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 <asm/uaccess.h>
53 #include <asm/mmu_context.h>
54 #include <asm/ioctls.h>
59 int compat_printk(const char *fmt, ...)
66 ret = vprintk(fmt, ap);
72 * Not all architectures have sys_utime, so implement this in terms
75 COMPAT_SYSCALL_DEFINE2(utime, const char __user *, filename,
76 struct compat_utimbuf __user *, t)
78 struct timespec tv[2];
81 if (get_user(tv[0].tv_sec, &t->actime) ||
82 get_user(tv[1].tv_sec, &t->modtime))
87 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
90 COMPAT_SYSCALL_DEFINE4(utimensat, unsigned int, dfd, const char __user *, filename, struct compat_timespec __user *, t, int, flags)
92 struct timespec tv[2];
95 if (compat_get_timespec(&tv[0], &t[0]) ||
96 compat_get_timespec(&tv[1], &t[1]))
99 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
102 return do_utimes(dfd, filename, t ? tv : NULL, flags);
105 COMPAT_SYSCALL_DEFINE3(futimesat, unsigned int, dfd, const char __user *, filename, struct compat_timeval __user *, t)
107 struct timespec tv[2];
110 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
111 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
112 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
113 get_user(tv[1].tv_nsec, &t[1].tv_usec))
115 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
116 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
118 tv[0].tv_nsec *= 1000;
119 tv[1].tv_nsec *= 1000;
121 return do_utimes(dfd, filename, t ? tv : NULL, 0);
124 COMPAT_SYSCALL_DEFINE2(utimes, const char __user *, filename, struct compat_timeval __user *, t)
126 return compat_sys_futimesat(AT_FDCWD, filename, t);
129 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
131 struct compat_stat tmp;
133 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
136 memset(&tmp, 0, sizeof(tmp));
137 tmp.st_dev = old_encode_dev(stat->dev);
138 tmp.st_ino = stat->ino;
139 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
141 tmp.st_mode = stat->mode;
142 tmp.st_nlink = stat->nlink;
143 if (tmp.st_nlink != stat->nlink)
145 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
146 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
147 tmp.st_rdev = old_encode_dev(stat->rdev);
148 if ((u64) stat->size > MAX_NON_LFS)
150 tmp.st_size = stat->size;
151 tmp.st_atime = stat->atime.tv_sec;
152 tmp.st_atime_nsec = stat->atime.tv_nsec;
153 tmp.st_mtime = stat->mtime.tv_sec;
154 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
155 tmp.st_ctime = stat->ctime.tv_sec;
156 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
157 tmp.st_blocks = stat->blocks;
158 tmp.st_blksize = stat->blksize;
159 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
162 COMPAT_SYSCALL_DEFINE2(newstat, const char __user *, filename,
163 struct compat_stat __user *, statbuf)
168 error = vfs_stat(filename, &stat);
171 return cp_compat_stat(&stat, statbuf);
174 COMPAT_SYSCALL_DEFINE2(newlstat, const char __user *, filename,
175 struct compat_stat __user *, statbuf)
180 error = vfs_lstat(filename, &stat);
183 return cp_compat_stat(&stat, statbuf);
186 #ifndef __ARCH_WANT_STAT64
187 COMPAT_SYSCALL_DEFINE4(newfstatat, unsigned int, dfd,
188 const char __user *, filename,
189 struct compat_stat __user *, statbuf, int, flag)
194 error = vfs_fstatat(dfd, filename, &stat, flag);
197 return cp_compat_stat(&stat, statbuf);
201 COMPAT_SYSCALL_DEFINE2(newfstat, unsigned int, fd,
202 struct compat_stat __user *, statbuf)
205 int error = vfs_fstat(fd, &stat);
208 error = cp_compat_stat(&stat, statbuf);
212 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
215 if (sizeof ubuf->f_blocks == 4) {
216 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
217 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
219 /* f_files and f_ffree may be -1; it's okay
220 * to stuff that into 32 bits */
221 if (kbuf->f_files != 0xffffffffffffffffULL
222 && (kbuf->f_files & 0xffffffff00000000ULL))
224 if (kbuf->f_ffree != 0xffffffffffffffffULL
225 && (kbuf->f_ffree & 0xffffffff00000000ULL))
228 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
229 __put_user(kbuf->f_type, &ubuf->f_type) ||
230 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
231 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
232 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
233 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
234 __put_user(kbuf->f_files, &ubuf->f_files) ||
235 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
236 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
237 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
238 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
239 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
