Thread (14 messages) flat view 14 messages, 5 authors, 2012-03-14

Re: TCP syn flood handling regression

From: Eric Dumazet <hidden>
Date: 2012-03-10 17:44:57

Le samedi 10 mars 2012 à 04:27 -0800, Simon Kirby a écrit :
Hello!

A typical port 80 SYN flood started up to one of our clusters, but this
time, it didn't work so well. Legitimate connections and trying to fetch
server-status via localhost would hang for ~30 seconds before responding,
even though though the box had plenty of spare cycles. An strace of all
Apache processes showed quite a bit of sleeping in accept4().

This was with 3.2.9, so I went back in kernel builds and found that 3.1
and 3.0 were also broken, while 2.6.39 works as I remember -- when syn
cookies are enabled, everything just works and is fast. The DoS kept up,
so I was able to feed a bit to a node to do some bisection.

Of course, the DoS stopped literally seconds before the last bisection
test, but I got it down to:

# good: [0e734419923bd8e599858f8fc196c7804bb85564] ipv4: Use inet_csk_route_child_sock() in DCCP and TCP.
# bad: [ea4fc0d6193ff56fcef39b0d2210d402a7acb5f0] ipv4: Don't use rt->rt_{src,dst} in ip_queue_xmit().

...leaving ea4fc0d6193ff56fcef39b0d2210d402a7acb5f0 and
d9d8da805dcb503ef8ee49918a94d49085060f23 as culprits.

I've stared at them but can't see what could be doing this.
Hi Simon, thanks a lot for this report.

This sounds as a SYNCOOKIE side effect

I could reproduce this with a typical DOS on ssh port.

Yes, it seems first messages are not sent at all, only after several tcp
timeouts :

Here my sshd write() its first frame (23 bytes) right after the 3WHS,
but we can see first frame leaves the machine 17 seconds later :

17:42:49.958432 IP 10.170.237.2.37550 > 192.168.20.110.22: S 3066771530:3066771530(0) win 14600 <mss 1460,sackOK,timestamp 19981015 0,nop,wscale 6>
17:42:49.958438 IP 192.168.20.110.22 > 10.170.237.2.37550: S 881027529:881027529(0) ack 3066771531 win 14480 <mss 1460,sackOK,timestamp 1550353622 19981015,nop,wscale 4>
17:42:50.137163 IP 10.170.237.2.37550 > 192.168.20.110.22: . ack 1 win 229 <nop,nop,timestamp 19981060 1550353622>

17:43:07.557096 IP 192.168.20.110.22 > 10.170.237.2.37550: P 1:24(23) ack 1 win 913 <nop,nop,timestamp 1550371264 19981060>
17:43:07.757096 IP 10.170.237.2.37550 > 192.168.20.110.22: . ack 24 win 229 <nop,nop,timestamp 19985459 1550371264>
17:43:07.936756 IP 10.170.237.2.37550 > 192.168.20.110.22: P 1:40(39) ack 24 win 229 <nop,nop,timestamp 19985459 1550371264>
17:43:07.936769 IP 192.168.20.110.22 > 10.170.237.2.37550: . ack 40 win 913 <nop,nop,timestamp 1550371643 19985459>
17:43:07.937384 IP 192.168.20.110.22 > 10.170.237.2.37550: P 24:664(640) ack 40 win 913 <nop,nop,timestamp 1550371644 19985459>
17:43:08.156252 IP 10.170.237.2.37550 > 192.168.20.110.22: P 40:1184(1144) ack 24 win 229 <nop,nop,timestamp 19985554 1550371643>

To check that its indeed the tcp retransmits that trigger and eventually send the frame :

$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      23                        ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(on,700ms,0) ino:11067765 sk:f3564780
	 mem:(r0,w2240,f1856,t0) ts sack bic wscale:6,4 rto:1209 rtt:403/201.5 cwnd:10 ssthresh:5 send 287.4Kbps rcv_space:14600
$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      23                        ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(on,1.529ms,1) ino:11067765 sk:f3564780
	 mem:(r0,w2240,f1856,t0) ts sack bic wscale:6,4 rto:2418 rtt:403/201.5 ssthresh:5 send 57.5Kbps rcv_space:14600
$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      23                        ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(on,4.005ms,2) ino:11067765 sk:f3564780
	 mem:(r0,w2240,f1856,t0) ts sack bic wscale:6,4 rto:4836 rtt:403/201.5 ssthresh:5 send 57.5Kbps rcv_space:14600
$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      23                        ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(on,1.773ms,2) ino:11067765 sk:f3564780
	 mem:(r0,w2240,f1856,t0) ts sack bic wscale:6,4 rto:4836 rtt:403/201.5 ssthresh:5 send 57.5Kbps rcv_space:14600
$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      23                        ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(on,8.574ms,3) ino:11067765 sk:f3564780
	 mem:(r0,w2240,f1856,t0) ts sack bic wscale:6,4 rto:9672 rtt:403/201.5 ssthresh:5 send 57.5Kbps rcv_space:14600
$ ss -emoi dst 10.170.237.2
State       Recv-Q Send-Q                            Local Address:Port                                Peer Address:Port   
ESTAB       0      0                         ::ffff:192.168.20.110:ssh                          ::ffff:10.170.237.2:37604    timer:(keepalive,48min,0) ino:11067765 sk:f3564780
	 mem:(r0,w0,f0,t0) ts sack bic wscale:6,4 rto:806 rtt:263.875/105.5 ato:40 cwnd:6 ssthresh:5 send 263.4Kbps rcv_rtt:181 rcv_space:14600
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help