MIPS: fix blast_icache32 on loongson2
[linux-2.6-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>
9800339b 8#include <linux/swap.h>
33806f06 9#include <linux/swapops.h>
00619bcc
JM
10#include <linux/mman.h>
11#include <linux/hugetlb.h>
12
96840aa0 13#include <asm/uaccess.h>
30992c97 14
6038def0
NK
15#include "internal.h"
16
a8d154b0 17#define CREATE_TRACE_POINTS
ad8d75ff 18#include <trace/events/kmem.h>
a8d154b0 19
30992c97 20/**
30992c97 21 * kstrdup - allocate space for and copy an existing string
30992c97
MM
22 * @s: the string to duplicate
23 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
24 */
25char *kstrdup(const char *s, gfp_t gfp)
26{
27 size_t len;
28 char *buf;
29
30 if (!s)
31 return NULL;
32
33 len = strlen(s) + 1;
1d2c8eea 34 buf = kmalloc_track_caller(len, gfp);
30992c97
MM
35 if (buf)
36 memcpy(buf, s, len);
37 return buf;
38}
39EXPORT_SYMBOL(kstrdup);
96840aa0 40
1e66df3e
JF
41/**
42 * kstrndup - allocate space for and copy an existing string
43 * @s: the string to duplicate
44 * @max: read at most @max chars from @s
45 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
46 */
47char *kstrndup(const char *s, size_t max, gfp_t gfp)
48{
49 size_t len;
50 char *buf;
51
52 if (!s)
53 return NULL;
54
55 len = strnlen(s, max);
56 buf = kmalloc_track_caller(len+1, gfp);
57 if (buf) {
58 memcpy(buf, s, len);
59 buf[len] = '\0';
60 }
61 return buf;
62}
63EXPORT_SYMBOL(kstrndup);
64
1a2f67b4
AD
65/**
66 * kmemdup - duplicate region of memory
67 *
68 * @src: memory region to duplicate
69 * @len: memory region length
70 * @gfp: GFP mask to use
71 */
72void *kmemdup(const void *src, size_t len, gfp_t gfp)
73{
74 void *p;
75
1d2c8eea 76 p = kmalloc_track_caller(len, gfp);
1a2f67b4
AD
77 if (p)
78 memcpy(p, src, len);
79 return p;
80}
81EXPORT_SYMBOL(kmemdup);
82
610a77e0
LZ
83/**
84 * memdup_user - duplicate memory region from user space
85 *
86 * @src: source address in user space
87 * @len: number of bytes to copy
88 *
89 * Returns an ERR_PTR() on failure.
90 */
91void *memdup_user(const void __user *src, size_t len)
92{
93 void *p;
94
95 /*
96 * Always use GFP_KERNEL, since copy_from_user() can sleep and
97 * cause pagefault, which makes it pointless to use GFP_NOFS
98 * or GFP_ATOMIC.
99 */
100 p = kmalloc_track_caller(len, GFP_KERNEL);
101 if (!p)
102 return ERR_PTR(-ENOMEM);
103
104 if (copy_from_user(p, src, len)) {
105 kfree(p);
106 return ERR_PTR(-EFAULT);
107 }
108
109 return p;
110}
111EXPORT_SYMBOL(memdup_user);
112
e21827aa
EG
113static __always_inline void *__do_krealloc(const void *p, size_t new_size,
114 gfp_t flags)
115{
116 void *ret;
117 size_t ks = 0;
118
119 if (p)
120 ks = ksize(p);
121
122 if (ks >= new_size)
123 return (void *)p;
124
125 ret = kmalloc_track_caller(new_size, flags);
126 if (ret && p)
127 memcpy(ret, p, ks);
128
129 return ret;
130}
131
ef2ad80c 132/**
93bc4e89 133 * __krealloc - like krealloc() but don't free @p.
ef2ad80c
CL
134 * @p: object to reallocate memory for.
135 * @new_size: how many bytes of memory are required.
136 * @flags: the type of memory to allocate.
137 *
93bc4e89
PE
138 * This function is like krealloc() except it never frees the originally
139 * allocated buffer. Use this if you don't want to free the buffer immediately
140 * like, for example, with RCU.
ef2ad80c 141 */
93bc4e89 142void *__krealloc(const void *p, size_t new_size, gfp_t flags)
ef2ad80c 143{
93bc4e89 144 if (unlikely(!new_size))
6cb8f913 145 return ZERO_SIZE_PTR;
ef2ad80c 146
e21827aa 147 return __do_krealloc(p, new_size, flags);
ef8b4520 148
93bc4e89
PE
149}
150EXPORT_SYMBOL(__krealloc);
151
152/**
153 * krealloc - reallocate memory. The contents will remain unchanged.
154 * @p: object to reallocate memory for.
155 * @new_size: how many bytes of memory are required.
156 * @flags: the type of memory to allocate.
157 *
158 * The contents of the object pointed to are preserved up to the
159 * lesser of the new and old sizes. If @p is %NULL, krealloc()
0db10c8e 160 * behaves exactly like kmalloc(). If @new_size is 0 and @p is not a
93bc4e89
PE
161 * %NULL pointer, the object pointed to is freed.
162 */
163void *krealloc(const void *p, size_t new_size, gfp_t flags)
164{
165 void *ret;
166
167 if (unlikely(!new_size)) {
ef2ad80c 168 kfree(p);
93bc4e89 169 return ZERO_SIZE_PTR;
ef2ad80c 170 }
93bc4e89 171
e21827aa 172 ret = __do_krealloc(p, new_size, flags);
93bc4e89
PE
173 if (ret && p != ret)
174 kfree(p);
175
ef2ad80c
CL
176 return ret;
177}
178EXPORT_SYMBOL(krealloc);
179
3ef0e5ba
JW
180/**
181 * kzfree - like kfree but zero memory
182 * @p: object to free memory of
183 *
184 * The memory of the object @p points to is zeroed before freed.
185 * If @p is %NULL, kzfree() does nothing.
a234bdc9
PE
186 *
187 * Note: this function zeroes the whole allocated buffer which can be a good
188 * deal bigger than the requested buffer size passed to kmalloc(). So be
189 * careful when using this function in performance sensitive code.
3ef0e5ba
JW
190 */
191void kzfree(const void *p)
192{
193 size_t ks;
194 void *mem = (void *)p;
195
196 if (unlikely(ZERO_OR_NULL_PTR(mem)))
197 return;
198 ks = ksize(mem);
199 memset(mem, 0, ks);
200 kfree(mem);
201}
202EXPORT_SYMBOL(kzfree);
203
96840aa0
DA
204/*
205 * strndup_user - duplicate an existing string from user space
96840aa0
DA
206 * @s: The string to duplicate
207 * @n: Maximum number of bytes to copy, including the trailing NUL.
208 */
209char *strndup_user(const char __user *s, long n)
210{
211 char *p;
212 long length;
213
214 length = strnlen_user(s, n);
215
216 if (!length)
217 return ERR_PTR(-EFAULT);
218
219 if (length > n)
220 return ERR_PTR(-EINVAL);
221
90d74045 222 p = memdup_user(s, length);
96840aa0 223
90d74045
JL
224 if (IS_ERR(p))
225 return p;
96840aa0
DA
226
227 p[length - 1] = '\0';
228
229 return p;
230}
231EXPORT_SYMBOL(strndup_user);
16d69265 232
6038def0
NK
233void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
234 struct vm_area_struct *prev, struct rb_node *rb_parent)
235{
236 struct vm_area_struct *next;
237
238 vma->vm_prev = prev;
239 if (prev) {
240 next = prev->vm_next;
241 prev->vm_next = vma;
242 } else {
243 mm->mmap = vma;
244 if (rb_parent)
245 next = rb_entry(rb_parent,
246 struct vm_area_struct, vm_rb);
247 else
248 next = NULL;
249 }
250 vma->vm_next = next;
251 if (next)
252 next->vm_prev = vma;
253}
254
b7643757
SP
255/* Check if the vma is being used as a stack by this task */
256static int vm_is_stack_for_task(struct task_struct *t,
257 struct vm_area_struct *vma)
258{
259 return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
260}
261
262/*
263 * Check if the vma is being used as a stack.
264 * If is_group is non-zero, check in the entire thread group or else
265 * just check in the current task. Returns the pid of the task that
266 * the vma is stack for.
267 */
268pid_t vm_is_stack(struct task_struct *task,
269 struct vm_area_struct *vma, int in_group)
270{
271 pid_t ret = 0;
272
273 if (vm_is_stack_for_task(task, vma))
274 return task->pid;
275
276 if (in_group) {
277 struct task_struct *t;
278 rcu_read_lock();
279 if (!pid_alive(task))
280 goto done;
281
282 t = task;
283 do {
284 if (vm_is_stack_for_task(t, vma)) {
285 ret = t->pid;
286 goto done;
287 }
288 } while_each_thread(task, t);
289done:
290 rcu_read_unlock();
291 }
292
293 return ret;
294}
295
efc1a3b1 296#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
16d69265
AM
297void arch_pick_mmap_layout(struct mm_struct *mm)
298{
299 mm->mmap_base = TASK_UNMAPPED_BASE;
300 mm->get_unmapped_area = arch_get_unmapped_area;
16d69265
AM
301}
302#endif
912985dc 303
45888a0c
XG
304/*
305 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
306 * back to the regular GUP.