240 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
241 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
247 * The following statfs calls are copies of code from fs/statfs.c and
248 * should be checked against those from time to time
250 COMPAT_SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct compat_statfs __user *, buf)
253 int error = user_statfs(pathname, &tmp);
255 error = put_compat_statfs(buf, &tmp);
259 COMPAT_SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct compat_statfs __user *, buf)
262 int error = fd_statfs(fd, &tmp);
264 error = put_compat_statfs(buf, &tmp);
268 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
270 if (sizeof ubuf->f_blocks == 4) {
271 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
272 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
274 /* f_files and f_ffree may be -1; it's okay
275 * to stuff that into 32 bits */
276 if (kbuf->f_files != 0xffffffffffffffffULL
277 && (kbuf->f_files & 0xffffffff00000000ULL))
279 if (kbuf->f_ffree != 0xffffffffffffffffULL
280 && (kbuf->f_ffree & 0xffffffff00000000ULL))
283 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
284 __put_user(kbuf->f_type, &ubuf->f_type) ||
285 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
286 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
287 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
288 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
289 __put_user(kbuf->f_files, &ubuf->f_files) ||
290 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
291 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
292 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
293 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
294 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
295 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
296 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
301 COMPAT_SYSCALL_DEFINE3(statfs64, const char __user *, pathname, compat_size_t, sz, struct compat_statfs64 __user *, buf)
306 if (sz != sizeof(*buf))
309 error = user_statfs(pathname, &tmp);
311 error = put_compat_statfs64(buf, &tmp);
315 COMPAT_SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, compat_size_t, sz, struct compat_statfs64 __user *, buf)
320 if (sz != sizeof(*buf))
323 error = fd_statfs(fd, &tmp);
325 error = put_compat_statfs64(buf, &tmp);
330 * This is a copy of sys_ustat, just dealing with a structure layout.
331 * Given how simple this syscall is that apporach is more maintainable
332 * than the various conversion hacks.
334 COMPAT_SYSCALL_DEFINE2(ustat, unsigned, dev, struct compat_ustat __user *, u)
336 struct compat_ustat tmp;
338 int err = vfs_ustat(new_decode_dev(dev), &sbuf);
342 memset(&tmp, 0, sizeof(struct compat_ustat));
343 tmp.f_tfree = sbuf.f_bfree;
344 tmp.f_tinode = sbuf.f_ffree;
345 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
350 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
352 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
353 __get_user(kfl->l_type, &ufl->l_type) ||
354 __get_user(kfl->l_whence, &ufl->l_whence) ||
355 __get_user(kfl->l_start, &ufl->l_start) ||
356 __get_user(kfl->l_len, &ufl->l_len) ||
357 __get_user(kfl->l_pid, &ufl->l_pid))
362 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
364 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
365 __put_user(kfl->l_type, &ufl->l_type) ||
366 __put_user(kfl->l_whence, &ufl->l_whence) ||
367 __put_user(kfl->l_start, &ufl->l_start) ||
368 __put_user(kfl->l_len, &ufl->l_len) ||
369 __put_user(kfl->l_pid, &ufl->l_pid))
374 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
375 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
377 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
378 __get_user(kfl->l_type, &ufl->l_type) ||
379 __get_user(kfl->l_whence, &ufl->l_whence) ||
380 __get_user(kfl->l_start, &ufl->l_start) ||
381 __get_user(kfl->l_len, &ufl->l_len) ||
382 __get_user(kfl->l_pid, &ufl->l_pid))
388 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
389 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
391 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
392 __put_user(kfl->l_type, &ufl->l_type) ||
393 __put_user(kfl->l_whence, &ufl->l_whence) ||
394 __put_user(kfl->l_start, &ufl->l_start) ||
395 __put_user(kfl->l_len, &ufl->l_len) ||
396 __put_user(kfl->l_pid, &ufl->l_pid))
402 COMPAT_SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
413 ret = get_compat_flock(&f, compat_ptr(arg));
418 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
420 if (cmd == F_GETLK && ret == 0) {
421 /* GETLK was successful and we need to return the data...
422 * but it needs to fit in the compat structure.
423 * l_start shouldn't be too big, unless the original
424 * start + end is greater than COMPAT_OFF_T_MAX, in which
425 * case the app was asking for trouble, so we return
426 * -EOVERFLOW in that case.
427 * l_len could be too big, in which case we just truncate it,
428 * and only allow the app to see that part of the conflicting
429 * lock that might make sense to it anyway
432 if (f.l_start > COMPAT_OFF_T_MAX)
434 if (f.l_len > COMPAT_OFF_T_MAX)
435 f.l_len = COMPAT_OFF_T_MAX;
437 ret = put_compat_flock(&f, compat_ptr(arg));
444 ret = get_compat_flock64(&f, compat_ptr(arg));
449 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
450 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
453 if (cmd == F_GETLK64 && ret == 0) {
454 /* need to return lock information - see above for commentary */
455 if (f.l_start > COMPAT_LOFF_T_MAX)
457 if (f.l_len > COMPAT_LOFF_T_MAX)
458 f.l_len = COMPAT_LOFF_T_MAX;
460 ret = put_compat_flock64(&f, compat_ptr(arg));
465 ret = sys_fcntl(fd, cmd, arg);
471 COMPAT_SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd,
474 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
476 return compat_sys_fcntl64(fd, cmd, arg);
479 COMPAT_SYSCALL_DEFINE2(io_setup, unsigned, nr_reqs, u32 __user *, ctx32p)
484 mm_segment_t oldfs = get_fs();
485 if (unlikely(get_user(ctx64, ctx32p)))
489 /* The __user pointer cast is valid because of the set_fs() */
490 ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
492 /* truncating is ok because it's a user address */
494 ret = put_user((u32) ctx64, ctx32p);
498 COMPAT_SYSCALL_DEFINE5(io_getevents, compat_aio_context_t, ctx_id,
499 compat_long_t, min_nr,
501 struct io_event __user *, events,
502 struct compat_timespec __user *, timeout)
505 struct timespec __user *ut = NULL;
508 if (compat_get_timespec(&t, timeout))
511 ut = compat_alloc_user_space(sizeof(*ut));
512 if (copy_to_user(ut, &t, sizeof(t)) )
515 return sys_io_getevents(ctx_id, min_nr, nr, events, ut);
518 /* A write operation does a read from user space and vice versa */
519 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
521 ssize_t compat_rw_copy_check_uvector(int type,
522 const struct compat_iovec __user *uvector, unsigned long nr_segs,
523 unsigned long fast_segs, struct iovec *fast_pointer,
524 struct iovec **ret_pointer)
526 compat_ssize_t tot_len;
527 struct iovec *iov = *ret_pointer = fast_pointer;
532 * SuS says "The readv() function *may* fail if the iovcnt argument
533 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
534 * traditionally returned zero for zero segments, so...
540 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
542 if (nr_segs > fast_segs) {
544 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
551 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
555 * Single unix specification:
556 * We should -EINVAL if an element length is not >= 0 and fitting an
559 * In Linux, the total length is limited to MAX_RW_COUNT, there is
560 * no overflow possibility.
564 for (seg = 0; seg < nr_segs; seg++) {
568 if (__get_user(len, &uvector->iov_len) ||
569 __get_user(buf, &uvector->iov_base)) {
573 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
576 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
580 if (len > MAX_RW_COUNT - tot_len)
581 len = MAX_RW_COUNT - tot_len;
583 iov->iov_base = compat_ptr(buf);
584 iov->iov_len = (compat_size_t) len;
595 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
600 for (i = 0; i < nr; ++i) {
601 if (get_user(uptr, ptr32 + i))
603 if (put_user(compat_ptr(uptr), ptr64 + i))
609 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
611 COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id,
612 int, nr, u32 __user *, iocb)
614 struct iocb __user * __user *iocb64;
617 if (unlikely(nr < 0))
620 if (nr > MAX_AIO_SUBMITS)
621 nr = MAX_AIO_SUBMITS;
623 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
624 ret = copy_iocb(nr, iocb, iocb64);
626 ret = do_io_submit(ctx_id, nr, iocb64, 1);
630 struct compat_ncp_mount_data {
631 compat_int_t version;
632 compat_uint_t ncp_fd;
633 __compat_uid_t mounted_uid;
634 compat_pid_t wdog_pid;
635 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
636 compat_uint_t time_out;
637 compat_uint_t retry_count;
641 compat_mode_t file_mode;
642 compat_mode_t dir_mode;
645 struct compat_ncp_mount_data_v4 {
646 compat_int_t version;
647 compat_ulong_t flags;
648 compat_ulong_t mounted_uid;
649 compat_long_t wdog_pid;
650 compat_uint_t ncp_fd;
651 compat_uint_t time_out;
652 compat_uint_t retry_count;
655 compat_ulong_t file_mode;
656 compat_ulong_t dir_mode;
659 static void *do_ncp_super_data_conv(void *raw_data)
661 int version = *(unsigned int *)raw_data;
664 struct compat_ncp_mount_data *c_n = raw_data;
665 struct ncp_mount_data *n = raw_data;
667 n->dir_mode = c_n->dir_mode;
668 n->file_mode = c_n->file_mode;
671 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
672 n->wdog_pid = c_n->wdog_pid;
673 n->mounted_uid = c_n->mounted_uid;
674 } else if (version == 4) {
675 struct compat_ncp_mount_data_v4 *c_n = raw_data;
676 struct ncp_mount_data_v4 *n = raw_data;
678 n->dir_mode = c_n->dir_mode;
679 n->file_mode = c_n->file_mode;
682 n->retry_count = c_n->retry_count;
683 n->time_out = c_n->time_out;
684 n->ncp_fd = c_n->ncp_fd;
685 n->wdog_pid = c_n->wdog_pid;
686 n->mounted_uid = c_n->mounted_uid;
687 n->flags = c_n->flags;
688 } else if (version != 5) {
696 struct compat_nfs_string {
701 static inline void compat_nfs_string(struct nfs_string *dst,
702 struct compat_nfs_string *src)
704 dst->data = compat_ptr(src->data);
708 struct compat_nfs4_mount_data_v1 {
709 compat_int_t version;
714 compat_int_t retrans;
715 compat_int_t acregmin;
716 compat_int_t acregmax;
717 compat_int_t acdirmin;
718 compat_int_t acdirmax;
719 struct compat_nfs_string client_addr;
720 struct compat_nfs_string mnt_path;
721 struct compat_nfs_string hostname;
722 compat_uint_t host_addrlen;
723 compat_uptr_t host_addr;
725 compat_int_t auth_flavourlen;
726 compat_uptr_t auth_flavours;
729 static int do_nfs4_super_data_conv(void *raw_data)
731 int version = *(compat_uint_t *) raw_data;
734 struct compat_nfs4_mount_data_v1 *raw = raw_data;
735 struct nfs4_mount_data *real = raw_data;
737 /* copy the fields backwards */
738 real->auth_flavours = compat_ptr(raw->auth_flavours);
739 real->auth_flavourlen = raw->auth_flavourlen;
740 real->proto = raw->proto;
741 real->host_addr = compat_ptr(raw->host_addr);
742 real->host_addrlen = raw->host_addrlen;
743 compat_nfs_string(&real->hostname, &raw->hostname);
744 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
745 compat_nfs_string(&real->client_addr, &raw->client_addr);
746 real->acdirmax = raw->acdirmax;
747 real->acdirmin = raw->acdirmin;
748 real->acregmax = raw->acregmax;
749 real->acregmin = raw->acregmin;
750 real->retrans = raw->retrans;
751 real->timeo = raw->timeo;
752 real->wsize = raw->wsize;
753 real->rsize = raw->rsize;
754 real->flags = raw->flags;
755 real->version = raw->version;
761 #define NCPFS_NAME "ncpfs"
762 #define NFS4_NAME "nfs4"
764 COMPAT_SYSCALL_DEFINE5(mount, const char __user *, dev_name,
765 const char __user *, dir_name,
766 const char __user *, type, compat_ulong_t, flags,
767 const void __user *, data)
770 unsigned long data_page;
772 struct filename *dir;
775 retval = copy_mount_string(type, &kernel_type);
779 dir = getname(dir_name);
780 retval = PTR_ERR(dir);
784 retval = copy_mount_string(dev_name, &kernel_dev);
788 retval = copy_mount_options(data, &data_page);
794 if (kernel_type && data_page) {
795 if (!strcmp(kernel_type, NCPFS_NAME)) {
