Linux Audio

Check our new training course

Loading...
v4.6
  1#include <linux/mm.h>
  2#include <linux/slab.h>
  3#include <linux/string.h>
  4#include <linux/compiler.h>
  5#include <linux/export.h>
  6#include <linux/err.h>
  7#include <linux/sched.h>
  8#include <linux/security.h>
  9#include <linux/swap.h>
 10#include <linux/swapops.h>
 11#include <linux/mman.h>
 12#include <linux/hugetlb.h>
 13#include <linux/vmalloc.h>
 14
 15#include <asm/sections.h>
 16#include <asm/uaccess.h>
 17
 18#include "internal.h"
 19
 20static 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
 39/**
 40 * kstrdup - allocate space for and copy an existing string
 41 * @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;
 53	buf = kmalloc_track_caller(len, gfp);
 54	if (buf)
 55		memcpy(buf, s, len);
 56	return buf;
 57}
 58EXPORT_SYMBOL(kstrdup);
 59
 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 */
 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/**
 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 */
 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/**
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 */
109void *kmemdup(const void *src, size_t len, gfp_t gfp)
110{
111	void *p;
112
113	p = kmalloc_track_caller(len, gfp);
114	if (p)
115		memcpy(p, src, len);
116	return p;
117}
118EXPORT_SYMBOL(kmemdup);
119
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 */
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
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
150/*
151 * strndup_user - duplicate an existing string from user space
152 * @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
168	p = memdup_user(s, length);
169
170	if (IS_ERR(p))
171		return p;
172
173	p[length - 1] = '\0';
174
175	return p;
176}
177EXPORT_SYMBOL(strndup_user);
178
179/**
180 * memdup_user_nul - duplicate memory region from user space and NUL-terminate
181 *
182 * @src: source address in user space
183 * @len: number of bytes to copy
184 *
185 * Returns an ERR_PTR() on failure.
186 */
187void *memdup_user_nul(const void __user *src, size_t len)
188{
189	char *p;
190
191	/*
192	 * Always use GFP_KERNEL, since copy_from_user() can sleep and
193	 * cause pagefault, which makes it pointless to use GFP_NOFS
194	 * or GFP_ATOMIC.
195	 */
196	p = kmalloc_track_caller(len + 1, GFP_KERNEL);
197	if (!p)
198		return ERR_PTR(-ENOMEM);
199
200	if (copy_from_user(p, src, len)) {
201		kfree(p);
202		return ERR_PTR(-EFAULT);
203	}
204	p[len] = '\0';
205
206	return p;
207}
208EXPORT_SYMBOL(memdup_user_nul);
209
210void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
211		struct vm_area_struct *prev, struct rb_node *rb_parent)
212{
213	struct vm_area_struct *next;
214
215	vma->vm_prev = prev;
216	if (prev) {
217		next = prev->vm_next;
218		prev->vm_next = vma;
219	} else {
220		mm->mmap = vma;
221		if (rb_parent)
222			next = rb_entry(rb_parent,
223					struct vm_area_struct, vm_rb);
224		else
225			next = NULL;
226	}
227	vma->vm_next = next;
228	if (next)
229		next->vm_prev = vma;
230}
231
232/* Check if the vma is being used as a stack by this task */
233int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t)
 
