mm/util.c

   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 <asm/uaccess.h>
  10
  11 #include "internal.h"
  12
  13 #define CREATE_TRACE_POINTS
  14 #include <trace/events/kmem.h>
  15
  16 /**
  17  * kstrdup - allocate space for and copy an existing string
  18  * @s: the string to duplicate
  19  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  20  */
  21 char *kstrdup(const char *s, gfp_t gfp)
  22 {
  23         size_t len;
  24         char *buf;
  25
  26         if (!s)
  27                 return NULL;
  28
  29         len = strlen(s) + 1;
  30         buf = kmalloc_track_caller(len, gfp);
  31         if (buf)
  32                 memcpy(buf, s, len);
  33         return buf;
  34 }
  35 EXPORT_SYMBOL(kstrdup);
  36
  37 /**
  38  * kstrndup - allocate space for and copy an existing string
  39  * @s: the string to duplicate
  40  * @max: read at most @max chars from @s
  41  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  42  */
  43 char *kstrndup(const char *s, size_t max, gfp_t gfp)
  44 {
  45         size_t len;
  46         char *buf;
  47
  48         if (!s)
  49                 return NULL;
  50
  51         len = strnlen(s, max);
  52         buf = kmalloc_track_caller(len+1, gfp);
  53         if (buf) {
  54                 memcpy(buf, s, len);
  55                 buf[len] = '\0';
  56         }
  57         return buf;
  58 }
  59 EXPORT_SYMBOL(kstrndup);
  60
  61 /**
  62  * kmemdup - duplicate region of memory
  63  *
  64  * @src: memory region to duplicate
  65  * @len: memory region length
  66  * @gfp: GFP mask to use
  67  */
  68 void *kmemdup(const void *src, size_t len, gfp_t gfp)
  69 {
  70         void *p;
  71
  72         p = kmalloc_track_caller(len, gfp);
  73         if (p)
  74                 memcpy(p, src, len);
  75         return p;
  76 }
  77 EXPORT_SYMBOL(kmemdup);
  78
  79 /**
  80  * memdup_user - duplicate memory region from user space
  81  *
  82  * @src: source address in user space
  83  * @len: number of bytes to copy
  84  *
  85  * Returns an ERR_PTR() on failure.
  86  */
  87 void *memdup_user(const void __user *src, size_t len)
  88 {
  89         void *p;
  90
  91         /*
  92          * Always use GFP_KERNEL, since copy_from_user() can sleep and
  93          * cause pagefault, which makes it pointless to use GFP_NOFS
  94          * or GFP_ATOMIC.
  95          */
  96         p = kmalloc_track_caller(len, GFP_KERNEL);
  97         if (!p)
  98                 return ERR_PTR(-ENOMEM);
  99
 100         if (copy_from_user(p, src, len)) {
 101                 kfree(p);
 102                 return ERR_PTR(-EFAULT);
 103         }
 104
 105         return p;
 106 }
 107 EXPORT_SYMBOL(memdup_user);
 108
 109 static __always_inline void *__do_krealloc(const void *p, size_t new_size,
 110                                            gfp_t flags)
 111 {
 112         void *ret;
 113         size_t ks = 0;
 114
 115         if (p)
 116                 ks = ksize(p);
 117
 118         if (ks >= new_size)
 119                 return (void *)p;
 120
 121         ret = kmalloc_track_caller(new_size, flags);
 122         if (ret && p)
 123                 memcpy(ret, p, ks);
 124
 125         return ret;
 126 }
 127
 128 /**
 129  * __krealloc - like krealloc() but don't free @p.
 130  * @p: object to reallocate memory for.
 131  * @new_size: how many bytes of memory are required.
 132  * @flags: the type of memory to allocate.
 133  *
 134  * This function is like krealloc() except it never frees the originally
 135  * allocated buffer. Use this if you don't want to free the buffer immediately
 136  * like, for example, with RCU.
 137  */
 138 void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 139 {
 140         if (unlikely(!new_size))
 141                 return ZERO_SIZE_PTR;
 142
 143         return __do_krealloc(p, new_size, flags);
 144
 145 }
 146 EXPORT_SYMBOL(__krealloc);
 147
 148 /**
 149  * krealloc - reallocate memory. The contents will remain unchanged.
 150  * @p: object to reallocate memory for.
 151  * @new_size: how many bytes of memory are required.
 152  * @flags: the type of memory to allocate.
 153  *
 154  * The contents of the object pointed to are preserved up to the
 155  * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 156  * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
 157  * %NULL pointer, the object pointed to is freed.
 158  */
 159 void *krealloc(const void *p, size_t new_size, gfp_t flags)
 160 {
 161         void *ret;
 162
 163         if (unlikely(!new_size)) {
 164                 kfree(p);
 165                 return ZERO_SIZE_PTR;
 166         }
 167
 168         ret = __do_krealloc(p, new_size, flags);
 169         if (ret && p != ret)
 170                 kfree(p);
 171
 172         return ret;
 173 }
 174 EXPORT_SYMBOL(krealloc);
 175
 176 /**
 177  * kzfree - like kfree but zero memory
 178  * @p: object to free memory of
 179  *
 180  * The memory of the object @p points to is zeroed before freed.
 181  * If @p is %NULL, kzfree() does nothing.
 182  *
 183  * Note: this function zeroes the whole allocated buffer which can be a good
 184  * deal bigger than the requested buffer size passed to kmalloc(). So be
 185  * careful when using this function in performance sensitive code.
 186  */
 187 void kzfree(const void *p)
 188 {
 189         size_t ks;
 190         void *mem = (void *)p;
 191
 192         if (unlikely(ZERO_OR_NULL_PTR(mem)))
 193                 return;
 194         ks = ksize(mem);
 195         memset(mem, 0, ks);
 196         kfree(mem);
 197 }
 198 EXPORT_SYMBOL(kzfree);
 199
 200 /*
 201  * strndup_user - duplicate an existing string from user space
 202  * @s: The string to duplicate
 203  * @n: Maximum number of bytes to copy, including the trailing NUL.
 204  */
 205 char *strndup_user(const char __user *s, long n)
 206 {
 207         char *p;
 208         long length;
 209
 210         length = strnlen_user(s, n);
 211
 212         if (!length)
 213                 return ERR_PTR(-EFAULT);
 214
 215         if (length > n)
 216                 return ERR_PTR(-EINVAL);
 217
 218         p = memdup_user(s, length);
 219
 220         if (IS_ERR(p))
 221                 return p;
 222
 223         p[length - 1] = '\0';
 224
 225         return p;
 226 }
 227 EXPORT_SYMBOL(strndup_user);
 228
 229 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 230                 struct vm_area_struct *prev, struct rb_node *rb_parent)
 231 {
 232         struct vm_area_struct *next;
 233
 234         vma->vm_prev = prev;
 235         if (prev) {
 236                 next = prev->vm_next;
 237                 prev->vm_next = vma;
 238         } else {
 239                 mm->mmap = vma;
 240                 if (rb_parent)
 241                         next = rb_entry(rb_parent,
 242                                         struct vm_area_struct, vm_rb);
 243                 else
 244                         next = NULL;
 245         }
 246         vma->vm_next = next;
 247         if (next)
 248                 next->vm_prev = vma;
 249 }
 250
 251 /* Check if the vma is being used as a stack by this task */
 252 static int vm_is_stack_for_task(struct task_struct *t,
 253                                 struct vm_area_struct *vma)
 254 {
 255         return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 256 }
 257
 258 /*
 259  * Check if the vma is being used as a stack.
 260  * If is_group is non-zero, check in the entire thread group or else
 261  * just check in the current task. Returns the pid of the task that
 262  * the vma is stack for.
 263  */
 264 pid_t vm_is_stack(struct task_struct *task,
 265                   struct vm_area_struct *vma, int in_group)
 266 {
 267         pid_t ret = 0;
 268
 269         if (vm_is_stack_for_task(task, vma))
 270                 return task->pid;
 271
 272         if (in_group) {
 273                 struct task_struct *t;
 274                 rcu_read_lock();
 275                 if (!pid_alive(task))
 276                         goto done;
 277
 278                 t = task;
 279                 do {
 280                         if (vm_is_stack_for_task(t, vma)) {
 281                                 ret = t->pid;
 282                                 goto done;
 283                         }
 284                 } while_each_thread(task, t);
 285 done:
 286                 rcu_read_unlock();
 287         }
 288
 289         return ret;
 290 }
 291
 292 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 293 void arch_pick_mmap_layout(struct mm_struct *mm)
 294 {
 295         mm->mmap_base = TASK_UNMAPPED_BASE;
 296         mm->get_unmapped_area = arch_get_unmapped_area;
 297         mm->unmap_area = arch_unmap_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  */
 307 int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 308                                  int nr_pages, int write, struct page **pages)
 309 {
 310         return 0;
 311 }
 312 EXPORT_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  */
 338 int __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 }
 351 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 352
 353 unsigned 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
 373 unsigned 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 }
 384 EXPORT_SYMBOL(vm_mmap);
 385
 386 struct address_space *page_mapping(struct page *page)
 387 {
 388         struct address_space *mapping = page->mapping;
 389
 390         VM_BUG_ON(PageSlab(page));
 391 #ifdef CONFIG_SWAP
 392         if (unlikely(PageSwapCache(page)))
 393                 mapping = &swapper_space;
 394         else
 395 #endif
 396         if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 397                 mapping = NULL;
 398         return mapping;
 399 }
 400
 401 /* Tracepoints definitions. */
 402 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 403 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 404 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 405 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 406 EXPORT_TRACEPOINT_SYMBOL(kfree);
 407 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);