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>
51 #include <asm/uaccess.h>
52 #include <asm/mmu_context.h>
53 #include <asm/ioctls.h>
58 int compat_printk(const char *fmt, ...)
65 ret = vprintk(fmt, ap);
70 #include "read_write.h"
73 * Not all architectures have sys_utime, so implement this in terms
76 asmlinkage long compat_sys_utime(const char __user *filename,
77 struct compat_utimbuf __user *t)
79 struct timespec tv[2];
82 if (get_user(tv[0].tv_sec, &t->actime) ||
83 get_user(tv[1].tv_sec, &t->modtime))
88 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
91 asmlinkage long compat_sys_utimensat(unsigned int dfd, const char __user *filename, struct compat_timespec __user *t, int flags)
93 struct timespec tv[2];
96 if (get_compat_timespec(&tv[0], &t[0]) ||
97 get_compat_timespec(&tv[1], &t[1]))
100 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
103 return do_utimes(dfd, filename, t ? tv : NULL, flags);
106 asmlinkage long compat_sys_futimesat(unsigned int dfd, const char __user *filename, struct compat_timeval __user *t)
108 struct timespec tv[2];
111 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
112 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
113 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
114 get_user(tv[1].tv_nsec, &t[1].tv_usec))
116 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
117 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
119 tv[0].tv_nsec *= 1000;
120 tv[1].tv_nsec *= 1000;
122 return do_utimes(dfd, filename, t ? tv : NULL, 0);
125 asmlinkage long compat_sys_utimes(const char __user *filename, struct compat_timeval __user *t)
127 return compat_sys_futimesat(AT_FDCWD, filename, t);
130 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
132 struct compat_stat tmp;
134 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
137 memset(&tmp, 0, sizeof(tmp));
138 tmp.st_dev = old_encode_dev(stat->dev);
139 tmp.st_ino = stat->ino;
140 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
142 tmp.st_mode = stat->mode;
143 tmp.st_nlink = stat->nlink;
144 if (tmp.st_nlink != stat->nlink)
146 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
147 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
148 tmp.st_rdev = old_encode_dev(stat->rdev);
149 if ((u64) stat->size > MAX_NON_LFS)
151 tmp.st_size = stat->size;
152 tmp.st_atime = stat->atime.tv_sec;
153 tmp.st_atime_nsec = stat->atime.tv_nsec;
154 tmp.st_mtime = stat->mtime.tv_sec;
155 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
156 tmp.st_ctime = stat->ctime.tv_sec;
157 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
158 tmp.st_blocks = stat->blocks;
159 tmp.st_blksize = stat->blksize;
160 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
163 asmlinkage long compat_sys_newstat(const char __user * filename,
164 struct compat_stat __user *statbuf)
169 error = vfs_stat(filename, &stat);
172 return cp_compat_stat(&stat, statbuf);
175 asmlinkage long compat_sys_newlstat(const char __user * filename,
176 struct compat_stat __user *statbuf)
181 error = vfs_lstat(filename, &stat);
184 return cp_compat_stat(&stat, statbuf);
187 #ifndef __ARCH_WANT_STAT64
188 asmlinkage long compat_sys_newfstatat(unsigned int dfd,
189 const char __user *filename,
190 struct compat_stat __user *statbuf, int flag)
195 error = vfs_fstatat(dfd, filename, &stat, flag);
198 return cp_compat_stat(&stat, statbuf);
202 asmlinkage long compat_sys_newfstat(unsigned int fd,
203 struct compat_stat __user * statbuf)
206 int error = vfs_fstat(fd, &stat);
209 error = cp_compat_stat(&stat, statbuf);
213 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
216 if (sizeof ubuf->f_blocks == 4) {
217 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
218 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
220 /* f_files and f_ffree may be -1; it's okay
221 * to stuff that into 32 bits */
222 if (kbuf->f_files != 0xffffffffffffffffULL
223 && (kbuf->f_files & 0xffffffff00000000ULL))
225 if (kbuf->f_ffree != 0xffffffffffffffffULL
226 && (kbuf->f_ffree & 0xffffffff00000000ULL))
229 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
230 __put_user(kbuf->f_type, &ubuf->f_type) ||
231 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
232 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
233 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
234 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
235 __put_user(kbuf->f_files, &ubuf->f_files) ||
236 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
237 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
238 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
239 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
240 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
241 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
242 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
248 * The following statfs calls are copies of code from fs/statfs.c and
249 * should be checked against those from time to time
251 asmlinkage long compat_sys_statfs(const char __user *pathname, struct compat_statfs __user *buf)
254 int error = user_statfs(pathname, &tmp);
256 error = put_compat_statfs(buf, &tmp);
260 asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
263 int error = fd_statfs(fd, &tmp);
265 error = put_compat_statfs(buf, &tmp);
269 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
271 if (sizeof ubuf->f_blocks == 4) {
272 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
273 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
275 /* f_files and f_ffree may be -1; it's okay
276 * to stuff that into 32 bits */
277 if (kbuf->f_files != 0xffffffffffffffffULL
278 && (kbuf->f_files & 0xffffffff00000000ULL))
280 if (kbuf->f_ffree != 0xffffffffffffffffULL
281 && (kbuf->f_ffree & 0xffffffff00000000ULL))
284 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
285 __put_user(kbuf->f_type, &ubuf->f_type) ||
286 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
287 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
288 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
289 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
290 __put_user(kbuf->f_files, &ubuf->f_files) ||
291 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
292 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
293 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
294 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
295 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
296 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
297 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
302 asmlinkage long compat_sys_statfs64(const char __user *pathname, compat_size_t sz, struct compat_statfs64 __user *buf)
307 if (sz != sizeof(*buf))
310 error = user_statfs(pathname, &tmp);
312 error = put_compat_statfs64(buf, &tmp);
316 asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
321 if (sz != sizeof(*buf))
324 error = fd_statfs(fd, &tmp);
326 error = put_compat_statfs64(buf, &tmp);
331 * This is a copy of sys_ustat, just dealing with a structure layout.
332 * Given how simple this syscall is that apporach is more maintainable
333 * than the various conversion hacks.
335 asmlinkage long compat_sys_ustat(unsigned dev, struct compat_ustat __user *u)
337 struct compat_ustat tmp;
339 int err = vfs_ustat(new_decode_dev(dev), &sbuf);
343 memset(&tmp, 0, sizeof(struct compat_ustat));
344 tmp.f_tfree = sbuf.f_bfree;
345 tmp.f_tinode = sbuf.f_ffree;
346 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
351 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
353 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
354 __get_user(kfl->l_type, &ufl->l_type) ||
355 __get_user(kfl->l_whence, &ufl->l_whence) ||
356 __get_user(kfl->l_start, &ufl->l_start) ||
357 __get_user(kfl->l_len, &ufl->l_len) ||
358 __get_user(kfl->l_pid, &ufl->l_pid))
363 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
365 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
366 __put_user(kfl->l_type, &ufl->l_type) ||
367 __put_user(kfl->l_whence, &ufl->l_whence) ||
368 __put_user(kfl->l_start, &ufl->l_start) ||
369 __put_user(kfl->l_len, &ufl->l_len) ||
370 __put_user(kfl->l_pid, &ufl->l_pid))
375 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
376 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
378 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
379 __get_user(kfl->l_type, &ufl->l_type) ||
380 __get_user(kfl->l_whence, &ufl->l_whence) ||
381 __get_user(kfl->l_start, &ufl->l_start) ||
382 __get_user(kfl->l_len, &ufl->l_len) ||
383 __get_user(kfl->l_pid, &ufl->l_pid))
389 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
390 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
392 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
393 __put_user(kfl->l_type, &ufl->l_type) ||
394 __put_user(kfl->l_whence, &ufl->l_whence) ||
395 __put_user(kfl->l_start, &ufl->l_start) ||
396 __put_user(kfl->l_len, &ufl->l_len) ||
397 __put_user(kfl->l_pid, &ufl->l_pid))
403 asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
414 ret = get_compat_flock(&f, compat_ptr(arg));
419 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
421 if (cmd == F_GETLK && ret == 0) {
422 /* GETLK was successful and we need to return the data...
423 * but it needs to fit in the compat structure.
424 * l_start shouldn't be too big, unless the original
425 * start + end is greater than COMPAT_OFF_T_MAX, in which
426 * case the app was asking for trouble, so we return
427 * -EOVERFLOW in that case.
428 * l_len could be too big, in which case we just truncate it,
429 * and only allow the app to see that part of the conflicting
430 * lock that might make sense to it anyway
433 if (f.l_start > COMPAT_OFF_T_MAX)
435 if (f.l_len > COMPAT_OFF_T_MAX)
436 f.l_len = COMPAT_OFF_T_MAX;
438 ret = put_compat_flock(&f, compat_ptr(arg));
445 ret = get_compat_flock64(&f, compat_ptr(arg));
450 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
451 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
454 if (cmd == F_GETLK64 && ret == 0) {
455 /* need to return lock information - see above for commentary */
456 if (f.l_start > COMPAT_LOFF_T_MAX)
458 if (f.l_len > COMPAT_LOFF_T_MAX)
459 f.l_len = COMPAT_LOFF_T_MAX;
461 ret = put_compat_flock64(&f, compat_ptr(arg));
466 ret = sys_fcntl(fd, cmd, arg);
472 asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
475 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
477 return compat_sys_fcntl64(fd, cmd, arg);
481 compat_sys_io_setup(unsigned nr_reqs, u32 __user *ctx32p)
486 mm_segment_t oldfs = get_fs();
487 if (unlikely(get_user(ctx64, ctx32p)))
491 /* The __user pointer cast is valid because of the set_fs() */
492 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
494 /* truncating is ok because it's a user address */
496 ret = put_user((u32) ctx64, ctx32p);
501 compat_sys_io_getevents(aio_context_t ctx_id,
502 unsigned long min_nr,
504 struct io_event __user *events,
505 struct compat_timespec __user *timeout)
509 struct timespec __user *ut = NULL;
512 if (unlikely(!access_ok(VERIFY_WRITE, events,
513 nr * sizeof(struct io_event))))
516 if (get_compat_timespec(&t, timeout))
519 ut = compat_alloc_user_space(sizeof(*ut));
520 if (copy_to_user(ut, &t, sizeof(t)) )
523 ret = sys_io_getevents(ctx_id, min_nr, nr, events, ut);
528 /* A write operation does a read from user space and vice versa */
529 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
531 ssize_t compat_rw_copy_check_uvector(int type,
532 const struct compat_iovec __user *uvector, unsigned long nr_segs,
533 unsigned long fast_segs, struct iovec *fast_pointer,
534 struct iovec **ret_pointer)
536 compat_ssize_t tot_len;
537 struct iovec *iov = *ret_pointer = fast_pointer;
542 * SuS says "The readv() function *may* fail if the iovcnt argument
543 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
544 * traditionally returned zero for zero segments, so...
550 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
552 if (nr_segs > fast_segs) {
554 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
561 * Single unix specification:
562 * We should -EINVAL if an element length is not >= 0 and fitting an
565 * In Linux, the total length is limited to MAX_RW_COUNT, there is
566 * no overflow possibility.
570 for (seg = 0; seg < nr_segs; seg++) {
574 if (__get_user(len, &uvector->iov_len) ||
575 __get_user(buf, &uvector->iov_base)) {
579 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
582 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
586 if (len > MAX_RW_COUNT - tot_len)
587 len = MAX_RW_COUNT - tot_len;
589 iov->iov_base = compat_ptr(buf);
590 iov->iov_len = (compat_size_t) len;
601 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
606 for (i = 0; i < nr; ++i) {
607 if (get_user(uptr, ptr32 + i))
609 if (put_user(compat_ptr(uptr), ptr64 + i))
615 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
618 compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
620 struct iocb __user * __user *iocb64;
623 if (unlikely(nr < 0))
626 if (nr > MAX_AIO_SUBMITS)
627 nr = MAX_AIO_SUBMITS;
629 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
630 ret = copy_iocb(nr, iocb, iocb64);
632 ret = do_io_submit(ctx_id, nr, iocb64, 1);
636 struct compat_ncp_mount_data {
637 compat_int_t version;
638 compat_uint_t ncp_fd;
639 __compat_uid_t mounted_uid;
640 compat_pid_t wdog_pid;
641 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
642 compat_uint_t time_out;
643 compat_uint_t retry_count;
647 compat_mode_t file_mode;
648 compat_mode_t dir_mode;
651 struct compat_ncp_mount_data_v4 {
652 compat_int_t version;
653 compat_ulong_t flags;
654 compat_ulong_t mounted_uid;
655 compat_long_t wdog_pid;
656 compat_uint_t ncp_fd;
657 compat_uint_t time_out;
658 compat_uint_t retry_count;
661 compat_ulong_t file_mode;
662 compat_ulong_t dir_mode;
665 static void *do_ncp_super_data_conv(void *raw_data)
667 int version = *(unsigned int *)raw_data;
670 struct compat_ncp_mount_data *c_n = raw_data;
671 struct ncp_mount_data *n = raw_data;
673 n->dir_mode = c_n->dir_mode;
674 n->file_mode = c_n->file_mode;
677 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
678 n->wdog_pid = c_n->wdog_pid;
679 n->mounted_uid = c_n->mounted_uid;
680 } else if (version == 4) {
681 struct compat_ncp_mount_data_v4 *c_n = raw_data;
682 struct ncp_mount_data_v4 *n = raw_data;
684 n->dir_mode = c_n->dir_mode;
685 n->file_mode = c_n->file_mode;
688 n->retry_count = c_n->retry_count;
689 n->time_out = c_n->time_out;
690 n->ncp_fd = c_n->ncp_fd;
691 n->wdog_pid = c_n->wdog_pid;
692 n->mounted_uid = c_n->mounted_uid;
693 n->flags = c_n->flags;
694 } else if (version != 5) {
702 struct compat_nfs_string {
707 static inline void compat_nfs_string(struct nfs_string *dst,
708 struct compat_nfs_string *src)
710 dst->data = compat_ptr(src->data);
714 struct compat_nfs4_mount_data_v1 {
715 compat_int_t version;
720 compat_int_t retrans;
721 compat_int_t acregmin;
722 compat_int_t acregmax;
723 compat_int_t acdirmin;
724 compat_int_t acdirmax;
725 struct compat_nfs_string client_addr;
726 struct compat_nfs_string mnt_path;
727 struct compat_nfs_string hostname;
728 compat_uint_t host_addrlen;
729 compat_uptr_t host_addr;
731 compat_int_t auth_flavourlen;
732 compat_uptr_t auth_flavours;
735 static int do_nfs4_super_data_conv(void *raw_data)
737 int version = *(compat_uint_t *) raw_data;
740 struct compat_nfs4_mount_data_v1 *raw = raw_data;
741 struct nfs4_mount_data *real = raw_data;
743 /* copy the fields backwards */
744 real->auth_flavours = compat_ptr(raw->auth_flavours);
745 real->auth_flavourlen = raw->auth_flavourlen;
746 real->proto = raw->proto;
747 real->host_addr = compat_ptr(raw->host_addr);
748 real->host_addrlen = raw->host_addrlen;
749 compat_nfs_string(&real->hostname, &raw->hostname);
750 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
751 compat_nfs_string(&real->client_addr, &raw->client_addr);
752 real->acdirmax = raw->acdirmax;
753 real->acdirmin = raw->acdirmin;
754 real->acregmax = raw->acregmax;
755 real->acregmin = raw->acregmin;
756 real->retrans = raw->retrans;
757 real->timeo = raw->timeo;
758 real->wsize = raw->wsize;
759 real->rsize = raw->rsize;
760 real->flags = raw->flags;
761 real->version = raw->version;
767 #define NCPFS_NAME "ncpfs"
768 #define NFS4_NAME "nfs4"
770 asmlinkage long compat_sys_mount(const char __user * dev_name,
771 const char __user * dir_name,
772 const char __user * type, unsigned long flags,
773 const void __user * data)
776 unsigned long data_page;
778 struct filename *dir;
781 retval = copy_mount_string(type, &kernel_type);
785 dir = getname(dir_name);
786 retval = PTR_ERR(dir);
790 retval = copy_mount_string(dev_name, &kernel_dev);
794 retval = copy_mount_options(data, &data_page);
800 if (kernel_type && data_page) {
801 if (!strcmp(kernel_type, NCPFS_NAME)) {
802 do_ncp_super_data_conv((void *)data_page);
803 } else if (!strcmp(kernel_type, NFS4_NAME)) {
804 if (do_nfs4_super_data_conv((void *) data_page))
809 retval = do_mount(kernel_dev, dir->name, kernel_type,
810 flags, (void*)data_page);
813 free_page(data_page);
824 struct compat_old_linux_dirent {
825 compat_ulong_t d_ino;
826 compat_ulong_t d_offset;
827 unsigned short d_namlen;
831 struct compat_readdir_callback {
832 struct compat_old_linux_dirent __user *dirent;
836 static int compat_fillonedir(void *__buf, const char *name, int namlen,
837 loff_t offset, u64 ino, unsigned int d_type)
839 struct compat_readdir_callback *buf = __buf;
840 struct compat_old_linux_dirent __user *dirent;
841 compat_ulong_t d_ino;
846 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
847 buf->result = -EOVERFLOW;
851 dirent = buf->dirent;
852 if (!access_ok(VERIFY_WRITE, dirent,
853 (unsigned long)(dirent->d_name + namlen + 1) -
854 (unsigned long)dirent))
856 if ( __put_user(d_ino, &dirent->d_ino) ||
857 __put_user(offset, &dirent->d_offset) ||
858 __put_user(namlen, &dirent->d_namlen) ||
859 __copy_to_user(dirent->d_name, name, namlen) ||
860 __put_user(0, dirent->d_name + namlen))
864 buf->result = -EFAULT;
868 asmlinkage long compat_sys_old_readdir(unsigned int fd,
869 struct compat_old_linux_dirent __user *dirent, unsigned int count)
872 struct fd f = fdget(fd);
873 struct compat_readdir_callback buf;
881 error = vfs_readdir(f.file, compat_fillonedir, &buf);
889 struct compat_linux_dirent {
890 compat_ulong_t d_ino;
891 compat_ulong_t d_off;
892 unsigned short d_reclen;
896 struct compat_getdents_callback {
897 struct compat_linux_dirent __user *current_dir;
898 struct compat_linux_dirent __user *previous;
903 static int compat_filldir(void *__buf, const char *name, int namlen,
904 loff_t offset, u64 ino, unsigned int d_type)
906 struct compat_linux_dirent __user * dirent;
907 struct compat_getdents_callback *buf = __buf;
908 compat_ulong_t d_ino;
909 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
910 namlen + 2, sizeof(compat_long_t));
912 buf->error = -EINVAL; /* only used if we fail.. */
913 if (reclen > buf->count)
916 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
917 buf->error = -EOVERFLOW;
920 dirent = buf->previous;
922 if (__put_user(offset, &dirent->d_off))
925 dirent = buf->current_dir;
926 if (__put_user(d_ino, &dirent->d_ino))
928 if (__put_user(reclen, &dirent->d_reclen))
930 if (copy_to_user(dirent->d_name, name, namlen))
932 if (__put_user(0, dirent->d_name + namlen))
934 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
936 buf->previous = dirent;
937 dirent = (void __user *)dirent + reclen;
938 buf->current_dir = dirent;
939 buf->count -= reclen;
942 buf->error = -EFAULT;
946 asmlinkage long compat_sys_getdents(unsigned int fd,
947 struct compat_linux_dirent __user *dirent, unsigned int count)
950 struct compat_linux_dirent __user * lastdirent;
951 struct compat_getdents_callback buf;
954 if (!access_ok(VERIFY_WRITE, dirent, count))
961 buf.current_dir = dirent;
966 error = vfs_readdir(f.file, compat_filldir, &buf);
969 lastdirent = buf.previous;
971 if (put_user(f.file->f_pos, &lastdirent->d_off))
974 error = count - buf.count;
980 #ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
982 struct compat_getdents_callback64 {
983 struct linux_dirent64 __user *current_dir;
984 struct linux_dirent64 __user *previous;
989 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
990 u64 ino, unsigned int d_type)
992 struct linux_dirent64 __user *dirent;
993 struct compat_getdents_callback64 *buf = __buf;
994 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
998 buf->error = -EINVAL; /* only used if we fail.. */
999 if (reclen > buf->count)
1001 dirent = buf->previous;
1004 if (__put_user_unaligned(offset, &dirent->d_off))
1007 dirent = buf->current_dir;
1008 if (__put_user_unaligned(ino, &dirent->d_ino))
1011 if (__put_user_unaligned(off, &dirent->d_off))
1013 if (__put_user(reclen, &dirent->d_reclen))
1015 if (__put_user(d_type, &dirent->d_type))
1017 if (copy_to_user(dirent->d_name, name, namlen))
1019 if (__put_user(0, dirent->d_name + namlen))
1021 buf->previous = dirent;
1022 dirent = (void __user *)dirent + reclen;
1023 buf->current_dir = dirent;
1024 buf->count -= reclen;
1027 buf->error = -EFAULT;
1031 asmlinkage long compat_sys_getdents64(unsigned int fd,
1032 struct linux_dirent64 __user * dirent, unsigned int count)
1035 struct linux_dirent64 __user * lastdirent;
1036 struct compat_getdents_callback64 buf;
1039 if (!access_ok(VERIFY_WRITE, dirent, count))
1046 buf.current_dir = dirent;
1047 buf.previous = NULL;
1051 error = vfs_readdir(f.file, compat_filldir64, &buf);
1054 lastdirent = buf.previous;
1056 typeof(lastdirent->d_off) d_off = f.file->f_pos;
1057 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1060 error = count - buf.count;
1065 #endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
1067 static ssize_t compat_do_readv_writev(int type, struct file *file,
1068 const struct compat_iovec __user *uvector,
1069 unsigned long nr_segs, loff_t *pos)
1071 compat_ssize_t tot_len;
1072 struct iovec iovstack[UIO_FASTIOV];
1073 struct iovec *iov = iovstack;
1083 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
1086 tot_len = compat_rw_copy_check_uvector(type, uvector, nr_segs,
1087 UIO_FASTIOV, iovstack, &iov);
1093 ret = rw_verify_area(type, file, pos, tot_len);
1099 fn = file->f_op->read;
1100 fnv = file->f_op->aio_read;
1102 fn = (io_fn_t)file->f_op->write;
1103 fnv = file->f_op->aio_write;
1107 ret = do_sync_readv_writev(file, iov, nr_segs, tot_len,
1110 ret = do_loop_readv_writev(file, iov, nr_segs, pos, fn);
1113 if (iov != iovstack)
1115 if ((ret + (type == READ)) > 0) {
1117 fsnotify_access(file);
1119 fsnotify_modify(file);
1124 static size_t compat_readv(struct file *file,
1125 const struct compat_iovec __user *vec,
1126 unsigned long vlen, loff_t *pos)
1128 ssize_t ret = -EBADF;
1130 if (!(file->f_mode & FMODE_READ))
1134 if (!file->f_op || (!file->f_op->aio_read && !file->f_op->read))
1137 ret = compat_do_readv_writev(READ, file, vec, vlen, pos);
1141 add_rchar(current, ret);
1147 compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec,
1150 struct fd f = fdget(fd);
1156 pos = f.file->f_pos;
1157 ret = compat_readv(f.file, vec, vlen, &pos);
1158 f.file->f_pos = pos;
1164 compat_sys_preadv64(unsigned long fd, const struct compat_iovec __user *vec,
1165 unsigned long vlen, loff_t pos)
1176 if (f.file->f_mode & FMODE_PREAD)
1177 ret = compat_readv(f.file, vec, vlen, &pos);
1183 compat_sys_preadv(unsigned long fd, const struct compat_iovec __user *vec,
1184 unsigned long vlen, u32 pos_low, u32 pos_high)
1186 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1187 return compat_sys_preadv64(fd, vec, vlen, pos);
1190 static size_t compat_writev(struct file *file,
1191 const struct compat_iovec __user *vec,
1192 unsigned long vlen, loff_t *pos)
1194 ssize_t ret = -EBADF;
1196 if (!(file->f_mode & FMODE_WRITE))
1200 if (!file->f_op || (!file->f_op->aio_write && !file->f_op->write))
1203 ret = compat_do_readv_writev(WRITE, file, vec, vlen, pos);
1207 add_wchar(current, ret);
1213 compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec,
1216 struct fd f = fdget(fd);
1222 pos = f.file->f_pos;
1223 ret = compat_writev(f.file, vec, vlen, &pos);
1224 f.file->f_pos = pos;
1230 compat_sys_pwritev64(unsigned long fd, const struct compat_iovec __user *vec,
1231 unsigned long vlen, loff_t pos)
1242 if (f.file->f_mode & FMODE_PWRITE)
1243 ret = compat_writev(f.file, vec, vlen, &pos);
1249 compat_sys_pwritev(unsigned long fd, const struct compat_iovec __user *vec,
1250 unsigned long vlen, u32 pos_low, u32 pos_high)
1252 loff_t pos = ((loff_t)pos_high << 32) | pos_low;
1253 return compat_sys_pwritev64(fd, vec, vlen, pos);
1257 compat_sys_vmsplice(int fd, const struct compat_iovec __user *iov32,
1258 unsigned int nr_segs, unsigned int flags)
1261 struct iovec __user *iov;
1262 if (nr_segs > UIO_MAXIOV)
1264 iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
1265 for (i = 0; i < nr_segs; i++) {
1266 struct compat_iovec v;
1267 if (get_user(v.iov_base, &iov32[i].iov_base) ||
1268 get_user(v.iov_len, &iov32[i].iov_len) ||
1269 put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
1270 put_user(v.iov_len, &iov[i].iov_len))
1273 return sys_vmsplice(fd, iov, nr_segs, flags);
1277 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1280 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1282 return do_sys_open(AT_FDCWD, filename, flags, mode);
1286 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1289 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1291 return do_sys_open(dfd, filename, flags, mode);
1294 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1296 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1297 int timeval, int ret)
1304 if (current->personality & STICKY_TIMEOUTS)
1307 /* No update for zero timeout */
1308 if (!end_time->tv_sec && !end_time->tv_nsec)
1312 ts = timespec_sub(*end_time, ts);
1314 ts.tv_sec = ts.tv_nsec = 0;
1317 struct compat_timeval rtv;
1319 rtv.tv_sec = ts.tv_sec;
1320 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1322 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1325 struct compat_timespec rts;
1327 rts.tv_sec = ts.tv_sec;
1328 rts.tv_nsec = ts.tv_nsec;
1330 if (!copy_to_user(p, &rts, sizeof(rts)))
1334 * If an application puts its timeval in read-only memory, we
1335 * don't want the Linux-specific update to the timeval to
1336 * cause a fault after the select has completed
1337 * successfully. However, because we're not updating the
1338 * timeval, we can't restart the system call.
1342 if (ret == -ERESTARTNOHAND)
1348 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1349 * 64-bit unsigned longs.
1352 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1353 unsigned long *fdset)
1355 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1359 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1366 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1369 *fdset++ = h << 32 | l;
1372 if (odd && __get_user(*fdset, ufdset))
1375 /* Tricky, must clear full unsigned long in the
1376 * kernel fdset at the end, this makes sure that
1379 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1385 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1386 unsigned long *fdset)
1389 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1400 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1405 if (odd && __put_user(*fdset, ufdset))
1412 * This is a virtual copy of sys_select from fs/select.c and probably
1413 * should be compared to it from time to time
1417 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1418 * like to be certain this leads to no problems. So I return
1419 * EINTR just for safety.
1421 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1422 * I'm trying ERESTARTNOHAND which restart only when you want to.
1424 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1425 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1426 struct timespec *end_time)
1430 int size, max_fds, ret = -EINVAL;
1431 struct fdtable *fdt;
1432 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1437 /* max_fds can increase, so grab it once to avoid race */
1439 fdt = files_fdtable(current->files);
1440 max_fds = fdt->max_fds;
1446 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1447 * since we used fdset we need to allocate memory in units of
1450 size = FDS_BYTES(n);
1452 if (size > sizeof(stack_fds) / 6) {
1453 bits = kmalloc(6 * size, GFP_KERNEL);
1458 fds.in = (unsigned long *) bits;
1459 fds.out = (unsigned long *) (bits + size);
1460 fds.ex = (unsigned long *) (bits + 2*size);
1461 fds.res_in = (unsigned long *) (bits + 3*size);
1462 fds.res_out = (unsigned long *) (bits + 4*size);
1463 fds.res_ex = (unsigned long *) (bits + 5*size);
1465 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1466 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1467 (ret = compat_get_fd_set(n, exp, fds.ex)))
1469 zero_fd_set(n, fds.res_in);
1470 zero_fd_set(n, fds.res_out);
1471 zero_fd_set(n, fds.res_ex);
1473 ret = do_select(n, &fds, end_time);
1478 ret = -ERESTARTNOHAND;
1479 if (signal_pending(current))
1484 if (compat_set_fd_set(n, inp, fds.res_in) ||
1485 compat_set_fd_set(n, outp, fds.res_out) ||
1486 compat_set_fd_set(n, exp, fds.res_ex))
1489 if (bits != stack_fds)
1495 asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
1496 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1497 struct compat_timeval __user *tvp)
1499 struct timespec end_time, *to = NULL;
1500 struct compat_timeval tv;
1504 if (copy_from_user(&tv, tvp, sizeof(tv)))
1508 if (poll_select_set_timeout(to,
1509 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1510 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1514 ret = compat_core_sys_select(n, inp, outp, exp, to);
1515 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1520 struct compat_sel_arg_struct {
1528 asmlinkage long compat_sys_old_select(struct compat_sel_arg_struct __user *arg)
1530 struct compat_sel_arg_struct a;
1532 if (copy_from_user(&a, arg, sizeof(a)))
1534 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1535 compat_ptr(a.exp), compat_ptr(a.tvp));
1538 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1539 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1540 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1541 compat_size_t sigsetsize)
1543 compat_sigset_t ss32;
1544 sigset_t ksigmask, sigsaved;
1545 struct compat_timespec ts;
1546 struct timespec end_time, *to = NULL;
1550 if (copy_from_user(&ts, tsp, sizeof(ts)))
1554 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1559 if (sigsetsize != sizeof(compat_sigset_t))
1561 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1563 sigset_from_compat(&ksigmask, &ss32);
1565 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1566 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1569 ret = compat_core_sys_select(n, inp, outp, exp, to);
1570 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1572 if (ret == -ERESTARTNOHAND) {
1574 * Don't restore the signal mask yet. Let do_signal() deliver
1575 * the signal on the way back to userspace, before the signal
1579 memcpy(¤t->saved_sigmask, &sigsaved,
1581 set_restore_sigmask();
1584 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1589 asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
1590 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1591 struct compat_timespec __user *tsp, void __user *sig)
1593 compat_size_t sigsetsize = 0;
1594 compat_uptr_t up = 0;
1597 if (!access_ok(VERIFY_READ, sig,
1598 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1599 __get_user(up, (compat_uptr_t __user *)sig) ||
1600 __get_user(sigsetsize,
1601 (compat_size_t __user *)(sig+sizeof(up))))
1604 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1608 asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1609 unsigned int nfds, struct compat_timespec __user *tsp,
1610 const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
1612 compat_sigset_t ss32;
1613 sigset_t ksigmask, sigsaved;
1614 struct compat_timespec ts;
1615 struct timespec end_time, *to = NULL;
1619 if (copy_from_user(&ts, tsp, sizeof(ts)))
1623 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1628 if (sigsetsize != sizeof(compat_sigset_t))
1630 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1632 sigset_from_compat(&ksigmask, &ss32);
1634 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1635 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1638 ret = do_sys_poll(ufds, nfds, to);
1640 /* We can restart this syscall, usually */
1641 if (ret == -EINTR) {
1643 * Don't restore the signal mask yet. Let do_signal() deliver
1644 * the signal on the way back to userspace, before the signal
1648 memcpy(¤t->saved_sigmask, &sigsaved,
1650 set_restore_sigmask();
1652 ret = -ERESTARTNOHAND;
1654 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1656 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1661 #ifdef CONFIG_FHANDLE
1663 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1664 * doesn't set the O_LARGEFILE flag.
1666 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1667 struct file_handle __user *, handle, int, flags)
1669 return do_handle_open(mountdirfd, handle, flags);