004e mm: numa: mark hu... |
1520 |
pmd = pmd_mkyoung(pmd); |
b191f9b10 mm: numa: preserv... |
1521 1522 |
if (was_writable) pmd = pmd_mkwrite(pmd); |
82b0f8c39 mm: join struct f... |
1523 1524 |
set_pmd_at(vma->vm_mm, haddr, vmf->pmd, pmd); update_mmu_cache_pmd(vma, vmf->address, vmf->pmd); |
a54a407fb mm: Close races b... |
1525 |
unlock_page(page); |
d10e63f29 mm: numa: Create ... |
1526 |
out_unlock: |
82b0f8c39 mm: join struct f... |
1527 |
spin_unlock(vmf->ptl); |
b8916634b mm: Prevent paral... |
1528 1529 1530 1531 |
out: if (anon_vma) page_unlock_anon_vma_read(anon_vma); |
8191acbd3 mm: numa: Sanitiz... |
1532 |
if (page_nid != -1) |
82b0f8c39 mm: join struct f... |
1533 |
task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR, |
9a8b300f2 mm/thp/autonuma: ... |
1534 |
flags); |
8191acbd3 mm: numa: Sanitiz... |
1535 |
|
d10e63f29 mm: numa: Create ... |
1536 1537 |
return 0; } |
319904ad4 mm, THP: clean up... |
1538 1539 1540 1541 1542 |
/* * Return true if we do MADV_FREE successfully on entire pmd page. * Otherwise, return false. */ bool madvise_free_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma, |
b8d3c4c30 mm/huge_memory.c:... |
1543 |
pmd_t *pmd, unsigned long addr, unsigned long next) |
b8d3c4c30 mm/huge_memory.c:... |
1544 1545 1546 1547 1548 |
{ spinlock_t *ptl; pmd_t orig_pmd; struct page *page; struct mm_struct *mm = tlb->mm; |
319904ad4 mm, THP: clean up... |
1549 |
bool ret = false; |
b8d3c4c30 mm/huge_memory.c:... |
1550 |
|
07e326610 mm: add tlb_remov... |
1551 |
tlb_remove_check_page_size_change(tlb, HPAGE_PMD_SIZE); |
b6ec57f4b thp: change pmd_t... |
1552 1553 |
ptl = pmd_trans_huge_lock(pmd, vma); if (!ptl) |
25eedabe0 vm: fix incorrect... |
1554 |
goto out_unlocked; |
b8d3c4c30 mm/huge_memory.c:... |
1555 1556 |
orig_pmd = *pmd; |
319904ad4 mm, THP: clean up... |
1557 |
if (is_huge_zero_pmd(orig_pmd)) |
b8d3c4c30 mm/huge_memory.c:... |
1558 |
goto out; |
b8d3c4c30 mm/huge_memory.c:... |
1559 |
|
84c3fc4e9 mm: thp: check pm... |
1560 1561 1562 1563 1564 |
if (unlikely(!pmd_present(orig_pmd))) { VM_BUG_ON(thp_migration_supported() && !is_pmd_migration_entry(orig_pmd)); goto out; } |
b8d3c4c30 mm/huge_memory.c:... |
1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 |
page = pmd_page(orig_pmd); /* * If other processes are mapping this page, we couldn't discard * the page unless they all do MADV_FREE so let's skip the page. */ if (page_mapcount(page) != 1) goto out; if (!trylock_page(page)) goto out; /* * If user want to discard part-pages of THP, split it so MADV_FREE * will deactivate only them. */ if (next - addr != HPAGE_PMD_SIZE) { get_page(page); spin_unlock(ptl); |
9818b8cde madvise_free, thp... |
1583 |
split_huge_page(page); |
b8d3c4c30 mm/huge_memory.c:... |
1584 |
unlock_page(page); |
bbf29ffc7 thp, mm: fix cras... |
1585 |
put_page(page); |
b8d3c4c30 mm/huge_memory.c:... |
1586 1587 1588 1589 1590 1591 |
goto out_unlocked; } if (PageDirty(page)) ClearPageDirty(page); unlock_page(page); |
b8d3c4c30 mm/huge_memory.c:... |
1592 |
if (pmd_young(orig_pmd) || pmd_dirty(orig_pmd)) { |
58ceeb6be thp: fix MADV_DON... |
1593 |
pmdp_invalidate(vma, addr, pmd); |
b8d3c4c30 mm/huge_memory.c:... |
1594 1595 1596 1597 1598 1599 |
orig_pmd = pmd_mkold(orig_pmd); orig_pmd = pmd_mkclean(orig_pmd); set_pmd_at(mm, addr, pmd, orig_pmd); tlb_remove_pmd_tlb_entry(tlb, pmd, addr); } |
802a3a92a mm: reclaim MADV_... |
1600 1601 |
mark_page_lazyfree(page); |
319904ad4 mm, THP: clean up... |
1602 |
ret = true; |
b8d3c4c30 mm/huge_memory.c:... |
1603 1604 1605 1606 1607 |
out: spin_unlock(ptl); out_unlocked: return ret; } |
953c66c2b mm: THP page cach... |
1608 1609 1610 1611 1612 1613 1614 1615 |
static inline void zap_deposited_table(struct mm_struct *mm, pmd_t *pmd) { pgtable_t pgtable; pgtable = pgtable_trans_huge_withdraw(mm, pmd); pte_free(mm, pgtable); atomic_long_dec(&mm->nr_ptes); } |
71e3aac07 thp: transparent ... |
1616 |
int zap_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma, |
f21760b15 thp: add tlb_remo... |
1617 |
pmd_t *pmd, unsigned long addr) |
71e3aac07 thp: transparent ... |
1618 |
{ |
da1467690 thp: fix zap_huge... |
1619 |
pmd_t orig_pmd; |
bf929152e mm, thp: change p... |
1620 |
spinlock_t *ptl; |
71e3aac07 thp: transparent ... |
1621 |
|
07e326610 mm: add tlb_remov... |
1622 |
tlb_remove_check_page_size_change(tlb, HPAGE_PMD_SIZE); |
b6ec57f4b thp: change pmd_t... |
1623 1624 |
ptl = __pmd_trans_huge_lock(pmd, vma); if (!ptl) |
da1467690 thp: fix zap_huge... |
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 |
return 0; /* * For architectures like ppc64 we look at deposited pgtable * when calling pmdp_huge_get_and_clear. So do the * pgtable_trans_huge_withdraw after finishing pmdp related * operations. */ orig_pmd = pmdp_huge_get_and_clear_full(tlb->mm, addr, pmd, tlb->fullmm); tlb_remove_pmd_tlb_entry(tlb, pmd, addr); if (vma_is_dax(vma)) { |
3b6521f53 mm/huge_memory.c:... |
1636 1637 |
if (arch_needs_pgtable_deposit()) zap_deposited_table(tlb->mm, pmd); |
da1467690 thp: fix zap_huge... |
1638 1639 |
spin_unlock(ptl); if (is_huge_zero_pmd(orig_pmd)) |
c0f2e176f mm: use the corre... |
1640 |
tlb_remove_page_size(tlb, pmd_page(orig_pmd), HPAGE_PMD_SIZE); |
da1467690 thp: fix zap_huge... |
1641 |
} else if (is_huge_zero_pmd(orig_pmd)) { |
c14a6eb44 mm/huge_memory.c:... |
1642 |
zap_deposited_table(tlb->mm, pmd); |
da1467690 thp: fix zap_huge... |
1643 |
spin_unlock(ptl); |
c0f2e176f mm: use the corre... |
1644 |
tlb_remove_page_size(tlb, pmd_page(orig_pmd), HPAGE_PMD_SIZE); |
da1467690 thp: fix zap_huge... |
1645 |
} else { |
616b83715 mm: thp: enable t... |
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 |
struct page *page = NULL; int flush_needed = 1; if (pmd_present(orig_pmd)) { page = pmd_page(orig_pmd); page_remove_rmap(page, true); VM_BUG_ON_PAGE(page_mapcount(page) < 0, page); VM_BUG_ON_PAGE(!PageHead(page), page); } else if (thp_migration_supported()) { swp_entry_t entry; VM_BUG_ON(!is_pmd_migration_entry(orig_pmd)); entry = pmd_to_swp_entry(orig_pmd); page = pfn_to_page(swp_offset(entry)); flush_needed = 0; } else WARN_ONCE(1, "Non present huge pmd without pmd migration enabled!"); |
b5072380e thp: support file... |
1663 |
if (PageAnon(page)) { |
c14a6eb44 mm/huge_memory.c:... |
1664 |
zap_deposited_table(tlb->mm, pmd); |
b5072380e thp: support file... |
1665 1666 |
add_mm_counter(tlb->mm, MM_ANONPAGES, -HPAGE_PMD_NR); } else { |
