Fedora kernel-2.6.17-1.2142_FC4 patched with stable patch-2.6.17.4-vs2.0.2-rc26.diff
[linux-2.6.git] / fs / compat.c
index 689fdd0..b1f6478 100644 (file)
 #include <linux/file.h>
 #include <linux/vfs.h>
 #include <linux/ioctl32.h>
+#include <linux/ioctl.h>
 #include <linux/init.h>
 #include <linux/sockios.h>     /* for SIOCDEVPRIVATE */
 #include <linux/smb.h>
 #include <linux/smb_mount.h>
 #include <linux/ncp_mount.h>
+#include <linux/nfs4_mount.h>
 #include <linux/smp_lock.h>
 #include <linux/syscalls.h>
 #include <linux/ctype.h>
 #include <linux/module.h>
-#include <linux/dnotify.h>
+#include <linux/dirent.h>
+#include <linux/fsnotify.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 <linux/rwsem.h>
+#include <linux/acct.h>
+#include <linux/mm.h>
 
 #include <net/sock.h>          /* siocdevprivate_ioctl */
 
 #include <asm/uaccess.h>
 #include <asm/mmu_context.h>
+#include <asm/ioctls.h>
+
+extern void sigset_from_compat(sigset_t *set, compat_sigset_t *compat);
 
 /*
  * Not all architectures have sys_utime, so implement this in terms
@@ -61,28 +70,33 @@ asmlinkage long compat_sys_utime(char __user *filename, struct compat_utimbuf __
                tv[0].tv_usec = 0;
                tv[1].tv_usec = 0;
        }
-       return do_utimes(filename, t ? tv : NULL);
+       return do_utimes(AT_FDCWD, filename, t ? tv : NULL);
 }
 
-asmlinkage long compat_sys_utimes(char __user *filename, struct compat_timeval __user *t)
+asmlinkage long compat_sys_futimesat(unsigned int dfd, char __user *filename, struct compat_timeval __user *t)
 {
        struct timeval tv[2];
 
-       if (t) { 
+       if (t) {
                if (get_user(tv[0].tv_sec, &t[0].tv_sec) ||
                    get_user(tv[0].tv_usec, &t[0].tv_usec) ||
                    get_user(tv[1].tv_sec, &t[1].tv_sec) ||
                    get_user(tv[1].tv_usec, &t[1].tv_usec))
-                       return -EFAULT; 
-       } 
-       return do_utimes(filename, t ? tv : NULL);
+                       return -EFAULT;
+       }
+       return do_utimes(dfd, filename, t ? tv : NULL);
+}
+
+asmlinkage long compat_sys_utimes(char __user *filename, struct compat_timeval __user *t)
+{
+       return compat_sys_futimesat(AT_FDCWD, filename, t);
 }
 
 asmlinkage long compat_sys_newstat(char __user * filename,
                struct compat_stat __user *statbuf)
 {
        struct kstat stat;
-       int error = vfs_stat(filename, &stat);
+       int error = vfs_stat_fd(AT_FDCWD, filename, &stat);
 
        if (!error)
                error = cp_compat_stat(&stat, statbuf);
@@ -93,12 +107,35 @@ asmlinkage long compat_sys_newlstat(char __user * filename,
                struct compat_stat __user *statbuf)
 {
        struct kstat stat;
-       int error = vfs_lstat(filename, &stat);
+       int error = vfs_lstat_fd(AT_FDCWD, filename, &stat);
+
+       if (!error)
+               error = cp_compat_stat(&stat, statbuf);
+       return error;
+}
+
+#ifndef __ARCH_WANT_STAT64
+asmlinkage long compat_sys_newfstatat(unsigned int dfd, char __user *filename,
+               struct compat_stat __user *statbuf, int flag)
+{
+       struct kstat stat;
+       int error = -EINVAL;
+
+       if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
+               goto out;
+
+       if (flag & AT_SYMLINK_NOFOLLOW)
+               error = vfs_lstat_fd(dfd, filename, &stat);
+       else
+               error = vfs_stat_fd(dfd, filename, &stat);
 
        if (!error)
                error = cp_compat_stat(&stat, statbuf);
+
+out:
        return error;
 }
+#endif
 
 asmlinkage long compat_sys_newfstat(unsigned int fd,
                struct compat_stat __user * statbuf)
@@ -115,12 +152,19 @@ static int put_compat_statfs(struct compat_statfs __user *ubuf, struct kstatfs *
 {
        
        if (sizeof ubuf->f_blocks == 4) {
-               if ((kbuf->f_blocks | kbuf->f_bfree |
-                    kbuf->f_bavail | kbuf->f_files | kbuf->f_ffree) &
+               if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail) &
                    0xffffffff00000000ULL)
                        return -EOVERFLOW;
+               /* f_files and f_ffree may be -1; it's okay
+                * to stuff that into 32 bits */
+               if (kbuf->f_files != 0xffffffffffffffffULL
+                && (kbuf->f_files & 0xffffffff00000000ULL))
+                       return -EOVERFLOW;
+               if (kbuf->f_ffree != 0xffffffffffffffffULL
+                && (kbuf->f_ffree & 0xffffffff00000000ULL))
+                       return -EOVERFLOW;
        }
-       if (verify_area(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
+       if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
            __put_user(kbuf->f_type, &ubuf->f_type) ||
            __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
            __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
@@ -154,8 +198,8 @@ asmlinkage long compat_sys_statfs(const char __user *path, struct compat_statfs
        if (!error) {
                struct kstatfs tmp;
                error = vfs_statfs(nd.dentry->d_inode->i_sb, &tmp);
-               if (!error && put_compat_statfs(buf, &tmp))
-                       error = -EFAULT;
+               if (!error)
+                       error = put_compat_statfs(buf, &tmp);
                path_release(&nd);
        }
        return error;
@@ -172,8 +216,8 @@ asmlinkage long compat_sys_fstatfs(unsigned int fd, struct compat_statfs __user
        if (!file)
                goto out;
        error = vfs_statfs(file->f_dentry->d_inode->i_sb, &tmp);
-       if (!error && put_compat_statfs(buf, &tmp))
-               error = -EFAULT;
+       if (!error)
+               error = put_compat_statfs(buf, &tmp);
        fput(file);
 out:
        return error;
@@ -182,12 +226,19 @@ out:
 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 |
-                    kbuf->f_bavail | kbuf->f_files | kbuf->f_ffree) &
+               if ((kbuf->f_blocks | kbuf->f_bfree | kbuf->f_bavail) &
                    0xffffffff00000000ULL)
                        return -EOVERFLOW;
+               /* f_files and f_ffree may be -1; it's okay
+                * to stuff that into 32 bits */
+               if (kbuf->f_files != 0xffffffffffffffffULL
+                && (kbuf->f_files & 0xffffffff00000000ULL))
+                       return -EOVERFLOW;
+               if (kbuf->f_ffree != 0xffffffffffffffffULL
+                && (kbuf->f_ffree & 0xffffffff00000000ULL))
+                       return -EOVERFLOW;
        }
-       if (verify_area(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
+       if (!access_ok(VERIFY_WRITE, ubuf, sizeof(*ubuf)) ||
            __put_user(kbuf->f_type, &ubuf->f_type) ||
            __put_user(kbuf->f_bsize, &ubuf->f_bsize) ||
            __put_user(kbuf->f_blocks, &ubuf->f_blocks) ||
@@ -203,7 +254,7 @@ static int put_compat_statfs64(struct compat_statfs64 __user *ubuf, struct kstat
        return 0;
 }
 
-asmlinkage long compat_statfs64(const char __user *path, compat_size_t sz, struct compat_statfs64 __user *buf)
+asmlinkage long compat_sys_statfs64(const char __user *path, compat_size_t sz, struct compat_statfs64 __user *buf)
 {
        struct nameidata nd;
        int error;
@@ -215,14 +266,14 @@ asmlinkage long compat_statfs64(const char __user *path, compat_size_t sz, struc
        if (!error) {
                struct kstatfs tmp;
                error = vfs_statfs(nd.dentry->d_inode->i_sb, &tmp);
-               if (!error && put_compat_statfs64(buf, &tmp))
-                       error = -EFAULT;
+               if (!error)
+                       error = put_compat_statfs64(buf, &tmp);
                path_release(&nd);
        }
        return error;
 }
 
-asmlinkage long compat_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
+asmlinkage long compat_sys_fstatfs64(unsigned int fd, compat_size_t sz, struct compat_statfs64 __user *buf)
 {
        struct file * file;
        struct kstatfs tmp;
@@ -236,8 +287,8 @@ asmlinkage long compat_fstatfs64(unsigned int fd, compat_size_t sz, struct compa
        if (!file)
                goto out;
        error = vfs_statfs(file->f_dentry->d_inode->i_sb, &tmp);
-       if (!error && put_compat_statfs64(buf, &tmp))
-               error = -EFAULT;
+       if (!error)
+               error = put_compat_statfs64(buf, &tmp);
        fput(file);
 out:
        return error;
@@ -246,7 +297,7 @@ out:
 /* ioctl32 stuff, used by sparc64, parisc, s390x, ppc64, x86_64, MIPS */
 
 #define IOCTL_HASHSIZE 256
-struct ioctl_trans *ioctl32_hash_table[IOCTL_HASHSIZE];
+static struct ioctl_trans *ioctl32_hash_table[IOCTL_HASHSIZE];
 
 extern struct ioctl_trans ioctl_start[];
 extern int ioctl_table_size;
@@ -268,7 +319,7 @@ static void ioctl32_insert_translation(struct ioctl_trans *trans)
                t = ioctl32_hash_table[hash];
                while (t->next)
                        t = t->next;
-               trans->next = 0;
+               trans->next = NULL;
                t->next = trans;
        }
 }
@@ -290,163 +341,115 @@ static int __init init_sys32_ioctl(void)
 
 __initcall(init_sys32_ioctl);
 
-int register_ioctl32_conversion(unsigned int cmd, int (*handler)(unsigned int,
-                               unsigned int, unsigned long, struct file *))
+static void compat_ioctl_error(struct file *filp, unsigned int fd,
+               unsigned int cmd, unsigned long arg)
 {
-       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 = ioctl32_hash_table[hash]; t; t = t->next) {
-               if (t->cmd == cmd) {
-                       printk(KERN_ERR "Trying to register duplicated ioctl32 "
-                                       "handler %x\n", cmd);
-                       unlock_kernel();
-                       kfree(new_t);
-                       return -EINVAL; 
-               }
+       char buf[10];
+       char *fn = "?";
+       char *path;
+
+       /* find the name of the device. */
+       path = (char *)__get_free_page(GFP_KERNEL);
+       if (path) {
+               fn = d_path(filp->f_dentry, filp->f_vfsmnt, path, PAGE_SIZE);
+               if (IS_ERR(fn))
+                       fn = "?";
        }
-       new_t->next = NULL;
-       new_t->cmd = cmd;
-       new_t->handler = handler;
-       ioctl32_insert_translation(new_t);
 
-       unlock_kernel();
-       return 0;
+       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);
 }
-EXPORT_SYMBOL(register_ioctl32_conversion);
 
-static inline int builtin_ioctl(struct ioctl_trans *t)
-{ 
-       return t >= ioctl_start && t < (ioctl_start + ioctl_table_size);
-} 
+asmlinkage long compat_sys_ioctl(unsigned int fd, unsigned int cmd,
+                               unsigned long arg)
+{
+       struct file *filp;
+       int error = -EBADF;
+       struct ioctl_trans *t;
+       int fput_needed;
 
-/* Problem: 
-   This function cannot unregister duplicate ioctls, because they are not
-   unique.
-   When they happen we need to extend the prototype to pass the handler too. */
+       filp = fget_light(fd, &fput_needed);
+       if (!filp)
+               goto out;
 
-int unregister_ioctl32_conversion(unsigned int cmd)
-{
-       unsigned long hash = ioctl32_hash(cmd);
-       struct ioctl_trans *t, *t1;
+       /* RED-PEN how should LSM module know it's handling 32bit? */
+       error = security_file_ioctl(filp, cmd, arg);
+       if (error)
+               goto out_fput;
 
-       lock_kernel(); 
+       /*
+        * To allow the compat_ioctl handlers to be self contained
+        * we need to check the common ioctls here first.
+        * Just handle them with the standard handlers below.
+        */
+       switch (cmd) {
+       case FIOCLEX:
+       case FIONCLEX:
+       case FIONBIO:
+       case FIOASYNC:
+       case FIOQSIZE:
+               break;
 
-       t = ioctl32_hash_table[hash];
-       if (!t) { 
-               unlock_kernel();
-               return -EINVAL;
-       } 
+       case FIBMAP:
+       case FIGETBSZ:
+       case FIONREAD:
+               if (S_ISREG(filp->f_dentry->d_inode->i_mode))
+                       break;
+               /*FALL THROUGH*/
 
-       if (t->cmd == cmd) { 
-               if (builtin_ioctl(t)) {
-                       printk("%p tried to unregister builtin ioctl %x\n",
-                              __builtin_return_address(0), cmd);
-               } else { 
-                       ioctl32_hash_table[hash] = t->next;
-                       unlock_kernel();
-                       kfree(t);
-                       return 0;
-               }
-       } 
-       while (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);
-                               goto out;
-                       } else { 
-                               t->next = t1->next;
-                               unlock_kernel();
-                               kfree(t1);
-                               return 0;
-                       }
+       default:
+               if (filp->f_op && filp->f_op->compat_ioctl) {
+                       error = filp->f_op->compat_ioctl(filp, cmd, arg);
+                       if (error != -ENOIOCTLCMD)
+                               goto out_fput;
                }
-               t = t1;
+
+               if (!filp->f_op ||
+                   (!filp->f_op->ioctl && !filp->f_op->unlocked_ioctl))
+                       goto do_ioctl;
+               break;
        }
-       printk(KERN_ERR "Trying to free unknown 32bit ioctl handler %x\n",
-                               cmd);
-out:
-       unlock_kernel();
-       return -EINVAL;
-}
-EXPORT_SYMBOL(unregister_ioctl32_conversion); 
 
