Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
From:   "NeilBrown" <neilb@suse.de>
To:     "Lorenzo Bianconi" <lorenzo@kernel.org>
Cc:     linux-nfs@vger.kernel.org, lorenzo.bianconi@redhat.com,
        chuck.lever@oracle.com, jlayton@kernel.org
Subject: Re: [PATCH v4 2/2] NFSD: add rpc_status entry in nfsd debug filesystem
In-reply-to: <a23a0482a465299ac06d07d191e0c9377a11a4d1.1690569488.git.lorenzo@kernel.org>
References: <cover.1690569488.git.lorenzo@kernel.org>,
 <a23a0482a465299ac06d07d191e0c9377a11a4d1.1690569488.git.lorenzo@kernel.org>
Date:   Sat, 29 Jul 2023 08:23:26 +1000
Message-id: <169058300693.32308.16341899855806134699@noble.neil.brown.name>
Precedence: bulk

On Sat, 29 Jul 2023, Lorenzo Bianconi wrote:
> Introduce rpc_status entry in nfsd debug filesystem in order to dump
> pending RPC requests debugging information.
>=20
> Link: https://bugzilla.linux-nfs.org/show_bug.cgi?id=3D366
> Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> ---
>  fs/nfsd/nfs4proc.c         |   4 +-
>  fs/nfsd/nfsctl.c           |  10 +++
>  fs/nfsd/nfsd.h             |   2 +
>  fs/nfsd/nfssvc.c           | 122 +++++++++++++++++++++++++++++++++++++
>  include/linux/sunrpc/svc.h |   1 +
>  net/sunrpc/svc.c           |   2 +-
>  6 files changed, 137 insertions(+), 4 deletions(-)
>=20
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index f0f318e78630..b7ad3081bc36 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -2497,8 +2497,6 @@ static inline void nfsd4_increment_op_stats(u32 opnum)
> =20
>  static const struct nfsd4_operation nfsd4_ops[];
> =20
> -static const char *nfsd4_op_name(unsigned opnum);
> -
>  /*
>   * Enforce NFSv4.1 COMPOUND ordering rules:
>   *
> @@ -3628,7 +3626,7 @@ void warn_on_nonidempotent_op(struct nfsd4_op *op)
>  	}
>  }
> =20
> -static const char *nfsd4_op_name(unsigned opnum)
> +const char *nfsd4_op_name(unsigned opnum)
>  {
>  	if (opnum < ARRAY_SIZE(nfsd4_ops))
>  		return nfsd4_ops[opnum].op_name;
> diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
> index 35d2e2cde1eb..f2e4f4b1e4d1 100644
> --- a/fs/nfsd/nfsctl.c
> +++ b/fs/nfsd/nfsctl.c
> @@ -57,6 +57,8 @@ enum {
>  	NFSD_RecoveryDir,
>  	NFSD_V4EndGrace,
>  #endif
> +	NFSD_Rpc_Status,

I think NFSD_Rpc_Status needs to come before the CONFIG_NFSD_V4 block.
Otherwise the comment above (which I apparently approved) makes even
less sense than it does now.
(Maybe just remove the comment??)

> +
>  	NFSD_MaxReserved
>  };
> =20
> @@ -195,6 +197,13 @@ static inline struct net *netns(struct file *file)
>  	return file_inode(file)->i_sb->s_fs_info;
>  }
> =20
> +static const struct file_operations nfsd_rpc_status_operations =3D {
> +	.open		=3D nfsd_rpc_status_open,
> +	.read		=3D seq_read,
> +	.llseek		=3D seq_lseek,
> +	.release	=3D nfsd_pool_stats_release,
> +};
> +
>  /*
>   * write_unlock_ip - Release all locks used by a client
>   *
> @@ -1400,6 +1409,7 @@ static int nfsd_fill_super(struct super_block *sb, st=
ruct fs_context *fc)
>  		[NFSD_RecoveryDir] =3D {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_=
IRUSR},
>  		[NFSD_V4EndGrace] =3D {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO=
},
>  #endif
> +		[NFSD_Rpc_Status] =3D {"rpc_status", &nfsd_rpc_status_operations, S_IRUG=
O},
>  		/* last one */ {""}
>  	};
> =20
> diff --git a/fs/nfsd/nfsd.h b/fs/nfsd/nfsd.h
> index d88498f8b275..75a3e1d55bc8 100644
> --- a/fs/nfsd/nfsd.h
> +++ b/fs/nfsd/nfsd.h
> @@ -94,6 +94,7 @@ int		nfsd_get_nrthreads(int n, int *, struct net *);
>  int		nfsd_set_nrthreads(int n, int *, struct net *);
>  int		nfsd_pool_stats_open(struct inode *, struct file *);
>  int		nfsd_pool_stats_release(struct inode *, struct file *);
> +int		nfsd_rpc_status_open(struct inode *inode, struct file *file);
>  void		nfsd_shutdown_threads(struct net *net);
> =20
>  void		nfsd_put(struct net *net);
> @@ -506,6 +507,7 @@ extern void nfsd4_ssc_init_umount_work(struct nfsd_net =
*nn);
> =20
>  extern void nfsd4_init_leases_net(struct nfsd_net *nn);
> =20
> +const char *nfsd4_op_name(unsigned opnum);
>  #else /* CONFIG_NFSD_V4 */
>  static inline int nfsd4_is_junction(struct dentry *dentry)
>  {
> diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
> index 97830e28c140..e9e954b5ae47 100644
> --- a/fs/nfsd/nfssvc.c
> +++ b/fs/nfsd/nfssvc.c
> @@ -1057,6 +1057,8 @@ int nfsd_dispatch(struct svc_rqst *rqstp)
>  	if (!proc->pc_decode(rqstp, &rqstp->rq_arg_stream))
>  		goto out_decode_err;
> =20
> +	atomic_inc(&rqstp->rq_status_counter);
> +
>  	rp =3D NULL;
>  	switch (nfsd_cache_lookup(rqstp, &rp)) {
>  	case RC_DOIT:
> @@ -1074,6 +1076,8 @@ int nfsd_dispatch(struct svc_rqst *rqstp)
>  	if (!proc->pc_encode(rqstp, &rqstp->rq_res_stream))
>  		goto out_encode_err;
> =20
> +	atomic_inc(&rqstp->rq_status_counter);
> +
>  	nfsd_cache_update(rqstp, rp, rqstp->rq_cachetype, statp + 1);
>  out_cached_reply:
>  	return 1;
> @@ -1149,3 +1153,121 @@ int nfsd_pool_stats_release(struct inode *inode, st=
ruct file *file)
>  	mutex_unlock(&nfsd_mutex);
>  	return ret;
>  }
> +
> +static int nfsd_rpc_status_show(struct seq_file *m, void *v)
> +{
> +	struct inode *inode =3D file_inode(m->file);
> +	struct nfsd_net *nn =3D net_generic(inode->i_sb->s_fs_info, nfsd_net_id);
> +	int i;
> +
> +	rcu_read_lock();
> +
> +	for (i =3D 0; i < nn->nfsd_serv->sv_nrpools; i++) {
> +		struct svc_rqst *rqstp;
> +
> +		list_for_each_entry_rcu(rqstp,
> +				&nn->nfsd_serv->sv_pools[i].sp_all_threads,
> +				rq_all) {
> +			struct nfsd_rpc_status_info {
> +				struct sockaddr daddr;
> +				struct sockaddr saddr;
> +				unsigned long rq_flags;
> +				__be32 rq_xid;
> +				u32 rq_prog;
> +				u32 rq_vers;
> +				const char *pc_name;
> +				ktime_t rq_stime;
> +				u32 opnum[NFSD_MAX_OPS_PER_COMPOUND]; /* NFSv4 compund */
> +			} rqstp_info;
> +			unsigned int status_counter;
> +			char buf[RPC_MAX_ADDRBUFLEN];
> +			int j, opcnt =3D 0;
> +
> +			if (!test_bit(RQ_BUSY, &rqstp->rq_flags))
> +				continue;
> +
> +			status_counter =3D atomic_read(&rqstp->rq_status_counter);
> +
> +			rqstp_info.rq_xid =3D rqstp->rq_xid;
> +			rqstp_info.rq_flags =3D rqstp->rq_flags;
> +			rqstp_info.rq_prog =3D rqstp->rq_prog;
> +			rqstp_info.rq_vers =3D rqstp->rq_vers;
> +			rqstp_info.pc_name =3D svc_proc_name(rqstp);
> +			rqstp_info.rq_stime =3D rqstp->rq_stime;
> +			memcpy(&rqstp_info.daddr, svc_daddr(rqstp),
> +			       sizeof(struct sockaddr));
> +			memcpy(&rqstp_info.saddr, svc_addr(rqstp),
> +			       sizeof(struct sockaddr));
> +
> +#ifdef CONFIG_NFSD_V4
> +			if (rqstp->rq_vers =3D=3D NFS4_VERSION &&
> +			    rqstp->rq_proc =3D=3D NFSPROC4_COMPOUND) {
> +				/* NFSv4 compund */
> +				struct nfsd4_compoundargs *args =3D rqstp->rq_argp;
> +
> +				opcnt =3D args->opcnt;
> +				for (j =3D 0; j < opcnt; j++) {
> +					struct nfsd4_op *op =3D &args->ops[j];
> +
> +					rqstp_info.opnum[j] =3D op->opnum;
> +				}
> +			}
> +#endif /* CONFIG_NFSD_V4 */
> +
> +			/* In order to detect if the RPC request is pending and
> +			 * RPC info are stable we check if rq_status_counter
> +			 * has been incremented during the handler processing.
> +			 */
> +			if (status_counter !=3D atomic_read(&rqstp->rq_status_counter))
> +				continue;
> +
> +			seq_printf(m,
> +				   "0x%08x, 0x%08lx, 0x%08x, NFSv%d, %s, %016lld,",

Please drop the commas.
It might be defensible to have commas and no spaces by comparing with
/proc/fs/nfsd/supported_krb5_enctypes, but the dominant pattern is to
use only spaces to separate fields on /proc files.

Thanks,
NeilBrown