public inbox for pve-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: "Daniel Kral" <d.kral@proxmox.com>
To: "David Riley" <d.riley@proxmox.com>, <pve-devel@lists.proxmox.com>
Subject: Re: [PATCH pve-network v4 5/5] fix #7520: config: prune orphaned ACLs and relocate moved VNets
Date: Thu, 02 Jul 2026 15:26:12 +0200	[thread overview]
Message-ID: <DJO4BC76N4VK.2ZWBWSQ45S51U@proxmox.com> (raw)
In-Reply-To: <20260626105258.56914-6-d.riley@proxmox.com>

On Fri Jun 26, 2026 at 12:52 PM CEST, David Riley wrote:
> Compare the running configuration with the newly compiled state during
> config commit to identify and prune orphaned SDN ACL entries. This
> ensures state consistency for manual applies via the UI/API as well as
> during automatic configuration reloads on system boot.
>
> Track configuration changes across:
> * Zones
> * VNets
> * Fabrics
> * Controllers
> * Route maps
> * Prefix lists
>
> Trigger a validation check when a VNet is moved between zones to
> mitigate potential ACL path conflicts. Abort the apply operation if a
> conflict is detected to prevent configuration overwrites. If the path
> is clear, relocate the ACLs to the updated zone path.
>
> Suggested-by: Stefan Hanreich <s.hanreich@proxmox.com>
> Signed-off-by: David Riley <d.riley@proxmox.com>
> ---
>  src/PVE/Network/SDN.pm | 141 +++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 141 insertions(+)
>
> diff --git a/src/PVE/Network/SDN.pm b/src/PVE/Network/SDN.pm
> index 33a3cf3..cd563d2 100644
> --- a/src/PVE/Network/SDN.pm
> +++ b/src/PVE/Network/SDN.pm
> @@ -10,6 +10,7 @@ use LWP::UserAgent;
>  use Net::SSLeay;
>  use UUID;
>  
> +use PVE::AccessControl;
>  use PVE::Cluster qw(cfs_read_file cfs_write_file cfs_lock_file);
>  use PVE::INotify;
>  use PVE::RESTEnvironment qw(log_warn);
> @@ -238,11 +239,151 @@ sub compile_running_cfg {
>  }
>  
>  sub commit_config {
> +    my $old_cfg = cfs_read_file($RUNNING_CFG_FILENAME);
>      my $cfg = compile_running_cfg();
>  
> +    cleanup($old_cfg, $cfg);
> +
>      cfs_write_file($RUNNING_CFG_FILENAME, $cfg);
>  }
>  
> +sub cleanup {

maybe a more telling name like cleanup_sdn_resource_acls(...) or
something similar? Otherwise, it might suggest that it cleans up
something in the SDN config.

> +    my ($old_cfg, $cfg) = @_;
> +    if (!defined($old_cfg) || !defined($cfg)) {
> +        return;
> +    }

AFAICT $old_cfg and $cfg are always defined as the cfs_read_file()
parser for $RUNNING_CFG_FILENAME will always at least return an empty
hash ref and compile_running_cfg() will also always return a hash ref,
so this should not be needed?

> +
> +    my $route_map_paths = diff_route_maps($old_cfg, $cfg);
> +    my $generic_paths = diff_generic_resources($old_cfg, $cfg);
> +    my ($vnet_delete_paths, $vnet_move_paths) = diff_vnets($old_cfg, $cfg);
> +
> +    if (@$vnet_move_paths) {
> +        PVE::AccessControl::migrate_sdn_resource_access($vnet_move_paths);
> +    }
> +
> +    my @paths_to_delete = (@$vnet_delete_paths, @$route_map_paths, @$generic_paths);
> +    if (@paths_to_delete) {
> +        PVE::AccessControl::remove_sdn_resource_access(\@paths_to_delete);
> +    }
> +
> +}
> +
> +sub diff_generic_resources {
> +    my ($old_cfg, $cfg) = @_;
> +
> +    my @paths_to_delete;

nit: initialize with empty array ()

> +
> +    my @types = qw(zones controllers fabrics prefix-lists);
> +    for my $type (@types) {
> +        if (!defined($old_cfg->{$type}) || !defined($old_cfg->{$type}->{ids})) {
> +            next;
> +        }
> +
> +        my $old_ids = $old_cfg->{$type}->{ids};
> +        my $new_ids = {};
> +
> +        if (defined($cfg->{$type}) && defined($cfg->{$type}->{ids})) {
> +            $new_ids = $cfg->{$type}->{ids};
> +        }

As compile_running_cfg() does always set the $cfg->{$type}->{ids} this
could probably be set unconditionally?

> +
> +        for my $id (keys %$old_ids) {
> +            if (defined($new_ids->{$id})) {
> +                next;
> +            }

nit: use post-if statement

> +
> +            push(@paths_to_delete, [$type, $id]);
> +        }
> +    }
> +
> +    return \@paths_to_delete;
> +}
> +
> +sub diff_route_maps {
> +    my ($old_cfg, $cfg) = @_;
> +
> +    my @paths_to_delete;

nit: initialize with empty array ()

> +
> +    if (defined($old_cfg->{'route-maps'}) && defined($old_cfg->{'route-maps'}->{ids})) {
> +        my $old_route_maps = $old_cfg->{'route-maps'}->{ids};
> +        my $route_map_suffix = qr/_[0-9]+$/;

nit: suffix with _re(gex)

> +
> +        my %active_route_maps;

nit: initialize with empty hash ()

> +        if (defined($cfg->{'route-maps'}) && defined($cfg->{'route-maps'}->{ids})) {
> +            for my $id (keys %{ $cfg->{'route-maps'}->{ids} }) {

nit: use postfix:    keys $cfg->{'route-maps'}->{ids}->%*

> +                (my $base_name = $id) =~ s/$route_map_suffix//;
> +                $active_route_maps{$base_name} = 1;
> +            }
> +        }
> +
> +        my %queued_route_maps;

nit: initialize with empty hash ()

> +        for my $id (keys %$old_route_maps) {
> +            if (defined($cfg->{'route-maps'}->{ids}->{$id})) {
> +                next;
> +            }

nit: use post-if statement

> +
> +            (my $base_name = $id) =~ s/$route_map_suffix//;
> +
> +            if ($active_route_maps{$base_name}) {
> +                next;
> +            }

nit: use post-if statement

> +
> +            if ($queued_route_maps{$base_name}) {
> +                next;
> +            }

nit: use post-if statement

> +
> +            push(@paths_to_delete, ['route-maps', $base_name]);
> +            $queued_route_maps{$base_name} = 1;
> +        }
> +    }
> +
> +    return \@paths_to_delete;
> +}
> +
> +sub diff_vnets {
> +    my ($old_cfg, $cfg) = @_;
> +
> +    my @paths_to_delete;
> +    my @paths_to_move;

nit: initialize with empty arrays ()

> +
> +    if (defined($old_cfg->{vnets}) && defined($old_cfg->{vnets}->{ids})) {
> +        my $old_vnets = $old_cfg->{vnets}->{ids};
> +        my $new_vnets = {};
> +
> +        if (defined($cfg->{vnets}) && defined($cfg->{vnets}->{ids})) {
> +            $new_vnets = $cfg->{vnets}->{ids};
> +        }

same as above, could be set unconditionally as compile_running_cfg()
always sets the $cfg->{vnets}->{ids} hash ref

> +
> +        for my $vnetid (keys %$old_vnets) {
> +            my $old_zone = $old_vnets->{$vnetid}->{zone};
> +
> +            if (!defined($new_vnets->{$vnetid})) {
> +                if (defined($old_zone)) {
> +                    push(@paths_to_delete, ['zones', $old_zone, $vnetid]);
> +                } else {
> +                    log_warn(
> +                        "SDN Cleanup: Could not find zone for deleted VNet $vnetid, skipping ACL"
> +                            . " cleanup");
> +                }
> +                next;
> +            }
> +
> +            my $new_zone = $new_vnets->{$vnetid}->{zone};
> +
> +            if (defined($old_zone) && defined($new_zone) && $old_zone ne $new_zone) {
> +                push(
> +                    @paths_to_move,
> +                    {
> +                        src_path => ['zones', $old_zone, $vnetid],
> +                        dest_path => ['zones', $new_zone, $vnetid],
> +                    },
> +                );
> +            }

Hm, there's a lot of testing whether the zone exists... Can this
actually happen besides users directly editing the sdn/vnets.cfg file
and removing the vnet property?

The API handler doesn't permit me to add a VNet without or with a
non-existing zone (only the zone 'abc' is existing; the error message is
a bit misleading, because the type is undefined here because the zone
does not exist at all):

# pvesh create /cluster/sdn/vnets --vnet aeaewk --zone '' --type 'vnet'
create sdn vnet object failed: cannot lookup undefined type! at /usr/share/perl5/PVE/API2/Network/SDN/Vnets.pm line 326.
# pvesh create /cluster/sdn/vnets --vnet aeaewk --zone 'abc' --type 'vnet'
# pvesh create /cluster/sdn/vnets --vnet aeaewk2 --zone 'def' --type 'vnet'
create sdn vnet object failed: cannot lookup undefined type! at /usr/share/perl5/PVE/API2/Network/SDN/Vnets.pm line 326.

> +        }
> +    }
> +
> +    return (\@paths_to_delete, \@paths_to_move);
> +}
> +
>  sub has_pending_changes {
>      my $running_cfg = PVE::Network::SDN::running_config();
>  






      reply	other threads:[~2026-07-02 13:26 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-06-26 10:52 [PATCH access-control/network v4 0/5] fix #7520: sdn: prune orphaned ACLs and handle VNet migrations David Riley
2026-06-26 10:52 ` [PATCH pve-access-control v4 1/5] fix: #7520: sdn: prune orphaned ACLs on resource deletion David Riley
2026-07-02 13:14   ` Daniel Kral
2026-06-26 10:52 ` [PATCH pve-access-control v4 2/5] fix #7520: test: add unit tests for sdn acl pruning logic David Riley
2026-07-02 13:14   ` Daniel Kral
2026-06-26 10:52 ` [PATCH pve-access-control v4 3/5] fix: #7520: sdn: add VNet ACL migration David Riley
2026-07-02 13:26   ` Daniel Kral
2026-06-26 10:52 ` [PATCH pve-access-control v4 4/5] fix #7520: test: add unit tests for sdn acl migration logic David Riley
2026-06-26 10:52 ` [PATCH pve-network v4 5/5] fix #7520: config: prune orphaned ACLs and relocate moved VNets David Riley
2026-07-02 13:26   ` Daniel Kral [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=DJO4BC76N4VK.2ZWBWSQ45S51U@proxmox.com \
    --to=d.kral@proxmox.com \
    --cc=d.riley@proxmox.com \
    --cc=pve-devel@lists.proxmox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal