blob: ab1424dbe2e6c9396ee66ab4446f1bc1cd557382 [file] [log] [blame]
Andrew Morton16d69262008-07-25 19:44:36 -07001#include <linux/mm.h>
Matt Mackall30992c92006-01-08 01:01:43 -08002#include <linux/slab.h>
3#include <linux/string.h>
Paul Gortmakerb95f1b312011-10-16 02:01:52 -04004#include <linux/export.h>
Davi Arnaut96840aa2006-03-24 03:18:42 -08005#include <linux/err.h>
Adrian Bunk3b8f14b2008-07-26 15:22:28 -07006#include <linux/sched.h>
Al Viroeb36c582012-05-30 20:17:35 -04007#include <linux/security.h>
Shaohua Li98003392013-02-22 16:34:35 -08008#include <linux/swap.h>
Shaohua Li33806f02013-02-22 16:34:37 -08009#include <linux/swapops.h>
Davi Arnaut96840aa2006-03-24 03:18:42 -080010#include <asm/uaccess.h>
Matt Mackall30992c92006-01-08 01:01:43 -080011
Namhyung Kim6038def2011-05-24 17:11:22 -070012#include "internal.h"
13
Steven Rostedta8d154b2009-04-10 09:36:00 -040014#define CREATE_TRACE_POINTS
Steven Rostedtad8d75f2009-04-14 19:39:12 -040015#include <trace/events/kmem.h>
Steven Rostedta8d154b2009-04-10 09:36:00 -040016
Matt Mackall30992c92006-01-08 01:01:43 -080017/**
Matt Mackall30992c92006-01-08 01:01:43 -080018 * kstrdup - allocate space for and copy an existing string
Matt Mackall30992c92006-01-08 01:01:43 -080019 * @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;
Christoph Hellwig1d2c8ee2006-10-04 02:15:25 -070031 buf = kmalloc_track_caller(len, gfp);
Matt Mackall30992c92006-01-08 01:01:43 -080032 if (buf)
33 memcpy(buf, s, len);
34 return buf;
35}
36EXPORT_SYMBOL(kstrdup);
Davi Arnaut96840aa2006-03-24 03:18:42 -080037
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -070038/**
Jeremy Fitzhardinge1e66df32007-07-17 18:37:02 -070039 * 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
62/**
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -070063 * 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
Christoph Hellwig1d2c8ee2006-10-04 02:15:25 -070073 p = kmalloc_track_caller(len, gfp);
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -070074 if (p)
75 memcpy(p, src, len);
76 return p;
77}
78EXPORT_SYMBOL(kmemdup);
79
Christoph Lameteref2ad802007-07-17 04:03:21 -070080/**
Li Zefan610a77e2009-03-31 15:23:16 -070081 * 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
Ezequiel Garciae21827a2012-08-14 09:55:21 -0300110static __always_inline void *__do_krealloc(const void *p, size_t new_size,
111 gfp_t flags)
Pekka Enberg93bc4e82008-07-26 17:49:33 -0700112{
113 void *ret;
114 size_t ks = 0;
115
Pekka Enberg93bc4e82008-07-26 17:49:33 -0700116 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}
Ezequiel Garciae21827a2012-08-14 09:55:21 -0300128
129/**
130 * __krealloc - like krealloc() but don't free @p.
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 *
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.
