~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/compat.c

Version: ~ [ linux-5.11 ] ~ [ linux-5.10.17 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.99 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.176 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.221 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.257 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.257 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/dirent.h>
 37 #include <linux/fsnotify.h>
 38 #include <linux/highuid.h>
 39 #include <linux/personality.h>
 40 #include <linux/rwsem.h>
 41 #include <linux/tsacct_kern.h>
 42 #include <linux/security.h>
 43 #include <linux/highmem.h>
 44 #include <linux/signal.h>
 45 #include <linux/poll.h>
 46 #include <linux/mm.h>
 47 #include <linux/fs_struct.h>
 48 #include <linux/slab.h>
 49 #include <linux/pagemap.h>
 50 #include <linux/aio.h>
 51 
 52 #include <asm/uaccess.h>
 53 #include <asm/mmu_context.h>
 54 #include <asm/ioctls.h>
 55 #include "internal.h"
 56 
 57 int compat_log = 1;
 58 
 59 int compat_printk(const char *fmt, ...)
 60 {
 61         va_list ap;
 62         int ret;
 63         if (!compat_log)
 64                 return 0;
 65         va_start(ap, fmt);
 66         ret = vprintk(fmt, ap);
 67         va_end(ap);
 68         return ret;
 69 }
 70 
 71 /*
 72  * Not all architectures have sys_utime, so implement this in terms
 73  * of sys_utimes.
 74  */
 75 COMPAT_SYSCALL_DEFINE2(utime, const char __user *, filename,
 76                        struct compat_utimbuf __user *, t)
 77 {
 78         struct timespec tv[2];
 79 
 80         if (t) {
 81                 if (get_user(tv[0].tv_sec, &t->actime) ||
 82                     get_user(tv[1].tv_sec, &t->modtime))
 83                         return -EFAULT;
 84                 tv[0].tv_nsec = 0;
 85                 tv[1].tv_nsec = 0;
 86         }
 87         return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
 88 }
 89 
 90 COMPAT_SYSCALL_DEFINE4(utimensat, unsigned int, dfd, const char __user *, filename, struct compat_timespec __user *, t, int, flags)
 91 {
 92         struct timespec tv[2];
 93 
 94         if  (t) {
 95                 if (compat_get_timespec(&tv[0], &t[0]) ||
 96                     compat_get_timespec(&tv[1], &t[1]))
 97                         return -EFAULT;
 98 
 99                 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
100                         return 0;
101         }
102         return do_utimes(dfd, filename, t ? tv : NULL, flags);
103 }
104 
105 COMPAT_SYSCALL_DEFINE3(futimesat, unsigned int, dfd, const char __user *, filename, struct compat_timeval __user *, t)
106 {
107         struct timespec tv[2];
108 
109         if (t) {
110                 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
111                     get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
112                     get_user(tv[1].tv_sec, &t[1].tv_sec) ||
113                     get_user(tv[1].tv_nsec, &t[1].tv_usec))
114                         return -EFAULT;
115                 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
116                     tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
117                         return -EINVAL;
118                 tv[0].tv_nsec *= 1000;
119                 tv[1].tv_nsec *= 1000;
120         }
121         return do_utimes(dfd, filename, t ? tv : NULL, 0);
122 }
123 
124 COMPAT_SYSCALL_DEFINE2(utimes, const char __user *, filename, struct compat_timeval __user *, t)
125 {
126         return compat_sys_futimesat(AT_FDCWD, filename, t);
127 }
128 
129 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
130 {
131         struct compat_stat tmp;
132 
133         if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
134                 return -EOVERFLOW;
135 
136         memset(&tmp, 0, sizeof(tmp));
137         tmp.st_dev = old_encode_dev(stat->dev);
138         tmp.st_ino = stat->ino;
139         if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
140                 return -EOVERFLOW;
141         tmp.st_mode = stat->mode;
142         tmp.st_nlink = stat->nlink;
143         if (tmp.st_nlink != stat->nlink)
144                 return -EOVERFLOW;
145         SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
146         SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
147         tmp.st_rdev = old_encode_dev(stat->rdev);
148         if ((u64) stat->size > MAX_NON_LFS)
149                 return -EOVERFLOW;
150         tmp.st_size = stat->size;
151         tmp.st_atime = stat->atime.tv_sec;
152         tmp.st_atime_nsec = stat->atime.tv_nsec;
153         tmp.st_mtime = stat->mtime.tv_sec;
154         tmp.st_mtime_nsec = stat->mtime.tv_nsec;
155         tmp.st_ctime = stat->ctime.tv_sec;
156         tmp.st_ctime_nsec = stat->ctime.tv_nsec;
157         tmp.st_blocks = stat->blocks;
158         tmp.st_blksize = stat->blksize;
159         return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
160 }
161 
162 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
163                        struct compat_stat __user *, statbuf)
164 {
165         struct kstat stat;
166         int error;
167 
168         error = vfs_stat(filename, &stat);
169         if (error)
170                 return error;
171         return cp_compat_stat(&stat, statbuf);
172 }
173 
174 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
175                        struct compat_stat __user *, statbuf)
176 {
177         struct kstat stat;
178         int error;
179 
180         error = vfs_lstat(filename, &stat);
181         if (error)
182                 return error;
183         return cp_compat_stat(&stat, statbuf);
184 }
185 
186 #ifndef __ARCH_WANT_STAT64
187 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
188                        const char __user *, filename,
189                        struct compat_stat __user *, statbuf, int, flag)
