tcp: fix tcp_init_transfer() to not reset icsk_ca_initialized
Subsystems (4)
PATCH v4 1/1 tcp: fix tcp_init_transfer() to not reset icsk_ca_initialized
This commit fixes a bug (found by syzkaller) that could cause spurious double-initializations for congestion control modules, which could cause memory leaks orother problems for congestion control modules (like CDG) that allocate memory in their init functions.
The buggy scenario constructed by syzkaller was something like:
(1) create a TCP socket
(2) initiate a TFO connect via sendto()
(3) while socket is in TCP_SYN_SENT, call setsockopt(TCP_CONGESTION),
which calls:
tcp_set_congestion_control() ->
tcp_reinit_congestion_control() ->
tcp_init_congestion_control()(4) receive ACK, connection is established, call tcp_init_transfer(), set icsk_ca_initialized=0 (without first calling cc->release()), call tcp_init_congestion_control() again.
Note that in this sequence tcp_init_congestion_control() is called twice without a cc->release() call in between. Thus, for CC modules that allocate memory in their init() function, e.g, CDG, a memory leak may occur. The syzkaller tool managed to find a reproducer that triggered such a leak in CDG.
The bug was introduced when that commit 8919a9b31eb4 ("tcp: Only init congestion control if not initialized already")
introduced icsk_ca_initialized and set icsk_ca_initialized to 0 in tcp_init_transfer(), missing the possibility for a sequence like the one above, where a process could call setsockopt(TCP_CONGESTION) in state TCP_SYN_SENT (i.e. after the connect() or TFO open sendmsg()), which would call tcp_init_congestion_control(). It did not intend to reset any initialization that the user had already explicitly made; it just missed the possibility of that particular sequence (which syzkaller managed to find).
---
V2: - Modify the Subject line.
- Adjust the commit message.
- Add Fixes: tag.
V3: - Fix netdev/verify_fixes format error.
V4: - Add blamed authors to receiver list.
net/ipv4/tcp_input.c | 1 - 1 file changed, 1 deletion(-)
net/ipv4/tcp_input.c +0 -1
@@ -5922,7 +5922,6 @@ void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) tp->snd_cwnd =tcp_init_cwnd (tp,__sk_dst_get (sk)); tp->snd_cwnd_stamp = tcp_jiffies32; - icsk->icsk_ca_initialized =0 ;bpf_skops_established (sk, bpf_op, skb);if (!icsk->icsk_ca_initialized)tcp_init_congestion_control (sk);
--
2.25.1
David Miller
maintainerFrom: Nguyen Dinh Phi <[email protected]>
Date: Sat, 3 Jul 2021 17:34:17 +0800
net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) · line 5927
Quoted 12 lines
net/ipv4/tcp_input.c +0
-1@@ -5922,7 +5922,6 @@ void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) tp->snd_cwnd =tcp_init_cwnd (tp,__sk_dst_get (sk)); tp->snd_cwnd_stamp = tcp_jiffies32; - icsk->icsk_ca_initialized =0 ;bpf_skops_established (sk, bpf_op, skb);if (!icsk->icsk_ca_initialized)tcp_init_congestion_control (sk);
Don't you have to make the tcp_init_congestion_control() call unconditional now?
_______________________________________________
Linux-kernel-mentees mailing list [email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 9 lines · David Miller
From: Nguyen Dinh Phi <[email protected]>
Date: Sat, 3 Jul 2021 17:34:17 +0800net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int
Quoted 5 lines
net/ipv4/tcp_input.c +0
-0@@ -5922,7 +5922,6 @@ void tcp_init_transfer(struct sock *sk, int
bpf_op, struct sk_buff *skb)
Quoted 9 lines
net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) · line 5927
Quoted 7 lines
net/ipv4/tcp_input.c +0
-1tp->snd_cwnd =tcp_init_cwnd (tp,__sk_dst_get (sk)); tp->snd_cwnd_stamp = tcp_jiffies32; - icsk->icsk_ca_initialized =0 ;bpf_skops_established (sk, bpf_op, skb);if (!icsk->icsk_ca_initialized)tcp_init_congestion_control (sk);Don't you have to make the
tcp_init_congestion_control()call
unconditional now?
I think we want to keep it conditional, to avoid double-initialization if the BPF code sets the congestion control algorithm and initializes it. But that's relatively new and subtle, so it might be nice for this patch to add a comment about that, since it's touching this part of the code anyway:
- icsk->icsk_ca_initialized = 0;
bpf_skops_established(sk, bpf_op, skb);
+ /* Initialize congestion control unless a BPF op initialized it
already: */
if (!icsk->icsk_ca_initialized)
tcp_init_congestion_control(sk);neal
_______________________________________________
Linux-kernel-mentees mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Quoted 26 lines · Neal Cardwell
Quoted 9 lines · On Sat, Jul 3, 2021 at 5:49 PM David Miller <[email protected]…
From: Nguyen Dinh Phi <[email protected] <mailto:[email protected]>>
Date: Sat, 3 Jul 2021 17:34:17 +0800net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int
Quoted 5 lines
net/ipv4/tcp_input.c +0
-0@@ -5922,7 +5922,6 @@ void tcp_init_transfer(struct sock *sk, int
bpf_op, structsk_buff*skb)Quoted 9 lines
net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) · line 5927
Quoted 7 lines
net/ipv4/tcp_input.c +0
-1tp->snd_cwnd = tcp_init_cwnd(tp, __sk_dst_get(sk)); tp->snd_cwnd_stamp = tcp_jiffies32; - icsk->icsk_ca_initialized = 0; bpf_skops_established(sk, bpf_op, skb); if (!icsk->icsk_ca_initialized) tcp_init_congestion_control(sk);Don't you have to make the
tcp_init_congestion_control()callunconditional now?
I think we want to keep it conditional, to avoid double-initialization if the BPF code sets the congestion control algorithm and initializes it. But that's relatively new and subtle, so it might be nice for this patch to add a comment about that, since it's touching this part of the code anyway:
Quoted 9 lines
- icsk->icsk_ca_initialized = 0;
bpf_skops_established(sk, bpf_op, skb);
+ /* Initialize congestion control unless a BPF op initialized it
already: */
if (!icsk->icsk_ca_initialized)tcp_init_congestion_control(sk);neal
I will make a new version with your comment.
Thank you very much for helping.
Neal Cardwell
maintainerQuoted 12 lines · David Miller
From: Nguyen Dinh Phi <[email protected]>
Date: Sat, 3 Jul 2021 17:34:17 +0800Quoted 8 lines
net/ipv4/tcp_input.c +0
-0@@ -5922,7 +5922,6 @@ void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) tp->snd_cwnd =tcp_init_cwnd (tp,__sk_dst_get (sk)); tp->snd_cwnd_stamp = tcp_jiffies32;
Quoted 6 lines
net/ipv4/tcp_input.c void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb) · line 5927
Quoted 4 lines
net/ipv4/tcp_input.c +0
-1- icsk->icsk_ca_initialized =0 ;bpf_skops_established (sk, bpf_op, skb);if (!icsk->icsk_ca_initialized)tcp_init_congestion_control (sk);Don't you have to make the
tcp_init_congestion_control()call unconditional now?
I think we want to keep it conditional, to avoid double-initialization if the BPF code sets the congestion control algorithm and initializes it. But that's relatively new and subtle, so it might be nice for this patch to add a comment about that, since it's touching this part of the code anyway:
- icsk->icsk_ca_initialized = 0;
bpf_skops_established(sk, bpf_op, skb);
+ /* Initialize congestion control unless BPF initialized it already: */
if (!icsk->icsk_ca_initialized)
tcp_init_congestion_control(sk);neal