Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/vapier...
[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>
4#include <linux/module.h>
96840aa0 5#include <linux/err.h>
3b8f14b4 6#include <linux/sched.h>
f8b72560
AV
7#include <linux/hugetlb.h>
8#include <linux/syscalls.h>
9#include <linux/mman.h>
10#include <linux/file.h>
96840aa0 11#include <asm/uaccess.h>
30992c97 12
a8d154b0 13#define CREATE_TRACE_POINTS
ad8d75ff 14#include <trace/events/kmem.h>
a8d154b0 15
30992c97 16/**
30992c97 17 * kstrdup - allocate space for and copy an existing string
30992c97
MM
18 * @s: the string to duplicate
19 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
20 */
21char *kstrdup(const char *s, gfp_t gfp)
22{
23 size_t len;
24 char *buf;
25
26 if (!s)
27 return NULL;
28
29 len = strlen(s) + 1;
1d2c8eea 30 buf = kmalloc_track_caller(len, gfp);
30992c97
MM
31 if (buf)
32 memcpy(buf, s, len);
33 return buf;
34}
35EXPORT_SYMBOL(kstrdup);
96840aa0 36
1e66df3e
JF
37/**
38 * kstrndup - allocate space for and copy an existing string
39 * @s: the string to duplicate
40 * @max: read at most @max chars from @s
41 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
42 */
43char *kstrndup(const char *s, size_t max, gfp_t gfp)
44{
45 size_t len;
46 char *buf;
47
48 if (!s)
49 return NULL;
50
51 len = strnlen(s, max);
52 buf = kmalloc_track_caller(len+1, gfp);
53 if (buf) {
54 memcpy(buf, s, len);
55 buf[len] = '\0';
56 }
57 return buf;
58}
59EXPORT_SYMBOL(kstrndup);
60
1a2f67b4
AD
61/**
62 * kmemdup - duplicate region of memory
63 *
64 * @src: memory region to duplicate
65 * @len: memory region length
66 * @gfp: GFP mask to use
67 */
68void *kmemdup(const void *src, size_t len, gfp_t gfp)
69{
70 void *p;
71
1d2c8eea 72 p = kmalloc_track_caller(len, gfp);
1a2f67b4
AD
73 if (p)
74 memcpy(p, src, len);
75 return p;
76}
77EXPORT_SYMBOL(kmemdup);
78
610a77e0
LZ
79/**
80 * memdup_user - duplicate memory region from user space
81 *
82 * @src: source address in user space
83 * @len: number of bytes to copy
84 *
85 * Returns an ERR_PTR() on failure.
86 */
87void *memdup_user(const void __user *src, size_t len)
88{
89 void *p;
90
91 /*
92 * Always use GFP_KERNEL, since copy_from_user() can sleep and
93 * cause pagefault, which makes it pointless to use GFP_NOFS
94 * or GFP_ATOMIC.
95 */
96 p = kmalloc_track_caller(len, GFP_KERNEL);
97 if (!p)
98 return ERR_PTR(-ENOMEM);
99
100 if (copy_from_user(p, src, len)) {
101 kfree(p);
102 return ERR_PTR(-EFAULT);
103 }
104
105 return p;
106}
107EXPORT_SYMBOL(memdup_user);
108
ef2ad80c 109/**
93bc4e89 110 * __krealloc - like krealloc() but don't free @p.
ef2ad80c
CL
111 * @p: object to reallocate memory for.
112 * @new_size: how many bytes of memory are required.
113 * @flags: the type of memory to allocate.
114 *
93bc4e89
PE
115 * This function is like krealloc() except it never frees the originally
116 * allocated buffer. Use this if you don't want to free the buffer immediately
117 * like, for example, with RCU.
ef2ad80c 118 */
93bc4e89 119void *__krealloc(const void *p, size_t new_size, gfp_t flags)
ef2ad80c
CL
120{
121 void *ret;
ef8b4520 122 size_t ks = 0;
ef2ad80c 123
93bc4e89 124 if (unlikely(!new_size))
6cb8f913 125 return ZERO_SIZE_PTR;
ef2ad80c 126
ef8b4520
CL
127 if (p)
128 ks = ksize(p);
129
ef2ad80c
CL
130 if (ks >= new_size)
131 return (void *)p;
132
133 ret = kmalloc_track_caller(new_size, flags);
93bc4e89 134 if (ret && p)
be21f0ab 135 memcpy(ret, p, ks);
93bc4e89
PE
136
137 return ret;
138}
139EXPORT_SYMBOL(__krealloc);
140
141/**
142 * krealloc - reallocate memory. The contents will remain unchanged.
143 * @p: object to reallocate memory for.
144 * @new_size: how many bytes of memory are required.
145 * @flags: the type of memory to allocate.
146 *
147 * The contents of the object pointed to are preserved up to the
148 * lesser of the new and old sizes. If @p is %NULL, krealloc()
149 * behaves exactly like kmalloc(). If @size is 0 and @p is not a
150 * %NULL pointer, the object pointed to is freed.
151 */
152void *krealloc(const void *p, size_t new_size, gfp_t flags)
153{
154 void *ret;
155
156 if (unlikely(!new_size)) {
ef2ad80c 157 kfree(p);
93bc4e89 158 return ZERO_SIZE_PTR;
ef2ad80c 159 }
93bc4e89
PE
160
161 ret = __krealloc(p, new_size, flags);
162 if (ret && p != ret)
163 kfree(p);
164
ef2ad80c
CL
165 return ret;
166}
167EXPORT_SYMBOL(krealloc);
168
3ef0e5ba
JW
169/**
170 * kzfree - like kfree but zero memory
171 * @p: object to free memory of
172 *
173 * The memory of the object @p points to is zeroed before freed.