190 {
191         struct kstat stat;
192         int error;
193 
194         error = vfs_fstatat(dfd, filename, &stat, flag);
195         if (error)
196                 return error;
197         return cp_compat_stat(&stat, statbuf);
198 }
199 #endif
200 
201 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
202                        struct compat_stat __user *, statbuf)
203 {
204         struct kstat stat;
205         int error = vfs_fstat(fd, &stat);
206 
207         if (!error)
208                 error = cp_compat_stat(&stat, statbuf);
209         return error;
210 }
211 
212 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
213 {
214         
215         if (sizeof ubuf->f_blocks == 4) {
216                 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
217                      kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
218                         return -EOVERFLOW;
219                 /* f_files and f_ffree may be -1; it's okay
220                  * to stuff that into 32 bits */
221                 if (kbuf->f_files != 0xffffffffffffffffULL
222                  && (kbuf->f_files & 0xffffffff00000000ULL))
223                         return -EOVERFLOW;
224                 if (kbuf->f_ffree != 0xffffffffffffffffULL
225                  && (kbuf->f_ffree & 0xffffffff00000000ULL))
226                         return -EOVERFLOW;
227         }
228         if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
229             __put_user(kbuf->f_type, &ubuf->f_type) ||
230             __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
231             __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
232             __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
233             __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
234             __put_user(kbuf->f_files, &ubuf->f_files) ||
235             __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
236             __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
237             __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
238             __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
239             __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
240             __put_user(kbuf->f_flags, &ubuf->f_flags) ||
241             __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
242                 return -EFAULT;
243         return 0;
244 }
245 
246 /*
247  * The following statfs calls are copies of code from fs/statfs.c and
248  * should be checked against those from time to time
249  */
250 COMPAT_SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct compat_statfs __user *, buf)
251 {
252         struct kstatfs tmp;
253         int error = user_statfs(pathname, &tmp);
254         if (!error)
255                 error = put_compat_statfs(buf, &tmp);
256         return error;
257 }
258 
259 COMPAT_SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct compat_statfs __user *, buf)
260 {
261         struct kstatfs tmp;
262         int error = fd_statfs(fd, &tmp);
263         if (!error)
264                 error = put_compat_statfs(buf, &tmp);
265         return error;
266 }
267 
268 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
269 {
270         if (sizeof ubuf->f_blocks == 4) {
271                 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
272                      kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
273                         return -EOVERFLOW;
274                 /* f_files and f_ffree may be -1; it's okay
275                  * to stuff that into 32 bits */
276                 if (kbuf->f_files != 0xffffffffffffffffULL
277                  && (kbuf->f_files & 0xffffffff00000000ULL))
278                         return -EOVERFLOW;
279                 if (kbuf->f_ffree != 0xffffffffffffffffULL
280                  && (kbuf->f_ffree & 0xffffffff00000000ULL))
281                         return -EOVERFLOW;
282         }
283         if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
284             __put_user(kbuf->f_type, &ubuf->f_type) ||
285             __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
286             __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
287             __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
288             __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
289             __put_user(kbuf->f_files, &ubuf->f_files) ||
290             __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
291             __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
292             __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
293             __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
294             __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
295             __put_user(kbuf->f_flags, &ubuf->f_flags) ||
296             __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
297                 return -EFAULT;
298         return 0;
299 }
300 
301 COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, sz, struct compat_statfs64 __user *, buf)
302 {
303         struct kstatfs tmp;
304         int error;
305 
306         if (sz != sizeof(*buf))
307                 return -EINVAL;
308 
309         error = user_statfs(pathname, &tmp);
310         if (!error)
311                 error = put_compat_statfs64(buf, &tmp);
312         return error;
313 }
314 
315 COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct compat_statfs64 __user *, buf)
316 {
317         struct kstatfs tmp;
318         int error;
319 
320         if (sz != sizeof(*buf))
321                 return -EINVAL;
322 
323         error = fd_statfs(fd, &tmp);
324         if (!error)
325                 error = put_compat_statfs64(buf, &tmp);
326         return error;
327 }
328 
329 /*
330  * This is a copy of sys_ustat, just dealing with a structure layout.
