patch-2_6_7-vs1_9_1_12
[linux-2.6.git] / fs / compat.c
index b628d2e..689fdd0 100644 (file)
 #include <linux/syscalls.h>
 #include <linux/ctype.h>
 #include <linux/module.h>
+#include <linux/dnotify.h>
+#include <linux/highuid.h>
+#include <linux/sunrpc/svc.h>
+#include <linux/nfsd/nfsd.h>
+#include <linux/nfsd/syscall.h>
+#include <linux/personality.h>
+
 #include <net/sock.h>          /* siocdevprivate_ioctl */
 
 #include <asm/uaccess.h>
+#include <asm/mmu_context.h>
 
 /*
  * Not all architectures have sys_utime, so implement this in terms
  * of sys_utimes.
  */
-asmlinkage long compat_sys_utime(char *filename, struct compat_utimbuf *t)
+asmlinkage long compat_sys_utime(char __user *filename, struct compat_utimbuf __user *t)
 {
        struct timeval tv[2];
 
@@ -56,7 +64,7 @@ asmlinkage long compat_sys_utime(char *filename, struct compat_utimbuf *t)
        return do_utimes(filename, t ? tv : NULL);
 }
 
-asmlinkage long compat_sys_utimes(char *filename, struct compat_timeval *t)
+asmlinkage long compat_sys_utimes(char __user *filename, struct compat_timeval __user *t)
 {
        struct timeval tv[2];
 
@@ -70,8 +78,8 @@ asmlinkage long compat_sys_utimes(char *filename, struct compat_timeval *t)
        return do_utimes(filename, t ? tv : NULL);
 }
 
-asmlinkage long compat_sys_newstat(char * filename,
-               struct compat_stat *statbuf)
+asmlinkage long compat_sys_newstat(char __user * filename,
+               struct compat_stat __user *statbuf)
 {
        struct kstat stat;
        int error = vfs_stat(filename, &stat);
@@ -81,8 +89,8 @@ asmlinkage long compat_sys_newstat(char * filename,
        return error;
 }
 
-asmlinkage long compat_sys_newlstat(char * filename,
-               struct compat_stat *statbuf)
+asmlinkage long compat_sys_newlstat(char __user * filename,
+               struct compat_stat __user *statbuf)
 {
        struct kstat stat;
        int error = vfs_lstat(filename, &stat);
@@ -93,7 +101,7 @@ asmlinkage long compat_sys_newlstat(char * filename,
 }
 
 asmlinkage long compat_sys_newfstat(unsigned int fd,
-               struct compat_stat * statbuf)
+               struct compat_stat __user * statbuf)
 {
        struct kstat stat;
        int error = vfs_fstat(fd, &stat);
@@ -103,7 +111,7 @@ asmlinkage long compat_sys_newfstat(unsigned int fd,
        return error;
 }
 
-static int put_compat_statfs(struct compat_statfs *ubuf, struct kstatfs *kbuf)
+static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *kbuf)
 {
        
        if (sizeof ubuf->f_blocks == 4) {
@@ -137,7 +145,7 @@ static int put_compat_statfs(struct compat_statfs *ubuf, struct kstatfs *kbuf)
  * The following statfs calls are copies of code from fs/open.c and
  * should be checked against those from time to time
  */
-asmlinkage long compat_sys_statfs(const char *path, struct compat_statfs *buf)
+asmlinkage long compat_sys_statfs(const char __user *path, struct compat_statfs __user *buf)
 {
        struct nameidata nd;
        int error;
@@ -153,7 +161,7 @@ asmlinkage long compat_sys_statfs(const char *path, struct compat_statfs *buf)
        return error;
 }
 
-asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs *buf)
+asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user *buf)
 {
        struct file * file;
        struct kstatfs tmp;
@@ -171,7 +179,7 @@ out:
        return error;
 }
 
-static int put_compat_statfs64(struct compat_statfs64 *ubuf, struct kstatfs *kbuf)
+static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstatfs *kbuf)
 {
        if (sizeof ubuf->f_blocks == 4) {
                if ((kbuf->f_blocks | kbuf->f_bfree |
@@ -195,7 +203,7 @@ static int put_compat_statfs64(struct compat_statfs64 *ubuf, struct kstatfs *kbu
        return 0;
 }
 
-asmlinkage long compat_statfs64(const char *path, compat_size_t sz, struct compat_statfs64 *buf)
+asmlinkage long compat_statfs64(const char __user *path, compat_size_t sz, struct compat_statfs64 __user *buf)
 {
        struct nameidata nd;
        int error;
@@ -214,7 +222,7 @@ asmlinkage long compat_statfs64(const char *path, compat_size_t sz, struct compa
        return error;
 }
 
-asmlinkage long compat_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 *buf)
+asmlinkage long compat_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
 {
        struct file * file;
        struct kstatfs tmp;
@@ -282,53 +290,36 @@ static int __init init_sys32_ioctl(void)
 
 __initcall(init_sys32_ioctl);
 
-static struct ioctl_trans *ioctl_free_list;
-
-/* Never free them really. This avoids SMP races. With a Read-Copy-Update
-   enabled kernel we could just use the RCU infrastructure for this. */
-static void free_ioctl(struct ioctl_trans *t) 
-{ 
-       t->cmd = 0; 
-       mb();
-       t->next = ioctl_free_list;
-       ioctl_free_list = t;
-} 
-
-int register_ioctl32_conversion(unsigned int cmd, int (*handler)(unsigned int, unsigned int, unsigned long, struct file *))
+int register_ioctl32_conversion(unsigned int cmd, int (*handler)(unsigned int,
+                               unsigned int, unsigned long, struct file *))
 {
        struct ioctl_trans *t;
+       struct ioctl_trans *new_t;
        unsigned long hash = ioctl32_hash(cmd);
 
+       new_t = kmalloc(sizeof(*new_t), GFP_KERNEL);
+       if (!new_t)
+               return -ENOMEM;
+
        lock_kernel(); 
-       for (t = (struct ioctl_trans *)ioctl32_hash_table[hash];
-            t;
-            t = t->next) { 
+       for (t = ioctl32_hash_table[hash]; t; t = t->next) {
                if (t->cmd == cmd) {
-                       printk("Trying to register duplicated ioctl32 handler %x\n", cmd);
+                       printk(KERN_ERR "Trying to register duplicated ioctl32 "
+                                       "handler %x\n", cmd);
                        unlock_kernel();
+                       kfree(new_t);
                        return -EINVAL; 
                }
-       } 
-
-       if (ioctl_free_list) { 
-               t = ioctl_free_list; 
-               ioctl_free_list = t->next; 
-       } else { 
-               t = kmalloc(sizeof(struct ioctl_trans), GFP_KERNEL); 
-               if (!t) { 
-                       unlock_kernel();
-                       return -ENOMEM;
-               }
        }
-       
-       t->next = NULL;
-       t->cmd = cmd;
-       t->handler = handler; 
-       ioctl32_insert_translation(t);
+       new_t->next = NULL;
+       new_t->cmd = cmd;
+       new_t->handler = handler;
+       ioctl32_insert_translation(new_t);
 
        unlock_kernel();
        return 0;
 }
+EXPORT_SYMBOL(register_ioctl32_conversion);
 
 static inline int builtin_ioctl(struct ioctl_trans *t)
 { 
@@ -347,7 +338,7 @@ int unregister_ioctl32_conversion(unsigned int cmd)
 
        lock_kernel(); 
 
-       t = (struct ioctl_trans *)ioctl32_hash_table[hash];
+       t = ioctl32_hash_table[hash];
        if (!t) { 
                unlock_kernel();
                return -EINVAL;
@@ -358,38 +349,39 @@ int unregister_ioctl32_conversion(unsigned int cmd)
                        printk("%p tried to unregister builtin ioctl %x\n",
                               __builtin_return_address(0), cmd);
                } else { 
-               ioctl32_hash_table[hash] = t->next;
-                       free_ioctl(t); 
+                       ioctl32_hash_table[hash] = t->next;
                        unlock_kernel();
-               return 0;
+                       kfree(t);
+                       return 0;
                }
        } 
        while (t->next) {
-               t1 = (struct ioctl_trans *)(long)t->next;
+               t1 = t->next;
                if (t1->cmd == cmd) { 
                        if (builtin_ioctl(t1)) {
-                               printk("%p tried to unregister builtin ioctl %x\n",
-                                      __builtin_return_address(0), cmd);
+                               printk("%p tried to unregister builtin "
+                                       "ioctl %x\n",
+                                       __builtin_return_address(0), cmd);
                                goto out;
                        } else { 
-                       t->next = t1->next;
-                               free_ioctl(t1); 
+                               t->next = t1->next;
                                unlock_kernel();
-                       return 0;
+                               kfree(t1);
+                               return 0;
                        }
                }
                t = t1;
        }
-       printk(KERN_ERR "Trying to free unknown 32bit ioctl handler %x\n", cmd);
- out:
+       printk(KERN_ERR "Trying to free unknown 32bit ioctl handler %x\n",
+                               cmd);
+out:
        unlock_kernel();
        return -EINVAL;
 }
-
-EXPORT_SYMBOL(register_ioctl32_conversion); 
 EXPORT_SYMBOL(unregister_ioctl32_conversion); 
 
-asmlinkage long compat_sys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
+asmlinkage long compat_sys_ioctl(unsigned int fd, unsigned int cmd,
+                               unsigned long arg)
 {
        struct file * filp;
        int error = -EBADF;
@@ -404,43 +396,53 @@ asmlinkage long compat_sys_ioctl(unsigned int fd, unsigned int cmd, unsigned lon
                goto out;
        }
 
-       t = (struct ioctl_trans *)ioctl32_hash_table [ioctl32_hash (cmd)];
+       lock_kernel();
+
+       t = ioctl32_hash_table[ioctl32_hash (cmd)];
 
        while (t && t->cmd != cmd)
-               t = (struct ioctl_trans *)t->next;
+               t = t->next;
        if (t) {
                if (t->handler) { 
-                       lock_kernel();
                        error = t->handler(fd, cmd, arg, filp);
                        unlock_kernel();
-               } else
+               } else {
+                       unlock_kernel();
                        error = sys_ioctl(fd, cmd, arg);
-       } else if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
-               error = siocdevprivate_ioctl(fd, cmd, arg);
+               }
        } else {
-               static int count;
-               if (++count <= 50) { 
-                       char buf[10];
-                       char *path = (char *)__get_free_page(GFP_KERNEL), *fn = "?"; 
-
-                       /* find the name of the device. */
-                       if (path) {
-                               fn = d_path(filp->f_dentry, filp->f_vfsmnt, 
-                                           path, PAGE_SIZE);
+               unlock_kernel();
+               if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
+                       error = siocdevprivate_ioctl(fd, cmd, arg);
+               } else {
+                       static int count;
+                       if (++count <= 50) {
+                               char buf[10];
+                               char *fn = "?";
+                               char *path;
+
+                               path = (char *)__get_free_page(GFP_KERNEL);
+
+                               /* find the name of the device. */
+                               if (path) {
+                                       fn = d_path(filp->f_dentry,
+                                               filp->f_vfsmnt, path,
+                                               PAGE_SIZE);
+                               }
+
+                               sprintf(buf,"'%c'", (cmd>>24) & 0x3f);
+                               if (!isprint(buf[1]))
+                                   sprintf(buf, "%02x", buf[1]);
+                               printk("ioctl32(%s:%d): Unknown cmd fd(%d) "
+                                       "cmd(%08x){%s} arg(%08x) on %s\n",
+                                       current->comm, current->pid,
+                                       (int)fd, (unsigned int)cmd, buf,
+                                       (unsigned int)arg, fn);
+                               if (path)
+                                       free_page((unsigned long)path);
                        }
-
-                       sprintf(buf,"'%c'", (cmd>>24) & 0x3f); 
-                       if (!isprint(buf[1]))
-                           sprintf(buf, "%02x", buf[1]);
-                       printk("ioctl32(%s:%d): Unknown cmd fd(%d) "
-                              "cmd(%08x){%s} arg(%08x) on %s\n",
-                              current->comm, current->pid,
-                              (int)fd, (unsigned int)cmd, buf, (unsigned int)arg,
-                              fn);
-                       if (path) 
-                               free_page((unsigned long)path); 
+                       error = -EINVAL;
                }
-               error = -EINVAL;
        }
 out:
        fput(filp);
@@ -448,7 +450,7 @@ out2:
        return error;
 }
 
-static int get_compat_flock(struct flock *kfl, struct compat_flock *ufl)
+static int get_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
 {
        if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
            __get_user(kfl->l_type, &ufl->l_type) ||
@@ -460,7 +462,7 @@ static int get_compat_flock(struct flock *kfl, struct compat_flock *ufl)
        return 0;
 }
 
-static int put_compat_flock(struct flock *kfl, struct compat_flock *ufl)
+static int put_compat_flock(struct flock *kfl, struct compat_flock __user *ufl)
 {
        if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
            __put_user(kfl->l_type, &ufl->l_type) ||
@@ -473,7 +475,7 @@ static int put_compat_flock(struct flock *kfl, struct compat_flock *ufl)
 }
 
 #ifndef HAVE_ARCH_GET_COMPAT_FLOCK64
-static int get_compat_flock64(struct flock *kfl, struct compat_flock64 *ufl)
+static int get_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
 {
        if (!access_ok(VERIFY_READ, ufl, sizeof(*ufl)) ||
            __get_user(kfl->l_type, &ufl->l_type) ||
@@ -487,7 +489,7 @@ static int get_compat_flock64(struct flock *kfl, struct compat_flock64 *ufl)
 #endif
 
 #ifndef HAVE_ARCH_PUT_COMPAT_FLOCK64
-static int put_compat_flock64(struct flock *kfl, struct compat_flock64 *ufl)
+static int put_compat_flock64(struct flock *kfl, struct compat_flock64 __user *ufl)
 {
        if (!access_ok(VERIFY_WRITE, ufl, sizeof(*ufl)) ||
            __put_user(kfl->l_type, &ufl->l_type) ||
@@ -564,7 +566,7 @@ asmlinkage long compat_sys_fcntl(unsigned int fd, unsigned int cmd,
 }
 
 asmlinkage long
-compat_sys_io_setup(unsigned nr_reqs, u32 *ctx32p)
+compat_sys_io_setup(unsigned nr_reqs, u32 __user *ctx32p)
 {
        long ret;
        aio_context_t ctx64;
@@ -574,7 +576,8 @@ compat_sys_io_setup(unsigned nr_reqs, u32 *ctx32p)
                return -EFAULT;
 
        set_fs(KERNEL_DS);
-       ret = sys_io_setup(nr_reqs, &ctx64);
+       /* The __user pointer cast is valid because of the set_fs() */
+       ret = sys_io_setup(nr_reqs, (aio_context_t __user *) &ctx64);
        set_fs(oldfs);
        /* truncating is ok because it's a user address */
        if (!ret)
@@ -586,12 +589,12 @@ asmlinkage long
 compat_sys_io_getevents(aio_context_t ctx_id,
                                 unsigned long min_nr,
                                 unsigned long nr,
-                                struct io_event *events,
-                                struct compat_timespec *timeout)
+                                struct io_event __user *events,
+                                struct compat_timespec __user *timeout)
 {
        long ret;
        struct timespec t;
-       struct timespec *ut = NULL;
+       struct timespec __user *ut = NULL;
 
        ret = -EFAULT;
        if (unlikely(!access_ok(VERIFY_WRITE, events, 
@@ -611,7 +614,7 @@ out:
 }
 
 static inline long
-copy_iocb(long nr, u32 *ptr32, u64 *ptr64)
+copy_iocb(long nr, u32 __user *ptr32, struct iocb __user * __user *ptr64)
 {
        compat_uptr_t uptr;
        int i;
@@ -619,7 +622,7 @@ copy_iocb(long nr, u32 *ptr32, u64 *ptr64)
        for (i = 0; i < nr; ++i) {
                if (get_user(uptr, ptr32 + i))
                        return -EFAULT;
-               if (put_user((u64)compat_ptr(uptr), ptr64 + i))
+               if (put_user(compat_ptr(uptr), ptr64 + i))
                        return -EFAULT;
        }
        return 0;
@@ -628,9 +631,9 @@ copy_iocb(long nr, u32 *ptr32, u64 *ptr64)
 #define MAX_AIO_SUBMITS        (PAGE_SIZE/sizeof(struct iocb *))
 
 asmlinkage long
-compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 *iocb)
+compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
 {
-       struct iocb **iocb64; 
+       struct iocb __user * __user *iocb64; 
        long ret;
 
        if (unlikely(nr < 0))
@@ -640,7 +643,7 @@ compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 *iocb)
                nr = MAX_AIO_SUBMITS;
        
        iocb64 = compat_alloc_user_space(nr * sizeof(*iocb64));
-       ret = copy_iocb(nr, iocb, (u64 *) iocb64);
+       ret = copy_iocb(nr, iocb, iocb64);
        if (!ret)
                ret = sys_io_submit(ctx_id, nr, iocb64);
        return ret;
@@ -741,9 +744,9 @@ extern int copy_mount_options (const void __user *, unsigned long *);
 #define SMBFS_NAME      "smbfs"
 #define NCPFS_NAME      "ncpfs"
 
-asmlinkage int compat_sys_mount(char __user * dev_name, char __user * dir_name,
-                               char __user * type, unsigned long flags,
-                               void __user * data)
+asmlinkage long compat_sys_mount(char __user * dev_name, char __user * dir_name,
+                                char __user * type, unsigned long flags,
+                                void __user * data)
 {
        unsigned long type_page;
        unsigned long data_page;
@@ -794,3 +797,840 @@ asmlinkage int compat_sys_mount(char __user * dev_name, char __user * dir_name,
        return retval;
 }
 
+static ssize_t compat_do_readv_writev(int type, struct file *file,
+                              const struct compat_iovec __user *uvector,
+                              unsigned long nr_segs, loff_t *pos)
+{
+       typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *);
+       typedef ssize_t (*iov_fn_t)(struct file *, const struct iovec *, unsigned long, loff_t *);
+
+       compat_ssize_t tot_len;
+       struct iovec iovstack[UIO_FASTIOV];
+       struct iovec *iov=iovstack, *vector;
+       ssize_t ret;
+       int seg;
+       io_fn_t fn;
+       iov_fn_t fnv;
+       struct inode *inode;
+
+       /*
+        * SuS says "The readv() function *may* fail if the iovcnt argument
+        * was less than or equal to 0, or greater than {IOV_MAX}.  Linux has
+        * traditionally returned zero for zero segments, so...
+        */
+       ret = 0;
+       if (nr_segs == 0)
+               goto out;
+
+       /*
+        * First get the "struct iovec" from user memory and
+        * verify all the pointers
+        */
+       ret = -EINVAL;
+       if ((nr_segs > UIO_MAXIOV) || (nr_segs <= 0))
+               goto out;
+       if (!file->f_op)
+               goto out;
+       if (nr_segs > UIO_FASTIOV) {
+               ret = -ENOMEM;
+               iov = kmalloc(nr_segs*sizeof(struct iovec), GFP_KERNEL);
+               if (!iov)
+                       goto out;
+       }
+       ret = -EFAULT;
+       if (verify_area(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
+               goto out;
+
+       /*
+        * Single unix specification:
+        * We should -EINVAL if an element length is not >= 0 and fitting an
+        * ssize_t.  The total length is fitting an ssize_t
+        *
+        * Be careful here because iov_len is a size_t not an ssize_t
+        */
+       tot_len = 0;
+       vector = iov;
+       ret = -EINVAL;
+       for (seg = 0 ; seg < nr_segs; seg++) {
+               compat_ssize_t tmp = tot_len;
+               compat_ssize_t len;
+               compat_uptr_t buf;
+
+               if (__get_user(len, &uvector->iov_len) ||
+                   __get_user(buf, &uvector->iov_base)) {
+                       ret = -EFAULT;
+                       goto out;
+               }
+               if (len < 0)    /* size_t not fitting an compat_ssize_t .. */
+                       goto out;
+               tot_len += len;
+               if (tot_len < tmp) /* maths overflow on the compat_ssize_t */
+                       goto out;
+               vector->iov_base = compat_ptr(buf);
+               vector->iov_len = (compat_size_t) len;
+               uvector++;
+               vector++;
+       }
+       if (tot_len == 0) {
+               ret = 0;
+               goto out;
+       }
+
+       inode = file->f_dentry->d_inode;
+       /* VERIFY_WRITE actually means a read, as we write to user space */
+       ret = locks_verify_area((type == READ
+                                ? FLOCK_VERIFY_READ : FLOCK_VERIFY_WRITE),
+                               inode, file, *pos, tot_len);
+       if (ret)
+               goto out;
+
+       fnv = NULL;
+       if (type == READ) {
+               fn = file->f_op->read;
+               fnv = file->f_op->readv;
+       } else {
+               fn = (io_fn_t)file->f_op->write;
+               fnv = file->f_op->writev;
+       }
+       if (fnv) {
+               ret = fnv(file, iov, nr_segs, pos);
+               goto out;
+       }
+
+       /* Do it by hand, with file-ops */
+       ret = 0;
+       vector = iov;
+       while (nr_segs > 0) {
+               void __user * base;
+               size_t len;
+               ssize_t nr;
+
+               base = vector->iov_base;
+               len = vector->iov_len;
+               vector++;
+               nr_segs--;
+
+               nr = fn(file, base, len, pos);
+
+               if (nr < 0) {
+                       if (!ret) ret = nr;
+                       break;
+               }
+               ret += nr;
+               if (nr != len)
+                       break;
+       }
+out:
+       if (iov != iovstack)
+               kfree(iov);
+       if ((ret + (type == READ)) > 0)
+               dnotify_parent(file->f_dentry,
+                               (type == READ) ? DN_ACCESS : DN_MODIFY);
+       return ret;
+}
+
+asmlinkage ssize_t
+compat_sys_readv(unsigned long fd, const struct compat_iovec __user *vec, unsigned long vlen)
+{
+       struct file *file;
+       ssize_t ret = -EBADF;
+
+       file = fget(fd);
+       if (!file)
+               return -EBADF;
+
+       if (!(file->f_mode & FMODE_READ))
+               goto out;
+
+       ret = -EINVAL;
+       if (!file->f_op || (!file->f_op->readv && !file->f_op->read))
+               goto out;
+
+       ret = compat_do_readv_writev(READ, file, vec, vlen, &file->f_pos);
+
+out:
+       fput(file);
+       return ret;
+}
+
+asmlinkage ssize_t
+compat_sys_writev(unsigned long fd, const struct compat_iovec __user *vec, unsigned long vlen)
+{
+       struct file *file;
+       ssize_t ret = -EBADF;
+
+       file = fget(fd);
+       if (!file)
+               return -EBADF;
+       if (!(file->f_mode & FMODE_WRITE))
+               goto out;
+
+       ret = -EINVAL;
+       if (!file->f_op || (!file->f_op->writev && !file->f_op->write))
+               goto out;
+
+       ret = compat_do_readv_writev(WRITE, file, vec, vlen, &file->f_pos);
+
+out:
+       fput(file);
+       return ret;
+}
+
+/*
+ * compat_count() counts the number of arguments/envelopes. It is basically
+ * a copy of count() from fs/exec.c, except that it works with 32 bit argv
+ * and envp pointers.
+ */
+static int compat_count(compat_uptr_t __user *argv, int max)
+{
+       int i = 0;
+
+       if (argv != NULL) {
+               for (;;) {
+                       compat_uptr_t p;
+
+                       if (get_user(p, argv))
+                               return -EFAULT;
+                       if (!p)
+                               break;
+                       argv++;
+                       if(++i > max)
+                               return -E2BIG;
+               }
+       }
+       return i;
+}
+
+/*
+ * compat_copy_strings() is basically a copy of copy_strings() from fs/exec.c
+ * except that it works with 32 bit argv and envp pointers.
+ */
+static int compat_copy_strings(int argc, compat_uptr_t __user *argv,
+                               struct linux_binprm *bprm)
+{
+       struct page *kmapped_page = NULL;
+       char *kaddr = NULL;
+       int ret;
+
+       while (argc-- > 0) {
+               compat_uptr_t str;
+               int len;
+               unsigned long pos;
+
+               if (get_user(str, argv+argc) ||
+                       !(len = strnlen_user(compat_ptr(str), bprm->p))) {
+                       ret = -EFAULT;
+                       goto out;
+               }
+
+               if (bprm->p < len)  {
+                       ret = -E2BIG;
+                       goto out;
+               }
+
+               bprm->p -= len;
+               /* XXX: add architecture specific overflow check here. */
+               pos = bprm->p;
+
+               while (len > 0) {
+                       int i, new, err;
+                       int offset, bytes_to_copy;
+                       struct page *page;
+
+                       offset = pos % PAGE_SIZE;
+                       i = pos/PAGE_SIZE;
+                       page = bprm->page[i];
+                       new = 0;
+                       if (!page) {
+                               page = alloc_page(GFP_HIGHUSER);
+                               bprm->page[i] = page;
+                               if (!page) {
+                                       ret = -ENOMEM;
+                                       goto out;
+                               }
+                               new = 1;
+                       }
+
+                       if (page != kmapped_page) {
+                               if (kmapped_page)
+                                       kunmap(kmapped_page);
+                               kmapped_page = page;
+                               kaddr = kmap(kmapped_page);
+                       }
+                       if (new && offset)
+                               memset(kaddr, 0, offset);
+                       bytes_to_copy = PAGE_SIZE - offset;
+                       if (bytes_to_copy > len) {
+                               bytes_to_copy = len;
+                               if (new)
+                                       memset(kaddr+offset+len, 0,
+                                               PAGE_SIZE-offset-len);
+                       }
+                       err = copy_from_user(kaddr+offset, compat_ptr(str),
+                                               bytes_to_copy);
+                       if (err) {
+                               ret = -EFAULT;
+                               goto out;
+                       }
+
+                       pos += bytes_to_copy;
+                       str += bytes_to_copy;
+                       len -= bytes_to_copy;
+               }
+       }
+       ret = 0;
+out:
+       if (kmapped_page)
+               kunmap(kmapped_page);
+       return ret;
+}
+
+#ifdef CONFIG_MMU
+
+#define free_arg_pages(bprm) do { } while (0)
+
+#else
+
+static inline void free_arg_pages(struct linux_binprm *bprm)
+{
+       int i;
+
+       for (i = 0; i < MAX_ARG_PAGES; i++) {
+               if (bprm->page[i])
+                       __free_page(bprm->page[i]);
+               bprm->page[i] = NULL;
+       }
+}
+
+#endif /* CONFIG_MMU */
+
+/*
+ * compat_do_execve() is mostly a copy of do_execve(), with the exception
+ * that it processes 32 bit argv and envp pointers.
+ */
+int compat_do_execve(char * filename,
+       compat_uptr_t __user *argv,
+       compat_uptr_t __user *envp,
+       struct pt_regs * regs)
+{
+       struct linux_binprm bprm;
+       struct file *file;
+       int retval;
+       int i;
+
+       sched_balance_exec();
+
+       file = open_exec(filename);
+
+       retval = PTR_ERR(file);
+       if (IS_ERR(file))
+               return retval;
+
+       bprm.p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
+       memset(bprm.page, 0, MAX_ARG_PAGES*sizeof(bprm.page[0]));
+
+       bprm.file = file;
+       bprm.filename = filename;
+       bprm.interp = filename;
+       bprm.sh_bang = 0;
+       bprm.loader = 0;
+       bprm.exec = 0;
+       bprm.security = NULL;
+       bprm.mm = mm_alloc();
+       retval = -ENOMEM;
+       if (!bprm.mm)
+               goto out_file;
+
+       retval = init_new_context(current, bprm.mm);
+       if (retval < 0)
+               goto out_mm;
+
+       bprm.argc = compat_count(argv, bprm.p / sizeof(compat_uptr_t));
+       if ((retval = bprm.argc) < 0)
+               goto out_mm;
+
+       bprm.envc = compat_count(envp, bprm.p / sizeof(compat_uptr_t));
+       if ((retval = bprm.envc) < 0)
+               goto out_mm;
+
+       retval = security_bprm_alloc(&bprm);
+       if (retval)
+               goto out;
+
+       retval = prepare_binprm(&bprm);
+       if (retval < 0)
+               goto out;
+
+       retval = copy_strings_kernel(1, &bprm.filename, &bprm);
+       if (retval < 0)
+               goto out;
+
+       bprm.exec = bprm.p;
+       retval = compat_copy_strings(bprm.envc, envp, &bprm);
+       if (retval < 0)
+               goto out;
+
+       retval = compat_copy_strings(bprm.argc, argv, &bprm);
+       if (retval < 0)
+               goto out;
+
+       retval = search_binary_handler(&bprm,regs);
+       if (retval >= 0) {
+               free_arg_pages(&bprm);
+
+               /* execve success */
+               security_bprm_free(&bprm);
+               return retval;
+       }
+
+out:
+       /* Something went wrong, return the inode and free the argument pages*/
+       for (i = 0 ; i < MAX_ARG_PAGES ; i++) {
+               struct page * page = bprm.page[i];
+               if (page)
+                       __free_page(page);
+       }
+
+       if (bprm.security)
+               security_bprm_free(&bprm);
+
+out_mm:
+       if (bprm.mm)
+               mmdrop(bprm.mm);
+
+out_file:
+       if (bprm.file) {
+               allow_write_access(bprm.file);
+               fput(bprm.file);
+       }
+
+       return retval;
+}
+
+#define __COMPAT_NFDBITS       (8 * sizeof(compat_ulong_t))
+
+#define ROUND_UP(x,y) (((x)+(y)-1)/(y))
+
+/*
+ * Ooo, nasty.  We need here to frob 32-bit unsigned longs to
+ * 64-bit unsigned longs.
+ */
+static inline
+int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
+                       unsigned long *fdset)
+{
+       nr = ROUND_UP(nr, __COMPAT_NFDBITS);
+       if (ufdset) {
+               unsigned long odd;
+
+               if (verify_area(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
+                       return -EFAULT;
+
+               odd = nr & 1UL;
+               nr &= ~1UL;
+               while (nr) {
+                       unsigned long h, l;
+                       __get_user(l, ufdset);
+                       __get_user(h, ufdset+1);
+                       ufdset += 2;
+                       *fdset++ = h << 32 | l;
+                       nr -= 2;
+               }
+               if (odd)
+                       __get_user(*fdset, ufdset);
+       } else {
+               /* Tricky, must clear full unsigned long in the
+                * kernel fdset at the end, this makes sure that
+                * actually happens.
+                */
+               memset(fdset, 0, ((nr + 1) & ~1)*sizeof(compat_ulong_t));
+       }
+       return 0;
+}
+
+static inline
+void compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
+                       unsigned long *fdset)
+{
+       unsigned long odd;
+       nr = ROUND_UP(nr, __COMPAT_NFDBITS);
+
+       if (!ufdset)
+               return;
+
+       odd = nr & 1UL;
+       nr &= ~1UL;
+       while (nr) {
+               unsigned long h, l;
+               l = *fdset++;
+               h = l >> 32;
+               __put_user(l, ufdset);
+               __put_user(h, ufdset+1);
+               ufdset += 2;
+               nr -= 2;
+       }
+       if (odd)
+               __put_user(*fdset, ufdset);
+}
+
+
+/*
+ * This is a virtual copy of sys_select from fs/select.c and probably
+ * should be compared to it from time to time
+ */
+static void *select_bits_alloc(int size)
+{
+       return kmalloc(6 * size, GFP_KERNEL);
+}
+
+static void select_bits_free(void *bits, int size)
+{
+       kfree(bits);
+}
+
+/*
+ * We can actually return ERESTARTSYS instead of EINTR, but I'd
+ * like to be certain this leads to no problems. So I return
+ * EINTR just for safety.
+ *
+ * Update: ERESTARTSYS breaks at least the xview clock binary, so
+ * I'm trying ERESTARTNOHAND which restart only when you want to.
+ */
+#define MAX_SELECT_SECONDS \
+       ((unsigned long) (MAX_SCHEDULE_TIMEOUT / HZ)-1)
+
+asmlinkage long
+compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp,
+               compat_ulong_t __user *exp, struct compat_timeval __user *tvp)
+{
+       fd_set_bits fds;
+       char *bits;
+       long timeout;
+       int ret, size, max_fdset;
+
+       timeout = MAX_SCHEDULE_TIMEOUT;
+       if (tvp) {
+               time_t sec, usec;
+
+               if ((ret = verify_area(VERIFY_READ, tvp, sizeof(*tvp)))
+                   || (ret = __get_user(sec, &tvp->tv_sec))
+                   || (ret = __get_user(usec, &tvp->tv_usec)))
+                       goto out_nofds;
+
+               ret = -EINVAL;
+               if (sec < 0 || usec < 0)
+                       goto out_nofds;
+
+               if ((unsigned long) sec < MAX_SELECT_SECONDS) {
+                       timeout = ROUND_UP(usec, 1000000/HZ);
+                       timeout += sec * (unsigned long) HZ;
+               }
+       }
+
+       ret = -EINVAL;
+       if (n < 0)
+               goto out_nofds;
+
+       /* max_fdset can increase, so grab it once to avoid race */
+       max_fdset = current->files->max_fdset;
+       if (n > max_fdset)
+               n = max_fdset;
+
+       /*
+        * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
+        * since we used fdset we need to allocate memory in units of
+        * long-words.
+        */
+       ret = -ENOMEM;
+       size = FDS_BYTES(n);
+       bits = select_bits_alloc(size);
+       if (!bits)
+               goto out_nofds;
+       fds.in      = (unsigned long *)  bits;
+       fds.out     = (unsigned long *) (bits +   size);
+       fds.ex      = (unsigned long *) (bits + 2*size);
+       fds.res_in  = (unsigned long *) (bits + 3*size);
+       fds.res_out = (unsigned long *) (bits + 4*size);
+       fds.res_ex  = (unsigned long *) (bits + 5*size);
+
+       if ((ret = compat_get_fd_set(n, inp, fds.in)) ||
+           (ret = compat_get_fd_set(n, outp, fds.out)) ||
+           (ret = compat_get_fd_set(n, exp, fds.ex)))
+               goto out;
+       zero_fd_set(n, fds.res_in);
+       zero_fd_set(n, fds.res_out);
+       zero_fd_set(n, fds.res_ex);
+
+       ret = do_select(n, &fds, &timeout);
+
+       if (tvp && !(current->personality & STICKY_TIMEOUTS)) {
+               time_t sec = 0, usec = 0;
+               if (timeout) {
+                       sec = timeout / HZ;
+                       usec = timeout % HZ;
+                       usec *= (1000000/HZ);
+               }
+               if (put_user(sec, &tvp->tv_sec) ||
+                   put_user(usec, &tvp->tv_usec))
+                       ret = -EFAULT;
+       }
+
+       if (ret < 0)
+               goto out;
+       if (!ret) {
+               ret = -ERESTARTNOHAND;
+               if (signal_pending(current))
+                       goto out;
+               ret = 0;
+       }
+
+       compat_set_fd_set(n, inp, fds.res_in);
+       compat_set_fd_set(n, outp, fds.res_out);
+       compat_set_fd_set(n, exp, fds.res_ex);
+
+out:
+       select_bits_free(bits, size);
+out_nofds:
+       return ret;
+}
+
+#if defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE)
+/* Stuff for NFS server syscalls... */
+struct compat_nfsctl_svc {
+       u16                     svc32_port;
+       s32                     svc32_nthreads;
+};
+
+struct compat_nfsctl_client {
+       s8                      cl32_ident[NFSCLNT_IDMAX+1];
+       s32                     cl32_naddr;
+       struct in_addr          cl32_addrlist[NFSCLNT_ADDRMAX];
+       s32                     cl32_fhkeytype;
+       s32                     cl32_fhkeylen;
+       u8                      cl32_fhkey[NFSCLNT_KEYMAX];
+};
+
+struct compat_nfsctl_export {
+       char            ex32_client[NFSCLNT_IDMAX+1];
+       char            ex32_path[NFS_MAXPATHLEN+1];
+       compat_dev_t    ex32_dev;
+       compat_ino_t    ex32_ino;
+       compat_int_t    ex32_flags;
+       compat_uid_t    ex32_anon_uid;
+       compat_gid_t    ex32_anon_gid;
+};
+
+struct compat_nfsctl_fdparm {
+       struct sockaddr         gd32_addr;
+       s8                      gd32_path[NFS_MAXPATHLEN+1];
+       compat_int_t            gd32_version;
+};
+
+struct compat_nfsctl_fsparm {
+       struct sockaddr         gd32_addr;
+       s8                      gd32_path[NFS_MAXPATHLEN+1];
+       compat_int_t            gd32_maxlen;
+};
+
+struct compat_nfsctl_arg {
+       compat_int_t            ca32_version;   /* safeguard */
+       union {
+               struct compat_nfsctl_svc        u32_svc;
+               struct compat_nfsctl_client     u32_client;
+               struct compat_nfsctl_export     u32_export;
+               struct compat_nfsctl_fdparm     u32_getfd;
+               struct compat_nfsctl_fsparm     u32_getfs;
+       } u;
+#define ca32_svc       u.u32_svc
+#define ca32_client    u.u32_client
+#define ca32_export    u.u32_export
+#define ca32_getfd     u.u32_getfd
+#define ca32_getfs     u.u32_getfs
+};
+
+union compat_nfsctl_res {
+       __u8                    cr32_getfh[NFS_FHSIZE];
+       struct knfsd_fh         cr32_getfs;
+};
+
+static int compat_nfs_svc_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+{
+       int err;
+
+       err = access_ok(VERIFY_READ, &arg->ca32_svc, sizeof(arg->ca32_svc));
+       err |= get_user(karg->ca_version, &arg->ca32_version);
+       err |= __get_user(karg->ca_svc.svc_port, &arg->ca32_svc.svc32_port);
+       err |= __get_user(karg->ca_svc.svc_nthreads, &arg->ca32_svc.svc32_nthreads);
+       return (err) ? -EFAULT : 0;
+}
+
+static int compat_nfs_clnt_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+{
+       int err;
+
+       err = access_ok(VERIFY_READ, &arg->ca32_client, sizeof(arg->ca32_client));
+       err |= get_user(karg->ca_version, &arg->ca32_version);
+       err |= __copy_from_user(&karg->ca_client.cl_ident[0],
+                         &arg->ca32_client.cl32_ident[0],
+                         NFSCLNT_IDMAX);
+       err |= __get_user(karg->ca_client.cl_naddr, &arg->ca32_client.cl32_naddr);
+       err |= __copy_from_user(&karg->ca_client.cl_addrlist[0],
+                         &arg->ca32_client.cl32_addrlist[0],
