Received-SPF: pass (google.com: domain of linux-nfs+bounces-2021-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk
MIME-Version: 1.0
From: "NeilBrown" <neilb@suse.de>
To: trondmy@kernel.org
Cc: "Chuck Lever" <chuck.lever@oracle.com>, linux-nfs@vger.kernel.org
Subject: Re: [PATCH v2 1/2] nfsd: Fix a regression in nfsd_setattr()
In-reply-to: <20240216012451.22725-2-trondmy@kernel.org>
References: <20240216012451.22725-1-trondmy@kernel.org>,
 <20240216012451.22725-2-trondmy@kernel.org>
Date: Mon, 19 Feb 2024 08:57:15 +1100
Message-id: <170829343514.1530.1077342787397990579@noble.neil.brown.name>

On Fri, 16 Feb 2024, trondmy@kernel.org wrote:
> From: Trond Myklebust <trond.myklebust@hammerspace.com>
>=20
> Commit bb4d53d66e4b broke the NFSv3 pre/post op attributes behaviour
> when doing a SETATTR rpc call by stripping out the calls to
> fh_fill_pre_attrs() and fh_fill_post_attrs().
>=20
> Fixes: bb4d53d66e4b ("NFSD: use (un)lock_inode instead of fh_(un)lock for f=
ile operations")
> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
> ---
>  fs/nfsd/nfs4proc.c | 4 ++++
>  fs/nfsd/vfs.c      | 9 +++++++--
>  2 files changed, 11 insertions(+), 2 deletions(-)
>=20
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index 14712fa08f76..e6d8624efc83 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -1143,6 +1143,7 @@ nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_co=
mpound_state *cstate,
>  	};
>  	struct inode *inode;
>  	__be32 status =3D nfs_ok;
> +	bool save_no_wcc;
>  	int err;
> =20
>  	if (setattr->sa_iattr.ia_valid & ATTR_SIZE) {
> @@ -1168,8 +1169,11 @@ nfsd4_setattr(struct svc_rqst *rqstp, struct nfsd4_c=
ompound_state *cstate,
> =20
>  	if (status)
>  		goto out;
> +	save_no_wcc =3D cstate->current_fh.fh_no_wcc;
> +	cstate->current_fh.fh_no_wcc =3D true;
>  	status =3D nfsd_setattr(rqstp, &cstate->current_fh, &attrs,
>  				0, (time64_t)0);
> +	cstate->current_fh.fh_no_wcc =3D save_no_wcc;

This looks clumsy.
I think the background is that NFSv3 needs atomic wcc attributes for
file operations, but NFSv4 doesn't - it only has them for directory ops.
So NFSv4, like NFSv2, doesn't want fh_fill_pre_attrs() to be called by
nfsd_setattr().

NFSv2 avoids it by always setting ->fh_no_wcc.  Here you temporarily set
fh_no_wcc to true for the same effect.  So the code is correct.
But it is not obvious to the casual reader why this is happening.

I would rather a "wcc_wanted" flag or similar, but that can be done in a
separate clean-up patch later.

Thanks for finding and fixing this regression of mine.

Reviewed-by: NeilBrown <neilb@suse.de>

NeilBrown


>  	if (!status)
>  		status =3D nfserrno(attrs.na_labelerr);
>  	if (!status)
> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> index 6e7e37192461..58fab461bc00 100644
> --- a/fs/nfsd/vfs.c
> +++ b/fs/nfsd/vfs.c
> @@ -497,7 +497,7 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp,
>  	int		accmode =3D NFSD_MAY_SATTR;
>  	umode_t		ftype =3D 0;
>  	__be32		err;
> -	int		host_err;
> +	int		host_err =3D 0;
>  	bool		get_write_count;
>  	bool		size_change =3D (iap->ia_valid & ATTR_SIZE);
>  	int		retries;
> @@ -555,6 +555,9 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp,
>  	}
> =20
>  	inode_lock(inode);
> +	err =3D fh_fill_pre_attrs(fhp);
> +	if (err)
> +		goto out_unlock;
>  	for (retries =3D 1;;) {
>  		struct iattr attrs;
> =20
> @@ -582,13 +585,15 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *f=
hp,
>  		attr->na_aclerr =3D set_posix_acl(&nop_mnt_idmap,
>  						dentry, ACL_TYPE_DEFAULT,
>  						attr->na_dpacl);
> +	fh_fill_post_attrs(fhp);
> +out_unlock:
>  	inode_unlock(inode);
>  	if (size_change)
>  		put_write_access(inode);
>  out:
>  	if (!host_err)
>  		host_err =3D commit_metadata(fhp);
> -	return nfserrno(host_err);
> +	return err !=3D 0 ? err : nfserrno(host_err);
>  }
> =20
>  #if defined(CONFIG_NFSD_V4)
> --=20
> 2.43.1
>=20
>=20
>=20