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