diff mbox series

[net-next,v3] tcp: Add logic to check for SYN w/ data in tcp_simple_retransmit

Message ID 160780498125.3272.15437756269539236825.stgit@localhost.localdomain (mailing list archive)
State Accepted
Commit c31b70c9968fe9c4194d1b5d06d07596a3b680de
Delegated to: Netdev Maintainers
Headers show
Series [net-next,v3] tcp: Add logic to check for SYN w/ data in tcp_simple_retransmit | expand

Checks

Context Check Description
netdev/cover_letter success Link
netdev/fixes_present success Link
netdev/patch_count success Link
netdev/tree_selection success Clearly marked for net-next
netdev/subject_prefix success Link
netdev/source_inline success Was 0 now: 0
netdev/verify_signedoff success Link
netdev/module_param success Was 0 now: 0
netdev/build_32bit success Errors and warnings before: 2 this patch: 2
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/verify_fixes success Link
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 23 lines checked
netdev/build_allmodconfig_warn success Errors and warnings before: 2 this patch: 2
netdev/header_inline success Link
netdev/stable success Stable not CCed

Commit Message

Alexander Duyck Dec. 12, 2020, 8:31 p.m. UTC
From: Alexander Duyck <alexanderduyck@fb.com>

There are cases where a fastopen SYN may trigger either a ICMP_TOOBIG
message in the case of IPv6 or a fragmentation request in the case of
IPv4. This results in the socket stalling for a second or more as it does
not respond to the message by retransmitting the SYN frame.

Normally a SYN frame should not be able to trigger a ICMP_TOOBIG or
ICMP_FRAG_NEEDED however in the case of fastopen we can have a frame that
makes use of the entire MSS. In the case of fastopen it does, and an
additional complication is that the retransmit queue doesn't contain the
original frames. As a result when tcp_simple_retransmit is called and
walks the list of frames in the queue it may not mark the frames as lost
because both the SYN and the data packet each individually are smaller than
the MSS size after the adjustment. This results in the socket being stalled
until the retransmit timer kicks in and forces the SYN frame out again
without the data attached.

In order to resolve this we can reduce the MSS the packets are compared
to in tcp_simple_retransmit to -1 for cases where we are still in the
TCP_SYN_SENT state for a fastopen socket. Doing this we will mark all of
the packets related to the fastopen SYN as lost.

Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
---

v2: Changed logic to invalidate all retransmit queue frames if fastopen SYN
v3: Updated commit message to reflect actual solution in 3rd paragraph

 net/ipv4/tcp_input.c |   17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

Comments

Eric Dumazet Dec. 14, 2020, 5:42 p.m. UTC | #1
On Sat, Dec 12, 2020 at 9:31 PM Alexander Duyck
<alexander.duyck@gmail.com> wrote:
>
> From: Alexander Duyck <alexanderduyck@fb.com>
>
> There are cases where a fastopen SYN may trigger either a ICMP_TOOBIG
> message in the case of IPv6 or a fragmentation request in the case of
> IPv4. This results in the socket stalling for a second or more as it does
> not respond to the message by retransmitting the SYN frame.
>
> Normally a SYN frame should not be able to trigger a ICMP_TOOBIG or
> ICMP_FRAG_NEEDED however in the case of fastopen we can have a frame that
> makes use of the entire MSS. In the case of fastopen it does, and an
> additional complication is that the retransmit queue doesn't contain the
> original frames. As a result when tcp_simple_retransmit is called and
> walks the list of frames in the queue it may not mark the frames as lost
> because both the SYN and the data packet each individually are smaller than
> the MSS size after the adjustment. This results in the socket being stalled
> until the retransmit timer kicks in and forces the SYN frame out again
> without the data attached.
>
> In order to resolve this we can reduce the MSS the packets are compared
> to in tcp_simple_retransmit to -1 for cases where we are still in the
> TCP_SYN_SENT state for a fastopen socket. Doing this we will mark all of
> the packets related to the fastopen SYN as lost.
>
> Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
> ---
>

SGTM, thanks !

