Skip to contentSkip to conversation

mptcp: Make mptcp initialization part of tcp initialization

PATCH In review Peter Krystad 3 participants 5 messages May 28, 2019 – May 31, 2019

Subsystems (3)
lore
Share & tools
lore

Peter Krystad

author
#0 lore raw

PATCH v3 1/1 mptcp: Make mptcp initialization part of tcp initialization

Remove remnant module initialization of MPTCP and make it part of the inet/tcp initialization sequence.

OK to squash but needs to be split up between "mptcp: Add MPTCP socket stubs" and
"mptcp: Associate MPTCP context with TCP socket"

v2 - ignore
v3 - fix function declaration

Signed-off-by Peter Krystad <peter.krystad(a)linux.intel.com>
---
 include/net/mptcp.h  |  6 ++++++
 net/ipv4/tcp.c       |  2 ++
 net/mptcp/protocol.c | 22 +++-------------------
 net/mptcp/subflow.c  | 15 ++++++++++++---
 4 files changed, 23 insertions(+), 22 deletions(-)
include/net/mptcp.h +6 -0
index ed67b140af33..65dd7f052e94 100644
@@ -49,6 +49,8 @@ struct mptcp_out_options {
 
 #ifdef CONFIG_MPTCP
 
+void mptcp_init(void);
+
 static inline bool sk_is_mptcp(const struct sock *sk)
 {
 	return tcp_sk(sk)->is_mptcp;
@@ -84,6 +86,10 @@ void mptcp_write_options(__be32 *ptr, struct mptcp_out_options *opts);
 
 #else
 
+static inline void mptcp_init(void)
+{
+}
+
 static inline bool sk_is_mptcp(const struct sock *sk)
 {
 	return false;
net/ipv4/tcp.c +2 -0
index cb072dce81a7..ab242851da3f 100644
@@ -275,6 +275,7 @@
 #include <net/icmp.h>
 #include <net/inet_common.h>
 #include <net/tcp.h>
+#include <net/mptcp.h>
 #include <net/xfrm.h>
 #include <net/ip.h>
 #include <net/sock.h>
@@ -3947,4 +3948,5 @@ void __init tcp_init(void)
 	tcp_metrics_init();
 	BUG_ON(tcp_register_congestion_control(&tcp_reno) != 0);
 	tcp_tasklet_init();
+	mptcp_init();
 }
net/mptcp/protocol.c +3 -19
index 06312efc0b90..d1c617852e0f 100644
@@ -1009,7 +1009,7 @@ static struct inet_protosw mptcp_protosw = {
 	.flags		= INET_PROTOSW_ICSK,
 };
 
-static int __init mptcp_init(void)
+void mptcp_init(void)
 {
 	int err;
 
@@ -1036,26 +1036,10 @@ static int __init mptcp_init(void)
 
 	inet_register_protosw(&mptcp_protosw);
 
-	return 0;
-
+	return;
 proto_failed:
 	subflow_exit();
 
 subflow_failed:
-	return err;
-}
-
-static void __exit mptcp_exit(void)
-{
-	inet_unregister_protosw(&mptcp_protosw);
-	proto_unregister(&mptcp_prot);
-
-	subflow_exit();
+	return;
 }
-
-module_init(mptcp_init);
-module_exit(mptcp_exit);
-
-MODULE_LICENSE("GPL");
-MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_MPTCP);
-MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_MPTCP);
net/mptcp/subflow.c +12 -3
index bc8e1a0658ec..e4ae5c5d177d 100644
@@ -294,6 +294,12 @@ static int subflow_ops_init(struct request_sock_ops *subflow_ops)
 	return 0;
 }
 
+static void subflow_ops_exit(void)
+{
+	kfree(subflow_request_sock_ops.slab_name);
+	kmem_cache_destroy(subflow_request_sock_ops.slab);
+}
+
 int subflow_init(void)
 {
 	int ret;
@@ -312,12 +318,15 @@ int subflow_init(void)
 	subflow_specific.sk_rx_dst_set = subflow_finish_connect;
 	subflow_specific.rebuild_header = subflow_rebuild_header;
 
-	return tcp_register_ulp(&subflow_ulp_ops);
+	ret = tcp_register_ulp(&subflow_ulp_ops);
+	if (ret != 0)
+		subflow_ops_exit();
+
+	return ret;
 }
 
 void subflow_exit(void)
 {
 	tcp_unregister_ulp(&subflow_ulp_ops);
+	subflow_ops_exit();
 }
-
-MODULE_LICENSE("GPL");
Signature · 1 line
2.17.2
#1 lore raw

Hi Peter, Florian,

Quoted 9 lines · Peter Krystad

Remove remnant module initialization of MPTCP and make it part of the inet/tcp initialization sequence.

OK to squash but needs to be split up between "mptcp: Add MPTCP socket stubs" and
"mptcp: Associate MPTCP context with TCP socket"

v2 - ignore
v3 - fix function declaration

Thank you for the patch, the new version and the review.

Florian: is this new version OK for you?

Speak to you later,
Matt

Signature · 5 lines
Matthieu Baerts | R&D Engineer
matthieu.baerts(a)tessares.net
Tessares SA | Hybrid Access Solutions
www.tessares.net
1 Avenue Jean Monnet, 1348 Louvain-la-Neuve, Belgium
#2 lore raw
Quoted 1 line · Matthieu Baerts

Florian: is this new version OK for you?

Sure.

#3 lore raw
Quoted 5 lines · Florian Westphal
Quoted 1 line · Matthieu Baerts

Florian: is this new version OK for you?

Sure.

Thank you!

Applying it right now. I have a few comments, I will quote the original email with the patch v3.

Cheers,
Matt

Signature · 5 lines
Matthieu Baerts | R&D Engineer
matthieu.baerts(a)tessares.net
Tessares SA | Hybrid Access Solutions
www.tessares.net
1 Avenue Jean Monnet, 1348 Louvain-la-Neuve, Belgium
#4 lore raw
In reply to Peter Krystad

Hi Peter,

Quoted 9 lines · Peter Krystad

Remove remnant module initialization of MPTCP and make it part of the inet/tcp initialization sequence.

OK to squash but needs to be split up between "mptcp: Add MPTCP socket stubs" and
"mptcp: Associate MPTCP context with TCP socket"

v2 - ignore
v3 - fix function declaration

(note: I have 3 extra comments inline below)

Thank you for the new version!

- 9499dac1f027: "squash" part 1 in "mptcp: Add MPTCP socket stubs"
- efdab8bae260: add signed-off
- c9810d1865f6: conflict 1
- f2d604e96056: conflict 2
- 00074d55450c: "squash" part 2 in "mptcp: Associate MPTCP context with TCP socket" (also removed an extra new line from this patch)
- 058d81e14f71: conflict 3
- 4757d74e2865: "squash" part 3 in "mptcp: Create SUBFLOW socket for incoming connections" (with a small fix, see below)
- result: 035b1eddfc37..f986a2c1c4e0

Note: it is not in "export" branch because net-next doesn't compile for the moment. Waiting for this patch to be applied:

https://patchwork.ozlabs.org/patch/1108255/

When applying your patch, I also found two parts of it that made me adding these three comments here below:

Quoted 45 lines
Signed-off-by Peter Krystad <peter.krystad(a)linux.intel.com>
---
 include/net/mptcp.h  |  6 ++++++
 net/ipv4/tcp.c       |  2 ++
 net/mptcp/protocol.c | 22 +++-------------------
 net/mptcp/subflow.c  | 15 ++++++++++++---
 4 files changed, 23 insertions(+), 22 deletions(-)
include/net/mptcp.h +6 -0
index ed67b140af33..65dd7f052e94 100644
@@ -49,6 +49,8 @@ struct mptcp_out_options {
 
 #ifdef CONFIG_MPTCP
 
+void mptcp_init(void);
+
 static inline bool sk_is_mptcp(const struct sock *sk)
 {
 	return tcp_sk(sk)->is_mptcp;
@@ -84,6 +86,10 @@ void mptcp_write_options(__be32 *ptr, struct mptcp_out_options *opts);
 
 #else
 
+static inline void mptcp_init(void)
+{
+}
+
 static inline bool sk_is_mptcp(const struct sock *sk)
 {
 	return false;
net/ipv4/tcp.c +1 -0
index cb072dce81a7..ab242851da3f 100644
@@ -275,6 +275,7 @@
 #include <net/icmp.h>
 #include <net/inet_common.h>
 #include <net/tcp.h>
+#include <net/mptcp.h>
 #include <net/xfrm.h>
 #include <net/ip.h>
 #include <net/sock.h>

net/ipv4/tcp.c void __init tcp_init(void) · line 3951

Quoted 4 lines
net/ipv4/tcp.c +1 -0
@@ -3947,4 +3948,5 @@ void __init tcp_init(void)
 	tcp_metrics_init();
 	BUG_ON(tcp_register_congestion_control(&tcp_reno) != 0);
 	tcp_tasklet_init();
+	mptcp_init();

I see that the other "_init" functions above either use "panic" inside or BUG_ON() here to deal with errors at startup. I can have a look at doing the same in mptcp_init() before this evening (morning) meeting to cleanly manage errors in this init part and avoid other problems later.

Quoted 48 lines
net/ipv4/tcp.c +0 -0
 }
net/mptcp/protocol.c +3 -19
index 06312efc0b90..d1c617852e0f 100644
@@ -1009,7 +1009,7 @@ static struct inet_protosw mptcp_protosw = {
 	.flags		= INET_PROTOSW_ICSK,
 };
 
-static int __init mptcp_init(void)
+void mptcp_init(void)
 {
 	int err;
 
@@ -1036,26 +1036,10 @@ static int __init mptcp_init(void)
 
 	inet_register_protosw(&mptcp_protosw);
 
-	return 0;
-
+	return;
 proto_failed:
 	subflow_exit();
 
 subflow_failed:
-	return err;
-}
-
-static void __exit mptcp_exit(void)
-{
-	inet_unregister_protosw(&mptcp_protosw);
-	proto_unregister(&mptcp_prot);
-
-	subflow_exit();
+	return;
 }
-
-module_init(mptcp_init);
-module_exit(mptcp_exit);
-
-MODULE_LICENSE("GPL");
-MODULE_ALIAS_NET_PF_PROTO(PF_INET, IPPROTO_MPTCP);
-MODULE_ALIAS_NET_PF_PROTO(PF_INET6, IPPROTO_MPTCP);
net/mptcp/subflow.c +0 -0
index bc8e1a0658ec..e4ae5c5d177d 100644

net/mptcp/subflow.c static int subflow_ops_init(struct request_sock_ops *subflow_ops) · line 299

Quoted 6 lines
net/mptcp/subflow.c +3 -0
@@ -294,6 +294,12 @@ static int subflow_ops_init(struct request_sock_ops *subflow_ops)
 	return 0;
 }
 
+static void subflow_ops_exit(void)
+{
+	kfree(subflow_request_sock_ops.slab_name);

This kfree() is no longer needed. See 9b5faeebe499 (net: mptcp: don't leak 'slab' and 'slab_name' on exit).

Quoted 12 lines
net/mptcp/subflow.c +4 -1
+	kmem_cache_destroy(subflow_request_sock_ops.slab);
+}
+
 int subflow_init(void)
 {
 	int ret;
@@ -312,12 +318,15 @@ int subflow_init(void)
 	subflow_specific.sk_rx_dst_set = subflow_finish_connect;
 	subflow_specific.rebuild_header = subflow_rebuild_header;
 
-	return tcp_register_ulp(&subflow_ulp_ops);
+	ret = tcp_register_ulp(&subflow_ulp_ops);

net/mptcp/subflow.c int subflow_init(void) · line 331

Quoted 10 lines
net/mptcp/subflow.c +5 -0
+	if (ret != 0)
+		subflow_ops_exit();
+
+	return ret;
 }
 
 void subflow_exit(void)
 {
 	tcp_unregister_ulp(&subflow_ulp_ops);
+	subflow_ops_exit();

It seems that you were not on top of the latest "for-review" branch (or "export" one) because kmem_cache_destroy() was called here.

Cheers,
Matt

Quoted 4 lines
net/mptcp/subflow.c +0 -2
 }
-
-MODULE_LICENSE("GPL");
Signature · 5 lines
Matthieu Baerts | R&D Engineer
matthieu.baerts(a)tessares.net
Tessares SA | Hybrid Access Solutions
www.tessares.net
1 Avenue Jean Monnet, 1348 Louvain-la-Neuve, Belgium