Re: Proposal: B_ARRIER (addresses wapbl performance?)

[Thor Lancelot Simon <tls%rek.tjls.com@localhost>]

On Sun, Nov 02, 2008 at 07:57:37AM -0800, Bill Stouder-Studenmund wrote:
> 
> So Thor, why are you so entrenched in this? If you're going to add a bit, 
> just add FUA. It does exactly what you want. It was designed to do what 
> you want.

We can't enqueue new commands while waiting for a cache flush to complete.
We *can* enqueue new commands while waiting for an ordered tag (and all
its prior simple tags) to complete.  This seems like a huge advantage to
me.  I also see other uses for disk write barriers beyond just solving
WAPBL's immediate problem.

The other thing is that if you look at other journalled filesystems
(particularly XFS) they manage to use write barriers to not do every
journal write synchronously (okay, pseudo-synchronously).  Given that we
don't support an external journal this seems particularly interesting to
me.

Also, it is clear to me how to implement B_BARRIER -- I know exactly what
to modify and how, except for the disk sorting code, which I'm reasonably
confident I can figure out.  I do not know how to implement FUA for each
kind of disk in the system -- if you do, by all means, be my guest!

Thor