mm/util.c

   1 #include <linux/mm.h>
   2 #include <linux/slab.h>
   3 #include <linux/string.h>
   4 #include <linux/compiler.h>
   5 #include <linux/export.h>
   6 #include <linux/err.h>
   7 #include <linux/sched.h>
   8 #include <linux/security.h>
   9 #include <linux/swap.h>
  10 #include <linux/swapops.h>
  11 #include <linux/mman.h>
  12 #include <linux/hugetlb.h>
  13 #include <linux/vmalloc.h>
  14
  15 #include <asm/uaccess.h>
  16
  17 #include "internal.h"
  18
  19 #define CREATE_TRACE_POINTS
  20 #include <trace/events/kmem.h>
  21
  22 /**
  23  * kstrdup - allocate space for and copy an existing string
  24  * @s: the string to duplicate
  25  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  26  */
  27 char *kstrdup(const char *s, gfp_t gfp)
  28 {
  29         size_t len;
  30         char *buf;
  31
  32         if (!s)
  33                 return NULL;
  34
  35         len = strlen(s) + 1;
  36         buf = kmalloc_track_caller(len, gfp);
  37         if (buf)
  38                 memcpy(buf, s, len);
  39         return buf;
  40 }
  41 EXPORT_SYMBOL(kstrdup);
  42
  43 /**
  44  * kstrndup - allocate space for and copy an existing string
  45  * @s: the string to duplicate
  46  * @max: read at most @max chars from @s
  47  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  48  */
  49 char *kstrndup(const char *s, size_t max, gfp_t gfp)
  50 {
  51         size_t len;
  52         char *buf;
  53
  54         if (!s)
  55                 return NULL;
  56
  57         len = strnlen(s, max);
  58         buf = kmalloc_track_caller(len+1, gfp);
  59         if (buf) {
  60                 memcpy(buf, s, len);
  61                 buf[len] = '\0';
  62         }
  63         return buf;
  64 }
  65 EXPORT_SYMBOL(kstrndup);
  66
  67 /**
  68  * kmemdup - duplicate region of memory
  69  *
  70  * @src: memory region to duplicate
  71  * @len: memory region length
  72  * @gfp: GFP mask to use
  73  */
  74 void *kmemdup(const void *src, size_t len, gfp_t gfp)
  75 {
  76         void *p;
  77
  78         p = kmalloc_track_caller(len, gfp);
  79         if (p)
  80                 memcpy(p, src, len);
  81         return p;
  82 }
  83 EXPORT_SYMBOL(kmemdup);
  84
  85 /**
  86  * memdup_user - duplicate memory region from user space
  87  *
  88  * @src: source address in user space
  89  * @len: number of bytes to copy
  90  *
  91  * Returns an ERR_PTR() on failure.
  92  */
  93 void *memdup_user(const void __user *src, size_t len)
  94 {
  95         void *p;
  96
  97         /*
  98          * Always use GFP_KERNEL, since copy_from_user() can sleep and
  99          * cause pagefault, which makes it pointless to use GFP_NOFS
 100          * or GFP_ATOMIC.
 101          */
 102         p = kmalloc_track_caller(len, GFP_KERNEL);
 103         if (!p)
 104                 return ERR_PTR(-ENOMEM);
 105
 106         if (copy_from_user(p, src, len)) {
 107                 kfree(p);
 108                 return ERR_PTR(-EFAULT);
 109         }
 110
 111         return p;
 112 }
 113 EXPORT_SYMBOL(memdup_user);
 114
 115 static __always_inline void *__do_krealloc(const void *p, size_t new_size,
 116                                            gfp_t flags)
 117 {
 118         void *ret;
 119         size_t ks = 0;
 120
 121         if (p)
 122                 ks = ksize(p);
 123
 124         if (ks >= new_size)
 125                 return (void *)p;
 126
 127         ret = kmalloc_track_caller(new_size, flags);
 128         if (ret && p)
 129                 memcpy(ret, p, ks);
 130
 131         return ret;
 132 }
 133
 134 /**
 135  * __krealloc - like krealloc() but don't free @p.
 136  * @p: object to reallocate memory for.
 137  * @new_size: how many bytes of memory are required.
 138  * @flags: the type of memory to allocate.
 139  *
 140  * This function is like krealloc() except it never frees the originally
 141  * allocated buffer. Use this if you don't want to free the buffer immediately
 142  * like, for example, with RCU.
 143  */
 144 void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 145 {
 146         if (unlikely(!new_size))
 147                 return ZERO_SIZE_PTR;
 148
 149         return __do_krealloc(p, new_size, flags);
 150
 151 }
 152 EXPORT_SYMBOL(__krealloc);
 153
 154 /**
 155  * krealloc - reallocate memory. The contents will remain unchanged.
 156  * @p: object to reallocate memory for.
 157  * @new_size: how many bytes of memory are required.
 158  * @flags: the type of memory to allocate.
 159  *
 160  * The contents of the object pointed to are preserved up to the
 161  * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 162  * behaves exactly like kmalloc().  If @new_size is 0 and @p is not a
 163  * %NULL pointer, the object pointed to is freed.
 164  */
 165 void *krealloc(const void *p, size_t new_size, gfp_t flags)
 166 {
 167         void *ret;
 168
 169         if (unlikely(!new_size)) {
 170                 kfree(p);
 171                 return ZERO_SIZE_PTR;
 172         }
 173
 174         ret = __do_krealloc(p, new_size, flags);
 175         if (ret && p != ret)
 176                 kfree(p);
 177
 178         return ret;
 179 }
 180 EXPORT_SYMBOL(krealloc);
 181
 182 /**
 183  * kzfree - like kfree but zero memory
 184  * @p: object to free memory of
 185  *
 186  * The memory of the object @p points to is zeroed before freed.
 187  * If @p is %NULL, kzfree() does nothing.
 188  *
 189  * Note: this function zeroes the whole allocated buffer which can be a good
 190  * deal bigger than the requested buffer size passed to kmalloc(). So be
 191  * careful when using this function in performance sensitive code.
 192  */
 193 void kzfree(const void *p)
 194 {
 195         size_t ks;
 196         void *mem = (void *)p;
 197
 198         if (unlikely(ZERO_OR_NULL_PTR(mem)))
 199                 return;
 200         ks = ksize(mem);
 201         memset(mem, 0, ks);
 202         kfree(mem);
 203 }
 204 EXPORT_SYMBOL(kzfree);
 205
 206 /*
 207  * strndup_user - duplicate an existing string from user space
 208  * @s: The string to duplicate
 209  * @n: Maximum number of bytes to copy, including the trailing NUL.
 210  */
 211 char *strndup_user(const char __user *s, long n)
 212 {
 213         char *p;
 214         long length;
 215
 216         length = strnlen_user(s, n);
 217
 218         if (!length)
 219                 return ERR_PTR(-EFAULT);
 220
 221         if (length > n)
 222                 return ERR_PTR(-EINVAL);
 223
 224         p = memdup_user(s, length);
 225
 226         if (IS_ERR(p))
 227                 return p;
 228
 229         p[length - 1] = '\0';
 230
 231         return p;
 232 }
 233 EXPORT_SYMBOL(strndup_user);
 234
 235 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 236                 struct vm_area_struct *prev, struct rb_node *rb_parent)
 237 {
 238         struct vm_area_struct *next;
 239
 240         vma->vm_prev = prev;
 241         if (prev) {
 242                 next = prev->vm_next;
 243                 prev->vm_next = vma;
 244         } else {
 245                 mm->mmap = vma;
 246                 if (rb_parent)
 247                         next = rb_entry(rb_parent,
 248                                         struct vm_area_struct, vm_rb);
 249                 else
 250                         next = NULL;
 251         }
 252         vma->vm_next = next;
 253         if (next)
 254                 next->vm_prev = vma;
 255 }
 256
 257 /* Check if the vma is being used as a stack by this task */
 258 static int vm_is_stack_for_task(struct task_struct *t,
 259                                 struct vm_area_struct *vma)
 260 {
 261         return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 262 }
 263
 264 /*
 265  * Check if the vma is being used as a stack.
 266  * If is_group is non-zero, check in the entire thread group or else
 267  * just check in the current task. Returns the pid of the task that
 268  * the vma is stack for.
 269  */
 270 pid_t vm_is_stack(struct task_struct *task,
 271                   struct vm_area_struct *vma, int in_group)
 272 {
 273         pid_t ret = 0;
 274
 275         if (vm_is_stack_for_task(task, vma))
 276                 return task->pid;
 277
 278         if (in_group) {
 279                 struct task_struct *t;
 280                 rcu_read_lock();
 281                 if (!pid_alive(task))
 282                         goto done;
 283
 284                 t = task;
 285                 do {
 286                         if (vm_is_stack_for_task(t, vma)) {
 287                                 ret = t->pid;
 288                                 goto done;
 289                         }
 290                 } while_each_thread(task, t);
 291 done:
 292                 rcu_read_unlock();
 293         }
 294
 295         return ret;
 296 }
 297
 298 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 299 void arch_pick_mmap_layout(struct mm_struct *mm)
 300 {
 301         mm->mmap_base = TASK_UNMAPPED_BASE;
 302         mm->get_unmapped_area = arch_get_unmapped_area;
 303 }
 304 #endif
 305
 306 /*
 307  * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 308  * back to the regular GUP.
 309  * If the architecture not support this function, simply return with no
 310  * page pinned
 311  */
 312 int __weak __get_user_pages_fast(unsigned long start,
 313                                  int nr_pages, int write, struct page **pages)
 314 {
 315         return 0;
 316 }
 317 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 318
 319 /**
 320  * get_user_pages_fast() - pin user pages in memory
 321  * @start:      starting user address
 322  * @nr_pages:   number of pages from start to pin
 323  * @write:      whether pages will be written to
 324  * @pages:      array that receives pointers to the pages pinned.
 325  *              Should be at least nr_pages long.
 326  *
 327  * Returns number of pages pinned. This may be fewer than the number
 328  * requested. If nr_pages is 0 or negative, returns 0. If no pages
 329  * were pinned, returns -errno.
 330  *
 331  * get_user_pages_fast provides equivalent functionality to get_user_pages,
 332  * operating on current and current->mm, with force=0 and vma=NULL. However
 333  * unlike get_user_pages, it must be called without mmap_sem held.
 334  *
 335  * get_user_pages_fast may take mmap_sem and page table locks, so no
 336  * assumptions can be made about lack of locking. get_user_pages_fast is to be
 337  * implemented in a way that is advantageous (vs get_user_pages()) when the
 338  * user memory area is already faulted in and present in ptes. However if the
 339  * pages have to be faulted in, it may turn out to be slightly slower so
 340  * callers need to carefully consider what to use. On many architectures,
 341  * get_user_pages_fast simply falls back to get_user_pages.
 342  */
 343 int __weak get_user_pages_fast(unsigned long start,
 344                                 int nr_pages, int write, struct page **pages)
 345 {
 346         struct mm_struct *mm = current->mm;
 347         int ret;
 348
 349         down_read(&mm->mmap_sem);
 350         ret = get_user_pages(current, mm, start, nr_pages,
 351                                         write, 0, pages, NULL);
 352         up_read(&mm->mmap_sem);
 353
 354         return ret;
 355 }
 356 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 357
 358 unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 359         unsigned long len, unsigned long prot,
 360         unsigned long flag, unsigned long pgoff)
 361 {
 362         unsigned long ret;
 363         struct mm_struct *mm = current->mm;
 364         unsigned long populate;
 365
 366         ret = security_mmap_file(file, prot, flag);
 367         if (!ret) {
 368                 down_write(&mm->mmap_sem);
 369                 ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,
 370                                     &populate);
 371                 up_write(&mm->mmap_sem);
 372                 if (populate)
 373                         mm_populate(ret, populate);
 374         }
 375         return ret;
 376 }
 377
 378 unsigned long vm_mmap(struct file *file, unsigned long addr,
 379         unsigned long len, unsigned long prot,
 380         unsigned long flag, unsigned long offset)
 381 {
 382         if (unlikely(offset + PAGE_ALIGN(len) < offset))
 383                 return -EINVAL;
 384         if (unlikely(offset & ~PAGE_MASK))
 385                 return -EINVAL;
 386
 387         return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
 388 }
 389 EXPORT_SYMBOL(vm_mmap);
 390
 391 void kvfree(const void *addr)
 392 {
 393         if (is_vmalloc_addr(addr))
 394                 vfree(addr);
 395         else
 396                 kfree(addr);
 397 }
 398 EXPORT_SYMBOL(kvfree);
 399
 400 struct address_space *page_mapping(struct page *page)
 401 {
 402         struct address_space *mapping = page->mapping;
 403
 404         /* This happens if someone calls flush_dcache_page on slab page */
 405         if (unlikely(PageSlab(page)))
 406                 return NULL;
 407
 408         if (unlikely(PageSwapCache(page))) {
 409                 swp_entry_t entry;
 410
 411                 entry.val = page_private(page);
 412                 mapping = swap_address_space(entry);
 413         } else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 414                 mapping = NULL;
 415         return mapping;
 416 }
 417
 418 int overcommit_ratio_handler(struct ctl_table *table, int write,
 419                              void __user *buffer, size_t *lenp,
 420                              loff_t *ppos)
 421 {
 422         int ret;
 423
 424         ret = proc_dointvec(table, write, buffer, lenp, ppos);
 425         if (ret == 0 && write)
 426                 sysctl_overcommit_kbytes = 0;
 427         return ret;
 428 }
 429
 430 int overcommit_kbytes_handler(struct ctl_table *table, int write,
 431                              void __user *buffer, size_t *lenp,
 432                              loff_t *ppos)
 433 {
 434         int ret;
 435
 436         ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos);
 437         if (ret == 0 && write)
 438                 sysctl_overcommit_ratio = 0;
 439         return ret;
 440 }
 441
 442 /*
 443  * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used
 444  */
 445 unsigned long vm_commit_limit(void)
 446 {
 447         unsigned long allowed;
 448
 449         if (sysctl_overcommit_kbytes)
 450                 allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10);
 451         else
 452                 allowed = ((totalram_pages - hugetlb_total_pages())
 453                            * sysctl_overcommit_ratio / 100);
 454         allowed += total_swap_pages;
 455
 456         return allowed;
 457 }
 458
 459 /**
 460  * get_cmdline() - copy the cmdline value to a buffer.
 461  * @task:     the task whose cmdline value to copy.
 462  * @buffer:   the buffer to copy to.
 463  * @buflen:   the length of the buffer. Larger cmdline values are truncated
 464  *            to this length.
 465  * Returns the size of the cmdline field copied. Note that the copy does
 466  * not guarantee an ending NULL byte.
 467  */
 468 int get_cmdline(struct task_struct *task, char *buffer, int buflen)
 469 {
 470         int res = 0;
 471         unsigned int len;
 472         struct mm_struct *mm = get_task_mm(task);
 473         if (!mm)
 474                 goto out;
 475         if (!mm->arg_end)
 476                 goto out_mm;    /* Shh! No looking before we're done */
 477
 478         len = mm->arg_end - mm->arg_start;
 479
 480         if (len > buflen)
 481                 len = buflen;
 482
 483         res = access_process_vm(task, mm->arg_start, buffer, len, 0);
 484
 485         /*
 486          * If the nul at the end of args has been overwritten, then
 487          * assume application is using setproctitle(3).
 488          */
 489         if (res > 0 && buffer[res-1] != '\0' && len < buflen) {
 490                 len = strnlen(buffer, res);
 491                 if (len < res) {
 492                         res = len;
 493                 } else {
 494                         len = mm->env_end - mm->env_start;
 495                         if (len > buflen - res)
 496                                 len = buflen - res;
 497                         res += access_process_vm(task, mm->env_start,
 498                                                  buffer+res, len, 0);
 499                         res = strnlen(buffer, res);
 500                 }
 501         }
 502 out_mm:
 503         mmput(mm);
 504 out:
 505         return res;
 506 }
 507
 508 /* Tracepoints definitions. */
 509 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 510 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 511 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 512 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 513 EXPORT_TRACEPOINT_SYMBOL(kfree);
 514 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);