blob: 9af1c12b310c7f092f030fc1a4303c287724b0a6 [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>
Gideon Israel Dsouza3b321232014-04-07 15:37:26 -07004#include <linux/compiler.h>
Paul Gortmakerb95f1b312011-10-16 02:01:52 -04005#include <linux/export.h>
Davi Arnaut96840aa2006-03-24 03:18:42 -08006#include <linux/err.h>
Adrian Bunk3b8f14b2008-07-26 15:22:28 -07007#include <linux/sched.h>
Al Viroeb36c582012-05-30 20:17:35 -04008#include <linux/security.h>
Shaohua Li98003392013-02-22 16:34:35 -08009#include <linux/swap.h>
Shaohua Li33806f02013-02-22 16:34:37 -080010#include <linux/swapops.h>
Jerome Marchand00619bc2013-11-12 15:08:31 -080011#include <linux/mman.h>
12#include <linux/hugetlb.h>
Al Viro39f1f782014-05-06 14:02:53 -040013#include <linux/vmalloc.h>
Jerome Marchand00619bc2013-11-12 15:08:31 -080014
Andrzej Hajdaa4bb1e42015-02-13 14:36:24 -080015#include <asm/sections.h>
Davi Arnaut96840aa2006-03-24 03:18:42 -080016#include <asm/uaccess.h>
Matt Mackall30992c92006-01-08 01:01:43 -080017
Namhyung Kim6038def2011-05-24 17:11:22 -070018#include "internal.h"
19
Andrzej Hajdaa4bb1e42015-02-13 14:36:24 -080020static 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 */
32void kfree_const(const void *x)
33{
34 if (!is_kernel_rodata((unsigned long)x))
35 kfree(x);
36}
37EXPORT_SYMBOL(kfree_const);
38
Matt Mackall30992c92006-01-08 01:01:43 -080039/**
Matt Mackall30992c92006-01-08 01:01:43 -080040 * kstrdup - allocate space for and copy an existing string
Matt Mackall30992c92006-01-08 01:01:43 -080041 * @s: the string to duplicate
42 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
43 */
44char *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;
Christoph Hellwig1d2c8ee2006-10-04 02:15:25 -070053 buf = kmalloc_track_caller(len, gfp);
Matt Mackall30992c92006-01-08 01:01:43 -080054 if (buf)
55 memcpy(buf, s, len);
56 return buf;
57}
58EXPORT_SYMBOL(kstrdup);
Davi Arnaut96840aa2006-03-24 03:18:42 -080059
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -070060/**
Andrzej Hajdaa4bb1e42015-02-13 14:36:24 -080061 * 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 */
69const 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}
76EXPORT_SYMBOL(kstrdup_const);
77
78/**
Jeremy Fitzhardinge1e66df32007-07-17 18:37:02 -070079 * 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 */
84char *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}
100EXPORT_SYMBOL(kstrndup);
101
102/**
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -0700103 * kmemdup - duplicate region of memory
104 *
105 * @src: memory region to duplicate
106 * @len: memory region length
107 * @gfp: GFP mask to use
108 */
109void *kmemdup(const void *src, size_t len, gfp_t gfp)
110{
111 void *p;
112
Christoph Hellwig1d2c8ee2006-10-04 02:15:25 -0700113 p = kmalloc_track_caller(len, gfp);
Alexey Dobriyan1a2f67b2006-09-30 23:27:20 -0700114 if (p)
115 memcpy(p, src, len);
116 return p;
117}
118EXPORT_SYMBOL(kmemdup);
119
Christoph Lameteref2ad802007-07-17 04:03:21 -0700120/**
Li Zefan610a77e2009-03-31 15:23:16 -0700121 * 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 */
128void *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}
148EXPORT_SYMBOL(memdup_user);
149
Davi Arnaut96840aa2006-03-24 03:18:42 -0800150/*
151 * strndup_user - duplicate an existing string from user space
Davi Arnaut96840aa2006-03-24 03:18:42 -0800152 * @s: The string to duplicate
153 * @n: Maximum number of bytes to copy, including the trailing NUL.
