Re: linux-next: nfs tree build failures

[Trond Myklebust <trond.myklebust@fys.uio.no>]

On Mon, 2009-04-06 at 10:43 +0300, Benny Halevy wrote:
> On Apr. 06, 2009, 9:29 +0300, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> > Hi Trond,
> > 
> > Today's linux-next build (powerpc ppc44x_config) failed like this:
> > 
> > fs/nfs/client.c: In function 'nfs_match_client':
> > fs/nfs/client.c:444: error: 'struct nfs_client' has no member named 'cl_minorversion'
> > 
> > Caused by commit 10d5a514705f0687cfbb6a080c3562b857e340fe ("nfs41: Use
> > mount minorversion option").  The existence of cl_minorversion is
> > controlled by CONFIG_NFS_V4 but the reference is not.
> 
> 
> Stephan,
> 
> First, thanks! I apologize for introducing this.
> This was caused by a late fix to Trond's review comment which
> apparently was under-tested, as you mentioned.
> 
> > 
> > I added the small patch below for today.
> 
> Based on Trond's review comment:
> http://linux-nfs.org/pipermail/pnfs/2009-March/006938.html
> 
> We'd actually want to keep cl_minorversion's usage in the clear
> and rather move its definition out of the #ifdef.
> 
> Please see PATCH 1/3 in reply to this message.
> 
> > 
> > The same build gets this warning as well:
> > 
> > fs/nfs/client.c:554: warning: 'nfs4_session_set_rwsize' defined but not used
> 
> This one needs to go under CONFIG_NFS_V4.
> See PATCH 2/2
> 
> > 
> > and then fails like this:
> > 
> > fs/built-in.o: In function `nfs_readpage_result_full':
> > read.c:(.text+0x70d24): undefined reference to `nfs4_restart_rpc'
> > fs/built-in.o: In function `nfs_readpage_result_partial':
> > read.c:(.text+0x70e3c): undefined reference to `nfs4_restart_rpc'
> > fs/built-in.o: In function `nfs_async_unlink_done':
> > unlink.c:(.text+0x71af0): undefined reference to `nfs4_restart_rpc'
> > fs/built-in.o: In function `nfs_writeback_done':
> > (.text+0x72c00): undefined reference to `nfs4_restart_rpc'
> > 
> > Caused by commit c105a9f16f11aaa9fcb9f75f8342b89c5d9665cf ("nfs41: use
> > rpc prepare call state for session reset") which I have reverted for
> > today.
> 
> See PATCH 3/3 that inlines nfs4_restart_rpc's definition.
> 
> Benny

All squashed into the relevant original patches.

> > 
> > Trond, is all this NFS v4.1 stuff that turned up yesterday really 2.6.30
> > material?  If not, it should be removed from linux-next until after
> > 2.6.30-rc1 is released.  (and clearly better tested ...)
> > 
> > Bruce, I guess the same question applies to the nfsd tree as well.

This is code that isn't supposed to significantly affect the existing
NFS tree unless CONFIG_NFS_V4_1 is set. Since the corresponding NFSd
code has been merged into 2.6.30, it would be nice to have the client
code too so that people can start testing, and so that development can
continue in the mainline tree instead of in the smoky backrooms of the
linux-nfs.org git trees.

Cheers
  Trond