Commit | Line | Data |
---|---|---|
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> |
3c2a0909 | 10 | #include <linux/vmalloc.h> |
96840aa0 | 11 | #include <asm/uaccess.h> |
30992c97 | 12 | |
6038def0 NK |
13 | #include "internal.h" |
14 | ||
a8d154b0 | 15 | #define CREATE_TRACE_POINTS |
ad8d75ff | 16 | #include <trace/events/kmem.h> |
a8d154b0 | 17 | |
30992c97 | 18 | /** |
30992c97 | 19 | * kstrdup - allocate space for and copy an existing string |
30992c97 MM |
20 | * @s: the string to duplicate |
21 | * @gfp: the GFP mask used in the kmalloc() call when allocating memory | |
22 | */ | |
23 | char *kstrdup(const char *s, gfp_t gfp) | |
24 | { | |
25 | size_t len; | |
26 | char *buf; | |
27 | ||
28 | if (!s) | |
29 | return NULL; | |
30 | ||
31 | len = strlen(s) + 1; | |
1d2c8eea | 32 | buf = kmalloc_track_caller(len, gfp); |
30992c97 MM |
33 | if (buf) |
34 | memcpy(buf, s, len); | |
35 | return buf; | |
36 | } | |
37 | EXPORT_SYMBOL(kstrdup); | |
96840aa0 | 38 | |
1e66df3e JF |
39 | /** |
40 | * kstrndup - allocate space for and copy an existing string | |
41 | * @s: the string to duplicate | |
42 | * @max: read at most @max chars from @s | |
43 | * @gfp: the GFP mask used in the kmalloc() call when allocating memory | |
44 | */ | |
45 | char *kstrndup(const char *s, size_t max, gfp_t gfp) | |
46 | { | |
47 | size_t len; | |
48 | char *buf; | |
49 | ||
50 | if (!s) | |
51 | return NULL; | |
52 | ||
53 | len = strnlen(s, max); | |
54 | buf = kmalloc_track_caller(len+1, gfp); | |
55 | if (buf) { | |
56 | memcpy(buf, s, len); | |
57 | buf[len] = '\0'; | |
58 | } | |
59 | return buf; | |
60 | } | |
61 | EXPORT_SYMBOL(kstrndup); | |
62 | ||
1a2f67b4 AD |
63 | /** |
64 | * kmemdup - duplicate region of memory | |
65 | * | |
66 | * @src: memory region to duplicate | |
67 | * @len: memory region length | |
68 | * @gfp: GFP mask to use | |
69 | */ | |
70 | void *kmemdup(const void *src, size_t len, gfp_t gfp) | |
71 | { | |
72 | void *p; | |
73 | ||
1d2c8eea | 74 | p = kmalloc_track_caller(len, gfp); |
1a2f67b4 AD |
75 | if (p) |
76 | memcpy(p, src, len); | |
77 | return p; | |
78 | } | |
79 | EXPORT_SYMBOL(kmemdup); | |
80 | ||
610a77e0 LZ |
81 | /** |
82 | * memdup_user - duplicate memory region from user space | |
83 | * | |
84 | * @src: source address in user space | |
85 | * @len: number of bytes to copy | |
86 | * | |
87 | * Returns an ERR_PTR() on failure. | |
88 | */ | |
89 | void *memdup_user(const void __user *src, size_t len) | |
90 | { | |
91 | void *p; | |
92 | ||
93 | /* | |
94 | * Always use GFP_KERNEL, since copy_from_user() can sleep and | |
95 | * cause pagefault, which makes it pointless to use GFP_NOFS | |
96 | * or GFP_ATOMIC. | |
97 | */ | |
98 | p = kmalloc_track_caller(len, GFP_KERNEL); | |
99 | if (!p) | |
100 | return ERR_PTR(-ENOMEM); | |
101 | ||
102 | if (copy_from_user(p, src, len)) { | |
103 | kfree(p); | |
104 | return ERR_PTR(-EFAULT); | |
105 | } | |
106 | ||
107 | return p; | |
108 | } | |
109 | EXPORT_SYMBOL(memdup_user); | |
110 | ||
e21827aa EG |
111 | static __always_inline void *__do_krealloc(const void *p, size_t new_size, |
112 | gfp_t flags) | |
113 | { | |
114 | void *ret; | |
115 | size_t ks = 0; | |
116 | ||
117 | if (p) | |
118 | ks = ksize(p); | |
119 | ||
120 | if (ks >= new_size) | |
121 | return (void *)p; | |
122 | ||
123 | ret = kmalloc_track_caller(new_size, flags); | |
124 | if (ret && p) | |
125 | memcpy(ret, p, ks); | |
126 | ||
127 | return ret; | |
128 | } | |
129 | ||
ef2ad80c | 130 | /** |
93bc4e89 | 131 | * __krealloc - like krealloc() but don't free @p. |
ef2ad80c CL |
132 | * @p: object to reallocate memory for. |
133 | * @new_size: how many bytes of memory are required. | |
134 | * @flags: the type of memory to allocate. | |
135 | * | |
93bc4e89 PE |
136 | * This function is like krealloc() except it never frees the originally |
137 | * allocated buffer. Use this if you don't want to free the buffer immediately | |
138 | * like, for example, with RCU. | |
ef2ad80c | 139 | */ |
93bc4e89 | 140 | void *__krealloc(const void *p, size_t new_size, gfp_t flags) |
ef2ad80c | 141 | { |
93bc4e89 | 142 | if (unlikely(!new_size)) |
6cb8f913 | 143 | return ZERO_SIZE_PTR; |
ef2ad80c | 144 | |
e21827aa | 145 | return __do_krealloc(p, new_size, flags); |
ef8b4520 | 146 | |
93bc4e89 PE |
147 | } |
148 | EXPORT_SYMBOL(__krealloc); | |
149 | ||
150 | /** | |
151 | * krealloc - reallocate memory. The contents will remain unchanged. | |
152 | * @p: object to reallocate memory for. | |
153 | * @new_size: how many bytes of memory are required. | |
154 | * @flags: the type of memory to allocate. | |
155 | * | |
156 | * The contents of the object pointed to are preserved up to the | |
157 | * lesser of the new and old sizes. If @p is %NULL, krealloc() | |
0db10c8e | 158 | * behaves exactly like kmalloc(). If @new_size is 0 and @p is not a |
93bc4e89 PE |
159 | * %NULL pointer, the object pointed to is freed. |
160 | */ | |
161 | void *krealloc(const void *p, size_t new_size, gfp_t flags) | |
162 | { | |
163 | void *ret; | |
164 | ||
165 | if (unlikely(!new_size)) { | |
ef2ad80c | 166 | kfree(p); |
93bc4e89 | 167 | return ZERO_SIZE_PTR; |
ef2ad80c | 168 | } |
93bc4e89 | 169 | |
e21827aa | 170 | ret = __do_krealloc(p, new_size, flags); |
93bc4e89 PE |
171 | if (ret && p != ret) |
172 | kfree(p); | |
173 | ||
ef2ad80c CL |
174 | return ret; |
175 | } | |
176 | EXPORT_SYMBOL(krealloc); | |
177 | ||
3ef0e5ba JW |
178 | /** |
179 | * kzfree - like kfree but zero memory | |
180 | * @p: object to free memory of | |
181 | * | |
182 | * The memory of the object @p points to is zeroed before freed. | |
183 | * If @p is %NULL, kzfree() does nothing. | |
a234bdc9 PE |
184 | * |
185 | * Note: this function zeroes the whole allocated buffer which can be a good | |
186 | * deal bigger than the requested buffer size passed to kmalloc(). So be | |
187 | * careful when using this function in performance sensitive code. | |
3ef0e5ba JW |
188 | */ |
189 | void kzfree(const void *p) | |
190 | { | |
191 | size_t ks; | |
192 | void *mem = (void *)p; | |
193 | ||
194 | if (unlikely(ZERO_OR_NULL_PTR(mem))) | |
195 | return; | |
196 | ks = ksize(mem); | |
197 | memset(mem, 0, ks); | |
198 | kfree(mem); | |
199 | } | |
200 | EXPORT_SYMBOL(kzfree); | |
201 | ||
96840aa0 DA |
202 | /* |
203 | * strndup_user - duplicate an existing string from user space | |
96840aa0 DA |
204 | * @s: The string to duplicate |
205 | * @n: Maximum number of bytes to copy, including the trailing NUL. | |
206 | */ | |
207 | char *strndup_user(const char __user *s, long n) | |
208 | { | |
209 | char *p; | |
210 | long length; | |
211 | ||
212 | length = strnlen_user(s, n); | |
213 | ||
214 | if (!length) | |
215 | return ERR_PTR(-EFAULT); | |
216 | ||
217 | if (length > n) | |
218 | return ERR_PTR(-EINVAL); | |
219 | ||
90d74045 | 220 | p = memdup_user(s, length); |
96840aa0 | 221 | |
90d74045 JL |
222 | if (IS_ERR(p)) |
223 | return p; | |
96840aa0 DA |
224 | |
225 | p[length - 1] = '\0'; | |
226 | ||
227 | return p; | |
228 | } | |
229 | EXPORT_SYMBOL(strndup_user); | |
16d69265 | 230 | |
6038def0 NK |
231 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, |
232 | struct vm_area_struct *prev, struct rb_node *rb_parent) | |
233 | { | |
234 | struct vm_area_struct *next; | |
235 | ||
236 | vma->vm_prev = prev; | |
237 | if (prev) { | |
238 | next = prev->vm_next; | |
239 | prev->vm_next = vma; | |
240 | } else { | |
241 | mm->mmap = vma; | |
242 | if (rb_parent) | |
243 | next = rb_entry(rb_parent, | |
244 | struct vm_area_struct, vm_rb); | |
245 | else | |
246 | next = NULL; | |
247 | } | |
248 | vma->vm_next = next; | |
249 | if (next) | |
250 | next->vm_prev = vma; | |
251 | } | |
252 | ||
b7643757 SP |
253 | /* Check if the vma is being used as a stack by this task */ |
254 | static int vm_is_stack_for_task(struct task_struct *t, | |
255 | struct vm_area_struct *vma) | |
256 | { | |
257 | return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t)); | |
258 | } | |
259 | ||
260 | /* | |
261 | * Check if the vma is being used as a stack. | |
262 | * If is_group is non-zero, check in the entire thread group or else | |
263 | * just check in the current task. Returns the pid of the task that | |
264 | * the vma is stack for. | |