Signed-off-by: Eric Dumazet <edumazet@google.com>

> v2: Changed logic to invalidate all retransmit queue frames if fastopen SYN
> v3: Updated commit message to reflect actual solution in 3rd paragraph
>
Yuchung Cheng Dec. 14, 2020, 6:52 p.m. UTC | #2
On Mon, Dec 14, 2020 at 9:42 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Sat, Dec 12, 2020 at 9:31 PM Alexander Duyck
> <alexander.duyck@gmail.com> wrote:
> >
> > From: Alexander Duyck <alexanderduyck@fb.com>
> >
> > There are cases where a fastopen SYN may trigger either a ICMP_TOOBIG
> > message in the case of IPv6 or a fragmentation request in the case of
> > IPv4. This results in the socket stalling for a second or more as it does
> > not respond to the message by retransmitting the SYN frame.
> >
> > Normally a SYN frame should not be able to trigger a ICMP_TOOBIG or
> > ICMP_FRAG_NEEDED however in the case of fastopen we can have a frame that
> > makes use of the entire MSS. In the case of fastopen it does, and an
> > additional complication is that the retransmit queue doesn't contain the
> > original frames. As a result when tcp_simple_retransmit is called and
> > walks the list of frames in the queue it may not mark the frames as lost
> > because both the SYN and the data packet each individually are smaller than
> > the MSS size after the adjustment. This results in the socket being stalled
> > until the retransmit timer kicks in and forces the SYN frame out again
> > without the data attached.
> >
> > In order to resolve this we can reduce the MSS the packets are compared
> > to in tcp_simple_retransmit to -1 for cases where we are still in the
> > TCP_SYN_SENT state for a fastopen socket. Doing this we will mark all of
> > the packets related to the fastopen SYN as lost.
> >
> > Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
> > ---
> >
>
> SGTM, thanks !
>
> Signed-off-by: Eric Dumazet <edumazet@google.com>
Nice work. I tested and verified it works with our packetdrill

Signed-off-by: Yuchung Cheng <ycheng@google.com>

>
> > v2: Changed logic to invalidate all retransmit queue frames if fastopen SYN
> > v3: Updated commit message to reflect actual solution in 3rd paragraph
> >
patchwork-bot+netdevbpf@kernel.org Dec. 15, 2020, 3:40 a.m. UTC | #3
Hello:

This patch was applied to netdev/net-next.git (refs/heads/master):

On Sat, 12 Dec 2020 12:31:24 -0800 you wrote:
> From: Alexander Duyck <alexanderduyck@fb.com>
> 
> There are cases where a fastopen SYN may trigger either a ICMP_TOOBIG
> message in the case of IPv6 or a fragmentation request in the case of
> IPv4. This results in the socket stalling for a second or more as it does
> not respond to the message by retransmitting the SYN frame.
> 
> [...]

Here is the summary with links:
  - [net-next,v3] tcp: Add logic to check for SYN w/ data in tcp_simple_retransmit
    https://git.kernel.org/netdev/net-next/c/c31b70c9968f

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
diff mbox series

Patch

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 9e8a6c1aa019..e44327a39a1f 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2688,7 +2688,22 @@  void tcp_simple_retransmit(struct sock *sk)
 	const struct inet_connection_sock *icsk = inet_csk(sk);
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct sk_buff *skb;
-	unsigned int mss = tcp_current_mss(sk);
+	int mss;
+
+	/* A fastopen SYN request is stored as two separate packets within
+	 * the retransmit queue, this is done by tcp_send_syn_data().
+	 * As a result simply checking the MSS of the frames in the queue
+	 * will not work for the SYN packet.
+	 *
+	 * Us being here is an indication of a path MTU issue so we can
+	 * assume that the fastopen SYN was lost and just mark all the
+	 * frames in the retransmit queue as lost. We will use an MSS of
+	 * -1 to mark all frames as lost, otherwise compute the current MSS.
+	 */
+	if (tp->syn_data && sk->sk_state == TCP_SYN_SENT)
+		mss = -1;
+	else
+		mss = tcp_current_mss(sk);
 
 	skb_rbtree_walk(skb, &sk->tcp_rtx_queue) {
 		if (tcp_skb_seglen(skb) > mss)