?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
compat: sync compat_stats with statfs.
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / compat.c
blob0f1ab468fa2d489f1a4f67fc5460d4dee5322961
1 /*
2 * linux/fs/compat.c
3 *
4 * Kernel compatibililty routines for e.g. 32 bit syscall support
5 * on 64 bit kernels.
6 *
7 * Copyright (C) 2002 Stephen Rothwell, IBM Corporation
8 * Copyright (C) 1997-2000 Jakub Jelinek (jakub@redhat.com)
9 * Copyright (C) 1998 Eddie C. Dost (ecd@skynet.be)
10 * Copyright (C) 2001,2002 Andi Kleen, SuSE Labs
11 * Copyright (C) 2003 Pavel Machek (pavel@ucw.cz)
12 *
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License version 2 as
15 * published by the Free Software Foundation.
16 */
17
18 #include <linux/stddef.h>
19 #include <linux/kernel.h>
20 #include <linux/linkage.h>
21 #include <linux/compat.h>
22 #include <linux/errno.h>
23 #include <linux/time.h>
24 #include <linux/fs.h>
25 #include <linux/fcntl.h>
26 #include <linux/namei.h>
27 #include <linux/file.h>
28 #include <linux/fdtable.h>
29 #include <linux/vfs.h>
30 #include <linux/ioctl.h>
31 #include <linux/init.h>
32 #include <linux/ncp_mount.h>
33 #include <linux/nfs4_mount.h>
34 #include <linux/syscalls.h>
35 #include <linux/ctype.h>
36 #include <linux/module.h>
37 #include <linux/dirent.h>
38 #include <linux/fsnotify.h>
39 #include <linux/highuid.h>
40 #include <linux/nfsd/syscall.h>
41 #include <linux/personality.h>
42 #include <linux/rwsem.h>
43 #include <linux/tsacct_kern.h>
44 #include <linux/security.h>
45 #include <linux/highmem.h>
46 #include <linux/signal.h>
47 #include <linux/poll.h>
48 #include <linux/mm.h>
49 #include <linux/eventpoll.h>
50 #include <linux/fs_struct.h>
51 #include <linux/slab.h>
52 #include <linux/pagemap.h>
53
54 #include <asm/uaccess.h>
55 #include <asm/mmu_context.h>
56 #include <asm/ioctls.h>
57 #include "internal.h"
58
59 int compat_log = 1;
60
61 int compat_printk(const char *fmt, ...)
62 {
63 va_list ap;
64 int ret;
65 if (!compat_log)
66 return 0;
67 va_start(ap, fmt);
68 ret = vprintk(fmt, ap);
69 va_end(ap);
70 return ret;
71 }
72
73 #include "read_write.h"
74
75 /*
76 * Not all architectures have sys_utime, so implement this in terms
77 * of sys_utimes.
78 */
79 asmlinkage long compat_sys_utime(const char __user *filename,
80 struct compat_utimbuf __user *t)
81 {
82 struct timespec tv[2];
83
84 if (t) {
85 if (get_user(tv[0].tv_sec, &t->actime) ||
86 get_user(tv[1].tv_sec, &t->modtime))
87 return -EFAULT;
88 tv[0].tv_nsec = 0;
89 tv[1].tv_nsec = 0;
90 }
91 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
92 }
93
94 asmlinkage long compat_sys_utimensat(unsigned int dfd, const char __user *filename, struct compat_timespec __user *t, int flags)
95 {
96 struct timespec tv[2];
97
98 if (t) {
99 if (get_compat_timespec(&tv[0], &t[0]) ||
100 get_compat_timespec(&tv[1], &t[1]))
101 return -EFAULT;
102
103 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
104 return 0;
105 }
106 return do_utimes(dfd, filename, t ? tv : NULL, flags);
107 }
108
109 asmlinkage long compat_sys_futimesat(unsigned int dfd, const char __user *filename, struct compat_timeval __user *t)
110 {
111 struct timespec tv[2];
112
113 if (t) {
114 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
115 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
116 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
117 get_user(tv[1].tv_nsec, &t[1].tv_usec))
118 return -EFAULT;
119 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
120 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
121 return -EINVAL;
122 tv[0].tv_nsec *= 1000;
123 tv[1].tv_nsec *= 1000;
124 }
125 return do_utimes(dfd, filename, t ? tv : NULL, 0);
126 }
127
128 asmlinkage long compat_sys_utimes(const char __user *filename, struct compat_timeval __user *t)
129 {
130 return compat_sys_futimesat(AT_FDCWD, filename, t);
131 }
132
133 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
134 {
135 compat_ino_t ino = stat->ino;
136 typeof(ubuf->st_uid) uid = 0;
137 typeof(ubuf->st_gid) gid = 0;
138 int err;
139
140 SET_UID(uid, stat->uid);
141 SET_GID(gid, stat->gid);
142
143 if ((u64) stat->size > MAX_NON_LFS ||
144 !old_valid_dev(stat->dev) ||
145 !old_valid_dev(stat->rdev))
146 return -EOVERFLOW;
147 if (sizeof(ino) < sizeof(stat->ino) && ino != stat->ino)
148 return -EOVERFLOW;
149
150 if (clear_user(ubuf, sizeof(*ubuf)))
151 return -EFAULT;
152
153 err = __put_user(old_encode_dev(stat->dev), &ubuf->st_dev);
154 err |= __put_user(ino, &ubuf->st_ino);
155 err |= __put_user(stat->mode, &ubuf->st_mode);
156 err |= __put_user(stat->nlink, &ubuf->st_nlink);
157 err |= __put_user(uid, &ubuf->st_uid);
158 err |= __put_user(gid, &ubuf->st_gid);
159 err |= __put_user(old_encode_dev(stat->rdev), &ubuf->st_rdev);
160 err |= __put_user(stat->size, &ubuf->st_size);
161 err |= __put_user(stat->atime.tv_sec, &ubuf->st_atime);
162 err |= __put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec);
163 err |= __put_user(stat->mtime.tv_sec, &ubuf->st_mtime);
164 err |= __put_user(stat->mtime.tv_nsec, &ubuf->st_mtime_nsec);
165 err |= __put_user(stat->ctime.tv_sec, &ubuf->st_ctime);
166 err |= __put_user(stat->ctime.tv_nsec, &ubuf->st_ctime_nsec);
167 err |= __put_user(stat->blksize, &ubuf->st_blksize);
168 err |= __put_user(stat->blocks, &ubuf->st_blocks);
169 return err;
170 }
171
172 asmlinkage long compat_sys_newstat(const char __user * filename,
173 struct compat_stat __user *statbuf)
174 {
175 struct kstat stat;
176 int error;
177
178 error = vfs_stat(filename, &stat);
179 if (error)
180 return error;
181 return cp_compat_stat(&stat, statbuf);
182 }
183
184 asmlinkage long compat_sys_newlstat(const char __user * filename,
185 struct compat_stat __user *statbuf)
186 {
187 struct kstat stat;
188 int error;
189
190 error = vfs_lstat(filename, &stat);
191 if (error)
192 return error;
193 return cp_compat_stat(&stat, statbuf);
194 }
195
196 #ifndef __ARCH_WANT_STAT64
197 asmlinkage long compat_sys_newfstatat(unsigned int dfd,
198 const char __user *filename,
199 struct compat_stat __user *statbuf, int flag)
200 {
201 struct kstat stat;
202 int error;
203
204 error = vfs_fstatat(dfd, filename, &stat, flag);
205 if (error)
206 return error;
207 return cp_compat_stat(&stat, statbuf);
208 }
209 #endif
210
211 asmlinkage long compat_sys_newfstat(unsigned int fd,
212 struct compat_stat __user * statbuf)
213 {
214 struct kstat stat;
215 int error = vfs_fstat(fd, &stat);
216
217 if (!error)
218 error = cp_compat_stat(&stat, statbuf);
219 return error;
220 }
221
222 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
223 {
224
225 if (sizeof ubuf->f_blocks == 4) {
226 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
227 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
228 return -EOVERFLOW;
229 /* f_files and f_ffree may be -1; it's okay
230 * to stuff that into 32 bits */
231 if (kbuf->f_files != 0xffffffffffffffffULL
232 && (kbuf->f_files & 0xffffffff00000000ULL))
233 return -EOVERFLOW;
234 if (kbuf->f_ffree != 0xffffffffffffffffULL
235 && (kbuf->f_ffree & 0xffffffff00000000ULL))
236 return -EOVERFLOW;
237 }
238 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
239 __put_user(kbuf->f_type, &ubuf->f_type) ||
240 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
241 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
242 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
243 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
244 __put_user(kbuf->f_files, &ubuf->f_files) ||
245 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
246 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
247 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
248 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
249 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
250 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
251 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
252 return -EFAULT;
253 return 0;
254 }
255
256 /*
257 * The following statfs calls are copies of code from fs/statfs.c and
258 * should be checked against those from time to time
259 */
260 asmlinkage long compat_sys_statfs(const char __user *pathname, struct compat_statfs __user *buf)
261 {
262 struct kstatfs tmp;
263 int error = user_statfs(pathname, &tmp);
264 if (!error)
265 error = put_compat_statfs(buf, &tmp);
266 return error;
267 }
268
269 asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
270 {
271 struct kstatfs tmp;
272 int error = fd_statfs(fd, &tmp);
273 if (!error)
274 error = put_compat_statfs(buf, &tmp);
275 return error;
276 }
277
278 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
279 {
280 if (sizeof ubuf->f_blocks == 4) {
281 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
282 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
283 return -EOVERFLOW;
284 /* f_files and f_ffree may be -1; it's okay
285 * to stuff that into 32 bits */
286 if (kbuf->f_files != 0xffffffffffffffffULL
287 && (kbuf->f_files & 0xffffffff00000000ULL))
288 return -EOVERFLOW;
289 if (kbuf->f_ffree != 0xffffffffffffffffULL
290 && (kbuf->f_ffree & 0xffffffff00000000ULL))
291 return -EOVERFLOW;
292 }
293 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
294 __put_user(kbuf->f_type, &ubuf->f_type) ||
295 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
296 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
297 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
298 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
299 __put_user(kbuf->f_files, &ubuf->f_files) ||
300 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
301 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
302 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
303 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
304 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
305 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
306 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
307 return -EFAULT;
308 return 0;
309 }
310
311 asmlinkage long compat_sys_statfs64(const char __user *pathname, compat_size_t sz, struct compat_statfs64 __user *buf)
312 {
313 struct kstatfs tmp;
314 int error;
315
316 if (sz != sizeof(*buf))
317 return -EINVAL;
318
319 error = user_statfs(pathname, &tmp);
320 if (!error)
321 error = put_compat_statfs64(buf, &tmp);
322 return error;
323 }
324
325 asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
326 {
327 struct kstatfs tmp;
328 int error;
329
330 if (sz != sizeof(*buf))
331 return -EINVAL;
332
333 error = fd_statfs(fd, &tmp);
334 if (!error)
335 error = put_compat_statfs64(buf, &tmp);
336 return error;
337 }
338
339 /*
340 * This is a copy of sys_ustat, just dealing with a structure layout.
341 * Given how simple this syscall is that apporach is more maintainable
342 * than the various conversion hacks.
343 */
344 asmlinkage long compat_sys_ustat(unsigned dev, struct compat_ustat __user *u)
345 {
346 struct super_block *sb;
347 struct compat_ustat tmp;
348 struct kstatfs sbuf;
349 int err;
350
351 sb = user_get_super(new_decode_dev(dev));
352 if (!sb)
353 return -EINVAL;
354 err = statfs_by_dentry(sb->s_root, &sbuf);
355 drop_super(sb);
356 if (err)
357 return err;
358
359 memset(&tmp, 0, sizeof(struct compat_ustat));
360 tmp.f_tfree = sbuf.f_bfree;
361 tmp.f_tinode = sbuf.f_ffree;
362 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
363 return -EFAULT;
364 return 0;
365 }
366
367 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
368 {
369 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
370 __get_user(kfl->l_type, &ufl->l_type) ||
371 __get_user(kfl->l_whence, &ufl->l_whence) ||
372 __get_user(kfl->l_start, &ufl->l_start) ||
373 __get_user(kfl->l_len, &ufl->l_len) ||
374 __get_user(kfl->l_pid, &ufl->l_pid))
375 return -EFAULT;
376 return 0;
377 }
378
379 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
380 {
381 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
382 __put_user(kfl->l_type, &ufl->l_type) ||
383 __put_user(kfl->l_whence, &ufl->l_whence) ||
384 __put_user(kfl->l_start, &ufl->l_start) ||
385 __put_user(kfl->l_len, &ufl->l_len) ||
386 __put_user(kfl->l_pid, &ufl->l_pid))
387 return -EFAULT;
388 return 0;
389 }
390
391 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
392 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
393 {
394 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
395 __get_user(kfl->l_type, &ufl->l_type) ||
396 __get_user(kfl->l_whence, &ufl->l_whence) ||
397 __get_user(kfl->l_start, &ufl->l_start) ||
398 __get_user(kfl->l_len, &ufl->l_len) ||
399 __get_user(kfl->l_pid, &ufl->l_pid))
400 return -EFAULT;
401 return 0;
402 }
403 #endif
404
405 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
406 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
407 {
408 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
409 __put_user(kfl->l_type, &ufl->l_type) ||
410 __put_user(kfl->l_whence, &ufl->l_whence) ||
411 __put_user(kfl->l_start, &ufl->l_start) ||
412 __put_user(kfl->l_len, &ufl->l_len) ||
413 __put_user(kfl->l_pid, &ufl->l_pid))
414 return -EFAULT;
415 return 0;
416 }
417 #endif
418
419 asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
420 unsigned long arg)
421 {
422 mm_segment_t old_fs;
423 struct flock f;
424 long ret;
425
426 switch (cmd) {
427 case F_GETLK:
428 case F_SETLK:
429 case F_SETLKW:
430 ret = get_compat_flock(&f, compat_ptr(arg));
431 if (ret != 0)
432 break;
433 old_fs = get_fs();
434 set_fs(KERNEL_DS);
435 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
436 set_fs(old_fs);
437 if (cmd == F_GETLK && ret == 0) {
438 /* GETLK was successful and we need to return the data...
439 * but it needs to fit in the compat structure.
440 * l_start shouldn't be too big, unless the original
441 * start + end is greater than COMPAT_OFF_T_MAX, in which
442 * case the app was asking for trouble, so we return
443 * -EOVERFLOW in that case.
444 * l_len could be too big, in which case we just truncate it,
445 * and only allow the app to see that part of the conflicting
446 * lock that might make sense to it anyway
447 */
448
449 if (f.l_start > COMPAT_OFF_T_MAX)
450 ret = -EOVERFLOW;
451 if (f.l_len > COMPAT_OFF_T_MAX)
452 f.l_len = COMPAT_OFF_T_MAX;
453 if (ret == 0)
454 ret = put_compat_flock(&f, compat_ptr(arg));
455 }
456 break;
457
458 case F_GETLK64:
459 case F_SETLK64:
460 case F_SETLKW64:
461 ret = get_compat_flock64(&f, compat_ptr(arg));
462 if (ret != 0)
463 break;
464 old_fs = get_fs();
465 set_fs(KERNEL_DS);
466 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
467 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
468 (unsigned long)&f);
469 set_fs(old_fs);
470 if (cmd == F_GETLK64 && ret == 0) {
471 /* need to return lock information - see above for commentary */
472 if (f.l_start > COMPAT_LOFF_T_MAX)
473 ret = -EOVERFLOW;
474 if (f.l_len > COMPAT_LOFF_T_MAX)
475 f.l_len = COMPAT_LOFF_T_MAX;
476 if (ret == 0)
477 ret = put_compat_flock64(&f, compat_ptr(arg));
478 }
479 break;
480
481 default:
482 ret = sys_fcntl(fd, cmd, arg);
483 break;
484 }
485 return ret;
486 }
487
488 asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
489 unsigned long arg)
490 {
491 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
492 return -EINVAL;
493 return compat_sys_fcntl64(fd, cmd, arg);
494 }
495
496 asmlinkage long
497 compat_sys_io_setup(unsigned nr_reqs, u32 __user *ctx32p)
498 {
499 long ret;
500 aio_context_t ctx64;
501
502 mm_segment_t oldfs = get_fs();
503 if (unlikely(get_user(ctx64, ctx32p)))
504 return -EFAULT;
505
506 set_fs(KERNEL_DS);
507 /* The __user pointer cast is valid because of the set_fs() */
508 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
509 set_fs(oldfs);
510 /* truncating is ok because it's a user address */
511 if (!ret)
512 ret = put_user((u32) ctx64, ctx32p);
513 return ret;
514 }
515
516 asmlinkage long
517 compat_sys_io_getevents(aio_context_t ctx_id,
518 unsigned long min_nr,
519 unsigned long nr,
520 struct io_event __user *events,
521 struct compat_timespec __user *timeout)
522 {
523 long ret;
524 struct timespec t;
525 struct timespec __user *ut = NULL;
526
527 ret = -EFAULT;
528 if (unlikely(!access_ok(VERIFY_WRITE, events,
529 nr * sizeof(struct io_event))))
530 goto out;
531 if (timeout) {
532 if (get_compat_timespec(&t, timeout))
533 goto out;
534
535 ut = compat_alloc_user_space(sizeof(*ut));
536 if (copy_to_user(ut, &t, sizeof(t)) )
537 goto out;
538 }
539 ret = sys_io_getevents(ctx_id, min_nr, nr, events, ut);
540 out:
541 return ret;
542 }
543
544 /* A write operation does a read from user space and vice versa */
545 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
546
547 ssize_t compat_rw_copy_check_uvector(int type,
548 const struct compat_iovec __user *uvector, unsigned long nr_segs,
549 unsigned long fast_segs, struct iovec *fast_pointer,
550 struct iovec **ret_pointer)
551 {
552 compat_ssize_t tot_len;
553 struct iovec *iov = *ret_pointer = fast_pointer;
554 ssize_t ret = 0;
555 int seg;
556
557 /*
558 * SuS says "The readv() function *may* fail if the iovcnt argument
559 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
560 * traditionally returned zero for zero segments, so...
561 */
562 if (nr_segs == 0)
563 goto out;
564
565 ret = -EINVAL;
566 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
567 goto out;
568 if (nr_segs > fast_segs) {
569 ret = -ENOMEM;
570 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
571 if (iov == NULL)
572 goto out;
573 }
574 *ret_pointer = iov;
575
576 /*
577 * Single unix specification:
578 * We should -EINVAL if an element length is not >= 0 and fitting an
579 * ssize_t.
580 *
581 * In Linux, the total length is limited to MAX_RW_COUNT, there is
582 * no overflow possibility.
583 */
584 tot_len = 0;
585 ret = -EINVAL;
586 for (seg = 0; seg < nr_segs; seg++) {
587 compat_uptr_t buf;
588 compat_ssize_t len;
589
590 if (__get_user(len, &uvector->iov_len) ||
591 __get_user(buf, &uvector->iov_base)) {
592 ret = -EFAULT;
593 goto out;
594 }
595 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
596 goto out;
597 if (!access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
598 ret = -EFAULT;
599 goto out;
600 }
601 if (len > MAX_RW_COUNT - tot_len)
602 len = MAX_RW_COUNT - tot_len;
603 tot_len += len;
604 iov->iov_base = compat_ptr(buf);
605 iov->iov_len = (compat_size_t) len;
606 uvector++;
607 iov++;
608 }
609 ret = tot_len;
610
611 out:
612 return ret;
613 }
614
615 static inline long
616 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
617 {
618 compat_uptr_t uptr;
619 int i;
620
621 for (i = 0; i < nr; ++i) {
622 if (get_user(uptr, ptr32 + i))
623 return -EFAULT;
624 if (put_user(compat_ptr(uptr), ptr64 + i))
625 return -EFAULT;
626 }
627 return 0;
628 }
629
630 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
631
632 asmlinkage long
633 compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
634 {
635 struct iocb __user * __user *iocb64;
636 long ret;
637
638 if (unlikely(nr < 0))
639 return -EINVAL;
640
641 if (nr > MAX_AIO_SUBMITS)
642 nr = MAX_AIO_SUBMITS;
643
644 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
645 ret = copy_iocb(nr, iocb, iocb64);
646 if (!ret)
647 ret = do_io_submit(ctx_id, nr, iocb64, 1);
648 return ret;
649 }
650
651 struct compat_ncp_mount_data {
652 compat_int_t version;
653 compat_uint_t ncp_fd;
654 __compat_uid_t mounted_uid;
655 compat_pid_t wdog_pid;
656 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
657 compat_uint_t time_out;
658 compat_uint_t retry_count;
659 compat_uint_t flags;
660 __compat_uid_t uid;
661 __compat_gid_t gid;
662 compat_mode_t file_mode;
663 compat_mode_t dir_mode;
664 };
665
666 struct compat_ncp_mount_data_v4 {
667 compat_int_t version;
668 compat_ulong_t flags;
669 compat_ulong_t mounted_uid;
670 compat_long_t wdog_pid;
671 compat_uint_t ncp_fd;
672 compat_uint_t time_out;
673 compat_uint_t retry_count;
674 compat_ulong_t uid;
675 compat_ulong_t gid;
676 compat_ulong_t file_mode;
677 compat_ulong_t dir_mode;
678 };
679
680 static void *do_ncp_super_data_conv(void *raw_data)
681 {
682 int version = *(unsigned int *)raw_data;
683
684 if (version == 3) {
685 struct compat_ncp_mount_data *c_n = raw_data;
686 struct ncp_mount_data *n = raw_data;
687
688 n->dir_mode = c_n->dir_mode;
689 n->file_mode = c_n->file_mode;
690 n->gid = c_n->gid;
691 n->uid = c_n->uid;
692 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
693 n->wdog_pid = c_n->wdog_pid;
694 n->mounted_uid = c_n->mounted_uid;
695 } else if (version == 4) {
696 struct compat_ncp_mount_data_v4 *c_n = raw_data;
697 struct ncp_mount_data_v4 *n = raw_data;
698
699 n->dir_mode = c_n->dir_mode;
700 n->file_mode = c_n->file_mode;
701 n->gid = c_n->gid;
702 n->uid = c_n->uid;
703 n->retry_count = c_n->retry_count;
704 n->time_out = c_n->time_out;
705 n->ncp_fd = c_n->ncp_fd;
706 n->wdog_pid = c_n->wdog_pid;
707 n->mounted_uid = c_n->mounted_uid;
708 n->flags = c_n->flags;
709 } else if (version != 5) {
710 return NULL;
711 }
712
713 return raw_data;
714 }
715
716
717 struct compat_nfs_string {
718 compat_uint_t len;
719 compat_uptr_t data;
720 };
721
722 static inline void compat_nfs_string(struct nfs_string *dst,
723 struct compat_nfs_string *src)
724 {
725 dst->data = compat_ptr(src->data);
726 dst->len = src->len;
727 }
728
729 struct compat_nfs4_mount_data_v1 {
730 compat_int_t version;
731 compat_int_t flags;
732 compat_int_t rsize;
733 compat_int_t wsize;
734 compat_int_t timeo;
735 compat_int_t retrans;
736 compat_int_t acregmin;
737 compat_int_t acregmax;
738 compat_int_t acdirmin;
739 compat_int_t acdirmax;
740 struct compat_nfs_string client_addr;
741 struct compat_nfs_string mnt_path;
742 struct compat_nfs_string hostname;
743 compat_uint_t host_addrlen;
744 compat_uptr_t host_addr;
745 compat_int_t proto;
746 compat_int_t auth_flavourlen;
747 compat_uptr_t auth_flavours;
748 };
749
750 static int do_nfs4_super_data_conv(void *raw_data)
751 {
752 int version = *(compat_uint_t *) raw_data;
753
754 if (version == 1) {
755 struct compat_nfs4_mount_data_v1 *raw = raw_data;
756 struct nfs4_mount_data *real = raw_data;
757
758 /* copy the fields backwards */
759 real->auth_flavours = compat_ptr(raw->auth_flavours);
760 real->auth_flavourlen = raw->auth_flavourlen;
761 real->proto = raw->proto;
762 real->host_addr = compat_ptr(raw->host_addr);
763 real->host_addrlen = raw->host_addrlen;
764 compat_nfs_string(&real->hostname, &raw->hostname);
765 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
766 compat_nfs_string(&real->client_addr, &raw->client_addr);
767 real->acdirmax = raw->acdirmax;
768 real->acdirmin = raw->acdirmin;
769 real->acregmax = raw->acregmax;
770 real->acregmin = raw->acregmin;
771 real->retrans = raw->retrans;
772 real->timeo = raw->timeo;
773 real->wsize = raw->wsize;
774 real->rsize = raw->rsize;
775 real->flags = raw->flags;
776 real->version = raw->version;
777 }
778
779 return 0;
780 }
781
782 #define NCPFS_NAME "ncpfs"
783 #define NFS4_NAME "nfs4"
784
785 asmlinkage long compat_sys_mount(const char __user * dev_name,
786 const char __user * dir_name,
787 const char __user * type, unsigned long flags,
788 const void __user * data)
789 {
790 char *kernel_type;
791 unsigned long data_page;
792 char *kernel_dev;
793 char *dir_page;
794 int retval;
795
796 retval = copy_mount_string(type, &kernel_type);
797 if (retval < 0)
798 goto out;
799
800 dir_page = getname(dir_name);
801 retval = PTR_ERR(dir_page);
802 if (IS_ERR(dir_page))
803 goto out1;
804
805 retval = copy_mount_string(dev_name, &kernel_dev);
806 if (retval < 0)
807 goto out2;
808
809 retval = copy_mount_options(data, &data_page);
810 if (retval < 0)
811 goto out3;
812
813 retval = -EINVAL;
814
815 if (kernel_type && data_page) {
816 if (!strcmp(kernel_type, NCPFS_NAME)) {
817 do_ncp_super_data_conv((void *)data_page);
818 } else if (!strcmp(kernel_type, NFS4_NAME)) {
819 if (do_nfs4_super_data_conv((void *) data_page))
820 goto out4;
821 }
822 }
823
824 retval = do_mount(kernel_dev, dir_page, kernel_type,
825 flags, (void*)data_page);
826
827 out4:
828 free_page(data_page);
829 out3:
830 kfree(kernel_dev);
831 out2:
832 putname(dir_page);
833 out1:
834 kfree(kernel_type);
835 out:
836 return retval;
837 }
838
839 struct compat_old_linux_dirent {
840 compat_ulong_t d_ino;
841 compat_ulong_t d_offset;
842 unsigned short d_namlen;
843 char d_name[1];
844 };
845
846 struct compat_readdir_callback {
847 struct compat_old_linux_dirent __user *dirent;
848 int result;
849 };
850
851 static int compat_fillonedir(void *__buf, const char *name, int namlen,
852 loff_t offset, u64 ino, unsigned int d_type)
853 {
854 struct compat_readdir_callback *buf = __buf;
855 struct compat_old_linux_dirent __user *dirent;
856 compat_ulong_t d_ino;
857
858 if (buf->result)
859 return -EINVAL;
860 d_ino = ino;
861 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
862 buf->result = -EOVERFLOW;
863 return -EOVERFLOW;
864 }
865 buf->result++;
866 dirent = buf->dirent;
867 if (!access_ok(VERIFY_WRITE, dirent,
868 (unsigned long)(dirent->d_name + namlen + 1) -
869 (unsigned long)dirent))
870 goto efault;
871 if ( __put_user(d_ino, &dirent->d_ino) ||
872 __put_user(offset, &dirent->d_offset) ||
873 __put_user(namlen, &dirent->d_namlen) ||
874 __copy_to_user(dirent->d_name, name, namlen) ||
875 __put_user(0, dirent->d_name + namlen))
876 goto efault;
877 return 0;
878 efault:
879 buf->result = -EFAULT;
880 return -EFAULT;
881 }
882
883 asmlinkage long compat_sys_old_readdir(unsigned int fd,
884 struct compat_old_linux_dirent __user *dirent, unsigned int count)
885 {
886 int error;
887 struct file *file;
888 struct compat_readdir_callback buf;
889
890 error = -EBADF;
891 file = fget(fd);
892 if (!file)
893 goto out;
894
895 buf.result = 0;
896 buf.dirent = dirent;
897
898 error = vfs_readdir(file, compat_fillonedir, &buf);
899 if (buf.result)
900 error = buf.result;
901
902 fput(file);
903 out:
904 return error;
905 }
906
907 struct compat_linux_dirent {
908 compat_ulong_t d_ino;
909 compat_ulong_t d_off;
910 unsigned short d_reclen;
911 char d_name[1];
912 };
913
914 struct compat_getdents_callback {
915 struct compat_linux_dirent __user *current_dir;
916 struct compat_linux_dirent __user *previous;
917 int count;
918 int error;
919 };
920
921 static int compat_filldir(void *__buf, const char *name, int namlen,
922 loff_t offset, u64 ino, unsigned int d_type)
923 {
924 struct compat_linux_dirent __user * dirent;
925 struct compat_getdents_callback *buf = __buf;
926 compat_ulong_t d_ino;
927 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
928 namlen + 2, sizeof(compat_long_t));
929
930 buf->error = -EINVAL; /* only used if we fail.. */
931 if (reclen > buf->count)
932 return -EINVAL;
933 d_ino = ino;
934 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
935 buf->error = -EOVERFLOW;
936 return -EOVERFLOW;
937 }
938 dirent = buf->previous;
939 if (dirent) {
940 if (__put_user(offset, &dirent->d_off))
941 goto efault;
942 }
943 dirent = buf->current_dir;
944 if (__put_user(d_ino, &dirent->d_ino))
945 goto efault;
946 if (__put_user(reclen, &dirent->d_reclen))
947 goto efault;
948 if (copy_to_user(dirent->d_name, name, namlen))
949 goto efault;
950 if (__put_user(0, dirent->d_name + namlen))
951 goto efault;
952 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
953 goto efault;
954 buf->previous = dirent;
955 dirent = (void __user *)dirent + reclen;
956 buf->current_dir = dirent;
957 buf->count -= reclen;
958 return 0;
959 efault:
960 buf->error = -EFAULT;
961 return -EFAULT;
962 }
963
964 asmlinkage long compat_sys_getdents(unsigned int fd,
965 struct compat_linux_dirent __user *dirent, unsigned int count)
966 {
967 struct file * file;
968 struct compat_linux_dirent __user * lastdirent;
969 struct compat_getdents_callback buf;
970 int error;
971
972 error = -EFAULT;
973 if (!access_ok(VERIFY_WRITE, dirent, count))
974 goto out;
975
976 error = -EBADF;
977 file = fget(fd);
978 if (!file)
979 goto out;
980
981 buf.current_dir = dirent;
982 buf.previous = NULL;
983 buf.count = count;
984 buf.error = 0;
985
986 error = vfs_readdir(file, compat_filldir, &buf);
987 if (error >= 0)
988 error = buf.error;
989 lastdirent = buf.previous;
990 if (lastdirent) {
991 if (put_user(file->f_pos, &lastdirent->d_off))
992 error = -EFAULT;
993 else
994 error = count - buf.count;
995 }
996 fput(file);
997 out:
998 return error;
999 }
1000
1001 #ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
1002
1003 struct compat_getdents_callback64 {
1004 struct linux_dirent64 __user *current_dir;
1005 struct linux_dirent64 __user *previous;
1006 int count;
1007 int error;
1008 };
1009
1010 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
1011 u64 ino, unsigned int d_type)
1012 {
1013 struct linux_dirent64 __user *dirent;
1014 struct compat_getdents_callback64 *buf = __buf;
1015 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
1016 sizeof(u64));
1017 u64 off;
1018
1019 buf->error = -EINVAL; /* only used if we fail.. */
1020 if (reclen > buf->count)
1021 return -EINVAL;
1022 dirent = buf->previous;
1023
1024 if (dirent) {
1025 if (__put_user_unaligned(offset, &dirent->d_off))
1026 goto efault;
1027 }
1028 dirent = buf->current_dir;
1029 if (__put_user_unaligned(ino, &dirent->d_ino))
1030 goto efault;
1031 off = 0;
1032 if (__put_user_unaligned(off, &dirent->d_off))
1033 goto efault;
1034 if (__put_user(reclen, &dirent->d_reclen))
1035 goto efault;
1036 if (__put_user(d_type, &dirent->d_type))
1037 goto efault;
1038 if (copy_to_user(dirent->d_name, name, namlen))
1039 goto efault;
1040 if (__put_user(0, dirent->d_name + namlen))
1041 goto efault;
1042 buf->previous = dirent;
1043 dirent = (void __user *)dirent + reclen;
1044 buf->current_dir = dirent;
1045 buf->count -= reclen;
1046 return 0;
1047 efault:
1048 buf->error = -EFAULT;
1049 return -EFAULT;
1050 }
1051
1052 asmlinkage long compat_sys_getdents64(unsigned int fd,
1053 struct linux_dirent64 __user * dirent, unsigned int count)
1054 {
1055 struct file * file;
1056 struct linux_dirent64 __user * lastdirent;
1057 struct compat_getdents_callback64 buf;
1058 int error;
1059
1060 error = -EFAULT;
1061 if (!access_ok(VERIFY_WRITE, dirent, count))
1062 goto out;
1063
1064 error = -EBADF;
1065 file = fget(fd);
1066 if (!file)
1067 goto out;
1068
1069 buf.current_dir = dirent;
1070 buf.previous = NULL;
1071 buf.count = count;
1072 buf.error = 0;
1073
1074 error = vfs_readdir(file, compat_filldir64, &buf);
1075 if (error >= 0)
1076 error = buf.error;
1077 lastdirent = buf.previous;
1078 if (lastdirent) {
1079 typeof(lastdirent->d_off) d_off = file->f_pos;
1080 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1081 error = -EFAULT;
1082 else
1083 error = count - buf.count;
1084 }
1085 fput(file);
1086 out:
1087 return error;
1088 }
1089 #endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
1090
1091 static ssize_t compat_do_readv_writev(int type, struct file *file,
1092 const struct compat_iovec __user *uvector,
1093 unsigned long nr_segs, loff_t *pos)
1094 {
1095 compat_ssize_t tot_len;
1096 struct iovec iovstack[UIO_FASTIOV];
1097 struct iovec *iov = iovstack;
1098 ssize_t ret;
1099 io_fn_t fn;
1100 iov_fn_t fnv;
1101
1102 ret = -EINVAL;
1103 if (!file->f_op)
1104 goto out;
1105
1106 ret = -EFAULT;
1107 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
1108 goto out;
1109
1110 tot_len = compat_rw_copy_check_uvector(type, uvector, nr_segs,
1111 UIO_FASTIOV, iovstack, &iov);
1112 if (tot_len == 0) {
1113 ret = 0;
1114 goto out;
1115 }
1116
1117 ret = rw_verify_area(type, file, pos, tot_len);
1118 if (ret < 0)
1119 goto out;
1120
1121 fnv = NULL;
1122 if (type == READ) {
1123 fn = file->f_op->read;
1124 fnv = file->f_op->aio_read;
1125 } else {
1126 fn = (io_fn_t)file->f_op->write;
1127 fnv = file->f_op->aio_write;
1128 }
1129
1130 if (fnv)
1131 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
1132 pos, fnv);
1133 else
1134 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
1135
1136 out:
1137 if (iov != iovstack)
1138 kfree(iov);
1139 if ((ret + (type == READ)) > 0) {
1140 if (type == READ)
1141 fsnotify_access(file);
1142 else
1143 fsnotify_modify(file);
1144 }
1145 return ret;
1146 }
1147
1148 static size_t compat_readv(struct file *file,
1149 const struct compat_iovec __user *vec,
1150 unsigned long vlen, loff_t *pos)
1151 {
1152 ssize_t ret = -EBADF;
1153
1154 if (!(file->f_mode & FMODE_READ))
1155 goto out;
1156
1157 ret = -EINVAL;
1158 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
1159 goto out;
1160
1161 ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
1162
1163 out:
1164 if (ret > 0)
1165 add_rchar(current, ret);
1166 inc_syscr(current);
1167 return ret;
1168 }
1169
1170 asmlinkage ssize_t
1171 compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec,
1172 unsigned long vlen)
1173 {
1174 struct file *file;
1175 int fput_needed;
1176 ssize_t ret;
1177
1178 file = fget_light(fd, &fput_needed);
1179 if (!file)
1180 return -EBADF;
1181 ret = compat_readv(file, vec, vlen, &file->f_pos);
1182 fput_light(file, fput_needed);
1183 return ret;
1184 }
1185
1186 asmlinkage ssize_t
1187 compat_sys_preadv(unsigned long fd, const struct compat_iovec __user *vec,
1188 unsigned long vlen, u32 pos_low, u32 pos_high)
1189 {
1190 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1191 struct file *file;
1192 int fput_needed;
1193 ssize_t ret;
1194
1195 if (pos < 0)
1196 return -EINVAL;
1197 file = fget_light(fd, &fput_needed);
1198 if (!file)
1199 return -EBADF;
1200 ret = -ESPIPE;
1201 if (file->f_mode & FMODE_PREAD)
1202 ret = compat_readv(file, vec, vlen, &pos);
1203 fput_light(file, fput_needed);
1204 return ret;
1205 }
1206
1207 static size_t compat_writev(struct file *file,
1208 const struct compat_iovec __user *vec,
1209 unsigned long vlen, loff_t *pos)
1210 {
1211 ssize_t ret = -EBADF;
1212
1213 if (!(file->f_mode & FMODE_WRITE))
1214 goto out;
1215
1216 ret = -EINVAL;
1217 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
1218 goto out;
1219
1220 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
1221
1222 out:
1223 if (ret > 0)
1224 add_wchar(current, ret);
1225 inc_syscw(current);
1226 return ret;
1227 }
1228
1229 asmlinkage ssize_t
1230 compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec,
1231 unsigned long vlen)
1232 {
1233 struct file *file;
1234 int fput_needed;
1235 ssize_t ret;
1236
1237 file = fget_light(fd, &fput_needed);
1238 if (!file)
1239 return -EBADF;
1240 ret = compat_writev(file, vec, vlen, &file->f_pos);
1241 fput_light(file, fput_needed);
1242 return ret;
1243 }
1244
1245 asmlinkage ssize_t
1246 compat_sys_pwritev(unsigned long fd, const struct compat_iovec __user *vec,
1247 unsigned long vlen, u32 pos_low, u32 pos_high)
1248 {
1249 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1250 struct file *file;
1251 int fput_needed;
1252 ssize_t ret;
1253
1254 if (pos < 0)
1255 return -EINVAL;
1256 file = fget_light(fd, &fput_needed);
1257 if (!file)
1258 return -EBADF;
1259 ret = -ESPIPE;
1260 if (file->f_mode & FMODE_PWRITE)
1261 ret = compat_writev(file, vec, vlen, &pos);
1262 fput_light(file, fput_needed);
1263 return ret;
1264 }
1265
1266 asmlinkage long
1267 compat_sys_vmsplice(int fd, const struct compat_iovec __user *iov32,
1268 unsigned int nr_segs, unsigned int flags)
1269 {
1270 unsigned i;
1271 struct iovec __user *iov;
1272 if (nr_segs > UIO_MAXIOV)
1273 return -EINVAL;
1274 iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
1275 for (i = 0; i < nr_segs; i++) {
1276 struct compat_iovec v;
1277 if (get_user(v.iov_base, &iov32[i].iov_base) ||
1278 get_user(v.iov_len, &iov32[i].iov_len) ||
1279 put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
1280 put_user(v.iov_len, &iov[i].iov_len))
1281 return -EFAULT;
1282 }
1283 return sys_vmsplice(fd, iov, nr_segs, flags);
1284 }
1285
1286 /*
1287 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1288 * O_LARGEFILE flag.
1289 */
1290 asmlinkage long
1291 compat_sys_open(const char __user *filename, int flags, int mode)
1292 {
1293 return do_sys_open(AT_FDCWD, filename, flags, mode);
1294 }
1295
1296 /*
1297 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1298 * O_LARGEFILE flag.
1299 */
1300 asmlinkage long
1301 compat_sys_openat(unsigned int dfd, const char __user *filename, int flags, int mode)
1302 {
1303 return do_sys_open(dfd, filename, flags, mode);
1304 }
1305
1306 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1307
1308 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1309 int timeval, int ret)
1310 {
1311 struct timespec ts;
1312
1313 if (!p)
1314 return ret;
1315
1316 if (current->personality & STICKY_TIMEOUTS)
1317 goto sticky;
1318
1319 /* No update for zero timeout */
1320 if (!end_time->tv_sec && !end_time->tv_nsec)
1321 return ret;
1322
1323 ktime_get_ts(&ts);
1324 ts = timespec_sub(*end_time, ts);
1325 if (ts.tv_sec < 0)
1326 ts.tv_sec = ts.tv_nsec = 0;
1327
1328 if (timeval) {
1329 struct compat_timeval rtv;
1330
1331 rtv.tv_sec = ts.tv_sec;
1332 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1333
1334 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1335 return ret;
1336 } else {
1337 struct compat_timespec rts;
1338
1339 rts.tv_sec = ts.tv_sec;
1340 rts.tv_nsec = ts.tv_nsec;
1341
1342 if (!copy_to_user(p, &rts, sizeof(rts)))
1343 return ret;
1344 }
1345 /*
1346 * If an application puts its timeval in read-only memory, we
1347 * don't want the Linux-specific update to the timeval to
1348 * cause a fault after the select has completed
1349 * successfully. However, because we're not updating the
1350 * timeval, we can't restart the system call.
1351 */
1352
1353 sticky:
1354 if (ret == -ERESTARTNOHAND)
1355 ret = -EINTR;
1356 return ret;
1357 }
1358
1359 /*
1360 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1361 * 64-bit unsigned longs.
1362 */
1363 static
1364 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1365 unsigned long *fdset)
1366 {
1367 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1368 if (ufdset) {
1369 unsigned long odd;
1370
1371 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1372 return -EFAULT;
1373
1374 odd = nr & 1UL;
1375 nr &= ~1UL;
1376 while (nr) {
1377 unsigned long h, l;
1378 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1379 return -EFAULT;
1380 ufdset += 2;
1381 *fdset++ = h << 32 | l;
1382 nr -= 2;
1383 }
1384 if (odd && __get_user(*fdset, ufdset))
1385 return -EFAULT;
1386 } else {
1387 /* Tricky, must clear full unsigned long in the
1388 * kernel fdset at the end, this makes sure that
1389 * actually happens.
1390 */
1391 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1392 }
1393 return 0;
1394 }
1395
1396 static
1397 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1398 unsigned long *fdset)
1399 {
1400 unsigned long odd;
1401 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1402
1403 if (!ufdset)
1404 return 0;
1405
1406 odd = nr & 1UL;
1407 nr &= ~1UL;
1408 while (nr) {
1409 unsigned long h, l;
1410 l = *fdset++;
1411 h = l >> 32;
1412 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1413 return -EFAULT;
1414 ufdset += 2;
1415 nr -= 2;
1416 }
1417 if (odd && __put_user(*fdset, ufdset))
1418 return -EFAULT;
1419 return 0;
1420 }
1421
1422
1423 /*
1424 * This is a virtual copy of sys_select from fs/select.c and probably
1425 * should be compared to it from time to time
1426 */
1427
1428 /*
1429 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1430 * like to be certain this leads to no problems. So I return
1431 * EINTR just for safety.
1432 *
1433 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1434 * I'm trying ERESTARTNOHAND which restart only when you want to.
1435 */
1436 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1437 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1438 struct timespec *end_time)
1439 {
1440 fd_set_bits fds;
1441 void *bits;
1442 int size, max_fds, ret = -EINVAL;
1443 struct fdtable *fdt;
1444 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1445
1446 if (n < 0)
1447 goto out_nofds;
1448
1449 /* max_fds can increase, so grab it once to avoid race */
1450 rcu_read_lock();
1451 fdt = files_fdtable(current->files);
1452 max_fds = fdt->max_fds;
1453 rcu_read_unlock();
1454 if (n > max_fds)
1455 n = max_fds;
1456
1457 /*
1458 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1459 * since we used fdset we need to allocate memory in units of
1460 * long-words.
1461 */
1462 size = FDS_BYTES(n);
1463 bits = stack_fds;
1464 if (size > sizeof(stack_fds) / 6) {
1465 bits = kmalloc(6 * size, GFP_KERNEL);
1466 ret = -ENOMEM;
1467 if (!bits)
1468 goto out_nofds;
1469 }
1470 fds.in = (unsigned long *) bits;
1471 fds.out = (unsigned long *) (bits + size);
1472 fds.ex = (unsigned long *) (bits + 2*size);
1473 fds.res_in = (unsigned long *) (bits + 3*size);
1474 fds.res_out = (unsigned long *) (bits + 4*size);
1475 fds.res_ex = (unsigned long *) (bits + 5*size);
1476
1477 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1478 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1479 (ret = compat_get_fd_set(n, exp, fds.ex)))
1480 goto out;
1481 zero_fd_set(n, fds.res_in);
1482 zero_fd_set(n, fds.res_out);
1483 zero_fd_set(n, fds.res_ex);
1484
1485 ret = do_select(n, &fds, end_time);
1486
1487 if (ret < 0)
1488 goto out;
1489 if (!ret) {
1490 ret = -ERESTARTNOHAND;
1491 if (signal_pending(current))
1492 goto out;
1493 ret = 0;
1494 }
1495
1496 if (compat_set_fd_set(n, inp, fds.res_in) ||
1497 compat_set_fd_set(n, outp, fds.res_out) ||
1498 compat_set_fd_set(n, exp, fds.res_ex))
1499 ret = -EFAULT;
1500 out:
1501 if (bits != stack_fds)
1502 kfree(bits);
1503 out_nofds:
1504 return ret;
1505 }
1506
1507 asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
1508 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1509 struct compat_timeval __user *tvp)
1510 {
1511 struct timespec end_time, *to = NULL;
1512 struct compat_timeval tv;
1513 int ret;
1514
1515 if (tvp) {
1516 if (copy_from_user(&tv, tvp, sizeof(tv)))
1517 return -EFAULT;
1518
1519 to = &end_time;
1520 if (poll_select_set_timeout(to,
1521 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1522 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1523 return -EINVAL;
1524 }
1525
1526 ret = compat_core_sys_select(n, inp, outp, exp, to);
1527 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1528
1529 return ret;
1530 }
1531
1532 struct compat_sel_arg_struct {
1533 compat_ulong_t n;
1534 compat_uptr_t inp;
1535 compat_uptr_t outp;
1536 compat_uptr_t exp;
1537 compat_uptr_t tvp;
1538 };
1539
1540 asmlinkage long compat_sys_old_select(struct compat_sel_arg_struct __user *arg)
1541 {
1542 struct compat_sel_arg_struct a;
1543
1544 if (copy_from_user(&a, arg, sizeof(a)))
1545 return -EFAULT;
1546 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1547 compat_ptr(a.exp), compat_ptr(a.tvp));
1548 }
1549
1550 #ifdef HAVE_SET_RESTORE_SIGMASK
1551 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1552 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1553 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1554 compat_size_t sigsetsize)
1555 {
1556 compat_sigset_t ss32;
1557 sigset_t ksigmask, sigsaved;
1558 struct compat_timespec ts;
1559 struct timespec end_time, *to = NULL;
1560 int ret;
1561
1562 if (tsp) {
1563 if (copy_from_user(&ts, tsp, sizeof(ts)))
1564 return -EFAULT;
1565
1566 to = &end_time;
1567 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1568 return -EINVAL;
1569 }
1570
1571 if (sigmask) {
1572 if (sigsetsize != sizeof(compat_sigset_t))
1573 return -EINVAL;
1574 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1575 return -EFAULT;
1576 sigset_from_compat(&ksigmask, &ss32);
1577
1578 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1579 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1580 }
1581
1582 ret = compat_core_sys_select(n, inp, outp, exp, to);
1583 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1584
1585 if (ret == -ERESTARTNOHAND) {
1586 /*
1587 * Don't restore the signal mask yet. Let do_signal() deliver
1588 * the signal on the way back to userspace, before the signal
1589 * mask is restored.
1590 */
1591 if (sigmask) {
1592 memcpy(&current->saved_sigmask, &sigsaved,
1593 sizeof(sigsaved));
1594 set_restore_sigmask();
1595 }
1596 } else if (sigmask)
1597 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1598
1599 return ret;
1600 }
1601
1602 asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
1603 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1604 struct compat_timespec __user *tsp, void __user *sig)
1605 {
1606 compat_size_t sigsetsize = 0;
1607 compat_uptr_t up = 0;
1608
1609 if (sig) {
1610 if (!access_ok(VERIFY_READ, sig,
1611 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1612 __get_user(up, (compat_uptr_t __user *)sig) ||
1613 __get_user(sigsetsize,
1614 (compat_size_t __user *)(sig+sizeof(up))))
1615 return -EFAULT;
1616 }
1617 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1618 sigsetsize);
1619 }
1620
1621 asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1622 unsigned int nfds, struct compat_timespec __user *tsp,
1623 const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
1624 {
1625 compat_sigset_t ss32;
1626 sigset_t ksigmask, sigsaved;
1627 struct compat_timespec ts;
1628 struct timespec end_time, *to = NULL;
1629 int ret;
1630
1631 if (tsp) {
1632 if (copy_from_user(&ts, tsp, sizeof(ts)))
1633 return -EFAULT;
1634
1635 to = &end_time;
1636 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1637 return -EINVAL;
1638 }
1639
1640 if (sigmask) {
1641 if (sigsetsize != sizeof(compat_sigset_t))
1642 return -EINVAL;
1643 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1644 return -EFAULT;
1645 sigset_from_compat(&ksigmask, &ss32);
1646
1647 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1648 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1649 }
1650
1651 ret = do_sys_poll(ufds, nfds, to);
1652
1653 /* We can restart this syscall, usually */
1654 if (ret == -EINTR) {
1655 /*
1656 * Don't restore the signal mask yet. Let do_signal() deliver
1657 * the signal on the way back to userspace, before the signal
1658 * mask is restored.
1659 */
1660 if (sigmask) {
1661 memcpy(&current->saved_sigmask, &sigsaved,
1662 sizeof(sigsaved));
1663 set_restore_sigmask();
1664 }
1665 ret = -ERESTARTNOHAND;
1666 } else if (sigmask)
1667 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1668
1669 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1670
1671 return ret;
1672 }
1673 #endif /* HAVE_SET_RESTORE_SIGMASK */
1674
1675 #ifdef CONFIG_EPOLL
1676
1677 #ifdef HAVE_SET_RESTORE_SIGMASK
1678 asmlinkage long compat_sys_epoll_pwait(int epfd,
1679 struct compat_epoll_event __user *events,
1680 int maxevents, int timeout,
1681 const compat_sigset_t __user *sigmask,
1682 compat_size_t sigsetsize)
1683 {
1684 long err;
1685 compat_sigset_t csigmask;
1686 sigset_t ksigmask, sigsaved;
1687
1688 /*
1689 * If the caller wants a certain signal mask to be set during the wait,
1690 * we apply it here.
1691 */
1692 if (sigmask) {
1693 if (sigsetsize != sizeof(compat_sigset_t))
1694 return -EINVAL;
1695 if (copy_from_user(&csigmask, sigmask, sizeof(csigmask)))
1696 return -EFAULT;
1697 sigset_from_compat(&ksigmask, &csigmask);
1698 sigdelsetmask(&ksigmask, sigmask(SIGKILL) | sigmask(SIGSTOP));
1699 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1700 }
1701
1702 err = sys_epoll_wait(epfd, events, maxevents, timeout);
1703
1704 /*
1705 * If we changed the signal mask, we need to restore the original one.
1706 * In case we've got a signal while waiting, we do not restore the
1707 * signal mask yet, and we allow do_signal() to deliver the signal on
1708 * the way back to userspace, before the signal mask is restored.
1709 */
1710 if (sigmask) {
1711 if (err == -EINTR) {
1712 memcpy(&current->saved_sigmask, &sigsaved,
1713 sizeof(sigsaved));
1714 set_restore_sigmask();
1715 } else
1716 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1717 }
1718
1719 return err;
1720 }
1721 #endif /* HAVE_SET_RESTORE_SIGMASK */
1722
1723 #endif /* CONFIG_EPOLL */
1724
1725 #ifdef CONFIG_SIGNALFD
1726
1727 asmlinkage long compat_sys_signalfd4(int ufd,
1728 const compat_sigset_t __user *sigmask,
1729 compat_size_t sigsetsize, int flags)
1730 {
1731 compat_sigset_t ss32;
1732 sigset_t tmp;
1733 sigset_t __user *ksigmask;
1734
1735 if (sigsetsize != sizeof(compat_sigset_t))
1736 return -EINVAL;
1737 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1738 return -EFAULT;
1739 sigset_from_compat(&tmp, &ss32);
1740 ksigmask = compat_alloc_user_space(sizeof(sigset_t));
1741 if (copy_to_user(ksigmask, &tmp, sizeof(sigset_t)))
1742 return -EFAULT;
1743
1744 return sys_signalfd4(ufd, ksigmask, sizeof(sigset_t), flags);
1745 }
1746
1747 asmlinkage long compat_sys_signalfd(int ufd,
1748 const compat_sigset_t __user *sigmask,
1749 compat_size_t sigsetsize)
1750 {
1751 return compat_sys_signalfd4(ufd, sigmask, sigsetsize, 0);
1752 }
1753 #endif /* CONFIG_SIGNALFD */
1754
1755 #ifdef CONFIG_TIMERFD
1756
1757 asmlinkage long compat_sys_timerfd_settime(int ufd, int flags,
1758 const struct compat_itimerspec __user *utmr,
1759 struct compat_itimerspec __user *otmr)
1760 {
1761 int error;
1762 struct itimerspec t;
1763 struct itimerspec __user *ut;
1764
1765 if (get_compat_itimerspec(&t, utmr))
1766 return -EFAULT;
1767 ut = compat_alloc_user_space(2 * sizeof(struct itimerspec));
1768 if (copy_to_user(&ut[0], &t, sizeof(t)))
1769 return -EFAULT;
1770 error = sys_timerfd_settime(ufd, flags, &ut[0], &ut[1]);
1771 if (!error && otmr)
1772 error = (copy_from_user(&t, &ut[1], sizeof(struct itimerspec)) ||
1773 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1774
1775 return error;
1776 }
1777
1778 asmlinkage long compat_sys_timerfd_gettime(int ufd,
1779 struct compat_itimerspec __user *otmr)
1780 {
1781 int error;
1782 struct itimerspec t;
1783 struct itimerspec __user *ut;
1784
1785 ut = compat_alloc_user_space(sizeof(struct itimerspec));
1786 error = sys_timerfd_gettime(ufd, ut);
1787 if (!error)
1788 error = (copy_from_user(&t, ut, sizeof(struct itimerspec)) ||
1789 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1790
1791 return error;
1792 }
1793
1794 #endif /* CONFIG_TIMERFD */
1795
1796 #ifdef CONFIG_FHANDLE
1797 /*
1798 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1799 * doesn't set the O_LARGEFILE flag.
1800 */
1801 asmlinkage long
1802 compat_sys_open_by_handle_at(int mountdirfd,
1803 struct file_handle __user *handle, int flags)
1804 {
1805 return do_handle_open(mountdirfd, handle, flags);
1806 }
1807 #endif
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree