From: Alexandre Derumier <alexandre.derumier@groupe-cyllene.com>
To: pve-devel@lists.proxmox.com
Cc: Alexandre Derumier <aderumier@groupe-cyllene.com>
Subject: SPAM: [RFC qemu-server 02/13] add D-Bus display support
Date: Tue, 25 Aug 2026 13:08:34 +0200 [thread overview]
Message-ID: <20260825110849.2967694-3-alexandre.derumier@groupe-cyllene.com> (raw)
In-Reply-To: <20260825110849.2967694-1-alexandre.derumier@groupe-cyllene.com>
From: Alexandre Derumier <aderumier@groupe-cyllene.com>
Add a private D-Bus daemon per VM for QEMU's -display dbus.
Signed-off-by: Alexandre Derumier <aderumier@groupe-cyllene.com>
---
src/PVE/QemuServer/DBusDisplay.pm | 100 ++++++++++++++++++++++++++++++
src/PVE/QemuServer/Helpers.pm | 6 ++
2 files changed, 106 insertions(+)
create mode 100644 src/PVE/QemuServer/DBusDisplay.pm
diff --git a/src/PVE/QemuServer/DBusDisplay.pm b/src/PVE/QemuServer/DBusDisplay.pm
new file mode 100644
index 0000000..b8684be
--- /dev/null
+++ b/src/PVE/QemuServer/DBusDisplay.pm
@@ -0,0 +1,100 @@
+package PVE::QemuServer::DBusDisplay;
+
+# A private D-Bus bus per VM for QEMU's -display dbus. QEMU connects to the
+# address rather than creating it, and every QEMU wants to own org.qemu, so the
+# session bus would cap a node at one VM. libvirt does the same.
+
+use strict;
+use warnings;
+
+use Time::HiRes qw(usleep);
+
+use PVE::ProcFSTools;
+use PVE::Tools qw(file_set_contents);
+use PVE::QemuServer::Helpers;
+
+sub config_file {
+ my ($vmid) = @_;
+ return "$PVE::QemuServer::Helpers::var_run_tmpdir/$vmid.dbusdisplay.conf";
+}
+
+sub pidfile {
+ my ($vmid) = @_;
+ return "$PVE::QemuServer::Helpers::var_run_tmpdir/$vmid.dbusdisplay.pid";
+}
+
+sub write_config {
+ my ($vmid) = @_;
+
+ my $socket = PVE::QemuServer::Helpers::dbus_socket($vmid);
+ my $pidfile = pidfile($vmid);
+
+ my $conf = <<"EOF";
+<!DOCTYPE busconfig PUBLIC "-//freedesktop//DTD D-Bus Bus Configuration 1.0//EN"
+ "http://www.freedesktop.org/standards/dbus/1.0/busconfig.dtd">
+<busconfig>
+ <type>org.qemu</type>
+ <listen>unix:path=$socket</listen>
+ <pidfile>$pidfile</pidfile>
+ <auth>EXTERNAL</auth>
+ <policy context="default">
+ <allow send_destination="*" eavesdrop="true"/>
+ <allow eavesdrop="true"/>
+ <allow own="*"/>
+ </policy>
+</busconfig>
+EOF
+
+ my $path = config_file($vmid);
+ file_set_contents($path, $conf, 0600);
+
+ return $path;
+}
+
+# Called inside the VM's systemd scope, so it dies with the VM, as swtpm does.
+sub start {
+ my ($vmid) = @_;
+
+ stop($vmid); # a survivor from an unclean stop still holds the socket
+
+ my $config = write_config($vmid);
+ unlink PVE::QemuServer::Helpers::dbus_socket($vmid);
+
+ PVE::Tools::run_command(
+ ['dbus-daemon', "--config-file=$config", '--fork'],
+ errmsg => "failed to start D-Bus daemon for VM $vmid",
+ );
+
+ return;
+}
+
+sub stop {
+ my ($vmid) = @_;
+
+ my $pidfile = pidfile($vmid);
+ if (my $pid = eval { PVE::Tools::file_read_firstline($pidfile) }) {
+ if ($pid =~ m/^(\d+)$/) {
+ $pid = $1;
+ kill('TERM', $pid);
+
+ # Waited for: a dying daemon unlinks the new socket and holds the VM's scope
+ # cgroup open, which blocks the next start.
+ for (my $waited = 0; $waited < 5; $waited += 0.05) {
+ last if !PVE::ProcFSTools::check_process_running($pid);
+ usleep(50_000);
+ }
+ if (PVE::ProcFSTools::check_process_running($pid)) {
+ warn "D-Bus daemon for VM $vmid did not exit, killing it\n";
+ kill('KILL', $pid);
+ }
+ }
+ }
+
+ unlink $pidfile;
+ unlink config_file($vmid);
+ unlink PVE::QemuServer::Helpers::dbus_socket($vmid);
+
+ return;
+}
+
+1;
diff --git a/src/PVE/QemuServer/Helpers.pm b/src/PVE/QemuServer/Helpers.pm
index dd17eef..816f7aa 100644
--- a/src/PVE/QemuServer/Helpers.pm
+++ b/src/PVE/QemuServer/Helpers.pm
@@ -140,6 +140,12 @@ sub vnc_socket {
return "${var_run_tmpdir}/$vmid.vnc";
}
+sub dbus_socket {
+ my ($vmid) = @_;
+ # dbusdisplay, not dbus: keep it apart from the dbus-vmstate helper's files.
+ return "${var_run_tmpdir}/$vmid.dbusdisplay";
+}
+
# Parse the cmdline of a running kvm/qemu-* process and return arguments as hash
sub parse_cmdline {
my ($pid) = @_;
--
2.55.0
next prev parent reply other threads:[~2026-08-25 11:09 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-25 11:08 SPAM: [RFC pve-http-server/qemu-server/pve-manager/pve-{qemu-kyber,kyberproxy, kyber-web,qemu-rdp,rdpproxy,rdp-web} 00/13] add rdp && kyber consoles for qemu over D-Bus display Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-http-server 01/13] anyevent : proxy a path prefix to a local http proxy Alexandre Derumier
2026-08-25 11:08 ` Alexandre Derumier [this message]
2026-08-25 11:08 ` SPAM: [RFC qemu-server 03/13] add kyber display Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC qemu-server 04/13] add rdp display Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC qemu-server 05/13] add experimental kyber-gl display Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-manager 06/13] ui: add kyber console Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-manager 07/13] ui: add rdp console Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-kyber-web 10/13] add pve-kyber-web: console's webassembly client Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-qemu-rdp 11/13] Add pve-qemu-rdp: an RDP server for the console Alexandre Derumier
2026-08-25 11:08 ` SPAM: [RFC pve-rdpproxy 12/13] Add pve-rdpproxy Alexandre Derumier
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=20260825110849.2967694-3-alexandre.derumier@groupe-cyllene.com \
--to=alexandre.derumier@groupe-cyllene.com \
--cc=aderumier@groupe-cyllene.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.