by Bruce Fields[permalink] [raw]
On Tue, Feb 12, 2008 at 04:47:16PM -0500, Trond Myklebust wrote:
> On Tue, 2008-02-12 at 16:23 -0500, J. Bruce Fields wrote:
> > On Mon, Feb 11, 2008 at 05:12:16PM -0500, Chuck Lever wrote:
> > > Recently, commit 440bcc59 added a reverse dependency to fs/Kconfig to
> > > ensure that PROC_FS was enabled if NFSD_V4 was enabled.
> > >
> > > There is a guideline in Documentation/kbuild/kconfig-language.txt that
> > > states "In general use select only for non-visible symbols (no prompts
> > > anywhere) and for symbols with no dependencies."
> > >
> > > A quick grep around other Kconfig files reveals that no entry currently
> > > uses "select PROC_FS" -- every one uses "depends on". Thus CONFIG_NFSD_V4
> > > should use "depends on PROC_FS" as well.
> > OK.
> > >
> > > For SUNRPC_GSS, it's a little more complex. Other entries can "select"
> > > SUNRPC_GSS, as it is non-visible. However, the guideline suggests an
> > > entry can't "select" it if it has a dependency (such as PROC_FS).
> > > So, we add forward dependencies on PROC_FS to RPCSEC_GSS_FOO instead.
> > This also makes the client dependent on PROC_FS, which it wasn't before.
> > I assume nobody cares?
> You assume wrongly. Patch NAKed...
Do you have a use case in mind? (Just curious.)
2008-02-12 22:58:00[permalink] [raw]
On Tue, 2008-02-12 at 17:32 -0500, J. Bruce Fields wrote:
> > > I assume nobody cares?
> > You assume wrongly. Patch NAKed...
> Do you have a use case in mind? (Just curious.)
Embedded systems often compile without sysctl and/or procfs support. Why
shouldn't they be able to run an NFS client with RPCSEC_GSS?
The other argument is that there is _nothing_ in the actual auth_gss
code that depends on procfs. The "dependency" (if there is one) is
between net/sunrpc/cache.c and procfs, and in fact, both sunrpc and the
auth_gss modules will compile just fine without procfs.