Difference between revisions of "Atomic rule replacement"

From nftables wiki
Jump to: navigation, search
m (Moar details about when table creation must be done manually)
(Cleanup article: delete bits about format (moved to scripting). Clarify about versions. Put a single 'Notes' section. Don't talk about specific Debian versions, etc.)
Line 25: Line 25:
 
</source>
 
</source>
  
'''Notes:'''
+
=== Notes ===
  
'''''Table Creation''''' - you may have to create the table with ''nft create table ip filter'' before you can load the file exported with ''nft list table filter > filter-table'' otherwise you will hit errors because the table does not exist. Table creation is implied in Debian Stretch and above, Debian Jessie (via backports) will require that you create the tables manually.
+
Please, take these notes into consideration:
  
'''''Comments''''' - You can also add comments to the ''filter-table'' file. Comments are bash style, starting with # and go to the end of the line.
+
* Table Creation: you may have to create the table with ''nft create table ip filter'' before you can load the file exported with ''nft list table filter > filter-table'' otherwise you will hit errors because the table does not exist. Latest releases behaves with more consistency regarding this.
  
'''''Duplicate Rules''''' - If you prepend the ''flush table filter'' line at the very beginning of the ''filter-table'' file, you achieve atomic rule-set replacement equivalent to what ''iptables-restore'' provides. The kernel handles the rule commands in the file in one single transaction, so basically the flushing and the load of the new rules happens in one single shot. If you choose not to flush your tables then you will see duplicate rules for each time you reloaded the config.
+
* Duplicate Rules: If you prepend the ''flush table filter'' line at the very beginning of the ''filter-table'' file, you achieve atomic rule-set replacement equivalent to what ''iptables-restore'' provides. The kernel handles the rule commands in the file in one single transaction, so basically the flushing and the load of the new rules happens in one single shot. If you choose not to flush your tables then you will see duplicate rules for each time you reloaded the config.
  
'''''Flushing Sets''''' - ''flush table filter'' will not flush any sets defined in that table. To flush sets as well, use ''flush ruleset'' (not available in Linux 3.16 or below) or delete the sets explicitly. Early versions (Linux <=3.16) do not allow you to import a set if it already exists, but this is allowed in later versions.
+
* Flushing Sets: ''flush table filter'' will not flush any sets defined in that table. To flush sets as well, use ''flush ruleset'' (available since Linux 3.17 ) or delete the sets explicitly. Early versions (Linux <=3.16) do not allow you to import a set if it already exists, but this is allowed in later versions.
  
'''''Nftables Config File Formats''''' - ''nft -f <filename>'' accepts 2 formats, the first is the format seen in the output of ''nft list table''. The second is [[Scripting]] and is the format you typically see on this website.
+
* What happens when you include 2 files which each have a statement for the filter table? If you have two included files both with statements for the filter table, but one adds a rule allowing traffic from 192.168.1.1 and the other allows traffic from 192.168.1.2 then both rules will be included in the chain, even if one or both files contains a flush statement.
  
Example of nftables output format:
+
* What about flush statements in either, or neither file? If there are any flush commands in any included file then those will be run at the moment the config swap is executed, not at the moment the file is loaded. If you do not include a flush statement in any included file, you will get duplicate rules. If you do include a flush statement, you will not get duplicate rules and the config from *both* files will be included.
 
 
<source lang="bash">
 
create table ip nat
 
#flush table nat
 
table ip nat {
 
chain prerouting {
 
type filter hook prerouting priority 0; policy accept;
 
}
 
 
 
chain postrouting {
 
type filter hook postrouting priority 100; policy accept;
 
}
 
}
 
</source>
 
 
 
Example of scripted config format:
 
 
 
<source lang="bash">
 
#!/usr/sbin/nft -f
 
 
 
define ntp_servers = { 84.77.40.132, 176.31.53.99, 81.19.96.148, 138.100.62.8 }
 
 
 
add table filter
 
add chain filter input { type filter hook input priority 0; }
 
add rule filter input ip saddr $ntp_servers counter
 
</source>
 
 
 
'''Combining multiple files'''
 
 
 
''What happens when you include 2 files which each have a statement for the filter table?''
 
 
 
If you have two files, both ''include''d, both with statements for the filter table, but one adds a rule allowing traffic from 192.168.1.1 and the other allows traffic from 192.168.1.2 then both rules will be included in the chain, even if one or both files contains a flush statement.
 
 
 
''What about flush statements in either, or neither file?''
 
 
 
If there are any flush commands in any included file then those will be run at the moment the config swap is executed, not at the moment the file is loaded. If you do not include a flush statement in any included file, you will get duplicate rules. If you do include a flush statement, you will not get duplicate rules and the config from *both* files will be included.
 
 
 
'''Converting between formats'''
 
 
 
The easiest way to convert from scripted config format to output-based is to run the commands and then show the nftables config with  ''nft list table ip filter'' and then copy/paste that into your config file.
 
The easiest way to convert from output-based to scripted based is to lookup the equivalent on this wiki.
 

Revision as of 12:34, 1 February 2018

Warning about Shell scripting + nftables

With iptables it was common to use a bash script comprised of multiple iptables commands to configure a firewall. This is sub-optimal because it is not atomic, that is to say that during the few fractions of a second that your bash script takes to run your firewall is in a partially configured state. Nftables introduces atomic rule replacement with the -f option. This is different from bash scripts because nftables will read in all of the included config files, create the config object in memory alongside the existing config, and then in one atomic operation it swaps the old config for the new one meaning there is no moment when the firewall is partially configured.

Atomic Rule Replacement

You can use the -f option to atomically update your rule-set:

% nft -f file

Where file contains your rule-set.

You can save your rule-set by storing the existing listing in a file, ie.

% nft list table filter > filter-table

Then you can restore it by using the -f option:

% nft -f filter-table

Notes

Please, take these notes into consideration:

  • Table Creation: you may have to create the table with nft create table ip filter before you can load the file exported with nft list table filter > filter-table otherwise you will hit errors because the table does not exist. Latest releases behaves with more consistency regarding this.
  • Duplicate Rules: If you prepend the flush table filter line at the very beginning of the filter-table file, you achieve atomic rule-set replacement equivalent to what iptables-restore provides. The kernel handles the rule commands in the file in one single transaction, so basically the flushing and the load of the new rules happens in one single shot. If you choose not to flush your tables then you will see duplicate rules for each time you reloaded the config.
  • Flushing Sets: flush table filter will not flush any sets defined in that table. To flush sets as well, use flush ruleset (available since Linux 3.17 ) or delete the sets explicitly. Early versions (Linux <=3.16) do not allow you to import a set if it already exists, but this is allowed in later versions.
  • What happens when you include 2 files which each have a statement for the filter table? If you have two included files both with statements for the filter table, but one adds a rule allowing traffic from 192.168.1.1 and the other allows traffic from 192.168.1.2 then both rules will be included in the chain, even if one or both files contains a flush statement.
  • What about flush statements in either, or neither file? If there are any flush commands in any included file then those will be run at the moment the config swap is executed, not at the moment the file is loaded. If you do not include a flush statement in any included file, you will get duplicate rules. If you do include a flush statement, you will not get duplicate rules and the config from *both* files will be included.