?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
net: rose: restore old recvmsg behavior
[linux/fpc-iii.git] / fs / compat.c
blob4bf082dbc273dcf9ad8bdeac084400229e478b7b
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/personality.h>
41 #include <linux/rwsem.h>
42 #include <linux/tsacct_kern.h>
43 #include <linux/security.h>
44 #include <linux/highmem.h>
45 #include <linux/signal.h>
46 #include <linux/poll.h>
47 #include <linux/mm.h>
48 #include <linux/eventpoll.h>
49 #include <linux/fs_struct.h>
50 #include <linux/slab.h>
51 #include <linux/pagemap.h>
52
53 #include <asm/uaccess.h>
54 #include <asm/mmu_context.h>
55 #include <asm/ioctls.h>
56 #include "internal.h"
57
58 int compat_log = 1;
59
60 int compat_printk(const char *fmt, ...)
61 {
62 va_list ap;
63 int ret;
64 if (!compat_log)
65 return 0;
66 va_start(ap, fmt);
67 ret = vprintk(fmt, ap);
68 va_end(ap);
69 return ret;
70 }
71
72 #include "read_write.h"
73
74 /*
75 * Not all architectures have sys_utime, so implement this in terms
76 * of sys_utimes.
77 */
78 asmlinkage long compat_sys_utime(const char __user *filename,
79 struct compat_utimbuf __user *t)
80 {
81 struct timespec tv[2];
82
83 if (t) {
84 if (get_user(tv[0].tv_sec, &t->actime) ||
85 get_user(tv[1].tv_sec, &t->modtime))
86 return -EFAULT;
87 tv[0].tv_nsec = 0;
88 tv[1].tv_nsec = 0;
89 }
90 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
91 }
92
93 asmlinkage long compat_sys_utimensat(unsigned int dfd, const char __user *filename, struct compat_timespec __user *t, int flags)
94 {
95 struct timespec tv[2];
96
97 if (t) {
98 if (get_compat_timespec(&tv[0], &t[0]) ||
99 get_compat_timespec(&tv[1], &t[1]))
100 return -EFAULT;
101
102 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
103 return 0;
104 }
105 return do_utimes(dfd, filename, t ? tv : NULL, flags);
106 }
107
108 asmlinkage long compat_sys_futimesat(unsigned int dfd, const char __user *filename, struct compat_timeval __user *t)
109 {
110 struct timespec tv[2];
111
112 if (t) {
113 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
114 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
115 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
116 get_user(tv[1].tv_nsec, &t[1].tv_usec))
117 return -EFAULT;
118 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
119 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
120 return -EINVAL;
121 tv[0].tv_nsec *= 1000;
122 tv[1].tv_nsec *= 1000;
123 }
124 return do_utimes(dfd, filename, t ? tv : NULL, 0);
125 }
126
127 asmlinkage long compat_sys_utimes(const char __user *filename, struct compat_timeval __user *t)
128 {
129 return compat_sys_futimesat(AT_FDCWD, filename, t);
130 }
131
132 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
133 {
134 compat_ino_t ino = stat->ino;
135 typeof(ubuf->st_uid) uid = 0;
136 typeof(ubuf->st_gid) gid = 0;
137 int err;
138
139 SET_UID(uid, stat->uid);
140 SET_GID(gid, stat->gid);
141
142 if ((u64) stat->size > MAX_NON_LFS ||
143 !old_valid_dev(stat->dev) ||
144 !old_valid_dev(stat->rdev))
145 return -EOVERFLOW;
146 if (sizeof(ino) < sizeof(stat->ino) && ino != stat->ino)
147 return -EOVERFLOW;
148
149 if (clear_user(ubuf, sizeof(*ubuf)))
150 return -EFAULT;
151
152 err = __put_user(old_encode_dev(stat->dev), &ubuf->st_dev);
153 err |= __put_user(ino, &ubuf->st_ino);
154 err |= __put_user(stat->mode, &ubuf->st_mode);
155 err |= __put_user(stat->nlink, &ubuf->st_nlink);
156 err |= __put_user(uid, &ubuf->st_uid);
157 err |= __put_user(gid, &ubuf->st_gid);
158 err |= __put_user(old_encode_dev(stat->rdev), &ubuf->st_rdev);
159 err |= __put_user(stat->size, &ubuf->st_size);
160 err |= __put_user(stat->atime.tv_sec, &ubuf->st_atime);
161 err |= __put_user(stat->atime.tv_nsec, &ubuf->st_atime_nsec);
162 err |= __put_user(stat->mtime.tv_sec, &ubuf->st_mtime);
163 err |= __put_user(stat->mtime.tv_nsec, &ubuf->st_mtime_nsec);
164 err |= __put_user(stat->ctime.tv_sec, &ubuf->st_ctime);
165 err |= __put_user(stat->ctime.tv_nsec, &ubuf->st_ctime_nsec);
166 err |= __put_user(stat->blksize, &ubuf->st_blksize);
167 err |= __put_user(stat->blocks, &ubuf->st_blocks);
168 return err;
169 }
170
171 asmlinkage long compat_sys_newstat(const char __user * filename,
172 struct compat_stat __user *statbuf)
173 {
174 struct kstat stat;
175 int error;
176
177 error = vfs_stat(filename, &stat);
178 if (error)
179 return error;
180 return cp_compat_stat(&stat, statbuf);
181 }
182
183 asmlinkage long compat_sys_newlstat(const char __user * filename,
184 struct compat_stat __user *statbuf)
185 {
186 struct kstat stat;
187 int error;
188
189 error = vfs_lstat(filename, &stat);
190 if (error)
191 return error;
192 return cp_compat_stat(&stat, statbuf);
193 }
194
195 #ifndef __ARCH_WANT_STAT64
196 asmlinkage long compat_sys_newfstatat(unsigned int dfd,
197 const char __user *filename,
198 struct compat_stat __user *statbuf, int flag)
199 {
200 struct kstat stat;
201 int error;
202
203 error = vfs_fstatat(dfd, filename, &stat, flag);
204 if (error)
205 return error;
206 return cp_compat_stat(&stat, statbuf);
207 }
208 #endif
209
210 asmlinkage long compat_sys_newfstat(unsigned int fd,
211 struct compat_stat __user * statbuf)
212 {
213 struct kstat stat;
214 int error = vfs_fstat(fd, &stat);
215
216 if (!error)
217 error = cp_compat_stat(&stat, statbuf);
218 return error;
219 }
220
221 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
222 {
223
224 if (sizeof ubuf->f_blocks == 4) {
225 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
226 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
227 return -EOVERFLOW;
228 /* f_files and f_ffree may be -1; it's okay
229 * to stuff that into 32 bits */
230 if (kbuf->f_files != 0xffffffffffffffffULL
231 && (kbuf->f_files & 0xffffffff00000000ULL))
232 return -EOVERFLOW;
233 if (kbuf->f_ffree != 0xffffffffffffffffULL
234 && (kbuf->f_ffree & 0xffffffff00000000ULL))
235 return -EOVERFLOW;
236 }
237 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
238 __put_user(kbuf->f_type, &ubuf->f_type) ||
239 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
240 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
241 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
242 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
243 __put_user(kbuf->f_files, &ubuf->f_files) ||
244 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
245 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
246 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
247 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
248 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
249 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
250 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
251 return -EFAULT;
252 return 0;
253 }
254
255 /*
256 * The following statfs calls are copies of code from fs/statfs.c and
257 * should be checked against those from time to time
258 */
259 asmlinkage long compat_sys_statfs(const char __user *pathname, struct compat_statfs __user *buf)
260 {
261 struct kstatfs tmp;
262 int error = user_statfs(pathname, &tmp);
263 if (!error)
264 error = put_compat_statfs(buf, &tmp);
265 return error;
266 }
267
268 asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
269 {
270 struct kstatfs tmp;
271 int error = fd_statfs(fd, &tmp);
272 if (!error)
273 error = put_compat_statfs(buf, &tmp);
274 return error;
275 }
276
277 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
278 {
279 if (sizeof ubuf->f_blocks == 4) {
280 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
281 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
282 return -EOVERFLOW;
283 /* f_files and f_ffree may be -1; it's okay
284 * to stuff that into 32 bits */
285 if (kbuf->f_files != 0xffffffffffffffffULL
286 && (kbuf->f_files & 0xffffffff00000000ULL))
287 return -EOVERFLOW;
288 if (kbuf->f_ffree != 0xffffffffffffffffULL
289 && (kbuf->f_ffree & 0xffffffff00000000ULL))
290 return -EOVERFLOW;
291 }
292 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
293 __put_user(kbuf->f_type, &ubuf->f_type) ||
294 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
295 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
296 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
297 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
298 __put_user(kbuf->f_files, &ubuf->f_files) ||
299 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
300 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
301 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
302 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
303 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
304 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
305 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
306 return -EFAULT;
307 return 0;
308 }
309
310 asmlinkage long compat_sys_statfs64(const char __user *pathname, compat_size_t sz, struct compat_statfs64 __user *buf)
311 {
312 struct kstatfs tmp;
313 int error;
314
315 if (sz != sizeof(*buf))
316 return -EINVAL;
317
318 error = user_statfs(pathname, &tmp);
319 if (!error)
320 error = put_compat_statfs64(buf, &tmp);
321 return error;
322 }
323
324 asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
325 {
326 struct kstatfs tmp;
327 int error;
328
329 if (sz != sizeof(*buf))
330 return -EINVAL;
331
332 error = fd_statfs(fd, &tmp);
333 if (!error)
334 error = put_compat_statfs64(buf, &tmp);
335 return error;
336 }
337
338 /*
339 * This is a copy of sys_ustat, just dealing with a structure layout.
340 * Given how simple this syscall is that apporach is more maintainable
341 * than the various conversion hacks.
342 */
343 asmlinkage long compat_sys_ustat(unsigned dev, struct compat_ustat __user *u)
344 {
345 struct super_block *sb;
346 struct compat_ustat tmp;
347 struct kstatfs sbuf;
348 int err;
349
350 sb = user_get_super(new_decode_dev(dev));
351 if (!sb)
352 return -EINVAL;
353 err = statfs_by_dentry(sb->s_root, &sbuf);
354 drop_super(sb);
355 if (err)
356 return err;
357
358 memset(&tmp, 0, sizeof(struct compat_ustat));
359 tmp.f_tfree = sbuf.f_bfree;
360 tmp.f_tinode = sbuf.f_ffree;
361 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
362 return -EFAULT;
363 return 0;
364 }
365
366 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
367 {
368 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
369 __get_user(kfl->l_type, &ufl->l_type) ||
370 __get_user(kfl->l_whence, &ufl->l_whence) ||
371 __get_user(kfl->l_start, &ufl->l_start) ||
372 __get_user(kfl->l_len, &ufl->l_len) ||
373 __get_user(kfl->l_pid, &ufl->l_pid))
374 return -EFAULT;
375 return 0;
376 }
377
378 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
379 {
380 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
381 __put_user(kfl->l_type, &ufl->l_type) ||
382 __put_user(kfl->l_whence, &ufl->l_whence) ||
383 __put_user(kfl->l_start, &ufl->l_start) ||
384 __put_user(kfl->l_len, &ufl->l_len) ||
385 __put_user(kfl->l_pid, &ufl->l_pid))
386 return -EFAULT;
387 return 0;
388 }
389
390 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
391 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
392 {
393 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
394 __get_user(kfl->l_type, &ufl->l_type) ||
395 __get_user(kfl->l_whence, &ufl->l_whence) ||
396 __get_user(kfl->l_start, &ufl->l_start) ||
397 __get_user(kfl->l_len, &ufl->l_len) ||
398 __get_user(kfl->l_pid, &ufl->l_pid))
399 return -EFAULT;
400 return 0;
401 }
402 #endif
403
404 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
405 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
406 {
407 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
408 __put_user(kfl->l_type, &ufl->l_type) ||
409 __put_user(kfl->l_whence, &ufl->l_whence) ||
410 __put_user(kfl->l_start, &ufl->l_start) ||
411 __put_user(kfl->l_len, &ufl->l_len) ||
412 __put_user(kfl->l_pid, &ufl->l_pid))
413 return -EFAULT;
414 return 0;
415 }
416 #endif
417
418 asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
419 unsigned long arg)
420 {
421 mm_segment_t old_fs;
422 struct flock f;
423 long ret;
424
425 switch (cmd) {
426 case F_GETLK:
427 case F_SETLK:
428 case F_SETLKW:
429 ret = get_compat_flock(&f, compat_ptr(arg));
430 if (ret != 0)
431 break;
432 old_fs = get_fs();
433 set_fs(KERNEL_DS);
434 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
435 set_fs(old_fs);
436 if (cmd == F_GETLK && ret == 0) {
437 /* GETLK was successful and we need to return the data...
438 * but it needs to fit in the compat structure.
439 * l_start shouldn't be too big, unless the original
440 * start + end is greater than COMPAT_OFF_T_MAX, in which
441 * case the app was asking for trouble, so we return
442 * -EOVERFLOW in that case.
443 * l_len could be too big, in which case we just truncate it,
444 * and only allow the app to see that part of the conflicting
445 * lock that might make sense to it anyway
446 */
447
448 if (f.l_start > COMPAT_OFF_T_MAX)
449 ret = -EOVERFLOW;
450 if (f.l_len > COMPAT_OFF_T_MAX)
451 f.l_len = COMPAT_OFF_T_MAX;
452 if (ret == 0)
453 ret = put_compat_flock(&f, compat_ptr(arg));
454 }
455 break;
456
457 case F_GETLK64:
458 case F_SETLK64:
459 case F_SETLKW64:
460 ret = get_compat_flock64(&f, compat_ptr(arg));
461 if (ret != 0)
462 break;
463 old_fs = get_fs();
464 set_fs(KERNEL_DS);
465 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
466 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
467 (unsigned long)&f);
468 set_fs(old_fs);
469 if (cmd == F_GETLK64 && ret == 0) {
470 /* need to return lock information - see above for commentary */
471 if (f.l_start > COMPAT_LOFF_T_MAX)
472 ret = -EOVERFLOW;
473 if (f.l_len > COMPAT_LOFF_T_MAX)
474 f.l_len = COMPAT_LOFF_T_MAX;
475 if (ret == 0)
476 ret = put_compat_flock64(&f, compat_ptr(arg));
477 }
478 break;
479
480 default:
481 ret = sys_fcntl(fd, cmd, arg);
482 break;
483 }
484 return ret;
485 }
486
487 asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
488 unsigned long arg)
489 {
490 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
491 return -EINVAL;
492 return compat_sys_fcntl64(fd, cmd, arg);
493 }
494
495 asmlinkage long
496 compat_sys_io_setup(unsigned nr_reqs, u32 __user *ctx32p)
497 {
498 long ret;
499 aio_context_t ctx64;
500
501 mm_segment_t oldfs = get_fs();
502 if (unlikely(get_user(ctx64, ctx32p)))
503 return -EFAULT;
504
505 set_fs(KERNEL_DS);
506 /* The __user pointer cast is valid because of the set_fs() */
507 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
508 set_fs(oldfs);
509 /* truncating is ok because it's a user address */
510 if (!ret)
511 ret = put_user((u32) ctx64, ctx32p);
512 return ret;
513 }
514
515 asmlinkage long
516 compat_sys_io_getevents(aio_context_t ctx_id,
517 unsigned long min_nr,
518 unsigned long nr,
519 struct io_event __user *events,
520 struct compat_timespec __user *timeout)
521 {
522 long ret;
523 struct timespec t;
524 struct timespec __user *ut = NULL;
525
526 ret = -EFAULT;
527 if (unlikely(!access_ok(VERIFY_WRITE, events,
528 nr * sizeof(struct io_event))))
529 goto out;
530 if (timeout) {
531 if (get_compat_timespec(&t, timeout))
532 goto out;
533
534 ut = compat_alloc_user_space(sizeof(*ut));
535 if (copy_to_user(ut, &t, sizeof(t)) )
536 goto out;
537 }
538 ret = sys_io_getevents(ctx_id, min_nr, nr, events, ut);
539 out:
540 return ret;
541 }
542
543 /* A write operation does a read from user space and vice versa */
544 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
545
546 ssize_t compat_rw_copy_check_uvector(int type,
547 const struct compat_iovec __user *uvector, unsigned long nr_segs,
548 unsigned long fast_segs, struct iovec *fast_pointer,
549 struct iovec **ret_pointer, int check_access)
550 {
551 compat_ssize_t tot_len;
552 struct iovec *iov = *ret_pointer = fast_pointer;
553 ssize_t ret = 0;
554 int seg;
555
556 /*
557 * SuS says "The readv() function *may* fail if the iovcnt argument
558 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
559 * traditionally returned zero for zero segments, so...
560 */
561 if (nr_segs == 0)
562 goto out;
563
564 ret = -EINVAL;
565 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
566 goto out;
567 if (nr_segs > fast_segs) {
568 ret = -ENOMEM;
569 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
570 if (iov == NULL)
571 goto out;
572 }
573 *ret_pointer = iov;
574
575 ret = -EFAULT;
576 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
577 goto out;
578
579 /*
580 * Single unix specification:
581 * We should -EINVAL if an element length is not >= 0 and fitting an
582 * ssize_t.
583 *
584 * In Linux, the total length is limited to MAX_RW_COUNT, there is
585 * no overflow possibility.
586 */
587 tot_len = 0;
588 ret = -EINVAL;
589 for (seg = 0; seg < nr_segs; seg++) {
590 compat_uptr_t buf;
591 compat_ssize_t len;
592
593 if (__get_user(len, &uvector->iov_len) ||
594 __get_user(buf, &uvector->iov_base)) {
595 ret = -EFAULT;
596 goto out;
597 }
598 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
599 goto out;
600 if (check_access &&
601 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
602 ret = -EFAULT;
603 goto out;
604 }
605 if (len > MAX_RW_COUNT - tot_len)
606 len = MAX_RW_COUNT - tot_len;
607 tot_len += len;
608 iov->iov_base = compat_ptr(buf);
609 iov->iov_len = (compat_size_t) len;
610 uvector++;
611 iov++;
612 }
613 ret = tot_len;
614
615 out:
616 return ret;
617 }
618
619 static inline long
620 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
621 {
622 compat_uptr_t uptr;
623 int i;
624
625 for (i = 0; i < nr; ++i) {
626 if (get_user(uptr, ptr32 + i))
627 return -EFAULT;
628 if (put_user(compat_ptr(uptr), ptr64 + i))
629 return -EFAULT;
630 }
631 return 0;
632 }
633
634 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
635
636 asmlinkage long
637 compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
638 {
639 struct iocb __user * __user *iocb64;
640 long ret;
641
642 if (unlikely(nr < 0))
643 return -EINVAL;
644
645 if (nr > MAX_AIO_SUBMITS)
646 nr = MAX_AIO_SUBMITS;
647
648 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
649 ret = copy_iocb(nr, iocb, iocb64);
650 if (!ret)
651 ret = do_io_submit(ctx_id, nr, iocb64, 1);
652 return ret;
653 }
654
655 struct compat_ncp_mount_data {
656 compat_int_t version;
657 compat_uint_t ncp_fd;
658 __compat_uid_t mounted_uid;
659 compat_pid_t wdog_pid;
660 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
661 compat_uint_t time_out;
662 compat_uint_t retry_count;
663 compat_uint_t flags;
664 __compat_uid_t uid;
665 __compat_gid_t gid;
666 compat_mode_t file_mode;
667 compat_mode_t dir_mode;
668 };
669
670 struct compat_ncp_mount_data_v4 {
671 compat_int_t version;
672 compat_ulong_t flags;
673 compat_ulong_t mounted_uid;
674 compat_long_t wdog_pid;
675 compat_uint_t ncp_fd;
676 compat_uint_t time_out;
677 compat_uint_t retry_count;
678 compat_ulong_t uid;
679 compat_ulong_t gid;
680 compat_ulong_t file_mode;
681 compat_ulong_t dir_mode;
682 };
683
684 static void *do_ncp_super_data_conv(void *raw_data)
685 {
686 int version = *(unsigned int *)raw_data;
687
688 if (version == 3) {
689 struct compat_ncp_mount_data *c_n = raw_data;
690 struct ncp_mount_data *n = raw_data;
691
692 n->dir_mode = c_n->dir_mode;
693 n->file_mode = c_n->file_mode;
694 n->gid = c_n->gid;
695 n->uid = c_n->uid;
696 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
697 n->wdog_pid = c_n->wdog_pid;
698 n->mounted_uid = c_n->mounted_uid;
699 } else if (version == 4) {
700 struct compat_ncp_mount_data_v4 *c_n = raw_data;
701 struct ncp_mount_data_v4 *n = raw_data;
702
703 n->dir_mode = c_n->dir_mode;
704 n->file_mode = c_n->file_mode;
705 n->gid = c_n->gid;
706 n->uid = c_n->uid;
707 n->retry_count = c_n->retry_count;
708 n->time_out = c_n->time_out;
709 n->ncp_fd = c_n->ncp_fd;
710 n->wdog_pid = c_n->wdog_pid;
711 n->mounted_uid = c_n->mounted_uid;
712 n->flags = c_n->flags;
713 } else if (version != 5) {
714 return NULL;
715 }
716
717 return raw_data;
718 }
719
720
721 struct compat_nfs_string {
722 compat_uint_t len;
723 compat_uptr_t data;
724 };
725
726 static inline void compat_nfs_string(struct nfs_string *dst,
727 struct compat_nfs_string *src)
728 {
729 dst->data = compat_ptr(src->data);
730 dst->len = src->len;
731 }
732
733 struct compat_nfs4_mount_data_v1 {
734 compat_int_t version;
735 compat_int_t flags;
736 compat_int_t rsize;
737 compat_int_t wsize;
738 compat_int_t timeo;
739 compat_int_t retrans;
740 compat_int_t acregmin;
741 compat_int_t acregmax;
742 compat_int_t acdirmin;
743 compat_int_t acdirmax;
744 struct compat_nfs_string client_addr;
745 struct compat_nfs_string mnt_path;
746 struct compat_nfs_string hostname;
747 compat_uint_t host_addrlen;
748 compat_uptr_t host_addr;
749 compat_int_t proto;
750 compat_int_t auth_flavourlen;
751 compat_uptr_t auth_flavours;
752 };
753
754 static int do_nfs4_super_data_conv(void *raw_data)
755 {
756 int version = *(compat_uint_t *) raw_data;
757
758 if (version == 1) {
759 struct compat_nfs4_mount_data_v1 *raw = raw_data;
760 struct nfs4_mount_data *real = raw_data;
761
762 /* copy the fields backwards */
763 real->auth_flavours = compat_ptr(raw->auth_flavours);
764 real->auth_flavourlen = raw->auth_flavourlen;
765 real->proto = raw->proto;
766 real->host_addr = compat_ptr(raw->host_addr);
767 real->host_addrlen = raw->host_addrlen;
768 compat_nfs_string(&real->hostname, &raw->hostname);
769 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
770 compat_nfs_string(&real->client_addr, &raw->client_addr);
771 real->acdirmax = raw->acdirmax;
772 real->acdirmin = raw->acdirmin;
773 real->acregmax = raw->acregmax;
774 real->acregmin = raw->acregmin;
775 real->retrans = raw->retrans;
776 real->timeo = raw->timeo;
777 real->wsize = raw->wsize;
778 real->rsize = raw->rsize;
779 real->flags = raw->flags;
780 real->version = raw->version;
781 }
782
783 return 0;
784 }
785
786 #define NCPFS_NAME "ncpfs"
787 #define NFS4_NAME "nfs4"
788
789 asmlinkage long compat_sys_mount(const char __user * dev_name,
790 const char __user * dir_name,
791 const char __user * type, unsigned long flags,
792 const void __user * data)
793 {
794 char *kernel_type;
795 unsigned long data_page;
796 char *kernel_dev;
797 char *dir_page;
798 int retval;
799
800 retval = copy_mount_string(type, &kernel_type);
801 if (retval < 0)
802 goto out;
803
804 dir_page = getname(dir_name);
805 retval = PTR_ERR(dir_page);
806 if (IS_ERR(dir_page))
807 goto out1;
808
809 retval = copy_mount_string(dev_name, &kernel_dev);
810 if (retval < 0)
811 goto out2;
812
813 retval = copy_mount_options(data, &data_page);
814 if (retval < 0)
815 goto out3;
816
817 retval = -EINVAL;
818
819 if (kernel_type && data_page) {
820 if (!strcmp(kernel_type, NCPFS_NAME)) {
821 do_ncp_super_data_conv((void *)data_page);
822 } else if (!strcmp(kernel_type, NFS4_NAME)) {
823 if (do_nfs4_super_data_conv((void *) data_page))
824 goto out4;
825 }
826 }
827
828 retval = do_mount(kernel_dev, dir_page, kernel_type,
829 flags, (void*)data_page);
830
831 out4:
832 free_page(data_page);
833 out3:
834 kfree(kernel_dev);
835 out2:
836 putname(dir_page);
837 out1:
838 kfree(kernel_type);
839 out:
840 return retval;
841 }
842
843 struct compat_old_linux_dirent {
844 compat_ulong_t d_ino;
845 compat_ulong_t d_offset;
846 unsigned short d_namlen;
847 char d_name[1];
848 };
849
850 struct compat_readdir_callback {
851 struct compat_old_linux_dirent __user *dirent;
852 int result;
853 };
854
855 static int compat_fillonedir(void *__buf, const char *name, int namlen,
856 loff_t offset, u64 ino, unsigned int d_type)
857 {
858 struct compat_readdir_callback *buf = __buf;
859 struct compat_old_linux_dirent __user *dirent;
860 compat_ulong_t d_ino;
861
862 if (buf->result)
863 return -EINVAL;
864 d_ino = ino;
865 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
866 buf->result = -EOVERFLOW;
867 return -EOVERFLOW;
868 }
869 buf->result++;
870 dirent = buf->dirent;
871 if (!access_ok(VERIFY_WRITE, dirent,
872 (unsigned long)(dirent->d_name + namlen + 1) -
873 (unsigned long)dirent))
874 goto efault;
875 if ( __put_user(d_ino, &dirent->d_ino) ||
876 __put_user(offset, &dirent->d_offset) ||
877 __put_user(namlen, &dirent->d_namlen) ||
878 __copy_to_user(dirent->d_name, name, namlen) ||
879 __put_user(0, dirent->d_name + namlen))
880 goto efault;
881 return 0;
882 efault:
883 buf->result = -EFAULT;
884 return -EFAULT;
885 }
886
887 asmlinkage long compat_sys_old_readdir(unsigned int fd,
888 struct compat_old_linux_dirent __user *dirent, unsigned int count)
889 {
890 int error;
891 struct file *file;
892 struct compat_readdir_callback buf;
893
894 error = -EBADF;
895 file = fget(fd);
896 if (!file)
897 goto out;
898
899 buf.result = 0;
900 buf.dirent = dirent;
901
902 error = vfs_readdir(file, compat_fillonedir, &buf);
903 if (buf.result)
904 error = buf.result;
905
906 fput(file);
907 out:
908 return error;
909 }
910
911 struct compat_linux_dirent {
912 compat_ulong_t d_ino;
913 compat_ulong_t d_off;
914 unsigned short d_reclen;
915 char d_name[1];
916 };
917
918 struct compat_getdents_callback {
919 struct compat_linux_dirent __user *current_dir;
920 struct compat_linux_dirent __user *previous;
921 int count;
922 int error;
923 };
924
925 static int compat_filldir(void *__buf, const char *name, int namlen,
926 loff_t offset, u64 ino, unsigned int d_type)
927 {
928 struct compat_linux_dirent __user * dirent;
929 struct compat_getdents_callback *buf = __buf;
930 compat_ulong_t d_ino;
931 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
932 namlen + 2, sizeof(compat_long_t));
933
934 buf->error = -EINVAL; /* only used if we fail.. */
935 if (reclen > buf->count)
936 return -EINVAL;
937 d_ino = ino;
938 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
939 buf->error = -EOVERFLOW;
940 return -EOVERFLOW;
941 }
942 dirent = buf->previous;
943 if (dirent) {
944 if (__put_user(offset, &dirent->d_off))
945 goto efault;
946 }
947 dirent = buf->current_dir;
948 if (__put_user(d_ino, &dirent->d_ino))
949 goto efault;
950 if (__put_user(reclen, &dirent->d_reclen))
951 goto efault;
952 if (copy_to_user(dirent->d_name, name, namlen))
953 goto efault;
954 if (__put_user(0, dirent->d_name + namlen))
955 goto efault;
956 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
957 goto efault;
958 buf->previous = dirent;
959 dirent = (void __user *)dirent + reclen;
960 buf->current_dir = dirent;
961 buf->count -= reclen;
962 return 0;
963 efault:
964 buf->error = -EFAULT;
965 return -EFAULT;
966 }
967
968 asmlinkage long compat_sys_getdents(unsigned int fd,
969 struct compat_linux_dirent __user *dirent, unsigned int count)
970 {
971 struct file * file;
972 struct compat_linux_dirent __user * lastdirent;
973 struct compat_getdents_callback buf;
974 int error;
975
976 error = -EFAULT;
977 if (!access_ok(VERIFY_WRITE, dirent, count))
978 goto out;
979
980 error = -EBADF;
981 file = fget(fd);
982 if (!file)
983 goto out;
984
985 buf.current_dir = dirent;
986 buf.previous = NULL;
987 buf.count = count;
988 buf.error = 0;
989
990 error = vfs_readdir(file, compat_filldir, &buf);
991 if (error >= 0)
992 error = buf.error;
993 lastdirent = buf.previous;
994 if (lastdirent) {
995 if (put_user(file->f_pos, &lastdirent->d_off))
996 error = -EFAULT;
997 else
998 error = count - buf.count;
999 }
1000 fput(file);
1001 out:
1002 return error;
1003 }
1004
1005 #ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
1006
1007 struct compat_getdents_callback64 {
1008 struct linux_dirent64 __user *current_dir;
1009 struct linux_dirent64 __user *previous;
1010 int count;
1011 int error;
1012 };
1013
1014 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
1015 u64 ino, unsigned int d_type)
1016 {
1017 struct linux_dirent64 __user *dirent;
1018 struct compat_getdents_callback64 *buf = __buf;
1019 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
1020 sizeof(u64));
1021 u64 off;
1022
1023 buf->error = -EINVAL; /* only used if we fail.. */
1024 if (reclen > buf->count)
1025 return -EINVAL;
1026 dirent = buf->previous;
1027
1028 if (dirent) {
1029 if (__put_user_unaligned(offset, &dirent->d_off))
1030 goto efault;
1031 }
1032 dirent = buf->current_dir;
1033 if (__put_user_unaligned(ino, &dirent->d_ino))
1034 goto efault;
1035 off = 0;
1036 if (__put_user_unaligned(off, &dirent->d_off))
1037 goto efault;
1038 if (__put_user(reclen, &dirent->d_reclen))
1039 goto efault;
1040 if (__put_user(d_type, &dirent->d_type))
1041 goto efault;
1042 if (copy_to_user(dirent->d_name, name, namlen))
1043 goto efault;
1044 if (__put_user(0, dirent->d_name + namlen))
1045 goto efault;
1046 buf->previous = dirent;
1047 dirent = (void __user *)dirent + reclen;
1048 buf->current_dir = dirent;
1049 buf->count -= reclen;
1050 return 0;
1051 efault:
1052 buf->error = -EFAULT;
1053 return -EFAULT;
1054 }
1055
1056 asmlinkage long compat_sys_getdents64(unsigned int fd,
1057 struct linux_dirent64 __user * dirent, unsigned int count)
1058 {
1059 struct file * file;
1060 struct linux_dirent64 __user * lastdirent;
1061 struct compat_getdents_callback64 buf;
1062 int error;
1063
1064 error = -EFAULT;
1065 if (!access_ok(VERIFY_WRITE, dirent, count))
1066 goto out;
1067
1068 error = -EBADF;
1069 file = fget(fd);
1070 if (!file)
1071 goto out;
1072
1073 buf.current_dir = dirent;
1074 buf.previous = NULL;
1075 buf.count = count;
1076 buf.error = 0;
1077
1078 error = vfs_readdir(file, compat_filldir64, &buf);
1079 if (error >= 0)
1080 error = buf.error;
1081 lastdirent = buf.previous;
1082 if (lastdirent) {
1083 typeof(lastdirent->d_off) d_off = file->f_pos;
1084 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1085 error = -EFAULT;
1086 else
1087 error = count - buf.count;
1088 }
1089 fput(file);
1090 out:
1091 return error;
1092 }
1093 #endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
1094
1095 static ssize_t compat_do_readv_writev(int type, struct file *file,
1096 const struct compat_iovec __user *uvector,
1097 unsigned long nr_segs, loff_t *pos)
1098 {
1099 compat_ssize_t tot_len;
1100 struct iovec iovstack[UIO_FASTIOV];
1101 struct iovec *iov = iovstack;
1102 ssize_t ret;
1103 io_fn_t fn;
1104 iov_fn_t fnv;
1105
1106 ret = -EINVAL;
1107 if (!file->f_op)
1108 goto out;
1109
1110 ret = compat_rw_copy_check_uvector(type, uvector, nr_segs,
1111 UIO_FASTIOV, iovstack, &iov, 1);
1112 if (ret <= 0)
1113 goto out;
1114
1115 tot_len = ret;
1116 ret = rw_verify_area(type, file, pos, tot_len);
1117 if (ret < 0)
1118 goto out;
1119
1120 fnv = NULL;
1121 if (type == READ) {
1122 fn = file->f_op->read;
1123 fnv = file->f_op->aio_read;
1124 } else {
1125 fn = (io_fn_t)file->f_op->write;
1126 fnv = file->f_op->aio_write;
1127 }
1128
1129 if (fnv)
1130 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
1131 pos, fnv);
1132 else
1133 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
1134
1135 out:
1136 if (iov != iovstack)
1137 kfree(iov);
1138 if ((ret + (type == READ)) > 0) {
1139 if (type == READ)
1140 fsnotify_access(file);
1141 else
1142 fsnotify_modify(file);
1143 }
1144 return ret;
1145 }
1146
1147 static size_t compat_readv(struct file *file,
1148 const struct compat_iovec __user *vec,
1149 unsigned long vlen, loff_t *pos)
1150 {
1151 ssize_t ret = -EBADF;
1152
1153 if (!(file->f_mode & FMODE_READ))
1154 goto out;
1155
1156 ret = -EINVAL;
1157 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
1158 goto out;
1159
1160 ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
1161
1162 out:
1163 if (ret > 0)
1164 add_rchar(current, ret);
1165 inc_syscr(current);
1166 return ret;
1167 }
1168
1169 asmlinkage ssize_t
1170 compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec,
1171 unsigned long vlen)
1172 {
1173 struct file *file;
1174 int fput_needed;
1175 ssize_t ret;
1176 loff_t pos;
1177
1178 file = fget_light(fd, &fput_needed);
1179 if (!file)
1180 return -EBADF;
1181 pos = file->f_pos;
1182 ret = compat_readv(file, vec, vlen, &pos);
1183 file->f_pos = pos;
1184 fput_light(file, fput_needed);
1185 return ret;
1186 }
1187
1188 asmlinkage ssize_t
1189 compat_sys_preadv(unsigned long fd, const struct compat_iovec __user *vec,
1190 unsigned long vlen, u32 pos_low, u32 pos_high)
1191 {
1192 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1193 struct file *file;
1194 int fput_needed;
1195 ssize_t ret;
1196
1197 if (pos < 0)
1198 return -EINVAL;
1199 file = fget_light(fd, &fput_needed);
1200 if (!file)
1201 return -EBADF;
1202 ret = -ESPIPE;
1203 if (file->f_mode & FMODE_PREAD)
1204 ret = compat_readv(file, vec, vlen, &pos);
1205 fput_light(file, fput_needed);
1206 return ret;
1207 }
1208
1209 static size_t compat_writev(struct file *file,
1210 const struct compat_iovec __user *vec,
1211 unsigned long vlen, loff_t *pos)
1212 {
1213 ssize_t ret = -EBADF;
1214
1215 if (!(file->f_mode & FMODE_WRITE))
1216 goto out;
1217
1218 ret = -EINVAL;
1219 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
1220 goto out;
1221
1222 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
1223
1224 out:
1225 if (ret > 0)
1226 add_wchar(current, ret);
1227 inc_syscw(current);
1228 return ret;
1229 }
1230
1231 asmlinkage ssize_t
1232 compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec,
1233 unsigned long vlen)
1234 {
1235 struct file *file;
1236 int fput_needed;
1237 ssize_t ret;
1238 loff_t pos;
1239
1240 file = fget_light(fd, &fput_needed);
1241 if (!file)
1242 return -EBADF;
1243 pos = file->f_pos;
1244 ret = compat_writev(file, vec, vlen, &pos);
1245 file->f_pos = pos;
1246 fput_light(file, fput_needed);
1247 return ret;
1248 }
1249
1250 asmlinkage ssize_t
1251 compat_sys_pwritev(unsigned long fd, const struct compat_iovec __user *vec,
1252 unsigned long vlen, u32 pos_low, u32 pos_high)
1253 {
1254 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1255 struct file *file;
1256 int fput_needed;
1257 ssize_t ret;
1258
1259 if (pos < 0)
1260 return -EINVAL;
1261 file = fget_light(fd, &fput_needed);
1262 if (!file)
1263 return -EBADF;
1264 ret = -ESPIPE;
1265 if (file->f_mode & FMODE_PWRITE)
1266 ret = compat_writev(file, vec, vlen, &pos);
1267 fput_light(file, fput_needed);
1268 return ret;
1269 }
1270
1271 asmlinkage long
1272 compat_sys_vmsplice(int fd, const struct compat_iovec __user *iov32,
1273 unsigned int nr_segs, unsigned int flags)
1274 {
1275 unsigned i;
1276 struct iovec __user *iov;
1277 if (nr_segs > UIO_MAXIOV)
1278 return -EINVAL;
1279 iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
1280 for (i = 0; i < nr_segs; i++) {
1281 struct compat_iovec v;
1282 if (get_user(v.iov_base, &iov32[i].iov_base) ||
1283 get_user(v.iov_len, &iov32[i].iov_len) ||
1284 put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
1285 put_user(v.iov_len, &iov[i].iov_len))
1286 return -EFAULT;
1287 }
1288 return sys_vmsplice(fd, iov, nr_segs, flags);
1289 }
1290
1291 /*
1292 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1293 * O_LARGEFILE flag.
1294 */
1295 asmlinkage long
1296 compat_sys_open(const char __user *filename, int flags, int mode)
1297 {
1298 return do_sys_open(AT_FDCWD, filename, flags, mode);
1299 }
1300
1301 /*
1302 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1303 * O_LARGEFILE flag.
1304 */
1305 asmlinkage long
1306 compat_sys_openat(unsigned int dfd, const char __user *filename, int flags, int mode)
1307 {
1308 return do_sys_open(dfd, filename, flags, mode);
1309 }
1310
1311 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1312
1313 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1314 int timeval, int ret)
1315 {
1316 struct timespec ts;
1317
1318 if (!p)
1319 return ret;
1320
1321 if (current->personality & STICKY_TIMEOUTS)
1322 goto sticky;
1323
1324 /* No update for zero timeout */
1325 if (!end_time->tv_sec && !end_time->tv_nsec)
1326 return ret;
1327
1328 ktime_get_ts(&ts);
1329 ts = timespec_sub(*end_time, ts);
1330 if (ts.tv_sec < 0)
1331 ts.tv_sec = ts.tv_nsec = 0;
1332
1333 if (timeval) {
1334 struct compat_timeval rtv;
1335
1336 rtv.tv_sec = ts.tv_sec;
1337 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1338
1339 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1340 return ret;
1341 } else {
1342 struct compat_timespec rts;
1343
1344 rts.tv_sec = ts.tv_sec;
1345 rts.tv_nsec = ts.tv_nsec;
1346
1347 if (!copy_to_user(p, &rts, sizeof(rts)))
1348 return ret;
1349 }
1350 /*
1351 * If an application puts its timeval in read-only memory, we
1352 * don't want the Linux-specific update to the timeval to
1353 * cause a fault after the select has completed
1354 * successfully. However, because we're not updating the
1355 * timeval, we can't restart the system call.
1356 */
1357
1358 sticky:
1359 if (ret == -ERESTARTNOHAND)
1360 ret = -EINTR;
1361 return ret;
1362 }
1363
1364 /*
1365 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1366 * 64-bit unsigned longs.
