4 * Kernel compatibililty routines for e.g. 32 bit syscall support
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)
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.
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>
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>
47 #include <linux/fs_struct.h>
48 #include <linux/slab.h>
49 #include <linux/pagemap.h>
50 #include <linux/aio.h>
52 #include <asm/uaccess.h>
53 #include <asm/mmu_context.h>
54 #include <asm/ioctls.h>
58 * Not all architectures have sys_utime, so implement this in terms
61 COMPAT_SYSCALL_DEFINE2(utime, const char __user *, filename,
62 struct compat_utimbuf __user *, t)
64 struct timespec tv[2];
67 if (get_user(tv[0].tv_sec, &t->actime) ||
68 get_user(tv[1].tv_sec, &t->modtime))
73 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
76 COMPAT_SYSCALL_DEFINE4(utimensat, unsigned int, dfd, const char __user *, filename, struct compat_timespec __user *, t, int, flags)
78 struct timespec tv[2];
81 if (compat_get_timespec(&tv[0], &t[0]) ||
82 compat_get_timespec(&tv[1], &t[1]))
85 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
88 return do_utimes(dfd, filename, t ? tv : NULL, flags);
91 COMPAT_SYSCALL_DEFINE3(futimesat, unsigned int, dfd, const char __user *, filename, struct compat_timeval __user *, t)
93 struct timespec tv[2];
96 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
97 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
98 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
99 get_user(tv[1].tv_nsec, &t[1].tv_usec))
101 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
102 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
104 tv[0].tv_nsec *= 1000;
105 tv[1].tv_nsec *= 1000;
107 return do_utimes(dfd, filename, t ? tv : NULL, 0);
110 COMPAT_SYSCALL_DEFINE2(utimes, const char __user *, filename, struct compat_timeval __user *, t)
112 return compat_sys_futimesat(AT_FDCWD, filename, t);
115 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
117 struct compat_stat tmp;
119 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
122 memset(&tmp, 0, sizeof(tmp));
123 tmp.st_dev = old_encode_dev(stat->dev);
124 tmp.st_ino = stat->ino;
125 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
127 tmp.st_mode = stat->mode;
128 tmp.st_nlink = stat->nlink;
129 if (tmp.st_nlink != stat->nlink)
131 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
132 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
133 tmp.st_rdev = old_encode_dev(stat->rdev);
134 if ((u64) stat->size > MAX_NON_LFS)
136 tmp.st_size = stat->size;
137 tmp.st_atime = stat->atime.tv_sec;
138 tmp.st_atime_nsec = stat->atime.tv_nsec;
139 tmp.st_mtime = stat->mtime.tv_sec;
140 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
141 tmp.st_ctime = stat->ctime.tv_sec;
142 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
143 tmp.st_blocks = stat->blocks;
144 tmp.st_blksize = stat->blksize;
145 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
148 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
149 struct compat_stat __user *, statbuf)
154 error = vfs_stat(filename, &stat);
157 return cp_compat_stat(&stat, statbuf);
160 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
161 struct compat_stat __user *, statbuf)
166 error = vfs_lstat(filename, &stat);
169 return cp_compat_stat(&stat, statbuf);
172 #ifndef __ARCH_WANT_STAT64
173 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
174 const char __user *, filename,
175 struct compat_stat __user *, statbuf, int, flag)
180 error = vfs_fstatat(dfd, filename, &stat, flag);
183 return cp_compat_stat(&stat, statbuf);
187 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
188 struct compat_stat __user *, statbuf)
191 int error = vfs_fstat(fd, &stat);
194 error = cp_compat_stat(&stat, statbuf);
198 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
201 if (sizeof ubuf->f_blocks == 4) {
202 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
203 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
205 /* f_files and f_ffree may be -1; it's okay
206 * to stuff that into 32 bits */
207 if (kbuf->f_files != 0xffffffffffffffffULL
208 && (kbuf->f_files & 0xffffffff00000000ULL))
210 if (kbuf->f_ffree != 0xffffffffffffffffULL
211 && (kbuf->f_ffree & 0xffffffff00000000ULL))
214 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
215 __put_user(kbuf->f_type, &ubuf->f_type) ||
216 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
217 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
218 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
219 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
220 __put_user(kbuf->f_files, &ubuf->f_files) ||
221 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
222 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
223 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
224 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
225 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
226 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
227 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
233 * The following statfs calls are copies of code from fs/statfs.c and
234 * should be checked against those from time to time
236 COMPAT_SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct compat_statfs __user *, buf)
239 int error = user_statfs(pathname, &tmp);
241 error = put_compat_statfs(buf, &tmp);
245 COMPAT_SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct compat_statfs __user *, buf)
248 int error = fd_statfs(fd, &tmp);
250 error = put_compat_statfs(buf, &tmp);
254 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
256 if (sizeof ubuf->f_blocks == 4) {
257 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
258 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
260 /* f_files and f_ffree may be -1; it's okay
261 * to stuff that into 32 bits */
262 if (kbuf->f_files != 0xffffffffffffffffULL
263 && (kbuf->f_files & 0xffffffff00000000ULL))
265 if (kbuf->f_ffree != 0xffffffffffffffffULL
266 && (kbuf->f_ffree & 0xffffffff00000000ULL))
269 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
270 __put_user(kbuf->f_type, &ubuf->f_type) ||
271 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
272 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
273 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
274 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
275 __put_user(kbuf->f_files, &ubuf->f_files) ||
276 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
277 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
278 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
279 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
280 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
281 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
282 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
287 COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, sz, struct compat_statfs64 __user *, buf)
292 if (sz != sizeof(*buf))
295 error = user_statfs(pathname, &tmp);
297 error = put_compat_statfs64(buf, &tmp);
301 COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct compat_statfs64 __user *, buf)
306 if (sz != sizeof(*buf))
309 error = fd_statfs(fd, &tmp);
311 error = put_compat_statfs64(buf, &tmp);
316 * This is a copy of sys_ustat, just dealing with a structure layout.
317 * Given how simple this syscall is that apporach is more maintainable
318 * than the various conversion hacks.
320 COMPAT_SYSCALL_DEFINE2(ustat, unsigned, dev, struct compat_ustat __user *, u)
322 struct compat_ustat tmp;
324 int err = vfs_ustat(new_decode_dev(dev), &sbuf);
328 memset(&tmp, 0, sizeof(struct compat_ustat));
329 tmp.f_tfree = sbuf.f_bfree;
330 tmp.f_tinode = sbuf.f_ffree;
331 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
336 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
338 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
339 __get_user(kfl->l_type, &ufl->l_type) ||
340 __get_user(kfl->l_whence, &ufl->l_whence) ||
341 __get_user(kfl->l_start, &ufl->l_start) ||
342 __get_user(kfl->l_len, &ufl->l_len) ||
343 __get_user(kfl->l_pid, &ufl->l_pid))
348 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
350 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
351 __put_user(kfl->l_type, &ufl->l_type) ||
352 __put_user(kfl->l_whence, &ufl->l_whence) ||
353 __put_user(kfl->l_start, &ufl->l_start) ||
354 __put_user(kfl->l_len, &ufl->l_len) ||
355 __put_user(kfl->l_pid, &ufl->l_pid))
360 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
361 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
363 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
364 __get_user(kfl->l_type, &ufl->l_type) ||
365 __get_user(kfl->l_whence, &ufl->l_whence) ||
366 __get_user(kfl->l_start, &ufl->l_start) ||
367 __get_user(kfl->l_len, &ufl->l_len) ||
368 __get_user(kfl->l_pid, &ufl->l_pid))
374 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
375 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
377 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
378 __put_user(kfl->l_type, &ufl->l_type) ||
379 __put_user(kfl->l_whence, &ufl->l_whence) ||
380 __put_user(kfl->l_start, &ufl->l_start) ||
381 __put_user(kfl->l_len, &ufl->l_len) ||
382 __put_user(kfl->l_pid, &ufl->l_pid))
389 convert_fcntl_cmd(unsigned int cmd)
403 COMPAT_SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
409 unsigned int conv_cmd;
415 ret = get_compat_flock(&f, compat_ptr(arg));
420 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
422 if (cmd == F_GETLK && ret == 0) {
423 /* GETLK was successful and we need to return the data...
424 * but it needs to fit in the compat structure.
425 * l_start shouldn't be too big, unless the original
426 * start + end is greater than COMPAT_OFF_T_MAX, in which
427 * case the app was asking for trouble, so we return
428 * -EOVERFLOW in that case.
429 * l_len could be too big, in which case we just truncate it,
430 * and only allow the app to see that part of the conflicting
431 * lock that might make sense to it anyway
434 if (f.l_start > COMPAT_OFF_T_MAX)
436 if (f.l_len > COMPAT_OFF_T_MAX)
437 f.l_len = COMPAT_OFF_T_MAX;
439 ret = put_compat_flock(&f, compat_ptr(arg));
449 ret = get_compat_flock64(&f, compat_ptr(arg));
454 conv_cmd = convert_fcntl_cmd(cmd);
455 ret = sys_fcntl(fd, conv_cmd, (unsigned long)&f);
457 if ((conv_cmd == F_GETLK || conv_cmd == F_OFD_GETLK) && ret == 0) {
458 /* need to return lock information - see above for commentary */
459 if (f.l_start > COMPAT_LOFF_T_MAX)
461 if (f.l_len > COMPAT_LOFF_T_MAX)
462 f.l_len = COMPAT_LOFF_T_MAX;
464 ret = put_compat_flock64(&f, compat_ptr(arg));
469 ret = sys_fcntl(fd, cmd, arg);
475 COMPAT_SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd,
487 return compat_sys_fcntl64(fd, cmd, arg);
490 COMPAT_SYSCALL_DEFINE2(io_setup, unsigned, nr_reqs, u32 __user *, ctx32p)
495 mm_segment_t oldfs = get_fs();
496 if (unlikely(get_user(ctx64, ctx32p)))
500 /* The __user pointer cast is valid because of the set_fs() */
501 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
503 /* truncating is ok because it's a user address */
505 ret = put_user((u32) ctx64, ctx32p);
509 COMPAT_SYSCALL_DEFINE5(io_getevents, compat_aio_context_t, ctx_id,
510 compat_long_t, min_nr,
512 struct io_event __user *, events,
513 struct compat_timespec __user *, timeout)
516 struct timespec __user *ut = NULL;
519 if (compat_get_timespec(&t, timeout))
522 ut = compat_alloc_user_space(sizeof(*ut));
523 if (copy_to_user(ut, &t, sizeof(t)) )
526 return sys_io_getevents(ctx_id, min_nr, nr, events, ut);
529 /* A write operation does a read from user space and vice versa */
530 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
532 ssize_t compat_rw_copy_check_uvector(int type,
533 const struct compat_iovec __user *uvector, unsigned long nr_segs,
534 unsigned long fast_segs, struct iovec *fast_pointer,
535 struct iovec **ret_pointer)
537 compat_ssize_t tot_len;
538 struct iovec *iov = *ret_pointer = fast_pointer;
543 * SuS says "The readv() function *may* fail if the iovcnt argument
544 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
545 * traditionally returned zero for zero segments, so...
551 if (nr_segs > UIO_MAXIOV)
553 if (nr_segs > fast_segs) {
555 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
562 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
566 * Single unix specification:
567 * We should -EINVAL if an element length is not >= 0 and fitting an
570 * In Linux, the total length is limited to MAX_RW_COUNT, there is
571 * no overflow possibility.
575 for (seg = 0; seg < nr_segs; seg++) {
579 if (__get_user(len, &uvector->iov_len) ||
580 __get_user(buf, &uvector->iov_base)) {
584 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
587 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
591 if (len > MAX_RW_COUNT - tot_len)
592 len = MAX_RW_COUNT - tot_len;
594 iov->iov_base = compat_ptr(buf);
595 iov->iov_len = (compat_size_t) len;
606 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
611 for (i = 0; i < nr; ++i) {
612 if (get_user(uptr, ptr32 + i))
614 if (put_user(compat_ptr(uptr), ptr64 + i))
620 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
622 COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id,
623 int, nr, u32 __user *, iocb)
625 struct iocb __user * __user *iocb64;
628 if (unlikely(nr < 0))
631 if (nr > MAX_AIO_SUBMITS)
632 nr = MAX_AIO_SUBMITS;
634 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
635 ret = copy_iocb(nr, iocb, iocb64);
637 ret = do_io_submit(ctx_id, nr, iocb64, 1);
641 struct compat_ncp_mount_data {
642 compat_int_t version;
643 compat_uint_t ncp_fd;
644 __compat_uid_t mounted_uid;
645 compat_pid_t wdog_pid;
646 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
647 compat_uint_t time_out;
648 compat_uint_t retry_count;
652 compat_mode_t file_mode;
653 compat_mode_t dir_mode;
656 struct compat_ncp_mount_data_v4 {
657 compat_int_t version;
658 compat_ulong_t flags;
659 compat_ulong_t mounted_uid;
660 compat_long_t wdog_pid;
661 compat_uint_t ncp_fd;
662 compat_uint_t time_out;
663 compat_uint_t retry_count;
666 compat_ulong_t file_mode;
667 compat_ulong_t dir_mode;
670 static void *do_ncp_super_data_conv(void *raw_data)
672 int version = *(unsigned int *)raw_data;
675 struct compat_ncp_mount_data *c_n = raw_data;
676 struct ncp_mount_data *n = raw_data;
678 n->dir_mode = c_n->dir_mode;
679 n->file_mode = c_n->file_mode;
682 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
683 n->wdog_pid = c_n->wdog_pid;
684 n->mounted_uid = c_n->mounted_uid;
685 } else if (version == 4) {
686 struct compat_ncp_mount_data_v4 *c_n = raw_data;
687 struct ncp_mount_data_v4 *n = raw_data;
689 n->dir_mode = c_n->dir_mode;
690 n->file_mode = c_n->file_mode;
693 n->retry_count = c_n->retry_count;
694 n->time_out = c_n->time_out;
695 n->ncp_fd = c_n->ncp_fd;
696 n->wdog_pid = c_n->wdog_pid;
697 n->mounted_uid = c_n->mounted_uid;
698 n->flags = c_n->flags;
699 } else if (version != 5) {
707 struct compat_nfs_string {
712 static inline void compat_nfs_string(struct nfs_string *dst,
713 struct compat_nfs_string *src)
715 dst->data = compat_ptr(src->data);
719 struct compat_nfs4_mount_data_v1 {
720 compat_int_t version;
725 compat_int_t retrans;
726 compat_int_t acregmin;
727 compat_int_t acregmax;
728 compat_int_t acdirmin;
729 compat_int_t acdirmax;
730 struct compat_nfs_string client_addr;
731 struct compat_nfs_string mnt_path;
732 struct compat_nfs_string hostname;
733 compat_uint_t host_addrlen;
734 compat_uptr_t host_addr;
736 compat_int_t auth_flavourlen;
737 compat_uptr_t auth_flavours;
740 static int do_nfs4_super_data_conv(void *raw_data)
742 int version = *(compat_uint_t *) raw_data;
745 struct compat_nfs4_mount_data_v1 *raw = raw_data;
746 struct nfs4_mount_data *real = raw_data;
748 /* copy the fields backwards */
749 real->auth_flavours = compat_ptr(raw->auth_flavours);
750 real->auth_flavourlen = raw->auth_flavourlen;
751 real->proto = raw->proto;
752 real->host_addr = compat_ptr(raw->host_addr);
753 real->host_addrlen = raw->host_addrlen;
754 compat_nfs_string(&real->hostname, &raw->hostname);
755 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
756 compat_nfs_string(&real->client_addr, &raw->client_addr);
757 real->acdirmax = raw->acdirmax;
758 real->acdirmin = raw->acdirmin;
759 real->acregmax = raw->acregmax;
760 real->acregmin = raw->acregmin;
761 real->retrans = raw->retrans;
762 real->timeo = raw->timeo;
763 real->wsize = raw->wsize;
764 real->rsize = raw->rsize;
765 real->flags = raw->flags;
766 real->version = raw->version;
772 #define NCPFS_NAME "ncpfs"
773 #define NFS4_NAME "nfs4"
775 COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name,
776 const char __user *, dir_name,
777 const char __user *, type, compat_ulong_t, flags,
778 const void __user *, data)
785 kernel_type = copy_mount_string(type);
786 retval = PTR_ERR(kernel_type);
787 if (IS_ERR(kernel_type))
790 kernel_dev = copy_mount_string(dev_name);
791 retval = PTR_ERR(kernel_dev);
792 if (IS_ERR(kernel_dev))
795 options = copy_mount_options(data);
796 retval = PTR_ERR(options);
800 if (kernel_type && options) {
801 if (!strcmp(kernel_type, NCPFS_NAME)) {
802 do_ncp_super_data_conv(options);
803 } else if (!strcmp(kernel_type, NFS4_NAME)) {
805 if (do_nfs4_super_data_conv(options))
810 retval = do_mount(kernel_dev, dir_name, kernel_type, flags, options);
822 struct compat_old_linux_dirent {
