2 #include <linux/slab.h>
3 #include <linux/string.h>
4 #include <linux/compiler.h>
5 #include <linux/export.h>
6 #include <linux/ctype.h>
8 #include <linux/sched.h>
9 #include <linux/security.h>
10 #include <linux/swap.h>
11 #include <linux/swapops.h>
12 #include <linux/mman.h>
13 #include <linux/hugetlb.h>
14 #include <linux/vmalloc.h>
16 #include <asm/sections.h>
17 #include <asm/uaccess.h>
21 static inline int is_kernel_rodata(unsigned long addr)
23 return addr >= (unsigned long)__start_rodata &&
24 addr < (unsigned long)__end_rodata;
28 * kfree_const - conditionally free memory
29 * @x: pointer to the memory
31 * Function calls kfree only if @x is not in .rodata section.
33 void kfree_const(const void *x)
35 if (!is_kernel_rodata((unsigned long)x))
38 EXPORT_SYMBOL(kfree_const);
41 * kstrdup - allocate space for and copy an existing string
42 * @s: the string to duplicate
43 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
45 char *kstrdup(const char *s, gfp_t gfp)
54 buf = kmalloc_track_caller(len, gfp);
59 EXPORT_SYMBOL(kstrdup);
62 * kstrdup_const - conditionally duplicate an existing const string
63 * @s: the string to duplicate
64 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
66 * Function returns source string if it is in .rodata section otherwise it
67 * fallbacks to kstrdup.
68 * Strings allocated by kstrdup_const should be freed by kfree_const.
70 const char *kstrdup_const(const char *s, gfp_t gfp)
72 if (is_kernel_rodata((unsigned long)s))
75 return kstrdup(s, gfp);
77 EXPORT_SYMBOL(kstrdup_const);
80 * kstrndup - allocate space for and copy an existing string
81 * @s: the string to duplicate
82 * @max: read at most @max chars from @s
83 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
85 char *kstrndup(const char *s, size_t max, gfp_t gfp)
93 len = strnlen(s, max);
94 buf = kmalloc_track_caller(len+1, gfp);
101 EXPORT_SYMBOL(kstrndup);
104 * kstrimdup - Trim and copy a %NUL terminated string.
105 * @s: the string to trim and duplicate
106 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
108 * Returns an address, which the caller must kfree, containing
109 * a duplicate of the passed string with leading and/or trailing
110 * whitespace (as defined by isspace) removed.
112 char *kstrimdup(const char *s, gfp_t gfp)
115 char *begin = skip_spaces(s);
116 size_t len = strlen(begin);
118 while (len && isspace(begin[len - 1]))
121 buf = kmalloc_track_caller(len + 1, gfp);
125 memcpy(buf, begin, len);
130 EXPORT_SYMBOL(kstrimdup);
133 * kmemdup - duplicate region of memory
135 * @src: memory region to duplicate
136 * @len: memory region length
137 * @gfp: GFP mask to use
139 void *kmemdup(const void *src, size_t len, gfp_t gfp)
143 p = kmalloc_track_caller(len, gfp);
148 EXPORT_SYMBOL(kmemdup);
151 * memdup_user - duplicate memory region from user space
153 * @src: source address in user space
154 * @len: number of bytes to copy
156 * Returns an ERR_PTR() on failure.
158 void *memdup_user(const void __user *src, size_t len)
163 * Always use GFP_KERNEL, since copy_from_user() can sleep and
164 * cause pagefault, which makes it pointless to use GFP_NOFS
167 p = kmalloc_track_caller(len, GFP_KERNEL);
169 return ERR_PTR(-ENOMEM);
171 if (copy_from_user(p, src, len)) {
173 return ERR_PTR(-EFAULT);
178 EXPORT_SYMBOL(memdup_user);
181 * strndup_user - duplicate an existing string from user space
182 * @s: The string to duplicate
183 * @n: Maximum number of bytes to copy, including the trailing NUL.
185 char *strndup_user(const char __user *s, long n)
190 length = strnlen_user(s, n);
193 return ERR_PTR(-EFAULT);
196 return ERR_PTR(-EINVAL);
198 p = memdup_user(s, length);
203 p[length - 1] = '\0';
207 EXPORT_SYMBOL(strndup_user);
210 * memdup_user_nul - duplicate memory region from user space and NUL-terminate
212 * @src: source address in user space
213 * @len: number of bytes to copy
215 * Returns an ERR_PTR() on failure.
217 void *memdup_user_nul(const void __user *src, size_t len)
222 * Always use GFP_KERNEL, since copy_from_user() can sleep and
223 * cause pagefault, which makes it pointless to use GFP_NOFS
226 p = kmalloc_track_caller(len + 1, GFP_KERNEL);
228 return ERR_PTR(-ENOMEM);
230 if (copy_from_user(p, src, len)) {
232 return ERR_PTR(-EFAULT);
238 EXPORT_SYMBOL(memdup_user_nul);
240 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
241 struct vm_area_struct *prev, struct rb_node *rb_parent)
243 struct vm_area_struct *next;
247 next = prev->vm_next;
252 next = rb_entry(rb_parent,
253 struct vm_area_struct, vm_rb);
262 /* Check if the vma is being used as a stack by this task */
263 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t)
265 return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
268 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
269 void arch_pick_mmap_layout(struct mm_struct *mm)
271 mm->mmap_base = TASK_UNMAPPED_BASE;
272 mm->get_unmapped_area = arch_get_unmapped_area;
277 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
278 * back to the regular GUP.
279 * If the architecture not support this function, simply return with no
282 int __weak __get_user_pages_fast(unsigned long start,
283 int nr_pages, int write, struct page **pages)
287 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
290 * get_user_pages_fast() - pin user pages in memory
291 * @start: starting user address
292 * @nr_pages: number of pages from start to pin
293 * @write: whether pages will be written to
294 * @pages: array that receives pointers to the pages pinned.
295 * Should be at least nr_pages long.
297 * Returns number of pages pinned. This may be fewer than the number
298 * requested. If nr_pages is 0 or negative, returns 0. If no pages
299 * were pinned, returns -errno.
301 * get_user_pages_fast provides equivalent functionality to get_user_pages,
302 * operating on current and current->mm, with force=0 and vma=NULL. However
303 * unlike get_user_pages, it must be called without mmap_sem held.
305 * get_user_pages_fast may take mmap_sem and page table locks, so no
306 * assumptions can be made about lack of locking. get_user_pages_fast is to be
307 * implemented in a way that is advantageous (vs get_user_pages()) when the
308 * user memory area is already faulted in and present in ptes. However if the
309 * pages have to be faulted in, it may turn out to be slightly slower so
310 * callers need to carefully consider what to use. On many architectures,
311 * get_user_pages_fast simply falls back to get_user_pages.