331  * Given how simple this syscall is that apporach is more maintainable
332  * than the various conversion hacks.
333  */
334 COMPAT_SYSCALL_DEFINE2(ustat, unsigned, dev, struct compat_ustat __user *, u)
335 {
336         struct compat_ustat tmp;
337         struct kstatfs sbuf;
338         int err = vfs_ustat(new_decode_dev(dev), &sbuf);
339         if (err)
340                 return err;
341 
342         memset(&tmp, 0, sizeof(struct compat_ustat));
343         tmp.f_tfree = sbuf.f_bfree;
344         tmp.f_tinode = sbuf.f_ffree;
345         if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
346                 return -EFAULT;
347         return 0;
348 }
349 
350 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
351 {
352         if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
353             __get_user(kfl->l_type, &ufl->l_type) ||
354             __get_user(kfl->l_whence, &ufl->l_whence) ||
355             __get_user(kfl->l_start, &ufl->l_start) ||
356             __get_user(kfl->l_len, &ufl->l_len) ||
357             __get_user(kfl->l_pid, &ufl->l_pid))
358                 return -EFAULT;
359         return 0;
360 }
361 
362 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
363 {
364         if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
365             __put_user(kfl->l_type, &ufl->l_type) ||
366             __put_user(kfl->l_whence, &ufl->l_whence) ||
367             __put_user(kfl->l_start, &ufl->l_start) ||
368             __put_user(kfl->l_len, &ufl->l_len) ||
369             __put_user(kfl->l_pid, &ufl->l_pid))
370                 return -EFAULT;
371         return 0;
372 }
373 
374 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
375 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
376 {
377         if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
378             __get_user(kfl->l_type, &ufl->l_type) ||
379             __get_user(kfl->l_whence, &ufl->l_whence) ||
380             __get_user(kfl->l_start, &ufl->l_start) ||
381             __get_user(kfl->l_len, &ufl->l_len) ||
382             __get_user(kfl->l_pid, &ufl->l_pid))
383                 return -EFAULT;
384         return 0;
385 }
386 #endif
387 
388 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
389 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
390 {
391         if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
392             __put_user(kfl->l_type, &ufl->l_type) ||
393             __put_user(kfl->l_whence, &ufl->l_whence) ||
394             __put_user(kfl->l_start, &ufl->l_start) ||
395             __put_user(kfl->l_len, &ufl->l_len) ||
396             __put_user(kfl->l_pid, &ufl->l_pid))
397                 return -EFAULT;
398         return 0;
399 }
400 #endif
401 
402 static unsigned int
403 convert_fcntl_cmd(unsigned int cmd)
404 {
405         switch (cmd) {
406         case F_GETLK64:
407                 return F_GETLK;
408         case F_SETLK64:
409                 return F_SETLK;
410         case F_SETLKW64:
411                 return F_SETLKW;
412         }
413 
414         return cmd;
415 }
416 
417 COMPAT_SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
418                        compat_ulong_t, arg)
419 {
420         mm_segment_t old_fs;
421         struct flock f;
422         long ret;
423         unsigned int conv_cmd;
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         case F_OFD_GETLK:
461         case F_OFD_SETLK:
462         case F_OFD_SETLKW:
463                 ret = get_compat_flock64(&f, compat_ptr(arg));
464                 if (ret != 0)
465                         break;
466                 old_fs = get_fs();
467                 set_fs(KERNEL_DS);
468                 conv_cmd = convert_fcntl_cmd(cmd);
469                 ret = sys_fcntl(fd, conv_cmd, (unsigned long)&f);
470                 set_fs(old_fs);
471                 if ((conv_cmd == F_GETLK || conv_cmd == F_OFD_GETLK) && ret == 0) {
472                         /* need to return lock information - see above for commentary */
473                         if (f.l_start > COMPAT_LOFF_T_MAX)
474                                 ret = -EOVERFLOW;
475                         if (f.l_len > COMPAT_LOFF_T_MAX)
476                                 f.l_len = COMPAT_LOFF_T_MAX;
477                         if (ret == 0)
478                                 ret = put_compat_flock64(&f, compat_ptr(arg));
479                 }
480                 break;
481 
482         default:
483                 ret = sys_fcntl(fd, cmd, arg);
484                 break;
485         }
486         return ret;
487 }
488 
489 COMPAT_SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd,
490                        compat_ulong_t, arg)
491 {
492         switch (cmd) {
493         case F_GETLK64:
494         case F_SETLK64:
495         case F_SETLKW64:
496         case F_OFD_GETLK:
497         case F_OFD_SETLK:
498         case F_OFD_SETLKW:
499                 return -EINVAL;
500         }
501         return compat_sys_fcntl64(fd, cmd, arg);
502 }
503 
504 COMPAT_SYSCALL_DEFINE2(io_setup, unsigned, nr_reqs, u32 __user *, ctx32p)
505 {
506         long ret;
507         aio_context_t ctx64;
508 
509         mm_segment_t oldfs = get_fs();
510         if (unlikely(get_user(ctx64, ctx32p)))
511                 return -EFAULT;
512 
513         set_fs(KERNEL_DS);
514         /* The __user pointer cast is valid because of the set_fs() */
515         ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
516         set_fs(oldfs);
517         /* truncating is ok because it's a user address */
518         if (!ret)
519                 ret = put_user((u32) ctx64, ctx32p);
520         return ret;
521 }
522 
523 COMPAT_SYSCALL_DEFINE5(io_getevents, compat_aio_context_t, ctx_id,
524                        compat_long_t, min_nr,
525                        compat_long_t, nr,
526                        struct io_event __user *, events,
527                        struct compat_timespec __user *, timeout)
528 {
529         struct timespec t;
530         struct timespec __user *ut = NULL;
531 
532         if (timeout) {
533                 if (compat_get_timespec(&t, timeout))
534                         return -EFAULT;
535 
536                 ut = compat_alloc_user_space(sizeof(*ut));
537                 if (copy_to_user(ut, &t, sizeof(t)) )
538                         return -EFAULT;
539         } 
540         return sys_io_getevents(ctx_id, min_nr, nr, events, ut);
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)
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 (type >= 0 &&
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 COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id,
637                        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 COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name,
790                        const char __user *, dir_name,
791                        const char __user *, type, compat_ulong_t, flags,
792                        const void __user *, data)
793 {
794         char *kernel_type;
795         unsigned long data_page;
796         char *kernel_dev;
797         struct filename *dir;
798         int retval;
799 
800         retval = copy_mount_string(type, &kernel_type);
801         if (retval < 0)
802                 goto out;
803 
804         dir = getname(dir_name);
805         retval = PTR_ERR(dir);
806         if (IS_ERR(dir))
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->name, 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);
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 dir_context ctx;
852         struct compat_old_linux_dirent __user *dirent;
853         int result;
854 };
855 
856 static int compat_fillonedir(void *__buf, const char *name, int namlen,
857                         loff_t offset, u64 ino, unsigned int d_type)
858 {
859         struct compat_readdir_callback *buf = __buf;
860         struct compat_old_linux_dirent __user *dirent;
861         compat_ulong_t d_ino;
862 
863         if (buf->result)
864                 return -EINVAL;
865         d_ino = ino;
866         if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
867                 buf->result = -EOVERFLOW;
868                 return -EOVERFLOW;
869         }
870         buf->result++;
871         dirent = buf->dirent;
872         if (!access_ok(VERIFY_WRITE, dirent,
873                         (unsigned long)(dirent->d_name + namlen + 1) -
874                                 (unsigned long)dirent))
875                 goto efault;
876         if (    __put_user(d_ino, &dirent->d_ino) ||
877                 __put_user(offset, &dirent->d_offset) ||
878                 __put_user(namlen, &dirent->d_namlen) ||
879                 __copy_to_user(dirent->d_name, name, namlen) ||
880                 __put_user(0, dirent->d_name + namlen))
881                 goto efault;
882         return 0;
883 efault:
884         buf->result = -EFAULT;
885         return -EFAULT;
886 }
887 
888 COMPAT_SYSCALL_DEFINE3(old_readdir, unsigned int, fd,
889                 struct compat_old_linux_dirent __user *, dirent, unsigned int, count)
890 {
891         int error;
892         struct fd f = fdget(fd);
893         struct compat_readdir_callback buf = {
894                 .ctx.actor = compat_fillonedir,
895                 .dirent = dirent
896         };
897 
898         if (!f.file)
899                 return -EBADF;
900 
901         error = iterate_dir(f.file, &buf.ctx);
902         if (buf.result)
903                 error = buf.result;
904 
905         fdput(f);
906         return error;
907 }
908 
909 struct compat_linux_dirent {
910         compat_ulong_t  d_ino;
911         compat_ulong_t  d_off;
912         unsigned short  d_reclen;
913         char            d_name[1];
914 };
915 
916 struct compat_getdents_callback {
917         struct dir_context ctx;
918         struct compat_linux_dirent __user *current_dir;
919         struct compat_linux_dirent __user *previous;
920         int count;
921         int error;
922 };
923 
924 static int compat_filldir(void *__buf, const char *name, int namlen,
925                 loff_t offset, u64 ino, unsigned int d_type)
926 {
927         struct compat_linux_dirent __user * dirent;
928         struct compat_getdents_callback *buf = __buf;
929         compat_ulong_t d_ino;
930         int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
931                 namlen + 2, sizeof(compat_long_t));
932 
933         buf->error = -EINVAL;   /* only used if we fail.. */
934         if (reclen > buf->count)
935                 return -EINVAL;
936         d_ino = ino;
937         if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
938                 buf->error = -EOVERFLOW;
939                 return -EOVERFLOW;
940         }
941         dirent = buf->previous;
942         if (dirent) {
943                 if (__put_user(offset, &dirent->d_off))
944                         goto efault;
945         }
946         dirent = buf->current_dir;
947         if (__put_user(d_ino, &dirent->d_ino))
948                 goto efault;
949         if (__put_user(reclen, &dirent->d_reclen))
950                 goto efault;
951         if (copy_to_user(dirent->d_name, name, namlen))
952                 goto efault;
953         if (__put_user(0, dirent->d_name + namlen))
954                 goto efault;
955         if (__put_user(d_type, (char  __user *) dirent + reclen - 1))
956                 goto efault;
957         buf->previous = dirent;
958         dirent = (void __user *)dirent + reclen;
959         buf->current_dir = dirent;
960         buf->count -= reclen;
961         return 0;
962 efault:
963         buf->error = -EFAULT;
964         return -EFAULT;
965 }
966 
967 COMPAT_SYSCALL_DEFINE3(getdents, unsigned int, fd,
968                 struct compat_linux_dirent __user *, dirent, unsigned int, count)
969 {
970         struct fd f;
971         struct compat_linux_dirent __user * lastdirent;
972         struct compat_getdents_callback buf = {
973                 .ctx.actor = compat_filldir,
974                 .current_dir = dirent,
975                 .count = count
976         };
977         int error;
978 
979         if (!access_ok(VERIFY_WRITE, dirent, count))
980                 return -EFAULT;
981 
982         f = fdget(fd);
983         if (!f.file)
984                 return -EBADF;
985 
986         error = iterate_dir(f.file, &buf.ctx);
987         if (error >= 0)
988                 error = buf.error;
989         lastdirent = buf.previous;
990         if (lastdirent) {
991                 if (put_user(buf.ctx.pos, &lastdirent->d_off))
992                         error = -EFAULT;
993                 else
994                         error = count - buf.count;
995         }
996         fdput(f);
997         return error;
998 }
999 
1000 #ifdef __ARCH_WANT_COMPAT_SYS_GETDENTS64
1001 
1002 struct compat_getdents_callback64 {
1003         struct dir_context ctx;
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 COMPAT_SYSCALL_DEFINE3(getdents64, unsigned int, fd,
1053                 struct linux_dirent64 __user *, dirent, unsigned int, count)
1054 {
1055         struct fd f;
1056         struct linux_dirent64 __user * lastdirent;
1057         struct compat_getdents_callback64 buf = {
1058                 .ctx.actor = compat_filldir64,
1059                 .current_dir = dirent,
1060                 .count = count
1061         };
1062         int error;
1063 
1064         if (!access_ok(VERIFY_WRITE, dirent, count))
1065                 return -EFAULT;
1066 
1067         f = fdget(fd);
1068         if (!f.file)
1069                 return -EBADF;
1070 
1071         error = iterate_dir(f.file, &buf.ctx);
1072         if (error >= 0)
1073                 error = buf.error;
1074         lastdirent = buf.previous;
1075         if (lastdirent) {
1076                 typeof(lastdirent->d_off) d_off = buf.ctx.pos;
1077                 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1078                         error = -EFAULT;
1079                 else
1080                         error = count - buf.count;
1081         }
1082         fdput(f);
1083         return error;
1084 }
1085 #endif /* __ARCH_WANT_COMPAT_SYS_GETDENTS64 */
1086 
1087 /*
1088  * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1089  * O_LARGEFILE flag.
1090  */
1091 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1092 {
1093         return do_sys_open(AT_FDCWD, filename, flags, mode);
1094 }
1095 
1096 /*
1097  * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1098  * O_LARGEFILE flag.
1099  */
1100 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1101 {
1102         return do_sys_open(dfd, filename, flags, mode);
1103 }
1104 
1105 #define __COMPAT_NFDBITS       (8 * sizeof(compat_ulong_t))
1106 
1107 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1108                                       int timeval, int ret)
1109 {
1110         struct timespec ts;
1111 
1112         if (!p)
1113                 return ret;
1114 
1115         if (current->personality & STICKY_TIMEOUTS)
1116                 goto sticky;
1117 
1118         /* No update for zero timeout */
1119         if (!end_time->tv_sec && !end_time->tv_nsec)
1120                 return ret;
1121 
1122         ktime_get_ts(&ts);
1123         ts = timespec_sub(*end_time, ts);
1124         if (ts.tv_sec < 0)
1125                 ts.tv_sec = ts.tv_nsec = 0;
1126 
1127         if (timeval) {
1128                 struct compat_timeval rtv;
1129 
1130                 rtv.tv_sec = ts.tv_sec;
1131                 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1132 
1133                 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1134                         return ret;
1135         } else {
1136                 struct compat_timespec rts;
1137 
1138                 rts.tv_sec = ts.tv_sec;
1139                 rts.tv_nsec = ts.tv_nsec;
1140 
1141                 if (!copy_to_user(p, &rts, sizeof(rts)))
1142                         return ret;
1143         }
1144         /*
1145          * If an application puts its timeval in read-only memory, we
1146          * don't want the Linux-specific update to the timeval to
1147          * cause a fault after the select has completed
1148          * successfully. However, because we're not updating the
1149          * timeval, we can't restart the system call.
1150          */
1151 
1152 sticky:
1153         if (ret == -ERESTARTNOHAND)
1154                 ret = -EINTR;
1155         return ret;
1156 }
1157 
1158 /*
1159  * Ooo, nasty.  We need here to frob 32-bit unsigned longs to
1160  * 64-bit unsigned longs.
1161  */
1162 static
1163 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1164                         unsigned long *fdset)
1165 {
1166         nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1167         if (ufdset) {
1168                 unsigned long odd;
1169 
1170                 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1171                         return -EFAULT;
1172 
1173                 odd = nr & 1UL;
1174                 nr &= ~1UL;
1175                 while (nr) {
1176                         unsigned long h, l;
1177                         if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1178                                 return -EFAULT;
1179                         ufdset += 2;
1180                         *fdset++ = h << 32 | l;
1181                         nr -= 2;
1182                 }
1183                 if (odd && __get_user(*fdset, ufdset))
1184                         return -EFAULT;
1185         } else {
1186                 /* Tricky, must clear full unsigned long in the
1187                  * kernel fdset at the end, this makes sure that
1188                  * actually happens.
1189                  */
1190                 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1191         }
1192         return 0;
1193 }
1194 
1195 static
1196 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1197                       unsigned long *fdset)
1198 {
1199         unsigned long odd;
1200         nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1201 
1202         if (!ufdset)
1203                 return 0;
1204 
1205         odd = nr & 1UL;
1206         nr &= ~1UL;
1207         while (nr) {
1208                 unsigned long h, l;
1209                 l = *fdset++;
1210                 h = l >> 32;
1211                 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1212                         return -EFAULT;
1213                 ufdset += 2;
1214                 nr -= 2;
1215         }
1216         if (odd && __put_user(*fdset, ufdset))
1217                 return -EFAULT;
1218         return 0;
1219 }
1220 
1221 
1222 /*
1223  * This is a virtual copy of sys_select from fs/select.c and probably
1224  * should be compared to it from time to time
1225  */
1226 
1227 /*
1228  * We can actually return ERESTARTSYS instead of EINTR, but I'd
1229  * like to be certain this leads to no problems. So I return
1230  * EINTR just for safety.
1231  *
1232  * Update: ERESTARTSYS breaks at least the xview clock binary, so
1233  * I'm trying ERESTARTNOHAND which restart only when you want to.
1234  */
1235 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1236         compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1237         struct timespec *end_time)
1238 {
1239         fd_set_bits fds;
1240         void *bits;
1241         int size, max_fds, ret = -EINVAL;
1242         struct fdtable *fdt;
1243         long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1244 
1245         if (n < 0)
1246                 goto out_nofds;
1247 
1248         /* max_fds can increase, so grab it once to avoid race */
1249         rcu_read_lock();
1250         fdt = files_fdtable(current->files);
1251         max_fds = fdt->max_fds;
1252         rcu_read_unlock();
1253         if (n > max_fds)
1254                 n = max_fds;
1255 
1256         /*
1257          * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1258          * since we used fdset we need to allocate memory in units of
1259          * long-words.
1260          */
1261         size = FDS_BYTES(n);
1262         bits = stack_fds;
1263         if (size > sizeof(stack_fds) / 6) {
1264                 bits = kmalloc(6 * size, GFP_KERNEL);
1265                 ret = -ENOMEM;
1266                 if (!bits)
1267                         goto out_nofds;
1268         }
1269         fds.in      = (unsigned long *)  bits;
1270         fds.out     = (unsigned long *) (bits +   size);
1271         fds.ex      = (unsigned long *) (bits + 2*size);
1272         fds.res_in  = (unsigned long *) (bits + 3*size);
1273         fds.res_out = (unsigned long *) (bits + 4*size);
1274         fds.res_ex  = (unsigned long *) (bits + 5*size);
1275 
1276         if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1277             (ret = compat_get_fd_set(n, outp, fds.out)) ||
1278             (ret = compat_get_fd_set(n, exp, fds.ex)))
1279                 goto out;
1280         zero_fd_set(n, fds.res_in);
1281         zero_fd_set(n, fds.res_out);
1282         zero_fd_set(n, fds.res_ex);
1283 
1284         ret = do_select(n, &fds, end_time);
1285 
1286         if (ret < 0)
1287                 goto out;
1288         if (!ret) {
1289                 ret = -ERESTARTNOHAND;
1290                 if (signal_pending(current))
1291                         goto out;
1292                 ret = 0;
1293         }
1294 
1295         if (compat_set_fd_set(n, inp, fds.res_in) ||
1296             compat_set_fd_set(n, outp, fds.res_out) ||
1297             compat_set_fd_set(n, exp, fds.res_ex))
1298                 ret = -EFAULT;
1299 out:
1300         if (bits != stack_fds)
1301                 kfree(bits);
1302 out_nofds:
1303         return ret;
1304 }
1305 
1306 COMPAT_SYSCALL_DEFINE5(select, int, n, compat_ulong_t __user *, inp,
1307         compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1308         struct compat_timeval __user *, tvp)
1309 {
1310         struct timespec end_time, *to = NULL;
1311         struct compat_timeval tv;
1312         int ret;
1313 
1314         if (tvp) {
1315                 if (copy_from_user(&tv, tvp, sizeof(tv)))
1316                         return -EFAULT;
1317 
1318                 to = &end_time;
1319                 if (poll_select_set_timeout(to,
1320                                 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1321                                 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1322                         return -EINVAL;
1323         }
1324 
1325         ret = compat_core_sys_select(n, inp, outp, exp, to);
1326         ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1327 
1328         return ret;
1329 }
1330 
1331 struct compat_sel_arg_struct {
1332         compat_ulong_t n;
1333         compat_uptr_t inp;
1334         compat_uptr_t outp;
1335         compat_uptr_t exp;
1336         compat_uptr_t tvp;
1337 };
1338 
1339 COMPAT_SYSCALL_DEFINE1(old_select, struct compat_sel_arg_struct __user *, arg)
1340 {
1341         struct compat_sel_arg_struct a;
1342 
1343         if (copy_from_user(&a, arg, sizeof(a)))
1344                 return -EFAULT;
1345         return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1346                                  compat_ptr(a.exp), compat_ptr(a.tvp));
1347 }
1348 
1349 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1350         compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1351         struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1352         compat_size_t sigsetsize)
1353 {
1354         compat_sigset_t ss32;
1355         sigset_t ksigmask, sigsaved;
1356         struct compat_timespec ts;
1357         struct timespec end_time, *to = NULL;
1358         int ret;
1359 
1360         if (tsp) {
1361                 if (copy_from_user(&ts, tsp, sizeof(ts)))
1362                         return -EFAULT;
1363 
1364                 to = &end_time;
1365                 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1366                         return -EINVAL;
1367         }
1368 
1369         if (sigmask) {
1370                 if (sigsetsize != sizeof(compat_sigset_t))
1371                         return -EINVAL;
1372                 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1373                         return -EFAULT;
1374                 sigset_from_compat(&ksigmask, &ss32);
1375 
1376                 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1377                 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1378         }
1379 
1380         ret = compat_core_sys_select(n, inp, outp, exp, to);
1381         ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1382 
1383         if (ret == -ERESTARTNOHAND) {
1384                 /*
1385                  * Don't restore the signal mask yet. Let do_signal() deliver
1386                  * the signal on the way back to userspace, before the signal
1387                  * mask is restored.
