netdev - [PATCH net 3/3] sctp: call sctp_auth_init_hmacs() in sctp_sock

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Sun,  3 Mar 2019 17:54:55 +0800
From:   Xin Long <lucien.xin@...il.com>
To:     network dev <netdev@...r.kernel.org>, linux-sctp@...r.kernel.org
Cc:     davem@...emloft.net,
        Marcelo Ricardo Leitner <marcelo.leitner@...il.com>,
        Neil Horman <nhorman@...driver.com>
Subject: [PATCH net 3/3] sctp: call sctp_auth_init_hmacs() in sctp_sock_migrate()

New ep's auth_hmacs should be set if old ep's is set, in case that
net->sctp.auth_enable has been changed to 0 by users and new ep's
auth_hmacs couldn't be set in sctp_endpoint_init().

It can even crash kernel by doing:

  1. on server: sysctl -w net.sctp.auth_enable=1,
                sysctl -w net.sctp.addip_enable=1,
                sysctl -w net.sctp.addip_noauth_enable=0,
                listen() on server,
                sysctl -w net.sctp.auth_enable=0.
  2. on client: connect() to server.
  3. on server: accept() the asoc,
                sysctl -w net.sctp.auth_enable=1.
  4. on client: send() asconf packet to server.

The call trace:

  [  245.280251] BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
  [  245.286872] RIP: 0010:sctp_auth_calculate_hmac+0xa3/0x140 [sctp]
  [  245.304572] Call Trace:
  [  245.305091]  <IRQ>
  [  245.311287]  sctp_sf_authenticate+0x110/0x160 [sctp]
  [  245.312311]  sctp_sf_eat_auth+0xf2/0x230 [sctp]
  [  245.313249]  sctp_do_sm+0x9a/0x2d0 [sctp]
  [  245.321483]  sctp_assoc_bh_rcv+0xed/0x1a0 [sctp]
  [  245.322495]  sctp_rcv+0xa66/0xc70 [sctp]

It's because the old ep->auth_hmacs wasn't copied to the new ep while
ep->auth_hmacs is used in sctp_auth_calculate_hmac() when processing
the incoming auth chunks, and it should have been done when migrating
sock.

Reported-by: Ying Xu <yinxu@...hat.com>
Signed-off-by: Xin Long <lucien.xin@...il.com>
---
 net/sctp/socket.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/net/sctp/socket.c b/net/sctp/socket.c
index 22adb8d..def3335 100644
--- a/net/sctp/socket.c
+++ b/net/sctp/socket.c
@@ -8978,6 +8978,16 @@ static int sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
 	if (err)
 		return err;
 
+	/* New ep's auth_hmacs should be set if old ep's is set, in case
+	 * that net->sctp.auth_enable has been changed to 0 by users and
+	 * new ep's auth_hmacs couldn't be set in sctp_endpoint_init().
+	 */
+	if (oldsp->ep->auth_hmacs) {
+		err = sctp_auth_init_hmacs(newsp->ep, GFP_KERNEL);
+		if (err)
+			return err;
+	}
+
 	/* Move any messages in the old socket's receive queue that are for the
 	 * peeled off association to the new socket's receive queue.
 	 */
-- 
2.1.0