[PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[Stanislav Kinsbursky]

New routine is responsible for service registration in specified network
context.
The idea is to separate service creation from per-net operations.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>

---
 fs/lockd/svc.c             |    6 ++++++
 fs/nfs/callback.c          |    8 ++++++++
 fs/nfsd/nfssvc.c           |    9 +++++++++
 include/linux/sunrpc/svc.h |    1 +
 net/sunrpc/svc.c           |   19 ++++++++++---------
 5 files changed, 34 insertions(+), 9 deletions(-)

diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index 1ead075..b7e92ed 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -324,6 +324,12 @@ int lockd_up(struct net *net)
 		goto out;
 	}
 
+	error = svc_bind(serv, net);
+	if (error < 0) {
+		printk(KERN_WARNING "lockd_up: bind service failed\n");
+		goto destroy_and_out;
+	}
+
 	error = make_socks(serv, net);
 	if (error < 0)
 		goto destroy_and_out;
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index eb95f50..26b38fb 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -17,6 +17,7 @@
 #include <linux/kthread.h>
 #include <linux/sunrpc/svcauth_gss.h>
 #include <linux/sunrpc/bc_xprt.h>
+#include <linux/nsproxy.h>
 
 #include <net/inet_sock.h>
 
@@ -253,6 +254,7 @@ int nfs_callback_up(u32 minorversion, struct rpc_xprt *xprt)
 	char svc_name[12];
 	int ret = 0;
 	int minorversion_setup;
+	struct net *net = current->nsproxy->net_ns;
 
 	mutex_lock(&nfs_callback_mutex);
 	if (cb_info->users++ || cb_info->task != NULL) {
@@ -265,6 +267,12 @@ int nfs_callback_up(u32 minorversion, struct rpc_xprt *xprt)
 		goto out_err;
 	}
 
+	ret = svc_bind(serv, net);
+	if (ret < 0) {
+		printk(KERN_WARNING "NFS: bind callback service failed\n");
+		goto out_err;
+	}
+
 	minorversion_setup =  nfs_minorversion_callback_svc_setup(minorversion,
 					serv, xprt, &rqstp, &callback_svc);
 	if (!minorversion_setup) {
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index cb4d51d..0762f3c 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/fs_struct.h>
 #include <linux/swap.h>
+#include <linux/nsproxy.h>
 
 #include <linux/sunrpc/stats.h>
 #include <linux/sunrpc/svcsock.h>
@@ -330,6 +331,8 @@ static int nfsd_get_default_max_blksize(void)
 
 int nfsd_create_serv(void)
 {
+	int error;
+
 	WARN_ON(!mutex_is_locked(&nfsd_mutex));
 	if (nfsd_serv) {
 		svc_get(nfsd_serv);
@@ -343,6 +346,12 @@ int nfsd_create_serv(void)
 	if (nfsd_serv == NULL)
 		return -ENOMEM;
 
+	error = svc_bind(nfsd_serv, current->nsproxy->net_ns);
+	if (error < 0) {
+		svc_destroy(nfsd_serv);
+		return error;
+	}
+
 	set_max_drc();
 	do_gettimeofday(&nfssvc_boot);		/* record boot time */
 	return 0;
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 51b29ac..2b43e02 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -416,6 +416,7 @@ struct svc_procedure {
  */
 int svc_rpcb_setup(struct svc_serv *serv, struct net *net);
 void svc_rpcb_cleanup(struct svc_serv *serv, struct net *net);
+int svc_bind(struct svc_serv *serv, struct net *net);
 struct svc_serv *svc_create(struct svc_program *, unsigned int,
 			    void (*shutdown)(struct svc_serv *, struct net *net));
 struct svc_rqst *svc_prepare_thread(struct svc_serv *serv,
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 4153846..e6d542c 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -407,6 +407,14 @@ static int svc_uses_rpcbind(struct svc_serv *serv)
 	return 0;
 }
 
+int svc_bind(struct svc_serv *serv, struct net *net)
+{
+	if (!svc_uses_rpcbind(serv))
+		return 0;
+	return svc_rpcb_setup(serv, net);
+}
+EXPORT_SYMBOL_GPL(svc_bind);
+
 /*
  * Create an RPC service
  */
@@ -471,15 +479,8 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
 		spin_lock_init(&pool->sp_lock);
 	}
 
-	if (svc_uses_rpcbind(serv)) {
-		if (svc_rpcb_setup(serv, current->nsproxy->net_ns) < 0) {
-			kfree(serv->sv_pools);
-			kfree(serv);
-			return NULL;
-		}
-		if (!serv->sv_shutdown)
-			serv->sv_shutdown = svc_rpcb_cleanup;
-	}
+	if (svc_uses_rpcbind(serv) && (!serv->sv_shutdown))
+		serv->sv_shutdown = svc_rpcb_cleanup;
 
 	return serv;
 }

3.4.z nfsd backports

[J. Bruce Fields]

These three patches are all upstream, but required some conflict
resolution to backport, so I'm sending backported versions.

--b.

[PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[J. Bruce Fields]

From: Stanislav Kinsbursky <skinsbursky@parallels.com>

commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.

This new routine is responsible for service registration in a specified
network context.

The idea is to separate service creation from per-net operations.

Note also: since registering service with svc_bind() can fail, the
service will be destroyed and during destruction it will try to
unregister itself from rpcbind. In this case unregistration has to be
skipped.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/lockd/svc.c             |    6 ++++++
 fs/nfs/callback.c          |    8 ++++++++
 fs/nfsd/nfssvc.c           |    9 +++++++++
 include/linux/sunrpc/svc.h |    1 +
 net/sunrpc/rpcb_clnt.c     |   12 +++++++-----
 net/sunrpc/svc.c           |   19 ++++++++++---------
 6 files changed, 41 insertions(+), 14 deletions(-)

diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index f49b9af..b1d0708 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -325,6 +325,12 @@ int lockd_up(void)
 		goto out;
 	}
 
+	error = svc_bind(serv, net);
+	if (error < 0) {
+		printk(KERN_WARNING "lockd_up: bind service failed\n");
+		goto destroy_and_out;
+	}
+
 	error = make_socks(serv, net);
 	if (error < 0)
 		goto destroy_and_out;
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index eb95f50..26b38fb 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -17,6 +17,7 @@
 #include <linux/kthread.h>
 #include <linux/sunrpc/svcauth_gss.h>
 #include <linux/sunrpc/bc_xprt.h>
+#include <linux/nsproxy.h>
 
 #include <net/inet_sock.h>
 
@@ -253,6 +254,7 @@ int nfs_callback_up(u32 minorversion, struct rpc_xprt *xprt)
 	char svc_name[12];
 	int ret = 0;
 	int minorversion_setup;
+	struct net *net = current->nsproxy->net_ns;
 
 	mutex_lock(&nfs_callback_mutex);
 	if (cb_info->users++ || cb_info->task != NULL) {
@@ -265,6 +267,12 @@ int nfs_callback_up(u32 minorversion, struct rpc_xprt *xprt)
 		goto out_err;
 	}
 
+	ret = svc_bind(serv, net);
+	if (ret < 0) {
+		printk(KERN_WARNING "NFS: bind callback service failed\n");
+		goto out_err;
+	}
+
 	minorversion_setup =  nfs_minorversion_callback_svc_setup(minorversion,
 					serv, xprt, &rqstp, &callback_svc);
 	if (!minorversion_setup) {
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index 28dfad3..a6461f3 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/fs_struct.h>
 #include <linux/swap.h>
+#include <linux/nsproxy.h>
 
 #include <linux/sunrpc/stats.h>
 #include <linux/sunrpc/svcsock.h>
@@ -330,6 +331,8 @@ static int nfsd_get_default_max_blksize(void)
 
 int nfsd_create_serv(void)
 {
+	int error;
+
 	WARN_ON(!mutex_is_locked(&nfsd_mutex));
 	if (nfsd_serv) {
 		svc_get(nfsd_serv);
@@ -343,6 +346,12 @@ int nfsd_create_serv(void)
 	if (nfsd_serv == NULL)
 		return -ENOMEM;
 
+	error = svc_bind(nfsd_serv, current->nsproxy->net_ns);
+	if (error < 0) {
+		svc_destroy(nfsd_serv);
+		return error;
+	}
+
 	set_max_drc();
 	do_gettimeofday(&nfssvc_boot);		/* record boot time */
 	return 0;
diff --git a/include/linux/sunrpc/svc.h b/include/linux/sunrpc/svc.h
index 51b29ac..2b43e02 100644
--- a/include/linux/sunrpc/svc.h
+++ b/include/linux/sunrpc/svc.h
@@ -416,6 +416,7 @@ struct svc_procedure {
  */
 int svc_rpcb_setup(struct svc_serv *serv, struct net *net);
 void svc_rpcb_cleanup(struct svc_serv *serv, struct net *net);
+int svc_bind(struct svc_serv *serv, struct net *net);
 struct svc_serv *svc_create(struct svc_program *, unsigned int,
 			    void (*shutdown)(struct svc_serv *, struct net *net));
 struct svc_rqst *svc_prepare_thread(struct svc_serv *serv,
diff --git a/net/sunrpc/rpcb_clnt.c b/net/sunrpc/rpcb_clnt.c
index 78ac39f..4c38b33 100644
--- a/net/sunrpc/rpcb_clnt.c
+++ b/net/sunrpc/rpcb_clnt.c
@@ -180,14 +180,16 @@ void rpcb_put_local(struct net *net)
 	struct sunrpc_net *sn = net_generic(net, sunrpc_net_id);
 	struct rpc_clnt *clnt = sn->rpcb_local_clnt;
 	struct rpc_clnt *clnt4 = sn->rpcb_local_clnt4;
-	int shutdown;
+	int shutdown = 0;
 
 	spin_lock(&sn->rpcb_clnt_lock);
-	if (--sn->rpcb_users == 0) {
-		sn->rpcb_local_clnt = NULL;
-		sn->rpcb_local_clnt4 = NULL;
+	if (sn->rpcb_users) {
+		if (--sn->rpcb_users == 0) {
+			sn->rpcb_local_clnt = NULL;
+			sn->rpcb_local_clnt4 = NULL;
+		}
+		shutdown = !sn->rpcb_users;
 	}
-	shutdown = !sn->rpcb_users;
 	spin_unlock(&sn->rpcb_clnt_lock);
 
 	if (shutdown) {
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 4153846..e6d542c 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -407,6 +407,14 @@ static int svc_uses_rpcbind(struct svc_serv *serv)
 	return 0;
 }
 
+int svc_bind(struct svc_serv *serv, struct net *net)
+{
+	if (!svc_uses_rpcbind(serv))
+		return 0;
+	return svc_rpcb_setup(serv, net);
+}
+EXPORT_SYMBOL_GPL(svc_bind);
+
 /*
  * Create an RPC service
  */
@@ -471,15 +479,8 @@ __svc_create(struct svc_program *prog, unsigned int bufsize, int npools,
 		spin_lock_init(&pool->sp_lock);
 	}
 
-	if (svc_uses_rpcbind(serv)) {
-		if (svc_rpcb_setup(serv, current->nsproxy->net_ns) < 0) {
-			kfree(serv->sv_pools);
-			kfree(serv);
-			return NULL;
-		}
-		if (!serv->sv_shutdown)
-			serv->sv_shutdown = svc_rpcb_cleanup;
-	}
+	if (svc_uses_rpcbind(serv) && (!serv->sv_shutdown))
+		serv->sv_shutdown = svc_rpcb_cleanup;
 
 	return serv;
 }
-- 
1.7.9.5

[PATCH 2/3] SUNRPC: move per-net operations from svc_destroy()

[J. Bruce Fields]

From: Stanislav Kinsbursky <skinsbursky@parallels.com>

commit 786185b5f8abefa6a8a16695bb4a59c164d5a071 upstream.

The idea is to separate service destruction and per-net operations,
because these are two different things and the mix looks ugly.

Notes:

1) For NFS server this patch looks ugly (sorry for that). But these
place will be rewritten soon during NFSd containerization.

2) LockD per-net counter increase int lockd_up() was moved prior to
make_socks() to make lockd_down_net() call safe in case of error.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/lockd/svc.c    |   27 +++++++++++++++------------
 fs/nfs/callback.c |    3 +++
 fs/nfsd/nfsctl.c  |   12 +++++++++---
 fs/nfsd/nfssvc.c  |   14 ++++++++++++++
 net/sunrpc/svc.c  |    4 ----
 5 files changed, 41 insertions(+), 19 deletions(-)

diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index b1d0708..f1b3cce 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -257,7 +257,7 @@ static int lockd_up_net(struct net *net)
 	struct svc_serv *serv = nlmsvc_rqst->rq_server;
 	int error;
 
-	if (ln->nlmsvc_users)
+	if (ln->nlmsvc_users++)
 		return 0;
 
 	error = svc_rpcb_setup(serv, net);
@@ -272,6 +272,7 @@ static int lockd_up_net(struct net *net)
 err_socks:
 	svc_rpcb_cleanup(serv, net);
 err_rpcb:
+	ln->nlmsvc_users--;
 	return error;
 }
 
@@ -300,6 +301,7 @@ int lockd_up(void)
 	struct svc_serv *serv;
 	int		error = 0;
 	struct net *net = current->nsproxy->net_ns;
+	struct lockd_net *ln = net_generic(net, lockd_net_id);
 
 	mutex_lock(&nlmsvc_mutex);
 	/*
@@ -331,9 +333,11 @@ int lockd_up(void)
 		goto destroy_and_out;
 	}
 
+	ln->nlmsvc_users++;
+
 	error = make_socks(serv, net);
 	if (error < 0)
-		goto destroy_and_out;
+		goto err_start;
 
 	/*
 	 * Create the kernel thread and wait for it to start.
@@ -345,7 +349,7 @@ int lockd_up(void)
 		printk(KERN_WARNING
 			"lockd_up: svc_rqst allocation failed, error=%d\n",
 			error);
-		goto destroy_and_out;
+		goto err_start;
 	}
 
 	svc_sock_update_bufs(serv);
@@ -359,7 +363,7 @@ int lockd_up(void)
 		nlmsvc_rqst = NULL;
 		printk(KERN_WARNING
 			"lockd_up: kthread_run failed, error=%d\n", error);
-		goto destroy_and_out;
+		goto err_start;
 	}
 
 	/*
@@ -369,14 +373,14 @@ int lockd_up(void)
 destroy_and_out:
 	svc_destroy(serv);
 out:
-	if (!error) {
-		struct lockd_net *ln = net_generic(net, lockd_net_id);
-
-		ln->nlmsvc_users++;
+	if (!error)
 		nlmsvc_users++;
-	}
 	mutex_unlock(&nlmsvc_mutex);
 	return error;
+
+err_start:
+	lockd_down_net(net);
+	goto destroy_and_out;
 }
 EXPORT_SYMBOL_GPL(lockd_up);
 
@@ -387,11 +391,10 @@ void
 lockd_down(void)
 {
 	mutex_lock(&nlmsvc_mutex);
+	lockd_down_net(current->nsproxy->net_ns);
 	if (nlmsvc_users) {
-		if (--nlmsvc_users) {
-			lockd_down_net(current->nsproxy->net_ns);
+		if (--nlmsvc_users)
 			goto out;
-		}
 	} else {
 		printk(KERN_ERR "lockd_down: no users! task=%p\n",
 			nlmsvc_task);
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index 26b38fb..cff3940 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -314,6 +314,8 @@ out_err:
 	dprintk("NFS: Couldn't create callback socket or server thread; "
 		"err = %d\n", ret);
 	cb_info->users--;
+	if (serv)
+		svc_shutdown_net(serv, net);
 	goto out;
 }
 
@@ -328,6 +330,7 @@ void nfs_callback_down(int minorversion)
 	cb_info->users--;
 	if (cb_info->users == 0 && cb_info->task != NULL) {
 		kthread_stop(cb_info->task);
+		svc_shutdown_net(cb_info->serv, current->nsproxy->net_ns);
 		svc_exit_thread(cb_info->rqst);
 		cb_info->serv = NULL;
 		cb_info->rqst = NULL;
diff --git a/fs/nfsd/nfsctl.c b/fs/nfsd/nfsctl.c
index 2c53be6..3ab12eb 100644
--- a/fs/nfsd/nfsctl.c
+++ b/fs/nfsd/nfsctl.c
@@ -651,6 +651,7 @@ static ssize_t __write_ports_addfd(char *buf)
 {
 	char *mesg = buf;
 	int fd, err;
+	struct net *net = &init_net;
 
 	err = get_int(&mesg, &fd);
 	if (err != 0 || fd < 0)
@@ -662,6 +663,8 @@ static ssize_t __write_ports_addfd(char *buf)
 
 	err = svc_addsock(nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT);
 	if (err < 0) {
+		if (nfsd_serv->sv_nrthreads == 1)
+			svc_shutdown_net(nfsd_serv, net);
 		svc_destroy(nfsd_serv);
 		return err;
 	}
@@ -699,6 +702,7 @@ static ssize_t __write_ports_addxprt(char *buf)
 	char transport[16];
 	struct svc_xprt *xprt;
 	int port, err;
+	struct net *net = &init_net;
 
 	if (sscanf(buf, "%15s %4u", transport, &port) != 2)
 		return -EINVAL;
@@ -710,12 +714,12 @@ static ssize_t __write_ports_addxprt(char *buf)
 	if (err != 0)
 		return err;
 
-	err = svc_create_xprt(nfsd_serv, transport, &init_net,
+	err = svc_create_xprt(nfsd_serv, transport, net,
 				PF_INET, port, SVC_SOCK_ANONYMOUS);
 	if (err < 0)
 		goto out_err;
 
-	err = svc_create_xprt(nfsd_serv, transport, &init_net,
+	err = svc_create_xprt(nfsd_serv, transport, net,
 				PF_INET6, port, SVC_SOCK_ANONYMOUS);
 	if (err < 0 && err != -EAFNOSUPPORT)
 		goto out_close;
@@ -724,12 +728,14 @@ static ssize_t __write_ports_addxprt(char *buf)
 	nfsd_serv->sv_nrthreads--;
 	return 0;
 out_close:
-	xprt = svc_find_xprt(nfsd_serv, transport, &init_net, PF_INET, port);
+	xprt = svc_find_xprt(nfsd_serv, transport, net, PF_INET, port);
 	if (xprt != NULL) {
 		svc_close_xprt(xprt);
 		svc_xprt_put(xprt);
 	}
 out_err:
+	if (nfsd_serv->sv_nrthreads == 1)
+		svc_shutdown_net(nfsd_serv, net);
 	svc_destroy(nfsd_serv);
 	return err;
 }
diff --git a/fs/nfsd/nfssvc.c b/fs/nfsd/nfssvc.c
index a6461f3..da50e1c 100644
--- a/fs/nfsd/nfssvc.c
+++ b/fs/nfsd/nfssvc.c
@@ -382,6 +382,7 @@ int nfsd_set_nrthreads(int n, int *nthreads)
 	int i = 0;
 	int tot = 0;
 	int err = 0;
+	struct net *net = &init_net;
 
 	WARN_ON(!mutex_is_locked(&nfsd_mutex));
 
@@ -426,6 +427,9 @@ int nfsd_set_nrthreads(int n, int *nthreads)
 		if (err)
 			break;
 	}
+
+	if (nfsd_serv->sv_nrthreads == 1)
+		svc_shutdown_net(nfsd_serv, net);
 	svc_destroy(nfsd_serv);
 
 	return err;
@@ -441,6 +445,7 @@ nfsd_svc(unsigned short port, int nrservs)
 {
 	int	error;
 	bool	nfsd_up_before;
+	struct net *net = &init_net;
 
 	mutex_lock(&nfsd_mutex);
 	dprintk("nfsd: creating service\n");
@@ -473,6 +478,8 @@ out_shutdown:
 	if (error < 0 && !nfsd_up_before)
 		nfsd_shutdown();
 out_destroy:
+	if (nfsd_serv->sv_nrthreads == 1)
+		svc_shutdown_net(nfsd_serv, net);
 	svc_destroy(nfsd_serv);		/* Release server */
 out:
 	mutex_unlock(&nfsd_mutex);
@@ -556,6 +563,9 @@ nfsd(void *vrqstp)
 	nfsdstats.th_cnt --;
 
 out:
+	if (rqstp->rq_server->sv_nrthreads == 1)
+		svc_shutdown_net(rqstp->rq_server, &init_net);
+
 	/* Release the thread */
 	svc_exit_thread(rqstp);
 
@@ -668,8 +678,12 @@ int nfsd_pool_stats_open(struct inode *inode, struct file *file)
 int nfsd_pool_stats_release(struct inode *inode, struct file *file)
 {
 	int ret = seq_release(inode, file);
+	struct net *net = &init_net;
+
 	mutex_lock(&nfsd_mutex);
 	/* this function really, really should have been called svc_put() */
+	if (nfsd_serv->sv_nrthreads == 1)
+		svc_shutdown_net(nfsd_serv, net);
 	svc_destroy(nfsd_serv);
 	mutex_unlock(&nfsd_mutex);
 	return ret;
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index e6d542c..b7210f5 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -537,8 +537,6 @@ EXPORT_SYMBOL_GPL(svc_shutdown_net);
 void
 svc_destroy(struct svc_serv *serv)
 {
-	struct net *net = current->nsproxy->net_ns;
-
 	dprintk("svc: svc_destroy(%s, %d)\n",
 				serv->sv_program->pg_name,
 				serv->sv_nrthreads);
@@ -553,8 +551,6 @@ svc_destroy(struct svc_serv *serv)
 
 	del_timer_sync(&serv->sv_temptimer);
 
-	svc_shutdown_net(serv, net);
-
 	/*
 	 * The last user is gone and thus all sockets have to be destroyed to
 	 * the point. Check this.
-- 
1.7.9.5

[PATCH 3/3] NFS: hard-code init_net for NFS callback transports

[J. Bruce Fields]

From: Stanislav Kinsbursky <skinsbursky@parallels.com>

commit 12918b10d59e975fd5241eef03ef9e6d5ea3dcfe upstream.

In case of destroying mount namespace on child reaper exit, nsproxy is zeroed
to the point already. So, dereferencing of it is invalid.
This patch hard-code "init_net" for all network namespace references for NFS
callback services. This will be fixed with proper NFS callback
containerization.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/nfs/callback.c |   11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
index cff3940..38a44c6 100644
--- a/fs/nfs/callback.c
+++ b/fs/nfs/callback.c
@@ -17,7 +17,6 @@
 #include <linux/kthread.h>
 #include <linux/sunrpc/svcauth_gss.h>
 #include <linux/sunrpc/bc_xprt.h>
-#include <linux/nsproxy.h>
 
 #include <net/inet_sock.h>
 
@@ -107,7 +106,7 @@ nfs4_callback_up(struct svc_serv *serv, struct rpc_xprt *xprt)
 {
 	int ret;
 
-	ret = svc_create_xprt(serv, "tcp", xprt->xprt_net, PF_INET,
+	ret = svc_create_xprt(serv, "tcp", &init_net, PF_INET,
 				nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
 	if (ret <= 0)
 		goto out_err;
@@ -115,7 +114,7 @@ nfs4_callback_up(struct svc_serv *serv, struct rpc_xprt *xprt)
 	dprintk("NFS: Callback listener port = %u (af %u)\n",
 			nfs_callback_tcpport, PF_INET);
 
-	ret = svc_create_xprt(serv, "tcp", xprt->xprt_net, PF_INET6,
+	ret = svc_create_xprt(serv, "tcp", &init_net, PF_INET6,
 				nfs_callback_set_tcpport, SVC_SOCK_ANONYMOUS);
 	if (ret > 0) {
 		nfs_callback_tcpport6 = ret;
@@ -184,7 +183,7 @@ nfs41_callback_up(struct svc_serv *serv, struct rpc_xprt *xprt)
 	 * fore channel connection.
 	 * Returns the input port (0) and sets the svc_serv bc_xprt on success
 	 */
-	ret = svc_create_xprt(serv, "tcp-bc", xprt->xprt_net, PF_INET, 0,
+	ret = svc_create_xprt(serv, "tcp-bc", &init_net, PF_INET, 0,
 			      SVC_SOCK_ANONYMOUS);
 	if (ret < 0) {
 		rqstp = ERR_PTR(ret);
@@ -254,7 +253,7 @@ int nfs_callback_up(u32 minorversion, struct rpc_xprt *xprt)
 	char svc_name[12];
 	int ret = 0;
 	int minorversion_setup;
-	struct net *net = current->nsproxy->net_ns;
+	struct net *net = &init_net;
 
 	mutex_lock(&nfs_callback_mutex);
 	if (cb_info->users++ || cb_info->task != NULL) {
@@ -330,7 +329,7 @@ void nfs_callback_down(int minorversion)
 	cb_info->users--;
 	if (cb_info->users == 0 && cb_info->task != NULL) {
 		kthread_stop(cb_info->task);
-		svc_shutdown_net(cb_info->serv, current->nsproxy->net_ns);
+		svc_shutdown_net(cb_info->serv, &init_net);
 		svc_exit_thread(cb_info->rqst);
 		cb_info->serv = NULL;
 		cb_info->rqst = NULL;
-- 
1.7.9.5

[PATCH 4/3] lockd: hard-code init_net in lockd case as well

[J. Bruce Fields]

From: "J. Bruce Fields" <bfields@redhat.com>

The previous patch missed lockd_down, which has the same problem.
Symptoms were crashes on unmount or reboot.

Note this was fixed in a different way upstream.

Acked-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
---
 fs/lockd/svc.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/lockd/svc.c b/fs/lockd/svc.c
index f1b3cce..c933df4 100644
--- a/fs/lockd/svc.c
+++ b/fs/lockd/svc.c
@@ -300,7 +300,7 @@ int lockd_up(void)
 {
 	struct svc_serv *serv;
 	int		error = 0;
-	struct net *net = current->nsproxy->net_ns;
+	struct net *net = &init_net;
 	struct lockd_net *ln = net_generic(net, lockd_net_id);
 
 	mutex_lock(&nlmsvc_mutex);
@@ -391,7 +391,7 @@ void
 lockd_down(void)
 {
 	mutex_lock(&nlmsvc_mutex);
-	lockd_down_net(current->nsproxy->net_ns);
+	lockd_down_net(&init_net);
 	if (nlmsvc_users) {
 		if (--nlmsvc_users)
 			goto out;
-- 
1.7.9.5

Re: [PATCH 4/3] lockd: hard-code init_net in lockd case as well

[Greg KH]

On Wed, Jun 20, 2012 at 09:27:38AM -0400, J. Bruce Fields wrote:
> From: "J. Bruce Fields" <bfields@redhat.com>
> 
> The previous patch missed lockd_down, which has the same problem.
> Symptoms were crashes on unmount or reboot.
> 
> Note this was fixed in a different way upstream.
> 
> Acked-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
> Signed-off-by: J. Bruce Fields <bfields@redhat.com>
> ---
>  fs/lockd/svc.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

<formletter>

This is not the correct way to submit patches for inclusion in the
stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
for how to do this properly.

</formletter>

Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[Greg KH]

On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> 
> commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.

Um, no, that's not what this commit id is at all.

Where did you get that from?  I kind of don't trust this series now,
care to fix these up with the correct git commit ids and resend them?

thanks,

greg k-h

Re: 3.4.z nfsd backports

[Greg KH]

On Mon, Jun 18, 2012 at 04:07:26PM -0400, J. Bruce Fields wrote:
> These three patches are all upstream, but required some conflict
> resolution to backport, so I'm sending backported versions.

Please resend all of these, with the correct git commit ids.

Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[J. Bruce Fields]

On Wed, Jun 20, 2012 at 09:34:27AM -0700, Greg KH wrote:
> On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> > From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> > 
> > commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.
> 
> Um, no, that's not what this commit id is at all.

Whoops, sorry!  Should have been 9793f7c88.  Upstream commits on #2 and
#3 are correct.

> Where did you get that from?  I kind of don't trust this series now,
> care to fix these up with the correct git commit ids and resend them?

Sure, I'll resend once I get the "4/3" patch sorted out.

(That got a form rejection--I assume it was the missing upstream commit
ID?  That's because there isn't one--3.4 requires a different fix.  But
I can point to where the bug got fixed upstream even if it isn't the
same.)

--b.

Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[Greg KH]

On Wed, Jun 20, 2012 at 02:09:29PM -0400, J. Bruce Fields wrote:
> On Wed, Jun 20, 2012 at 09:34:27AM -0700, Greg KH wrote:
> > On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> > > From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> > > 
> > > commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.
> > 
> > Um, no, that's not what this commit id is at all.
> 
> Whoops, sorry!  Should have been 9793f7c88.  Upstream commits on #2 and
> #3 are correct.
> 
> > Where did you get that from?  I kind of don't trust this series now,
> > care to fix these up with the correct git commit ids and resend them?
> 
> Sure, I'll resend once I get the "4/3" patch sorted out.
> 
> (That got a form rejection--I assume it was the missing upstream commit
> ID?  That's because there isn't one--3.4 requires a different fix.  But
> I can point to where the bug got fixed upstream even if it isn't the
> same.)

You need to be really specific when sending a patch that is not
upstream, detailing why it isn't, and what different fix went in
upstream, if possible.

thanks,

greg k-h

Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[J. Bruce Fields]

On Wed, Jun 20, 2012 at 11:18:35AM -0700, Greg KH wrote:
> On Wed, Jun 20, 2012 at 02:09:29PM -0400, J. Bruce Fields wrote:
> > On Wed, Jun 20, 2012 at 09:34:27AM -0700, Greg KH wrote:
> > > On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> > > > From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> > > > 
> > > > commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.
> > > 
> > > Um, no, that's not what this commit id is at all.
> > 
> > Whoops, sorry!  Should have been 9793f7c88.  Upstream commits on #2 and
> > #3 are correct.
> > 
> > > Where did you get that from?  I kind of don't trust this series now,
> > > care to fix these up with the correct git commit ids and resend them?
> > 
> > Sure, I'll resend once I get the "4/3" patch sorted out.
> > 
> > (That got a form rejection--I assume it was the missing upstream commit
> > ID?  That's because there isn't one--3.4 requires a different fix.  But
> > I can point to where the bug got fixed upstream even if it isn't the
> > same.)
> 
> You need to be really specific when sending a patch that is not
> upstream, detailing why it isn't, and what different fix went in
> upstream, if possible.

OK, looking at it closer I notice we can actually just take a different
fix directly from upstream, and then there's no problem.  But it took me
a few days to get around to setting up a test-case to make sure that
still did the job....

I'll pass along the revised series in a moment.--b.

Re: [PATCH 1/3] SUNRPC: new svc_bind() routine introduced

[J. Bruce Fields]

On Mon, Jun 25, 2012 at 04:33:32PM -0400, bfields wrote:
> On Wed, Jun 20, 2012 at 11:18:35AM -0700, Greg KH wrote:
> > On Wed, Jun 20, 2012 at 02:09:29PM -0400, J. Bruce Fields wrote:
> > > On Wed, Jun 20, 2012 at 09:34:27AM -0700, Greg KH wrote:
> > > > On Mon, Jun 18, 2012 at 04:07:27PM -0400, J. Bruce Fields wrote:
> > > > > From: Stanislav Kinsbursky <skinsbursky@parallels.com>
> > > > > 
> > > > > commit cd96891d48a945ca2011fbeceda73813d6286195 upstream.
> > > > 
> > > > Um, no, that's not what this commit id is at all.
> > > 
> > > Whoops, sorry!  Should have been 9793f7c88.  Upstream commits on #2 and
> > > #3 are correct.
> > > 
> > > > Where did you get that from?  I kind of don't trust this series now,
> > > > care to fix these up with the correct git commit ids and resend them?
> > > 
> > > Sure, I'll resend once I get the "4/3" patch sorted out.
> > > 
> > > (That got a form rejection--I assume it was the missing upstream commit
> > > ID?  That's because there isn't one--3.4 requires a different fix.  But
> > > I can point to where the bug got fixed upstream even if it isn't the
> > > same.)
> > 
> > You need to be really specific when sending a patch that is not
> > upstream, detailing why it isn't, and what different fix went in
> > upstream, if possible.
> 
> OK, looking at it closer I notice we can actually just take a different
> fix directly from upstream, and then there's no problem.  But it took me
> a few days to get around to setting up a test-case to make sure that
> still did the job....

(Oh, and my reproducer in case anyone wonders: build the following, run

	./namespace-exec /bin/bash

and

	mount -tproc proc /proc
	ps ax

just to verify you're in a new pid/mount namespace, then

	mount -tnfs yourserver:/somepath /mnt

then ^D out of the shell, and it should crash.)

--b.


#define _GNU_SOURCE
#include <unistd.h>
#include <signal.h>
#include <sched.h>
#include <sys/syscall.h>
#include <err.h>

int main(int argc, char *argv[])
{
	int pid;
	int flags = SIGCHLD;

	if (argc < 2)
		errx(1, "usage: %s command arg1 arg2 ...", argv[0]);
	flags |= CLONE_NEWNS|CLONE_NEWPID;
	pid = syscall(__NR_clone, flags, NULL);
	if (pid < 0)
		err(1, "clone failed");
	if (pid == 0) {
		/* child */
		int ret;

		ret = execv(argv[1], argv+1);
		if (ret)
			err(1,"failed to exec %s", argv[1]);
	}
	wait();
}