-asmlinkage long compat_sys_ioctl(unsigned int fd, unsigned int cmd,
-                               unsigned long arg)
-{
-       struct file * filp;
-       int error = -EBADF;
-       struct ioctl_trans *t;
+       for (t = ioctl32_hash_table[ioctl32_hash(cmd)]; t; t = t->next) {
+               if (t->cmd == cmd)
+                       goto found_handler;
+       }
 
-       filp = fget(fd);
-       if(!filp)
-               goto out2;
+       if (S_ISSOCK(filp->f_dentry->d_inode->i_mode) &&
+           cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
+               error = siocdevprivate_ioctl(fd, cmd, arg);
+       } else {
+               static int count;
 
-       if (!filp->f_op || !filp->f_op->ioctl) {
-               error = sys_ioctl (fd, cmd, arg);
-               goto out;
+               if (++count <= 50)
+                       compat_ioctl_error(filp, fd, cmd, arg);
+               error = -EINVAL;
        }
 
-       lock_kernel();
-
-       t = ioctl32_hash_table[ioctl32_hash (cmd)];
+       goto out_fput;
 
-       while (t && t->cmd != cmd)
-               t = t->next;
-       if (t) {
-               if (t->handler) { 
-                       error = t->handler(fd, cmd, arg, filp);
-                       unlock_kernel();
-               } else {
-                       unlock_kernel();
-                       error = sys_ioctl(fd, cmd, arg);
-               }
-       } else {
+ found_handler:
+       if (t->handler) {
+               lock_kernel();
+               error = t->handler(fd, cmd, arg, filp);
                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);
-                       }
-                       error = -EINVAL;
-               }
+               goto out_fput;
        }
-out:
-       fput(filp);
-out2:
+
+ do_ioctl:
+       error = vfs_ioctl(filp, fd, cmd, arg);
+ out_fput:
+       fput_light(filp, fput_needed);
+ out:
        return error;
 }
 
@@ -520,10 +523,22 @@ asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
                set_fs(KERNEL_DS);
                ret = sys_fcntl(fd, cmd, (unsigned long)&f);
                set_fs(old_fs);
-               if ((cmd == F_GETLK) && (ret == 0)) {
-                       if ((f.l_start >= COMPAT_OFF_T_MAX) ||
-                           ((f.l_start + f.l_len) >= COMPAT_OFF_T_MAX))
+               if (cmd == F_GETLK && ret == 0) {
+                       /* GETLK was successfule and we need to return the data...
+                        * but it needs to fit in the compat structure.
+                        * l_start shouldn't be too big, unless the original
+                        * start + end is greater than COMPAT_OFF_T_MAX, in which
+                        * case the app was asking for trouble, so we return
+                        * -EOVERFLOW in that case.
+                        * l_len could be too big, in which case we just truncate it,
+                        * and only allow the app to see that part of the conflicting
+                        * lock that might make sense to it anyway
+                        */
+
+                       if (f.l_start > COMPAT_OFF_T_MAX)
                                ret = -EOVERFLOW;
+                       if (f.l_len > COMPAT_OFF_T_MAX)
+                               f.l_len = COMPAT_OFF_T_MAX;
                        if (ret == 0)
                                ret = put_compat_flock(&f, compat_ptr(arg));
                }
@@ -541,10 +556,12 @@ asmlinkage long compat_sys_fcntl64(unsigned int fd, unsigned int cmd,
                                ((cmd == F_SETLK64) ? F_SETLK : F_SETLKW),
                                (unsigned long)&f);
                set_fs(old_fs);
-               if ((cmd == F_GETLK64) && (ret == 0)) {
-                       if ((f.l_start >= COMPAT_LOFF_T_MAX) ||
-                           ((f.l_start + f.l_len) >= COMPAT_LOFF_T_MAX))
+               if (cmd == F_GETLK64 && ret == 0) {
+                       /* need to return lock information - see above for commentary */
+                       if (f.l_start > COMPAT_LOFF_T_MAX)
                                ret = -EOVERFLOW;
+                       if (f.l_len > COMPAT_LOFF_T_MAX)
+                               f.l_len = COMPAT_LOFF_T_MAX;
                        if (ret == 0)
                                ret = put_compat_flock64(&f, compat_ptr(arg));
                }
@@ -652,14 +669,14 @@ compat_sys_io_submit(aio_context_t ctx_id, int nr, u32 __user *iocb)
 struct compat_ncp_mount_data {
        compat_int_t version;
        compat_uint_t ncp_fd;
-       compat_uid_t mounted_uid;
+       __compat_uid_t mounted_uid;
        compat_pid_t wdog_pid;
        unsigned char mounted_vol[NCP_VOLNAME_LEN + 1];
        compat_uint_t time_out;
        compat_uint_t retry_count;
        compat_uint_t flags;
-       compat_uid_t uid;
-       compat_gid_t gid;
+       __compat_uid_t uid;
+       __compat_gid_t gid;
        compat_mode_t file_mode;
        compat_mode_t dir_mode;
 };
@@ -716,9 +733,9 @@ static void *do_ncp_super_data_conv(void *raw_data)
 
 struct compat_smb_mount_data {
        compat_int_t version;
-       compat_uid_t mounted_uid;
-       compat_uid_t uid;
-       compat_gid_t gid;
+       __compat_uid_t mounted_uid;
+       __compat_uid_t uid;
+       __compat_gid_t gid;
        compat_mode_t file_mode;
        compat_mode_t dir_mode;
 };
@@ -739,10 +756,79 @@ static void *do_smb_super_data_conv(void *raw_data)
        return raw_data;
 }
 
+struct compat_nfs_string {
+       compat_uint_t len;
+       compat_uptr_t data;
+};
+
+static inline void compat_nfs_string(struct nfs_string *dst,
+                                    struct compat_nfs_string *src)
+{
+       dst->data = compat_ptr(src->data);
+       dst->len = src->len;
+}
+
+struct compat_nfs4_mount_data_v1 {
+       compat_int_t version;
+       compat_int_t flags;
+       compat_int_t rsize;
+       compat_int_t wsize;
+       compat_int_t timeo;
+       compat_int_t retrans;
+       compat_int_t acregmin;
+       compat_int_t acregmax;
+       compat_int_t acdirmin;
+       compat_int_t acdirmax;
+       struct compat_nfs_string client_addr;
+       struct compat_nfs_string mnt_path;
+       struct compat_nfs_string hostname;
+       compat_uint_t host_addrlen;
+       compat_uptr_t host_addr;
+       compat_int_t proto;
+       compat_int_t auth_flavourlen;
+       compat_uptr_t auth_flavours;
+};
+
+static int do_nfs4_super_data_conv(void *raw_data)
+{
+       int version = *(compat_uint_t *) raw_data;
+
+       if (version == 1) {
+               struct compat_nfs4_mount_data_v1 *raw = raw_data;
+               struct nfs4_mount_data *real = raw_data;
+
+               /* copy the fields backwards */
+               real->auth_flavours = compat_ptr(raw->auth_flavours);
+               real->auth_flavourlen = raw->auth_flavourlen;
+               real->proto = raw->proto;
+               real->host_addr = compat_ptr(raw->host_addr);
+               real->host_addrlen = raw->host_addrlen;
+               compat_nfs_string(&real->hostname, &raw->hostname);
+               compat_nfs_string(&real->mnt_path, &raw->mnt_path);
+               compat_nfs_string(&real->client_addr, &raw->client_addr);
+               real->acdirmax = raw->acdirmax;
+               real->acdirmin = raw->acdirmin;
+               real->acregmax = raw->acregmax;
+               real->acregmin = raw->acregmin;
+               real->retrans = raw->retrans;
+               real->timeo = raw->timeo;
+               real->wsize = raw->wsize;
+               real->rsize = raw->rsize;
+               real->flags = raw->flags;
+               real->version = raw->version;
+       }
+       else {
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
 extern int copy_mount_options (const void __user *, unsigned long *);
 
 #define SMBFS_NAME      "smbfs"
 #define NCPFS_NAME      "ncpfs"
+#define NFS4_NAME      "nfs4"
 
 asmlinkage long compat_sys_mount(char __user * dev_name, char __user * dir_name,
                                 char __user * type, unsigned long flags,
@@ -778,6 +864,9 @@ asmlinkage long compat_sys_mount(char __user * dev_name, char __user * dir_name,
                        do_smb_super_data_conv((void *)data_page);
                } else if (!strcmp((char *)type_page, NCPFS_NAME)) {
                        do_ncp_super_data_conv((void *)data_page);
+               } else if (!strcmp((char *)type_page, NFS4_NAME)) {
+                       if (do_nfs4_super_data_conv((void *) data_page))
+                               goto out4;
                }
        }
 
@@ -786,6 +875,7 @@ asmlinkage long compat_sys_mount(char __user * dev_name, char __user * dir_name,
                        flags, (void*)data_page);
        unlock_kernel();
 
+ out4:
        free_page(data_page);
  out3:
        free_page(dev_page);
@@ -797,6 +887,254 @@ asmlinkage long compat_sys_mount(char __user * dev_name, char __user * dir_name,
        return retval;
 }
 
+#define NAME_OFFSET(de) ((int) ((de)->d_name - (char __user *) (de)))
+#define COMPAT_ROUND_UP(x) (((x)+sizeof(compat_long_t)-1) & \
+                               ~(sizeof(compat_long_t)-1))
+
+struct compat_old_linux_dirent {
+       compat_ulong_t  d_ino;
+       compat_ulong_t  d_offset;
+       unsigned short  d_namlen;
+       char            d_name[1];
+};
+
+struct compat_readdir_callback {
+       struct compat_old_linux_dirent __user *dirent;
+       int result;
+};
+
+static int compat_fillonedir(void *__buf, const char *name, int namlen,
+                       loff_t offset, ino_t ino, unsigned int d_type)
+{
+       struct compat_readdir_callback *buf = __buf;
+       struct compat_old_linux_dirent __user *dirent;
+
+       if (buf->result)
+               return -EINVAL;
+       buf->result++;
+       dirent = buf->dirent;
+       if (!access_ok(VERIFY_WRITE, dirent,
+                       (unsigned long)(dirent->d_name + namlen + 1) -
+                               (unsigned long)dirent))
+               goto efault;
+       if (    __put_user(ino, &dirent->d_ino) ||
+               __put_user(offset, &dirent->d_offset) ||
+               __put_user(namlen, &dirent->d_namlen) ||
+               __copy_to_user(dirent->d_name, name, namlen) ||
+               __put_user(0, dirent->d_name + namlen))
+               goto efault;
+       return 0;
+efault:
+       buf->result = -EFAULT;
+       return -EFAULT;
+}
+
+asmlinkage long compat_sys_old_readdir(unsigned int fd,
+       struct compat_old_linux_dirent __user *dirent, unsigned int count)
+{
+       int error;
+       struct file *file;
+       struct compat_readdir_callback buf;
+
+       error = -EBADF;
+       file = fget(fd);
+       if (!file)
+               goto out;
+
+       buf.result = 0;
+       buf.dirent = dirent;
+
+       error = vfs_readdir(file, compat_fillonedir, &buf);
+       if (error >= 0)
+               error = buf.result;
+
+       fput(file);
+out:
+       return error;
+}
+
+struct compat_linux_dirent {
+       compat_ulong_t  d_ino;
+       compat_ulong_t  d_off;
+       unsigned short  d_reclen;
+       char            d_name[1];
+};
+
+struct compat_getdents_callback {
+       struct compat_linux_dirent __user *current_dir;
+       struct compat_linux_dirent __user *previous;
+       int count;
+       int error;
+};
+
+static int compat_filldir(void *__buf, const char *name, int namlen,
+               loff_t offset, ino_t ino, unsigned int d_type)
+{
+       struct compat_linux_dirent __user * dirent;
+       struct compat_getdents_callback *buf = __buf;
+       int reclen = COMPAT_ROUND_UP(NAME_OFFSET(dirent) + namlen + 2);
+
+       buf->error = -EINVAL;   /* only used if we fail.. */
+       if (reclen > buf->count)
+               return -EINVAL;
+       dirent = buf->previous;
+       if (dirent) {
+               if (__put_user(offset, &dirent->d_off))
+                       goto efault;
+       }
+       dirent = buf->current_dir;
+       if (__put_user(ino, &dirent->d_ino))
+               goto efault;
+       if (__put_user(reclen, &dirent->d_reclen))
+               goto efault;
+       if (copy_to_user(dirent->d_name, name, namlen))
+               goto efault;
+       if (__put_user(0, dirent->d_name + namlen))
+               goto efault;
+       if (__put_user(d_type, (char  __user *) dirent + reclen - 1))
+               goto efault;
+       buf->previous = dirent;
+       dirent = (void __user *)dirent + reclen;
+       buf->current_dir = dirent;
+       buf->count -= reclen;
+       return 0;
+efault:
+       buf->error = -EFAULT;
+       return -EFAULT;
+}
+
+asmlinkage long compat_sys_getdents(unsigned int fd,
+               struct compat_linux_dirent __user *dirent, unsigned int count)
+{
+       struct file * file;
+       struct compat_linux_dirent __user * lastdirent;
+       struct compat_getdents_callback buf;
+       int error;
+
+       error = -EFAULT;
+       if (!access_ok(VERIFY_WRITE, dirent, count))
+               goto out;
+
+       error = -EBADF;
+       file = fget(fd);
+       if (!file)
+               goto out;
+
+       buf.current_dir = dirent;
+       buf.previous = NULL;
+       buf.count = count;
+       buf.error = 0;
+
+       error = vfs_readdir(file, compat_filldir, &buf);
+       if (error < 0)
+               goto out_putf;
+       error = buf.error;
+       lastdirent = buf.previous;
+       if (lastdirent) {
+               if (put_user(file->f_pos, &lastdirent->d_off))
+                       error = -EFAULT;
+               else
+                       error = count - buf.count;
+       }
+
+out_putf:
+       fput(file);
+out:
+       return error;
+}
+
+#ifndef __ARCH_OMIT_COMPAT_SYS_GETDENTS64
+#define COMPAT_ROUND_UP64(x) (((x)+sizeof(u64)-1) & ~(sizeof(u64)-1))
+
+struct compat_getdents_callback64 {
+       struct linux_dirent64 __user *current_dir;
+       struct linux_dirent64 __user *previous;
+       int count;
+       int error;
+};
+
+static int compat_filldir64(void * __buf, const char * name, int namlen, loff_t offset,
+                    ino_t ino, unsigned int d_type)
+{
+       struct linux_dirent64 __user *dirent;
+       struct compat_getdents_callback64 *buf = __buf;
+       int jj = NAME_OFFSET(dirent);
+       int reclen = COMPAT_ROUND_UP64(jj + namlen + 1);
+       u64 off;
+
+       buf->error = -EINVAL;   /* only used if we fail.. */
+       if (reclen > buf->count)
+               return -EINVAL;
+       dirent = buf->previous;
+
+       if (dirent) {
+               if (__put_user_unaligned(offset, &dirent->d_off))
+                       goto efault;
+       }
+       dirent = buf->current_dir;
+       if (__put_user_unaligned(ino, &dirent->d_ino))
+               goto efault;
+       off = 0;
+       if (__put_user_unaligned(off, &dirent->d_off))
+               goto efault;
+       if (__put_user(reclen, &dirent->d_reclen))
+               goto efault;
+       if (__put_user(d_type, &dirent->d_type))
+               goto efault;
+       if (copy_to_user(dirent->d_name, name, namlen))
+               goto efault;
+       if (__put_user(0, dirent->d_name + namlen))
+               goto efault;
+       buf->previous = dirent;
+       dirent = (void __user *)dirent + reclen;
+       buf->current_dir = dirent;
+       buf->count -= reclen;
+       return 0;
+efault:
+       buf->error = -EFAULT;
+       return -EFAULT;
+}
+
+asmlinkage long compat_sys_getdents64(unsigned int fd,
+               struct linux_dirent64 __user * dirent, unsigned int count)
+{
+       struct file * file;
+       struct linux_dirent64 __user * lastdirent;
+       struct compat_getdents_callback64 buf;
+       int error;
+
+       error = -EFAULT;
+       if (!access_ok(VERIFY_WRITE, dirent, count))
+               goto out;
+
+       error = -EBADF;
+       file = fget(fd);
+       if (!file)
+               goto out;
+
+       buf.current_dir = dirent;
+       buf.previous = NULL;
+       buf.count = count;
+       buf.error = 0;
+
+       error = vfs_readdir(file, compat_filldir64, &buf);
+       if (error < 0)
+               goto out_putf;
+       error = buf.error;
+       lastdirent = buf.previous;
+       if (lastdirent) {
+               typeof(lastdirent->d_off) d_off = file->f_pos;
+               __put_user_unaligned(d_off, &lastdirent->d_off);
+               error = count - buf.count;
+       }
+
+out_putf:
+       fput(file);
+out:
+       return error;
+}
+#endif /* ! __ARCH_OMIT_COMPAT_SYS_GETDENTS64 */
+
 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)
@@ -811,7 +1149,6 @@ static ssize_t compat_do_readv_writev(int type, struct file *file,
        int seg;
        io_fn_t fn;
        iov_fn_t fnv;
-       struct inode *inode;
 
        /*
         * SuS says "The readv() function *may* fail if the iovcnt argument
@@ -838,7 +1175,7 @@ static ssize_t compat_do_readv_writev(int type, struct file *file,
                        goto out;
        }
        ret = -EFAULT;
-       if (verify_area(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
+       if (!access_ok(VERIFY_READ, uvector, nr_segs*sizeof(*uvector)))
                goto out;
 
        /*
@@ -876,11 +1213,11 @@ static ssize_t compat_do_readv_writev(int type, struct file *file,
                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);
+       ret = rw_verify_area(type, file, pos, tot_len);
+       if (ret < 0)
+               goto out;
+
+       ret = security_file_permission(file, type == READ ? MAY_READ:MAY_WRITE);
        if (ret)
                goto out;
 
@@ -923,9 +1260,13 @@ static ssize_t compat_do_readv_writev(int type, struct file *file,
 out:
        if (iov != iovstack)
                kfree(iov);
-       if ((ret + (type == READ)) > 0)
-               dnotify_parent(file->f_dentry,
-                               (type == READ) ? DN_ACCESS : DN_MODIFY);
+       if ((ret + (type == READ)) > 0) {
+               struct dentry *dentry = file->f_dentry;
+               if (type == READ)
+                       fsnotify_access(dentry);
+               else
+                       fsnotify_modify(dentry);
+       }
        return ret;
 }
 
@@ -976,6 +1317,46 @@ out:
        return ret;
 }
 
+asmlinkage long
+compat_sys_vmsplice(int fd, const struct compat_iovec __user *iov32,
+                   unsigned int nr_segs, unsigned int flags)
+{
+       unsigned i;
+       struct iovec *iov;
+       if (nr_segs > UIO_MAXIOV)
+               return -EINVAL;
+       iov = compat_alloc_user_space(nr_segs * sizeof(struct iovec));
+       for (i = 0; i < nr_segs; i++) {
+               struct compat_iovec v;
+               if (get_user(v.iov_base, &iov32[i].iov_base) ||
+                   get_user(v.iov_len, &iov32[i].iov_len) ||
+                   put_user(compat_ptr(v.iov_base), &iov[i].iov_base) ||
+                   put_user(v.iov_len, &iov[i].iov_len))
+                       return -EFAULT;
+       }
+       return sys_vmsplice(fd, iov, nr_segs, flags);
+}
+
+/*
+ * Exactly like fs/open.c:sys_open(), except that it doesn't set the
+ * O_LARGEFILE flag.
+ */
+asmlinkage long
+compat_sys_open(const char __user *filename, int flags, int mode)
+{
+       return do_sys_open(AT_FDCWD, filename, flags, mode);
+}
+
+/*
+ * Exactly like fs/open.c:sys_openat(), except that it doesn't set the
+ * O_LARGEFILE flag.
+ */
+asmlinkage long
+compat_sys_openat(unsigned int dfd, const char __user *filename, int flags, int mode)
+{
+       return do_sys_open(dfd, filename, flags, mode);
+}
+
 /*
  * 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
@@ -1113,97 +1494,101 @@ int compat_do_execve(char * filename,
        compat_uptr_t __user *envp,
        struct pt_regs * regs)
 {
-       struct linux_binprm bprm;
+       struct linux_binprm *bprm;
        struct file *file;
        int retval;
        int i;
 
-       sched_balance_exec();
+       retval = -ENOMEM;
+       bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
+       if (!bprm)
+               goto out_ret;
 
        file = open_exec(filename);
-
        retval = PTR_ERR(file);
        if (IS_ERR(file))
-               return retval;
+               goto out_kfree;
 
-       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();
+       sched_exec();
+
+       bprm->p = PAGE_SIZE*MAX_ARG_PAGES-sizeof(void *);
+       bprm->file = file;
+       bprm->filename = filename;
+       bprm->interp = filename;
+       bprm->mm = mm_alloc();
        retval = -ENOMEM;
-       if (!bprm.mm)
+       if (!bprm->mm)
                goto out_file;
 
-       retval = init_new_context(current, bprm.mm);
+       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)
+       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)
+       bprm->envc = compat_count(envp, bprm->p / sizeof(compat_uptr_t));
+       if ((retval = bprm->envc) < 0)
                goto out_mm;
 
-       retval = security_bprm_alloc(&bprm);
+       retval = security_bprm_alloc(bprm);
        if (retval)
                goto out;
 
-       retval = prepare_binprm(&bprm);
+       retval = prepare_binprm(bprm);
        if (retval < 0)
                goto out;
 
-       retval = copy_strings_kernel(1, &bprm.filename, &bprm);
+       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);
+       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);
+       retval = compat_copy_strings(bprm->argc, argv, bprm);
        if (retval < 0)
                goto out;
 
-       retval = search_binary_handler(&bprm,regs);
+       retval = search_binary_handler(bprm, regs);
        if (retval >= 0) {
-               free_arg_pages(&bprm);
+               free_arg_pages(bprm);
 
                /* execve success */
-               security_bprm_free(&bprm);
+               security_bprm_free(bprm);
+               acct_update_integrals(current);
+               kfree(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];
+               struct page * page = bprm->page[i];
                if (page)
                        __free_page(page);
        }
 
-       if (bprm.security)
-               security_bprm_free(&bprm);
+       if (bprm->security)
+               security_bprm_free(bprm);
 
 out_mm:
-       if (bprm.mm)
-               mmdrop(bprm.mm);
+       if (bprm->mm)
+               mmdrop(bprm->mm);
 
 out_file:
-       if (bprm.file) {
-               allow_write_access(bprm.file);
-               fput(bprm.file);
+       if (bprm->file) {
+               allow_write_access(bprm->file);
+               fput(bprm->file);
        }
 
+out_kfree:
+       kfree(bprm);
+
+out_ret:
        return retval;
 }
 
@@ -1215,7 +1600,7 @@ out_file:
  * Ooo, nasty.  We need here to frob 32-bit unsigned longs to
  * 64-bit unsigned longs.
  */
-static inline
+static
 int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
                        unsigned long *fdset)
 {
@@ -1223,7 +1608,7 @@ int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
        if (ufdset) {
                unsigned long odd;
 
-               if (verify_area(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
+               if (!access_ok(VERIFY_WRITE, ufdset, nr*sizeof(compat_ulong_t)))
                        return -EFAULT;
 
                odd = nr & 1UL;
@@ -1248,7 +1633,7 @@ int compat_get_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
        return 0;
 }
 
-static inline
+static
 void compat_set_fd_set(unsigned long nr, compat_ulong_t __user *ufdset,
                        unsigned long *fdset)
 {
@@ -1278,15 +1663,6 @@ void compat_set_fd_set(unsigned long nr, compat_ulong_t __user *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
@@ -1299,40 +1675,22 @@ static void select_bits_free(void *bits, int size)
 #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)
+int compat_core_sys_select(int n, compat_ulong_t __user *inp,
+       compat_ulong_t __user *outp, compat_ulong_t __user *exp, s64 *timeout)
 {
        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;
+       int size, max_fdset, ret = -EINVAL;
+       struct fdtable *fdt;
 
-               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;
+       rcu_read_lock();
+       fdt = files_fdtable(current->files);
+       max_fdset = fdt->max_fdset;
+       rcu_read_unlock();
        if (n > max_fdset)
                n = max_fdset;
 
@@ -1343,7 +1701,7 @@ compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp
         */
        ret = -ENOMEM;
        size = FDS_BYTES(n);
-       bits = select_bits_alloc(size);
+       bits = kmalloc(6 * size, GFP_KERNEL);
        if (!bits)
                goto out_nofds;
        fds.in      = (unsigned long *)  bits;
@@ -1361,19 +1719,7 @@ compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp
        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;
-       }
+       ret = do_select(n, &fds, timeout);
 
        if (ret < 0)
                goto out;
@@ -1389,11 +1735,242 @@ compat_sys_select(int n, compat_ulong_t __user *inp, compat_ulong_t __user *outp
        compat_set_fd_set(n, exp, fds.res_ex);
 
 out:
-       select_bits_free(bits, size);
+       kfree(bits);
 out_nofds:
        return ret;
 }
 
+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)
+{
+       s64 timeout = -1;
+       struct compat_timeval tv;
+       int ret;
+
+       if (tvp) {
+               if (copy_from_user(&tv, tvp, sizeof(tv)))
+                       return -EFAULT;
+
+               if (tv.tv_sec < 0 || tv.tv_usec < 0)
+                       return -EINVAL;
+
+               /* Cast to u64 to make GCC stop complaining */
+               if ((u64)tv.tv_sec >= (u64)MAX_INT64_SECONDS)
+                       timeout = -1;   /* infinite */
+               else {
+                       timeout = ROUND_UP(tv.tv_usec, 1000000/HZ);
+                       timeout += tv.tv_sec * HZ;
+               }
+       }
+
+       ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+       if (tvp) {
+               struct compat_timeval rtv;
+
+               if (current->personality & STICKY_TIMEOUTS)
+                       goto sticky;
+               rtv.tv_usec = jiffies_to_usecs(do_div((*(u64*)&timeout), HZ));
+               rtv.tv_sec = timeout;
+               if (compat_timeval_compare(&rtv, &tv) >= 0)
+                       rtv = tv;
+               if (copy_to_user(tvp, &rtv, sizeof(rtv))) {
+sticky:
+                       /*
+                        * If an application puts its timeval in read-only
+                        * memory, we don't want the Linux-specific update to
+                        * the timeval to cause a fault after the select has
+                        * completed successfully. However, because we're not
+                        * updating the timeval, we can't restart the system
+                        * call.
+                        */
+                       if (ret == -ERESTARTNOHAND)
+                               ret = -EINTR;
+               }
+       }
+
+       return ret;
+}
+
+#ifdef TIF_RESTORE_SIGMASK
+asmlinkage long compat_sys_pselect7(int n, compat_ulong_t __user *inp,
+       compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+       struct compat_timespec __user *tsp, compat_sigset_t __user *sigmask,
+       compat_size_t sigsetsize)
+{
+       compat_sigset_t ss32;
+       sigset_t ksigmask, sigsaved;
+       s64 timeout = MAX_SCHEDULE_TIMEOUT;
+       struct compat_timespec ts;
+       int ret;
+
+       if (tsp) {
+               if (copy_from_user(&ts, tsp, sizeof(ts)))
+                       return -EFAULT;
+
+               if (ts.tv_sec < 0 || ts.tv_nsec < 0)
+                       return -EINVAL;
+       }
+
+       if (sigmask) {
+               if (sigsetsize != sizeof(compat_sigset_t))
+                       return -EINVAL;
+               if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
+                       return -EFAULT;
+               sigset_from_compat(&ksigmask, &ss32);
+
+               sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+               sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+       }
+
+       do {
+               if (tsp) {
+                       if ((unsigned long)ts.tv_sec < MAX_SELECT_SECONDS) {
+                               timeout = ROUND_UP(ts.tv_nsec, 1000000000/HZ);
+                               timeout += ts.tv_sec * (unsigned long)HZ;
+                               ts.tv_sec = 0;
+                               ts.tv_nsec = 0;
+                       } else {
+                               ts.tv_sec -= MAX_SELECT_SECONDS;
+                               timeout = MAX_SELECT_SECONDS * HZ;
+                       }
+               }
+
+               ret = compat_core_sys_select(n, inp, outp, exp, &timeout);
+
+       } while (!ret && !timeout && tsp && (ts.tv_sec || ts.tv_nsec));
+
+       if (tsp && !(current->personality & STICKY_TIMEOUTS)) {
+               struct compat_timespec rts;
+
+               rts.tv_sec = timeout / HZ;
+               rts.tv_nsec = (timeout % HZ) * (NSEC_PER_SEC/HZ);
+               if (rts.tv_nsec >= NSEC_PER_SEC) {
+                       rts.tv_sec++;
+                       rts.tv_nsec -= NSEC_PER_SEC;
+               }
+               if (compat_timespec_compare(&rts, &ts) >= 0)
+                       rts = ts;
+               copy_to_user(tsp, &rts, sizeof(rts));
+       }
+
+       if (ret == -ERESTARTNOHAND) {
+               /*
+                * Don't restore the signal mask yet. Let do_signal() deliver
+                * the signal on the way back to userspace, before the signal
+                * mask is restored.
+                */
+               if (sigmask) {
+                       memcpy(&current->saved_sigmask, &sigsaved,
+                                       sizeof(sigsaved));
+                       set_thread_flag(TIF_RESTORE_SIGMASK);
+               }
+       } else if (sigmask)
+               sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+       return ret;
+}
+
+asmlinkage long compat_sys_pselect6(int n, compat_ulong_t __user *inp,
+       compat_ulong_t __user *outp, compat_ulong_t __user *exp,
+       struct compat_timespec __user *tsp, void __user *sig)
+{
+       compat_size_t sigsetsize = 0;
+       compat_uptr_t up = 0;
+
+       if (sig) {
+               if (!access_ok(VERIFY_READ, sig,
+                               sizeof(compat_uptr_t)+sizeof(compat_size_t)) ||
+                       __get_user(up, (compat_uptr_t __user *)sig) ||
+                       __get_user(sigsetsize,
+                               (compat_size_t __user *)(sig+sizeof(up))))
+                       return -EFAULT;
+       }
+       return compat_sys_pselect7(n, inp, outp, exp, tsp, compat_ptr(up),
+                                       sigsetsize);
+}
+
+asmlinkage long compat_sys_ppoll(struct pollfd __user *ufds,
+       unsigned int nfds, struct compat_timespec __user *tsp,
+       const compat_sigset_t __user *sigmask, compat_size_t sigsetsize)
+{
+       compat_sigset_t ss32;
+       sigset_t ksigmask, sigsaved;
+       struct compat_timespec ts;
+       s64 timeout = -1;
+       int ret;
+
+       if (tsp) {
+               if (copy_from_user(&ts, tsp, sizeof(ts)))
+                       return -EFAULT;
+
+               /* We assume that ts.tv_sec is always lower than
+                  the number of seconds that can be expressed in
+                  an s64. Otherwise the compiler bitches at us */
+               timeout = ROUND_UP(ts.tv_nsec, 1000000000/HZ);
+               timeout += ts.tv_sec * HZ;
+       }
+
+       if (sigmask) {
+               if (sigsetsize != sizeof(compat_sigset_t))
+                       return -EINVAL;
+               if (copy_from_user(&ss32, sigmask, sizeof(ss32)))
+                       return -EFAULT;
+               sigset_from_compat(&ksigmask, &ss32);
+
+               sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
+               sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
+       }
+
+       ret = do_sys_poll(ufds, nfds, &timeout);
+
+       /* We can restart this syscall, usually */
+       if (ret == -EINTR) {
+               /*
+                * Don't restore the signal mask yet. Let do_signal() deliver
+                * the signal on the way back to userspace, before the signal
+                * mask is restored.
+                */
+               if (sigmask) {
+                       memcpy(&current->saved_sigmask, &sigsaved,
+                               sizeof(sigsaved));
+                       set_thread_flag(TIF_RESTORE_SIGMASK);
+               }
+               ret = -ERESTARTNOHAND;
+       } else if (sigmask)
+               sigprocmask(SIG_SETMASK, &sigsaved, NULL);
+
+       if (tsp && timeout >= 0) {
+               struct compat_timespec rts;
+
+               if (current->personality & STICKY_TIMEOUTS)
+                       goto sticky;
+               /* Yes, we know it's actually an s64, but it's also positive. */
+               rts.tv_nsec = jiffies_to_usecs(do_div((*(u64*)&timeout), HZ)) *
+                                       1000;
+               rts.tv_sec = timeout;
+               if (compat_timespec_compare(&rts, &ts) >= 0)
+                       rts = ts;
+               if (copy_to_user(tsp, &rts, sizeof(rts))) {
+sticky:
+                       /*
+                        * If an application puts its timeval in read-only
+                        * memory, we don't want the Linux-specific update to
+                        * the timeval to cause a fault after the select has
+                        * completed successfully. However, because we're not
+                        * updating the timeval, we can't restart the system
+                        * call.
+                        */
+                       if (ret == -ERESTARTNOHAND && timeout >= 0)
+                               ret = -EINTR;
+               }
+       }
+
+       return ret;
+}
+#endif /* TIF_RESTORE_SIGMASK */
+
 #if defined(CONFIG_NFSD) || defined(CONFIG_NFSD_MODULE)
 /* Stuff for NFS server syscalls... */
 struct compat_nfsctl_svc {
@@ -1416,8 +1993,8 @@ struct compat_nfsctl_export {
        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;
+       __compat_uid_t  ex32_anon_uid;
+       __compat_gid_t  ex32_anon_gid;
 };
 
 struct compat_nfsctl_fdparm {
@@ -1453,109 +2030,115 @@ union compat_nfsctl_res {
        struct knfsd_fh         cr32_getfs;
 };
 
-static int compat_nfs_svc_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+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;
+       if (!access_ok(VERIFY_READ, &arg->ca32_svc, sizeof(arg->ca32_svc)) ||
+               get_user(karg->ca_version, &arg->ca32_version) ||
+               __get_user(karg->ca_svc.svc_port, &arg->ca32_svc.svc32_port) ||
+               __get_user(karg->ca_svc.svc_nthreads,
+                               &arg->ca32_svc.svc32_nthreads))
+               return -EFAULT;
+       return 0;
 }
 
-static int compat_nfs_clnt_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+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);
+       if (!access_ok(VERIFY_READ, &arg->ca32_client,
+                       sizeof(arg->ca32_client)) ||
+               get_user(karg->ca_version, &arg->ca32_version) ||
+               __copy_from_user(&karg->ca_client.cl_ident[0],
+                               &arg->ca32_client.cl32_ident[0],
+                               NFSCLNT_IDMAX) ||
+               __get_user(karg->ca_client.cl_naddr,
+                               &arg->ca32_client.cl32_naddr) ||
+               __copy_from_user(&karg->ca_client.cl_addrlist[0],
+                               &arg->ca32_client.cl32_addrlist[0],
+                               (sizeof(struct in_addr) * NFSCLNT_ADDRMAX)) ||
+               __get_user(karg->ca_client.cl_fhkeytype,
+                               &arg->ca32_client.cl32_fhkeytype) ||
+               __get_user(karg->ca_client.cl_fhkeylen,
+                               &arg->ca32_client.cl32_fhkeylen) ||
+               __copy_from_user(&karg->ca_client.cl_fhkey[0],
+                               &arg->ca32_client.cl32_fhkey[0],
+                               NFSCLNT_KEYMAX))
+               return -EFAULT;
 
-       return (err) ? -EFAULT : 0;
+       return 0;
 }
 
-static int compat_nfs_exp_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+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);
+       if (!access_ok(VERIFY_READ, &arg->ca32_export,
+                               sizeof(arg->ca32_export)) ||
+               get_user(karg->ca_version, &arg->ca32_version) ||
+               __copy_from_user(&karg->ca_export.ex_client[0],
+                               &arg->ca32_export.ex32_client[0],
+                               NFSCLNT_IDMAX) ||
+               __copy_from_user(&karg->ca_export.ex_path[0],
+                               &arg->ca32_export.ex32_path[0],
+                               NFS_MAXPATHLEN) ||
+               __get_user(karg->ca_export.ex_dev,
+                               &arg->ca32_export.ex32_dev) ||
+               __get_user(karg->ca_export.ex_ino,
+                               &arg->ca32_export.ex32_ino) ||
+               __get_user(karg->ca_export.ex_flags,
+                               &arg->ca32_export.ex32_flags) ||
+               __get_user(karg->ca_export.ex_anon_uid,
+                               &arg->ca32_export.ex32_anon_uid) ||
+               __get_user(karg->ca_export.ex_anon_gid,
+                               &arg->ca32_export.ex32_anon_gid))
+               return -EFAULT;
        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;
+       return 0;
 }
 
-static int compat_nfs_getfd_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+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);
+       if (!access_ok(VERIFY_READ, &arg->ca32_getfd,
+                       sizeof(arg->ca32_getfd)) ||
+               get_user(karg->ca_version, &arg->ca32_version) ||
+               __copy_from_user(&karg->ca_getfd.gd_addr,
+                               &arg->ca32_getfd.gd32_addr,
+                               (sizeof(struct sockaddr))) ||
+               __copy_from_user(&karg->ca_getfd.gd_path,
+                               &arg->ca32_getfd.gd32_path,
+                               (NFS_MAXPATHLEN+1)) ||
+               __get_user(karg->ca_getfd.gd_version,
+                               &arg->ca32_getfd.gd32_version))
+               return -EFAULT;
 
-       return (err) ? -EFAULT : 0;
+       return 0;
 }
 
-static int compat_nfs_getfs_trans(struct nfsctl_arg *karg, struct compat_nfsctl_arg __user *arg)
+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);
+       if (!access_ok(VERIFY_READ,&arg->ca32_getfs,sizeof(arg->ca32_getfs)) ||
+               get_user(karg->ca_version, &arg->ca32_version) ||
+               __copy_from_user(&karg->ca_getfs.gd_addr,
+                               &arg->ca32_getfs.gd32_addr,
+                               (sizeof(struct sockaddr))) ||
+               __copy_from_user(&karg->ca_getfs.gd_path,
+                               &arg->ca32_getfs.gd32_path,
+                               (NFS_MAXPATHLEN+1)) ||
+               __get_user(karg->ca_getfs.gd_maxlen,
+                               &arg->ca32_getfs.gd32_maxlen))
+               return -EFAULT;
 
-       return (err) ? -EFAULT : 0;
+       return 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)
+static int compat_nfs_getfh_res_trans(union nfsctl_res *kres,
+                               union compat_nfsctl_res __user *res)
 {
        int err;
 
@@ -1564,8 +2147,9 @@ static int compat_nfs_getfh_res_trans(union nfsctl_res *kres, union compat_nfsct
        return (err) ? -EFAULT : 0;
 }
 
-asmlinkage long compat_sys_nfsservctl(int cmd, struct compat_nfsctl_arg __user *arg,
-                                       union compat_nfsctl_res __user *res)
+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;
@@ -1607,9 +2191,12 @@ asmlinkage long compat_sys_nfsservctl(int cmd, struct compat_nfsctl_arg __user *
 
        default:
                err = -EINVAL;
-               goto done;
+               break;
        }
 
+       if (err)
+               goto done;
+
        oldfs = get_fs();
        set_fs(KERNEL_DS);
        /* The __user pointer casts are valid because of the set_fs() */