[PATCH net] tcp/dccp: block bh before arming time_wait timer

[PATCH net] tcp/dccp: block bh before arming time_wait timer

[Eric Dumazet]

From: Eric Dumazet <edumazet@google.com>

Maciej Żenczykowski reported some panics in tcp_twsk_destructor()
that might be caused by the following bug.

timewait timer is pinned to the cpu, because we want to transition
timwewait refcount from 0 to 4 in one go, once everything has been
initialized.

At the time commit ed2e92394589 ("tcp/dccp: fix timewait races in timer
handling") was merged, TCP was always running from BH habdler.

After commit 5413d1babe8f ("net: do not block BH while processing
socket backlog") we definitely can run tcp_time_wait() from process
context.

We need to block BH in the critical section so that the pinned timer
has still its purpose.

This bug is more likely to happen under stress and when very small RTO
are used in datacenter flows.

Fixes: 5413d1babe8f ("net: do not block BH while processing socket backlog")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Maciej Żenczykowski <maze@google.com>
---
 net/dccp/minisocks.c     |    6 ++++++
 net/ipv4/tcp_minisocks.c |    6 ++++++
 2 files changed, 12 insertions(+)


diff --git a/net/dccp/minisocks.c b/net/dccp/minisocks.c
index abd07a443219853b022bef41cb072e90ff8f07f0..178bb9833311f83205317b07fe64cb2e45a9f734 100644
--- a/net/dccp/minisocks.c
+++ b/net/dccp/minisocks.c
@@ -57,10 +57,16 @@ void dccp_time_wait(struct sock *sk, int state, int timeo)
 		if (state == DCCP_TIME_WAIT)
 			timeo = DCCP_TIMEWAIT_LEN;
 
+		/* tw_timer is pinned, so we need to make sure BH are disabled
+		 * in following section, otherwise timer handler could run before
+		 * we complete the initialization.
+		 */
+		local_bh_disable();
 		inet_twsk_schedule(tw, timeo);
 		/* Linkage updates. */
 		__inet_twsk_hashdance(tw, sk, &dccp_hashinfo);
 		inet_twsk_put(tw);
+		local_bh_enable();
 	} else {
 		/* Sorry, if we're out of memory, just CLOSE this
 		 * socket up.  We've got bigger problems than
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index e36eff0403f4e80c4f7291a70614f40125652133..b079b619b60ca577d5ef20a5065fce87acecd96c 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -310,10 +310,16 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 		if (state == TCP_TIME_WAIT)
 			timeo = TCP_TIMEWAIT_LEN;
 
+		/* tw_timer is pinned, so we need to make sure BH are disabled
+		 * in following section, otherwise timer handler could run before
+		 * we complete the initialization.
+		 */
+		local_bh_disable();
 		inet_twsk_schedule(tw, timeo);
 		/* Linkage updates. */
 		__inet_twsk_hashdance(tw, sk, &tcp_hashinfo);
 		inet_twsk_put(tw);
+		local_bh_enable();
 	} else {
 		/* Sorry, if we're out of memory, just CLOSE this
 		 * socket up.  We've got bigger problems than

Re: [PATCH net] tcp/dccp: block bh before arming time_wait timer

[Maciej Żenczykowski]

Acked-by: Maciej Żenczykowski <maze@google.com>

Re: [PATCH net] tcp/dccp: block bh before arming time_wait timer

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Fri, 01 Dec 2017 10:06:56 -0800

> From: Eric Dumazet <edumazet@google.com>
> 
> Maciej Żenczykowski reported some panics in tcp_twsk_destructor()
> that might be caused by the following bug.
> 
> timewait timer is pinned to the cpu, because we want to transition
> timwewait refcount from 0 to 4 in one go, once everything has been
> initialized.
> 
> At the time commit ed2e92394589 ("tcp/dccp: fix timewait races in timer
> handling") was merged, TCP was always running from BH habdler.
> 
> After commit 5413d1babe8f ("net: do not block BH while processing
> socket backlog") we definitely can run tcp_time_wait() from process
> context.
> 
> We need to block BH in the critical section so that the pinned timer
> has still its purpose.
> 
> This bug is more likely to happen under stress and when very small RTO
> are used in datacenter flows.
> 
> Fixes: 5413d1babe8f ("net: do not block BH while processing socket backlog")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Reported-by: Maciej Żenczykowski <maze@google.com>

Applied and queued up for -stable, thanks Eric.

Re: [PATCH net] tcp/dccp: block bh before arming time_wait timer

[Eric Dumazet]

On Fri, 2017-12-01 at 15:12 -0500, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Fri, 01 Dec 2017 10:06:56 -0800
> 
> > From: Eric Dumazet <edumazet@google.com>
> > 
> > Maciej Żenczykowski reported some panics in tcp_twsk_destructor()
> > that might be caused by the following bug.
> > 
> > timewait timer is pinned to the cpu, because we want to transition
> > timwewait refcount from 0 to 4 in one go, once everything has been
> > initialized.
> > 
> > At the time commit ed2e92394589 ("tcp/dccp: fix timewait races in
> timer
> > handling") was merged, TCP was always running from BH habdler.
> > 
> > After commit 5413d1babe8f ("net: do not block BH while processing
> > socket backlog") we definitely can run tcp_time_wait() from process
> > context.
> > 
> > We need to block BH in the critical section so that the pinned
> timer
> > has still its purpose.
> > 
> > This bug is more likely to happen under stress and when very small
> RTO
> > are used in datacenter flows.
> > 
> > Fixes: 5413d1babe8f ("net: do not block BH while processing socket
> backlog")
> > Signed-off-by: Eric Dumazet <edumazet@google.com>
> > Reported-by: Maciej Żenczykowski <maze@google.com>
> 
> Applied and queued up for -stable, thanks Eric.

It just occurred to me that we can now revert 614bdd4d6e61d26
("tcp: must block bh in __inet_twsk_hashdance()")