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 <linux/swapops.h>
  10 #include <linux/vmalloc.h>
  11 #include <asm/uaccess.h>
  12
  13 #include "internal.h"
  14
  15 #define CREATE_TRACE_POINTS
  16 #include <trace/events/kmem.h>
  17
  18 /**
  19  * kstrdup - allocate space for and copy an existing string
  20  * @s: the string to duplicate
  21  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  22  */
  23 char *kstrdup(const char *s, gfp_t gfp)
  24 {
  25         size_t len;
  26         char *buf;
  27
  28         if (!s)
  29                 return NULL;
  30
  31         len = strlen(s) + 1;
  32         buf = kmalloc_track_caller(len, gfp);
  33         if (buf)
  34                 memcpy(buf, s, len);
  35         return buf;
  36 }
  37 EXPORT_SYMBOL(kstrdup);
  38
  39 /**
  40  * kstrndup - allocate space for and copy an existing string
  41  * @s: the string to duplicate
  42  * @max: read at most @max chars from @s
  43  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  44  */
  45 char *kstrndup(const char *s, size_t max, gfp_t gfp)
  46 {
  47         size_t len;
  48         char *buf;
  49
  50         if (!s)
  51                 return NULL;
  52
  53         len = strnlen(s, max);
  54         buf = kmalloc_track_caller(len+1, gfp);
  55         if (buf) {
  56                 memcpy(buf, s, len);
  57                 buf[len] = '\0';
  58         }
  59         return buf;
  60 }
  61 EXPORT_SYMBOL(kstrndup);
  62
  63 /**
  64  * kmemdup - duplicate region of memory
  65  *
  66  * @src: memory region to duplicate
  67  * @len: memory region length
  68  * @gfp: GFP mask to use
  69  */
  70 void *kmemdup(const void *src, size_t len, gfp_t gfp)
  71 {
  72         void *p;
  73
  74         p = kmalloc_track_caller(len, gfp);
  75         if (p)
  76                 memcpy(p, src, len);
  77         return p;
  78 }
  79 EXPORT_SYMBOL(kmemdup);
  80
  81 /**
  82  * memdup_user - duplicate memory region from user space
  83  *
  84  * @src: source address in user space
  85  * @len: number of bytes to copy
  86  *
  87  * Returns an ERR_PTR() on failure.
  88  */
  89 void *memdup_user(const void __user *src, size_t len)
  90 {
  91         void *p;
  92
  93         /*
  94          * Always use GFP_KERNEL, since copy_from_user() can sleep and
  95          * cause pagefault, which makes it pointless to use GFP_NOFS
  96          * or GFP_ATOMIC.
  97          */
  98         p = kmalloc_track_caller(len, GFP_KERNEL);
  99         if (!p)
 100                 return ERR_PTR(-ENOMEM);
 101
 102         if (copy_from_user(p, src, len)) {
 103                 kfree(p);
 104                 return ERR_PTR(-EFAULT);
 105         }
 106
 107         return p;
 108 }
 109 EXPORT_SYMBOL(memdup_user);
 110
 111 static __always_inline void *__do_krealloc(const void *p, size_t new_size,
 112                                            gfp_t flags)
 113 {
 114         void *ret;
 115         size_t ks = 0;
 116
 117         if (p)
 118                 ks = ksize(p);
 119
 120         if (ks >= new_size)
 121                 return (void *)p;
 122
 123         ret = kmalloc_track_caller(new_size, flags);
 124         if (ret && p)
 125                 memcpy(ret, p, ks);
 126
 127         return ret;
 128 }
 129
 130 /**
 131  * __krealloc - like krealloc() but don't free @p.
 132  * @p: object to reallocate memory for.
 133  * @new_size: how many bytes of memory are required.
 134  * @flags: the type of memory to allocate.
 135  *
 136  * This function is like krealloc() except it never frees the originally
 137  * allocated buffer. Use this if you don't want to free the buffer immediately
 138  * like, for example, with RCU.
 139  */
 140 void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 141 {
 142         if (unlikely(!new_size))
 143                 return ZERO_SIZE_PTR;
 144
 145         return __do_krealloc(p, new_size, flags);
 146
 147 }
 148 EXPORT_SYMBOL(__krealloc);
 149
 150 /**
 151  * krealloc - reallocate memory. The contents will remain unchanged.