154 */
155char *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
Julia Lawall90d74042010-08-09 17:18:26 -0700168 p = memdup_user(s, length);
Davi Arnaut96840aa2006-03-24 03:18:42 -0800169
Julia Lawall90d74042010-08-09 17:18:26 -0700170 if (IS_ERR(p))
171 return p;
Davi Arnaut96840aa2006-03-24 03:18:42 -0800172
173 p[length - 1] = '\0';
174
175 return p;
176}
177EXPORT_SYMBOL(strndup_user);
Andrew Morton16d69262008-07-25 19:44:36 -0700178
Namhyung Kim6038def2011-05-24 17:11:22 -0700179void __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
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700201/* Check if the vma is being used as a stack by this task */
202static int vm_is_stack_for_task(struct task_struct *t,
203 struct vm_area_struct *vma)
204{
205 return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
206}
207
208/*
209 * Check if the vma is being used as a stack.
210 * If is_group is non-zero, check in the entire thread group or else
Oleg Nesterov58cb6542014-10-09 15:25:54 -0700211 * just check in the current task. Returns the task_struct of the task
212 * that the vma is stack for. Must be called under rcu_read_lock().
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700213 */
Oleg Nesterov58cb6542014-10-09 15:25:54 -0700214struct task_struct *task_of_stack(struct task_struct *task,
215 struct vm_area_struct *vma, bool in_group)
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700216{
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700217 if (vm_is_stack_for_task(task, vma))
Oleg Nesterov58cb6542014-10-09 15:25:54 -0700218 return task;
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700219
220 if (in_group) {
221 struct task_struct *t;
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700222
Oleg Nesterov4449a512014-08-08 14:19:17 -0700223 for_each_thread(task, t) {
Oleg Nesterov58cb6542014-10-09 15:25:54 -0700224 if (vm_is_stack_for_task(t, vma))
225 return t;
Oleg Nesterov4449a512014-08-08 14:19:17 -0700226 }
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700227 }
228
Oleg Nesterov58cb6542014-10-09 15:25:54 -0700229 return NULL;
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -0700230}
231
David Howellsefc1a3b2010-01-15 17:01:35 -0800232#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
Andrew Morton16d69262008-07-25 19:44:36 -0700233void arch_pick_mmap_layout(struct mm_struct *mm)
234{
235 mm->mmap_base = TASK_UNMAPPED_BASE;
236 mm->get_unmapped_area = arch_get_unmapped_area;
Andrew Morton16d69262008-07-25 19:44:36 -0700237}
238#endif
Rusty Russell912985d2008-08-12 17:52:52 -0500239
Xiao Guangrong45888a02010-08-22 19:08:57 +0800240/*
241 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
242 * back to the regular GUP.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300243 * If the architecture not support this function, simply return with no
Xiao Guangrong45888a02010-08-22 19:08:57 +0800244 * page pinned
245 */
Gideon Israel Dsouza3b321232014-04-07 15:37:26 -0700246int __weak __get_user_pages_fast(unsigned long start,
Xiao Guangrong45888a02010-08-22 19:08:57 +0800247 int nr_pages, int write, struct page **pages)
248{
249 return 0;
250}
251EXPORT_SYMBOL_GPL(__get_user_pages_fast);
252
Andy Grover9de100d2009-04-13 14:40:05 -0700253/**
254 * get_user_pages_fast() - pin user pages in memory
255 * @start: starting user address
256 * @nr_pages: number of pages from start to pin
257 * @write: whether pages will be written to
258 * @pages: array that receives pointers to the pages pinned.
259 * Should be at least nr_pages long.
260 *
Andy Grover9de100d2009-04-13 14:40:05 -0700261 * Returns number of pages pinned. This may be fewer than the number
262 * requested. If nr_pages is 0 or negative, returns 0. If no pages
263 * were pinned, returns -errno.
Nick Piggind2bf6be2009-06-16 15:31:39 -0700264 *
265 * get_user_pages_fast provides equivalent functionality to get_user_pages,
266 * operating on current and current->mm, with force=0 and vma=NULL. However
267 * unlike get_user_pages, it must be called without mmap_sem held.
268 *
269 * get_user_pages_fast may take mmap_sem and page table locks, so no
270 * assumptions can be made about lack of locking. get_user_pages_fast is to be
271 * implemented in a way that is advantageous (vs get_user_pages()) when the
272 * user memory area is already faulted in and present in ptes. However if the
273 * pages have to be faulted in, it may turn out to be slightly slower so
274 * callers need to carefully consider what to use. On many architectures,
275 * get_user_pages_fast simply falls back to get_user_pages.
Andy Grover9de100d2009-04-13 14:40:05 -0700276 */
Gideon Israel Dsouza3b321232014-04-07 15:37:26 -0700277int __weak get_user_pages_fast(unsigned long start,
Rusty Russell912985d2008-08-12 17:52:52 -0500278 int nr_pages, int write, struct page **pages)
279{
280 struct mm_struct *mm = current->mm;
Andrea Arcangelia7b78072015-02-11 15:27:23 -0800281 return get_user_pages_unlocked(current, mm, start, nr_pages,
282 write, 0, pages);
Rusty Russell912985d2008-08-12 17:52:52 -0500283}
284EXPORT_SYMBOL_GPL(get_user_pages_fast);
Eduard - Gabriel Munteanuca2b84cb2009-03-23 15:12:24 +0200285
Al Viroeb36c582012-05-30 20:17:35 -0400286unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
287 unsigned long len, unsigned long prot,
288 unsigned long flag, unsigned long pgoff)
289{
290 unsigned long ret;
291 struct mm_struct *mm = current->mm;
Michel Lespinasse41badc12013-02-22 16:32:47 -0800292 unsigned long populate;
Al Viroeb36c582012-05-30 20:17:35 -0400293
294 ret = security_mmap_file(file, prot, flag);
295 if (!ret) {
296 down_write(&mm->mmap_sem);
Michel Lespinassebebeb3d2013-02-22 16:32:37 -0800297 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
298 &populate);
Al Viroeb36c582012-05-30 20:17:35 -0400299 up_write(&mm->mmap_sem);
Michel Lespinasse41badc12013-02-22 16:32:47 -0800300 if (populate)
301 mm_populate(ret, populate);
Al Viroeb36c582012-05-30 20:17:35 -0400302 }
303 return ret;
304}
305
306unsigned long vm_mmap(struct file *file, unsigned long addr,
307 unsigned long len, unsigned long prot,
308 unsigned long flag, unsigned long offset)
309{
310 if (unlikely(offset + PAGE_ALIGN(len) < offset))
311 return -EINVAL;
Alexander Kuleshovea53cde2015-11-05 18:46:46 -0800312 if (unlikely(offset_in_page(offset)))
Al Viroeb36c582012-05-30 20:17:35 -0400313 return -EINVAL;
314
315 return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
316}
317EXPORT_SYMBOL(vm_mmap);
318
Al Viro39f1f782014-05-06 14:02:53 -0400319void kvfree(const void *addr)
320{
321 if (is_vmalloc_addr(addr))
322 vfree(addr);
323 else
324 kfree(addr);
325}
326EXPORT_SYMBOL(kvfree);
327
Kirill A. Shutemove39155e2015-04-15 16:14:53 -0700328static inline void *__page_rmapping(struct page *page)
329{
330 unsigned long mapping;
331
332 mapping = (unsigned long)page->mapping;
333 mapping &= ~PAGE_MAPPING_FLAGS;
334
335 return (void *)mapping;
336}
337
338/* Neutral page->mapping pointer to address_space or anon_vma or other */
339void *page_rmapping(struct page *page)
340{
341 page = compound_head(page);
342 return __page_rmapping(page);
343}
344
345struct anon_vma *page_anon_vma(struct page *page)
346{
347 unsigned long mapping;
348
349 page = compound_head(page);
350 mapping = (unsigned long)page->mapping;
351 if ((mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
352 return NULL;
353 return __page_rmapping(page);
354}
355
Shaohua Li98003392013-02-22 16:34:35 -0800356struct address_space *page_mapping(struct page *page)
357{
Kirill A. Shutemove39155e2015-04-15 16:14:53 -0700358 unsigned long mapping;
Shaohua Li98003392013-02-22 16:34:35 -0800359
Mikulas Patocka03e5ac22014-01-14 17:56:40 -0800360 /* This happens if someone calls flush_dcache_page on slab page */
361 if (unlikely(PageSlab(page)))
362 return NULL;
363
Shaohua Li33806f02013-02-22 16:34:37 -0800364 if (unlikely(PageSwapCache(page))) {
365 swp_entry_t entry;
366
367 entry.val = page_private(page);
Kirill A. Shutemove39155e2015-04-15 16:14:53 -0700368 return swap_address_space(entry);
369 }
370
371 mapping = (unsigned long)page->mapping;
372 if (mapping & PAGE_MAPPING_FLAGS)
373 return NULL;
374 return page->mapping;
Shaohua Li98003392013-02-22 16:34:35 -0800375}
376
Jerome Marchand49f0ce52014-01-21 15:49:14 -0800377int overcommit_ratio_handler(struct ctl_table *table, int write,
378 void __user *buffer, size_t *lenp,
379 loff_t *ppos)
380{
381 int ret;
382
383 ret = proc_dointvec(table, write, buffer, lenp, ppos);
384 if (ret == 0 && write)
385 sysctl_overcommit_kbytes = 0;
386 return ret;
387}
388
389int overcommit_kbytes_handler(struct ctl_table *table, int write,
390 void __user *buffer, size_t *lenp,
391 loff_t *ppos)
392{
393 int ret;
394
395 ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
396 if (ret == 0 && write)
397 sysctl_overcommit_ratio = 0;
398 return ret;
399}
400
Jerome Marchand00619bc2013-11-12 15:08:31 -0800401/*
402 * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
403 */
404unsigned long vm_commit_limit(void)
405{
Jerome Marchand49f0ce52014-01-21 15:49:14 -0800406 unsigned long allowed;
407
408 if (sysctl_overcommit_kbytes)
409 allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
410 else
411 allowed = ((totalram_pages - hugetlb_total_pages())
412 * sysctl_overcommit_ratio / 100);
413 allowed += total_swap_pages;
414
415 return allowed;
Jerome Marchand00619bc2013-11-12 15:08:31 -0800416}
417
William Robertsa9090252014-02-11 10:11:59 -0800418/**
419 * get_cmdline() - copy the cmdline value to a buffer.
420 * @task: the task whose cmdline value to copy.
421 * @buffer: the buffer to copy to.
422 * @buflen: the length of the buffer. Larger cmdline values are truncated
423 * to this length.
424 * Returns the size of the cmdline field copied. Note that the copy does
425 * not guarantee an ending NULL byte.
426 */
427int get_cmdline(struct task_struct *task, char *buffer, int buflen)
428{
429 int res = 0;
430 unsigned int len;
431 struct mm_struct *mm = get_task_mm(task);
432 if (!mm)
433 goto out;
434 if (!mm->arg_end)
435 goto out_mm; /* Shh! No looking before we're done */
436
437 len = mm->arg_end - mm->arg_start;
438
439 if (len > buflen)
440 len = buflen;
441
442 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
443
444 /*
445 * If the nul at the end of args has been overwritten, then
446 * assume application is using setproctitle(3).
447 */
448 if (res > 0 && buffer[res-1] != '\0' && len < buflen) {
449 len = strnlen(buffer, res);
450 if (len < res) {
451 res = len;
452 } else {
453 len = mm->env_end - mm->env_start;
454 if (len > buflen - res)
455 len = buflen - res;
456 res += access_process_vm(task, mm->env_start,
457 buffer+res, len, 0);
458 res = strnlen(buffer, res);
459 }
460 }
461out_mm:
462 mmput(mm);
463out:
464 return res;
465}