Re: [PATCH qemu-server v4 9/9] hotplug: remove iothread if adding drive device failed

[Fiona Ebner <f.ebner@proxmox.com>]

Am 23.10.26 um 3:04 PM schrieb Dominik Csapak:
> diff --git a/src/PVE/QemuServer.pm b/src/PVE/QemuServer.pm
> index 5be49090..95a8c6c1 100644
> --- a/src/PVE/QemuServer.pm
> +++ b/src/PVE/QemuServer.pm
> @@ -3913,6 +3913,8 @@ sub vm_deviceplug {

Above here are more calls that could fail and leave the iothread object
behind. We should expand the eval block and maybe have variables to
track which delete operations to do in the error case based on how far
we got. Or just attempt the delete operations always, shouldn't hurt either?

>          if (my $err = $@) {
>              eval { qemu_drivedel($vmid, $deviceid); };
>              warn $@ if $@;
> +            eval { qemu_iothread_del($vmid, $deviceid, $device) };
> +            warn $@ if $@;
>              die $err;
>          }
>      } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
> @@ -3932,7 +3934,14 @@ sub vm_deviceplug {
>          }
>  
>          qemu_deviceadd($vmid, $devicefull);

This call should be in the eval too.

> -        qemu_deviceaddverify($vmid, $deviceid);
> +        eval { qemu_deviceaddverify($vmid, $deviceid) };
> +        if (my $err = $@) {
> +            if ($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
> +                eval { qemu_iothread_del($vmid, $deviceid, $device) };
> +                warn $@ if $@;
> +            }
> +            die $err;
> +        }
>      } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
>          qemu_findorcreatescsihw($storecfg, $conf, $vmid, $device, $arch, $machine_type);
>          qemu_driveadd($storecfg, $vmid, $device);