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 <asm/uaccess.h>
   8
   9 #include "internal.h"
  10
  11 #define CREATE_TRACE_POINTS
  12 #include <trace/events/kmem.h>
  13
  14 /**
  15  * kstrdup - allocate space for and copy an existing string
  16  * @s: the string to duplicate
  17  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  18  */
  19 char *kstrdup(const char *s, gfp_t gfp)
  20 {
  21         size_t len;
  22         char *buf;
  23
  24         if (!s)
  25                 return NULL;
  26
  27         len = strlen(s) + 1;
  28         buf = kmalloc_track_caller(len, gfp);
  29         if (buf)
  30                 memcpy(buf, s, len);
  31         return buf;
  32 }
  33 EXPORT_SYMBOL(kstrdup);
  34
  35 /**
  36  * kstrndup - allocate space for and copy an existing string
  37  * @s: the string to duplicate
  38  * @max: read at most @max chars from @s
  39  * @gfp: the GFP mask used in the kmalloc() call when allocating memory
  40  */
  41 char *kstrndup(const char *s, size_t max, gfp_t gfp)
  42 {
  43         size_t len;
  44         char *buf;
  45
  46         if (!s)
  47                 return NULL;
  48
  49         len = strnlen(s, max);
  50         buf = kmalloc_track_caller(len+1, gfp);
  51         if (buf) {
  52                 memcpy(buf, s, len);
  53                 buf[len] = '\0';
  54         }
  55         return buf;
  56 }
  57 EXPORT_SYMBOL(kstrndup);
  58
  59 /**
  60  * kmemdup - duplicate region of memory
  61  *
  62  * @src: memory region to duplicate
  63  * @len: memory region length
  64  * @gfp: GFP mask to use
  65  */
  66 void *kmemdup(const void *src, size_t len, gfp_t gfp)
  67 {
  68         void *p;
  69
  70         p = kmalloc_track_caller(len, gfp);
  71         if (p)
  72                 memcpy(p, src, len);
  73         return p;
  74 }
  75 EXPORT_SYMBOL(kmemdup);
  76
  77 /**
  78  * memdup_user - duplicate memory region from user space
  79  *
  80  * @src: source address in user space
  81  * @len: number of bytes to copy
  82  *
  83  * Returns an ERR_PTR() on failure.
  84  */
  85 void *memdup_user(const void __user *src, size_t len)
  86 {
  87         void *p;
  88
  89         /*
  90          * Always use GFP_KERNEL, since copy_from_user() can sleep and
  91          * cause pagefault, which makes it pointless to use GFP_NOFS
  92          * or GFP_ATOMIC.
  93          */
  94         p = kmalloc_track_caller(len, GFP_KERNEL);
  95         if (!p)
  96                 return ERR_PTR(-ENOMEM);
  97
  98         if (copy_from_user(p, src, len)) {
  99                 kfree(p);
 100                 return ERR_PTR(-EFAULT);
 101         }
 102
 103         return p;
 104 }
 105 EXPORT_SYMBOL(memdup_user);
 106
 107 /**
 108  * __krealloc - like krealloc() but don't free @p.
 109  * @p: object to reallocate memory for.
 110  * @new_size: how many bytes of memory are required.
 111  * @flags: the type of memory to allocate.
 112  *
 113  * This function is like krealloc() except it never frees the originally
 114  * allocated buffer. Use this if you don't want to free the buffer immediately
 115  * like, for example, with RCU.
 116  */
 117 void *__krealloc(const void *p, size_t new_size, gfp_t flags)
 118 {
 119         void *ret;
 120         size_t ks = 0;
 121
 122         if (unlikely(!new_size))
 123                 return ZERO_SIZE_PTR;
 124
 125         if (p)
 126                 ks = ksize(p);
 127
 128         if (ks >= new_size)
 129                 return (void *)p;
 130
 131         ret = kmalloc_track_caller(new_size, flags);
 132         if (ret && p)
 133                 memcpy(ret, p, ks);
 134
 135         return ret;
 136 }
 137 EXPORT_SYMBOL(__krealloc);
 138
 139 /**
 140  * krealloc - reallocate memory. The contents will remain unchanged.
 141  * @p: object to reallocate memory for.
 142  * @new_size: how many bytes of memory are required.
 143  * @flags: the type of memory to allocate.
 144  *
 145  * The contents of the object pointed to are preserved up to the
 146  * lesser of the new and old sizes.  If @p is %NULL, krealloc()
 147  * behaves exactly like kmalloc().  If @size is 0 and @p is not a
 148  * %NULL pointer, the object pointed to is freed.
 149  */
 150 void *krealloc(const void *p, size_t new_size, gfp_t flags)
 151 {
 152         void *ret;
 153
 154         if (unlikely(!new_size)) {
 155                 kfree(p);
 156                 return ZERO_SIZE_PTR;
 157         }
 158
 159         ret = __krealloc(p, new_size, flags);
 160         if (ret && p != ret)
 161                 kfree(p);
 162
 163         return ret;
 164 }
 165 EXPORT_SYMBOL(krealloc);
 166
 167 /**
 168  * kzfree - like kfree but zero memory
 169  * @p: object to free memory of
 170  *
 171  * The memory of the object @p points to is zeroed before freed.
 172  * If @p is %NULL, kzfree() does nothing.
 173  *
 174  * Note: this function zeroes the whole allocated buffer which can be a good
 175  * deal bigger than the requested buffer size passed to kmalloc(). So be
 176  * careful when using this function in performance sensitive code.
 177  */
 178 void kzfree(const void *p)
 179 {
 180         size_t ks;
 181         void *mem = (void *)p;
 182
 183         if (unlikely(ZERO_OR_NULL_PTR(mem)))
 184                 return;
 185         ks = ksize(mem);
 186         memset(mem, 0, ks);
 187         kfree(mem);
 188 }
 189 EXPORT_SYMBOL(kzfree);
 190
 191 /*
 192  * strndup_user - duplicate an existing string from user space
 193  * @s: The string to duplicate
 194  * @n: Maximum number of bytes to copy, including the trailing NUL.
 195  */
 196 char *strndup_user(const char __user *s, long n)
 197 {
 198         char *p;
 199         long length;
 200
 201         length = strnlen_user(s, n);
 202
 203         if (!length)
 204                 return ERR_PTR(-EFAULT);
 205
 206         if (length > n)
 207                 return ERR_PTR(-EINVAL);
 208
 209         p = memdup_user(s, length);
 210
 211         if (IS_ERR(p))
 212                 return p;
 213
 214         p[length - 1] = '\0';
 215
 216         return p;
 217 }
 218 EXPORT_SYMBOL(strndup_user);
 219
 220 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 221                 struct vm_area_struct *prev, struct rb_node *rb_parent)
 222 {
 223         struct vm_area_struct *next;
 224
 225         vma->vm_prev = prev;
 226         if (prev) {
 227                 next = prev->vm_next;
 228                 prev->vm_next = vma;
 229         } else {
 230                 mm->mmap = vma;
 231                 if (rb_parent)
 232                         next = rb_entry(rb_parent,
 233                                         struct vm_area_struct, vm_rb);
 234                 else
 235                         next = NULL;
 236         }
 237         vma->vm_next = next;
 238         if (next)
 239                 next->vm_prev = vma;
 240 }
 241
 242 /* Check if the vma is being used as a stack by this task */
 243 static int vm_is_stack_for_task(struct task_struct *t,
 244                                 struct vm_area_struct *vma)
 245 {
 246         return (vma->vm_start <= KSTK_ESP(t) && vma->vm_end >= KSTK_ESP(t));
 247 }
 248
 249 /*
 250  * Check if the vma is being used as a stack.
 251  * If is_group is non-zero, check in the entire thread group or else
 252  * just check in the current task. Returns the pid of the task that
 253  * the vma is stack for.
 254  */
 255 pid_t vm_is_stack(struct task_struct *task,
 256                   struct vm_area_struct *vma, int in_group)
 257 {
 258         pid_t ret = 0;
 259
 260         if (vm_is_stack_for_task(task, vma))
 261                 return task->pid;
 262
 263         if (in_group) {
 264                 struct task_struct *t;
 265                 rcu_read_lock();
 266                 if (!pid_alive(task))
 267                         goto done;
 268
 269                 t = task;
 270                 do {
 271                         if (vm_is_stack_for_task(t, vma)) {
 272                                 ret = t->pid;
 273                                 goto done;
 274                         }
 275                 } while_each_thread(task, t);
 276 done:
 277                 rcu_read_unlock();
 278         }
 279
 280         return ret;
 281 }
 282
 283 #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT)
 284 void arch_pick_mmap_layout(struct mm_struct *mm)
 285 {
 286         mm->mmap_base = TASK_UNMAPPED_BASE;
 287         mm->get_unmapped_area = arch_get_unmapped_area;
 288         mm->unmap_area = arch_unmap_area;
 289 }
 290 #endif
 291
 292 /*
 293  * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 294  * back to the regular GUP.
 295  * If the architecture not support this function, simply return with no
 296  * page pinned
 297  */
 298 int __attribute__((weak)) __get_user_pages_fast(unsigned long start,
 299                                  int nr_pages, int write, struct page **pages)
 300 {
 301         return 0;
 302 }
 303 EXPORT_SYMBOL_GPL(__get_user_pages_fast);
 304
 305 /**
 306  * get_user_pages_fast() - pin user pages in memory
 307  * @start:      starting user address
 308  * @nr_pages:   number of pages from start to pin
 309  * @write:      whether pages will be written to
 310  * @pages:      array that receives pointers to the pages pinned.
 311  *              Should be at least nr_pages long.
 312  *
 313  * Returns number of pages pinned. This may be fewer than the number
 314  * requested. If nr_pages is 0 or negative, returns 0. If no pages
 315  * were pinned, returns -errno.
 316  *
 317  * get_user_pages_fast provides equivalent functionality to get_user_pages,
 318  * operating on current and current->mm, with force=0 and vma=NULL. However
 319  * unlike get_user_pages, it must be called without mmap_sem held.
 320  *
 321  * get_user_pages_fast may take mmap_sem and page table locks, so no
 322  * assumptions can be made about lack of locking. get_user_pages_fast is to be
 323  * implemented in a way that is advantageous (vs get_user_pages()) when the
 324  * user memory area is already faulted in and present in ptes. However if the
 325  * pages have to be faulted in, it may turn out to be slightly slower so
 326  * callers need to carefully consider what to use. On many architectures,
 327  * get_user_pages_fast simply falls back to get_user_pages.
 328  */
 329 int __attribute__((weak)) get_user_pages_fast(unsigned long start,
 330                                 int nr_pages, int write, struct page **pages)
 331 {
 332         struct mm_struct *mm = current->mm;
 333         int ret;
 334
 335         down_read(&mm->mmap_sem);
 336         ret = get_user_pages(current, mm, start, nr_pages,
 337                                         write, 0, pages, NULL);
 338         up_read(&mm->mmap_sem);
 339
 340         return ret;
 341 }
 342 EXPORT_SYMBOL_GPL(get_user_pages_fast);
 343
 344 /* Tracepoints definitions. */
 345 EXPORT_TRACEPOINT_SYMBOL(kmalloc);
 346 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc);
 347 EXPORT_TRACEPOINT_SYMBOL(kmalloc_node);
 348 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_alloc_node);
 349 EXPORT_TRACEPOINT_SYMBOL(kfree);
 350 EXPORT_TRACEPOINT_SYMBOL(kmem_cache_free);