Subject: Re: letting userland issue FUA writes
To: None <pavel.cahyna@st.mff.cuni.cz, tech-kern@netbsd.org>
From: Bill Studenmund <wrstuden@netbsd.org>
List: tech-kern
Date: 03/15/2006 16:30:59
--pE2VAHO2njSJCslu
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 15, 2006 at 05:23:51PM +0100, Pavel Cahyna wrote:
> On Tue, Mar 14, 2006 at 06:35:46PM -0800, Bill Studenmund wrote:
> > I'm interested in adding infrastructure so that applications can issue=
=20
> > writes that have the FUA (Force Unit Access) bit set. SCSI CDBs handle=
=20
> > this, as do some RAID cards.
> >=20
> > The main question I have is what kind of user-level interfaces exist fo=
r=20
> > this? I certainly could cook up a new system call, but I would much rat=
her=20
> > use prior art. My simplistic attempt at googling for this failed. Does=
=20
> > anyone else have suggestions?
>=20
> O_DSYNC is documented as: "each write will wait for the file data=20
> to be committed to stable storage."
>=20
> Is this semantics different from FUA?

Not really. However I really want a system call where I can request this=20
on a per-call basis. Otherwise I have to have two open file descriptors,=20
one with and one without this flag set. That seems kinda silly.

Take care,

Bill

--pE2VAHO2njSJCslu
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFEGLHDWz+3JHUci9cRAgKvAJ48bkdpfw5oUQvdIoNvUSoq8J/E1wCfd8+Z
lcKeOZcn4GJveRGH+C+NNqI=
=xp8T
-----END PGP SIGNATURE-----

--pE2VAHO2njSJCslu--