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/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>
48 #include <linux/eventpoll.h>
49 #include <linux/fs_struct.h>
50 #include <linux/slab.h>
51 #include <linux/pagemap.h>
53 #include <asm/uaccess.h>
54 #include <asm/mmu_context.h>
55 #include <asm/ioctls.h>
60 int compat_printk(const char *fmt, ...)
67 ret = vprintk(fmt, ap);
72 #include "read_write.h"
75 * Not all architectures have sys_utime, so implement this in terms
78 asmlinkage long compat_sys_utime(const char __user *filename,
79 struct compat_utimbuf __user *t)
81 struct timespec tv[2];
84 if (get_user(tv[0].tv_sec, &t->actime) ||
85 get_user(tv[1].tv_sec, &t->modtime))
90 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
93 asmlinkage long compat_sys_utimensat(unsigned int dfd, const char __user *filename, struct compat_timespec __user *t, int flags)
95 struct timespec tv[2];
98 if (get_compat_timespec(&tv[0], &t[0]) ||
99 get_compat_timespec(&tv[1], &t[1]))
102 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
105 return do_utimes(dfd, filename, t ? tv : NULL, flags);
108 asmlinkage long compat_sys_futimesat(unsigned int dfd, const char __user *filename, struct compat_timeval __user *t)
110 struct timespec tv[2];
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))
118 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
119 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
121 tv[0].tv_nsec *= 1000;
122 tv[1].tv_nsec *= 1000;
124 return do_utimes(dfd, filename, t ? tv : NULL, 0);
127 asmlinkage long compat_sys_utimes(const char __user *filename, struct compat_timeval __user *t)
129 return compat_sys_futimesat(AT_FDCWD, filename, t);
132 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
134 compat_ino_t ino = stat->ino;
135 typeof(ubuf->st_uid) uid = 0;
136 typeof(ubuf->st_gid) gid = 0;
139 SET_UID(uid, stat->uid);
140 SET_GID(gid, stat->gid);
142 if ((u64) stat->size > MAX_NON_LFS ||
143 !old_valid_dev(stat->dev) ||
144 !old_valid_dev(stat->rdev))
146 if (sizeof(ino) < sizeof(stat->ino) && ino != stat->ino)
149 if (clear_user(ubuf, sizeof(*ubuf)))
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);
171 asmlinkage long compat_sys_newstat(const char __user * filename,
172 struct compat_stat __user *statbuf)
177 error = vfs_stat(filename, &stat);
180 return cp_compat_stat(&stat, statbuf);
183 asmlinkage long compat_sys_newlstat(const char __user * filename,
184 struct compat_stat __user *statbuf)
189 error = vfs_lstat(filename, &stat);
192 return cp_compat_stat(&stat, statbuf);
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)
203 error = vfs_fstatat(dfd, filename, &stat, flag);
206 return cp_compat_stat(&stat, statbuf);
210 asmlinkage long compat_sys_newfstat(unsigned int fd,
211 struct compat_stat __user * statbuf)
214 int error = vfs_fstat(fd, &stat);
217 error = cp_compat_stat(&stat, statbuf);
221 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
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)
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))
233 if (kbuf->f_ffree != 0xffffffffffffffffULL
234 && (kbuf->f_ffree & 0xffffffff00000000ULL))
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)))
256 * The following statfs calls are copies of code from fs/statfs.c and
257 * should be checked against those from time to time
259 asmlinkage long compat_sys_statfs(const char __user *pathname, struct compat_statfs __user *buf)
262 int error = user_statfs(pathname, &tmp);
264 error = put_compat_statfs(buf, &tmp);
268 asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
271 int error = fd_statfs(fd, &tmp);
273 error = put_compat_statfs(buf, &tmp);
277 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
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)
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))
288 if (kbuf->f_ffree != 0xffffffffffffffffULL
289 && (kbuf->f_ffree & 0xffffffff00000000ULL))
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)))
310 asmlinkage long compat_sys_statfs64(const char __user *pathname, compat_size_t sz, struct compat_statfs64 __user *buf)
315 if (sz != sizeof(*buf))
318 error = user_statfs(pathname, &tmp);
320 error = put_compat_statfs64(buf, &tmp);
324 asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
329 if (sz != sizeof(*buf))
332 error = fd_statfs(fd, &tmp);
334 error = put_compat_statfs64(buf, &tmp);
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.
343 asmlinkage long compat_sys_ustat(unsigned dev, struct compat_ustat __user *u)
345 struct super_block *sb;
346 struct compat_ustat tmp;
350 sb = user_get_super(new_decode_dev(dev));
353 err = statfs_by_dentry(sb->s_root, &sbuf);
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)))
366 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
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))
378 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
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))
390 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
391 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
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))
404 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
405 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
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))
418 asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
429 ret = get_compat_flock(&f, compat_ptr(arg));
434 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
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
448 if (f.l_start > COMPAT_OFF_T_MAX)
450 if (f.l_len > COMPAT_OFF_T_MAX)
451 f.l_len = COMPAT_OFF_T_MAX;
453 ret = put_compat_flock(&f, compat_ptr(arg));
460 ret = get_compat_flock64(&f, compat_ptr(arg));
465 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
466 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
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)
473 if (f.l_len > COMPAT_LOFF_T_MAX)
474 f.l_len = COMPAT_LOFF_T_MAX;
476 ret = put_compat_flock64(&f, compat_ptr(arg));
481 ret = sys_fcntl(fd, cmd, arg);
487 asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
490 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
492 return compat_sys_fcntl64(fd, cmd, arg);
496 compat_sys_io_setup(unsigned nr_reqs, u32 __user *ctx32p)
501 mm_segment_t oldfs = get_fs();
502 if (unlikely(get_user(ctx64, ctx32p)))
506 /* The __user pointer cast is valid because of the set_fs() */
507 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
509 /* truncating is ok because it's a user address */
511 ret = put_user((u32) ctx64, ctx32p);
516 compat_sys_io_getevents(aio_context_t ctx_id,
517 unsigned long min_nr,
519 struct io_event __user *events,
520 struct compat_timespec __user *timeout)
524 struct timespec __user *ut = NULL;
527 if (unlikely(!access_ok(VERIFY_WRITE, events,
528 nr * sizeof(struct io_event))))
531 if (get_compat_timespec(&t, timeout))
534 ut = compat_alloc_user_space(sizeof(*ut));
535 if (copy_to_user(ut, &t, sizeof(t)) )
538 ret = sys_io_getevents(ctx_id, min_nr, nr, events, ut);
543 /* A write operation does a read from user space and vice versa */
544 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
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)
551 compat_ssize_t tot_len;
552 struct iovec *iov = *ret_pointer = fast_pointer;
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...
565 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
567 if (nr_segs > fast_segs) {
569 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
576 * Single unix specification:
577 * We should -EINVAL if an element length is not >= 0 and fitting an
580 * In Linux, the total length is limited to MAX_RW_COUNT, there is
581 * no overflow possibility.
585 for (seg = 0; seg < nr_segs; seg++) {
589 if (__get_user(len, &uvector->iov_len) ||
590 __get_user(buf, &uvector->iov_base)) {
594 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
597 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
601 if (len > MAX_RW_COUNT - tot_len)
602 len = MAX_RW_COUNT - tot_len;
604 iov->iov_base = compat_ptr(buf);
605 iov->iov_len = (compat_size_t) len;
616 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
621 for (i = 0; i < nr; ++i) {
622 if (get_user(uptr, ptr32 + i))
624 if (put_user(compat_ptr(uptr), ptr64 + i))
630 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
633 compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
635 struct iocb __user * __user *iocb64;
638 if (unlikely(nr < 0))
641 if (nr > MAX_AIO_SUBMITS)
642 nr = MAX_AIO_SUBMITS;
644 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
645 ret = copy_iocb(nr, iocb, iocb64);
647 ret = do_io_submit(ctx_id, nr, iocb64, 1);
651 struct compat_ncp_mount_data {
652 compat_int_t version;
653 compat_uint_t ncp_fd;
654 __compat_uid_t mounted_uid;
655 compat_pid_t wdog_pid;
656 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
657 compat_uint_t time_out;
658 compat_uint_t retry_count;
662 compat_mode_t file_mode;
663 compat_mode_t dir_mode;
666 struct compat_ncp_mount_data_v4 {
667 compat_int_t version;
668 compat_ulong_t flags;
669 compat_ulong_t mounted_uid;
670 compat_long_t wdog_pid;
671 compat_uint_t ncp_fd;
672 compat_uint_t time_out;
673 compat_uint_t retry_count;
676 compat_ulong_t file_mode;
677 compat_ulong_t dir_mode;
680 static void *do_ncp_super_data_conv(void *raw_data)
682 int version = *(unsigned int *)raw_data;
685 struct compat_ncp_mount_data *c_n = raw_data;
686 struct ncp_mount_data *n = raw_data;
688 n->dir_mode = c_n->dir_mode;
689 n->file_mode = c_n->file_mode;
692 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
693 n->wdog_pid = c_n->wdog_pid;
694 n->mounted_uid = c_n->mounted_uid;
695 } else if (version == 4) {
696 struct compat_ncp_mount_data_v4 *c_n = raw_data;
697 struct ncp_mount_data_v4 *n = raw_data;
699 n->dir_mode = c_n->dir_mode;
700 n->file_mode = c_n->file_mode;
703 n->retry_count = c_n->retry_count;
704 n->time_out = c_n->time_out;
705 n->ncp_fd = c_n->ncp_fd;
706 n->wdog_pid = c_n->wdog_pid;
707 n->mounted_uid = c_n->mounted_uid;
708 n->flags = c_n->flags;
709 } else if (version != 5) {
717 struct compat_nfs_string {
722 static inline void compat_nfs_string(struct nfs_string *dst,
723 struct compat_nfs_string *src)
725 dst->data = compat_ptr(src->data);
729 struct compat_nfs4_mount_data_v1 {
730 compat_int_t version;
735 compat_int_t retrans;
736 compat_int_t acregmin;
737 compat_int_t acregmax;
738 compat_int_t acdirmin;
739 compat_int_t acdirmax;
740 struct compat_nfs_string client_addr;
741 struct compat_nfs_string mnt_path;
742 struct compat_nfs_string hostname;
743 compat_uint_t host_addrlen;
744 compat_uptr_t host_addr;
746 compat_int_t auth_flavourlen;
747 compat_uptr_t auth_flavours;
750 static int do_nfs4_super_data_conv(void *raw_data)
752 int version = *(compat_uint_t *) raw_data;
755 struct compat_nfs4_mount_data_v1 *raw = raw_data;
756 struct nfs4_mount_data *real = raw_data;
758 /* copy the fields backwards */
759 real->auth_flavours = compat_ptr(raw->auth_flavours);
760 real->auth_flavourlen = raw->auth_flavourlen;
761 real->proto = raw->proto;
762 real->host_addr = compat_ptr(raw->host_addr);
763 real->host_addrlen = raw->host_addrlen;
764 compat_nfs_string(&real->hostname, &raw->hostname);
765 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
766 compat_nfs_string(&real->client_addr, &raw->client_addr);
767 real->acdirmax = raw->acdirmax;
768 real->acdirmin = raw->acdirmin;
769 real->acregmax = raw->acregmax;
770 real->acregmin = raw->acregmin;
771 real->retrans = raw->retrans;
772 real->timeo = raw->timeo;
773 real->wsize = raw->wsize;
774 real->rsize = raw->rsize;
775 real->flags = raw->flags;
776 real->version = raw->version;
782 #define NCPFS_NAME "ncpfs"
783 #define NFS4_NAME "nfs4"
785 asmlinkage long compat_sys_mount(const char __user * dev_name,
786 const char __user * dir_name,
787 const char __user * type, unsigned long flags,
788 const void __user * data)
791 unsigned long data_page;
796 retval = copy_mount_string(type, &kernel_type);
800 dir_page = getname(dir_name);
801 retval = PTR_ERR(dir_page);
802 if (IS_ERR(dir_page))
805 retval = copy_mount_string(dev_name, &kernel_dev);
809 retval = copy_mount_options(data, &data_page);
815 if (kernel_type && data_page) {
816 if (!strcmp(kernel_type, NCPFS_NAME)) {
817 do_ncp_super_data_conv((void *)data_page);
818 } else if (!strcmp(kernel_type, NFS4_NAME)) {
819 if (do_nfs4_super_data_conv((void *) data_page))
824 retval = do_mount(kernel_dev, dir_page, kernel_type,
825 flags, (void*)data_page);
828 free_page(data_page);
839 struct compat_old_linux_dirent {
840 compat_ulong_t d_ino;
841 compat_ulong_t d_offset;
842 unsigned short d_namlen;
846 struct compat_readdir_callback {
847 struct compat_old_linux_dirent __user *dirent;
851 static int compat_fillonedir(void *__buf, const char *name, int namlen,
852 loff_t offset, u64 ino, unsigned int d_type)
854 struct compat_readdir_callback *buf = __buf;
855 struct compat_old_linux_dirent __user *dirent;
856 compat_ulong_t d_ino;
861 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
862 buf->result = -EOVERFLOW;
866 dirent = buf->dirent;
867 if (!access_ok(VERIFY_WRITE, dirent,
868 (unsigned long)(dirent->d_name + namlen + 1) -
869 (unsigned long)dirent))
871 if ( __put_user(d_ino, &dirent->d_ino) ||
872 __put_user(offset, &dirent->d_offset) ||
873 __put_user(namlen, &dirent->d_namlen) ||
874 __copy_to_user(dirent->d_name, name, namlen) ||
875 __put_user(0, dirent->d_name + namlen))
879 buf->result = -EFAULT;
883 asmlinkage long compat_sys_old_readdir(unsigned int fd,
884 struct compat_old_linux_dirent __user *dirent, unsigned int count)
888 struct compat_readdir_callback buf;
898 error = vfs_readdir(file, compat_fillonedir, &buf);
907 struct compat_linux_dirent {
908 compat_ulong_t d_ino;
909 compat_ulong_t d_off;
910 unsigned short d_reclen;
914 struct compat_getdents_callback {
915 struct compat_linux_dirent __user *current_dir;
916 struct compat_linux_dirent __user *previous;
921 static int compat_filldir(void *__buf, const char *name, int namlen,
922 loff_t offset, u64 ino, unsigned int d_type)
924 struct compat_linux_dirent __user * dirent;
925 struct compat_getdents_callback *buf = __buf;
926 compat_ulong_t d_ino;
927 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
928 namlen + 2, sizeof(compat_long_t));
930 buf->error = -EINVAL; /* only used if we fail.. */
931 if (reclen > buf->count)
934 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
935 buf->error = -EOVERFLOW;
938 dirent = buf->previous;
940 if (__put_user(offset, &dirent->d_off))
943 dirent = buf->current_dir;
944 if (__put_user(d_ino, &dirent->d_ino))
946 if (__put_user(reclen, &dirent->d_reclen))
948 if (copy_to_user(dirent->d_name, name, namlen))
950 if (__put_user(0, dirent->d_name + namlen))
952 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
954 buf->previous = dirent;
955 dirent = (void __user *)dirent + reclen;
956 buf->current_dir = dirent;
957 buf->count -= reclen;
960 buf->error = -EFAULT;
964 asmlinkage long compat_sys_getdents(unsigned int fd,
965 struct compat_linux_dirent __user *dirent, unsigned int count)
968 struct compat_linux_dirent __user * lastdirent;
969 struct compat_getdents_callback buf;
973 if (!access_ok(VERIFY_WRITE, dirent, count))
981 buf.current_dir = dirent;
986 error = vfs_readdir(file, compat_filldir, &buf);
989 lastdirent = buf.previous;
991 if (put_user(file->f_pos, &lastdirent->d_off))
994 error = count - buf.count;
1001 #ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
1003 struct compat_getdents_callback64 {
1004 struct linux_dirent64 __user *current_dir;
1005 struct linux_dirent64 __user *previous;
1010 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
1011 u64 ino, unsigned int d_type)
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,
1019 buf->error = -EINVAL; /* only used if we fail.. */
1020 if (reclen > buf->count)
1022 dirent = buf->previous;
1025 if (__put_user_unaligned(offset, &dirent->d_off))
1028 dirent = buf->current_dir;
1029 if (__put_user_unaligned(ino, &dirent->d_ino))
1032 if (__put_user_unaligned(off, &dirent->d_off))
1034 if (__put_user(reclen, &dirent->d_reclen))
1036 if (__put_user(d_type, &dirent->d_type))
1038 if (copy_to_user(dirent->d_name, name, namlen))
1040 if (__put_user(0, dirent->d_name + namlen))
1042 buf->previous = dirent;
1043 dirent = (void __user *)dirent + reclen;
1044 buf->current_dir = dirent;
1045 buf->count -= reclen;
1048 buf->error = -EFAULT;
1052 asmlinkage long compat_sys_getdents64(unsigned int fd,
1053 struct linux_dirent64 __user * dirent, unsigned int count)
1056 struct linux_dirent64 __user * lastdirent;
1057 struct compat_getdents_callback64 buf;
1061 if (!access_ok(VERIFY_WRITE, dirent, count))
1069 buf.current_dir = dirent;
1070 buf.previous = NULL;
1074 error = vfs_readdir(file, compat_filldir64, &buf);
1077 lastdirent = buf.previous;
1079 typeof(lastdirent->d_off) d_off = file->f_pos;
1080 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1083 error = count - buf.count;
1089 #endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
1091 static ssize_t compat_do_readv_writev(int type, struct file *file,
1092 const struct compat_iovec __user *uvector,
1093 unsigned long nr_segs, loff_t *pos)
1095 compat_ssize_t tot_len;
1096 struct iovec iovstack[UIO_FASTIOV];
1097 struct iovec *iov = iovstack;
1107 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
1110 tot_len = compat_rw_copy_check_uvector(type, uvector, nr_segs,
1111 UIO_FASTIOV, iovstack, &iov, 1);
1117 ret = rw_verify_area(type, file, pos, tot_len);
1123 fn = file->f_op->read;
1124 fnv = file->f_op->aio_read;
1126 fn = (io_fn_t)file->f_op->write;
1127 fnv = file->f_op->aio_write;
1131 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
1134 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
1137 if (iov != iovstack)
1139 if ((ret + (type == READ)) > 0) {
1141 fsnotify_access(file);
1143 fsnotify_modify(file);
1148 static size_t compat_readv(struct file *file,
1149 const struct compat_iovec __user *vec,
1150 unsigned long vlen, loff_t *pos)
1152 ssize_t ret = -EBADF;
1154 if (!(file->f_mode & FMODE_READ))
1158 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
1161 ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
1165 add_rchar(current, ret);
1171 compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec,
1178 file = fget_light(fd, &fput_needed);
1181 ret = compat_readv(file, vec, vlen, &file->f_pos);
1182 fput_light(file, fput_needed);
1187 compat_sys_preadv(unsigned long fd, const struct compat_iovec __user *vec,
1188 unsigned long vlen, u32 pos_low, u32 pos_high)
1190 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1197 file = fget_light(fd, &fput_needed);
1201 if (file->f_mode & FMODE_PREAD)
1202 ret = compat_readv(file, vec, vlen, &pos);
1203 fput_light(file, fput_needed);
1207 static size_t compat_writev(struct file *file,
1208 const struct compat_iovec __user *vec,
1209 unsigned long vlen, loff_t *pos)
1211 ssize_t ret = -EBADF;
1213 if (!(file->f_mode & FMODE_WRITE))
1217 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
1220 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
1224 add_wchar(current, ret);
1230 compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec,
1237 file = fget_light(fd, &fput_needed);
1240 ret = compat_writev(file, vec, vlen, &file->f_pos);
1241 fput_light(file, fput_needed);
1246 compat_sys_pwritev(unsigned long fd, const struct compat_iovec __user *vec,
1247 unsigned long vlen, u32 pos_low, u32 pos_high)
1249 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1256 file = fget_light(fd, &fput_needed);
1260 if (file->f_mode & FMODE_PWRITE)
1261 ret = compat_writev(file, vec, vlen, &pos);
1262 fput_light(file, fput_needed);
1267 compat_sys_vmsplice(int fd, const struct compat_iovec __user *iov32,
1268 unsigned int nr_segs, unsigned int flags)
1271 struct iovec __user *iov;
1272 if (nr_segs > UIO_MAXIOV)
1274 iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
1275 for (i = 0; i < nr_segs; i++) {
1276 struct compat_iovec v;
1277 if (get_user(v.iov_base, &iov32[i].iov_base) ||
1278 get_user(v.iov_len, &iov32[i].iov_len) ||
1279 put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
1280 put_user(v.iov_len, &iov[i].iov_len))
1283 return sys_vmsplice(fd, iov, nr_segs, flags);
1287 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1291 compat_sys_open(const char __user *filename, int flags, int mode)
1293 return do_sys_open(AT_FDCWD, filename, flags, mode);
1297 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1301 compat_sys_openat(unsigned int dfd, const char __user *filename, int flags, int mode)
1303 return do_sys_open(dfd, filename, flags, mode);
1306 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1308 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1309 int timeval, int ret)
1316 if (current->personality & STICKY_TIMEOUTS)
1319 /* No update for zero timeout */
1320 if (!end_time->tv_sec && !end_time->tv_nsec)
1324 ts = timespec_sub(*end_time, ts);
1326 ts.tv_sec = ts.tv_nsec = 0;
1329 struct compat_timeval rtv;
1331 rtv.tv_sec = ts.tv_sec;
1332 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1334 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1337 struct compat_timespec rts;
1339 rts.tv_sec = ts.tv_sec;
1340 rts.tv_nsec = ts.tv_nsec;
1342 if (!copy_to_user(p, &rts, sizeof(rts)))
1346 * If an application puts its timeval in read-only memory, we
1347 * don't want the Linux-specific update to the timeval to
1348 * cause a fault after the select has completed
1349 * successfully. However, because we're not updating the
1350 * timeval, we can't restart the system call.
1354 if (ret == -ERESTARTNOHAND)
1360 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1361 * 64-bit unsigned longs.
1364 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1365 unsigned long *fdset)
1367 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1371 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1378 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1381 *fdset++ = h << 32 | l;
1384 if (odd && __get_user(*fdset, ufdset))
1387 /* Tricky, must clear full unsigned long in the
1388 * kernel fdset at the end, this makes sure that
1391 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1397 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1398 unsigned long *fdset)
1401 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1412 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1417 if (odd && __put_user(*fdset, ufdset))
1424 * This is a virtual copy of sys_select from fs/select.c and probably
1425 * should be compared to it from time to time
1429 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1430 * like to be certain this leads to no problems. So I return
1431 * EINTR just for safety.
1433 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1434 * I'm trying ERESTARTNOHAND which restart only when you want to.
1436 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1437 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1438 struct timespec *end_time)
1442 int size, max_fds, ret = -EINVAL;
1443 struct fdtable *fdt;
1444 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1449 /* max_fds can increase, so grab it once to avoid race */
1451 fdt = files_fdtable(current->files);
1452 max_fds = fdt->max_fds;
1458 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1459 * since we used fdset we need to allocate memory in units of
1462 size = FDS_BYTES(n);
1464 if (size > sizeof(stack_fds) / 6) {
1465 bits = kmalloc(6 * size, GFP_KERNEL);
1470 fds.in = (unsigned long *) bits;
1471 fds.out = (unsigned long *) (bits + size);
1472 fds.ex = (unsigned long *) (bits + 2*size);
1473 fds.res_in = (unsigned long *) (bits + 3*size);
1474 fds.res_out = (unsigned long *) (bits + 4*size);
1475 fds.res_ex = (unsigned long *) (bits + 5*size);
1477 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1478 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1479 (ret = compat_get_fd_set(n, exp, fds.ex)))
1481 zero_fd_set(n, fds.res_in);
1482 zero_fd_set(n, fds.res_out);
1483 zero_fd_set(n, fds.res_ex);
1485 ret = do_select(n, &fds, end_time);
1490 ret = -ERESTARTNOHAND;
1491 if (signal_pending(current))
1496 if (compat_set_fd_set(n, inp, fds.res_in) ||
1497 compat_set_fd_set(n, outp, fds.res_out) ||
1498 compat_set_fd_set(n, exp, fds.res_ex))
1501 if (bits != stack_fds)
1507 asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
1508 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1509 struct compat_timeval __user *tvp)
1511 struct timespec end_time, *to = NULL;
1512 struct compat_timeval tv;
1516 if (copy_from_user(&tv, tvp, sizeof(tv)))
1520 if (poll_select_set_timeout(to,
1521 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1522 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1526 ret = compat_core_sys_select(n, inp, outp, exp, to);
1527 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1532 struct compat_sel_arg_struct {
1540 asmlinkage long compat_sys_old_select(struct compat_sel_arg_struct __user *arg)
1542 struct compat_sel_arg_struct a;
1544 if (copy_from_user(&a, arg, sizeof(a)))
1546 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1547 compat_ptr(a.exp), compat_ptr(a.tvp));
1550 #ifdef HAVE_SET_RESTORE_SIGMASK
1551 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1552 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1553 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1554 compat_size_t sigsetsize)
1556 compat_sigset_t ss32;
1557 sigset_t ksigmask, sigsaved;
1558 struct compat_timespec ts;
1559 struct timespec end_time, *to = NULL;
1563 if (copy_from_user(&ts, tsp, sizeof(ts)))
1567 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1572 if (sigsetsize != sizeof(compat_sigset_t))
1574 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1576 sigset_from_compat(&ksigmask, &ss32);
1578 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1579 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1582 ret = compat_core_sys_select(n, inp, outp, exp, to);
1583 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1585 if (ret == -ERESTARTNOHAND) {
1587 * Don't restore the signal mask yet. Let do_signal() deliver
1588 * the signal on the way back to userspace, before the signal
1592 memcpy(¤t->saved_sigmask, &sigsaved,
1594 set_restore_sigmask();
1597 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1602 asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
1603 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1604 struct compat_timespec __user *tsp, void __user *sig)
1606 compat_size_t sigsetsize = 0;
1607 compat_uptr_t up = 0;
1610 if (!access_ok(VERIFY_READ, sig,
1611 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1612 __get_user(up, (compat_uptr_t __user *)sig) ||
1613 __get_user(sigsetsize,
1614 (compat_size_t __user *)(sig+sizeof(up))))
1617 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1621 asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1622 unsigned int nfds, struct compat_timespec __user *tsp,
1623 const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
1625 compat_sigset_t ss32;
1626 sigset_t ksigmask, sigsaved;
1627 struct compat_timespec ts;
1628 struct timespec end_time, *to = NULL;
1632 if (copy_from_user(&ts, tsp, sizeof(ts)))
1636 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1641 if (sigsetsize != sizeof(compat_sigset_t))
1643 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1645 sigset_from_compat(&ksigmask, &ss32);
1647 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1648 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1651 ret = do_sys_poll(ufds, nfds, to);
1653 /* We can restart this syscall, usually */
1654 if (ret == -EINTR) {
1656 * Don't restore the signal mask yet. Let do_signal() deliver
1657 * the signal on the way back to userspace, before the signal
1661 memcpy(¤t->saved_sigmask, &sigsaved,
1663 set_restore_sigmask();
1665 ret = -ERESTARTNOHAND;
1667 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1669 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1673 #endif /* HAVE_SET_RESTORE_SIGMASK */
1677 #ifdef HAVE_SET_RESTORE_SIGMASK
1678 asmlinkage long compat_sys_epoll_pwait(int epfd,
1679 struct compat_epoll_event __user *events,
1680 int maxevents, int timeout,
1681 const compat_sigset_t __user *sigmask,
1682 compat_size_t sigsetsize)
1685 compat_sigset_t csigmask;
1686 sigset_t ksigmask, sigsaved;
1689 * If the caller wants a certain signal mask to be set during the wait,
1693 if (sigsetsize != sizeof(compat_sigset_t))
1695 if (copy_from_user(&csigmask, sigmask, sizeof(csigmask)))
1697 sigset_from_compat(&ksigmask, &csigmask);
1698 sigdelsetmask(&ksigmask, sigmask(SIGKILL) | sigmask(SIGSTOP));
1699 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1702 err = sys_epoll_wait(epfd, events, maxevents, timeout);
1705 * If we changed the signal mask, we need to restore the original one.
1706 * In case we've got a signal while waiting, we do not restore the
1707 * signal mask yet, and we allow do_signal() to deliver the signal on
1708 * the way back to userspace, before the signal mask is restored.
1711 if (err == -EINTR) {
1712 memcpy(¤t->saved_sigmask, &sigsaved,
1714 set_restore_sigmask();
1716 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1721 #endif /* HAVE_SET_RESTORE_SIGMASK */
1723 #endif /* CONFIG_EPOLL */
1725 #ifdef CONFIG_SIGNALFD
1727 asmlinkage long compat_sys_signalfd4(int ufd,
1728 const compat_sigset_t __user *sigmask,
1729 compat_size_t sigsetsize, int flags)
1731 compat_sigset_t ss32;
1733 sigset_t __user *ksigmask;
1735 if (sigsetsize != sizeof(compat_sigset_t))
1737 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1739 sigset_from_compat(&tmp, &ss32);
1740 ksigmask = compat_alloc_user_space(sizeof(sigset_t));
1741 if (copy_to_user(ksigmask, &tmp, sizeof(sigset_t)))
1744 return sys_signalfd4(ufd, ksigmask, sizeof(sigset_t), flags);
1747 asmlinkage long compat_sys_signalfd(int ufd,
1748 const compat_sigset_t __user *sigmask,
1749 compat_size_t sigsetsize)
1751 return compat_sys_signalfd4(ufd, sigmask, sigsetsize, 0);
1753 #endif /* CONFIG_SIGNALFD */
1755 #ifdef CONFIG_TIMERFD
1757 asmlinkage long compat_sys_timerfd_settime(int ufd, int flags,
1758 const struct compat_itimerspec __user *utmr,
1759 struct compat_itimerspec __user *otmr)
1762 struct itimerspec t;
1763 struct itimerspec __user *ut;
1765 if (get_compat_itimerspec(&t, utmr))
1767 ut = compat_alloc_user_space(2 * sizeof(struct itimerspec));
1768 if (copy_to_user(&ut[0], &t, sizeof(t)))
1770 error = sys_timerfd_settime(ufd, flags, &ut[0], &ut[1]);
1772 error = (copy_from_user(&t, &ut[1], sizeof(struct itimerspec)) ||
1773 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1778 asmlinkage long compat_sys_timerfd_gettime(int ufd,
1779 struct compat_itimerspec __user *otmr)
1782 struct itimerspec t;
1783 struct itimerspec __user *ut;
1785 ut = compat_alloc_user_space(sizeof(struct itimerspec));
1786 error = sys_timerfd_gettime(ufd, ut);
1788 error = (copy_from_user(&t, ut, sizeof(struct itimerspec)) ||
1789 put_compat_itimerspec(otmr, &t)) ? -EFAULT: 0;
1794 #endif /* CONFIG_TIMERFD */
1796 #ifdef CONFIG_FHANDLE
1798 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1799 * doesn't set the O_LARGEFILE flag.
1802 compat_sys_open_by_handle_at(int mountdirfd,
1803 struct file_handle __user *handle, int flags)
1805 return do_handle_open(mountdirfd, handle, flags);