313 int __weak get_user_pages_fast(unsigned long start,
314 int nr_pages, int write, struct page **pages)
316 struct mm_struct *mm = current->mm;
317 return get_user_pages_unlocked(current, mm, start, nr_pages,
320 EXPORT_SYMBOL_GPL(get_user_pages_fast);
322 unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
323 unsigned long len, unsigned long prot,
324 unsigned long flag, unsigned long pgoff)
327 struct mm_struct *mm = current->mm;
328 unsigned long populate;
330 ret = security_mmap_file(file, prot, flag);
332 down_write(&mm->mmap_sem);
333 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
335 up_write(&mm->mmap_sem);
337 mm_populate(ret, populate);
342 unsigned long vm_mmap(struct file *file, unsigned long addr,
343 unsigned long len, unsigned long prot,
344 unsigned long flag, unsigned long offset)
346 if (unlikely(offset + PAGE_ALIGN(len) < offset))
348 if (unlikely(offset_in_page(offset)))
351 return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
353 EXPORT_SYMBOL(vm_mmap);
355 void kvfree(const void *addr)
357 if (is_vmalloc_addr(addr))
362 EXPORT_SYMBOL(kvfree);
364 static inline void *__page_rmapping(struct page *page)
366 unsigned long mapping;
368 mapping = (unsigned long)page->mapping;
369 mapping &= ~PAGE_MAPPING_FLAGS;
371 return (void *)mapping;
374 /* Neutral page->mapping pointer to address_space or anon_vma or other */
375 void *page_rmapping(struct page *page)
377 page = compound_head(page);
378 return __page_rmapping(page);
381 struct anon_vma *page_anon_vma(struct page *page)
383 unsigned long mapping;
385 page = compound_head(page);
386 mapping = (unsigned long)page->mapping;
387 if ((mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
389 return __page_rmapping(page);
392 struct address_space *page_mapping(struct page *page)
394 struct address_space *mapping;
396 page = compound_head(page);
398 /* This happens if someone calls flush_dcache_page on slab page */
399 if (unlikely(PageSlab(page)))
402 if (unlikely(PageSwapCache(page))) {
405 entry.val = page_private(page);
406 return swap_address_space(entry);
409 mapping = page->mapping;
410 if ((unsigned long)mapping & PAGE_MAPPING_FLAGS)
415 /* Slow path of page_mapcount() for compound pages */
416 int __page_mapcount(struct page *page)
420 ret = atomic_read(&page->_mapcount) + 1;
421 page = compound_head(page);
422 ret += atomic_read(compound_mapcount_ptr(page)) + 1;
423 if (PageDoubleMap(page))
427 EXPORT_SYMBOL_GPL(__page_mapcount);
429 int overcommit_ratio_handler(struct ctl_table *table, int write,
430 void __user *buffer, size_t *lenp,
435 ret = proc_dointvec(table, write, buffer, lenp, ppos);
436 if (ret == 0 && write)
437 sysctl_overcommit_kbytes = 0;
441 int overcommit_kbytes_handler(struct ctl_table *table, int write,
442 void __user *buffer, size_t *lenp,
447 ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
448 if (ret == 0 && write)
449 sysctl_overcommit_ratio = 0;
454 * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
456 unsigned long vm_commit_limit(void)
458 unsigned long allowed;
460 if (sysctl_overcommit_kbytes)
461 allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
463 allowed = ((totalram_pages - hugetlb_total_pages())
464 * sysctl_overcommit_ratio / 100);
465 allowed += total_swap_pages;
471 * get_cmdline() - copy the cmdline value to a buffer.
472 * @task: the task whose cmdline value to copy.
473 * @buffer: the buffer to copy to.
474 * @buflen: the length of the buffer. Larger cmdline values are truncated
476 * Returns the size of the cmdline field copied. Note that the copy does
477 * not guarantee an ending NULL byte.
479 int get_cmdline(struct task_struct *task, char *buffer, int buflen)
483 struct mm_struct *mm = get_task_mm(task);
484 unsigned long arg_start, arg_end, env_start, env_end;
488 goto out_mm; /* Shh! No looking before we're done */
490 down_read(&mm->mmap_sem);
491 arg_start = mm->arg_start;
492 arg_end = mm->arg_end;
493 env_start = mm->env_start;
494 env_end = mm->env_end;
495 up_read(&mm->mmap_sem);
497 len = arg_end - arg_start;
502 res = access_process_vm(task, arg_start, buffer, len, 0);
505 * If the nul at the end of args has been overwritten, then
506 * assume application is using setproctitle(3).
508 if (res > 0 && buffer[res-1] != '\0' && len < buflen) {
509 len = strnlen(buffer, res);
513 len = env_end - env_start;
514 if (len > buflen - res)
516 res += access_process_vm(task, env_start,
518 res = strnlen(buffer, res);