4 * Kernel compatibililty routines for e.g. 32 bit syscall support
7 * Copyright (C) 2002 Stephen Rothwell, IBM Corporation
10 * Copyright (C) 2001,2002 Andi Kleen, SuSE Labs
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 <linux/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_bsize) == 4) {
257 if ((kbuf->f_type | kbuf->f_bsize | kbuf->f_namelen |
258 kbuf->f_frsize | kbuf->f_flags) & 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 /* A write operation does a read from user space and vice versa */
491 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
493 ssize_t compat_rw_copy_check_uvector(int type,
494 const struct compat_iovec __user *uvector, unsigned long nr_segs,
495 unsigned long fast_segs, struct iovec *fast_pointer,
496 struct iovec **ret_pointer)
498 compat_ssize_t tot_len;
499 struct iovec *iov = *ret_pointer = fast_pointer;
504 * SuS says "The readv() function *may* fail if the iovcnt argument
505 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
506 * traditionally returned zero for zero segments, so...
512 if (nr_segs > UIO_MAXIOV)
514 if (nr_segs > fast_segs) {
516 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
523 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
527 * Single unix specification:
528 * We should -EINVAL if an element length is not >= 0 and fitting an
531 * In Linux, the total length is limited to MAX_RW_COUNT, there is
532 * no overflow possibility.
536 for (seg = 0; seg < nr_segs; seg++) {
540 if (__get_user(len, &uvector->iov_len) ||
541 __get_user(buf, &uvector->iov_base)) {
545 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
548 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
552 if (len > MAX_RW_COUNT - tot_len)
553 len = MAX_RW_COUNT - tot_len;
555 iov->iov_base = compat_ptr(buf);
556 iov->iov_len = (compat_size_t) len;
566 struct compat_ncp_mount_data {
567 compat_int_t version;
568 compat_uint_t ncp_fd;
569 __compat_uid_t mounted_uid;
570 compat_pid_t wdog_pid;
571 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
572 compat_uint_t time_out;
573 compat_uint_t retry_count;
577 compat_mode_t file_mode;
578 compat_mode_t dir_mode;
581 struct compat_ncp_mount_data_v4 {
582 compat_int_t version;
583 compat_ulong_t flags;
584 compat_ulong_t mounted_uid;
585 compat_long_t wdog_pid;
586 compat_uint_t ncp_fd;
587 compat_uint_t time_out;
588 compat_uint_t retry_count;
591 compat_ulong_t file_mode;
592 compat_ulong_t dir_mode;
595 static void *do_ncp_super_data_conv(void *raw_data)
597 int version = *(unsigned int *)raw_data;
600 struct compat_ncp_mount_data *c_n = raw_data;
601 struct ncp_mount_data *n = raw_data;
603 n->dir_mode = c_n->dir_mode;
604 n->file_mode = c_n->file_mode;
607 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
608 n->wdog_pid = c_n->wdog_pid;
609 n->mounted_uid = c_n->mounted_uid;
610 } else if (version == 4) {
611 struct compat_ncp_mount_data_v4 *c_n = raw_data;
612 struct ncp_mount_data_v4 *n = raw_data;
614 n->dir_mode = c_n->dir_mode;
615 n->file_mode = c_n->file_mode;
618 n->retry_count = c_n->retry_count;
619 n->time_out = c_n->time_out;
620 n->ncp_fd = c_n->ncp_fd;
621 n->wdog_pid = c_n->wdog_pid;
622 n->mounted_uid = c_n->mounted_uid;
623 n->flags = c_n->flags;
624 } else if (version != 5) {
632 struct compat_nfs_string {
637 static inline void compat_nfs_string(struct nfs_string *dst,
638 struct compat_nfs_string *src)
640 dst->data = compat_ptr(src->data);
644 struct compat_nfs4_mount_data_v1 {
645 compat_int_t version;
650 compat_int_t retrans;
651 compat_int_t acregmin;
652 compat_int_t acregmax;
653 compat_int_t acdirmin;
654 compat_int_t acdirmax;
655 struct compat_nfs_string client_addr;
656 struct compat_nfs_string mnt_path;
657 struct compat_nfs_string hostname;
658 compat_uint_t host_addrlen;
659 compat_uptr_t host_addr;
661 compat_int_t auth_flavourlen;
662 compat_uptr_t auth_flavours;
665 static int do_nfs4_super_data_conv(void *raw_data)
667 int version = *(compat_uint_t *) raw_data;
670 struct compat_nfs4_mount_data_v1 *raw = raw_data;
671 struct nfs4_mount_data *real = raw_data;
673 /* copy the fields backwards */
674 real->auth_flavours = compat_ptr(raw->auth_flavours);
675 real->auth_flavourlen = raw->auth_flavourlen;
676 real->proto = raw->proto;
677 real->host_addr = compat_ptr(raw->host_addr);
678 real->host_addrlen = raw->host_addrlen;
679 compat_nfs_string(&real->hostname, &raw->hostname);
680 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
681 compat_nfs_string(&real->client_addr, &raw->client_addr);
682 real->acdirmax = raw->acdirmax;
683 real->acdirmin = raw->acdirmin;
684 real->acregmax = raw->acregmax;
685 real->acregmin = raw->acregmin;
686 real->retrans = raw->retrans;
687 real->timeo = raw->timeo;
688 real->wsize = raw->wsize;
689 real->rsize = raw->rsize;
690 real->flags = raw->flags;
691 real->version = raw->version;
697 #define NCPFS_NAME "ncpfs"
698 #define NFS4_NAME "nfs4"
700 COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name,
701 const char __user *, dir_name,
702 const char __user *, type, compat_ulong_t, flags,
703 const void __user *, data)
710 kernel_type = copy_mount_string(type);
711 retval = PTR_ERR(kernel_type);
712 if (IS_ERR(kernel_type))
715 kernel_dev = copy_mount_string(dev_name);
716 retval = PTR_ERR(kernel_dev);
717 if (IS_ERR(kernel_dev))
720 options = copy_mount_options(data);
721 retval = PTR_ERR(options);
725 if (kernel_type && options) {
726 if (!strcmp(kernel_type, NCPFS_NAME)) {
727 do_ncp_super_data_conv(options);
728 } else if (!strcmp(kernel_type, NFS4_NAME)) {
730 if (do_nfs4_super_data_conv(options))
735 retval = do_mount(kernel_dev, dir_name, kernel_type, flags, options);
747 struct compat_old_linux_dirent {
748 compat_ulong_t d_ino;
749 compat_ulong_t d_offset;
750 unsigned short d_namlen;
754 struct compat_readdir_callback {
755 struct dir_context ctx;
756 struct compat_old_linux_dirent __user *dirent;
760 static int compat_fillonedir(struct dir_context *ctx, const char *name,
761 int namlen, loff_t offset, u64 ino,
764 struct compat_readdir_callback *buf =
765 container_of(ctx, struct compat_readdir_callback, ctx);
766 struct compat_old_linux_dirent __user *dirent;
767 compat_ulong_t d_ino;
772 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
773 buf->result = -EOVERFLOW;
777 dirent = buf->dirent;
778 if (!access_ok(VERIFY_WRITE, dirent,
779 (unsigned long)(dirent->d_name + namlen + 1) -
780 (unsigned long)dirent))
782 if ( __put_user(d_ino, &dirent->d_ino) ||
783 __put_user(offset, &dirent->d_offset) ||
784 __put_user(namlen, &dirent->d_namlen) ||
785 __copy_to_user(dirent->d_name, name, namlen) ||
786 __put_user(0, dirent->d_name + namlen))
790 buf->result = -EFAULT;
794 COMPAT_SYSCALL_DEFINE3(old_readdir, unsigned int, fd,
795 struct compat_old_linux_dirent __user *, dirent, unsigned int, count)
798 struct fd f = fdget_pos(fd);
799 struct compat_readdir_callback buf = {
800 .ctx.actor = compat_fillonedir,
807 error = iterate_dir(f.file, &buf.ctx);
815 struct compat_linux_dirent {
816 compat_ulong_t d_ino;
817 compat_ulong_t d_off;
818 unsigned short d_reclen;
822 struct compat_getdents_callback {
823 struct dir_context ctx;
824 struct compat_linux_dirent __user *current_dir;
825 struct compat_linux_dirent __user *previous;
830 static int compat_filldir(struct dir_context *ctx, const char *name, int namlen,
831 loff_t offset, u64 ino, unsigned int d_type)
833 struct compat_linux_dirent __user * dirent;
834 struct compat_getdents_callback *buf =
835 container_of(ctx, struct compat_getdents_callback, ctx);
836 compat_ulong_t d_ino;
837 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
838 namlen + 2, sizeof(compat_long_t));
840 buf->error = -EINVAL; /* only used if we fail.. */
841 if (reclen > buf->count)
844 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
845 buf->error = -EOVERFLOW;
848 dirent = buf->previous;
850 if (signal_pending(current))
852 if (__put_user(offset, &dirent->d_off))
855 dirent = buf->current_dir;
856 if (__put_user(d_ino, &dirent->d_ino))
858 if (__put_user(reclen, &dirent->d_reclen))
860 if (copy_to_user(dirent->d_name, name, namlen))
862 if (__put_user(0, dirent->d_name + namlen))
864 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
866 buf->previous = dirent;
867 dirent = (void __user *)dirent + reclen;
868 buf->current_dir = dirent;
869 buf->count -= reclen;
872 buf->error = -EFAULT;
876 COMPAT_SYSCALL_DEFINE3(getdents, unsigned int, fd,
877 struct compat_linux_dirent __user *, dirent, unsigned int, count)
880 struct compat_linux_dirent __user * lastdirent;
881 struct compat_getdents_callback buf = {
882 .ctx.actor = compat_filldir,
883 .current_dir = dirent,
888 if (!access_ok(VERIFY_WRITE, dirent, count))
895 error = iterate_dir(f.file, &buf.ctx);
898 lastdirent = buf.previous;
900 if (put_user(buf.ctx.pos, &lastdirent->d_off))
903 error = count - buf.count;
909 #ifdef __ARCH_WANT_COMPAT_SYS_GETDENTS64
911 struct compat_getdents_callback64 {
912 struct dir_context ctx;
913 struct linux_dirent64 __user *current_dir;
914 struct linux_dirent64 __user *previous;
919 static int compat_filldir64(struct dir_context *ctx, const char *name,
920 int namlen, loff_t offset, u64 ino,
923 struct linux_dirent64 __user *dirent;
924 struct compat_getdents_callback64 *buf =
925 container_of(ctx, struct compat_getdents_callback64, ctx);
926 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
930 buf->error = -EINVAL; /* only used if we fail.. */
931 if (reclen > buf->count)
933 dirent = buf->previous;
936 if (signal_pending(current))
938 if (__put_user_unaligned(offset, &dirent->d_off))
941 dirent = buf->current_dir;
942 if (__put_user_unaligned(ino, &dirent->d_ino))
945 if (__put_user_unaligned(off, &dirent->d_off))
947 if (__put_user(reclen, &dirent->d_reclen))
949 if (__put_user(d_type, &dirent->d_type))
951 if (copy_to_user(dirent->d_name, name, namlen))
953 if (__put_user(0, dirent->d_name + namlen))
955 buf->previous = dirent;
956 dirent = (void __user *)dirent + reclen;
957 buf->current_dir = dirent;
958 buf->count -= reclen;
961 buf->error = -EFAULT;
965 COMPAT_SYSCALL_DEFINE3(getdents64, unsigned int, fd,
966 struct linux_dirent64 __user *, dirent, unsigned int, count)
969 struct linux_dirent64 __user * lastdirent;
970 struct compat_getdents_callback64 buf = {
971 .ctx.actor = compat_filldir64,
972 .current_dir = dirent,
977 if (!access_ok(VERIFY_WRITE, dirent, count))
984 error = iterate_dir(f.file, &buf.ctx);
987 lastdirent = buf.previous;
989 typeof(lastdirent->d_off) d_off = buf.ctx.pos;
990 if (__put_user_unaligned(d_off, &lastdirent->d_off))
993 error = count - buf.count;
998 #endif /* __ARCH_WANT_COMPAT_SYS_GETDENTS64 */
1001 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1004 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1006 return do_sys_open(AT_FDCWD, filename, flags, mode);
1010 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1013 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1015 return do_sys_open(dfd, filename, flags, mode);
1018 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1020 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1021 int timeval, int ret)
1028 if (current->personality & STICKY_TIMEOUTS)
1031 /* No update for zero timeout */
1032 if (!end_time->tv_sec && !end_time->tv_nsec)
1036 ts = timespec_sub(*end_time, ts);
1038 ts.tv_sec = ts.tv_nsec = 0;
1041 struct compat_timeval rtv;
1043 rtv.tv_sec = ts.tv_sec;
1044 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1046 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1049 struct compat_timespec rts;
1051 rts.tv_sec = ts.tv_sec;
1052 rts.tv_nsec = ts.tv_nsec;
1054 if (!copy_to_user(p, &rts, sizeof(rts)))
1058 * If an application puts its timeval in read-only memory, we
1059 * don't want the Linux-specific update to the timeval to
1060 * cause a fault after the select has completed
1061 * successfully. However, because we're not updating the
1062 * timeval, we can't restart the system call.
1066 if (ret == -ERESTARTNOHAND)
1072 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1073 * 64-bit unsigned longs.
1076 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1077 unsigned long *fdset)
1079 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1083 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1090 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1093 *fdset++ = h << 32 | l;
1096 if (odd && __get_user(*fdset, ufdset))
1099 /* Tricky, must clear full unsigned long in the
1100 * kernel fdset at the end, this makes sure that
1103 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1109 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1110 unsigned long *fdset)
1113 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1124 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1129 if (odd && __put_user(*fdset, ufdset))
1136 * This is a virtual copy of sys_select from fs/select.c and probably
1137 * should be compared to it from time to time
1141 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1142 * like to be certain this leads to no problems. So I return
1143 * EINTR just for safety.
1145 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1146 * I'm trying ERESTARTNOHAND which restart only when you want to.
1148 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1149 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1150 struct timespec *end_time)
1154 int size, max_fds, ret = -EINVAL;
1155 struct fdtable *fdt;
1156 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1161 /* max_fds can increase, so grab it once to avoid race */
1163 fdt = files_fdtable(current->files);
1164 max_fds = fdt->max_fds;
1170 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1171 * since we used fdset we need to allocate memory in units of
1174 size = FDS_BYTES(n);
1176 if (size > sizeof(stack_fds) / 6) {
1177 bits = kmalloc(6 * size, GFP_KERNEL);
1182 fds.in = (unsigned long *) bits;
1183 fds.out = (unsigned long *) (bits + size);
1184 fds.ex = (unsigned long *) (bits + 2*size);
1185 fds.res_in = (unsigned long *) (bits + 3*size);
1186 fds.res_out = (unsigned long *) (bits + 4*size);
1187 fds.res_ex = (unsigned long *) (bits + 5*size);
1189 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1190 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1191 (ret = compat_get_fd_set(n, exp, fds.ex)))
1193 zero_fd_set(n, fds.res_in);
1194 zero_fd_set(n, fds.res_out);
1195 zero_fd_set(n, fds.res_ex);
1197 ret = do_select(n, &fds, end_time);
1202 ret = -ERESTARTNOHAND;
1203 if (signal_pending(current))
1208 if (compat_set_fd_set(n, inp, fds.res_in) ||
1209 compat_set_fd_set(n, outp, fds.res_out) ||
1210 compat_set_fd_set(n, exp, fds.res_ex))
1213 if (bits != stack_fds)
1219 COMPAT_SYSCALL_DEFINE5(select, int, n, compat_ulong_t __user *, inp,
1220 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1221 struct compat_timeval __user *, tvp)
1223 struct timespec end_time, *to = NULL;
1224 struct compat_timeval tv;
1228 if (copy_from_user(&tv, tvp, sizeof(tv)))
1232 if (poll_select_set_timeout(to,
1233 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1234 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1238 ret = compat_core_sys_select(n, inp, outp, exp, to);
1239 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1244 struct compat_sel_arg_struct {
1252 COMPAT_SYSCALL_DEFINE1(old_select, struct compat_sel_arg_struct __user *, arg)
1254 struct compat_sel_arg_struct a;
1256 if (copy_from_user(&a, arg, sizeof(a)))
1258 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1259 compat_ptr(a.exp), compat_ptr(a.tvp));
1262 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1263 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1264 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1265 compat_size_t sigsetsize)
1267 compat_sigset_t ss32;
1268 sigset_t ksigmask, sigsaved;
1269 struct compat_timespec ts;
1270 struct timespec end_time, *to = NULL;
1274 if (copy_from_user(&ts, tsp, sizeof(ts)))
1278 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1283 if (sigsetsize != sizeof(compat_sigset_t))
1285 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1287 sigset_from_compat(&ksigmask, &ss32);
1289 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1290 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1293 ret = compat_core_sys_select(n, inp, outp, exp, to);
1294 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1296 if (ret == -ERESTARTNOHAND) {
1298 * Don't restore the signal mask yet. Let do_signal() deliver
1299 * the signal on the way back to userspace, before the signal
1303 memcpy(¤t->saved_sigmask, &sigsaved,
1305 set_restore_sigmask();
1308 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1313 COMPAT_SYSCALL_DEFINE6(pselect6, int, n, compat_ulong_t __user *, inp,
1314 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1315 struct compat_timespec __user *, tsp, void __user *, sig)
1317 compat_size_t sigsetsize = 0;
1318 compat_uptr_t up = 0;
1321 if (!access_ok(VERIFY_READ, sig,
1322 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1323 __get_user(up, (compat_uptr_t __user *)sig) ||
1324 __get_user(sigsetsize,
1325 (compat_size_t __user *)(sig+sizeof(up))))
1328 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1332 COMPAT_SYSCALL_DEFINE5(ppoll, struct pollfd __user *, ufds,
1333 unsigned int, nfds, struct compat_timespec __user *, tsp,
1334 const compat_sigset_t __user *, sigmask, compat_size_t, sigsetsize)
1336 compat_sigset_t ss32;
1337 sigset_t ksigmask, sigsaved;
1338 struct compat_timespec ts;
1339 struct timespec end_time, *to = NULL;
1343 if (copy_from_user(&ts, tsp, sizeof(ts)))
1347 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1352 if (sigsetsize != sizeof(compat_sigset_t))
1354 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1356 sigset_from_compat(&ksigmask, &ss32);
1358 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1359 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1362 ret = do_sys_poll(ufds, nfds, to);
1364 /* We can restart this syscall, usually */
1365 if (ret == -EINTR) {
1367 * Don't restore the signal mask yet. Let do_signal() deliver
1368 * the signal on the way back to userspace, before the signal
1372 memcpy(¤t->saved_sigmask, &sigsaved,
1374 set_restore_sigmask();
1376 ret = -ERESTARTNOHAND;
1378 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1380 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1385 #ifdef CONFIG_FHANDLE
1387 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1388 * doesn't set the O_LARGEFILE flag.
1390 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1391 struct file_handle __user *, handle, int, flags)
1393 return do_handle_open(mountdirfd, handle, flags);