ceph: clean up on forwarded aborted mds request
[safe/jmp/linux-2.6] / fs / ext4 / file.c
index 0b622c0..d0776e4 100644 (file)
 #include <linux/time.h>
 #include <linux/fs.h>
 #include <linux/jbd2.h>
-#include <linux/ext4_fs.h>
-#include <linux/ext4_jbd2.h>
+#include <linux/mount.h>
+#include <linux/path.h>
+#include <linux/quotaops.h>
+#include "ext4.h"
+#include "ext4_jbd2.h"
 #include "xattr.h"
 #include "acl.h"
 
  * from ext4_file_open: open gets called at every open, but release
  * gets called only when /all/ the files are closed.
  */
-static int ext4_release_file (struct inode * inode, struct file * filp)
+static int ext4_release_file(struct inode *inode, struct file *filp)
 {
+       if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE)) {
+               ext4_alloc_da_blocks(inode);
+               ext4_clear_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE);
+       }
        /* if we are the last writer on the inode, drop the block reservation */
        if ((filp->f_mode & FMODE_WRITE) &&
-                       (atomic_read(&inode->i_writecount) == 1))
+                       (atomic_read(&inode->i_writecount) == 1) &&
+                       !EXT4_I(inode)->i_reserved_data_blocks)
        {
-               mutex_lock(&EXT4_I(inode)->truncate_mutex);
-               ext4_discard_reservation(inode);
-               mutex_unlock(&EXT4_I(inode)->truncate_mutex);
+               down_write(&EXT4_I(inode)->i_data_sem);
+               ext4_discard_preallocations(inode);
+               up_write(&EXT4_I(inode)->i_data_sem);
        }
        if (is_dx(inode) && filp->private_data)
                ext4_htree_free_dir_info(filp->private_data);
@@ -51,59 +59,74 @@ static ssize_t
 ext4_file_write(struct kiocb *iocb, const struct iovec *iov,
                unsigned long nr_segs, loff_t pos)
 {
-       struct file *file = iocb->ki_filp;
-       struct inode *inode = file->f_dentry->d_inode;
-       ssize_t ret;
-       int err;
-
-       ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
+       struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
 
        /*
-        * Skip flushing if there was an error, or if nothing was written.
+        * If we have encountered a bitmap-format file, the size limit
+        * is smaller than s_maxbytes, which is for extent-mapped files.
         */
-       if (ret <= 0)
-               return ret;
 
-       /*
-        * If the inode is IS_SYNC, or is O_SYNC and we are doing data
-        * journalling then we need to make sure that we force the transaction
-        * to disk to keep all metadata uptodate synchronously.
-        */
-       if (file->f_flags & O_SYNC) {
-               /*
-                * If we are non-data-journaled, then the dirty data has
-                * already been flushed to backing store by generic_osync_inode,
-                * and the inode has been flushed too if there have been any
-                * modifications other than mere timestamp updates.
-                *
-                * Open question --- do we care about flushing timestamps too
-                * if the inode is IS_SYNC?
-                */
-               if (!ext4_should_journal_data(inode))
-                       return ret;
+       if (!(EXT4_I(inode)->i_flags & EXT4_EXTENTS_FL)) {
+               struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+               size_t length = iov_length(iov, nr_segs);
 
-               goto force_commit;
+               if (pos > sbi->s_bitmap_maxbytes)
+                       return -EFBIG;
+
+               if (pos + length > sbi->s_bitmap_maxbytes) {
+                       nr_segs = iov_shorten((struct iovec *)iov, nr_segs,
+                                             sbi->s_bitmap_maxbytes - pos);
+               }
        }
 
-       /*
-        * So we know that there has been no forced data flush.  If the inode
-        * is marked IS_SYNC, we need to force one ourselves.
-        */
-       if (!IS_SYNC(inode))
-               return ret;
+       return generic_file_aio_write(iocb, iov, nr_segs, pos);
+}
 
