All of lore.kernel.org
 help / color / mirror / Atom feed
* remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments

^ permalink raw reply	[flat|nested] 165+ messages in thread

* remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments

^ permalink raw reply	[flat|nested] 165+ messages in thread

* remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments

^ permalink raw reply	[flat|nested] 165+ messages in thread

* remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments

^ permalink raw reply	[flat|nested] 165+ messages in thread

* [Cluster-devel] remove most callers of kernel_setsockopt v3
@ 2020-05-28  5:12 ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Hi Dave,

this series removes most callers of the kernel_setsockopt functions, and
instead switches their users to small functions that implement setting a
sockopt directly using a normal kernel function call with type safety and
all the other benefits of not having a function call.

In some cases these functions seem pretty heavy handed as they do
a lock_sock even for just setting a single variable, but this mirrors
the real setsockopt implementation unlike a few drivers that just set
set the fields directly.


Changes since v2:
 - drop the separately merged kernel_getopt_removal
 - drop the sctp patches, as there is conflicting cleanup going on
 - add an additional ACK for the rxrpc changes

Changes since v1:
 - use ->getname for sctp sockets in dlm
 - add a new ->bind_add struct proto method for dlm/sctp
 - switch the ipv6 and remaining sctp helpers to inline function so that
   the ipv6 and sctp modules are not pulled in by any module that could
   potentially use ipv6 or sctp connections
 - remove arguments to various sock_* helpers that are always used with
   the same constant arguments



^ permalink raw reply	[flat|nested] 165+ messages in thread

* [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family == AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family == AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family = AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family = AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Acked-by: Sagi Grimberg <sagi-NQWnxTmZq1alnMjI0IkVqw@public.gmane.org>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family == AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family == AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, linux-kernel, linux-nvme, Jon Maloy,
	linux-rdma, Eric Dumazet, target-devel, tipc-discussion,
	linux-nfs, Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs,
	ocfs2-devel, drbd-dev

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family == AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family == AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family == AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family == AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 01/28] net: add sock_set_reuseaddr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_REUSEADDR sockopt from kernel space
without going through a fake uaccess.

For this the iscsi target now has to formally depend on inet to avoid
a mostly theoretical compile failure.  For actual operation it already
did depend on having ipv4 or ipv6 support.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/infiniband/sw/siw/siw_cm.c        | 18 +++++-------------
 drivers/nvme/target/tcp.c                 |  8 +-------
 drivers/target/iscsi/Kconfig              |  2 +-
 drivers/target/iscsi/iscsi_target_login.c |  9 +--------
 fs/dlm/lowcomms.c                         |  6 +-----
 include/net/sock.h                        |  2 ++
 net/core/sock.c                           |  8 ++++++++
 7 files changed, 19 insertions(+), 34 deletions(-)

diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index 559e5fd3bad8b..d1860f3e87401 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -1312,17 +1312,14 @@ static void siw_cm_llp_state_change(struct sock *sk)
 static int kernel_bindconnect(struct socket *s, struct sockaddr *laddr,
 			      struct sockaddr *raddr)
 {
-	int rv, flags = 0, s_val = 1;
+	int rv, flags = 0;
 	size_t size = laddr->sa_family == AF_INET ?
 		sizeof(struct sockaddr_in) : sizeof(struct sockaddr_in6);
 
 	/*
 	 * Make address available again asap.
 	 */
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv < 0)
-		return rv;
+	sock_set_reuseaddr(s->sk);
 
 	rv = s->ops->bind(s, laddr, size);
 	if (rv < 0)
@@ -1781,7 +1778,7 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	struct siw_cep *cep = NULL;
 	struct siw_device *sdev = to_siw_dev(id->device);
 	int addr_family = id->local_addr.ss_family;
-	int rv = 0, s_val;
+	int rv = 0;
 
 	if (addr_family != AF_INET && addr_family != AF_INET6)
 		return -EAFNOSUPPORT;
@@ -1793,13 +1790,8 @@ int siw_create_listen(struct iw_cm_id *id, int backlog)
 	/*
 	 * Allow binding local port when still in TIME_WAIT from last close.
 	 */
-	s_val = 1;
-	rv = kernel_setsockopt(s, SOL_SOCKET, SO_REUSEADDR, (char *)&s_val,
-			       sizeof(s_val));
-	if (rv) {
-		siw_dbg(id->device, "setsockopt error: %d\n", rv);
-		goto error;
-	}
+	sock_set_reuseaddr(s->sk);
+
 	if (addr_family == AF_INET) {
 		struct sockaddr_in *laddr = &to_sockaddr_in(id->local_addr);
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f40..40757a63f4553 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1632,6 +1632,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->sock->sk->sk_user_data = port;
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
+	sock_set_reuseaddr(port->sock->sk);
 
 	opt = 1;
 	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
@@ -1641,13 +1642,6 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set SO_REUSEADDR sock opt %d\n", ret);
-		goto err_sock;
-	}
-
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
 				(char *)&so_priority, sizeof(so_priority));
diff --git a/drivers/target/iscsi/Kconfig b/drivers/target/iscsi/Kconfig
index 1f93ea3813536..922484ea4e304 100644
--- a/drivers/target/iscsi/Kconfig
+++ b/drivers/target/iscsi/Kconfig
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config ISCSI_TARGET
 	tristate "Linux-iSCSI.org iSCSI Target Mode Stack"
-	depends on NET
+	depends on INET
 	select CRYPTO
 	select CRYPTO_CRC32C
 	select CRYPTO_CRC32C_INTEL if X86
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 731ee67fe914b..91acb3f07b4cc 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -909,14 +909,7 @@ int iscsit_setup_np(
 		}
 	}
 
-	/* FIXME: Someone please explain why this is endian-safe */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for SO_REUSEADDR"
-			" failed\n");
-		goto fail;
-	}
+	sock_set_reuseaddr(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index cdfaf4f0e11a0..b801e77e3e596 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1244,12 +1244,8 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
 			  sizeof(one));
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
-				   (char *)&one, sizeof(one));
+	sock_set_reuseaddr(sock->sk);
 
-	if (result < 0) {
-		log_print("Failed to set SO_REUSEADDR on socket: %d", result);
-	}
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = con;
 	save_listen_callbacks(sock);
diff --git a/include/net/sock.h b/include/net/sock.h
index 3e8c6d4b4b59f..2ec085044790c 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,4 +2688,6 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_set_reuseaddr(struct sock *sk);
+
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index fd85e651ce284..18eb84fdf5fbe 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -712,6 +712,14 @@ bool sk_mc_loop(struct sock *sk)
 }
 EXPORT_SYMBOL(sk_mc_loop);
 
+void sock_set_reuseaddr(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuse = SK_CAN_REUSE;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseaddr);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Acked-by: Sagi Grimberg <sagi-NQWnxTmZq1alnMjI0IkVqw@public.gmane.org>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, linux-kernel, linux-nvme, Jon Maloy,
	linux-rdma, Eric Dumazet, target-devel, tipc-discussion,
	linux-nfs, Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs,
	ocfs2-devel, drbd-dev

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 02/28] net: add sock_no_linger
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_LINGER sockopt from kernel space
with onoff set to true and a linger time of 0 without going through a
fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   |  9 +--------
 drivers/nvme/target/tcp.c |  6 +-----
 include/net/sock.h        |  1 +
 net/core/sock.c           |  9 +++++++++
 net/rds/tcp.h             |  1 -
 net/rds/tcp_connect.c     |  2 +-
 net/rds/tcp_listen.c      | 13 +------------
 net/sunrpc/svcsock.c      | 12 ++----------
 8 files changed, 16 insertions(+), 37 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index c15a92163c1f7..e72d87482eb78 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,6 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret, opt, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
@@ -1361,13 +1360,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set SO_LINGER sock opt %d\n", ret);
-		goto err_sock;
-	}
+	sock_no_linger(queue->sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 40757a63f4553..e0801494b097f 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1429,7 +1429,6 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 {
 	struct socket *sock = queue->sock;
 	struct inet_sock *inet = inet_sk(sock->sk);
-	struct linger sol = { .l_onoff = 1, .l_linger = 0 };
 	int ret;
 
 	ret = kernel_getsockname(sock,
@@ -1447,10 +1446,7 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 * close. This is done to prevent stale data from being sent should
 	 * the network connection be restored before TCP times out.
 	 */
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			(char *)&sol, sizeof(sol));
-	if (ret)
-		return ret;
+	sock_no_linger(sock->sk);
 
 	if (so_priority > 0) {
 		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
diff --git a/include/net/sock.h b/include/net/sock.h
index 2ec085044790c..6ed00bf009bbe 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+void sock_no_linger(struct sock *sk);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 18eb84fdf5fbe..f0f09524911c8 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -720,6 +720,15 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_no_linger(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_lingertime = 0;
+	sock_set_flag(sk, SOCK_LINGER);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_no_linger);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index 3c69361d21c73..d640e210b97b6 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -73,7 +73,6 @@ void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
 int rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
-void rds_tcp_set_linger(struct socket *sock);
 
 /* tcp_recv.c */
 int rds_tcp_recv_init(void);
diff --git a/net/rds/tcp_connect.c b/net/rds/tcp_connect.c
index 008f50fb25dd2..4e64598176b05 100644
--- a/net/rds/tcp_connect.c
+++ b/net/rds/tcp_connect.c
@@ -207,7 +207,7 @@ void rds_tcp_conn_path_shutdown(struct rds_conn_path *cp)
 
 	if (sock) {
 		if (rds_destroy_pending(cp->cp_conn))
-			rds_tcp_set_linger(sock);
+			sock_no_linger(sock->sk);
 		sock->ops->shutdown(sock, RCV_SHUTDOWN | SEND_SHUTDOWN);
 		lock_sock(sock->sk);
 		rds_tcp_restore_callbacks(sock, tc); /* tc->tc_sock = NULL */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 810a3a49e9474..bbb31b9c0b391 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -111,17 +111,6 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	return NULL;
 }
 
-void rds_tcp_set_linger(struct socket *sock)
-{
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
-}
-
 int rds_tcp_accept_one(struct socket *sock)
 {
 	struct socket *new_sock = NULL;
@@ -241,7 +230,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	 * be pending on it. By setting linger, we achieve the side-effect
 	 * of avoiding TIME_WAIT state on new_sock.
 	 */
-	rds_tcp_set_linger(new_sock);
+	sock_no_linger(new_sock->sk);
 	kernel_sock_shutdown(new_sock, SHUT_RDWR);
 	ret = 0;
 out:
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 023514e392b31..6773dacc64d8e 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -323,17 +323,9 @@ static int svc_tcp_has_wspace(struct svc_xprt *xprt)
 
 static void svc_tcp_kill_temp_xprt(struct svc_xprt *xprt)
 {
-	struct svc_sock *svsk;
-	struct socket *sock;
-	struct linger no_linger = {
-		.l_onoff = 1,
-		.l_linger = 0,
-	};
+	struct svc_sock *svsk = container_of(xprt, struct svc_sock, sk_xprt);
 
-	svsk = container_of(xprt, struct svc_sock, sk_xprt);
-	sock = svsk->sk_sock;
-	kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
-			  (char *)&no_linger, sizeof(no_linger));
+	sock_no_linger(svsk->sk_sock->sk);
 }
 
 /*
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 03/28] net: add sock_set_priority
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the SO_PRIORITY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 12 ++----------
 drivers/nvme/target/tcp.c | 18 ++++--------------
 include/net/sock.h        |  1 +
 net/core/sock.c           |  8 ++++++++
 4 files changed, 15 insertions(+), 24 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index e72d87482eb78..a307972d33a02 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1362,16 +1362,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 */
 	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (nctrl->opts->tos >= 0) {
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index e0801494b097f..f3088156d01da 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1448,12 +1448,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 */
 	sock_no_linger(sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret)
-			return ret;
-	}
+	if (so_priority > 0)
+		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (inet->rcv_tos > 0) {
@@ -1638,14 +1634,8 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			pr_err("failed to set SO_PRIORITY sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(port->sock->sk, so_priority);
 
 	ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr,
 			sizeof(port->addr));
diff --git a/include/net/sock.h b/include/net/sock.h
index 6ed00bf009bbe..a3a43141a4be2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 void sock_no_linger(struct sock *sk);
+void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index f0f09524911c8..ceda1a9248b3e 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_no_linger(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_no_linger);
 
+void sock_set_priority(struct sock *sk, u32 priority)
+{
+	lock_sock(sk);
+	sk->sk_priority = priority;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_priority);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 03/28] net: add sock_set_priority
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the SO_PRIORITY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 12 ++----------
 drivers/nvme/target/tcp.c | 18 ++++--------------
 include/net/sock.h        |  1 +
 net/core/sock.c           |  8 ++++++++
 4 files changed, 15 insertions(+), 24 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index e72d87482eb78..a307972d33a02 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1362,16 +1362,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 */
 	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (nctrl->opts->tos >= 0) {
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index e0801494b097f..f3088156d01da 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1448,12 +1448,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 */
 	sock_no_linger(sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret)
-			return ret;
-	}
+	if (so_priority > 0)
+		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (inet->rcv_tos > 0) {
@@ -1638,14 +1634,8 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			pr_err("failed to set SO_PRIORITY sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(port->sock->sk, so_priority);
 
 	ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr,
 			sizeof(port->addr));
diff --git a/include/net/sock.h b/include/net/sock.h
index 6ed00bf009bbe..a3a43141a4be2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 void sock_no_linger(struct sock *sk);
+void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index f0f09524911c8..ceda1a9248b3e 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_no_linger(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_no_linger);
 
+void sock_set_priority(struct sock *sk, u32 priority)
+{
+	lock_sock(sk);
+	sk->sk_priority = priority;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_priority);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 03/28] net: add sock_set_priority
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, linux-kernel, linux-nvme, Jon Maloy,
	linux-rdma, Eric Dumazet, target-devel, tipc-discussion,
	linux-nfs, Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs,
	ocfs2-devel, drbd-dev

Add a helper to directly set the SO_PRIORITY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 12 ++----------
 drivers/nvme/target/tcp.c | 18 ++++--------------
 include/net/sock.h        |  1 +
 net/core/sock.c           |  8 ++++++++
 4 files changed, 15 insertions(+), 24 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index e72d87482eb78..a307972d33a02 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1362,16 +1362,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 */
 	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (nctrl->opts->tos >= 0) {
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index e0801494b097f..f3088156d01da 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1448,12 +1448,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 */
 	sock_no_linger(sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret)
-			return ret;
-	}
+	if (so_priority > 0)
+		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (inet->rcv_tos > 0) {
@@ -1638,14 +1634,8 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			pr_err("failed to set SO_PRIORITY sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(port->sock->sk, so_priority);
 
 	ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr,
 			sizeof(port->addr));
diff --git a/include/net/sock.h b/include/net/sock.h
index 6ed00bf009bbe..a3a43141a4be2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 void sock_no_linger(struct sock *sk);
+void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index f0f09524911c8..ceda1a9248b3e 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_no_linger(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_no_linger);
 
+void sock_set_priority(struct sock *sk, u32 priority)
+{
+	lock_sock(sk);
+	sk->sk_priority = priority;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_priority);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 03/28] net: add sock_set_priority
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the SO_PRIORITY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 12 ++----------
 drivers/nvme/target/tcp.c | 18 ++++--------------
 include/net/sock.h        |  1 +
 net/core/sock.c           |  8 ++++++++
 4 files changed, 15 insertions(+), 24 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index e72d87482eb78..a307972d33a02 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1362,16 +1362,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 */
 	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (nctrl->opts->tos >= 0) {
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index e0801494b097f..f3088156d01da 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1448,12 +1448,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 */
 	sock_no_linger(sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret)
-			return ret;
-	}
+	if (so_priority > 0)
+		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (inet->rcv_tos > 0) {
@@ -1638,14 +1634,8 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			pr_err("failed to set SO_PRIORITY sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(port->sock->sk, so_priority);
 
 	ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr,
 			sizeof(port->addr));
diff --git a/include/net/sock.h b/include/net/sock.h
index 6ed00bf009bbe..a3a43141a4be2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 void sock_no_linger(struct sock *sk);
+void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index f0f09524911c8..ceda1a9248b3e 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_no_linger(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_no_linger);
 
+void sock_set_priority(struct sock *sk, u32 priority)
+{
+	lock_sock(sk);
+	sk->sk_priority = priority;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_priority);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 03/28] net: add sock_set_priority
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_PRIORITY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 12 ++----------
 drivers/nvme/target/tcp.c | 18 ++++--------------
 include/net/sock.h        |  1 +
 net/core/sock.c           |  8 ++++++++
 4 files changed, 15 insertions(+), 24 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index e72d87482eb78..a307972d33a02 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1362,16 +1362,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	 */
 	sock_no_linger(queue->sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(queue->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			dev_err(ctrl->ctrl.device,
-				"failed to set SO_PRIORITY sock opt, ret %d\n",
-				ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (nctrl->opts->tos >= 0) {
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index e0801494b097f..f3088156d01da 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1448,12 +1448,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 	 */
 	sock_no_linger(sock->sk);
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret)
-			return ret;
-	}
+	if (so_priority > 0)
+		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
 	if (inet->rcv_tos > 0) {
@@ -1638,14 +1634,8 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 		goto err_sock;
 	}
 
-	if (so_priority > 0) {
-		ret = kernel_setsockopt(port->sock, SOL_SOCKET, SO_PRIORITY,
-				(char *)&so_priority, sizeof(so_priority));
-		if (ret) {
-			pr_err("failed to set SO_PRIORITY sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (so_priority > 0)
+		sock_set_priority(port->sock->sk, so_priority);
 
 	ret = kernel_bind(port->sock, (struct sockaddr *)&port->addr,
 			sizeof(port->addr));
diff --git a/include/net/sock.h b/include/net/sock.h
index 6ed00bf009bbe..a3a43141a4be2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 void sock_no_linger(struct sock *sk);
+void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index f0f09524911c8..ceda1a9248b3e 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_no_linger(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_no_linger);
 
+void sock_set_priority(struct sock *sk, u32 priority)
+{
+	lock_sock(sk);
+	sk->sk_priority = priority;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_priority);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 04/28] net: add sock_set_sndtimeo
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_SNDTIMEO_NEW sockopt from kernel
space without going through a fake uaccess.  The interface is
simplified to only pass the seconds value, as that is the only
thing needed at the moment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  8 ++------
 include/net/sock.h |  1 +
 net/core/sock.c    | 11 +++++++++++
 3 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b801e77e3e596..b4d491122814b 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1035,7 +1035,6 @@ static void sctp_connect_to_sock(struct connection *con)
 	int result;
 	int addr_len;
 	struct socket *sock;
-	struct __kernel_sock_timeval tv = { .tv_sec = 5, .tv_usec = 0 };
 
 	if (con->nodeid == 0) {
 		log_print("attempt to connect sock 0 foiled");
@@ -1087,13 +1086,10 @@ static void sctp_connect_to_sock(struct connection *con)
 	 * since O_NONBLOCK argument in connect() function does not work here,
 	 * then, we should restore the default value of this attribute.
 	 */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 5);
 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
 				   0);
-	memset(&tv, 0, sizeof(tv));
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 0);
 
 	if (result == -EINPROGRESS)
 		result = 0;
diff --git a/include/net/sock.h b/include/net/sock.h
index a3a43141a4be2..9a7b9e98685ac 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,5 +2691,6 @@ void sock_def_readable(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index ceda1a9248b3e..d3b1d61e4f768 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -737,6 +737,17 @@ void sock_set_priority(struct sock *sk, u32 priority)
 }
 EXPORT_SYMBOL(sock_set_priority);
 
+void sock_set_sndtimeo(struct sock *sk, s64 secs)
+{
+	lock_sock(sk);
+	if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
+		sk->sk_sndtimeo = secs * HZ;
+	else
+		sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_sndtimeo);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 04/28] net: add sock_set_sndtimeo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_SNDTIMEO_NEW sockopt from kernel
space without going through a fake uaccess.  The interface is
simplified to only pass the seconds value, as that is the only
thing needed at the moment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  8 ++------
 include/net/sock.h |  1 +
 net/core/sock.c    | 11 +++++++++++
 3 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b801e77e3e596..b4d491122814b 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1035,7 +1035,6 @@ static void sctp_connect_to_sock(struct connection *con)
 	int result;
 	int addr_len;
 	struct socket *sock;
-	struct __kernel_sock_timeval tv = { .tv_sec = 5, .tv_usec = 0 };
 
 	if (con->nodeid = 0) {
 		log_print("attempt to connect sock 0 foiled");
@@ -1087,13 +1086,10 @@ static void sctp_connect_to_sock(struct connection *con)
 	 * since O_NONBLOCK argument in connect() function does not work here,
 	 * then, we should restore the default value of this attribute.
 	 */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 5);
 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
 				   0);
-	memset(&tv, 0, sizeof(tv));
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 0);
 
 	if (result = -EINPROGRESS)
 		result = 0;
diff --git a/include/net/sock.h b/include/net/sock.h
index a3a43141a4be2..9a7b9e98685ac 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,5 +2691,6 @@ void sock_def_readable(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index ceda1a9248b3e..d3b1d61e4f768 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -737,6 +737,17 @@ void sock_set_priority(struct sock *sk, u32 priority)
 }
 EXPORT_SYMBOL(sock_set_priority);
 
+void sock_set_sndtimeo(struct sock *sk, s64 secs)
+{
+	lock_sock(sk);
+	if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
+		sk->sk_sndtimeo = secs * HZ;
+	else
+		sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_sndtimeo);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 04/28] net: add sock_set_sndtimeo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the SO_SNDTIMEO_NEW sockopt from kernel
space without going through a fake uaccess.  The interface is
simplified to only pass the seconds value, as that is the only
thing needed at the moment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  8 ++------
 include/net/sock.h |  1 +
 net/core/sock.c    | 11 +++++++++++
 3 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b801e77e3e596..b4d491122814b 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1035,7 +1035,6 @@ static void sctp_connect_to_sock(struct connection *con)
 	int result;
 	int addr_len;
 	struct socket *sock;
-	struct __kernel_sock_timeval tv = { .tv_sec = 5, .tv_usec = 0 };
 
 	if (con->nodeid == 0) {
 		log_print("attempt to connect sock 0 foiled");
@@ -1087,13 +1086,10 @@ static void sctp_connect_to_sock(struct connection *con)
 	 * since O_NONBLOCK argument in connect() function does not work here,
 	 * then, we should restore the default value of this attribute.
 	 */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 5);
 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
 				   0);
-	memset(&tv, 0, sizeof(tv));
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 0);
 
 	if (result == -EINPROGRESS)
 		result = 0;
diff --git a/include/net/sock.h b/include/net/sock.h
index a3a43141a4be2..9a7b9e98685ac 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,5 +2691,6 @@ void sock_def_readable(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index ceda1a9248b3e..d3b1d61e4f768 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -737,6 +737,17 @@ void sock_set_priority(struct sock *sk, u32 priority)
 }
 EXPORT_SYMBOL(sock_set_priority);
 
+void sock_set_sndtimeo(struct sock *sk, s64 secs)
+{
+	lock_sock(sk);
+	if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
+		sk->sk_sndtimeo = secs * HZ;
+	else
+		sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_sndtimeo);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 04/28] net: add sock_set_sndtimeo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_SNDTIMEO_NEW sockopt from kernel
space without going through a fake uaccess.  The interface is
simplified to only pass the seconds value, as that is the only
thing needed at the moment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  8 ++------
 include/net/sock.h |  1 +
 net/core/sock.c    | 11 +++++++++++
 3 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b801e77e3e596..b4d491122814b 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1035,7 +1035,6 @@ static void sctp_connect_to_sock(struct connection *con)
 	int result;
 	int addr_len;
 	struct socket *sock;
-	struct __kernel_sock_timeval tv = { .tv_sec = 5, .tv_usec = 0 };
 
 	if (con->nodeid == 0) {
 		log_print("attempt to connect sock 0 foiled");
@@ -1087,13 +1086,10 @@ static void sctp_connect_to_sock(struct connection *con)
 	 * since O_NONBLOCK argument in connect() function does not work here,
 	 * then, we should restore the default value of this attribute.
 	 */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 5);
 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
 				   0);
-	memset(&tv, 0, sizeof(tv));
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 0);
 
 	if (result == -EINPROGRESS)
 		result = 0;
diff --git a/include/net/sock.h b/include/net/sock.h
index a3a43141a4be2..9a7b9e98685ac 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,5 +2691,6 @@ void sock_def_readable(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index ceda1a9248b3e..d3b1d61e4f768 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -737,6 +737,17 @@ void sock_set_priority(struct sock *sk, u32 priority)
 }
 EXPORT_SYMBOL(sock_set_priority);
 
+void sock_set_sndtimeo(struct sock *sk, s64 secs)
+{
+	lock_sock(sk);
+	if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
+		sk->sk_sndtimeo = secs * HZ;
+	else
+		sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_sndtimeo);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 04/28] net: add sock_set_sndtimeo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_SNDTIMEO_NEW sockopt from kernel
space without going through a fake uaccess.  The interface is
simplified to only pass the seconds value, as that is the only
thing needed at the moment.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  8 ++------
 include/net/sock.h |  1 +
 net/core/sock.c    | 11 +++++++++++
 3 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b801e77e3e596..b4d491122814b 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1035,7 +1035,6 @@ static void sctp_connect_to_sock(struct connection *con)
 	int result;
 	int addr_len;
 	struct socket *sock;
-	struct __kernel_sock_timeval tv = { .tv_sec = 5, .tv_usec = 0 };
 
 	if (con->nodeid == 0) {
 		log_print("attempt to connect sock 0 foiled");
@@ -1087,13 +1086,10 @@ static void sctp_connect_to_sock(struct connection *con)
 	 * since O_NONBLOCK argument in connect() function does not work here,
 	 * then, we should restore the default value of this attribute.
 	 */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 5);
 	result = sock->ops->connect(sock, (struct sockaddr *)&daddr, addr_len,
 				   0);
-	memset(&tv, 0, sizeof(tv));
-	kernel_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO_NEW, (char *)&tv,
-			  sizeof(tv));
+	sock_set_sndtimeo(sock->sk, 0);
 
 	if (result == -EINPROGRESS)
 		result = 0;
diff --git a/include/net/sock.h b/include/net/sock.h
index a3a43141a4be2..9a7b9e98685ac 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,5 +2691,6 @@ void sock_def_readable(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index ceda1a9248b3e..d3b1d61e4f768 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -737,6 +737,17 @@ void sock_set_priority(struct sock *sk, u32 priority)
 }
 EXPORT_SYMBOL(sock_set_priority);
 
+void sock_set_sndtimeo(struct sock *sk, s64 secs)
+{
+	lock_sock(sk);
+	if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
+		sk->sk_sndtimeo = secs * HZ;
+	else
+		sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_sndtimeo);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 05/28] net: add sock_bindtoindex
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_BINDTOIFINDEX sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h        |  1 +
 net/core/sock.c           | 21 +++++++++++++++------
 net/ipv4/udp_tunnel.c     |  4 +---
 net/ipv6/ip6_udp_tunnel.c |  4 +---
 4 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 9a7b9e98685ac..cdec7bc055d5b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index d3b1d61e4f768..23f80880fbb2c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -566,7 +566,7 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
 }
 EXPORT_SYMBOL(sk_dst_check);
 
-static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
+static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 {
 	int ret = -ENOPROTOOPT;
 #ifdef CONFIG_NETDEVICES
@@ -594,6 +594,18 @@ static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
 	return ret;
 }
 
+int sock_bindtoindex(struct sock *sk, int ifindex)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = sock_bindtoindex_locked(sk, ifindex);
+	release_sock(sk);
+
+	return ret;
+}
+EXPORT_SYMBOL(sock_bindtoindex);
+
 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 				int optlen)
 {
@@ -634,10 +646,7 @@ static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 			goto out;
 	}
 
-	lock_sock(sk);
-	ret = sock_setbindtodevice_locked(sk, index);
-	release_sock(sk);
-
+	return sock_bindtoindex(sk, index);
 out:
 #endif
 
@@ -1216,7 +1225,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_BINDTOIFINDEX:
-		ret = sock_setbindtodevice_locked(sk, val);
+		ret = sock_bindtoindex_locked(sk, val);
 		break;
 
 	default:
diff --git a/net/ipv4/udp_tunnel.c b/net/ipv4/udp_tunnel.c
index 150e6f0fdbf59..2158e8bddf41c 100644
--- a/net/ipv4/udp_tunnel.c
+++ b/net/ipv4/udp_tunnel.c
@@ -22,9 +22,7 @@ int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 58956a6b66a21..6523609516d25 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -33,9 +33,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 			goto error;
 	}
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 05/28] net: add sock_bindtoindex
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_BINDTOIFINDEX sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h        |  1 +
 net/core/sock.c           | 21 +++++++++++++++------
 net/ipv4/udp_tunnel.c     |  4 +---
 net/ipv6/ip6_udp_tunnel.c |  4 +---
 4 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 9a7b9e98685ac..cdec7bc055d5b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index d3b1d61e4f768..23f80880fbb2c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -566,7 +566,7 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
 }
 EXPORT_SYMBOL(sk_dst_check);
 
-static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
+static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 {
 	int ret = -ENOPROTOOPT;
 #ifdef CONFIG_NETDEVICES
@@ -594,6 +594,18 @@ static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
 	return ret;
 }
 
+int sock_bindtoindex(struct sock *sk, int ifindex)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = sock_bindtoindex_locked(sk, ifindex);
+	release_sock(sk);
+
+	return ret;
+}
+EXPORT_SYMBOL(sock_bindtoindex);
+
 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 				int optlen)
 {
@@ -634,10 +646,7 @@ static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 			goto out;
 	}
 
-	lock_sock(sk);
-	ret = sock_setbindtodevice_locked(sk, index);
-	release_sock(sk);
-
+	return sock_bindtoindex(sk, index);
 out:
 #endif
 
@@ -1216,7 +1225,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_BINDTOIFINDEX:
-		ret = sock_setbindtodevice_locked(sk, val);
+		ret = sock_bindtoindex_locked(sk, val);
 		break;
 
 	default:
diff --git a/net/ipv4/udp_tunnel.c b/net/ipv4/udp_tunnel.c
index 150e6f0fdbf59..2158e8bddf41c 100644
--- a/net/ipv4/udp_tunnel.c
+++ b/net/ipv4/udp_tunnel.c
@@ -22,9 +22,7 @@ int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 58956a6b66a21..6523609516d25 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -33,9 +33,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 			goto error;
 	}
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 05/28] net: add sock_bindtoindex
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the SO_BINDTOIFINDEX sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h        |  1 +
 net/core/sock.c           | 21 +++++++++++++++------
 net/ipv4/udp_tunnel.c     |  4 +---
 net/ipv6/ip6_udp_tunnel.c |  4 +---
 4 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 9a7b9e98685ac..cdec7bc055d5b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index d3b1d61e4f768..23f80880fbb2c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -566,7 +566,7 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
 }
 EXPORT_SYMBOL(sk_dst_check);
 
-static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
+static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 {
 	int ret = -ENOPROTOOPT;
 #ifdef CONFIG_NETDEVICES
@@ -594,6 +594,18 @@ static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
 	return ret;
 }
 
+int sock_bindtoindex(struct sock *sk, int ifindex)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = sock_bindtoindex_locked(sk, ifindex);
+	release_sock(sk);
+
+	return ret;
+}
+EXPORT_SYMBOL(sock_bindtoindex);
+
 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 				int optlen)
 {
@@ -634,10 +646,7 @@ static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 			goto out;
 	}
 
-	lock_sock(sk);
-	ret = sock_setbindtodevice_locked(sk, index);
-	release_sock(sk);
-
+	return sock_bindtoindex(sk, index);
 out:
 #endif
 
@@ -1216,7 +1225,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_BINDTOIFINDEX:
-		ret = sock_setbindtodevice_locked(sk, val);
+		ret = sock_bindtoindex_locked(sk, val);
 		break;
 
 	default:
diff --git a/net/ipv4/udp_tunnel.c b/net/ipv4/udp_tunnel.c
index 150e6f0fdbf59..2158e8bddf41c 100644
--- a/net/ipv4/udp_tunnel.c
+++ b/net/ipv4/udp_tunnel.c
@@ -22,9 +22,7 @@ int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 58956a6b66a21..6523609516d25 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -33,9 +33,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 			goto error;
 	}
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 05/28] net: add sock_bindtoindex
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_BINDTOIFINDEX sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h        |  1 +
 net/core/sock.c           | 21 +++++++++++++++------
 net/ipv4/udp_tunnel.c     |  4 +---
 net/ipv6/ip6_udp_tunnel.c |  4 +---
 4 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 9a7b9e98685ac..cdec7bc055d5b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index d3b1d61e4f768..23f80880fbb2c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -566,7 +566,7 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
 }
 EXPORT_SYMBOL(sk_dst_check);
 
-static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
+static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 {
 	int ret = -ENOPROTOOPT;
 #ifdef CONFIG_NETDEVICES
@@ -594,6 +594,18 @@ static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
 	return ret;
 }
 
+int sock_bindtoindex(struct sock *sk, int ifindex)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = sock_bindtoindex_locked(sk, ifindex);
+	release_sock(sk);
+
+	return ret;
+}
+EXPORT_SYMBOL(sock_bindtoindex);
+
 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 				int optlen)
 {
@@ -634,10 +646,7 @@ static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 			goto out;
 	}
 
-	lock_sock(sk);
-	ret = sock_setbindtodevice_locked(sk, index);
-	release_sock(sk);
-
+	return sock_bindtoindex(sk, index);
 out:
 #endif
 
@@ -1216,7 +1225,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_BINDTOIFINDEX:
-		ret = sock_setbindtodevice_locked(sk, val);
+		ret = sock_bindtoindex_locked(sk, val);
 		break;
 
 	default:
diff --git a/net/ipv4/udp_tunnel.c b/net/ipv4/udp_tunnel.c
index 150e6f0fdbf59..2158e8bddf41c 100644
--- a/net/ipv4/udp_tunnel.c
+++ b/net/ipv4/udp_tunnel.c
@@ -22,9 +22,7 @@ int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 58956a6b66a21..6523609516d25 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -33,9 +33,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 			goto error;
 	}
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 05/28] net: add sock_bindtoindex
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_BINDTOIFINDEX sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h        |  1 +
 net/core/sock.c           | 21 +++++++++++++++------
 net/ipv4/udp_tunnel.c     |  4 +---
 net/ipv6/ip6_udp_tunnel.c |  4 +---
 4 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 9a7b9e98685ac..cdec7bc055d5b 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2688,6 +2688,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 
 void sock_def_readable(struct sock *sk);
 
+int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index d3b1d61e4f768..23f80880fbb2c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -566,7 +566,7 @@ struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
 }
 EXPORT_SYMBOL(sk_dst_check);
 
-static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
+static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 {
 	int ret = -ENOPROTOOPT;
 #ifdef CONFIG_NETDEVICES
@@ -594,6 +594,18 @@ static int sock_setbindtodevice_locked(struct sock *sk, int ifindex)
 	return ret;
 }
 
+int sock_bindtoindex(struct sock *sk, int ifindex)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = sock_bindtoindex_locked(sk, ifindex);
+	release_sock(sk);
+
+	return ret;
+}
+EXPORT_SYMBOL(sock_bindtoindex);
+
 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 				int optlen)
 {
@@ -634,10 +646,7 @@ static int sock_setbindtodevice(struct sock *sk, char __user *optval,
 			goto out;
 	}
 
-	lock_sock(sk);
-	ret = sock_setbindtodevice_locked(sk, index);
-	release_sock(sk);
-
+	return sock_bindtoindex(sk, index);
 out:
 #endif
 
@@ -1216,7 +1225,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_BINDTOIFINDEX:
-		ret = sock_setbindtodevice_locked(sk, val);
+		ret = sock_bindtoindex_locked(sk, val);
 		break;
 
 	default:
diff --git a/net/ipv4/udp_tunnel.c b/net/ipv4/udp_tunnel.c
index 150e6f0fdbf59..2158e8bddf41c 100644
--- a/net/ipv4/udp_tunnel.c
+++ b/net/ipv4/udp_tunnel.c
@@ -22,9 +22,7 @@ int udp_sock_create4(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 58956a6b66a21..6523609516d25 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -33,9 +33,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 			goto error;
 	}
 	if (cfg->bind_ifindex) {
-		err = kernel_setsockopt(sock, SOL_SOCKET, SO_BINDTOIFINDEX,
-					(void *)&cfg->bind_ifindex,
-					sizeof(cfg->bind_ifindex));
+		err = sock_bindtoindex(sock->sk, cfg->bind_ifindex);
 		if (err < 0)
 			goto error;
 	}
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname == SO_TIMESTAMP_NEW || optname == SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname == SO_TIMESTAMP_OLD || optname == SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname = SO_TIMESTAMP_NEW || optname = SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname = SO_TIMESTAMP_OLD || optname = SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname == SO_TIMESTAMP_NEW || optname == SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname == SO_TIMESTAMP_OLD || optname == SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname == SO_TIMESTAMP_NEW || optname == SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname == SO_TIMESTAMP_OLD || optname == SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname == SO_TIMESTAMP_NEW || optname == SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname == SO_TIMESTAMP_OLD || optname == SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 06/28] net: add sock_enable_timestamps
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly enable timestamps instead of setting the
SO_TIMESTAMP* sockopts from kernel space and going through a fake
uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h       |  1 +
 net/core/sock.c          | 47 +++++++++++++++++++++++++---------------
 net/rxrpc/local_object.c |  8 +------
 3 files changed, 31 insertions(+), 25 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index cdec7bc055d5b..99ef43508d2b5 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2689,6 +2689,7 @@ static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
 void sock_def_readable(struct sock *sk);
 
 int sock_bindtoindex(struct sock *sk, int ifindex);
+void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
diff --git a/net/core/sock.c b/net/core/sock.c
index 23f80880fbb2c..e4a4dd2b3d8b3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -757,6 +757,28 @@ void sock_set_sndtimeo(struct sock *sk, s64 secs)
 }
 EXPORT_SYMBOL(sock_set_sndtimeo);
 
+static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
+{
+	if (val)  {
+		sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
+		sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
+		sock_set_flag(sk, SOCK_RCVTSTAMP);
+		sock_enable_timestamp(sk, SOCK_TIMESTAMP);
+	} else {
+		sock_reset_flag(sk, SOCK_RCVTSTAMP);
+		sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
+		sock_reset_flag(sk, SOCK_TSTAMP_NEW);
+	}
+}
+
+void sock_enable_timestamps(struct sock *sk)
+{
+	lock_sock(sk);
+	__sock_set_timestamps(sk, true, false, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_enable_timestamps);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -948,28 +970,17 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		break;
 
 	case SO_TIMESTAMP_OLD:
+		__sock_set_timestamps(sk, valbool, false, false);
+		break;
 	case SO_TIMESTAMP_NEW:
+		__sock_set_timestamps(sk, valbool, true, false);
+		break;
 	case SO_TIMESTAMPNS_OLD:
+		__sock_set_timestamps(sk, valbool, false, true);
+		break;
 	case SO_TIMESTAMPNS_NEW:
-		if (valbool)  {
-			if (optname == SO_TIMESTAMP_NEW || optname == SO_TIMESTAMPNS_NEW)
-				sock_set_flag(sk, SOCK_TSTAMP_NEW);
-			else
-				sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-
-			if (optname == SO_TIMESTAMP_OLD || optname == SO_TIMESTAMP_NEW)
-				sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			else
-				sock_set_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_set_flag(sk, SOCK_RCVTSTAMP);
-			sock_enable_timestamp(sk, SOCK_TIMESTAMP);
-		} else {
-			sock_reset_flag(sk, SOCK_RCVTSTAMP);
-			sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
-			sock_reset_flag(sk, SOCK_TSTAMP_NEW);
-		}
+		__sock_set_timestamps(sk, valbool, true, true);
 		break;
-
 	case SO_TIMESTAMPING_NEW:
 		sock_set_flag(sk, SOCK_TSTAMP_NEW);
 		/* fall through */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 01135e54d95d2..5ea2bd01fdd59 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -189,13 +189,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		}
 
 		/* We want receive timestamps. */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
-					(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		sock_enable_timestamps(local->socket->sk);
 		break;
 
 	default:
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 07/28] net: add sock_set_keepalive
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_KEEPALIVE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c     |  6 +-----
 include/net/sock.h    |  1 +
 net/core/sock.c       | 10 ++++++++++
 net/rds/tcp_listen.c  |  6 +-----
 net/sunrpc/xprtsock.c |  4 +---
 5 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b4d491122814b..138009c6a2ee1 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1259,11 +1259,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 		con->sock = NULL;
 		goto create_out;
 	}
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				 (char *)&one, sizeof(one));
-	if (result < 0) {
-		log_print("Set keepalive failed: %d", result);
-	}
+	sock_set_keepalive(sock->sk);
 
 	result = sock->ops->listen(sock, 5);
 	if (result < 0) {
diff --git a/include/net/sock.h b/include/net/sock.h
index 99ef43508d2b5..dc08c176238fd 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,6 +2691,7 @@ void sock_def_readable(struct sock *sk);
 int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
+void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
diff --git a/net/core/sock.c b/net/core/sock.c
index e4a4dd2b3d8b3..728f5fb156a0c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -779,6 +779,16 @@ void sock_enable_timestamps(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_enable_timestamps);
 
+void sock_set_keepalive(struct sock *sk)
+{
+	lock_sock(sk);
+	if (sk->sk_prot->keepalive)
+		sk->sk_prot->keepalive(sk, true);
+	sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_keepalive);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index bbb31b9c0b391..d8bd132769594 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -43,13 +43,9 @@ int rds_tcp_keepalive(struct socket *sock)
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int keepalive = 1;
 	int ret = 0;
 
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				(char *)&keepalive, sizeof(keepalive));
-	if (ret < 0)
-		goto bail;
+	sock_set_keepalive(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
 				(char *)&keepcnt, sizeof(keepcnt));
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 845d0be805ece..30082cd039960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2110,7 +2110,6 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 	unsigned int keepidle;
 	unsigned int keepcnt;
-	unsigned int opt_on = 1;
 	unsigned int timeo;
 
 	spin_lock(&xprt->transport_lock);
@@ -2122,8 +2121,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	spin_unlock(&xprt->transport_lock);
 
 	/* TCP Keepalive options */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-			(char *)&opt_on, sizeof(opt_on));
+	sock_set_keepalive(sock->sk);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 07/28] net: add sock_set_keepalive
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_KEEPALIVE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c     |  6 +-----
 include/net/sock.h    |  1 +
 net/core/sock.c       | 10 ++++++++++
 net/rds/tcp_listen.c  |  6 +-----
 net/sunrpc/xprtsock.c |  4 +---
 5 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b4d491122814b..138009c6a2ee1 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1259,11 +1259,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 		con->sock = NULL;
 		goto create_out;
 	}
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				 (char *)&one, sizeof(one));
-	if (result < 0) {
-		log_print("Set keepalive failed: %d", result);
-	}
+	sock_set_keepalive(sock->sk);
 
 	result = sock->ops->listen(sock, 5);
 	if (result < 0) {
diff --git a/include/net/sock.h b/include/net/sock.h
index 99ef43508d2b5..dc08c176238fd 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,6 +2691,7 @@ void sock_def_readable(struct sock *sk);
 int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
+void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
diff --git a/net/core/sock.c b/net/core/sock.c
index e4a4dd2b3d8b3..728f5fb156a0c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -779,6 +779,16 @@ void sock_enable_timestamps(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_enable_timestamps);
 
+void sock_set_keepalive(struct sock *sk)
+{
+	lock_sock(sk);
+	if (sk->sk_prot->keepalive)
+		sk->sk_prot->keepalive(sk, true);
+	sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_keepalive);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index bbb31b9c0b391..d8bd132769594 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -43,13 +43,9 @@ int rds_tcp_keepalive(struct socket *sock)
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int keepalive = 1;
 	int ret = 0;
 
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				(char *)&keepalive, sizeof(keepalive));
-	if (ret < 0)
-		goto bail;
+	sock_set_keepalive(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
 				(char *)&keepcnt, sizeof(keepcnt));
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 845d0be805ece..30082cd039960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2110,7 +2110,6 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 	unsigned int keepidle;
 	unsigned int keepcnt;
-	unsigned int opt_on = 1;
 	unsigned int timeo;
 
 	spin_lock(&xprt->transport_lock);
@@ -2122,8 +2121,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	spin_unlock(&xprt->transport_lock);
 
 	/* TCP Keepalive options */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-			(char *)&opt_on, sizeof(opt_on));
+	sock_set_keepalive(sock->sk);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 07/28] net: add sock_set_keepalive
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the SO_KEEPALIVE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c     |  6 +-----
 include/net/sock.h    |  1 +
 net/core/sock.c       | 10 ++++++++++
 net/rds/tcp_listen.c  |  6 +-----
 net/sunrpc/xprtsock.c |  4 +---
 5 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b4d491122814b..138009c6a2ee1 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1259,11 +1259,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 		con->sock = NULL;
 		goto create_out;
 	}
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				 (char *)&one, sizeof(one));
-	if (result < 0) {
-		log_print("Set keepalive failed: %d", result);
-	}
+	sock_set_keepalive(sock->sk);
 
 	result = sock->ops->listen(sock, 5);
 	if (result < 0) {
diff --git a/include/net/sock.h b/include/net/sock.h
index 99ef43508d2b5..dc08c176238fd 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,6 +2691,7 @@ void sock_def_readable(struct sock *sk);
 int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
+void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
diff --git a/net/core/sock.c b/net/core/sock.c
index e4a4dd2b3d8b3..728f5fb156a0c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -779,6 +779,16 @@ void sock_enable_timestamps(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_enable_timestamps);
 
+void sock_set_keepalive(struct sock *sk)
+{
+	lock_sock(sk);
+	if (sk->sk_prot->keepalive)
+		sk->sk_prot->keepalive(sk, true);
+	sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_keepalive);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index bbb31b9c0b391..d8bd132769594 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -43,13 +43,9 @@ int rds_tcp_keepalive(struct socket *sock)
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int keepalive = 1;
 	int ret = 0;
 
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				(char *)&keepalive, sizeof(keepalive));
-	if (ret < 0)
-		goto bail;
+	sock_set_keepalive(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
 				(char *)&keepcnt, sizeof(keepcnt));
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 845d0be805ece..30082cd039960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2110,7 +2110,6 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 	unsigned int keepidle;
 	unsigned int keepcnt;
-	unsigned int opt_on = 1;
 	unsigned int timeo;
 
 	spin_lock(&xprt->transport_lock);
@@ -2122,8 +2121,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	spin_unlock(&xprt->transport_lock);
 
 	/* TCP Keepalive options */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-			(char *)&opt_on, sizeof(opt_on));
+	sock_set_keepalive(sock->sk);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 07/28] net: add sock_set_keepalive
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_KEEPALIVE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c     |  6 +-----
 include/net/sock.h    |  1 +
 net/core/sock.c       | 10 ++++++++++
 net/rds/tcp_listen.c  |  6 +-----
 net/sunrpc/xprtsock.c |  4 +---
 5 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b4d491122814b..138009c6a2ee1 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1259,11 +1259,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 		con->sock = NULL;
 		goto create_out;
 	}
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				 (char *)&one, sizeof(one));
-	if (result < 0) {
-		log_print("Set keepalive failed: %d", result);
-	}
+	sock_set_keepalive(sock->sk);
 
 	result = sock->ops->listen(sock, 5);
 	if (result < 0) {
diff --git a/include/net/sock.h b/include/net/sock.h
index 99ef43508d2b5..dc08c176238fd 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,6 +2691,7 @@ void sock_def_readable(struct sock *sk);
 int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
+void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
diff --git a/net/core/sock.c b/net/core/sock.c
index e4a4dd2b3d8b3..728f5fb156a0c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -779,6 +779,16 @@ void sock_enable_timestamps(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_enable_timestamps);
 
+void sock_set_keepalive(struct sock *sk)
+{
+	lock_sock(sk);
+	if (sk->sk_prot->keepalive)
+		sk->sk_prot->keepalive(sk, true);
+	sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_keepalive);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index bbb31b9c0b391..d8bd132769594 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -43,13 +43,9 @@ int rds_tcp_keepalive(struct socket *sock)
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int keepalive = 1;
 	int ret = 0;
 
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				(char *)&keepalive, sizeof(keepalive));
-	if (ret < 0)
-		goto bail;
+	sock_set_keepalive(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
 				(char *)&keepcnt, sizeof(keepcnt));
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 845d0be805ece..30082cd039960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2110,7 +2110,6 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 	unsigned int keepidle;
 	unsigned int keepcnt;
-	unsigned int opt_on = 1;
 	unsigned int timeo;
 
 	spin_lock(&xprt->transport_lock);
@@ -2122,8 +2121,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	spin_unlock(&xprt->transport_lock);
 
 	/* TCP Keepalive options */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-			(char *)&opt_on, sizeof(opt_on));
+	sock_set_keepalive(sock->sk);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 07/28] net: add sock_set_keepalive
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_KEEPALIVE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c     |  6 +-----
 include/net/sock.h    |  1 +
 net/core/sock.c       | 10 ++++++++++
 net/rds/tcp_listen.c  |  6 +-----
 net/sunrpc/xprtsock.c |  4 +---
 5 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index b4d491122814b..138009c6a2ee1 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1259,11 +1259,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 		con->sock = NULL;
 		goto create_out;
 	}
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				 (char *)&one, sizeof(one));
-	if (result < 0) {
-		log_print("Set keepalive failed: %d", result);
-	}
+	sock_set_keepalive(sock->sk);
 
 	result = sock->ops->listen(sock, 5);
 	if (result < 0) {
diff --git a/include/net/sock.h b/include/net/sock.h
index 99ef43508d2b5..dc08c176238fd 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2691,6 +2691,7 @@ void sock_def_readable(struct sock *sk);
 int sock_bindtoindex(struct sock *sk, int ifindex);
 void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
+void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
diff --git a/net/core/sock.c b/net/core/sock.c
index e4a4dd2b3d8b3..728f5fb156a0c 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -779,6 +779,16 @@ void sock_enable_timestamps(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_enable_timestamps);
 
+void sock_set_keepalive(struct sock *sk)
+{
+	lock_sock(sk);
+	if (sk->sk_prot->keepalive)
+		sk->sk_prot->keepalive(sk, true);
+	sock_valbool_flag(sk, SOCK_KEEPOPEN, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_keepalive);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index bbb31b9c0b391..d8bd132769594 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -43,13 +43,9 @@ int rds_tcp_keepalive(struct socket *sock)
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int keepalive = 1;
 	int ret = 0;
 
-	ret = kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-				(char *)&keepalive, sizeof(keepalive));
-	if (ret < 0)
-		goto bail;
+	sock_set_keepalive(sock->sk);
 
 	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
 				(char *)&keepcnt, sizeof(keepcnt));
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 845d0be805ece..30082cd039960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2110,7 +2110,6 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 	unsigned int keepidle;
 	unsigned int keepcnt;
-	unsigned int opt_on = 1;
 	unsigned int timeo;
 
 	spin_lock(&xprt->transport_lock);
@@ -2122,8 +2121,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	spin_unlock(&xprt->transport_lock);
 
 	/* TCP Keepalive options */
-	kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
-			(char *)&opt_on, sizeof(opt_on));
+	sock_set_keepalive(sock->sk);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA, Hideaki YOSHIFUJI,
	netdev-u79uwXL29TY76Z2rM5mHXA,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r, Jon Maloy,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA, Eric Dumazet,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA, Ying Xue, Alexey Kuznetsov,
	ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	drbd-dev-cunTk1MwBs8qoQakbn7OcQ

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 08/28] net: add sock_set_rcvbuf
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_RCVBUFFORCE sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/dlm/lowcomms.c  |  7 +-----
 include/net/sock.h |  1 +
 net/core/sock.c    | 59 +++++++++++++++++++++++++---------------------
 3 files changed, 34 insertions(+), 33 deletions(-)

diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 138009c6a2ee1..45c37f572c9d2 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1297,7 +1297,6 @@ static int sctp_listen_for_all(void)
 	struct socket *sock = NULL;
 	int result = -EINVAL;
 	struct connection *con = nodeid2con(0, GFP_NOFS);
-	int bufsize = NEEDED_RMEM;
 	int one = 1;
 
 	if (!con)
@@ -1312,11 +1311,7 @@ static int sctp_listen_for_all(void)
 		goto out;
 	}
 
-	result = kernel_setsockopt(sock, SOL_SOCKET, SO_RCVBUFFORCE,
-				 (char *)&bufsize, sizeof(bufsize));
-	if (result)
-		log_print("Error increasing buffer space on socket %d", result);
-
+	sock_set_rcvbuf(sock->sk, NEEDED_RMEM);
 	result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
 				   sizeof(one));
 	if (result < 0)
diff --git a/include/net/sock.h b/include/net/sock.h
index dc08c176238fd..c997289aabbf9 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2693,6 +2693,7 @@ void sock_enable_timestamps(struct sock *sk);
 void sock_no_linger(struct sock *sk);
 void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
+void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
diff --git a/net/core/sock.c b/net/core/sock.c
index 728f5fb156a0c..3c6ebf952e9ad 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -789,6 +789,35 @@ void sock_set_keepalive(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_keepalive);
 
+static void __sock_set_rcvbuf(struct sock *sk, int val)
+{
+	/* Ensure val * 2 fits into an int, to prevent max_t() from treating it
+	 * as a negative value.
+	 */
+	val = min_t(int, val, INT_MAX / 2);
+	sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
+
+	/* We double it on the way in to account for "struct sk_buff" etc.
+	 * overhead.   Applications assume that the SO_RCVBUF setting they make
+	 * will allow that much actual data to be received on that socket.
+	 *
+	 * Applications are unaware that "struct sk_buff" and other overheads
+	 * allocate from the receive buffer during socket buffer allocation.
+	 *
+	 * And after considering the possible alternatives, returning the value
+	 * we actually used in getsockopt is the most desirable behavior.
+	 */
+	WRITE_ONCE(sk->sk_rcvbuf, max_t(int, val * 2, SOCK_MIN_RCVBUF));
+}
+
+void sock_set_rcvbuf(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__sock_set_rcvbuf(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_rcvbuf);
+
 /*
  *	This is meant for all protocols to use and covers goings on
  *	at the socket level. Everything here is generic.
@@ -885,30 +914,7 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		 * play 'guess the biggest size' games. RCVBUF/SNDBUF
 		 * are treated in BSD as hints
 		 */
-		val = min_t(u32, val, sysctl_rmem_max);
-set_rcvbuf:
-		/* Ensure val * 2 fits into an int, to prevent max_t()
-		 * from treating it as a negative value.
-		 */
-		val = min_t(int, val, INT_MAX / 2);
-		sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
-		/*
-		 * We double it on the way in to account for
-		 * "struct sk_buff" etc. overhead.   Applications
-		 * assume that the SO_RCVBUF setting they make will
-		 * allow that much actual data to be received on that
-		 * socket.
-		 *
-		 * Applications are unaware that "struct sk_buff" and
-		 * other overheads allocate from the receive buffer
-		 * during socket buffer allocation.
-		 *
-		 * And after considering the possible alternatives,
-		 * returning the value we actually used in getsockopt
-		 * is the most desirable behavior.
-		 */
-		WRITE_ONCE(sk->sk_rcvbuf,
-			   max_t(int, val * 2, SOCK_MIN_RCVBUF));
+		__sock_set_rcvbuf(sk, min_t(u32, val, sysctl_rmem_max));
 		break;
 
 	case SO_RCVBUFFORCE:
@@ -920,9 +926,8 @@ int sock_setsockopt(struct socket *sock, int level, int optname,
 		/* No negative values (to prevent underflow, as val will be
 		 * multiplied by 2).
 		 */
-		if (val < 0)
-			val = 0;
-		goto set_rcvbuf;
+		__sock_set_rcvbuf(sk, max(val, 0));
+		break;
 
 	case SO_KEEPALIVE:
 		if (sk->sk_prot->keepalive)
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 09/28] net: add sock_set_reuseport
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_REUSEPORT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h    |  1 +
 net/core/sock.c       |  8 ++++++++
 net/sunrpc/xprtsock.c | 17 +----------------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c997289aabbf9..d994daa418ec2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2695,6 +2695,7 @@ void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_reuseport(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 3c6ebf952e9ad..2ca3425b519c0 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_set_reuseport(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuseport = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseport);
+
 void sock_no_linger(struct sock *sk)
 {
 	lock_sock(sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 30082cd039960..399848c2bcb29 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1594,21 +1594,6 @@ static int xs_get_random_port(void)
 	return rand + min;
 }
 
-/**
- * xs_set_reuseaddr_port - set the socket's port and address reuse options
- * @sock: socket
- *
- * Note that this function has to be called on all sockets that share the
- * same port, and it must be called before binding.
- */
-static void xs_sock_set_reuseport(struct socket *sock)
-{
-	int opt = 1;
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
-			(char *)&opt, sizeof(opt));
-}
-
 static unsigned short xs_sock_getport(struct socket *sock)
 {
 	struct sockaddr_storage buf;
@@ -1801,7 +1786,7 @@ static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 	xs_reclassify_socket(family, sock);
 
 	if (reuseport)
-		xs_sock_set_reuseport(sock);
+		sock_set_reuseport(sock->sk);
 
 	err = xs_bind(transport, sock);
 	if (err) {
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 09/28] net: add sock_set_reuseport
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_REUSEPORT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h    |  1 +
 net/core/sock.c       |  8 ++++++++
 net/sunrpc/xprtsock.c | 17 +----------------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c997289aabbf9..d994daa418ec2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2695,6 +2695,7 @@ void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_reuseport(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 3c6ebf952e9ad..2ca3425b519c0 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_set_reuseport(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuseport = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseport);
+
 void sock_no_linger(struct sock *sk)
 {
 	lock_sock(sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 30082cd039960..399848c2bcb29 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1594,21 +1594,6 @@ static int xs_get_random_port(void)
 	return rand + min;
 }
 
-/**
- * xs_set_reuseaddr_port - set the socket's port and address reuse options
- * @sock: socket
- *
- * Note that this function has to be called on all sockets that share the
- * same port, and it must be called before binding.
- */
-static void xs_sock_set_reuseport(struct socket *sock)
-{
-	int opt = 1;
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
-			(char *)&opt, sizeof(opt));
-}
-
 static unsigned short xs_sock_getport(struct socket *sock)
 {
 	struct sockaddr_storage buf;
@@ -1801,7 +1786,7 @@ static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 	xs_reclassify_socket(family, sock);
 
 	if (reuseport)
-		xs_sock_set_reuseport(sock);
+		sock_set_reuseport(sock->sk);
 
 	err = xs_bind(transport, sock);
 	if (err) {
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 09/28] net: add sock_set_reuseport
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the SO_REUSEPORT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h    |  1 +
 net/core/sock.c       |  8 ++++++++
 net/sunrpc/xprtsock.c | 17 +----------------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c997289aabbf9..d994daa418ec2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2695,6 +2695,7 @@ void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_reuseport(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 3c6ebf952e9ad..2ca3425b519c0 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_set_reuseport(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuseport = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseport);
+
 void sock_no_linger(struct sock *sk)
 {
 	lock_sock(sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 30082cd039960..399848c2bcb29 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1594,21 +1594,6 @@ static int xs_get_random_port(void)
 	return rand + min;
 }
 
-/**
- * xs_set_reuseaddr_port - set the socket's port and address reuse options
- * @sock: socket
- *
- * Note that this function has to be called on all sockets that share the
- * same port, and it must be called before binding.
- */
-static void xs_sock_set_reuseport(struct socket *sock)
-{
-	int opt = 1;
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
-			(char *)&opt, sizeof(opt));
-}
-
 static unsigned short xs_sock_getport(struct socket *sock)
 {
 	struct sockaddr_storage buf;
@@ -1801,7 +1786,7 @@ static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 	xs_reclassify_socket(family, sock);
 
 	if (reuseport)
-		xs_sock_set_reuseport(sock);
+		sock_set_reuseport(sock->sk);
 
 	err = xs_bind(transport, sock);
 	if (err) {
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 09/28] net: add sock_set_reuseport
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the SO_REUSEPORT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h    |  1 +
 net/core/sock.c       |  8 ++++++++
 net/sunrpc/xprtsock.c | 17 +----------------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c997289aabbf9..d994daa418ec2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2695,6 +2695,7 @@ void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_reuseport(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 3c6ebf952e9ad..2ca3425b519c0 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_set_reuseport(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuseport = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseport);
+
 void sock_no_linger(struct sock *sk)
 {
 	lock_sock(sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 30082cd039960..399848c2bcb29 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1594,21 +1594,6 @@ static int xs_get_random_port(void)
 	return rand + min;
 }
 
-/**
- * xs_set_reuseaddr_port - set the socket's port and address reuse options
- * @sock: socket
- *
- * Note that this function has to be called on all sockets that share the
- * same port, and it must be called before binding.
- */
-static void xs_sock_set_reuseport(struct socket *sock)
-{
-	int opt = 1;
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
-			(char *)&opt, sizeof(opt));
-}
-
 static unsigned short xs_sock_getport(struct socket *sock)
 {
 	struct sockaddr_storage buf;
@@ -1801,7 +1786,7 @@ static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 	xs_reclassify_socket(family, sock);
 
 	if (reuseport)
-		xs_sock_set_reuseport(sock);
+		sock_set_reuseport(sock->sk);
 
 	err = xs_bind(transport, sock);
 	if (err) {
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 09/28] net: add sock_set_reuseport
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the SO_REUSEPORT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/sock.h    |  1 +
 net/core/sock.c       |  8 ++++++++
 net/sunrpc/xprtsock.c | 17 +----------------
 3 files changed, 10 insertions(+), 16 deletions(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index c997289aabbf9..d994daa418ec2 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -2695,6 +2695,7 @@ void sock_set_keepalive(struct sock *sk);
 void sock_set_priority(struct sock *sk, u32 priority);
 void sock_set_rcvbuf(struct sock *sk, int val);
 void sock_set_reuseaddr(struct sock *sk);
+void sock_set_reuseport(struct sock *sk);
 void sock_set_sndtimeo(struct sock *sk, s64 secs);
 
 #endif	/* _SOCK_H */
diff --git a/net/core/sock.c b/net/core/sock.c
index 3c6ebf952e9ad..2ca3425b519c0 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -729,6 +729,14 @@ void sock_set_reuseaddr(struct sock *sk)
 }
 EXPORT_SYMBOL(sock_set_reuseaddr);
 
+void sock_set_reuseport(struct sock *sk)
+{
+	lock_sock(sk);
+	sk->sk_reuseport = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(sock_set_reuseport);
+
 void sock_no_linger(struct sock *sk)
 {
 	lock_sock(sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 30082cd039960..399848c2bcb29 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1594,21 +1594,6 @@ static int xs_get_random_port(void)
 	return rand + min;
 }
 
-/**
- * xs_set_reuseaddr_port - set the socket's port and address reuse options
- * @sock: socket
- *
- * Note that this function has to be called on all sockets that share the
- * same port, and it must be called before binding.
- */
-static void xs_sock_set_reuseport(struct socket *sock)
-{
-	int opt = 1;
-
-	kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
-			(char *)&opt, sizeof(opt));
-}
-
 static unsigned short xs_sock_getport(struct socket *sock)
 {
 	struct sockaddr_storage buf;
@@ -1801,7 +1786,7 @@ static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 	xs_reclassify_socket(family, sock);
 
 	if (reuseport)
-		xs_sock_set_reuseport(sock);
+		sock_set_reuseport(sock->sk);
 
 	err = xs_bind(transport, sock);
 	if (err) {
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 10/28] tcp: add tcp_sock_set_cork
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_CORK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      | 14 --------
 drivers/block/drbd/drbd_receiver.c |  4 +--
 drivers/block/drbd/drbd_worker.c   |  6 ++--
 fs/cifs/transport.c                |  8 ++---
 include/linux/tcp.h                |  2 ++
 net/ipv4/tcp.c                     | 51 +++++++++++++++++++-----------
 net/rds/tcp_send.c                 |  9 ++----
 7 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index aae99a2d7bd40..3550adc93c68b 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_cork(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
-	int val = 0;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_nodelay(struct socket *sock)
 {
 	int val = 1;
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index c15e7083b13a6..55ea907ad33cb 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	rcu_read_unlock();
 
 	if (tcp_cork)
-		drbd_tcp_cork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, true);
 
 	err = drbd_finish_peer_reqs(device);
 	kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	}
 
 	if (tcp_cork)
-		drbd_tcp_uncork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, false);
 
 	return;
 }
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 0dc019da1f8d0..2b89c9f2ca707 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	if (uncork) {
 		mutex_lock(&connection->data.mutex);
 		if (connection->data.socket)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 		mutex_unlock(&connection->data.mutex);
 	}
 
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	mutex_lock(&connection->data.mutex);
 	if (connection->data.socket) {
 		if (cork)
-			drbd_tcp_cork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, true);
 		else if (!uncork)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 	}
 	mutex_unlock(&connection->data.mutex);
 }
diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index c97570eb2c180..99760063e0006 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	size_t total_len = 0, sent, size;
 	struct socket *ssocket = server->ssocket;
 	struct msghdr smb_msg;
-	int val = 1;
 	__be32 rfc1002_marker;
 
 	if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* cork the socket */
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, true);
 
 	for (j = 0; j < num_rqst; j++)
 		send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* uncork it */
-	val = 0;
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, false);
 
 	if ((total_len > 0) && (total_len != send_length)) {
 		cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index bf44e85d709dc..889eeb2256c2d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 9700649963773..e6cf702e16d66 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
 	}
 }
 
+/* When set indicates to always queue non-full frames.  Later the user clears
+ * this option and we transmit any pending partial frames in the queue.  This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (on) {
+		tp->nonagle |= TCP_NAGLE_CORK;
+	} else {
+		tp->nonagle &= ~TCP_NAGLE_CORK;
+		if (tp->nonagle & TCP_NAGLE_OFF)
+			tp->nonagle |= TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	}
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	lock_sock(sk);
+	__tcp_sock_set_cork(sk, on);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_CORK:
-		/* When set indicates to always queue non-full frames.
-		 * Later the user clears this option and we transmit
-		 * any pending partial frames in the queue.  This is
-		 * meant to be used alongside sendfile() to get properly
-		 * filled frames when the user (for example) must write
-		 * out headers with a write() call first and then use
-		 * sendfile to send out the data parts.
-		 *
-		 * TCP_CORK can be set together with TCP_NODELAY and it is
-		 * stronger than TCP_NODELAY.
-		 */
-		if (val) {
-			tp->nonagle |= TCP_NAGLE_CORK;
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_CORK;
-			if (tp->nonagle&TCP_NAGLE_OFF)
-				tp->nonagle |= TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		}
+		__tcp_sock_set_cork(sk, val);
 		break;
 
 	case TCP_KEEPIDLE:
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 78a2554a44979..8c4d1d6e9249d 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -38,23 +38,18 @@
 #include "rds.h"
 #include "tcp.h"
 
-static void rds_tcp_cork(struct socket *sock, int val)
-{
-	kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
 void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 1);
+	tcp_sock_set_cork(tc->t_sock->sk, true);
 }
 
 void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 0);
+	tcp_sock_set_cork(tc->t_sock->sk, false);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 10/28] tcp: add tcp_sock_set_cork
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_CORK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      | 14 --------
 drivers/block/drbd/drbd_receiver.c |  4 +--
 drivers/block/drbd/drbd_worker.c   |  6 ++--
 fs/cifs/transport.c                |  8 ++---
 include/linux/tcp.h                |  2 ++
 net/ipv4/tcp.c                     | 51 +++++++++++++++++++-----------
 net/rds/tcp_send.c                 |  9 ++----
 7 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index aae99a2d7bd40..3550adc93c68b 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_cork(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
-	int val = 0;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_nodelay(struct socket *sock)
 {
 	int val = 1;
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index c15e7083b13a6..55ea907ad33cb 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	rcu_read_unlock();
 
 	if (tcp_cork)
-		drbd_tcp_cork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, true);
 
 	err = drbd_finish_peer_reqs(device);
 	kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	}
 
 	if (tcp_cork)
-		drbd_tcp_uncork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, false);
 
 	return;
 }
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 0dc019da1f8d0..2b89c9f2ca707 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	if (uncork) {
 		mutex_lock(&connection->data.mutex);
 		if (connection->data.socket)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 		mutex_unlock(&connection->data.mutex);
 	}
 
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	mutex_lock(&connection->data.mutex);
 	if (connection->data.socket) {
 		if (cork)
-			drbd_tcp_cork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, true);
 		else if (!uncork)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 	}
 	mutex_unlock(&connection->data.mutex);
 }
diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index c97570eb2c180..99760063e0006 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	size_t total_len = 0, sent, size;
 	struct socket *ssocket = server->ssocket;
 	struct msghdr smb_msg;
-	int val = 1;
 	__be32 rfc1002_marker;
 
 	if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* cork the socket */
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, true);
 
 	for (j = 0; j < num_rqst; j++)
 		send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* uncork it */
-	val = 0;
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, false);
 
 	if ((total_len > 0) && (total_len != send_length)) {
 		cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index bf44e85d709dc..889eeb2256c2d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 9700649963773..e6cf702e16d66 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
 	}
 }
 
+/* When set indicates to always queue non-full frames.  Later the user clears
+ * this option and we transmit any pending partial frames in the queue.  This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (on) {
+		tp->nonagle |= TCP_NAGLE_CORK;
+	} else {
+		tp->nonagle &= ~TCP_NAGLE_CORK;
+		if (tp->nonagle & TCP_NAGLE_OFF)
+			tp->nonagle |= TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	}
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	lock_sock(sk);
+	__tcp_sock_set_cork(sk, on);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_CORK:
-		/* When set indicates to always queue non-full frames.
-		 * Later the user clears this option and we transmit
-		 * any pending partial frames in the queue.  This is
-		 * meant to be used alongside sendfile() to get properly
-		 * filled frames when the user (for example) must write
-		 * out headers with a write() call first and then use
-		 * sendfile to send out the data parts.
-		 *
-		 * TCP_CORK can be set together with TCP_NODELAY and it is
-		 * stronger than TCP_NODELAY.
-		 */
-		if (val) {
-			tp->nonagle |= TCP_NAGLE_CORK;
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_CORK;
-			if (tp->nonagle&TCP_NAGLE_OFF)
-				tp->nonagle |= TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		}
+		__tcp_sock_set_cork(sk, val);
 		break;
 
 	case TCP_KEEPIDLE:
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 78a2554a44979..8c4d1d6e9249d 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -38,23 +38,18 @@
 #include "rds.h"
 #include "tcp.h"
 
-static void rds_tcp_cork(struct socket *sock, int val)
-{
-	kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
 void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 1);
+	tcp_sock_set_cork(tc->t_sock->sk, true);
 }
 
 void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 0);
+	tcp_sock_set_cork(tc->t_sock->sk, false);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 10/28] tcp: add tcp_sock_set_cork
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_CORK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      | 14 --------
 drivers/block/drbd/drbd_receiver.c |  4 +--
 drivers/block/drbd/drbd_worker.c   |  6 ++--
 fs/cifs/transport.c                |  8 ++---
 include/linux/tcp.h                |  2 ++
 net/ipv4/tcp.c                     | 51 +++++++++++++++++++-----------
 net/rds/tcp_send.c                 |  9 ++----
 7 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index aae99a2d7bd40..3550adc93c68b 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_cork(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
-	int val = 0;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_nodelay(struct socket *sock)
 {
 	int val = 1;
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index c15e7083b13a6..55ea907ad33cb 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	rcu_read_unlock();
 
 	if (tcp_cork)
-		drbd_tcp_cork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, true);
 
 	err = drbd_finish_peer_reqs(device);
 	kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	}
 
 	if (tcp_cork)
-		drbd_tcp_uncork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, false);
 
 	return;
 }
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 0dc019da1f8d0..2b89c9f2ca707 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	if (uncork) {
 		mutex_lock(&connection->data.mutex);
 		if (connection->data.socket)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 		mutex_unlock(&connection->data.mutex);
 	}
 
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	mutex_lock(&connection->data.mutex);
 	if (connection->data.socket) {
 		if (cork)
-			drbd_tcp_cork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, true);
 		else if (!uncork)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 	}
 	mutex_unlock(&connection->data.mutex);
 }
diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index c97570eb2c180..99760063e0006 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	size_t total_len = 0, sent, size;
 	struct socket *ssocket = server->ssocket;
 	struct msghdr smb_msg;
-	int val = 1;
 	__be32 rfc1002_marker;
 
 	if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* cork the socket */
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, true);
 
 	for (j = 0; j < num_rqst; j++)
 		send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* uncork it */
-	val = 0;
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, false);
 
 	if ((total_len > 0) && (total_len != send_length)) {
 		cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index bf44e85d709dc..889eeb2256c2d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 9700649963773..e6cf702e16d66 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
 	}
 }
 
+/* When set indicates to always queue non-full frames.  Later the user clears
+ * this option and we transmit any pending partial frames in the queue.  This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (on) {
+		tp->nonagle |= TCP_NAGLE_CORK;
+	} else {
+		tp->nonagle &= ~TCP_NAGLE_CORK;
+		if (tp->nonagle & TCP_NAGLE_OFF)
+			tp->nonagle |= TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	}
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	lock_sock(sk);
+	__tcp_sock_set_cork(sk, on);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_CORK:
-		/* When set indicates to always queue non-full frames.
-		 * Later the user clears this option and we transmit
-		 * any pending partial frames in the queue.  This is
-		 * meant to be used alongside sendfile() to get properly
-		 * filled frames when the user (for example) must write
-		 * out headers with a write() call first and then use
-		 * sendfile to send out the data parts.
-		 *
-		 * TCP_CORK can be set together with TCP_NODELAY and it is
-		 * stronger than TCP_NODELAY.
-		 */
-		if (val) {
-			tp->nonagle |= TCP_NAGLE_CORK;
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_CORK;
-			if (tp->nonagle&TCP_NAGLE_OFF)
-				tp->nonagle |= TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		}
+		__tcp_sock_set_cork(sk, val);
 		break;
 
 	case TCP_KEEPIDLE:
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 78a2554a44979..8c4d1d6e9249d 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -38,23 +38,18 @@
 #include "rds.h"
 #include "tcp.h"
 
-static void rds_tcp_cork(struct socket *sock, int val)
-{
-	kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
 void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 1);
+	tcp_sock_set_cork(tc->t_sock->sk, true);
 }
 
 void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 0);
+	tcp_sock_set_cork(tc->t_sock->sk, false);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 10/28] tcp: add tcp_sock_set_cork
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_CORK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      | 14 --------
 drivers/block/drbd/drbd_receiver.c |  4 +--
 drivers/block/drbd/drbd_worker.c   |  6 ++--
 fs/cifs/transport.c                |  8 ++---
 include/linux/tcp.h                |  2 ++
 net/ipv4/tcp.c                     | 51 +++++++++++++++++++-----------
 net/rds/tcp_send.c                 |  9 ++----
 7 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index aae99a2d7bd40..3550adc93c68b 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_cork(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
-	int val = 0;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_nodelay(struct socket *sock)
 {
 	int val = 1;
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index c15e7083b13a6..55ea907ad33cb 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	rcu_read_unlock();
 
 	if (tcp_cork)
-		drbd_tcp_cork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, true);
 
 	err = drbd_finish_peer_reqs(device);
 	kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	}
 
 	if (tcp_cork)
-		drbd_tcp_uncork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, false);
 
 	return;
 }
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 0dc019da1f8d0..2b89c9f2ca707 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	if (uncork) {
 		mutex_lock(&connection->data.mutex);
 		if (connection->data.socket)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 		mutex_unlock(&connection->data.mutex);
 	}
 
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	mutex_lock(&connection->data.mutex);
 	if (connection->data.socket) {
 		if (cork)
-			drbd_tcp_cork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, true);
 		else if (!uncork)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 	}
 	mutex_unlock(&connection->data.mutex);
 }
diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index c97570eb2c180..99760063e0006 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	size_t total_len = 0, sent, size;
 	struct socket *ssocket = server->ssocket;
 	struct msghdr smb_msg;
-	int val = 1;
 	__be32 rfc1002_marker;
 
 	if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* cork the socket */
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, true);
 
 	for (j = 0; j < num_rqst; j++)
 		send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* uncork it */
-	val = 0;
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, false);
 
 	if ((total_len > 0) && (total_len != send_length)) {
 		cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index bf44e85d709dc..889eeb2256c2d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 9700649963773..e6cf702e16d66 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
 	}
 }
 
+/* When set indicates to always queue non-full frames.  Later the user clears
+ * this option and we transmit any pending partial frames in the queue.  This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (on) {
+		tp->nonagle |= TCP_NAGLE_CORK;
+	} else {
+		tp->nonagle &= ~TCP_NAGLE_CORK;
+		if (tp->nonagle & TCP_NAGLE_OFF)
+			tp->nonagle |= TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	}
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	lock_sock(sk);
+	__tcp_sock_set_cork(sk, on);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_CORK:
-		/* When set indicates to always queue non-full frames.
-		 * Later the user clears this option and we transmit
-		 * any pending partial frames in the queue.  This is
-		 * meant to be used alongside sendfile() to get properly
-		 * filled frames when the user (for example) must write
-		 * out headers with a write() call first and then use
-		 * sendfile to send out the data parts.
-		 *
-		 * TCP_CORK can be set together with TCP_NODELAY and it is
-		 * stronger than TCP_NODELAY.
-		 */
-		if (val) {
-			tp->nonagle |= TCP_NAGLE_CORK;
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_CORK;
-			if (tp->nonagle&TCP_NAGLE_OFF)
-				tp->nonagle |= TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		}
+		__tcp_sock_set_cork(sk, val);
 		break;
 
 	case TCP_KEEPIDLE:
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 78a2554a44979..8c4d1d6e9249d 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -38,23 +38,18 @@
 #include "rds.h"
 #include "tcp.h"
 
-static void rds_tcp_cork(struct socket *sock, int val)
-{
-	kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
 void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 1);
+	tcp_sock_set_cork(tc->t_sock->sk, true);
 }
 
 void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 0);
+	tcp_sock_set_cork(tc->t_sock->sk, false);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 10/28] tcp: add tcp_sock_set_cork
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_CORK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      | 14 --------
 drivers/block/drbd/drbd_receiver.c |  4 +--
 drivers/block/drbd/drbd_worker.c   |  6 ++--
 fs/cifs/transport.c                |  8 ++---
 include/linux/tcp.h                |  2 ++
 net/ipv4/tcp.c                     | 51 +++++++++++++++++++-----------
 net/rds/tcp_send.c                 |  9 ++----
 7 files changed, 43 insertions(+), 51 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index aae99a2d7bd40..3550adc93c68b 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,20 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_cork(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
-static inline void drbd_tcp_uncork(struct socket *sock)
-{
-	int val = 0;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_nodelay(struct socket *sock)
 {
 	int val = 1;
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index c15e7083b13a6..55ea907ad33cb 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -6162,7 +6162,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	rcu_read_unlock();
 
 	if (tcp_cork)
-		drbd_tcp_cork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, true);
 
 	err = drbd_finish_peer_reqs(device);
 	kref_put(&device->kref, drbd_destroy_device);
@@ -6175,7 +6175,7 @@ void drbd_send_acks_wf(struct work_struct *ws)
 	}
 
 	if (tcp_cork)
-		drbd_tcp_uncork(connection->meta.socket);
+		tcp_sock_set_cork(connection->meta.socket->sk, false);
 
 	return;
 }
diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c
index 0dc019da1f8d0..2b89c9f2ca707 100644
--- a/drivers/block/drbd/drbd_worker.c
+++ b/drivers/block/drbd/drbd_worker.c
@@ -2098,7 +2098,7 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	if (uncork) {
 		mutex_lock(&connection->data.mutex);
 		if (connection->data.socket)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 		mutex_unlock(&connection->data.mutex);
 	}
 
@@ -2153,9 +2153,9 @@ static void wait_for_work(struct drbd_connection *connection, struct list_head *
 	mutex_lock(&connection->data.mutex);
 	if (connection->data.socket) {
 		if (cork)
-			drbd_tcp_cork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, true);
 		else if (!uncork)
-			drbd_tcp_uncork(connection->data.socket);
+			tcp_sock_set_cork(connection->data.socket->sk, false);
 	}
 	mutex_unlock(&connection->data.mutex);
 }
diff --git a/fs/cifs/transport.c b/fs/cifs/transport.c
index c97570eb2c180..99760063e0006 100644
--- a/fs/cifs/transport.c
+++ b/fs/cifs/transport.c
@@ -325,7 +325,6 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	size_t total_len = 0, sent, size;
 	struct socket *ssocket = server->ssocket;
 	struct msghdr smb_msg;
-	int val = 1;
 	__be32 rfc1002_marker;
 
 	if (cifs_rdma_enabled(server)) {
@@ -345,8 +344,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* cork the socket */
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, true);
 
 	for (j = 0; j < num_rqst; j++)
 		send_length += smb_rqst_len(server, &rqst[j]);
@@ -435,9 +433,7 @@ __smb_send_rqst(struct TCP_Server_Info *server, int num_rqst,
 	}
 
 	/* uncork it */
-	val = 0;
-	kernel_setsockopt(ssocket, SOL_TCP, TCP_CORK,
-				(char *)&val, sizeof(val));
+	tcp_sock_set_cork(ssocket->sk, false);
 
 	if ((total_len > 0) && (total_len != send_length)) {
 		cifs_dbg(FYI, "partial send (wanted=%u sent=%zu): terminating session\n",
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index bf44e85d709dc..889eeb2256c2d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -497,4 +497,6 @@ static inline u16 tcp_mss_clamp(const struct tcp_sock *tp, u16 mss)
 int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
+void tcp_sock_set_cork(struct sock *sk, bool on);
+
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 9700649963773..e6cf702e16d66 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2801,6 +2801,37 @@ static void tcp_enable_tx_delay(void)
 	}
 }
 
+/* When set indicates to always queue non-full frames.  Later the user clears
+ * this option and we transmit any pending partial frames in the queue.  This is
+ * meant to be used alongside sendfile() to get properly filled frames when the
+ * user (for example) must write out headers with a write() call first and then
+ * use sendfile to send out the data parts.
+ *
+ * TCP_CORK can be set together with TCP_NODELAY and it is stronger than
+ * TCP_NODELAY.
+ */
+static void __tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (on) {
+		tp->nonagle |= TCP_NAGLE_CORK;
+	} else {
+		tp->nonagle &= ~TCP_NAGLE_CORK;
+		if (tp->nonagle & TCP_NAGLE_OFF)
+			tp->nonagle |= TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	}
+}
+
+void tcp_sock_set_cork(struct sock *sk, bool on)
+{
+	lock_sock(sk);
+	__tcp_sock_set_cork(sk, on);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_cork);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2979,25 +3010,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_CORK:
-		/* When set indicates to always queue non-full frames.
-		 * Later the user clears this option and we transmit
-		 * any pending partial frames in the queue.  This is
-		 * meant to be used alongside sendfile() to get properly
-		 * filled frames when the user (for example) must write
-		 * out headers with a write() call first and then use
-		 * sendfile to send out the data parts.
-		 *
-		 * TCP_CORK can be set together with TCP_NODELAY and it is
-		 * stronger than TCP_NODELAY.
-		 */
-		if (val) {
-			tp->nonagle |= TCP_NAGLE_CORK;
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_CORK;
-			if (tp->nonagle&TCP_NAGLE_OFF)
-				tp->nonagle |= TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		}
+		__tcp_sock_set_cork(sk, val);
 		break;
 
 	case TCP_KEEPIDLE:
diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
index 78a2554a44979..8c4d1d6e9249d 100644
--- a/net/rds/tcp_send.c
+++ b/net/rds/tcp_send.c
@@ -38,23 +38,18 @@
 #include "rds.h"
 #include "tcp.h"
 
-static void rds_tcp_cork(struct socket *sock, int val)
-{
-	kernel_setsockopt(sock, SOL_TCP, TCP_CORK, (void *)&val, sizeof(val));
-}
-
 void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 1);
+	tcp_sock_set_cork(tc->t_sock->sk, true);
 }
 
 void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
 {
 	struct rds_tcp_connection *tc = cp->cp_transport_data;
 
-	rds_tcp_cork(tc->t_sock, 0);
+	tcp_sock_set_cork(tc->t_sock->sk, false);
 }
 
 /* the core send_sem serializes this with other xmit and shutdown */
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg, Jason Gunthorpe

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport == ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport == ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid == 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family == AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg,
	Jason Gunthorpe

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd = P_PING || cmd = P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle = false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle = false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle = false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle = false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport = ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport = ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid = 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family = AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg,
	Jason Gunthorpe

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Acked-by: Sagi Grimberg <sagi-NQWnxTmZq1alnMjI0IkVqw@public.gmane.org>
Acked-by: Jason Gunthorpe <jgg-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport == ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport == ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid == 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family == AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, Jason Gunthorpe, linux-kernel,
	linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet, target-devel,
	tipc-discussion, linux-nfs, Ying Xue, Alexey Kuznetsov,
	ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport == ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport == ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid == 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family == AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg,
	Jason Gunthorpe

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport == ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport == ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid == 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family == AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 11/28] tcp: add tcp_sock_set_nodelay
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_NODELAY sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
Acked-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/block/drbd/drbd_int.h             |  7 ----
 drivers/block/drbd/drbd_main.c            |  2 +-
 drivers/block/drbd/drbd_receiver.c        |  4 +--
 drivers/infiniband/sw/siw/siw_cm.c        | 24 +++-----------
 drivers/nvme/host/tcp.c                   |  9 +-----
 drivers/nvme/target/tcp.c                 | 12 ++-----
 drivers/target/iscsi/iscsi_target_login.c | 15 ++-------
 fs/cifs/connect.c                         | 10 ++----
 fs/dlm/lowcomms.c                         |  8 ++---
 fs/ocfs2/cluster/tcp.c                    | 20 ++----------
 include/linux/tcp.h                       |  1 +
 net/ceph/messenger.c                      | 11 ++-----
 net/ipv4/tcp.c                            | 39 +++++++++++++++--------
 net/rds/tcp.c                             | 11 +------
 net/rds/tcp.h                             |  1 -
 net/rds/tcp_listen.c                      |  2 +-
 16 files changed, 49 insertions(+), 127 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 3550adc93c68b..e24bba87c8e02 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_nodelay(struct socket *sock)
-{
-	int val = 1;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-			(char*)&val, sizeof(val));
-}
-
 static inline void drbd_tcp_quickack(struct socket *sock)
 {
 	int val = 2;
diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c
index c094c3c2c5d4d..45fbd526c453b 100644
--- a/drivers/block/drbd/drbd_main.c
+++ b/drivers/block/drbd/drbd_main.c
@@ -660,7 +660,7 @@ static int __send_command(struct drbd_connection *connection, int vnr,
 	/* DRBD protocol "pings" are latency critical.
 	 * This is supposed to trigger tcp_push_pending_frames() */
 	if (!err && (cmd == P_PING || cmd == P_PING_ACK))
-		drbd_tcp_nodelay(sock->socket);
+		tcp_sock_set_nodelay(sock->socket->sk);
 
 	return err;
 }
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 55ea907ad33cb..20a5e94494acd 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1051,8 +1051,8 @@ static int conn_connect(struct drbd_connection *connection)
 
 	/* we don't want delays.
 	 * we use TCP_CORK where appropriate, though */
-	drbd_tcp_nodelay(sock.socket);
-	drbd_tcp_nodelay(msock.socket);
+	tcp_sock_set_nodelay(sock.socket->sk);
+	tcp_sock_set_nodelay(msock.socket->sk);
 
 	connection->data.socket = sock.socket;
 	connection->meta.socket = msock.socket;
diff --git a/drivers/infiniband/sw/siw/siw_cm.c b/drivers/infiniband/sw/siw/siw_cm.c
index d1860f3e87401..1662216be66df 100644
--- a/drivers/infiniband/sw/siw/siw_cm.c
+++ b/drivers/infiniband/sw/siw/siw_cm.c
@@ -947,16 +947,8 @@ static void siw_accept_newconn(struct siw_cep *cep)
 	siw_cep_get(new_cep);
 	new_s->sk->sk_user_data = new_cep;
 
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(new_s, SOL_TCP, TCP_NODELAY,
-				       (char *)&val, sizeof(val));
-		if (rv) {
-			siw_dbg_cep(cep, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(new_s->sk);
 	new_cep->state = SIW_EPSTATE_AWAIT_MPAREQ;
 
 	rv = siw_cm_queue_work(new_cep, SIW_CM_WORK_MPATIMEOUT);
@@ -1386,16 +1378,8 @@ int siw_connect(struct iw_cm_id *id, struct iw_cm_conn_param *params)
 		siw_dbg_qp(qp, "kernel_bindconnect: error %d\n", rv);
 		goto error;
 	}
-	if (siw_tcp_nagle == false) {
-		int val = 1;
-
-		rv = kernel_setsockopt(s, SOL_TCP, TCP_NODELAY, (char *)&val,
-				       sizeof(val));
-		if (rv) {
-			siw_dbg_qp(qp, "setsockopt NODELAY error: %d\n", rv);
-			goto error;
-		}
-	}
+	if (siw_tcp_nagle == false)
+		tcp_sock_set_nodelay(s->sk);
 	cep = siw_cep_alloc(sdev);
 	if (!cep) {
 		rv = -ENOMEM;
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index a307972d33a02..4e4a750ecdb97 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1346,14 +1346,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Set TCP no delay */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_nodelay(queue->sock->sk);
 
 	/*
 	 * Cleanup whatever is sitting in the TCP transmit queue on socket
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f3088156d01da..55bc4c3c0a74a 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1580,7 +1580,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 {
 	struct nvmet_tcp_port *port;
 	__kernel_sa_family_t af;
-	int opt, ret;
+	int ret;
 
 	port = kzalloc(sizeof(*port), GFP_KERNEL);
 	if (!port)
@@ -1625,15 +1625,7 @@ static int nvmet_tcp_add_port(struct nvmet_port *nport)
 	port->data_ready = port->sock->sk->sk_data_ready;
 	port->sock->sk->sk_data_ready = nvmet_tcp_listen_data_ready;
 	sock_set_reuseaddr(port->sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(port->sock, IPPROTO_TCP,
-			TCP_NODELAY, (char *)&opt, sizeof(opt));
-	if (ret) {
-		pr_err("failed to set TCP_NODELAY sock opt %d\n", ret);
-		goto err_sock;
-	}
-
+	tcp_sock_set_nodelay(port->sock->sk);
 	if (so_priority > 0)
 		sock_set_priority(port->sock->sk, so_priority);
 
diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index 91acb3f07b4cc..b561b07a869a0 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -897,20 +897,11 @@ int iscsit_setup_np(
 	/*
 	 * Set SO_REUSEADDR, and disable Nagel Algorithm with TCP_NODELAY.
 	 */
-	/* FIXME: Someone please explain why this is endian-safe */
-	opt = 1;
-	if (np->np_network_transport == ISCSI_TCP) {
-		ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_NODELAY,
-				(char *)&opt, sizeof(opt));
-		if (ret < 0) {
-			pr_err("kernel_setsockopt() for TCP_NODELAY"
-				" failed: %d\n", ret);
-			goto fail;
-		}
-	}
-
+	if (np->np_network_transport == ISCSI_TCP)
+		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
 
+	opt = 1;
 	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
 			(char *)&opt, sizeof(opt));
 	if (ret < 0) {
diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
index 28268ed461b82..ad8fb53b36827 100644
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -3929,14 +3929,8 @@ generic_ip_connect(struct TCP_Server_Info *server)
 			socket->sk->sk_rcvbuf = 140 * 1024;
 	}
 
-	if (server->tcp_nodelay) {
-		int val = 1;
-		rc = kernel_setsockopt(socket, SOL_TCP, TCP_NODELAY,
-				(char *)&val, sizeof(val));
-		if (rc)
-			cifs_dbg(FYI, "set TCP_NODELAY socket option error %d\n",
-				 rc);
-	}
+	if (server->tcp_nodelay)
+		tcp_sock_set_nodelay(socket->sk);
 
 	cifs_dbg(FYI, "sndbuf %d rcvbuf %d rcvtimeo 0x%lx\n",
 		 socket->sk->sk_sndbuf,
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 45c37f572c9d2..749f9bc548d2a 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -1128,7 +1128,6 @@ static void tcp_connect_to_sock(struct connection *con)
 	struct sockaddr_storage saddr, src_addr;
 	int addr_len;
 	struct socket *sock = NULL;
-	int one = 1;
 	int result;
 
 	if (con->nodeid == 0) {
@@ -1177,8 +1176,7 @@ static void tcp_connect_to_sock(struct connection *con)
 	log_print("connecting to %d", con->nodeid);
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
 				   O_NONBLOCK);
@@ -1220,7 +1218,6 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 {
 	struct socket *sock = NULL;
 	int result = 0;
-	int one = 1;
 	int addr_len;
 
 	if (dlm_local_addr[0]->ss_family == AF_INET)
@@ -1237,8 +1234,7 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
 	}
 
 	/* Turn off Nagle's algorithm */
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
-			  sizeof(one));
+	tcp_sock_set_nodelay(sock->sk);
 
 	sock_set_reuseaddr(sock->sk);
 
diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 2c512b40a940e..4c70fe9d19ab2 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_nodelay(struct socket *sock)
-{
-	int val = 1;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-				    (void *)&val, sizeof(val));
-}
-
 static int o2net_set_usertimeout(struct socket *sock)
 {
 	int user_timeout = O2NET_TCP_USER_TIMEOUT;
@@ -1636,11 +1628,7 @@ static void o2net_start_connect(struct work_struct *work)
 		goto out;
 	}
 
-	ret = o2net_set_nodelay(sc->sc_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(sc->sc_sock->sk);
 
 	ret = o2net_set_usertimeout(sock);
 	if (ret) {
@@ -1832,11 +1820,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	*more = 1;
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
-	ret = o2net_set_nodelay(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_nodelay(new_sock->sk);
 
 	ret = o2net_set_usertimeout(new_sock);
 	if (ret) {
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 889eeb2256c2d..9e42c7fe50a8b 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,5 +498,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+void tcp_sock_set_nodelay(struct sock *sk);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index f8ca5edc5f2c9..27d6ab11f9ee8 100644
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -490,15 +490,8 @@ static int ceph_tcp_connect(struct ceph_connection *con)
 		return ret;
 	}
 
-	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY)) {
-		int optval = 1;
-
-		ret = kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
-					(char *)&optval, sizeof(optval));
-		if (ret)
-			pr_err("kernel_setsockopt(TCP_NODELAY) failed: %d",
-			       ret);
-	}
+	if (ceph_test_opt(from_msgr(con->msgr), TCP_NODELAY))
+		tcp_sock_set_nodelay(sock->sk);
 
 	con->sock = sock;
 	return 0;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index e6cf702e16d66..a65f293a19fac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2832,6 +2832,30 @@ void tcp_sock_set_cork(struct sock *sk, bool on)
 }
 EXPORT_SYMBOL(tcp_sock_set_cork);
 
+/* TCP_NODELAY is weaker than TCP_CORK, so that this option on corked socket is
+ * remembered, but it is not activated until cork is cleared.
+ *
+ * However, when TCP_NODELAY is set we make an explicit push, which overrides
+ * even TCP_CORK for currently queued segments.
+ */
+static void __tcp_sock_set_nodelay(struct sock *sk, bool on)
+{
+	if (on) {
+		tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
+		tcp_push_pending_frames(sk);
+	} else {
+		tcp_sk(sk)->nonagle &= ~TCP_NAGLE_OFF;
+	}
+}
+
+void tcp_sock_set_nodelay(struct sock *sk)
+{
+	lock_sock(sk);
+	__tcp_sock_set_nodelay(sk, true);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_nodelay);
+
 /*
  *	Socket option code for TCP.
  */
@@ -2929,20 +2953,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_NODELAY:
-		if (val) {
-			/* TCP_NODELAY is weaker than TCP_CORK, so that
-			 * this option on corked socket is remembered, but
-			 * it is not activated until cork is cleared.
-			 *
-			 * However, when TCP_NODELAY is set we make
-			 * an explicit push, which overrides even TCP_CORK
-			 * for currently queued segments.
-			 */
-			tp->nonagle |= TCP_NAGLE_OFF|TCP_NAGLE_PUSH;
-			tcp_push_pending_frames(sk);
-		} else {
-			tp->nonagle &= ~TCP_NAGLE_OFF;
-		}
+		__tcp_sock_set_nodelay(sk, val);
 		break;
 
 	case TCP_THIN_LINEAR_TIMEOUTS:
diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index 46782fac4c162..43db0eca911fa 100644
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -89,15 +89,6 @@ static struct ctl_table rds_tcp_sysctl_table[] = {
 	{ }
 };
 
-/* doing it this way avoids calling tcp_sk() */
-void rds_tcp_nonagle(struct socket *sock)
-{
-	int val = 1;
-
-	kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (void *)&val,
-			      sizeof(val));
-}
-
 u32 rds_tcp_write_seq(struct rds_tcp_connection *tc)
 {
 	/* seq# of the last byte of data in tcp send buffer */
@@ -502,7 +493,7 @@ void rds_tcp_tune(struct socket *sock)
 	struct net *net = sock_net(sk);
 	struct rds_tcp_net *rtn = net_generic(net, rds_tcp_netid);
 
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 	lock_sock(sk);
 	if (rtn->sndbuf_size > 0) {
 		sk->sk_sndbuf = rtn->sndbuf_size;
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index d640e210b97b6..f6d75d8cb167a 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -50,7 +50,6 @@ struct rds_tcp_statistics {
 
 /* tcp.c */
 void rds_tcp_tune(struct socket *sock);
-void rds_tcp_nonagle(struct socket *sock);
 void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_reset_callbacks(struct socket *sock, struct rds_conn_path *cp);
 void rds_tcp_restore_callbacks(struct socket *sock,
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index d8bd132769594..6f90ea077adcd 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -288,7 +288,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
 	}
 
 	sock->sk->sk_reuse = SK_CAN_REUSE;
-	rds_tcp_nonagle(sock);
+	tcp_sock_set_nodelay(sock->sk);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = sock->sk->sk_data_ready;
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 12/28] tcp: add tcp_sock_set_quickack
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_QUICKACK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      |  7 ------
 drivers/block/drbd/drbd_receiver.c |  5 ++--
 include/linux/tcp.h                |  1 +
 net/ipv4/tcp.c                     | 39 ++++++++++++++++++++----------
 4 files changed, 29 insertions(+), 23 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e24bba87c8e02..14345a87c7cc5 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_quickack(struct socket *sock)
-{
-	int val = 2;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
-			(char*)&val, sizeof(val));
-}
-
 /* sets the number of 512 byte sectors of our virtual device */
 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 20a5e94494acd..3a3f2b6a821f3 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1223,7 +1223,7 @@ static int drbd_recv_header_maybe_unplug(struct drbd_connection *connection, str
 		 * quickly as possible, and let remote TCP know what we have
 		 * received so far. */
 		if (err == -EAGAIN) {
-			drbd_tcp_quickack(connection->data.socket);
+			tcp_sock_set_quickack(connection->data.socket->sk, 2);
 			drbd_unplug_all_devices(connection);
 		}
 		if (err > 0) {
@@ -4959,8 +4959,7 @@ static int receive_UnplugRemote(struct drbd_connection *connection, struct packe
 {
 	/* Make sure we've acked all the TCP data associated
 	 * with the data requests being unplugged */
-	drbd_tcp_quickack(connection->data.socket);
-
+	tcp_sock_set_quickack(connection->data.socket->sk, 2);
 	return 0;
 }
 
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 9e42c7fe50a8b..2eaf8320b9db0 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,5 +499,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
+void tcp_sock_set_quickack(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a65f293a19fac..27b5e7a4e2ef9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2856,6 +2856,31 @@ void tcp_sock_set_nodelay(struct sock *sk)
 }
 EXPORT_SYMBOL(tcp_sock_set_nodelay);
 
+static void __tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	if (!val) {
+		inet_csk_enter_pingpong_mode(sk);
+		return;
+	}
+
+	inet_csk_exit_pingpong_mode(sk);
+	if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
+	    inet_csk_ack_scheduled(sk)) {
+		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED;
+		tcp_cleanup_rbuf(sk, 1);
+		if (!(val & 1))
+			inet_csk_enter_pingpong_mode(sk);
+	}
+}
+
+void tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__tcp_sock_set_quickack(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_quickack);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3096,19 +3121,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_QUICKACK:
-		if (!val) {
-			inet_csk_enter_pingpong_mode(sk);
-		} else {
-			inet_csk_exit_pingpong_mode(sk);
-			if ((1 << sk->sk_state) &
-			    (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
-			    inet_csk_ack_scheduled(sk)) {
-				icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
-				tcp_cleanup_rbuf(sk, 1);
-				if (!(val & 1))
-					inet_csk_enter_pingpong_mode(sk);
-			}
-		}
+		__tcp_sock_set_quickack(sk, val);
 		break;
 
 #ifdef CONFIG_TCP_MD5SIG
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 12/28] tcp: add tcp_sock_set_quickack
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_QUICKACK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      |  7 ------
 drivers/block/drbd/drbd_receiver.c |  5 ++--
 include/linux/tcp.h                |  1 +
 net/ipv4/tcp.c                     | 39 ++++++++++++++++++++----------
 4 files changed, 29 insertions(+), 23 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e24bba87c8e02..14345a87c7cc5 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_quickack(struct socket *sock)
-{
-	int val = 2;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
-			(char*)&val, sizeof(val));
-}
-
 /* sets the number of 512 byte sectors of our virtual device */
 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 20a5e94494acd..3a3f2b6a821f3 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1223,7 +1223,7 @@ static int drbd_recv_header_maybe_unplug(struct drbd_connection *connection, str
 		 * quickly as possible, and let remote TCP know what we have
 		 * received so far. */
 		if (err = -EAGAIN) {
-			drbd_tcp_quickack(connection->data.socket);
+			tcp_sock_set_quickack(connection->data.socket->sk, 2);
 			drbd_unplug_all_devices(connection);
 		}
 		if (err > 0) {
@@ -4959,8 +4959,7 @@ static int receive_UnplugRemote(struct drbd_connection *connection, struct packe
 {
 	/* Make sure we've acked all the TCP data associated
 	 * with the data requests being unplugged */
-	drbd_tcp_quickack(connection->data.socket);
-
+	tcp_sock_set_quickack(connection->data.socket->sk, 2);
 	return 0;
 }
 
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 9e42c7fe50a8b..2eaf8320b9db0 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,5 +499,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
+void tcp_sock_set_quickack(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a65f293a19fac..27b5e7a4e2ef9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2856,6 +2856,31 @@ void tcp_sock_set_nodelay(struct sock *sk)
 }
 EXPORT_SYMBOL(tcp_sock_set_nodelay);
 
+static void __tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	if (!val) {
+		inet_csk_enter_pingpong_mode(sk);
+		return;
+	}
+
+	inet_csk_exit_pingpong_mode(sk);
+	if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
+	    inet_csk_ack_scheduled(sk)) {
+		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED;
+		tcp_cleanup_rbuf(sk, 1);
+		if (!(val & 1))
+			inet_csk_enter_pingpong_mode(sk);
+	}
+}
+
+void tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__tcp_sock_set_quickack(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_quickack);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3096,19 +3121,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_QUICKACK:
-		if (!val) {
-			inet_csk_enter_pingpong_mode(sk);
-		} else {
-			inet_csk_exit_pingpong_mode(sk);
-			if ((1 << sk->sk_state) &
-			    (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
-			    inet_csk_ack_scheduled(sk)) {
-				icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
-				tcp_cleanup_rbuf(sk, 1);
-				if (!(val & 1))
-					inet_csk_enter_pingpong_mode(sk);
-			}
-		}
+		__tcp_sock_set_quickack(sk, val);
 		break;
 
 #ifdef CONFIG_TCP_MD5SIG
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 12/28] tcp: add tcp_sock_set_quickack
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_QUICKACK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      |  7 ------
 drivers/block/drbd/drbd_receiver.c |  5 ++--
 include/linux/tcp.h                |  1 +
 net/ipv4/tcp.c                     | 39 ++++++++++++++++++++----------
 4 files changed, 29 insertions(+), 23 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e24bba87c8e02..14345a87c7cc5 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_quickack(struct socket *sock)
-{
-	int val = 2;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
-			(char*)&val, sizeof(val));
-}
-
 /* sets the number of 512 byte sectors of our virtual device */
 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 20a5e94494acd..3a3f2b6a821f3 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1223,7 +1223,7 @@ static int drbd_recv_header_maybe_unplug(struct drbd_connection *connection, str
 		 * quickly as possible, and let remote TCP know what we have
 		 * received so far. */
 		if (err == -EAGAIN) {
-			drbd_tcp_quickack(connection->data.socket);
+			tcp_sock_set_quickack(connection->data.socket->sk, 2);
 			drbd_unplug_all_devices(connection);
 		}
 		if (err > 0) {
@@ -4959,8 +4959,7 @@ static int receive_UnplugRemote(struct drbd_connection *connection, struct packe
 {
 	/* Make sure we've acked all the TCP data associated
 	 * with the data requests being unplugged */
-	drbd_tcp_quickack(connection->data.socket);
-
+	tcp_sock_set_quickack(connection->data.socket->sk, 2);
 	return 0;
 }
 
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 9e42c7fe50a8b..2eaf8320b9db0 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,5 +499,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
+void tcp_sock_set_quickack(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a65f293a19fac..27b5e7a4e2ef9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2856,6 +2856,31 @@ void tcp_sock_set_nodelay(struct sock *sk)
 }
 EXPORT_SYMBOL(tcp_sock_set_nodelay);
 
+static void __tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	if (!val) {
+		inet_csk_enter_pingpong_mode(sk);
+		return;
+	}
+
+	inet_csk_exit_pingpong_mode(sk);
+	if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
+	    inet_csk_ack_scheduled(sk)) {
+		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED;
+		tcp_cleanup_rbuf(sk, 1);
+		if (!(val & 1))
+			inet_csk_enter_pingpong_mode(sk);
+	}
+}
+
+void tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__tcp_sock_set_quickack(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_quickack);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3096,19 +3121,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_QUICKACK:
-		if (!val) {
-			inet_csk_enter_pingpong_mode(sk);
-		} else {
-			inet_csk_exit_pingpong_mode(sk);
-			if ((1 << sk->sk_state) &
-			    (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
-			    inet_csk_ack_scheduled(sk)) {
-				icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
-				tcp_cleanup_rbuf(sk, 1);
-				if (!(val & 1))
-					inet_csk_enter_pingpong_mode(sk);
-			}
-		}
+		__tcp_sock_set_quickack(sk, val);
 		break;
 
 #ifdef CONFIG_TCP_MD5SIG
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 12/28] tcp: add tcp_sock_set_quickack
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_QUICKACK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      |  7 ------
 drivers/block/drbd/drbd_receiver.c |  5 ++--
 include/linux/tcp.h                |  1 +
 net/ipv4/tcp.c                     | 39 ++++++++++++++++++++----------
 4 files changed, 29 insertions(+), 23 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e24bba87c8e02..14345a87c7cc5 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_quickack(struct socket *sock)
-{
-	int val = 2;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
-			(char*)&val, sizeof(val));
-}
-
 /* sets the number of 512 byte sectors of our virtual device */
 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 20a5e94494acd..3a3f2b6a821f3 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1223,7 +1223,7 @@ static int drbd_recv_header_maybe_unplug(struct drbd_connection *connection, str
 		 * quickly as possible, and let remote TCP know what we have
 		 * received so far. */
 		if (err == -EAGAIN) {
-			drbd_tcp_quickack(connection->data.socket);
+			tcp_sock_set_quickack(connection->data.socket->sk, 2);
 			drbd_unplug_all_devices(connection);
 		}
 		if (err > 0) {
@@ -4959,8 +4959,7 @@ static int receive_UnplugRemote(struct drbd_connection *connection, struct packe
 {
 	/* Make sure we've acked all the TCP data associated
 	 * with the data requests being unplugged */
-	drbd_tcp_quickack(connection->data.socket);
-
+	tcp_sock_set_quickack(connection->data.socket->sk, 2);
 	return 0;
 }
 
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 9e42c7fe50a8b..2eaf8320b9db0 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,5 +499,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
+void tcp_sock_set_quickack(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a65f293a19fac..27b5e7a4e2ef9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2856,6 +2856,31 @@ void tcp_sock_set_nodelay(struct sock *sk)
 }
 EXPORT_SYMBOL(tcp_sock_set_nodelay);
 
+static void __tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	if (!val) {
+		inet_csk_enter_pingpong_mode(sk);
+		return;
+	}
+
+	inet_csk_exit_pingpong_mode(sk);
+	if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
+	    inet_csk_ack_scheduled(sk)) {
+		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED;
+		tcp_cleanup_rbuf(sk, 1);
+		if (!(val & 1))
+			inet_csk_enter_pingpong_mode(sk);
+	}
+}
+
+void tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__tcp_sock_set_quickack(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_quickack);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3096,19 +3121,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_QUICKACK:
-		if (!val) {
-			inet_csk_enter_pingpong_mode(sk);
-		} else {
-			inet_csk_exit_pingpong_mode(sk);
-			if ((1 << sk->sk_state) &
-			    (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
-			    inet_csk_ack_scheduled(sk)) {
-				icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
-				tcp_cleanup_rbuf(sk, 1);
-				if (!(val & 1))
-					inet_csk_enter_pingpong_mode(sk);
-			}
-		}
+		__tcp_sock_set_quickack(sk, val);
 		break;
 
 #ifdef CONFIG_TCP_MD5SIG
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 12/28] tcp: add tcp_sock_set_quickack
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_QUICKACK sockopt from kernel space
without going through a fake uaccess.  Cleanup the callers to avoid
pointless wrappers now that this is a simple function call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/drbd/drbd_int.h      |  7 ------
 drivers/block/drbd/drbd_receiver.c |  5 ++--
 include/linux/tcp.h                |  1 +
 net/ipv4/tcp.c                     | 39 ++++++++++++++++++++----------
 4 files changed, 29 insertions(+), 23 deletions(-)

diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index e24bba87c8e02..14345a87c7cc5 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1570,13 +1570,6 @@ extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
 extern int drbd_connected(struct drbd_peer_device *);
 
-static inline void drbd_tcp_quickack(struct socket *sock)
-{
-	int val = 2;
-	(void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
-			(char*)&val, sizeof(val));
-}
-
 /* sets the number of 512 byte sectors of our virtual device */
 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
 
diff --git a/drivers/block/drbd/drbd_receiver.c b/drivers/block/drbd/drbd_receiver.c
index 20a5e94494acd..3a3f2b6a821f3 100644
--- a/drivers/block/drbd/drbd_receiver.c
+++ b/drivers/block/drbd/drbd_receiver.c
@@ -1223,7 +1223,7 @@ static int drbd_recv_header_maybe_unplug(struct drbd_connection *connection, str
 		 * quickly as possible, and let remote TCP know what we have
 		 * received so far. */
 		if (err == -EAGAIN) {
-			drbd_tcp_quickack(connection->data.socket);
+			tcp_sock_set_quickack(connection->data.socket->sk, 2);
 			drbd_unplug_all_devices(connection);
 		}
 		if (err > 0) {
@@ -4959,8 +4959,7 @@ static int receive_UnplugRemote(struct drbd_connection *connection, struct packe
 {
 	/* Make sure we've acked all the TCP data associated
 	 * with the data requests being unplugged */
-	drbd_tcp_quickack(connection->data.socket);
-
+	tcp_sock_set_quickack(connection->data.socket->sk, 2);
 	return 0;
 }
 
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 9e42c7fe50a8b..2eaf8320b9db0 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,5 +499,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
+void tcp_sock_set_quickack(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index a65f293a19fac..27b5e7a4e2ef9 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2856,6 +2856,31 @@ void tcp_sock_set_nodelay(struct sock *sk)
 }
 EXPORT_SYMBOL(tcp_sock_set_nodelay);
 
+static void __tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	if (!val) {
+		inet_csk_enter_pingpong_mode(sk);
+		return;
+	}
+
+	inet_csk_exit_pingpong_mode(sk);
+	if ((1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
+	    inet_csk_ack_scheduled(sk)) {
+		inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_PUSHED;
+		tcp_cleanup_rbuf(sk, 1);
+		if (!(val & 1))
+			inet_csk_enter_pingpong_mode(sk);
+	}
+}
+
+void tcp_sock_set_quickack(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__tcp_sock_set_quickack(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_quickack);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3096,19 +3121,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_QUICKACK:
-		if (!val) {
-			inet_csk_enter_pingpong_mode(sk);
-		} else {
-			inet_csk_exit_pingpong_mode(sk);
-			if ((1 << sk->sk_state) &
-			    (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT) &&
-			    inet_csk_ack_scheduled(sk)) {
-				icsk->icsk_ack.pending |= ICSK_ACK_PUSHED;
-				tcp_cleanup_rbuf(sk, 1);
-				if (!(val & 1))
-					inet_csk_enter_pingpong_mode(sk);
-			}
-		}
+		__tcp_sock_set_quickack(sk, val);
 		break;
 
 #ifdef CONFIG_TCP_MD5SIG
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 13/28] tcp: add tcp_sock_set_syncnt
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the TCP_SYNCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c |  9 +--------
 include/linux/tcp.h     |  1 +
 net/ipv4/tcp.c          | 12 ++++++++++++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4e4a750ecdb97..2872584f52f63 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1336,14 +1336,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
 	tcp_sock_set_nodelay(queue->sock->sk);
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2eaf8320b9db0..6aa4ae5ebf3d5 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -500,5 +500,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
+int tcp_sock_set_syncnt(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 27b5e7a4e2ef9..d2c67ae1da07a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2881,6 +2881,18 @@ void tcp_sock_set_quickack(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_quickack);
 
+int tcp_sock_set_syncnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_SYNCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	inet_csk(sk)->icsk_syn_retries = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_syncnt);
+
 /*
  *	Socket option code for TCP.
  */
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 13/28] tcp: add tcp_sock_set_syncnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the TCP_SYNCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c |  9 +--------
 include/linux/tcp.h     |  1 +
 net/ipv4/tcp.c          | 12 ++++++++++++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4e4a750ecdb97..2872584f52f63 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1336,14 +1336,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
 	tcp_sock_set_nodelay(queue->sock->sk);
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2eaf8320b9db0..6aa4ae5ebf3d5 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -500,5 +500,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
+int tcp_sock_set_syncnt(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 27b5e7a4e2ef9..d2c67ae1da07a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2881,6 +2881,18 @@ void tcp_sock_set_quickack(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_quickack);
 
+int tcp_sock_set_syncnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_SYNCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	inet_csk(sk)->icsk_syn_retries = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_syncnt);
+
 /*
  *	Socket option code for TCP.
  */
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 13/28] tcp: add tcp_sock_set_syncnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, linux-kernel, linux-nvme, Jon Maloy,
	linux-rdma, Eric Dumazet, target-devel, tipc-discussion,
	linux-nfs, Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs,
	ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_SYNCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c |  9 +--------
 include/linux/tcp.h     |  1 +
 net/ipv4/tcp.c          | 12 ++++++++++++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4e4a750ecdb97..2872584f52f63 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1336,14 +1336,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
 	tcp_sock_set_nodelay(queue->sock->sk);
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2eaf8320b9db0..6aa4ae5ebf3d5 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -500,5 +500,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
+int tcp_sock_set_syncnt(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 27b5e7a4e2ef9..d2c67ae1da07a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2881,6 +2881,18 @@ void tcp_sock_set_quickack(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_quickack);
 
+int tcp_sock_set_syncnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_SYNCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	inet_csk(sk)->icsk_syn_retries = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_syncnt);
+
 /*
  *	Socket option code for TCP.
  */
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 13/28] tcp: add tcp_sock_set_syncnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the TCP_SYNCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c |  9 +--------
 include/linux/tcp.h     |  1 +
 net/ipv4/tcp.c          | 12 ++++++++++++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4e4a750ecdb97..2872584f52f63 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1336,14 +1336,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
 	tcp_sock_set_nodelay(queue->sock->sk);
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2eaf8320b9db0..6aa4ae5ebf3d5 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -500,5 +500,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
+int tcp_sock_set_syncnt(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 27b5e7a4e2ef9..d2c67ae1da07a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2881,6 +2881,18 @@ void tcp_sock_set_quickack(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_quickack);
 
+int tcp_sock_set_syncnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_SYNCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	inet_csk(sk)->icsk_syn_retries = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_syncnt);
+
 /*
  *	Socket option code for TCP.
  */
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 13/28] tcp: add tcp_sock_set_syncnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_SYNCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c |  9 +--------
 include/linux/tcp.h     |  1 +
 net/ipv4/tcp.c          | 12 ++++++++++++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 4e4a750ecdb97..2872584f52f63 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1336,14 +1336,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 	}
 
 	/* Single syn retry */
-	opt = 1;
-	ret = kernel_setsockopt(queue->sock, IPPROTO_TCP, TCP_SYNCNT,
-			(char *)&opt, sizeof(opt));
-	if (ret) {
-		dev_err(nctrl->device,
-			"failed to set TCP_SYNCNT sock opt %d\n", ret);
-		goto err_sock;
-	}
+	tcp_sock_set_syncnt(queue->sock->sk, 1);
 
 	/* Set TCP no delay */
 	tcp_sock_set_nodelay(queue->sock->sk);
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2eaf8320b9db0..6aa4ae5ebf3d5 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -500,5 +500,6 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
+int tcp_sock_set_syncnt(struct sock *sk, int val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 27b5e7a4e2ef9..d2c67ae1da07a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2881,6 +2881,18 @@ void tcp_sock_set_quickack(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_quickack);
 
+int tcp_sock_set_syncnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_SYNCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	inet_csk(sk)->icsk_syn_retries = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_syncnt);
+
 /*
  *	Socket option code for TCP.
  */
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 14/28] tcp: add tcp_sock_set_user_timeout
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_USER_TIMEOUT sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/ocfs2/cluster/tcp.c | 22 ++--------------------
 include/linux/tcp.h    |  1 +
 net/ipv4/tcp.c         |  8 ++++++++
 net/sunrpc/xprtsock.c  |  3 +--
 4 files changed, 12 insertions(+), 22 deletions(-)

diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c
index 4c70fe9d19ab2..79a2317194600 100644
--- a/fs/ocfs2/cluster/tcp.c
+++ b/fs/ocfs2/cluster/tcp.c
@@ -1441,14 +1441,6 @@ static void o2net_rx_until_empty(struct work_struct *work)
 	sc_put(sc);
 }
 
-static int o2net_set_usertimeout(struct socket *sock)
-{
-	int user_timeout = O2NET_TCP_USER_TIMEOUT;
-
-	return kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-				(void *)&user_timeout, sizeof(user_timeout));
-}
-
 static void o2net_initialize_handshake(void)
 {
 	o2net_hand->o2hb_heartbeat_timeout_ms = cpu_to_be32(
@@ -1629,12 +1621,7 @@ static void o2net_start_connect(struct work_struct *work)
 	}
 
 	tcp_sock_set_nodelay(sc->sc_sock->sk);
-
-	ret = o2net_set_usertimeout(sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	o2net_register_callbacks(sc->sc_sock->sk, sc);
 
@@ -1821,12 +1808,7 @@ static int o2net_accept_one(struct socket *sock, int *more)
 	new_sock->sk->sk_allocation = GFP_ATOMIC;
 
 	tcp_sock_set_nodelay(new_sock->sk);
-
-	ret = o2net_set_usertimeout(new_sock);
-	if (ret) {
-		mlog(ML_ERROR, "set TCP_USER_TIMEOUT failed with %d\n", ret);
-		goto out;
-	}
+	tcp_sock_set_user_timeout(new_sock->sk, O2NET_TCP_USER_TIMEOUT);
 
 	ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin, 1);
 	if (ret < 0)
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 6aa4ae5ebf3d5..de682143efe4d 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -501,5 +501,6 @@ void tcp_sock_set_cork(struct sock *sk, bool on);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val);
 
 #endif	/* _LINUX_TCP_H */
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d2c67ae1da07a..0004bd9ae7b0a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2893,6 +2893,14 @@ int tcp_sock_set_syncnt(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_syncnt);
 
+void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
+{
+	lock_sock(sk);
+	inet_csk(sk)->icsk_user_timeout = val;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(tcp_sock_set_user_timeout);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 399848c2bcb29..231fd6162f68d 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2115,8 +2115,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 			(char *)&keepcnt, sizeof(keepcnt));
 
 	/* TCP user timeout (see RFC5482) */
-	kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
-			(char *)&timeo, sizeof(timeo));
+	tcp_sock_set_user_timeout(sock->sk, timeo);
 }
 
 static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt,
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 15/28] tcp: add tcp_sock_set_keepidle
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_KEEP_IDLE sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 49 ++++++++++++++++++++++++++++++-------------
 net/rds/tcp_listen.c  |  5 +----
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 37 insertions(+), 21 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index de682143efe4d..5724dd84a85ed 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepidle(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 0004bd9ae7b0a..bdf0ff9333514 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2901,6 +2901,39 @@ void tcp_sock_set_user_timeout(struct sock *sk, u32 val)
 }
 EXPORT_SYMBOL(tcp_sock_set_user_timeout);
 
+static int __tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	if (val < 1 || val > MAX_TCP_KEEPIDLE)
+		return -EINVAL;
+
+	tp->keepalive_time = val * HZ;
+	if (sock_flag(sk, SOCK_KEEPOPEN) &&
+	    !((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_LISTEN))) {
+		u32 elapsed = keepalive_time_elapsed(tp);
+
+		if (tp->keepalive_time > elapsed)
+			elapsed = tp->keepalive_time - elapsed;
+		else
+			elapsed = 0;
+		inet_csk_reset_keepalive_timer(sk, elapsed);
+	}
+
+	return 0;
+}
+
+int tcp_sock_set_keepidle(struct sock *sk, int val)
+{
+	int err;
+
+	lock_sock(sk);
+	err = __tcp_sock_set_keepidle(sk, val);
+	release_sock(sk);
+	return err;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepidle);
+
 /*
  *	Socket option code for TCP.
  */
@@ -3070,21 +3103,7 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
 		break;
 
 	case TCP_KEEPIDLE:
-		if (val < 1 || val > MAX_TCP_KEEPIDLE)
-			err = -EINVAL;
-		else {
-			tp->keepalive_time = val * HZ;
-			if (sock_flag(sk, SOCK_KEEPOPEN) &&
-			    !((1 << sk->sk_state) &
-			      (TCPF_CLOSE | TCPF_LISTEN))) {
-				u32 elapsed = keepalive_time_elapsed(tp);
-				if (tp->keepalive_time > elapsed)
-					elapsed = tp->keepalive_time - elapsed;
-				else
-					elapsed = 0;
-				inet_csk_reset_keepalive_timer(sk, elapsed);
-			}
-		}
+		err = __tcp_sock_set_keepidle(sk, val);
 		break;
 	case TCP_KEEPINTVL:
 		if (val < 1 || val > MAX_TCP_KEEPINTVL)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 6f90ea077adcd..79f9adc008114 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -52,10 +52,7 @@ int rds_tcp_keepalive(struct socket *sock)
 	if (ret < 0)
 		goto bail;
 
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPIDLE,
-				(char *)&keepidle, sizeof(keepidle));
-	if (ret < 0)
-		goto bail;
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 231fd6162f68d..473290f7c5c0a 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2107,8 +2107,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepidle(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
 			(char *)&keepidle, sizeof(keepidle));
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 16/28] tcp: add tcp_sock_set_keepintvl
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_KEEPINTVL sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp_listen.c  |  4 +---
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 5724dd84a85ed..1f9bada00faab 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,6 +499,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
+int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bdf0ff9333514..7eb083e09786a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2934,6 +2934,18 @@ int tcp_sock_set_keepidle(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepidle);
 
+int tcp_sock_set_keepintvl(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPINTVL)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_intvl = val * HZ;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepintvl);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 79f9adc008114..9ad555c48d15d 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -53,12 +53,10 @@ int rds_tcp_keepalive(struct socket *sock)
 		goto bail;
 
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
-				(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 bail:
 	return ret;
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 473290f7c5c0a..5ca64e12af0c5 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2108,8 +2108,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
 			(char *)&keepcnt, sizeof(keepcnt));
 
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 16/28] tcp: add tcp_sock_set_keepintvl
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_KEEPINTVL sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp_listen.c  |  4 +---
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 5724dd84a85ed..1f9bada00faab 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,6 +499,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
+int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bdf0ff9333514..7eb083e09786a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2934,6 +2934,18 @@ int tcp_sock_set_keepidle(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepidle);
 
+int tcp_sock_set_keepintvl(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPINTVL)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_intvl = val * HZ;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepintvl);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 79f9adc008114..9ad555c48d15d 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -53,12 +53,10 @@ int rds_tcp_keepalive(struct socket *sock)
 		goto bail;
 
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
-				(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 bail:
 	return ret;
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 473290f7c5c0a..5ca64e12af0c5 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2108,8 +2108,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
 			(char *)&keepcnt, sizeof(keepcnt));
 
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 16/28] tcp: add tcp_sock_set_keepintvl
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_KEEPINTVL sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp_listen.c  |  4 +---
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 5724dd84a85ed..1f9bada00faab 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,6 +499,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
+int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bdf0ff9333514..7eb083e09786a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2934,6 +2934,18 @@ int tcp_sock_set_keepidle(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepidle);
 
+int tcp_sock_set_keepintvl(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPINTVL)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_intvl = val * HZ;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepintvl);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 79f9adc008114..9ad555c48d15d 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -53,12 +53,10 @@ int rds_tcp_keepalive(struct socket *sock)
 		goto bail;
 
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
-				(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 bail:
 	return ret;
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 473290f7c5c0a..5ca64e12af0c5 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2108,8 +2108,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
 			(char *)&keepcnt, sizeof(keepcnt));
 
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 16/28] tcp: add tcp_sock_set_keepintvl
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_KEEPINTVL sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp_listen.c  |  4 +---
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 5724dd84a85ed..1f9bada00faab 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,6 +499,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
+int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bdf0ff9333514..7eb083e09786a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2934,6 +2934,18 @@ int tcp_sock_set_keepidle(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepidle);
 
+int tcp_sock_set_keepintvl(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPINTVL)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_intvl = val * HZ;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepintvl);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 79f9adc008114..9ad555c48d15d 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -53,12 +53,10 @@ int rds_tcp_keepalive(struct socket *sock)
 		goto bail;
 
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
-				(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 bail:
 	return ret;
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 473290f7c5c0a..5ca64e12af0c5 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2108,8 +2108,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
 			(char *)&keepcnt, sizeof(keepcnt));
 
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 16/28] tcp: add tcp_sock_set_keepintvl
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_KEEPINTVL sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp_listen.c  |  4 +---
 net/sunrpc/xprtsock.c |  3 +--
 4 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 5724dd84a85ed..1f9bada00faab 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -499,6 +499,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
+int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
 void tcp_sock_set_quickack(struct sock *sk, int val);
 int tcp_sock_set_syncnt(struct sock *sk, int val);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bdf0ff9333514..7eb083e09786a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2934,6 +2934,18 @@ int tcp_sock_set_keepidle(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepidle);
 
+int tcp_sock_set_keepintvl(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPINTVL)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_intvl = val * HZ;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepintvl);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 79f9adc008114..9ad555c48d15d 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -53,12 +53,10 @@ int rds_tcp_keepalive(struct socket *sock)
 		goto bail;
 
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPINTVL,
-				(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 bail:
 	return ret;
 }
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 473290f7c5c0a..5ca64e12af0c5 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2108,8 +2108,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	/* TCP Keepalive options */
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
-			(char *)&keepidle, sizeof(keepidle));
+	tcp_sock_set_keepintvl(sock->sk, keepidle);
 	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
 			(char *)&keepcnt, sizeof(keepcnt));
 
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 17/28] tcp: add tcp_sock_set_keepcnt
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the TCP_KEEPCNT sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/tcp.h   |  1 +
 net/ipv4/tcp.c        | 12 ++++++++++++
 net/rds/tcp.h         |  2 +-
 net/rds/tcp_listen.c  | 17 +++--------------
 net/sunrpc/xprtsock.c |  3 +--
 5 files changed, 18 insertions(+), 17 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 1f9bada00faab..9aac824c523cf 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -498,6 +498,7 @@ int tcp_skb_shift(struct sk_buff *to, struct sk_buff *from, int pcount,
 		  int shiftlen);
 
 void tcp_sock_set_cork(struct sock *sk, bool on);
+int tcp_sock_set_keepcnt(struct sock *sk, int val);
 int tcp_sock_set_keepidle(struct sock *sk, int val);
 int tcp_sock_set_keepintvl(struct sock *sk, int val);
 void tcp_sock_set_nodelay(struct sock *sk);
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 7eb083e09786a..15d47d5e79510 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2946,6 +2946,18 @@ int tcp_sock_set_keepintvl(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(tcp_sock_set_keepintvl);
 
+int tcp_sock_set_keepcnt(struct sock *sk, int val)
+{
+	if (val < 1 || val > MAX_TCP_KEEPCNT)
+		return -EINVAL;
+
+	lock_sock(sk);
+	tcp_sk(sk)->keepalive_probes = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(tcp_sock_set_keepcnt);
+
 /*
  *	Socket option code for TCP.
  */
diff --git a/net/rds/tcp.h b/net/rds/tcp.h
index f6d75d8cb167a..bad9cf49d5657 100644
--- a/net/rds/tcp.h
+++ b/net/rds/tcp.h
@@ -70,7 +70,7 @@ struct socket *rds_tcp_listen_init(struct net *net, bool isv6);
 void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor);
 void rds_tcp_listen_data_ready(struct sock *sk);
 int rds_tcp_accept_one(struct socket *sock);
-int rds_tcp_keepalive(struct socket *sock);
+void rds_tcp_keepalive(struct socket *sock);
 void *rds_tcp_listen_sock_def_readable(struct net *net);
 
 /* tcp_recv.c */
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 9ad555c48d15d..101cf14215a0b 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -38,27 +38,19 @@
 #include "rds.h"
 #include "tcp.h"
 
-int rds_tcp_keepalive(struct socket *sock)
+void rds_tcp_keepalive(struct socket *sock)
 {
 	/* values below based on xs_udp_default_timeout */
 	int keepidle = 5; /* send a probe 'keepidle' secs after last data */
 	int keepcnt = 5; /* number of unack'ed probes before declaring dead */
-	int ret = 0;
 
 	sock_set_keepalive(sock->sk);
-
-	ret = kernel_setsockopt(sock, IPPROTO_TCP, TCP_KEEPCNT,
-				(char *)&keepcnt, sizeof(keepcnt));
-	if (ret < 0)
-		goto bail;
-
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	/* KEEPINTVL is the interval between successive probes. We follow
 	 * the model in xs_tcp_finish_connecting() and re-use keepidle.
 	 */
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-bail:
-	return ret;
 }
 
 /* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
@@ -140,10 +132,7 @@ int rds_tcp_accept_one(struct socket *sock)
 	new_sock->ops = sock->ops;
 	__module_get(new_sock->ops->owner);
 
-	ret = rds_tcp_keepalive(new_sock);
-	if (ret < 0)
-		goto out;
-
+	rds_tcp_keepalive(new_sock);
 	rds_tcp_tune(new_sock);
 
 	inet = inet_sk(new_sock->sk);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 5ca64e12af0c5..0d3ec055bc12f 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2109,8 +2109,7 @@ static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt,
 	sock_set_keepalive(sock->sk);
 	tcp_sock_set_keepidle(sock->sk, keepidle);
 	tcp_sock_set_keepintvl(sock->sk, keepidle);
-	kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
-			(char *)&keepcnt, sizeof(keepcnt));
+	tcp_sock_set_keepcnt(sock->sk, keepcnt);
 
 	/* TCP user timeout (see RFC5482) */
 	tcp_sock_set_user_timeout(sock->sk, timeo);
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	Sagi Grimberg

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type == SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type == SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type = SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type = SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
Acked-by: Sagi Grimberg <sagi-NQWnxTmZq1alnMjI0IkVqw@public.gmane.org>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type == SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type == SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Sagi Grimberg,
	Hideaki YOSHIFUJI, netdev, linux-kernel, linux-nvme, Jon Maloy,
	linux-rdma, Eric Dumazet, target-devel, tipc-discussion,
	linux-nfs, Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs,
	ocfs2-devel, drbd-dev

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type == SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type == SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Sagi Grimberg

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type == SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type == SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 18/28] ipv4: add ip_sock_set_tos
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IP_TOS sockopt from kernel space without
going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Sagi Grimberg <sagi@grimberg.me>
---
 drivers/nvme/host/tcp.c   | 14 +++-----------
 drivers/nvme/target/tcp.c | 10 ++--------
 include/net/ip.h          |  2 ++
 net/ipv4/ip_sockglue.c    | 30 +++++++++++++++++++++---------
 4 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
index 2872584f52f63..4c972d8abf317 100644
--- a/drivers/nvme/host/tcp.c
+++ b/drivers/nvme/host/tcp.c
@@ -1313,7 +1313,7 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 {
 	struct nvme_tcp_ctrl *ctrl = to_tcp_ctrl(nctrl);
 	struct nvme_tcp_queue *queue = &ctrl->queues[qid];
-	int ret, opt, rcv_pdu_size;
+	int ret, rcv_pdu_size;
 
 	queue->ctrl = ctrl;
 	INIT_LIST_HEAD(&queue->send_list);
@@ -1352,16 +1352,8 @@ static int nvme_tcp_alloc_queue(struct nvme_ctrl *nctrl,
 		sock_set_priority(queue->sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (nctrl->opts->tos >= 0) {
-		opt = nctrl->opts->tos;
-		ret = kernel_setsockopt(queue->sock, SOL_IP, IP_TOS,
-				(char *)&opt, sizeof(opt));
-		if (ret) {
-			dev_err(nctrl->device,
-				"failed to set IP_TOS sock opt %d\n", ret);
-			goto err_sock;
-		}
-	}
+	if (nctrl->opts->tos >= 0)
+		ip_sock_set_tos(queue->sock->sk, nctrl->opts->tos);
 
 	queue->sock->sk->sk_allocation = GFP_ATOMIC;
 	nvme_tcp_set_queue_io_cpu(queue);
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 55bc4c3c0a74a..4546049a96b37 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -1452,14 +1452,8 @@ static int nvmet_tcp_set_queue_sock(struct nvmet_tcp_queue *queue)
 		sock_set_priority(sock->sk, so_priority);
 
 	/* Set socket type of service */
-	if (inet->rcv_tos > 0) {
-		int tos = inet->rcv_tos;
-
-		ret = kernel_setsockopt(sock, SOL_IP, IP_TOS,
-				(char *)&tos, sizeof(tos));
-		if (ret)
-			return ret;
-	}
+	if (inet->rcv_tos > 0)
+		ip_sock_set_tos(sock->sk, inet->rcv_tos);
 
 	write_lock_bh(&sock->sk->sk_callback_lock);
 	sock->sk->sk_user_data = queue;
diff --git a/include/net/ip.h b/include/net/ip.h
index 5b317c9f4470a..2fc52e26fa88b 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,4 +765,6 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_tos(struct sock *sk, int val);
+
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index f43d5f12aa86a..b43a29e11f4a5 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -560,6 +560,26 @@ int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 	return err;
 }
 
+static void __ip_sock_set_tos(struct sock *sk, int val)
+{
+	if (sk->sk_type == SOCK_STREAM) {
+		val &= ~INET_ECN_MASK;
+		val |= inet_sk(sk)->tos & INET_ECN_MASK;
+	}
+	if (inet_sk(sk)->tos != val) {
+		inet_sk(sk)->tos = val;
+		sk->sk_priority = rt_tos2priority(val);
+		sk_dst_reset(sk);
+	}
+}
+
+void ip_sock_set_tos(struct sock *sk, int val)
+{
+	lock_sock(sk);
+	__ip_sock_set_tos(sk, val);
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_tos);
 
 /*
  *	Socket option code for IP. This is the end of the line after any
@@ -823,15 +843,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
 			inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE;
 		break;
 	case IP_TOS:	/* This sets both TOS and Precedence */
-		if (sk->sk_type == SOCK_STREAM) {
-			val &= ~INET_ECN_MASK;
-			val |= inet->tos & INET_ECN_MASK;
-		}
-		if (inet->tos != val) {
-			inet->tos = val;
-			sk->sk_priority = rt_tos2priority(val);
-			sk_dst_reset(sk);
-		}
+		__ip_sock_set_tos(sk, val);
 		break;
 	case IP_TTL:
 		if (optlen < 1)
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport == ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport = ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport == ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport == ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev-cunTk1MwBs8qoQakbn7OcQ,
	linux-kernel-u79uwXL29TY76Z2rM5mHXA,
	linux-rdma-u79uwXL29TY76Z2rM5mHXA,
	linux-nvme-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	target-devel-u79uwXL29TY76Z2rM5mHXA,
	linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
	linux-cifs-u79uwXL29TY76Z2rM5mHXA,
	cluster-devel-H+wXaHxf7aLQT0dZR+AlfA,
	ocfs2-devel-N0ozoZBvEnrZJqsBc5GL+g,
	netdev-u79uwXL29TY76Z2rM5mHXA, ceph-devel-u79uwXL29TY76Z2rM5mHXA,
	rds-devel-N0ozoZBvEnrZJqsBc5GL+g,
	linux-nfs-u79uwXL29TY76Z2rM5mHXA,
	tipc-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport == ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 19/28] ipv4: add ip_sock_set_freebind
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IP_FREEBIND sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/target/iscsi/iscsi_target_login.c | 13 +++----------
 include/net/ip.h                          |  1 +
 net/ipv4/ip_sockglue.c                    |  8 ++++++++
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target_login.c b/drivers/target/iscsi/iscsi_target_login.c
index b561b07a869a0..85748e3388582 100644
--- a/drivers/target/iscsi/iscsi_target_login.c
+++ b/drivers/target/iscsi/iscsi_target_login.c
@@ -15,6 +15,7 @@
 #include <linux/sched/signal.h>
 #include <linux/idr.h>
 #include <linux/tcp.h>        /* TCP_NODELAY */
+#include <net/ip.h>
 #include <net/ipv6.h>         /* ipv6_addr_v4mapped() */
 #include <scsi/iscsi_proto.h>
 #include <target/target_core_base.h>
@@ -855,7 +856,7 @@ int iscsit_setup_np(
 	struct sockaddr_storage *sockaddr)
 {
 	struct socket *sock = NULL;
-	int backlog = ISCSIT_TCP_BACKLOG, ret, opt = 0, len;
+	int backlog = ISCSIT_TCP_BACKLOG, ret, len;
 
 	switch (np->np_network_transport) {
 	case ISCSI_TCP:
@@ -900,15 +901,7 @@ int iscsit_setup_np(
 	if (np->np_network_transport == ISCSI_TCP)
 		tcp_sock_set_nodelay(sock->sk);
 	sock_set_reuseaddr(sock->sk);
-
-	opt = 1;
-	ret = kernel_setsockopt(sock, IPPROTO_IP, IP_FREEBIND,
-			(char *)&opt, sizeof(opt));
-	if (ret < 0) {
-		pr_err("kernel_setsockopt() for IP_FREEBIND"
-			" failed\n");
-		goto fail;
-	}
+	ip_sock_set_freebind(sock->sk);
 
 	ret = kernel_bind(sock, (struct sockaddr *)&np->np_sockaddr, len);
 	if (ret < 0) {
diff --git a/include/net/ip.h b/include/net/ip.h
index 2fc52e26fa88b..5f5d8226b6abc 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -765,6 +765,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 	return likely(mtu >= IPV4_MIN_MTU);
 }
 
+void ip_sock_set_freebind(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index b43a29e11f4a5..767838d2030d8 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -581,6 +581,14 @@ void ip_sock_set_tos(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_tos);
 
+void ip_sock_set_freebind(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->freebind = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_freebind);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 20/28] ipv4: add ip_sock_set_recverr
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IP_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ip.h         | 1 +
 net/ipv4/ip_sockglue.c   | 8 ++++++++
 net/rxrpc/local_object.c | 8 +-------
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 5f5d8226b6abc..f063a491b9063 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 767838d2030d8..aca6b81da9bae 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -589,6 +589,14 @@ void ip_sock_set_freebind(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_freebind);
 
+void ip_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_recverr);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 5ea2bd01fdd59..4c0e8fe5ec1fb 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -171,13 +171,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		/* Fall through */
 	case AF_INET:
 		/* we want to receive ICMP errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
 		opt = IP_PMTUDISC_DO;
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 20/28] ipv4: add ip_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IP_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ip.h         | 1 +
 net/ipv4/ip_sockglue.c   | 8 ++++++++
 net/rxrpc/local_object.c | 8 +-------
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 5f5d8226b6abc..f063a491b9063 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 767838d2030d8..aca6b81da9bae 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -589,6 +589,14 @@ void ip_sock_set_freebind(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_freebind);
 
+void ip_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_recverr);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 5ea2bd01fdd59..4c0e8fe5ec1fb 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -171,13 +171,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		/* Fall through */
 	case AF_INET:
 		/* we want to receive ICMP errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
 		opt = IP_PMTUDISC_DO;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 20/28] ipv4: add ip_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	drbd-dev

Add a helper to directly set the IP_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ip.h         | 1 +
 net/ipv4/ip_sockglue.c   | 8 ++++++++
 net/rxrpc/local_object.c | 8 +-------
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 5f5d8226b6abc..f063a491b9063 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 767838d2030d8..aca6b81da9bae 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -589,6 +589,14 @@ void ip_sock_set_freebind(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_freebind);
 
+void ip_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_recverr);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 5ea2bd01fdd59..4c0e8fe5ec1fb 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -171,13 +171,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		/* Fall through */
 	case AF_INET:
 		/* we want to receive ICMP errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
 		opt = IP_PMTUDISC_DO;
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 20/28] ipv4: add ip_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IP_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ip.h         | 1 +
 net/ipv4/ip_sockglue.c   | 8 ++++++++
 net/rxrpc/local_object.c | 8 +-------
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 5f5d8226b6abc..f063a491b9063 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 767838d2030d8..aca6b81da9bae 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -589,6 +589,14 @@ void ip_sock_set_freebind(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_freebind);
 
+void ip_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_recverr);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 5ea2bd01fdd59..4c0e8fe5ec1fb 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -171,13 +171,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		/* Fall through */
 	case AF_INET:
 		/* we want to receive ICMP errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
 		opt = IP_PMTUDISC_DO;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 20/28] ipv4: add ip_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IP_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ip.h         | 1 +
 net/ipv4/ip_sockglue.c   | 8 ++++++++
 net/rxrpc/local_object.c | 8 +-------
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 5f5d8226b6abc..f063a491b9063 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
 #endif	/* _IP_H */
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index 767838d2030d8..aca6b81da9bae 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -589,6 +589,14 @@ void ip_sock_set_freebind(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_freebind);
 
+void ip_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_recverr);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 5ea2bd01fdd59..4c0e8fe5ec1fb 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -171,13 +171,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		/* Fall through */
 	case AF_INET:
 		/* we want to receive ICMP errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
 		opt = IP_PMTUDISC_DO;
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (3 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	drbd-dev, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	linux-rdma

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	drbd-dev, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	linux-rdma

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	drbd-dev

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	drbd-dev, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	linux-rdma

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 21/28] ipv4: add ip_sock_set_mtu_discover
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IP_MTU_DISCOVER sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com> [rxrpc bits]
---
 include/net/ip.h         |  1 +
 net/ipv4/ip_sockglue.c   | 11 +++++++++++
 net/rxrpc/local_object.c |  8 +-------
 net/rxrpc/output.c       | 14 +++++---------
 4 files changed, 18 insertions(+), 16 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index f063a491b9063..d3649c49dd333 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -766,6 +766,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 }
 
 void ip_sock_set_freebind(struct sock *sk);
+int ip_sock_set_mtu_discover(struct sock *sk, int val);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aca6b81da9bae..aa115be11dcfb 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -597,6 +597,17 @@ void ip_sock_set_recverr(struct sock *sk)
 }
 EXPORT_SYMBOL(ip_sock_set_recverr);
 
+int ip_sock_set_mtu_discover(struct sock *sk, int val)
+{
+	if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
+		return -EINVAL;
+	lock_sock(sk);
+	inet_sk(sk)->pmtudisc = val;
+	release_sock(sk);
+	return 0;
+}
+EXPORT_SYMBOL(ip_sock_set_mtu_discover);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 4c0e8fe5ec1fb..6f4e6b4817cf2 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -174,13 +174,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 		ip_sock_set_recverr(local->socket->sk);
 
 		/* we want to set the don't fragment bit */
-		opt = IP_PMTUDISC_DO;
-		ret = kernel_setsockopt(local->socket, SOL_IP, IP_MTU_DISCOVER,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip_sock_set_mtu_discover(local->socket->sk, IP_PMTUDISC_DO);
 
 		/* We want receive timestamps. */
 		sock_enable_timestamps(local->socket->sk);
diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c
index f8b632a5c6197..1ba43c3df4adb 100644
--- a/net/rxrpc/output.c
+++ b/net/rxrpc/output.c
@@ -321,7 +321,7 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	struct kvec iov[2];
 	rxrpc_serial_t serial;
 	size_t len;
-	int ret, opt;
+	int ret;
 
 	_enter(",{%d}", skb->len);
 
@@ -473,18 +473,14 @@ int rxrpc_send_data_packet(struct rxrpc_call *call, struct sk_buff *skb,
 	switch (conn->params.local->srx.transport.family) {
 	case AF_INET6:
 	case AF_INET:
-		opt = IP_PMTUDISC_DONT;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DONT);
 		ret = kernel_sendmsg(conn->params.local->socket, &msg,
 				     iov, 2, len);
 		conn->params.peer->last_tx_at = ktime_get_seconds();
 
-		opt = IP_PMTUDISC_DO;
-		kernel_setsockopt(conn->params.local->socket,
-				  SOL_IP, IP_MTU_DISCOVER,
-				  (char *)&opt, sizeof(opt));
+		ip_sock_set_mtu_discover(conn->params.local->socket->sk,
+				IP_PMTUDISC_DO);
 		break;
 
 	default:
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (3 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, drbd-dev,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, linux-rdma

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, drbd-dev,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, linux-rdma

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, Ying Xue, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, drbd-dev,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, linux-rdma

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 22/28] ipv4: add ip_sock_set_pktinfo
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IP_PKTINFO sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ip.h       | 1 +
 net/ipv4/ip_sockglue.c | 8 ++++++++
 net/sunrpc/svcsock.c   | 5 ++---
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index d3649c49dd333..04ebe7bf54c6a 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -767,6 +767,7 @@ static inline bool inetdev_valid_mtu(unsigned int mtu)
 
 void ip_sock_set_freebind(struct sock *sk);
 int ip_sock_set_mtu_discover(struct sock *sk, int val);
+void ip_sock_set_pktinfo(struct sock *sk);
 void ip_sock_set_recverr(struct sock *sk);
 void ip_sock_set_tos(struct sock *sk, int val);
 
diff --git a/net/ipv4/ip_sockglue.c b/net/ipv4/ip_sockglue.c
index aa115be11dcfb..84ec3703c9091 100644
--- a/net/ipv4/ip_sockglue.c
+++ b/net/ipv4/ip_sockglue.c
@@ -608,6 +608,14 @@ int ip_sock_set_mtu_discover(struct sock *sk, int val)
 }
 EXPORT_SYMBOL(ip_sock_set_mtu_discover);
 
+void ip_sock_set_pktinfo(struct sock *sk)
+{
+	lock_sock(sk);
+	inet_sk(sk)->cmsg_flags |= IP_CMSG_PKTINFO;
+	release_sock(sk);
+}
+EXPORT_SYMBOL(ip_sock_set_pktinfo);
+
 /*
  *	Socket option code for IP. This is the end of the line after any
  *	TCP,UDP etc options on an IP socket.
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 6773dacc64d8e..7a805d165689c 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -616,9 +616,8 @@ static void svc_udp_init(struct svc_sock *svsk, struct svc_serv *serv)
 	/* make sure we get destination address info */
 	switch (svsk->sk_sk->sk_family) {
 	case AF_INET:
-		level = SOL_IP;
-		optname = IP_PKTINFO;
-		break;
+		ip_sock_set_pktinfo(svsk->sk_sock->sk);
+		return;
 	case AF_INET6:
 		level = SOL_IPV6;
 		optname = IPV6_RECVPKTINFO;
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 23/28] ipv6: add ip6_sock_set_v6only
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IPV6_V6ONLY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h        | 11 +++++++++++
 net/ipv6/ip6_udp_tunnel.c |  5 +----
 net/sunrpc/svcsock.c      |  6 +-----
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 39a00d3ef5e22..9b91188c9a74c 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1177,4 +1177,15 @@ int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 			  const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
 		      const struct in6_addr *addr);
+
+static inline int ip6_sock_set_v6only(struct sock *sk)
+{
+	if (inet_sk(sk)->inet_num)
+		return -EINVAL;
+	lock_sock(sk);
+	sk->sk_ipv6only = true;
+	release_sock(sk);
+	return 0;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 6523609516d25..2e0ad1bc84a83 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -25,10 +25,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->ipv6_v6only) {
-		int val = 1;
-
-		err = kernel_setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY,
-					(char *) &val, sizeof(val));
+		err = ip6_sock_set_v6only(sock->sk);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7a805d165689c..a391892977cd2 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1328,7 +1328,6 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	struct sockaddr *newsin = (struct sockaddr *)&addr;
 	int		newlen;
 	int		family;
-	int		val;
 	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 
 	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1364,11 +1363,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	 * getting requests from IPv4 remotes.  Those should
 	 * be shunted to a PF_INET listener via rpcbind.
 	 */
-	val = 1;
 	if (family == PF_INET6)
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
-					(char *)&val, sizeof(val));
-
+		ip6_sock_set_v6only(sock->sk);
 	if (type == SOCK_STREAM)
 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
 	error = kernel_bind(sock, sin, len);
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 23/28] ipv6: add ip6_sock_set_v6only
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IPV6_V6ONLY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h        | 11 +++++++++++
 net/ipv6/ip6_udp_tunnel.c |  5 +----
 net/sunrpc/svcsock.c      |  6 +-----
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 39a00d3ef5e22..9b91188c9a74c 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1177,4 +1177,15 @@ int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 			  const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
 		      const struct in6_addr *addr);
+
+static inline int ip6_sock_set_v6only(struct sock *sk)
+{
+	if (inet_sk(sk)->inet_num)
+		return -EINVAL;
+	lock_sock(sk);
+	sk->sk_ipv6only = true;
+	release_sock(sk);
+	return 0;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 6523609516d25..2e0ad1bc84a83 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -25,10 +25,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->ipv6_v6only) {
-		int val = 1;
-
-		err = kernel_setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY,
-					(char *) &val, sizeof(val));
+		err = ip6_sock_set_v6only(sock->sk);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7a805d165689c..a391892977cd2 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1328,7 +1328,6 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	struct sockaddr *newsin = (struct sockaddr *)&addr;
 	int		newlen;
 	int		family;
-	int		val;
 	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 
 	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1364,11 +1363,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	 * getting requests from IPv4 remotes.  Those should
 	 * be shunted to a PF_INET listener via rpcbind.
 	 */
-	val = 1;
 	if (family = PF_INET6)
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
-					(char *)&val, sizeof(val));
-
+		ip6_sock_set_v6only(sock->sk);
 	if (type = SOCK_STREAM)
 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
 	error = kernel_bind(sock, sin, len);
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 23/28] ipv6: add ip6_sock_set_v6only
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the IPV6_V6ONLY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h        | 11 +++++++++++
 net/ipv6/ip6_udp_tunnel.c |  5 +----
 net/sunrpc/svcsock.c      |  6 +-----
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 39a00d3ef5e22..9b91188c9a74c 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1177,4 +1177,15 @@ int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 			  const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
 		      const struct in6_addr *addr);
+
+static inline int ip6_sock_set_v6only(struct sock *sk)
+{
+	if (inet_sk(sk)->inet_num)
+		return -EINVAL;
+	lock_sock(sk);
+	sk->sk_ipv6only = true;
+	release_sock(sk);
+	return 0;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 6523609516d25..2e0ad1bc84a83 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -25,10 +25,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->ipv6_v6only) {
-		int val = 1;
-
-		err = kernel_setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY,
-					(char *) &val, sizeof(val));
+		err = ip6_sock_set_v6only(sock->sk);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7a805d165689c..a391892977cd2 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1328,7 +1328,6 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	struct sockaddr *newsin = (struct sockaddr *)&addr;
 	int		newlen;
 	int		family;
-	int		val;
 	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 
 	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1364,11 +1363,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	 * getting requests from IPv4 remotes.  Those should
 	 * be shunted to a PF_INET listener via rpcbind.
 	 */
-	val = 1;
 	if (family == PF_INET6)
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
-					(char *)&val, sizeof(val));
-
+		ip6_sock_set_v6only(sock->sk);
 	if (type == SOCK_STREAM)
 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
 	error = kernel_bind(sock, sin, len);
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 23/28] ipv6: add ip6_sock_set_v6only
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IPV6_V6ONLY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h        | 11 +++++++++++
 net/ipv6/ip6_udp_tunnel.c |  5 +----
 net/sunrpc/svcsock.c      |  6 +-----
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 39a00d3ef5e22..9b91188c9a74c 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1177,4 +1177,15 @@ int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 			  const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
 		      const struct in6_addr *addr);
+
+static inline int ip6_sock_set_v6only(struct sock *sk)
+{
+	if (inet_sk(sk)->inet_num)
+		return -EINVAL;
+	lock_sock(sk);
+	sk->sk_ipv6only = true;
+	release_sock(sk);
+	return 0;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 6523609516d25..2e0ad1bc84a83 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -25,10 +25,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->ipv6_v6only) {
-		int val = 1;
-
-		err = kernel_setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY,
-					(char *) &val, sizeof(val));
+		err = ip6_sock_set_v6only(sock->sk);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7a805d165689c..a391892977cd2 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1328,7 +1328,6 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	struct sockaddr *newsin = (struct sockaddr *)&addr;
 	int		newlen;
 	int		family;
-	int		val;
 	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 
 	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1364,11 +1363,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	 * getting requests from IPv4 remotes.  Those should
 	 * be shunted to a PF_INET listener via rpcbind.
 	 */
-	val = 1;
 	if (family == PF_INET6)
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
-					(char *)&val, sizeof(val));
-
+		ip6_sock_set_v6only(sock->sk);
 	if (type == SOCK_STREAM)
 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
 	error = kernel_bind(sock, sin, len);
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 23/28] ipv6: add ip6_sock_set_v6only
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IPV6_V6ONLY sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h        | 11 +++++++++++
 net/ipv6/ip6_udp_tunnel.c |  5 +----
 net/sunrpc/svcsock.c      |  6 +-----
 3 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 39a00d3ef5e22..9b91188c9a74c 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1177,4 +1177,15 @@ int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
 			  const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
 		      const struct in6_addr *addr);
+
+static inline int ip6_sock_set_v6only(struct sock *sk)
+{
+	if (inet_sk(sk)->inet_num)
+		return -EINVAL;
+	lock_sock(sk);
+	sk->sk_ipv6only = true;
+	release_sock(sk);
+	return 0;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ip6_udp_tunnel.c b/net/ipv6/ip6_udp_tunnel.c
index 6523609516d25..2e0ad1bc84a83 100644
--- a/net/ipv6/ip6_udp_tunnel.c
+++ b/net/ipv6/ip6_udp_tunnel.c
@@ -25,10 +25,7 @@ int udp_sock_create6(struct net *net, struct udp_port_cfg *cfg,
 		goto error;
 
 	if (cfg->ipv6_v6only) {
-		int val = 1;
-
-		err = kernel_setsockopt(sock, IPPROTO_IPV6, IPV6_V6ONLY,
-					(char *) &val, sizeof(val));
+		err = ip6_sock_set_v6only(sock->sk);
 		if (err < 0)
 			goto error;
 	}
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index 7a805d165689c..a391892977cd2 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -1328,7 +1328,6 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	struct sockaddr *newsin = (struct sockaddr *)&addr;
 	int		newlen;
 	int		family;
-	int		val;
 	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
 
 	dprintk("svc: svc_create_socket(%s, %d, %s)\n",
@@ -1364,11 +1363,8 @@ static struct svc_xprt *svc_create_socket(struct svc_serv *serv,
 	 * getting requests from IPv4 remotes.  Those should
 	 * be shunted to a PF_INET listener via rpcbind.
 	 */
-	val = 1;
 	if (family == PF_INET6)
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY,
-					(char *)&val, sizeof(val));
-
+		ip6_sock_set_v6only(sock->sk);
 	if (type == SOCK_STREAM)
 		sock->sk->sk_reuse = SK_CAN_REUSE; /* allow address reuse */
 	error = kernel_bind(sock, sin, len);
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 24/28] ipv6: add ip6_sock_set_recverr
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IPV6_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ipv6.h       |  7 +++++++
 net/rxrpc/local_object.c | 10 ++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9b91188c9a74c..49c4abf991489 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1188,4 +1188,11 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 	return 0;
 }
 
+static inline void ip6_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet6_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 6f4e6b4817cf2..c8b2097f499c0 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -107,7 +107,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
 	struct sock *usk;
-	int ret, opt;
+	int ret;
 
 	_enter("%p{%d,%d}",
 	       local, local->srx.transport_type, local->srx.transport.family);
@@ -157,13 +157,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	switch (local->srx.transport.family) {
 	case AF_INET6:
 		/* we want to receive ICMPv6 errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip6_sock_set_recverr(local->socket->sk);
 
 		/* Fall through and set IPv4 options too otherwise we don't get
 		 * errors from IPv4 packets sent through the IPv6 socket.
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 24/28] ipv6: add ip6_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IPV6_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ipv6.h       |  7 +++++++
 net/rxrpc/local_object.c | 10 ++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9b91188c9a74c..49c4abf991489 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1188,4 +1188,11 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 	return 0;
 }
 
+static inline void ip6_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet6_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 6f4e6b4817cf2..c8b2097f499c0 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -107,7 +107,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
 	struct sock *usk;
-	int ret, opt;
+	int ret;
 
 	_enter("%p{%d,%d}",
 	       local, local->srx.transport_type, local->srx.transport.family);
@@ -157,13 +157,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	switch (local->srx.transport.family) {
 	case AF_INET6:
 		/* we want to receive ICMPv6 errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip6_sock_set_recverr(local->socket->sk);
 
 		/* Fall through and set IPv4 options too otherwise we don't get
 		 * errors from IPv4 packets sent through the IPv6 socket.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 24/28] ipv6: add ip6_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, David Howells,
	Ying Xue, Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel,
	drbd-dev

Add a helper to directly set the IPV6_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ipv6.h       |  7 +++++++
 net/rxrpc/local_object.c | 10 ++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9b91188c9a74c..49c4abf991489 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1188,4 +1188,11 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 	return 0;
 }
 
+static inline void ip6_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet6_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 6f4e6b4817cf2..c8b2097f499c0 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -107,7 +107,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
 	struct sock *usk;
-	int ret, opt;
+	int ret;
 
 	_enter("%p{%d,%d}",
 	       local, local->srx.transport_type, local->srx.transport.family);
@@ -157,13 +157,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	switch (local->srx.transport.family) {
 	case AF_INET6:
 		/* we want to receive ICMPv6 errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip6_sock_set_recverr(local->socket->sk);
 
 		/* Fall through and set IPv4 options too otherwise we don't get
 		 * errors from IPv4 packets sent through the IPv6 socket.
-- 
2.26.2


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Ocfs2-devel] [PATCH 24/28] ipv6: add ip6_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion,
	David Howells

Add a helper to directly set the IPV6_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ipv6.h       |  7 +++++++
 net/rxrpc/local_object.c | 10 ++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9b91188c9a74c..49c4abf991489 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1188,4 +1188,11 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 	return 0;
 }
 
+static inline void ip6_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet6_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 6f4e6b4817cf2..c8b2097f499c0 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -107,7 +107,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
 	struct sock *usk;
-	int ret, opt;
+	int ret;
 
 	_enter("%p{%d,%d}",
 	       local, local->srx.transport_type, local->srx.transport.family);
@@ -157,13 +157,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	switch (local->srx.transport.family) {
 	case AF_INET6:
 		/* we want to receive ICMPv6 errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip6_sock_set_recverr(local->socket->sk);
 
 		/* Fall through and set IPv4 options too otherwise we don't get
 		 * errors from IPv4 packets sent through the IPv6 socket.
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [Cluster-devel] [PATCH 24/28] ipv6: add ip6_sock_set_recverr
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: cluster-devel.redhat.com

Add a helper to directly set the IPV6_RECVERR sockopt from kernel space
without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
---
 include/net/ipv6.h       |  7 +++++++
 net/rxrpc/local_object.c | 10 ++--------
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9b91188c9a74c..49c4abf991489 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1188,4 +1188,11 @@ static inline int ip6_sock_set_v6only(struct sock *sk)
 	return 0;
 }
 
+static inline void ip6_sock_set_recverr(struct sock *sk)
+{
+	lock_sock(sk);
+	inet6_sk(sk)->recverr = true;
+	release_sock(sk);
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 6f4e6b4817cf2..c8b2097f499c0 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -107,7 +107,7 @@ static struct rxrpc_local *rxrpc_alloc_local(struct rxrpc_net *rxnet,
 static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 {
 	struct sock *usk;
-	int ret, opt;
+	int ret;
 
 	_enter("%p{%d,%d}",
 	       local, local->srx.transport_type, local->srx.transport.family);
@@ -157,13 +157,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	switch (local->srx.transport.family) {
 	case AF_INET6:
 		/* we want to receive ICMPv6 errors */
-		opt = 1;
-		ret = kernel_setsockopt(local->socket, SOL_IPV6, IPV6_RECVERR,
-					(char *) &opt, sizeof(opt));
-		if (ret < 0) {
-			_debug("setsockopt failed");
-			goto error;
-		}
+		ip6_sock_set_recverr(local->socket->sk);
 
 		/* Fall through and set IPv4 options too otherwise we don't get
 		 * errors from IPv4 packets sent through the IPv6 socket.
-- 
2.26.2



^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 25/28] ipv6: add ip6_sock_set_addr_preferences
  2020-05-28  5:12 ` Christoph Hellwig
                     ` (2 preceding siblings ...)
  (?)
@ 2020-05-28  5:12   ` Christoph Hellwig
  -1 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IPV6_ADD_PREFERENCES sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h       | 67 ++++++++++++++++++++++++++++++++++++++++
 net/ipv6/ipv6_sockglue.c | 59 +----------------------------------
 net/sunrpc/xprtsock.c    |  7 +++--
 3 files changed, 72 insertions(+), 61 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 49c4abf991489..9a90759830162 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1195,4 +1195,71 @@ static inline void ip6_sock_set_recverr(struct sock *sk)
 	release_sock(sk);
 }
 
+static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
+{
+	unsigned int pref = 0;
+	unsigned int prefmask = ~0;
+
+	/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
+	switch (val & (IPV6_PREFER_SRC_PUBLIC |
+		       IPV6_PREFER_SRC_TMP |
+		       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
+	case IPV6_PREFER_SRC_PUBLIC:
+		pref |= IPV6_PREFER_SRC_PUBLIC;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_TMP:
+		pref |= IPV6_PREFER_SRC_TMP;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check HOME/COA conflicts */
+	switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
+	case IPV6_PREFER_SRC_HOME:
+		prefmask &= ~IPV6_PREFER_SRC_COA;
+		break;
+	case IPV6_PREFER_SRC_COA:
+		pref |= IPV6_PREFER_SRC_COA;
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check CGA/NONCGA conflicts */
+	switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
+	case IPV6_PREFER_SRC_CGA:
+	case IPV6_PREFER_SRC_NONCGA:
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	inet6_sk(sk)->srcprefs = (inet6_sk(sk)->srcprefs & prefmask) | pref;
+	return 0;
+}
+
+static inline int ip6_sock_set_addr_preferences(struct sock *sk, bool val)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = __ip6_sock_set_addr_preferences(sk, val);
+	release_sock(sk);
+	return ret;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index e10258c2210e8..adbfed6adf11c 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -845,67 +845,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case IPV6_ADDR_PREFERENCES:
-	    {
-		unsigned int pref = 0;
-		unsigned int prefmask = ~0;
-
 		if (optlen < sizeof(int))
 			goto e_inval;
-
-		retv = -EINVAL;
-
-		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
-		switch (val & (IPV6_PREFER_SRC_PUBLIC|
-			       IPV6_PREFER_SRC_TMP|
-			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
-		case IPV6_PREFER_SRC_PUBLIC:
-			pref |= IPV6_PREFER_SRC_PUBLIC;
-			break;
-		case IPV6_PREFER_SRC_TMP:
-			pref |= IPV6_PREFER_SRC_TMP;
-			break;
-		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
-			break;
-		case 0:
-			goto pref_skip_pubtmp;
-		default:
-			goto e_inval;
-		}
-
-		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
-			      IPV6_PREFER_SRC_TMP);
-pref_skip_pubtmp:
-
-		/* check HOME/COA conflicts */
-		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
-		case IPV6_PREFER_SRC_HOME:
-			break;
-		case IPV6_PREFER_SRC_COA:
-			pref |= IPV6_PREFER_SRC_COA;
-		case 0:
-			goto pref_skip_coa;
-		default:
-			goto e_inval;
-		}
-
-		prefmask &= ~IPV6_PREFER_SRC_COA;
-pref_skip_coa:
-
-		/* check CGA/NONCGA conflicts */
-		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
-		case IPV6_PREFER_SRC_CGA:
-		case IPV6_PREFER_SRC_NONCGA:
-		case 0:
-			break;
-		default:
-			goto e_inval;
-		}
-
-		np->srcprefs = (np->srcprefs & prefmask) | pref;
-		retv = 0;
-
+		retv = __ip6_sock_set_addr_preferences(sk, val);
 		break;
-	    }
 	case IPV6_MINHOPCOUNT:
 		if (optlen < sizeof(int))
 			goto e_inval;
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 0d3ec055bc12f..3a143e250b9ac 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2150,7 +2150,6 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 
 	if (!transport->inet) {
 		struct sock *sk = sock->sk;
-		unsigned int addr_pref = IPV6_PREFER_SRC_PUBLIC;
 
 		/* Avoid temporary address, they are bad for long-lived
 		 * connections such as NFS mounts.
@@ -2159,8 +2158,10 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 		 *    knowledge about the normal duration of connections,
 		 *    MAY override this as appropriate.
 		 */
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_ADDR_PREFERENCES,
-				(char *)&addr_pref, sizeof(addr_pref));
+		if (xs_addr(xprt)->sa_family == PF_INET6) {
+			ip6_sock_set_addr_preferences(sk,
+				IPV6_PREFER_SRC_PUBLIC);
+		}
 
 		xs_tcp_set_socket_timeouts(xprt, sock);
 
-- 
2.26.2


^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 25/28] ipv6: add ip6_sock_set_addr_preferences
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: Eric Dumazet, Alexey Kuznetsov, Hideaki YOSHIFUJI, Jon Maloy,
	Ying Xue, drbd-dev, linux-kernel, linux-rdma, linux-nvme,
	target-devel, linux-afs, linux-cifs, cluster-devel, ocfs2-devel,
	netdev, ceph-devel, rds-devel, linux-nfs, tipc-discussion

Add a helper to directly set the IPV6_ADD_PREFERENCES sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h       | 67 ++++++++++++++++++++++++++++++++++++++++
 net/ipv6/ipv6_sockglue.c | 59 +----------------------------------
 net/sunrpc/xprtsock.c    |  7 +++--
 3 files changed, 72 insertions(+), 61 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 49c4abf991489..9a90759830162 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1195,4 +1195,71 @@ static inline void ip6_sock_set_recverr(struct sock *sk)
 	release_sock(sk);
 }
 
+static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
+{
+	unsigned int pref = 0;
+	unsigned int prefmask = ~0;
+
+	/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
+	switch (val & (IPV6_PREFER_SRC_PUBLIC |
+		       IPV6_PREFER_SRC_TMP |
+		       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
+	case IPV6_PREFER_SRC_PUBLIC:
+		pref |= IPV6_PREFER_SRC_PUBLIC;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_TMP:
+		pref |= IPV6_PREFER_SRC_TMP;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check HOME/COA conflicts */
+	switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
+	case IPV6_PREFER_SRC_HOME:
+		prefmask &= ~IPV6_PREFER_SRC_COA;
+		break;
+	case IPV6_PREFER_SRC_COA:
+		pref |= IPV6_PREFER_SRC_COA;
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check CGA/NONCGA conflicts */
+	switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
+	case IPV6_PREFER_SRC_CGA:
+	case IPV6_PREFER_SRC_NONCGA:
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	inet6_sk(sk)->srcprefs = (inet6_sk(sk)->srcprefs & prefmask) | pref;
+	return 0;
+}
+
+static inline int ip6_sock_set_addr_preferences(struct sock *sk, bool val)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = __ip6_sock_set_addr_preferences(sk, val);
+	release_sock(sk);
+	return ret;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index e10258c2210e8..adbfed6adf11c 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -845,67 +845,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case IPV6_ADDR_PREFERENCES:
-	    {
-		unsigned int pref = 0;
-		unsigned int prefmask = ~0;
-
 		if (optlen < sizeof(int))
 			goto e_inval;
-
-		retv = -EINVAL;
-
-		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
-		switch (val & (IPV6_PREFER_SRC_PUBLIC|
-			       IPV6_PREFER_SRC_TMP|
-			       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
-		case IPV6_PREFER_SRC_PUBLIC:
-			pref |= IPV6_PREFER_SRC_PUBLIC;
-			break;
-		case IPV6_PREFER_SRC_TMP:
-			pref |= IPV6_PREFER_SRC_TMP;
-			break;
-		case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
-			break;
-		case 0:
-			goto pref_skip_pubtmp;
-		default:
-			goto e_inval;
-		}
-
-		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
-			      IPV6_PREFER_SRC_TMP);
-pref_skip_pubtmp:
-
-		/* check HOME/COA conflicts */
-		switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
-		case IPV6_PREFER_SRC_HOME:
-			break;
-		case IPV6_PREFER_SRC_COA:
-			pref |= IPV6_PREFER_SRC_COA;
-		case 0:
-			goto pref_skip_coa;
-		default:
-			goto e_inval;
-		}
-
-		prefmask &= ~IPV6_PREFER_SRC_COA;
-pref_skip_coa:
-
-		/* check CGA/NONCGA conflicts */
-		switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
-		case IPV6_PREFER_SRC_CGA:
-		case IPV6_PREFER_SRC_NONCGA:
-		case 0:
-			break;
-		default:
-			goto e_inval;
-		}
-
-		np->srcprefs = (np->srcprefs & prefmask) | pref;
-		retv = 0;
-
+		retv = __ip6_sock_set_addr_preferences(sk, val);
 		break;
-	    }
 	case IPV6_MINHOPCOUNT:
 		if (optlen < sizeof(int))
 			goto e_inval;
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 0d3ec055bc12f..3a143e250b9ac 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2150,7 +2150,6 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 
 	if (!transport->inet) {
 		struct sock *sk = sock->sk;
-		unsigned int addr_pref = IPV6_PREFER_SRC_PUBLIC;
 
 		/* Avoid temporary address, they are bad for long-lived
 		 * connections such as NFS mounts.
@@ -2159,8 +2158,10 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 		 *    knowledge about the normal duration of connections,
 		 *    MAY override this as appropriate.
 		 */
-		kernel_setsockopt(sock, SOL_IPV6, IPV6_ADDR_PREFERENCES,
-				(char *)&addr_pref, sizeof(addr_pref));
+		if (xs_addr(xprt)->sa_family = PF_INET6) {
+			ip6_sock_set_addr_preferences(sk,
+				IPV6_PREFER_SRC_PUBLIC);
+		}
 
 		xs_tcp_set_socket_timeouts(xprt, sock);
 
-- 
2.26.2

^ permalink raw reply related	[flat|nested] 165+ messages in thread

* [PATCH 25/28] ipv6: add ip6_sock_set_addr_preferences
@ 2020-05-28  5:12   ` Christoph Hellwig
  0 siblings, 0 replies; 165+ messages in thread
From: Christoph Hellwig @ 2020-05-28  5:12 UTC (permalink / raw)
  To: David S. Miller, Jakub Kicinski
  Cc: linux-cifs, rds-devel, cluster-devel, Hideaki YOSHIFUJI, netdev,
	linux-kernel, linux-nvme, Jon Maloy, linux-rdma, Eric Dumazet,
	target-devel, tipc-discussion, linux-nfs, Ying Xue,
	Alexey Kuznetsov, ceph-devel, linux-afs, ocfs2-devel, drbd-dev

Add a helper to directly set the IPV6_ADD_PREFERENCES sockopt from kernel
space without going through a fake uaccess.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/net/ipv6.h       | 67 ++++++++++++++++++++++++++++++++++++++++
 net/ipv6/ipv6_sockglue.c | 59 +----------------------------------
 net/sunrpc/xprtsock.c    |  7 +++--
 3 files changed, 72 insertions(+), 61 deletions(-)

diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 49c4abf991489..9a90759830162 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -1195,4 +1195,71 @@ static inline void ip6_sock_set_recverr(struct sock *sk)
 	release_sock(sk);
 }
 
+static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
+{
+	unsigned int pref = 0;
+	unsigned int prefmask = ~0;
+
+	/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
+	switch (val & (IPV6_PREFER_SRC_PUBLIC |
+		       IPV6_PREFER_SRC_TMP |
+		       IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
+	case IPV6_PREFER_SRC_PUBLIC:
+		pref |= IPV6_PREFER_SRC_PUBLIC;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_TMP:
+		pref |= IPV6_PREFER_SRC_TMP;
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
+		prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
+			      IPV6_PREFER_SRC_TMP);
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check HOME/COA conflicts */
+	switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
+	case IPV6_PREFER_SRC_HOME:
+		prefmask &= ~IPV6_PREFER_SRC_COA;
+		break;
+	case IPV6_PREFER_SRC_COA:
+		pref |= IPV6_PREFER_SRC_COA;
+		break;
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check CGA/NONCGA conflicts */
+	switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
+	case IPV6_PREFER_SRC_CGA:
+	case IPV6_PREFER_SRC_NONCGA:
+	case 0:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	inet6_sk(sk)->srcprefs = (inet6_sk(sk)->srcprefs & prefmask) | pref;
+	return 0;
+}
+
+static inline int ip6_sock_set_addr_preferences(struct sock *sk, bool val)
+{
+	int ret;
+
+	lock_sock(sk);
+	ret = __ip6_sock_set_addr_preferences(sk, val);
+	release_sock(sk);
+	return ret;
+}
+
 #endif /* _NET_IPV6_H */
diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c
index e10258c2210e8..adbfed6adf11c 100644
--- a/net/ipv6/ipv6_sockglue.c
+++ b/net/ipv6/ipv6_sockglue.c
@@ -845,67 +845,10 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 		break;
 
 	case IPV6_ADDR_PREFERENCES:
-	    {
-		unsigned int pref = 0;
-		unsigned int prefmask = ~0;
-
 		if (optlen < sizeof(int))
 			goto e_inval;
-
-		retv = -EINVAL;
-
-		/* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
-		sw