| 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 task_struct of the task |
| 174 | * that the vma is stack for. Must be called under rcu_read_lock(). |
| 175 | */ |
| 176 | struct task_struct *task_of_stack(struct task_struct *task, |
| 177 | struct vm_area_struct *vma, bool in_group) |
| 178 | { |
| 179 | if (vm_is_stack_for_task(task, vma)) |
| 180 | return task; |
| 181 | |
| 182 | if (in_group) { |
| 183 | struct task_struct *t; |
| 184 | |
| 185 | for_each_thread(task, t) { |
| 186 | if (vm_is_stack_for_task(t, vma)) |
| 187 | return t; |
| 188 | } |
| 189 | } |
| 190 | |
| 191 | return NULL; |
| 192 | } |
| 193 | |
| 194 | #if defined(CONFIG_MMU) && !defined(HAVE_ARCH_PICK_MMAP_LAYOUT) |
| 195 | void arch_pick_mmap_layout(struct mm_struct *mm) |
| 196 | { |
| 197 | mm->mmap_base = TASK_UNMAPPED_BASE; |
| 198 | mm->get_unmapped_area = arch_get_unmapped_area; |
| 199 | } |
| 200 | #endif |
| 201 | |
| 202 | /* |
| 203 | * Like get_user_pages_fast() except its IRQ-safe in that it won't fall |
| 204 | * back to the regular GUP. |
| 205 | * If the architecture not support this function, simply return with no |
| 206 | * page pinned |
| 207 | */ |
| 208 | int __weak __get_user_pages_fast(unsigned long start, |
| 209 | int nr_pages, int write, struct page **pages) |
| 210 | { |
| 211 | return 0; |
| 212 | } |
| 213 | EXPORT_SYMBOL_GPL(__get_user_pages_fast); |
| 214 | |
| 215 | /** |
| 216 | * get_user_pages_fast() - pin user pages in memory |
| 217 | * @start: starting user address |
| 218 | * @nr_pages: number of pages from start to pin |
| 219 | * @write: whether pages will be written to |
| 220 | * @pages: array that receives pointers to the pages pinned. |
| 221 | * Should be at least nr_pages long. |
| 222 | * |
| 223 | * Returns number of pages pinned. This may be fewer than the number |
| 224 | * requested. If nr_pages is 0 or negative, returns 0. If no pages |
| 225 | * were pinned, returns -errno. |
| 226 | * |
| 227 | * get_user_pages_fast provides equivalent functionality to get_user_pages, |
| 228 | * operating on current and current->mm, with force=0 and vma=NULL. However |
| 229 | * unlike get_user_pages, it must be called without mmap_sem held. |
| 230 | * |
| 231 | * get_user_pages_fast may take mmap_sem and page table locks, so no |
| 232 | * assumptions can be made about lack of locking. get_user_pages_fast is to be |
| 233 | * implemented in a way that is advantageous (vs get_user_pages()) when the |
| 234 | * user memory area is already faulted in and present in ptes. However if the |
| 235 | * pages have to be faulted in, it may turn out to be slightly slower so |
| 236 | * callers need to carefully consider what to use. On many architectures, |
| 237 | * get_user_pages_fast simply falls back to get_user_pages. |
| 238 | */ |
| 239 | int __weak get_user_pages_fast(unsigned long start, |
| 240 | int nr_pages, int write, struct page **pages) |
| 241 | { |
| 242 | struct mm_struct *mm = current->mm; |
| 243 | int ret; |
| 244 | |
| 245 | down_read(&mm->mmap_sem); |
| 246 | ret = get_user_pages(current, mm, start, nr_pages, |
| 247 | write, 0, pages, NULL); |
| 248 | up_read(&mm->mmap_sem); |
| 249 | |
| 250 | return ret; |
| 251 | } |
| 252 | EXPORT_SYMBOL_GPL(get_user_pages_fast); |
| 253 | |
| 254 | unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr, |
| 255 | unsigned long len, unsigned long prot, |
| 256 | unsigned long flag, unsigned long pgoff) |
| 257 | { |
| 258 | unsigned long ret; |
| 259 | struct mm_struct *mm = current->mm; |
| 260 | unsigned long populate; |
| 261 | |
| 262 | ret = security_mmap_file(file, prot, flag); |
| 263 | if (!ret) { |
| 264 | down_write(&mm->mmap_sem); |
| 265 | ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff, |
| 266 | &populate); |
| 267 | up_write(&mm->mmap_sem); |
| 268 | if (populate) |
| 269 | mm_populate(ret, populate); |
| 270 | } |
| 271 | return ret; |
| 272 | } |
| 273 | |
| 274 | unsigned long vm_mmap(struct file *file, unsigned long addr, |
| 275 | unsigned long len, unsigned long prot, |
| 276 | unsigned long flag, unsigned long offset) |
| 277 | { |
| 278 | if (unlikely(offset + PAGE_ALIGN(len) < offset)) |
| 279 | return -EINVAL; |
| 280 | if (unlikely(offset & ~PAGE_MASK)) |
| 281 | return -EINVAL; |
| 282 | |
| 283 | return vm_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT); |
| 284 | } |
| 285 | EXPORT_SYMBOL(vm_mmap); |
| 286 | |
| 287 | void kvfree(const void *addr) |
| 288 | { |
| 289 | if (is_vmalloc_addr(addr)) |
| 290 | vfree(addr); |
| 291 | else |
| 292 | kfree(addr); |
| 293 | } |
| 294 | EXPORT_SYMBOL(kvfree); |
| 295 | |
| 296 | struct address_space *page_mapping(struct page *page) |
| 297 | { |
| 298 | struct address_space *mapping = page->mapping; |
| 299 | |
| 300 | /* This happens if someone calls flush_dcache_page on slab page */ |
| 301 | if (unlikely(PageSlab(page))) |
| 302 | return NULL; |
| 303 | |
| 304 | if (unlikely(PageSwapCache(page))) { |
| 305 | swp_entry_t entry; |
| 306 | |
| 307 | entry.val = page_private(page); |
| 308 | mapping = swap_address_space(entry); |
| 309 | } else if ((unsigned long)mapping & PAGE_MAPPING_ANON) |
| 310 | mapping = NULL; |
| 311 | return mapping; |
| 312 | } |
| 313 | |
| 314 | int overcommit_ratio_handler(struct ctl_table *table, int write, |
| 315 | void __user *buffer, size_t *lenp, |
| 316 | loff_t *ppos) |
| 317 | { |
| 318 | int ret; |
| 319 | |
| 320 | ret = proc_dointvec(table, write, buffer, lenp, ppos); |
| 321 | if (ret == 0 && write) |
| 322 | sysctl_overcommit_kbytes = 0; |
| 323 | return ret; |
| 324 | } |
| 325 | |
| 326 | int overcommit_kbytes_handler(struct ctl_table *table, int write, |
| 327 | void __user *buffer, size_t *lenp, |
| 328 | loff_t *ppos) |
| 329 | { |
| 330 | int ret; |
| 331 | |
| 332 | ret = proc_doulongvec_minmax(table, write, buffer, lenp, ppos); |
| 333 | if (ret == 0 && write) |
| 334 | sysctl_overcommit_ratio = 0; |
| 335 | return ret; |
| 336 | } |
| 337 | |
| 338 | /* |
| 339 | * Committed memory limit enforced when OVERCOMMIT_NEVER policy is used |
| 340 | */ |
| 341 | unsigned long vm_commit_limit(void) |
| 342 | { |
| 343 | unsigned long allowed; |
| 344 | |
| 345 | if (sysctl_overcommit_kbytes) |
| 346 | allowed = sysctl_overcommit_kbytes >> (PAGE_SHIFT - 10); |
| 347 | else |
| 348 | allowed = ((totalram_pages - hugetlb_total_pages()) |
| 349 | * sysctl_overcommit_ratio / 100); |
| 350 | allowed += total_swap_pages; |
| 351 | |
| 352 | return allowed; |
| 353 | } |
| 354 | |
| 355 | /** |
| 356 | * get_cmdline() - copy the cmdline value to a buffer. |
| 357 | * @task: the task whose cmdline value to copy. |
| 358 | * @buffer: the buffer to copy to. |
| 359 | * @buflen: the length of the buffer. Larger cmdline values are truncated |
| 360 | * to this length. |
| 361 | * Returns the size of the cmdline field copied. Note that the copy does |
| 362 | * not guarantee an ending NULL byte. |
| 363 | */ |
| 364 | int get_cmdline(struct task_struct *task, char *buffer, int buflen) |
| 365 | { |
| 366 | int res = 0; |
| 367 | unsigned int len; |
| 368 | struct mm_struct *mm = get_task_mm(task); |
| 369 | if (!mm) |
| 370 | goto out; |
| 371 | if (!mm->arg_end) |
| 372 | goto out_mm; /* Shh! No looking before we're done */ |
| 373 | |
| 374 | len = mm->arg_end - mm->arg_start; |
| 375 | |
| 376 | if (len > buflen) |
| 377 | len = buflen; |
| 378 | |
| 379 | res = access_process_vm(task, mm->arg_start, buffer, len, 0); |
| 380 | |
| 381 | /* |
| 382 | * If the nul at the end of args has been overwritten, then |
| 383 | * assume application is using setproctitle(3). |
| 384 | */ |
| 385 | if (res > 0 && buffer[res-1] != '\0' && len < buflen) { |
| 386 | len = strnlen(buffer, res); |
| 387 | if (len < res) { |
| 388 | res = len; |
| 389 | } else { |
| 390 | len = mm->env_end - mm->env_start; |
| 391 | if (len > buflen - res) |
| 392 | len = buflen - res; |
| 393 | res += access_process_vm(task, mm->env_start, |
| 394 | buffer+res, len, 0); |
| 395 | res = strnlen(buffer, res); |
| 396 | } |
| 397 | } |
| 398 | out_mm: |
| 399 | mmput(mm); |
| 400 | out: |
| 401 | return res; |
| 402 | } |