[PATCH qemu-server/docs/manager/ha-manager 0/5] add pvpanic device support

[PATCH qemu-server/docs/manager/ha-manager 0/5] add pvpanic device support

[Kaiyang Wu]

This patchset adds pvpanic device support to QEMU virtual machines to
provide extra virtual machine status (guest-panicked) when virtual
machines panic [0]. The HA stack now restarts a "started" VM when it
panics.

[0]: https://www.qemu.org/docs/master/specs/pvpanic.html

v1: https://lore.proxmox.com/pve-devel/20260827103529.393388-1-wukaiyang@loongfans.cn/

Changes v1 -> v2:
- Remove the pvpanic ISA device support to simplify the code
- Move the pvpanic setting from Hardware to Options
- Restart a HA-managed VM when it panics to restore the "started" state


qemu-server:

Kaiyang Wu (1):
  qemuserver: add pvpanic device

 src/PVE/API2/Qemu.pm      |  1 +
 src/PVE/QemuServer.pm     | 11 +++++++++++
 src/PVE/QemuServer/PCI.pm |  1 +
 3 files changed, 13 insertions(+)


docs:

Kaiyang Wu (1):
  qm: add document section for the pvpanic device

 qm.adoc | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)


manager:

Kaiyang Wu (1):
  ui: qemu: add pvpanic device support

 www/manager6/qemu/Options.js | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)


ha-manager:

Kaiyang Wu (2):
  vm resource: return running status -1 for panicked VMs
  lrm: stop and restart panicked VMs to keep the "started" state

 src/PVE/HA/LRM.pm             | 9 ++++++++-
 src/PVE/HA/Resources/PVEVM.pm | 5 ++++-
 2 files changed, 12 insertions(+), 2 deletions(-)

-- 
2.55.0

[PATCH qemu-server 1/5] qemuserver: add pvpanic device

[Kaiyang Wu]

Add pvpanic PCI device to monitor guest system panics.

Suggested-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Kaiyang Wu <wukaiyang@loongfans.cn>
---
 src/PVE/API2/Qemu.pm      |  1 +
 src/PVE/QemuServer.pm     | 11 +++++++++++
 src/PVE/QemuServer/PCI.pm |  1 +
 3 files changed, 13 insertions(+)

diff --git a/src/PVE/API2/Qemu.pm b/src/PVE/API2/Qemu.pm
index 714c87a9..d60e90a8 100644
--- a/src/PVE/API2/Qemu.pm
+++ b/src/PVE/API2/Qemu.pm
@@ -818,6 +818,7 @@ my $hwtypeoptions = {
     'watchdog' => 1,
     'audio0' => 1,
     'rng0' => 1,
+    'pvpanic' => 1,
 };
 
 my $generaloptions = {
diff --git a/src/PVE/QemuServer.pm b/src/PVE/QemuServer.pm
index 5849392e..facbb06a 100644
--- a/src/PVE/QemuServer.pm
+++ b/src/PVE/QemuServer.pm
@@ -740,6 +740,12 @@ EODESCR
         optional => 1,
         default => 1,
     },
+    pvpanic => {
+        optional => 1,
+        type => 'boolean',
+        description => "Configure a pvpanic device to monitor guest panics",
+        default => 0,
+    },
 };
 
 my $cicustom_fmt = {
@@ -3439,6 +3445,11 @@ sub config_to_command {
         push @$cmd, '-nographic';
     }
 
+    if (defined($conf->{pvpanic}) && $conf->{pvpanic}) {
+        my $pvpanicpciaddr = print_pci_addr("pvpanic", $arch);
+        push @$devices, '-device', "pvpanic-pci,id=pvpanic$pvpanicpciaddr";
+    }
+
     # For now, handles only specific parts, but the final goal is to cover everything.
     my $cfg2cmd_opts = { forcemachine => $forcemachine };
     my $cfg2cmd = PVE::QemuServer::Cfg2Cmd->new($conf, $defaults, $version_guard, $cfg2cmd_opts);
diff --git a/src/PVE/QemuServer/PCI.pm b/src/PVE/QemuServer/PCI.pm
index 35e18ad9..faaeffca 100644
--- a/src/PVE/QemuServer/PCI.pm
+++ b/src/PVE/QemuServer/PCI.pm
@@ -220,6 +220,7 @@ sub get_pci_addr_map {
         'pci.4' => { bus => 1, addr => 28 },
         'rng0' => { bus => 1, addr => 29 },
         'pci.2-igd' => { bus => 1, addr => 30 }, # replaces pci.2 in case a legacy IGD device is passed through
+        'pvpanic' => { bus => 1, addr => 31 },
         'virtio6' => { bus => 2, addr => 1 },
         'virtio7' => { bus => 2, addr => 2 },
         'virtio8' => { bus => 2, addr => 3 },
-- 
2.55.0

[PATCH docs 2/5] qm: add document section for the pvpanic device

[Kaiyang Wu]

Document the pvpanic device and corresponding driver requirements on
different VMs.

Signed-off-by: Kaiyang Wu <wukaiyang@loongfans.cn>
---
 qm.adoc | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/qm.adoc b/qm.adoc
index 18a4fbe..49376ec 100644
--- a/qm.adoc
+++ b/qm.adoc
@@ -1476,6 +1476,28 @@ tag (name used to mount the device on the guest).
 For more information on available virtiofsd parameters, see the
 https://gitlab.com/virtio-fs/virtiofsd[GitLab virtiofsd project page].
 
+[[qm_pvpanic]]
+Pvpanic
+~~~~~~~
+
+Pvpanic is a device to send system panic events to QEMU. It allows to
+provide extra VM status when the VM system crashes.
+
+To add a pvpanic PCI device run the following command:
+
+----
+qm set <vmid> -pvpanic 1
+----
+
+Linux VMs with kernel >=5.2 support pvpanic PCI device by default
+(https://www.kernelconfig.io/CONFIG_PVPANIC_PCI[pvpanic PCI driver]).
+
+Windows VMs require
+https://github.com/virtio-win/kvm-guest-drivers-windows/[kvm-guest-drivers-windows]
+to use pvpanic.
+
+FreeBSD VMs do not support pvpanic.
+
 [[qm_bootorder]]
 Device Boot Order
 ~~~~~~~~~~~~~~~~~
-- 
2.55.0

[PATCH manager 3/5] ui: qemu: add pvpanic device support

[Kaiyang Wu]

Add basic structure for adding a pvpanic device, allowing to monitor
guest system panics through the pvpanic device.

Suggested-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Kaiyang Wu <wukaiyang@loongfans.cn>
---
 www/manager6/qemu/Options.js | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/www/manager6/qemu/Options.js b/www/manager6/qemu/Options.js
index 73e27500..157ab7b6 100644
--- a/www/manager6/qemu/Options.js
+++ b/www/manager6/qemu/Options.js
@@ -313,6 +313,26 @@ Ext.define('PVE.qemu.Options', {
                       }
                     : undefined,
             },
+            pvpanic: {
+                header: gettext('Panic Monitor'),
+                defaultValue: false,
+                renderer: Proxmox.Utils.format_boolean,
+                editor: caps.vms['VM.Config.HWType']
+                    ? {
+                          xtype: 'proxmoxWindowEdit',
+                          subject: gettext('Panic Monitor'),
+                          onlineHelp: 'qm_pvpanic',
+                          items: {
+                              xtype: 'proxmoxcheckbox',
+                              name: 'pvpanic',
+                              uncheckedValue: 0,
+                              defaultValue: 0,
+                              deleteDefaultValue: true,
+                              fieldLabel: gettext('Panic Monitor'),
+                          },
+                      }
+                    : undefined,
+            },
             protection: {
                 header: gettext('Protection'),
                 defaultValue: false,
-- 
2.55.0

[PATCH ha-manager 4/5] vm resource: return running status -1 for panicked VMs

[Kaiyang Wu]

Indicate the panicked status used to stop the VM before restarting it to
keep the 'started' state. Keep the return value truthy to shutdown the
VM when needed (e.g. 'stopped' state).

Signed-off-by: Kaiyang Wu <wukaiyang@loongfans.cn>
---
 src/PVE/HA/Resources/PVEVM.pm | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/src/PVE/HA/Resources/PVEVM.pm b/src/PVE/HA/Resources/PVEVM.pm
index 8753271..51c7bac 100644
--- a/src/PVE/HA/Resources/PVEVM.pm
+++ b/src/PVE/HA/Resources/PVEVM.pm
@@ -143,10 +143,13 @@ sub check_running {
     my $nodename = $haenv->nodename();
 
     if (PVE::QemuServer::check_running($vmid, 1, $nodename)) {
+        # return -1 to indicate the panicked status while keeping it truthy
+        my $qmpstatus = eval { PVE::QemuServer::Monitor::mon_cmd($vmid, 'query-status') };
+        return -1 if defined($qmpstatus) && $qmpstatus->{status} eq 'guest-panicked';
+
         # do not count VMs which are suspended for a backup job as running
         my $conf = PVE::QemuConfig->load_config($vmid, $nodename);
         if (defined($conf->{lock}) && $conf->{lock} eq 'backup') {
-            my $qmpstatus = eval { PVE::QemuServer::Monitor::mon_cmd($vmid, 'query-status') };
             $haenv->log('warning', "$@") if $@;
 
             return 0 if defined($qmpstatus) && $qmpstatus->{status} eq 'prelaunch';
-- 
2.55.0

[PATCH ha-manager 5/5] lrm: stop and restart panicked VMs to keep the "started" state

[Kaiyang Wu]

When a VM panics, 'check_running()' returns -1 to indicate the panicked
state. Stop the service first since panicked VMs are still running, then
restart the service to keep the "started" state.

Suggested-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Kaiyang Wu <wukaiyang@loongfans.cn>
---
 src/PVE/HA/LRM.pm | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/src/PVE/HA/LRM.pm b/src/PVE/HA/LRM.pm
index 72e37e6..8fa1ff6 100644
--- a/src/PVE/HA/LRM.pm
+++ b/src/PVE/HA/LRM.pm
@@ -982,7 +982,14 @@ sub exec_resource_agent {
 
     if ($cmd eq 'started') {
 
-        return SUCCESS if $running;
+        return SUCCESS if $running == 1; # running normally
+
+        # stop the panicked service before restarting it
+        if ($running == -1) {
+            $haenv->log("info", "stopping panicked service $sid");
+
+            $plugin->shutdown($haenv, $id, 0);
+        }
 
         $haenv->log("info", "starting service $sid");
 
-- 
2.55.0