[PATCH net-next 1/2] net: dccp: initialize (addr,port) listening hashtable

[PATCH net-next 1/2] net: dccp: initialize (addr,port) listening hashtable

[Peter Oskolkov]

Commit d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
removes port-only listener lookups. This caused segfaults in DCCP
lookups because DCCP did not initialize the (addr,port) hashtable.

This patch adds said initialization.

The only non-trivial issue here is the size of the new hashtable.
It seemed reasonable to make it match the size of the port-only
hashtable (= INET_LHTABLE_SIZE) that was used previously. Other
parameters to inet_hashinfo2_init() match those used in TCP.

Tested: syzcaller issues fixed; the second patch in the patchset
        tests that DCCP lookups work correctly.

Fixes: d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
Reported-by: syzcaller <syzkaller@googlegroups.com>
Signed-off-by: Peter Oskolkov <posk@google.com>
---
 net/dccp/proto.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/net/dccp/proto.c b/net/dccp/proto.c
index 658cd32bb7b37..be0b223aa8625 100644
--- a/net/dccp/proto.c
+++ b/net/dccp/proto.c
@@ -1141,6 +1141,9 @@ static int __init dccp_init(void)
 		goto out_fail;
 	rc = -ENOBUFS;
 	inet_hashinfo_init(&dccp_hashinfo);
+	inet_hashinfo2_init(&dccp_hashinfo, "dccp_listen_portaddr_hash",
+			    INET_LHTABLE_SIZE, 21,  /* one slot per 2 MB*/
+			    0, 64 * 1024);
 	dccp_hashinfo.bind_bucket_cachep =
 		kmem_cache_create("dccp_bind_bucket",
 				  sizeof(struct inet_bind_bucket), 0,
-- 
2.20.0.405.gbc1bbc6f85-goog

[PATCH net-next 2/2] selftests: net: reuseport_addr_any: add DCCP

[Peter Oskolkov]

This patch adds coverage of DCCP to reuseport_addr_any selftest.

Signed-off-by: Peter Oskolkov <posk@google.com>
---
 .../selftests/net/reuseport_addr_any.c        | 49 ++++++++++++++++++-
 1 file changed, 47 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/net/reuseport_addr_any.c b/tools/testing/selftests/net/reuseport_addr_any.c
index f5e01d989519d..4ac3e24b7cfcf 100644
--- a/tools/testing/selftests/net/reuseport_addr_any.c
+++ b/tools/testing/selftests/net/reuseport_addr_any.c
@@ -9,6 +9,7 @@
 #include <arpa/inet.h>
 #include <errno.h>
 #include <error.h>
+#include <linux/dccp.h>
 #include <linux/in.h>
 #include <linux/unistd.h>
 #include <stdbool.h>
@@ -75,7 +76,16 @@ static void build_rcv_fd(int family, int proto, int *rcv_fds, int count,
 			error(1, errno, "failed to bind receive socket");
 
 		if (proto == SOCK_STREAM && listen(rcv_fds[i], 10))
-			error(1, errno, "failed to listen on receive port");
+			error(1, errno, "tcp: failed to listen on receive port");
+		else if (proto == SOCK_DCCP) {
+			if (setsockopt(rcv_fds[i], SOL_DCCP,
+					DCCP_SOCKOPT_SERVICE,
+					&(int) {htonl(42)}, sizeof(int)))
+				error(1, errno, "failed to setsockopt");
+
+			if (listen(rcv_fds[i], 10))
+				error(1, errno, "dccp: failed to listen on receive port");
+		}
 	}
 }
 
@@ -124,6 +134,11 @@ static int connect_and_send(int family, int proto)
 	if (fd < 0)
 		error(1, errno, "failed to create send socket");
 
+	if (proto == SOCK_DCCP &&
+		setsockopt(fd, SOL_DCCP, DCCP_SOCKOPT_SERVICE,
+				&(int){htonl(42)}, sizeof(int)))
+		error(1, errno, "failed to setsockopt");
+
 	if (bind(fd, saddr, sz))
 		error(1, errno, "failed to bind send socket");
 
@@ -146,7 +161,7 @@ static int receive_once(int epfd, int proto)
 	if (i < 0)
 		error(1, errno, "epoll_wait failed");
 
-	if (proto == SOCK_STREAM) {
+	if (proto == SOCK_STREAM || proto == SOCK_DCCP) {
 		fd = accept(ev.data.fd, NULL, NULL);
 		if (fd < 0)
 			error(1, errno, "failed to accept");
@@ -259,6 +274,36 @@ int main(void)
 	for (i = 0; i < 9; ++i)
 		close(rcv_fds[i]);
 
+	fprintf(stderr, "---- DCCP IPv4 ----\n");
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 2, 2, NULL);
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds + 4, 1, IP4_ADDR);
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds + 5, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 7, 2, NULL);
+	test(rcv_fds, 9, AF_INET, SOCK_DCCP, rcv_fds[4]);
+	for (i = 0; i < 9; ++i)
+		close(rcv_fds[i]);
+
+	fprintf(stderr, "---- DCCP IPv6 ----\n");
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 2, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 4, 1, IP6_ADDR);
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds + 5, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 7, 2, NULL);
+	test(rcv_fds, 9, AF_INET6, SOCK_DCCP, rcv_fds[4]);
+	for (i = 0; i < 9; ++i)
+		close(rcv_fds[i]);
+
+	fprintf(stderr, "---- DCCP IPv4 mapped to IPv6 ----\n");
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 2, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 4, 1, IP4_MAPPED6);
+	build_rcv_fd(AF_INET, SOCK_DCCP, rcv_fds + 5, 2, NULL);
+	build_rcv_fd(AF_INET6, SOCK_DCCP, rcv_fds + 7, 2, NULL);
+	test(rcv_fds, 9, AF_INET, SOCK_DCCP, rcv_fds[4]);
+	for (i = 0; i < 9; ++i)
+		close(rcv_fds[i]);
+
 	fprintf(stderr, "SUCCESS\n");
 	return 0;
 }
-- 
2.20.0.405.gbc1bbc6f85-goog

Re: [PATCH net-next 1/2] net: dccp: initialize (addr,port) listening hashtable

[David Miller]

From: Peter Oskolkov <posk@google.com>
Date: Sat, 15 Dec 2018 14:27:23 -0800

> Commit d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
> removes port-only listener lookups. This caused segfaults in DCCP
> lookups because DCCP did not initialize the (addr,port) hashtable.
> 
> This patch adds said initialization.
> 
> The only non-trivial issue here is the size of the new hashtable.
> It seemed reasonable to make it match the size of the port-only
> hashtable (= INET_LHTABLE_SIZE) that was used previously. Other
> parameters to inet_hashinfo2_init() match those used in TCP.
> 
> Tested: syzcaller issues fixed; the second patch in the patchset
>         tests that DCCP lookups work correctly.
> 
> Fixes: d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
> Reported-by: syzcaller <syzkaller@googlegroups.com>
> Signed-off-by: Peter Oskolkov <posk@google.com>

Applied.

Re: [PATCH net-next 2/2] selftests: net: reuseport_addr_any: add DCCP

[David Miller]

From: Peter Oskolkov <posk@google.com>
Date: Sat, 15 Dec 2018 14:27:24 -0800

> This patch adds coverage of DCCP to reuseport_addr_any selftest.
> 
> Signed-off-by: Peter Oskolkov <posk@google.com>

Applied.

Re: [PATCH net-next 1/2] net: dccp: initialize (addr,port) listening hashtable

[David Miller]

From: David Miller <davem@davemloft.net>
Date: Sun, 16 Dec 2018 12:14:50 -0800 (PST)

> From: Peter Oskolkov <posk@google.com>
> Date: Sat, 15 Dec 2018 14:27:23 -0800
> 
>> Commit d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
>> removes port-only listener lookups. This caused segfaults in DCCP
>> lookups because DCCP did not initialize the (addr,port) hashtable.
>> 
>> This patch adds said initialization.
>> 
>> The only non-trivial issue here is the size of the new hashtable.
>> It seemed reasonable to make it match the size of the port-only
>> hashtable (= INET_LHTABLE_SIZE) that was used previously. Other
>> parameters to inet_hashinfo2_init() match those used in TCP.
>> 
>> Tested: syzcaller issues fixed; the second patch in the patchset
>>         tests that DCCP lookups work correctly.
>> 
>> Fixes: d9fbc7f6431f "net: tcp: prefer listeners bound to an address"
>> Reported-by: syzcaller <syzkaller@googlegroups.com>
>> Signed-off-by: Peter Oskolkov <posk@google.com>
> 
> Applied.

I had to revert, please test with dccp being modular:

ERROR: "inet_hashinfo2_init" [net/dccp/dccp.ko] undefined!
make[1]: *** [scripts/Makefile.modpost:92: __modpost] Error 1