265 | */ | |
266 | pid_t vm_is_stack(struct task_struct *task, | |
267 | struct vm_area_struct *vma, int in_group) | |
268 | { | |
269 | pid_t ret = 0; | |
270 | ||
271 | if (vm_is_stack_for_task(task, vma)) | |
272 | return task->pid; | |
273 | ||
274 | if (in_group) { | |
275 | struct task_struct *t; | |
b7643757 | 276 | |
35818329 ON |
277 | rcu_read_lock(); |
278 | for_each_thread(task, t) { | |
b7643757 SP |
279 | if (vm_is_stack_for_task(t, vma)) { |
280 | ret = t->pid; | |
281 | goto done; | |
282 | } | |
35818329 | 283 | } |
b7643757 SP |
284 | done: |
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 |
292 | void 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 | */ | |
306 | int __attribute__((weak)) __get_user_pages_fast(unsigned long start, | |
307 | int nr_pages, int write, struct page **pages) | |
308 | { | |
309 | return 0; | |
310 | } | |
311 | EXPORT_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 |
337 | int __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 | } | |
350 | EXPORT_SYMBOL_GPL(get_user_pages_fast); | |
ca2b84cb | 351 | |
eb36c587 AV |
352 | unsigned 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 | ||
372 | unsigned 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 | } | |
383 | EXPORT_SYMBOL(vm_mmap); | |
384 | ||
3c2a0909 S |
385 | void kvfree(const void *addr) |
386 | { | |
387 | if (is_vmalloc_addr(addr)) | |
388 | vfree(addr); | |
389 | else | |
390 | kfree(addr); | |
391 | } | |
392 | EXPORT_SYMBOL(kvfree); | |
393 | ||
9800339b SL |
394 | struct address_space *page_mapping(struct page *page) |
395 | { | |
396 | struct address_space *mapping = page->mapping; | |
397 | ||
398 | VM_BUG_ON(PageSlab(page)); | |
399 | #ifdef CONFIG_SWAP | |
33806f06 SL |
400 | if (unlikely(PageSwapCache(page))) { |
401 | swp_entry_t entry; | |
402 | ||
403 | entry.val = page_private(page); | |
404 | mapping = swap_address_space(entry); | |
405 | } else | |
9800339b SL |
406 | #endif |
407 | if ((unsigned long)mapping & PAGE_MAPPING_ANON) | |
408 | mapping = NULL; | |
409 | return mapping; | |
410 | } | |
411 | ||
3c2a0909 S |
412 | /** |
413 | * get_cmdline() - copy the cmdline value to a buffer. | |
414 | * @task: the task whose cmdline value to copy. | |
415 | * @buffer: the buffer to copy to. | |
416 | * @buflen: the length of the buffer. Larger cmdline values are truncated | |
417 | * to this length. | |
418 | * Returns the size of the cmdline field copied. Note that the copy does | |
419 | * not guarantee an ending NULL byte. | |
420 | */ | |
421 | int get_cmdline(struct task_struct *task, char *buffer, int buflen) | |
422 | { | |
423 | int res = 0; | |
424 | unsigned int len; | |
425 | struct mm_struct *mm = get_task_mm(task); | |
426 | if (!mm) | |
427 | goto out; | |
428 | if (!mm->arg_end) | |
429 | goto out_mm; /* Shh! No looking before we're done */ | |
430 | ||
431 | len = mm->arg_end - mm->arg_start; | |
432 | ||
433 | if (len > buflen) | |
434 | len = buflen; | |
435 | ||
436 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
437 | ||
438 | /* | |
439 | * If the nul at the end of args has been overwritten, then | |
440 | * assume application is using setproctitle(3). | |
441 | */ | |
442 | if (res > 0 && buffer[res-1] != '\0' && len < buflen) { | |
443 | len = strnlen(buffer, res); | |
444 | if (len < res) { | |
445 | res = len; | |
446 | } else { | |
447 | len = mm->env_end - mm->env_start; | |
448 | if (len > buflen - res) | |
449 | len = buflen - res; | |
450 | res += access_process_vm(task, mm->env_start, | |
451 | buffer+res, len, 0); | |
452 | res = strnlen(buffer, res); | |
453 | } | |
454 | } | |
455 | out_mm: | |
456 | mmput(mm); | |
457 | out: | |
458 | return res; | |
459 | } | |
460 | ||
ca2b84cb | 461 | /* Tracepoints definitions. */ |
ca2b84cb EGM |
462 | EXPORT_TRACEPOINT_SYMBOL(kmalloc); |
463 | EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc); | |
464 | EXPORT_TRACEPOINT_SYMBOL(kmalloc_node); | |
465 | EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node); | |
466 | EXPORT_TRACEPOINT_SYMBOL(kfree); | |
467 | EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free); |