[pve-devel] pve-firewall: default loglevel behaviour change

[Thomas Lamprecht]

On 5/16/19 1:18 AM, Alexandre DERUMIER wrote:
> Thanks !
> 
> Also,
> about new log ratelimit
> https://git.proxmox.com/?p=pve-firewall.git;a=blobdiff;f=src/PVE/Firewall.pm;h=48e630004bf424c2db578aad3e60eb61db42f8e6;hp=ba1cb5f5924af094fe647f98d40e2703eca96ff4;hb=cc37e0005d0617e405883fd12b80faf65cedb153;hpb=95ed90a9a8418173ab7cac2d0c1f6ddd04c376cc
> 
> it's seem that the default is now
> +my $global_log_ratelimit = '--limit 1/sec';
> 
> 
> (If don't have the option "log_ratelimit" in cluster.fw)
> 
> 
>  
> +my $set_global_log_ratelimit = sub {
> +    my $cluster_opts = shift;
> +
> +    $global_log_ratelimit = '--limit 1/sec';
> +    if (defined(my $log_rlimit = $cluster_opts->{log_ratelimit})) {
> +       my $ll_format = $cluster_option_properties->{log_ratelimit}->{format};
> +       my $limit = PVE::JSONSchema::parse_property_string($ll_format, $log_rlimit);
> +
> +       if ($limit->{enable}) {
> +           if (my $rate = $limit->{rate}) {
> +               $global_log_ratelimit = "--limit $rate";
> +           }
> +           if (my $burst = $limit->{burst}) {
> +               $global_log_ratelimit .= " --limit-burst $burst";
> +           }
> +       } else {
> +           $global_log_ratelimit = undef;
> +       }
> +    }
> +};
> 
> 
> but it's correctly disabled with
> "log_ratelimit: enable=0" in cluster.fw
> 
> I think the behaviour should be also the same when option is not defined.

When we added this we had some discussion about the sensible default, IIRC,
I initially was for no ratelimit per default, but Christian argued that with
the bucket burst limit we have a sensible balance between spamming the logs
(if something is off, and a lot of packets get generated) but still seeing
bursts. FYI, you can also just set log_ratelimit to 0 (enabled is the default
key), but not sure if we want tot change this to opt-out, I mean either way
one only needs to change it once per cluster...

> 
> 
> ----- Mail original -----
> De: "Christian Ebner" <c.ebner at proxmox.com>
> À: "pve-devel" <pve-devel at pve.proxmox.com>, "aderumier" <aderumier at odiso.com>
> Envoyé: Mercredi 15 Mai 2019 17:03:13
> Objet: Re: [pve-devel] pve-firewall: default loglevel behaviour change
> 
> Hi Alexandre, 
> 
> you are right, logging the packets passing the firewall bridge seems a bit overkill and redundant. 
> 
> Will send a patch to fix this. 
> 
>> On May 15, 2019 at 4:08 PM Alexandre DERUMIER <aderumier at odiso.com> wrote: 
>>
>>
>> Hi, 
>>
>> since this commit 
>>
>>
>> https://git.proxmox.com/?p=pve-firewall.git;a=blobdiff;f=src/PVE/Firewall.pm;h=ccc5d7ffb62f1ef4c4d59e363192f512c88742c9;hp=6ac303831a62f630d437ef0e0508decd2b72f5ac;hb=3489f8a2b9597201fe8e42fa5832507f96169619;hpb=33efd363ab32d3f8f6a9c49e481cb62a8da40b5d 
>>
>> @@ -2341,10 +2348,10 @@ sub generate_tap_rules_direction { 
>> # plug the tap chain to bridge chain 
>> if ($direction eq 'IN') { 
>> ruleset_addrule($ruleset, "PVEFW-FWBR-IN", 
>> - "-m physdev --physdev-is-bridged --physdev-out $iface", "-j $tapchain"); 
>> + "-m physdev --physdev-is-bridged --physdev-out $iface", "-j $tapchain", $loglevel, 'FWBR-IN: ', $vmid); 
>> } else { 
>> ruleset_addrule($ruleset, "PVEFW-FWBR-OUT", 
>> - "-m physdev --physdev-is-bridged --physdev-in $iface", "-j $tapchain"); 
>> + "-m physdev --physdev-is-bridged --physdev-in $iface", "-j $tapchain", $loglevel, 'FWBR-OUT: ', $vmid); 
>> } 
>> } 
>>
>>
>>
>> The default loglevel of a vm is logging all packets comming to the tap chain 
>> ex: 
>> -A PVEFW-FWBR-IN -m physdev --physdev-out tap135i0 --physdev-is-bridged -m limit --limit 1/sec -j NFLOG --nflog-prefix ":135:6:PVEFW-FWBR-IN: FWBR-IN: " 
>>
>> Previously, it was only for the final DROP/REJECT/ACCEPT action 
>>
>>
>> Here a example with a connection to port 53, with default action input/output policy = reject, loglevelin:info, and no rule to allow port 
>> [OPTIONS] 
>>
>> log_level_out: info 
>> macfilter: 1 
>> policy_out: REJECT 
>> enable: 1 
>> policy_in: REJECT 
>> dhcp: 0 
>> log_level_in: info 
>>
>> [RULES] 
>>
>>
>>
>>
>>
>> 135 6 PVEFW-FWBR-IN 15/May/2019:15:53:23 +0200 FWBR-IN: IN=fwbr135i0 OUT=fwbr135i0 PHYSIN=fwln135i0 PHYSOUT=tap135i0 MAC=56:af:a2:0d:53:9b:58:49:3b:80:fb:24:08:00 SRC=10.11.53.33 DST=10.3.95.29 LEN=283 TOS=0x00 PREC=0x00 TTL=127 ID=20498 PROTO=UDP SPT=54689 DPT=389 LEN=263 
>> 135 6 tap135i0-IN 15/May/2019:15:53:23 +0200 policy REJECT: IN=fwbr135i0 OUT=fwbr135i0 PHYSIN=fwln135i0 PHYSOUT=tap135i0 MAC=56:af:a2:0d:53:9b:58:49:3b:80:fb:24:08:00 SRC=10.11.53.33 DST=10.3.95.29 LEN=283 TOS=0x00 PREC=0x00 TTL=127 ID=20498 PROTO=UDP SPT=54689 DPT=389 LEN=263 
>>
>>
>> As you see, it's logged twice. (once when coming to tap chain, one when final action is reject) 
>>
>> Same if you make a rule in the vm, activating log on the rule, it's loggued twice. 
>>
>> I would like to be able to remove this logging in FWBR-IN, and only log on rules or default input/output policy action. 
>> But currently, if I disable the loglevel=nolog, it's disabling too the log for default action. 
>>
>>
>> I'm not sure, but do we really need this log on FWBR-IN ? (I mean, we don't have any info if it's drop/accept/reject, so it's pretty useless, we only now that a packet is coming). 
>> Or maybe could we have a different loglevel option for default input/output policy ? 
>>
>> _______________________________________________ 
>> pve-devel mailing list 
>> pve-devel at pve.proxmox.com 
>> https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
> 
> _______________________________________________
> pve-devel mailing list
> pve-devel at pve.proxmox.com
> https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
>