823 compat_ulong_t d_ino;
824 compat_ulong_t d_offset;
825 unsigned short d_namlen;
829 struct compat_readdir_callback {
830 struct dir_context ctx;
831 struct compat_old_linux_dirent __user *dirent;
835 static int compat_fillonedir(struct dir_context *ctx, const char *name,
836 int namlen, loff_t offset, u64 ino,
839 struct compat_readdir_callback *buf =
840 container_of(ctx, struct compat_readdir_callback, ctx);
841 struct compat_old_linux_dirent __user *dirent;
842 compat_ulong_t d_ino;
847 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
848 buf->result = -EOVERFLOW;
852 dirent = buf->dirent;
853 if (!access_ok(VERIFY_WRITE, dirent,
854 (unsigned long)(dirent->d_name + namlen + 1) -
855 (unsigned long)dirent))
857 if ( __put_user(d_ino, &dirent->d_ino) ||
858 __put_user(offset, &dirent->d_offset) ||
859 __put_user(namlen, &dirent->d_namlen) ||
860 __copy_to_user(dirent->d_name, name, namlen) ||
861 __put_user(0, dirent->d_name + namlen))
865 buf->result = -EFAULT;
869 COMPAT_SYSCALL_DEFINE3(old_readdir, unsigned int, fd,
870 struct compat_old_linux_dirent __user *, dirent, unsigned int, count)
873 struct fd f = fdget_pos(fd);
874 struct compat_readdir_callback buf = {
875 .ctx.actor = compat_fillonedir,
882 error = iterate_dir(f.file, &buf.ctx);
890 struct compat_linux_dirent {
891 compat_ulong_t d_ino;
892 compat_ulong_t d_off;
893 unsigned short d_reclen;
897 struct compat_getdents_callback {
898 struct dir_context ctx;
899 struct compat_linux_dirent __user *current_dir;
900 struct compat_linux_dirent __user *previous;
905 static int compat_filldir(struct dir_context *ctx, const char *name, int namlen,
906 loff_t offset, u64 ino, unsigned int d_type)
908 struct compat_linux_dirent __user * dirent;
909 struct compat_getdents_callback *buf =
910 container_of(ctx, struct compat_getdents_callback, ctx);
911 compat_ulong_t d_ino;
912 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
913 namlen + 2, sizeof(compat_long_t));
915 buf->error = -EINVAL; /* only used if we fail.. */
916 if (reclen > buf->count)
919 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
920 buf->error = -EOVERFLOW;
923 dirent = buf->previous;
925 if (signal_pending(current))
927 if (__put_user(offset, &dirent->d_off))
930 dirent = buf->current_dir;
931 if (__put_user(d_ino, &dirent->d_ino))
933 if (__put_user(reclen, &dirent->d_reclen))
935 if (copy_to_user(dirent->d_name, name, namlen))
937 if (__put_user(0, dirent->d_name + namlen))
939 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
941 buf->previous = dirent;
942 dirent = (void __user *)dirent + reclen;
943 buf->current_dir = dirent;
944 buf->count -= reclen;
947 buf->error = -EFAULT;
951 COMPAT_SYSCALL_DEFINE3(getdents, unsigned int, fd,
952 struct compat_linux_dirent __user *, dirent, unsigned int, count)
955 struct compat_linux_dirent __user * lastdirent;
956 struct compat_getdents_callback buf = {
957 .ctx.actor = compat_filldir,
958 .current_dir = dirent,
963 if (!access_ok(VERIFY_WRITE, dirent, count))
970 error = iterate_dir(f.file, &buf.ctx);
973 lastdirent = buf.previous;
975 if (put_user(buf.ctx.pos, &lastdirent->d_off))
978 error = count - buf.count;
984 #ifdef __ARCH_WANT_COMPAT_SYS_GETDENTS64
986 struct compat_getdents_callback64 {
987 struct dir_context ctx;
988 struct linux_dirent64 __user *current_dir;
989 struct linux_dirent64 __user *previous;
994 static int compat_filldir64(struct dir_context *ctx, const char *name,
995 int namlen, loff_t offset, u64 ino,
998 struct linux_dirent64 __user *dirent;
999 struct compat_getdents_callback64 *buf =
1000 container_of(ctx, struct compat_getdents_callback64, ctx);
1001 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
1005 buf->error = -EINVAL; /* only used if we fail.. */
1006 if (reclen > buf->count)