1388                  */
1389                 if (sigmask) {
1390                         memcpy(&current->saved_sigmask, &sigsaved,
1391                                         sizeof(sigsaved));
1392                         set_restore_sigmask();
1393                 }
1394         } else if (sigmask)
1395                 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1396 
1397         return ret;
1398 }
1399 
1400 COMPAT_SYSCALL_DEFINE6(pselect6, int, n, compat_ulong_t __user *, inp,
1401         compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1402         struct compat_timespec __user *, tsp, void __user *, sig)
1403 {
1404         compat_size_t sigsetsize = 0;
1405         compat_uptr_t up = 0;
1406 
1407         if (sig) {
1408                 if (!access_ok(VERIFY_READ, sig,
1409                                 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1410                         __get_user(up, (compat_uptr_t __user *)sig) ||
1411                         __get_user(sigsetsize,
1412                                 (compat_size_t __user *)(sig+sizeof(up))))
1413                         return -EFAULT;
1414         }
1415         return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1416                                  sigsetsize);
1417 }
1418 
1419 COMPAT_SYSCALL_DEFINE5(ppoll, struct pollfd __user *, ufds,
1420         unsigned int,  nfds, struct compat_timespec __user *, tsp,
1421         const compat_sigset_t __user *, sigmask, compat_size_t, sigsetsize)
1422 {
1423         compat_sigset_t ss32;
1424         sigset_t ksigmask, sigsaved;
1425         struct compat_timespec ts;
1426         struct timespec end_time, *to = NULL;
1427         int ret;
1428 
1429         if (tsp) {
1430                 if (copy_from_user(&ts, tsp, sizeof(ts)))
1431                         return -EFAULT;
1432 
1433                 to = &end_time;
1434                 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1435                         return -EINVAL;
1436         }
1437 
1438         if (sigmask) {
1439                 if (sigsetsize != sizeof(compat_sigset_t))
1440                         return -EINVAL;
1441                 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1442                         return -EFAULT;
1443                 sigset_from_compat(&ksigmask, &ss32);
1444 
1445                 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1446                 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1447         }
1448 
1449         ret = do_sys_poll(ufds, nfds, to);
1450 
1451         /* We can restart this syscall, usually */
1452         if (ret == -EINTR) {
1453                 /*
1454                  * Don't restore the signal mask yet. Let do_signal() deliver
1455                  * the signal on the way back to userspace, before the signal
1456                  * mask is restored.
1457                  */
1458                 if (sigmask) {
1459                         memcpy(&current->saved_sigmask, &sigsaved,
1460                                 sizeof(sigsaved));
1461                         set_restore_sigmask();
1462                 }
1463                 ret = -ERESTARTNOHAND;
1464         } else if (sigmask)
1465                 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1466 
1467         ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1468 
1469         return ret;
1470 }
1471 
1472 #ifdef CONFIG_FHANDLE
1473 /*
1474  * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1475  * doesn't set the O_LARGEFILE flag.
1476  */
1477 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1478                              struct file_handle __user *, handle, int, flags)
1479 {
1480         return do_handle_open(mountdirfd, handle, flags);
1481 }
1482 #endif
1483 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp