Skip to contentSkip to conversation

exposing knfsd state to userspace

PATCH New J. Bruce Fields 2 participants 16 messages May 16, 2019 – May 18, 2019

Subsystems (3)
lore
Share & tools
lore
#1 lore raw

PATCH 1/2 nfsd: allow fh_want_write to be called twice

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

A fuzzer recently triggered lockdep warnings about potential sb_writers deadlocks caused by fh_want_write().

Looks like we aren't careful to pair each fh_want_write() with an fh_drop_write().

It's not normally a problem since fh_put() will call fh_drop_write() for us. And was OK for NFSv3 where we'd do one operation that might call fh_want_write(), and then put the filehandle.

But an NFSv4 protocol fuzzer can do weird things like call unlink twice in a compound, and then we get into trouble.

I'm a little worried about this approach of just leaving everything to fh_put(). But I think there are probably a lot of fh_want_write()/fh_drop_write() imbalances so for now I think we need it to be more forgiving.

---
 fs/nfsd/vfs.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
fs/nfsd/vfs.h +4 -1
index a7e107309f76..db351247892d 100644
@@ -120,8 +120,11 @@ void		nfsd_put_raparams(struct file *file, struct raparms *ra);
 
 static inline int fh_want_write(struct svc_fh *fh)
 {
-	int ret = mnt_want_write(fh->fh_export->ex_path.mnt);
+	int ret;
 
+	if (fh->fh_want_write)
+		return 0;
+	ret = mnt_want_write(fh->fh_export->ex_path.mnt);
 	if (!ret)
 		fh->fh_want_write = true;
 	return ret;
Signature · 1 line
2.21.0
#3 lore raw

PATCH 1/12 nfsd: persist nfsd filesystem across mounts

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

Keep around one internal mount of the nfsd filesystem so that we can e.g. add stuff to it when clients come and go, regardless of whether anyone has it mounted.

---
 fs/nfsd/netns.h  |  3 +++
 fs/nfsd/nfsctl.c | 13 +++++++++++++
 2 files changed, 16 insertions(+)
fs/nfsd/netns.h +3 -0
index 32cb8c027483..cce335e1ec98 100644
@@ -55,6 +55,9 @@ struct nfsd_net {
 	bool grace_ended;
 	time_t boot_time;
 
+	/* internal mount of the "nfsd" pseudofilesystem: */
+	struct vfsmount *nfsd_mnt;
+
 	/*
 	 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
 	 * used in reboot/reset lease grace period processing
fs/nfsd/nfsctl.c +13 -0
index f2feb2d11bae..8d2062428569 100644
@@ -1231,6 +1231,7 @@ unsigned int nfsd_net_id;
 static __net_init int nfsd_init_net(struct net *net)
 {
 	int retval;
+	struct vfsmount *mnt;
 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
 	retval = nfsd_export_init(net);
@@ -1248,8 +1249,17 @@ static __net_init int nfsd_init_net(struct net *net)
 
 	atomic_set(&nn->ntf_refcnt, 0);
 	init_waitqueue_head(&nn->ntf_wq);
+
+	mnt =  vfs_kern_mount(&nfsd_fs_type, SB_KERNMOUNT, "nfsd", NULL);
+	if (IS_ERR(mnt)) {
+		retval = PTR_ERR(mnt);
+		goto out_mount_err;
+	}
+	nn->nfsd_mnt = mnt;
 	return 0;
 
+out_mount_err:
+	nfsd_idmap_shutdown(net);
 out_idmap_error:
 	nfsd_export_shutdown(net);
 out_export_error:
@@ -1258,6 +1268,9 @@ static __net_init int nfsd_init_net(struct net *net)
 
 static __net_exit void nfsd_exit_net(struct net *net)
 {
+	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
+
+	mntput(nn->nfsd_mnt);
 	nfsd_idmap_shutdown(net);
 	nfsd_export_shutdown(net);
 }
Signature · 1 line
2.21.0
#11 lore raw

PATCH 9/12 nfsd4: add file to display list of client's opens

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

Add a nfsd/clients/#/opens file to list some information about all the opens held by the given client, including open modes, device numbers, inode numbers, and open owners.

Open owners are totally opaque but seem to sometimes have some useful ascii strings included, so passing through printable ascii characters and escaping the rest seems useful while still being machine-readable.

---
 fs/nfsd/nfs4state.c | 149 +++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 147 insertions(+), 2 deletions(-)
fs/nfsd/nfs4state.c +147 -2
index 876c035d2a91..4025b3c692fd 100644
@@ -685,7 +685,8 @@ struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *sla
 
 	idr_preload(GFP_KERNEL);
 	spin_lock(&cl->cl_lock);
-	new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
+	/* Reserving 0 for start of file in nfsdfs "states" file: */
+	new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 1, 0, GFP_NOWAIT);
 	spin_unlock(&cl->cl_lock);
 	idr_preload_end();
 	if (new_id < 0)
@@ -2241,9 +2242,153 @@ static const struct file_operations client_info_fops = {
 	.release	= single_release,
 };
 
+static void *states_start(struct seq_file *s, loff_t *pos)
+	__acquires(&clp->cl_lock)
+{
+	struct nfs4_client *clp = s->private;
+	unsigned long id = *pos;
+	void *ret;
+
+	spin_lock(&clp->cl_lock);
+	ret = idr_get_next_ul(&clp->cl_stateids, &id);
+	*pos = id;
+	return ret;
+}
+
+static void *states_next(struct seq_file *s, void *v, loff_t *pos)
+{
+	struct nfs4_client *clp = s->private;
+	unsigned long id = *pos;
+	void *ret;
+
+	id = *pos;
+	id++;
+	ret = idr_get_next_ul(&clp->cl_stateids, &id);
+	*pos = id;
+	return ret;
+}
+
+static void states_stop(struct seq_file *s, void *v)
+	__releases(&clp->cl_lock)
+{
+	struct nfs4_client *clp = s->private;
+
+	spin_unlock(&clp->cl_lock);
+}
+
+static void nfs4_show_superblock(struct seq_file *s, struct file *f)
+{
+	struct inode *inode = file_inode(f);
+
+	seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
+					MAJOR(inode->i_sb->s_dev),
+					 MINOR(inode->i_sb->s_dev),
+					 inode->i_ino);
+}
+
+static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
+{
+	seq_printf(s, "owner: ");
+	seq_quote_mem(s, oo->so_owner.data, oo->so_owner.len);
+}
+
+static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
+{
+	struct nfs4_ol_stateid *ols;
+	struct nfs4_file *nf;
+	struct file *file;
+	struct nfs4_stateowner *oo;
+	unsigned int access, deny;
+
+	if (st->sc_type != NFS4_OPEN_STID && st->sc_type != NFS4_LOCK_STID)
+		return 0; /* XXX: or SEQ_SKIP? */
+	ols = openlockstateid(st);
+	oo = ols->st_stateowner;
+	nf = st->sc_file;
+	file = find_any_file(nf);
+
+	seq_printf(s, "- 0x%16phN: { type: open, ", &st->sc_stateid);
+
+	access = bmap_to_share_mode(ols->st_access_bmap);
+	deny   = bmap_to_share_mode(ols->st_deny_bmap);
+
+	seq_printf(s, "access: \%s\%s, ",
+		access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
+		access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
+	seq_printf(s, "deny: \%s\%s, ",
+		deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
+		deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
+
+	nfs4_show_superblock(s, file);
+	seq_printf(s, ", ");
+	nfs4_show_owner(s, oo);
+	seq_printf(s, " }\n");
+
+	fput(file);
+
+	return 0;
+}
+
+static int states_show(struct seq_file *s, void *v)
+{
+	struct nfs4_stid *st = v;
+
+	switch (st->sc_type) {
+	case NFS4_OPEN_STID:
+		return nfs4_show_open(s, st);
+	default:
+		return 0; /* XXX: or SEQ_SKIP? */
+	}
+}
+
+static struct seq_operations states_seq_ops = {
+	.start = states_start,
+	.next = states_next,
+	.stop = states_stop,
+	.show = states_show
+};
+
+static int client_states_open(struct inode *inode, struct file *file)
+{
+	struct nfsdfs_client *nc;
+	struct seq_file *s;
+	struct nfs4_client *clp;
+	int ret;
+
+	nc = get_nfsdfs_client(inode);
+	if (!nc)
+		return -ENXIO;
+	clp = container_of(nc, struct nfs4_client, cl_nfsdfs);
+
+	ret = seq_open(file, &states_seq_ops);
+	if (ret)
+		return ret;
+	s = file->private_data;
+	s->private = clp;
+	return 0;
+}
+
+static int client_opens_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *m = file->private_data;
+	struct nfs4_client *clp = m->private;
+
+	/* XXX: alternatively, we could get/drop in seq start/stop */
+	drop_client(clp);
+	return 0;
+}
+
+static const struct file_operations client_states_fops = {
+	.open		= client_states_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= client_opens_release,
+};
+
 static const struct tree_descr client_files[] = {
 	[0] = {"info", &client_info_fops, S_IRUSR},
-	[1] = {""},
+	[1] = {"states", &client_states_fops, S_IRUSR},
+	[3] = {""},
 };
 
 static struct nfs4_client *create_client(struct xdr_netobj name,
Signature · 1 line
2.21.0
#6 lore raw

PATCH 4/12 nfsd: add nfsd/clients directory

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

I plan to expose some information about nfsv4 clients here.

---
 fs/nfsd/netns.h     |   2 +
 fs/nfsd/nfs4state.c |  23 ++++++----
 fs/nfsd/nfsctl.c    | 108 +++++++++++++++++++++++++++++++++++++++++++-
 fs/nfsd/nfsd.h      |   9 ++++
 fs/nfsd/state.h     |   6 ++-
 5 files changed, 138 insertions(+), 10 deletions(-)
fs/nfsd/netns.h +2 -0
index cce335e1ec98..46d956676480 100644
@@ -58,6 +58,8 @@ struct nfsd_net {
 	/* internal mount of the "nfsd" pseudofilesystem: */
 	struct vfsmount *nfsd_mnt;
 
+	struct dentry *nfsd_client_dir;
+
 	/*
 	 * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
 	 * used in reboot/reset lease grace period processing
fs/nfsd/nfs4state.c +15 -8
index 83d0ee329e14..dfcb90743861 100644
@@ -1871,20 +1871,19 @@ static struct nfs4_client *alloc_client(struct xdr_netobj name)
 
 static void __free_client(struct kref *k)
 {
-	struct nfs4_client *clp = container_of(k, struct nfs4_client, cl_ref);
+	struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
+	struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
 
 	free_svc_cred(&clp->cl_cred);
 	kfree(clp->cl_ownerstr_hashtbl);
 	kfree(clp->cl_name.data);
 	idr_destroy(&clp->cl_stateids);
-	if (clp->cl_nfsd_dentry)
-		nfsd_client_rmdir(clp->cl_nfsd_dentry);
 	kmem_cache_free(client_slab, clp);
 }
 
 void drop_client(struct nfs4_client *clp)
 {
-	kref_put(&clp->cl_ref, __free_client);
+	kref_put(&clp->cl_nfsdfs.cl_ref, __free_client);
 }
 
 static void
@@ -1899,6 +1898,8 @@ free_client(struct nfs4_client *clp)
 		free_session(ses);
 	}
 	rpc_destroy_wait_queue(&clp->cl_cb_waitq);
+	if (clp->cl_nfsd_dentry)
+		nfsd_client_rmdir(clp->cl_nfsd_dentry);
 	drop_client(clp);
 }
 
@@ -2199,6 +2200,7 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 	struct sockaddr *sa = svc_addr(rqstp);
 	int ret;
 	struct net *net = SVC_NET(rqstp);
+	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
 
 	clp = alloc_client(name);
 	if (clp == NULL)
@@ -2209,8 +2211,8 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 		free_client(clp);
 		return NULL;
 	}
-
-	kref_init(&clp->cl_ref);
+	gen_clid(clp, nn);
+	kref_init(&clp->cl_nfsdfs.cl_ref);
 	nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
 	clp->cl_time = get_seconds();
 	clear_bit(0, &clp->cl_cb_slot_busy);
@@ -2218,6 +2220,12 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 	rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
 	clp->cl_cb_session = NULL;
 	clp->net = net;
+	clp->cl_nfsd_dentry = nfsd_client_mkdir(nn, &clp->cl_nfsdfs,
+						clp->cl_clientid.cl_id);
+	if (!clp->cl_nfsd_dentry) {
+		free_client(clp);
+		return NULL;
+	}
 	return clp;
 }
 
@@ -2661,7 +2669,6 @@ nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 	new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
 	new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
 
-	gen_clid(new, nn);
 	add_to_unconfirmed(new);
 	swap(new, conf);
 out_copy:
@@ -3404,7 +3411,7 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 		copy_clid(new, conf);
 		gen_confirm(new, nn);
 	} else /* case 4 (new client) or cases 2, 3 (client reboot): */
-		gen_clid(new, nn);
+		;
 	new->cl_minorversion = 0;
 	gen_callback(new, setclid, rqstp);
 	add_to_unconfirmed(new);
fs/nfsd/nfsctl.c +107 -1

134 lines not on this page · raw message

fs/nfsd/nfsd.h +9 -0
index 066899929863..fe7418c2b88f 100644
@@ -21,6 +21,7 @@
 
 #include <uapi/linux/nfsd/debug.h>
 
+#include "netns.h"
 #include "stats.h"
 #include "export.h"
 
@@ -85,6 +86,14 @@ int		nfsd_pool_stats_release(struct inode *, struct file *);
 
 void		nfsd_destroy(struct net *net);
 
+struct nfsdfs_client {
+	struct kref cl_ref;
+	void (*cl_release)(struct kref *kref);
+};
+
+struct dentry *nfsd_client_mkdir(struct nfsd_net *nn, struct nfsdfs_client *ncl, u32 id);
+void nfsd_client_rmdir(struct dentry *dentry);
+
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
 #ifdef CONFIG_NFSD_V2_ACL
 extern const struct svc_version nfsd_acl_version2;
fs/nfsd/state.h +5 -1
index aa26ae520fb6..5a999e4b766e 100644
@@ -39,6 +39,7 @@
 #include <linux/refcount.h>
 #include <linux/sunrpc/svc_xprt.h>
 #include "nfsfh.h"
+#include "nfsd.h"
 
 typedef struct {
 	u32             cl_boot;
@@ -347,9 +348,12 @@ struct nfs4_client {
 	u32			cl_exchange_flags;
 	/* number of rpc's in progress over an associated session: */
 	atomic_t		cl_rpc_users;
-	struct kref		cl_ref;
+	struct nfsdfs_client	cl_nfsdfs;
 	struct nfs4_op_map      cl_spo_must_allow;
 
+	/* debugging info directory under nfsd/clients/ : */
+	struct dentry		*cl_nfsd_dentry;
+
 	/* for nfs41 callbacks */
 	/* We currently support a single back channel with a single slot */
 	unsigned long		cl_cb_slot_busy;
Signature · 1 line
2.21.0
#14 lore raw

PATCH 12/12 nfsd: allow forced expiration of NFSv4 clients

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

NFSv4 clients are automatically expired and all their locks removed if they don't contact the server for a certain amount of time (the lease period, 90 seconds by default).

There can still be situations where that's not enough, so allow userspace to force expiry by writing "expire\n" to the new nfsd/client/#/ctl file.

(The generic "ctl" name is because I expect we may want to allow other operations on clients in the future.)

The write will not return until the client is expired and all of its locks and other state removed.

The fault injection code also provides a way of expiring clients, but it fails if there are any in-progress RPC's referencing the client. Also, its method of selecting a client to expire is a little more primitive--it uses an IP address, which can't always uniquely specify an NFSv4 client.

---
 fs/nfsd/nfs4state.c | 91 ++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 82 insertions(+), 9 deletions(-)
fs/nfsd/nfs4state.c +82 -9
index e36c6e64e448..5ef27e9cbed3 100644
@@ -99,6 +99,13 @@ enum nfsd4_st_mutex_lock_subclass {
  */
 static DECLARE_WAIT_QUEUE_HEAD(close_wq);
 
+/*
+ * A waitqueue where a writer to clients/#/ctl destroying a client can
+ * wait for cl_rpc_users to drop to 0 and then for the client to be
+ * unhashed.
+ */
+static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
+
 static struct kmem_cache *client_slab;
 static struct kmem_cache *openowner_slab;
 static struct kmem_cache *lockowner_slab;
@@ -174,6 +181,8 @@ static void put_client_renew_locked(struct nfs4_client *clp)
 		return;
 	if (!is_client_expired(clp))
 		renew_client_locked(clp);
+	else
+		wake_up_all(&expiry_wq);
 }
 
 static void put_client_renew(struct nfs4_client *clp)
@@ -184,6 +193,8 @@ static void put_client_renew(struct nfs4_client *clp)
 		return;
 	if (!is_client_expired(clp))
 		renew_client_locked(clp);
+	else
+		wake_up_all(&expiry_wq);
 	spin_unlock(&nn->client_lock);
 }
 
@@ -1900,8 +1911,11 @@ free_client(struct nfs4_client *clp)
 		free_session(ses);
 	}
 	rpc_destroy_wait_queue(&clp->cl_cb_waitq);
-	if (clp->cl_nfsd_dentry)
+	if (clp->cl_nfsd_dentry) {
 		nfsd_client_rmdir(clp->cl_nfsd_dentry);
+		clp->cl_nfsd_dentry = NULL;
+		wake_up_all(&expiry_wq);
+	}
 	drop_client(clp);
 }
 
@@ -1996,6 +2010,7 @@ __destroy_client(struct nfs4_client *clp)
 	if (clp->cl_cb_conn.cb_xprt)
 		svc_xprt_put(clp->cl_cb_conn.cb_xprt);
 	free_client(clp);
+	wake_up_all(&expiry_wq);
 }
 
 static void
@@ -2195,6 +2210,15 @@ find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
 	return s;
 }
 
+static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
+{
+	struct nfsdfs_client *nc;
+	nc = get_nfsdfs_client(inode);
+	if (!nc)
+		return NULL;
+	return container_of(nc, struct nfs4_client, cl_nfsdfs);
+}
+
 static void seq_escape_quotable_mem(struct seq_file *m, char *data, int len)
 {
 	/* List taken from kstrdup_quotable: */
@@ -2211,14 +2235,12 @@ static void seq_quote_mem(struct seq_file *m, char *data, int len)
 static int client_info_show(struct seq_file *m, void *v)
 {
 	struct inode *inode = m->private;
-	struct nfsdfs_client *nc;
 	struct nfs4_client *clp;
 	u64 clid;
 
-	nc = get_nfsdfs_client(inode);
-	if (!nc)
+	clp = get_nfsdfs_clp(inode);
+	if (!clp)
 		return -ENXIO;
-	clp = container_of(nc, struct nfs4_client, cl_nfsdfs);
 	memcpy(&clid, &clp->cl_clientid, sizeof(clid));
 	seq_printf(m, "clientid: %llx\n", clid);
 	seq_printf(m, "address: %pISpc\n", (struct sockaddr *)&clp->cl_addr);
@@ -2429,15 +2451,13 @@ static struct seq_operations states_seq_ops = {
 
 static int client_states_open(struct inode *inode, struct file *file)
 {
-	struct nfsdfs_client *nc;
 	struct seq_file *s;
 	struct nfs4_client *clp;
 	int ret;
 
-	nc = get_nfsdfs_client(inode);
-	if (!nc)
+	clp = get_nfsdfs_clp(inode);
+	if (!clp)
 		return -ENXIO;
-	clp = container_of(nc, struct nfs4_client, cl_nfsdfs);
 
 	ret = seq_open(file, &states_seq_ops);
 	if (ret)
@@ -2464,9 +2484,62 @@ static const struct file_operations client_states_fops = {
 	.release	= client_opens_release,
 };
 
+/*
+ * Normally we refuse to destroy clients that are in use, but here the
+ * administrator is telling us to just do it.  We also want to wait
+ * so the caller has a guarantee that the client's locks are gone by
+ * the time the write returns:
+ */
+void force_expire_client(struct nfs4_client *clp)
+{
+	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
+	bool already_expired;
+
+	spin_lock(&clp->cl_lock);
+	clp->cl_time = 0;
+	spin_unlock(&clp->cl_lock);
+
+	wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
+	spin_lock(&nn->client_lock);
+	already_expired = list_empty(&clp->cl_lru);
+	if (!already_expired)
+		unhash_client_locked(clp);
+	spin_unlock(&nn->client_lock);
+
+	if (!already_expired)
+		expire_client(clp);
+	else
+		wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
+}
+
+static ssize_t client_ctl_write(struct file *file, const char __user *buf,
+				   size_t size, loff_t *pos)
+{
+	char *data;
+	struct nfs4_client *clp;
+
+	data = simple_transaction_get(file, buf, size);
+	if (IS_ERR(data))
+		return PTR_ERR(data);
+	if (size != 7 || 0 != memcmp(data, "expire\n", 7))
+		return -EINVAL;
+	clp = get_nfsdfs_clp(file_inode(file));
+	if (!clp)
+		return -ENXIO;
+	force_expire_client(clp);
+	drop_client(clp);
+	return 7;
+}
+
+static const struct file_operations client_ctl_fops = {
+	.write		= client_ctl_write,
+	.release	= simple_transaction_release,
+};
+
 static const struct tree_descr client_files[] = {
 	[0] = {"info", &client_info_fops, S_IRUSR},
 	[1] = {"states", &client_states_fops, S_IRUSR},
+	[2] = {"ctl", &client_ctl_fops, S_IRUSR|S_IWUSR},
 	[3] = {""},
 };
 
Signature · 1 line
2.21.0
#13 lore raw

PATCH 11/12 nfsd4: show layout stateids

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

These are also minimal for now, I'm not sure what information would be useful.

---
 fs/nfsd/nfs4state.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)
fs/nfsd/nfs4state.c +20 -0
index 652e0544e058..e36c6e64e448 100644
@@ -2383,6 +2383,24 @@ static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
 	return 0;
 }
 
+static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
+{
+	struct nfs4_layout_stateid *ls;
+	struct file *file;
+
+	ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
+	file = ls->ls_file;
+
+	seq_printf(s, "- 0x%16phN: { type: layout, ", &st->sc_stateid);
+
+	/* XXX: What else would be useful? */
+
+	nfs4_show_superblock(s, file);
+	seq_printf(s, " }\n");
+
+	return 0;
+}
+
 static int states_show(struct seq_file *s, void *v)
 {
 	struct nfs4_stid *st = v;
@@ -2394,6 +2412,8 @@ static int states_show(struct seq_file *s, void *v)
 		return nfs4_show_lock(s, st);
 	case NFS4_DELEG_STID:
 		return nfs4_show_deleg(s, st);
+	case NFS4_LAYOUT_STID:
+		return nfs4_show_layout(s, st);
 	default:
 		return 0; /* XXX: or SEQ_SKIP? */
 	}
Signature · 1 line
2.21.0
#5 lore raw

PATCH 3/12 nfsd4: use reference count to free client

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

Keep a second reference count which is what is really used to decide when to free the client's memory. File objects under nfsd/client/ will hold these references.

---
 fs/nfsd/nfs4state.c | 26 +++++++++++++++++++++-----
 fs/nfsd/state.h     |  1 +
 2 files changed, 22 insertions(+), 5 deletions(-)
fs/nfsd/nfs4state.c +21 -5
index 9dab61bbd256..83d0ee329e14 100644
@@ -1869,6 +1869,24 @@ static struct nfs4_client *alloc_client(struct xdr_netobj name)
 	return NULL;
 }
 
+static void __free_client(struct kref *k)
+{
+	struct nfs4_client *clp = container_of(k, struct nfs4_client, cl_ref);
+
+	free_svc_cred(&clp->cl_cred);
+	kfree(clp->cl_ownerstr_hashtbl);
+	kfree(clp->cl_name.data);
+	idr_destroy(&clp->cl_stateids);
+	if (clp->cl_nfsd_dentry)
+		nfsd_client_rmdir(clp->cl_nfsd_dentry);
+	kmem_cache_free(client_slab, clp);
+}
+
+void drop_client(struct nfs4_client *clp)
+{
+	kref_put(&clp->cl_ref, __free_client);
+}
+
 static void
 free_client(struct nfs4_client *clp)
 {
@@ -1881,11 +1899,7 @@ free_client(struct nfs4_client *clp)
 		free_session(ses);
 	}
 	rpc_destroy_wait_queue(&clp->cl_cb_waitq);
-	free_svc_cred(&clp->cl_cred);
-	kfree(clp->cl_ownerstr_hashtbl);
-	kfree(clp->cl_name.data);
-	idr_destroy(&clp->cl_stateids);
-	kmem_cache_free(client_slab, clp);
+	drop_client(clp);
 }
 
 /* must be called under the client_lock */
@@ -2195,6 +2209,8 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 		free_client(clp);
 		return NULL;
 	}
+
+	kref_init(&clp->cl_ref);
 	nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
 	clp->cl_time = get_seconds();
 	clear_bit(0, &clp->cl_cb_slot_busy);
fs/nfsd/state.h +1 -0
index aa9f1676e88a..aa26ae520fb6 100644
@@ -347,6 +347,7 @@ struct nfs4_client {
 	u32			cl_exchange_flags;
 	/* number of rpc's in progress over an associated session: */
 	atomic_t		cl_rpc_users;
+	struct kref		cl_ref;
 	struct nfs4_op_map      cl_spo_must_allow;
 
 	/* for nfs41 callbacks */
Signature · 1 line
2.21.0
#8 lore raw

PATCH 6/12 nfsd4: add a client info file

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

Add a new nfsd/clients/#/info file with some basic information about each NFSv4 client.

---
 fs/nfsd/nfs4state.c |  38 ++++++++++++++-
 fs/nfsd/nfsctl.c    | 115 +++++++++++++++++++++++++++++++++++++++++---
 fs/nfsd/nfsd.h      |   4 +-
 3 files changed, 148 insertions(+), 9 deletions(-)
fs/nfsd/nfs4state.c +37 -1
index de99bfe3e6e2..928705fc8ff5 100644
@@ -2193,6 +2193,41 @@ find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
 	return s;
 }
 
+static int client_info_show(struct seq_file *m, void *v)
+{
+	struct inode *inode = m->private;
+	struct nfsdfs_client *nc;
+	struct nfs4_client *clp;
+	u64 clid;
+
+	nc = get_nfsdfs_client(inode);
+	if (!nc)
+		return -ENXIO;
+	clp = container_of(nc, struct nfs4_client, cl_nfsdfs);
+	memcpy(&clid, &clp->cl_clientid, sizeof(clid));
+	seq_printf(m, "clientid: %llx\n", clid);
+	drop_client(clp);
+
+	return 0;
+}
+
+static int client_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, client_info_show, inode);
+}
+
+static const struct file_operations client_info_fops = {
+	.open		= client_info_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static const struct tree_descr client_files[] = {
+	[0] = {"info", &client_info_fops, S_IRUSR},
+	[1] = {""},
+};
+
 static struct nfs4_client *create_client(struct xdr_netobj name,
 		struct svc_rqst *rqstp, nfs4_verifier *verf)
 {
@@ -2221,7 +2256,8 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 	clp->cl_cb_session = NULL;
 	clp->net = net;
 	clp->cl_nfsd_dentry = nfsd_client_mkdir(nn, &clp->cl_nfsdfs,
-			clp->cl_clientid.cl_id - nn->clientid_base);
+			clp->cl_clientid.cl_id - nn->clientid_base,
+			client_files);
 	if (!clp->cl_nfsd_dentry) {
 		free_client(clp);
 		return NULL;
fs/nfsd/nfsctl.c +108 -7

145 lines not on this page · raw message

fs/nfsd/nfsd.h +3 -1
index fe7418c2b88f..a60a2ee2e8c1 100644
@@ -91,7 +91,9 @@ struct nfsdfs_client {
 	void (*cl_release)(struct kref *kref);
 };
 
-struct dentry *nfsd_client_mkdir(struct nfsd_net *nn, struct nfsdfs_client *ncl, u32 id);
+struct nfsdfs_client *get_nfsdfs_client(struct inode *);
+struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
+		struct nfsdfs_client *ncl, u32 id, const struct tree_descr *);
 void nfsd_client_rmdir(struct dentry *dentry);
 
 #if defined(CONFIG_NFSD_V2_ACL) || defined(CONFIG_NFSD_V3_ACL)
Signature · 1 line
2.21.0
#7 lore raw

PATCH 5/12 nfsd: make client/ directory names small ints

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

We want clientid's on the wire to be randomized for reasons explained in ebd7c72c63ac "nfsd: randomize SETCLIENTID reply to help distinguish servers". But I'd rather have mostly small integers for the clients/ directory.

---
 fs/nfsd/netns.h     | 1 +
 fs/nfsd/nfs4state.c | 2 +-
 fs/nfsd/nfsctl.c    | 5 +++--
 3 files changed, 5 insertions(+), 3 deletions(-)
fs/nfsd/netns.h +1 -0
index 46d956676480..888b2891960e 100644
@@ -121,6 +121,7 @@ struct nfsd_net {
 	 */
 	unsigned int max_connections;
 
+	u32 clientid_base;
 	u32 clientid_counter;
 	u32 clverifier_counter;
 
fs/nfsd/nfs4state.c +1 -1
index dfcb90743861..de99bfe3e6e2 100644
@@ -2221,7 +2221,7 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 	clp->cl_cb_session = NULL;
 	clp->net = net;
 	clp->cl_nfsd_dentry = nfsd_client_mkdir(nn, &clp->cl_nfsdfs,
-						clp->cl_clientid.cl_id);
+			clp->cl_clientid.cl_id - nn->clientid_base);
 	if (!clp->cl_nfsd_dentry) {
 		free_client(clp);
 		return NULL;
fs/nfsd/nfsctl.c +3 -2
index 8a2261cdefee..edf4dada42bf 100644
@@ -1212,7 +1212,7 @@ struct dentry *nfsd_client_mkdir(struct nfsd_net *nn, struct nfsdfs_client *ncl,
 {
 	char name[11];
 
-	sprintf(name, "%d", id++);
+	sprintf(name, "%u", id);
 
 	return nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
 }
@@ -1350,7 +1350,8 @@ static __net_init int nfsd_init_net(struct net *net)
 	nn->nfsd4_grace = 90;
 	nn->somebody_reclaimed = false;
 	nn->clverifier_counter = prandom_u32();
-	nn->clientid_counter = prandom_u32();
+	nn->clientid_base = prandom_u32();
+	nn->clientid_counter = nn->clientid_base + 1;
 	nn->s2s_cp_cl_id = nn->clientid_counter++;
 
 	atomic_set(&nn->ntf_refcnt, 0);
Signature · 1 line
2.21.0
#10 lore raw

PATCH 8/12 nfsd: add more information to client info file

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

Add ip address, full client-provided identifier, and minor version. There's much more that could possibly be useful but this is a start.

As with open owners, client identifiers are opaque binary data, but some clients happen to include useful information in ascii.

---
 fs/nfsd/nfs4state.c      | 18 ++++++++++++++++++
 fs/seq_file.c            | 17 +++++++++++++++++
 include/linux/seq_file.h |  2 ++
 3 files changed, 37 insertions(+)
fs/nfsd/nfs4state.c +18 -0
index c665ce9af773..876c035d2a91 100644
@@ -42,6 +42,7 @@
 #include <linux/sunrpc/svcauth_gss.h>
 #include <linux/sunrpc/addr.h>
 #include <linux/jhash.h>
+#include <linux/string_helpers.h>
 #include "xdr4.h"
 #include "xdr4cb.h"
 #include "vfs.h"
@@ -2193,6 +2194,19 @@ find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
 	return s;
 }
 
+static void seq_escape_quotable_mem(struct seq_file *m, char *data, int len)
+{
+	/* List taken from kstrdup_quotable: */
+	seq_escape_mem(m, data, len, ESCAPE_HEX, "\f\n\r\t\v\a\e\\\"");
+}
+
+static void seq_quote_mem(struct seq_file *m, char *data, int len)
+{
+	seq_printf(m, "\"");
+	seq_escape_quotable_mem(m, data, len);
+	seq_printf(m, "\"");
+}
+
 static int client_info_show(struct seq_file *m, void *v)
 {
 	struct inode *inode = m->private;
@@ -2206,6 +2220,10 @@ static int client_info_show(struct seq_file *m, void *v)
 	clp = container_of(nc, struct nfs4_client, cl_nfsdfs);
 	memcpy(&clid, &clp->cl_clientid, sizeof(clid));
 	seq_printf(m, "clientid: %llx\n", clid);
+	seq_printf(m, "address: %pISpc\n", (struct sockaddr *)&clp->cl_addr);
+	seq_printf(m, "name: ");
+	seq_quote_mem(m, clp->cl_name.data, clp->cl_name.len);
+	seq_printf(m, "\nminor version: %d\n", clp->cl_minorversion);
 	drop_client(clp);
 
 	return 0;
fs/seq_file.c +17 -0
index 1dea7a8a5255..bb646e0c5b9c 100644
@@ -383,6 +383,23 @@ void seq_escape(struct seq_file *m, const char *s, const char *esc)
 }
 EXPORT_SYMBOL(seq_escape);
 
+/**
+ * seq_escape_mem - copy data to buffer, escaping some characters
+ *
+ * See string_escape_mem for explanations of the arguments.
+ */
+void seq_escape_mem(struct seq_file *m, const char *src, size_t isz,
+		    unsigned int flags, const char *only)
+{
+	char *buf;
+	size_t size = seq_get_buf(m, &buf);
+	int ret;
+
+	ret = string_escape_mem(src, isz, buf, size, flags, only);
+	seq_commit(m, ret < size ? ret : -1);
+}
+EXPORT_SYMBOL(seq_escape_mem);
+
 void seq_vprintf(struct seq_file *m, const char *f, va_list args)
 {
 	int len;
include/linux/seq_file.h +2 -0
index a121982af0f5..de6ace9e5e6b 100644
@@ -127,6 +127,8 @@ void seq_put_hex_ll(struct seq_file *m, const char *delimiter,
 		    unsigned long long v, unsigned int width);
 
 void seq_escape(struct seq_file *m, const char *s, const char *esc);
+void seq_escape_mem(struct seq_file *m, const char *src, size_t isz,
+		    unsigned int flags, const char *only);
 
 void seq_hex_dump(struct seq_file *m, const char *prefix_str, int prefix_type,
 		  int rowsize, int groupsize, const void *buf, size_t len,
Signature · 1 line
2.21.0
#12 lore raw

PATCH 10/12 nfsd: show lock and deleg stateids

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

These entries are pretty minimal for now.

---
 fs/nfsd/nfs4state.c | 59 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 59 insertions(+)
fs/nfsd/nfs4state.c +59 -0
index 4025b3c692fd..652e0544e058 100644
@@ -2323,12 +2323,66 @@ static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
 	seq_printf(s, ", ");
 	nfs4_show_owner(s, oo);
 	seq_printf(s, " }\n");
+	fput(file);
+
+	return 0;
+}
+
+static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
+{
+	struct nfs4_ol_stateid *ols;
+	struct nfs4_file *nf;
+	struct file *file;
+	struct nfs4_stateowner *oo;
 
+	ols = openlockstateid(st);
+	oo = ols->st_stateowner;
+	nf = st->sc_file;
+	file = find_any_file(nf);
+
+	seq_printf(s, "- 0x%16phN: { type: lock, ", &st->sc_stateid);
+
+	/*
+	 * Note: a lock stateid isn't really the same thing as a lock,
+	 * it's the locking state held by one owner on a file, and there
+	 * may be multiple (or no) lock ranges associated with it.
+	 * (Same for the matter is true of open stateids.)
+	 */
+
+	nfs4_show_superblock(s, file);
+	/* XXX: open stateid? */
+	seq_printf(s, ", ");
+	nfs4_show_owner(s, oo);
+	seq_printf(s, " }\n");
 	fput(file);
 
 	return 0;
 }
 
+static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
+{
+	struct nfs4_delegation *ds;
+	struct nfs4_file *nf;
+	struct file *file;
+
+	ds = delegstateid(st);
+	nf = st->sc_file;
+	file = nf->fi_deleg_file;
+
+	seq_printf(s, "- 0x%16phN: { type: deleg, ", &st->sc_stateid);
+
+	/* Kinda dead code as long as we only support read delegs: */
+	seq_printf(s, "access: %s, ",
+		ds->dl_type == NFS4_OPEN_DELEGATE_READ ? "r" : "w");
+
+	/* XXX: lease time, whether it's being recalled. */
+
+	nfs4_show_superblock(s, file);
+	seq_printf(s, " }\n");
+
+	return 0;
+}
+
 static int states_show(struct seq_file *s, void *v)
 {
 	struct nfs4_stid *st = v;
@@ -2336,9 +2390,14 @@ static int states_show(struct seq_file *s, void *v)
 	switch (st->sc_type) {
 	case NFS4_OPEN_STID:
 		return nfs4_show_open(s, st);
+	case NFS4_LOCK_STID:
+		return nfs4_show_lock(s, st);
+	case NFS4_DELEG_STID:
+		return nfs4_show_deleg(s, st);
 	default:
 		return 0; /* XXX: or SEQ_SKIP? */
 	}
+	/* XXX: copy stateids? */
 }
 
 static struct seq_operations states_seq_ops = {
Signature · 1 line
2.21.0
#9 lore raw

PATCH 7/12 nfsd: copy client's address including port number to cl_addr

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

rpc_copy_addr() copies only the IP address and misses any port numbers. It seems potentially useful to keep the port number around too.

---
 fs/nfsd/nfs4state.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/nfsd/nfs4state.c +1 -1
index 928705fc8ff5..c665ce9af773 100644
@@ -2252,7 +2252,7 @@ static struct nfs4_client *create_client(struct xdr_netobj name,
 	clp->cl_time = get_seconds();
 	clear_bit(0, &clp->cl_cb_slot_busy);
 	copy_verf(clp, verf);
-	rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
+	memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
 	clp->cl_cb_session = NULL;
 	clp->net = net;
 	clp->cl_nfsd_dentry = nfsd_client_mkdir(nn, &clp->cl_nfsdfs,
Signature · 1 line
2.21.0
#0 lore raw

cover exposing knfsd state to userspace

From: "J. Bruce Fields" <[email protected]>

This is still a little rough, but maybe closer. Changes since last time:
- renamed the "opens" file to "states" and added some (minimal) information about lock, delegation, and layout stateids as well as opens.
- converted the states file to a YAML-like format.
- added the ability to remove a client's state by writing "expire\n" into a new nfsd/client/#/ctl file.

Recapping discussion from last time:

The following patches expose information about NFSv4 state held by knfsd on behalf of NFSv4 clients. This especially important for opens, which are currently invisible to userspace on the server, unlike locks (/proc/locks) and local processes' opens (under /proc/<pid>/).

The approach is to add a new directory /proc/fs/nfsd/clients/ with subdirectories for each active NFSv4 client. Each subdirectory has an "info" file with some basic information to help identify the client and a "states" directory that lists the open state held by that client.

Currently these pseudofiles look like:

  # find /proc/fs/nfsd/clients -type f|xargs tail 
  ==> /proc/fs/nfsd/clients/3/opens <==
 - 0x020000006a5fdc5c4ad09d9e01000000: { type: open, access: rw, deny: --, superblock: "fd:10:13649", owner: "open id:\x00\x00\x00&\x00\x00\x00\x00\x00\x0046��QH " }
 - 0x010000006a5fdc5c4ad09d9e03000000: { type: open, access: r-, deny: --, superblock: "fd:10:13650", owner: "open id:\x00\x00\x00&\x00\x00\x00\x00\x00\x0046��QH" }
 - 0x010000006a5fdc5c4ad09d9e04000000: { type: deleg, access: r, superblock: "fd:10:13650" }
 - 0x010000006a5fdc5c4ad09d9e06000000: { type: lock, superblock: "fd:10:13649", owner: "lock id:\x00\x00\x00&\x00\x00\x00\x00\x00\x00\x00\x00" }

==> /proc/fs/nfsd/clients/3/info <==
clientid: 6debfb505cc0cd36
address: 192.168.122.36:907
name: "Linux NFSv4.2 test2.fieldses.org"
minor version: 2

I'm conflicted about how I'm representing stateowners and client names, both opaque byte-streams in the protocol but that often include human-readable ascii.

Possibly also todo:
- add some information about krb5 principals to the clients file.
- add information about the stateids used to represent asynchronous copies. They're a little different from the other stateids and might end up in a separate "copies" file,
- this duplicates some functionality of the little-used fault injection code; could we replace it entirely?
- some of the bits of filesystem code could probably be shared with rpc_pipefs and libfs.

--b.

J. Bruce Fields (10):
nfsd: persist nfsd filesystem across mounts
nfsd: rename cl_refcount
nfsd4: use reference count to free client
nfsd: add nfsd/clients directory
nfsd: make client/ directory names small ints
rpc: replace rpc_filelist by tree_descr
nfsd4: add a client info file
nfsd4: add file to display list of client's opens
nfsd: expose some more information about NFSv4 opens
nfsd: add more information to client info file

 fs/nfsd/netns.h                |   6 +
 fs/nfsd/nfs4state.c            | 228 ++++++++++++++++++++++++++++++---
 fs/nfsd/nfsctl.c               | 225 +++++++++++++++++++++++++++++++-
 fs/nfsd/nfsd.h                 |  11 ++
 fs/nfsd/state.h                |   9 +-
 fs/seq_file.c                  |  17 +++
 include/linux/seq_file.h       |   2 +
 include/linux/string_helpers.h |   1 +
 lib/string_helpers.c           |   5 +-
 net/sunrpc/rpc_pipe.c          |  37 ++----
 10 files changed, 491 insertions(+), 50 deletions(-)
Signature · 28 lines
2.20.1


J. Bruce Fields (12):
  nfsd: persist nfsd filesystem across mounts
  nfsd: rename cl_refcount
  nfsd4: use reference count to free client
  nfsd: add nfsd/clients directory
  nfsd: make client/ directory names small ints
  nfsd4: add a client info file
  nfsd: copy client's address including port number to cl_addr
  nfsd: add more information to client info file
  nfsd4: add file to display list of client's opens
  nfsd: show lock and deleg stateids
  nfsd4: show layout stateids
  nfsd: allow forced expiration of NFSv4 clients

 fs/nfsd/netns.h          |   6 +
 fs/nfsd/nfs4state.c      | 418 ++++++++++++++++++++++++++++++++++++---
 fs/nfsd/nfsctl.c         | 225 ++++++++++++++++++++-
 fs/nfsd/nfsd.h           |  11 ++
 fs/nfsd/state.h          |   7 +-
 fs/seq_file.c            |  17 ++
 include/linux/seq_file.h |   2 +
 7 files changed, 661 insertions(+), 25 deletions(-)

-- 
2.21.0
#4 lore raw

PATCH 2/12 nfsd: rename cl_refcount

From: "J. Bruce Fields" <[email protected]>

Rename this to a more descriptive name.

Also, I'm going to want a second refcount with a slightly different use.

---
 fs/nfsd/nfs4state.c | 26 +++++++++++++-------------
 fs/nfsd/state.h     |  2 +-
 2 files changed, 14 insertions(+), 14 deletions(-)
fs/nfsd/nfs4state.c +13 -13
index 6a45fb00c5fc..9dab61bbd256 100644
@@ -137,7 +137,7 @@ static __be32 get_client_locked(struct nfs4_client *clp)
 
 	if (is_client_expired(clp))
 		return nfserr_expired;
-	atomic_inc(&clp->cl_refcount);
+	atomic_inc(&clp->cl_rpc_users);
 	return nfs_ok;
 }
 
@@ -169,7 +169,7 @@ static void put_client_renew_locked(struct nfs4_client *clp)
 
 	lockdep_assert_held(&nn->client_lock);
 
-	if (!atomic_dec_and_test(&clp->cl_refcount))
+	if (!atomic_dec_and_test(&clp->cl_rpc_users))
 		return;
 	if (!is_client_expired(clp))
 		renew_client_locked(clp);
@@ -179,7 +179,7 @@ static void put_client_renew(struct nfs4_client *clp)
 {
 	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 
-	if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
+	if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
 		return;
 	if (!is_client_expired(clp))
 		renew_client_locked(clp);
@@ -1847,7 +1847,7 @@ static struct nfs4_client *alloc_client(struct xdr_netobj name)
 	clp->cl_name.len = name.len;
 	INIT_LIST_HEAD(&clp->cl_sessions);
 	idr_init(&clp->cl_stateids);
-	atomic_set(&clp->cl_refcount, 0);
+	atomic_set(&clp->cl_rpc_users, 0);
 	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
 	INIT_LIST_HEAD(&clp->cl_idhash);
 	INIT_LIST_HEAD(&clp->cl_openowners);
@@ -1926,7 +1926,7 @@ unhash_client(struct nfs4_client *clp)
 
 static __be32 mark_client_expired_locked(struct nfs4_client *clp)
 {
-	if (atomic_read(&clp->cl_refcount))
+	if (atomic_read(&clp->cl_rpc_users))
 		return nfserr_jukebox;
 	unhash_client_locked(clp);
 	return nfs_ok;
@@ -4066,7 +4066,7 @@ static __be32 lookup_clientid(clientid_t *clid,
 		spin_unlock(&nn->client_lock);
 		return nfserr_expired;
 	}
-	atomic_inc(&found->cl_refcount);
+	atomic_inc(&found->cl_rpc_users);
 	spin_unlock(&nn->client_lock);
 
 	/* Cache the nfs4_client in cstate! */
@@ -6550,7 +6550,7 @@ nfs4_check_open_reclaim(clientid_t *clid,
 static inline void
 put_client(struct nfs4_client *clp)
 {
-	atomic_dec(&clp->cl_refcount);
+	atomic_dec(&clp->cl_rpc_users);
 }
 
 static struct nfs4_client *
@@ -6668,7 +6668,7 @@ nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid *lst,
 		return;
 
 	lockdep_assert_held(&nn->client_lock);
-	atomic_inc(&clp->cl_refcount);
+	atomic_inc(&clp->cl_rpc_users);
 	list_add(&lst->st_locks, collect);
 }
 
@@ -6697,7 +6697,7 @@ static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
 				 * Despite the fact that these functions deal
 				 * with 64-bit integers for "count", we must
 				 * ensure that it doesn't blow up the
-				 * clp->cl_refcount. Throw a warning if we
+				 * clp->cl_rpc_users. Throw a warning if we
 				 * start to approach INT_MAX here.
 				 */
 				WARN_ON_ONCE(count == (INT_MAX / 2));
@@ -6821,7 +6821,7 @@ nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
 		if (func) {
 			func(oop);
 			if (collect) {
-				atomic_inc(&clp->cl_refcount);
+				atomic_inc(&clp->cl_rpc_users);
 				list_add(&oop->oo_perclient, collect);
 			}
 		}
@@ -6829,7 +6829,7 @@ nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
 		/*
 		 * Despite the fact that these functions deal with
 		 * 64-bit integers for "count", we must ensure that
-		 * it doesn't blow up the clp->cl_refcount. Throw a
+		 * it doesn't blow up the clp->cl_rpc_users. Throw a
 		 * warning if we start to approach INT_MAX here.
 		 */
 		WARN_ON_ONCE(count == (INT_MAX / 2));
@@ -6959,7 +6959,7 @@ static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
 			if (dp->dl_time != 0)
 				continue;
 
-			atomic_inc(&clp->cl_refcount);
+			atomic_inc(&clp->cl_rpc_users);
 			WARN_ON(!unhash_delegation_locked(dp));
 			list_add(&dp->dl_recall_lru, victims);
 		}
@@ -6967,7 +6967,7 @@ static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
 		/*
 		 * Despite the fact that these functions deal with
 		 * 64-bit integers for "count", we must ensure that
-		 * it doesn't blow up the clp->cl_refcount. Throw a
+		 * it doesn't blow up the clp->cl_rpc_users. Throw a
 		 * warning if we start to approach INT_MAX here.
 		 */
 		WARN_ON_ONCE(count == (INT_MAX / 2));
fs/nfsd/state.h +1 -1
index 396c76755b03..aa9f1676e88a 100644
@@ -346,7 +346,7 @@ struct nfs4_client {
 	struct nfsd4_clid_slot	cl_cs_slot;	/* create_session slot */
 	u32			cl_exchange_flags;
 	/* number of rpc's in progress over an associated session: */
-	atomic_t		cl_refcount;
+	atomic_t		cl_rpc_users;
 	struct nfs4_op_map      cl_spo_must_allow;
 
 	/* for nfs41 callbacks */
Signature · 1 line
2.21.0
#15 lore raw

PATCH 2/2 nfsd: fh_drop_write in nfsd_unlink

In reply to J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

fh_want_write() can now be called twice, but I'm also fixing up the callers not to do that.

Other cases include setattr and create.

---
 fs/nfsd/vfs.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
fs/nfsd/vfs.c +5 -3
index 7dc98e14655d..fc24ee47eab5 100644
@@ -1786,12 +1786,12 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 	rdentry = lookup_one_len(fname, dentry, flen);
 	host_err = PTR_ERR(rdentry);
 	if (IS_ERR(rdentry))
-		goto out_nfserr;
+		goto out_drop_write;
 
 	if (d_really_is_negative(rdentry)) {
 		dput(rdentry);
-		err = nfserr_noent;
-		goto out;
+		host_err = -ENOENT;
+		goto out_drop_write;
 	}
 
 	if (!type)
@@ -1805,6 +1805,8 @@ nfsd_unlink(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 		host_err = commit_metadata(fhp);
 	dput(rdentry);
 
+out_drop_write:
+	fh_drop_write(fhp);
 out_nfserr:
 	err = nfserrno(host_err);
 out:
Signature · 1 line
2.21.0
#2 lore raw

Re: [PATCH 1/2] nfsd: allow fh_want_write to be called twice

In reply to J. Bruce Fields

Ugh, sorry, ignore the two old patches that got sent with the new series.

--b.

Quoted 44 lines · J. Bruce Fields

From: "J. Bruce Fields" <[email protected]>

A fuzzer recently triggered lockdep warnings about potential sb_writers deadlocks caused by fh_want_write().

Looks like we aren't careful to pair each fh_want_write() with an fh_drop_write().

It's not normally a problem since fh_put() will call fh_drop_write() for us. And was OK for NFSv3 where we'd do one operation that might call fh_want_write(), and then put the filehandle.

But an NFSv4 protocol fuzzer can do weird things like call unlink twice in a compound, and then we get into trouble.

I'm a little worried about this approach of just leaving everything to fh_put(). But I think there are probably a lot of fh_want_write()/fh_drop_write() imbalances so for now I think we need it to be more forgiving.

---
 fs/nfsd/vfs.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
fs/nfsd/vfs.h +4 -1
index a7e107309f76..db351247892d 100644
@@ -120,8 +120,11 @@ void		nfsd_put_raparams(struct file *file, struct raparms *ra);
 
 static inline int fh_want_write(struct svc_fh *fh)
 {
-	int ret = mnt_want_write(fh->fh_export->ex_path.mnt);
+	int ret;
 
+	if (fh->fh_want_write)
+		return 0;
+	ret = mnt_want_write(fh->fh_export->ex_path.mnt);
 	if (!ret)
 		fh->fh_want_write = true;
 	return ret;
Signature · 1 line
2.21.0