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 <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  */
  22 char *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 }
  36 EXPORT_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  */
  44 char *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 }
  60 EXPORT_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  */
  69 void *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 }
  78 EXPORT_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  */
  88 void *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 }
 108 EXPORT_SYMBOL(memdup_user);
 109
 110 static __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  */
 139 void *__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 }
 147 EXPORT_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  */
 160 void *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 }
 175 EXPORT_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  */
 188 void 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 }
 199 EXPORT_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  */
 206 char *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 }
 228 EXPORT_SYMBOL(strndup_user);
 229
 230 void __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 */
 253 static 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  */
 265 pid_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);
 286 done:
 287                 rcu_read_unlock();
 288         }
 289
 290         return ret;
 291 }
 292
 293 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 294 void 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         mm->unmap_area = arch_unmap_area;
 299 }
 300 #endif
 301
 302 /*
 303  * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 304  * back to the regular GUP.
 305  * If the architecture not support this function, simply return with no
 306  * page pinned
 307  */
 308 int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 309                                  int nr_pages, int write, struct page **pages)
 310 {
 311         return 0;
 312 }
 313 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 314
 315 /**
 316  * get_user_pages_fast() - pin user pages in memory
 317  * @start:      starting user address
 318  * @nr_pages:   number of pages from start to pin
 319  * @write:      whether pages will be written to
 320  * @pages:      array that receives pointers to the pages pinned.
 321  *              Should be at least nr_pages long.
 322  *
 323  * Returns number of pages pinned. This may be fewer than the number
 324  * requested. If nr_pages is 0 or negative, returns 0. If no pages
 325  * were pinned, returns -errno.
 326  *
 327  * get_user_pages_fast provides equivalent functionality to get_user_pages,
 328  * operating on current and current->mm, with force=0 and vma=NULL. However
 329  * unlike get_user_pages, it must be called without mmap_sem held.
 330  *
 331  * get_user_pages_fast may take mmap_sem and page table locks, so no
 332  * assumptions can be made about lack of locking. get_user_pages_fast is to be
 333  * implemented in a way that is advantageous (vs get_user_pages()) when the
 334  * user memory area is already faulted in and present in ptes. However if the
 335  * pages have to be faulted in, it may turn out to be slightly slower so
 336  * callers need to carefully consider what to use. On many architectures,
 337  * get_user_pages_fast simply falls back to get_user_pages.
 338  */
 339 int __attribute__((weak)) get_user_pages_fast(unsigned long start,
 340                                 int nr_pages, int write, struct page **pages)
 341 {
 342         struct mm_struct *mm = current->mm;
 343         int ret;
 344
 345         down_read(&mm->mmap_sem);
 346         ret = get_user_pages(current, mm, start, nr_pages,
 347                                         write, 0, pages, NULL);
 348         up_read(&mm->mmap_sem);
 349
 350         return ret;
 351 }
 352 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 353
 354 unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 355         unsigned long len, unsigned long prot,
 356         unsigned long flag, unsigned long pgoff)
 357 {
 358         unsigned long ret;
 359         struct mm_struct *mm = current->mm;
 360         unsigned long populate;
 361
 362         ret = security_mmap_file(file, prot, flag);
 363         if (!ret) {
 364                 down_write(&mm->mmap_sem);
 365                 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
 366                                     &populate);
 367                 up_write(&mm->mmap_sem);
 368                 if (populate)
 369                         mm_populate(ret, populate);
 370         }
 371         return ret;
 372 }
 373
 374 unsigned long vm_mmap(struct file *file, unsigned long addr,
 375         unsigned long len, unsigned long prot,
 376         unsigned long flag, unsigned long offset)
 377 {
 378         if (unlikely(offset + PAGE_ALIGN(len) < offset))
 379                 return -EINVAL;
 380         if (unlikely(offset & ~PAGE_MASK))
 381                 return -EINVAL;
 382
 383         return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
 384 }
 385 EXPORT_SYMBOL(vm_mmap);
 386
 387 struct address_space *page_mapping(struct page *page)
 388 {
 389         struct address_space *mapping = page->mapping;
 390
 391         VM_BUG_ON(PageSlab(page));
 392 #ifdef CONFIG_SWAP
 393         if (unlikely(PageSwapCache(page))) {
 394                 swp_entry_t entry;
 395
 396                 entry.val = page_private(page);
 397                 mapping = swap_address_space(entry);
 398         } else
 399 #endif
 400         if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 401                 mapping = NULL;
 402         return mapping;
 403 }
 404
 405 /* Tracepoints definitions. */
 406 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 407 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 408 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 409 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 410 EXPORT_TRACEPOINT_SYMBOL(kfree);
 411 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);