174 * If @p is %NULL, kzfree() does nothing.
a234bdc9
PE
175 *
176 * Note: this function zeroes the whole allocated buffer which can be a good
177 * deal bigger than the requested buffer size passed to kmalloc(). So be
178 * careful when using this function in performance sensitive code.
3ef0e5ba
JW
179 */
180void kzfree(const void *p)
181{
182 size_t ks;
183 void *mem = (void *)p;
184
185 if (unlikely(ZERO_OR_NULL_PTR(mem)))
186 return;
187 ks = ksize(mem);
188 memset(mem, 0, ks);
189 kfree(mem);
190}
191EXPORT_SYMBOL(kzfree);
192
96840aa0
DA
193/*
194 * strndup_user - duplicate an existing string from user space
96840aa0
DA
195 * @s: The string to duplicate
196 * @n: Maximum number of bytes to copy, including the trailing NUL.
197 */
198char *strndup_user(const char __user *s, long n)
199{
200 char *p;
201 long length;
202
203 length = strnlen_user(s, n);
204
205 if (!length)
206 return ERR_PTR(-EFAULT);
207
208 if (length > n)
209 return ERR_PTR(-EINVAL);
210
211 p = kmalloc(length, GFP_KERNEL);
212
213 if (!p)
214 return ERR_PTR(-ENOMEM);
215
216 if (copy_from_user(p, s, length)) {
217 kfree(p);
218 return ERR_PTR(-EFAULT);
219 }
220
221 p[length - 1] = '\0';
222
223 return p;
224}
225EXPORT_SYMBOL(strndup_user);
16d69265
AM
226
227#ifndef HAVE_ARCH_PICK_MMAP_LAYOUT
228void arch_pick_mmap_layout(struct mm_struct *mm)
229{
230 mm->mmap_base = TASK_UNMAPPED_BASE;
231 mm->get_unmapped_area = arch_get_unmapped_area;
232 mm->unmap_area = arch_unmap_area;
233}
234#endif
912985dc 235
9de100d0
AG
236/**
237 * get_user_pages_fast() - pin user pages in memory
238 * @start: starting user address
239 * @nr_pages: number of pages from start to pin
240 * @write: whether pages will be written to
241 * @pages: array that receives pointers to the pages pinned.
242 * Should be at least nr_pages long.
243 *
9de100d0
AG
244 * Returns number of pages pinned. This may be fewer than the number
245 * requested. If nr_pages is 0 or negative, returns 0. If no pages
246 * were pinned, returns -errno.
d2bf6be8
NP
247 *
248 * get_user_pages_fast provides equivalent functionality to get_user_pages,
249 * operating on current and current->mm, with force=0 and vma=NULL. However
250 * unlike get_user_pages, it must be called without mmap_sem held.
251 *
252 * get_user_pages_fast may take mmap_sem and page table locks, so no
253 * assumptions can be made about lack of locking. get_user_pages_fast is to be
254 * implemented in a way that is advantageous (vs get_user_pages()) when the
255 * user memory area is already faulted in and present in ptes. However if the
256 * pages have to be faulted in, it may turn out to be slightly slower so
257 * callers need to carefully consider what to use. On many architectures,
258 * get_user_pages_fast simply falls back to get_user_pages.
9de100d0 259 */
912985dc
RR
260int __attribute__((weak)) get_user_pages_fast(unsigned long start,
261 int nr_pages, int write, struct page **pages)
262{
263 struct mm_struct *mm = current->mm;
264 int ret;
265
266 down_read(&mm->mmap_sem);
267 ret = get_user_pages(current, mm, start, nr_pages,
268 write, 0, pages, NULL);
269 up_read(&mm->mmap_sem);
270
271 return ret;
272}
273EXPORT_SYMBOL_GPL(get_user_pages_fast);
ca2b84cb 274
f8b72560
AV
275SYSCALL_DEFINE6(mmap_pgoff, unsigned long, addr, unsigned long, len,
276 unsigned long, prot, unsigned long, flags,
277 unsigned long, fd, unsigned long, pgoff)
278{
279 struct file * file = NULL;
280 unsigned long retval = -EBADF;
281
282 if (!(flags & MAP_ANONYMOUS)) {
8c7b49b3
AV
283 if (unlikely(flags & MAP_HUGETLB))
284 return -EINVAL;
f8b72560
AV
285 file = fget(fd);
286 if (!file)
287 goto out;
8c7b49b3
AV
288 } else if (flags & MAP_HUGETLB) {
289 struct user_struct *user = NULL;
290 /*
291 * VM_NORESERVE is used because the reservations will be
292 * taken when vm_ops->mmap() is called
293 * A dummy user value is used because we are not locking
294 * memory so no accounting is necessary
295 */
296 len = ALIGN(len, huge_page_size(&default_hstate));
297 file = hugetlb_file_setup(HUGETLB_ANON_FILE, len, VM_NORESERVE,
298 &user, HUGETLB_ANONHUGE_INODE);
299 if (IS_ERR(file))
300 return PTR_ERR(file);
f8b72560
AV
301 }
302
303 flags &= ~(MAP_EXECUTABLE | MAP_DENYWRITE);
304
305 down_write(&current->mm->mmap_sem);
306 retval = do_mmap_pgoff(file, addr, len, prot, flags, pgoff);
307 up_write(&current->mm->mmap_sem);
308
309 if (file)
310 fput(file);
311out:
312 return retval;
313}
314
ca2b84cb 315/* Tracepoints definitions. */
ca2b84cb
EGM
316EXPORT_TRACEPOINT_SYMBOL(kmalloc);
317EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
318EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
319EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
320EXPORT_TRACEPOINT_SYMBOL(kfree);
321EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);