From: Dominik Csapak <d.csapak@proxmox.com>
To: Elias Huhsovitz <e.huhsovitz@proxmox.com>, pve-devel@lists.proxmox.com
Subject: Re: [PATCH manager v2] fix #6735: api: pci: allow mdevscan access via mapping permissions.
Date: Tue, 1 Sep 2026 14:27:34 +0200 [thread overview]
Message-ID: <f79a99cb-c6b2-4a98-9e7f-f96bdc6b5720@proxmox.com> (raw)
In-Reply-To: <20260827112525.154445-1-e.huhsovitz@proxmox.com>
some comments inline
On 8/27/26 1:25 PM, Elias Huhsovitz wrote:
> The mdevscan endpoint requires Sys.Audit or Sys.Modify on '/'. This
> blocks non-admin users from listing mediated device types for a PCI
> mapping, even when they hold Mapping.Use on that mapping.
>
> Check the permission in the API handler based on the parameter type: For
> a raw PCI ID, require Sys.Audit or Sys.Modify on '/'. For a mapping,
> require Sys.Audit or Sys.Modify on '/', or fall back to requiring
> Mapping.Use, Mapping.Modify or Mapping.Audit on the specific mapping
> path.
>
> Set the endpoint permission to 'user => all' so the handler performs the
> type-dependent check. This keeps raw PCI IDs, which are not valid ACL
> paths, out of the declarative ACL evaluation.
>
> Reduce indentation by removing redundant else statement. (return
> provides implicit branching)
>
> Signed-off-by: Elias Huhsovitz <e.huhsovitz@proxmox.com>
> ---
> v1: https://lore.proxmox.com/pve-devel/20260824112610.148089-1-e.huhsovitz@proxmox.com/
>
> Changes v1->v2
> --------------
> * Allow all users in the declarative API permissions.
> * Implement ACL check in API handler by calling
> check_any and raise_perm_exc.
> * Remove else statement in API handler, since return statement
> provides implicit branching
> * update commit message
>
> Side Note
> ---------
> Since this pattern of manual permission checks appear quite frequently,
> it might be of interest to provide generic way to deal with such cases
> inside of the permissions block (Similar to overriding the
> hasPermission method in Java Spring Boot)
>
> For example like this:
> permissions => {
> user => 'all',
> custom => my_perm_sub(),
> },
>
> my_perm_sub would simply return a boolean. This would be an optional
> extension, without requiring major changes in ACL Logic.
>
> Let me know if this is something you would consider worthwhile.
>
> Changes
> -------
> PVE/API2/Hardware/PCI.pm | 71 +++++++++++++++++++++++++++-------------
> 1 file changed, 48 insertions(+), 23 deletions(-)
>
> diff --git a/PVE/API2/Hardware/PCI.pm b/PVE/API2/Hardware/PCI.pm
> index 36b9741b..47c4b29b 100644
> --- a/PVE/API2/Hardware/PCI.pm
> +++ b/PVE/API2/Hardware/PCI.pm
> @@ -3,10 +3,12 @@ package PVE::API2::Hardware::PCI;
> use strict;
> use warnings;
>
> +use PVE::Exception qw(raise raise_perm_exc);
> use PVE::JSONSchema qw(get_standard_option);
>
> use PVE::QemuServer::PCI::Mdev;
> use PVE::RESTHandler;
> +use PVE::RPCEnvironment;
>
> use base qw(PVE::RESTHandler);
>
> @@ -180,7 +182,11 @@ __PACKAGE__->register_method({
> protected => 1,
> proxyto => "node",
> permissions => {
> - check => ['perm', '/', ['Sys.Audit', 'Sys.Modify'], any => 1],
> + description =>
> + "For a PCI ID, requires 'Sys.Audit' or 'Sys.Modify' on '/'. For a mapping,"
> + . " requires the same global permissions, or 'Mapping.Use', 'Mapping.Modify'"
> + . ", or'Mapping.Audit' on '/mapping/pci/<id>'.",
> + user => 'all',
> },
> parameters => {
> additionalProperties => 0,
> @@ -222,31 +228,50 @@ __PACKAGE__->register_method({
> code => sub {
> my ($param) = @_;
>
> - if ($param->{'pci-id-or-mapping'} =~
> - m/^(?:[0-9a-fA-F]{4}:)?[0-9a-fA-F]{2}:[0-9a-fA-F]{2}\.[0-9a-fA-F]$/
> - ) {
> - return PVE::QemuServer::PCI::Mdev::get_mdev_types($param->{'pci-id-or-mapping'}); # PCI ID
> - } else {
> - my $mapping = $param->{'pci-id-or-mapping'};
> -
> - my $types = {};
> - my $devices = PVE::Mapping::PCI::find_on_current_node($mapping);
> - for my $device ($devices->@*) {
> - my $id = $device->{path};
> - next if $id =~ m/;/; # mdev not supported for multifunction devices
> -
> - my $device_types = PVE::QemuServer::PCI::Mdev::get_mdev_types($id);
> -
> - for my $type_definition ($device_types->@*) {
> - my $type = $type_definition->{type};
> - if (!defined($types->{$type})) {
> - $types->{$type} = $type_definition;
> - }
> + my $id = $param->{'pci-id-or-mapping'};
> + my $rpcenv = PVE::RPCEnvironment::get();
> + my $authuser = $rpcenv->get_user();
> +
> + my $is_pci_id =
> + $id =~ m/^(?:[0-9a-fA-F]{4}:)?[0-9a-fA-F]{2}:[0-9a-fA-F]{2}\.[0-9a-fA-F]$/;
> + my $has_global_perms =
> + $rpcenv->check_any($authuser, '/', ['Sys.Audit', 'Sys.Modify'], 1);
> +
> + if (!$has_global_perms) {
> + raise_perm_exc("/, " . join("|", ['Sys.Audit', 'Sys.Modify'])) if ($is_pci_id);
> + $rpcenv->check_any(
> + $authuser,
> + "/mapping/pci/$id",
> + ['Mapping.Use', 'Mapping.Modify', 'Mapping.Audit'],
> + );
> + }
> +
> + return PVE::QemuServer::PCI::Mdev::get_mdev_types($id) if ($is_pci_id);
> +
> + if (!$rpcenv->check_any($authuser, '/', ['Sys.Audit', 'Sys.Modify'], 1)) {
isn't this just $has_global_perms again?
> + $rpcenv->check_any(
> + $authuser,
> + "/mapping/pci/$id",
> + ['Mapping.Use', 'Mapping.Modify', 'Mapping.Audit'],
> + );
this was also already checked?
IMO these happen because the code flow is not very obvious. Instead
of having a bunch of postif clauses, I'd prefer a simpler approach:
my $id = ...
my $rpcenv = ...
my $authuser = ...
my $has_global_perms = ...
my $is_pci_id = ...
if ($is_pci_id) {
raise_perm_exc if !$has_global_perms;
...
} else {
raise_perm_exc if !$has_global_perms && !check_any(mapping_perms);
...
}
with such code, the existing flow and indentation stays the same,
and it's clear which branch takes which permissions
does that make sense?
> + }
> +
> + my $types = {};
> + my $devices = PVE::Mapping::PCI::find_on_current_node($id);
> + for my $device ($devices->@*) {
> + my $dev_id = $device->{path};
> + next if $dev_id =~ m/;/; # mdev not supported for multifunction devices
> +
> + my $device_types = PVE::QemuServer::PCI::Mdev::get_mdev_types($dev_id);
> +
> + for my $type_definition ($device_types->@*) {
> + my $type = $type_definition->{type};
> + if (!defined($types->{$type})) {
> + $types->{$type} = $type_definition;
> }
> }
> -
> - return [sort { $a->{type} cmp $b->{type} } values($types->%*)];
> }
>
> + return [sort { $a->{type} cmp $b->{type} } values($types->%*)];
> },
> });
next prev parent reply other threads:[~2026-09-01 12:27 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-27 11:25 [PATCH manager v2] fix #6735: api: pci: allow mdevscan access via mapping permissions Elias Huhsovitz
2026-09-01 12:27 ` Dominik Csapak [this message]
2026-09-02 8:51 ` Elias Huhsovitz
2026-09-02 10:56 ` Dominik Csapak
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=f79a99cb-c6b2-4a98-9e7f-f96bdc6b5720@proxmox.com \
--to=d.csapak@proxmox.com \
--cc=e.huhsovitz@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