linux/mm/util.c
<<
>>
Prefs
   1#include <linux/mm.h>
   2#include <linux/slab.h>
   3#include <linux/string.h>
   4#include <linux/export.h>
   5#include <linux/err.h>
   6#include <linux/sched.h>
   7#include <linux/security.h>
   8#include <linux/swap.h>
   9#include <linux/swapops.h>
  10#include <asm/uaccess.h>
  11
  12#include "internal.h"
  13
  14#define CREATE_TRACE_POINTS
  15#include <trace/events/kmem.h>
  16
  17/**
  18 * kstrdup - allocate space for and copy an existing string
  19 * @s: the string to duplicate
  20 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  21 */
  22char *kstrdup(const char *s, gfp_t gfp)
  23{
  24        size_t len;
  25        char *buf;
  26
  27        if (!s)
  28                return NULL;
  29
  30        len = strlen(s) + 1;
  31        buf = kmalloc_track_caller(len, gfp);
  32        if (buf)
  33                memcpy(buf, s, len);
  34        return buf;
  35}
  36EXPORT_SYMBOL(kstrdup);
  37
  38/**
  39 * kstrndup - allocate space for and copy an existing string
  40 * @s: the string to duplicate
  41 * @max: read at most @max chars from @s
  42 * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  43 */
  44char *kstrndup(const char *s, size_t max, gfp_t gfp)
  45{
  46        size_t len;
  47        char *buf;
  48
  49        if (!s)
  50                return NULL;
  51
  52        len = strnlen(s, max);
  53        buf = kmalloc_track_caller(len+1, gfp);
  54        if (buf) {
  55                memcpy(buf, s, len);
  56                buf[len] = '\0';
  57        }
  58        return buf;
  59}
  60EXPORT_SYMBOL(kstrndup);
  61
  62/**
  63 * kmemdup - duplicate region of memory
  64 *
  65 * @src: memory region to duplicate
  66 * @len: memory region length
  67 * @gfp: GFP mask to use
  68 */
  69void *kmemdup(const void *src, size_t len, gfp_t gfp)
  70{
  71        void *p;
  72
  73        p = kmalloc_track_caller(len, gfp);
  74        if (p)
  75                memcpy(p, src, len);
  76        return p;
  77}
  78EXPORT_SYMBOL(kmemdup);
  79
  80/**
  81 * memdup_user - duplicate memory region from user space
  82 *
  83 * @src: source address in user space
  84 * @len: number of bytes to copy
  85 *
  86 * Returns an ERR_PTR() on failure.
  87 */
  88void *memdup_user(const void __user *src, size_t len)
  89{
  90        void *p;
  91
  92        /*
  93         * Always use GFP_KERNEL, since copy_from_user() can sleep and
  94         * cause pagefault, which makes it pointless to use GFP_NOFS
  95         * or GFP_ATOMIC.
  96         */
  97        p = kmalloc_track_caller(len, GFP_KERNEL);
  98        if (!p)
  99                return ERR_PTR(-ENOMEM);
 100
 101        if (copy_from_user(p, src, len)) {
 102                kfree(p);
 103                return ERR_PTR(-EFAULT);
 104        }
 105
 106        return p;
 107}
 108EXPORT_SYMBOL(memdup_user);
 109
 110static __always_inline void *__do_krealloc(const void *p, size_t new_size,
 111                                           gfp_t flags)
 112{
 113        void *ret;
 114        size_t ks = 0;
 115
 116        if (p)
 117                ks = ksize(p);
 118
 119        if (ks >= new_size)
 120                return (void *)p;
 121
 122        ret = kmalloc_track_caller(new_size, flags);
 123        if (ret && p)
 124                memcpy(ret, p, ks);
 125
 126        return ret;
 127}
 128
 129/**
 130 * __krealloc - like krealloc() but don't free @p.
 131 * @p: object to reallocate memory for.
 132 * @new_size: how many bytes of memory are required.
 133 * @flags: the type of memory to allocate.
 134 *
 135 * This function is like krealloc() except it never frees the originally
 136 * allocated buffer. Use this if you don't want to free the buffer immediately
 137 * like, for example, with RCU.
 138 */
 139void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 140{
 141        if (unlikely(!new_size))
 142                return ZERO_SIZE_PTR;
 143
 144        return __do_krealloc(p, new_size, flags);
 145
 146}
 147EXPORT_SYMBOL(__krealloc);
 148
 149/**
 150 * krealloc - reallocate memory. The contents will remain unchanged.
 151 * @p: object to reallocate memory for.
 152 * @new_size: how many bytes of memory are required.
 153 * @flags: the type of memory to allocate.
 154 *
 155 * The contents of the object pointed to are preserved up to the
 156 * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 157 * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
 158 * %NULL pointer, the object pointed to is freed.
 159 */
 160void *krealloc(const void *p, size_t new_size, gfp_t flags)
 161{
 162        void *ret;
 163
 164        if (unlikely(!new_size)) {
 165                kfree(p);
 166                return ZERO_SIZE_PTR;
 167        }
 168
 169        ret = __do_krealloc(p, new_size, flags);
 170        if (ret && p != ret)
 171                kfree(p);
 172
 173        return ret;
 174}
 175EXPORT_SYMBOL(krealloc);
 176
 177/**
 178 * kzfree - like kfree but zero memory
 179 * @p: object to free memory of
 180 *
 181 * The memory of the object @p points to is zeroed before freed.
 182 * If @p is %NULL, kzfree() does nothing.
 183 *
 184 * Note: this function zeroes the whole allocated buffer which can be a good
 185 * deal bigger than the requested buffer size passed to kmalloc(). So be
 186 * careful when using this function in performance sensitive code.
 187 */
 188void kzfree(const void *p)
 189{
 190        size_t ks;
 191        void *mem = (void *)p;
 192
 193        if (unlikely(ZERO_OR_NULL_PTR(mem)))
 194                return;
 195        ks = ksize(mem);
 196        memset(mem, 0, ks);
 197        kfree(mem);
 198}
 199EXPORT_SYMBOL(kzfree);
 200
 201/*
 202 * strndup_user - duplicate an existing string from user space
 203 * @s: The string to duplicate
 204 * @n: Maximum number of bytes to copy, including the trailing NUL.
 205 */
 206char *strndup_user(const char __user *s, long n)
 207{
 208        char *p;
 209        long length;
 210
 211        length = strnlen_user(s, n);
 212
 213        if (!length)
 214                return ERR_PTR(-EFAULT);
 215
 216        if (length > n)
 217                return ERR_PTR(-EINVAL);
 218
 219        p = memdup_user(s, length);
 220
 221        if (IS_ERR(p))
 222                return p;
 223
 224        p[length - 1] = '\0';
 225
 226        return p;
 227}
 228EXPORT_SYMBOL(strndup_user);
 229
 230void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 231                struct vm_area_struct *prev, struct rb_node *rb_parent)
 232{
 233        struct vm_area_struct *next;
 234
 235        vma->vm_prev = prev;
 236        if (prev) {
 237                next = prev->vm_next;
 238                prev->vm_next = vma;
 239        } else {
 240                mm->mmap = vma;
 241                if (rb_parent)
 242                        next = rb_entry(rb_parent,
 243                                        struct vm_area_struct, vm_rb);
 244                else
 245                        next = NULL;
 246        }
 247        vma->vm_next = next;
 248        if (next)
 249                next->vm_prev = vma;
 250}
 251
 252/* Check if the vma is being used as a stack by this task */
 253static int vm_is_stack_for_task(struct task_struct *t,
 254                                struct vm_area_struct *vma)
 255{
 256        return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 257}
 258
 259/*
 260 * Check if the vma is being used as a stack.
 261 * If is_group is non-zero, check in the entire thread group or else
 262 * just check in the current task. Returns the pid of the task that
 263 * the vma is stack for.
 264 */
 265pid_t vm_is_stack(struct task_struct *task,
 266                  struct vm_area_struct *vma, int in_group)
 267{
 268        pid_t ret = 0;
 269
 270        if (vm_is_stack_for_task(task, vma))
 271                return task->pid;
 272
 273        if (in_group) {
 274                struct task_struct *t;
 275                rcu_read_lock();
 276                if (!pid_alive(task))
 277                        goto done;
 278
 279                t = task;
 280                do {
 281                        if (vm_is_stack_for_task(t, vma)) {
 282                                ret = t->pid;
 283                                goto done;
 284                        }
 285                } while_each_thread(task, t);
 286done:
 287                rcu_read_unlock();
 288        }
 289
 290        return ret;
 291}
 292
 293#if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 294void arch_pick_mmap_layout(struct mm_struct *mm)
 295{
 296        mm->mmap_base = TASK_UNMAPPED_BASE;
 297        mm->get_unmapped_area = arch_get_unmapped_area;
 298}
 299#endif
 300
 301/*
 302 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 303 * back to the regular GUP.
 304 * If the architecture not support this function, simply return with no
 305 * page pinned
 306 */
 307int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 308                                 int nr_pages, int write, struct page **pages)
 309{
 310        return 0;
 311}
 312EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 313
 314/**
 315 * get_user_pages_fast() - pin user pages in memory
 316 * @start:      starting user address
 317 * @nr_pages:   number of pages from start to pin
 318 * @write:      whether pages will be written to
 319 * @pages:      array that receives pointers to the pages pinned.
 320 *              Should be at least nr_pages long.
 321 *
 322 * Returns number of pages pinned. This may be fewer than the number
 323 * requested. If nr_pages is 0 or negative, returns 0. If no pages
 324 * were pinned, returns -errno.
 325 *
 326 * get_user_pages_fast provides equivalent functionality to get_user_pages,
 327 * operating on current and current->mm, with force=0 and vma=NULL. However
 328 * unlike get_user_pages, it must be called without mmap_sem held.
 329 *
 330 * get_user_pages_fast may take mmap_sem and page table locks, so no
 331 * assumptions can be made about lack of locking. get_user_pages_fast is to be
 332 * implemented in a way that is advantageous (vs get_user_pages()) when the
 333 * user memory area is already faulted in and present in ptes. However if the
 334 * pages have to be faulted in, it may turn out to be slightly slower so
 335 * callers need to carefully consider what to use. On many architectures,
 336 * get_user_pages_fast simply falls back to get_user_pages.
 337 */
 338int __attribute__((weak)) get_user_pages_fast(unsigned long start,
 339                                int nr_pages, int write, struct page **pages)
 340{
 341        struct mm_struct *mm = current->mm;
 342        int ret;
 343
 344        down_read(&mm->mmap_sem);
 345        ret = get_user_pages(current, mm, start, nr_pages,
 346                                        write, 0, pages, NULL);
 347        up_read(&mm->mmap_sem);
 348
 349        return ret;
 350}
 351EXPORT_SYMBOL_GPL(get_user_pages_fast);
 352
 353unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 354        unsigned long len, unsigned long prot,
 355        unsigned long flag, unsigned long pgoff)
 356{
 357        unsigned long ret;
 358        struct mm_struct *mm = current->mm;
 359        unsigned long populate;
 360
 361        ret = security_mmap_file(file, prot, flag);
 362        if (!ret) {
 363                down_write(&mm->mmap_sem);
 364                ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
 365                                    &populate);
 366                up_write(&mm->mmap_sem);
 367                if (populate)
 368                        mm_populate(ret, populate);
 369        }
 370        return ret;
 371}
 372
 373unsigned long vm_mmap(struct file *file, unsigned long addr,
 374        unsigned long len, unsigned long prot,
 375        unsigned long flag, unsigned long offset)
 376{
 377        if (unlikely(offset + PAGE_ALIGN(len) < offset))
 378                return -EINVAL;
 379        if (unlikely(offset & ~PAGE_MASK))
 380                return -EINVAL;
 381
 382        return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
 383}
 384EXPORT_SYMBOL(vm_mmap);
 385
 386struct address_space *page_mapping(struct page *page)
 387{
 388        struct address_space *mapping = page->mapping;
 389
 390        VM_BUG_ON(PageSlab(page));
 391        if (unlikely(PageSwapCache(page))) {
 392                swp_entry_t entry;
 393
 394                entry.val = page_private(page);
 395                mapping = swap_address_space(entry);
 396        } else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 397                mapping = NULL;
 398        return mapping;
 399}
 400
 401/* Tracepoints definitions. */
 402EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 403EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 404EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 405EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 406EXPORT_TRACEPOINT_SYMBOL(kfree);
 407EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);
 408