Hi,
I have also seen "connect(): connection timeout for basically ipv6 " behavior
for networking/stress/tcp testcase
i.e.
tcp6-multi-sameport12 0 TINFO : - Test duration is 120 [sec]
tcp6-multi-sameport12 0 TINFO : - Target number of the connection is 128
tcp6-multi-sameport12 0 TINFO : - Version of IP is IPv6
tcp6-multi-sameport12 0 TINFO : - Network delay is 600ms +/- 200ms
tcp6-multi-sameport12 0 TINFO : - IPsec [ ESP / tunnel ]
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
connect(): Connection timed out
This behavior is not due testcase but its due to kernel. So I did some code
changes in Linux kernel 3.2 i.e. mentioned below
In /net/ipv6/ip6_output.c
int ip6_xmit(struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
dst->dev, dst_output);
}
- net_dbg_ratelimited("IPv6: sending pkt_too_big to self\n");
skb->dev = dst->dev;
- icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
+ ipv6_local_error(sk, EMSGSIZE, fl6, mtu);
IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_FRAGFAILS);
kfree_skb(skb);
return -EMSGSIZE;
After above mentioned code changes all networking/stress/tcp testcase gets PASS.
Thanks and Regards,
Bhushan
-----Original Message-----
From: Hangbin Liu [mailto:[email protected]]
Sent: Monday, July 01, 2013 2:01 PM
To: Jan Stancek
Cc: ltp-list
Subject: Re: [LTP] [PATCH] networking/stress/tcp: make sure we will kill the
test case after timeout
Hi Jan,
The problem is the connection won't fail, but only need too much time. For
example:
On a guest: on cpu process, 2G memory
# for i in `seq 4000`; do ./ns-tcpclient -b -f 4 -S 10.66.87.216 -p 9999; echo
"The $i connection : `date`"; done The 1 connection : Mon Jul 1 04:03:05 EDT
2013 The 2 connection : Mon Jul 1 04:03:05 EDT 2013 The 3 connection : Mon Jul
1 04:03:05 EDT 2013 The 4 connection : Mon Jul 1 04:03:06 EDT 2013 <snip> The
19 connection : Mon Jul 1 04:04:46 EDT 2013 <snip> The 30 connection : Mon Jul
1 04:06:34 EDT 2013 The 31 connection : Mon Jul 1 04:07:52 EDT 2013 <snip>
The 45 connection : Mon Jul 1 04:10:11 EDT 2013 <snip> The 50 connection : Mon
Jul 1 04:12:05 EDT 2013
You can saw we took 9 minutes to set up 50 tcp connections.
But on a powerful physical machine,
The 1 connection : Mon Jul 1 04:21:08 EDT 2013 The 2 connection : Mon Jul 1
04:21:08 EDT 2013 <snip> The 499 connection : Mon Jul 1 04:24:35 EDT 2013 The
500 connection : Mon Jul 1 04:24:35 EDT 2013
It only use 3 minutest to set up 500 connections.
So we can't just continue on success, but should check the time after some
connections.
Thanks and Best regards
Hangbin Liu
2013/6/28 Jan Stancek <[email protected]>:
>
>
> ----- Original Message -----
>> From: "Hangbin Liu" <[email protected]>
>> To: "Jan Stancek" <[email protected]>
>> Sent: Friday, 28 June, 2013 11:30:02 AM
>> Subject: Re: [LTP] [PATCH] networking/stress/tcp: make sure we will
>> kill the test case after timeout
>>
>
> Hangbin,
>
> just fyi, your replies are not going to list.
>
>> Opps, My mistake. Didn't find there is a sleep 1...
>>
>> I will send another patch for this issue. How about check
>> $connection_num every 100 times. like
>
> I'd go with new variable to count each attempt, not just successful ones.
>
> Or something like this (on top your patch):
> Skip the sleep and continue on success, if we hit issue sleep and retry as
> watchdog allows.
>
> diff --git
> a/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport01
> b/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport01
> index 206a0d1..649c60b 100644
> ---
> a/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport01
> +++ b/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport0
> +++ 1
> @@ -351,9 +351,10 @@ while true ; do
> fi
> # Failed to start a client
> tst_resm TINFO "$connection_num seems the maximum
> number of the client"
> - break
> + else
> + connection_num=`expr $connection_num + 1`
> + continue;
> fi
> - connection_num=`expr $connection_num + 1`
> fi
>
> current_epoc=`date +%s`
>
> Regards,
> Jan
>
>>
>> ---
>> a/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport01
>> +++ b/testcases/network/stress/tcp/multi-sameport/tcp4-multi-sameport
>> +++ 01
>> @@ -340,7 +340,7 @@ connection_num=0
>> start_epoc=`date +%s`
>> while true ; do
>> # Making connections
>> - if [ $connection_num -lt $CONNECTION_TOTAL ]; then
>> + while [ $connection_num -lt $CONNECTION_TOTAL ]; do
>> # Run a client
>> ret=`$LTP_RSH $RHOST
>> ${LTPROOT}/testcases/bin/ns-tcpclient -b -f $IP_VER -S $lhost_addr -p
>> $server_port' ; ec
>> if [ $ret -ne 0 ]; then @@ -354,7 +354,12 @@ while
>> true ; do
>> break
>> fi
>> connection_num=`expr $connection_num + 1`
>> - fi
>> +
>> + # check timeout every 100 times
>> + if [ $(($connection_num % 100)) -eq 0 ];then
>> + break
>> + fi
>> + done
>>
>> current_epoc=`date +%s`
>> elapse_epoc=`expr $current_epoc - $start_epoc`
>>
>> 2013/6/28 Jan Stancek <[email protected]>:
>> >
>> >
>> >
>> >
>> > ----- Original Message -----
>> >> From: "Hangbin Liu" <[email protected]>
>> >> To: "Jan Stancek" <[email protected]>
>> >> Sent: Friday, 28 June, 2013 8:36:05 AM
>> >> Subject: Re: [LTP] [PATCH] networking/stress/tcp: make sure we
>> >> will kill the test case after timeout
>> >>
>> >> No, we can set up $CONNECTION_TOTAL connections as soon as
>> >> possible, and also transmit tcp data at the same time.
>> >
>> > Where does this happen?
>> >
>> > I'm looking at the while loop, and I see it creates 1 connection
>> > each iteration, then it sleeps for 1 second.
>> >
>> > Am I missing something here?
>> >
>> > while true ; do
>> > # Making connections
>> > if [ $connection_num -lt $CONNECTION_TOTAL ]; then
>> > # make new connection
>> > ...
>> > connection_num=`expr $connection_num + 1`
>> > fi
>> >
>> > # check time
>> > ...
>> > sleep 1
>> > done
>> >
>> > Regards,
>> > Jan
>>
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
