nfsd: move most of nfsfh.h to fs/nfsd
[safe/jmp/linux-2.6] / fs / ext4 / ioctl.c
index 25b13ed..c1cdf61 100644 (file)
 #include <linux/fs.h>
 #include <linux/jbd2.h>
 #include <linux/capability.h>
-#include <linux/ext4_fs.h>
-#include <linux/ext4_jbd2.h>
 #include <linux/time.h>
 #include <linux/compat.h>
-#include <linux/smp_lock.h>
 #include <linux/mount.h>
+#include <linux/file.h>
 #include <asm/uaccess.h>
+#include "ext4_jbd2.h"
+#include "ext4.h"
 
-int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
-               unsigned long arg)
+long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
+       struct inode *inode = filp->f_dentry->d_inode;
        struct ext4_inode_info *ei = EXT4_I(inode);
        unsigned int flags;
-       unsigned short rsv_window_size;
 
-       ext4_debug ("cmd = %u, arg = %lu\n", cmd, arg);
+       ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
 
        switch (cmd) {
        case EXT4_IOC_GETFLAGS:
@@ -34,7 +33,7 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
                return put_user(flags, (int __user *) arg);
        case EXT4_IOC_SETFLAGS: {
                handle_t *handle = NULL;
-               int err;
+               int err, migrate = 0;
                struct ext4_iloc iloc;
                unsigned int oldflags;
                unsigned int jflag;
@@ -49,8 +48,7 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
                if (err)
                        return err;
 
-               if (!S_ISDIR(inode->i_mode))
-                       flags &= ~EXT4_DIRSYNC_FL;
+               flags = ext4_mask_flags(inode->i_mode, flags);
 
                err = -EPERM;
                mutex_lock(&inode->i_mutex);
@@ -82,6 +80,17 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
                        if (!capable(CAP_SYS_RESOURCE))
                                goto flags_out;
                }
+               if (oldflags & EXT4_EXTENTS_FL) {
+                       /* We don't support clearning extent flags */
+                       if (!(flags & EXT4_EXTENTS_FL)) {
+                               err = -EOPNOTSUPP;
+                               goto flags_out;
+                       }
+               } else if (flags & EXT4_EXTENTS_FL) {
+                       /* migrate the file */
+                       migrate = 1;
+                       flags &= ~EXT4_EXTENTS_FL;
+               }
 
                handle = ext4_journal_start(inode, 1);
                if (IS_ERR(handle)) {
@@ -89,7 +98,7 @@ int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd,
                        goto flags_out;
                }
                if (IS_SYNC(inode))
-                       handle->h_sync = 1;
+                       ext4_handle_sync(handle);
                err = ext4_reserve_inode_write(handle, inode, &iloc);
                if (err)
                        goto flags_err;
@@ -109,6 +118,10 @@ flags_err:
 
                if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL))
                        err = ext4_change_inode_journal_flag(inode, jflag);
+               if (err)
+                       goto flags_out;
+               if (migrate)
+                       err = ext4_ext_migrate(inode);
 flags_out:
                mutex_unlock(&inode->i_mutex);
                mnt_drop_write(filp->f_path.mnt);
@@ -175,53 +188,10 @@ setversion_out:
                        return ret;
                }
 #endif
-       case EXT4_IOC_GETRSVSZ:
-               if (test_opt(inode->i_sb, RESERVATION)
-                       && S_ISREG(inode->i_mode)
-                       && ei->i_block_alloc_info) {
-                       rsv_window_size = ei->i_block_alloc_info->rsv_window_node.rsv_goal_size;
-                       return put_user(rsv_window_size, (int __user *)arg);
-               }
-               return -ENOTTY;
-       case EXT4_IOC_SETRSVSZ: {
-               int err;
-
-               if (!test_opt(inode->i_sb, RESERVATION) ||!S_ISREG(inode->i_mode))
-                       return -ENOTTY;
-
-               if (!is_owner_or_cap(inode))
-                       return -EACCES;
-
-               if (get_user(rsv_window_size, (int __user *)arg))
-                       return -EFAULT;
-
-               err = mnt_want_write(filp->f_path.mnt);
-               if (err)
-                       return err;
-
-               if (rsv_window_size > EXT4_MAX_RESERVE_BLOCKS)
-                       rsv_window_size = EXT4_MAX_RESERVE_BLOCKS;
-
-               /*
-                * need to allocate reservation structure for this inode
-                * before set the window size
-                */
-               down_write(&ei->i_data_sem);
-               if (!ei->i_block_alloc_info)
-                       ext4_init_block_alloc_info(inode);
-
-               if (ei->i_block_alloc_info){
-                       struct ext4_reserve_window_node *rsv = &ei->i_block_alloc_info->rsv_window_node;
-                       rsv->rsv_goal_size = rsv_window_size;
-               }
-               up_write(&ei->i_data_sem);
-               mnt_drop_write(filp->f_path.mnt);
-               return 0;
-       }
        case EXT4_IOC_GROUP_EXTEND: {
                ext4_fsblk_t n_blocks_count;
                struct super_block *sb = inode->i_sb;
-               int err;
+               int err, err2=0;
 
                if (!capable(CAP_SYS_RESOURCE))
                        return -EPERM;
@@ -234,17 +204,55 @@ setversion_out:
                        return err;
 
                err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
-               jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
-               jbd2_journal_flush(EXT4_SB(sb)->s_journal);
-               jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
+               if (EXT4_SB(sb)->s_journal) {
+                       jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
+                       err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
+                       jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
+               }
+               if (err == 0)
+                       err = err2;
                mnt_drop_write(filp->f_path.mnt);
 
                return err;
        }
+
+       case EXT4_IOC_MOVE_EXT: {
+               struct move_extent me;
+               struct file *donor_filp;
+               int err;
+
+               if (copy_from_user(&me,
+                       (struct move_extent __user *)arg, sizeof(me)))
+                       return -EFAULT;
+
+               donor_filp = fget(me.donor_fd);
+               if (!donor_filp)
+                       return -EBADF;
+
+               if (!capable(CAP_DAC_OVERRIDE)) {
+                       if ((current->real_cred->fsuid != inode->i_uid) ||
+                               !(inode->i_mode & S_IRUSR) ||
+                               !(donor_filp->f_dentry->d_inode->i_mode &
+                               S_IRUSR)) {
+                               fput(donor_filp);
+                               return -EACCES;
+                       }
+               }
+
+               err = ext4_move_extents(filp, donor_filp, me.orig_start,
+                                       me.donor_start, me.len, &me.moved_len);
+               fput(donor_filp);
+
+               if (copy_to_user((struct move_extent *)arg, &me, sizeof(me)))
+                       return -EFAULT;
+
+               return err;
+       }
+
        case EXT4_IOC_GROUP_ADD: {
                struct ext4_new_group_data input;
                struct super_block *sb = inode->i_sb;
-               int err;
+               int err, err2=0;
 
                if (!capable(CAP_SYS_RESOURCE))
                        return -EPERM;
@@ -258,16 +266,53 @@ setversion_out:
                        return err;
 
                err = ext4_group_add(sb, &input);
-               jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
-               jbd2_journal_flush(EXT4_SB(sb)->s_journal);
-               jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
+               if (EXT4_SB(sb)->s_journal) {
+                       jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
+                       err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
+                       jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
+               }
+               if (err == 0)
+                       err = err2;
                mnt_drop_write(filp->f_path.mnt);
 
                return err;
        }
 
        case EXT4_IOC_MIGRATE:
-               return ext4_ext_migrate(inode, filp, cmd, arg);
+       {
+               int err;
+               if (!is_owner_or_cap(inode))
+                       return -EACCES;
+
+               err = mnt_want_write(filp->f_path.mnt);
+               if (err)
+                       return err;
+               /*
+                * inode_mutex prevent write and truncate on the file.
+                * Read still goes through. We take i_data_sem in
+                * ext4_ext_swap_inode_data before we switch the
+                * inode format to prevent read.
+                */
+               mutex_lock(&(inode->i_mutex));
+               err = ext4_ext_migrate(inode);
+               mutex_unlock(&(inode->i_mutex));
+               mnt_drop_write(filp->f_path.mnt);
+               return err;
+       }
+
+       case EXT4_IOC_ALLOC_DA_BLKS:
+       {
+               int err;
+               if (!is_owner_or_cap(inode))
+                       return -EACCES;
+
+               err = mnt_want_write(filp->f_path.mnt);
+               if (err)
+                       return err;
+               err = ext4_alloc_da_blocks(inode);
+               mnt_drop_write(filp->f_path.mnt);
+               return err;
+       }
 
        default:
                return -ENOTTY;
@@ -277,9 +322,6 @@ setversion_out:
 #ifdef CONFIG_COMPAT
 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
-       struct inode *inode = file->f_path.dentry->d_inode;
-       int ret;
-
        /* These are just misnamed, they actually get/put from/to user an int */
        switch (cmd) {
        case EXT4_IOC32_GETFLAGS:
@@ -319,9 +361,6 @@ long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
        default:
                return -ENOIOCTLCMD;
        }
-       lock_kernel();
-       ret = ext4_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
-       unlock_kernel();
-       return ret;
+       return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
 }
 #endif