tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Proposal: B_ARRIER (addresses wapbl performance?)

On Tue, Dec 09, 2008 at 03:34:17PM -0800, Jason Thorpe wrote:
> Journal writes really only need to be constrained with respect to the  
> metadata they're journaling, and themselves (i.e. entry hits oxide  
> before

or after, see joerg's mail

> the header update that makes it valid does). 
> Doesn't matter  
> for data writes (assuming you're not journaling data, which we don't).

We're not journaling data yet, but I guess we'll want to do it in the
future (this is a nice feature too). We have to account for this too.
And for data, the write cache being on or off doesn't make much difference
when we have tagged queuing. Where it does is for multiple writes
at the same place (where the write-back cache can aggregate them before
sending to platters), and this is typically the case where we would use
FUA ...

Manuel Bouyer, LIP6, Universite Paris VI.  
     NetBSD: 26 ans d'experience feront toujours la difference

Home | Main Index | Thread Index | Old Index