138 */
139void *__krealloc(const void *p, size_t new_size, gfp_t flags)
140{
141 if (unlikely(!new_size))
142 return ZERO_SIZE_PTR;
143
144 return __do_krealloc(p, new_size, flags);
145
146}
Pekka Enberg93bc4e82008-07-26 17:49:33 -0700147EXPORT_SYMBOL(__krealloc);
148
149/**
Christoph Lameteref2ad802007-07-17 04:03:21 -0700150 * 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()
Borislav Petkov0db10c82012-10-11 21:05:10 +0200157 * behaves exactly like kmalloc(). If @new_size is 0 and @p is not a
Christoph Lameteref2ad802007-07-17 04:03:21 -0700158 * %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;
Christoph Lameteref2ad802007-07-17 04:03:21 -0700163
164 if (unlikely(!new_size)) {
165 kfree(p);
Christoph Lameter6cb8f912007-07-17 04:03:22 -0700166 return ZERO_SIZE_PTR;
Christoph Lameteref2ad802007-07-17 04:03:21 -0700167 }
168
Ezequiel Garciae21827a2012-08-14 09:55:21 -0300169 ret = __do_krealloc(p, new_size, flags);
Pekka Enberg93bc4e82008-07-26 17:49:33 -0700170 if (ret && p != ret)
Christoph Lameteref2ad802007-07-17 04:03:21 -0700171 kfree(p);
Pekka Enberg93bc4e82008-07-26 17:49:33 -0700172
Christoph Lameteref2ad802007-07-17 04:03:21 -0700173 return ret;
174}
175EXPORT_SYMBOL(krealloc);
176
Johannes Weiner3ef0e5b2009-02-20 15:38:41 -0800177/**
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.
Pekka Enberga234bdc2009-05-31 13:50:38 +0300183 *
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.
Johannes Weiner3ef0e5b2009-02-20 15:38:41 -0800187 */
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
Davi Arnaut96840aa2006-03-24 03:18:42 -0800201/*
202 * strndup_user - duplicate an existing string from user space
Davi Arnaut96840aa2006-03-24 03:18:42 -0800203 * @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
Julia Lawall90d74042010-08-09 17:18:26 -0700219 p = memdup_user(s, length);
Davi Arnaut96840aa2006-03-24 03:18:42 -0800220
Julia Lawall90d74042010-08-09 17:18:26 -0700221 if (IS_ERR(p))
222 return p;
Davi Arnaut96840aa2006-03-24 03:18:42 -0800223
224 p[length - 1] = '\0';
225
226 return p;
227}
228EXPORT_SYMBOL(strndup_user);
Andrew Morton16d69262008-07-25 19:44:36 -0700229
Namhyung Kim6038def2011-05-24 17:11:22 -0700230void __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
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700252/* 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;
275 rcu_read_lock();
276 if (!pid_alive(task))
277 goto done;
278
279 t = task;
280 do {
281 if (vm_is_stack_for_task(t, vma)) {
282 ret = t->pid;
283 goto done;
284 }
285 } while_each_thread(task, t);
286done:
287 rcu_read_unlock();
288 }
289
290 return ret;
291}
292
David Howellsefc1a3b2010-01-15 17:01:35 -0800293#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
Andrew Morton16d69262008-07-25 19:44:36 -0700294void arch_pick_mmap_layout(struct mm_struct *mm)
295{
296 mm->mmap_base = TASK_UNMAPPED_BASE;
297 mm->get_unmapped_area = arch_get_unmapped_area;
298 mm->unmap_area = arch_unmap_area;
299}
300#endif
Rusty Russell912985d2008-08-12 17:52:52 -0500301
Xiao Guangrong45888a02010-08-22 19:08:57 +0800302/*
303 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
304 * back to the regular GUP.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300305 * If the architecture not support this function, simply return with no
Xiao Guangrong45888a02010-08-22 19:08:57 +0800306 * page pinned
307 */
308int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
309 int nr_pages, int write, struct page **pages)
310{
311 return 0;
312}
313EXPORT_SYMBOL_GPL(__get_user_pages_fast);
314
Andy Grover9de100d2009-04-13 14:40:05 -0700315/**
316 * get_user_pages_fast() - pin user pages in memory
317 * @start: starting user address
318 * @nr_pages: number of pages from start to pin
319 * @write: whether pages will be written to
320 * @pages: array that receives pointers to the pages pinned.
321 * Should be at least nr_pages long.
322 *
Andy Grover9de100d2009-04-13 14:40:05 -0700323 * Returns number of pages pinned. This may be fewer than the number
324 * requested. If nr_pages is 0 or negative, returns 0. If no pages
325 * were pinned, returns -errno.
Nick Piggind2bf6be2009-06-16 15:31:39 -0700326 *
327 * get_user_pages_fast provides equivalent functionality to get_user_pages,
328 * operating on current and current->mm, with force=0 and vma=NULL. However
329 * unlike get_user_pages, it must be called without mmap_sem held.
330 *
331 * get_user_pages_fast may take mmap_sem and page table locks, so no
332 * assumptions can be made about lack of locking. get_user_pages_fast is to be
333 * implemented in a way that is advantageous (vs get_user_pages()) when the
334 * user memory area is already faulted in and present in ptes. However if the
335 * pages have to be faulted in, it may turn out to be slightly slower so
336 * callers need to carefully consider what to use. On many architectures,
337 * get_user_pages_fast simply falls back to get_user_pages.
Andy Grover9de100d2009-04-13 14:40:05 -0700338 */
Rusty Russell912985d2008-08-12 17:52:52 -0500339int __attribute__((weak)) get_user_pages_fast(unsigned long start,
340 int nr_pages, int write, struct page **pages)
341{
342 struct mm_struct *mm = current->mm;
343 int ret;
344
345 down_read(&mm->mmap_sem);
346 ret = get_user_pages(current, mm, start, nr_pages,
347 write, 0, pages, NULL);
348 up_read(&mm->mmap_sem);
349
350 return ret;
351}
352EXPORT_SYMBOL_GPL(get_user_pages_fast);
Eduard - Gabriel Munteanuca2b84cb2009-03-23 15:12:24 +0200353
Al Viroeb36c582012-05-30 20:17:35 -0400354unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
355 unsigned long len, unsigned long prot,
356 unsigned long flag, unsigned long pgoff)
357{
358 unsigned long ret;
359 struct mm_struct *mm = current->mm;
Michel Lespinasse41badc12013-02-22 16:32:47 -0800360 unsigned long populate;
Al Viroeb36c582012-05-30 20:17:35 -0400361
362 ret = security_mmap_file(file, prot, flag);
363 if (!ret) {
364 down_write(&mm->mmap_sem);
Michel Lespinassebebeb3d2013-02-22 16:32:37 -0800365 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
366 &populate);
Al Viroeb36c582012-05-30 20:17:35 -0400367 up_write(&mm->mmap_sem);
Michel Lespinasse41badc12013-02-22 16:32:47 -0800368 if (populate)
369 mm_populate(ret, populate);
Al Viroeb36c582012-05-30 20:17:35 -0400370 }
371 return ret;
372}
373
374unsigned long vm_mmap(struct file *file, unsigned long addr,
375 unsigned long len, unsigned long prot,
376 unsigned long flag, unsigned long offset)
377{
378 if (unlikely(offset + PAGE_ALIGN(len) < offset))
379 return -EINVAL;
380 if (unlikely(offset & ~PAGE_MASK))
381 return -EINVAL;
382
383 return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
384}
385EXPORT_SYMBOL(vm_mmap);
386
Shaohua Li98003392013-02-22 16:34:35 -0800387struct address_space *page_mapping(struct page *page)
388{
389 struct address_space *mapping = page->mapping;
390
391 VM_BUG_ON(PageSlab(page));
392#ifdef CONFIG_SWAP
Shaohua Li33806f02013-02-22 16:34:37 -0800393 if (unlikely(PageSwapCache(page))) {
394 swp_entry_t entry;
395
396 entry.val = page_private(page);
397 mapping = swap_address_space(entry);
398 } else
Shaohua Li98003392013-02-22 16:34:35 -0800399#endif
400 if ((unsigned long)mapping & PAGE_MAPPING_ANON)
401 mapping = NULL;
402 return mapping;
403}
404
Eduard - Gabriel Munteanuca2b84cb2009-03-23 15:12:24 +0200405/* Tracepoints definitions. */
Eduard - Gabriel Munteanuca2b84cb2009-03-23 15:12:24 +0200406EXPORT_TRACEPOINT_SYMBOL(kmalloc);
407EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
408EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
409EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
410EXPORT_TRACEPOINT_SYMBOL(kfree);
411EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);