Re: ext4: Can we talk about bforget() and metadata blocks

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

On Thu, Sep 10, 2009 at 08:46:41AM -0700, Curt Wohlgemuth wrote:
> On Wed, Sep 9, 2009 at 11:54 PM, Aneesh Kumar
> K.V<aneesh.kumar@linux.vnet.ibm.com> wrote:
> > On Wed, Sep 09, 2009 at 09:35:40PM -0400, Theodore Tso wrote:
> >> On Wed, Sep 09, 2009 at 05:07:28PM -0700, Curt Wohlgemuth wrote:
> >> >
> >> > First, ext4_journal_forget() is called from ext4_forget() only when
> >> > we're journalling; without a journal, ext4_journal_forget() is only
> >> > called for various non-extent paths.  ext4_forget() could be changed,
> >> > of course...
> >>
> >> Ext4_forget() calls either ext4_journal_forget() or
> >> ext4_journal_revoke().  So we need to fix up both functions.
> >>
> >>                                         - Ted
> >>
> >> commit 4afdf0958f6f7b878e6d85cb4e0c0c12a0bd74e2
> >> Author: Theodore Ts'o <tytso@mit.edu>
> >> Date:   Wed Sep 9 21:32:41 2009 -0400
> >>
> >>     ext4: Use bforget() in no journal mode for ext4_journal_{forget,revoke}()
> >>
> >>     When ext4 is using a journal, a metadata block which is deallocated
> >>     must be passed into the journal layer so it can be dropped from the
> >>     current transaction and/or revoked.  This is done by calling the
> >>     functions ext4_journal_forget() and ext4_journal_revoke(), which call
> >>     jbd2_journal_forget(), and jbd2_journal_revoke(), respectively.
> >>
> >>     Since the jbd2_journal_forget() and jbd2_journal_revoke() call
> >>     bforget(), if ext4 is not using a journal, ext4_journal_forget() and
> >>     ext4_journal_revoke() must call bforget() to avoid a dirty metadata
> >>     block overwriting a block after it has been reallocated and reused for
> >>     another inode's data block.
> >>
> >
> > I am sure i am missing something. But where are we adding the buffer_head
> > to the mapping->private_list ?. For ext2 when we allocate meta data blocks
> > we do mark_buffer_dirty_inode which add the buffer_head to the inodes
> > private_list. Shouldn't we do something similar with Ext4 without journal ?
> 
> As Ted explained to me, all buffer heads pointing to metadata blocks
> are attached to the block device inode.  So pdflush writes of these
> pages go through the block device address space ops.  Explicit
> sync_dirty_buffer() calls for the metadata buffer heads still work, of
> course.

But how would it work for fsync ? I mean 

I would expect for no journal mode ext4_sync_file  should be doing
simple_fsync(). That should be forcing the metadata buffer_heads
via sync_mapping_buffers. And if we reuse these meta buffers we
drop them the inode->mapping->private_list using bforget.

But I don't see any of the above in code

-aneesh
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html