+                         (sizeof(struct in_addr) * NFSCLNT_ADDRMAX));
+       err |= __get_user(karg->ca_client.cl_fhkeytype,
+                     &arg->ca32_client.cl32_fhkeytype);
+       err |= __get_user(karg->ca_client.cl_fhkeylen,
+                     &arg->ca32_client.cl32_fhkeylen);
+       err |= __copy_from_user(&karg->ca_client.cl_fhkey[0],
+                         &arg->ca32_client.cl32_fhkey[0],
+                         NFSCLNT_KEYMAX);
+
+       return (err) ? -EFAULT : 0;
+}
+
+static int compat_nfs_exp_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+{
+       int err;
+
+       err = access_ok(VERIFY_READ, &arg->ca32_export, sizeof(arg->ca32_export));
+       err |= get_user(karg->ca_version, &arg->ca32_version);
+       err |= __copy_from_user(&karg->ca_export.ex_client[0],
+                         &arg->ca32_export.ex32_client[0],
+                         NFSCLNT_IDMAX);
+       err |= __copy_from_user(&karg->ca_export.ex_path[0],
+                         &arg->ca32_export.ex32_path[0],
+                         NFS_MAXPATHLEN);
+       err |= __get_user(karg->ca_export.ex_dev,
+                     &arg->ca32_export.ex32_dev);
+       err |= __get_user(karg->ca_export.ex_ino,
+                     &arg->ca32_export.ex32_ino);
+       err |= __get_user(karg->ca_export.ex_flags,
+                     &arg->ca32_export.ex32_flags);
+       err |= __get_user(karg->ca_export.ex_anon_uid,
+                     &arg->ca32_export.ex32_anon_uid);
+       err |= __get_user(karg->ca_export.ex_anon_gid,
+                     &arg->ca32_export.ex32_anon_gid);
+       SET_UID(karg->ca_export.ex_anon_uid, karg->ca_export.ex_anon_uid);
+       SET_GID(karg->ca_export.ex_anon_gid, karg->ca_export.ex_anon_gid);
+
+       return (err) ? -EFAULT : 0;
+}
+
+static int compat_nfs_getfd_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+{
+       int err;
+
+       err = access_ok(VERIFY_READ, &arg->ca32_getfd, sizeof(arg->ca32_getfd));
+       err |= get_user(karg->ca_version, &arg->ca32_version);
+       err |= __copy_from_user(&karg->ca_getfd.gd_addr,
+                         &arg->ca32_getfd.gd32_addr,
+                         (sizeof(struct sockaddr)));
+       err |= __copy_from_user(&karg->ca_getfd.gd_path,
+                         &arg->ca32_getfd.gd32_path,
+                         (NFS_MAXPATHLEN+1));
+       err |= __get_user(karg->ca_getfd.gd_version,
+                     &arg->ca32_getfd.gd32_version);
+
+       return (err) ? -EFAULT : 0;
+}
+
+static int compat_nfs_getfs_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+{
+       int err;
+
+       err = access_ok(VERIFY_READ, &arg->ca32_getfs, sizeof(arg->ca32_getfs));
+       err |= get_user(karg->ca_version, &arg->ca32_version);
+       err |= __copy_from_user(&karg->ca_getfs.gd_addr,
+                         &arg->ca32_getfs.gd32_addr,
+                         (sizeof(struct sockaddr)));
+       err |= __copy_from_user(&karg->ca_getfs.gd_path,
+                         &arg->ca32_getfs.gd32_path,
+                         (NFS_MAXPATHLEN+1));
+       err |= __get_user(karg->ca_getfs.gd_maxlen,
+                     &arg->ca32_getfs.gd32_maxlen);
+
+       return (err) ? -EFAULT : 0;
+}
+
+/* This really doesn't need translations, we are only passing
+ * back a union which contains opaque nfs file handle data.
+ */
+static int compat_nfs_getfh_res_trans(union nfsctl_res *kres, union compat_nfsctl_res __user *res)
+{
+       int err;
+
+       err = copy_to_user(res, kres, sizeof(*res));
+
+       return (err) ? -EFAULT : 0;
+}
+
+asmlinkage long compat_sys_nfsservctl(int cmd, struct compat_nfsctl_arg __user *arg,
+                                       union compat_nfsctl_res __user *res)
+{
+       struct nfsctl_arg *karg;
+       union nfsctl_res *kres;
+       mm_segment_t oldfs;
+       int err;
+
+       karg = kmalloc(sizeof(*karg), GFP_USER);
+       kres = kmalloc(sizeof(*kres), GFP_USER);
+       if(!karg || !kres) {
+               err = -ENOMEM;
+               goto done;
+       }
+
+       switch(cmd) {
+       case NFSCTL_SVC:
+               err = compat_nfs_svc_trans(karg, arg);
+               break;
+
+       case NFSCTL_ADDCLIENT:
+               err = compat_nfs_clnt_trans(karg, arg);
+               break;
+
+       case NFSCTL_DELCLIENT:
+               err = compat_nfs_clnt_trans(karg, arg);
+               break;
+
+       case NFSCTL_EXPORT:
+       case NFSCTL_UNEXPORT:
+               err = compat_nfs_exp_trans(karg, arg);
+               break;
+
+       case NFSCTL_GETFD:
+               err = compat_nfs_getfd_trans(karg, arg);
+               break;
+
+       case NFSCTL_GETFS:
+               err = compat_nfs_getfs_trans(karg, arg);
+               break;
+
+       default:
+               err = -EINVAL;
+               goto done;
+       }
+
+       oldfs = get_fs();
+       set_fs(KERNEL_DS);
+       /* The __user pointer casts are valid because of the set_fs() */
+       err = sys_nfsservctl(cmd, (void __user *) karg, (void __user *) kres);
+       set_fs(oldfs);
+
+       if (err)
+               goto done;
+
+       if((cmd == NFSCTL_GETFD) ||
+          (cmd == NFSCTL_GETFS))
+               err = compat_nfs_getfh_res_trans(kres, res);
+
+done:
+       kfree(karg);
+       kfree(kres);
+       return err;
+}
+#else /* !NFSD */
+long asmlinkage compat_sys_nfsservctl(int cmd, void *notused, void *notused2)
+{
+       return sys_ni_syscall();
+}
+#endif