Re: [PATCH] barrier patch set

[Jens Axboe <axboe@suse.de>]

On Sat, Mar 20 2004, Bartlomiej Zolnierkiewicz wrote:
> On Saturday 20 of March 2004 18:10, Chris Mason wrote:
> > On Sat, 2004-03-20 at 12:05, Bartlomiej Zolnierkiewicz wrote:
> > > On Saturday 20 of March 2004 17:32, Chris Mason wrote:
> > > > On Sat, 2004-03-20 at 11:23, Bartlomiej Zolnierkiewicz wrote:
> > > > > > > - why are we doing pre-flush?
> > > >
> > > > The journaled filesystems need this.  We need to make sure that before
> > > > we write the commit block for a transaction, all the previous log
> > > > blocks we're written are safely on media.  Then we also need to make
> > > > sure the commit block is on media.
> > >
> > > For low-level driver it shouldn't really matter whether sectors to be
> > > written are the commit block for a transaction or the previous log blocks
> > > and in the current implementation it does matter.
> >
> > As Jens said, it depends on how you define barrier ;-)  I define it as
> > this io will be written after all the previous io and before any later
> > io.   It was originally written with scsi tags in mind as well, the FS
> > side was the same for both.
> 
> Yes, thanks for explaining this.
> 
> I took a quick look at fs/jbd/ and now I think I understand the way barriers
> currently work.  I assume that SCSI handles barriers by ordered tags, right?

That's the idea, yes. Needs some SCSI error handling work though, to
always ensure correct ordering.

-- 
Jens Axboe