]>
Commit | Line | Data |
---|---|---|
fcf63409 CY |
1 | /* |
2 | * linux/mm/process_vm_access.c | |
3 | * | |
4 | * Copyright (C) 2010-2011 Christopher Yeoh <cyeoh@au1.ibm.com>, IBM Corp. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | */ | |
11 | ||
12 | #include <linux/mm.h> | |
13 | #include <linux/uio.h> | |
14 | #include <linux/sched.h> | |
15 | #include <linux/highmem.h> | |
16 | #include <linux/ptrace.h> | |
17 | #include <linux/slab.h> | |
18 | #include <linux/syscalls.h> | |
19 | ||
20 | #ifdef CONFIG_COMPAT | |
21 | #include <linux/compat.h> | |
22 | #endif | |
23 | ||
24 | /** | |
25 | * process_vm_rw_pages - read/write pages from task specified | |
26 | * @task: task to read/write from | |
27 | * @mm: mm for task | |
28 | * @process_pages: struct pages area that can store at least | |
29 | * nr_pages_to_copy struct page pointers | |
30 | * @pa: address of page in task to start copying from/to | |
31 | * @start_offset: offset in page to start copying from/to | |
32 | * @len: number of bytes to copy | |
33 | * @lvec: iovec array specifying where to copy to/from | |
34 | * @lvec_cnt: number of elements in iovec array | |
35 | * @lvec_current: index in iovec array we are up to | |
36 | * @lvec_offset: offset in bytes from current iovec iov_base we are up to | |
37 | * @vm_write: 0 means copy from, 1 means copy to | |
38 | * @nr_pages_to_copy: number of pages to copy | |
39 | * @bytes_copied: returns number of bytes successfully copied | |
40 | * Returns 0 on success, error code otherwise | |
41 | */ | |
42 | static int process_vm_rw_pages(struct task_struct *task, | |
43 | struct mm_struct *mm, | |
44 | struct page **process_pages, | |
45 | unsigned long pa, | |
46 | unsigned long start_offset, | |
47 | unsigned long len, | |
9f78bdfa | 48 | struct iov_iter *iter, |
fcf63409 CY |
49 | int vm_write, |
50 | unsigned int nr_pages_to_copy, | |
51 | ssize_t *bytes_copied) | |
52 | { | |
53 | int pages_pinned; | |
54 | void *target_kaddr; | |
55 | int pgs_copied = 0; | |
56 | int j; | |
57 | int ret; | |
58 | ssize_t bytes_to_copy; | |
59 | ssize_t rc = 0; | |
9f78bdfa | 60 | const struct iovec *iov = iter->iov; |
fcf63409 CY |
61 | |
62 | *bytes_copied = 0; | |
63 | ||
64 | /* Get the pages we're interested in */ | |
65 | down_read(&mm->mmap_sem); | |
66 | pages_pinned = get_user_pages(task, mm, pa, | |
67 | nr_pages_to_copy, | |
68 | vm_write, 0, process_pages, NULL); | |
69 | up_read(&mm->mmap_sem); | |
70 | ||
71 | if (pages_pinned != nr_pages_to_copy) { | |
72 | rc = -EFAULT; | |
73 | goto end; | |
74 | } | |
75 | ||
76 | /* Do the copy for each page */ | |
77 | for (pgs_copied = 0; | |
9f78bdfa | 78 | (pgs_copied < nr_pages_to_copy) && iter->nr_segs; |
fcf63409 CY |
79 | pgs_copied++) { |
80 | /* Make sure we have a non zero length iovec */ | |
9f78bdfa | 81 | while (iter->nr_segs && iov->iov_len == 0) { |
480402e1 | 82 | iov++; |
9f78bdfa | 83 | iter->nr_segs--; |
480402e1 | 84 | } |
9f78bdfa | 85 | if (!iter->nr_segs) |
fcf63409 CY |
86 | break; |
87 | ||
88 | /* | |
89 | * Will copy smallest of: | |
90 | * - bytes remaining in page | |
91 | * - bytes remaining in destination iovec | |
92 | */ | |
93 | bytes_to_copy = min_t(ssize_t, PAGE_SIZE - start_offset, | |
94 | len - *bytes_copied); | |
95 | bytes_to_copy = min_t(ssize_t, bytes_to_copy, | |
480402e1 | 96 | iov->iov_len |
9f78bdfa | 97 | - iter->iov_offset); |
fcf63409 CY |
98 | |
99 | target_kaddr = kmap(process_pages[pgs_copied]) + start_offset; | |
100 | ||
101 | if (vm_write) | |
102 | ret = copy_from_user(target_kaddr, | |
480402e1 | 103 | iov->iov_base |
9f78bdfa | 104 | + iter->iov_offset, |
fcf63409 CY |
105 | bytes_to_copy); |
106 | else | |
480402e1 | 107 | ret = copy_to_user(iov->iov_base |
9f78bdfa | 108 | + iter->iov_offset, |
fcf63409 CY |
109 | target_kaddr, bytes_to_copy); |
110 | kunmap(process_pages[pgs_copied]); | |
111 | if (ret) { | |
112 | *bytes_copied += bytes_to_copy - ret; | |
113 | pgs_copied++; | |
114 | rc = -EFAULT; | |
115 | goto end; | |
116 | } | |
117 | *bytes_copied += bytes_to_copy; | |
9f78bdfa AV |
118 | iter->iov_offset += bytes_to_copy; |
119 | if (iter->iov_offset == iov->iov_len) { | |
fcf63409 CY |
120 | /* |
121 | * Need to copy remaining part of page into the | |
122 | * next iovec if there are any bytes left in page | |
123 | */ | |
9f78bdfa | 124 | iter->nr_segs--; |
480402e1 | 125 | iov++; |
9f78bdfa | 126 | iter->iov_offset = 0; |
fcf63409 CY |
127 | start_offset = (start_offset + bytes_to_copy) |
128 | % PAGE_SIZE; | |
129 | if (start_offset) | |
130 | pgs_copied--; | |
131 | } else { | |
132 | start_offset = 0; | |
133 | } | |
134 | } | |
135 | ||
136 | end: | |
137 | if (vm_write) { | |
138 | for (j = 0; j < pages_pinned; j++) { | |
139 | if (j < pgs_copied) | |
140 | set_page_dirty_lock(process_pages[j]); | |
141 | put_page(process_pages[j]); | |
142 | } | |
143 | } else { | |
144 | for (j = 0; j < pages_pinned; j++) | |
145 | put_page(process_pages[j]); | |
146 | } | |
147 | ||
9f78bdfa | 148 | iter->iov = iov; |
fcf63409 CY |
149 | return rc; |
150 | } | |
151 | ||
152 | /* Maximum number of pages kmalloc'd to hold struct page's during copy */ | |
153 | #define PVM_MAX_KMALLOC_PAGES (PAGE_SIZE * 2) | |
154 | ||
155 | /** | |
156 | * process_vm_rw_single_vec - read/write pages from task specified | |
157 | * @addr: start memory address of target process | |
158 | * @len: size of area to copy to/from | |
159 | * @lvec: iovec array specifying where to copy to/from locally | |
160 | * @lvec_cnt: number of elements in iovec array | |
161 | * @lvec_current: index in iovec array we are up to | |
162 | * @lvec_offset: offset in bytes from current iovec iov_base we are up to | |
163 | * @process_pages: struct pages area that can store at least | |
164 | * nr_pages_to_copy struct page pointers | |
165 | * @mm: mm for task | |
166 | * @task: task to read/write from | |
167 | * @vm_write: 0 means copy from, 1 means copy to | |
168 | * @bytes_copied: returns number of bytes successfully copied | |
169 | * Returns 0 on success or on failure error code | |
170 | */ | |
171 | static int process_vm_rw_single_vec(unsigned long addr, | |
172 | unsigned long len, | |
9f78bdfa | 173 | struct iov_iter *iter, |
fcf63409 CY |
174 | struct page **process_pages, |
175 | struct mm_struct *mm, | |
176 | struct task_struct *task, | |
177 | int vm_write, | |
178 | ssize_t *bytes_copied) | |
179 | { | |
180 | unsigned long pa = addr & PAGE_MASK; | |
181 | unsigned long start_offset = addr - pa; | |
182 | unsigned long nr_pages; | |
183 | ssize_t bytes_copied_loop; | |
184 | ssize_t rc = 0; | |
185 | unsigned long nr_pages_copied = 0; | |
186 | unsigned long nr_pages_to_copy; | |
187 | unsigned long max_pages_per_loop = PVM_MAX_KMALLOC_PAGES | |
188 | / sizeof(struct pages *); | |
189 | ||
190 | *bytes_copied = 0; | |
191 | ||
192 | /* Work out address and page range required */ | |
193 | if (len == 0) | |
194 | return 0; | |
195 | nr_pages = (addr + len - 1) / PAGE_SIZE - addr / PAGE_SIZE + 1; | |
196 | ||
9f78bdfa | 197 | while ((nr_pages_copied < nr_pages) && iter->nr_segs) { |
fcf63409 CY |
198 | nr_pages_to_copy = min(nr_pages - nr_pages_copied, |
199 | max_pages_per_loop); | |
200 | ||
201 | rc = process_vm_rw_pages(task, mm, process_pages, pa, | |
9f78bdfa | 202 | start_offset, len, iter, |
fcf63409 CY |
203 | vm_write, nr_pages_to_copy, |
204 | &bytes_copied_loop); | |
205 | start_offset = 0; | |
206 | *bytes_copied += bytes_copied_loop; | |
207 | ||
208 | if (rc < 0) { | |
209 | return rc; | |
210 | } else { | |
211 | len -= bytes_copied_loop; | |
212 | nr_pages_copied += nr_pages_to_copy; | |
213 | pa += nr_pages_to_copy * PAGE_SIZE; | |
214 | } | |
215 | } | |
216 | ||
217 | return rc; | |
218 | } | |
219 | ||
220 | /* Maximum number of entries for process pages array | |
221 | which lives on stack */ | |
222 | #define PVM_MAX_PP_ARRAY_COUNT 16 | |
223 | ||
224 | /** | |
225 | * process_vm_rw_core - core of reading/writing pages from task specified | |
226 | * @pid: PID of process to read/write from/to | |
227 | * @lvec: iovec array specifying where to copy to/from locally | |
228 | * @liovcnt: size of lvec array | |
229 | * @rvec: iovec array specifying where to copy to/from in the other process | |
230 | * @riovcnt: size of rvec array | |
231 | * @flags: currently unused | |
232 | * @vm_write: 0 if reading from other process, 1 if writing to other process | |
233 | * Returns the number of bytes read/written or error code. May | |
234 | * return less bytes than expected if an error occurs during the copying | |
235 | * process. | |
236 | */ | |
9f78bdfa | 237 | static ssize_t process_vm_rw_core(pid_t pid, struct iov_iter *iter, |
fcf63409 CY |
238 | const struct iovec *rvec, |
239 | unsigned long riovcnt, | |
240 | unsigned long flags, int vm_write) | |
241 | { | |
242 | struct task_struct *task; | |
243 | struct page *pp_stack[PVM_MAX_PP_ARRAY_COUNT]; | |
244 | struct page **process_pages = pp_stack; | |
245 | struct mm_struct *mm; | |
246 | unsigned long i; | |
247 | ssize_t rc = 0; | |
248 | ssize_t bytes_copied_loop; | |
249 | ssize_t bytes_copied = 0; | |
250 | unsigned long nr_pages = 0; | |
251 | unsigned long nr_pages_iov; | |
fcf63409 CY |
252 | ssize_t iov_len; |
253 | ||
254 | /* | |
255 | * Work out how many pages of struct pages we're going to need | |
256 | * when eventually calling get_user_pages | |
257 | */ | |
258 | for (i = 0; i < riovcnt; i++) { | |
259 | iov_len = rvec[i].iov_len; | |
260 | if (iov_len > 0) { | |
261 | nr_pages_iov = ((unsigned long)rvec[i].iov_base | |
262 | + iov_len) | |
263 | / PAGE_SIZE - (unsigned long)rvec[i].iov_base | |
264 | / PAGE_SIZE + 1; | |
265 | nr_pages = max(nr_pages, nr_pages_iov); | |
266 | } | |
267 | } | |
268 | ||
269 | if (nr_pages == 0) | |
270 | return 0; | |
271 | ||
272 | if (nr_pages > PVM_MAX_PP_ARRAY_COUNT) { | |
273 | /* For reliability don't try to kmalloc more than | |
274 | 2 pages worth */ | |
275 | process_pages = kmalloc(min_t(size_t, PVM_MAX_KMALLOC_PAGES, | |
276 | sizeof(struct pages *)*nr_pages), | |
277 | GFP_KERNEL); | |
278 | ||
279 | if (!process_pages) | |
280 | return -ENOMEM; | |
281 | } | |
282 | ||
283 | /* Get process information */ | |
284 | rcu_read_lock(); | |
285 | task = find_task_by_vpid(pid); | |
286 | if (task) | |
287 | get_task_struct(task); | |
288 | rcu_read_unlock(); | |
289 | if (!task) { | |
290 | rc = -ESRCH; | |
291 | goto free_proc_pages; | |
292 | } | |
293 | ||
8cdb878d CY |
294 | mm = mm_access(task, PTRACE_MODE_ATTACH); |
295 | if (!mm || IS_ERR(mm)) { | |
296 | rc = IS_ERR(mm) ? PTR_ERR(mm) : -ESRCH; | |
297 | /* | |
298 | * Explicitly map EACCES to EPERM as EPERM is a more a | |
299 | * appropriate error code for process_vw_readv/writev | |
300 | */ | |
301 | if (rc == -EACCES) | |
302 | rc = -EPERM; | |
fcf63409 CY |
303 | goto put_task_struct; |
304 | } | |
305 | ||
9f78bdfa | 306 | for (i = 0; i < riovcnt && iter->nr_segs; i++) { |
fcf63409 CY |
307 | rc = process_vm_rw_single_vec( |
308 | (unsigned long)rvec[i].iov_base, rvec[i].iov_len, | |
9f78bdfa AV |
309 | iter, process_pages, mm, task, vm_write, |
310 | &bytes_copied_loop); | |
fcf63409 CY |
311 | bytes_copied += bytes_copied_loop; |
312 | if (rc != 0) { | |
313 | /* If we have managed to copy any data at all then | |
314 | we return the number of bytes copied. Otherwise | |
315 | we return the error code */ | |
316 | if (bytes_copied) | |
317 | rc = bytes_copied; | |
318 | goto put_mm; | |
319 | } | |
320 | } | |
321 | ||
322 | rc = bytes_copied; | |
323 | put_mm: | |
324 | mmput(mm); | |
325 | ||
326 | put_task_struct: | |
327 | put_task_struct(task); | |
328 | ||
329 | free_proc_pages: | |
330 | if (process_pages != pp_stack) | |
331 | kfree(process_pages); | |
332 | return rc; | |
333 | } | |
334 | ||
335 | /** | |
336 | * process_vm_rw - check iovecs before calling core routine | |
337 | * @pid: PID of process to read/write from/to | |
338 | * @lvec: iovec array specifying where to copy to/from locally | |
339 | * @liovcnt: size of lvec array | |
340 | * @rvec: iovec array specifying where to copy to/from in the other process | |
341 | * @riovcnt: size of rvec array | |
342 | * @flags: currently unused | |
343 | * @vm_write: 0 if reading from other process, 1 if writing to other process | |
344 | * Returns the number of bytes read/written or error code. May | |
345 | * return less bytes than expected if an error occurs during the copying | |
346 | * process. | |
347 | */ | |
348 | static ssize_t process_vm_rw(pid_t pid, | |
349 | const struct iovec __user *lvec, | |
350 | unsigned long liovcnt, | |
351 | const struct iovec __user *rvec, | |
352 | unsigned long riovcnt, | |
353 | unsigned long flags, int vm_write) | |
354 | { | |
355 | struct iovec iovstack_l[UIO_FASTIOV]; | |
356 | struct iovec iovstack_r[UIO_FASTIOV]; | |
357 | struct iovec *iov_l = iovstack_l; | |
358 | struct iovec *iov_r = iovstack_r; | |
9f78bdfa | 359 | struct iov_iter iter; |
fcf63409 CY |
360 | ssize_t rc; |
361 | ||
362 | if (flags != 0) | |
363 | return -EINVAL; | |
364 | ||
365 | /* Check iovecs */ | |
366 | if (vm_write) | |
367 | rc = rw_copy_check_uvector(WRITE, lvec, liovcnt, UIO_FASTIOV, | |
ac34ebb3 | 368 | iovstack_l, &iov_l); |
fcf63409 CY |
369 | else |
370 | rc = rw_copy_check_uvector(READ, lvec, liovcnt, UIO_FASTIOV, | |
ac34ebb3 | 371 | iovstack_l, &iov_l); |
fcf63409 CY |
372 | if (rc <= 0) |
373 | goto free_iovecs; | |
374 | ||
9f78bdfa AV |
375 | iov_iter_init(&iter, iov_l, liovcnt, rc, 0); |
376 | ||
ac34ebb3 CY |
377 | rc = rw_copy_check_uvector(CHECK_IOVEC_ONLY, rvec, riovcnt, UIO_FASTIOV, |
378 | iovstack_r, &iov_r); | |
fcf63409 CY |
379 | if (rc <= 0) |
380 | goto free_iovecs; | |
381 | ||
9f78bdfa | 382 | rc = process_vm_rw_core(pid, &iter, iov_r, riovcnt, flags, vm_write); |
fcf63409 CY |
383 | |
384 | free_iovecs: | |
385 | if (iov_r != iovstack_r) | |
386 | kfree(iov_r); | |
387 | if (iov_l != iovstack_l) | |
388 | kfree(iov_l); | |
389 | ||
390 | return rc; | |
391 | } | |
392 | ||
393 | SYSCALL_DEFINE6(process_vm_readv, pid_t, pid, const struct iovec __user *, lvec, | |
394 | unsigned long, liovcnt, const struct iovec __user *, rvec, | |
395 | unsigned long, riovcnt, unsigned long, flags) | |
396 | { | |
397 | return process_vm_rw(pid, lvec, liovcnt, rvec, riovcnt, flags, 0); | |
398 | } | |
399 | ||
400 | SYSCALL_DEFINE6(process_vm_writev, pid_t, pid, | |
401 | const struct iovec __user *, lvec, | |
402 | unsigned long, liovcnt, const struct iovec __user *, rvec, | |
403 | unsigned long, riovcnt, unsigned long, flags) | |
404 | { | |
405 | return process_vm_rw(pid, lvec, liovcnt, rvec, riovcnt, flags, 1); | |
406 | } | |
407 | ||
408 | #ifdef CONFIG_COMPAT | |
409 | ||
410 | asmlinkage ssize_t | |
411 | compat_process_vm_rw(compat_pid_t pid, | |
412 | const struct compat_iovec __user *lvec, | |
413 | unsigned long liovcnt, | |
414 | const struct compat_iovec __user *rvec, | |
415 | unsigned long riovcnt, | |
416 | unsigned long flags, int vm_write) | |
417 | { | |
418 | struct iovec iovstack_l[UIO_FASTIOV]; | |
419 | struct iovec iovstack_r[UIO_FASTIOV]; | |
420 | struct iovec *iov_l = iovstack_l; | |
421 | struct iovec *iov_r = iovstack_r; | |
9f78bdfa | 422 | struct iov_iter iter; |
fcf63409 CY |
423 | ssize_t rc = -EFAULT; |
424 | ||
425 | if (flags != 0) | |
426 | return -EINVAL; | |
427 | ||
fcf63409 CY |
428 | if (vm_write) |
429 | rc = compat_rw_copy_check_uvector(WRITE, lvec, liovcnt, | |
430 | UIO_FASTIOV, iovstack_l, | |
ac34ebb3 | 431 | &iov_l); |
fcf63409 CY |
432 | else |
433 | rc = compat_rw_copy_check_uvector(READ, lvec, liovcnt, | |
434 | UIO_FASTIOV, iovstack_l, | |
ac34ebb3 | 435 | &iov_l); |
fcf63409 CY |
436 | if (rc <= 0) |
437 | goto free_iovecs; | |
9f78bdfa | 438 | iov_iter_init(&iter, iov_l, liovcnt, rc, 0); |
ac34ebb3 | 439 | rc = compat_rw_copy_check_uvector(CHECK_IOVEC_ONLY, rvec, riovcnt, |
fcf63409 | 440 | UIO_FASTIOV, iovstack_r, |
ac34ebb3 | 441 | &iov_r); |
fcf63409 CY |
442 | if (rc <= 0) |
443 | goto free_iovecs; | |
444 | ||
9f78bdfa | 445 | rc = process_vm_rw_core(pid, &iter, iov_r, riovcnt, flags, vm_write); |
fcf63409 CY |
446 | |
447 | free_iovecs: | |
448 | if (iov_r != iovstack_r) | |
449 | kfree(iov_r); | |
450 | if (iov_l != iovstack_l) | |
451 | kfree(iov_l); | |
fcf63409 CY |
452 | return rc; |
453 | } | |
454 | ||
455 | asmlinkage ssize_t | |
456 | compat_sys_process_vm_readv(compat_pid_t pid, | |
457 | const struct compat_iovec __user *lvec, | |
458 | unsigned long liovcnt, | |
459 | const struct compat_iovec __user *rvec, | |
460 | unsigned long riovcnt, | |
461 | unsigned long flags) | |
462 | { | |
463 | return compat_process_vm_rw(pid, lvec, liovcnt, rvec, | |
464 | riovcnt, flags, 0); | |
465 | } | |
466 | ||
467 | asmlinkage ssize_t | |
468 | compat_sys_process_vm_writev(compat_pid_t pid, | |
469 | const struct compat_iovec __user *lvec, | |
470 | unsigned long liovcnt, | |
471 | const struct compat_iovec __user *rvec, | |
472 | unsigned long riovcnt, | |
473 | unsigned long flags) | |
474 | { | |
475 | return compat_process_vm_rw(pid, lvec, liovcnt, rvec, | |
476 | riovcnt, flags, 1); | |
477 | } | |
478 | ||
479 | #endif |