953c66c2b mm: THP page cach... |
1667 1668 |
if (arch_needs_pgtable_deposit()) zap_deposited_table(tlb->mm, pmd); |
b5072380e thp: support file... |
1669 1670 |
add_mm_counter(tlb->mm, MM_FILEPAGES, -HPAGE_PMD_NR); } |
616b83715 mm: thp: enable t... |
1671 |
|
da1467690 thp: fix zap_huge... |
1672 |
spin_unlock(ptl); |
616b83715 mm: thp: enable t... |
1673 1674 |
if (flush_needed) tlb_remove_page_size(tlb, page, HPAGE_PMD_SIZE); |
025c5b245 thp: optimize awa... |
1675 |
} |
da1467690 thp: fix zap_huge... |
1676 |
return 1; |
71e3aac07 thp: transparent ... |
1677 |
} |
1dd38b6c2 mm: move vma_is_a... |
1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 |
#ifndef pmd_move_must_withdraw static inline int pmd_move_must_withdraw(spinlock_t *new_pmd_ptl, spinlock_t *old_pmd_ptl, struct vm_area_struct *vma) { /* * With split pmd lock we also need to move preallocated * PTE page table if new_pmd is on different PMD page table. * * We also don't deposit and withdraw tables for file pages. */ return (new_pmd_ptl != old_pmd_ptl) && vma_is_anonymous(vma); } #endif |
ab6e3d093 mm: soft-dirty: k... |
1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 |
static pmd_t move_soft_dirty_pmd(pmd_t pmd) { #ifdef CONFIG_MEM_SOFT_DIRTY if (unlikely(is_pmd_migration_entry(pmd))) pmd = pmd_swp_mksoft_dirty(pmd); else if (pmd_present(pmd)) pmd = pmd_mksoft_dirty(pmd); #endif return pmd; } |
bf8616d5f huge mm: move_hug... |
1702 |
bool move_huge_pmd(struct vm_area_struct *vma, unsigned long old_addr, |
37a1c49a9 thp: mremap suppo... |
1703 |
unsigned long new_addr, unsigned long old_end, |
541500abf mremap: properly ... |
1704 |
pmd_t *old_pmd, pmd_t *new_pmd) |
37a1c49a9 thp: mremap suppo... |
1705 |
{ |
bf929152e mm, thp: change p... |
1706 |
spinlock_t *old_ptl, *new_ptl; |
37a1c49a9 thp: mremap suppo... |
1707 |
pmd_t pmd; |
37a1c49a9 thp: mremap suppo... |
1708 |
struct mm_struct *mm = vma->vm_mm; |
5d1904204 mremap: fix race ... |
1709 |
bool force_flush = false; |
37a1c49a9 thp: mremap suppo... |
1710 1711 1712 |
if ((old_addr & ~HPAGE_PMD_MASK) || (new_addr & ~HPAGE_PMD_MASK) || |
bf8616d5f huge mm: move_hug... |
1713 |
old_end - old_addr < HPAGE_PMD_SIZE) |
4b471e889 mm, thp: remove i... |
1714 |
return false; |
37a1c49a9 thp: mremap suppo... |
1715 1716 1717 1718 1719 1720 1721 |
/* * 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)); |
4b471e889 mm, thp: remove i... |
1722 |
return false; |
37a1c49a9 thp: mremap suppo... |
1723 |
} |
bf929152e mm, thp: change p... |
1724 1725 1726 1727 |
/* * We don't have to worry about the ordering of src and dst * ptlocks because exclusive mmap_sem prevents deadlock. */ |
b6ec57f4b thp: change pmd_t... |
1728 1729 |
old_ptl = __pmd_trans_huge_lock(old_pmd, vma); if (old_ptl) { |
bf929152e mm, thp: change p... |
1730 1731 1732 |
new_ptl = pmd_lockptr(mm, new_pmd); if (new_ptl != old_ptl) spin_lock_nested(new_ptl, SINGLE_DEPTH_NESTING); |
8809aa2d2 mm: clarify that ... |
1733 |
pmd = pmdp_huge_get_and_clear(mm, old_addr, old_pmd); |
541500abf mremap: properly ... |
1734 |
if (pmd_present(pmd)) |
a2ce2666a mremap: move_ptes... |
1735 |
force_flush = true; |
025c5b245 thp: optimize awa... |
1736 |
VM_BUG_ON(!pmd_none(*new_pmd)); |
3592806cf thp: move preallo... |
1737 |
|
1dd38b6c2 mm: move vma_is_a... |
1738 |
if (pmd_move_must_withdraw(new_ptl, old_ptl, vma)) { |
b3084f4db powerpc/thp: Fix ... |
1739 |
pgtable_t pgtable; |
3592806cf thp: move preallo... |
1740 1741 |
pgtable = pgtable_trans_huge_withdraw(mm, old_pmd); pgtable_trans_huge_deposit(mm, new_pmd, pgtable); |
3592806cf thp: move preallo... |
1742 |
} |
ab6e3d093 mm: soft-dirty: k... |
1743 1744 |
pmd = move_soft_dirty_pmd(pmd); set_pmd_at(mm, new_addr, new_pmd, pmd); |
5d1904204 mremap: fix race ... |
1745 1746 |
if (force_flush) flush_tlb_range(vma, old_addr, old_addr + PMD_SIZE); |
541500abf mremap: properly ... |
1747 1748 |
if (new_ptl != old_ptl) spin_unlock(new_ptl); |
bf929152e mm, thp: change p... |
1749 |
spin_unlock(old_ptl); |
4b471e889 mm, thp: remove i... |
1750 |
return true; |
37a1c49a9 thp: mremap suppo... |
1751 |
} |
4b471e889 mm, thp: remove i... |
1752 |
return false; |
37a1c49a9 thp: mremap suppo... |
1753 |
} |
f123d74ab mm: Only flush TL... |
1754 1755 1756 1757 1758 1759 |
/* * 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... |
1760 |
int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, |
e944fd67b mm: numa: do not ... |
1761 |
unsigned long addr, pgprot_t newprot, int prot_numa) |
cd7548ab3 thp: mprotect: tr... |
1762 1763 |
{ struct mm_struct *mm = vma->vm_mm; |
bf929152e mm, thp: change p... |
1764 |
spinlock_t *ptl; |
0a85e51d3 thp: reduce inden... |
1765 1766 1767 |
pmd_t entry; bool preserve_write; int ret; |
cd7548ab3 thp: mprotect: tr... |
1768 |
|
b6ec57f4b thp: change pmd_t... |
1769 |
ptl = __pmd_trans_huge_lock(pmd, vma); |
0a85e51d3 thp: reduce inden... |
1770 1771 |
if (!ptl) return 0; |
e944fd67b mm: numa: do not ... |
1772 |
|
0a85e51d3 thp: reduce inden... |
1773 1774 |
preserve_write = prot_numa && pmd_write(*pmd); ret = 1; |
e944fd67b mm: numa: do not ... |
1775 |
|
84c3fc4e9 mm: thp: check pm... |
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 |
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION if (is_swap_pmd(*pmd)) { swp_entry_t entry = pmd_to_swp_entry(*pmd); VM_BUG_ON(!is_pmd_migration_entry(*pmd)); if (is_write_migration_entry(entry)) { pmd_t newpmd; /* * A protection check is difficult so * just be safe and disable write */ make_migration_entry_read(&entry); newpmd = swp_entry_to_pmd(entry); |
ab6e3d093 mm: soft-dirty: k... |
1789 1790 |
if (pmd_swp_soft_dirty(*pmd)) newpmd = pmd_swp_mksoft_dirty(newpmd); |
84c3fc4e9 mm: thp: check pm... |
1791 1792 1793 1794 1795 |
set_pmd_at(mm, addr, pmd, newpmd); } goto unlock; } #endif |
0a85e51d3 thp: reduce inden... |
1796 1797 1798 1799 1800 1801 1802 |
/* * Avoid trapping faults against the zero page. The read-only * data is likely to be read-cached on the local CPU and * local/remote hits to the zero page are not interesting. */ if (prot_numa && is_huge_zero_pmd(*pmd)) goto unlock; |
025c5b245 thp: optimize awa... |
1803 |
|
0a85e51d3 thp: reduce inden... |
1804 1805 |
if (prot_numa && pmd_protnone(*pmd)) goto unlock; |
ced108037 thp: fix MADV_DON... |
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 |
/* * In case prot_numa, we are under down_read(mmap_sem). It's critical * to not clear pmd intermittently to avoid race with MADV_DONTNEED * which is also under down_read(mmap_sem): * * CPU0: CPU1: * change_huge_pmd(prot_numa=1) * pmdp_huge_get_and_clear_notify() * madvise_dontneed() * zap_pmd_range() * pmd_trans_huge(*pmd) == 0 (without ptl) * // skip the pmd * set_pmd_at(); * // pmd is re-established * * The race makes MADV_DONTNEED miss the huge pmd and don't clear it * which may break userspace. * * pmdp_invalidate() is required to make sure we don't miss * dirty/young flags set by hardware. */ entry = *pmd; pmdp_invalidate(vma, addr, pmd); /* * Recover dirty/young flags. It relies on pmdp_invalidate to not * corrupt them. */ if (pmd_dirty(*pmd)) entry = pmd_mkdirty(entry); if (pmd_young(*pmd)) entry = pmd_mkyoung(entry); |
0a85e51d3 thp: reduce inden... |
1838 1839 1840 1841 1842 1843 1844 1845 |
entry = pmd_modify(entry, newprot); if (preserve_write) entry = pmd_mk_savedwrite(entry); ret = HPAGE_PMD_NR; set_pmd_at(mm, addr, pmd, entry); BUG_ON(vma_is_anonymous(vma) && !preserve_write && pmd_write(entry)); unlock: spin_unlock(ptl); |
025c5b245 thp: optimize awa... |
1846 1847 1848 1849 |
return ret; } /* |
8f19b0c05 thp: fix comments... |
1850 |
* Returns page table lock pointer if a given pmd maps a thp, NULL otherwise. |
025c5b245 thp: optimize awa... |
1851 |
* |
8f19b0c05 thp: fix comments... |
1852 1853 |
* Note that if it returns page table lock pointer, this routine returns without * unlocking page table lock. So callers must unlock it. |
025c5b245 thp: optimize awa... |
1854 |
*/ |
b6ec57f4b thp: change pmd_t... |
1855 |
spinlock_t *__pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma) |
025c5b245 thp: optimize awa... |
1856 |
{ |
b6ec57f4b thp: change pmd_t... |
1857 1858 |
spinlock_t *ptl; ptl = pmd_lock(vma->vm_mm, pmd); |
84c3fc4e9 mm: thp: check pm... |
1859 1860 |
if (likely(is_swap_pmd(*pmd) || pmd_trans_huge(*pmd) || pmd_devmap(*pmd))) |
b6ec57f4b thp: change pmd_t... |
1861 1862 1863 |
return ptl; spin_unlock(ptl); return NULL; |
cd7548ab3 thp: mprotect: tr... |
1864 |
} |
a00cc7d9d mm, x86: add supp... |
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 |
/* * Returns true if a given pud maps a thp, false otherwise. * * Note that if it returns true, this routine returns without unlocking page * table lock. So callers must unlock it. */ spinlock_t *__pud_trans_huge_lock(pud_t *pud, struct vm_area_struct *vma) { spinlock_t *ptl; ptl = pud_lock(vma->vm_mm, pud); if (likely(pud_trans_huge(*pud) || pud_devmap(*pud))) return ptl; spin_unlock(ptl); return NULL; } #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD int zap_huge_pud(struct mmu_gather *tlb, struct vm_area_struct *vma, pud_t *pud, unsigned long addr) { pud_t orig_pud; spinlock_t *ptl; ptl = __pud_trans_huge_lock(pud, vma); if (!ptl) return 0; /* * For architectures like ppc64 we look at deposited pgtable * when calling pudp_huge_get_and_clear. So do the * pgtable_trans_huge_withdraw after finishing pudp related * operations. */ orig_pud = pudp_huge_get_and_clear_full(tlb->mm, addr, pud, tlb->fullmm); tlb_remove_pud_tlb_entry(tlb, pud, addr); if (vma_is_dax(vma)) { spin_unlock(ptl); /* No zero page support yet */ } else { /* No support for anonymous PUD pages yet */ BUG(); } return 1; } static void __split_huge_pud_locked(struct vm_area_struct *vma, pud_t *pud, unsigned long haddr) { VM_BUG_ON(haddr & ~HPAGE_PUD_MASK); VM_BUG_ON_VMA(vma->vm_start > haddr, vma); VM_BUG_ON_VMA(vma->vm_end < haddr + HPAGE_PUD_SIZE, vma); VM_BUG_ON(!pud_trans_huge(*pud) && !pud_devmap(*pud)); |
ce9311cf9 mm/vmstats: add t... |
1918 |
count_vm_event(THP_SPLIT_PUD); |
a00cc7d9d mm, x86: add supp... |
1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 |
pudp_huge_clear_flush_notify(vma, haddr, pud); } void __split_huge_pud(struct vm_area_struct *vma, pud_t *pud, unsigned long address) { spinlock_t *ptl; struct mm_struct *mm = vma->vm_mm; unsigned long haddr = address & HPAGE_PUD_MASK; mmu_notifier_invalidate_range_start(mm, haddr, haddr + HPAGE_PUD_SIZE); ptl = pud_lock(mm, pud); if (unlikely(!pud_trans_huge(*pud) && !pud_devmap(*pud))) goto out; __split_huge_pud_locked(vma, pud, haddr); out: spin_unlock(ptl); mmu_notifier_invalidate_range_end(mm, haddr, haddr + HPAGE_PUD_SIZE); } #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */ |
eef1b3ba0 thp: implement sp... |
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 |
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; /* leave pmd empty until pte is filled */ pmdp_huge_clear_flush_notify(vma, haddr, pmd); pgtable = pgtable_trans_huge_withdraw(mm, pmd); 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); |
eef1b3ba0 thp: implement sp... |
1966 1967 1968 |
} static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd, |
ba9882808 thp: add option t... |
1969 |
unsigned long haddr, bool freeze) |
eef1b3ba0 thp: implement sp... |
1970 1971 1972 1973 1974 |
{ struct mm_struct *mm = vma->vm_mm; struct page *page; pgtable_t pgtable; pmd_t _pmd; |
84c3fc4e9 mm: thp: check pm... |
1975 |
bool young, write, dirty, soft_dirty, pmd_migration = false; |
2ac015e29 thp: call pmdp_in... |
1976 |
unsigned long addr; |
eef1b3ba0 thp: implement sp... |
1977 1978 1979 1980 1981 |
int i; VM_BUG_ON(haddr & ~HPAGE_PMD_MASK); VM_BUG_ON_VMA(vma->vm_start > haddr, vma); VM_BUG_ON_VMA(vma->vm_end < haddr + HPAGE_PMD_SIZE, vma); |
84c3fc4e9 mm: thp: check pm... |
1982 1983 |
VM_BUG_ON(!is_pmd_migration_entry(*pmd) && !pmd_trans_huge(*pmd) && !pmd_devmap(*pmd)); |
eef1b3ba0 thp: implement sp... |
1984 1985 |
count_vm_event(THP_SPLIT_PMD); |
d21b9e57c thp: handle file ... |
1986 1987 |
if (!vma_is_anonymous(vma)) { _pmd = pmdp_huge_clear_flush_notify(vma, haddr, pmd); |
953c66c2b mm: THP page cach... |
1988 1989 1990 1991 1992 1993 |
/* * We are going to unmap this huge page. So * just go ahead and zap it */ if (arch_needs_pgtable_deposit()) zap_deposited_table(mm, pmd); |
d21b9e57c thp: handle file ... |
1994 1995 1996 |
if (vma_is_dax(vma)) return; page = pmd_page(_pmd); |
70ef1db1f mm/huge_memory.c:... |
1997 1998 |
if (!PageDirty(page) && pmd_dirty(_pmd)) set_page_dirty(page); |
d21b9e57c thp: handle file ... |
1999 2000 2001 2002 2003 |
if (!PageReferenced(page) && pmd_young(_pmd)) SetPageReferenced(page); page_remove_rmap(page, true); put_page(page); add_mm_counter(mm, MM_FILEPAGES, -HPAGE_PMD_NR); |
eef1b3ba0 thp: implement sp... |
2004 2005 2006 2007 |
return; } else if (is_huge_zero_pmd(*pmd)) { return __split_huge_zero_page_pmd(vma, haddr, pmd); } |
84c3fc4e9 mm: thp: check pm... |
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 |
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION pmd_migration = is_pmd_migration_entry(*pmd); if (pmd_migration) { swp_entry_t entry; entry = pmd_to_swp_entry(*pmd); page = pfn_to_page(swp_offset(entry)); } else #endif page = pmd_page(*pmd); |
eef1b3ba0 thp: implement sp... |
2018 |
VM_BUG_ON_PAGE(!page_count(page), page); |
fe896d187 mm: introduce pag... |
2019 |
page_ref_add(page, HPAGE_PMD_NR - 1); |
eef1b3ba0 thp: implement sp... |
2020 2021 |
write = pmd_write(*pmd); young = pmd_young(*pmd); |
b8d3c4c30 mm/huge_memory.c:... |
2022 |
dirty = pmd_dirty(*pmd); |
804dd1504 soft_dirty: fix s... |
2023 |
soft_dirty = pmd_soft_dirty(*pmd); |
eef1b3ba0 thp: implement sp... |
2024 |
|
c777e2a8b powerpc/mm: Fix M... |
2025 |
pmdp_huge_split_prepare(vma, haddr, pmd); |
eef1b3ba0 thp: implement sp... |
2026 2027 |
pgtable = pgtable_trans_huge_withdraw(mm, pmd); pmd_populate(mm, &_pmd, pgtable); |
2ac015e29 thp: call pmdp_in... |
2028 |
for (i = 0, addr = haddr; i < HPAGE_PMD_NR; i++, addr += PAGE_SIZE) { |
eef1b3ba0 thp: implement sp... |
2029 2030 2031 2032 2033 2034 |
pte_t entry, *pte; /* * Note that NUMA hinting access restrictions are not * transferred to avoid any possibility of altering * permissions across VMAs. */ |
84c3fc4e9 mm: thp: check pm... |
2035 |
if (freeze || pmd_migration) { |
ba9882808 thp: add option t... |
2036 2037 2038 |
swp_entry_t swp_entry; swp_entry = make_migration_entry(page + i, write); entry = swp_entry_to_pte(swp_entry); |
804dd1504 soft_dirty: fix s... |
2039 2040 |
if (soft_dirty) entry = pte_swp_mksoft_dirty(entry); |
ba9882808 thp: add option t... |
2041 |
} else { |
6d2329f88 mm: vm_page_prot:... |
2042 |
entry = mk_pte(page + i, READ_ONCE(vma->vm_page_prot)); |
b8d3c4c30 mm/huge_memory.c:... |
2043 |
entry = maybe_mkwrite(entry, vma); |
ba9882808 thp: add option t... |
2044 2045 2046 2047 |
if (!write) entry = pte_wrprotect(entry); if (!young) entry = pte_mkold(entry); |
804dd1504 soft_dirty: fix s... |
2048 2049 |
if (soft_dirty) entry = pte_mksoft_dirty(entry); |
ba9882808 thp: add option t... |
2050 |
} |
b8d3c4c30 mm/huge_memory.c:... |
2051 2052 |
if (dirty) SetPageDirty(page + i); |
2ac015e29 thp: call pmdp_in... |
2053 |
pte = pte_offset_map(&_pmd, addr); |
eef1b3ba0 thp: implement sp... |
2054 |
BUG_ON(!pte_none(*pte)); |
2ac015e29 thp: call pmdp_in... |
2055 |
set_pte_at(mm, addr, pte, entry); |
eef1b3ba0 thp: implement sp... |
2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 |
atomic_inc(&page[i]._mapcount); pte_unmap(pte); } /* * Set PG_double_map before dropping compound_mapcount to avoid * false-negative page_mapped(). */ if (compound_mapcount(page) > 1 && !TestSetPageDoubleMap(page)) { for (i = 0; i < HPAGE_PMD_NR; i++) atomic_inc(&page[i]._mapcount); } if (atomic_add_negative(-1, compound_mapcount_ptr(page))) { /* Last compound_mapcount is gone. */ |
11fb99898 mm: move most fil... |
2071 |
__dec_node_page_state(page, NR_ANON_THPS); |
eef1b3ba0 thp: implement sp... |
2072 2073 2074 2075 2076 2077 2078 2079 |
if (TestClearPageDoubleMap(page)) { /* No need in mapcount reference anymore */ for (i = 0; i < HPAGE_PMD_NR; i++) atomic_dec(&page[i]._mapcount); } } smp_wmb(); /* make pte visible before pmd */ |
e9b61f198 thp: reintroduce ... |
2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 |
/* * 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_pmd_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. */ pmdp_invalidate(vma, haddr, pmd); |
eef1b3ba0 thp: implement sp... |
2102 |
pmd_populate(mm, pmd, pgtable); |
e9b61f198 thp: reintroduce ... |
2103 2104 |
if (freeze) { |
2ac015e29 thp: call pmdp_in... |
2105 |
for (i = 0; i < HPAGE_PMD_NR; i++) { |
e9b61f198 thp: reintroduce ... |
2106 2107 2108 2109 |
page_remove_rmap(page + i, false); put_page(page + i); } } |
eef1b3ba0 thp: implement sp... |
2110 2111 2112 |
} void __split_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, |
33f4751e9 mm: thp: move pmd... |
2113 |
unsigned long address, bool freeze, struct page *page) |
eef1b3ba0 thp: implement sp... |
2114 2115 2116 2117 2118 2119 2120 |
{ spinlock_t *ptl; struct mm_struct *mm = vma->vm_mm; unsigned long haddr = address & HPAGE_PMD_MASK; mmu_notifier_invalidate_range_start(mm, haddr, haddr + HPAGE_PMD_SIZE); ptl = pmd_lock(mm, pmd); |
33f4751e9 mm: thp: move pmd... |
2121 2122 2123 2124 2125 2126 2127 2128 |
/* * If caller asks to setup a migration entries, we need a page to check * pmd against. Otherwise we can end up replacing wrong page. */ VM_BUG_ON(freeze && !page); if (page && page != pmd_page(*pmd)) goto out; |
5c7fb56e5 mm, dax: dax-pmd ... |
2129 |
if (pmd_trans_huge(*pmd)) { |
33f4751e9 mm: thp: move pmd... |
2130 |
page = pmd_page(*pmd); |
5c7fb56e5 mm, dax: dax-pmd ... |
2131 |
if (PageMlocked(page)) |
5f7377147 thp: fix deadlock... |
2132 |
clear_page_mlock(page); |
84c3fc4e9 mm: thp: check pm... |
2133 |
} else if (!(pmd_devmap(*pmd) || is_pmd_migration_entry(*pmd))) |
e90309c9f thp: allow mlocke... |
2134 |
goto out; |
fec89c109 thp: rewrite free... |
2135 |
__split_huge_pmd_locked(vma, pmd, haddr, freeze); |
e90309c9f thp: allow mlocke... |
2136 |
out: |
eef1b3ba0 thp: implement sp... |
2137 2138 2139 |
spin_unlock(ptl); mmu_notifier_invalidate_range_end(mm, haddr, haddr + HPAGE_PMD_SIZE); } |
fec89c109 thp: rewrite free... |
2140 2141 |
void split_huge_pmd_address(struct vm_area_struct *vma, unsigned long address, bool freeze, struct page *page) |
94fcc585f thp: avoid breaki... |
2142 |
{ |
f72e7dcdd mm: let mm_find_p... |
2143 |
pgd_t *pgd; |
c2febafc6 mm: convert gener... |
2144 |
p4d_t *p4d; |
f72e7dcdd mm: let mm_find_p... |
2145 |
pud_t *pud; |
94fcc585f thp: avoid breaki... |
2146 |
pmd_t *pmd; |
78ddc5347 thp: rename split... |
2147 |
pgd = pgd_offset(vma->vm_mm, address); |
f72e7dcdd mm: let mm_find_p... |
2148 2149 |
if (!pgd_present(*pgd)) return; |
c2febafc6 mm: convert gener... |
2150 2151 2152 2153 2154 |
p4d = p4d_offset(pgd, address); if (!p4d_present(*p4d)) return; pud = pud_offset(p4d, address); |
f72e7dcdd mm: let mm_find_p... |
2155 2156 2157 2158 |
if (!pud_present(*pud)) return; pmd = pmd_offset(pud, address); |
fec89c109 thp: rewrite free... |
2159 |
|
33f4751e9 mm: thp: move pmd... |
2160 |
__split_huge_pmd(vma, pmd, address, freeze, page); |
94fcc585f thp: avoid breaki... |
2161 |
} |
e1b9996b8 thp: vma_adjust_t... |
2162 |
void vma_adjust_trans_huge(struct vm_area_struct *vma, |
94fcc585f thp: avoid breaki... |
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 |
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) |
fec89c109 thp: rewrite free... |
2175 |
split_huge_pmd_address(vma, start, false, NULL); |
94fcc585f thp: avoid breaki... |
2176 2177 2178 2179 2180 2181 2182 2183 2184 |
/* * 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) |
fec89c109 thp: rewrite free... |
2185 |
split_huge_pmd_address(vma, end, false, NULL); |
94fcc585f thp: avoid breaki... |
2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 |
/* * 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) |
fec89c109 thp: rewrite free... |
2199 |
split_huge_pmd_address(next, nstart, false, NULL); |
94fcc585f thp: avoid breaki... |
2200 2201 |
} } |
e9b61f198 thp: reintroduce ... |
2202 |
|
e12b67d81 mm/huge_memory: r... |
2203 |
static void unmap_page(struct page *page) |
e9b61f198 thp: reintroduce ... |
2204 |
{ |
baa355fd3 thp: file pages s... |
2205 |
enum ttu_flags ttu_flags = TTU_IGNORE_MLOCK | TTU_IGNORE_ACCESS | |
c7ab0d2fd mm: convert try_t... |
2206 |
TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD; |
666e5a406 mm: make ttu's re... |
2207 |
bool unmap_success; |
e9b61f198 thp: reintroduce ... |
2208 2209 |
VM_BUG_ON_PAGE(!PageHead(page), page); |
baa355fd3 thp: file pages s... |
2210 |
if (PageAnon(page)) |
b5ff8161e mm: thp: introduc... |
2211 |
ttu_flags |= TTU_SPLIT_FREEZE; |
baa355fd3 thp: file pages s... |
2212 |
|
666e5a406 mm: make ttu's re... |
2213 2214 |
unmap_success = try_to_unmap(page, ttu_flags); VM_BUG_ON_PAGE(!unmap_success, page); |
e9b61f198 thp: reintroduce ... |
2215 |
} |
e12b67d81 mm/huge_memory: r... |
2216 |
static void remap_page(struct page *page) |
e9b61f198 thp: reintroduce ... |
2217 |
{ |
fec89c109 thp: rewrite free... |
2218 |
int i; |
ace71a19c mm: introduce pag... |
2219 2220 2221 2222 2223 2224 |
if (PageTransHuge(page)) { remove_migration_ptes(page, page, true); } else { for (i = 0; i < HPAGE_PMD_NR; i++) remove_migration_ptes(page + i, page + i, true); } |
e9b61f198 thp: reintroduce ... |
2225 |
} |
8df651c70 thp: cleanup spli... |
2226 |
static void __split_huge_page_tail(struct page *head, int tail, |
e9b61f198 thp: reintroduce ... |
2227 2228 |
struct lruvec *lruvec, struct list_head *list) { |
e9b61f198 thp: reintroduce ... |
2229 |
struct page *page_tail = head + tail; |
8df651c70 thp: cleanup spli... |
2230 |
VM_BUG_ON_PAGE(atomic_read(&page_tail->_mapcount) != -1, page_tail); |
e9b61f198 thp: reintroduce ... |
2231 2232 |
/* |
30241d721 mm/huge_memory.c:... |
2233 2234 2235 2236 |
* Clone page flags before unfreezing refcount. * * After successful get_page_unless_zero() might follow flags change, * for exmaple lock_page() which set PG_waiters. |
e9b61f198 thp: reintroduce ... |
2237 |
*/ |
e9b61f198 thp: reintroduce ... |
2238 2239 2240 2241 |
page_tail->flags &= ~PAGE_FLAGS_CHECK_AT_PREP; page_tail->flags |= (head->flags & ((1L << PG_referenced) | (1L << PG_swapbacked) | |
38d8b4e6b mm, THP, swap: de... |
2242 |
(1L << PG_swapcache) | |
e9b61f198 thp: reintroduce ... |
2243 2244 2245 2246 |
(1L << PG_mlocked) | (1L << PG_uptodate) | (1L << PG_active) | (1L << PG_locked) | |
b8d3c4c30 mm/huge_memory.c:... |
2247 2248 |
(1L << PG_unevictable) | (1L << PG_dirty))); |
e9b61f198 thp: reintroduce ... |
2249 |
|
16d07443b mm/huge_memory: s... |
2250 2251 2252 2253 2254 |
/* ->mapping in first tail page is compound_mapcount */ VM_BUG_ON_PAGE(tail > 2 && page_tail->mapping != TAIL_MAPPING, page_tail); page_tail->mapping = head->mapping; page_tail->index = head->index + tail; |
30241d721 mm/huge_memory.c:... |
2255 |
/* Page flags must be visible before we make the page non-compound. */ |
e9b61f198 thp: reintroduce ... |
2256 |
smp_wmb(); |
30241d721 mm/huge_memory.c:... |
2257 2258 2259 2260 2261 2262 |
/* * Clear PageTail before unfreezing page refcount. * * After successful get_page_unless_zero() might follow put_page() * which needs correct compound_head(). */ |
e9b61f198 thp: reintroduce ... |
2263 |
clear_compound_head(page_tail); |
30241d721 mm/huge_memory.c:... |
2264 2265 2266 |
/* Finally unfreeze refcount. Additional reference from page cache. */ page_ref_unfreeze(page_tail, 1 + (!PageAnon(head) || PageSwapCache(head))); |
e9b61f198 thp: reintroduce ... |
2267 2268 2269 2270 |
if (page_is_young(head)) set_page_young(page_tail); if (page_is_idle(head)) set_page_idle(page_tail); |
e9b61f198 thp: reintroduce ... |
2271 2272 |
page_cpupid_xchg_last(page_tail, page_cpupid_last(head)); lru_add_page_tail(head, page_tail, lruvec, list); |
e9b61f198 thp: reintroduce ... |
2273 |
} |
baa355fd3 thp: file pages s... |
2274 |
static void __split_huge_page(struct page *page, struct list_head *list, |
6f75a0983 mm/huge_memory: f... |
2275 |
pgoff_t end, unsigned long flags) |
e9b61f198 thp: reintroduce ... |
2276 2277 2278 2279 |
{ struct page *head = compound_head(page); struct zone *zone = page_zone(head); struct lruvec *lruvec; |
8df651c70 thp: cleanup spli... |
2280 |
int i; |
e9b61f198 thp: reintroduce ... |
2281 |
|
599d0c954 mm, vmscan: move ... |
2282 |
lruvec = mem_cgroup_page_lruvec(head, zone->zone_pgdat); |
e9b61f198 thp: reintroduce ... |
2283 2284 2285 |
/* complete memcg works before add pages to LRU */ mem_cgroup_split_huge_fixup(head); |
baa355fd3 thp: file pages s... |
2286 |
for (i = HPAGE_PMD_NR - 1; i >= 1; i--) { |
8df651c70 thp: cleanup spli... |
2287 |
__split_huge_page_tail(head, i, lruvec, list); |
baa355fd3 thp: file pages s... |
2288 2289 |
/* Some pages can be beyond i_size: drop them from page cache */ if (head[i].index >= end) { |
a7027b7d6 mm/huge_memory.c:... |
2290 |
ClearPageDirty(head + i); |
baa355fd3 thp: file pages s... |
2291 |
__delete_from_page_cache(head + i, NULL); |
800d8c63b shmem: add huge p... |
2292 2293 |
if (IS_ENABLED(CONFIG_SHMEM) && PageSwapBacked(head)) shmem_uncharge(head->mapping->host, 1); |
baa355fd3 thp: file pages s... |
2294 2295 2296 |
put_page(head + i); } } |
e9b61f198 thp: reintroduce ... |
2297 2298 |
ClearPageCompound(head); |
baa355fd3 thp: file pages s... |
2299 2300 |
/* See comment in __split_huge_page_tail() */ if (PageAnon(head)) { |
38d8b4e6b mm, THP, swap: de... |
2301 2302 2303 2304 2305 |
/* Additional pin to radix tree of swap cache */ if (PageSwapCache(head)) page_ref_add(head, 2); else page_ref_inc(head); |
baa355fd3 thp: file pages s... |
2306 2307 2308 2309 2310 |
} else { /* Additional pin to radix tree */ page_ref_add(head, 2); spin_unlock(&head->mapping->tree_lock); } |
a52633d8e mm, vmscan: move ... |
2311 |
spin_unlock_irqrestore(zone_lru_lock(page_zone(head)), flags); |
e9b61f198 thp: reintroduce ... |
2312 |
|
e12b67d81 mm/huge_memory: r... |
2313 |
remap_page(head); |
e9b61f198 thp: reintroduce ... |
2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 |
for (i = 0; i < HPAGE_PMD_NR; i++) { struct page *subpage = head + i; if (subpage == page) continue; unlock_page(subpage); /* * Subpages 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(subpage); } } |
b20ce5e03 mm: prepare page_... |
2331 2332 |
int total_mapcount(struct page *page) { |
dd78fedde rmap: support fil... |
2333 |
int i, compound, ret; |
b20ce5e03 mm: prepare page_... |
2334 2335 2336 2337 2338 |
VM_BUG_ON_PAGE(PageTail(page), page); if (likely(!PageCompound(page))) return atomic_read(&page->_mapcount) + 1; |
dd78fedde rmap: support fil... |
2339 |
compound = compound_mapcount(page); |
b20ce5e03 mm: prepare page_... |
2340 |
if (PageHuge(page)) |
dd78fedde rmap: support fil... |
2341 2342 |
return compound; ret = compound; |
b20ce5e03 mm: prepare page_... |
2343 2344 |
for (i = 0; i < HPAGE_PMD_NR; i++) ret += atomic_read(&page[i]._mapcount) + 1; |
dd78fedde rmap: support fil... |
2345 2346 2347 |
/* File pages has compound_mapcount included in _mapcount */ if (!PageAnon(page)) return ret - compound * HPAGE_PMD_NR; |
b20ce5e03 mm: prepare page_... |
2348 2349 2350 2351 |
if (PageDoubleMap(page)) ret -= HPAGE_PMD_NR; return ret; } |
e9b61f198 thp: reintroduce ... |
2352 |
/* |
6d0a07edd mm: thp: calculat... |
2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 |
* This calculates accurately how many mappings a transparent hugepage * has (unlike page_mapcount() which isn't fully accurate). This full * accuracy is primarily needed to know if copy-on-write faults can * reuse the page and change the mapping to read-write instead of * copying them. At the same time this returns the total_mapcount too. * * The function returns the highest mapcount any one of the subpages * has. If the return value is one, even if different processes are * mapping different subpages of the transparent hugepage, they can * all reuse it, because each process is reusing a different subpage. * * The total_mapcount is instead counting all virtual mappings of the * subpages. If the total_mapcount is equal to "one", it tells the * caller all mappings belong to the same "mm" and in turn the * anon_vma of the transparent hugepage can become the vma->anon_vma * local one as no other process may be mapping any of the subpages. * * It would be more accurate to replace page_mapcount() with * page_trans_huge_mapcount(), however we only use * page_trans_huge_mapcount() in the copy-on-write faults where we * need full accuracy to avoid breaking page pinning, because * page_trans_huge_mapcount() is slower than page_mapcount(). */ int page_trans_huge_mapcount(struct page *page, int *total_mapcount) { int i, ret, _total_mapcount, mapcount; /* hugetlbfs shouldn't call it */ VM_BUG_ON_PAGE(PageHuge(page), page); if (likely(!PageTransCompound(page))) { mapcount = atomic_read(&page->_mapcount) + 1; if (total_mapcount) *total_mapcount = mapcount; return mapcount; } page = compound_head(page); _total_mapcount = ret = 0; for (i = 0; i < HPAGE_PMD_NR; i++) { mapcount = atomic_read(&page[i]._mapcount) + 1; ret = max(ret, mapcount); _total_mapcount += mapcount; } if (PageDoubleMap(page)) { ret -= 1; _total_mapcount -= HPAGE_PMD_NR; } mapcount = compound_mapcount(page); ret += mapcount; _total_mapcount += mapcount; if (total_mapcount) *total_mapcount = _total_mapcount; return ret; } |
b8f593cd0 mm, THP, swap: ch... |
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 |
/* Racy check whether the huge page can be split */ bool can_split_huge_page(struct page *page, int *pextra_pins) { int extra_pins; /* Additional pins from radix tree */ if (PageAnon(page)) extra_pins = PageSwapCache(page) ? HPAGE_PMD_NR : 0; else extra_pins = HPAGE_PMD_NR; if (pextra_pins) *pextra_pins = extra_pins; return total_mapcount(page) == page_count(page) - extra_pins - 1; } |
6d0a07edd mm: thp: calculat... |
2423 |
/* |
e9b61f198 thp: reintroduce ... |
2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 |
* This function splits huge page into normal pages. @page can point to any * subpage of huge page to split. Split doesn't change the position of @page. * * Only caller must hold pin on the @page, otherwise split fails with -EBUSY. * The huge page must be locked. * * 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. * * GUP pin and PG_locked transferred to @page. Rest subpages can be freed if * they are not mapped. * * Returns 0 if the hugepage is split successfully. * Returns -EBUSY if the page is pinned or if anon_vma disappeared from under * us. */ int split_huge_page_to_list(struct page *page, struct list_head *list) { struct page *head = compound_head(page); |
a3d0a9185 thp: make split_q... |
2445 |
struct pglist_data *pgdata = NODE_DATA(page_to_nid(head)); |
baa355fd3 thp: file pages s... |
2446 2447 2448 |
struct anon_vma *anon_vma = NULL; struct address_space *mapping = NULL; int count, mapcount, extra_pins, ret; |
d96543223 thp: increase spl... |
2449 |
bool mlocked; |
0b9b6fff7 thp: fix interrup... |
2450 |
unsigned long flags; |
6f75a0983 mm/huge_memory: f... |
2451 |
pgoff_t end; |
e9b61f198 thp: reintroduce ... |
2452 2453 |
VM_BUG_ON_PAGE(is_huge_zero_page(page), page); |
e9b61f198 thp: reintroduce ... |
2454 |
VM_BUG_ON_PAGE(!PageLocked(page), page); |
e9b61f198 thp: reintroduce ... |
2455 |
VM_BUG_ON_PAGE(!PageCompound(page), page); |
59807685a mm, THP, swap: su... |
2456 2457 |
if (PageWriteback(page)) return -EBUSY; |
baa355fd3 thp: file pages s... |
2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 |
if (PageAnon(head)) { /* * 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(head); if (!anon_vma) { ret = -EBUSY; goto out; } |
6f75a0983 mm/huge_memory: f... |
2472 |
end = -1; |
baa355fd3 thp: file pages s... |
2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 |
mapping = NULL; anon_vma_lock_write(anon_vma); } else { mapping = head->mapping; /* Truncated ? */ if (!mapping) { ret = -EBUSY; goto out; } |
baa355fd3 thp: file pages s... |
2483 2484 |
anon_vma = NULL; i_mmap_lock_read(mapping); |
6f75a0983 mm/huge_memory: f... |
2485 2486 2487 2488 2489 2490 2491 2492 2493 |
/* *__split_huge_page() may need to trim off pages beyond EOF: * but on 32-bit, i_size_read() takes an irq-unsafe seqlock, * which cannot be nested inside the page tree lock. So note * end now: i_size itself may be changed at any moment, but * head page lock is good enough to serialize the trimming. */ end = DIV_ROUND_UP(i_size_read(mapping->host), PAGE_SIZE); |
e9b61f198 thp: reintroduce ... |
2494 |
} |
e9b61f198 thp: reintroduce ... |
2495 2496 |
/* |
e12b67d81 mm/huge_memory: r... |
2497 |
* Racy check if we can split the page, before unmap_page() will |
e9b61f198 thp: reintroduce ... |
2498 2499 |
* split PMDs */ |
b8f593cd0 mm, THP, swap: ch... |
2500 |
if (!can_split_huge_page(head, &extra_pins)) { |
e9b61f198 thp: reintroduce ... |
2501 2502 2503 |
ret = -EBUSY; goto out_unlock; } |
d96543223 thp: increase spl... |
2504 |
mlocked = PageMlocked(page); |
e12b67d81 mm/huge_memory: r... |
2505 |
unmap_page(head); |
e9b61f198 thp: reintroduce ... |
2506 |
VM_BUG_ON_PAGE(compound_mapcount(head), head); |
d96543223 thp: increase spl... |
2507 2508 2509 |
/* Make sure the page is not on per-CPU pagevec as it takes pin */ if (mlocked) lru_add_drain(); |
baa355fd3 thp: file pages s... |
2510 |
/* prevent PageLRU to go away from under us, and freeze lru stats */ |
a52633d8e mm, vmscan: move ... |
2511 |
spin_lock_irqsave(zone_lru_lock(page_zone(head)), flags); |
baa355fd3 thp: file pages s... |
2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 |
if (mapping) { void **pslot; spin_lock(&mapping->tree_lock); pslot = radix_tree_lookup_slot(&mapping->page_tree, page_index(head)); /* * Check if the head page is present in radix tree. * We assume all tail are present too, if head is there. */ if (radix_tree_deref_slot_protected(pslot, &mapping->tree_lock) != head) goto fail; } |
0139aa7b7 mm: rename _count... |
2527 |
/* Prevent deferred_split_scan() touching ->_refcount */ |
baa355fd3 thp: file pages s... |
2528 |
spin_lock(&pgdata->split_queue_lock); |
e9b61f198 thp: reintroduce ... |
2529 2530 |
count = page_count(head); mapcount = total_mapcount(head); |
baa355fd3 thp: file pages s... |
2531 |
if (!mapcount && page_ref_freeze(head, 1 + extra_pins)) { |
9a982250f thp: introduce de... |
2532 |
if (!list_empty(page_deferred_list(head))) { |
a3d0a9185 thp: make split_q... |
2533 |
pgdata->split_queue_len--; |
9a982250f thp: introduce de... |
2534 2535 |
list_del(page_deferred_list(head)); } |
65c453778 mm, rmap: account... |
2536 |
if (mapping) |
11fb99898 mm: move most fil... |
2537 |
__dec_node_page_state(page, NR_SHMEM_THPS); |
baa355fd3 thp: file pages s... |
2538 |
spin_unlock(&pgdata->split_queue_lock); |
6f75a0983 mm/huge_memory: f... |
2539 |
__split_huge_page(page, list, end, flags); |
59807685a mm, THP, swap: su... |
2540 2541 2542 2543 2544 2545 |
if (PageSwapCache(head)) { swp_entry_t entry = { .val = page_private(head) }; ret = split_swap_cluster(entry); } else ret = 0; |
e9b61f198 thp: reintroduce ... |
2546 |
} else { |
baa355fd3 thp: file pages s... |
2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 |
if (IS_ENABLED(CONFIG_DEBUG_VM) && mapcount) { pr_alert("total_mapcount: %u, page_count(): %u ", mapcount, count); if (PageTail(page)) dump_page(head, NULL); dump_page(page, "total_mapcount(head) > 0"); BUG(); } spin_unlock(&pgdata->split_queue_lock); fail: if (mapping) spin_unlock(&mapping->tree_lock); |
a52633d8e mm, vmscan: move ... |
2559 |
spin_unlock_irqrestore(zone_lru_lock(page_zone(head)), flags); |
e12b67d81 mm/huge_memory: r... |
2560 |
remap_page(head); |
e9b61f198 thp: reintroduce ... |
2561 2562 2563 2564 |
ret = -EBUSY; } out_unlock: |
baa355fd3 thp: file pages s... |
2565 2566 2567 2568 2569 2570 |
if (anon_vma) { anon_vma_unlock_write(anon_vma); put_anon_vma(anon_vma); } if (mapping) i_mmap_unlock_read(mapping); |
e9b61f198 thp: reintroduce ... |
2571 2572 2573 2574 |
out: count_vm_event(!ret ? THP_SPLIT_PAGE : THP_SPLIT_PAGE_FAILED); return ret; } |
9a982250f thp: introduce de... |
2575 2576 2577 |
void free_transhuge_page(struct page *page) { |
a3d0a9185 thp: make split_q... |
2578 |
struct pglist_data *pgdata = NODE_DATA(page_to_nid(page)); |
9a982250f thp: introduce de... |
2579 |
unsigned long flags; |
a3d0a9185 thp: make split_q... |
2580 |
spin_lock_irqsave(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2581 |
if (!list_empty(page_deferred_list(page))) { |
a3d0a9185 thp: make split_q... |
2582 |
pgdata->split_queue_len--; |
9a982250f thp: introduce de... |
2583 2584 |
list_del(page_deferred_list(page)); } |
a3d0a9185 thp: make split_q... |
2585 |
spin_unlock_irqrestore(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2586 2587 2588 2589 2590 |
free_compound_page(page); } void deferred_split_huge_page(struct page *page) { |
a3d0a9185 thp: make split_q... |
2591 |
struct pglist_data *pgdata = NODE_DATA(page_to_nid(page)); |
9a982250f thp: introduce de... |
2592 2593 2594 |
unsigned long flags; VM_BUG_ON_PAGE(!PageTransHuge(page), page); |
a3d0a9185 thp: make split_q... |
2595 |
spin_lock_irqsave(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2596 |
if (list_empty(page_deferred_list(page))) { |
f9719a03d thp, vmstats: cou... |
2597 |
count_vm_event(THP_DEFERRED_SPLIT_PAGE); |
a3d0a9185 thp: make split_q... |
2598 2599 |
list_add_tail(page_deferred_list(page), &pgdata->split_queue); pgdata->split_queue_len++; |
9a982250f thp: introduce de... |
2600 |
} |
a3d0a9185 thp: make split_q... |
2601 |
spin_unlock_irqrestore(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2602 2603 2604 2605 2606 |
} static unsigned long deferred_split_count(struct shrinker *shrink, struct shrink_control *sc) { |
a3d0a9185 thp: make split_q... |
2607 |
struct pglist_data *pgdata = NODE_DATA(sc->nid); |
cb8d68ec1 thp: change defer... |
2608 |
return ACCESS_ONCE(pgdata->split_queue_len); |
9a982250f thp: introduce de... |
2609 2610 2611 2612 2613 |
} static unsigned long deferred_split_scan(struct shrinker *shrink, struct shrink_control *sc) { |
a3d0a9185 thp: make split_q... |
2614 |
struct pglist_data *pgdata = NODE_DATA(sc->nid); |
9a982250f thp: introduce de... |
2615 2616 2617 2618 |
unsigned long flags; LIST_HEAD(list), *pos, *next; struct page *page; int split = 0; |
a3d0a9185 thp: make split_q... |
2619 |
spin_lock_irqsave(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2620 |
/* Take pin on all head pages to avoid freeing them under us */ |
ae026204a thp: make deferre... |
2621 |
list_for_each_safe(pos, next, &pgdata->split_queue) { |
9a982250f thp: introduce de... |
2622 2623 |
page = list_entry((void *)pos, struct page, mapping); page = compound_head(page); |
e3ae19535 thp: limit number... |
2624 2625 2626 2627 |
if (get_page_unless_zero(page)) { list_move(page_deferred_list(page), &list); } else { /* We lost race with put_compound_page() */ |
9a982250f thp: introduce de... |
2628 |
list_del_init(page_deferred_list(page)); |
a3d0a9185 thp: make split_q... |
2629 |
pgdata->split_queue_len--; |
9a982250f thp: introduce de... |
2630 |
} |
e3ae19535 thp: limit number... |
2631 2632 |
if (!--sc->nr_to_scan) break; |
9a982250f thp: introduce de... |
2633 |
} |
a3d0a9185 thp: make split_q... |
2634 |
spin_unlock_irqrestore(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2635 2636 2637 |
list_for_each_safe(pos, next, &list) { page = list_entry((void *)pos, struct page, mapping); |
b6b6783c8 mm/thp: do not wa... |
2638 2639 |
if (!trylock_page(page)) goto next; |
9a982250f thp: introduce de... |
2640 2641 2642 2643 |
/* split_huge_page() removes page from list on success */ if (!split_huge_page(page)) split++; unlock_page(page); |
b6b6783c8 mm/thp: do not wa... |
2644 |
next: |
9a982250f thp: introduce de... |
2645 2646 |
put_page(page); } |
a3d0a9185 thp: make split_q... |
2647 2648 2649 |
spin_lock_irqsave(&pgdata->split_queue_lock, flags); list_splice_tail(&list, &pgdata->split_queue); spin_unlock_irqrestore(&pgdata->split_queue_lock, flags); |
9a982250f thp: introduce de... |
2650 |
|
cb8d68ec1 thp: change defer... |
2651 2652 2653 2654 2655 2656 2657 |
/* * Stop shrinker if we didn't split any page, but the queue is empty. * This can happen if pages were freed under us. */ if (!split && list_empty(&pgdata->split_queue)) return SHRINK_STOP; return split; |
9a982250f thp: introduce de... |
2658 2659 2660 2661 2662 2663 |
} static struct shrinker deferred_split_shrinker = { .count_objects = deferred_split_count, .scan_objects = deferred_split_scan, .seeks = DEFAULT_SEEKS, |
a3d0a9185 thp: make split_q... |
2664 |
.flags = SHRINKER_NUMA_AWARE, |
9a982250f thp: introduce de... |
2665 |
}; |
49071d436 thp: add debugfs ... |
2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 |
#ifdef CONFIG_DEBUG_FS static int split_huge_pages_set(void *data, u64 val) { struct zone *zone; struct page *page; unsigned long pfn, max_zone_pfn; unsigned long total = 0, split = 0; if (val != 1) return -EINVAL; for_each_populated_zone(zone) { max_zone_pfn = zone_end_pfn(zone); for (pfn = zone->zone_start_pfn; pfn < max_zone_pfn; pfn++) { if (!pfn_valid(pfn)) continue; page = pfn_to_page(pfn); if (!get_page_unless_zero(page)) continue; if (zone != page_zone(page)) goto next; |
baa355fd3 thp: file pages s... |
2690 |
if (!PageHead(page) || PageHuge(page) || !PageLRU(page)) |
49071d436 thp: add debugfs ... |
2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 |
goto next; total++; lock_page(page); if (!split_huge_page(page)) split++; unlock_page(page); next: put_page(page); } } |
145bdaa15 mm: thp: correct ... |
2702 2703 |
pr_info("%lu of %lu THP split ", split, total); |
49071d436 thp: add debugfs ... |
2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 |
return 0; } DEFINE_SIMPLE_ATTRIBUTE(split_huge_pages_fops, NULL, split_huge_pages_set, "%llu "); static int __init split_huge_pages_debugfs(void) { void *ret; |
145bdaa15 mm: thp: correct ... |
2714 |
ret = debugfs_create_file("split_huge_pages", 0200, NULL, NULL, |
49071d436 thp: add debugfs ... |
2715 2716 2717 2718 2719 2720 2721 |
&split_huge_pages_fops); if (!ret) pr_warn("Failed to create split_huge_pages in debugfs"); return 0; } late_initcall(split_huge_pages_debugfs); #endif |
616b83715 mm: thp: enable t... |
2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 |
#ifdef CONFIG_ARCH_ENABLE_THP_MIGRATION void set_pmd_migration_entry(struct page_vma_mapped_walk *pvmw, struct page *page) { struct vm_area_struct *vma = pvmw->vma; struct mm_struct *mm = vma->vm_mm; unsigned long address = pvmw->address; pmd_t pmdval; swp_entry_t entry; |
ab6e3d093 mm: soft-dirty: k... |
2732 |
pmd_t pmdswp; |
616b83715 mm: thp: enable t... |
2733 2734 2735 |
if (!(pvmw->pmd && !pvmw->pte)) return; |
616b83715 mm: thp: enable t... |
2736 2737 2738 2739 2740 2741 |
flush_cache_range(vma, address, address + HPAGE_PMD_SIZE); pmdval = *pvmw->pmd; pmdp_invalidate(vma, address, pvmw->pmd); if (pmd_dirty(pmdval)) set_page_dirty(page); entry = make_migration_entry(page, pmd_write(pmdval)); |
ab6e3d093 mm: soft-dirty: k... |
2742 2743 2744 2745 |
pmdswp = swp_entry_to_pmd(entry); if (pmd_soft_dirty(pmdval)) pmdswp = pmd_swp_mksoft_dirty(pmdswp); set_pmd_at(mm, address, pvmw->pmd, pmdswp); |
616b83715 mm: thp: enable t... |
2746 2747 |
page_remove_rmap(page, true); put_page(page); |
616b83715 mm: thp: enable t... |
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 |
} void remove_migration_pmd(struct page_vma_mapped_walk *pvmw, struct page *new) { struct vm_area_struct *vma = pvmw->vma; struct mm_struct *mm = vma->vm_mm; unsigned long address = pvmw->address; unsigned long mmun_start = address & HPAGE_PMD_MASK; pmd_t pmde; swp_entry_t entry; if (!(pvmw->pmd && !pvmw->pte)) return; entry = pmd_to_swp_entry(*pvmw->pmd); get_page(new); pmde = pmd_mkold(mk_huge_pmd(new, vma->vm_page_prot)); |
ab6e3d093 mm: soft-dirty: k... |
2765 2766 |
if (pmd_swp_soft_dirty(*pvmw->pmd)) pmde = pmd_mksoft_dirty(pmde); |
616b83715 mm: thp: enable t... |
2767 2768 2769 2770 2771 2772 |
if (is_write_migration_entry(entry)) pmde = maybe_pmd_mkwrite(pmde, vma); flush_cache_range(vma, mmun_start, mmun_start + HPAGE_PMD_SIZE); page_add_anon_rmap(new, vma, mmun_start, true); set_pmd_at(mm, mmun_start, pvmw->pmd, pmde); |
a2e0493f9 mm, thp: fix mloc... |
2773 |
if ((vma->vm_flags & VM_LOCKED) && !PageDoubleMap(new)) |
616b83715 mm: thp: enable t... |
2774 2775 2776 2777 |
mlock_vma_page(new); update_mmu_cache_pmd(vma, address, pvmw->pmd); } #endif |