Skip to contentSkip to conversation

mptcp: fix for ipv6-related oops

PATCH In review Paolo Abeni 2 participants 9 messages Nov 21, 2019

Subsystems (3)
lore
Share & tools
lore

Paolo Abeni

author
#0 lore raw

cover mptcp: fix for ipv6-related oops

We need to allocate a larger struct sock for mptcp v6 master sockets, otherwise inet6_sk() will access mptcp related data instead of struct ipv6_pinfo.

This should fix the crash reported by Matt.

The protocol.c code changes are split in 2 patches to simplify the squashing.
(yep, I know we somewhat agreed to big giant rebase, but this specific should be trivial).

Additionally adds some fixlet for the ipv6 tests

Paolo Abeni (3):
selftests: fixup for IPv6 tests
mptcp: use ipv6-specific proto struct
mptcp/ipv6: mptcp_v6_prot needs costom destroy.

 net/mptcp/protocol.c                          | 17 ++++++++++--
 .../selftests/net/mptcp/mptcp_connect.sh      | 27 +++++++++----------
 2 files changed, 27 insertions(+), 17 deletions(-)
Signature · 1 line
2.21.0
author
#1 lore raw

PATCH 1/3 selftests: fixup for IPv6 tests

Some distros still have the 'ping6' command, take care of that. Addiotnally, setting 'noadad' on ipv6 addresses will avoid explicit waits later

Squash-to: "mptcp: add basic kselftest for mptcp"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

---
 .../selftests/net/mptcp/mptcp_connect.sh      | 27 +++++++++----------
 1 file changed, 12 insertions(+), 15 deletions(-)
tools/testing/selftests/net/mptcp/mptcp_connect.sh +12 -15
index a03760d12eff..9a065183545c 100755
@@ -17,6 +17,7 @@ ethtool_random_on=true
 tc_delay="$((RANDOM%400))"
 tc_loss=$((RANDOM%101))
 tc_reorder=""
+which ping6 > /dev/null 2>&1 && ping6=$(which ping6) || ping6=$(which ping)
 
 if [ $tc_loss -eq 100 ];then
 	tc_loss=1%
@@ -124,18 +125,18 @@ ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3"
 ip link add ns3eth4 netns "$ns3" type veth peer name ns4eth3 netns "$ns4"
 
 ip -net "$ns1" addr add 10.0.1.1/24 dev ns1eth2
-ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2
+ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2 nodad
 
 ip -net "$ns1" link set ns1eth2 up
 ip -net "$ns1" route add default via 10.0.1.2
 ip -net "$ns1" route add default via dead:beef:1::2
 
 ip -net "$ns2" addr add 10.0.1.2/24 dev ns2eth1
-ip -net "$ns2" addr add dead:beef:1::2/64 dev ns2eth1
+ip -net "$ns2" addr add dead:beef:1::2/64 dev ns2eth1 nodad
 ip -net "$ns2" link set ns2eth1 up
 
 ip -net "$ns2" addr add 10.0.2.1/24 dev ns2eth3
-ip -net "$ns2" addr add dead:beef:2::1/64 dev ns2eth3
+ip -net "$ns2" addr add dead:beef:2::1/64 dev ns2eth3 nodad
 ip -net "$ns2" link set ns2eth3 up
 ip -net "$ns2" route add default via 10.0.2.2
 ip -net "$ns2" route add default via dead:beef:2::2
@@ -143,11 +144,11 @@ ip netns exec "$ns2" sysctl -q net.ipv4.ip_forward=1
 ip netns exec "$ns2" sysctl -q net.ipv6.conf.all.forwarding=1
 
 ip -net "$ns3" addr add 10.0.2.2/24 dev ns3eth2
-ip -net "$ns3" addr add dead:beef:2::2/64 dev ns3eth2
+ip -net "$ns3" addr add dead:beef:2::2/64 dev ns3eth2 nodad
 ip -net "$ns3" link set ns3eth2 up
 
 ip -net "$ns3" addr add 10.0.3.2/24 dev ns3eth4
-ip -net "$ns3" addr add dead:beef:3::2/64 dev ns3eth4
+ip -net "$ns3" addr add dead:beef:3::2/64 dev ns3eth4 nodad
 ip -net "$ns3" link set ns3eth4 up
 ip -net "$ns3" route add default via 10.0.2.1
 ip -net "$ns3" route add default via dead:beef:2::1
@@ -155,7 +156,7 @@ ip netns exec "$ns3" sysctl -q net.ipv4.ip_forward=1
 ip netns exec "$ns3" sysctl -q net.ipv6.conf.all.forwarding=1
 
 ip -net "$ns4" addr add 10.0.3.1/24 dev ns4eth3
-ip -net "$ns4" addr add dead:beef:3::1/64 dev ns4eth3
+ip -net "$ns4" addr add dead:beef:3::1/64 dev ns4eth3 nodad
 ip -net "$ns4" link set ns4eth3 up
 ip -net "$ns4" route add default via 10.0.3.2
 ip -net "$ns4" route add default via dead:beef:3::2
@@ -277,12 +278,14 @@ do_ping()
 	local listener_ns="$1"
 	local connector_ns="$2"
 	local connect_addr="$3"
+	local ping_cmd=ping
 
-	if ! $ipv6 && is_v6 "${connect_addr}"; then
-		return 0
+	if is_v6 "${connect_addr}"; then
+		$ipv6 || return 0
+		ping_cmd=${ping6}
 	fi
 
-	ip netns exec ${connector_ns} ping -q -c 1 $connect_addr >/dev/null
+	ip netns exec ${connector_ns} ${ping_cmd} -q -c 1 $connect_addr >/dev/null
 	if [ $? -ne 0 ] ; then
 		echo "$listener_ns -> $connect_addr connectivity [ FAIL ]" 1>&2
 		ret=1
@@ -493,12 +496,6 @@ show_all_ipv6()
 	done
 }
 
-# Allow DAD to finish
-for dad in $(seq 20); do
-	show_all_ipv6 | grep -q -e tentative -e temporary || break
-	sleep 0.1
-done
-
 echo "INFO: validating network environment with pings"
 for sender in "$ns1" "$ns2" "$ns3" "$ns4";do
 	do_ping "$ns1" $sender 10.0.1.1
Signature · 1 line
2.21.0
#2 lore raw

[MPTCP] Re: [PATCH 1/3] selftests: fixup for IPv6 tests

Hi Paolo,

Quoted 15 lines · Paolo Abeni

Some distros still have the 'ping6' command, take care of that. Addiotnally, setting 'noadad' on ipv6 addresses will avoid explicit waits later

Squash-to: "mptcp: add basic kselftest for mptcp"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

---
  .../selftests/net/mptcp/mptcp_connect.sh      | 27 +++++++++----------
  1 file changed, 12 insertions(+), 15 deletions(-)
tools/testing/selftests/net/mptcp/mptcp_connect.sh +0 -0
index a03760d12eff..9a065183545c 100755

tools/testing/selftests/net/mptcp/mptcp_connect.sh ethtool_random_on=true · line 20

Quoted 4 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +1 -0
@@ -17,6 +17,7 @@ ethtool_random_on=true
  tc_delay="$((RANDOM%400))"
  tc_loss=$((RANDOM%101))
  tc_reorder=""
+which ping6 > /dev/null 2>&1 && ping6=$(which ping6) || ping6=$(which ping)

Just out of curiosity: why is it not OK to use 'ping -6' on these distro?

Note that you can also use:

   ping6=$(which ping6 || which ping)

(I don't think 'which' is supposed to print something if the program is not available)

tools/testing/selftests/net/mptcp/mptcp_connect.sh ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3" · line 128

Quoted 8 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +1 -1
  if [ $tc_loss -eq 100 ];then
  	tc_loss=1%
@@ -124,18 +125,18 @@ ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3"
  ip link add ns3eth4 netns "$ns3" type veth peer name ns4eth3 netns "$ns4"
 
  ip -net "$ns1" addr add 10.0.1.1/24 dev ns1eth2
-ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2
+ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2 nodad

What a sad option :'(

But good idea (Luke)!

[...]

tools/testing/selftests/net/mptcp/mptcp_connect.sh show_all_ipv6() · line 498

Quoted 7 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +0 -3
@@ -493,12 +496,6 @@ show_all_ipv6()
  	done
  }
 
-# Allow DAD to finish
-for dad in $(seq 20); do
-	show_all_ipv6 | grep -q -e tentative -e temporary || break

A detail but I guess we can also remove the show_all_ipv6() function declared just above that.

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
author
#3 lore raw

[MPTCP] Re: [PATCH 1/3] selftests: fixup for IPv6 tests

Quoted 25 lines · Matthieu Baerts

Hi Paolo,

Quoted 15 lines · Paolo Abeni

Some distros still have the 'ping6' command, take care of that. Addiotnally, setting 'noadad' on ipv6 addresses will avoid explicit waits later

Squash-to: "mptcp: add basic kselftest for mptcp"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

---
  .../selftests/net/mptcp/mptcp_connect.sh      | 27 +++++++++----------
  1 file changed, 12 insertions(+), 15 deletions(-)
tools/testing/selftests/net/mptcp/mptcp_connect.sh +0 -0
index a03760d12eff..9a065183545c 100755

tools/testing/selftests/net/mptcp/mptcp_connect.sh ethtool_random_on=true · line 20

Quoted 4 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +1 -0
@@ -17,6 +17,7 @@ ethtool_random_on=true
  tc_delay="$((RANDOM%400))"
  tc_loss=$((RANDOM%101))
  tc_reorder=""
+which ping6 > /dev/null 2>&1 && ping6=$(which ping6) || ping6=$(which ping)

Just out of curiosity: why is it not OK to use 'ping -6' on these distro?

Quoted 3 lines

Note that you can also use:

   ping6=$(which ping6 || which ping)

I shamefully copied the existing code from other self-tests script to deal with that. I think adding '-6' for ipv6 address should work everywhere, let me try that.

Quoted 23 lines

(I don't think 'which' is supposed to print something if the program is not available)

tools/testing/selftests/net/mptcp/mptcp_connect.sh ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3" · line 128

Quoted 8 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +1 -1
  if [ $tc_loss -eq 100 ];then
  	tc_loss=1%
@@ -124,18 +125,18 @@ ip link add ns2eth3 netns "$ns2" type veth peer name ns3eth2 netns "$ns3"
  ip link add ns3eth4 netns "$ns3" type veth peer name ns4eth3 netns "$ns4"
 
  ip -net "$ns1" addr add 10.0.1.1/24 dev ns1eth2
-ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2
+ip -net "$ns1" addr add dead:beef:1::1/64 dev ns1eth2 nodad

What a sad option :'(

But good idea (Luke)!

[...]

Quoted 4 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +0 -0
@@ -493,12 +496,6 @@ show_all_ipv6()
  	done
  }
 
Quoted 6 lines

tools/testing/selftests/net/mptcp/mptcp_connect.sh show_all_ipv6() · line 498

Quoted 3 lines
tools/testing/selftests/net/mptcp/mptcp_connect.sh +0 -3
-# Allow DAD to finish
-for dad in $(seq 20); do
-	show_all_ipv6 | grep -q -e tentative -e temporary || break

A detail but I guess we can also remove the show_all_ipv6() function declared just above that.

Yep, will drop in next iteration.

Thanks,

Paolo

author
#4 lore raw

PATCH 2/3 mptcp: use ipv6-specific proto struct

In reply to Paolo Abeni

Because it needs a different (greater) obj_size.

Squash-to: "mptcp: Add MPTCP socket stubs"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

---
 net/mptcp/protocol.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
net/mptcp/protocol.c +8 -2
index 40da2fae65ac..128822e54672 100644
@@ -689,7 +689,7 @@ static struct proto mptcp_prot = {
 	.unhash		= inet_unhash,
 	.get_port	= mptcp_get_port,
 	.obj_size	= sizeof(struct mptcp_sock),
-	.no_autobind	= 1,
+	.no_autobind	= true,
 };
 
 static struct socket *mptcp_socket_create_get(struct mptcp_sock *msk)
@@ -972,11 +972,12 @@ void mptcp_proto_init(void)
 
 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
 static struct proto_ops mptcp_v6_stream_ops;
+static struct proto mptcp_v6_prot;
 
 static struct inet_protosw mptcp_v6_protosw = {
 	.type		= SOCK_STREAM,
 	.protocol	= IPPROTO_MPTCP,
-	.prot		= &mptcp_prot,
+	.prot		= &mptcp_v6_prot,
 	.ops		= &mptcp_v6_stream_ops,
 	.flags		= INET_PROTOSW_ICSK,
 };
@@ -985,6 +986,11 @@ int mptcp_proto_v6_init(void)
 {
 	int err;
 
+	mptcp_v6_prot = mptcp_prot;
+	strcpy(mptcp_v6_prot.name, "TCPv6");
+	mptcp_v6_prot.obj_size = sizeof(struct mptcp_sock) +
+				 sizeof(struct ipv6_pinfo);
+
 	mptcp_v6_stream_ops = inet6_stream_ops;
 	mptcp_v6_stream_ops.bind = mptcp_bind;
 	mptcp_v6_stream_ops.connect = mptcp_stream_connect;
Signature · 1 line
2.21.0
#5 lore raw

[MPTCP] Re: [PATCH 2/3] mptcp: use ipv6-specific proto struct

Quoted 4 lines · Paolo Abeni

Because it needs a different (greater) obj_size.

Squash-to: "mptcp: Add MPTCP socket stubs"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

[...]

net/mptcp/protocol.c int mptcp_proto_v6_init(void) · line 990

Quoted 6 lines
net/mptcp/protocol.c +2 -0
@@ -985,6 +986,11 @@ int mptcp_proto_v6_init(void)
  {
  	int err;
 
+	mptcp_v6_prot = mptcp_prot;
+	strcpy(mptcp_v6_prot.name, "TCPv6");

Should we call it "MPTCPv6" (mptcp_prot is called "MPTCP")

net/mptcp/protocol.c int mptcp_proto_v6_init(void) · line 992

Quoted 2 lines
net/mptcp/protocol.c +2 -0
+	mptcp_v6_prot.obj_size = sizeof(struct mptcp_sock) +
+				 sizeof(struct ipv6_pinfo);

Very good catch!

Why did we not have the crash before?

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
author
#6 lore raw

[MPTCP] Re: [PATCH 2/3] mptcp: use ipv6-specific proto struct

Quoted 8 lines · Matthieu Baerts
Quoted 4 lines · Paolo Abeni

Because it needs a different (greater) obj_size.

Squash-to: "mptcp: Add MPTCP socket stubs"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

[...]

Quoted 8 lines

net/mptcp/protocol.c int mptcp_proto_v6_init(void) · line 990

Quoted 6 lines
net/mptcp/protocol.c +2 -0
@@ -985,6 +986,11 @@ int mptcp_proto_v6_init(void)
  {
  	int err;
 
+	mptcp_v6_prot = mptcp_prot;
+	strcpy(mptcp_v6_prot.name, "TCPv6");

Should we call it "MPTCPv6" (mptcp_prot is called "MPTCP")

Indeed! I rushed too much to assemble this patches, sorry.

Quoted 7 lines

net/mptcp/protocol.c int mptcp_proto_v6_init(void) · line 992

Quoted 3 lines
net/mptcp/protocol.c +2 -0
+	mptcp_v6_prot.obj_size = sizeof(struct mptcp_sock) +
+				 sizeof(struct ipv6_pinfo);

Very good catch!

Why did we not have the crash before?

Just a matter of luck !?! It depends on how the trailing part of mptcp_sock struct overlap with ipv6_pinfo. If we have NULL values at the 'correct' places, we can't observe the issue without setsockopt()

Possibly mptcp_struct layout changed a bit with the rebase (e.g. fields swapped)??? I haven't double checked that.

Will send v2 soon.

Paolo

p.s. we likely want to add some more self-tests doing [almost random] setsockopt on the mptcp socket before the transfer...

author
#7 lore raw

PATCH 3/3 mptcp/ipv6: mptcp_v6_prot needs costom destroy.

In reply to Paolo Abeni

Otherwise we can leak memory on close, if user space does some setsockopt before connectiong/accepting.

Note: the squash tag points to the commit currently introducing mptcp_destroy, but we likely want to move such function in another one - it's currently added as empty

Squash-to: "mptcp: Add key generation and token tree"
Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>

---
 net/mptcp/protocol.c | 7 +++++++
 1 file changed, 7 insertions(+)
net/mptcp/protocol.c +7 -0
index 128822e54672..d1ea0ba480ed 100644
@@ -974,6 +974,12 @@ void mptcp_proto_init(void)
 static struct proto_ops mptcp_v6_stream_ops;
 static struct proto mptcp_v6_prot;
 
+static void mptcp_v6_destroy(struct sock *sk)
+{
+	mptcp_destroy(sk);
+	inet6_destroy_sock(sk);
+}
+
 static struct inet_protosw mptcp_v6_protosw = {
 	.type		= SOCK_STREAM,
 	.protocol	= IPPROTO_MPTCP,
@@ -988,6 +994,7 @@ int mptcp_proto_v6_init(void)
 
 	mptcp_v6_prot = mptcp_prot;
 	strcpy(mptcp_v6_prot.name, "TCPv6");
+	mptcp_v6_prot.destroy = mptcp_v6_destroy;
 	mptcp_v6_prot.obj_size = sizeof(struct mptcp_sock) +
 				 sizeof(struct ipv6_pinfo);
 
Signature · 1 line
2.21.0
#8 lore raw
In reply to Paolo Abeni

Hi Paolo,

Quoted 5 lines · Paolo Abeni

We need to allocate a larger struct sock for mptcp v6 master sockets, otherwise inet6_sk() will access mptcp related data instead of struct ipv6_pinfo.

This should fix the crash reported by Matt.

Thank you for the quick fix!

Quoted 4 lines

The protocol.c code changes are split in 2 patches to simplify the squashing.
(yep, I know we somewhat agreed to big giant rebase, but this specific should be trivial).

That's a good idea because for the moment, the CI is notifying me each time that the kselftests are failing (when testing the commit introducing them)!

Quoted 1 line

Additionally adds some fixlet for the ipv6 tests

I have a few questions in the 2 first patches.

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