Commit | Line | Data |
---|---|---|
16d69265 | 1 | #include <linux/mm.h> |
30992c97 MM |
2 | #include <linux/slab.h> |
3 | #include <linux/string.h> | |
3b32123d | 4 | #include <linux/compiler.h> |
b95f1b31 | 5 | #include <linux/export.h> |
96840aa0 | 6 | #include <linux/err.h> |
3b8f14b4 | 7 | #include <linux/sched.h> |
eb36c587 | 8 | #include <linux/security.h> |
9800339b | 9 | #include <linux/swap.h> |
33806f06 | 10 | #include <linux/swapops.h> |
00619bcc JM |
11 | #include <linux/mman.h> |
12 | #include <linux/hugetlb.h> | |
39f1f78d | 13 | #include <linux/vmalloc.h> |
00619bcc | 14 | |
a4bb1e43 | 15 | #include <asm/sections.h> |
96840aa0 | 16 | #include <asm/uaccess.h> |
30992c97 | 17 | |
6038def0 NK |
18 | #include "internal.h" |
19 | ||
a4bb1e43 AH |
20 | static inline int is_kernel_rodata(unsigned long addr) |
21 | { | |
22 | return addr >= (unsigned long)__start_rodata && | |
23 | addr < (unsigned long)__end_rodata; | |
24 | } | |
25 | ||
26 | /** | |
27 | * kfree_const - conditionally free memory | |
28 | * @x: pointer to the memory | |
29 | * | |
30 | * Function calls kfree only if @x is not in .rodata section. | |
31 | */ | |
32 | void kfree_const(const void *x) | |
33 | { | |
34 | if (!is_kernel_rodata((unsigned long)x)) | |
35 | kfree(x); | |
36 | } | |
37 | EXPORT_SYMBOL(kfree_const); | |
38 | ||
30992c97 | 39 | /** |
30992c97 | 40 | * kstrdup - allocate space for and copy an existing string |
30992c97 MM |
41 | * @s: the string to duplicate |
42 | * @gfp: the GFP mask used in the kmalloc() call when allocating memory | |
43 | */ | |
44 | char *kstrdup(const char *s, gfp_t gfp) | |
45 | { | |
46 | size_t len; | |
47 | char *buf; | |
48 | ||
49 | if (!s) | |
50 | return NULL; | |
51 | ||
52 | len = strlen(s) + 1; | |
1d2c8eea | 53 | buf = kmalloc_track_caller(len, gfp); |
30992c97 MM |
54 | if (buf) |
55 | memcpy(buf, s, len); | |
56 | return buf; | |
57 | } | |
58 | EXPORT_SYMBOL(kstrdup); | |
96840aa0 | 59 | |
a4bb1e43 AH |
60 | /** |
61 | * kstrdup_const - conditionally duplicate an existing const string | |
62 | * @s: the string to duplicate | |
63 | * @gfp: the GFP mask used in the kmalloc() call when allocating memory | |
64 | * | |
65 | * Function returns source string if it is in .rodata section otherwise it | |
66 | * fallbacks to kstrdup. | |
67 | * Strings allocated by kstrdup_const should be freed by kfree_const. | |
68 | */ | |
69 | const char *kstrdup_const(const char *s, gfp_t gfp) | |
70 | { | |
71 | if (is_kernel_rodata((unsigned long)s)) | |
72 | return s; | |
73 | ||
74 | return kstrdup(s, gfp); | |
75 | } | |
76 | EXPORT_SYMBOL(kstrdup_const); | |
77 | ||
1e66df3e JF |
78 | /** |
79 | * kstrndup - allocate space for and copy an existing string | |
80 | * @s: the string to duplicate | |
81 | * @max: read at most @max chars from @s | |
82 | * @gfp: the GFP mask used in the kmalloc() call when allocating memory | |
83 | */ | |
84 | char *kstrndup(const char *s, size_t max, gfp_t gfp) | |
85 | { | |
86 | size_t len; | |
87 | char *buf; | |
88 | ||
89 | if (!s) | |
90 | return NULL; | |
91 | ||
92 | len = strnlen(s, max); | |
93 | buf = kmalloc_track_caller(len+1, gfp); | |
94 | if (buf) { | |
95 | memcpy(buf, s, len); | |
96 | buf[len] = '\0'; | |
97 | } | |
98 | return buf; | |
99 | } | |
100 | EXPORT_SYMBOL(kstrndup); | |
101 | ||
1a2f67b4 AD |
102 | /** |
103 | * kmemdup - duplicate region of memory | |
104 | * | |
105 | * @src: memory region to duplicate | |
106 | * @len: memory region length | |
107 | * @gfp: GFP mask to use | |
108 | */ | |
109 | void *kmemdup(const void *src, size_t len, gfp_t gfp) | |
110 | { | |
111 | void *p; | |
112 | ||
1d2c8eea | 113 | p = kmalloc_track_caller(len, gfp); |
1a2f67b4 AD |
114 | if (p) |
115 | memcpy(p, src, len); | |
116 | return p; | |
117 | } | |
