PVE/QemuServer.pm

   1 package PVE::QemuServer;
   2
   3 use strict;
   4 use warnings;
   5
   6 use Cwd 'abs_path';
   7 use Digest::SHA;
   8 use Fcntl ':flock';
   9 use Fcntl;
  10 use File::Basename;
  11 use File::Copy qw(copy);
  12 use File::Path;
  13 use File::stat;
  14 use Getopt::Long;
  15 use IO::Dir;
  16 use IO::File;
  17 use IO::Handle;
  18 use IO::Select;
  19 use IO::Socket::UNIX;
  20 use IPC::Open3;
  21 use JSON;
  22 use MIME::Base64;
  23 use POSIX;
  24 use Storable qw(dclone);
  25 use Time::HiRes qw(gettimeofday);
  26 use URI::Escape;
  27 use UUID;
  28
  29 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file);
  30 use PVE::DataCenterConfig;
  31 use PVE::Exception qw(raise raise_param_exc);
  32 use PVE::GuestHelpers qw(safe_string_ne safe_num_ne safe_boolean_ne);
  33 use PVE::INotify;
  34 use PVE::JSONSchema qw(get_standard_option parse_property_string);
  35 use PVE::ProcFSTools;
  36 use PVE::RPCEnvironment;
  37 use PVE::Storage;
  38 use PVE::SysFSTools;
  39 use PVE::Systemd;
  40 use PVE::Tools qw(run_command file_read_firstline file_get_contents dir_glob_foreach get_host_arch $IPV6RE);
  41
  42 use PVE::QMPClient;
  43 use PVE::QemuConfig;
  44 use PVE::QemuServer::Helpers qw(min_version config_aware_timeout);
  45 use PVE::QemuServer::Cloudinit;
  46 use PVE::QemuServer::CPUConfig qw(print_cpu_device get_cpu_options);
  47 use PVE::QemuServer::Drive qw(is_valid_drivename drive_is_cloudinit drive_is_cdrom parse_drive print_drive);
  48 use PVE::QemuServer::Machine;
  49 use PVE::QemuServer::Memory;
  50 use PVE::QemuServer::Monitor qw(mon_cmd);
  51 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr print_pcie_root_port parse_hostpci);
  52 use PVE::QemuServer::USB qw(parse_usb_device);
  53
  54 my $have_sdn;
  55 eval {
  56     require PVE::Network::SDN::Zones;
  57     $have_sdn = 1;
  58 };
  59
  60 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
  61 my $OVMF = {
  62     x86_64 => [
  63         "$EDK2_FW_BASE/OVMF_CODE.fd",
  64         "$EDK2_FW_BASE/OVMF_VARS.fd"
  65     ],
  66     aarch64 => [
  67         "$EDK2_FW_BASE/AAVMF_CODE.fd",
  68         "$EDK2_FW_BASE/AAVMF_VARS.fd"
  69     ],
  70 };
  71
  72 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
  73
  74 # Note about locking: we use flock on the config file protect
  75 # against concurent actions.
  76 # Aditionaly, we have a 'lock' setting in the config file. This
  77 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
  78 # allowed when such lock is set. But you can ignore this kind of
  79 # lock with the --skiplock flag.
  80
  81 cfs_register_file('/qemu-server/',
  82                   \&parse_vm_config,
  83                   \&write_vm_config);
  84
  85 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
  86     description => "Some command save/restore state from this location.",
  87     type => 'string',
  88     maxLength => 128,
  89     optional => 1,
  90 });
  91
  92 PVE::JSONSchema::register_standard_option('pve-qemu-machine', {
  93         description => "Specifies the Qemu machine type.",
  94         type => 'string',
  95         pattern => '(pc|pc(-i440fx)?-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|q35|pc-q35-\d+(\.\d+)+(\+pve\d+)?(\.pxe)?|virt(?:-\d+(\.\d+)+)?(\+pve\d+)?)',
  96         maxLength => 40,
  97         optional => 1,
  98 });
  99
 100
 101 sub map_storage {
 102     my ($map, $source) = @_;
 103
 104     return $source if !defined($map);
 105
 106     return $map->{entries}->{$source}
 107         if $map->{entries} && defined($map->{entries}->{$source});
 108
 109     return $map->{default} if $map->{default};
 110
 111     # identity (fallback)
 112     return $source;
 113 }
 114
 115 PVE::JSONSchema::register_standard_option('pve-targetstorage', {
 116     description => "Mapping from source to target storages. Providing only a single storage ID maps all source storages to that storage. Providing the special value '1' will map each source storage to itself.",
 117     type => 'string',
 118     format => 'storagepair-list',
 119     optional => 1,
 120 });
 121
 122 #no warnings 'redefine';
 123
 124 sub cgroups_write {
 125    my ($controller, $vmid, $option, $value) = @_;
 126
 127    my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
 128    PVE::ProcFSTools::write_proc_entry($path, $value);
 129
 130 }
 131
 132 my $nodename_cache;
 133 sub nodename {
 134     $nodename_cache //= PVE::INotify::nodename();
 135     return $nodename_cache;
 136 }
 137
 138 my $watchdog_fmt = {
 139     model => {
 140         default_key => 1,
 141         type => 'string',
 142         enum => [qw(i6300esb ib700)],
 143         description => "Watchdog type to emulate.",
 144         default => 'i6300esb',
 145         optional => 1,
 146     },
 147     action => {
 148         type => 'string',
 149         enum => [qw(reset shutdown poweroff pause debug none)],
 150         description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
 151         optional => 1,
 152     },
 153 };
 154 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
 155
 156 my $agent_fmt = {
 157     enabled => {
 158         description => "Enable/disable Qemu GuestAgent.",
 159         type => 'boolean',
 160         default => 0,
 161         default_key => 1,
 162     },
 163     fstrim_cloned_disks => {
 164         description => "Run fstrim after cloning/moving a disk.",
 165         type => 'boolean',
 166         optional => 1,
 167         default => 0
 168     },
 169     type => {
 170         description => "Select the agent type",
 171         type => 'string',
 172         default => 'virtio',
 173         optional => 1,
 174         enum => [qw(virtio isa)],
 175     },
 176 };
 177
 178 my $vga_fmt = {
 179     type => {
 180         description => "Select the VGA type.",
 181         type => 'string',
 182         default => 'std',
 183         optional => 1,
 184         default_key => 1,
 185         enum => [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio vmware)],
 186     },
 187     memory => {
 188         description => "Sets the VGA memory (in MiB). Has no effect with serial display.",
 189         type => 'integer',
 190         optional => 1,
 191         minimum => 4,
 192         maximum => 512,
 193     },
 194 };
 195
 196 my $ivshmem_fmt = {
 197     size => {
 198         type => 'integer',
 199         minimum => 1,
 200         description => "The size of the file in MB.",
 201     },
 202     name => {
 203         type => 'string',
 204         pattern => '[a-zA-Z0-9\-]+',
 205         optional => 1,
 206         format_description => 'string',
 207         description => "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
 208     },
 209 };
 210
 211 my $audio_fmt = {
 212     device => {
 213         type => 'string',
 214         enum => [qw(ich9-intel-hda intel-hda AC97)],
 215         description =>  "Configure an audio device."
 216     },
 217     driver =>  {
 218         type => 'string',
 219         enum => ['spice'],
 220         default => 'spice',
 221         optional => 1,
 222         description => "Driver backend for the audio device."
 223     },
 224 };
 225
 226 my $spice_enhancements_fmt = {
 227     foldersharing => {
 228         type => 'boolean',
 229         optional => 1,
 230         default => '0',
 231         description =>  "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
 232     },
 233     videostreaming =>  {
 234         type => 'string',
 235         enum => ['off', 'all', 'filter'],
 236         default => 'off',
 237         optional => 1,
 238         description => "Enable video streaming. Uses compression for detected video streams."
 239     },
 240 };
 241
 242 my $rng_fmt = {
 243     source => {
 244         type => 'string',
 245         enum => ['/dev/urandom', '/dev/random', '/dev/hwrng'],
 246         default_key => 1,
 247         description => "The file on the host to gather entropy from. In most"
 248                      . " cases /dev/urandom should be preferred over /dev/random"
 249                      . " to avoid entropy-starvation issues on the host. Using"
 250                      . " urandom does *not* decrease security in any meaningful"
 251                      . " way, as it's still seeded from real entropy, and the"
 252                      . " bytes provided will most likely be mixed with real"
 253                      . " entropy on the guest as well. /dev/hwrng can be used"
 254                      . " to pass through a hardware RNG from the host.",
 255     },
 256     max_bytes => {
 257         type => 'integer',
 258         description => "Maximum bytes of entropy injected into the guest every"
 259                      . " 'period' milliseconds. Prefer a lower value when using"
 260                      . " /dev/random as source. Use 0 to disable limiting"
 261                      . " (potentially dangerous!).",
 262         optional => 1,
 263
 264         # default is 1 KiB/s, provides enough entropy to the guest to avoid
 265         # boot-starvation issues (e.g. systemd etc...) while allowing no chance
 266         # of overwhelming the host, provided we're reading from /dev/urandom
 267         default => 1024,
 268     },
 269     period => {
 270         type => 'integer',
 271         description => "Every 'period' milliseconds the entropy-injection quota"
 272                      . " is reset, allowing the guest to retrieve another"
 273                      . " 'max_bytes' of entropy.",
 274         optional => 1,
 275         default => 1000,
 276     },
 277 };
 278
 279 my $confdesc = {
 280     onboot => {
 281         optional => 1,
 282         type => 'boolean',
 283         description => "Specifies whether a VM will be started during system bootup.",
 284         default => 0,
 285     },
 286     autostart => {
 287         optional => 1,
 288         type => 'boolean',
 289         description => "Automatic restart after crash (currently ignored).",
 290         default => 0,
 291     },
 292     hotplug => {
 293         optional => 1,
 294         type => 'string', format => 'pve-hotplug-features',
 295         description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
 296         default => 'network,disk,usb',
 297     },
 298     reboot => {
 299         optional => 1,
 300         type => 'boolean',
 301         description => "Allow reboot. If set to '0' the VM exit on reboot.",
 302         default => 1,
 303     },
 304     lock => {
 305         optional => 1,
 306         type => 'string',
 307         description => "Lock/unlock the VM.",
 308         enum => [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
 309     },
 310     cpulimit => {
 311         optional => 1,
 312         type => 'number',
 313         description => "Limit of CPU usage.",
 314         verbose_description => "Limit of CPU usage.\n\nNOTE: If the computer has 2 CPUs, it has total of '2' CPU time. Value '0' indicates no CPU limit.",
 315         minimum => 0,
 316         maximum => 128,
 317         default => 0,
 318     },
 319     cpuunits => {
 320         optional => 1,
 321         type => 'integer',
 322         description => "CPU weight for a VM.",
 323         verbose_description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.",
 324         minimum => 2,
 325         maximum => 262144,
 326         default => 1024,
 327     },
 328     memory => {
 329         optional => 1,
 330         type => 'integer',
 331         description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
 332         minimum => 16,
 333         default => 512,
 334     },
 335     balloon => {
 336         optional => 1,
 337         type => 'integer',
 338         description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
 339         minimum => 0,
 340     },
 341     shares => {
 342         optional => 1,
 343         type => 'integer',
 344         description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning. Auto-ballooning is done by pvestatd.",
 345         minimum => 0,
 346         maximum => 50000,
 347         default => 1000,
 348     },
 349     keyboard => {
 350         optional => 1,
 351         type => 'string',
 352         description => "Keybord layout for vnc server. Default is read from the '/etc/pve/datacenter.cfg' configuration file.".
 353                        "It should not be necessary to set it.",
 354         enum => PVE::Tools::kvmkeymaplist(),
 355         default => undef,
 356     },
 357     name => {
 358         optional => 1,
 359         type => 'string', format => 'dns-name',
 360         description => "Set a name for the VM. Only used on the configuration web interface.",
 361     },
 362     scsihw => {
 363         optional => 1,
 364         type => 'string',
 365         description => "SCSI controller model",
 366         enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
 367         default => 'lsi',
 368     },
 369     description => {
 370         optional => 1,
 371         type => 'string',
 372         description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
 373     },
 374     ostype => {
 375         optional => 1,
 376         type => 'string',
 377         enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 l24 l26 solaris)],
 378         description => "Specify guest operating system.",
 379         verbose_description => <<EODESC,
 380 Specify guest operating system. This is used to enable special
 381 optimization/features for specific operating systems:
 382
 383 [horizontal]
 384 other;; unspecified OS
 385 wxp;; Microsoft Windows XP
 386 w2k;; Microsoft Windows 2000
 387 w2k3;; Microsoft Windows 2003
 388 w2k8;; Microsoft Windows 2008
 389 wvista;; Microsoft Windows Vista
 390 win7;; Microsoft Windows 7
 391 win8;; Microsoft Windows 8/2012/2012r2
 392 win10;; Microsoft Windows 10/2016
 393 l24;; Linux 2.4 Kernel
 394 l26;; Linux 2.6 - 5.X Kernel
 395 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
 396 EODESC
 397     },
 398     boot => {
 399         optional => 1,
 400         type => 'string', format => 'pve-qm-boot',
 401         description => "Specify guest boot order. Use with 'order=', usage with"
 402                      . " no key or 'legacy=' is deprecated.",
 403     },
 404     bootdisk => {
 405         optional => 1,
 406         type => 'string', format => 'pve-qm-bootdisk',
 407         description => "Enable booting from specified disk. Deprecated: Use 'boot: order=foo;bar' instead.",
 408         pattern => '(ide|sata|scsi|virtio)\d+',
 409     },
 410     smp => {
 411         optional => 1,
 412         type => 'integer',
 413         description => "The number of CPUs. Please use option -sockets instead.",
 414         minimum => 1,
 415         default => 1,
 416     },
 417     sockets => {
 418         optional => 1,
 419         type => 'integer',
 420         description => "The number of CPU sockets.",
 421         minimum => 1,
 422         default => 1,
 423     },
 424     cores => {
 425         optional => 1,
 426         type => 'integer',
 427         description => "The number of cores per socket.",
 428         minimum => 1,
 429         default => 1,
 430     },
 431     numa => {
 432         optional => 1,
 433         type => 'boolean',
 434         description => "Enable/disable NUMA.",
 435         default => 0,
 436     },
 437     hugepages => {
 438         optional => 1,
 439         type => 'string',
 440         description => "Enable/disable hugepages memory.",
 441         enum => [qw(any 2 1024)],
 442     },
 443     keephugepages => {
 444         optional => 1,
 445         type => 'boolean',
 446         default => 0,
 447         description => "Use together with hugepages. If enabled, hugepages will not not be deleted"
 448             ." after VM shutdown and can be used for subsequent starts.",
 449     },
 450     vcpus => {
 451         optional => 1,
 452         type => 'integer',
 453         description => "Number of hotplugged vcpus.",
 454         minimum => 1,
 455         default => 0,
 456     },
 457     acpi => {
 458         optional => 1,
 459         type => 'boolean',
 460         description => "Enable/disable ACPI.",
 461         default => 1,
 462     },
 463     agent => {
 464         optional => 1,
 465         description => "Enable/disable Qemu GuestAgent and its properties.",
 466         type => 'string',
 467         format => $agent_fmt,
 468     },
 469     kvm => {
 470         optional => 1,
 471         type => 'boolean',
 472         description => "Enable/disable KVM hardware virtualization.",
 473         default => 1,
 474     },
 475     tdf => {
 476         optional => 1,
 477         type => 'boolean',
 478         description => "Enable/disable time drift fix.",
 479         default => 0,
 480     },
 481     localtime => {
 482         optional => 1,
 483         type => 'boolean',
 484         description => "Set the real time clock to local time. This is enabled by default if ostype"
 485             ." indicates a Microsoft OS.",
 486     },
 487     freeze => {
 488         optional => 1,
 489         type => 'boolean',
 490         description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
 491     },
 492     vga => {
 493         optional => 1,
 494         type => 'string', format => $vga_fmt,
 495         description => "Configure the VGA hardware.",
 496         verbose_description => "Configure the VGA Hardware. If you want to use high resolution"
 497             ." modes (>= 1280x1024x16) you may need to increase the vga memory option. Since QEMU"
 498             ." 2.9 the default VGA display type is 'std' for all OS types besides some Windows"
 499             ." versions (XP and older) which use 'cirrus'. The 'qxl' option enables the SPICE"
 500             ." display server. For win* OS you can select how many independent displays you want,"
 501             ." Linux guests can add displays them self.\nYou can also run without any graphic card,"
 502             ." using a serial device as terminal.",
 503     },
 504     watchdog => {
 505         optional => 1,
 506         type => 'string', format => 'pve-qm-watchdog',
 507         description => "Create a virtual hardware watchdog device.",
 508         verbose_description => "Create a virtual hardware watchdog device. Once enabled (by a guest"
 509             ." action), the watchdog must be periodically polled by an agent inside the guest or"
 510             ." else the watchdog will reset the guest (or execute the respective action specified)",
 511     },
 512     startdate => {
 513         optional => 1,
 514         type => 'string',
 515         typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
 516         description => "Set the initial date of the real time clock. Valid format for date are:"
 517             ."'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
 518         pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
 519         default => 'now',
 520     },
 521     startup =>  get_standard_option('pve-startup-order'),
 522     template => {
 523         optional => 1,
 524         type => 'boolean',
 525         description => "Enable/disable Template.",
 526         default => 0,
 527     },
 528     args => {
 529         optional => 1,
 530         type => 'string',
 531         description => "Arbitrary arguments passed to kvm.",
 532         verbose_description => <<EODESCR,
 533 Arbitrary arguments passed to kvm, for example:
 534
 535 args: -no-reboot -no-hpet
 536
 537 NOTE: this option is for experts only.
 538 EODESCR
 539     },
 540     tablet => {
 541         optional => 1,
 542         type => 'boolean',
 543         default => 1,
 544         description => "Enable/disable the USB tablet device.",
 545         verbose_description => "Enable/disable the USB tablet device. This device is usually needed"
 546             ." to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with"
 547             ." normal VNC clients. If you're running lots of console-only guests on one host, you"
 548             ." may consider disabling this to save some context switches. This is turned off by"
 549             ." default if you use spice (`qm set <vmid> --vga qxl`).",
 550     },
 551     migrate_speed => {
 552         optional => 1,
 553         type => 'integer',
 554         description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
 555         minimum => 0,
 556         default => 0,
 557     },
 558     migrate_downtime => {
 559         optional => 1,
 560         type => 'number',
 561         description => "Set maximum tolerated downtime (in seconds) for migrations.",
 562         minimum => 0,
 563         default => 0.1,
 564     },
 565     cdrom => {
 566         optional => 1,
 567         type => 'string', format => 'pve-qm-ide',
 568         typetext => '<volume>',
 569         description => "This is an alias for option -ide2",
 570     },
 571     cpu => {
 572         optional => 1,
 573         description => "Emulated CPU type.",
 574         type => 'string',
 575         format => 'pve-vm-cpu-conf',
 576     },
 577     parent => get_standard_option('pve-snapshot-name', {
 578         optional => 1,
 579         description => "Parent snapshot name. This is used internally, and should not be modified.",
 580     }),
 581     snaptime => {
 582         optional => 1,
 583         description => "Timestamp for snapshots.",
 584         type => 'integer',
 585         minimum => 0,
 586     },
 587     vmstate => {
 588         optional => 1,
 589         type => 'string', format => 'pve-volume-id',
 590         description => "Reference to a volume which stores the VM state. This is used internally"
 591             ." for snapshots.",
 592     },
 593     vmstatestorage => get_standard_option('pve-storage-id', {
 594         description => "Default storage for VM state volumes/files.",
 595         optional => 1,
 596     }),
 597     runningmachine => get_standard_option('pve-qemu-machine', {
 598         description => "Specifies the QEMU machine type of the running vm. This is used internally"
 599             ." for snapshots.",
 600     }),
 601     runningcpu => {
 602         description => "Specifies the QEMU '-cpu' parameter of the running vm. This is used"
 603             ." internally for snapshots.",
 604         optional => 1,
 605         type => 'string',
 606         pattern => $PVE::QemuServer::CPUConfig::qemu_cmdline_cpu_re,
 607         format_description => 'QEMU -cpu parameter'
 608     },
 609     machine => get_standard_option('pve-qemu-machine'),
 610     arch => {
 611         description => "Virtual processor architecture. Defaults to the host.",
 612         optional => 1,
 613         type => 'string',
 614         enum => [qw(x86_64 aarch64)],
 615     },
 616     smbios1 => {
 617         description => "Specify SMBIOS type 1 fields.",
 618         type => 'string', format => 'pve-qm-smbios1',
 619         maxLength => 512,
 620         optional => 1,
 621     },
 622     protection => {
 623         optional => 1,
 624         type => 'boolean',
 625         description => "Sets the protection flag of the VM. This will disable the remove VM and"
 626             ." remove disk operations.",
 627         default => 0,
 628     },
 629     bios => {
 630         optional => 1,
 631         type => 'string',
 632         enum => [ qw(seabios ovmf) ],
 633         description => "Select BIOS implementation.",
 634         default => 'seabios',
 635     },
 636     vmgenid => {
 637         type => 'string',
 638         pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
 639         format_description => 'UUID',
 640         description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0'"
 641             ." to disable explicitly.",
 642         verbose_description => "The VM generation ID (vmgenid) device exposes a 128-bit integer"
 643             ." value identifier to the guest OS. This allows to notify the guest operating system"
 644             ." when the virtual machine is executed with a different configuration (e.g. snapshot"
 645             ." execution or creation from a template). The guest operating system notices the"
 646             ." change, and is then able to react as appropriate by marking its copies of"
 647             ." distributed databases as dirty, re-initializing its random number generator, etc.\n"
 648             ."Note that auto-creation only works when done through API/CLI create or update methods"
 649             .", but not when manually editing the config file.",
 650         default => "1 (autogenerated)",
 651         optional => 1,
 652     },
 653     hookscript => {
 654         type => 'string',
 655         format => 'pve-volume-id',
 656         optional => 1,
 657         description => "Script that will be executed during various steps in the vms lifetime.",
 658     },
 659     ivshmem => {
 660         type => 'string',
 661         format => $ivshmem_fmt,
 662         description => "Inter-VM shared memory. Useful for direct communication between VMs, or to"
 663             ." the host.",
 664         optional => 1,
 665     },
 666     audio0 => {
 667         type => 'string',
 668         format => $audio_fmt,
 669         description => "Configure a audio device, useful in combination with QXL/Spice.",
 670         optional => 1
 671     },
 672     spice_enhancements => {
 673         type => 'string',
 674         format => $spice_enhancements_fmt,
 675         description => "Configure additional enhancements for SPICE.",
 676         optional => 1
 677     },
 678     tags => {
 679         type => 'string', format => 'pve-tag-list',
 680         description => 'Tags of the VM. This is only meta information.',
 681         optional => 1,
 682     },
 683     rng0 => {
 684         type => 'string',
 685         format => $rng_fmt,
 686         description => "Configure a VirtIO-based Random Number Generator.",
 687         optional => 1,
 688     },
 689 };
 690
 691 my $cicustom_fmt = {
 692     meta => {
 693         type => 'string',
 694         optional => 1,
 695         description => 'Specify a custom file containing all meta data passed to the VM via"
 696             ." cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
 697         format => 'pve-volume-id',
 698         format_description => 'volume',
 699     },
 700     network => {
 701         type => 'string',
 702         optional => 1,
 703         description => 'Specify a custom file containing all network data passed to the VM via'
 704             .' cloud-init.',
 705         format => 'pve-volume-id',
 706         format_description => 'volume',
 707     },
 708     user => {
 709         type => 'string',
 710         optional => 1,
 711         description => 'Specify a custom file containing all user data passed to the VM via'
 712             .' cloud-init.',
 713         format => 'pve-volume-id',
 714         format_description => 'volume',
 715     },
 716 };
 717 PVE::JSONSchema::register_format('pve-qm-cicustom', $cicustom_fmt);
 718
 719 my $confdesc_cloudinit = {
 720     citype => {
 721         optional => 1,
 722         type => 'string',
 723         description => 'Specifies the cloud-init configuration format. The default depends on the'
 724             .' configured operating system type (`ostype`. We use the `nocloud` format for Linux,'
 725             .' and `configdrive2` for windows.',
 726         enum => ['configdrive2', 'nocloud'],
 727     },
 728     ciuser => {
 729         optional => 1,
 730         type => 'string',
 731         description => "cloud-init: User name to change ssh keys and password for instead of the"
 732             ." image's configured default user.",
 733     },
 734     cipassword => {
 735         optional => 1,
 736         type => 'string',
 737         description => 'cloud-init: Password to assign the user. Using this is generally not'
 738             .' recommended. Use ssh keys instead. Also note that older cloud-init versions do not'
 739             .' support hashed passwords.',
 740     },
 741     cicustom => {
 742         optional => 1,
 743         type => 'string',
 744         description => 'cloud-init: Specify custom files to replace the automatically generated'
 745             .' ones at start.',
 746         format => 'pve-qm-cicustom',
 747     },
 748     searchdomain => {
 749         optional => 1,
 750         type => 'string',
 751         description => "cloud-init: Sets DNS search domains for a container. Create will'
 752             .' automatically use the setting from the host if neither searchdomain nor nameserver'
 753             .' are set.",
 754     },
 755     nameserver => {
 756         optional => 1,
 757         type => 'string', format => 'address-list',
 758         description => "cloud-init: Sets DNS server IP address for a container. Create will'
 759             .' automatically use the setting from the host if neither searchdomain nor nameserver'
 760             .' are set.",
 761     },
 762     sshkeys => {
 763         optional => 1,
 764         type => 'string',
 765         format => 'urlencoded',
 766         description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
 767     },
 768 };
 769
 770 # what about other qemu settings ?
 771 #cpu => 'string',
 772 #machine => 'string',
 773 #fda => 'file',
 774 #fdb => 'file',
 775 #mtdblock => 'file',
 776 #sd => 'file',
 777 #pflash => 'file',
 778 #snapshot => 'bool',
 779 #bootp => 'file',
 780 ##tftp => 'dir',
 781 ##smb => 'dir',
 782 #kernel => 'file',
 783 #append => 'string',
 784 #initrd => 'file',
 785 ##soundhw => 'string',
 786
 787 while (my ($k, $v) = each %$confdesc) {
 788     PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
 789 }
 790
 791 my $MAX_USB_DEVICES = 5;
 792 my $MAX_NETS = 32;
 793 my $MAX_SERIAL_PORTS = 4;
 794 my $MAX_PARALLEL_PORTS = 3;
 795 my $MAX_NUMA = 8;
 796
 797 my $numa_fmt = {
 798     cpus => {
 799         type => "string",
 800         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 801         description => "CPUs accessing this NUMA node.",
 802         format_description => "id[-id];...",
 803     },
 804     memory => {
 805         type => "number",
 806         description => "Amount of memory this NUMA node provides.",
 807         optional => 1,
 808     },
 809     hostnodes => {
 810         type => "string",
 811         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 812         description => "Host NUMA nodes to use.",
 813         format_description => "id[-id];...",
 814         optional => 1,
 815     },
 816     policy => {
 817         type => 'string',
 818         enum => [qw(preferred bind interleave)],
 819         description => "NUMA allocation policy.",
 820         optional => 1,
 821     },
 822 };
 823 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
 824 my $numadesc = {
 825     optional => 1,
 826     type => 'string', format => $numa_fmt,
 827     description => "NUMA topology.",
 828 };
 829 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
 830
 831 for (my $i = 0; $i < $MAX_NUMA; $i++)  {
 832     $confdesc->{"numa$i"} = $numadesc;
 833 }
 834
 835 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000',  'pcnet',  'virtio',
 836                       'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
 837                       'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
 838 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
 839
 840 my $net_fmt_bridge_descr = <<__EOD__;
 841 Bridge to attach the network device to. The Proxmox VE standard bridge
 842 is called 'vmbr0'.
