ext4: Use new framework for data=ordered mode in JBD2
authorJan Kara <jack@suse.cz>
Fri, 11 Jul 2008 23:27:31 +0000 (19:27 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Fri, 11 Jul 2008 23:27:31 +0000 (19:27 -0400)
This patch makes ext4 use inode-based implementation of data=ordered mode
in JBD2. It allows us to unify some data=ordered and data=writeback paths
(especially writepage since we don't have to start a transaction anymore)
and remove some buffer walking.

Updated fix from Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
to fix file system hang due to corrupt jinode values.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/ext4_i.h
fs/ext4/ext4_jbd2.h
fs/ext4/inode.c
fs/ext4/super.c

index abf2744164e07202d64acc462910c8f65cdc776d..c2903ef72159f0d435127e3dfc3703dff701eabd 100644 (file)
@@ -150,6 +150,7 @@ struct ext4_inode_info {
         */
        struct rw_semaphore i_data_sem;
        struct inode vfs_inode;
+       struct jbd2_inode jinode;
 
        unsigned long i_ext_generation;
        struct ext4_ext_cache i_cached_extent;
index d0aa9ee20f88fc999f5e009864b96b5a9005ecb3..eb8bc3afe6e9f590f491bf9098980f39f27bf245 100644 (file)
@@ -154,8 +154,6 @@ int __ext4_journal_dirty_metadata(const char *where,
 #define ext4_journal_forget(handle, bh) \
        __ext4_journal_forget(__func__, (handle), (bh))
 
-int ext4_journal_dirty_data(handle_t *handle, struct buffer_head *bh);
-
 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks);
 int __ext4_journal_stop(const char *where, handle_t *handle);
 
@@ -192,6 +190,11 @@ static inline int ext4_journal_force_commit(journal_t *journal)
        return jbd2_journal_force_commit(journal);
 }
 
+static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
+{
+       return jbd2_journal_file_inode(handle, &EXT4_I(inode)->jinode);
+}
+
 /* super.c */
 int ext4_force_commit(struct super_block *sb);
 
index 320acb6c35bfc356243b30a3ff8f72e66d4b913d..7b9569179fdf51041e08e4982189dea26d1ad6cf 100644 (file)
 #include "xattr.h"
 #include "acl.h"
 
+static inline int ext4_begin_ordered_truncate(struct inode *inode,
+                                             loff_t new_size)
+{
+       return jbd2_journal_begin_ordered_truncate(&EXT4_I(inode)->jinode,
+                                                  new_size);
+}
+
 /*
  * Test whether an inode is a fast symlink.
  */
@@ -181,6 +188,8 @@ void ext4_delete_inode (struct inode * inode)
 {
        handle_t *handle;
 
+       if (ext4_should_order_data(inode))
+               ext4_begin_ordered_truncate(inode, 0);
        truncate_inode_pages(&inode->i_data, 0);
 
        if (is_bad_inode(inode))
@@ -1273,15 +1282,6 @@ out:
        return ret;
 }
 
-int ext4_journal_dirty_data(handle_t *handle, struct buffer_head *bh)
-{
-       int err = jbd2_journal_dirty_data(handle, bh);
-       if (err)
-               ext4_journal_abort_handle(__func__, __func__,
-                                               bh, handle, err);
-       return err;
-}
-
 /* For write_end() in data=journal mode */
 static int write_end_fn(handle_t *handle, struct buffer_head *bh)
 {
@@ -1311,8 +1311,7 @@ static int ext4_ordered_write_end(struct file *file,
        from = pos & (PAGE_CACHE_SIZE - 1);
        to = from + len;
 
-       ret = walk_page_buffers(handle, page_buffers(page),
-               from, to, NULL, ext4_journal_dirty_data);
+       ret = ext4_jbd2_file_inode(handle, inode);
 
        if (ret == 0) {
                /*
@@ -1472,25 +1471,22 @@ static int bput_one(handle_t *handle, struct buffer_head *bh)
        return 0;
 }
 
-static int jbd2_journal_dirty_data_fn(handle_t *handle, struct buffer_head *bh)
-{
-       if (buffer_mapped(bh))
-               return ext4_journal_dirty_data(handle, bh);
-       return 0;
-}
-
 static int ext4_bh_unmapped_or_delay(handle_t *handle, struct buffer_head *bh)
 {
        return !buffer_mapped(bh) || buffer_delay(bh);
 }
 
 /*
- * Note that we don't need to start a transaction unless we're journaling
- * data because we should have holes filled from ext4_page_mkwrite(). If
- * we are journaling data, we cannot start transaction directly because
- * transaction start ranks above page lock so we have to do some magic...
+ * Note that we don't need to start a transaction unless we're journaling data
+ * because we should have holes filled from ext4_page_mkwrite(). We even don't
+ * need to file the inode to the transaction's list in ordered mode because if
+ * we are writing back data added by write(), the inode is already there and if
+ * we are writing back data modified via mmap(), noone guarantees in which
+ * transaction the data will hit the disk. In case we are journaling data, we
+ * cannot start transaction directly because transaction start ranks above page
+ * lock so we have to do some magic.
  *
- * In all journalling modes block_write_full_page() will start the I/O.
+ * In all journaling modes block_write_full_page() will start the I/O.
  *
  * Problem:
  *
@@ -1533,86 +1529,7 @@ static int ext4_bh_unmapped_or_delay(handle_t *handle, struct buffer_head *bh)
  * us.
  *
  */
-static int __ext4_ordered_writepage(struct page *page,
-                               struct writeback_control *wbc)
-{
-       struct inode *inode = page->mapping->host;
-       struct buffer_head *page_bufs;
-       handle_t *handle = NULL;
-       int ret = 0;
-       int err;
-
-       if (!page_has_buffers(page)) {
-               create_empty_buffers(page, inode->i_sb->s_blocksize,
-                               (1 << BH_Dirty)|(1 << BH_Uptodate));
-       }
-       page_bufs = page_buffers(page);
-       walk_page_buffers(handle, page_bufs, 0,
-                       PAGE_CACHE_SIZE, NULL, bget_one);
-
-       ret = block_write_full_page(page, ext4_get_block, wbc);
-
-       /*
-        * The page can become unlocked at any point now, and
-        * truncate can then come in and change things.  So we
-        * can't touch *page from now on.  But *page_bufs is
-        * safe due to elevated refcount.
-        */
-
-       /*
-        * And attach them to the current transaction.  But only if
-        * block_write_full_page() succeeded.  Otherwise they are unmapped,
-        * and generally junk.
-        */
-       if (ret == 0) {
-               handle = ext4_journal_start(inode,
-                                       ext4_writepage_trans_blocks(inode));
-               if (IS_ERR(handle)) {
-                       ret = PTR_ERR(handle);
-                       goto out_put;
-               }
-
-               ret = walk_page_buffers(handle, page_bufs, 0, PAGE_CACHE_SIZE,
-                                       NULL, jbd2_journal_dirty_data_fn);
-               err = ext4_journal_stop(handle);
-               if (!ret)
-                       ret = err;
-       }
-out_put:
-       walk_page_buffers(handle, page_bufs, 0, PAGE_CACHE_SIZE, NULL,
-                         bput_one);
-       return ret;
-}
-
-static int ext4_ordered_writepage(struct page *page,
-                               struct writeback_control *wbc)
-{
-       struct inode *inode = page->mapping->host;
-       loff_t size = i_size_read(inode);
-       loff_t len;
-
-       J_ASSERT(PageLocked(page));
-       J_ASSERT(page_has_buffers(page));
-       if (page->index == size >> PAGE_CACHE_SHIFT)
-               len = size & ~PAGE_CACHE_MASK;
-       else
-               len = PAGE_CACHE_SIZE;
-       BUG_ON(walk_page_buffers(NULL, page_buffers(page), 0, len, NULL,
-                                ext4_bh_unmapped_or_delay));
-
-       /*
-        * We give up here if we're reentered, because it might be for a
-        * different filesystem.
-        */
-       if (!ext4_journal_current_handle())
-               return __ext4_ordered_writepage(page, wbc);
-
-       redirty_page_for_writepage(wbc, page);
-       unlock_page(page);
-       return 0;
-}
-
-static int __ext4_writeback_writepage(struct page *page,
+static int __ext4_normal_writepage(struct page *page,
                                struct writeback_control *wbc)
 {
        struct inode *inode = page->mapping->host;
@@ -1624,7 +1541,7 @@ static int __ext4_writeback_writepage(struct page *page,
 }
 
 
-static int ext4_writeback_writepage(struct page *page,
+static int ext4_normal_writepage(struct page *page,
                                struct writeback_control *wbc)
 {
        struct inode *inode = page->mapping->host;
@@ -1641,7 +1558,7 @@ static int ext4_writeback_writepage(struct page *page,
                                 ext4_bh_unmapped_or_delay));
 
        if (!ext4_journal_current_handle())
-               return __ext4_writeback_writepage(page, wbc);
+               return __ext4_normal_writepage(page, wbc);
 
        redirty_page_for_writepage(wbc, page);
        unlock_page(page);
@@ -1877,7 +1794,7 @@ static int ext4_journalled_set_page_dirty(struct page *page)
 static const struct address_space_operations ext4_ordered_aops = {
        .readpage       = ext4_readpage,
        .readpages      = ext4_readpages,
-       .writepage      = ext4_ordered_writepage,
+       .writepage      = ext4_normal_writepage,
        .sync_page      = block_sync_page,
        .write_begin    = ext4_write_begin,
        .write_end      = ext4_ordered_write_end,
@@ -1891,7 +1808,7 @@ static const struct address_space_operations ext4_ordered_aops = {
 static const struct address_space_operations ext4_writeback_aops = {
        .readpage       = ext4_readpage,
        .readpages      = ext4_readpages,
-       .writepage      = ext4_writeback_writepage,
+       .writepage      = ext4_normal_writepage,
        .sync_page      = block_sync_page,
        .write_begin    = ext4_write_begin,
        .write_end      = ext4_writeback_write_end,
@@ -2019,7 +1936,7 @@ int ext4_block_truncate_page(handle_t *handle,
                err = ext4_journal_dirty_metadata(handle, bh);
        } else {
                if (ext4_should_order_data(inode))
-                       err = ext4_journal_dirty_data(handle, bh);
+                       err = ext4_jbd2_file_inode(handle, inode);
                mark_buffer_dirty(bh);
        }
 
@@ -3171,7 +3088,14 @@ int ext4_write_inode(struct inode *inode, int wait)
  * be freed, so we have a strong guarantee that no future commit will
  * leave these blocks visible to the user.)
  *
- * Called with inode->sem down.
+ * Another thing we have to assure is that if we are in ordered mode
+ * and inode is still attached to the committing transaction, we must
+ * we start writeout of all the dirty pages which are being truncated.
+ * This way we are sure that all the data written in the previous
+ * transaction are already on disk (truncate waits for pages under
+ * writeback).
+ *
+ * Called with inode->i_mutex down.
  */
 int ext4_setattr(struct dentry *dentry, struct iattr *attr)
 {
@@ -3237,6 +3161,22 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
                if (!error)
                        error = rc;
                ext4_journal_stop(handle);
+
+               if (ext4_should_order_data(inode)) {
+                       error = ext4_begin_ordered_truncate(inode,
+                                                           attr->ia_size);
+                       if (error) {
+                               /* Do as much error cleanup as possible */
+                               handle = ext4_journal_start(inode, 3);
+                               if (IS_ERR(handle)) {
+                                       ext4_orphan_del(NULL, inode);
+                                       goto err_out;
+                               }
+                               ext4_orphan_del(handle, inode);
+                               ext4_journal_stop(handle);
+                               goto err_out;
+                       }
+               }
        }
 
        rc = inode_setattr(inode, attr);
index 1b330cd71ca842b44b53adc350cd65bcbdef188d..629d0fa27e3a1a0c20f56f564c3734c35d2ee9b4 100644 (file)
@@ -573,6 +573,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
        memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
        INIT_LIST_HEAD(&ei->i_prealloc_list);
        spin_lock_init(&ei->i_prealloc_lock);
+       jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
        return &ei->vfs_inode;
 }
 
@@ -637,6 +638,8 @@ static void ext4_clear_inode(struct inode *inode)
        EXT4_I(inode)->i_block_alloc_info = NULL;
        if (unlikely(rsv))
                kfree(rsv);
+       jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
+                                      &EXT4_I(inode)->jinode);
 }
 
 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
@@ -3378,7 +3381,7 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type,
                        err = ext4_journal_dirty_metadata(handle, bh);
                else {
                        /* Always do at least ordered writes for quotas */
-                       err = ext4_journal_dirty_data(handle, bh);
+                       err = ext4_jbd2_file_inode(handle, inode);
                        mark_buffer_dirty(bh);
                }
                brelse(bh);