Subject: Re: letting userland issue FUA writes
To: Thor Lancelot Simon <>
From: Bill Studenmund <>
List: tech-kern
Date: 03/17/2006 15:34:09
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Fri, Mar 17, 2006 at 04:37:23PM -0500, Thor Lancelot Simon wrote:
> On Fri, Mar 17, 2006 at 10:34:11PM +0100, Joachim Koenig-Baltes wrote:
> Unfortunately, there are a few different incompatible APIs for this basic
> operation in different operating systems.  IIRC, the one in HP/UX works
> exactly as you suggest, with an offset per iovec.  I think it might even
> be named pwrite/pwritev, too, which opens up a whole different can of
> worms. :-/

I wasn't able to find this on their website. Do you have a reference?

> For us, for this purpose, now, a new API with flags and an offset per
> chunk might be the best way to go.

Please enlighten me on the offset. What would need it?

I really worry about it as a per-chunk offset would make it really hard to=
maintain POSIX atomicity. Everything else we've done with these calls so=20
far hasn't touched that. And it'd be messy; our lower-level APIs either=20
would need to radically change (more of a change than I feel I can=20
justify) or we would need an itterator.

Take care,


Content-Type: application/pgp-signature
Content-Disposition: inline

Version: GnuPG v1.2.3 (NetBSD)