 152  * @p: object to reallocate memory for.
 153  * @new_size: how many bytes of memory are required.
 154  * @flags: the type of memory to allocate.
 155  *
 156  * The contents of the object pointed to are preserved up to the
 157  * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 158  * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
 159  * %NULL pointer, the object pointed to is freed.
 160  */
 161 void *krealloc(const void *p, size_t new_size, gfp_t flags)
 162 {
 163         void *ret;
 164
 165         if (unlikely(!new_size)) {
 166                 kfree(p);
 167                 return ZERO_SIZE_PTR;
 168         }
 169
 170         ret = __do_krealloc(p, new_size, flags);
 171         if (ret && p != ret)
 172                 kfree(p);
 173
 174         return ret;
 175 }
 176 EXPORT_SYMBOL(krealloc);
 177
 178 /**
 179  * kzfree - like kfree but zero memory
 180  * @p: object to free memory of
 181  *
 182  * The memory of the object @p points to is zeroed before freed.
 183  * If @p is %NULL, kzfree() does nothing.
 184  *
 185  * Note: this function zeroes the whole allocated buffer which can be a good
 186  * deal bigger than the requested buffer size passed to kmalloc(). So be
 187  * careful when using this function in performance sensitive code.
 188  */
 189 void kzfree(const void *p)
 190 {
 191         size_t ks;
 192         void *mem = (void *)p;
 193
 194         if (unlikely(ZERO_OR_NULL_PTR(mem)))
 195                 return;
 196         ks = ksize(mem);
 197         memset(mem, 0, ks);
 198         kfree(mem);
 199 }
 200 EXPORT_SYMBOL(kzfree);
 201
 202 /*
 203  * strndup_user - duplicate an existing string from user space
 204  * @s: The string to duplicate
 205  * @n: Maximum number of bytes to copy, including the trailing NUL.
 206  */
 207 char *strndup_user(const char __user *s, long n)
 208 {
 209         char *p;
 210         long length;
 211
 212         length = strnlen_user(s, n);
 213
 214         if (!length)
 215                 return ERR_PTR(-EFAULT);
 216
 217         if (length > n)
 218                 return ERR_PTR(-EINVAL);
 219
 220         p = memdup_user(s, length);
 221
 222         if (IS_ERR(p))
 223                 return p;
 224
 225         p[length - 1] = '\0';
 226
 227         return p;
 228 }
 229 EXPORT_SYMBOL(strndup_user);
 230
 231 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 232                 struct vm_area_struct *prev, struct rb_node *rb_parent)
 233 {
 234         struct vm_area_struct *next;
 235
 236         vma->vm_prev = prev;
 237         if (prev) {
 238                 next = prev->vm_next;
 239                 prev->vm_next = vma;
 240         } else {
 241                 mm->mmap = vma;
 242                 if (rb_parent)
 243                         next = rb_entry(rb_parent,
 244                                         struct vm_area_struct, vm_rb);
 245                 else
 246                         next = NULL;
 247         }
 248         vma->vm_next = next;
 249         if (next)
 250                 next->vm_prev = vma;
 251 }
 252
 253 /* Check if the vma is being used as a stack by this task */
 254 static int vm_is_stack_for_task(struct task_struct *t,
 255                                 struct vm_area_struct *vma)
 256 {
 257         return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 258 }
 259
 260 /*
 261  * Check if the vma is being used as a stack.
 262  * If is_group is non-zero, check in the entire thread group or else
 263  * just check in the current task. Returns the pid of the task that
 264  * the vma is stack for.
 265  */
 266 pid_t vm_is_stack(struct task_struct *task,
 267                   struct vm_area_struct *vma, int in_group)
 268 {
 269         pid_t ret = 0;
 270
 271         if (vm_is_stack_for_task(task, vma))
 272                 return task->pid;
 273
 274         if (in_group) {
 275                 struct task_struct *t;
 276
 277                 rcu_read_lock();
 278                 for_each_thread(task, t) {
 279                         if (vm_is_stack_for_task(t, vma)) {
 280                                 ret = t->pid;
 281                                 goto done;
 282                         }
 283                 }
 284 done:
 285                 rcu_read_unlock();
 286         }
 287
 288         return ret;
 289 }
 290
 291 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 292 void arch_pick_mmap_layout(struct mm_struct *mm)
 293 {
 294         mm->mmap_base = TASK_UNMAPPED_BASE;
 295         mm->get_unmapped_area = arch_get_unmapped_area;
 296 }
 297 #endif
 298
 299 /*
 300  * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 301  * back to the regular GUP.
 302  * If the architecture not support this function, simply return with no
 303  * page pinned
 304  */
 305 int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 306                                  int nr_pages, int write, struct page **pages)
 307 {
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 311
 312 /**
 313  * get_user_pages_fast() - pin user pages in memory
 314  * @start:      starting user address
 315  * @nr_pages:   number of pages from start to pin
 316  * @write:      whether pages will be written to
 317  * @pages:      array that receives pointers to the pages pinned.
 318  *              Should be at least nr_pages long.
 319  *
 320  * Returns number of pages pinned. This may be fewer than the number
 321  * requested. If nr_pages is 0 or negative, returns 0. If no pages
 322  * were pinned, returns -errno.
 323  *
 324  * get_user_pages_fast provides equivalent functionality to get_user_pages,
 325  * operating on current and current->mm, with force=0 and vma=NULL. However
 326  * unlike get_user_pages, it must be called without mmap_sem held.
 327  *
 328  * get_user_pages_fast may take mmap_sem and page table locks, so no
 329  * assumptions can be made about lack of locking. get_user_pages_fast is to be
 330  * implemented in a way that is advantageous (vs get_user_pages()) when the
 331  * user memory area is already faulted in and present in ptes. However if the
 332  * pages have to be faulted in, it may turn out to be slightly slower so
 333  * callers need to carefully consider what to use. On many architectures,
 334  * get_user_pages_fast simply falls back to get_user_pages.
 335  */
 336 int __attribute__((weak)) get_user_pages_fast(unsigned long start,
 337                                 int nr_pages, int write, struct page **pages)
 338 {
 339         struct mm_struct *mm = current->mm;
 340         int ret;
 341
 342         down_read(&mm->mmap_sem);
 343         ret = get_user_pages(current, mm, start, nr_pages,
 344                                         write, 0, pages, NULL);
 345         up_read(&mm->mmap_sem);
 346
 347         return ret;
 348 }
 349 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 350
 351 unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 352         unsigned long len, unsigned long prot,
 353         unsigned long flag, unsigned long pgoff)
 354 {
 355         unsigned long ret;
 356         struct mm_struct *mm = current->mm;
 357         unsigned long populate;
 358
 359         ret = security_mmap_file(file, prot, flag);
 360         if (!ret) {
 361                 down_write(&mm->mmap_sem);
 362                 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
 363                                     &populate);
 364                 up_write(&mm->mmap_sem);
 365                 if (populate)
 366                         mm_populate(ret, populate);
 367         }
 368         return ret;
 369 }
 370
 371 unsigned long vm_mmap(struct file *file, unsigned long addr,
 372         unsigned long len, unsigned long prot,
 373         unsigned long flag, unsigned long offset)
 374 {
 375         if (unlikely(offset + PAGE_ALIGN(len) < offset))
 376                 return -EINVAL;
 377         if (unlikely(offset & ~PAGE_MASK))
 378                 return -EINVAL;
 379
 380         return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
 381 }
 382 EXPORT_SYMBOL(vm_mmap);
 383
 384 void kvfree(const void *addr)
 385 {
 386         if (is_vmalloc_addr(addr))
 387                 vfree(addr);
 388         else
 389                 kfree(addr);
 390 }
 391 EXPORT_SYMBOL(kvfree);
 392
 393 struct address_space *page_mapping(struct page *page)
 394 {
 395         struct address_space *mapping = page->mapping;
 396
 397         /* This happens if someone calls flush_dcache_page on slab page */
 398         if (unlikely(PageSlab(page)))
 399                 return NULL;
 400
 401         if (unlikely(PageSwapCache(page))) {
 402                 swp_entry_t entry;
 403
 404                 entry.val = page_private(page);
 405                 mapping = swap_address_space(entry);
 406         } else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 407                 mapping = NULL;
 408         return mapping;
 409 }
 410
 411 /* Tracepoints definitions. */
 412 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 413 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 414 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 415 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 416 EXPORT_TRACEPOINT_SYMBOL(kfree);
 417 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);