problems with suricata 6.0.0

bmeeks

The 5.0.3 binary is installed and running, but with the 6.0.0 package GUI code. So the few small GUI bug fixes are still in place, but the suricata binary is reverted to the older version. Will let this combo run for a while and make sure alerting still works.

Tested stopping an interface with rules defined and it worked properly. Suricata immediately shutdown and the process disappeared as it should. No duplicate process created when restarting.

Update: tested alerts after 30 minutes of runtime and still working fine.

kiokoman

with 5.0.3 or with 6.0.0 ?
i'm still doing some test with 6 and i have a long run of alert if i disable all Logging Settings
nope still not working, false alarm
i tried out of curiosity inline mode against igb0 and i had the same problem of the other thread, wan goes offline as soon as truss show "nanosleep"
inline stop workin, nothing is logged anymore and kill the wan
legacy mode stop working and nothing is logged anymore but wan still working

bmeeks

@kiokoman said in problems with suricata 6.0.0:

with 5.0.3 or with 6.0.0 ?
i'm still doing some test with 6 and i have a long run of alert if i disable all Logging Settings
nope still not working, false alarm
i tried out of curiosity inline mode against igb0 and i had the same problem of the other thread, wan goes offline as soon as truss show "nanosleep"

I installed 5.0.3 on my test VM. I have my own private pfSense packages repository server, so I can build any variation of binary and GUI package versions I need to. So I built the 5.0.3 binary and paired it with the 6.0.0 PHP GUI code. I wanted to prove to myself the issues were within the 6.0.0 binary, and I'm now pretty sure they are.

The GUI serves no function with regards to blocking and alerting. All of that happens in the binary piece. The GUI simply generates the suricata.yaml conf file for the interface and copies the rules file to a specified directory. It has nothing to do with detection and actual alerting.

kiokoman

yeah i'm with you, the problem is the binary

bmeeks

The upstream Suricata team has made many changes to the binary after deciding to require Rust and rewrite major parts of Suricata to use Rust. That started with the 5.x branch and has greatly intensified in 6.0.0. There were also a bunch of other changes in the source code for 6.0.0.

Tomorrow I will build the 5.0.4 binary and see how it performs. Mainly to see if it contains the same "no alerts after a time" bug.

bmeeks

@kiokoman said in problems with suricata 6.0.0:

with 5.0.3 or with 6.0.0 ?
i'm still doing some test with 6 and i have a long run of alert if i disable all Logging Settings
nope still not working, false alarm
i tried out of curiosity inline mode against igb0 and i had the same problem of the other thread, wan goes offline as soon as truss show "nanosleep"
inline stop workin, nothing is logged anymore and kill the wan
legacy mode stop working and nothing is logged anymore but wan still working

One more test you can try, if you have not already, is to disable all blocking and run in just plain IDS mode. See if alerts still cease being logged. I suspect they will, but would be nice to be sure.

kiokoman

IDS mode 5-minute running and I'm still receiving alerts
idk how long it will last but it's a new record
i'll tell you in a few hours, 2:30am here, it's time to hit the sack ..

same behavior 5 minutes only

bmeeks

My 5.0.3 binary VM with the 6.0.0 GUI is still working fine after running all night. So the problems are definitely within the new 6.0.0 binary.

I will test the newer 5.0.4 binary today to see if it has the same issues. Depending on what I find there, I will work with the pfSense team to pull down the 6.0.0 binary and replace it with either 5.0.4 (if it works okay) or 5.0.3.

I have already closed the pull request that was sitting in the pfSense-2.4.5_p1 RELEASE branch, so we will not be deploying the 6.0.0 binary there. The RELEASE branch is still on the older 5.0.3 package (both binary and GUI).

kiokoman

nice ! .. my grafana maps is empty without suricata

bmeeks

So far my testing with the 5.0.4 binary has been good. Still getting alerts after an hour. Will let it go a bit longer in Legacy Mode blocking, then I will test Inline IPS Mode.

Have traded emails with a member of the pfSense team. The 6.0.0 binary will be reverted shortly to 5.0.3 for now. Depending on how my 5.0.4 testing goes, we will either stay at 5.0.3 or update to 5.0.4. But we are not going to push out 6.0.0 for now. Will wait at least until 6.0.1 is released upstream.

kiokoman

fair enough
if you need me to do some testing when the time comes, i will be happy to help.

bmeeks

@kiokoman said in problems with suricata 6.0.0:

fair enough
if you need me to do some testing when the time comes, i will be happy to help.

I had done testing with 6.0.0 before I submitted it, but admittedly just a very quick verification of basic functionality. I did not let it run a long time and thus I did not catch the fact it stopped alerting after a period.

But times like this are why we push things to the DEVEL snapshot branch first so we can find issues.

bmeeks

Another update:

The 5.0.4 binary is behaving just fine. No loss of alerting after several hours. Now running with Inline IPS Mode using the netmap device on the e1000 virtual NIC, so using the em driver in FreeBSD.

Before I installed the 5.0.4 binary, I updated the pfSense virtual machine to the latest 2.5 snapshot from this morning.

Looks like the 6.0.0 Suricata binary has issues, so I'll abandon it for now in the package update and use the 5.0.4 binary instead. That is the latest version for the 5.x Suricata branch.

kiokoman

yup I saw it here https://www.freshports.org/security/suricata
5.0.4 seem to be the latest
only a few resolved bug https://redmine.openinfosecfoundation.org/versions/149

kiokoman

@bmeeks i noticed a problem with pppoe and suricata

i had a loss of alerting after this event, i had to restart suricata on that interface, but i suppose i can try to bind suricata to igb0 instead of the pppoe interface
5.0.4 legacy mode

Nov 12 10:34:25 pfSense ppp[16779]: [wan] IFACE: Removing IPv4 address from pppoe0 failed(IGNORING for now. This should be only for PPPoE friendly!): Can't assign requested address
Nov 12 10:34:25 pfSense ppp[16779]: [wan] IFACE: Down event
Nov 12 10:34:25 pfSense ppp[16779]: [wan] IFACE: Rename interface pppoe0 to pppoe0
Nov 12 10:34:26 pfSense php-fpm[800]: /rc.newwanip: rc.newwanip: Info: starting on igb0.
Nov 12 10:34:26 pfSense php-fpm[800]: /rc.newwanip: rc.newwanip: on (IP address: 192.168.2.2) (interface: MODEM[opt2]) (real interface: igb0).
Nov 12 10:34:26 pfSense ppp[11122]: waiting for process 16779 to die...
Nov 12 10:34:26 pfSense php-fpm[28978]: /rc.dyndns.update: Dynamic DNS: updatedns() starting
Nov 12 10:34:27 pfSense ppp[11122]: waiting for process 16779 to die...
Nov 12 10:34:27 pfSense ppp[16779]: [wan] Bundle: Shutdown
Nov 12 10:34:27 pfSense kernel: pppoe0: promiscuous mode disabled
Nov 12 10:34:27 pfSense suricata[28601]: [100652] <Error> -- [ERRCODE:SC_ERR_PCAP_DISPATCH(20)] - error code -1 The interface disappeared 
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Nov 12 10:34:27 pfSense ppp[16779]: [wan_link0] Link: Shutdown
Nov 12 10:34:27 pfSense ppp[16779]: process 16779 terminated
Nov 12 10:34:28 pfSense ppp[11122]: web: web is not running
Nov 12 10:34:28 pfSense ppp[11122]: [wan] Bundle: Interface ng0 created
Nov 12 10:34:28 pfSense ppp[11122]: [wan_link0] Link: OPEN event
Nov 12 10:34:28 pfSense kernel: ng0: changing name to 'pppoe0'
Nov 12 10:34:28 pfSense ppp[11122]: [wan_link0] LCP: Open event
Nov 12 10:34:28 pfSense ppp[11122]: [wan_link0] LCP: state change Initial --> Starting
Nov 12 10:34:28 pfSense ppp[11122]: [wan_link0] LCP: LayerStart
Nov 12 10:34:28 pfSense ppp[11122]: [wan_link0] PPPoE: Connecting to ''

bmeeks

The PPPoE interface is a virtual thing in FreeBSD. It has always been a little flaky. In the past, Suricata would not work at all with a PPPoE interface. Some changes were made to accomodate PPPoE on FreeBSD a few years ago. Since then it has worked fairly well so far as I know.

In FreeBSD-12, the entire networking subsytem underwent a big change with the move to iflib for NIC drivers. I assume, based on the forum sub-topic where this thread is located, that you are running pfSense-2.5. That is a FreeBSD-12 OS, and so it contains the new iflib stuff. I've seen some weirdness with the new API.

kiokoman

yes, it's pfSense 2.5