118 | EXPORT_SYMBOL(kmemdup); | |
119 | ||
610a77e0 LZ |
120 | /** |
121 | * memdup_user - duplicate memory region from user space | |
122 | * | |
123 | * @src: source address in user space | |
124 | * @len: number of bytes to copy | |
125 | * | |
126 | * Returns an ERR_PTR() on failure. | |
127 | */ | |
128 | void *memdup_user(const void __user *src, size_t len) | |
129 | { | |
130 | void *p; | |
131 | ||
132 | /* | |
133 | * Always use GFP_KERNEL, since copy_from_user() can sleep and | |
134 | * cause pagefault, which makes it pointless to use GFP_NOFS | |
135 | * or GFP_ATOMIC. | |
136 | */ | |
137 | p = kmalloc_track_caller(len, GFP_KERNEL); | |
138 | if (!p) | |
139 | return ERR_PTR(-ENOMEM); | |
140 | ||
141 | if (copy_from_user(p, src, len)) { | |
142 | kfree(p); | |
143 | return ERR_PTR(-EFAULT); | |
144 | } | |
145 | ||
146 | return p; | |
147 | } | |
148 | EXPORT_SYMBOL(memdup_user); | |
149 | ||
96840aa0 DA |
150 | /* |
151 | * strndup_user - duplicate an existing string from user space | |
96840aa0 DA |
152 | * @s: The string to duplicate |
153 | * @n: Maximum number of bytes to copy, including the trailing NUL. | |
154 | */ | |
155 | char *strndup_user(const char __user *s, long n) | |
156 | { | |
157 | char *p; | |
158 | long length; | |
159 | ||
160 | length = strnlen_user(s, n); | |
161 | ||
162 | if (!length) | |
163 | return ERR_PTR(-EFAULT); | |
164 | ||
165 | if (length > n) | |
166 | return ERR_PTR(-EINVAL); | |
167 | ||
90d74045 | 168 | p = memdup_user(s, length); |
96840aa0 | 169 | |
90d74045 JL |
170 | if (IS_ERR(p)) |
171 | return p; | |
96840aa0 DA |
172 | |
173 | p[length - 1] = '\0'; | |
174 | ||
175 | return p; | |
176 | } | |
177 | EXPORT_SYMBOL(strndup_user); | |
16d69265 | 178 | |
6038def0 NK |
179 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, |
180 | struct vm_area_struct *prev, struct rb_node *rb_parent) | |
181 | { | |
182 | struct vm_area_struct *next; | |
183 | ||
184 | vma->vm_prev = prev; | |
185 | if (prev) { | |
186 | next = prev->vm_next; | |
187 | prev->vm_next = vma; | |
188 | } else { | |
189 | mm->mmap = vma; | |
190 | if (rb_parent) | |
191 | next = rb_entry(rb_parent, | |
192 | struct vm_area_struct, vm_rb); | |
193 | else | |
194 | next = NULL; | |
195 | } | |
196 | vma->vm_next = next; | |
197 | if (next) | |
198 | next->vm_prev = vma; | |
199 | } | |
200 | ||
b7643757 | 201 | /* Check if the vma is being used as a stack by this task */ |
f3de8fbe | 202 | int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t) |
b7643757 SP |
203 | { |
204 | return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t)); | |
205 | } | |
206 | ||
efc1a3b1 | 207 | #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT) |
16d69265 AM |
208 | void arch_pick_mmap_layout(struct mm_struct *mm) |
209 | { | |
210 | mm->mmap_base = TASK_UNMAPPED_BASE; | |
211 | mm->get_unmapped_area = arch_get_unmapped_area; | |
16d69265 AM |
212 | } |
213 | #endif | |
912985dc | 214 | |
45888a0c XG |
215 | /* |
216 | * Like get_user_pages_fast() except its IRQ-safe in that it won't fall | |
217 | * back to the regular GUP. | |
25985edc | 218 | * If the architecture not support this function, simply return with no |
45888a0c XG |
219 | * page pinned |
220 | */ | |
3b32123d | 221 | int __weak __get_user_pages_fast(unsigned long start, |
45888a0c XG |
222 | int nr_pages, int write, struct page **pages) |
223 | { | |
224 | return 0; | |
225 | } | |
226 | EXPORT_SYMBOL_GPL(__get_user_pages_fast); | |
227 | ||
9de100d0 AG |
228 | /** |
229 | * get_user_pages_fast() - pin user pages in memory | |
230 | * @start: starting user address | |
231 | * @nr_pages: number of pages from start to pin | |
232 | * @write: whether pages will be written to | |
233 | * @pages: array that receives pointers to the pages pinned. | |
234 | * Should be at least nr_pages long. | |
235 | * | |
9de100d0 AG |
236 | * Returns number of pages pinned. This may be fewer than the number |
237 | * requested. If nr_pages is 0 or negative, returns 0. If no pages | |
238 | * were pinned, returns -errno. | |
d2bf6be8 NP |
239 | * |
240 | * get_user_pages_fast provides equivalent functionality to get_user_pages, | |
241 | * operating on current and current->mm, with force=0 and vma=NULL. However | |
242 | * unlike get_user_pages, it must be called without mmap_sem held. | |
243 | * | |
244 | * get_user_pages_fast may take mmap_sem and page table locks, so no | |
245 | * assumptions can be made about lack of locking. get_user_pages_fast is to be | |
246 | * implemented in a way that is advantageous (vs get_user_pages()) when the | |
247 | * user memory area is already faulted in and present in ptes. However if the | |
248 | * pages have to be faulted in, it may turn out to be slightly slower so | |
249 | * callers need to carefully consider what to use. On many architectures, | |
250 | * get_user_pages_fast simply falls back to get_user_pages. | |
9de100d0 | 251 | */ |
3b32123d | 252 | int __weak get_user_pages_fast(unsigned long start, |
912985dc RR |
253 | int nr_pages, int write, struct page **pages) |
254 | { | |
255 | struct mm_struct *mm = current->mm; | |
a7b78075 AA |
256 | return get_user_pages_unlocked(current, mm, start, nr_pages, |
257 | write, 0, pages); | |
912985dc RR |
258 | } |
259 | EXPORT_SYMBOL_GPL(get_user_pages_fast); | |
ca2b84cb | 260 | |
eb36c587 AV |
261 | unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr, |
262 | unsigned long len, unsigned long prot, | |
263 | unsigned long flag, unsigned long pgoff) | |
264 | { | |
265 | unsigned long ret; | |
266 | struct mm_struct *mm = current->mm; | |
41badc15 | 267 | unsigned long populate; |
eb36c587 AV |
268 | |
269 | ret = security_mmap_file(file, prot, flag); | |
270 | if (!ret) { | |
271 | down_write(&mm->mmap_sem); | |
bebeb3d6 ML |
272 | ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff, |
273 | &populate); | |
eb36c587 | 274 | up_write(&mm->mmap_sem); |
41badc15 ML |
275 | if (populate) |
276 | mm_populate(ret, populate); | |
eb36c587 AV |
277 | } |
278 | return ret; | |
279 | } | |
280 | ||
281 | unsigned long vm_mmap(struct file *file, unsigned long addr, | |
282 | unsigned long len, unsigned long prot, | |
283 | unsigned long flag, unsigned long offset) | |
284 | { | |
285 | if (unlikely(offset + PAGE_ALIGN(len) < offset)) | |
286 | return -EINVAL; | |
ea53cde0 | 287 | if (unlikely(offset_in_page(offset))) |
eb36c587 AV |
288 | return -EINVAL; |
289 | ||
290 | return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); | |
291 | } | |
292 | EXPORT_SYMBOL(vm_mmap); | |
293 | ||
39f1f78d AV |
294 | void kvfree(const void *addr) |
295 | { | |
296 | if (is_vmalloc_addr(addr)) | |
297 | vfree(addr); | |
298 | else | |
299 | kfree(addr); | |
300 | } | |
301 | EXPORT_SYMBOL(kvfree); | |
302 | ||
e39155ea KS |
303 | static inline void *__page_rmapping(struct page *page) |
304 | { | |
305 | unsigned long mapping; | |
306 | ||
307 | mapping = (unsigned long)page->mapping; | |
308 | mapping &= ~PAGE_MAPPING_FLAGS; | |
309 | ||
310 | return (void *)mapping; | |
311 | } | |
312 | ||
313 | /* Neutral page->mapping pointer to address_space or anon_vma or other */ | |
314 | void *page_rmapping(struct page *page) | |
315 | { | |
316 | page = compound_head(page); | |
317 | return __page_rmapping(page); | |
318 | } | |
319 | ||
320 | struct anon_vma *page_anon_vma(struct page *page) | |
321 | { | |
322 | unsigned long mapping; | |
323 | ||
324 | page = compound_head(page); | |
325 | mapping = (unsigned long)page->mapping; | |
326 | if ((mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON) | |
327 | return NULL; | |
328 | return __page_rmapping(page); | |
329 | } | |
330 | ||
9800339b SL |
331 | struct address_space *page_mapping(struct page *page) |
332 | { | |
e39155ea | 333 | unsigned long mapping; |
9800339b | 334 | |
03e5ac2f MP |
335 | /* This happens if someone calls flush_dcache_page on slab page */ |
336 | if (unlikely(PageSlab(page))) | |
337 | return NULL; | |
338 | ||
33806f06 SL |
339 | if (unlikely(PageSwapCache(page))) { |
340 | swp_entry_t entry; | |
341 | ||
342 | entry.val = page_private(page); | |
e39155ea KS |
343 | return swap_address_space(entry); |
344 | } | |
345 | ||
346 | mapping = (unsigned long)page->mapping; | |
347 | if (mapping & PAGE_MAPPING_FLAGS) | |
348 | return NULL; | |
349 | return page->mapping; | |
9800339b | 350 | } |
13f00235 | 351 | EXPORT_SYMBOL(page_mapping); |
9800339b | 352 | |
49f0ce5f JM |
353 | int overcommit_ratio_handler(struct ctl_table *table, int write, |
354 | void __user *buffer, size_t *lenp, | |
355 | loff_t *ppos) | |
356 | { | |
357 | int ret; | |
358 | ||
359 | ret = proc_dointvec(table, write, buffer, lenp, ppos); | |
360 | if (ret == 0 && write) | |
361 | sysctl_overcommit_kbytes = 0; | |
362 | return ret; | |
363 | } | |
364 | ||
365 | int overcommit_kbytes_handler(struct ctl_table *table, int write, | |
366 | void __user *buffer, size_t *lenp, | |
367 | loff_t *ppos) | |
368 | { | |
369 | int ret; | |
370 | ||
371 | ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos); | |
372 | if (ret == 0 && write) | |
373 | sysctl_overcommit_ratio = 0; | |
374 | return ret; | |
375 | } | |
376 | ||
00619bcc JM |
377 | /* |
378 | * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used | |
379 | */ | |
380 | unsigned long vm_commit_limit(void) | |
381 | { | |
49f0ce5f JM |
382 | unsigned long allowed; |
383 | ||
384 | if (sysctl_overcommit_kbytes) | |
385 | allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10); | |
386 | else | |
387 | allowed = ((totalram_pages - hugetlb_total_pages()) | |
388 | * sysctl_overcommit_ratio / 100); | |
389 | allowed += total_swap_pages; | |
390 | ||
391 | return allowed; | |
00619bcc JM |
392 | } |
393 | ||
a9090253 WR |
394 | /** |
395 | * get_cmdline() - copy the cmdline value to a buffer. | |
396 | * @task: the task whose cmdline value to copy. | |
397 | * @buffer: the buffer to copy to. | |
398 | * @buflen: the length of the buffer. Larger cmdline values are truncated | |
399 | * to this length. | |
400 | * Returns the size of the cmdline field copied. Note that the copy does | |
401 | * not guarantee an ending NULL byte. | |
402 | */ | |
403 | int get_cmdline(struct task_struct *task, char *buffer, int buflen) | |
404 | { | |
405 | int res = 0; | |
406 | unsigned int len; | |
407 | struct mm_struct *mm = get_task_mm(task); | |
408 | if (!mm) | |
409 | goto out; | |
410 | if (!mm->arg_end) | |
411 | goto out_mm; /* Shh! No looking before we're done */ | |
412 | ||
413 | len = mm->arg_end - mm->arg_start; | |
414 | ||
415 | if (len > buflen) | |
416 | len = buflen; | |
417 | ||
418 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); | |
419 | ||
420 | /* | |
421 | * If the nul at the end of args has been overwritten, then | |
422 | * assume application is using setproctitle(3). | |
423 | */ | |
424 | if (res > 0 && buffer[res-1] != '\0' && len < buflen) { | |
425 | len = strnlen(buffer, res); | |
426 | if (len < res) { | |
427 | res = len; | |
428 | } else { | |
429 | len = mm->env_end - mm->env_start; | |
430 | if (len > buflen - res) | |
431 | len = buflen - res; | |
432 | res += access_process_vm(task, mm->env_start, | |
433 | buffer+res, len, 0); | |
434 | res = strnlen(buffer, res); | |
435 | } | |
436 | } | |
437 | out_mm: | |
438 | mmput(mm); | |
439 | out: | |
440 | return res; | |
441 | } |