1008 dirent = buf->previous;
1011 if (signal_pending(current))
1013 if (__put_user_unaligned(offset, &dirent->d_off))
1016 dirent = buf->current_dir;
1017 if (__put_user_unaligned(ino, &dirent->d_ino))
1020 if (__put_user_unaligned(off, &dirent->d_off))
1022 if (__put_user(reclen, &dirent->d_reclen))
1024 if (__put_user(d_type, &dirent->d_type))
1026 if (copy_to_user(dirent->d_name, name, namlen))
1028 if (__put_user(0, dirent->d_name + namlen))
1030 buf->previous = dirent;
1031 dirent = (void __user *)dirent + reclen;
1032 buf->current_dir = dirent;
1033 buf->count -= reclen;
1036 buf->error = -EFAULT;
1040 COMPAT_SYSCALL_DEFINE3(getdents64, unsigned int, fd,
1041 struct linux_dirent64 __user *, dirent, unsigned int, count)
1044 struct linux_dirent64 __user * lastdirent;
1045 struct compat_getdents_callback64 buf = {
1046 .ctx.actor = compat_filldir64,
1047 .current_dir = dirent,
1052 if (!access_ok(VERIFY_WRITE, dirent, count))
1059 error = iterate_dir(f.file, &buf.ctx);
1062 lastdirent = buf.previous;
1064 typeof(lastdirent->d_off) d_off = buf.ctx.pos;
1065 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1068 error = count - buf.count;
1073 #endif /* __ARCH_WANT_COMPAT_SYS_GETDENTS64 */
1076 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1079 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1081 return do_sys_open(AT_FDCWD, filename, flags, mode);
1085 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1088 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1090 return do_sys_open(dfd, filename, flags, mode);
1093 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1095 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1096 int timeval, int ret)
1103 if (current->personality & STICKY_TIMEOUTS)
1106 /* No update for zero timeout */
1107 if (!end_time->tv_sec && !end_time->tv_nsec)
1111 ts = timespec_sub(*end_time, ts);
1113 ts.tv_sec = ts.tv_nsec = 0;
1116 struct compat_timeval rtv;
1118 rtv.tv_sec = ts.tv_sec;
1119 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1121 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1124 struct compat_timespec rts;
1126 rts.tv_sec = ts.tv_sec;
1127 rts.tv_nsec = ts.tv_nsec;
1129 if (!copy_to_user(p, &rts, sizeof(rts)))
1133 * If an application puts its timeval in read-only memory, we
1134 * don't want the Linux-specific update to the timeval to
1135 * cause a fault after the select has completed
1136 * successfully. However, because we're not updating the
1137 * timeval, we can't restart the system call.
1141 if (ret == -ERESTARTNOHAND)
1147 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1148 * 64-bit unsigned longs.
1151 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1152 unsigned long *fdset)
1154 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1158 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1165 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1168 *fdset++ = h << 32 | l;
1171 if (odd && __get_user(*fdset, ufdset))
1174 /* Tricky, must clear full unsigned long in the
1175 * kernel fdset at the end, this makes sure that
1178 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1184 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1185 unsigned long *fdset)
1188 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1199 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1204 if (odd && __put_user(*fdset, ufdset))
1211 * This is a virtual copy of sys_select from fs/select.c and probably
1212 * should be compared to it from time to time
1216 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1217 * like to be certain this leads to no problems. So I return
1218 * EINTR just for safety.
1220 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1221 * I'm trying ERESTARTNOHAND which restart only when you want to.
1223 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1224 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1225 struct timespec *end_time)
1229 int size, max_fds, ret = -EINVAL;
1230 struct fdtable *fdt;
1231 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1236 /* max_fds can increase, so grab it once to avoid race */
1238 fdt = files_fdtable(current->files);
1239 max_fds = fdt->max_fds;
1245 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1246 * since we used fdset we need to allocate memory in units of
1249 size = FDS_BYTES(n);
1251 if (size > sizeof(stack_fds) / 6) {
1252 bits = kmalloc(6 * size, GFP_KERNEL);
1257 fds.in = (unsigned long *) bits;
1258 fds.out = (unsigned long *) (bits + size);
1259 fds.ex = (unsigned long *) (bits + 2*size);
1260 fds.res_in = (unsigned long *) (bits + 3*size);
1261 fds.res_out = (unsigned long *) (bits + 4*size);
1262 fds.res_ex = (unsigned long *) (bits + 5*size);
1264 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1265 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1266 (ret = compat_get_fd_set(n, exp, fds.ex)))
1268 zero_fd_set(n, fds.res_in);
1269 zero_fd_set(n, fds.res_out);
1270 zero_fd_set(n, fds.res_ex);
1272 ret = do_select(n, &fds, end_time);
1277 ret = -ERESTARTNOHAND;
1278 if (signal_pending(current))
1283 if (compat_set_fd_set(n, inp, fds.res_in) ||
1284 compat_set_fd_set(n, outp, fds.res_out) ||
1285 compat_set_fd_set(n, exp, fds.res_ex))
1288 if (bits != stack_fds)
1294 COMPAT_SYSCALL_DEFINE5(select, int, n, compat_ulong_t __user *, inp,
1295 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1296 struct compat_timeval __user *, tvp)
1298 struct timespec end_time, *to = NULL;
1299 struct compat_timeval tv;
1303 if (copy_from_user(&tv, tvp, sizeof(tv)))
1307 if (poll_select_set_timeout(to,
1308 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1309 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1313 ret = compat_core_sys_select(n, inp, outp, exp, to);
1314 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1319 struct compat_sel_arg_struct {
1327 COMPAT_SYSCALL_DEFINE1(old_select, struct compat_sel_arg_struct __user *, arg)
1329 struct compat_sel_arg_struct a;
1331 if (copy_from_user(&a, arg, sizeof(a)))
1333 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1334 compat_ptr(a.exp), compat_ptr(a.tvp));
1337 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1338 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1339 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1340 compat_size_t sigsetsize)
1342 compat_sigset_t ss32;
1343 sigset_t ksigmask, sigsaved;
1344 struct compat_timespec ts;
1345 struct timespec end_time, *to = NULL;
1349 if (copy_from_user(&ts, tsp, sizeof(ts)))
1353 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1358 if (sigsetsize != sizeof(compat_sigset_t))
1360 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1362 sigset_from_compat(&ksigmask, &ss32);
1364 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1365 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1368 ret = compat_core_sys_select(n, inp, outp, exp, to);
1369 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1371 if (ret == -ERESTARTNOHAND) {
1373 * Don't restore the signal mask yet. Let do_signal() deliver
1374 * the signal on the way back to userspace, before the signal
1378 memcpy(¤t->saved_sigmask, &sigsaved,
1380 set_restore_sigmask();
1383 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1388 COMPAT_SYSCALL_DEFINE6(pselect6, int, n, compat_ulong_t __user *, inp,
1389 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1390 struct compat_timespec __user *, tsp, void __user *, sig)
1392 compat_size_t sigsetsize = 0;
1393 compat_uptr_t up = 0;
1396 if (!access_ok(VERIFY_READ, sig,
1397 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1398 __get_user(up, (compat_uptr_t __user *)sig) ||
1399 __get_user(sigsetsize,
1400 (compat_size_t __user *)(sig+sizeof(up))))
1403 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1407 COMPAT_SYSCALL_DEFINE5(ppoll, struct pollfd __user *, ufds,
1408 unsigned int, nfds, struct compat_timespec __user *, tsp,
1409 const compat_sigset_t __user *, sigmask, compat_size_t, sigsetsize)
1411 compat_sigset_t ss32;
1412 sigset_t ksigmask, sigsaved;
1413 struct compat_timespec ts;
1414 struct timespec end_time, *to = NULL;
1418 if (copy_from_user(&ts, tsp, sizeof(ts)))
1422 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1427 if (sigsetsize != sizeof(compat_sigset_t))
1429 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1431 sigset_from_compat(&ksigmask, &ss32);
1433 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1434 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1437 ret = do_sys_poll(ufds, nfds, to);
1439 /* We can restart this syscall, usually */
1440 if (ret == -EINTR) {
1442 * Don't restore the signal mask yet. Let do_signal() deliver
1443 * the signal on the way back to userspace, before the signal
1447 memcpy(¤t->saved_sigmask, &sigsaved,
1449 set_restore_sigmask();
1451 ret = -ERESTARTNOHAND;
1453 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1455 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1460 #ifdef CONFIG_FHANDLE
1462 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1463 * doesn't set the O_LARGEFILE flag.
1465 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1466 struct file_handle __user *, handle, int, flags)
1468 return do_handle_open(mountdirfd, handle, flags);