public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: npiggin@kernel.dk
To: linux-fsdevel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Subject: [patch 5/7] fs: ext2 inode sync fix
Date: Wed, 24 Nov 2010 01:06:15 +1100	[thread overview]
Message-ID: <20101123140707.991528524@kernel.dk> (raw)
In-Reply-To: 20101123140610.292941494@kernel.dk

[-- Attachment #1: ext2-sync-fixes.patch --]
[-- Type: text/plain, Size: 4142 bytes --]

There is a big fuckup with inode metadata writeback (I suspect in a lot
of filesystems, but I've only dared to look at a couple as yet).

ext2 relies on ->write_inode being called from sync_inode_metadata in
fsync in order to sync the inode. However I_DIRTY_SYNC gets cleared
after a call to this guy, and it doesn't actually write back and wait
on the inode block unless it is called for sync. This means that write_inode
from background writeback can kill the inode dirty bits without the data
getting to disk. Fsync will subsequently miss it.

The fix is for ->write_inode to dirty the buffers/cache, and then ->fsync to
write back the dirty data. In the full filesystem sync case, buffercache
writeback in the generic code will write back the dirty data. Other
filesystems could use ->sync_fs for this.

Signed-off-by: Nick Piggin <npiggin@kernel.dk>

Index: linux-2.6/fs/ext2/inode.c
===================================================================
--- linux-2.6.orig/fs/ext2/inode.c	2010-11-23 21:56:05.000000000 +1100
+++ linux-2.6/fs/ext2/inode.c	2010-11-23 22:05:02.000000000 +1100
@@ -1211,7 +1211,7 @@ static int ext2_setsize(struct inode *in
 	return 0;
 }
 
-static struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
+struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
 					struct buffer_head **p)
 {
 	struct buffer_head * bh;
@@ -1505,16 +1505,8 @@ static int __ext2_write_inode(struct ino
 	} else for (n = 0; n < EXT2_N_BLOCKS; n++)
 		raw_inode->i_block[n] = ei->i_data[n];
 	mark_buffer_dirty(bh);
-	if (do_sync) {
-		sync_dirty_buffer(bh);
-		if (buffer_req(bh) && !buffer_uptodate(bh)) {
-			printk ("IO error syncing ext2 inode [%s:%08lx]\n",
-				sb->s_id, (unsigned long) ino);
-			err = -EIO;
-		}
-	}
-	ei->i_state &= ~EXT2_STATE_NEW;
 	brelse (bh);
+	ei->i_state &= ~EXT2_STATE_NEW;
 	return err;
 }
 
Index: linux-2.6/fs/ext2/file.c
===================================================================
--- linux-2.6.orig/fs/ext2/file.c	2010-11-23 21:56:05.000000000 +1100
+++ linux-2.6/fs/ext2/file.c	2010-11-23 22:05:02.000000000 +1100
@@ -21,6 +21,7 @@
 #include <linux/time.h>
 #include <linux/pagemap.h>
 #include <linux/quotaops.h>
+#include <linux/buffer_head.h>
 #include "ext2.h"
 #include "xattr.h"
 #include "acl.h"
@@ -43,16 +44,33 @@ static int ext2_release_file (struct ino
 int ext2_fsync(struct file *file, int datasync)
 {
 	int ret;
-	struct super_block *sb = file->f_mapping->host->i_sb;
-	struct address_space *mapping = sb->s_bdev->bd_inode->i_mapping;
+	struct inode *inode = file->f_mapping->host;
+	ino_t ino = inode->i_ino;
+	struct super_block *sb = inode->i_sb;
+	struct address_space *sb_mapping = sb->s_bdev->bd_inode->i_mapping;
+	struct buffer_head *bh;
+	struct ext2_inode *raw_inode;
 
 	ret = generic_file_fsync(file, datasync);
-	if (ret == -EIO || test_and_clear_bit(AS_EIO, &mapping->flags)) {
+	if (ret == -EIO || test_and_clear_bit(AS_EIO, &sb_mapping->flags)) {
 		/* We don't really know where the IO error happened... */
 		ext2_error(sb, __func__,
 			   "detected IO error when writing metadata buffers");
+		return -EIO;
+	}
+
+	raw_inode = ext2_get_inode(sb, ino, &bh);
+	if (IS_ERR(raw_inode))
+ 		return -EIO;
+
+	sync_dirty_buffer(bh);
+	if (buffer_req(bh) && !buffer_uptodate(bh)) {
+		printk ("IO error syncing ext2 inode [%s:%08lx]\n",
+			sb->s_id, (unsigned long) ino);
 		ret = -EIO;
 	}
+	brelse (bh);
+
 	return ret;
 }
 
Index: linux-2.6/fs/ext2/ext2.h
===================================================================
--- linux-2.6.orig/fs/ext2/ext2.h	2010-11-23 21:56:05.000000000 +1100
+++ linux-2.6/fs/ext2/ext2.h	2010-11-23 22:05:02.000000000 +1100
@@ -124,6 +124,8 @@ extern int ext2_get_block(struct inode *
 extern int ext2_setattr (struct dentry *, struct iattr *);
 extern void ext2_set_inode_flags(struct inode *inode);
 extern void ext2_get_inode_flags(struct ext2_inode_info *);
+extern struct ext2_inode *ext2_get_inode(struct super_block *sb, ino_t ino,
+					struct buffer_head **p);
 extern int ext2_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		       u64 start, u64 len);
 



  parent reply	other threads:[~2010-11-23 14:12 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-11-23 14:06 [patch 0/7] icache dirty / sync fixes npiggin
2010-11-23 14:06 ` [patch 1/7] fs: mark_inode_dirty barrier fix npiggin
2010-11-23 14:06 ` [patch 2/7] fs: simple fsync race fix npiggin
2010-11-29 14:58   ` Christoph Hellwig
2010-11-30  0:05     ` Nick Piggin
2010-11-23 14:06 ` [patch 3/7] fs: introduce inode writeback helpers npiggin
2010-11-29 15:13   ` Christoph Hellwig
2010-11-30  0:22     ` Nick Piggin
2010-11-23 14:06 ` [patch 4/7] fs: preserve inode dirty bits on failed metadata writeback npiggin
2010-11-29 14:59   ` Christoph Hellwig
2010-11-30  0:08     ` Nick Piggin
2010-11-23 14:06 ` npiggin [this message]
2010-11-30 11:26   ` [patch 5/7] fs: ext2 inode sync fix Boaz Harrosh
2010-11-23 14:06 ` [patch 6/7] fs: fsync optimisations npiggin
2010-11-29 15:03   ` Christoph Hellwig
2010-11-30  0:11     ` Nick Piggin
2010-11-23 14:06 ` [patch 7/7] fs: fix or note I_DIRTY handling bugs in filesystems npiggin
2010-11-23 15:04   ` Steven Whitehouse
2010-11-23 22:51   ` Dave Chinner
2010-11-24  0:23     ` Nick Piggin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20101123140707.991528524@kernel.dk \
    --to=npiggin@kernel.dk \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox