[pve-devel] [PATCH container 1/2] fix #2044: add openwrt support

Oguz Bektas o.bektas at proxmox.com
Tue Nov 9 11:57:42 CET 2021


hi,

thanks for the comments :)

On Mon, Nov 08, 2021 at 04:11:06PM +0100, Thomas Lamprecht wrote:
> no sign-off and zero commit message stating thoughts about design choices
> (like just adding the last named configured vNIC which as single interface
> to the CT network config?!) what openWRT is based on (e.g., that it basically
> just uses busybox) and a, at least basic, pve-docs patch would be also good
> to have..
> 
> On 08.11.21 14:43, Oguz Bektas wrote:
> > ---
> >  src/PVE/LXC/Setup.pm         |  4 ++
> >  src/PVE/LXC/Setup/Makefile   |  1 +
> >  src/PVE/LXC/Setup/OpenWrt.pm | 82 ++++++++++++++++++++++++++++++++++++
> 
> file/module name should be cased OpenWRT

hmm, they call it 'OpenWrt' everywhere on their website though, also on
wikipedia etc. it's spelled in this way ;)

> 
> >  3 files changed, 87 insertions(+)
> >  create mode 100644 src/PVE/LXC/Setup/OpenWrt.pm
> > 
> > diff --git a/src/PVE/LXC/Setup.pm b/src/PVE/LXC/Setup.pm
> > index 5cc56af..e89886f 100644
> > --- a/src/PVE/LXC/Setup.pm
> > +++ b/src/PVE/LXC/Setup.pm
> > @@ -17,6 +17,7 @@ use PVE::LXC::Setup::Fedora;
> >  use PVE::LXC::Setup::Gentoo;
> >  use PVE::LXC::Setup::SUSE;
> >  use PVE::LXC::Setup::Ubuntu;
> > +use PVE::LXC::Setup::OpenWrt;
> >  use PVE::LXC::Setup::Unmanaged;
> >  
> >  my $plugins = {
> > @@ -28,6 +29,7 @@ my $plugins = {
> >      fedora    => 'PVE::LXC::Setup::Fedora',
> >      gentoo    => 'PVE::LXC::Setup::Gentoo',
> >      opensuse  => 'PVE::LXC::Setup::SUSE',
> > +    openwrt   => 'PVE::LXC::Setup::OpenWrt',
> >      ubuntu    => 'PVE::LXC::Setup::Ubuntu',
> >      unmanaged => 'PVE::LXC::Setup::Unmanaged',
> >  };
> > @@ -75,6 +77,8 @@ my $autodetect_type = sub {
> >  	return "alpine";
> >      } elsif (-f  "$rootdir/etc/gentoo-release") {
> >  	return "gentoo";
> > +    } elsif (-f "$rootdir/etc/openwrt_release") {
> > +	return "openwrt";
> >      } elsif (-f "$rootdir/etc/os-release") {
> >  	die "unable to detect OS distribution\n";
> >      } else {
> > diff --git a/src/PVE/LXC/Setup/Makefile b/src/PVE/LXC/Setup/Makefile
> > index 04ee2e4..22702ca 100644
> > --- a/src/PVE/LXC/Setup/Makefile
> > +++ b/src/PVE/LXC/Setup/Makefile
> > @@ -9,6 +9,7 @@ SOURCES=\
> >      Fedora.pm		\
> >      Gentoo.pm		\
> >      SUSE.pm		\
> > +    OpenWrt.pm		\
> >      Ubuntu.pm		\
> >      Unmanaged.pm	\
> >  
> > diff --git a/src/PVE/LXC/Setup/OpenWrt.pm b/src/PVE/LXC/Setup/OpenWrt.pm
> > new file mode 100644
> > index 0000000..89ea740
> > --- /dev/null
> > +++ b/src/PVE/LXC/Setup/OpenWrt.pm
> > @@ -0,0 +1,82 @@
> > +package PVE::LXC::Setup::OpenWrt;
> > +
> > +use strict;
> > +use warnings;
> > +
> > +use PVE::LXC;
> > +use PVE::Network;
> > +use File::Path;
> > +
> > +use PVE::LXC::Setup::Base;
> > +use base qw(PVE::LXC::Setup::Base);
> > +
> > +my $known_versions = {
> > +    '21.02.1' => 1,
> 
> so a simple upgrade from 21.02.1 -> 21.02.2 may already break CT start?
> 
> Busybox based CTs tend to be quite forgiving and not that likely to change,
> so I'd not fatally die on an unknown (newer) version, maybe just checking
> for > 21.02 would be enough of a safety net.

yes that makes sense, i was mainly concerned with the current version
working but should be fine to allow future newer versions as well.

> 
> > +};
> > +
> > +sub new {
> > +    my ($class, $conf, $rootdir) = @_;
> > +
> > +    my $version;
> > +    my $release = PVE::Tools::file_get_contents("$rootdir/etc/openwrt_release");
> > +    if ($release =~ m/^DISTRIB_RELEASE=\'(\d+\.\d+\.\d+)\'$/mi) {
> > +	$version = $1;
> > +    }
> > +    die "unsupported OpenWrt version '$version'\n"
> > +	if !$known_versions->{$version};
> > +
> > +    my $self = { conf => $conf, rootdir => $rootdir, version => $version };
> > +
> > +    $conf->{ostype} = "openwrt";
> > +
> > +    return bless $self, $class;
> > +}
> > +
> > +sub template_fixup {
> > +    my ($self, $conf) = @_;
> > +}
> > +
> > +sub setup_init {
> > +    my ($self, $conf) = @_;
> > +}
> > +
> > +sub setup_network {
> > +    my ($self, $conf) = @_;
> > +
> > +    my $d;
> > +    foreach my $k (keys %$conf) {
> > +	next if $k !~ m/^net(\d+)$/;
> > +	$d = PVE::LXC::Config->parse_lxc_network($conf->{$k});
> > +	next if !$d->{name};
> 
> what is this and why does it selects it a random (perl's `keys` is actively shuffling
> stuff around) different netX config to use on each new start?!
> 
> I'd figure that such nondeterministic behavior would be rather confusing and undesired
> for an user..


ah yes. i guess it would make sense to sort the keys before. i wanted to
add the interfaces in a loop but then got lost, i'll fix that in v2.

> 
> > +    }
> > +
> > +    my $proto = ($d->{ip} eq 'dhcp') ? 'dhcp' : 'static';
> > +    my $ip = "";
> > +    if ($proto eq 'static') {
> > +	$ip = $d->{ip};
> > +    }
> > +
> > +    my $data = <<"DATA";
> > +config interface 'loopback'
> > +	option proto 'static'
> > +	option ipaddr '127.0.0.1'
> > +	option netmask '255.0.0.0'
> > +	option device 'lo'
> > +
> > +config interface 'wan'
> > +	option proto '$proto'
> > +	option device '$d->{name}'
> > +	option ipaddr '$ip'
> > +	option netmask '255.255.255.0'
> 
> so no IPv6 and not multiple interfaces and no dhcp (for testing)? I'd imagine that some
> user may not be too happy with that, especially as they can configure such things without
> any complaint from the backend, at least some good reasoning in a commit message and short
> hints in code comments would be good _if_ we go for that overly simple config printer, I
> mean we surely do not need to support all of openWRT, nor can't we, but at least a minimum
> basic set of common used stuff would be nice, as would be multiple interfaces.

dhcp works (rather hackily but it does :D), see the line with 'proto'
above. we just leave $ip empty in case of dhcp, otherwise it takes the
'static' param in there.

ipv6 can be configured in the same way by adding 'wan6' interface and
doing the same.





More information about the pve-devel mailing list