25985edc 307 * If the architecture not support this function, simply return with no
45888a0c
XG
308 * page pinned
309 */
310int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
311 int nr_pages, int write, struct page **pages)
312{
313 return 0;
314}
315EXPORT_SYMBOL_GPL(__get_user_pages_fast);
316
9de100d0
AG
317/**
318 * get_user_pages_fast() - pin user pages in memory
319 * @start: starting user address
320 * @nr_pages: number of pages from start to pin
321 * @write: whether pages will be written to
322 * @pages: array that receives pointers to the pages pinned.
323 * Should be at least nr_pages long.
324 *
9de100d0
AG
325 * Returns number of pages pinned. This may be fewer than the number
326 * requested. If nr_pages is 0 or negative, returns 0. If no pages
327 * were pinned, returns -errno.
d2bf6be8
NP
328 *
329 * get_user_pages_fast provides equivalent functionality to get_user_pages,
330 * operating on current and current->mm, with force=0 and vma=NULL. However
331 * unlike get_user_pages, it must be called without mmap_sem held.
332 *
333 * get_user_pages_fast may take mmap_sem and page table locks, so no
334 * assumptions can be made about lack of locking. get_user_pages_fast is to be
335 * implemented in a way that is advantageous (vs get_user_pages()) when the
336 * user memory area is already faulted in and present in ptes. However if the
337 * pages have to be faulted in, it may turn out to be slightly slower so
338 * callers need to carefully consider what to use. On many architectures,
339 * get_user_pages_fast simply falls back to get_user_pages.
9de100d0 340 */
912985dc
RR
341int __attribute__((weak)) get_user_pages_fast(unsigned long start,
342 int nr_pages, int write, struct page **pages)
343{
344 struct mm_struct *mm = current->mm;
345 int ret;
346
347 down_read(&mm->mmap_sem);
348 ret = get_user_pages(current, mm, start, nr_pages,
349 write, 0, pages, NULL);
350 up_read(&mm->mmap_sem);
351
352 return ret;
353}
354EXPORT_SYMBOL_GPL(get_user_pages_fast);
ca2b84cb 355
eb36c587
AV
356unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
357 unsigned long len, unsigned long prot,
358 unsigned long flag, unsigned long pgoff)
359{
360 unsigned long ret;
361 struct mm_struct *mm = current->mm;
41badc15 362 unsigned long populate;
eb36c587
AV
363
364 ret = security_mmap_file(file, prot, flag);
365 if (!ret) {
366 down_write(&mm->mmap_sem);
bebeb3d6
ML
367 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
368 &populate);
eb36c587 369 up_write(&mm->mmap_sem);
41badc15
ML
370 if (populate)
371 mm_populate(ret, populate);
eb36c587
AV
372 }
373 return ret;
374}
375
376unsigned long vm_mmap(struct file *file, unsigned long addr,
377 unsigned long len, unsigned long prot,
378 unsigned long flag, unsigned long offset)
379{
380 if (unlikely(offset + PAGE_ALIGN(len) < offset))
381 return -EINVAL;
382 if (unlikely(offset & ~PAGE_MASK))
383 return -EINVAL;
384
385 return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
386}
387EXPORT_SYMBOL(vm_mmap);
388
9800339b
SL
389struct address_space *page_mapping(struct page *page)
390{
391 struct address_space *mapping = page->mapping;
392
393 VM_BUG_ON(PageSlab(page));
33806f06
SL
394 if (unlikely(PageSwapCache(page))) {
395 swp_entry_t entry;
396
397 entry.val = page_private(page);
398 mapping = swap_address_space(entry);
d2cf5ad6 399 } else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
9800339b
SL
400 mapping = NULL;
401 return mapping;
402}
403
00619bcc
JM
404/*
405 * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
406 */
407unsigned long vm_commit_limit(void)
408{
409 return ((totalram_pages - hugetlb_total_pages())
410 * sysctl_overcommit_ratio / 100) + total_swap_pages;
411}
412
413
ca2b84cb 414/* Tracepoints definitions. */
ca2b84cb
EGM
415EXPORT_TRACEPOINT_SYMBOL(kmalloc);
416EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
417EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
418EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
419EXPORT_TRACEPOINT_SYMBOL(kfree);
420EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);