234{
235	return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
236}
237
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
238#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
239void arch_pick_mmap_layout(struct mm_struct *mm)
240{
241	mm->mmap_base = TASK_UNMAPPED_BASE;
242	mm->get_unmapped_area = arch_get_unmapped_area;
243}
244#endif
245
246/*
247 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
248 * back to the regular GUP.
249 * If the architecture not support this function, simply return with no
250 * page pinned
251 */
252int __weak __get_user_pages_fast(unsigned long start,
253				 int nr_pages, int write, struct page **pages)
254{
255	return 0;
256}
257EXPORT_SYMBOL_GPL(__get_user_pages_fast);
258
259/**
260 * get_user_pages_fast() - pin user pages in memory
261 * @start:	starting user address
262 * @nr_pages:	number of pages from start to pin
263 * @write:	whether pages will be written to
264 * @pages:	array that receives pointers to the pages pinned.
265 *		Should be at least nr_pages long.
266 *
267 * Returns number of pages pinned. This may be fewer than the number
268 * requested. If nr_pages is 0 or negative, returns 0. If no pages
269 * were pinned, returns -errno.
270 *
271 * get_user_pages_fast provides equivalent functionality to get_user_pages,
272 * operating on current and current->mm, with force=0 and vma=NULL. However
273 * unlike get_user_pages, it must be called without mmap_sem held.
274 *
275 * get_user_pages_fast may take mmap_sem and page table locks, so no
276 * assumptions can be made about lack of locking. get_user_pages_fast is to be
277 * implemented in a way that is advantageous (vs get_user_pages()) when the
278 * user memory area is already faulted in and present in ptes. However if the
279 * pages have to be faulted in, it may turn out to be slightly slower so
280 * callers need to carefully consider what to use. On many architectures,
281 * get_user_pages_fast simply falls back to get_user_pages.
282 */
283int __weak get_user_pages_fast(unsigned long start,
284				int nr_pages, int write, struct page **pages)
285{
286	return get_user_pages_unlocked(start, nr_pages, write, 0, pages);
 
 
 
 
 
 
 
 
287}
288EXPORT_SYMBOL_GPL(get_user_pages_fast);
289
290unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
291	unsigned long len, unsigned long prot,
292	unsigned long flag, unsigned long pgoff)
293{
294	unsigned long ret;
295	struct mm_struct *mm = current->mm;
296	unsigned long populate;
297
298	ret = security_mmap_file(file, prot, flag);
299	if (!ret) {
300		down_write(&mm->mmap_sem);
301		ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
302				    &populate);
303		up_write(&mm->mmap_sem);
304		if (populate)
305			mm_populate(ret, populate);
306	}
307	return ret;
308}
309
310unsigned long vm_mmap(struct file *file, unsigned long addr,
311	unsigned long len, unsigned long prot,
312	unsigned long flag, unsigned long offset)
313{
314	if (unlikely(offset + PAGE_ALIGN(len) < offset))
315		return -EINVAL;
316	if (unlikely(offset_in_page(offset)))
317		return -EINVAL;
318
319	return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
320}
321EXPORT_SYMBOL(vm_mmap);
322
323void kvfree(const void *addr)
324{
325	if (is_vmalloc_addr(addr))
326		vfree(addr);
327	else
328		kfree(addr);
329}
330EXPORT_SYMBOL(kvfree);
331
332static inline void *__page_rmapping(struct page *page)
333{
334	unsigned long mapping;
335
336	mapping = (unsigned long)page->mapping;
337	mapping &= ~PAGE_MAPPING_FLAGS;
338
339	return (void *)mapping;
340}
341
342/* Neutral page->mapping pointer to address_space or anon_vma or other */
343void *page_rmapping(struct page *page)
344{
345	page = compound_head(page);
346	return __page_rmapping(page);
347}
348
349struct anon_vma *page_anon_vma(struct page *page)
350{
351	unsigned long mapping;
352
353	page = compound_head(page);
354	mapping = (unsigned long)page->mapping;
355	if ((mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
356		return NULL;
357	return __page_rmapping(page);
358}
359
360struct address_space *page_mapping(struct page *page)
361{
362	struct address_space *mapping;
363
364	page = compound_head(page);
365
366	/* This happens if someone calls flush_dcache_page on slab page */
367	if (unlikely(PageSlab(page)))
368		return NULL;
369
370	if (unlikely(PageSwapCache(page))) {
371		swp_entry_t entry;
372
373		entry.val = page_private(page);
374		return swap_address_space(entry);
375	}
376
377	mapping = page->mapping;
378	if ((unsigned long)mapping & PAGE_MAPPING_FLAGS)
379		return NULL;
380	return mapping;
381}
382
383/* Slow path of page_mapcount() for compound pages */
384int __page_mapcount(struct page *page)
385{
386	int ret;
387
388	ret = atomic_read(&page->_mapcount) + 1;
389	page = compound_head(page);
390	ret += atomic_read(compound_mapcount_ptr(page)) + 1;
391	if (PageDoubleMap(page))
392		ret--;
393	return ret;
394}
395EXPORT_SYMBOL_GPL(__page_mapcount);
396
397int sysctl_overcommit_memory __read_mostly = OVERCOMMIT_GUESS;
398int sysctl_overcommit_ratio __read_mostly = 50;
399unsigned long sysctl_overcommit_kbytes __read_mostly;
400int sysctl_max_map_count __read_mostly = DEFAULT_MAX_MAP_COUNT;
401unsigned long sysctl_user_reserve_kbytes __read_mostly = 1UL << 17; /* 128MB */
402unsigned long sysctl_admin_reserve_kbytes __read_mostly = 1UL << 13; /* 8MB */
403
404int overcommit_ratio_handler(struct ctl_table *table, int write,
405			     void __user *buffer, size_t *lenp,
406			     loff_t *ppos)
407{
408	int ret;
409
410	ret = proc_dointvec(table, write, buffer, lenp, ppos);
411	if (ret == 0 && write)
412		sysctl_overcommit_kbytes = 0;
413	return ret;
414}
415
416int overcommit_kbytes_handler(struct ctl_table *table, int write,
417			     void __user *buffer, size_t *lenp,
418			     loff_t *ppos)
419{
420	int ret;
421
422	ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
423	if (ret == 0 && write)
424		sysctl_overcommit_ratio = 0;
425	return ret;
426}
427
428/*
429 * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
430 */
431unsigned long vm_commit_limit(void)
432{
433	unsigned long allowed;
434
435	if (sysctl_overcommit_kbytes)
436		allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
437	else
438		allowed = ((totalram_pages - hugetlb_total_pages())
439			   * sysctl_overcommit_ratio / 100);
440	allowed += total_swap_pages;
441
442	return allowed;
443}
444
445/*
446 * Make sure vm_committed_as in one cacheline and not cacheline shared with
447 * other variables. It can be updated by several CPUs frequently.
448 */
449struct percpu_counter vm_committed_as ____cacheline_aligned_in_smp;
450
451/*
452 * The global memory commitment made in the system can be a metric
453 * that can be used to drive ballooning decisions when Linux is hosted
454 * as a guest. On Hyper-V, the host implements a policy engine for dynamically
455 * balancing memory across competing virtual machines that are hosted.
456 * Several metrics drive this policy engine including the guest reported
457 * memory commitment.
458 */
459unsigned long vm_memory_committed(void)
460{
461	return percpu_counter_read_positive(&vm_committed_as);
462}
463EXPORT_SYMBOL_GPL(vm_memory_committed);
464
465/*
466 * Check that a process has enough memory to allocate a new virtual
467 * mapping. 0 means there is enough memory for the allocation to
468 * succeed and -ENOMEM implies there is not.
469 *
470 * We currently support three overcommit policies, which are set via the
471 * vm.overcommit_memory sysctl.  See Documentation/vm/overcommit-accounting
472 *
473 * Strict overcommit modes added 2002 Feb 26 by Alan Cox.
474 * Additional code 2002 Jul 20 by Robert Love.
475 *
476 * cap_sys_admin is 1 if the process has admin privileges, 0 otherwise.
477 *
478 * Note this is a helper function intended to be used by LSMs which
479 * wish to use this logic.
480 */
481int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin)
482{
483	long free, allowed, reserve;
484
485	VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) <
486			-(s64)vm_committed_as_batch * num_online_cpus(),
487			"memory commitment underflow");
488
489	vm_acct_memory(pages);
490
491	/*
492	 * Sometimes we want to use more memory than we have
493	 */
494	if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS)
495		return 0;
496
497	if (sysctl_overcommit_memory == OVERCOMMIT_GUESS) {
498		free = global_page_state(NR_FREE_PAGES);
499		free += global_page_state(NR_FILE_PAGES);
500
501		/*
502		 * shmem pages shouldn't be counted as free in this
503		 * case, they can't be purged, only swapped out, and
504		 * that won't affect the overall amount of available
505		 * memory in the system.
506		 */
507		free -= global_page_state(NR_SHMEM);
508
509		free += get_nr_swap_pages();
510
511		/*
512		 * Any slabs which are created with the
513		 * SLAB_RECLAIM_ACCOUNT flag claim to have contents
514		 * which are reclaimable, under pressure.  The dentry
515		 * cache and most inode caches should fall into this
516		 */
517		free += global_page_state(NR_SLAB_RECLAIMABLE);
518
519		/*
520		 * Leave reserved pages. The pages are not for anonymous pages.
521		 */
522		if (free <= totalreserve_pages)
523			goto error;
524		else
525			free -= totalreserve_pages;
526
527		/*
528		 * Reserve some for root
529		 */
530		if (!cap_sys_admin)
531			free -= sysctl_admin_reserve_kbytes >> (PAGE_SHIFT - 10);
532
533		if (free > pages)
534			return 0;
535
536		goto error;
537	}
538
539	allowed = vm_commit_limit();
540	/*
541	 * Reserve some for root
542	 */
543	if (!cap_sys_admin)
544		allowed -= sysctl_admin_reserve_kbytes >> (PAGE_SHIFT - 10);
545
546	/*
547	 * Don't let a single process grow so big a user can't recover
548	 */
549	if (mm) {
550		reserve = sysctl_user_reserve_kbytes >> (PAGE_SHIFT - 10);
551		allowed -= min_t(long, mm->total_vm / 32, reserve);
552	}
553
554	if (percpu_counter_read_positive(&vm_committed_as) < allowed)
555		return 0;
556error:
557	vm_unacct_memory(pages);
558
559	return -ENOMEM;
560}
561
562/**
563 * get_cmdline() - copy the cmdline value to a buffer.
564 * @task:     the task whose cmdline value to copy.
565 * @buffer:   the buffer to copy to.
566 * @buflen:   the length of the buffer. Larger cmdline values are truncated
567 *            to this length.
568 * Returns the size of the cmdline field copied. Note that the copy does
569 * not guarantee an ending NULL byte.
570 */
571int get_cmdline(struct task_struct *task, char *buffer, int buflen)
572{
573	int res = 0;
574	unsigned int len;
575	struct mm_struct *mm = get_task_mm(task);
576	unsigned long arg_start, arg_end, env_start, env_end;
577	if (!mm)
578		goto out;
579	if (!mm->arg_end)
580		goto out_mm;	/* Shh! No looking before we're done */
581
582	down_read(&mm->mmap_sem);
583	arg_start = mm->arg_start;
584	arg_end = mm->arg_end;
585	env_start = mm->env_start;
586	env_end = mm->env_end;
587	up_read(&mm->mmap_sem);
588
589	len = arg_end - arg_start;
590
591	if (len > buflen)
592		len = buflen;
593
594	res = access_process_vm(task, arg_start, buffer, len, 0);
595
596	/*
597	 * If the nul at the end of args has been overwritten, then
598	 * assume application is using setproctitle(3).
599	 */
600	if (res > 0 && buffer[res-1] != '\0' && len < buflen) {
601		len = strnlen(buffer, res);
602		if (len < res) {
603			res = len;
604		} else {
605			len = env_end - env_start;
606			if (len > buflen - res)
607				len = buflen - res;
608			res += access_process_vm(task, env_start,
609						 buffer+res, len, 0);
610			res = strnlen(buffer, res);
611		}
612	}
613out_mm:
614	mmput(mm);
615out:
616	return res;
617}
v3.15
  1#include <linux/mm.h>
  2#include <linux/slab.h>
  3#include <linux/string.h>
  4#include <linux/compiler.h>
  5#include <linux/export.h>
  6#include <linux/err.h>
  7#include <linux/sched.h>
  8#include <linux/security.h>
  9#include <linux/swap.h>
 10#include <linux/swapops.h>
 11#include <linux/mman.h>
 12#include <linux/hugetlb.h>
 13#include <linux/vmalloc.h>
 14
 
 15#include <asm/uaccess.h>
 16
 17#include "internal.h"
 18
 19#define CREATE_TRACE_POINTS
 20#include <trace/events/kmem.h>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 21
 22/**
 23 * kstrdup - allocate space for and copy an existing string
 24 * @s: the string to duplicate
 25 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
 26 */
 27char *kstrdup(const char *s, gfp_t gfp)
 28{
 29	size_t len;
 30	char *buf;
 31
 32	if (!s)
 33		return NULL;
 34
 35	len = strlen(s) + 1;
 36	buf = kmalloc_track_caller(len, gfp);
 37	if (buf)
 38		memcpy(buf, s, len);
 39	return buf;
 40}
 41EXPORT_SYMBOL(kstrdup);
 42
 43/**
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 44 * kstrndup - allocate space for and copy an existing string
 45 * @s: the string to duplicate
 46 * @max: read at most @max chars from @s
 47 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
 48 */
 49char *kstrndup(const char *s, size_t max, gfp_t gfp)
 50{
 51	size_t len;
 52	char *buf;
 53
 54	if (!s)
 55		return NULL;
 56
 57	len = strnlen(s, max);
 58	buf = kmalloc_track_caller(len+1, gfp);
 59	if (buf) {
 60		memcpy(buf, s, len);
 61		buf[len] = '\0';
 62	}
 63	return buf;
 64}
 65EXPORT_SYMBOL(kstrndup);
 66
 67/**
 68 * kmemdup - duplicate region of memory
 69 *
 70 * @src: memory region to duplicate
 71 * @len: memory region length
 72 * @gfp: GFP mask to use
 73 */
 74void *kmemdup(const void *src, size_t len, gfp_t gfp)
 75{
 76	void *p;
 77
 78	p = kmalloc_track_caller(len, gfp);
 79	if (p)
 80		memcpy(p, src, len);
 81	return p;
 82}
 83EXPORT_SYMBOL(kmemdup);
 84
 85/**
 86 * memdup_user - duplicate memory region from user space
 87 *
 88 * @src: source address in user space
 89 * @len: number of bytes to copy
 90 *
 91 * Returns an ERR_PTR() on failure.
 92 */
 93void *memdup_user(const void __user *src, size_t len)
 94{
 95	void *p;
 96
 97	/*
 98	 * Always use GFP_KERNEL, since copy_from_user() can sleep and
 99	 * cause pagefault, which makes it pointless to use GFP_NOFS
100	 * or GFP_ATOMIC.
101	 */
102	p = kmalloc_track_caller(len, GFP_KERNEL);
103	if (!p)
104		return ERR_PTR(-ENOMEM);
105
106	if (copy_from_user(p, src, len)) {
107		kfree(p);
108		return ERR_PTR(-EFAULT);
109	}
110
111	return p;
112}
113EXPORT_SYMBOL(memdup_user);
114
115static __always_inline void *__do_krealloc(const void *p, size_t new_size,
116					   gfp_t flags)
117{
118	void *ret;
119	size_t ks = 0;
120
121	if (p)
122		ks = ksize(p);
123
124	if (ks >= new_size)
125		return (void *)p;
126
127	ret = kmalloc_track_caller(new_size, flags);
128	if (ret && p)
129		memcpy(ret, p, ks);
130
131	return ret;
132}
133
134/**
135 * __krealloc - like krealloc() but don't free @p.
136 * @p: object to reallocate memory for.
137 * @new_size: how many bytes of memory are required.
138 * @flags: the type of memory to allocate.
139 *
140 * This function is like krealloc() except it never frees the originally
141 * allocated buffer. Use this if you don't want to free the buffer immediately
142 * like, for example, with RCU.
143 */
144void *__krealloc(const void *p, size_t new_size, gfp_t flags)
145{
146	if (unlikely(!new_size))
147		return ZERO_SIZE_PTR;
148
149	return __do_krealloc(p, new_size, flags);
150
151}
152EXPORT_SYMBOL(__krealloc);
153
154/**
155 * krealloc - reallocate memory. The contents will remain unchanged.
156 * @p: object to reallocate memory for.
157 * @new_size: how many bytes of memory are required.
158 * @flags: the type of memory to allocate.
159 *
160 * The contents of the object pointed to are preserved up to the
161 * lesser of the new and old sizes.  If @p is %NULL, krealloc()
162 * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
163 * %NULL pointer, the object pointed to is freed.
164 */
165void *krealloc(const void *p, size_t new_size, gfp_t flags)
166{
167	void *ret;
168
169	if (unlikely(!new_size)) {
170		kfree(p);
171		return ZERO_SIZE_PTR;
172	}
173
174	ret = __do_krealloc(p, new_size, flags);
175	if (ret && p != ret)
176		kfree(p);
177
178	return ret;
179}
180EXPORT_SYMBOL(krealloc);
181
182/**
183 * kzfree - like kfree but zero memory
184 * @p: object to free memory of
185 *
186 * The memory of the object @p points to is zeroed before freed.
187 * If @p is %NULL, kzfree() does nothing.
188 *
189 * Note: this function zeroes the whole allocated buffer which can be a good
190 * deal bigger than the requested buffer size passed to kmalloc(). So be
191 * careful when using this function in performance sensitive code.
192 */
193void kzfree(const void *p)
194{
195	size_t ks;
196	void *mem = (void *)p;
197
198	if (unlikely(ZERO_OR_NULL_PTR(mem)))
199		return;
200	ks = ksize(mem);
201	memset(mem, 0, ks);
202	kfree(mem);
203}
204EXPORT_SYMBOL(kzfree);
205
206/*
207 * strndup_user - duplicate an existing string from user space
208 * @s: The string to duplicate
209 * @n: Maximum number of bytes to copy, including the trailing NUL.
210 */
211char *strndup_user(const char __user *s, long n)
212{
213	char *p;
214	long length;
215
216	length = strnlen_user(s, n);
217
218	if (!length)
219		return ERR_PTR(-EFAULT);
220
221	if (length > n)
222		return ERR_PTR(-EINVAL);
223
224	p = memdup_user(s, length);
225
226	if (IS_ERR(p))
227		return p;
228
229	p[length - 1] = '\0';
230
231	return p;
232}
233EXPORT_SYMBOL(strndup_user);
234
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
235void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
236		struct vm_area_struct *prev, struct rb_node *rb_parent)
237{
238	struct vm_area_struct *next;
239
240	vma->vm_prev = prev;
241	if (prev) {
242		next = prev->vm_next;
243		prev->vm_next = vma;
244	} else {
245		mm->mmap = vma;
246		if (rb_parent)
247			next = rb_entry(rb_parent,
248					struct vm_area_struct, vm_rb);
249		else
250			next = NULL;
251	}
252	vma->vm_next = next;
253	if (next)
254		next->vm_prev = vma;
255}
256
257/* Check if the vma is being used as a stack by this task */
258static int vm_is_stack_for_task(struct task_struct *t,
259				struct vm_area_struct *vma)
260{
261	return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
262}
263
264/*
265 * Check if the vma is being used as a stack.
266 * If is_group is non-zero, check in the entire thread group or else
267 * just check in the current task. Returns the pid of the task that
268 * the vma is stack for.
269 */
270pid_t vm_is_stack(struct task_struct *task,
271		  struct vm_area_struct *vma, int in_group)
272{
273	pid_t ret = 0;
274
275	if (vm_is_stack_for_task(task, vma))
276		return task->pid;
277
278	if (in_group) {
279		struct task_struct *t;
280		rcu_read_lock();
281		if (!pid_alive(task))
282			goto done;
283
284		t = task;
285		do {
286			if (vm_is_stack_for_task(t, vma)) {
287				ret = t->pid;
288				goto done;
289			}
290		} while_each_thread(task, t);
291done:
292		rcu_read_unlock();
293	}
294
295	return ret;
296}
297
298#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
299void arch_pick_mmap_layout(struct mm_struct *mm)
300{
301	mm->mmap_base = TASK_UNMAPPED_BASE;
302	mm->get_unmapped_area = arch_get_unmapped_area;
303}
304#endif
305
306/*
307 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
308 * back to the regular GUP.
309 * If the architecture not support this function, simply return with no
310 * page pinned
311 */
312int __weak __get_user_pages_fast(unsigned long start,
313				 int nr_pages, int write, struct page **pages)
314{
315	return 0;
316}
317EXPORT_SYMBOL_GPL(__get_user_pages_fast);
318
319/**
320 * get_user_pages_fast() - pin user pages in memory
321 * @start:	starting user address
322 * @nr_pages:	number of pages from start to pin
323 * @write:	whether pages will be written to
324 * @pages:	array that receives pointers to the pages pinned.
325 *		Should be at least nr_pages long.
326 *
327 * Returns number of pages pinned. This may be fewer than the number
328 * requested. If nr_pages is 0 or negative, returns 0. If no pages
329 * were pinned, returns -errno.
330 *
331 * get_user_pages_fast provides equivalent functionality to get_user_pages,
332 * operating on current and current->mm, with force=0 and vma=NULL. However
333 * unlike get_user_pages, it must be called without mmap_sem held.
334 *
335 * get_user_pages_fast may take mmap_sem and page table locks, so no
336 * assumptions can be made about lack of locking. get_user_pages_fast is to be
337 * implemented in a way that is advantageous (vs get_user_pages()) when the
338 * user memory area is already faulted in and present in ptes. However if the
339 * pages have to be faulted in, it may turn out to be slightly slower so
340 * callers need to carefully consider what to use. On many architectures,
341 * get_user_pages_fast simply falls back to get_user_pages.
342 */
343int __weak get_user_pages_fast(unsigned long start,
344				int nr_pages, int write, struct page **pages)
345{
346	struct mm_struct *mm = current->mm;
347	int ret;
348
349	down_read(&mm->mmap_sem);
350	ret = get_user_pages(current, mm, start, nr_pages,
351					write, 0, pages, NULL);
352	up_read(&mm->mmap_sem);
353
354	return ret;
355}
356EXPORT_SYMBOL_GPL(get_user_pages_fast);
357
358unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
359	unsigned long len, unsigned long prot,
360	unsigned long flag, unsigned long pgoff)
361{
362	unsigned long ret;
363	struct mm_struct *mm = current->mm;
364	unsigned long populate;
365
366	ret = security_mmap_file(file, prot, flag);
367	if (!ret) {
368		down_write(&mm->mmap_sem);
369		ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
370				    &populate);
371		up_write(&mm->mmap_sem);
372		if (populate)
373			mm_populate(ret, populate);
374	}
375	return ret;
376}
377
378unsigned long vm_mmap(struct file *file, unsigned long addr,
379	unsigned long len, unsigned long prot,
380	unsigned long flag, unsigned long offset)
381{
382	if (unlikely(offset + PAGE_ALIGN(len) < offset))
383		return -EINVAL;
384	if (unlikely(offset & ~PAGE_MASK))
385		return -EINVAL;
386
387	return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
388}
389EXPORT_SYMBOL(vm_mmap);
390
391void kvfree(const void *addr)
392{
393	if (is_vmalloc_addr(addr))
394		vfree(addr);
395	else
396		kfree(addr);
397}
398EXPORT_SYMBOL(kvfree);
399
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
400struct address_space *page_mapping(struct page *page)
401{
402	struct address_space *mapping = page->mapping;
 
 
403
404	/* This happens if someone calls flush_dcache_page on slab page */
405	if (unlikely(PageSlab(page)))
406		return NULL;
407
408	if (unlikely(PageSwapCache(page))) {
409		swp_entry_t entry;
410
411		entry.val = page_private(page);
412		mapping = swap_address_space(entry);
413	} else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
414		mapping = NULL;
 
 
 
415	return mapping;
416}
417
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
418int overcommit_ratio_handler(struct ctl_table *table, int write,
419			     void __user *buffer, size_t *lenp,
420			     loff_t *ppos)
421{
422	int ret;
423
424	ret = proc_dointvec(table, write, buffer, lenp, ppos);
425	if (ret == 0 && write)
426		sysctl_overcommit_kbytes = 0;
427	return ret;
428}
429
430int overcommit_kbytes_handler(struct ctl_table *table, int write,
431			     void __user *buffer, size_t *lenp,
432			     loff_t *ppos)
433{
434	int ret;
435
436	ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
437	if (ret == 0 && write)
438		sysctl_overcommit_ratio = 0;
439	return ret;
440}
441
442/*
443 * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
444 */
445unsigned long vm_commit_limit(void)
446{
447	unsigned long allowed;
448
449	if (sysctl_overcommit_kbytes)
450		allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
451	else
452		allowed = ((totalram_pages - hugetlb_total_pages())
453			   * sysctl_overcommit_ratio / 100);
454	allowed += total_swap_pages;
455
456	return allowed;
457}
458
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
459/**
460 * get_cmdline() - copy the cmdline value to a buffer.
461 * @task:     the task whose cmdline value to copy.
462 * @buffer:   the buffer to copy to.
463 * @buflen:   the length of the buffer. Larger cmdline values are truncated
464 *            to this length.
465 * Returns the size of the cmdline field copied. Note that the copy does
466 * not guarantee an ending NULL byte.
467 */
468int get_cmdline(struct task_struct *task, char *buffer, int buflen)
469{
470	int res = 0;
471	unsigned int len;
472	struct mm_struct *mm = get_task_mm(task);
 
473	if (!mm)
474		goto out;
475	if (!mm->arg_end)
476		goto out_mm;	/* Shh! No looking before we're done */
477
478	len = mm->arg_end - mm->arg_start;
 
 
 
 
 
 
 
479
480	if (len > buflen)
481		len = buflen;
482
483	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
484
485	/*
486	 * If the nul at the end of args has been overwritten, then
487	 * assume application is using setproctitle(3).
488	 */
489	if (res > 0 && buffer[res-1] != '\0' && len < buflen) {
490		len = strnlen(buffer, res);
491		if (len < res) {
492			res = len;
493		} else {
494			len = mm->env_end - mm->env_start;
495			if (len > buflen - res)
496				len = buflen - res;
497			res += access_process_vm(task, mm->env_start,
498						 buffer+res, len, 0);
499			res = strnlen(buffer, res);
500		}
501	}
502out_mm:
503	mmput(mm);
504out:
505	return res;
506}
507
508/* Tracepoints definitions. */
509EXPORT_TRACEPOINT_SYMBOL(kmalloc);
510EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
511EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
512EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
513EXPORT_TRACEPOINT_SYMBOL(kfree);
514EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);