Robin,
I have been trying the same test on OSOL. I have finally narrowed it
down to this set of conditions:
with a data_list of '1 1 1' and proto_list of just 'UDP' and with MAXQ.auto
test running just 'rx' traffic , the first MAXQ.auto test for 1byte UDP Traffic
in 'rx' path alone on the SUT seems to fail ..( Get a msg saying 'Finished: 26
out of 30 sessions failed'). Although eventually the test seems to pass (as in
the 'client_alive' test passes) .. This is with 1G igb driver as the client ...
Now though the test passes, is this THE real problem that i need to be
focussing at ? or is it OK for MAXQ.auto test to fail sometimes ?
Thanks
Som
________________________________
From: tian robin luo <robin....@sun.com>
To: Gireesh D. Nagabhushana <dngire...@gmail.com>
Cc: Somnath kotur <so...@yahoo.com>; solaris driver
<driver-discuss@opensolaris.org>
Sent: Wed, November 4, 2009 3:20:41 PM
Subject: Re: Unpredictable failure on NICDRV test15
It doesn't unplumb nic driver in the netstress test, this failure is
caused by the interface hung after transferring UDP packets with small
size payload for about 7000 seconds. All rsh or ping operations through
this interface failed. This netstress is to verify the driver can work
well under heave traffic for a long time.
Thanks,
Robin
Gireesh D. Nagabhushana:
These lines are from your log file.
>
>>stderr| + expr 7200 + 30
>>stderr| + sleep 7230
>>stderr| + [ X10.20.43.0 = X ]
>>stderr| + touch /tmp/netperf.sut /tmp/netperf.cnt /tmp/netperf.summary
>>stderr| + [ Xbi != Xrx ]
>>stderr| + cat /tmp/netperf.sut0 /tmp/netperf.sut1 /tmp/netperf.sut10
>/tmp/netperf.sut11 /tmp/netperf.sut12 /tmp/netperf.sut13
>/tmp/netperf.sut14 /tmp/netperf.sut15 /tmp/netperf.sut16
>/tmp/netperf.sut17 /tmp/netperf.sut18 /tmp/netperf.sut19
>/tmp/netperf.sut2 /tmp/netperf.sut20 /tmp/netperf.sut21
>/tmp/netperf.sut22 /tmp/netperf.sut23 /tmp/netperf.sut24
>/tmp/netperf.sut25 /tmp/netperf.sut26 /tmp/netperf.sut27
>/tmp/netperf.sut28 /tmp/netperf.sut29 /tmp/netperf.sut3
>/tmp/netperf.sut4 /tmp/netperf.sut5 /tmp/netperf.sut6 /tmp/netperf.sut7
>/tmp/netperf.sut8 /tmp/netperf.sut9
>>stderr| + awk {x+=$5}END{printf"%d",x}
>>stderr| + rsh -n 10.20.40.140 ls -l /tmp/netperf.10.20.40.140.cnt?*
>>stderr| ::ffff:10.20.40.140:
>Connection timed out
>>stderr| cli_wait=0
>>stderr| + [ 0 -eq 0 ]
>>stdout| waiting 60 seconds for client results
>>stderr| + echo waiting 60 seconds for client results
>>stderr| + sleep 60
>>stderr| + [ Xbi != Xtx ]
>>stderr| + rsh -n 10.20.40.140 cat /tmp/netperf.10.20.40.140.cnt?*
>>stdout| : FAILED
>>stderr| ::ffff:10.20.40.140:
>Connection timed out
>>stderr| + verify_rsh 1
>>stderr| + [ 1 -ne 0 ]
>>stderr| + echo : FAILED
>
>>After starting netperf clients (and after waiting for 7230 seconds),
>test tried "ls -l /tmp/netperf.10.20.40.140.cnt?*" in remote through
>rsh. This failed (connection timed out). Try running similar test
>manually on driver.
>
>>_fini won't be called during unplumb. Here are the lines from
>documentation:
>_fini() prepares a loadable module for unloading.
>It
>is called when the system wants to unload a module. If the module
>determines that it can be unloaded, then _fini() >returns the value returned
>by mod_remove(9F).
>Upon successful return from _fini() no other routine
>in the module will be called before _init() is
>called.
>
>>-- Gireesh
>
>
>On Wed, Nov 4, 2009 at 11:04 AM, Somnath
>kotur <so...@yahoo.com> wrote:
>
>Hi Robin,
>>>> I know this might be a long shot ,but am currently facing
>>some unpredictable problems while running the above test,and so far the
>>only success ive had is that i think i have nailed it down to the UDP
>>Stream portion of the test
>>>>1) This is where the test fails somewhere towards the end , (dumping
>>snippet of log below, have also attached the same)
>>
>>>>stdout| MAXQ.auto 30 sessions UDP_STREAM testing
>>>>stdout| Payload is 65000, runtime: 7000
>>>>stdout| Enable promiscuous mode 10 times on local host
>>>>stdout| Enable promiscuous mode 10 times on remote host
>>>>stdout| Enable promiscuous mode 10 times on local host
>>>>stdout| Enable promiscuous mode 10 times on remote host
>>>>stdout| MAXQ.auto 30 sessions UDP_STREAM testing
>>>>stdout| Payload is 1460, runtime: 7000
>>>>stdout| MAXQ.auto 30 sessions UDP_STREAM testing
>>>>stdout| Payload is 1, runtime: 7000
>>>>stdout| Killing snoop processes...
>>>>stderr| /opt/SUNWstc-nicdrv/tests/stress/netstress/runme[142]: 9016
>>Terminated
>>>>stdout| LAST dmesg:
>>>>stdout| no answer from 10.20.40.140
>>>>stdout| no answer from 10.20.40.140
>>>>stdout| no answer from 10.20.40.140
>>
>>>>When i modifed the runtime from 7000 to 3600s the test seemed to pass
>>?? I modified the script with 'set -x ' option enabled on all the
>>important scripts like 'runme' under the 'netstress' folder and
>>MAXQ.auto functions and i am attaching the results file which is really
>>quite huge, would you be able to take a look at it and tell me where
>>/what is the failure point?
>>
>>>> i also see in my /var/adm/messages that at times my driver's _fini()
>>is called , which makes me wonder if the test randomly even unplumb the
>>driver or is this something i should be worried about ??
>>
>>
>>>>Thanks
>>>>Som
>>
>>
>>
>>
>
_______________________________________________
driver-discuss mailing list
driver-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/driver-discuss