 843
 844 If you do not specify a bridge, we create a kvm user (NATed) network
 845 device, which provides DHCP and DNS services. The following addresses
 846 are used:
 847
 848  10.0.2.2   Gateway
 849  10.0.2.3   DNS Server
 850  10.0.2.4   SMB Server
 851
 852 The DHCP server assign addresses to the guest starting from 10.0.2.15.
 853 __EOD__
 854
 855 my $net_fmt = {
 856     macaddr  => get_standard_option('mac-addr', {
 857         description => "MAC address. That address must be unique withing your network. This is"
 858             ." automatically generated if not specified.",
 859     }),
 860     model => {
 861         type => 'string',
 862         description => "Network Card Model. The 'virtio' model provides the best performance with"
 863             ." very low CPU overhead. If your guest does not support this driver, it is usually"
 864             ." best to use 'e1000'.",
 865         enum => $nic_model_list,
 866         default_key => 1,
 867     },
 868     (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
 869     bridge => {
 870         type => 'string',
 871         description => $net_fmt_bridge_descr,
 872         format_description => 'bridge',
 873         pattern => '[-_.\w\d]+',
 874         optional => 1,
 875     },
 876     queues => {
 877         type => 'integer',
 878         minimum => 0, maximum => 16,
 879         description => 'Number of packet queues to be used on the device.',
 880         optional => 1,
 881     },
 882     rate => {
 883         type => 'number',
 884         minimum => 0,
 885         description => "Rate limit in mbps (megabytes per second) as floating point number.",
 886         optional => 1,
 887     },
 888     tag => {
 889         type => 'integer',
 890         minimum => 1, maximum => 4094,
 891         description => 'VLAN tag to apply to packets on this interface.',
 892         optional => 1,
 893     },
 894     trunks => {
 895         type => 'string',
 896         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 897         description => 'VLAN trunks to pass through this interface.',
 898         format_description => 'vlanid[;vlanid...]',
 899         optional => 1,
 900     },
 901     firewall => {
 902         type => 'boolean',
 903         description => 'Whether this interface should be protected by the firewall.',
 904         optional => 1,
 905     },
 906     link_down => {
 907         type => 'boolean',
 908         description => 'Whether this interface should be disconnected (like pulling the plug).',
 909         optional => 1,
 910     },
 911     mtu => {
 912         type => 'integer',
 913         minimum => 1, maximum => 65520,
 914         description => "Force MTU, for VirtIO only. Set to '1' to use the bridge MTU",
 915         optional => 1,
 916     },
 917 };
 918
 919 my $netdesc = {
 920     optional => 1,
 921     type => 'string', format => $net_fmt,
 922     description => "Specify network devices.",
 923 };
 924
 925 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
 926
 927 my $ipconfig_fmt = {
 928     ip => {
 929         type => 'string',
 930         format => 'pve-ipv4-config',
 931         format_description => 'IPv4Format/CIDR',
 932         description => 'IPv4 address in CIDR format.',
 933         optional => 1,
 934         default => 'dhcp',
 935     },
 936     gw => {
 937         type => 'string',
 938         format => 'ipv4',
 939         format_description => 'GatewayIPv4',
 940         description => 'Default gateway for IPv4 traffic.',
 941         optional => 1,
 942         requires => 'ip',
 943     },
 944     ip6 => {
 945         type => 'string',
 946         format => 'pve-ipv6-config',
 947         format_description => 'IPv6Format/CIDR',
 948         description => 'IPv6 address in CIDR format.',
 949         optional => 1,
 950         default => 'dhcp',
 951     },
 952     gw6 => {
 953         type => 'string',
 954         format => 'ipv6',
 955         format_description => 'GatewayIPv6',
 956         description => 'Default gateway for IPv6 traffic.',
 957         optional => 1,
 958         requires => 'ip6',
 959     },
 960 };
 961 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
 962 my $ipconfigdesc = {
 963     optional => 1,
 964     type => 'string', format => 'pve-qm-ipconfig',
 965     description => <<'EODESCR',
 966 cloud-init: Specify IP addresses and gateways for the corresponding interface.
 967
 968 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
 969
 970 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit
 971 gateway should be provided.
 972 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
 973
 974 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using
 975 dhcp on IPv4.
 976 EODESCR
 977 };
 978 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
 979
 980 for (my $i = 0; $i < $MAX_NETS; $i++)  {
 981     $confdesc->{"net$i"} = $netdesc;
 982     $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
 983 }
 984
 985 foreach my $key (keys %$confdesc_cloudinit) {
 986     $confdesc->{$key} = $confdesc_cloudinit->{$key};
 987 }
 988
 989 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
 990 sub verify_volume_id_or_qm_path {
 991     my ($volid, $noerr) = @_;
 992
 993     if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
 994         return $volid;
 995     }
 996
 997     # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
 998     $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
 999     if ($@) {
1000         return if $noerr;
1001         die $@;
1002     }
1003     return $volid;
1004 }
1005
1006 my $usb_fmt = {
1007     host => {
1008         default_key => 1,
1009         type => 'string', format => 'pve-qm-usb-device',
1010         format_description => 'HOSTUSBDEVICE|spice',
1011         description => <<EODESCR,
1012 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1013
1014  'bus-port(.port)*' (decimal numbers) or
1015  'vendor_id:product_id' (hexadeciaml numbers) or
1016  'spice'
1017
1018 You can use the 'lsusb -t' command to list existing usb devices.
1019
1020 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1021 machines - use with special care.
1022
1023 The value 'spice' can be used to add a usb redirection devices for spice.
1024 EODESCR
1025     },
1026     usb3 => {
1027         optional => 1,
1028         type => 'boolean',
1029         description => "Specifies whether if given host option is a USB3 device or port.",
1030         default => 0,
1031     },
1032 };
1033
1034 my $usbdesc = {
1035     optional => 1,
1036     type => 'string', format => $usb_fmt,
1037     description => "Configure an USB device (n is 0 to 4).",
1038 };
1039 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1040
1041 my $serialdesc = {
1042         optional => 1,
1043         type => 'string',
1044         pattern => '(/dev/.+|socket)',
1045         description =>  "Create a serial device inside the VM (n is 0 to 3)",
1046         verbose_description =>  <<EODESCR,
1047 Create a serial device inside the VM (n is 0 to 3), and pass through a
1048 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1049 host side (use 'qm terminal' to open a terminal connection).
1050
1051 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines -
1052 use with special care.
1053
1054 CAUTION: Experimental! User reported problems with this option.
1055 EODESCR
1056 };
1057
1058 my $paralleldesc= {
1059         optional => 1,
1060         type => 'string',
1061         pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1062         description =>  "Map host parallel devices (n is 0 to 2).",
1063         verbose_description =>  <<EODESCR,
1064 Map host parallel devices (n is 0 to 2).
1065
1066 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such
1067 machines - use with special care.
1068
1069 CAUTION: Experimental! User reported problems with this option.
1070 EODESCR
1071 };
1072
1073 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
1074     $confdesc->{"parallel$i"} = $paralleldesc;
1075 }
1076
1077 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
1078     $confdesc->{"serial$i"} = $serialdesc;
1079 }
1080
1081 for (my $i = 0; $i < $PVE::QemuServer::PCI::MAX_HOSTPCI_DEVICES; $i++)  {
1082     $confdesc->{"hostpci$i"} = $PVE::QemuServer::PCI::hostpcidesc;
1083 }
1084
1085 for my $key (keys %{$PVE::QemuServer::Drive::drivedesc_hash}) {
1086     $confdesc->{$key} = $PVE::QemuServer::Drive::drivedesc_hash->{$key};
1087 }
1088
1089 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++)  {
1090     $confdesc->{"usb$i"} = $usbdesc;
1091 }
1092
1093 my $boot_fmt = {
1094     legacy => {
1095         optional => 1,
1096         default_key => 1,
1097         type => 'string',
1098         description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n)."
1099                      . " Deprecated, use 'order=' instead.",
1100         pattern => '[acdn]{1,4}',
1101         format_description => "[acdn]{1,4}",
1102
1103         # note: this is also the fallback if boot: is not given at all
1104         default => 'cdn',
1105     },
1106     order => {
1107         optional => 1,
1108         type => 'string',
1109         format => 'pve-qm-bootdev-list',
1110         format_description => "device[;device...]",
1111         description => <<EODESC,
1112 The guest will attempt to boot from devices in the order they appear here.
1113
1114 Disks, optical drives and passed-through storage USB devices will be directly
1115 booted from, NICs will load PXE, and PCIe devices will either behave like disks
1116 (e.g. NVMe) or load an option ROM (e.g. RAID controller, hardware NIC).
1117
1118 Note that only devices in this list will be marked as bootable and thus loaded
1119 by the guest firmware (BIOS/UEFI). If you require multiple disks for booting
1120 (e.g. software-raid), you need to specify all of them here.
1121
1122 Overrides the deprecated 'legacy=[acdn]*' value when given.
1123 EODESC
1124     },
1125 };
1126 PVE::JSONSchema::register_format('pve-qm-boot', $boot_fmt);
1127
1128 PVE::JSONSchema::register_format('pve-qm-bootdev', \&verify_bootdev);
1129 sub verify_bootdev {
1130     my ($dev, $noerr) = @_;
1131
1132     return $dev if PVE::QemuServer::Drive::is_valid_drivename($dev) && $dev !~ m/^efidisk/;
1133
1134     my $check = sub {
1135         my ($base) = @_;
1136         return 0 if $dev !~ m/^$base\d+$/;
1137         return 0 if !$confdesc->{$dev};
1138         return 1;
1139     };
1140
1141     return $dev if $check->("net");
1142     return $dev if $check->("usb");
1143     return $dev if $check->("hostpci");
1144
1145     return if $noerr;
1146     die "invalid boot device '$dev'\n";
1147 }
1148
1149 sub print_bootorder {
1150     my ($devs) = @_;
1151     return "" if !@$devs;
1152     my $data = { order => join(';', @$devs) };
1153     return PVE::JSONSchema::print_property_string($data, $boot_fmt);
1154 }
1155
1156 my $kvm_api_version = 0;
1157
1158 sub kvm_version {
1159     return $kvm_api_version if $kvm_api_version;
1160
1161     open my $fh, '<', '/dev/kvm' or return;
1162
1163     # 0xae00 => KVM_GET_API_VERSION
1164     $kvm_api_version = ioctl($fh, 0xae00, 0);
1165     close($fh);
1166
1167     return $kvm_api_version;
1168 }
1169
1170 my $kvm_user_version = {};
1171 my $kvm_mtime = {};
1172
1173 sub kvm_user_version {
1174     my ($binary) = @_;
1175
1176     $binary //= get_command_for_arch(get_host_arch()); # get the native arch by default
1177     my $st = stat($binary);
1178
1179     my $cachedmtime = $kvm_mtime->{$binary} // -1;
1180     return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1181         $cachedmtime == $st->mtime;
1182
1183     $kvm_user_version->{$binary} = 'unknown';
1184     $kvm_mtime->{$binary} = $st->mtime;
1185
1186     my $code = sub {
1187         my $line = shift;
1188         if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1189             $kvm_user_version->{$binary} = $2;
1190         }
1191     };
1192
1193     eval { run_command([$binary, '--version'], outfunc => $code); };
1194     warn $@ if $@;
1195
1196     return $kvm_user_version->{$binary};
1197
1198 }
1199 my sub extract_version {
1200     my ($machine_type, $version) = @_;
1201     $version = kvm_user_version() if !defined($version);
1202     PVE::QemuServer::Machine::extract_version($machine_type, $version)
1203 }
1204
1205 sub kernel_has_vhost_net {
1206     return -c '/dev/vhost-net';
1207 }
1208
1209 sub option_exists {
1210     my $key = shift;
1211     return defined($confdesc->{$key});
1212 }
1213
1214 my $cdrom_path;
1215 sub get_cdrom_path {
1216
1217     return  $cdrom_path if $cdrom_path;
1218
1219     return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1220     return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1221     return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1222 }
1223
1224 sub get_iso_path {
1225     my ($storecfg, $vmid, $cdrom) = @_;
1226
1227     if ($cdrom eq 'cdrom') {
1228         return get_cdrom_path();
1229     } elsif ($cdrom eq 'none') {
1230         return '';
1231     } elsif ($cdrom =~ m|^/|) {
1232         return $cdrom;
1233     } else {
1234         return PVE::Storage::path($storecfg, $cdrom);
1235     }
1236 }
1237
1238 # try to convert old style file names to volume IDs
1239 sub filename_to_volume_id {
1240     my ($vmid, $file, $media) = @_;
1241
1242      if (!($file eq 'none' || $file eq 'cdrom' ||
1243           $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1244
1245         return if $file =~ m|/|;
1246
1247         if ($media && $media eq 'cdrom') {
1248             $file = "local:iso/$file";
1249         } else {
1250             $file = "local:$vmid/$file";
1251         }
1252     }
1253
1254     return $file;
1255 }
1256
1257 sub verify_media_type {
1258     my ($opt, $vtype, $media) = @_;
1259
1260     return if !$media;
1261
1262     my $etype;
1263     if ($media eq 'disk') {
1264         $etype = 'images';
1265     } elsif ($media eq 'cdrom') {
1266         $etype = 'iso';
1267     } else {
1268         die "internal error";
1269     }
1270
1271     return if ($vtype eq $etype);
1272
1273     raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1274 }
1275
1276 sub cleanup_drive_path {
1277     my ($opt, $storecfg, $drive) = @_;
1278
1279     # try to convert filesystem paths to volume IDs
1280
1281     if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1282         ($drive->{file} !~ m|^/dev/.+|) &&
1283         ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1284         ($drive->{file} !~ m/^\d+$/)) {
1285         my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1286         raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"})
1287             if !$vtype;
1288         $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1289         verify_media_type($opt, $vtype, $drive->{media});
1290         $drive->{file} = $volid;
1291     }
1292
1293     $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1294 }
1295
1296 sub parse_hotplug_features {
1297     my ($data) = @_;
1298
1299     my $res = {};
1300
1301     return $res if $data eq '0';
1302
1303     $data = $confdesc->{hotplug}->{default} if $data eq '1';
1304
1305     foreach my $feature (PVE::Tools::split_list($data)) {
1306         if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1307             $res->{$1} = 1;
1308         } else {
1309             die "invalid hotplug feature '$feature'\n";
1310         }
1311     }
1312     return $res;
1313 }
1314
1315 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1316 sub pve_verify_hotplug_features {
1317     my ($value, $noerr) = @_;
1318
1319     return $value if parse_hotplug_features($value);
1320
1321     return if $noerr;
1322
1323     die "unable to parse hotplug option\n";
1324 }
1325
1326 sub scsi_inquiry {
1327     my($fh, $noerr) = @_;
1328
1329     my $SG_IO = 0x2285;
1330     my $SG_GET_VERSION_NUM = 0x2282;
1331
1332     my $versionbuf = "\x00" x 8;
1333     my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1334     if (!$ret) {
1335         die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1336         return;
1337     }
1338     my $version = unpack("I", $versionbuf);
1339     if ($version < 30000) {
1340         die "scsi generic interface too old\n"  if !$noerr;
1341         return;
1342     }
1343
1344     my $buf = "\x00" x 36;
1345     my $sensebuf = "\x00" x 8;
1346     my $cmd = pack("C x3 C x1", 0x12, 36);
1347
1348     # see /usr/include/scsi/sg.h
1349     my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1350
1351     my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1352                       length($sensebuf), 0, length($buf), $buf,
1353                       $cmd, $sensebuf, 6000);
1354
1355     $ret = ioctl($fh, $SG_IO, $packet);
1356     if (!$ret) {
1357         die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1358         return;
1359     }
1360
1361     my @res = unpack($sg_io_hdr_t, $packet);
1362     if ($res[17] || $res[18]) {
1363         die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1364         return;
1365     }
1366
1367     my $res = {};
1368     (my $byte0, my $byte1, $res->{vendor},
1369      $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1370
1371     $res->{removable} = $byte1 & 128 ? 1 : 0;
1372     $res->{type} = $byte0 & 31;
1373
1374     return $res;
1375 }
1376
1377 sub path_is_scsi {
1378     my ($path) = @_;
1379
1380     my $fh = IO::File->new("+<$path") || return;
1381     my $res = scsi_inquiry($fh, 1);
1382     close($fh);
1383
1384     return $res;
1385 }
1386
1387 sub print_tabletdevice_full {
1388     my ($conf, $arch) = @_;
1389
1390     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
1391
1392     # we use uhci for old VMs because tablet driver was buggy in older qemu
1393     my $usbbus;
1394     if (PVE::QemuServer::Machine::machine_type_is_q35($conf) || $arch eq 'aarch64') {
1395         $usbbus = 'ehci';
1396     } else {
1397         $usbbus = 'uhci';
1398     }
1399
1400     return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1401 }
1402
1403 sub print_keyboarddevice_full {
1404     my ($conf, $arch, $machine) = @_;
1405
1406     return if $arch ne 'aarch64';
1407
1408     return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1409 }
1410
1411 sub print_drivedevice_full {
1412     my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1413
1414     my $device = '';
1415     my $maxdev = 0;
1416
1417     my $drive_id = "$drive->{interface}$drive->{index}";
1418     if ($drive->{interface} eq 'virtio') {
1419         my $pciaddr = print_pci_addr("$drive_id", $bridges, $arch, $machine_type);
1420         $device = "virtio-blk-pci,drive=drive-$drive_id,id=${drive_id}${pciaddr}";
1421         $device .= ",iothread=iothread-$drive_id" if $drive->{iothread};
1422     } elsif ($drive->{interface} eq 'scsi') {
1423
1424         my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1425         my $unit = $drive->{index} % $maxdev;
1426         my $devicetype = 'hd';
1427         my $path = '';
1428         if (drive_is_cdrom($drive)) {
1429             $devicetype = 'cd';
1430         } else {
1431             if ($drive->{file} =~ m|^/|) {
1432                 $path = $drive->{file};
1433                 if (my $info = path_is_scsi($path)) {
1434                     if ($info->{type} == 0 && $drive->{scsiblock}) {
1435                         $devicetype = 'block';
1436                     } elsif ($info->{type} == 1) { # tape
1437                         $devicetype = 'generic';
1438                     }
1439                 }
1440             } else {
1441                  $path = PVE::Storage::path($storecfg, $drive->{file});
1442             }
1443
1444             # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1445             my $version = extract_version($machine_type, kvm_user_version());
1446             if ($path =~ m/^iscsi\:\/\// &&
1447                !min_version($version, 4, 1)) {
1448                 $devicetype = 'generic';
1449             }
1450         }
1451
1452         if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1453             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit";
1454         } else {
1455             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0"
1456                 .",lun=$drive->{index}";
1457         }
1458         $device .= ",drive=drive-$drive_id,id=$drive_id";
1459
1460         if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1461             $device .= ",rotation_rate=1";
1462         }
1463         $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1464
1465     } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1466         my $maxdev = ($drive->{interface} eq 'sata') ? $PVE::QemuServer::Drive::MAX_SATA_DISKS : 2;
1467         my $controller = int($drive->{index} / $maxdev);
1468         my $unit = $drive->{index} % $maxdev;
1469         my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1470
1471         $device = "ide-$devicetype";
1472         if ($drive->{interface} eq 'ide') {
1473             $device .= ",bus=ide.$controller,unit=$unit";
1474         } else {
1475             $device .= ",bus=ahci$controller.$unit";
1476         }
1477         $device .= ",drive=drive-$drive_id,id=$drive_id";
1478
1479         if ($devicetype eq 'hd') {
1480             if (my $model = $drive->{model}) {
1481                 $model = URI::Escape::uri_unescape($model);
1482                 $device .= ",model=$model";
1483             }
1484             if ($drive->{ssd}) {
1485                 $device .= ",rotation_rate=1";
1486             }
1487         }
1488         $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1489     } elsif ($drive->{interface} eq 'usb') {
1490         die "implement me";
1491         #  -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1492     } else {
1493         die "unsupported interface type";
1494     }
1495
1496     $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1497
1498     if (my $serial = $drive->{serial}) {
1499         $serial = URI::Escape::uri_unescape($serial);
1500         $device .= ",serial=$serial";
1501     }
1502
1503
1504     return $device;
1505 }
1506
1507 sub get_initiator_name {
1508     my $initiator;
1509
1510     my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return;
1511     while (defined(my $line = <$fh>)) {
1512         next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1513         $initiator = $1;
1514         last;
1515     }
1516     $fh->close();
1517
1518     return $initiator;
1519 }
1520
1521 sub print_drive_commandline_full {
1522     my ($storecfg, $vmid, $drive) = @_;
1523
1524     my $path;
1525     my $volid = $drive->{file};
1526     my $format;
1527
1528     if (drive_is_cdrom($drive)) {
1529         $path = get_iso_path($storecfg, $vmid, $volid);
1530     } else {
1531         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1532         if ($storeid) {
1533             $path = PVE::Storage::path($storecfg, $volid);
1534             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1535             $format = qemu_img_format($scfg, $volname);
1536         } else {
1537             $path = $volid;
1538             $format = "raw";
1539         }
1540    }
1541
1542     my $opts = '';
1543     my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1544     foreach my $o (@qemu_drive_options) {
1545         $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1546     }
1547
1548     # snapshot only accepts on|off
1549     if (defined($drive->{snapshot})) {
1550         my $v = $drive->{snapshot} ? 'on' : 'off';
1551         $opts .= ",snapshot=$v";
1552     }
1553
1554     foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
1555         my ($dir, $qmpname) = @$type;
1556         if (my $v = $drive->{"mbps$dir"}) {
1557             $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1558         }
1559         if (my $v = $drive->{"mbps${dir}_max"}) {
1560             $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1561         }
1562         if (my $v = $drive->{"bps${dir}_max_length"}) {
1563             $opts .= ",throttling.bps$qmpname-max-length=$v";
1564         }
1565         if (my $v = $drive->{"iops${dir}"}) {
1566             $opts .= ",throttling.iops$qmpname=$v";
1567         }
1568         if (my $v = $drive->{"iops${dir}_max"}) {
1569             $opts .= ",throttling.iops$qmpname-max=$v";
1570         }
1571         if (my $v = $drive->{"iops${dir}_max_length"}) {
1572             $opts .= ",throttling.iops$qmpname-max-length=$v";
1573         }
1574     }
1575
1576     $opts .= ",format=$format" if $format && !$drive->{format};
1577
1578     my $cache_direct = 0;
1579
1580     if (my $cache = $drive->{cache}) {
1581         $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1582     } elsif (!drive_is_cdrom($drive)) {
1583         $opts .= ",cache=none";
1584         $cache_direct = 1;
1585     }
1586
1587     # aio native works only with O_DIRECT
1588     if (!$drive->{aio}) {
1589         if($cache_direct) {
1590             $opts .= ",aio=native";
1591         } else {
1592             $opts .= ",aio=threads";
1593         }
1594     }
1595
1596     if (!drive_is_cdrom($drive)) {
1597         my $detectzeroes;
1598         if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
1599             $detectzeroes = 'off';
1600         } elsif ($drive->{discard}) {
1601             $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
1602         } else {
1603             # This used to be our default with discard not being specified:
1604             $detectzeroes = 'on';
1605         }
1606         $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
1607     }
1608
1609     my $pathinfo = $path ? "file=$path," : '';
1610
1611     return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1612 }
1613
1614 sub print_netdevice_full {
1615     my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
1616
1617     my $device = $net->{model};
1618     if ($net->{model} eq 'virtio') {
1619          $device = 'virtio-net-pci';
1620      };
1621
1622     my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
1623     my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1624     if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1625         # Consider we have N queues, the number of vectors needed is 2 * N + 2, i.e., one per in
1626         # and out of each queue plus one config interrupt and control vector queue
1627         my $vectors = $net->{queues} * 2 + 2;
1628         $tmpstr .= ",vectors=$vectors,mq=on";
1629     }
1630     $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1631
1632     if (my $mtu = $net->{mtu}) {
1633         if ($net->{model} eq 'virtio' && $net->{bridge}) {
1634             my $bridge_mtu = PVE::Network::read_bridge_mtu($net->{bridge});
1635             if ($mtu == 1) {
1636                  $mtu = $bridge_mtu;
1637             } elsif ($mtu < 576) {
1638                 die "netdev $netid: MTU '$mtu' is smaller than the IP minimum MTU '576'\n";
1639             } elsif ($mtu > $bridge_mtu) {
1640                 die "netdev $netid: MTU '$mtu' is bigger than the bridge MTU '$bridge_mtu'\n";
1641             }
1642             $tmpstr .= ",host_mtu=$mtu";
1643         } else {
1644             warn "WARN: netdev $netid: ignoring MTU '$mtu', not using VirtIO or no bridge configured.\n";
1645         }
1646     }
1647
1648     if ($use_old_bios_files) {
1649         my $romfile;
1650         if ($device eq 'virtio-net-pci') {
1651             $romfile = 'pxe-virtio.rom';
1652         } elsif ($device eq 'e1000') {
1653             $romfile = 'pxe-e1000.rom';
1654         } elsif ($device eq 'ne2k') {
1655             $romfile = 'pxe-ne2k_pci.rom';
1656         } elsif ($device eq 'pcnet') {
1657             $romfile = 'pxe-pcnet.rom';
1658         } elsif ($device eq 'rtl8139') {
1659             $romfile = 'pxe-rtl8139.rom';
1660         }
1661         $tmpstr .= ",romfile=$romfile" if $romfile;
1662     }
1663
1664     return $tmpstr;
1665 }
1666
1667 sub print_netdev_full {
1668     my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1669
1670     my $i = '';
1671     if ($netid =~ m/^net(\d+)$/) {
1672         $i = int($1);
1673     }
1674
1675     die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1676
1677     my $ifname = "tap${vmid}i$i";
1678
1679     # kvm uses TUNSETIFF ioctl, and that limits ifname length
1680     die "interface name '$ifname' is too long (max 15 character)\n"
1681         if length($ifname) >= 16;
1682
1683     my $vhostparam = '';
1684     if (is_native($arch)) {
1685         $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
1686     }
1687
1688     my $vmname = $conf->{name} || "vm$vmid";
1689
1690     my $netdev = "";
1691     my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
1692
1693     if ($net->{bridge}) {
1694         $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script"
1695             .",downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1696     } else {
1697         $netdev = "type=user,id=$netid,hostname=$vmname";
1698     }
1699
1700     $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1701
1702     return $netdev;
1703 }
1704
1705 my $vga_map = {
1706     'cirrus' => 'cirrus-vga',
1707     'std' => 'VGA',
1708     'vmware' => 'vmware-svga',
1709     'virtio' => 'virtio-vga',
1710 };
1711
1712 sub print_vga_device {
1713     my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1714
1715     my $type = $vga_map->{$vga->{type}};
1716     if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1717         $type = 'virtio-gpu';
1718     }
1719     my $vgamem_mb = $vga->{memory};
1720
1721     my $max_outputs = '';
1722     if ($qxlnum) {
1723         $type = $id ? 'qxl' : 'qxl-vga';
1724
1725         if (!$conf->{ostype} || $conf->{ostype} =~ m/^(?:l\d\d)|(?:other)$/) {
1726             # set max outputs so linux can have up to 4 qxl displays with one device
1727             if (min_version($machine_version, 4, 1)) {
1728                 $max_outputs = ",max_outputs=4";
1729             }
1730         }
1731     }
1732
1733     die "no devicetype for $vga->{type}\n" if !$type;
1734
1735     my $memory = "";
1736     if ($vgamem_mb) {
1737         if ($vga->{type} eq 'virtio') {
1738             my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
1739             $memory = ",max_hostmem=$bytes";
1740         } elsif ($qxlnum) {
1741             # from https://www.spice-space.org/multiple-monitors.html
1742             $memory = ",vgamem_mb=$vga->{memory}";
1743             my $ram = $vgamem_mb * 4;
1744             my $vram = $vgamem_mb * 2;
1745             $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1746         } else {
1747             $memory = ",vgamem_mb=$vga->{memory}";
1748         }
1749     } elsif ($qxlnum && $id) {
1750         $memory = ",ram_size=67108864,vram_size=33554432";
1751     }
1752
1753     my $edidoff = "";
1754     if ($type eq 'VGA' && windows_version($conf->{ostype})) {
1755         $edidoff=",edid=off" if (!defined($conf->{bios}) || $conf->{bios} ne 'ovmf');
1756     }
1757
1758     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
1759     my $vgaid = "vga" . ($id // '');
1760     my $pciaddr;
1761
1762     if ($q35 && $vgaid eq 'vga') {
1763         # the first display uses pcie.0 bus on q35 machines
1764         $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
1765     } else {
1766         $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
1767     }
1768
1769     return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}${edidoff}";
1770 }
1771
1772 sub parse_number_sets {
1773     my ($set) = @_;
1774     my $res = [];
1775     foreach my $part (split(/;/, $set)) {
1776         if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1777             die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1778             push @$res, [ $1, $2 ];
1779         } else {
1780             die "invalid range: $part\n";
1781         }
1782     }
1783     return $res;
1784 }
1785
1786 sub parse_numa {
1787     my ($data) = @_;
1788
1789     my $res = parse_property_string($numa_fmt, $data);
1790     $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
1791     $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
1792     return $res;
1793 }
1794
1795 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1796 sub parse_net {
1797     my ($data) = @_;
1798
1799     my $res = eval { parse_property_string($net_fmt, $data) };
1800     if ($@) {
1801         warn $@;
1802         return;
1803     }
1804     if (!defined($res->{macaddr})) {
1805         my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
1806         $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
1807     }
1808     return $res;
1809 }
1810
1811 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1812 sub parse_ipconfig {
1813     my ($data) = @_;
1814
1815     my $res = eval { parse_property_string($ipconfig_fmt, $data) };
1816     if ($@) {
1817         warn $@;
1818         return;
1819     }
1820
1821     if ($res->{gw} && !$res->{ip}) {
1822         warn 'gateway specified without specifying an IP address';
1823         return;
1824     }
1825     if ($res->{gw6} && !$res->{ip6}) {
1826         warn 'IPv6 gateway specified without specifying an IPv6 address';
1827         return;
1828     }
1829     if ($res->{gw} && $res->{ip} eq 'dhcp') {
1830         warn 'gateway specified together with DHCP';
1831         return;
1832     }
1833     if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
1834         # gw6 + auto/dhcp
1835         warn "IPv6 gateway specified together with $res->{ip6} address";
1836         return;
1837     }
1838
1839     if (!$res->{ip} && !$res->{ip6}) {
1840         return { ip => 'dhcp', ip6 => 'dhcp' };
1841     }
1842
1843     return $res;
1844 }
1845
1846 sub print_net {
1847     my $net = shift;
1848
1849     return PVE::JSONSchema::print_property_string($net, $net_fmt);
1850 }
1851
1852 sub add_random_macs {
1853     my ($settings) = @_;
1854
1855     foreach my $opt (keys %$settings) {
1856         next if $opt !~ m/^net(\d+)$/;
1857         my $net = parse_net($settings->{$opt});
1858         next if !$net;
1859         $settings->{$opt} = print_net($net);
1860     }
1861 }
1862
1863 sub vm_is_volid_owner {
1864     my ($storecfg, $vmid, $volid) = @_;
1865
1866     if ($volid !~  m|^/|) {
1867         my ($path, $owner);
1868         eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1869         if ($owner && ($owner == $vmid)) {
1870             return 1;
1871         }
1872     }
1873
1874     return;
1875 }
1876
1877 sub vmconfig_register_unused_drive {
1878     my ($storecfg, $vmid, $conf, $drive) = @_;
1879
1880     if (drive_is_cloudinit($drive)) {
1881         eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
1882         warn $@ if $@;
1883     } elsif (!drive_is_cdrom($drive)) {
1884         my $volid = $drive->{file};
1885         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1886             PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
1887         }
1888     }
1889 }
1890
1891 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
1892 my $smbios1_fmt = {
1893     uuid => {
1894         type => 'string',
1895         pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1896         format_description => 'UUID',
1897         description => "Set SMBIOS1 UUID.",
1898         optional => 1,
1899     },
1900     version => {
1901         type => 'string',
1902         pattern => '[A-Za-z0-9+\/]+={0,2}',
1903         format_description => 'Base64 encoded string',
1904         description => "Set SMBIOS1 version.",
1905         optional => 1,
1906     },
1907     serial => {
1908         type => 'string',
1909         pattern => '[A-Za-z0-9+\/]+={0,2}',
1910         format_description => 'Base64 encoded string',
1911         description => "Set SMBIOS1 serial number.",
1912         optional => 1,
1913     },
1914     manufacturer => {
1915         type => 'string',
1916         pattern => '[A-Za-z0-9+\/]+={0,2}',
1917         format_description => 'Base64 encoded string',
1918         description => "Set SMBIOS1 manufacturer.",
1919         optional => 1,
1920     },
1921     product => {
1922         type => 'string',
1923         pattern => '[A-Za-z0-9+\/]+={0,2}',
1924         format_description => 'Base64 encoded string',
1925         description => "Set SMBIOS1 product ID.",
1926         optional => 1,
1927     },
1928     sku => {
1929         type => 'string',
1930         pattern => '[A-Za-z0-9+\/]+={0,2}',
1931         format_description => 'Base64 encoded string',
1932         description => "Set SMBIOS1 SKU string.",
1933         optional => 1,
1934     },
1935     family => {
1936         type => 'string',
1937         pattern => '[A-Za-z0-9+\/]+={0,2}',
1938         format_description => 'Base64 encoded string',
1939         description => "Set SMBIOS1 family string.",
1940         optional => 1,
1941     },
1942     base64 => {
1943         type => 'boolean',
1944         description => 'Flag to indicate that the SMBIOS values are base64 encoded',
1945         optional => 1,
1946     },
1947 };
1948
1949 sub parse_smbios1 {
1950     my ($data) = @_;
1951
1952     my $res = eval { parse_property_string($smbios1_fmt, $data) };
1953     warn $@ if $@;
1954     return $res;
1955 }
1956
1957 sub print_smbios1 {
1958     my ($smbios1) = @_;
1959     return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
1960 }
1961
1962 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
1963
1964 sub parse_watchdog {
1965     my ($value) = @_;
1966
1967     return if !$value;
1968
1969     my $res = eval { parse_property_string($watchdog_fmt, $value) };
1970     warn $@ if $@;
1971     return $res;
1972 }
1973
1974 sub parse_guest_agent {
1975     my ($value) = @_;
1976
1977     return {} if !defined($value->{agent});
1978
1979     my $res = eval { parse_property_string($agent_fmt, $value->{agent}) };
1980     warn $@ if $@;
1981
1982     # if the agent is disabled ignore the other potentially set properties
1983     return {} if !$res->{enabled};
1984     return $res;
1985 }
1986
1987 sub parse_vga {
1988     my ($value) = @_;
1989
1990     return {} if !$value;
1991     my $res = eval { parse_property_string($vga_fmt, $value) };
1992     warn $@ if $@;
1993     return $res;
1994 }
1995
1996 sub parse_rng {
1997     my ($value) = @_;
1998
1999     return if !$value;
2000
2001     my $res = eval { parse_property_string($rng_fmt, $value) };
2002     warn $@ if $@;
2003     return $res;
2004 }
2005
2006 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2007 sub verify_usb_device {
2008     my ($value, $noerr) = @_;
2009
2010     return $value if parse_usb_device($value);
2011
2012     return if $noerr;
2013
2014     die "unable to parse usb device\n";
2015 }
2016
2017 # add JSON properties for create and set function
2018 sub json_config_properties {
2019     my $prop = shift;
2020
2021     foreach my $opt (keys %$confdesc) {
2022         next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' ||
2023             $opt eq 'runningmachine' || $opt eq 'runningcpu';
2024         $prop->{$opt} = $confdesc->{$opt};
2025     }
2026
2027     return $prop;
2028 }
2029
2030 # return copy of $confdesc_cloudinit to generate documentation
2031 sub cloudinit_config_properties {
2032
2033     return dclone($confdesc_cloudinit);
2034 }
2035
2036 sub check_type {
2037     my ($key, $value) = @_;
2038
2039     die "unknown setting '$key'\n" if !$confdesc->{$key};
2040
2041     my $type = $confdesc->{$key}->{type};
2042
2043     if (!defined($value)) {
2044         die "got undefined value\n";
2045     }
2046
2047     if ($value =~ m/[\n\r]/) {
2048         die "property contains a line feed\n";
2049     }
2050
2051     if ($type eq 'boolean') {
2052         return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2053         return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2054         die "type check ('boolean') failed - got '$value'\n";
2055     } elsif ($type eq 'integer') {
2056         return int($1) if $value =~ m/^(\d+)$/;
2057         die "type check ('integer') failed - got '$value'\n";
2058     } elsif ($type eq 'number') {
2059         return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2060         die "type check ('number') failed - got '$value'\n";
2061     } elsif ($type eq 'string') {
2062         if (my $fmt = $confdesc->{$key}->{format}) {
2063             PVE::JSONSchema::check_format($fmt, $value);
2064             return $value;
2065         }
2066         $value =~ s/^\"(.*)\"$/$1/;
2067         return $value;
2068     } else {
2069         die "internal error"
2070     }
2071 }
2072
2073 sub destroy_vm {
2074     my ($storecfg, $vmid, $skiplock, $replacement_conf) = @_;
2075
2076     my $conf = PVE::QemuConfig->load_config($vmid);
2077
2078     PVE::QemuConfig->check_lock($conf) if !$skiplock;
2079
2080     if ($conf->{template}) {
2081         # check if any base image is still used by a linked clone
2082         PVE::QemuConfig->foreach_volume($conf, sub {
2083                 my ($ds, $drive) = @_;
2084                 return if drive_is_cdrom($drive);
2085
2086                 my $volid = $drive->{file};
2087                 return if !$volid || $volid =~ m|^/|;
2088
2089                 die "base volume '$volid' is still in use by linked cloned\n"
2090                     if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2091
2092         });
2093     }
2094
2095     # only remove disks owned by this VM
2096     PVE::QemuConfig->foreach_volume($conf, sub {
2097         my ($ds, $drive) = @_;
2098         return if drive_is_cdrom($drive, 1);
2099
2100         my $volid = $drive->{file};
2101         return if !$volid || $volid =~ m|^/|;
2102
2103         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2104         return if !$path || !$owner || ($owner != $vmid);
2105
2106         eval { PVE::Storage::vdisk_free($storecfg, $volid) };
2107         warn "Could not remove disk '$volid', check manually: $@" if $@;
2108     });
2109
2110     # also remove unused disk
2111     my $vmdisks = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2112     PVE::Storage::foreach_volid($vmdisks, sub {
2113         my ($volid, $sid, $volname, $d) = @_;
2114         eval { PVE::Storage::vdisk_free($storecfg, $volid) };
2115         warn $@ if $@;
2116     });
2117
2118     if (defined $replacement_conf) {
2119         PVE::QemuConfig->write_config($vmid, $replacement_conf);
2120     } else {
2121         PVE::QemuConfig->destroy_config($vmid);
2122     }
2123 }
2124
2125 sub parse_vm_config {
2126     my ($filename, $raw) = @_;
2127
2128     return if !defined($raw);
2129
2130     my $res = {
2131         digest => Digest::SHA::sha1_hex($raw),
2132         snapshots => {},
2133         pending => {},
2134     };
2135
2136     $filename =~ m|/qemu-server/(\d+)\.conf$|
2137         || die "got strange filename '$filename'";
2138
2139     my $vmid = $1;
2140
2141     my $conf = $res;
2142     my $descr;
2143     my $section = '';
2144
2145     my @lines = split(/\n/, $raw);
2146     foreach my $line (@lines) {
2147         next if $line =~ m/^\s*$/;
2148
2149         if ($line =~ m/^\[PENDING\]\s*$/i) {
2150             $section = 'pending';
2151             if (defined($descr)) {
2152                 $descr =~ s/\s+$//;
2153                 $conf->{description} = $descr;
2154             }
2155             $descr = undef;
2156             $conf = $res->{$section} = {};
2157             next;
2158
2159         } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2160             $section = $1;
2161             if (defined($descr)) {
2162                 $descr =~ s/\s+$//;
2163                 $conf->{description} = $descr;
2164             }
2165             $descr = undef;
2166             $conf = $res->{snapshots}->{$section} = {};
2167             next;
2168         }
2169
2170         if ($line =~ m/^\#(.*)\s*$/) {
2171             $descr = '' if !defined($descr);
2172             $descr .= PVE::Tools::decode_text($1) . "\n";
2173             next;
2174         }
2175
2176         if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2177             $descr = '' if !defined($descr);
2178             $descr .= PVE::Tools::decode_text($2);
2179         } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2180             $conf->{snapstate} = $1;
2181         } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2182             my $key = $1;
2183             my $value = $2;
2184             $conf->{$key} = $value;
2185         } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2186             my $value = $1;
2187             if ($section eq 'pending') {
2188                 $conf->{delete} = $value; # we parse this later
2189             } else {
2190                 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2191             }
2192         } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2193             my $key = $1;
2194             my $value = $2;
2195             eval { $value = check_type($key, $value); };
2196             if ($@) {
2197                 warn "vm $vmid - unable to parse value of '$key' - $@";
2198             } else {
2199                 $key = 'ide2' if $key eq 'cdrom';
2200                 my $fmt = $confdesc->{$key}->{format};
2201                 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2202                     my $v = parse_drive($key, $value);
2203                     if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2204                         $v->{file} = $volid;
2205                         $value = print_drive($v);
2206                     } else {
2207                         warn "vm $vmid - unable to parse value of '$key'\n";
2208                         next;
2209                     }
2210                 }
2211
2212                 $conf->{$key} = $value;
2213             }
2214         }
2215     }
2216
2217     if (defined($descr)) {
2218         $descr =~ s/\s+$//;
2219         $conf->{description} = $descr;
2220     }
2221     delete $res->{snapstate}; # just to be sure
2222
2223     return $res;
2224 }
2225
2226 sub write_vm_config {
2227     my ($filename, $conf) = @_;
2228
2229     delete $conf->{snapstate}; # just to be sure
2230
2231     if ($conf->{cdrom}) {
2232         die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2233         $conf->{ide2} = $conf->{cdrom};
2234         delete $conf->{cdrom};
2235     }
2236
2237     # we do not use 'smp' any longer
2238     if ($conf->{sockets}) {
2239         delete $conf->{smp};
2240     } elsif ($conf->{smp}) {
2241         $conf->{sockets} = $conf->{smp};
2242         delete $conf->{cores};
2243         delete $conf->{smp};
2244     }
2245
2246     my $used_volids = {};
2247
2248     my $cleanup_config = sub {
2249         my ($cref, $pending, $snapname) = @_;
2250
2251         foreach my $key (keys %$cref) {
2252             next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2253                 $key eq 'snapstate' || $key eq 'pending';
2254             my $value = $cref->{$key};
2255             if ($key eq 'delete') {
2256                 die "propertry 'delete' is only allowed in [PENDING]\n"
2257                     if !$pending;
2258                 # fixme: check syntax?
2259                 next;
2260             }
2261             eval { $value = check_type($key, $value); };
2262             die "unable to parse value of '$key' - $@" if $@;
2263
2264             $cref->{$key} = $value;
2265
2266             if (!$snapname && is_valid_drivename($key)) {
2267                 my $drive = parse_drive($key, $value);
2268                 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2269             }
2270         }
2271     };
2272
2273     &$cleanup_config($conf);
2274
2275     &$cleanup_config($conf->{pending}, 1);
2276
2277     foreach my $snapname (keys %{$conf->{snapshots}}) {
2278         die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2279         &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2280     }
2281
2282     # remove 'unusedX' settings if we re-add a volume
2283     foreach my $key (keys %$conf) {
2284         my $value = $conf->{$key};
2285         if ($key =~ m/^unused/ && $used_volids->{$value}) {
2286             delete $conf->{$key};
2287         }
2288     }
2289
2290     my $generate_raw_config = sub {
2291         my ($conf, $pending) = @_;
2292
2293         my $raw = '';
2294
2295         # add description as comment to top of file
2296         if (defined(my $descr = $conf->{description})) {
2297             if ($descr) {
2298                 foreach my $cl (split(/\n/, $descr)) {
2299                     $raw .= '#' .  PVE::Tools::encode_text($cl) . "\n";
2300                 }
2301             } else {
2302                 $raw .= "#\n" if $pending;
2303             }
2304         }
2305
2306         foreach my $key (sort keys %$conf) {
2307             next if $key =~ /^(digest|description|pending|snapshots)$/;
2308             $raw .= "$key: $conf->{$key}\n";
2309         }
2310         return $raw;
2311     };
2312
2313     my $raw = &$generate_raw_config($conf);
2314
2315     if (scalar(keys %{$conf->{pending}})){
2316         $raw .= "\n[PENDING]\n";
2317         $raw .= &$generate_raw_config($conf->{pending}, 1);
2318     }
2319
2320     foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2321         $raw .= "\n[$snapname]\n";
2322         $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2323     }
2324
2325     return $raw;
2326 }
2327
2328 sub load_defaults {
2329
2330     my $res = {};
2331
2332     # we use static defaults from our JSON schema configuration
2333     foreach my $key (keys %$confdesc) {
2334         if (defined(my $default = $confdesc->{$key}->{default})) {
2335             $res->{$key} = $default;
2336         }
2337     }
2338
2339     return $res;
2340 }
2341
2342 sub config_list {
2343     my $vmlist = PVE::Cluster::get_vmlist();
2344     my $res = {};
2345     return $res if !$vmlist || !$vmlist->{ids};
2346     my $ids = $vmlist->{ids};
2347     my $nodename = nodename();
2348
2349     foreach my $vmid (keys %$ids) {
2350         my $d = $ids->{$vmid};
2351         next if !$d->{node} || $d->{node} ne $nodename;
2352         next if !$d->{type} || $d->{type} ne 'qemu';
2353         $res->{$vmid}->{exists} = 1;
2354     }
2355     return $res;
2356 }
2357
2358 # test if VM uses local resources (to prevent migration)
2359 sub check_local_resources {
2360     my ($conf, $noerr) = @_;
2361
2362     my @loc_res = ();
2363
2364     push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2365     push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2366
2367     push @loc_res, "ivshmem" if $conf->{ivshmem};
2368
2369     foreach my $k (keys %$conf) {
2370         next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2371         # sockets are safe: they will recreated be on the target side post-migrate
2372         next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2373         push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2374     }
2375
2376     die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2377
2378     return \@loc_res;
2379 }
2380
2381 # check if used storages are available on all nodes (use by migrate)
2382 sub check_storage_availability {
2383     my ($storecfg, $conf, $node) = @_;
2384
2385     PVE::QemuConfig->foreach_volume($conf, sub {
2386         my ($ds, $drive) = @_;
2387
2388         my $volid = $drive->{file};
2389         return if !$volid;
2390
2391         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2392         return if !$sid;
2393
2394         # check if storage is available on both nodes
2395         my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2396         PVE::Storage::storage_check_node($storecfg, $sid, $node);
2397    });
2398 }
2399
2400 # list nodes where all VM images are available (used by has_feature API)
2401 sub shared_nodes {
2402     my ($conf, $storecfg) = @_;
2403
2404     my $nodelist = PVE::Cluster::get_nodelist();
2405     my $nodehash = { map { $_ => 1 } @$nodelist };
2406     my $nodename = nodename();
2407
2408     PVE::QemuConfig->foreach_volume($conf, sub {
2409         my ($ds, $drive) = @_;
2410
2411         my $volid = $drive->{file};
2412         return if !$volid;
2413
2414         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2415         if ($storeid) {
2416             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2417             if ($scfg->{disable}) {
2418                 $nodehash = {};
2419             } elsif (my $avail = $scfg->{nodes}) {
2420                 foreach my $node (keys %$nodehash) {
2421                     delete $nodehash->{$node} if !$avail->{$node};
2422                 }
2423             } elsif (!$scfg->{shared}) {
2424                 foreach my $node (keys %$nodehash) {
2425                     delete $nodehash->{$node} if $node ne $nodename
2426                 }
2427             }
2428         }
2429     });
2430
2431     return $nodehash
2432 }
2433
2434 sub check_local_storage_availability {
2435     my ($conf, $storecfg) = @_;
2436
2437     my $nodelist = PVE::Cluster::get_nodelist();
2438     my $nodehash = { map { $_ => {} } @$nodelist };
2439
2440     PVE::QemuConfig->foreach_volume($conf, sub {
2441         my ($ds, $drive) = @_;
2442
2443         my $volid = $drive->{file};
2444         return if !$volid;
2445
2446         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2447         if ($storeid) {
2448             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2449
2450             if ($scfg->{disable}) {
2451                 foreach my $node (keys %$nodehash) {
2452                     $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2453                 }
2454             } elsif (my $avail = $scfg->{nodes}) {
2455                 foreach my $node (keys %$nodehash) {
2456                     if (!$avail->{$node}) {
2457                         $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2458                     }
2459                 }
2460             }
2461         }
2462     });
2463
2464     foreach my $node (values %$nodehash) {
2465         if (my $unavail = $node->{unavailable_storages}) {
2466             $node->{unavailable_storages} = [ sort keys %$unavail ];
2467         }
2468     }
2469
2470     return $nodehash
2471 }
2472
2473 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2474 sub check_running {
2475     my ($vmid, $nocheck, $node) = @_;
2476
2477     PVE::QemuConfig::assert_config_exists_on_node($vmid, $node) if !$nocheck;
2478     return PVE::QemuServer::Helpers::vm_running_locally($vmid);
2479 }
2480
2481 sub vzlist {
2482
2483     my $vzlist = config_list();
2484
2485     my $fd = IO::Dir->new($PVE::QemuServer::Helpers::var_run_tmpdir) || return $vzlist;
2486
2487     while (defined(my $de = $fd->read)) {
2488         next if $de !~ m/^(\d+)\.pid$/;
2489         my $vmid = $1;
2490         next if !defined($vzlist->{$vmid});
2491         if (my $pid = check_running($vmid)) {
2492             $vzlist->{$vmid}->{pid} = $pid;
2493         }
2494     }
2495
2496     return $vzlist;
2497 }
2498
2499 our $vmstatus_return_properties = {
2500     vmid => get_standard_option('pve-vmid'),
2501     status => {
2502         description => "Qemu process status.",
2503         type => 'string',
2504         enum => ['stopped', 'running'],
2505     },
2506     maxmem => {
2507         description => "Maximum memory in bytes.",
2508         type => 'integer',
2509         optional => 1,
2510         renderer => 'bytes',
2511     },
2512     maxdisk => {
2513         description => "Root disk size in bytes.",
2514         type => 'integer',
2515         optional => 1,
2516         renderer => 'bytes',
2517     },
2518     name => {
2519         description => "VM name.",
2520         type => 'string',
2521         optional => 1,
2522     },
2523     qmpstatus => {
2524         description => "Qemu QMP agent status.",
2525         type => 'string',
2526         optional => 1,
2527     },
2528     pid => {
2529         description => "PID of running qemu process.",
2530         type => 'integer',
2531         optional => 1,
2532     },
2533     uptime => {
2534         description => "Uptime.",
2535         type => 'integer',
2536         optional => 1,
2537         renderer => 'duration',
2538     },
2539     cpus => {
2540         description => "Maximum usable CPUs.",
2541         type => 'number',
2542         optional => 1,
2543     },
2544     lock => {
2545         description => "The current config lock, if any.",
2546         type => 'string',
2547         optional => 1,
2548     },
2549     tags => {
2550         description  => "The current configured tags, if any",
2551         type => 'string',
2552         optional => 1,
2553     },
2554 };
2555
2556 my $last_proc_pid_stat;
2557
2558 # get VM status information
2559 # This must be fast and should not block ($full == false)
2560 # We only query KVM using QMP if $full == true (this can be slow)
2561 sub vmstatus {
2562     my ($opt_vmid, $full) = @_;
2563
2564     my $res = {};
2565
2566     my $storecfg = PVE::Storage::config();
2567
2568     my $list = vzlist();
2569     my $defaults = load_defaults();
2570
2571     my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2572
2573     my $cpucount = $cpuinfo->{cpus} || 1;
2574
2575     foreach my $vmid (keys %$list) {
2576         next if $opt_vmid && ($vmid ne $opt_vmid);
2577
2578         my $conf = PVE::QemuConfig->load_config($vmid);
2579
2580         my $d = { vmid => $vmid };
2581         $d->{pid} = $list->{$vmid}->{pid};
2582
2583         # fixme: better status?
2584         $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2585
2586         my $size = PVE::QemuServer::Drive::bootdisk_size($storecfg, $conf);
2587         if (defined($size)) {
2588             $d->{disk} = 0; # no info available
2589             $d->{maxdisk} = $size;
2590         } else {
2591             $d->{disk} = 0;
2592             $d->{maxdisk} = 0;
2593         }
2594
2595         $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
2596             * ($conf->{cores} || $defaults->{cores});
2597         $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2598         $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2599
2600         $d->{name} = $conf->{name} || "VM $vmid";
2601         $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
2602             : $defaults->{memory}*(1024*1024);
2603
2604         if ($conf->{balloon}) {
2605             $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2606             $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
2607                 : $defaults->{shares};
2608         }
2609
2610         $d->{uptime} = 0;
2611         $d->{cpu} = 0;
2612         $d->{mem} = 0;
2613
2614         $d->{netout} = 0;
2615         $d->{netin} = 0;
2616
2617         $d->{diskread} = 0;
2618         $d->{diskwrite} = 0;
2619
2620         $d->{template} = PVE::QemuConfig->is_template($conf);
2621
2622         $d->{serial} = 1 if conf_has_serial($conf);
2623         $d->{lock} = $conf->{lock} if $conf->{lock};
2624         $d->{tags} = $conf->{tags} if defined($conf->{tags});
2625
2626         $res->{$vmid} = $d;
2627     }
2628
2629     my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2630     foreach my $dev (keys %$netdev) {
2631         next if $dev !~ m/^tap([1-9]\d*)i/;
2632         my $vmid = $1;
2633         my $d = $res->{$vmid};
2634         next if !$d;
2635
2636         $d->{netout} += $netdev->{$dev}->{receive};
2637         $d->{netin} += $netdev->{$dev}->{transmit};
2638
2639         if ($full) {
2640             $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
2641             $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
2642         }
2643
2644     }
2645
2646     my $ctime = gettimeofday;
2647
2648     foreach my $vmid (keys %$list) {
2649
2650         my $d = $res->{$vmid};
2651         my $pid = $d->{pid};
2652         next if !$pid;
2653
2654         my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2655         next if !$pstat; # not running
2656
2657         my $used = $pstat->{utime} + $pstat->{stime};
2658
2659         $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2660
2661         if ($pstat->{vsize}) {
2662             $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2663         }
2664
2665         my $old = $last_proc_pid_stat->{$pid};
2666         if (!$old) {
2667             $last_proc_pid_stat->{$pid} = {
2668                 time => $ctime,
2669                 used => $used,
2670                 cpu => 0,
2671             };
2672             next;
2673         }
2674
2675         my $dtime = ($ctime -  $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2676
2677         if ($dtime > 1000) {
2678             my $dutime = $used -  $old->{used};
2679
2680             $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2681             $last_proc_pid_stat->{$pid} = {
2682                 time => $ctime,
2683                 used => $used,
2684                 cpu => $d->{cpu},
2685             };
2686         } else {
2687             $d->{cpu} = $old->{cpu};
2688         }
2689     }
2690
2691     return $res if !$full;
2692
2693     my $qmpclient = PVE::QMPClient->new();
2694
2695     my $ballooncb = sub {
2696         my ($vmid, $resp) = @_;
2697
2698         my $info = $resp->{'return'};
2699         return if !$info->{max_mem};
2700
2701         my $d = $res->{$vmid};
2702
2703         # use memory assigned to VM
2704         $d->{maxmem} = $info->{max_mem};
2705         $d->{balloon} = $info->{actual};
2706
2707         if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2708             $d->{mem} = $info->{total_mem} - $info->{free_mem};
2709             $d->{freemem} = $info->{free_mem};
2710         }
2711
2712         $d->{ballooninfo} = $info;
2713     };
2714
2715     my $blockstatscb = sub {
2716         my ($vmid, $resp) = @_;
2717         my $data = $resp->{'return'} || [];
2718         my $totalrdbytes = 0;
2719         my $totalwrbytes = 0;
2720
2721         for my $blockstat (@$data) {
2722             $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2723             $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2724
2725             $blockstat->{device} =~ s/drive-//;
2726             $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
2727         }
2728         $res->{$vmid}->{diskread} = $totalrdbytes;
2729         $res->{$vmid}->{diskwrite} = $totalwrbytes;
2730     };
2731
2732     my $statuscb = sub {
2733         my ($vmid, $resp) = @_;
2734
2735         $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2736         # this fails if ballon driver is not loaded, so this must be
2737         # the last commnand (following command are aborted if this fails).
2738         $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2739
2740         my $status = 'unknown';
2741         if (!defined($status = $resp->{'return'}->{status})) {
2742             warn "unable to get VM status\n";
2743             return;
2744         }
2745
2746         $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2747     };
2748
2749     foreach my $vmid (keys %$list) {
2750         next if $opt_vmid && ($vmid ne $opt_vmid);
2751         next if !$res->{$vmid}->{pid}; # not running
2752         $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2753     }
2754
2755     $qmpclient->queue_execute(undef, 2);
2756
2757     foreach my $vmid (keys %$list) {
2758         next if $opt_vmid && ($vmid ne $opt_vmid);
2759         $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2760     }
2761
2762     return $res;
2763 }
2764
2765 sub conf_has_serial {
2766     my ($conf) = @_;
2767
2768     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
2769         if ($conf->{"serial$i"}) {
2770             return 1;
2771         }
2772     }
2773
2774     return 0;
2775 }
2776
2777 sub conf_has_audio {
2778     my ($conf, $id) = @_;
2779
2780     $id //= 0;
2781     my $audio = $conf->{"audio$id"};
2782     return if !defined($audio);
2783
2784     my $audioproperties = parse_property_string($audio_fmt, $audio);
2785     my $audiodriver = $audioproperties->{driver} // 'spice';
2786
2787     return {
2788         dev => $audioproperties->{device},
2789         dev_id => "audiodev$id",
2790         backend => $audiodriver,
2791         backend_id => "$audiodriver-backend${id}",
2792     };
2793 }
2794
2795 sub audio_devs {
2796     my ($audio, $audiopciaddr, $machine_version) = @_;
2797
2798     my $devs = [];
2799
2800     my $id = $audio->{dev_id};
2801     my $audiodev = "";
2802     if (min_version($machine_version, 4, 2)) {
2803         $audiodev = ",audiodev=$audio->{backend_id}";
2804     }
2805
2806     if ($audio->{dev} eq 'AC97') {
2807         push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
2808     } elsif ($audio->{dev} =~ /intel\-hda$/) {
2809         push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
2810         push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
2811         push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
2812     } else {
2813         die "unkown audio device '$audio->{dev}', implement me!";
2814     }
2815
2816     push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
2817
2818     return $devs;
2819 }
2820
2821 sub vga_conf_has_spice {
2822     my ($vga) = @_;
2823
2824     my $vgaconf = parse_vga($vga);
2825     my $vgatype = $vgaconf->{type};
2826     return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
2827
2828     return $1 || 1;
2829 }
2830
2831 sub is_native($) {
2832     my ($arch) = @_;
2833     return get_host_arch() eq $arch;
2834 }
2835
2836 sub get_vm_arch {
2837     my ($conf) = @_;
2838     return $conf->{arch} // get_host_arch();
2839 }
2840
2841 my $default_machines = {
2842     x86_64 => 'pc',
2843     aarch64 => 'virt',
2844 };
2845
2846 sub get_vm_machine {
2847     my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
2848
2849     my $machine = $forcemachine || $conf->{machine};
2850
2851     if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
2852         $arch //= 'x86_64';
2853         $machine ||= $default_machines->{$arch};
2854         if ($add_pve_version) {
2855             $kvmversion //= kvm_user_version();
2856             my $pvever = PVE::QemuServer::Machine::get_pve_version($kvmversion);
2857             $machine .= "+pve$pvever";
2858         }
2859     }
2860
2861     if ($add_pve_version && $machine !~ m/\+pve\d+$/) {
2862         # for version-pinned machines that do not include a pve-version (e.g.
2863         # pc-q35-4.1), we assume 0 to keep them stable in case we bump
2864         $machine .= '+pve0';
2865     }
2866
2867     return $machine;
2868 }
2869
2870 sub get_ovmf_files($) {
2871     my ($arch) = @_;
2872
2873     my $ovmf = $OVMF->{$arch}
2874         or die "no OVMF images known for architecture '$arch'\n";
2875
2876     return @$ovmf;
2877 }
2878
2879 my $Arch2Qemu = {
2880     aarch64 => '/usr/bin/qemu-system-aarch64',
2881     x86_64 => '/usr/bin/qemu-system-x86_64',
2882 };
2883 sub get_command_for_arch($) {
2884     my ($arch) = @_;
2885     return '/usr/bin/kvm' if is_native($arch);
2886
2887     my $cmd = $Arch2Qemu->{$arch}
2888         or die "don't know how to emulate architecture '$arch'\n";
2889     return $cmd;
2890 }
2891
2892 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
2893 # to use in a QEMU command line (-cpu element), first array_intersect the result
2894 # of query_supported_ with query_understood_. This is necessary because:
2895 #
2896 # a) query_understood_ returns flags the host cannot use and
2897 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
2898 #    flags, but CPU settings - with most of them being flags. Those settings
2899 #    (and some flags, curiously) cannot be specified as a "-cpu" argument.
2900 #
2901 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
2902 # expensive. If you need the value returned from this, you can get it much
2903 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
2904 # $accel being 'kvm' or 'tcg'.
2905 #
2906 # pvestatd calls this function on startup and whenever the QEMU/KVM version
2907 # changes, automatically populating pmxcfs.
2908 #
2909 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
2910 # since kvm and tcg machines support different flags
2911 #
2912 sub query_supported_cpu_flags {
2913     my ($arch) = @_;
2914
2915     $arch //= get_host_arch();
2916     my $default_machine = $default_machines->{$arch};
2917
2918     my $flags = {};
2919
2920     # FIXME: Once this is merged, the code below should work for ARM as well:
2921     # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
2922     die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
2923         $arch eq "aarch64";
2924
2925     my $kvm_supported = defined(kvm_version());
2926     my $qemu_cmd = get_command_for_arch($arch);
2927     my $fakevmid = -1;
2928     my $pidfile = PVE::QemuServer::Helpers::pidfile_name($fakevmid);
2929
2930     # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
2931     my $query_supported_run_qemu = sub {
2932         my ($kvm) = @_;
2933
2934         my $flags = {};
2935         my $cmd = [
2936             $qemu_cmd,
2937             '-machine', $default_machine,
2938             '-display', 'none',
2939             '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server,nowait",
2940             '-mon', 'chardev=qmp,mode=control',
2941             '-pidfile', $pidfile,
2942             '-S', '-daemonize'
2943         ];
2944
2945         if (!$kvm) {
2946             push @$cmd, '-accel', 'tcg';
2947         }
2948
2949         my $rc = run_command($cmd, noerr => 1, quiet => 0);
2950         die "QEMU flag querying VM exited with code " . $rc if $rc;
2951
2952         eval {
2953             my $cmd_result = mon_cmd(
2954                 $fakevmid,
2955                 'query-cpu-model-expansion',
2956                 type => 'full',
2957                 model => { name => 'host' }
2958             );
2959
2960             my $props = $cmd_result->{model}->{props};
2961             foreach my $prop (keys %$props) {
2962                 next if $props->{$prop} ne '1';
2963                 # QEMU returns some flags multiple times, with '_', '.' or '-'
2964                 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
2965                 # We only keep those with underscores, to match /proc/cpuinfo
2966                 $prop =~ s/\.|-/_/g;
2967                 $flags->{$prop} = 1;
2968             }
2969         };
2970         my $err = $@;
2971
2972         # force stop with 10 sec timeout and 'nocheck'
2973         # always stop, even if QMP failed
2974         vm_stop(undef, $fakevmid, 1, 1, 10, 0, 1);
2975
2976         die $err if $err;
2977
2978         return [ sort keys %$flags ];
2979     };
2980
2981     # We need to query QEMU twice, since KVM and TCG have different supported flags
2982     PVE::QemuConfig->lock_config($fakevmid, sub {
2983         $flags->{tcg} = eval { $query_supported_run_qemu->(0) };
2984         warn "warning: failed querying supported tcg flags: $@\n" if $@;
2985
2986         if ($kvm_supported) {
2987             $flags->{kvm} = eval { $query_supported_run_qemu->(1) };
2988             warn "warning: failed querying supported kvm flags: $@\n" if $@;
2989         }
2990     });
2991
2992     return $flags;
2993 }
2994
2995 # Understood CPU flags are written to a file at 'pve-qemu' compile time
2996 my $understood_cpu_flag_dir = "/usr/share/kvm";
2997 sub query_understood_cpu_flags {
2998     my $arch = get_host_arch();
2999     my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
3000
3001     die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
3002         if ! -e $filepath;
3003
3004     my $raw = file_get_contents($filepath);
3005     $raw =~ s/^\s+|\s+$//g;
3006     my @flags = split(/\s+/, $raw);
3007
3008     return \@flags;
3009 }
3010
3011 sub config_to_command {
3012     my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu) = @_;
3013
3014     my $cmd = [];
3015     my $globalFlags = [];
3016     my $machineFlags = [];
3017     my $rtcFlags = [];
3018     my $devices = [];
3019     my $pciaddr = '';
3020     my $bridges = {};
3021     my $ostype = $conf->{ostype};
3022     my $winversion = windows_version($ostype);
3023     my $kvm = $conf->{kvm};
3024     my $nodename = nodename();
3025
3026     my $arch = get_vm_arch($conf);
3027     my $kvm_binary = get_command_for_arch($arch);
3028     my $kvmver = kvm_user_version($kvm_binary);
3029
3030     if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
3031         $kvmver //= "undefined";
3032         die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
3033     }
3034
3035     my $add_pve_version = min_version($kvmver, 4, 1);
3036
3037     my $machine_type = get_vm_machine($conf, $forcemachine, $arch, $add_pve_version);
3038     my $machine_version = extract_version($machine_type, $kvmver);
3039     $kvm //= 1 if is_native($arch);
3040
3041     $machine_version =~ m/(\d+)\.(\d+)/;
3042     my ($machine_major, $machine_minor) = ($1, $2);
3043
3044     if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3045         warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3046     } elsif (!min_version($kvmver, $machine_major, $machine_minor)) {
3047         die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type',"
3048             ." please upgrade node '$nodename'\n"
3049     } elsif (!PVE::QemuServer::Machine::can_run_pve_machine_version($machine_version, $kvmver)) {
3050         my $max_pve_version = PVE::QemuServer::Machine::get_pve_version($machine_version);
3051         die "Installed qemu-server (max feature level for $machine_major.$machine_minor is"
3052             ." pve$max_pve_version) is too old to run machine type '$machine_type', please upgrade"
3053             ." node '$nodename'\n";
3054     }
3055
3056     # if a specific +pve version is required for a feature, use $version_guard
3057     # instead of min_version to allow machines to be run with the minimum
3058     # required version
3059     my $required_pve_version = 0;
3060     my $version_guard = sub {
3061         my ($major, $minor, $pve) = @_;
3062         return 0 if !min_version($machine_version, $major, $minor, $pve);
3063         my $max_pve = PVE::QemuServer::Machine::get_pve_version("$major.$minor");
3064         return 1 if min_version($machine_version, $major, $minor, $max_pve+1);
3065         $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3066         return 1;
3067     };
3068
3069     if ($kvm && !defined kvm_version()) {
3070         die "KVM virtualisation configured, but not available. Either disable in VM configuration"
3071             ." or enable in BIOS.\n";
3072     }
3073
3074     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3075     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3076     my $use_old_bios_files = undef;
3077     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3078
3079     my $cpuunits = defined($conf->{cpuunits}) ?
3080             $conf->{cpuunits} : $defaults->{cpuunits};
3081
3082     push @$cmd, $kvm_binary;
3083
3084     push @$cmd, '-id', $vmid;
3085
3086     my $vmname = $conf->{name} || "vm$vmid";
3087
3088     push @$cmd, '-name', $vmname;
3089
3090     push @$cmd, '-no-shutdown';
3091
3092     my $use_virtio = 0;
3093
3094     my $qmpsocket = PVE::QemuServer::Helpers::qmp_socket($vmid);
3095     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3096     push @$cmd, '-mon', "chardev=qmp,mode=control";
3097
3098     if (min_version($machine_version, 2, 12)) {
3099         push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3100         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3101     }
3102
3103     push @$cmd, '-pidfile' , PVE::QemuServer::Helpers::pidfile_name($vmid);
3104
3105     push @$cmd, '-daemonize';
3106
3107     if ($conf->{smbios1}) {
3108         my $smbios_conf = parse_smbios1($conf->{smbios1});
3109         if ($smbios_conf->{base64}) {
3110             # Do not pass base64 flag to qemu
3111             delete $smbios_conf->{base64};
3112             my $smbios_string = "";
3113             foreach my $key (keys %$smbios_conf) {
3114                 my $value;
3115                 if ($key eq "uuid") {
3116                     $value = $smbios_conf->{uuid}
3117                 } else {
3118                     $value = decode_base64($smbios_conf->{$key});
3119                 }
3120                 # qemu accepts any binary data, only commas need escaping by double comma
3121                 $value =~ s/,/,,/g;
3122                 $smbios_string .= "," . $key . "=" . $value if $value;
3123             }
3124             push @$cmd, '-smbios', "type=1" . $smbios_string;
3125         } else {
3126             push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3127         }
3128     }
3129
3130     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3131         my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3132         die "uefi base image '$ovmf_code' not found\n" if ! -f $ovmf_code;
3133
3134         my ($path, $format);
3135         if (my $efidisk = $conf->{efidisk0}) {
3136             my $d = parse_drive('efidisk0', $efidisk);
3137             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3138             $format = $d->{format};
3139             if ($storeid) {
3140                 $path = PVE::Storage::path($storecfg, $d->{file});
3141                 if (!defined($format)) {
3142                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3143                     $format = qemu_img_format($scfg, $volname);
3144                 }
3145             } else {
3146                 $path = $d->{file};
3147                 die "efidisk format must be specified\n"
3148                     if !defined($format);
3149             }
3150         } else {
3151             warn "no efidisk configured! Using temporary efivars disk.\n";
3152             $path = "/tmp/$vmid-ovmf.fd";
3153             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3154             $format = 'raw';
3155         }
3156
3157         my $size_str = "";
3158
3159         if ($format eq 'raw' && $version_guard->(4, 1, 2)) {
3160             $size_str = ",size=" . (-s $ovmf_vars);
3161         }
3162
3163         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3164         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0$size_str,file=$path";
3165     }
3166
3167     # load q35 config
3168     if ($q35) {
3169         # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3170         if (min_version($machine_version, 4, 0)) {
3171             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3172         } else {
3173             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3174         }
3175     }
3176
3177     if ($conf->{vmgenid}) {
3178         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3179     }
3180
3181     # add usb controllers
3182     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers(
3183         $conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3184     push @$devices, @usbcontrollers if @usbcontrollers;
3185     my $vga = parse_vga($conf->{vga});
3186
3187     my $qxlnum = vga_conf_has_spice($conf->{vga});
3188     $vga->{type} = 'qxl' if $qxlnum;
3189
3190     if (!$vga->{type}) {
3191         if ($arch eq 'aarch64') {
3192             $vga->{type} = 'virtio';
3193         } elsif (min_version($machine_version, 2, 9)) {
3194             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3195         } else {
3196             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3197         }
3198     }
3199
3200     # enable absolute mouse coordinates (needed by vnc)
3201     my $tablet;
3202     if (defined($conf->{tablet})) {
3203         $tablet = $conf->{tablet};
3204     } else {
3205         $tablet = $defaults->{tablet};
3206         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3207         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3208     }
3209
3210     if ($tablet) {
3211         push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3212         my $kbd = print_keyboarddevice_full($conf, $arch);
3213         push @$devices, '-device', $kbd if defined($kbd);
3214     }
3215
3216     my $bootorder = device_bootorder($conf);
3217
3218     # host pci device passthrough
3219     my ($kvm_off, $gpu_passthrough, $legacy_igd) = PVE::QemuServer::PCI::print_hostpci_devices(
3220         $vmid, $conf, $devices, $vga, $winversion, $q35, $bridges, $arch, $machine_type, $bootorder);
3221
3222     # usb devices
3223     my $usb_dev_features = {};
3224     $usb_dev_features->{spice_usb3} = 1 if min_version($machine_version, 4, 0);
3225
3226     my @usbdevices = PVE::QemuServer::USB::get_usb_devices(
3227         $conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features, $bootorder);
3228     push @$devices, @usbdevices if @usbdevices;
3229
3230     # serial devices
3231     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3232         if (my $path = $conf->{"serial$i"}) {
3233             if ($path eq 'socket') {
3234                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3235                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3236                 # On aarch64, serial0 is the UART device. Qemu only allows
3237                 # connecting UART devices via the '-serial' command line, as
3238                 # the device has a fixed slot on the hardware...
3239                 if ($arch eq 'aarch64' && $i == 0) {
3240                     push @$devices, '-serial', "chardev:serial$i";
3241                 } else {
3242                     push @$devices, '-device', "isa-serial,chardev=serial$i";
3243                 }
3244             } else {
3245                 die "no such serial device\n" if ! -c $path;
3246                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3247                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3248             }
3249         }
3250     }
3251
3252     # parallel devices
3253     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3254         if (my $path = $conf->{"parallel$i"}) {
3255             die "no such parallel device\n" if ! -c $path;
3256             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3257             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3258             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3259         }
3260     }
3261
3262     if (min_version($machine_version, 4, 0) && (my $audio = conf_has_audio($conf))) {
3263         my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3264         my $audio_devs = audio_devs($audio, $audiopciaddr, $machine_version);
3265         push @$devices, @$audio_devs;
3266     }
3267
3268     my $sockets = 1;
3269     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3270     $sockets = $conf->{sockets} if  $conf->{sockets};
3271
3272     my $cores = $conf->{cores} || 1;
3273
3274     my $maxcpus = $sockets * $cores;
3275
3276     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3277
3278     my $allowed_vcpus = $cpuinfo->{cpus};
3279
3280     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3281         if ($allowed_vcpus < $maxcpus);
3282
3283     if($hotplug_features->{cpu} && min_version($machine_version, 2, 7)) {
3284
3285         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3286         for (my $i = 2; $i <= $vcpus; $i++)  {
3287             my $cpustr = print_cpu_device($conf,$i);
3288             push @$cmd, '-device', $cpustr;
3289         }
3290
3291     } else {
3292
3293         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3294     }
3295     push @$cmd, '-nodefaults';
3296
3297     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3298
3299     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3300
3301     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3302
3303     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3304         push @$devices, '-device', print_vga_device(
3305             $conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3306         my $socket = PVE::QemuServer::Helpers::vnc_socket($vmid);
3307         push @$cmd,  '-vnc', "unix:$socket,password";
3308     } else {
3309         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3310         push @$cmd, '-nographic';
3311     }
3312
3313     # time drift fix
3314     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3315     my $useLocaltime = $conf->{localtime};
3316
3317     if ($winversion >= 5) { # windows
3318         $useLocaltime = 1 if !defined($conf->{localtime});
3319
3320         # use time drift fix when acpi is enabled
3321         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3322             $tdf = 1 if !defined($conf->{tdf});
3323         }
3324     }
3325
3326     if ($winversion >= 6) {
3327         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3328         push @$cmd, '-no-hpet';
3329     }
3330
3331     push @$rtcFlags, 'driftfix=slew' if $tdf;
3332
3333     if ($conf->{startdate} && $conf->{startdate} ne 'now') {
3334         push @$rtcFlags, "base=$conf->{startdate}";
3335     } elsif ($useLocaltime) {
3336         push @$rtcFlags, 'base=localtime';
3337     }
3338
3339     if ($forcecpu) {
3340         push @$cmd, '-cpu', $forcecpu;
3341     } else {
3342         push @$cmd, get_cpu_options($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3343     }
3344
3345     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3346
3347     push @$cmd, '-S' if $conf->{freeze};
3348
3349     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3350
3351     my $guest_agent = parse_guest_agent($conf);
3352
3353     if ($guest_agent->{enabled}) {
3354         my $qgasocket = PVE::QemuServer::Helpers::qmp_socket($vmid, 1);
3355         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3356
3357         if (!$guest_agent->{type} || $guest_agent->{type} eq 'virtio') {
3358             my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3359             push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3360             push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3361         } elsif ($guest_agent->{type} eq 'isa') {
3362             push @$devices, '-device', "isa-serial,chardev=qga0";
3363         }
3364     }
3365
3366     my $rng = $conf->{rng0} ? parse_rng($conf->{rng0}) : undef;
3367     if ($rng && $version_guard->(4, 1, 2)) {
3368         check_rng_source($rng->{source});
3369
3370         my $max_bytes = $rng->{max_bytes} // $rng_fmt->{max_bytes}->{default};
3371         my $period = $rng->{period} // $rng_fmt->{period}->{default};
3372         my $limiter_str = "";
3373         if ($max_bytes) {
3374             $limiter_str = ",max-bytes=$max_bytes,period=$period";
3375         }
3376
3377         my $rng_addr = print_pci_addr("rng0", $bridges, $arch, $machine_type);
3378         push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3379         push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3380     }
3381
3382     my $spice_port;
3383
3384     if ($qxlnum) {
3385         if ($qxlnum > 1) {
3386             if ($winversion){
3387                 for (my $i = 1; $i < $qxlnum; $i++){
3388                     push @$devices, '-device', print_vga_device(
3389                         $conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3390                 }
3391             } else {
3392                 # assume other OS works like Linux
3393                 my ($ram, $vram) = ("134217728", "67108864");
3394                 if ($vga->{memory}) {
3395                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3396                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3397                 }
3398                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3399                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3400             }
3401         }
3402
3403         my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
3404
3405         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3406         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3407         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3408
3409         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3410         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3411         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3412
3413         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3414         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3415
3416         my $spice_enhancement_str = $conf->{spice_enhancements} // '';
3417         my $spice_enhancement = parse_property_string($spice_enhancements_fmt, $spice_enhancement_str);
3418         if ($spice_enhancement->{foldersharing}) {
3419             push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3420             push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3421         }
3422
3423         my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3424         $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}"
3425             if $spice_enhancement->{videostreaming};
3426
3427         push @$devices, '-spice', "$spice_opts";
3428     }
3429
3430     # enable balloon by default, unless explicitly disabled
3431     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3432         $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
3433         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3434     }
3435
3436     if ($conf->{watchdog}) {
3437         my $wdopts = parse_watchdog($conf->{watchdog});
3438         $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
3439         my $watchdog = $wdopts->{model} || 'i6300esb';
3440         push @$devices, '-device', "$watchdog$pciaddr";
3441         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3442     }
3443
3444     my $vollist = [];
3445     my $scsicontroller = {};
3446     my $ahcicontroller = {};
3447     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3448
3449     # Add iscsi initiator name if available
3450     if (my $initiator = get_initiator_name()) {
3451         push @$devices, '-iscsi', "initiator-name=$initiator";
3452     }
3453
3454     PVE::QemuConfig->foreach_volume($conf, sub {
3455         my ($ds, $drive) = @_;
3456
3457         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3458             push @$vollist, $drive->{file};
3459         }
3460
3461         # ignore efidisk here, already added in bios/fw handling code above
3462         return if $drive->{interface} eq 'efidisk';
3463
3464         $use_virtio = 1 if $ds =~ m/^virtio/;
3465
3466         $drive->{bootindex} = $bootorder->{$ds} if $bootorder->{$ds};
3467
3468         if ($drive->{interface} eq 'virtio'){
3469            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3470         }
3471
3472         if ($drive->{interface} eq 'scsi') {
3473
3474             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3475
3476             die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
3477                 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
3478
3479             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
3480             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3481
3482             my $iothread = '';
3483             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3484                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3485                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3486             } elsif ($drive->{iothread}) {
3487                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3488             }
3489
3490             my $queues = '';
3491             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3492                 $queues = ",num_queues=$drive->{queues}";
3493             }
3494
3495             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues"
3496                 if !$scsicontroller->{$controller};
3497             $scsicontroller->{$controller}=1;
3498         }
3499
3500         if ($drive->{interface} eq 'sata') {
3501             my $controller = int($drive->{index} / $PVE::QemuServer::Drive::MAX_SATA_DISKS);
3502             $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
3503             push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr"
3504                 if !$ahcicontroller->{$controller};
3505             $ahcicontroller->{$controller}=1;
3506         }
3507
3508         my $drive_cmd = print_drive_commandline_full($storecfg, $vmid, $drive);
3509         $drive_cmd .= ',readonly' if PVE::QemuConfig->is_template($conf);
3510
3511         push @$devices, '-drive',$drive_cmd;
3512         push @$devices, '-device', print_drivedevice_full(
3513             $storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
3514     });
3515
3516     for (my $i = 0; $i < $MAX_NETS; $i++) {
3517         my $netname = "net$i";
3518
3519         next if !$conf->{$netname};
3520         my $d = parse_net($conf->{$netname});
3521         next if !$d;
3522
3523         $use_virtio = 1 if $d->{model} eq 'virtio';
3524
3525         $d->{bootindex} = $bootorder->{$netname} if $bootorder->{$netname};
3526
3527         my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, $netname);
3528         push @$devices, '-netdev', $netdevfull;
3529
3530         my $netdevicefull = print_netdevice_full(
3531             $vmid, $conf, $d, $netname, $bridges, $use_old_bios_files, $arch, $machine_type);
3532
3533         push @$devices, '-device', $netdevicefull;
3534     }
3535
3536     if ($conf->{ivshmem}) {
3537         my $ivshmem = parse_property_string($ivshmem_fmt, $conf->{ivshmem});
3538
3539         my $bus;
3540         if ($q35) {
3541             $bus = print_pcie_addr("ivshmem");
3542         } else {
3543             $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
3544         }
3545
3546         my $ivshmem_name = $ivshmem->{name} // $vmid;
3547         my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
3548
3549         push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
3550         push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path"
3551             .",size=$ivshmem->{size}M";
3552     }
3553
3554     # pci.4 is nested in pci.1
3555     $bridges->{1} = 1 if $bridges->{4};
3556
3557     if (!$q35) {
3558         # add pci bridges
3559         if (min_version($machine_version, 2, 3)) {
3560            $bridges->{1} = 1;
3561            $bridges->{2} = 1;
3562         }
3563
3564         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3565
3566     }
3567
3568     for my $k (sort {$b cmp $a} keys %$bridges) {
3569         next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
3570
3571         my $k_name = $k;
3572         if ($k == 2 && $legacy_igd) {
3573             $k_name = "$k-igd";
3574         }
3575         $pciaddr = print_pci_addr("pci.$k_name", undef, $arch, $machine_type);
3576
3577         my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
3578         if ($q35) {
3579             # add after -readconfig pve-q35.cfg
3580             splice @$devices, 2, 0, '-device', $devstr;
3581         } else {
3582             unshift @$devices, '-device', $devstr if $k > 0;
3583         }
3584     }
3585
3586     if (!$kvm) {
3587         push @$machineFlags, 'accel=tcg';
3588     }
3589
3590     my $machine_type_min = $machine_type;
3591     if ($add_pve_version) {
3592         $machine_type_min =~ s/\+pve\d+$//;
3593         $machine_type_min .= "+pve$required_pve_version";
3594     }
3595     push @$machineFlags, "type=${machine_type_min}";
3596
3597     push @$cmd, @$devices;
3598     push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
3599     push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
3600     push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
3601
3602     if (my $vmstate = $conf->{vmstate}) {
3603         my $statepath = PVE::Storage::path($storecfg, $vmstate);
3604         push @$vollist, $vmstate;
3605         push @$cmd, '-loadstate', $statepath;
3606         print "activating and using '$vmstate' as vmstate\n";
3607     }
3608
3609     # add custom args
3610     if ($conf->{args}) {
3611         my $aa = PVE::Tools::split_args($conf->{args});
3612         push @$cmd, @$aa;
3613     }
3614
3615     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3616 }
3617
3618 sub check_rng_source {
3619     my ($source) = @_;
3620
3621     # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
3622     die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
3623         if ! -e $source;
3624
3625     my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
3626     if ($source eq '/dev/hwrng' && file_read_firstline($rng_current) eq 'none') {
3627         # Needs to abort, otherwise QEMU crashes on first rng access. Note that rng_current cannot
3628         # be changed to 'none' manually, so once the VM is past this point, it's no longer an issue.
3629         die "Cannot start VM with passed-through RNG device: '/dev/hwrng' exists, but"
3630             ." '$rng_current' is set to 'none'. Ensure that a compatible hardware-RNG is attached"
3631             ." to the host.\n";
3632     }
3633 }
3634
3635 sub spice_port {
3636     my ($vmid) = @_;
3637
3638     my $res = mon_cmd($vmid, 'query-spice');
3639
3640     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3641 }
3642
3643 sub vm_devices_list {
3644     my ($vmid) = @_;
3645
3646     my $res = mon_cmd($vmid, 'query-pci');
3647     my $devices_to_check = [];
3648     my $devices = {};
3649     foreach my $pcibus (@$res) {
3650         push @$devices_to_check, @{$pcibus->{devices}},
3651     }
3652
3653     while (@$devices_to_check) {
3654         my $to_check = [];
3655         for my $d (@$devices_to_check) {
3656             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3657             next if !$d->{'pci_bridge'};
3658
3659             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3660             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3661         }
3662         $devices_to_check = $to_check;
3663     }
3664
3665     my $resblock = mon_cmd($vmid, 'query-block');
3666     foreach my $block (@$resblock) {
3667         if($block->{device} =~ m/^drive-(\S+)/){
3668                 $devices->{$1} = 1;
3669         }
3670     }
3671
3672     my $resmice = mon_cmd($vmid, 'query-mice');
3673     foreach my $mice (@$resmice) {
3674         if ($mice->{name} eq 'QEMU HID Tablet') {
3675             $devices->{tablet} = 1;
3676             last;
3677         }
3678     }
3679
3680     # for usb devices there is no query-usb
3681     # but we can iterate over the entries in
3682     # qom-list path=/machine/peripheral
3683     my $resperipheral = mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3684     foreach my $per (@$resperipheral) {
3685         if ($per->{name} =~ m/^usb\d+$/) {
3686             $devices->{$per->{name}} = 1;
3687         }
3688     }
3689
3690     return $devices;
3691 }
3692
3693 sub vm_deviceplug {
3694     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
3695
3696     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3697
3698     my $devices_list = vm_devices_list($vmid);
3699     return 1 if defined($devices_list->{$deviceid});
3700
3701     # add PCI bridge if we need it for the device
3702     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type);
3703
3704     if ($deviceid eq 'tablet') {
3705
3706         qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
3707
3708     } elsif ($deviceid eq 'keyboard') {
3709
3710         qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
3711
3712     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3713
3714         die "usb hotplug currently not reliable\n";
3715         # since we can't reliably hot unplug all added usb devices and usb
3716         # passthrough breaks live migration we disable usb hotplugging for now
3717         #qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3718
3719     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3720
3721         qemu_iothread_add($vmid, $deviceid, $device);
3722
3723         qemu_driveadd($storecfg, $vmid, $device);
3724         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
3725
3726         qemu_deviceadd($vmid, $devicefull);
3727         eval { qemu_deviceaddverify($vmid, $deviceid); };
3728         if (my $err = $@) {
3729             eval { qemu_drivedel($vmid, $deviceid); };
3730             warn $@ if $@;
3731             die $err;
3732         }
3733
3734     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3735
3736
3737         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3738         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3739         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3740
3741         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3742
3743         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3744             qemu_iothread_add($vmid, $deviceid, $device);
3745             $devicefull .= ",iothread=iothread-$deviceid";
3746         }
3747
3748         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3749             $devicefull .= ",num_queues=$device->{queues}";
3750         }
3751
3752         qemu_deviceadd($vmid, $devicefull);
3753         qemu_deviceaddverify($vmid, $deviceid);
3754
3755     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3756
3757         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
3758         qemu_driveadd($storecfg, $vmid, $device);
3759
3760         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, undef, $arch, $machine_type);
3761         eval { qemu_deviceadd($vmid, $devicefull); };
3762         if (my $err = $@) {
3763             eval { qemu_drivedel($vmid, $deviceid); };
3764             warn $@ if $@;
3765             die $err;
3766         }
3767
3768     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3769
3770         return if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
3771
3772         my $machine_type = PVE::QemuServer::Machine::qemu_machine_pxe($vmid, $conf);
3773         my $use_old_bios_files = undef;
3774         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3775
3776         my $netdevicefull = print_netdevice_full(
3777             $vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
3778         qemu_deviceadd($vmid, $netdevicefull);
3779         eval {
3780             qemu_deviceaddverify($vmid, $deviceid);
3781             qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
3782         };
3783         if (my $err = $@) {
3784             eval { qemu_netdevdel($vmid, $deviceid); };
3785             warn $@ if $@;
3786             die $err;
3787         }
3788
3789     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3790
3791         my $bridgeid = $2;
3792         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3793         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3794
3795         qemu_deviceadd($vmid, $devicefull);
3796         qemu_deviceaddverify($vmid, $deviceid);
3797
3798     } else {
3799         die "can't hotplug device '$deviceid'\n";
3800     }
3801
3802     return 1;
3803 }
3804
3805 # fixme: this should raise exceptions on error!
3806 sub vm_deviceunplug {
3807     my ($vmid, $conf, $deviceid) = @_;
3808
3809     my $devices_list = vm_devices_list($vmid);
3810     return 1 if !defined($devices_list->{$deviceid});
3811
3812     my $bootdisks = PVE::QemuServer::Drive::get_bootdisks($conf);
3813     die "can't unplug bootdisk '$deviceid'\n" if grep {$_ eq $deviceid} @$bootdisks;
3814
3815     if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
3816
3817         qemu_devicedel($vmid, $deviceid);
3818
3819     } elsif ($deviceid =~ m/^usb\d+$/) {
3820
3821         die "usb hotplug currently not reliable\n";
3822         # when unplugging usb devices this way, there may be remaining usb
3823         # controllers/hubs so we disable it for now
3824         #qemu_devicedel($vmid, $deviceid);
3825         #qemu_devicedelverify($vmid, $deviceid);
3826
3827     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3828
3829         qemu_devicedel($vmid, $deviceid);
3830         qemu_devicedelverify($vmid, $deviceid);
3831         qemu_drivedel($vmid, $deviceid);
3832         qemu_iothread_del($conf, $vmid, $deviceid);
3833
3834     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3835
3836         qemu_devicedel($vmid, $deviceid);
3837         qemu_devicedelverify($vmid, $deviceid);
3838         qemu_iothread_del($conf, $vmid, $deviceid);
3839
3840     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3841
3842         qemu_devicedel($vmid, $deviceid);
3843         qemu_drivedel($vmid, $deviceid);
3844         qemu_deletescsihw($conf, $vmid, $deviceid);
3845
3846     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3847
3848         qemu_devicedel($vmid, $deviceid);
3849         qemu_devicedelverify($vmid, $deviceid);
3850         qemu_netdevdel($vmid, $deviceid);
3851
3852     } else {
3853         die "can't unplug device '$deviceid'\n";
3854     }
3855
3856     return 1;
3857 }
3858
3859 sub qemu_deviceadd {
3860     my ($vmid, $devicefull) = @_;
3861
3862     $devicefull = "driver=".$devicefull;
3863     my %options =  split(/[=,]/, $devicefull);
3864
3865     mon_cmd($vmid, "device_add" , %options);
3866 }
3867
3868 sub qemu_devicedel {
3869     my ($vmid, $deviceid) = @_;
3870
3871     my $ret = mon_cmd($vmid, "device_del", id => $deviceid);
3872 }
3873
3874 sub qemu_iothread_add {
3875     my($vmid, $deviceid, $device) = @_;
3876
3877     if ($device->{iothread}) {
3878         my $iothreads = vm_iothreads_list($vmid);
3879         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3880     }
3881 }
3882
3883 sub qemu_iothread_del {
3884     my($conf, $vmid, $deviceid) = @_;
3885
3886     my $confid = $deviceid;
3887     if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
3888         $confid = 'scsi' . $1;
3889     }
3890     my $device = parse_drive($confid, $conf->{$confid});
3891     if ($device->{iothread}) {
3892         my $iothreads = vm_iothreads_list($vmid);
3893         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3894     }
3895 }
3896
3897 sub qemu_objectadd {
3898     my($vmid, $objectid, $qomtype) = @_;
3899
3900     mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3901
3902     return 1;
3903 }
3904
3905 sub qemu_objectdel {
3906     my($vmid, $objectid) = @_;
3907
3908     mon_cmd($vmid, "object-del", id => $objectid);
3909
3910     return 1;
3911 }
3912
3913 sub qemu_driveadd {
3914     my ($storecfg, $vmid, $device) = @_;
3915
3916     my $drive = print_drive_commandline_full($storecfg, $vmid, $device);
3917     $drive =~ s/\\/\\\\/g;
3918     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_add auto \"$drive\"");
3919
3920     # If the command succeeds qemu prints: "OK"
3921     return 1 if $ret =~ m/OK/s;
3922
3923     die "adding drive failed: $ret\n";
3924 }
3925
3926 sub qemu_drivedel {
3927     my($vmid, $deviceid) = @_;
3928
3929     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-$deviceid");
3930     $ret =~ s/^\s+//;
3931
3932     return 1 if $ret eq "";
3933
3934     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3935     return 1 if $ret =~ m/Device \'.*?\' not found/s;
3936
3937     die "deleting drive $deviceid failed : $ret\n";
3938 }
3939
3940 sub qemu_deviceaddverify {
3941     my ($vmid, $deviceid) = @_;
3942
3943     for (my $i = 0; $i <= 5; $i++) {
3944          my $devices_list = vm_devices_list($vmid);
3945          return 1 if defined($devices_list->{$deviceid});
3946          sleep 1;
3947     }
3948
3949     die "error on hotplug device '$deviceid'\n";
3950 }
3951
3952
3953 sub qemu_devicedelverify {
3954     my ($vmid, $deviceid) = @_;
3955
3956     # need to verify that the device is correctly removed as device_del
3957     # is async and empty return is not reliable
3958
3959     for (my $i = 0; $i <= 5; $i++) {
3960          my $devices_list = vm_devices_list($vmid);
3961          return 1 if !defined($devices_list->{$deviceid});
3962          sleep 1;
3963     }
3964
3965     die "error on hot-unplugging device '$deviceid'\n";
3966 }
3967
3968 sub qemu_findorcreatescsihw {
3969     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
3970
3971     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3972
3973     my $scsihwid="$controller_prefix$controller";
3974     my $devices_list = vm_devices_list($vmid);
3975
3976     if(!defined($devices_list->{$scsihwid})) {
3977         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
3978     }
3979
3980     return 1;
3981 }
3982
3983 sub qemu_deletescsihw {
3984     my ($conf, $vmid, $opt) = @_;
3985
3986     my $device = parse_drive($opt, $conf->{$opt});
3987
3988     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
3989         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3990         return 1;
3991     }
3992
3993     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3994
3995     my $devices_list = vm_devices_list($vmid);
3996     foreach my $opt (keys %{$devices_list}) {
3997         if (is_valid_drivename($opt)) {
3998             my $drive = parse_drive($opt, $conf->{$opt});
3999             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4000                 return 1;
4001             }
4002         }
4003     }
4004
4005     my $scsihwid="scsihw$controller";
4006
4007     vm_deviceunplug($vmid, $conf, $scsihwid);
4008
4009     return 1;
4010 }
4011
4012 sub qemu_add_pci_bridge {
4013     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4014
4015     my $bridges = {};
4016
4017     my $bridgeid;
4018
4019     print_pci_addr($device, $bridges, $arch, $machine_type);
4020
4021     while (my ($k, $v) = each %$bridges) {
4022         $bridgeid = $k;
4023     }
4024     return 1 if !defined($bridgeid) || $bridgeid < 1;
4025
4026     my $bridge = "pci.$bridgeid";
4027     my $devices_list = vm_devices_list($vmid);
4028
4029     if (!defined($devices_list->{$bridge})) {
4030         vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4031     }
4032
4033     return 1;
4034 }
4035
4036 sub qemu_set_link_status {
4037     my ($vmid, $device, $up) = @_;
4038
4039     mon_cmd($vmid, "set_link", name => $device,
4040                up => $up ? JSON::true : JSON::false);
4041 }
4042
4043 sub qemu_netdevadd {
4044     my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4045
4046     my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4047     my %options =  split(/[=,]/, $netdev);
4048
4049     if (defined(my $vhost = $options{vhost})) {
4050         $options{vhost} = JSON::boolean(PVE::JSONSchema::parse_boolean($vhost));
4051     }
4052
4053     if (defined(my $queues = $options{queues})) {
4054         $options{queues} = $queues + 0;
4055     }
4056
4057     mon_cmd($vmid, "netdev_add",  %options);
4058     return 1;
4059 }
4060
4061 sub qemu_netdevdel {
4062     my ($vmid, $deviceid) = @_;
4063
4064     mon_cmd($vmid, "netdev_del", id => $deviceid);
4065 }
4066
4067 sub qemu_usb_hotplug {
4068     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4069
4070     return if !$device;
4071
4072     # remove the old one first
4073     vm_deviceunplug($vmid, $conf, $deviceid);
4074
4075     # check if xhci controller is necessary and available
4076     if ($device->{usb3}) {
4077
4078         my $devicelist = vm_devices_list($vmid);
4079
4080         if (!$devicelist->{xhci}) {
4081             my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4082             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4083         }
4084     }
4085     my $d = parse_usb_device($device->{host});
4086     $d->{usb3} = $device->{usb3};
4087
4088     # add the new one
4089     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4090 }
4091
4092 sub qemu_cpu_hotplug {
4093     my ($vmid, $conf, $vcpus) = @_;
4094
4095     my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4096
4097     my $sockets = 1;
4098     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4099     $sockets = $conf->{sockets} if  $conf->{sockets};
4100     my $cores = $conf->{cores} || 1;
4101     my $maxcpus = $sockets * $cores;
4102
4103     $vcpus = $maxcpus if !$vcpus;
4104
4105     die "you can't add more vcpus than maxcpus\n"
4106         if $vcpus > $maxcpus;
4107
4108     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4109
4110     if ($vcpus < $currentvcpus) {
4111
4112         if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4113
4114             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4115                 qemu_devicedel($vmid, "cpu$i");
4116                 my $retry = 0;
4117                 my $currentrunningvcpus = undef;
4118                 while (1) {
4119                     $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4120                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4121                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4122                     $retry++;
4123                     sleep 1;
4124                 }
4125                 #update conf after each succesfull cpu unplug
4126                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4127                 PVE::QemuConfig->write_config($vmid, $conf);
4128             }
4129         } else {
4130             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4131         }
4132
4133         return;
4134     }
4135
4136     my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4137     die "vcpus in running vm does not match its configuration\n"
4138         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4139
4140     if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4141
4142         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4143             my $cpustr = print_cpu_device($conf, $i);
4144             qemu_deviceadd($vmid, $cpustr);
4145
4146             my $retry = 0;
4147             my $currentrunningvcpus = undef;
4148             while (1) {
4149                 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4150                 last if scalar(@{$currentrunningvcpus}) == $i;
4151                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4152                 sleep 1;
4153                 $retry++;
4154             }
4155             #update conf after each succesfull cpu hotplug
4156             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4157             PVE::QemuConfig->write_config($vmid, $conf);
4158         }
4159     } else {
4160
4161         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4162             mon_cmd($vmid, "cpu-add", id => int($i));
4163         }
4164     }
4165 }
4166
4167 sub qemu_block_set_io_throttle {
4168     my ($vmid, $deviceid,
4169         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4170         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4171         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4172         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4173
4174     return if !check_running($vmid) ;
4175
4176     mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4177         bps => int($bps),
4178         bps_rd => int($bps_rd),
4179         bps_wr => int($bps_wr),
4180         iops => int($iops),
4181         iops_rd => int($iops_rd),
4182         iops_wr => int($iops_wr),
4183         bps_max => int($bps_max),
4184         bps_rd_max => int($bps_rd_max),
4185         bps_wr_max => int($bps_wr_max),
4186         iops_max => int($iops_max),
4187         iops_rd_max => int($iops_rd_max),
4188         iops_wr_max => int($iops_wr_max),
4189         bps_max_length => int($bps_max_length),
4190         bps_rd_max_length => int($bps_rd_max_length),
4191         bps_wr_max_length => int($bps_wr_max_length),
4192         iops_max_length => int($iops_max_length),
4193         iops_rd_max_length => int($iops_rd_max_length),
4194         iops_wr_max_length => int($iops_wr_max_length),
4195     );
4196
4197 }
4198
4199 sub qemu_block_resize {
4200     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4201
4202     my $running = check_running($vmid);
4203
4204     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4205
4206     return if !$running;
4207
4208     my $padding = (1024 - $size % 1024) % 1024;
4209     $size = $size + $padding;
4210
4211     mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4212
4213 }
4214
4215 sub qemu_volume_snapshot {
4216     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4217
4218     my $running = check_running($vmid);
4219
4220     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4221         mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4222     } else {
4223         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4224     }
4225 }
4226
4227 sub qemu_volume_snapshot_delete {
4228     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4229
4230     my $running = check_running($vmid);
4231
4232     if($running) {
4233
4234         $running = undef;
4235         my $conf = PVE::QemuConfig->load_config($vmid);
4236         PVE::QemuConfig->foreach_volume($conf, sub {
4237             my ($ds, $drive) = @_;
4238             $running = 1 if $drive->{file} eq $volid;
4239         });
4240     }
4241
4242     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4243         mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4244     } else {
4245         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4246     }
4247 }
4248
4249 sub set_migration_caps {
4250     my ($vmid) = @_;
4251
4252     my $qemu_support = eval { mon_cmd($vmid, "query-proxmox-support") };
4253
4254     my $cap_ref = [];
4255
4256     my $enabled_cap = {
4257         "auto-converge" => 1,
4258         "xbzrle" => 1,
4259         "x-rdma-pin-all" => 0,
4260         "zero-blocks" => 0,
4261         "compress" => 0,
4262         "dirty-bitmaps" => $qemu_support->{'pbs-dirty-bitmap-migration'} ? 1 : 0,
4263     };
4264
4265     my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4266
4267     for my $supported_capability (@$supported_capabilities) {
4268         push @$cap_ref, {
4269             capability => $supported_capability->{capability},
4270             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4271         };
4272     }
4273
4274     mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4275 }
4276
4277 sub foreach_volid {
4278     my ($conf, $func, @param) = @_;
4279
4280     my $volhash = {};
4281
4282     my $test_volid = sub {
4283         my ($key, $drive, $snapname) = @_;
4284
4285         my $volid = $drive->{file};
4286         return if !$volid;
4287
4288         $volhash->{$volid}->{cdrom} //= 1;
4289         $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4290
4291         my $replicate = $drive->{replicate} // 1;
4292         $volhash->{$volid}->{replicate} //= 0;
4293         $volhash->{$volid}->{replicate} = 1 if $replicate;
4294
4295         $volhash->{$volid}->{shared} //= 0;
4296         $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4297
4298         $volhash->{$volid}->{referenced_in_config} //= 0;
4299         $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4300
4301         $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4302             if defined($snapname);
4303
4304         my $size = $drive->{size};
4305         $volhash->{$volid}->{size} //= $size if $size;
4306
4307         $volhash->{$volid}->{is_vmstate} //= 0;
4308         $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate';
4309
4310         $volhash->{$volid}->{is_unused} //= 0;
4311         $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4312     };
4313
4314     my $include_opts = {
4315         extra_keys => ['vmstate'],
4316         include_unused => 1,
4317     };
4318
4319     PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4320     foreach my $snapname (keys %{$conf->{snapshots}}) {
4321         my $snap = $conf->{snapshots}->{$snapname};
4322         PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4323     }
4324
4325     foreach my $volid (keys %$volhash) {
4326         &$func($volid, $volhash->{$volid}, @param);
4327     }
4328 }
4329
4330 my $fast_plug_option = {
4331     'lock' => 1,
4332     'name' => 1,
4333     'onboot' => 1,
4334     'shares' => 1,
4335     'startup' => 1,
4336     'description' => 1,
4337     'protection' => 1,
4338     'vmstatestorage' => 1,
4339     'hookscript' => 1,
4340     'tags' => 1,
4341 };
4342
4343 # hotplug changes in [PENDING]
4344 # $selection hash can be used to only apply specified options, for
4345 # example: { cores => 1 } (only apply changed 'cores')
4346 # $errors ref is used to return error messages
4347 sub vmconfig_hotplug_pending {
4348     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4349
4350     my $defaults = load_defaults();
4351     my $arch = get_vm_arch($conf);
4352     my $machine_type = get_vm_machine($conf, undef, $arch);
4353
4354     # commit values which do not have any impact on running VM first
4355     # Note: those option cannot raise errors, we we do not care about
4356     # $selection and always apply them.
4357
4358     my $add_error = sub {
4359         my ($opt, $msg) = @_;
4360         $errors->{$opt} = "hotplug problem - $msg";
4361     };
4362
4363     my $changes = 0;
4364     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4365         if ($fast_plug_option->{$opt}) {
4366             $conf->{$opt} = $conf->{pending}->{$opt};
4367             delete $conf->{pending}->{$opt};
4368             $changes = 1;
4369         }
4370     }
4371
4372     if ($changes) {
4373         PVE::QemuConfig->write_config($vmid, $conf);
4374     }
4375
4376     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4377
4378     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4379     foreach my $opt (sort keys %$pending_delete_hash) {
4380         next if $selection && !$selection->{$opt};
4381         my $force = $pending_delete_hash->{$opt}->{force};
4382         eval {
4383             if ($opt eq 'hotplug') {
4384                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4385             } elsif ($opt eq 'tablet') {
4386                 die "skip\n" if !$hotplug_features->{usb};
4387                 if ($defaults->{tablet}) {
4388                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4389                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4390                         if $arch eq 'aarch64';
4391                 } else {
4392                     vm_deviceunplug($vmid, $conf, 'tablet');
4393                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4394                 }
4395             } elsif ($opt =~ m/^usb\d+/) {
4396                 die "skip\n";
4397                 # since we cannot reliably hot unplug usb devices we are disabling it
4398                 #die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4399                 #vm_deviceunplug($vmid, $conf, $opt);
4400             } elsif ($opt eq 'vcpus') {
4401                 die "skip\n" if !$hotplug_features->{cpu};
4402                 qemu_cpu_hotplug($vmid, $conf, undef);
4403             } elsif ($opt eq 'balloon') {
4404                 # enable balloon device is not hotpluggable
4405                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4406                 # here we reset the ballooning value to memory
4407                 my $balloon = $conf->{memory} || $defaults->{memory};
4408                 mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4409             } elsif ($fast_plug_option->{$opt}) {
4410                 # do nothing
4411             } elsif ($opt =~ m/^net(\d+)$/) {
4412                 die "skip\n" if !$hotplug_features->{network};
4413                 vm_deviceunplug($vmid, $conf, $opt);
4414             } elsif (is_valid_drivename($opt)) {
4415                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4416                 vm_deviceunplug($vmid, $conf, $opt);
4417                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4418             } elsif ($opt =~ m/^memory$/) {
4419                 die "skip\n" if !$hotplug_features->{memory};
4420                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4421             } elsif ($opt eq 'cpuunits') {
4422                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4423             } elsif ($opt eq 'cpulimit') {
4424                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4425             } else {
4426                 die "skip\n";
4427             }
4428         };
4429         if (my $err = $@) {
4430             &$add_error($opt, $err) if $err ne "skip\n";
4431         } else {
4432             delete $conf->{$opt};
4433             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4434         }
4435     }
4436
4437     my ($apply_pending_cloudinit, $apply_pending_cloudinit_done);
4438     $apply_pending_cloudinit = sub {
4439         return if $apply_pending_cloudinit_done; # once is enough
4440         $apply_pending_cloudinit_done = 1; # once is enough
4441
4442         my ($key, $value) = @_;
4443
4444         my @cloudinit_opts = keys %$confdesc_cloudinit;
4445         foreach my $opt (keys %{$conf->{pending}}) {
4446             next if !grep { $_ eq $opt } @cloudinit_opts;
4447             $conf->{$opt} = delete $conf->{pending}->{$opt};
4448         }
4449
4450         my $new_conf = { %$conf };
4451         $new_conf->{$key} = $value;
4452         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4453     };
4454
4455     foreach my $opt (keys %{$conf->{pending}}) {
4456         next if $selection && !$selection->{$opt};
4457         my $value = $conf->{pending}->{$opt};
4458         eval {
4459             if ($opt eq 'hotplug') {
4460                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4461             } elsif ($opt eq 'tablet') {
4462                 die "skip\n" if !$hotplug_features->{usb};
4463                 if ($value == 1) {
4464                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4465                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4466                         if $arch eq 'aarch64';
4467                 } elsif ($value == 0) {
4468                     vm_deviceunplug($vmid, $conf, 'tablet');
4469                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4470                 }
4471             } elsif ($opt =~ m/^usb\d+$/) {
4472                 die "skip\n";
4473                 # since we cannot reliably hot unplug usb devices we disable it for now
4474                 #die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4475                 #my $d = eval { parse_property_string($usbdesc->{format}, $value) };
4476                 #die "skip\n" if !$d;
4477                 #qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
4478             } elsif ($opt eq 'vcpus') {
4479                 die "skip\n" if !$hotplug_features->{cpu};
4480                 qemu_cpu_hotplug($vmid, $conf, $value);
4481             } elsif ($opt eq 'balloon') {
4482                 # enable/disable balloning device is not hotpluggable
4483                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4484                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4485                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4486
4487                 # allow manual ballooning if shares is set to zero
4488                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4489                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4490                     mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4491                 }
4492             } elsif ($opt =~ m/^net(\d+)$/) {
4493                 # some changes can be done without hotplug
4494                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4495                                     $vmid, $opt, $value, $arch, $machine_type);
4496             } elsif (is_valid_drivename($opt)) {
4497                 die "skip\n" if $opt eq 'efidisk0';
4498                 # some changes can be done without hotplug
4499                 my $drive = parse_drive($opt, $value);
4500                 if (drive_is_cloudinit($drive)) {
4501                     &$apply_pending_cloudinit($opt, $value);
4502                 }
4503                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4504                                      $vmid, $opt, $value, $arch, $machine_type);
4505             } elsif ($opt =~ m/^memory$/) { #dimms
4506                 die "skip\n" if !$hotplug_features->{memory};
4507                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4508             } elsif ($opt eq 'cpuunits') {
4509                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4510             } elsif ($opt eq 'cpulimit') {
4511                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4512                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4513             } else {
4514                 die "skip\n";  # skip non-hot-pluggable options
4515             }
4516         };
4517         if (my $err = $@) {
4518             &$add_error($opt, $err) if $err ne "skip\n";
4519         } else {
4520             $conf->{$opt} = $value;
4521             delete $conf->{pending}->{$opt};
4522         }
4523     }
4524
4525     PVE::QemuConfig->write_config($vmid, $conf);
4526 }
4527
4528 sub try_deallocate_drive {
4529     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4530
4531     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4532         my $volid = $drive->{file};
4533         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4534             my $sid = PVE::Storage::parse_volume_id($volid);
4535             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4536
4537             # check if the disk is really unused
4538             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4539                 if PVE::QemuServer::Drive::is_volume_in_use($storecfg, $conf, $key, $volid);
4540             PVE::Storage::vdisk_free($storecfg, $volid);
4541             return 1;
4542         } else {
4543             # If vm is not owner of this disk remove from config
4544             return 1;
4545         }
4546     }
4547
4548     return;
4549 }
4550
4551 sub vmconfig_delete_or_detach_drive {
4552     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4553
4554     my $drive = parse_drive($opt, $conf->{$opt});
4555
4556     my $rpcenv = PVE::RPCEnvironment::get();
4557     my $authuser = $rpcenv->get_user();
4558
4559     if ($force) {
4560         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4561         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4562     } else {
4563         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4564     }
4565 }
4566
4567
4568
4569 sub vmconfig_apply_pending {
4570     my ($vmid, $conf, $storecfg, $errors) = @_;
4571
4572     my $add_apply_error = sub {
4573         my ($opt, $msg) = @_;
4574         my $err_msg = "unable to apply pending change $opt : $msg";
4575         $errors->{$opt} = $err_msg;
4576         warn $err_msg;
4577     };
4578
4579     # cold plug
4580
4581     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4582     foreach my $opt (sort keys %$pending_delete_hash) {
4583         my $force = $pending_delete_hash->{$opt}->{force};
4584         eval {
4585             if ($opt =~ m/^unused/) {
4586                 die "internal error";
4587             } elsif (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4588                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4589             }
4590         };
4591         if (my $err = $@) {
4592             $add_apply_error->($opt, $err);
4593         } else {
4594             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4595             delete $conf->{$opt};
4596         }
4597     }
4598
4599     PVE::QemuConfig->cleanup_pending($conf);
4600
4601     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4602         next if $opt eq 'delete'; # just to be sure
4603         eval {
4604             if (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4605                 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4606             }
4607         };
4608         if (my $err = $@) {
4609             $add_apply_error->($opt, $err);
4610         } else {
4611             $conf->{$opt} = delete $conf->{pending}->{$opt};
4612         }
4613     }
4614
4615     # write all changes at once to avoid unnecessary i/o
4616     PVE::QemuConfig->write_config($vmid, $conf);
4617 }
4618
4619 sub vmconfig_update_net {
4620     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4621
4622     my $newnet = parse_net($value);
4623
4624     if ($conf->{$opt}) {
4625         my $oldnet = parse_net($conf->{$opt});
4626
4627         if (safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4628             safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4629             safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4630             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4631
4632             # for non online change, we try to hot-unplug
4633             die "skip\n" if !$hotplug;
4634             vm_deviceunplug($vmid, $conf, $opt);
4635         } else {
4636
4637             die "internal error" if $opt !~ m/net(\d+)/;
4638             my $iface = "tap${vmid}i$1";
4639
4640             if (safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4641                 safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4642                 safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4643                 safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4644                 PVE::Network::tap_unplug($iface);
4645
4646                 if ($have_sdn) {
4647                     PVE::Network::SDN::Zones::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4648                 } else {
4649                     PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4650                 }
4651             } elsif (safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4652                 # Rate can be applied on its own but any change above needs to
4653                 # include the rate in tap_plug since OVS resets everything.
4654                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4655             }
4656
4657             if (safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4658                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4659             }
4660
4661             return 1;
4662         }
4663     }
4664
4665     if ($hotplug) {
4666         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
4667     } else {
4668         die "skip\n";
4669     }
4670 }
4671
4672 sub vmconfig_update_disk {
4673     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4674
4675     my $drive = parse_drive($opt, $value);
4676
4677     if ($conf->{$opt} && (my $old_drive = parse_drive($opt, $conf->{$opt}))) {
4678         my $media = $drive->{media} || 'disk';
4679         my $oldmedia = $old_drive->{media} || 'disk';
4680         die "unable to change media type\n" if $media ne $oldmedia;
4681
4682         if (!drive_is_cdrom($old_drive)) {
4683
4684             if ($drive->{file} ne $old_drive->{file}) {
4685
4686                 die "skip\n" if !$hotplug;
4687
4688                 # unplug and register as unused
4689                 vm_deviceunplug($vmid, $conf, $opt);
4690                 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4691
4692             } else {
4693                 # update existing disk
4694
4695                 # skip non hotpluggable value
4696                 if (safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4697                     safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4698                     safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4699                     safe_string_ne($drive->{cache}, $old_drive->{cache}) ||
4700                     safe_string_ne($drive->{ssd}, $old_drive->{ssd})) {
4701                     die "skip\n";
4702                 }
4703
4704                 # apply throttle
4705                 if (safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4706                     safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4707                     safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4708                     safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4709                     safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4710                     safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4711                     safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4712                     safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4713                     safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4714                     safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4715                     safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4716                     safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4717                     safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4718                     safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4719                     safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4720                     safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4721                     safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4722                     safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4723
4724                     qemu_block_set_io_throttle(
4725                         $vmid,"drive-$opt",
4726                         ($drive->{mbps} || 0)*1024*1024,
4727                         ($drive->{mbps_rd} || 0)*1024*1024,
4728                         ($drive->{mbps_wr} || 0)*1024*1024,
4729                          $drive->{iops} || 0,
4730                          $drive->{iops_rd} || 0,
4731                          $drive->{iops_wr} || 0,
4732                         ($drive->{mbps_max} || 0)*1024*1024,
4733                         ($drive->{mbps_rd_max} || 0)*1024*1024,
4734                         ($drive->{mbps_wr_max} || 0)*1024*1024,
4735                          $drive->{iops_max} || 0,
4736                          $drive->{iops_rd_max} || 0,
4737                          $drive->{iops_wr_max} || 0,
4738                          $drive->{bps_max_length} || 1,
4739                          $drive->{bps_rd_max_length} || 1,
4740                          $drive->{bps_wr_max_length} || 1,
4741                          $drive->{iops_max_length} || 1,
4742                          $drive->{iops_rd_max_length} || 1,
4743                          $drive->{iops_wr_max_length} || 1,
4744                     );
4745
4746                 }
4747
4748                 return 1;
4749             }
4750
4751         } else { # cdrom
4752
4753             if ($drive->{file} eq 'none') {
4754                 mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4755                 if (drive_is_cloudinit($old_drive)) {
4756                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4757                 }
4758             } else {
4759                 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4760
4761                 # force eject if locked
4762                 mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4763
4764                 if ($path) {
4765                     mon_cmd($vmid, "blockdev-change-medium",
4766                         id => "$opt", filename => "$path");
4767                 }
4768             }
4769
4770             return 1;
4771         }
4772     }
4773
4774     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4775     # hotplug new disks
4776     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4777     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
4778 }
4779
4780 # called in locked context by incoming migration
4781 sub vm_migrate_get_nbd_disks {
4782     my ($storecfg, $conf, $replicated_volumes) = @_;
4783
4784     my $local_volumes = {};
4785     PVE::QemuConfig->foreach_volume($conf, sub {
4786         my ($ds, $drive) = @_;
4787
4788         return if drive_is_cdrom($drive);
4789
4790         my $volid = $drive->{file};
4791
4792         return if !$volid;
4793
4794         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
4795
4796         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4797         return if $scfg->{shared};
4798
4799         # replicated disks re-use existing state via bitmap
4800         my $use_existing = $replicated_volumes->{$volid} ? 1 : 0;
4801         $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
4802     });
4803     return $local_volumes;
4804 }
4805
4806 # called in locked context by incoming migration
4807 sub vm_migrate_alloc_nbd_disks {
4808     my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
4809
4810     my $format = undef;
4811
4812     my $nbd = {};
4813     foreach my $opt (sort keys %$source_volumes) {
4814         my ($volid, $storeid, $volname, $drive, $use_existing) = @{$source_volumes->{$opt}};
4815
4816         if ($use_existing) {
4817             $nbd->{$opt}->{drivestr} = print_drive($drive);
4818             $nbd->{$opt}->{volid} = $volid;
4819             $nbd->{$opt}->{replicated} = 1;
4820             next;
4821         }
4822
4823         # If a remote storage is specified and the format of the original
4824         # volume is not available there, fall back to the default format.
4825         # Otherwise use the same format as the original.
4826         if (!$storagemap->{identity}) {
4827             $storeid = map_storage($storagemap, $storeid);
4828             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
4829             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4830             my $fileFormat = qemu_img_format($scfg, $volname);
4831             $format = (grep {$fileFormat eq $_} @{$validFormats}) ? $fileFormat : $defFormat;
4832         } else {
4833             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4834             $format = qemu_img_format($scfg, $volname);
4835         }
4836
4837         my $size = $drive->{size} / 1024;
4838         my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, $size);
4839         my $newdrive = $drive;
4840         $newdrive->{format} = $format;
4841         $newdrive->{file} = $newvolid;
4842         my $drivestr = print_drive($newdrive);
4843         $nbd->{$opt}->{drivestr} = $drivestr;
4844         $nbd->{$opt}->{volid} = $newvolid;
4845     }
4846
4847     return $nbd;
4848 }
4849
4850 # see vm_start_nolock for parameters, additionally:
4851 # migrate_opts:
4852 #   storagemap = parsed storage map for allocating NBD disks
4853 sub vm_start {
4854     my ($storecfg, $vmid, $params, $migrate_opts) = @_;
4855
4856     return PVE::QemuConfig->lock_config($vmid, sub {
4857         my $conf = PVE::QemuConfig->load_config($vmid, $migrate_opts->{migratedfrom});
4858
4859         die "you can't start a vm if it's a template\n"
4860             if !$params->{skiptemplate} && PVE::QemuConfig->is_template($conf);
4861
4862         my $has_suspended_lock = PVE::QemuConfig->has_lock($conf, 'suspended');
4863
4864         PVE::QemuConfig->check_lock($conf)
4865             if !($params->{skiplock} || $has_suspended_lock);
4866
4867         $params->{resume} = $has_suspended_lock || defined($conf->{vmstate});
4868
4869         die "VM $vmid already running\n" if check_running($vmid, undef, $migrate_opts->{migratedfrom});
4870
4871         if (my $storagemap = $migrate_opts->{storagemap}) {
4872             my $replicated = $migrate_opts->{replicated_volumes};
4873             my $disks = vm_migrate_get_nbd_disks($storecfg, $conf, $replicated);
4874             $migrate_opts->{nbd} = vm_migrate_alloc_nbd_disks($storecfg, $vmid, $disks, $storagemap);
4875
4876             foreach my $opt (keys %{$migrate_opts->{nbd}}) {
4877                 $conf->{$opt} = $migrate_opts->{nbd}->{$opt}->{drivestr};
4878             }
4879         }
4880
4881         return vm_start_nolock($storecfg, $vmid, $conf, $params, $migrate_opts);
4882     });
4883 }
4884
4885
4886 # params:
4887 #   statefile => 'tcp', 'unix' for migration or path/volid for RAM state
4888 #   skiplock => 0/1, skip checking for config lock
4889 #   skiptemplate => 0/1, skip checking whether VM is template
4890 #   forcemachine => to force Qemu machine (rollback/migration)
4891 #   forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
4892 #   timeout => in seconds
4893 #   paused => start VM in paused state (backup)
4894 #   resume => resume from hibernation
4895 # migrate_opts:
4896 #   nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
4897 #   migratedfrom => source node
4898 #   spice_ticket => used for spice migration, passed via tunnel/stdin
4899 #   network => CIDR of migration network
4900 #   type => secure/insecure - tunnel over encrypted connection or plain-text
4901 #   nbd_proto_version => int, 0 for TCP, 1 for UNIX
4902 #   replicated_volumes = which volids should be re-used with bitmaps for nbd migration
4903 sub vm_start_nolock {
4904     my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
4905
4906     my $statefile = $params->{statefile};
4907     my $resume = $params->{resume};
4908
4909     my $migratedfrom = $migrate_opts->{migratedfrom};
4910     my $migration_type = $migrate_opts->{type};
4911
4912     my $res = {};
4913
4914     # clean up leftover reboot request files
4915     eval { clear_reboot_request($vmid); };
4916     warn $@ if $@;
4917
4918     if (!$statefile && scalar(keys %{$conf->{pending}})) {
4919         vmconfig_apply_pending($vmid, $conf, $storecfg);
4920         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4921     }
4922
4923     PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4924
4925     my $defaults = load_defaults();
4926
4927     # set environment variable useful inside network script
4928     $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4929
4930     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
4931
4932     my $forcemachine = $params->{forcemachine};
4933     my $forcecpu = $params->{forcecpu};
4934     if ($resume) {
4935         # enforce machine and CPU type on suspended vm to ensure HW compatibility
4936         $forcemachine = $conf->{runningmachine};
4937         $forcecpu = $conf->{runningcpu};
4938         print "Resuming suspended VM\n";
4939     }
4940
4941     my ($cmd, $vollist, $spice_port) =
4942         config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
4943
4944     my $migration_ip;
4945     my $get_migration_ip = sub {
4946         my ($nodename) = @_;
4947
4948         return $migration_ip if defined($migration_ip);
4949
4950         my $cidr = $migrate_opts->{network};
4951
4952         if (!defined($cidr)) {
4953             my $dc_conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4954             $cidr = $dc_conf->{migration}->{network};
4955         }
4956
4957         if (defined($cidr)) {
4958             my $ips = PVE::Network::get_local_ip_from_cidr($cidr);
4959
4960             die "could not get IP: no address configured on local " .
4961                 "node for network '$cidr'\n" if scalar(@$ips) == 0;
4962
4963             die "could not get IP: multiple addresses configured on local " .
4964                 "node for network '$cidr'\n" if scalar(@$ips) > 1;
4965
4966             $migration_ip = @$ips[0];
4967         }
4968
4969         $migration_ip = PVE::Cluster::remote_node_ip($nodename, 1)
4970             if !defined($migration_ip);
4971
4972         return $migration_ip;
4973     };
4974
4975     my $migrate_uri;
4976     if ($statefile) {
4977         if ($statefile eq 'tcp') {
4978             my $localip = "localhost";
4979             my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4980             my $nodename = nodename();
4981
4982             if (!defined($migration_type)) {
4983                 if (defined($datacenterconf->{migration}->{type})) {
4984                     $migration_type = $datacenterconf->{migration}->{type};
4985                 } else {
4986                     $migration_type = 'secure';
4987                 }
4988             }
4989
4990             if ($migration_type eq 'insecure') {
4991                 $localip = $get_migration_ip->($nodename);
4992                 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
4993             }
4994
4995             my $pfamily = PVE::Tools::get_host_address_family($nodename);
4996             my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
4997             $migrate_uri = "tcp:${localip}:${migrate_port}";
4998             push @$cmd, '-incoming', $migrate_uri;
4999             push @$cmd, '-S';
5000
5001         } elsif ($statefile eq 'unix') {
5002             # should be default for secure migrations as a ssh TCP forward
5003             # tunnel is not deterministic reliable ready and fails regurarly
5004             # to set up in time, so use UNIX socket forwards
5005             my $socket_addr = "/run/qemu-server/$vmid.migrate";
5006             unlink $socket_addr;
5007
5008             $migrate_uri = "unix:$socket_addr";
5009
5010             push @$cmd, '-incoming', $migrate_uri;
5011             push @$cmd, '-S';
5012
5013         } elsif (-e $statefile) {
5014             push @$cmd, '-loadstate', $statefile;
5015         } else {
5016             my $statepath = PVE::Storage::path($storecfg, $statefile);
5017             push @$vollist, $statefile;
5018             push @$cmd, '-loadstate', $statepath;
5019         }
5020     } elsif ($params->{paused}) {
5021         push @$cmd, '-S';
5022     }
5023
5024     # host pci devices
5025     for (my $i = 0; $i < $PVE::QemuServer::PCI::MAX_HOSTPCI_DEVICES; $i++)  {
5026       my $d = parse_hostpci($conf->{"hostpci$i"});
5027       next if !$d;
5028       my $pcidevices = $d->{pciid};
5029       foreach my $pcidevice (@$pcidevices) {
5030             my $pciid = $pcidevice->{id};
5031
5032             my $info = PVE::SysFSTools::pci_device_info("$pciid");
5033             die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5034             die "no pci device info for device '$pciid'\n" if !$info;
5035
5036             if ($d->{mdev}) {
5037                 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5038                 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5039             } else {
5040                 die "can't unbind/bind PCI group to VFIO '$pciid'\n"
5041                     if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5042                 die "can't reset PCI device '$pciid'\n"
5043                     if $info->{has_fl_reset} && !PVE::SysFSTools::pci_dev_reset($info);
5044             }
5045       }
5046     }
5047
5048     PVE::Storage::activate_volumes($storecfg, $vollist);
5049
5050     eval {
5051         run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5052             outfunc => sub {}, errfunc => sub {});
5053     };
5054     # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5055     # timeout should be more than enough here...
5056     PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5057
5058     my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5059                                               : $defaults->{cpuunits};
5060
5061     my $start_timeout = $params->{timeout} // config_aware_timeout($conf, $resume);
5062     my %run_params = (
5063         timeout => $statefile ? undef : $start_timeout,
5064         umask => 0077,
5065         noerr => 1,
5066     );
5067
5068     # when migrating, prefix QEMU output so other side can pick up any
5069     # errors that might occur and show the user
5070     if ($migratedfrom) {
5071         $run_params{quiet} = 1;
5072         $run_params{logfunc} = sub { print "QEMU: $_[0]\n" };
5073     }
5074
5075     my %properties = (
5076         Slice => 'qemu.slice',
5077         KillMode => 'none',
5078         CPUShares => $cpuunits
5079     );
5080
5081     if (my $cpulimit = $conf->{cpulimit}) {
5082         $properties{CPUQuota} = int($cpulimit * 100);
5083     }
5084     $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5085
5086     my $run_qemu = sub {
5087         PVE::Tools::run_fork sub {
5088             PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5089
5090             my $exitcode = run_command($cmd, %run_params);
5091             die "QEMU exited with code $exitcode\n" if $exitcode;
5092         };
5093     };
5094
5095     if ($conf->{hugepages}) {
5096
5097         my $code = sub {
5098             my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5099             my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5100
5101             PVE::QemuServer::Memory::hugepages_mount();
5102             PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5103
5104             eval { $run_qemu->() };
5105             if (my $err = $@) {
5106                 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology)
5107                     if !$conf->{keephugepages};
5108                 die $err;
5109             }
5110
5111             PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology)
5112                 if !$conf->{keephugepages};
5113         };
5114         eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5115
5116     } else {
5117         eval { $run_qemu->() };
5118     }
5119
5120     if (my $err = $@) {
5121         # deactivate volumes if start fails
5122         eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5123         die "start failed: $err";
5124     }
5125
5126     print "migration listens on $migrate_uri\n" if $migrate_uri;
5127     $res->{migrate_uri} = $migrate_uri;
5128
5129     if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5130         eval { mon_cmd($vmid, "cont"); };
5131         warn $@ if $@;
5132     }
5133
5134     #start nbd server for storage migration
5135     if (my $nbd = $migrate_opts->{nbd}) {
5136         my $nbd_protocol_version = $migrate_opts->{nbd_proto_version} // 0;
5137
5138         my $migrate_storage_uri;
5139         # nbd_protocol_version > 0 for unix socket support
5140         if ($nbd_protocol_version > 0 && $migration_type eq 'secure') {
5141             my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5142             mon_cmd($vmid, "nbd-server-start", addr => { type => 'unix', data => { path => $socket_path } } );
5143             $migrate_storage_uri = "nbd:unix:$socket_path";
5144         } else {
5145             my $nodename = nodename();
5146             my $localip = $get_migration_ip->($nodename);
5147             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5148             my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5149
5150             mon_cmd($vmid, "nbd-server-start", addr => {
5151                 type => 'inet',
5152                 data => {
5153                     host => "${localip}",
5154                     port => "${storage_migrate_port}",
5155                 },
5156             });
5157             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5158             $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5159         }
5160
5161         $res->{migrate_storage_uri} = $migrate_storage_uri;
5162
5163         foreach my $opt (sort keys %$nbd) {
5164             my $drivestr = $nbd->{$opt}->{drivestr};
5165             my $volid = $nbd->{$opt}->{volid};
5166             mon_cmd($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5167             my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
5168             print "storage migration listens on $nbd_uri volume:$drivestr\n";
5169             print "re-using replicated volume: $opt - $volid\n"
5170                 if $nbd->{$opt}->{replicated};
5171
5172             $res->{drives}->{$opt} = $nbd->{$opt};
5173             $res->{drives}->{$opt}->{nbd_uri} = $nbd_uri;
5174         }
5175     }
5176
5177     if ($migratedfrom) {
5178         eval {
5179             set_migration_caps($vmid);
5180         };
5181         warn $@ if $@;
5182
5183         if ($spice_port) {
5184             print "spice listens on port $spice_port\n";
5185             $res->{spice_port} = $spice_port;
5186             if ($migrate_opts->{spice_ticket}) {
5187                 mon_cmd($vmid, "set_password", protocol => 'spice', password =>
5188                     $migrate_opts->{spice_ticket});
5189                 mon_cmd($vmid, "expire_password", protocol => 'spice', time => "+30");
5190             }
5191         }
5192
5193     } else {
5194         mon_cmd($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5195             if !$statefile && $conf->{balloon};
5196
5197         foreach my $opt (keys %$conf) {
5198             next if $opt !~  m/^net\d+$/;
5199             my $nicconf = parse_net($conf->{$opt});
5200             qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5201         }
5202     }
5203
5204     mon_cmd($vmid, 'qom-set',
5205                 path => "machine/peripheral/balloon0",
5206                 property => "guest-stats-polling-interval",
5207                 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5208
5209     if ($resume) {
5210         print "Resumed VM, removing state\n";
5211         if (my $vmstate = $conf->{vmstate}) {
5212             PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5213             PVE::Storage::vdisk_free($storecfg, $vmstate);
5214         }
5215         delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
5216         PVE::QemuConfig->write_config($vmid, $conf);
5217     }
5218
5219     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5220
5221     return $res;
5222 }
5223
5224 sub vm_commandline {
5225     my ($storecfg, $vmid, $snapname) = @_;
5226
5227     my $conf = PVE::QemuConfig->load_config($vmid);
5228     my $forcemachine;
5229     my $forcecpu;
5230
5231     if ($snapname) {
5232         my $snapshot = $conf->{snapshots}->{$snapname};
5233         die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5234
5235         # check for machine or CPU overrides in snapshot
5236         $forcemachine = $snapshot->{runningmachine};
5237         $forcecpu = $snapshot->{runningcpu};
5238
5239         $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5240
5241         $conf = $snapshot;
5242     }
5243
5244     my $defaults = load_defaults();
5245
5246     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults,
5247         $forcemachine, $forcecpu);
5248
5249     return PVE::Tools::cmd2string($cmd);
5250 }
5251
5252 sub vm_reset {
5253     my ($vmid, $skiplock) = @_;
5254
5255     PVE::QemuConfig->lock_config($vmid, sub {
5256
5257         my $conf = PVE::QemuConfig->load_config($vmid);
5258
5259         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5260
5261         mon_cmd($vmid, "system_reset");
5262     });
5263 }
5264
5265 sub get_vm_volumes {
5266     my ($conf) = @_;
5267
5268     my $vollist = [];
5269     foreach_volid($conf, sub {
5270         my ($volid, $attr) = @_;
5271
5272         return if $volid =~ m|^/|;
5273
5274         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5275         return if !$sid;
5276
5277         push @$vollist, $volid;
5278     });
5279
5280     return $vollist;
5281 }
5282
5283 sub vm_stop_cleanup {
5284     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5285
5286     eval {
5287
5288         if (!$keepActive) {
5289             my $vollist = get_vm_volumes($conf);
5290             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5291         }
5292
5293         foreach my $ext (qw(mon qmp pid vnc qga)) {
5294             unlink "/var/run/qemu-server/${vmid}.$ext";
5295         }
5296
5297         if ($conf->{ivshmem}) {
5298             my $ivshmem = parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5299             # just delete it for now, VMs which have this already open do not
5300             # are affected, but new VMs will get a separated one. If this
5301             # becomes an issue we either add some sort of ref-counting or just
5302             # add a "don't delete on stop" flag to the ivshmem format.
5303             unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5304         }
5305
5306         foreach my $key (keys %$conf) {
5307             next if $key !~ m/^hostpci(\d+)$/;
5308             my $hostpciindex = $1;
5309             my $d = parse_hostpci($conf->{$key});
5310             my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5311
5312             foreach my $pci (@{$d->{pciid}}) {
5313                 my $pciid = $pci->{id};
5314                 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5315             }
5316         }
5317
5318         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5319     };
5320     warn $@ if $@; # avoid errors - just warn
5321 }
5322
5323 # call only in locked context
5324 sub _do_vm_stop {
5325     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5326
5327     my $pid = check_running($vmid, $nocheck);
5328     return if !$pid;
5329
5330     my $conf;
5331     if (!$nocheck) {
5332         $conf = PVE::QemuConfig->load_config($vmid);
5333         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5334         if (!defined($timeout) && $shutdown && $conf->{startup}) {
5335             my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5336             $timeout = $opts->{down} if $opts->{down};
5337         }
5338         PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5339     }
5340
5341     eval {
5342         if ($shutdown) {
5343             if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5344                 mon_cmd($vmid, "guest-shutdown", timeout => $timeout);
5345             } else {
5346                 mon_cmd($vmid, "system_powerdown");
5347             }
5348         } else {
5349             mon_cmd($vmid, "quit");
5350         }
5351     };
5352     my $err = $@;
5353
5354     if (!$err) {
5355         $timeout = 60 if !defined($timeout);
5356
5357         my $count = 0;
5358         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5359             $count++;
5360             sleep 1;
5361         }
5362
5363         if ($count >= $timeout) {
5364             if ($force) {
5365                 warn "VM still running - terminating now with SIGTERM\n";
5366                 kill 15, $pid;
5367             } else {
5368                 die "VM quit/powerdown failed - got timeout\n";
5369             }
5370         } else {
5371             vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5372             return;
5373         }
5374     } else {
5375        if (!check_running($vmid, $nocheck)) {
5376             warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
5377             return;
5378        }
5379        if ($force) {
5380             warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5381             kill 15, $pid;
5382         } else {
5383             die "VM quit/powerdown failed\n";
5384         }
5385     }
5386
5387     # wait again
5388     $timeout = 10;
5389
5390     my $count = 0;
5391     while (($count < $timeout) && check_running($vmid, $nocheck)) {
5392         $count++;
5393         sleep 1;
5394     }
5395
5396     if ($count >= $timeout) {
5397         warn "VM still running - terminating now with SIGKILL\n";
5398         kill 9, $pid;
5399         sleep 1;
5400     }
5401
5402     vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5403 }
5404
5405 # Note: use $nocheck to skip tests if VM configuration file exists.
5406 # We need that when migration VMs to other nodes (files already moved)
5407 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5408 sub vm_stop {
5409     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5410
5411     $force = 1 if !defined($force) && !$shutdown;
5412
5413     if ($migratedfrom){
5414         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5415         kill 15, $pid if $pid;
5416         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5417         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5418         return;
5419     }
5420
5421     PVE::QemuConfig->lock_config($vmid, sub {
5422         _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5423    });
5424 }
5425
5426 sub vm_reboot {
5427     my ($vmid, $timeout) = @_;
5428
5429     PVE::QemuConfig->lock_config($vmid, sub {
5430         eval {
5431
5432             # only reboot if running, as qmeventd starts it again on a stop event
5433             return if !check_running($vmid);
5434
5435             create_reboot_request($vmid);
5436
5437             my $storecfg = PVE::Storage::config();
5438             _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5439
5440         };
5441         if (my $err = $@) {
5442             # avoid that the next normal shutdown will be confused for a reboot
5443             clear_reboot_request($vmid);
5444             die $err;
5445         }
5446    });
5447 }
5448
5449 # note: if using the statestorage parameter, the caller has to check privileges
5450 sub vm_suspend {
5451     my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5452
5453     my $conf;
5454     my $path;
5455     my $storecfg;
5456     my $vmstate;
5457
5458     PVE::QemuConfig->lock_config($vmid, sub {
5459
5460         $conf = PVE::QemuConfig->load_config($vmid);
5461
5462         my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5463         PVE::QemuConfig->check_lock($conf)
5464             if !($skiplock || $is_backing_up);
5465
5466         die "cannot suspend to disk during backup\n"
5467             if $is_backing_up && $includestate;
5468
5469         if ($includestate) {
5470             $conf->{lock} = 'suspending';
5471             my $date = strftime("%Y-%m-%d", localtime(time()));
5472             $storecfg = PVE::Storage::config();
5473             if (!$statestorage) {
5474                 $statestorage = find_vmstate_storage($conf, $storecfg);
5475                 # check permissions for the storage
5476                 my $rpcenv = PVE::RPCEnvironment::get();
5477                 if ($rpcenv->{type} ne 'cli') {
5478                     my $authuser = $rpcenv->get_user();
5479                     $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
5480                 }
5481             }
5482
5483
5484             $vmstate = PVE::QemuConfig->__snapshot_save_vmstate(
5485                 $vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5486             $path = PVE::Storage::path($storecfg, $vmstate);
5487             PVE::QemuConfig->write_config($vmid, $conf);
5488         } else {
5489             mon_cmd($vmid, "stop");
5490         }
5491     });
5492
5493     if ($includestate) {
5494         # save vm state
5495         PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5496
5497         eval {
5498             mon_cmd($vmid, "savevm-start", statefile => $path);
5499             for(;;) {
5500                 my $state = mon_cmd($vmid, "query-savevm");
5501                 if (!$state->{status}) {
5502                     die "savevm not active\n";
5503                 } elsif ($state->{status} eq 'active') {
5504                     sleep(1);
5505                     next;
5506                 } elsif ($state->{status} eq 'completed') {
5507                     print "State saved, quitting\n";
5508                     last;
5509                 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5510                     die "query-savevm failed with error '$state->{error}'\n"
5511                 } else {
5512                     die "query-savevm returned status '$state->{status}'\n";
5513                 }
5514             }
5515         };
5516         my $err = $@;
5517
5518         PVE::QemuConfig->lock_config($vmid, sub {
5519             $conf = PVE::QemuConfig->load_config($vmid);
5520             if ($err) {
5521                 # cleanup, but leave suspending lock, to indicate something went wrong
5522                 eval {
5523                     mon_cmd($vmid, "savevm-end");
5524                     PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5525                     PVE::Storage::vdisk_free($storecfg, $vmstate);
5526                     delete $conf->@{qw(vmstate runningmachine runningcpu)};
5527                     PVE::QemuConfig->write_config($vmid, $conf);
5528                 };
5529                 warn $@ if $@;
5530                 die $err;
5531             }
5532
5533             die "lock changed unexpectedly\n"
5534                 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5535
5536             mon_cmd($vmid, "quit");
5537             $conf->{lock} = 'suspended';
5538             PVE::QemuConfig->write_config($vmid, $conf);
5539         });
5540     }
5541 }
5542
5543 sub vm_resume {
5544     my ($vmid, $skiplock, $nocheck) = @_;
5545
5546     PVE::QemuConfig->lock_config($vmid, sub {
5547         my $res = mon_cmd($vmid, 'query-status');
5548         my $resume_cmd = 'cont';
5549
5550         if ($res->{status} && $res->{status} eq 'suspended') {
5551             $resume_cmd = 'system_wakeup';
5552         }
5553
5554         if (!$nocheck) {
5555
5556             my $conf = PVE::QemuConfig->load_config($vmid);
5557
5558             PVE::QemuConfig->check_lock($conf)
5559                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5560         }
5561
5562         mon_cmd($vmid, $resume_cmd);
5563     });
5564 }
5565
5566 sub vm_sendkey {
5567     my ($vmid, $skiplock, $key) = @_;
5568
5569     PVE::QemuConfig->lock_config($vmid, sub {
5570
5571         my $conf = PVE::QemuConfig->load_config($vmid);
5572
5573         # there is no qmp command, so we use the human monitor command
5574         my $res = PVE::QemuServer::Monitor::hmp_cmd($vmid, "sendkey $key");
5575         die $res if $res ne '';
5576     });
5577 }
5578
5579 # vzdump restore implementaion
5580
5581 sub tar_archive_read_firstfile {
5582     my $archive = shift;
5583
5584     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5585
5586     # try to detect archive type first
5587     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5588         die "unable to open file '$archive'\n";
5589     my $firstfile = <$fh>;
5590     kill 15, $pid;
5591     close $fh;
5592
5593     die "ERROR: archive contaions no data\n" if !$firstfile;
5594     chomp $firstfile;
5595
5596     return $firstfile;
5597 }
5598
5599 sub tar_restore_cleanup {
5600     my ($storecfg, $statfile) = @_;
5601
5602     print STDERR "starting cleanup\n";
5603
5604     if (my $fd = IO::File->new($statfile, "r")) {
5605         while (defined(my $line = <$fd>)) {
5606             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5607                 my $volid = $2;
5608                 eval {
5609                     if ($volid =~ m|^/|) {
5610                         unlink $volid || die 'unlink failed\n';
5611                     } else {
5612                         PVE::Storage::vdisk_free($storecfg, $volid);
5613                     }
5614                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5615                 };
5616                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5617             } else {
5618                 print STDERR "unable to parse line in statfile - $line";
5619             }
5620         }
5621         $fd->close();
5622     }
5623 }
5624
5625 sub restore_file_archive {
5626     my ($archive, $vmid, $user, $opts) = @_;
5627
5628     return restore_vma_archive($archive, $vmid, $user, $opts)
5629         if $archive eq '-';
5630
5631     my $info = PVE::Storage::archive_info($archive);
5632     my $format = $opts->{format} // $info->{format};
5633     my $comp = $info->{compression};
5634
5635     # try to detect archive format
5636     if ($format eq 'tar') {
5637         return restore_tar_archive($archive, $vmid, $user, $opts);
5638     } else {
5639         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5640     }
5641 }
5642
5643 # hepler to remove disks that will not be used after restore
5644 my $restore_cleanup_oldconf = sub {
5645     my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
5646
5647     PVE::QemuConfig->foreach_volume($oldconf, sub {
5648         my ($ds, $drive) = @_;
5649
5650         return if drive_is_cdrom($drive, 1);
5651
5652         my $volid = $drive->{file};
5653         return if !$volid || $volid =~ m|^/|;
5654
5655         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
5656         return if !$path || !$owner || ($owner != $vmid);
5657
5658         # Note: only delete disk we want to restore
5659         # other volumes will become unused
5660         if ($virtdev_hash->{$ds}) {
5661             eval { PVE::Storage::vdisk_free($storecfg, $volid); };
5662             if (my $err = $@) {
5663                 warn $err;
5664             }
5665         }
5666     });
5667
5668     # delete vmstate files, after the restore we have no snapshots anymore
5669     foreach my $snapname (keys %{$oldconf->{snapshots}}) {
5670         my $snap = $oldconf->{snapshots}->{$snapname};
5671         if ($snap->{vmstate}) {
5672             eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5673             if (my $err = $@) {
5674                 warn $err;
5675             }
5676         }
5677     }
5678 };
5679
5680 # Helper to parse vzdump backup device hints
5681 #
5682 # $rpcenv: Environment, used to ckeck storage permissions
5683 # $user: User ID, to check storage permissions
5684 # $storecfg: Storage configuration
5685 # $fh: the file handle for reading the configuration
5686 # $devinfo: should contain device sizes for all backu-up'ed devices
5687 # $options: backup options (pool, default storage)
5688 #
5689 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
5690 my $parse_backup_hints = sub {
5691     my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
5692
5693     my $virtdev_hash = {};
5694
5695     while (defined(my $line = <$fh>)) {
5696         if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5697             my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5698             die "archive does not contain data for drive '$virtdev'\n"
5699                 if !$devinfo->{$devname};
5700
5701             if (defined($options->{storage})) {
5702                 $storeid = $options->{storage} || 'local';
5703             } elsif (!$storeid) {
5704                 $storeid = 'local';
5705             }
5706             $format = 'raw' if !$format;
5707             $devinfo->{$devname}->{devname} = $devname;
5708             $devinfo->{$devname}->{virtdev} = $virtdev;
5709             $devinfo->{$devname}->{format} = $format;
5710             $devinfo->{$devname}->{storeid} = $storeid;
5711
5712             # check permission on storage
5713             my $pool = $options->{pool}; # todo: do we need that?
5714             if ($user ne 'root@pam') {
5715                 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5716             }
5717
5718             $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5719         } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
5720             my $virtdev = $1;
5721             my $drive = parse_drive($virtdev, $2);
5722             if (drive_is_cloudinit($drive)) {
5723                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5724                 $storeid = $options->{storage} if defined ($options->{storage});
5725                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5726                 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
5727
5728                 $virtdev_hash->{$virtdev} = {
5729                     format => $format,
5730                     storeid => $storeid,
5731                     size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
5732                     is_cloudinit => 1,
5733                 };
5734             }
5735         }
5736     }
5737
5738     return $virtdev_hash;
5739 };
5740
5741 # Helper to allocate and activate all volumes required for a restore
5742 #
5743 # $storecfg: Storage configuration
5744 # $virtdev_hash: as returned by parse_backup_hints()
5745 #
5746 # Returns: { $virtdev => $volid }
5747 my $restore_allocate_devices = sub {
5748     my ($storecfg, $virtdev_hash, $vmid) = @_;
5749
5750     my $map = {};
5751     foreach my $virtdev (sort keys %$virtdev_hash) {
5752         my $d = $virtdev_hash->{$virtdev};
5753         my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5754         my $storeid = $d->{storeid};
5755         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5756
5757         # test if requested format is supported
5758         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5759         my $supported = grep { $_ eq $d->{format} } @$validFormats;
5760         $d->{format} = $defFormat if !$supported;
5761
5762         my $name;
5763         if ($d->{is_cloudinit}) {
5764             $name = "vm-$vmid-cloudinit";
5765             $name .= ".$d->{format}" if $d->{format} ne 'raw';
5766         }
5767
5768         my $volid = PVE::Storage::vdisk_alloc(
5769             $storecfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
5770
5771         print STDERR "new volume ID is '$volid'\n";
5772         $d->{volid} = $volid;
5773
5774         PVE::Storage::activate_volumes($storecfg, [$volid]);
5775
5776         $map->{$virtdev} = $volid;
5777     }
5778
5779     return $map;
5780 };
5781
5782 my $restore_update_config_line = sub {
5783     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5784
5785     return if $line =~ m/^\#qmdump\#/;
5786     return if $line =~ m/^\#vzdump\#/;
5787     return if $line =~ m/^lock:/;
5788     return if $line =~ m/^unused\d+:/;
5789     return if $line =~ m/^parent:/;
5790
5791     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5792     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5793         # try to convert old 1.X settings
5794         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5795         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5796             my ($model, $macaddr) = split(/\=/, $devconfig);
5797             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5798             my $net = {
5799                 model => $model,
5800                 bridge => "vmbr$ind",
5801                 macaddr => $macaddr,
5802             };
5803             my $netstr = print_net($net);
5804
5805             print $outfd "net$cookie->{netcount}: $netstr\n";
5806             $cookie->{netcount}++;
5807         }
5808     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5809         my ($id, $netstr) = ($1, $2);
5810         my $net = parse_net($netstr);
5811         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5812         $netstr = print_net($net);
5813         print $outfd "$id: $netstr\n";
5814     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5815         my $virtdev = $1;
5816         my $value = $3;
5817         my $di = parse_drive($virtdev, $value);
5818         if (defined($di->{backup}) && !$di->{backup}) {
5819             print $outfd "#$line";
5820         } elsif ($map->{$virtdev}) {
5821             delete $di->{format}; # format can change on restore
5822             $di->{file} = $map->{$virtdev};
5823             $value = print_drive($di);
5824             print $outfd "$virtdev: $value\n";
5825         } else {
5826             print $outfd $line;
5827         }
5828     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5829         my $vmgenid = $1;
5830         if ($vmgenid ne '0') {
5831             # always generate a new vmgenid if there was a valid one setup
5832             $vmgenid = generate_uuid();
5833         }
5834         print $outfd "vmgenid: $vmgenid\n";
5835     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5836         my ($uuid, $uuid_str);
5837         UUID::generate($uuid);
5838         UUID::unparse($uuid, $uuid_str);
5839         my $smbios1 = parse_smbios1($2);
5840         $smbios1->{uuid} = $uuid_str;
5841         print $outfd $1.print_smbios1($smbios1)."\n";
5842     } else {
5843         print $outfd $line;
5844     }
5845 };
5846
5847 my $restore_deactivate_volumes = sub {
5848     my ($storecfg, $devinfo) = @_;
5849
5850     my $vollist = [];
5851     foreach my $devname (keys %$devinfo) {
5852         my $volid = $devinfo->{$devname}->{volid};
5853         push @$vollist, $volid if $volid;
5854     }
5855
5856     PVE::Storage::deactivate_volumes($storecfg, $vollist);
5857 };
5858
5859 my $restore_destroy_volumes = sub {
5860     my ($storecfg, $devinfo) = @_;
5861
5862     foreach my $devname (keys %$devinfo) {
5863         my $volid = $devinfo->{$devname}->{volid};
5864         next if !$volid;
5865         eval {
5866             if ($volid =~ m|^/|) {
5867                 unlink $volid || die 'unlink failed\n';
5868             } else {
5869                 PVE::Storage::vdisk_free($storecfg, $volid);
5870             }
5871             print STDERR "temporary volume '$volid' sucessfuly removed\n";
5872         };
5873         print STDERR "unable to cleanup '$volid' - $@" if $@;
5874     }
5875 };
5876
5877 sub scan_volids {
5878     my ($cfg, $vmid) = @_;
5879
5880     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5881
5882     my $volid_hash = {};
5883     foreach my $storeid (keys %$info) {
5884         foreach my $item (@{$info->{$storeid}}) {
5885             next if !($item->{volid} && $item->{size});
5886             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5887             $volid_hash->{$item->{volid}} = $item;
5888         }
5889     }
5890
5891     return $volid_hash;
5892 }
5893
5894 sub update_disk_config {
5895     my ($vmid, $conf, $volid_hash) = @_;
5896
5897     my $changes;
5898     my $prefix = "VM $vmid";
5899
5900     # used and unused disks
5901     my $referenced = {};
5902
5903     # Note: it is allowed to define multiple storages with same path (alias), so
5904     # we need to check both 'volid' and real 'path' (two different volid can point
5905     # to the same path).
5906
5907     my $referencedpath = {};
5908
5909     # update size info
5910     PVE::QemuConfig->foreach_volume($conf, sub {
5911         my ($opt, $drive) = @_;
5912
5913         my $volid = $drive->{file};
5914         return if !$volid;
5915         my $volume = $volid_hash->{$volid};
5916
5917         # mark volid as "in-use" for next step
5918         $referenced->{$volid} = 1;
5919         if ($volume && (my $path = $volume->{path})) {
5920             $referencedpath->{$path} = 1;
5921         }
5922
5923         return if drive_is_cdrom($drive);
5924         return if !$volume;
5925
5926         my ($updated, $msg) = PVE::QemuServer::Drive::update_disksize($drive, $volume->{size});
5927         if (defined($updated)) {
5928             $changes = 1;
5929             $conf->{$opt} = print_drive($updated);
5930             print "$prefix ($opt): $msg\n";
5931         }
5932     });
5933
5934     # remove 'unusedX' entry if volume is used
5935     PVE::QemuConfig->foreach_unused_volume($conf, sub {
5936         my ($opt, $drive) = @_;
5937
5938         my $volid = $drive->{file};
5939         return if !$volid;
5940
5941         my $path;
5942         $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5943         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5944             print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
5945             $changes = 1;
5946             delete $conf->{$opt};
5947         }
5948
5949         $referenced->{$volid} = 1;
5950         $referencedpath->{$path} = 1 if $path;
5951     });
5952
5953     foreach my $volid (sort keys %$volid_hash) {
5954         next if $volid =~ m/vm-$vmid-state-/;
5955         next if $referenced->{$volid};
5956         my $path = $volid_hash->{$volid}->{path};
5957         next if !$path; # just to be sure
5958         next if $referencedpath->{$path};
5959         $changes = 1;
5960         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5961         print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
5962         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5963     }
5964
5965     return $changes;
5966 }
5967
5968 sub rescan {
5969     my ($vmid, $nolock, $dryrun) = @_;
5970
5971     my $cfg = PVE::Storage::config();
5972
5973     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5974     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
5975     foreach my $stor (keys %{$cfg->{ids}}) {
5976         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
5977     }
5978
5979     print "rescan volumes...\n";
5980     my $volid_hash = scan_volids($cfg, $vmid);
5981
5982     my $updatefn =  sub {
5983         my ($vmid) = @_;
5984
5985         my $conf = PVE::QemuConfig->load_config($vmid);
5986
5987         PVE::QemuConfig->check_lock($conf);
5988
5989         my $vm_volids = {};
5990         foreach my $volid (keys %$volid_hash) {
5991             my $info = $volid_hash->{$volid};
5992             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5993         }
5994
5995         my $changes = update_disk_config($vmid, $conf, $vm_volids);
5996
5997         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
5998     };
5999
6000     if (defined($vmid)) {
6001         if ($nolock) {
6002             &$updatefn($vmid);
6003         } else {
6004             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6005         }
6006     } else {
6007         my $vmlist = config_list();
6008         foreach my $vmid (keys %$vmlist) {
6009             if ($nolock) {
6010                 &$updatefn($vmid);
6011             } else {
6012                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6013             }
6014         }
6015     }
6016 }
6017
6018 sub restore_proxmox_backup_archive {
6019     my ($archive, $vmid, $user, $options) = @_;
6020
6021     my $storecfg = PVE::Storage::config();
6022
6023     my ($storeid, $volname) = PVE::Storage::parse_volume_id($archive);
6024     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6025
6026     my $server = $scfg->{server};
6027     my $datastore = $scfg->{datastore};
6028     my $username = $scfg->{username} // 'root@pam';
6029     my $fingerprint = $scfg->{fingerprint};
6030     my $keyfile = PVE::Storage::PBSPlugin::pbs_encryption_key_file_name($storecfg, $storeid);
6031
6032     my $repo = "$username\@$server:$datastore";
6033
6034     # This is only used for `pbs-restore`!
6035     my $password = PVE::Storage::PBSPlugin::pbs_get_password($scfg, $storeid);
6036     local $ENV{PBS_PASSWORD} = $password;
6037     local $ENV{PBS_FINGERPRINT} = $fingerprint if defined($fingerprint);
6038
6039     my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6040         PVE::Storage::parse_volname($storecfg, $archive);
6041
6042     die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6043
6044     die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6045
6046     my $tmpdir = "/var/tmp/vzdumptmp$$";
6047     rmtree $tmpdir;
6048     mkpath $tmpdir;
6049
6050     my $conffile = PVE::QemuConfig->config_file($vmid);
6051     my $tmpfn = "$conffile.$$.tmp";
6052      # disable interrupts (always do cleanups)
6053     local $SIG{INT} =
6054         local $SIG{TERM} =
6055         local $SIG{QUIT} =
6056         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6057
6058     # Note: $oldconf is undef if VM does not exists
6059     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6060     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6061
6062     my $rpcenv = PVE::RPCEnvironment::get();
6063     my $devinfo = {};
6064
6065     eval {
6066         # enable interrupts
6067         local $SIG{INT} =
6068             local $SIG{TERM} =
6069             local $SIG{QUIT} =
6070             local $SIG{HUP} =
6071             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6072
6073         my $cfgfn = "$tmpdir/qemu-server.conf";
6074         my $firewall_config_fn = "$tmpdir/fw.conf";
6075         my $index_fn = "$tmpdir/index.json";
6076
6077         my $cmd = "restore";
6078
6079         my $param = [$pbs_backup_name, "index.json", $index_fn];
6080         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6081         my $index = PVE::Tools::file_get_contents($index_fn);
6082         $index = decode_json($index);
6083
6084         # print Dumper($index);
6085         foreach my $info (@{$index->{files}}) {
6086             if ($info->{filename} =~ m/^(drive-\S+).img.fidx$/) {
6087                 my $devname = $1;
6088                 if ($info->{size} =~ m/^(\d+)$/) { # untaint size
6089                     $devinfo->{$devname}->{size} = $1;
6090                 } else {
6091                     die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
6092                 }
6093             }
6094         }
6095
6096         my $is_qemu_server_backup = scalar(
6097             grep { $_->{filename} eq 'qemu-server.conf.blob' } @{$index->{files}}
6098         );
6099         if (!$is_qemu_server_backup) {
6100             die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
6101         }
6102         my $has_firewall_config = scalar(grep { $_->{filename} eq 'fw.conf.blob' } @{$index->{files}});
6103
6104         $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
6105         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6106
6107         if ($has_firewall_config) {
6108             $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
6109             PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6110
6111             my $pve_firewall_dir = '/etc/pve/firewall';
6112             mkdir $pve_firewall_dir; # make sure the dir exists
6113             PVE::Tools::file_copy($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
6114         }
6115
6116         my $fh = IO::File->new($cfgfn, "r") ||
6117             die "unable to read qemu-server.conf - $!\n";
6118
6119         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
6120
6121         # fixme: rate limit?
6122
6123         # create empty/temp config
6124         PVE::Tools::file_set_contents($conffile, "memory: 128\nlock: create");
6125
6126         $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
6127
6128         # allocate volumes
6129         my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
6130
6131         foreach my $virtdev (sort keys %$virtdev_hash) {
6132             my $d = $virtdev_hash->{$virtdev};
6133             next if $d->{is_cloudinit}; # no need to restore cloudinit
6134
6135             my $volid = $d->{volid};
6136
6137             my $path = PVE::Storage::path($storecfg, $volid);
6138
6139             # This is the ONLY user of the PBS_ env vars set on top of this function!
6140             my $pbs_restore_cmd = [
6141                 '/usr/bin/pbs-restore',
6142                 '--repository', $repo,
6143                 $pbs_backup_name,
6144                 "$d->{devname}.img.fidx",
6145                 $path,
6146                 '--verbose',
6147                 ];
6148
6149             push @$pbs_restore_cmd, '--format', $d->{format} if $d->{format};
6150             push @$pbs_restore_cmd, '--keyfile', $keyfile if -e $keyfile;
6151
6152             if (PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $volid)) {
6153                 push @$pbs_restore_cmd, '--skip-zero';
6154             }
6155
6156             my $dbg_cmdstring = PVE::Tools::cmd2string($pbs_restore_cmd);
6157             print "restore proxmox backup image: $dbg_cmdstring\n";
6158             run_command($pbs_restore_cmd);
6159         }
6160
6161         $fh->seek(0, 0) || die "seek failed - $!\n";
6162
6163         my $outfd = IO::File->new($tmpfn, "w") || die "unable to write config for VM $vmid\n";
6164
6165         my $cookie = { netcount => 0 };
6166         while (defined(my $line = <$fh>)) {
6167             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $options->{unique});
6168         }
6169
6170         $fh->close();
6171         $outfd->close();
6172     };
6173     my $err = $@;
6174
6175     $restore_deactivate_volumes->($storecfg, $devinfo);
6176
6177     rmtree $tmpdir;
6178
6179     if ($err) {
6180         unlink $tmpfn;
6181         $restore_destroy_volumes->($storecfg, $devinfo);
6182         die $err;
6183     }
6184
6185     rename($tmpfn, $conffile) ||
6186         die "unable to commit configuration file '$conffile'\n";
6187
6188     PVE::Cluster::cfs_update(); # make sure we read new file
6189
6190     eval { rescan($vmid, 1); };
6191     warn $@ if $@;
6192 }
6193
6194 sub restore_vma_archive {
6195     my ($archive, $vmid, $user, $opts, $comp) = @_;
6196
6197     my $readfrom = $archive;
6198
6199     my $cfg = PVE::Storage::config();
6200     my $commands = [];
6201     my $bwlimit = $opts->{bwlimit};
6202
6203     my $dbg_cmdstring = '';
6204     my $add_pipe = sub {
6205         my ($cmd) = @_;
6206         push @$commands, $cmd;
6207         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6208         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6209         $readfrom = '-';
6210     };
6211
6212     my $input = undef;
6213     if ($archive eq '-') {
6214         $input = '<&STDIN';
6215     } else {
6216         # If we use a backup from a PVE defined storage we also consider that
6217         # storage's rate limit:
6218         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6219         if (defined($volid)) {
6220             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6221             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6222             if ($readlimit) {
6223                 print STDERR "applying read rate limit: $readlimit\n";
6224                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6225                 $add_pipe->($cstream);
6226             }
6227         }
6228     }
6229
6230     if ($comp) {
6231         my $info = PVE::Storage::decompressor_info('vma', $comp);
6232         my $cmd = $info->{decompressor};
6233         push @$cmd, $readfrom;
6234         $add_pipe->($cmd);
6235     }
6236
6237     my $tmpdir = "/var/tmp/vzdumptmp$$";
6238     rmtree $tmpdir;
6239
6240     # disable interrupts (always do cleanups)
6241     local $SIG{INT} =
6242         local $SIG{TERM} =
6243         local $SIG{QUIT} =
6244         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6245
6246     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6247     POSIX::mkfifo($mapfifo, 0600);
6248     my $fifofh;
6249     my $openfifo = sub { open($fifofh, '>', $mapfifo) or die $! };
6250
6251     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6252
6253     my $oldtimeout;
6254     my $timeout = 5;
6255
6256     my $devinfo = {};
6257
6258     my $rpcenv = PVE::RPCEnvironment::get();
6259
6260     my $conffile = PVE::QemuConfig->config_file($vmid);
6261     my $tmpfn = "$conffile.$$.tmp";
6262
6263     # Note: $oldconf is undef if VM does not exist
6264     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6265     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6266
6267     my %storage_limits;
6268
6269     my $print_devmap = sub {
6270         my $cfgfn = "$tmpdir/qemu-server.conf";
6271
6272         # we can read the config - that is already extracted
6273         my $fh = IO::File->new($cfgfn, "r") ||
6274             die "unable to read qemu-server.conf - $!\n";
6275
6276         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6277         if (-f $fwcfgfn) {
6278             my $pve_firewall_dir = '/etc/pve/firewall';
6279             mkdir $pve_firewall_dir; # make sure the dir exists
6280             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6281         }
6282
6283         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
6284
6285         foreach my $key (keys %storage_limits) {
6286             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6287             next if !$limit;
6288             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6289             $storage_limits{$key} = $limit * 1024;
6290         }
6291
6292         foreach my $devname (keys %$devinfo) {
6293             die "found no device mapping information for device '$devname'\n"
6294                 if !$devinfo->{$devname}->{virtdev};
6295         }
6296
6297         # create empty/temp config
6298         if ($oldconf) {
6299             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6300             $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
6301         }
6302
6303         # allocate volumes
6304         my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
6305
6306         # print restore information to $fifofh
6307         foreach my $virtdev (sort keys %$virtdev_hash) {
6308             my $d = $virtdev_hash->{$virtdev};
6309             next if $d->{is_cloudinit}; # no need to restore cloudinit
6310
6311             my $storeid = $d->{storeid};
6312             my $volid = $d->{volid};
6313
6314             my $map_opts = '';
6315             if (my $limit = $storage_limits{$storeid}) {
6316                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6317             }
6318
6319             my $write_zeros = 1;
6320             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6321                 $write_zeros = 0;
6322             }
6323
6324             my $path = PVE::Storage::path($cfg, $volid);
6325
6326             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6327
6328             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6329         }
6330
6331         $fh->seek(0, 0) || die "seek failed - $!\n";
6332
6333         my $outfd = IO::File->new($tmpfn, "w") || die "unable to write config for VM $vmid\n";
6334
6335         my $cookie = { netcount => 0 };
6336         while (defined(my $line = <$fh>)) {
6337             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6338         }
6339
6340         $fh->close();
6341         $outfd->close();
6342     };
6343
6344     eval {
6345         # enable interrupts
6346         local $SIG{INT} =
6347             local $SIG{TERM} =
6348             local $SIG{QUIT} =
6349             local $SIG{HUP} =
6350             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6351         local $SIG{ALRM} = sub { die "got timeout\n"; };
6352
6353         $oldtimeout = alarm($timeout);
6354
6355         my $parser = sub {
6356             my $line = shift;
6357
6358             print "$line\n";
6359
6360             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6361                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6362                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6363             } elsif ($line =~ m/^CTIME: /) {
6364                 # we correctly received the vma config, so we can disable
6365                 # the timeout now for disk allocation (set to 10 minutes, so
6366                 # that we always timeout if something goes wrong)
6367                 alarm(600);
6368                 &$print_devmap();
6369                 print $fifofh "done\n";
6370                 my $tmp = $oldtimeout || 0;
6371                 $oldtimeout = undef;
6372                 alarm($tmp);
6373                 close($fifofh);
6374                 $fifofh = undef;
6375             }
6376         };
6377
6378         print "restore vma archive: $dbg_cmdstring\n";
6379         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6380     };
6381     my $err = $@;
6382
6383     alarm($oldtimeout) if $oldtimeout;
6384
6385     $restore_deactivate_volumes->($cfg, $devinfo);
6386
6387     close($fifofh) if $fifofh;
6388     unlink $mapfifo;
6389     rmtree $tmpdir;
6390
6391     if ($err) {
6392         unlink $tmpfn;
6393         $restore_destroy_volumes->($cfg, $devinfo);
6394         die $err;
6395     }
6396
6397     rename($tmpfn, $conffile) ||
6398         die "unable to commit configuration file '$conffile'\n";
6399
6400     PVE::Cluster::cfs_update(); # make sure we read new file
6401
6402     eval { rescan($vmid, 1); };
6403     warn $@ if $@;
6404 }
6405
6406 sub restore_tar_archive {
6407     my ($archive, $vmid, $user, $opts) = @_;
6408
6409     if ($archive ne '-') {
6410         my $firstfile = tar_archive_read_firstfile($archive);
6411         die "ERROR: file '$archive' does not look like a QemuServer vzdump backup\n"
6412             if $firstfile ne 'qemu-server.conf';
6413     }
6414
6415     my $storecfg = PVE::Storage::config();
6416
6417     # avoid zombie disks when restoring over an existing VM -> cleanup first
6418     # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
6419     # skiplock=1 because qmrestore has set the 'create' lock itself already
6420     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6421     destroy_vm($storecfg, $vmid, 1, { lock => 'restore' }) if -f $vmcfgfn;
6422
6423     my $tocmd = "/usr/lib/qemu-server/qmextract";
6424
6425     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6426     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6427     $tocmd .= ' --prealloc' if $opts->{prealloc};
6428     $tocmd .= ' --info' if $opts->{info};
6429
6430     # tar option "xf" does not autodetect compression when read from STDIN,
6431     # so we pipe to zcat
6432     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6433         PVE::Tools::shellquote("--to-command=$tocmd");
6434
6435     my $tmpdir = "/var/tmp/vzdumptmp$$";
6436     mkpath $tmpdir;
6437
6438     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6439     local $ENV{VZDUMP_VMID} = $vmid;
6440     local $ENV{VZDUMP_USER} = $user;
6441
6442     my $conffile = PVE::QemuConfig->config_file($vmid);
6443     my $tmpfn = "$conffile.$$.tmp";
6444
6445     # disable interrupts (always do cleanups)
6446     local $SIG{INT} =
6447         local $SIG{TERM} =
6448         local $SIG{QUIT} =
6449         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6450
6451     eval {
6452         # enable interrupts
6453         local $SIG{INT} =
6454             local $SIG{TERM} =
6455             local $SIG{QUIT} =
6456             local $SIG{HUP} =
6457             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6458
6459         if ($archive eq '-') {
6460             print "extracting archive from STDIN\n";
6461             run_command($cmd, input => "<&STDIN");
6462         } else {
6463             print "extracting archive '$archive'\n";
6464             run_command($cmd);
6465         }
6466
6467         return if $opts->{info};
6468
6469         # read new mapping
6470         my $map = {};
6471         my $statfile = "$tmpdir/qmrestore.stat";
6472         if (my $fd = IO::File->new($statfile, "r")) {
6473             while (defined (my $line = <$fd>)) {
6474                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6475                     $map->{$1} = $2 if $1;
6476                 } else {
6477                     print STDERR "unable to parse line in statfile - $line\n";
6478                 }
6479             }
6480             $fd->close();
6481         }
6482
6483         my $confsrc = "$tmpdir/qemu-server.conf";
6484
6485         my $srcfd = IO::File->new($confsrc, "r") || die "unable to open file '$confsrc'\n";
6486
6487         my $outfd = IO::File->new($tmpfn, "w") || die "unable to write config for VM $vmid\n";
6488
6489         my $cookie = { netcount => 0 };
6490         while (defined (my $line = <$srcfd>)) {
6491             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6492         }
6493
6494         $srcfd->close();
6495         $outfd->close();
6496     };
6497     if (my $err = $@) {
6498         unlink $tmpfn;
6499         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6500         die $err;
6501     }
6502
6503     rmtree $tmpdir;
6504
6505     rename $tmpfn, $conffile ||
6506         die "unable to commit configuration file '$conffile'\n";
6507
6508     PVE::Cluster::cfs_update(); # make sure we read new file
6509
6510     eval { rescan($vmid, 1); };
6511     warn $@ if $@;
6512 };
6513
6514 sub foreach_storage_used_by_vm {
6515     my ($conf, $func) = @_;
6516
6517     my $sidhash = {};
6518
6519     PVE::QemuConfig->foreach_volume($conf, sub {
6520         my ($ds, $drive) = @_;
6521         return if drive_is_cdrom($drive);
6522
6523         my $volid = $drive->{file};
6524
6525         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6526         $sidhash->{$sid} = $sid if $sid;
6527     });
6528
6529     foreach my $sid (sort keys %$sidhash) {
6530         &$func($sid);
6531     }
6532 }
6533
6534 my $qemu_snap_storage = {
6535     rbd => 1,
6536 };
6537 sub do_snapshots_with_qemu {
6538     my ($storecfg, $volid) = @_;
6539
6540     my $storage_name = PVE::Storage::parse_volume_id($volid);
6541     my $scfg = $storecfg->{ids}->{$storage_name};
6542     die "could not find storage '$storage_name'\n" if !defined($scfg);
6543
6544     if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6545         return 1;
6546     }
6547
6548     if ($volid =~ m/\.(qcow2|qed)$/){
6549         return 1;
6550     }
6551
6552     return;
6553 }
6554
6555 sub qga_check_running {
6556     my ($vmid, $nowarn) = @_;
6557
6558     eval { mon_cmd($vmid, "guest-ping", timeout => 3); };
6559     if ($@) {
6560         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6561         return 0;
6562     }
6563     return 1;
6564 }
6565
6566 sub template_create {
6567     my ($vmid, $conf, $disk) = @_;
6568
6569     my $storecfg = PVE::Storage::config();
6570
6571     PVE::QemuConfig->foreach_volume($conf, sub {
6572         my ($ds, $drive) = @_;
6573
6574         return if drive_is_cdrom($drive);
6575         return if $disk && $ds ne $disk;
6576
6577         my $volid = $drive->{file};
6578         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6579
6580         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6581         $drive->{file} = $voliddst;
6582         $conf->{$ds} = print_drive($drive);
6583         PVE::QemuConfig->write_config($vmid, $conf);
6584     });
6585 }
6586
6587 sub convert_iscsi_path {
6588     my ($path) = @_;
6589
6590     if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6591         my $portal = $1;
6592         my $target = $2;
6593         my $lun = $3;
6594
6595         my $initiator_name = get_initiator_name();
6596
6597         return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6598                "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6599     }
6600
6601     die "cannot convert iscsi path '$path', unkown format\n";
6602 }
6603
6604 sub qemu_img_convert {
6605     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6606
6607     my $storecfg = PVE::Storage::config();
6608     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6609     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6610
6611     die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6612
6613     my $cachemode;
6614     my $src_path;
6615     my $src_is_iscsi = 0;
6616     my $src_format;
6617
6618     if ($src_storeid) {
6619         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6620         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6621         $src_format = qemu_img_format($src_scfg, $src_volname);
6622         $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6623         $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6624         $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6625     } elsif (-f $src_volid) {
6626         $src_path = $src_volid;
6627         if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6628             $src_format = $1;
6629         }
6630     }
6631
6632     die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6633
6634     my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6635     my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6636     my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6637     my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6638
6639     my $cmd = [];
6640     push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6641     push @$cmd, '-l', "snapshot.name=$snapname"
6642         if $snapname && $src_format && $src_format eq "qcow2";
6643     push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6644     push @$cmd, '-T', $cachemode if defined($cachemode);
6645
6646     if ($src_is_iscsi) {
6647         push @$cmd, '--image-opts';
6648         $src_path = convert_iscsi_path($src_path);
6649     } elsif ($src_format) {
6650         push @$cmd, '-f', $src_format;
6651     }
6652
6653     if ($dst_is_iscsi) {
6654         push @$cmd, '--target-image-opts';
6655         $dst_path = convert_iscsi_path($dst_path);
6656     } else {
6657         push @$cmd, '-O', $dst_format;
6658     }
6659
6660     push @$cmd, $src_path;
6661
6662     if (!$dst_is_iscsi && $is_zero_initialized) {
6663         push @$cmd, "zeroinit:$dst_path";
6664     } else {
6665         push @$cmd, $dst_path;
6666     }
6667
6668     my $parser = sub {
6669         my $line = shift;
6670         if($line =~ m/\((\S+)\/100\%\)/){
6671             my $percent = $1;
6672             my $transferred = int($size * $percent / 100);
6673             my $remaining = $size - $transferred;
6674
6675             print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6676         }
6677
6678     };
6679
6680     eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6681     my $err = $@;
6682     die "copy failed: $err" if $err;
6683 }
6684
6685 sub qemu_img_format {
6686     my ($scfg, $volname) = @_;
6687
6688     if ($scfg->{path} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6689         return $1;
6690     } else {
6691         return "raw";
6692     }
6693 }
6694
6695 sub qemu_drive_mirror {
6696     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
6697
6698     $jobs = {} if !$jobs;
6699
6700     my $qemu_target;
6701     my $format;
6702     $jobs->{"drive-$drive"} = {};
6703
6704     if ($dst_volid =~ /^nbd:/) {
6705         $qemu_target = $dst_volid;
6706         $format = "nbd";
6707     } else {
6708         my $storecfg = PVE::Storage::config();
6709         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6710
6711         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6712
6713         $format = qemu_img_format($dst_scfg, $dst_volname);
6714
6715         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6716
6717         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6718     }
6719
6720     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6721     $opts->{format} = $format if $format;
6722
6723     if (defined($src_bitmap)) {
6724         $opts->{sync} = 'incremental';
6725         $opts->{bitmap} = $src_bitmap;
6726         print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
6727     }
6728
6729     if (defined($bwlimit)) {
6730         $opts->{speed} = $bwlimit * 1024;
6731         print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
6732     } else {
6733         print "drive mirror is starting for drive-$drive\n";
6734     }
6735
6736     # if a job already runs for this device we get an error, catch it for cleanup
6737     eval { mon_cmd($vmid, "drive-mirror", %$opts); };
6738     if (my $err = $@) {
6739         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6740         warn "$@\n" if $@;
6741         die "mirroring error: $err\n";
6742     }
6743
6744     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $completion, $qga);
6745 }
6746
6747 # $completion can be either
6748 # 'complete': wait until all jobs are ready, block-job-complete them (default)
6749 # 'cancel': wait until all jobs are ready, block-job-cancel them
6750 # 'skip': wait until all jobs are ready, return with block jobs in ready state
6751 sub qemu_drive_mirror_monitor {
6752     my ($vmid, $vmiddst, $jobs, $completion, $qga) = @_;
6753
6754     $completion //= 'complete';
6755
6756     eval {
6757         my $err_complete = 0;
6758
6759         while (1) {
6760             die "storage migration timed out\n" if $err_complete > 300;
6761
6762             my $stats = mon_cmd($vmid, "query-block-jobs");
6763
6764             my $running_mirror_jobs = {};
6765             foreach my $stat (@$stats) {
6766                 next if $stat->{type} ne 'mirror';
6767                 $running_mirror_jobs->{$stat->{device}} = $stat;
6768             }
6769
6770             my $readycounter = 0;
6771
6772             foreach my $job (keys %$jobs) {
6773
6774                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6775                     print "$job : finished\n";
6776                     delete $jobs->{$job};
6777                     next;
6778                 }
6779
6780                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6781
6782                 my $busy = $running_mirror_jobs->{$job}->{busy};
6783                 my $ready = $running_mirror_jobs->{$job}->{ready};
6784                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6785                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6786                     my $remaining = $total - $transferred;
6787                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6788
6789                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6790                 }
6791
6792                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6793             }
6794
6795             last if scalar(keys %$jobs) == 0;
6796
6797             if ($readycounter == scalar(keys %$jobs)) {
6798                 print "all mirroring jobs are ready \n";
6799                 last if $completion eq 'skip'; #do the complete later
6800
6801                 if ($vmiddst && $vmiddst != $vmid) {
6802                     my $agent_running = $qga && qga_check_running($vmid);
6803                     if ($agent_running) {
6804                         print "freeze filesystem\n";
6805                         eval { mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6806                     } else {
6807                         print "suspend vm\n";
6808                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6809                     }
6810
6811                     # if we clone a disk for a new target vm, we don't switch the disk
6812                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6813
6814                     if ($agent_running) {
6815                         print "unfreeze filesystem\n";
6816                         eval { mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6817                     } else {
6818                         print "resume vm\n";
6819                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6820                     }
6821
6822                     last;
6823                 } else {
6824
6825                     foreach my $job (keys %$jobs) {
6826                         # try to switch the disk if source and destination are on the same guest
6827                         print "$job: Completing block job...\n";
6828
6829                         my $op;
6830                         if ($completion eq 'complete') {
6831                             $op = 'block-job-complete';
6832                         } elsif ($completion eq 'cancel') {
6833                             $op = 'block-job-cancel';
6834                         } else {
6835                             die "invalid completion value: $completion\n";
6836                         }
6837                         eval { mon_cmd($vmid, $op, device => $job) };
6838                         if ($@ =~ m/cannot be completed/) {
6839                             print "$job: Block job cannot be completed, try again.\n";
6840                             $err_complete++;
6841                         }else {
6842                             print "$job: Completed successfully.\n";
6843                             $jobs->{$job}->{complete} = 1;
6844                         }
6845                     }
6846                 }
6847             }
6848             sleep 1;
6849         }
6850     };
6851     my $err = $@;
6852
6853     if ($err) {
6854         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6855         die "mirroring error: $err";
6856     }
6857
6858 }
6859
6860 sub qemu_blockjobs_cancel {
6861     my ($vmid, $jobs) = @_;
6862
6863     foreach my $job (keys %$jobs) {
6864         print "$job: Cancelling block job\n";
6865         eval { mon_cmd($vmid, "block-job-cancel", device => $job); };
6866         $jobs->{$job}->{cancel} = 1;
6867     }
6868
6869     while (1) {
6870         my $stats = mon_cmd($vmid, "query-block-jobs");
6871
6872         my $running_jobs = {};
6873         foreach my $stat (@$stats) {
6874             $running_jobs->{$stat->{device}} = $stat;
6875         }
6876
6877         foreach my $job (keys %$jobs) {
6878
6879             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6880                 print "$job: Done.\n";
6881                 delete $jobs->{$job};
6882             }
6883         }
6884
6885         last if scalar(keys %$jobs) == 0;
6886
6887         sleep 1;
6888     }
6889 }
6890
6891 sub clone_disk {
6892     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6893         $newvmid, $storage, $format, $full, $newvollist, $jobs, $completion, $qga, $bwlimit, $conf) = @_;
6894
6895     my $newvolid;
6896
6897     if (!$full) {
6898         print "create linked clone of drive $drivename ($drive->{file})\n";
6899         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6900         push @$newvollist, $newvolid;
6901     } else {
6902
6903         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6904         $storeid = $storage if $storage;
6905
6906         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6907
6908         print "create full clone of drive $drivename ($drive->{file})\n";
6909         my $name = undef;
6910         my $size = undef;
6911         if (drive_is_cloudinit($drive)) {
6912             $name = "vm-$newvmid-cloudinit";
6913             $name .= ".$dst_format" if $dst_format ne 'raw';
6914             $snapname = undef;
6915             $size = PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE;
6916         } elsif ($drivename eq 'efidisk0') {
6917             $size = get_efivars_size($conf);
6918         } else {
6919             ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6920         }
6921         $size /= 1024;
6922         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, $size);
6923         push @$newvollist, $newvolid;
6924
6925         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6926
6927         if (drive_is_cloudinit($drive)) {
6928             goto no_data_clone;
6929         }
6930
6931         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6932         if (!$running || $snapname) {
6933             # TODO: handle bwlimits
6934             if ($drivename eq 'efidisk0') {
6935                 # the relevant data on the efidisk may be smaller than the source
6936                 # e.g. on RBD/ZFS, so we use dd to copy only the amount
6937                 # that is given by the OVMF_VARS.fd
6938                 my $src_path = PVE::Storage::path($storecfg, $drive->{file});
6939                 my $dst_path = PVE::Storage::path($storecfg, $newvolid);
6940                 run_command(['qemu-img', 'dd', '-n', '-O', $dst_format, "bs=1", "count=$size",
6941                     "if=$src_path", "of=$dst_path"]);
6942             } else {
6943                 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6944             }
6945         } else {
6946
6947             my $kvmver = get_running_qemu_version ($vmid);
6948             if (!min_version($kvmver, 2, 7)) {
6949                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6950                     if $drive->{iothread};
6951             }
6952
6953             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs,
6954                 $completion, $qga, $bwlimit);
6955         }
6956     }
6957
6958 no_data_clone:
6959     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6960
6961     my $disk = $drive;
6962     $disk->{format} = undef;
6963     $disk->{file} = $newvolid;
6964     $disk->{size} = $size;
6965
6966     return $disk;
6967 }
6968
6969 sub get_running_qemu_version {
6970     my ($vmid) = @_;
6971     my $res = mon_cmd($vmid, "query-version");
6972     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6973 }
6974
6975 sub qemu_use_old_bios_files {
6976     my ($machine_type) = @_;
6977
6978     return if !$machine_type;
6979
6980     my $use_old_bios_files = undef;
6981
6982     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6983         $machine_type = $1;
6984         $use_old_bios_files = 1;
6985     } else {
6986         my $version = extract_version($machine_type, kvm_user_version());
6987         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6988         # load new efi bios files on migration. So this hack is required to allow
6989         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6990         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6991         $use_old_bios_files = !min_version($version, 2, 4);
6992     }
6993
6994     return ($use_old_bios_files, $machine_type);
6995 }
6996
6997 sub get_efivars_size {
6998     my ($conf) = @_;
6999     my $arch = get_vm_arch($conf);
7000     my (undef, $ovmf_vars) = get_ovmf_files($arch);
7001     die "uefi vars image '$ovmf_vars' not found\n" if ! -f $ovmf_vars;
7002     return -s $ovmf_vars;
7003 }
7004
7005 sub update_efidisk_size {
7006     my ($conf) = @_;
7007
7008     return if !defined($conf->{efidisk0});
7009
7010     my $disk = PVE::QemuServer::parse_drive('efidisk0', $conf->{efidisk0});
7011     $disk->{size} = get_efivars_size($conf);
7012     $conf->{efidisk0} = print_drive($disk);
7013
7014     return;
7015 }
7016
7017 sub create_efidisk($$$$$) {
7018     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7019
7020     my (undef, $ovmf_vars) = get_ovmf_files($arch);
7021     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7022
7023     my $vars_size_b = -s $ovmf_vars;
7024     my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7025     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7026     PVE::Storage::activate_volumes($storecfg, [$volid]);
7027
7028     qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7029     my ($size) = PVE::Storage::volume_size_info($storecfg, $volid, 3);
7030
7031     return ($volid, $size/1024);
7032 }
7033
7034 sub vm_iothreads_list {
7035     my ($vmid) = @_;
7036
7037     my $res = mon_cmd($vmid, 'query-iothreads');
7038
7039     my $iothreads = {};
7040     foreach my $iothread (@$res) {
7041         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7042     }
7043
7044     return $iothreads;
7045 }
7046
7047 sub scsihw_infos {
7048     my ($conf, $drive) = @_;
7049
7050     my $maxdev = 0;
7051
7052     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7053         $maxdev = 7;
7054     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7055         $maxdev = 1;
7056     } else {
7057         $maxdev = 256;
7058     }
7059
7060     my $controller = int($drive->{index} / $maxdev);
7061     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single')
7062         ? "virtioscsi"
7063         : "scsihw";
7064
7065     return ($maxdev, $controller, $controller_prefix);
7066 }
7067
7068 sub windows_version {
7069     my ($ostype) = @_;
7070
7071     return 0 if !$ostype;
7072
7073     my $winversion = 0;
7074
7075     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7076         $winversion = 5;
7077     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7078         $winversion = 6;
7079     } elsif ($ostype =~ m/^win(\d+)$/) {
7080         $winversion = $1;
7081     }
7082
7083     return $winversion;
7084 }
7085
7086 sub resolve_dst_disk_format {
7087         my ($storecfg, $storeid, $src_volname, $format) = @_;
7088         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7089
7090         if (!$format) {
7091             # if no target format is specified, use the source disk format as hint
7092             if ($src_volname) {
7093                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7094                 $format = qemu_img_format($scfg, $src_volname);
7095             } else {
7096                 return $defFormat;
7097             }
7098         }
7099
7100         # test if requested format is supported - else use default
7101         my $supported = grep { $_ eq $format } @$validFormats;
7102         $format = $defFormat if !$supported;
7103         return $format;
7104 }
7105
7106 # NOTE: if this logic changes, please update docs & possibly gui logic
7107 sub find_vmstate_storage {
7108     my ($conf, $storecfg) = @_;
7109
7110     # first, return storage from conf if set
7111     return $conf->{vmstatestorage} if $conf->{vmstatestorage};
7112
7113     my ($target, $shared, $local);
7114
7115     foreach_storage_used_by_vm($conf, sub {
7116         my ($sid) = @_;
7117         my $scfg = PVE::Storage::storage_config($storecfg, $sid);
7118         my $dst = $scfg->{shared} ? \$shared : \$local;
7119         $$dst = $sid if !$$dst || $scfg->{path}; # prefer file based storage
7120     });
7121
7122     # second, use shared storage where VM has at least one disk
7123     # third, use local storage where VM has at least one disk
7124     # fall back to local storage
7125     $target = $shared // $local // 'local';
7126
7127     return $target;
7128 }
7129
7130 sub generate_uuid {
7131     my ($uuid, $uuid_str);
7132     UUID::generate($uuid);
7133     UUID::unparse($uuid, $uuid_str);
7134     return $uuid_str;
7135 }
7136
7137 sub generate_smbios1_uuid {
7138     return "uuid=".generate_uuid();
7139 }
7140
7141 sub nbd_stop {
7142     my ($vmid) = @_;
7143
7144     mon_cmd($vmid, 'nbd-server-stop');
7145 }
7146
7147 sub create_reboot_request {
7148     my ($vmid) = @_;
7149     open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7150         or die "failed to create reboot trigger file: $!\n";
7151     close($fh);
7152 }
7153
7154 sub clear_reboot_request {
7155     my ($vmid) = @_;
7156     my $path = "/run/qemu-server/$vmid.reboot";
7157     my $res = 0;
7158
7159     $res = unlink($path);
7160     die "could not remove reboot request for $vmid: $!"
7161         if !$res && $! != POSIX::ENOENT;
7162
7163     return $res;
7164 }
7165
7166 sub bootorder_from_legacy {
7167     my ($conf, $bootcfg) = @_;
7168
7169     my $boot = $bootcfg->{legacy} || $boot_fmt->{legacy}->{default};
7170     my $bootindex_hash = {};
7171     my $i = 1;
7172     foreach my $o (split(//, $boot)) {
7173         $bootindex_hash->{$o} = $i*100;
7174         $i++;
7175     }
7176
7177     my $bootorder = {};
7178
7179     PVE::QemuConfig->foreach_volume($conf, sub {
7180         my ($ds, $drive) = @_;
7181
7182         if (drive_is_cdrom ($drive, 1)) {
7183             if ($bootindex_hash->{d}) {
7184                 $bootorder->{$ds} = $bootindex_hash->{d};
7185                 $bootindex_hash->{d} += 1;
7186             }
7187         } elsif ($bootindex_hash->{c}) {
7188             $bootorder->{$ds} = $bootindex_hash->{c}
7189                 if $conf->{bootdisk} && $conf->{bootdisk} eq $ds;
7190             $bootindex_hash->{c} += 1;
7191         }
7192     });
7193
7194     if ($bootindex_hash->{n}) {
7195         for (my $i = 0; $i < $MAX_NETS; $i++) {
7196             my $netname = "net$i";
7197             next if !$conf->{$netname};
7198             $bootorder->{$netname} = $bootindex_hash->{n};
7199             $bootindex_hash->{n} += 1;
7200         }
7201     }
7202
7203     return $bootorder;
7204 }
7205
7206 # Generate default device list for 'boot: order=' property. Matches legacy
7207 # default boot order, but with explicit device names. This is important, since
7208 # the fallback for when neither 'order' nor the old format is specified relies
7209 # on 'bootorder_from_legacy' above, and it would be confusing if this diverges.
7210 sub get_default_bootdevices {
7211     my ($conf) = @_;
7212
7213     my @ret = ();
7214
7215     # harddisk
7216     my $first = PVE::QemuServer::Drive::resolve_first_disk($conf, 0);
7217     push @ret, $first if $first;
7218
7219     # cdrom
7220     $first = PVE::QemuServer::Drive::resolve_first_disk($conf, 1);
7221     push @ret, $first if $first;
7222
7223     # network
7224     for (my $i = 0; $i < $MAX_NETS; $i++) {
7225         my $netname = "net$i";
7226         next if !$conf->{$netname};
7227         push @ret, $netname;
7228         last;
7229     }
7230
7231     return \@ret;
7232 }
7233
7234 sub device_bootorder {
7235     my ($conf) = @_;
7236
7237     return bootorder_from_legacy($conf) if !defined($conf->{boot});
7238
7239     my $boot = parse_property_string($boot_fmt, $conf->{boot});
7240
7241     my $bootorder = {};
7242     if (!defined($boot) || $boot->{legacy}) {
7243         $bootorder = bootorder_from_legacy($conf, $boot);
7244     } elsif ($boot->{order}) {
7245         my $i = 100; # start at 100 to allow user to insert devices before us with -args
7246         for my $dev (PVE::Tools::split_list($boot->{order})) {
7247             $bootorder->{$dev} = $i++;
7248         }
7249     }
7250
7251     return $bootorder;
7252 }
7253
7254 # bash completion helper
7255
7256 sub complete_backup_archives {
7257     my ($cmdname, $pname, $cvalue) = @_;
7258
7259     my $cfg = PVE::Storage::config();
7260
7261     my $storeid;
7262
7263     if ($cvalue =~ m/^([^:]+):/) {
7264         $storeid = $1;
7265     }
7266
7267     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7268
7269     my $res = [];
7270     foreach my $id (keys %$data) {
7271         foreach my $item (@{$data->{$id}}) {
7272             next if $item->{format} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
7273             push @$res, $item->{volid} if defined($item->{volid});
7274         }
7275     }
7276
7277     return $res;
7278 }
7279
7280 my $complete_vmid_full = sub {
7281     my ($running) = @_;
7282
7283     my $idlist = vmstatus();
7284
7285     my $res = [];
7286
7287     foreach my $id (keys %$idlist) {
7288         my $d = $idlist->{$id};
7289         if (defined($running)) {
7290             next if $d->{template};
7291             next if $running && $d->{status} ne 'running';
7292             next if !$running && $d->{status} eq 'running';
7293         }
7294         push @$res, $id;
7295
7296     }
7297     return $res;
7298 };
7299
7300 sub complete_vmid {
7301     return &$complete_vmid_full();
7302 }
7303
7304 sub complete_vmid_stopped {
7305     return &$complete_vmid_full(0);
7306 }
7307
7308 sub complete_vmid_running {
7309     return &$complete_vmid_full(1);
7310 }
7311
7312 sub complete_storage {
7313
7314     my $cfg = PVE::Storage::config();
7315     my $ids = $cfg->{ids};
7316
7317     my $res = [];
7318     foreach my $sid (keys %$ids) {
7319         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7320         next if !$ids->{$sid}->{content}->{images};
7321         push @$res, $sid;
7322     }
7323
7324     return $res;
7325 }
7326
7327 sub complete_migration_storage {
7328     my ($cmd, $param, $current_value, $all_args) = @_;
7329
7330     my $targetnode = @$all_args[1];
7331
7332     my $cfg = PVE::Storage::config();
7333     my $ids = $cfg->{ids};
7334
7335     my $res = [];
7336     foreach my $sid (keys %$ids) {
7337         next if !PVE::Storage::storage_check_enabled($cfg, $sid, $targetnode, 1);
7338         next if !$ids->{$sid}->{content}->{images};
7339         push @$res, $sid;
7340     }
7341
7342     return $res;
7343 }
7344
7345 1;