lkml.org 
[lkml]   [2006]   [Mar]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: ext3_ordered_writepage() questions
> On Thu, 2006-03-16 at 16:04 -0500, Theodore Ts'o wrote:
> > On Thu, Mar 16, 2006 at 10:22:40AM -0800, Badari Pulavarty wrote:
> > > > However, if what we are doing is overwriting our own data with more an
> > > > updated, more recent version of the data block, do we guarantee that
> > > > any ordering semantics apply? For example, what if we write a data
> > > > block, and then follow it up with some kind of metadata update (say we
> > > > touch atime, or add an extended attribute). Do we guarantee that if
> > > > the metadata update is committed, that the data block will have made
> > > > it to disk as well?
> > >
> > > I don't see how we do this today. Yes. Metadata updates are jounalled,
> > > but I don't see how current adding buffers through journal_dirty_data
> > > (bh) call can guarantee that these buffers get added to metadata-update
> > > transaction ?
> >
> > Even though there aren't any updates to any metadata blocks that take
> > place between the journal_start() and journal_stop() calls, if
> > journal_dirty_data() is called (for example in ordered_writepage),
> > those buffers will be associated with the currently open transaction,
> > so they will be guaranteed to be written before the transaction is
> > allowed to commit.
> >
> > Remember, journal_start and journal_stop do not delineate a full
> > ext3/jbd transaction, but rather an operation, where a large number of
> > operations are bundled together to form a transaction. When you call
> > journal_start, and request a certain number of credits (number of
> > buffers that you maximally intend to dirty), that opens up an
> > operation. If the operation turns out not to dirty any metadata
> > blocks at the time of journal_stop(), all of the credits that were
> > reserved by jouranl_start() are returned to the currently open
> > transaction. However, any data blocks which are marked via
> > journal_dirty_data() are still going to be associated with the
> > currently open transaction, and they will still be forced out before
> > the transaction is allowed to commit.
> >
> > Does that make sense?
>
> Makes perfect sense, except that it doesn't match what I see through
> "debugfs" - logdump :(
>
> I wrote a testcase to re-write same blocks again & again - there is
> absolutely nothing showed up in log. Which implied to me that, all
> the jorunal_dirty_data we did on all those buffers, did nothing -
> since there is no current transaction. What am I missing ?
The data buffers are not journaled. The buffers are just attached to the
transaction and when the transaction is committed, they are written
directly to their final location. This ensures the ordering but no data
goes via the log... I guess you should see empty transactions in the log
which are eventually commited when they become too old.

Honza

--
Jan Kara <jack@suse.cz>
SuSE CR Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-03-16 23:08    [W:0.083 / U:0.508 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site