[PATCH qemu-server v3 0/3] improve guest cleanup handling

[PATCH qemu-server v3 0/3] improve guest cleanup handling

[Dominik Csapak]

this is a combination of my previous patches[0][1] (v3 because [0] was
already v2)

First we make the cleanup handling more consistent (1/3)
then we check explicitely for the backup lock to improve the error
message for stop backup mode (2/3)
and then we fix #7119 by waiting up to 30s for a possibly still running
guest to stop (e.g. this can occur when using usb passthrouh) (3/3)

changes from v2/RFC:
* use 'vm_running_locally' for getting the pid
* improve error messages
* use a 'use_old_cleanup' flag that will be auto-removed by a reboot
  to signal if we can use the new cleanup logic or the old

0: https://lore.proxmox.com/pve-devel/20260210111612.2017883-1-d.csapak@proxmox.com/
1: https://lore.proxmox.com/pve-devel/20260223105542.1525232-1-d.csapak@proxmox.com/

Dominik Csapak (3):
  cleanup: refactor to make cleanup flow consistent
  qm cleanup: die early when encountering a running stop mode backup
  fix #7119: qm cleanup: wait for process exiting for up to 30 seconds

 debian/preinst                   | 16 +++++++++++++++
 src/PVE/CLI/qm.pm                | 35 +++++++++++++++++++++++++++-----
 src/PVE/QemuServer.pm            | 14 +++++++++++++
 src/PVE/QemuServer/RunState.pm   | 28 +++++++++++++++++++++++++
 src/test/MigrationTest/QmMock.pm | 16 +++++++++++++++
 5 files changed, 104 insertions(+), 5 deletions(-)
 create mode 100755 debian/preinst

-- 
2.47.3

[PATCH qemu-server v3 1/3] cleanup: refactor to make cleanup flow consistent

[Dominik Csapak]

There are two ways a cleanup can be triggered:

* When a guest is stopped/shutdown via the API, 'vm_stop' calls 'vm_stop_cleanup'.
* When the guest process disconnects from qmeventd, 'qm cleanup' is
  called, which in turn also tries to call 'vm_stop_cleanup'.

Both of these happen under a qemu config lock, so there is no direct
race condition that it will be called out of order, but it could happen
that the 'qm cleanup' call happened in addition so cleanup was called
twice. Which could be a problem when the shutdown was called with
'keepActive' which 'qm cleanup' would simply know nothing of and ignore.

Also the post-stop hook might not be triggered in case e.g. a stop-mode
backup was done, since that was only happening via qm cleanup and this
would sometimes detect the now again running guest and abort.

To improve the situation we move the exec_hookscript call at the end
of vm_stop_cleanup. At this point we know the vm is stopped and we still
have the config lock.

To prevent a double cleanup, create a new cleanup flag on vm startup and
only cleanup when this is still there, then delete it at the end.

This means we can now drop the check for clean/guest shutdown to do
that.

Since this can only work for guests started after this logic is
introduced, create an additional flag that we must use the old logic
which will be cleared on reboot. After that the new logic is used.

That flag will be created with a 'preinst' script, so we create it when
we come from an older version before the code is actually used.

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
---
 debian/preinst                   | 16 ++++++++++++++++
 src/PVE/CLI/qm.pm                | 12 +++++++++---
 src/PVE/QemuServer.pm            | 14 ++++++++++++++
 src/PVE/QemuServer/RunState.pm   | 28 ++++++++++++++++++++++++++++
 src/test/MigrationTest/QmMock.pm | 16 ++++++++++++++++
 5 files changed, 83 insertions(+), 3 deletions(-)
 create mode 100755 debian/preinst

diff --git a/debian/preinst b/debian/preinst
new file mode 100755
index 00000000..1697020f
--- /dev/null
+++ b/debian/preinst
@@ -0,0 +1,16 @@
+#!/bin/sh
+
+set -e
+
+case "$1" in
+    upgrade)
+        if dpkg --compare-versions "$2" 'lt' '9.1.5'; then
+            # set use_old_cleanup flag file so the old cleanup code will be used until reboot
+            touch /run/qemu-server/use_old_cleanup
+        fi
+    ;;
+    install|abort-install)
+    ;;
+esac
+
+exit 0
diff --git a/src/PVE/CLI/qm.pm b/src/PVE/CLI/qm.pm
index bdae9641..b7bc4d9f 100755
--- a/src/PVE/CLI/qm.pm
+++ b/src/PVE/CLI/qm.pm
@@ -1120,12 +1120,18 @@ __PACKAGE__->register_method({
                     }
                 }
 
-                if (!$clean || $guest) {
-                    # vm was shutdown from inside the guest or crashed, doing api cleanup
+                my $can_use_cleanup_flag = PVE::QemuServer::RunState::can_use_cleanup_flag();
+
+                if (!$clean || $guest || $can_use_cleanup_flag) {
+                    # either we can use the new mechanism to check if cleanup is done, or
+                    # vm was shutdown from inside the guest or crashed
                     PVE::QemuServer::vm_stop_cleanup($storecfg, $vmid, $conf, 0, 0, 1);
                 }
 
-                PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-stop');
+                if (!$can_use_cleanup_flag) {
+                    # if the new cleanup mechanism is used, this will be called from 'vm_stop_cleanup'
+                    PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-stop');
+                }
 
                 $restart = eval { PVE::QemuServer::clear_reboot_request($vmid) };
                 warn $@ if $@;
diff --git a/src/PVE/QemuServer.pm b/src/PVE/QemuServer.pm
index 545758dc..2f1d5c11 100644
--- a/src/PVE/QemuServer.pm
+++ b/src/PVE/QemuServer.pm
@@ -5817,6 +5817,8 @@ sub vm_start_nolock {
 
     syslog("info", "VM $vmid started with PID $pid.");
 
+    PVE::QemuServer::RunState::create_cleanup_flag($vmid);
+
     if (defined(my $migrate = $res->{migrate})) {
         if ($migrate->{proto} eq 'tcp') {
             my $nodename = nodename();
@@ -6124,6 +6126,11 @@ sub cleanup_pci_devices {
 sub vm_stop_cleanup {
     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes, $noerr) = @_;
 
+    my $can_use_cleanup_flag = PVE::QemuServer::RunState::can_use_cleanup_flag();
+    if ($can_use_cleanup_flag) {
+        return if !PVE::QemuServer::RunState::cleanup_flag_exists($vmid);
+    }
+
     eval {
         PVE::QemuServer::QSD::quit($vmid)
             if PVE::QemuServer::Helpers::qsd_running_locally($vmid);
@@ -6158,6 +6165,13 @@ sub vm_stop_cleanup {
         die $err if !$noerr;
         warn $err;
     }
+
+    if ($can_use_cleanup_flag) {
+        # if the old cleanup mechanism is in place, this will be called by 'qm cleanup'
+        PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-stop');
+    }
+
+    PVE::QemuServer::RunState::clear_cleanup_flag($vmid);
 }
 
 # call only in locked context
diff --git a/src/PVE/QemuServer/RunState.pm b/src/PVE/QemuServer/RunState.pm
index 6a5fdbd7..e2bd4d94 100644
--- a/src/PVE/QemuServer/RunState.pm
+++ b/src/PVE/QemuServer/RunState.pm
@@ -6,6 +6,7 @@ use warnings;
 use POSIX qw(strftime);
 
 use PVE::Cluster;
+use PVE::File;
 use PVE::RPCEnvironment;
 use PVE::Storage;
 
@@ -183,4 +184,31 @@ sub vm_resume {
     );
 }
 
+my sub get_cleanup_path {
+    my ($vmid) = @_;
+    return "/var/run/qemu-server/$vmid.cleanup";
+}
+
+sub create_cleanup_flag {
+    my ($vmid) = @_;
+    PVE::File::file_set_contents(get_cleanup_path($vmid), time());
+}
+
+sub clear_cleanup_flag {
+    my ($vmid) = @_;
+    my $path = get_cleanup_path($vmid);
+    unlink $path or $! == POSIX::ENOENT or die "removing cleanup flag for $vmid failed: $!\n";
+}
+
+sub cleanup_flag_exists {
+    my ($vmid) = @_;
+    return -f get_cleanup_path($vmid);
+}
+
+# checks if /run/qemu-server/use_old_cleanup exists that will be created on
+# package update and cleared on bootup so we can be sure the guests were
+# started recently enough
+sub can_use_cleanup_flag {
+    !-f "/run/qemu-server/use_old_cleanup";
+}
 1;
diff --git a/src/test/MigrationTest/QmMock.pm b/src/test/MigrationTest/QmMock.pm
index 78be47d3..4d5fa3ac 100644
--- a/src/test/MigrationTest/QmMock.pm
+++ b/src/test/MigrationTest/QmMock.pm
@@ -77,6 +77,22 @@ $qemu_server_helpers_module->mock(
     },
 );
 
+my $qemu_server_runstate_module = Test::MockModule->new("PVE::QemuServer::RunState");
+$qemu_server_runstate_module->mock(
+    create_cleanup_flag => sub {
+        return undef;
+    },
+    clear_cleanup_flag => sub {
+        return undef;
+    },
+    cleanup_flag_exists => sub {
+        return 1;
+    },
+    can_use_cleanup_flag => sub {
+        return 1;
+    },
+);
+
 our $qemu_server_machine_module = Test::MockModule->new("PVE::QemuServer::Machine");
 $qemu_server_machine_module->mock(
     get_current_qemu_machine => sub {
-- 
2.47.3

[PATCH qemu-server v3 2/3] qm cleanup: die early when encountering a running stop mode backup

[Dominik Csapak]

this is an expected situation, so abort here early with a better message
than 'vm still running'

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
---
 src/PVE/CLI/qm.pm | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/src/PVE/CLI/qm.pm b/src/PVE/CLI/qm.pm
index b7bc4d9f..6aff5b7a 100755
--- a/src/PVE/CLI/qm.pm
+++ b/src/PVE/CLI/qm.pm
@@ -1102,6 +1102,13 @@ __PACKAGE__->register_method({
             sub {
                 my $conf = PVE::QemuConfig->load_config($vmid);
                 my $pid = PVE::QemuServer::check_running($vmid);
+
+                # With a stop mode backup, we might run here into a running vm with a backup
+                # lock, but this already did the cleanup and is an expected state, so abort
+                # here with a good message
+                die "skipping cleanup - 'backup' lock is present and vm is running again\n"
+                    if $pid && $clean && $conf->{lock} && $conf->{lock} eq 'backup';
+
                 die "vm still running\n" if $pid;
 
                 # Rollback already does cleanup when preparing and afterwards temporarily drops the
-- 
2.47.3

[PATCH qemu-server v3 3/3] fix #7119: qm cleanup: wait for process exiting for up to 30 seconds

[Dominik Csapak]

When qmeventd detects a vm exiting, it starts 'qm cleanup'.

Since the vm process exits is sometimes not instant, wait up to 30
seconds here to start the cleanup process instead of immediately
aborting if the pid still exits. This prevented executing the hookscript
on the 'post-stop' phase when either
* the cleanup mechanism is still the old one
* the guest was powered down from inside, not via the API

This can be reproduced by e.g. passing through a usb device, which
delays the qemu process exit for a few seconds (for most devices).

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
---
 src/PVE/CLI/qm.pm | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/src/PVE/CLI/qm.pm b/src/PVE/CLI/qm.pm
index 6aff5b7a..ee3ccedd 100755
--- a/src/PVE/CLI/qm.pm
+++ b/src/PVE/CLI/qm.pm
@@ -1101,7 +1101,7 @@ __PACKAGE__->register_method({
             60,
             sub {
                 my $conf = PVE::QemuConfig->load_config($vmid);
-                my $pid = PVE::QemuServer::check_running($vmid);
+                my $pid = PVE::QemuServer::Helpers::vm_running_locally($vmid);
 
                 # With a stop mode backup, we might run here into a running vm with a backup
                 # lock, but this already did the cleanup and is an expected state, so abort
@@ -1109,7 +1109,19 @@ __PACKAGE__->register_method({
                 die "skipping cleanup - 'backup' lock is present and vm is running again\n"
                     if $pid && $clean && $conf->{lock} && $conf->{lock} eq 'backup';
 
-                die "vm still running\n" if $pid;
+                # wait for some time until the QEMU process exits after the QMP
+                # 'SHUTDOWN' event, since this might not be instant
+                my $timeout = 30;
+                my $starttime = time();
+                warn "QEMU process $pid for VM $vmid still running (or newly started)\n"
+                    if $pid;
+
+                while ($pid && (time() - $starttime) < $timeout) {
+                    sleep(1);
+                    $pid = PVE::QemuServer::check_running($vmid);
+                }
+
+                die "vm still running after timeout - aborting cleanup\n" if $pid;
 
                 # Rollback already does cleanup when preparing and afterwards temporarily drops the
                 # lock on the configuration file to rollback the volumes. Deactivating volumes here
-- 
2.47.3