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