-       /*
-        * Open question #2 --- should we force data to disk here too?  If we
-        * don't, the only impact is that data=writeback filesystems won't
-        * flush data to disk automatically on IS_SYNC, only metadata (but
-        * historically, that is what ext2 has done.)
-        */
+static const struct vm_operations_struct ext4_file_vm_ops = {
+       .fault          = filemap_fault,
+       .page_mkwrite   = ext4_page_mkwrite,
+};
+
+static int ext4_file_mmap(struct file *file, struct vm_area_struct *vma)
+{
+       struct address_space *mapping = file->f_mapping;
+
+       if (!mapping->a_ops->readpage)
+               return -ENOEXEC;
+       file_accessed(file);
+       vma->vm_ops = &ext4_file_vm_ops;
+       vma->vm_flags |= VM_CAN_NONLINEAR;
+       return 0;
+}
 
-force_commit:
-       err = ext4_force_commit(inode->i_sb);
-       if (err)
-               return err;
-       return ret;
+static int ext4_file_open(struct inode * inode, struct file * filp)
+{
+       struct super_block *sb = inode->i_sb;
+       struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+       struct vfsmount *mnt = filp->f_path.mnt;
+       struct path path;
+       char buf[64], *cp;
+
+       if (unlikely(!(sbi->s_mount_flags & EXT4_MF_MNTDIR_SAMPLED) &&
+                    !(sb->s_flags & MS_RDONLY))) {
+               sbi->s_mount_flags |= EXT4_MF_MNTDIR_SAMPLED;
+               /*
+                * Sample where the filesystem has been mounted and
+                * store it in the superblock for sysadmin convenience
+                * when trying to sort through large numbers of block
+                * devices or filesystem images.
+                */
+               memset(buf, 0, sizeof(buf));
+               path.mnt = mnt;
+               path.dentry = mnt->mnt_root;
+               cp = d_path(&path, buf, sizeof(buf));
+               if (!IS_ERR(cp)) {
+                       memcpy(sbi->s_es->s_last_mounted, cp,
+                              sizeof(sbi->s_es->s_last_mounted));
+                       sb->s_dirt = 1;
+               }
+       }
+       return dquot_file_open(inode, filp);
 }
 
 const struct file_operations ext4_file_operations = {
@@ -112,28 +135,30 @@ const struct file_operations ext4_file_operations = {
        .write          = do_sync_write,
        .aio_read       = generic_file_aio_read,
        .aio_write      = ext4_file_write,
-       .ioctl          = ext4_ioctl,
+       .unlocked_ioctl = ext4_ioctl,
 #ifdef CONFIG_COMPAT
        .compat_ioctl   = ext4_compat_ioctl,
 #endif
-       .mmap           = generic_file_mmap,
-       .open           = generic_file_open,
+       .mmap           = ext4_file_mmap,
+       .open           = ext4_file_open,
        .release        = ext4_release_file,
        .fsync          = ext4_sync_file,
-       .sendfile       = generic_file_sendfile,
        .splice_read    = generic_file_splice_read,
        .splice_write   = generic_file_splice_write,
 };
 
-struct inode_operations ext4_file_inode_operations = {
+const struct inode_operations ext4_file_inode_operations = {
        .truncate       = ext4_truncate,
        .setattr        = ext4_setattr,
-#ifdef CONFIG_EXT4DEV_FS_XATTR
+       .getattr        = ext4_getattr,
+#ifdef CONFIG_EXT4_FS_XATTR
        .setxattr       = generic_setxattr,
        .getxattr       = generic_getxattr,
        .listxattr      = ext4_listxattr,
        .removexattr    = generic_removexattr,
 #endif
-       .permission     = ext4_permission,
+       .check_acl      = ext4_check_acl,
+       .fallocate      = ext4_fallocate,
+       .fiemap         = ext4_fiemap,
 };