[net PATCH V2] net: fix divide by zero in tcp algorithm illinois

Subsystems: networking [general], networking [tcp], the rest

STALE5091d

5 messages, 4 authors, 2012-11-01 · open the first message on its own page

[net PATCH V2] net: fix divide by zero in tcp algorithm illinois

From: Jesper Dangaard Brouer <hidden>
Date: 2012-10-31 12:44:25

Reading TCP stats when using TCP Illinois congestion control algorithm
can cause a divide by zero kernel oops.

The division by zero occur in tcp_illinois_info() at:
 do_div(t, ca->cnt_rtt);
where ca->cnt_rtt can become zero (when rtt_reset is called)

Steps to Reproduce:
 1. Register tcp_illinois:
     # sysctl -w net.ipv4.tcp_congestion_control=illinois
 2. Monitor internal TCP information via command "ss -i"
     # watch -d ss -i
 3. Establish new TCP conn to machine

Either it fails at the initial conn, or else it needs to wait
for a loss or a reset.

This is only related to reading stats.  The function avg_delay() also
performs the same divide, but is guarded with a (ca->cnt_rtt > 0) at its
calling point in update_params().  Thus, simply fix tcp_illinois_info().

Function tcp_illinois_info() / get_info() is called without
socket lock.  Thus, eliminate any race condition on ca->cnt_rtt
by using a local stack variable.  Simply reuse info.tcpv_rttcnt,
as its already set to ca->cnt_rtt.
Function avg_delay() is not affected by this race condition, as
its called with the socket lock.

Cc: Petr Matousek <redacted>
Signed-off-by: Jesper Dangaard Brouer <redacted>

---
V2:
 Address Eric Dumazets input:
  - Save 2 bytes of stack, by using info.tcpv_rttcnt.
  - Help compiler, and define "u64 t" inside if() lexical scope.


 net/ipv4/tcp_illinois.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/tcp_illinois.c b/net/ipv4/tcp_illinois.c
index 813b43a..834857f 100644
--- a/net/ipv4/tcp_illinois.c
+++ b/net/ipv4/tcp_illinois.c
@@ -313,11 +313,13 @@ static void tcp_illinois_info(struct sock *sk, u32 ext,
 			.tcpv_rttcnt = ca->cnt_rtt,
 			.tcpv_minrtt = ca->base_rtt,
 		};
-		u64 t = ca->sum_rtt;
 
-		do_div(t, ca->cnt_rtt);
-		info.tcpv_rtt = t;
+		if (info.tcpv_rttcnt > 0) {
+			u64 t = ca->sum_rtt;
 
+			do_div(t, info.tcpv_rttcnt);
+			info.tcpv_rtt = t;
+		}
 		nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
 	}
 }

Re: [net PATCH V2] net: fix divide by zero in tcp algorithm illinois

From: Eric Dumazet <hidden>
Date: 2012-10-31 17:14:43

On Wed, 2012-10-31 at 13:45 +0100, Jesper Dangaard Brouer wrote:
Reading TCP stats when using TCP Illinois congestion control algorithm
can cause a divide by zero kernel oops.

The division by zero occur in tcp_illinois_info() at:
 do_div(t, ca->cnt_rtt);
where ca->cnt_rtt can become zero (when rtt_reset is called)
Cc: Petr Matousek <redacted>
Signed-off-by: Jesper Dangaard Brouer <redacted>

---
Acked-by: Eric Dumazet <redacted>

Re: [net PATCH V2] net: fix divide by zero in tcp algorithm illinois

From: Stephen Hemminger <hidden>
Date: 2012-10-31 20:48:45

On Wed, 31 Oct 2012 13:45:32 +0100
Jesper Dangaard Brouer [off-list ref] wrote:
quoted hunk
Reading TCP stats when using TCP Illinois congestion control algorithm
can cause a divide by zero kernel oops.

The division by zero occur in tcp_illinois_info() at:
 do_div(t, ca->cnt_rtt);
where ca->cnt_rtt can become zero (when rtt_reset is called)

Steps to Reproduce:
 1. Register tcp_illinois:
     # sysctl -w net.ipv4.tcp_congestion_control=illinois
 2. Monitor internal TCP information via command "ss -i"
     # watch -d ss -i
 3. Establish new TCP conn to machine

Either it fails at the initial conn, or else it needs to wait
for a loss or a reset.

This is only related to reading stats.  The function avg_delay() also
performs the same divide, but is guarded with a (ca->cnt_rtt > 0) at its
calling point in update_params().  Thus, simply fix tcp_illinois_info().

Function tcp_illinois_info() / get_info() is called without
socket lock.  Thus, eliminate any race condition on ca->cnt_rtt
by using a local stack variable.  Simply reuse info.tcpv_rttcnt,
as its already set to ca->cnt_rtt.
Function avg_delay() is not affected by this race condition, as
its called with the socket lock.

Cc: Petr Matousek <redacted>
Signed-off-by: Jesper Dangaard Brouer <redacted>

---
V2:
 Address Eric Dumazets input:
  - Save 2 bytes of stack, by using info.tcpv_rttcnt.
  - Help compiler, and define "u64 t" inside if() lexical scope.


 net/ipv4/tcp_illinois.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/tcp_illinois.c b/net/ipv4/tcp_illinois.c
index 813b43a..834857f 100644
--- a/net/ipv4/tcp_illinois.c
+++ b/net/ipv4/tcp_illinois.c
@@ -313,11 +313,13 @@ static void tcp_illinois_info(struct sock *sk, u32 ext,
 			.tcpv_rttcnt = ca->cnt_rtt,
 			.tcpv_minrtt = ca->base_rtt,
 		};
-		u64 t = ca->sum_rtt;
 
-		do_div(t, ca->cnt_rtt);
-		info.tcpv_rtt = t;
+		if (info.tcpv_rttcnt > 0) {
+			u64 t = ca->sum_rtt;
 
+			do_div(t, info.tcpv_rttcnt);
+			info.tcpv_rtt = t;
+		}
 		nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);
 	}
 }
Acked-by: Stephen Hemminger <redacted>

Re: [net PATCH V2] net: fix divide by zero in tcp algorithm illinois

From: Jesper Dangaard Brouer <hidden>
Date: 2012-10-31 22:16:03

On Wed, 2012-10-31 at 13:48 -0700, Stephen Hemminger wrote:
On Wed, 31 Oct 2012 13:45:32 +0100
Jesper Dangaard Brouer [off-list ref] wrote:
quoted
Reading TCP stats when using TCP Illinois congestion control algorithm
can cause a divide by zero kernel oops.
[...]
Acked-by: Stephen Hemminger <redacted>
When DaveM is without power, people even have/get time to review and ack
my patches ;-). Thanks Stephen and Eric.

--Jesper

Re: [net PATCH V2] net: fix divide by zero in tcp algorithm illinois

From: David Miller <davem@davemloft.net>
Date: 2012-11-01 15:56:34

From: Eric Dumazet <redacted>
Date: Wed, 31 Oct 2012 18:14:38 +0100
On Wed, 2012-10-31 at 13:45 +0100, Jesper Dangaard Brouer wrote:
quoted
Reading TCP stats when using TCP Illinois congestion control algorithm
can cause a divide by zero kernel oops.

The division by zero occur in tcp_illinois_info() at:
 do_div(t, ca->cnt_rtt);
where ca->cnt_rtt can become zero (when rtt_reset is called)
quoted
Cc: Petr Matousek <redacted>
Signed-off-by: Jesper Dangaard Brouer <redacted>

---
Acked-by: Eric Dumazet <redacted>
Applied, thanks everyone.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help