1367 */
1368 static
1369 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1370 unsigned long *fdset)
1371 {
1372 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1373 if (ufdset) {
1374 unsigned long odd;
1375
1376 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1377 return -EFAULT;
1378
1379 odd = nr & 1UL;
1380 nr &= ~1UL;
1381 while (nr) {
1382 unsigned long h, l;
1383 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1384 return -EFAULT;
1385 ufdset += 2;
1386 *fdset++ = h << 32 | l;
1387 nr -= 2;
1388 }
1389 if (odd && __get_user(*fdset, ufdset))
1390 return -EFAULT;
1391 } else {
1392 /* Tricky, must clear full unsigned long in the
1393 * kernel fdset at the end, this makes sure that
1394 * actually happens.
1395 */
1396 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1397 }
1398 return 0;
1399 }
1400
1401 static
1402 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1403 unsigned long *fdset)
1404 {
1405 unsigned long odd;
1406 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1407
1408 if (!ufdset)
1409 return 0;
1410
1411 odd = nr & 1UL;
1412 nr &= ~1UL;
1413 while (nr) {
1414 unsigned long h, l;
1415 l = *fdset++;
1416 h = l >> 32;
1417 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1418 return -EFAULT;
1419 ufdset += 2;
1420 nr -= 2;
1421 }
1422 if (odd && __put_user(*fdset, ufdset))
1423 return -EFAULT;
1424 return 0;
1425 }
1426
1427
1428 /*
1429 * This is a virtual copy of sys_select from fs/select.c and probably
1430 * should be compared to it from time to time
1431 */
1432
1433 /*
1434 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1435 * like to be certain this leads to no problems. So I return
1436 * EINTR just for safety.
1437 *
1438 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1439 * I'm trying ERESTARTNOHAND which restart only when you want to.
1440 */
1441 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1442 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1443 struct timespec *end_time)
1444 {
1445 fd_set_bits fds;
1446 void *bits;
1447 int size, max_fds, ret = -EINVAL;
1448 struct fdtable *fdt;
1449 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1450
1451 if (n < 0)
1452 goto out_nofds;
1453
1454 /* max_fds can increase, so grab it once to avoid race */
1455 rcu_read_lock();
1456 fdt = files_fdtable(current->files);
1457 max_fds = fdt->max_fds;
1458 rcu_read_unlock();
1459 if (n > max_fds)
1460 n = max_fds;
1461
1462 /*
1463 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1464 * since we used fdset we need to allocate memory in units of
1465 * long-words.
1466 */
1467 size = FDS_BYTES(n);
1468 bits = stack_fds;
1469 if (size > sizeof(stack_fds) / 6) {
1470 bits = kmalloc(6 * size, GFP_KERNEL);
1471 ret = -ENOMEM;
1472 if (!bits)
1473 goto out_nofds;
1474 }
1475 fds.in = (unsigned long *) bits;
1476 fds.out = (unsigned long *) (bits + size);
1477 fds.ex = (unsigned long *) (bits + 2*size);
1478 fds.res_in = (unsigned long *) (bits + 3*size);
1479 fds.res_out = (unsigned long *) (bits + 4*size);
1480 fds.res_ex = (unsigned long *) (bits + 5*size);
1481
1482 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1483 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1484 (ret = compat_get_fd_set(n, exp, fds.ex)))
1485 goto out;
1486 zero_fd_set(n, fds.res_in);
1487 zero_fd_set(n, fds.res_out);
1488 zero_fd_set(n, fds.res_ex);
1489
1490 ret = do_select(n, &fds, end_time);
1491
1492 if (ret < 0)
1493 goto out;
1494 if (!ret) {
1495 ret = -ERESTARTNOHAND;
1496 if (signal_pending(current))
1497 goto out;
1498 ret = 0;
1499 }
1500
1501 if (compat_set_fd_set(n, inp, fds.res_in) ||
1502 compat_set_fd_set(n, outp, fds.res_out) ||
1503 compat_set_fd_set(n, exp, fds.res_ex))
1504 ret = -EFAULT;
1505 out:
1506 if (bits != stack_fds)
1507 kfree(bits);
1508 out_nofds:
1509 return ret;
1510 }
1511
1512 asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
1513 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1514 struct compat_timeval __user *tvp)
1515 {
1516 struct timespec end_time, *to = NULL;
1517 struct compat_timeval tv;
1518 int ret;
1519
1520 if (tvp) {
1521 if (copy_from_user(&tv, tvp, sizeof(tv)))
1522 return -EFAULT;
1523
1524 to = &end_time;
1525 if (poll_select_set_timeout(to,
1526 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1527 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1528 return -EINVAL;
1529 }
1530
1531 ret = compat_core_sys_select(n, inp, outp, exp, to);
1532 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1533
1534 return ret;
1535 }
1536
1537 struct compat_sel_arg_struct {
1538 compat_ulong_t n;
1539 compat_uptr_t inp;
1540 compat_uptr_t outp;
1541 compat_uptr_t exp;
1542 compat_uptr_t tvp;
1543 };
1544
1545 asmlinkage long compat_sys_old_select(struct compat_sel_arg_struct __user *arg)
1546 {
1547 struct compat_sel_arg_struct a;
1548
1549 if (copy_from_user(&a, arg, sizeof(a)))
1550 return -EFAULT;
1551 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1552 compat_ptr(a.exp), compat_ptr(a.tvp));
1553 }
1554
1555 #ifdef HAVE_SET_RESTORE_SIGMASK
1556 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1557 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1558 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1559 compat_size_t sigsetsize)
1560 {
1561 compat_sigset_t ss32;
1562 sigset_t ksigmask, sigsaved;
1563 struct compat_timespec ts;
1564 struct timespec end_time, *to = NULL;
1565 int ret;
1566
1567 if (tsp) {
1568 if (copy_from_user(&ts, tsp, sizeof(ts)))
1569 return -EFAULT;
1570
1571 to = &end_time;
1572 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1573 return -EINVAL;
1574 }
1575
1576 if (sigmask) {
1577 if (sigsetsize != sizeof(compat_sigset_t))
1578 return -EINVAL;
1579 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1580 return -EFAULT;
1581 sigset_from_compat(&ksigmask, &ss32);
1582
1583 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1584 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1585 }
1586
1587 ret = compat_core_sys_select(n, inp, outp, exp, to);
1588 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1589
1590 if (ret == -ERESTARTNOHAND) {
1591 /*
1592 * Don't restore the signal mask yet. Let do_signal() deliver
1593 * the signal on the way back to userspace, before the signal
1594 * mask is restored.
1595 */
1596 if (sigmask) {
1597 memcpy(&current->saved_sigmask, &sigsaved,
1598 sizeof(sigsaved));
1599 set_restore_sigmask();
1600 }
1601 } else if (sigmask)
1602 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1603
1604 return ret;
1605 }
1606
1607 asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
1608 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1609 struct compat_timespec __user *tsp, void __user *sig)
1610 {
1611 compat_size_t sigsetsize = 0;
1612 compat_uptr_t up = 0;
1613
1614 if (sig) {
1615 if (!access_ok(VERIFY_READ, sig,
1616 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1617 __get_user(up, (compat_uptr_t __user *)sig) ||
1618 __get_user(sigsetsize,
1619 (compat_size_t __user *)(sig+sizeof(up))))
1620 return -EFAULT;
1621 }
1622 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1623 sigsetsize);
1624 }
1625
1626 asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1627 unsigned int nfds, struct compat_timespec __user *tsp,
1628 const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
1629 {
1630 compat_sigset_t ss32;
1631 sigset_t ksigmask, sigsaved;
1632 struct compat_timespec ts;
1633 struct timespec end_time, *to = NULL;
1634 int ret;
1635
1636 if (tsp) {
1637 if (copy_from_user(&ts, tsp, sizeof(ts)))
1638 return -EFAULT;
1639
1640 to = &end_time;
1641 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1642 return -EINVAL;
1643 }
1644
1645 if (sigmask) {
1646 if (sigsetsize != sizeof(compat_sigset_t))
1647 return -EINVAL;
1648 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1649 return -EFAULT;
1650 sigset_from_compat(&ksigmask, &ss32);
1651
1652 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1653 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1654 }
1655
1656 ret = do_sys_poll(ufds, nfds, to);
1657
1658 /* We can restart this syscall, usually */
1659 if (ret == -EINTR) {
1660 /*
1661 * Don't restore the signal mask yet. Let do_signal() deliver
1662 * the signal on the way back to userspace, before the signal
1663 * mask is restored.
1664 */
1665 if (sigmask) {
1666 memcpy(&current->saved_sigmask, &sigsaved,
1667 sizeof(sigsaved));
1668 set_restore_sigmask();
1669 }
1670 ret = -ERESTARTNOHAND;
1671 } else if (sigmask)
1672 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1673
1674 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1675
1676 return ret;
1677 }
1678 #endif /* HAVE_SET_RESTORE_SIGMASK */
1679
1680 #ifdef CONFIG_EPOLL
1681
1682 #ifdef HAVE_SET_RESTORE_SIGMASK
1683 asmlinkage long compat_sys_epoll_pwait(int epfd,
1684 struct compat_epoll_event __user *events,
1685 int maxevents, int timeout,
1686 const compat_sigset_t __user *sigmask,
1687 compat_size_t sigsetsize)
1688 {
1689 long err;
1690 compat_sigset_t csigmask;
1691 sigset_t ksigmask, sigsaved;
1692
1693 /*
1694 * If the caller wants a certain signal mask to be set during the wait,
1695 * we apply it here.
1696 */
1697 if (sigmask) {
1698 if (sigsetsize != sizeof(compat_sigset_t))
1699 return -EINVAL;
1700 if (copy_from_user(&csigmask, sigmask, sizeof(csigmask)))
1701 return -EFAULT;
1702 sigset_from_compat(&ksigmask, &csigmask);
1703 sigdelsetmask(&ksigmask, sigmask(SIGKILL) | sigmask(SIGSTOP));
1704 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1705 }
1706
1707 err = sys_epoll_wait(epfd, events, maxevents, timeout);
1708
1709 /*
1710 * If we changed the signal mask, we need to restore the original one.
1711 * In case we've got a signal while waiting, we do not restore the
1712 * signal mask yet, and we allow do_signal() to deliver the signal on
1713 * the way back to userspace, before the signal mask is restored.
1714 */
1715 if (sigmask) {
1716 if (err == -EINTR) {
1717 memcpy(&current->saved_sigmask, &sigsaved,
1718 sizeof(sigsaved));
1719 set_restore_sigmask();
1720 } else
1721 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1722 }
1723
1724 return err;
1725 }
1726 #endif /* HAVE_SET_RESTORE_SIGMASK */
1727
1728 #endif /* CONFIG_EPOLL */
1729
1730 #ifdef CONFIG_SIGNALFD
1731
1732 asmlinkage long compat_sys_signalfd4(int ufd,
1733 const compat_sigset_t __user *sigmask,
1734 compat_size_t sigsetsize, int flags)
1735 {
1736 compat_sigset_t ss32;
1737 sigset_t tmp;
1738 sigset_t __user *ksigmask;
1739
1740 if (sigsetsize != sizeof(compat_sigset_t))
1741 return -EINVAL;
1742 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1743 return -EFAULT;
1744 sigset_from_compat(&tmp, &ss32);
1745 ksigmask = compat_alloc_user_space(sizeof(sigset_t));
1746 if (copy_to_user(ksigmask, &tmp, sizeof(sigset_t)))
1747 return -EFAULT;
1748
1749 return sys_signalfd4(ufd, ksigmask, sizeof(sigset_t), flags);
1750 }
1751
1752 asmlinkage long compat_sys_signalfd(int ufd,
1753 const compat_sigset_t __user *sigmask,
1754 compat_size_t sigsetsize)
1755 {
1756 return compat_sys_signalfd4(ufd, sigmask, sigsetsize, 0);
1757 }
1758 #endif /* CONFIG_SIGNALFD */
1759
1760 #ifdef CONFIG_TIMERFD
1761
1762 asmlinkage long compat_sys_timerfd_settime(int ufd, int flags,
1763 const struct compat_itimerspec __user *utmr,
1764 struct compat_itimerspec __user *otmr)
1765 {
1766 int error;
1767 struct itimerspec t;
1768 struct itimerspec __user *ut;
1769
1770 if (get_compat_itimerspec(&t, utmr))
1771 return -EFAULT;
1772 ut = compat_alloc_user_space(2 * sizeof(struct itimerspec));
1773 if (copy_to_user(&ut[0], &t, sizeof(t)))
1774 return -EFAULT;
1775 error = sys_timerfd_settime(ufd, flags, &ut[0], &ut[1]);
1776 if (!error && otmr)
1777 error = (copy_from_user(&t, &ut[1], sizeof(struct itimerspec)) ||
1778 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1779
1780 return error;
1781 }
1782
1783 asmlinkage long compat_sys_timerfd_gettime(int ufd,
1784 struct compat_itimerspec __user *otmr)
1785 {
1786 int error;
1787 struct itimerspec t;
1788 struct itimerspec __user *ut;
1789
1790 ut = compat_alloc_user_space(sizeof(struct itimerspec));
1791 error = sys_timerfd_gettime(ufd, ut);
1792 if (!error)
1793 error = (copy_from_user(&t, ut, sizeof(struct itimerspec)) ||
1794 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1795
1796 return error;
1797 }
1798
1799 #endif /* CONFIG_TIMERFD */
1800
1801 #ifdef CONFIG_FHANDLE
1802 /*
1803 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1804 * doesn't set the O_LARGEFILE flag.
1805 */
1806 asmlinkage long
1807 compat_sys_open_by_handle_at(int mountdirfd,
1808 struct file_handle __user *handle, int flags)
1809 {
1810 return do_handle_open(mountdirfd, handle, flags);
1811 }
1812 #endif
Linux with added FPC-III support