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/mman.h>
  11 #include <linux/hugetlb.h>
  12
  13 #include <asm/uaccess.h>
  14
  15 #include "internal.h"
  16
  17 #define CREATE_TRACE_POINTS
  18 #include <trace/events/kmem.h>
  19
  20 /**
  21  * kstrdup - allocate space for and copy an existing string
  22  * @s: the string to duplicate
  23  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  24  */
  25 char *kstrdup(const char *s, gfp_t gfp)
  26 {
  27         size_t len;
  28         char *buf;
  29
  30         if (!s)
  31                 return NULL;
  32
  33         len = strlen(s) + 1;
  34         buf = kmalloc_track_caller(len, gfp);
  35         if (buf)
  36                 memcpy(buf, s, len);
  37         return buf;
  38 }
  39 EXPORT_SYMBOL(kstrdup);
  40
  41 /**
  42  * kstrndup - allocate space for and copy an existing string
  43  * @s: the string to duplicate
  44  * @max: read at most @max chars from @s
  45  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  46  */
  47 char *kstrndup(const char *s, size_t max, gfp_t gfp)
  48 {
  49         size_t len;
  50         char *buf;
  51
  52         if (!s)
  53                 return NULL;
  54
  55         len = strnlen(s, max);
  56         buf = kmalloc_track_caller(len+1, gfp);
  57         if (buf) {
  58                 memcpy(buf, s, len);
  59                 buf[len] = '\0';
  60         }
  61         return buf;
  62 }
  63 EXPORT_SYMBOL(kstrndup);
  64
  65 /**
  66  * kmemdup - duplicate region of memory
  67  *
  68  * @src: memory region to duplicate
  69  * @len: memory region length
  70  * @gfp: GFP mask to use
  71  */
  72 void *kmemdup(const void *src, size_t len, gfp_t gfp)
  73 {
  74         void *p;
  75
  76         p = kmalloc_track_caller(len, gfp);
  77         if (p)
  78                 memcpy(p, src, len);
  79         return p;
  80 }
  81 EXPORT_SYMBOL(kmemdup);
  82
  83 /**
  84  * memdup_user - duplicate memory region from user space
  85  *
  86  * @src: source address in user space
  87  * @len: number of bytes to copy
  88  *
  89  * Returns an ERR_PTR() on failure.
  90  */
  91 void *memdup_user(const void __user *src, size_t len)
  92 {
  93         void *p;
  94
  95         /*
  96          * Always use GFP_KERNEL, since copy_from_user() can sleep and
  97          * cause pagefault, which makes it pointless to use GFP_NOFS
  98          * or GFP_ATOMIC.
  99          */
 100         p = kmalloc_track_caller(len, GFP_KERNEL);
 101         if (!p)
 102                 return ERR_PTR(-ENOMEM);
 103
 104         if (copy_from_user(p, src, len)) {
 105                 kfree(p);
 106                 return ERR_PTR(-EFAULT);
 107         }
 108
 109         return p;
 110 }
 111 EXPORT_SYMBOL(memdup_user);
 112
 113 static __always_inline void *__do_krealloc(const void *p, size_t new_size,
 114                                            gfp_t flags)
 115 {
 116         void *ret;
 117         size_t ks = 0;
 118
 119         if (p)
 120                 ks = ksize(p);
 121
 122         if (ks >= new_size)
 123                 return (void *)p;
 124
 125         ret = kmalloc_track_caller(new_size, flags);
 126         if (ret && p)
 127                 memcpy(ret, p, ks);
 128
 129         return ret;
 130 }
 131
 132 /**
 133  * __krealloc - like krealloc() but don't free @p.
 134  * @p: object to reallocate memory for.
 135  * @new_size: how many bytes of memory are required.
 136  * @flags: the type of memory to allocate.
 137  *
 138  * This function is like krealloc() except it never frees the originally
 139  * allocated buffer. Use this if you don't want to free the buffer immediately
 140  * like, for example, with RCU.
 141  */
 142 void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 143 {
 144         if (unlikely(!new_size))
 145                 return ZERO_SIZE_PTR;
 146
 147         return __do_krealloc(p, new_size, flags);
 148
 149 }
 150 EXPORT_SYMBOL(__krealloc);
 151
 152 /**
 153  * krealloc - reallocate memory. The contents will remain unchanged.
 154  * @p: object to reallocate memory for.
 155  * @new_size: how many bytes of memory are required.
 156  * @flags: the type of memory to allocate.
 157  *
 158  * The contents of the object pointed to are preserved up to the
 159  * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 160  * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
 161  * %NULL pointer, the object pointed to is freed.
 162  */
 163 void *krealloc(const void *p, size_t new_size, gfp_t flags)
 164 {
 165         void *ret;
 166
 167         if (unlikely(!new_size)) {
 168                 kfree(p);
 169                 return ZERO_SIZE_PTR;
 170         }
 171
 172         ret = __do_krealloc(p, new_size, flags);
 173         if (ret && p != ret)
 174                 kfree(p);
 175
 176         return ret;
 177 }
 178 EXPORT_SYMBOL(krealloc);
 179
 180 /**
 181  * kzfree - like kfree but zero memory
 182  * @p: object to free memory of
 183  *
 184  * The memory of the object @p points to is zeroed before freed.
 185  * If @p is %NULL, kzfree() does nothing.
 186  *
 187  * Note: this function zeroes the whole allocated buffer which can be a good
 188  * deal bigger than the requested buffer size passed to kmalloc(). So be
 189  * careful when using this function in performance sensitive code.
 190  */
 191 void kzfree(const void *p)
 192 {
 193         size_t ks;
 194         void *mem = (void *)p;
 195
 196         if (unlikely(ZERO_OR_NULL_PTR(mem)))
 197                 return;
 198         ks = ksize(mem);
 199         memset(mem, 0, ks);
 200         kfree(mem);
 201 }
 202 EXPORT_SYMBOL(kzfree);
 203
 204 /*
 205  * strndup_user - duplicate an existing string from user space
 206  * @s: The string to duplicate
 207  * @n: Maximum number of bytes to copy, including the trailing NUL.
 208  */
 209 char *strndup_user(const char __user *s, long n)
 210 {
 211         char *p;
 212         long length;
 213
 214         length = strnlen_user(s, n);
 215
 216         if (!length)
 217                 return ERR_PTR(-EFAULT);
 218
 219         if (length > n)
 220                 return ERR_PTR(-EINVAL);
 221
 222         p = memdup_user(s, length);
 223
 224         if (IS_ERR(p))
 225                 return p;
 226
 227         p[length - 1] = '\0';
 228
 229         return p;
 230 }
 231 EXPORT_SYMBOL(strndup_user);
 232
 233 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 234                 struct vm_area_struct *prev, struct rb_node *rb_parent)
 235 {
 236         struct vm_area_struct *next;
 237
 238         vma->vm_prev = prev;
 239         if (prev) {
 240                 next = prev->vm_next;
 241                 prev->vm_next = vma;
 242         } else {
 243                 mm->mmap = vma;
 244                 if (rb_parent)
 245                         next = rb_entry(rb_parent,
 246                                         struct vm_area_struct, vm_rb);
 247                 else
 248                         next = NULL;
 249         }
 250         vma->vm_next = next;
 251         if (next)
 252                 next->vm_prev = vma;
 253 }
 254
 255 /* Check if the vma is being used as a stack by this task */
 256 static int vm_is_stack_for_task(struct task_struct *t,
 257                                 struct vm_area_struct *vma)
 258 {
 259         return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 260 }
 261
 262 /*
 263  * Check if the vma is being used as a stack.
 264  * If is_group is non-zero, check in the entire thread group or else
 265  * just check in the current task. Returns the pid of the task that
 266  * the vma is stack for.
 267  */
 268 pid_t vm_is_stack(struct task_struct *task,
 269                   struct vm_area_struct *vma, int in_group)
 270 {
 271         pid_t ret = 0;
 272
 273         if (vm_is_stack_for_task(task, vma))
 274                 return task->pid;
 275
 276         if (in_group) {
 277                 struct task_struct *t;
 278                 rcu_read_lock();
 279                 if (!pid_alive(task))
 280                         goto done;
 281
 282                 t = task;
 283                 do {
 284                         if (vm_is_stack_for_task(t, vma)) {
 285                                 ret = t->pid;
 286                                 goto done;
 287                         }
 288                 } while_each_thread(task, t);
 289 done:
 290                 rcu_read_unlock();
 291         }
 292
 293         return ret;
 294 }
 295
 296 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 297 void arch_pick_mmap_layout(struct mm_struct *mm)
 298 {
 299         mm->mmap_base = TASK_UNMAPPED_BASE;
 300         mm->get_unmapped_area = arch_get_unmapped_area;
 301 }
 302 #endif
 303
 304 /*
 305  * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 306  * back to the regular GUP.
 307  * If the architecture not support this function, simply return with no
 308  * page pinned
 309  */
 310 int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 311                                  int nr_pages, int write, struct page **pages)
 312 {
 313         return 0;
 314 }
 315 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 316
 317 /**
 318  * get_user_pages_fast() - pin user pages in memory
 319  * @start:      starting user address
 320  * @nr_pages:   number of pages from start to pin
 321  * @write:      whether pages will be written to
 322  * @pages:      array that receives pointers to the pages pinned.
 323  *              Should be at least nr_pages long.
 324  *
 325  * Returns number of pages pinned. This may be fewer than the number
 326  * requested. If nr_pages is 0 or negative, returns 0. If no pages
 327  * were pinned, returns -errno.
 328  *
 329  * get_user_pages_fast provides equivalent functionality to get_user_pages,
 330  * operating on current and current->mm, with force=0 and vma=NULL. However
 331  * unlike get_user_pages, it must be called without mmap_sem held.
 332  *
 333  * get_user_pages_fast may take mmap_sem and page table locks, so no
 334  * assumptions can be made about lack of locking. get_user_pages_fast is to be
 335  * implemented in a way that is advantageous (vs get_user_pages()) when the
 336  * user memory area is already faulted in and present in ptes. However if the
 337  * pages have to be faulted in, it may turn out to be slightly slower so
 338  * callers need to carefully consider what to use. On many architectures,
 339  * get_user_pages_fast simply falls back to get_user_pages.
 340  */
 341 int __attribute__((weak)) get_user_pages_fast(unsigned long start,
 342                                 int nr_pages, int write, struct page **pages)
 343 {
 344         struct mm_struct *mm = current->mm;
 345         int ret;
 346
 347         down_read(&mm->mmap_sem);
 348         ret = get_user_pages(current, mm, start, nr_pages,
 349                                         write, 0, pages, NULL);
 350         up_read(&mm->mmap_sem);
 351
 352         return ret;
 353 }
 354 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 355
 356 unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 357         unsigned long len, unsigned long prot,
 358         unsigned long flag, unsigned long pgoff)
 359 {
 360         unsigned long ret;
 361         struct mm_struct *mm = current->mm;
 362         unsigned long populate;
 363
 364         ret = security_mmap_file(file, prot, flag);
 365         if (!ret) {
 366                 down_write(&mm->mmap_sem);
 367                 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
 368                                     &populate);
 369                 up_write(&mm->mmap_sem);
 370                 if (populate)
 371                         mm_populate(ret, populate);
 372         }
 373         return ret;
 374 }
 375
 376 unsigned long vm_mmap(struct file *file, unsigned long addr,
 377         unsigned long len, unsigned long prot,
 378         unsigned long flag, unsigned long offset)
 379 {
 380         if (unlikely(offset + PAGE_ALIGN(len) < offset))
 381                 return -EINVAL;
 382         if (unlikely(offset & ~PAGE_MASK))
 383                 return -EINVAL;
 384
 385         return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
 386 }
 387 EXPORT_SYMBOL(vm_mmap);
 388
 389 struct address_space *page_mapping(struct page *page)
 390 {
 391         struct address_space *mapping = page->mapping;
 392
 393         /* This happens if someone calls flush_dcache_page on slab page */
 394         if (unlikely(PageSlab(page)))
 395                 return NULL;
 396
 397         if (unlikely(PageSwapCache(page))) {
 398                 swp_entry_t entry;
 399
 400                 entry.val = page_private(page);
 401                 mapping = swap_address_space(entry);
 402         } else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 403                 mapping = NULL;
 404         return mapping;
 405 }
 406
 407 int overcommit_ratio_handler(struct ctl_table *table, int write,
 408                              void __user *buffer, size_t *lenp,
 409                              loff_t *ppos)
 410 {
 411         int ret;
 412
 413         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 414         if (ret == 0 && write)
 415                 sysctl_overcommit_kbytes = 0;
 416         return ret;
 417 }
 418
 419 int overcommit_kbytes_handler(struct ctl_table *table, int write,
 420                              void __user *buffer, size_t *lenp,
 421                              loff_t *ppos)
 422 {
 423         int ret;
 424
 425         ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
 426         if (ret == 0 && write)
 427                 sysctl_overcommit_ratio = 0;
 428         return ret;
 429 }
 430
 431 /*
 432  * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
 433  */
 434 unsigned long vm_commit_limit(void)
 435 {
 436         unsigned long allowed;
 437
 438         if (sysctl_overcommit_kbytes)
 439                 allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
 440         else
 441                 allowed = ((totalram_pages - hugetlb_total_pages())
 442                            * sysctl_overcommit_ratio / 100);
 443         allowed += total_swap_pages;
 444
 445         return allowed;
 446 }
 447
 448
 449 /* Tracepoints definitions. */
 450 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 451 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 452 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 453 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 454 EXPORT_TRACEPOINT_SYMBOL(kfree);
 455 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);