Re: [REGRESSION PATCH v2] NFS: let NFS_V4 and NFSD_V4 enforce CRYPTO

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

On Wed, Sep 01, 2010 at 09:17:04AM -0400, Trond Myklebust wrote:
> On Wed, 2010-09-01 at 10:52 +0200, Uwe Kleine-König wrote:
> > Hello,
> > 
> > here comes a v2 of the patch that improves the commit log with a more
> > detailed analysis of the breakage introduced by df486a2
> > (= v2.6.36-rc2~34^2~1 BTW) and additionally undoes the "default y" for
> > RPCSEC_GSS_KRB5.
> > 
> > So compared to the state before df486a2 the changes are:
> > 
> >         NFS_V4 selects CRYPTO
> >         NFSD_V4 selects CRYPTO
> > 	RPCSEC_GSS_KRB5 doesn't depend on EXPERIMENTAL anymore
> > 	RPCSEC_GSS_KRB5 now depends on CRYPTO instead of selecting it
> > 
> > Best regards
> > Uwe
> > 
> > ----------------------------->8----------------------------
> > 
> > This is a follow up to
> > 
> > 	df486a2 (NFS: Fix the selection of security flavours in Kconfig)
> > 
> > Before df486a2 NFS_V4 selected RPCSEC_GSS_KRB5 but didn't enforce the
> > latter's dependency EXPERIMENTAL.  df486a2 removed RPCSEC_GSS_KRB5's
> > dependency on EXPERIMENTAL but additionally let it depend on CRYPTO
> > (instead of select CRYPTO before).  So it was still possible to have a
> > config that has NFS_V4 but not RPCSEC_GSS_KRB5.  Moreover df486a2
> > changed the dependency of NFS_V4 and NFSD_V4 on RPCSEC_GSS_KRB5 from
> 
> As I said, the fix is to remove that dependency. I have a fix for the
> NFS client, but the server has more insidious dependencies on RPCSEC_GSS
> due to a poorly designed SECINFO implementation.
Yes, I still remember this, so I suggest to take my patch before 2.6.36
and you can fix it in the merge window for 2.6.37, no?

Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |