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);
71 * Not all architectures have sys_utime, so implement this in terms
74 asmlinkage long compat_sys_utime(const char __user *filename,
75 struct compat_utimbuf __user *t)
77 struct timespec tv[2];
80 if (get_user(tv[0].tv_sec, &t->actime) ||
81 get_user(tv[1].tv_sec, &t->modtime))
86 return do_utimes(AT_FDCWD, filename, t ? tv : NULL, 0);
89 asmlinkage long compat_sys_utimensat(unsigned int dfd, const char __user *filename, struct compat_timespec __user *t, int flags)
91 struct timespec tv[2];
94 if (get_compat_timespec(&tv[0], &t[0]) ||
95 get_compat_timespec(&tv[1], &t[1]))
98 if (tv[0].tv_nsec == UTIME_OMIT && tv[1].tv_nsec == UTIME_OMIT)
101 return do_utimes(dfd, filename, t ? tv : NULL, flags);
104 asmlinkage long compat_sys_futimesat(unsigned int dfd, const char __user *filename, struct compat_timeval __user *t)
106 struct timespec tv[2];
109 if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
110 get_user(tv[0].tv_nsec, &t[0].tv_usec) ||
111 get_user(tv[1].tv_sec, &t[1].tv_sec) ||
112 get_user(tv[1].tv_nsec, &t[1].tv_usec))
114 if (tv[0].tv_nsec >= 1000000 || tv[0].tv_nsec < 0 ||
115 tv[1].tv_nsec >= 1000000 || tv[1].tv_nsec < 0)
117 tv[0].tv_nsec *= 1000;
118 tv[1].tv_nsec *= 1000;
120 return do_utimes(dfd, filename, t ? tv : NULL, 0);
123 asmlinkage long compat_sys_utimes(const char __user *filename, struct compat_timeval __user *t)
125 return compat_sys_futimesat(AT_FDCWD, filename, t);
128 static int cp_compat_stat(struct kstat *stat, struct compat_stat __user *ubuf)
130 struct compat_stat tmp;
132 if (!old_valid_dev(stat->dev) || !old_valid_dev(stat->rdev))
135 memset(&tmp, 0, sizeof(tmp));
136 tmp.st_dev = old_encode_dev(stat->dev);
137 tmp.st_ino = stat->ino;
138 if (sizeof(tmp.st_ino) < sizeof(stat->ino) && tmp.st_ino != stat->ino)
140 tmp.st_mode = stat->mode;
141 tmp.st_nlink = stat->nlink;
142 if (tmp.st_nlink != stat->nlink)
144 SET_UID(tmp.st_uid, from_kuid_munged(current_user_ns(), stat->uid));
145 SET_GID(tmp.st_gid, from_kgid_munged(current_user_ns(), stat->gid));
146 tmp.st_rdev = old_encode_dev(stat->rdev);
147 if ((u64) stat->size > MAX_NON_LFS)
149 tmp.st_size = stat->size;
150 tmp.st_atime = stat->atime.tv_sec;
151 tmp.st_atime_nsec = stat->atime.tv_nsec;
152 tmp.st_mtime = stat->mtime.tv_sec;
153 tmp.st_mtime_nsec = stat->mtime.tv_nsec;
154 tmp.st_ctime = stat->ctime.tv_sec;
155 tmp.st_ctime_nsec = stat->ctime.tv_nsec;
156 tmp.st_blocks = stat->blocks;
157 tmp.st_blksize = stat->blksize;
158 return copy_to_user(ubuf, &tmp, sizeof(tmp)) ? -EFAULT : 0;
161 asmlinkage long compat_sys_newstat(const char __user * filename,
162 struct compat_stat __user *statbuf)
167 error = vfs_stat(filename, &stat);
170 return cp_compat_stat(&stat, statbuf);
173 asmlinkage long compat_sys_newlstat(const char __user * filename,
174 struct compat_stat __user *statbuf)
179 error = vfs_lstat(filename, &stat);
182 return cp_compat_stat(&stat, statbuf);
185 #ifndef __ARCH_WANT_STAT64
186 asmlinkage long compat_sys_newfstatat(unsigned int dfd,
187 const char __user *filename,
188 struct compat_stat __user *statbuf, int flag)
193 error = vfs_fstatat(dfd, filename, &stat, flag);
196 return cp_compat_stat(&stat, statbuf);
200 asmlinkage long compat_sys_newfstat(unsigned int fd,
201 struct compat_stat __user * statbuf)
204 int error = vfs_fstat(fd, &stat);
207 error = cp_compat_stat(&stat, statbuf);
211 static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
214 if (sizeof ubuf->f_blocks == 4) {
215 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
216 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
218 /* f_files and f_ffree may be -1; it's okay
219 * to stuff that into 32 bits */
220 if (kbuf->f_files != 0xffffffffffffffffULL
221 && (kbuf->f_files & 0xffffffff00000000ULL))
223 if (kbuf->f_ffree != 0xffffffffffffffffULL
224 && (kbuf->f_ffree & 0xffffffff00000000ULL))
227 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
228 __put_user(kbuf->f_type, &ubuf->f_type) ||
229 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
230 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
231 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
232 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
233 __put_user(kbuf->f_files, &ubuf->f_files) ||
234 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
235 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
236 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
237 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
238 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
239 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
240 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
246 * The following statfs calls are copies of code from fs/statfs.c and
247 * should be checked against those from time to time
249 asmlinkage long compat_sys_statfs(const char __user *pathname, struct compat_statfs __user *buf)
252 int error = user_statfs(pathname, &tmp);
254 error = put_compat_statfs(buf, &tmp);
258 asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
261 int error = fd_statfs(fd, &tmp);
263 error = put_compat_statfs(buf, &tmp);
267 static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
269 if (sizeof ubuf->f_blocks == 4) {
270 if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail |
271 kbuf->f_bsize | kbuf->f_frsize) & 0xffffffff00000000ULL)
273 /* f_files and f_ffree may be -1; it's okay
274 * to stuff that into 32 bits */
275 if (kbuf->f_files != 0xffffffffffffffffULL
276 && (kbuf->f_files & 0xffffffff00000000ULL))
278 if (kbuf->f_ffree != 0xffffffffffffffffULL
279 && (kbuf->f_ffree & 0xffffffff00000000ULL))
282 if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
283 __put_user(kbuf->f_type, &ubuf->f_type) ||
284 __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
285 __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
286 __put_user(kbuf->f_bfree, &ubuf->f_bfree) ||
287 __put_user(kbuf->f_bavail, &ubuf->f_bavail) ||
288 __put_user(kbuf->f_files, &ubuf->f_files) ||
289 __put_user(kbuf->f_ffree, &ubuf->f_ffree) ||
290 __put_user(kbuf->f_namelen, &ubuf->f_namelen) ||
291 __put_user(kbuf->f_fsid.val[0], &ubuf->f_fsid.val[0]) ||
292 __put_user(kbuf->f_fsid.val[1], &ubuf->f_fsid.val[1]) ||
293 __put_user(kbuf->f_frsize, &ubuf->f_frsize) ||
294 __put_user(kbuf->f_flags, &ubuf->f_flags) ||
295 __clear_user(ubuf->f_spare, sizeof(ubuf->f_spare)))
300 asmlinkage long compat_sys_statfs64(const char __user *pathname, compat_size_t sz, struct compat_statfs64 __user *buf)
305 if (sz != sizeof(*buf))
308 error = user_statfs(pathname, &tmp);
310 error = put_compat_statfs64(buf, &tmp);
314 asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
319 if (sz != sizeof(*buf))
322 error = fd_statfs(fd, &tmp);
324 error = put_compat_statfs64(buf, &tmp);
329 * This is a copy of sys_ustat, just dealing with a structure layout.
330 * Given how simple this syscall is that apporach is more maintainable
331 * than the various conversion hacks.
333 asmlinkage long compat_sys_ustat(unsigned dev, struct compat_ustat __user *u)
335 struct compat_ustat tmp;
337 int err = vfs_ustat(new_decode_dev(dev), &sbuf);
341 memset(&tmp, 0, sizeof(struct compat_ustat));
342 tmp.f_tfree = sbuf.f_bfree;
343 tmp.f_tinode = sbuf.f_ffree;
344 if (copy_to_user(u, &tmp, sizeof(struct compat_ustat)))
349 static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
351 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
352 __get_user(kfl->l_type, &ufl->l_type) ||
353 __get_user(kfl->l_whence, &ufl->l_whence) ||
354 __get_user(kfl->l_start, &ufl->l_start) ||
355 __get_user(kfl->l_len, &ufl->l_len) ||
356 __get_user(kfl->l_pid, &ufl->l_pid))
361 static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
363 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
364 __put_user(kfl->l_type, &ufl->l_type) ||
365 __put_user(kfl->l_whence, &ufl->l_whence) ||
366 __put_user(kfl->l_start, &ufl->l_start) ||
367 __put_user(kfl->l_len, &ufl->l_len) ||
368 __put_user(kfl->l_pid, &ufl->l_pid))
373 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
374 static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
376 if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
377 __get_user(kfl->l_type, &ufl->l_type) ||
378 __get_user(kfl->l_whence, &ufl->l_whence) ||
379 __get_user(kfl->l_start, &ufl->l_start) ||
380 __get_user(kfl->l_len, &ufl->l_len) ||
381 __get_user(kfl->l_pid, &ufl->l_pid))
387 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
388 static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
390 if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
391 __put_user(kfl->l_type, &ufl->l_type) ||
392 __put_user(kfl->l_whence, &ufl->l_whence) ||
393 __put_user(kfl->l_start, &ufl->l_start) ||
394 __put_user(kfl->l_len, &ufl->l_len) ||
395 __put_user(kfl->l_pid, &ufl->l_pid))
401 asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
412 ret = get_compat_flock(&f, compat_ptr(arg));
417 ret = sys_fcntl(fd, cmd, (unsigned long)&f);
419 if (cmd == F_GETLK && ret == 0) {
420 /* GETLK was successful and we need to return the data...
421 * but it needs to fit in the compat structure.
422 * l_start shouldn't be too big, unless the original
423 * start + end is greater than COMPAT_OFF_T_MAX, in which
424 * case the app was asking for trouble, so we return
425 * -EOVERFLOW in that case.
426 * l_len could be too big, in which case we just truncate it,
427 * and only allow the app to see that part of the conflicting
428 * lock that might make sense to it anyway
431 if (f.l_start > COMPAT_OFF_T_MAX)
433 if (f.l_len > COMPAT_OFF_T_MAX)
434 f.l_len = COMPAT_OFF_T_MAX;
436 ret = put_compat_flock(&f, compat_ptr(arg));
443 ret = get_compat_flock64(&f, compat_ptr(arg));
448 ret = sys_fcntl(fd, (cmd == F_GETLK64) ? F_GETLK :
449 ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
452 if (cmd == F_GETLK64 && ret == 0) {
453 /* need to return lock information - see above for commentary */
454 if (f.l_start > COMPAT_LOFF_T_MAX)
456 if (f.l_len > COMPAT_LOFF_T_MAX)
457 f.l_len = COMPAT_LOFF_T_MAX;
459 ret = put_compat_flock64(&f, compat_ptr(arg));
464 ret = sys_fcntl(fd, cmd, arg);
470 asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
473 if ((cmd == F_GETLK64) || (cmd == F_SETLK64) || (cmd == F_SETLKW64))
475 return compat_sys_fcntl64(fd, cmd, arg);
479 compat_sys_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);
499 compat_sys_io_getevents(aio_context_t ctx_id,
500 unsigned long min_nr,
502 struct io_event __user *events,
503 struct compat_timespec __user *timeout)
507 struct timespec __user *ut = NULL;
510 if (unlikely(!access_ok(VERIFY_WRITE, events,
511 nr * sizeof(struct io_event))))
514 if (get_compat_timespec(&t, timeout))
517 ut = compat_alloc_user_space(sizeof(*ut));
518 if (copy_to_user(ut, &t, sizeof(t)) )
521 ret = sys_io_getevents(ctx_id, min_nr, nr, events, ut);
526 /* A write operation does a read from user space and vice versa */
527 #define vrfy_dir(type) ((type) == READ ? VERIFY_WRITE : VERIFY_READ)
529 ssize_t compat_rw_copy_check_uvector(int type,
530 const struct compat_iovec __user *uvector, unsigned long nr_segs,
531 unsigned long fast_segs, struct iovec *fast_pointer,
532 struct iovec **ret_pointer)
534 compat_ssize_t tot_len;
535 struct iovec *iov = *ret_pointer = fast_pointer;
540 * SuS says "The readv() function *may* fail if the iovcnt argument
541 * was less than or equal to 0, or greater than {IOV_MAX}. Linux has
542 * traditionally returned zero for zero segments, so...
548 if (nr_segs > UIO_MAXIOV || nr_segs < 0)
550 if (nr_segs > fast_segs) {
552 iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
559 if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
563 * Single unix specification:
564 * We should -EINVAL if an element length is not >= 0 and fitting an
567 * In Linux, the total length is limited to MAX_RW_COUNT, there is
568 * no overflow possibility.
572 for (seg = 0; seg < nr_segs; seg++) {
576 if (__get_user(len, &uvector->iov_len) ||
577 __get_user(buf, &uvector->iov_base)) {
581 if (len < 0) /* size_t not fitting in compat_ssize_t .. */
584 !access_ok(vrfy_dir(type), compat_ptr(buf), len)) {
588 if (len > MAX_RW_COUNT - tot_len)
589 len = MAX_RW_COUNT - tot_len;
591 iov->iov_base = compat_ptr(buf);
592 iov->iov_len = (compat_size_t) len;
603 copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
608 for (i = 0; i < nr; ++i) {
609 if (get_user(uptr, ptr32 + i))
611 if (put_user(compat_ptr(uptr), ptr64 + i))
617 #define MAX_AIO_SUBMITS (PAGE_SIZE/sizeof(struct iocb *))
620 compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
622 struct iocb __user * __user *iocb64;
625 if (unlikely(nr < 0))
628 if (nr > MAX_AIO_SUBMITS)
629 nr = MAX_AIO_SUBMITS;
631 iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
632 ret = copy_iocb(nr, iocb, iocb64);
634 ret = do_io_submit(ctx_id, nr, iocb64, 1);
638 struct compat_ncp_mount_data {
639 compat_int_t version;
640 compat_uint_t ncp_fd;
641 __compat_uid_t mounted_uid;
642 compat_pid_t wdog_pid;
643 unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
644 compat_uint_t time_out;
645 compat_uint_t retry_count;
649 compat_mode_t file_mode;
650 compat_mode_t dir_mode;
653 struct compat_ncp_mount_data_v4 {
654 compat_int_t version;
655 compat_ulong_t flags;
656 compat_ulong_t mounted_uid;
657 compat_long_t wdog_pid;
658 compat_uint_t ncp_fd;
659 compat_uint_t time_out;
660 compat_uint_t retry_count;
663 compat_ulong_t file_mode;
664 compat_ulong_t dir_mode;
667 static void *do_ncp_super_data_conv(void *raw_data)
669 int version = *(unsigned int *)raw_data;
672 struct compat_ncp_mount_data *c_n = raw_data;
673 struct ncp_mount_data *n = raw_data;
675 n->dir_mode = c_n->dir_mode;
676 n->file_mode = c_n->file_mode;
679 memmove (n->mounted_vol, c_n->mounted_vol, (sizeof (c_n->mounted_vol) + 3 * sizeof (unsigned int)));
680 n->wdog_pid = c_n->wdog_pid;
681 n->mounted_uid = c_n->mounted_uid;
682 } else if (version == 4) {
683 struct compat_ncp_mount_data_v4 *c_n = raw_data;
684 struct ncp_mount_data_v4 *n = raw_data;
686 n->dir_mode = c_n->dir_mode;
687 n->file_mode = c_n->file_mode;
690 n->retry_count = c_n->retry_count;
691 n->time_out = c_n->time_out;
692 n->ncp_fd = c_n->ncp_fd;
693 n->wdog_pid = c_n->wdog_pid;
694 n->mounted_uid = c_n->mounted_uid;
695 n->flags = c_n->flags;
696 } else if (version != 5) {
704 struct compat_nfs_string {
709 static inline void compat_nfs_string(struct nfs_string *dst,
710 struct compat_nfs_string *src)
712 dst->data = compat_ptr(src->data);
716 struct compat_nfs4_mount_data_v1 {
717 compat_int_t version;
722 compat_int_t retrans;
723 compat_int_t acregmin;
724 compat_int_t acregmax;
725 compat_int_t acdirmin;
726 compat_int_t acdirmax;
727 struct compat_nfs_string client_addr;
728 struct compat_nfs_string mnt_path;
729 struct compat_nfs_string hostname;
730 compat_uint_t host_addrlen;
731 compat_uptr_t host_addr;
733 compat_int_t auth_flavourlen;
734 compat_uptr_t auth_flavours;
737 static int do_nfs4_super_data_conv(void *raw_data)
739 int version = *(compat_uint_t *) raw_data;
742 struct compat_nfs4_mount_data_v1 *raw = raw_data;
743 struct nfs4_mount_data *real = raw_data;
745 /* copy the fields backwards */
746 real->auth_flavours = compat_ptr(raw->auth_flavours);
747 real->auth_flavourlen = raw->auth_flavourlen;
748 real->proto = raw->proto;
749 real->host_addr = compat_ptr(raw->host_addr);
750 real->host_addrlen = raw->host_addrlen;
751 compat_nfs_string(&real->hostname, &raw->hostname);
752 compat_nfs_string(&real->mnt_path, &raw->mnt_path);
753 compat_nfs_string(&real->client_addr, &raw->client_addr);
754 real->acdirmax = raw->acdirmax;
755 real->acdirmin = raw->acdirmin;
756 real->acregmax = raw->acregmax;
757 real->acregmin = raw->acregmin;
758 real->retrans = raw->retrans;
759 real->timeo = raw->timeo;
760 real->wsize = raw->wsize;
761 real->rsize = raw->rsize;
762 real->flags = raw->flags;
763 real->version = raw->version;
769 #define NCPFS_NAME "ncpfs"
770 #define NFS4_NAME "nfs4"
772 asmlinkage long compat_sys_mount(const char __user * dev_name,
773 const char __user * dir_name,
774 const char __user * type, unsigned long flags,
775 const void __user * data)
778 unsigned long data_page;
780 struct filename *dir;
783 retval = copy_mount_string(type, &kernel_type);
787 dir = getname(dir_name);
788 retval = PTR_ERR(dir);
792 retval = copy_mount_string(dev_name, &kernel_dev);
796 retval = copy_mount_options(data, &data_page);
802 if (kernel_type && data_page) {
803 if (!strcmp(kernel_type, NCPFS_NAME)) {
804 do_ncp_super_data_conv((void *)data_page);
805 } else if (!strcmp(kernel_type, NFS4_NAME)) {
806 if (do_nfs4_super_data_conv((void *) data_page))
811 retval = do_mount(kernel_dev, dir->name, kernel_type,
812 flags, (void*)data_page);
815 free_page(data_page);
826 struct compat_old_linux_dirent {
827 compat_ulong_t d_ino;
828 compat_ulong_t d_offset;
829 unsigned short d_namlen;
833 struct compat_readdir_callback {
834 struct compat_old_linux_dirent __user *dirent;
838 static int compat_fillonedir(void *__buf, const char *name, int namlen,
839 loff_t offset, u64 ino, unsigned int d_type)
841 struct compat_readdir_callback *buf = __buf;
842 struct compat_old_linux_dirent __user *dirent;
843 compat_ulong_t d_ino;
848 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
849 buf->result = -EOVERFLOW;
853 dirent = buf->dirent;
854 if (!access_ok(VERIFY_WRITE, dirent,
855 (unsigned long)(dirent->d_name + namlen + 1) -
856 (unsigned long)dirent))
858 if ( __put_user(d_ino, &dirent->d_ino) ||
859 __put_user(offset, &dirent->d_offset) ||
860 __put_user(namlen, &dirent->d_namlen) ||
861 __copy_to_user(dirent->d_name, name, namlen) ||
862 __put_user(0, dirent->d_name + namlen))
866 buf->result = -EFAULT;
870 asmlinkage long compat_sys_old_readdir(unsigned int fd,
871 struct compat_old_linux_dirent __user *dirent, unsigned int count)
874 struct fd f = fdget(fd);
875 struct compat_readdir_callback buf;
883 error = vfs_readdir(f.file, compat_fillonedir, &buf);
891 struct compat_linux_dirent {
892 compat_ulong_t d_ino;
893 compat_ulong_t d_off;
894 unsigned short d_reclen;
898 struct compat_getdents_callback {
899 struct compat_linux_dirent __user *current_dir;
900 struct compat_linux_dirent __user *previous;
905 static int compat_filldir(void *__buf, const char *name, int namlen,
906 loff_t offset, u64 ino, unsigned int d_type)
908 struct compat_linux_dirent __user * dirent;
909 struct compat_getdents_callback *buf = __buf;
910 compat_ulong_t d_ino;
911 int reclen = ALIGN(offsetof(struct compat_linux_dirent, d_name) +
912 namlen + 2, sizeof(compat_long_t));
914 buf->error = -EINVAL; /* only used if we fail.. */
915 if (reclen > buf->count)
918 if (sizeof(d_ino) < sizeof(ino) && d_ino != ino) {
919 buf->error = -EOVERFLOW;
922 dirent = buf->previous;
924 if (__put_user(offset, &dirent->d_off))
927 dirent = buf->current_dir;
928 if (__put_user(d_ino, &dirent->d_ino))
930 if (__put_user(reclen, &dirent->d_reclen))
932 if (copy_to_user(dirent->d_name, name, namlen))
934 if (__put_user(0, dirent->d_name + namlen))
936 if (__put_user(d_type, (char __user *) dirent + reclen - 1))
938 buf->previous = dirent;
939 dirent = (void __user *)dirent + reclen;
940 buf->current_dir = dirent;
941 buf->count -= reclen;
944 buf->error = -EFAULT;
948 asmlinkage long compat_sys_getdents(unsigned int fd,
949 struct compat_linux_dirent __user *dirent, unsigned int count)
952 struct compat_linux_dirent __user * lastdirent;
953 struct compat_getdents_callback buf;
956 if (!access_ok(VERIFY_WRITE, dirent, count))
963 buf.current_dir = dirent;
968 error = vfs_readdir(f.file, compat_filldir, &buf);
971 lastdirent = buf.previous;
973 if (put_user(f.file->f_pos, &lastdirent->d_off))
976 error = count - buf.count;
982 #ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
984 struct compat_getdents_callback64 {
985 struct linux_dirent64 __user *current_dir;
986 struct linux_dirent64 __user *previous;
991 static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
992 u64 ino, unsigned int d_type)
994 struct linux_dirent64 __user *dirent;
995 struct compat_getdents_callback64 *buf = __buf;
996 int reclen = ALIGN(offsetof(struct linux_dirent64, d_name) + namlen + 1,
1000 buf->error = -EINVAL; /* only used if we fail.. */
1001 if (reclen > buf->count)
1003 dirent = buf->previous;
1006 if (__put_user_unaligned(offset, &dirent->d_off))
1009 dirent = buf->current_dir;
1010 if (__put_user_unaligned(ino, &dirent->d_ino))
1013 if (__put_user_unaligned(off, &dirent->d_off))
1015 if (__put_user(reclen, &dirent->d_reclen))
1017 if (__put_user(d_type, &dirent->d_type))
1019 if (copy_to_user(dirent->d_name, name, namlen))
1021 if (__put_user(0, dirent->d_name + namlen))
1023 buf->previous = dirent;
1024 dirent = (void __user *)dirent + reclen;
1025 buf->current_dir = dirent;
1026 buf->count -= reclen;
1029 buf->error = -EFAULT;
1033 asmlinkage long compat_sys_getdents64(unsigned int fd,
1034 struct linux_dirent64 __user * dirent, unsigned int count)
1037 struct linux_dirent64 __user * lastdirent;
1038 struct compat_getdents_callback64 buf;
1041 if (!access_ok(VERIFY_WRITE, dirent, count))
1048 buf.current_dir = dirent;
1049 buf.previous = NULL;
1053 error = vfs_readdir(f.file, compat_filldir64, &buf);
1056 lastdirent = buf.previous;
1058 typeof(lastdirent->d_off) d_off = f.file->f_pos;
1059 if (__put_user_unaligned(d_off, &lastdirent->d_off))
1062 error = count - buf.count;
1067 #endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
1070 * Exactly like fs/open.c:sys_open(), except that it doesn't set the
1073 COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1075 return do_sys_open(AT_FDCWD, filename, flags, mode);
1079 * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
1082 COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode)
1084 return do_sys_open(dfd, filename, flags, mode);
1087 #define __COMPAT_NFDBITS (8 * sizeof(compat_ulong_t))
1089 static int poll_select_copy_remaining(struct timespec *end_time, void __user *p,
1090 int timeval, int ret)
1097 if (current->personality & STICKY_TIMEOUTS)
1100 /* No update for zero timeout */
1101 if (!end_time->tv_sec && !end_time->tv_nsec)
1105 ts = timespec_sub(*end_time, ts);
1107 ts.tv_sec = ts.tv_nsec = 0;
1110 struct compat_timeval rtv;
1112 rtv.tv_sec = ts.tv_sec;
1113 rtv.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
1115 if (!copy_to_user(p, &rtv, sizeof(rtv)))
1118 struct compat_timespec rts;
1120 rts.tv_sec = ts.tv_sec;
1121 rts.tv_nsec = ts.tv_nsec;
1123 if (!copy_to_user(p, &rts, sizeof(rts)))
1127 * If an application puts its timeval in read-only memory, we
1128 * don't want the Linux-specific update to the timeval to
1129 * cause a fault after the select has completed
1130 * successfully. However, because we're not updating the
1131 * timeval, we can't restart the system call.
1135 if (ret == -ERESTARTNOHAND)
1141 * Ooo, nasty. We need here to frob 32-bit unsigned longs to
1142 * 64-bit unsigned longs.
1145 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1146 unsigned long *fdset)
1148 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1152 if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
1159 if (__get_user(l, ufdset) || __get_user(h, ufdset+1))
1162 *fdset++ = h << 32 | l;
1165 if (odd && __get_user(*fdset, ufdset))
1168 /* Tricky, must clear full unsigned long in the
1169 * kernel fdset at the end, this makes sure that
1172 memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
1178 int compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
1179 unsigned long *fdset)
1182 nr = DIV_ROUND_UP(nr, __COMPAT_NFDBITS);
1193 if (__put_user(l, ufdset) || __put_user(h, ufdset+1))
1198 if (odd && __put_user(*fdset, ufdset))
1205 * This is a virtual copy of sys_select from fs/select.c and probably
1206 * should be compared to it from time to time
1210 * We can actually return ERESTARTSYS instead of EINTR, but I'd
1211 * like to be certain this leads to no problems. So I return
1212 * EINTR just for safety.
1214 * Update: ERESTARTSYS breaks at least the xview clock binary, so
1215 * I'm trying ERESTARTNOHAND which restart only when you want to.
1217 int compat_core_sys_select(int n, compat_ulong_t __user *inp,
1218 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1219 struct timespec *end_time)
1223 int size, max_fds, ret = -EINVAL;
1224 struct fdtable *fdt;
1225 long stack_fds[SELECT_STACK_ALLOC/sizeof(long)];
1230 /* max_fds can increase, so grab it once to avoid race */
1232 fdt = files_fdtable(current->files);
1233 max_fds = fdt->max_fds;
1239 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
1240 * since we used fdset we need to allocate memory in units of
1243 size = FDS_BYTES(n);
1245 if (size > sizeof(stack_fds) / 6) {
1246 bits = kmalloc(6 * size, GFP_KERNEL);
1251 fds.in = (unsigned long *) bits;
1252 fds.out = (unsigned long *) (bits + size);
1253 fds.ex = (unsigned long *) (bits + 2*size);
1254 fds.res_in = (unsigned long *) (bits + 3*size);
1255 fds.res_out = (unsigned long *) (bits + 4*size);
1256 fds.res_ex = (unsigned long *) (bits + 5*size);
1258 if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
1259 (ret = compat_get_fd_set(n, outp, fds.out)) ||
1260 (ret = compat_get_fd_set(n, exp, fds.ex)))
1262 zero_fd_set(n, fds.res_in);
1263 zero_fd_set(n, fds.res_out);
1264 zero_fd_set(n, fds.res_ex);
1266 ret = do_select(n, &fds, end_time);
1271 ret = -ERESTARTNOHAND;
1272 if (signal_pending(current))
1277 if (compat_set_fd_set(n, inp, fds.res_in) ||
1278 compat_set_fd_set(n, outp, fds.res_out) ||
1279 compat_set_fd_set(n, exp, fds.res_ex))
1282 if (bits != stack_fds)
1288 asmlinkage long compat_sys_select(int n, compat_ulong_t __user *inp,
1289 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1290 struct compat_timeval __user *tvp)
1292 struct timespec end_time, *to = NULL;
1293 struct compat_timeval tv;
1297 if (copy_from_user(&tv, tvp, sizeof(tv)))
1301 if (poll_select_set_timeout(to,
1302 tv.tv_sec + (tv.tv_usec / USEC_PER_SEC),
1303 (tv.tv_usec % USEC_PER_SEC) * NSEC_PER_USEC))
1307 ret = compat_core_sys_select(n, inp, outp, exp, to);
1308 ret = poll_select_copy_remaining(&end_time, tvp, 1, ret);
1313 struct compat_sel_arg_struct {
1321 asmlinkage long compat_sys_old_select(struct compat_sel_arg_struct __user *arg)
1323 struct compat_sel_arg_struct a;
1325 if (copy_from_user(&a, arg, sizeof(a)))
1327 return compat_sys_select(a.n, compat_ptr(a.inp), compat_ptr(a.outp),
1328 compat_ptr(a.exp), compat_ptr(a.tvp));
1331 static long do_compat_pselect(int n, compat_ulong_t __user *inp,
1332 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1333 struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
1334 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 = compat_core_sys_select(n, inp, outp, exp, to);
1363 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1365 if (ret == -ERESTARTNOHAND) {
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();
1377 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1382 asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
1383 compat_ulong_t __user *outp, compat_ulong_t __user *exp,
1384 struct compat_timespec __user *tsp, void __user *sig)
1386 compat_size_t sigsetsize = 0;
1387 compat_uptr_t up = 0;
1390 if (!access_ok(VERIFY_READ, sig,
1391 sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
1392 __get_user(up, (compat_uptr_t __user *)sig) ||
1393 __get_user(sigsetsize,
1394 (compat_size_t __user *)(sig+sizeof(up))))
1397 return do_compat_pselect(n, inp, outp, exp, tsp, compat_ptr(up),
1401 asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
1402 unsigned int nfds, struct compat_timespec __user *tsp,
1403 const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
1405 compat_sigset_t ss32;
1406 sigset_t ksigmask, sigsaved;
1407 struct compat_timespec ts;
1408 struct timespec end_time, *to = NULL;
1412 if (copy_from_user(&ts, tsp, sizeof(ts)))
1416 if (poll_select_set_timeout(to, ts.tv_sec, ts.tv_nsec))
1421 if (sigsetsize != sizeof(compat_sigset_t))
1423 if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
1425 sigset_from_compat(&ksigmask, &ss32);
1427 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1428 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1431 ret = do_sys_poll(ufds, nfds, to);
1433 /* We can restart this syscall, usually */
1434 if (ret == -EINTR) {
1436 * Don't restore the signal mask yet. Let do_signal() deliver
1437 * the signal on the way back to userspace, before the signal
1441 memcpy(¤t->saved_sigmask, &sigsaved,
1443 set_restore_sigmask();
1445 ret = -ERESTARTNOHAND;
1447 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1449 ret = poll_select_copy_remaining(&end_time, tsp, 0, ret);
1454 #ifdef CONFIG_FHANDLE
1456 * Exactly like fs/open.c:sys_open_by_handle_at(), except that it
1457 * doesn't set the O_LARGEFILE flag.
1459 COMPAT_SYSCALL_DEFINE3(open_by_handle_at, int, mountdirfd,
1460 struct file_handle __user *, handle, int, flags)
1462 return do_handle_open(mountdirfd, handle, flags);