796 do_ncp_super_data_conv((void *)data_page);
797 } else if (!strcmp(kernel_type, NFS4_NAME)) {
798 if (do_nfs4_super_data_conv((void *) data_page))
803 retval = do_mount(kernel_dev, dir->name, kernel_type,
804 flags, (void*)data_page);
807 free_page(data_page);
818 struct compat_old_linux_dirent {
819 compat_ulong_t d_ino;
820 compat_ulong_t d_offset;
821 unsigned short d_namlen;
825 struct compat_readdir_callback {
826 struct dir_context ctx;
827 struct compat_old_linux_dirent __user *dirent;
831 static int compat_fillonedir(void *__buf, const char *name, int namlen,
832 loff_t offset, u64 ino, unsigned int d_type)
834 struct compat_readdir_callback *buf = __buf;
835 struct compat_old_linux_dirent __user *dirent;
836 compat_ulong_t d_ino;
841 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
842 buf->result = -EOVERFLOW;
846 dirent = buf->dirent;
847 if (!access_ok(VERIFY_WRITE, dirent,
848 (unsigned long)(dirent->d_name + namlen + 1) -
849 (unsigned long)dirent))
851 if ( __put_user(d_ino, &dirent->d_ino) ||
852 __put_user(offset, &dirent->d_offset) ||
853 __put_user(namlen, &dirent->d_namlen) ||
854 __copy_to_user(dirent->d_name, name, namlen) ||
855 __put_user(0, dirent->d_name + namlen))
859 buf->result = -EFAULT;
863 COMPAT_SYSCALL_DEFINE3(old_readdir, unsigned int, fd,
864 struct compat_old_linux_dirent __user *, dirent, unsigned int, count)
867 struct fd f = fdget(fd);
868 struct compat_readdir_callback buf = {
869 .ctx.actor = compat_fillonedir,
876 error = iterate_dir(f.file, &buf.ctx);
884 struct compat_linux_dirent {
885 compat_ulong_t d_ino;
886 compat_ulong_t d_off;
887 unsigned short d_reclen;
891 struct compat_getdents_callback {
892 struct dir_context ctx;
893 struct compat_linux_dirent __user *current_dir;
894 struct compat_linux_dirent __user *previous;
899 static int compat_filldir(void *__buf, const char *name, int namlen,
900 loff_t offset, u64 ino, unsigned int d_type)
902 struct compat_linux_dirent __user * dirent;
903 struct compat_getdents_callback *buf = __buf;
904 compat_ulong_t d_ino;
905 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
906 namlen + 2, sizeof(compat_long_t));
908 buf->error = -EINVAL; /* only used if we fail.. */
909 if (reclen > buf->count)
912 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
913 buf->error = -EOVERFLOW;
916 dirent = buf->previous;
918 if (__put_user(offset, &dirent->d_off))
921 dirent = buf->current_dir;
922 if (__put_user(d_ino, &dirent->d_ino))
924 if (__put_user(reclen, &dirent->d_reclen))
926 if (copy_to_user(dirent->d_name, name, namlen))
928 if (__put_user(0, dirent->d_name + namlen))
930 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
932 buf->previous = dirent;
933 dirent = (void __user *)dirent + reclen;
934 buf->current_dir = dirent;
935 buf->count -= reclen;
938 buf->error = -EFAULT;
942 COMPAT_SYSCALL_DEFINE3(getdents, unsigned int, fd,
943 struct compat_linux_dirent __user *, dirent, unsigned int, count)
946 struct compat_linux_dirent __user * lastdirent;
947 struct compat_getdents_callback buf = {
948 .ctx.actor = compat_filldir,
949 .current_dir = dirent,
954 if (!access_ok(VERIFY_WRITE, dirent, count))
961 error = iterate_dir(f.file, &buf.ctx);
964 lastdirent = buf.previous;
966 if (put_user(buf.ctx.pos, &lastdirent->d_off))
969 error = count - buf.count;
975 #ifdef __ARCH_WANT_COMPAT_SYS_GETDENTS64
977 struct compat_getdents_callback64 {
978 struct dir_context ctx;
979 struct linux_dirent64 __user *current_dir;
980 struct linux_dirent64 __user *previous;
985 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
986 u64 ino, unsigned int d_type)
988 struct linux_dirent64 __user *dirent;
989 struct compat_getdents_callback64 *buf = __buf;
990 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
994 buf->error = -EINVAL; /* only used if we fail.. */
995 if (reclen > buf->count)
997 dirent = buf->previous;
1000 if (__put_user_unaligned(offset, &dirent->d_off))
1003 dirent = buf->current_dir;
1004 if (__put_user_unaligned(ino, &dirent->d_ino))
1007 if (__put_user_unaligned(off, &dirent->d_off))
1009 if (__put_user(reclen, &dirent->d_reclen))
1011 if (__put_user(d_type, &dirent->d_type))
1013 if (copy_to_user(dirent->d_name, name, namlen))
1015 if (__put_user(0, dirent->d_name + namlen))
1017 buf->previous = dirent;
1018 dirent = (void __user *)dirent + reclen;
1019 buf->current_dir = dirent;
1020 buf->count -= reclen;
1023 buf->error = -EFAULT;
1027 COMPAT_SYSCALL_DEFINE3(getdents64, unsigned int, fd,
1028 struct linux_dirent64 __user *, dirent, unsigned int, count)
1031 struct linux_dirent64 __user * lastdirent;
1032 struct compat_getdents_callback64 buf = {
1033 .ctx.actor = compat_filldir64,
1034 .current_dir = dirent,
1039 if (!access_ok(VERIFY_WRITE, dirent, count))
1046 error = iterate_dir(f.file, &buf.ctx);
1049 lastdirent = buf.previous;
1051 typeof(lastdirent->d_off) d_off = buf.ctx.pos;
1052 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1055 error = count - buf.count;
1060 #endif /* __ARCH_WANT_COMPAT_SYS_GETDENTS64 */
1063 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1066 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1068 return do_sys_open(AT_FDCWD, filename, flags, mode);
1072 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1075 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1077 return do_sys_open(dfd, filename, flags, mode);
1080 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1082 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1083 int timeval, int ret)
1090 if (current->personality & STICKY_TIMEOUTS)
1093 /* No update for zero timeout */
1094 if (!end_time->tv_sec && !end_time->tv_nsec)
1098 ts = timespec_sub(*end_time, ts);
1100 ts.tv_sec = ts.tv_nsec = 0;
1103 struct compat_timeval rtv;
1105 rtv.tv_sec = ts.tv_sec;
1106 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1108 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1111 struct compat_timespec rts;
1113 rts.tv_sec = ts.tv_sec;
1114 rts.tv_nsec = ts.tv_nsec;
1116 if (!copy_to_user(p, &rts, sizeof(rts)))
1120 * If an application puts its timeval in read-only memory, we
1121 * don't want the Linux-specific update to the timeval to
1122 * cause a fault after the select has completed
1123 * successfully. However, because we're not updating the
1124 * timeval, we can't restart the system call.
1128 if (ret == -ERESTARTNOHAND)
1134 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1135 * 64-bit unsigned longs.
1138 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1139 unsigned long *fdset)
1141 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1145 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1152 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1155 *fdset++ = h << 32 | l;
1158 if (odd && __get_user(*fdset, ufdset))
1161 /* Tricky, must clear full unsigned long in the
1162 * kernel fdset at the end, this makes sure that
1165 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1171 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1172 unsigned long *fdset)
1175 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1186 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1191 if (odd && __put_user(*fdset, ufdset))
1198 * This is a virtual copy of sys_select from fs/select.c and probably
1199 * should be compared to it from time to time
1203 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1204 * like to be certain this leads to no problems. So I return
1205 * EINTR just for safety.
1207 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1208 * I'm trying ERESTARTNOHAND which restart only when you want to.
1210 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1211 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1212 struct timespec *end_time)
1216 int size, max_fds, ret = -EINVAL;
1217 struct fdtable *fdt;
1218 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1223 /* max_fds can increase, so grab it once to avoid race */
1225 fdt = files_fdtable(current->files);
1226 max_fds = fdt->max_fds;
1232 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1233 * since we used fdset we need to allocate memory in units of
1236 size = FDS_BYTES(n);
1238 if (size > sizeof(stack_fds) / 6) {
1239 bits = kmalloc(6 * size, GFP_KERNEL);
1244 fds.in = (unsigned long *) bits;
1245 fds.out = (unsigned long *) (bits + size);
1246 fds.ex = (unsigned long *) (bits + 2*size);
1247 fds.res_in = (unsigned long *) (bits + 3*size);
1248 fds.res_out = (unsigned long *) (bits + 4*size);
1249 fds.res_ex = (unsigned long *) (bits + 5*size);
1251 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1252 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1253 (ret = compat_get_fd_set(n, exp, fds.ex)))
1255 zero_fd_set(n, fds.res_in);
1256 zero_fd_set(n, fds.res_out);
1257 zero_fd_set(n, fds.res_ex);
1259 ret = do_select(n, &fds, end_time);
1264 ret = -ERESTARTNOHAND;
1265 if (signal_pending(current))
1270 if (compat_set_fd_set(n, inp, fds.res_in) ||
1271 compat_set_fd_set(n, outp, fds.res_out) ||
1272 compat_set_fd_set(n, exp, fds.res_ex))
1275 if (bits != stack_fds)
1281 COMPAT_SYSCALL_DEFINE5(select, int, n, compat_ulong_t __user *, inp,
1282 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1283 struct compat_timeval __user *, tvp)
1285 struct timespec end_time, *to = NULL;
1286 struct compat_timeval tv;
1290 if (copy_from_user(&tv, tvp, sizeof(tv)))
1294 if (poll_select_set_timeout(to,
1295 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1296 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1300 ret = compat_core_sys_select(n, inp, outp, exp, to);
1301 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1306 struct compat_sel_arg_struct {
1314 COMPAT_SYSCALL_DEFINE1(old_select, struct compat_sel_arg_struct __user *, arg)
1316 struct compat_sel_arg_struct a;
1318 if (copy_from_user(&a, arg, sizeof(a)))
1320 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1321 compat_ptr(a.exp), compat_ptr(a.tvp));
1324 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1325 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1326 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1327 compat_size_t sigsetsize)
1329 compat_sigset_t ss32;
1330 sigset_t ksigmask, sigsaved;
1331 struct compat_timespec ts;
1332 struct timespec end_time, *to = NULL;
1336 if (copy_from_user(&ts, tsp, sizeof(ts)))
1340 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1345 if (sigsetsize != sizeof(compat_sigset_t))
1347 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1349 sigset_from_compat(&ksigmask, &ss32);
1351 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1352 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1355 ret = compat_core_sys_select(n, inp, outp, exp, to);
1356 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1358 if (ret == -ERESTARTNOHAND) {
1360 * Don't restore the signal mask yet. Let do_signal() deliver
1361 * the signal on the way back to userspace, before the signal
1365 memcpy(¤t->saved_sigmask, &sigsaved,
1367 set_restore_sigmask();
1370 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1375 COMPAT_SYSCALL_DEFINE6(pselect6, int, n, compat_ulong_t __user *, inp,
1376 compat_ulong_t __user *, outp, compat_ulong_t __user *, exp,
1377 struct compat_timespec __user *, tsp, void __user *, sig)
1379 compat_size_t sigsetsize = 0;
1380 compat_uptr_t up = 0;
1383 if (!access_ok(VERIFY_READ, sig,
1384 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1385 __get_user(up, (compat_uptr_t __user *)sig) ||
1386 __get_user(sigsetsize,
1387 (compat_size_t __user *)(sig+sizeof(up))))
1390 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1394 COMPAT_SYSCALL_DEFINE5(ppoll, struct pollfd __user *, ufds,
1395 unsigned int, nfds, struct compat_timespec __user *, tsp,
1396 const compat_sigset_t __user *, sigmask, compat_size_t, sigsetsize)
1398 compat_sigset_t ss32;
1399 sigset_t ksigmask, sigsaved;
1400 struct compat_timespec ts;
1401 struct timespec end_time, *to = NULL;
1405 if (copy_from_user(&ts, tsp, sizeof(ts)))
1409 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1414 if (sigsetsize != sizeof(compat_sigset_t))
1416 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1418 sigset_from_compat(&ksigmask, &ss32);
1420 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1421 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1424 ret = do_sys_poll(ufds, nfds, to);
1426 /* We can restart this syscall, usually */
1427 if (ret == -EINTR) {
1429 * Don't restore the signal mask yet. Let do_signal() deliver
1430 * the signal on the way back to userspace, before the signal
1434 memcpy(¤t->saved_sigmask, &sigsaved,
1436 set_restore_sigmask();
1438 ret = -ERESTARTNOHAND;
1440 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1442 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1447 #ifdef CONFIG_FHANDLE
1449 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1450 * doesn't set the O_LARGEFILE flag.
1452 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1453 struct file_handle __user *, handle, int, flags)
1455 return do_handle_open(mountdirfd, handle, flags);