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