[Oisf-users] [EXT] Re: Packet Fanout on CentOS 7?

Tue Nov 19 08:29:40 UTC 2019

On Tue, Nov 19, 2019 at 2:11 AM Michał Purzyński
<michalpurzynski1 at gmail.com> wrote:
>
> Nah, any af_packet app will start even if the underlying hashing mechanism is not configured correctly. You will just have a hidden packet loss ;)
>

older versions use to do that  :)  , not 5.0 though (was fixed) , just
double checked - so it is not that for sure then...

> Can you attach your suricata.yaml and the exact command line you start Suricata with? It looks to me like problems with permissions.
>
> As a bonus can you try strace -o out <your suricata command goes here> and look for errors / access / permission messages in the resulting 'out' file?
>
> On Mon, Nov 18, 2019 at 1:44 PM Peter Manev <petermanev at gmail.com> wrote:
>>
>> On Mon, Nov 18, 2019 at 2:51 PM Cloherty, Sean E <scloherty at mitre.org> wrote:
>> >
>> > The NIC is a 10G dual port Intel Corporation 82599ES 10-Gigabit SFI/SFP+ Network Connection (rev 01).  The driver is ixgbe 5.6.3.  I’ve attached the output in a text file.  The script used to start it includes the following from the class  –
>> >
>> >
>> >
>>
>> I've seen similar errs in a couple of situations:
>> 1 - when there is another Suricata running on the same interface
>> 2 - when using XDP (hw mode) with the wrong/not correct firmware
>> version of the NIC
>>
>> From what I gather , it seems you don't have any the situation above
>> (if I am not mistaken)
>> Off note: For the cluster_qm part it is known that i40 can do the
>> symmetric hashing with the specified low entropy key  for ixgbe i am
>> not sure as i've seen both expected and unexpected results so ti is
>> something that needs to be further checked.
>> However there is  a bit more to that even if set up correct - it is
>> traffic dependent in some cases as well - it's been tracked here
>> https://redmine.openinfosecfoundation.org/issues/2725
>>
>> I dont have that much experience with CentOS but will be curious to
>> see if you have similar issue with CentOS 8
>>
>>
>> > ##### NIC #####
>> >
>> >
>> >
>> > ### ens1f1
>> >
>> >
>> >
>> > ifconfig ens1f1 down
>> >
>> > ifconfig ens1f1 up
>> >
>> >
>> >
>> > /usr/sbin/ethtool -L ens1f1 combined 16
>> >
>> > /usr/sbin/ethtool -K ens1f1 rxhash on
>> >
>> > /usr/sbin/ethtool -K ens1f1 ntuple on
>> >
>> > /usr/sbin/ethtool -X ens1f1 hkey 6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A:6D:5A equal 16
>> >
>> > /usr/sbin/ethtool -A ens1f1 rx off
>> >
>> > /usr/sbin/ethtool -C ens1f1 adaptive-rx off adaptive-tx off rx-usecs 125
>> >
>> > /usr/sbin/ethtool -G ens1f1 rx 1024
>> >
>> >
>> >
>> > for proto in tcp4 udp4 tcp6 udp6 ; do /usr/local/sbin/ethtool -N ens1f1 rx-flow-hash $proto sdfn ; done
>> >
>> >
>> >
>> > ##### START #####
>> >
>> > LD_PRELOAD="/usr/lib64/libtcmalloc_minimal.so.4" /usr/bin/suricata -c /etc/suricata/qm.yaml --af-packet=ens1f1 -vvv -D
>> >
>> >
>> >
>> >
>> >
>> > From: Peter Manev <petermanev at gmail.com>
>> > Sent: Sunday, November 17, 2019 1:59 AM
>> > To: Cloherty, Sean E <scloherty at mitre.org>
>> > Cc: oisf-users at lists.openinfosecfoundation.org
>> > Subject: [EXT] Re: [Oisf-users] Packet Fanout on CentOS 7?
>> >
>> >
>> >
>> > On 17 Nov 2019, at 02:46, Cloherty, Sean E <scloherty at mitre.org> wrote:
>> >
>> >
>> >
>> > 
>> >
>> > After taking the Advanced Deployment and Architecture class I was fired up with ideas for improvements in my own environment.  I want to use the cluster_qm mode and match worker/CPUs/RSS queues in CentOS 7.  Has anyone been able to get this working on CentOS 7 or should I start migrating to CentOS 8?  The kernel is 3.10.0-1062.4.1.el7.x86_64, running Suricata 5.0.0 and I was able to set the hash key and the hash functions correctly (I think) -
>> >
>> >
>> >
>> > RSS hash key:
>> >
>> > 6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a:6d:5a
>> >
>> > RSS hash function:
>> >
>> >     toeplitz: on
>> >
>> >     xor: off
>> >
>> >     crc32: off
>> >
>> >
>> >
>> >
>> >
>> > When I start Suricata all the messages look good until it gets to AFP when it gives me the message –
>> >
>> >
>> >
>> > 7/11/2019 -- 14:10:43 - <Notice> - all 16 packet processing threads, 4 management threads initialized, engine started.
>> >
>> > 7/11/2019 -- 14:10:43 - <Error> - [ERRCODE: SC_ERR_AFP_CREATE(190)] - Couldn't set fanout mode, error Invalid argument
>> >
>> > 7/11/2019 -- 14:10:43 - <Error> - [ERRCODE: SC_ERR_AFP_CREATE(190)] - Couldn't init AF_PACKET socket, fatal error
>> >
>> > 7/11/2019 -- 14:10:43 - <Error> - [ERRCODE: SC_ERR_FATAL(171)] - thread W#01-ens1f1 failed
>> >
>> >
>> >
>> >
>> >
>> > I’ve never tried it CentOS 7 but would recommend trying  8.
>> >
>> > What NIC is that ? (think you mentioned Intel but wasn’t sure what driver/model)
>> >
>> >
>> >
>> > Just to be in the safe side - “-T” (test) run passes ok too? (And there are no other Suri running on the same interface ?)
>> >
>> >
>> >
>> >
>> >
>> > I did take a look at the test script “can-i-use-afpacket-fanout” but my sensors have no internet connection so I am not able to use it via GO. Is there another way to run this if I downloading it manually ?
>> >
>> >
>> >
>> >
>> >
>> > Thanks,
>> >
>> >
>> >
>> > Sean
>> >
>> > _______________________________________________
>> > Suricata IDS Users mailing list: oisf-users at openinfosecfoundation.org
>> > Site: http://suricata-ids.org | Support: http://suricata-ids.org/support/
>> > List: https://lists.openinfosecfoundation.org/mailman/listinfo/oisf-users
>> >
>> > Conference: https://suricon.net
>> > Trainings: https://suricata-ids.org/training/
>>
>>
>>
>> --
>> Regards,
>> Peter Manev
>> _______________________________________________
>> Suricata IDS Users mailing list: oisf-users at openinfosecfoundation.org
>> Site: http://suricata-ids.org | Support: http://suricata-ids.org/support/
>> List: https://lists.openinfosecfoundation.org/mailman/listinfo/oisf-users
>>
>> Conference: https://suricon.net
>> Trainings: https://suricata-ids.org/training/

-- 
Regards,
Peter Manev