mm/memory-failure: call shake_page() when error hits thp tail page
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / mm / util.c
CommitLineData
16d69265 1#include <linux/mm.h>
30992c97
MM
2#include <linux/slab.h>
3#include <linux/string.h>
b95f1b31 4#include <linux/export.h>
96840aa0 5#include <linux/err.h>
3b8f14b4 6#include <linux/sched.h>
eb36c587 7#include <linux/security.h>
9800339b 8#include <linux/swap.h>
33806f06 9#include <linux/swapops.h>
96840aa0 10#include <asm/uaccess.h>
30992c97 11
6038def0
NK
12#include "internal.h"
13
a8d154b0 14#define CREATE_TRACE_POINTS
ad8d75ff 15#include <trace/events/kmem.h>
a8d154b0 16
30992c97 17/**
30992c97 18 * kstrdup - allocate space for and copy an existing string
30992c97
MM
19 * @s: the string to duplicate
20 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
21 */
22char *kstrdup(const char *s, gfp_t gfp)
23{
24 size_t len;
25 char *buf;
26
27 if (!s)
28 return NULL;
29
30 len = strlen(s) + 1;
1d2c8eea 31 buf = kmalloc_track_caller(len, gfp);
30992c97
MM
32 if (buf)
33 memcpy(buf, s, len);
34 return buf;
35}
36EXPORT_SYMBOL(kstrdup);
96840aa0 37
1e66df3e
JF
38/**
39 * kstrndup - allocate space for and copy an existing string
40 * @s: the string to duplicate
41 * @max: read at most @max chars from @s
42 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
43 */
44char *kstrndup(const char *s, size_t max, gfp_t gfp)
45{
46 size_t len;
47 char *buf;
48
49 if (!s)
50 return NULL;
51
52 len = strnlen(s, max);
53 buf = kmalloc_track_caller(len+1, gfp);
54 if (buf) {
55 memcpy(buf, s, len);
56 buf[len] = '\0';
57 }
58 return buf;
59}
60EXPORT_SYMBOL(kstrndup);
61
1a2f67b4
AD
62/**
63 * kmemdup - duplicate region of memory
64 *
65 * @src: memory region to duplicate
66 * @len: memory region length
67 * @gfp: GFP mask to use
68 */
69void *kmemdup(const void *src, size_t len, gfp_t gfp)
70{
71 void *p;
72
1d2c8eea 73 p = kmalloc_track_caller(len, gfp);
1a2f67b4
AD
74 if (p)
75 memcpy(p, src, len);
76 return p;
77}
78EXPORT_SYMBOL(kmemdup);
79
610a77e0
LZ
80/**
81 * memdup_user - duplicate memory region from user space
82 *
83 * @src: source address in user space
84 * @len: number of bytes to copy
85 *
86 * Returns an ERR_PTR() on failure.
87 */
88void *memdup_user(const void __user *src, size_t len)
89{
90 void *p;
91
92 /*
93 * Always use GFP_KERNEL, since copy_from_user() can sleep and
94 * cause pagefault, which makes it pointless to use GFP_NOFS
95 * or GFP_ATOMIC.
96 */
97 p = kmalloc_track_caller(len, GFP_KERNEL);
98 if (!p)
99 return ERR_PTR(-ENOMEM);
100
101 if (copy_from_user(p, src, len)) {
102 kfree(p);
103 return ERR_PTR(-EFAULT);
104 }
105
106 return p;
107}
108EXPORT_SYMBOL(memdup_user);
109
e21827aa
EG
110static __always_inline void *__do_krealloc(const void *p, size_t new_size,
111 gfp_t flags)
112{
113 void *ret;
114 size_t ks = 0;
115
116 if (p)
117 ks = ksize(p);
118
119 if (ks >= new_size)
120 return (void *)p;
121
122 ret = kmalloc_track_caller(new_size, flags);
123 if (ret && p)
124 memcpy(ret, p, ks);
125
126 return ret;
127}
128
ef2ad80c 129/**
93bc4e89 130 * __krealloc - like krealloc() but don't free @p.
ef2ad80c
CL
131 * @p: object to reallocate memory for.
132 * @new_size: how many bytes of memory are required.
133 * @flags: the type of memory to allocate.
134 *
93bc4e89
PE
135 * This function is like krealloc() except it never frees the originally
136 * allocated buffer. Use this if you don't want to free the buffer immediately
137 * like, for example, with RCU.
ef2ad80c 138 */
93bc4e89 139void *__krealloc(const void *p, size_t new_size, gfp_t flags)
ef2ad80c 140{
93bc4e89 141 if (unlikely(!new_size))
6cb8f913 142 return ZERO_SIZE_PTR;
ef2ad80c 143
e21827aa 144 return __do_krealloc(p, new_size, flags);
ef8b4520 145
93bc4e89
PE
146}
147EXPORT_SYMBOL(__krealloc);
148
149/**
150 * krealloc - reallocate memory. The contents will remain unchanged.
151 * @p: object to reallocate memory for.
152 * @new_size: how many bytes of memory are required.
153 * @flags: the type of memory to allocate.
154 *
155 * The contents of the object pointed to are preserved up to the
156 * lesser of the new and old sizes. If @p is %NULL, krealloc()
0db10c8e 157 * behaves exactly like kmalloc(). If @new_size is 0 and @p is not a
93bc4e89
PE
158 * %NULL pointer, the object pointed to is freed.
159 */
160void *krealloc(const void *p, size_t new_size, gfp_t flags)
161{
162 void *ret;
163
164 if (unlikely(!new_size)) {
ef2ad80c 165 kfree(p);
93bc4e89 166 return ZERO_SIZE_PTR;
ef2ad80c 167 }
93bc4e89 168
e21827aa 169 ret = __do_krealloc(p, new_size, flags);
93bc4e89
PE
170 if (ret && p != ret)
171 kfree(p);
172
ef2ad80c
CL
173 return ret;
174}
175EXPORT_SYMBOL(krealloc);
176
3ef0e5ba
JW
177/**
178 * kzfree - like kfree but zero memory
179 * @p: object to free memory of
180 *
181 * The memory of the object @p points to is zeroed before freed.
182 * If @p is %NULL, kzfree() does nothing.
a234bdc9
PE
183 *
184 * Note: this function zeroes the whole allocated buffer which can be a good
185 * deal bigger than the requested buffer size passed to kmalloc(). So be
186 * careful when using this function in performance sensitive code.
3ef0e5ba
JW
187 */
188void kzfree(const void *p)
189{
190 size_t ks;
191 void *mem = (void *)p;
192
193 if (unlikely(ZERO_OR_NULL_PTR(mem)))
194 return;
195 ks = ksize(mem);
196 memset(mem, 0, ks);
197 kfree(mem);
198}
199EXPORT_SYMBOL(kzfree);
200
96840aa0
DA
201/*
202 * strndup_user - duplicate an existing string from user space
96840aa0
DA
203 * @s: The string to duplicate
204 * @n: Maximum number of bytes to copy, including the trailing NUL.
205 */
206char *strndup_user(const char __user *s, long n)
207{
208 char *p;
209 long length;
210
211 length = strnlen_user(s, n);
212
213 if (!length)
214 return ERR_PTR(-EFAULT);
215
216 if (length > n)
217 return ERR_PTR(-EINVAL);
218
90d74045 219 p = memdup_user(s, length);
96840aa0 220
90d74045
JL
221 if (IS_ERR(p))
222 return p;
96840aa0
DA
223
224 p[length - 1] = '\0';
225
226 return p;
227}
228EXPORT_SYMBOL(strndup_user);
16d69265 229
6038def0
NK
230void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
231 struct vm_area_struct *prev, struct rb_node *rb_parent)
232{
233 struct vm_area_struct *next;
234
235 vma->vm_prev = prev;
236 if (prev) {
237 next = prev->vm_next;
238 prev->vm_next = vma;
239 } else {
240 mm->mmap = vma;
241 if (rb_parent)
242 next = rb_entry(rb_parent,
243 struct vm_area_struct, vm_rb);
244 else
245 next = NULL;
246 }
247 vma->vm_next = next;
248 if (next)
249 next->vm_prev = vma;
250}
251
b7643757
SP
252/* Check if the vma is being used as a stack by this task */
253static int vm_is_stack_for_task(struct task_struct *t,
254 struct vm_area_struct *vma)
255{
256 return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
257}
258
259/*
260 * Check if the vma is being used as a stack.
261 * If is_group is non-zero, check in the entire thread group or else
262 * just check in the current task. Returns the pid of the task that
263 * the vma is stack for.
264 */
265pid_t vm_is_stack(struct task_struct *task,
266 struct vm_area_struct *vma, int in_group)
267{
268 pid_t ret = 0;
269
270 if (vm_is_stack_for_task(task, vma))
271 return task->pid;
272
273 if (in_group) {
274 struct task_struct *t;
b7643757 275
35818329
ON
276 rcu_read_lock();
277 for_each_thread(task, t) {
b7643757
SP
278 if (vm_is_stack_for_task(t, vma)) {
279 ret = t->pid;
280 goto done;
281 }
35818329 282 }
b7643757
SP
283done:
284 rcu_read_unlock();
285 }
286
287 return ret;
288}
289
efc1a3b1 290#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
16d69265
AM
291void arch_pick_mmap_layout(struct mm_struct *mm)
292{
293 mm->mmap_base = TASK_UNMAPPED_BASE;
294 mm->get_unmapped_area = arch_get_unmapped_area;
295 mm->unmap_area = arch_unmap_area;
296}
297#endif
912985dc 298
45888a0c
XG
299/*
300 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
301 * back to the regular GUP.
25985edc 302 * If the architecture not support this function, simply return with no
45888a0c
XG
303 * page pinned
304 */
305int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
306 int nr_pages, int write, struct page **pages)
307{
308 return 0;
309}
310EXPORT_SYMBOL_GPL(__get_user_pages_fast);
311
9de100d0
AG
312/**
313 * get_user_pages_fast() - pin user pages in memory
314 * @start: starting user address
315 * @nr_pages: number of pages from start to pin
316 * @write: whether pages will be written to
317 * @pages: array that receives pointers to the pages pinned.
318 * Should be at least nr_pages long.
319 *
9de100d0
AG
320 * Returns number of pages pinned. This may be fewer than the number
321 * requested. If nr_pages is 0 or negative, returns 0. If no pages
322 * were pinned, returns -errno.
d2bf6be8
NP
323 *
324 * get_user_pages_fast provides equivalent functionality to get_user_pages,
325 * operating on current and current->mm, with force=0 and vma=NULL. However
326 * unlike get_user_pages, it must be called without mmap_sem held.
327 *
328 * get_user_pages_fast may take mmap_sem and page table locks, so no
329 * assumptions can be made about lack of locking. get_user_pages_fast is to be
330 * implemented in a way that is advantageous (vs get_user_pages()) when the
331 * user memory area is already faulted in and present in ptes. However if the
332 * pages have to be faulted in, it may turn out to be slightly slower so
333 * callers need to carefully consider what to use. On many architectures,
334 * get_user_pages_fast simply falls back to get_user_pages.
9de100d0 335 */
912985dc
RR
336int __attribute__((weak)) get_user_pages_fast(unsigned long start,
337 int nr_pages, int write, struct page **pages)
338{
339 struct mm_struct *mm = current->mm;
340 int ret;
341
342 down_read(&mm->mmap_sem);
343 ret = get_user_pages(current, mm, start, nr_pages,
344 write, 0, pages, NULL);
345 up_read(&mm->mmap_sem);
346
347 return ret;
348}
349EXPORT_SYMBOL_GPL(get_user_pages_fast);
ca2b84cb 350
eb36c587
AV
351unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
352 unsigned long len, unsigned long prot,
353 unsigned long flag, unsigned long pgoff)
354{
355 unsigned long ret;
356 struct mm_struct *mm = current->mm;
41badc15 357 unsigned long populate;
eb36c587
AV
358
359 ret = security_mmap_file(file, prot, flag);
360 if (!ret) {
361 down_write(&mm->mmap_sem);
bebeb3d6
ML
362 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
363 &populate);
eb36c587 364 up_write(&mm->mmap_sem);
41badc15
ML
365 if (populate)
366 mm_populate(ret, populate);
eb36c587
AV
367 }
368 return ret;
369}
370
371unsigned long vm_mmap(struct file *file, unsigned long addr,
372 unsigned long len, unsigned long prot,
373 unsigned long flag, unsigned long offset)
374{
375 if (unlikely(offset + PAGE_ALIGN(len) < offset))
376 return -EINVAL;
377 if (unlikely(offset & ~PAGE_MASK))
378 return -EINVAL;
379
380 return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
381}
382EXPORT_SYMBOL(vm_mmap);
383
9800339b
SL
384struct address_space *page_mapping(struct page *page)
385{
386 struct address_space *mapping = page->mapping;
387
388 VM_BUG_ON(PageSlab(page));
389#ifdef CONFIG_SWAP
33806f06
SL
390 if (unlikely(PageSwapCache(page))) {
391 swp_entry_t entry;
392
393 entry.val = page_private(page);
394 mapping = swap_address_space(entry);
395 } else
9800339b
SL
396#endif
397 if ((unsigned long)mapping & PAGE_MAPPING_ANON)
398 mapping = NULL;
399 return mapping;
400}
401
ca2b84cb 402/* Tracepoints definitions. */
ca2b84cb
EGM
403EXPORT_TRACEPOINT_SYMBOL(kmalloc);
404EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
405EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
406EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
407EXPORT_TRACEPOINT_SYMBOL(kfree);
408EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);