Is anyone using the following attributes? How are you using them? I
believe some of these have been deprecated.
&rotate_interval
&rotate_size
&mergeable
&synchronize (I think there was a post earlier last month about this one)
&persistent
&group
&add_func
&delete_func
&encrypt (applying this to a file causes Bro to "elegantly terminate" for me)
bro -Ci eth0 -e 'global f1: file = open("f.out") &encrypt'
What is the purpose of the unknown port type? And why do they only
range from 0-255? Compare the results of the following commands.
bro -e 'print 0/unknown; print 255/unknown;'
bro -e 'print 0/unknown; print 255/unknown; print 256/unknown;'
How are ports flattened? See the results of the following command.
bro -e 'print 0/udp; print |0/udp|; print |32/tcp|; print |11/tcp|;
print |132/unknown|; print 132/unknown;'
It should mostly be used internally to signify an uninitialized/invalid transport protocol. I don’t think it’s common for that to actually be exposed to the scripting-layer for practical usage.
And why do they only
range from 0-255? Compare the results of the following commands.
bro -e 'print 0/unknown; print 255/unknown;'
bro -e 'print 0/unknown; print 255/unknown; print 256/unknown;’
Likely arbitrary and just due to copy-paste of the code that parses ICMP port literals (for ICMP, Bro uses 0-255 to correspond to the Type/Code values).
How are ports flattened? See the results of the following command.
bro -e 'print 0/udp; print |0/udp|; print |32/tcp|; print |11/tcp|;
print |132/unknown|; print 132/unknown;'
Internally, a port is a single uint64 with some of the high-bits set to indicate which port-space it belongs to. You’re seeing that value here. E.g.
bro \-e 'print |32/tcp| == 0x10000 \+ 32'
T
bro -e 'print |37/udp| == 0x20000 + 37’
T
It's pretty new (though maybe it's actually where your questiosns are
coming from
To add a bit to that:
&rotate_interval
&rotate_size
This used to be primary log rotation mechanism before we switched to
the new logging system/format. I've been wondering if we should just
remove these attributes.
&mergeable
&synchronize (I think there was a post earlier last month about this one)
&persistent
These are going to go away, but we aren't there yet. We may start
deprecating them with the next release, which is scheduled to ship
with a first version of their replacement, the new Broker library.
&group
A bit of an obscure feature, originally added to toggle selected sets
of analysis dynamically from BroControl. Don't think that's used
anywhere and I'm inclined to remove it.
&add_func
&delete_func
These aren't used very often, but can be useful in individual cases.
&encrypt (applying this to a file causes Bro to "elegantly terminate" for me)
bro -Ci eth0 -e 'global f1: file = open("f.out") &encrypt'
Another relict from old-style logging, although the new framework
doesn't have any equivalent functionality yet.
Mind filing a ticket for the crash? We should either fix it or remove
the attribute.
That page is exactly where my questions are coming from. I tried using each of the attributes in a few toy scripts and was wondering if people are using them in production as I could not find some of them used in base or policy. Thanks for the insight, Robin.
