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 cfs_lock_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);
  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 lock_file lock_file_full 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 foreach_drive foreach_volid);
  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);
  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',
 401         description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
 402         pattern => '[acdn]{1,4}',
 403         default => 'cdn',
 404     },
 405     bootdisk => {
 406         optional => 1,
 407         type => 'string', format => 'pve-qm-bootdisk',
 408         description => "Enable booting from specified disk.",
 409         pattern => '(ide|sata|scsi|virtio)\d+',
 410     },
 411     smp => {
 412         optional => 1,
 413         type => 'integer',
 414         description => "The number of CPUs. Please use option -sockets instead.",
 415         minimum => 1,
 416         default => 1,
 417     },
 418     sockets => {
 419         optional => 1,
 420         type => 'integer',
 421         description => "The number of CPU sockets.",
 422         minimum => 1,
 423         default => 1,
 424     },
 425     cores => {
 426         optional => 1,
 427         type => 'integer',
 428         description => "The number of cores per socket.",
 429         minimum => 1,
 430         default => 1,
 431     },
 432     numa => {
 433         optional => 1,
 434         type => 'boolean',
 435         description => "Enable/disable NUMA.",
 436         default => 0,
 437     },
 438     hugepages => {
 439         optional => 1,
 440         type => 'string',
 441         description => "Enable/disable hugepages memory.",
 442         enum => [qw(any 2 1024)],
 443     },
 444     vcpus => {
 445         optional => 1,
 446         type => 'integer',
 447         description => "Number of hotplugged vcpus.",
 448         minimum => 1,
 449         default => 0,
 450     },
 451     acpi => {
 452         optional => 1,
 453         type => 'boolean',
 454         description => "Enable/disable ACPI.",
 455         default => 1,
 456     },
 457     agent => {
 458         optional => 1,
 459         description => "Enable/disable Qemu GuestAgent and its properties.",
 460         type => 'string',
 461         format => $agent_fmt,
 462     },
 463     kvm => {
 464         optional => 1,
 465         type => 'boolean',
 466         description => "Enable/disable KVM hardware virtualization.",
 467         default => 1,
 468     },
 469     tdf => {
 470         optional => 1,
 471         type => 'boolean',
 472         description => "Enable/disable time drift fix.",
 473         default => 0,
 474     },
 475     localtime => {
 476         optional => 1,
 477         type => 'boolean',
 478         description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
 479     },
 480     freeze => {
 481         optional => 1,
 482         type => 'boolean',
 483         description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
 484     },
 485     vga => {
 486         optional => 1,
 487         type => 'string', format => $vga_fmt,
 488         description => "Configure the VGA hardware.",
 489         verbose_description => "Configure the VGA Hardware. If you want to use ".
 490             "high resolution modes (>= 1280x1024x16) you may need to increase " .
 491             "the vga memory option. Since QEMU 2.9 the default VGA display type " .
 492             "is 'std' for all OS types besides some Windows versions (XP and " .
 493             "older) which use 'cirrus'. The 'qxl' option enables the SPICE " .
 494             "display server. For win* OS you can select how many independent " .
 495             "displays you want, Linux guests can add displays them self.\n".
 496             "You can also run without any graphic card, using a serial device as terminal.",
 497     },
 498     watchdog => {
 499         optional => 1,
 500         type => 'string', format => 'pve-qm-watchdog',
 501         description => "Create a virtual hardware watchdog device.",
 502         verbose_description => "Create a virtual hardware watchdog device. Once enabled" .
 503             " (by a guest action), the watchdog must be periodically polled " .
 504             "by an agent inside the guest or else the watchdog will reset " .
 505             "the guest (or execute the respective action specified)",
 506     },
 507     startdate => {
 508         optional => 1,
 509         type => 'string',
 510         typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
 511         description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
 512         pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
 513         default => 'now',
 514     },
 515     startup =>  get_standard_option('pve-startup-order'),
 516     template => {
 517         optional => 1,
 518         type => 'boolean',
 519         description => "Enable/disable Template.",
 520         default => 0,
 521     },
 522     args => {
 523         optional => 1,
 524         type => 'string',
 525         description => "Arbitrary arguments passed to kvm.",
 526         verbose_description => <<EODESCR,
 527 Arbitrary arguments passed to kvm, for example:
 528
 529 args: -no-reboot -no-hpet
 530
 531 NOTE: this option is for experts only.
 532 EODESCR
 533     },
 534     tablet => {
 535         optional => 1,
 536         type => 'boolean',
 537         default => 1,
 538         description => "Enable/disable the USB tablet device.",
 539         verbose_description => "Enable/disable the USB tablet device. This device is " .
 540             "usually needed to allow absolute mouse positioning with VNC. " .
 541             "Else the mouse runs out of sync with normal VNC clients. " .
 542             "If you're running lots of console-only guests on one host, " .
 543             "you may consider disabling this to save some context switches. " .
 544             "This is turned off by default if you use spice (-vga=qxl).",
 545     },
 546     migrate_speed => {
 547         optional => 1,
 548         type => 'integer',
 549         description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
 550         minimum => 0,
 551         default => 0,
 552     },
 553     migrate_downtime => {
 554         optional => 1,
 555         type => 'number',
 556         description => "Set maximum tolerated downtime (in seconds) for migrations.",
 557         minimum => 0,
 558         default => 0.1,
 559     },
 560     cdrom => {
 561         optional => 1,
 562         type => 'string', format => 'pve-qm-ide',
 563         typetext => '<volume>',
 564         description => "This is an alias for option -ide2",
 565     },
 566     cpu => {
 567         optional => 1,
 568         description => "Emulated CPU type.",
 569         type => 'string',
 570         format => 'pve-vm-cpu-conf',
 571     },
 572     parent => get_standard_option('pve-snapshot-name', {
 573         optional => 1,
 574         description => "Parent snapshot name. This is used internally, and should not be modified.",
 575     }),
 576     snaptime => {
 577         optional => 1,
 578         description => "Timestamp for snapshots.",
 579         type => 'integer',
 580         minimum => 0,
 581     },
 582     vmstate => {
 583         optional => 1,
 584         type => 'string', format => 'pve-volume-id',
 585         description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
 586     },
 587     vmstatestorage => get_standard_option('pve-storage-id', {
 588         description => "Default storage for VM state volumes/files.",
 589         optional => 1,
 590     }),
 591     runningmachine => get_standard_option('pve-qemu-machine', {
 592         description => "Specifies the QEMU machine type of the running vm. This is used internally for snapshots.",
 593     }),
 594     runningcpu => {
 595         description => "Specifies the QEMU '-cpu' parameter of the running vm. This is used internally for snapshots.",
 596         optional => 1,
 597         type => 'string',
 598         pattern => $PVE::QemuServer::CPUConfig::qemu_cmdline_cpu_re,
 599         format_description => 'QEMU -cpu parameter'
 600     },
 601     machine => get_standard_option('pve-qemu-machine'),
 602     arch => {
 603         description => "Virtual processor architecture. Defaults to the host.",
 604         optional => 1,
 605         type => 'string',
 606         enum => [qw(x86_64 aarch64)],
 607     },
 608     smbios1 => {
 609         description => "Specify SMBIOS type 1 fields.",
 610         type => 'string', format => 'pve-qm-smbios1',
 611         maxLength => 512,
 612         optional => 1,
 613     },
 614     protection => {
 615         optional => 1,
 616         type => 'boolean',
 617         description => "Sets the protection flag of the VM. This will disable the remove VM and remove disk operations.",
 618         default => 0,
 619     },
 620     bios => {
 621         optional => 1,
 622         type => 'string',
 623         enum => [ qw(seabios ovmf) ],
 624         description => "Select BIOS implementation.",
 625         default => 'seabios',
 626     },
 627     vmgenid => {
 628         type => 'string',
 629         pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
 630         format_description => 'UUID',
 631         description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0' to disable explicitly.",
 632         verbose_description => "The VM generation ID (vmgenid) device exposes a".
 633             " 128-bit integer value identifier to the guest OS. This allows to".
 634             " notify the guest operating system when the virtual machine is".
 635             " executed with a different configuration (e.g. snapshot execution".
 636             " or creation from a template). The guest operating system notices".
 637             " the change, and is then able to react as appropriate by marking".
 638             " its copies of distributed databases as dirty, re-initializing its".
 639             " random number generator, etc.\n".
 640             "Note that auto-creation only works when done throug API/CLI create".
 641             " or update methods, but not when manually editing the config file.",
 642         default => "1 (autogenerated)",
 643         optional => 1,
 644     },
 645     hookscript => {
 646         type => 'string',
 647         format => 'pve-volume-id',
 648         optional => 1,
 649         description => "Script that will be executed during various steps in the vms lifetime.",
 650     },
 651     ivshmem => {
 652         type => 'string',
 653         format => $ivshmem_fmt,
 654         description => "Inter-VM shared memory. Useful for direct communication between VMs, or to the host.",
 655         optional => 1,
 656     },
 657     audio0 => {
 658         type => 'string',
 659         format => $audio_fmt,
 660         description => "Configure a audio device, useful in combination with QXL/Spice.",
 661         optional => 1
 662     },
 663     spice_enhancements => {
 664         type => 'string',
 665         format => $spice_enhancements_fmt,
 666         description => "Configure additional enhancements for SPICE.",
 667         optional => 1
 668     },
 669     tags => {
 670         type => 'string', format => 'pve-tag-list',
 671         description => 'Tags of the VM. This is only meta information.',
 672         optional => 1,
 673     },
 674     rng0 => {
 675         type => 'string',
 676         format => $rng_fmt,
 677         description => "Configure a VirtIO-based Random Number Generator.",
 678         optional => 1,
 679     },
 680 };
 681
 682 my $cicustom_fmt = {
 683     meta => {
 684         type => 'string',
 685         optional => 1,
 686         description => 'Specify a custom file containing all meta data passed to the VM via cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
 687         format => 'pve-volume-id',
 688         format_description => 'volume',
 689     },
 690     network => {
 691         type => 'string',
 692         optional => 1,
 693         description => 'Specify a custom file containing all network data passed to the VM via cloud-init.',
 694         format => 'pve-volume-id',
 695         format_description => 'volume',
 696     },
 697     user => {
 698         type => 'string',
 699         optional => 1,
 700         description => 'Specify a custom file containing all user data passed to the VM via cloud-init.',
 701         format => 'pve-volume-id',
 702         format_description => 'volume',
 703     },
 704 };
 705 PVE::JSONSchema::register_format('pve-qm-cicustom', $cicustom_fmt);
 706
 707 my $confdesc_cloudinit = {
 708     citype => {
 709         optional => 1,
 710         type => 'string',
 711         description => 'Specifies the cloud-init configuration format. The default depends on the configured operating system type (`ostype`. We use the `nocloud` format for Linux, and `configdrive2` for windows.',
 712         enum => ['configdrive2', 'nocloud'],
 713     },
 714     ciuser => {
 715         optional => 1,
 716         type => 'string',
 717         description => "cloud-init: User name to change ssh keys and password for instead of the image's configured default user.",
 718     },
 719     cipassword => {
 720         optional => 1,
 721         type => 'string',
 722         description => 'cloud-init: Password to assign the user. Using this is generally not recommended. Use ssh keys instead. Also note that older cloud-init versions do not support hashed passwords.',
 723     },
 724     cicustom => {
 725         optional => 1,
 726         type => 'string',
 727         description => 'cloud-init: Specify custom files to replace the automatically generated ones at start.',
 728         format => 'pve-qm-cicustom',
 729     },
 730     searchdomain => {
 731         optional => 1,
 732         type => 'string',
 733         description => "cloud-init: Sets DNS search domains for a container. Create will automatically use the setting from the host if neither searchdomain nor nameserver are set.",
 734     },
 735     nameserver => {
 736         optional => 1,
 737         type => 'string', format => 'address-list',
 738         description => "cloud-init: Sets DNS server IP address for a container. Create will automatically use the setting from the host if neither searchdomain nor nameserver are set.",
 739     },
 740     sshkeys => {
 741         optional => 1,
 742         type => 'string',
 743         format => 'urlencoded',
 744         description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
 745     },
 746 };
 747
 748 # what about other qemu settings ?
 749 #cpu => 'string',
 750 #machine => 'string',
 751 #fda => 'file',
 752 #fdb => 'file',
 753 #mtdblock => 'file',
 754 #sd => 'file',
 755 #pflash => 'file',
 756 #snapshot => 'bool',
 757 #bootp => 'file',
 758 ##tftp => 'dir',
 759 ##smb => 'dir',
 760 #kernel => 'file',
 761 #append => 'string',
 762 #initrd => 'file',
 763 ##soundhw => 'string',
 764
 765 while (my ($k, $v) = each %$confdesc) {
 766     PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
 767 }
 768
 769 my $MAX_USB_DEVICES = 5;
 770 my $MAX_NETS = 32;
 771 my $MAX_HOSTPCI_DEVICES = 16;
 772 my $MAX_SERIAL_PORTS = 4;
 773 my $MAX_PARALLEL_PORTS = 3;
 774 my $MAX_NUMA = 8;
 775
 776 my $numa_fmt = {
 777     cpus => {
 778         type => "string",
 779         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 780         description => "CPUs accessing this NUMA node.",
 781         format_description => "id[-id];...",
 782     },
 783     memory => {
 784         type => "number",
 785         description => "Amount of memory this NUMA node provides.",
 786         optional => 1,
 787     },
 788     hostnodes => {
 789         type => "string",
 790         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 791         description => "Host NUMA nodes to use.",
 792         format_description => "id[-id];...",
 793         optional => 1,
 794     },
 795     policy => {
 796         type => 'string',
 797         enum => [qw(preferred bind interleave)],
 798         description => "NUMA allocation policy.",
 799         optional => 1,
 800     },
 801 };
 802 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
 803 my $numadesc = {
 804     optional => 1,
 805     type => 'string', format => $numa_fmt,
 806     description => "NUMA topology.",
 807 };
 808 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
 809
 810 for (my $i = 0; $i < $MAX_NUMA; $i++)  {
 811     $confdesc->{"numa$i"} = $numadesc;
 812 }
 813
 814 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000',  'pcnet',  'virtio',
 815                       'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
 816                       'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
 817 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
 818
 819 my $net_fmt_bridge_descr = <<__EOD__;
 820 Bridge to attach the network device to. The Proxmox VE standard bridge
 821 is called 'vmbr0'.
 822
 823 If you do not specify a bridge, we create a kvm user (NATed) network
 824 device, which provides DHCP and DNS services. The following addresses
 825 are used:
 826
 827  10.0.2.2   Gateway
 828  10.0.2.3   DNS Server
 829  10.0.2.4   SMB Server
 830
 831 The DHCP server assign addresses to the guest starting from 10.0.2.15.
 832 __EOD__
 833
 834 my $net_fmt = {
 835     macaddr  => get_standard_option('mac-addr', {
 836         description => "MAC address. That address must be unique withing your network. This is automatically generated if not specified.",
 837     }),
 838     model => {
 839         type => 'string',
 840         description => "Network Card Model. The 'virtio' model provides the best performance with very low CPU overhead. If your guest does not support this driver, it is usually best to use 'e1000'.",
 841         enum => $nic_model_list,
 842         default_key => 1,
 843     },
 844     (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
 845     bridge => {
 846         type => 'string',
 847         description => $net_fmt_bridge_descr,
 848         format_description => 'bridge',
 849         pattern => '[-_.\w\d]+',
 850         optional => 1,
 851     },
 852     queues => {
 853         type => 'integer',
 854         minimum => 0, maximum => 16,
 855         description => 'Number of packet queues to be used on the device.',
 856         optional => 1,
 857     },
 858     rate => {
 859         type => 'number',
 860         minimum => 0,
 861         description => "Rate limit in mbps (megabytes per second) as floating point number.",
 862         optional => 1,
 863     },
 864     tag => {
 865         type => 'integer',
 866         minimum => 1, maximum => 4094,
 867         description => 'VLAN tag to apply to packets on this interface.',
 868         optional => 1,
 869     },
 870     trunks => {
 871         type => 'string',
 872         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 873         description => 'VLAN trunks to pass through this interface.',
 874         format_description => 'vlanid[;vlanid...]',
 875         optional => 1,
 876     },
 877     firewall => {
 878         type => 'boolean',
 879         description => 'Whether this interface should be protected by the firewall.',
 880         optional => 1,
 881     },
 882     link_down => {
 883         type => 'boolean',
 884         description => 'Whether this interface should be disconnected (like pulling the plug).',
 885         optional => 1,
 886     },
 887 };
 888
 889 my $netdesc = {
 890     optional => 1,
 891     type => 'string', format => $net_fmt,
 892     description => "Specify network devices.",
 893 };
 894
 895 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
 896
 897 my $ipconfig_fmt = {
 898     ip => {
 899         type => 'string',
 900         format => 'pve-ipv4-config',
 901         format_description => 'IPv4Format/CIDR',
 902         description => 'IPv4 address in CIDR format.',
 903         optional => 1,
 904         default => 'dhcp',
 905     },
 906     gw => {
 907         type => 'string',
 908         format => 'ipv4',
 909         format_description => 'GatewayIPv4',
 910         description => 'Default gateway for IPv4 traffic.',
 911         optional => 1,
 912         requires => 'ip',
 913     },
 914     ip6 => {
 915         type => 'string',
 916         format => 'pve-ipv6-config',
 917         format_description => 'IPv6Format/CIDR',
 918         description => 'IPv6 address in CIDR format.',
 919         optional => 1,
 920         default => 'dhcp',
 921     },
 922     gw6 => {
 923         type => 'string',
 924         format => 'ipv6',
 925         format_description => 'GatewayIPv6',
 926         description => 'Default gateway for IPv6 traffic.',
 927         optional => 1,
 928         requires => 'ip6',
 929     },
 930 };
 931 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
 932 my $ipconfigdesc = {
 933     optional => 1,
 934     type => 'string', format => 'pve-qm-ipconfig',
 935     description => <<'EODESCR',
 936 cloud-init: Specify IP addresses and gateways for the corresponding interface.
 937
 938 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
 939
 940 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit gateway should be provided.
 941 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
 942
 943 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using dhcp on IPv4.
 944 EODESCR
 945 };
 946 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
 947
 948 for (my $i = 0; $i < $MAX_NETS; $i++)  {
 949     $confdesc->{"net$i"} = $netdesc;
 950     $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
 951 }
 952
 953 foreach my $key (keys %$confdesc_cloudinit) {
 954     $confdesc->{$key} = $confdesc_cloudinit->{$key};
 955 }
 956
 957 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
 958 sub verify_volume_id_or_qm_path {
 959     my ($volid, $noerr) = @_;
 960
 961     if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
 962         return $volid;
 963     }
 964
 965     # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
 966     $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
 967     if ($@) {
 968         return undef if $noerr;
 969         die $@;
 970     }
 971     return $volid;
 972 }
 973
 974 my $usb_fmt = {
 975     host => {
 976         default_key => 1,
 977         type => 'string', format => 'pve-qm-usb-device',
 978         format_description => 'HOSTUSBDEVICE|spice',
 979         description => <<EODESCR,
 980 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
 981
 982  'bus-port(.port)*' (decimal numbers) or
 983  'vendor_id:product_id' (hexadeciaml numbers) or
 984  'spice'
 985
 986 You can use the 'lsusb -t' command to list existing usb devices.
 987
 988 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
 989
 990 The value 'spice' can be used to add a usb redirection devices for spice.
 991 EODESCR
 992     },
 993     usb3 => {
 994         optional => 1,
 995         type => 'boolean',
 996         description => "Specifies whether if given host option is a USB3 device or port.",
 997         default => 0,
 998     },
 999 };
1000
1001 my $usbdesc = {
1002     optional => 1,
1003     type => 'string', format => $usb_fmt,
1004     description => "Configure an USB device (n is 0 to 4).",
1005 };
1006 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1007
1008 my $PCIRE = qr/([a-f0-9]{4}:)?[a-f0-9]{2}:[a-f0-9]{2}(?:\.[a-f0-9])?/;
1009 my $hostpci_fmt = {
1010     host => {
1011         default_key => 1,
1012         type => 'string',
1013         pattern => qr/$PCIRE(;$PCIRE)*/,
1014         format_description => 'HOSTPCIID[;HOSTPCIID2...]',
1015         description => <<EODESCR,
1016 Host PCI device pass through. The PCI ID of a host's PCI device or a list
1017 of PCI virtual functions of the host. HOSTPCIID syntax is:
1018
1019 'bus:dev.func' (hexadecimal numbers)
1020
1021 You can us the 'lspci' command to list existing PCI devices.
1022 EODESCR
1023     },
1024     rombar => {
1025         type => 'boolean',
1026         description =>  "Specify whether or not the device's ROM will be visible in the guest's memory map.",
1027         optional => 1,
1028         default => 1,
1029     },
1030     romfile => {
1031         type => 'string',
1032         pattern => '[^,;]+',
1033         format_description => 'string',
1034         description => "Custom pci device rom filename (must be located in /usr/share/kvm/).",
1035         optional => 1,
1036     },
1037     pcie => {
1038         type => 'boolean',
1039         description =>  "Choose the PCI-express bus (needs the 'q35' machine model).",
1040         optional => 1,
1041         default => 0,
1042     },
1043     'x-vga' => {
1044         type => 'boolean',
1045         description =>  "Enable vfio-vga device support.",
1046         optional => 1,
1047         default => 0,
1048     },
1049     'mdev' => {
1050         type => 'string',
1051         format_description => 'string',
1052         pattern => '[^/\.:]+',
1053         optional => 1,
1054         description => <<EODESCR
1055 The type of mediated device to use.
1056 An instance of this type will be created on startup of the VM and
1057 will be cleaned up when the VM stops.
1058 EODESCR
1059     }
1060 };
1061 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
1062
1063 my $hostpcidesc = {
1064         optional => 1,
1065         type => 'string', format => 'pve-qm-hostpci',
1066         description => "Map host PCI devices into guest.",
1067         verbose_description =>  <<EODESCR,
1068 Map host PCI devices into guest.
1069
1070 NOTE: This option allows direct access to host hardware. So it is no longer
1071 possible to migrate such machines - use with special care.
1072
1073 CAUTION: Experimental! User reported problems with this option.
1074 EODESCR
1075 };
1076 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
1077
1078 my $serialdesc = {
1079         optional => 1,
1080         type => 'string',
1081         pattern => '(/dev/.+|socket)',
1082         description =>  "Create a serial device inside the VM (n is 0 to 3)",
1083         verbose_description =>  <<EODESCR,
1084 Create a serial device inside the VM (n is 0 to 3), and pass through a
1085 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1086 host side (use 'qm terminal' to open a terminal connection).
1087
1088 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
1089
1090 CAUTION: Experimental! User reported problems with this option.
1091 EODESCR
1092 };
1093
1094 my $paralleldesc= {
1095         optional => 1,
1096         type => 'string',
1097         pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1098         description =>  "Map host parallel devices (n is 0 to 2).",
1099         verbose_description =>  <<EODESCR,
1100 Map host parallel devices (n is 0 to 2).
1101
1102 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1103
1104 CAUTION: Experimental! User reported problems with this option.
1105 EODESCR
1106 };
1107
1108 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
1109     $confdesc->{"parallel$i"} = $paralleldesc;
1110 }
1111
1112 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
1113     $confdesc->{"serial$i"} = $serialdesc;
1114 }
1115
1116 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
1117     $confdesc->{"hostpci$i"} = $hostpcidesc;
1118 }
1119
1120 for my $key (keys %{$PVE::QemuServer::Drive::drivedesc_hash}) {
1121     $confdesc->{$key} = $PVE::QemuServer::Drive::drivedesc_hash->{$key};
1122 }
1123
1124 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++)  {
1125     $confdesc->{"usb$i"} = $usbdesc;
1126 }
1127
1128 my $kvm_api_version = 0;
1129
1130 sub kvm_version {
1131     return $kvm_api_version if $kvm_api_version;
1132
1133     open my $fh, '<', '/dev/kvm'
1134         or return undef;
1135
1136     # 0xae00 => KVM_GET_API_VERSION
1137     $kvm_api_version = ioctl($fh, 0xae00, 0);
1138
1139     return $kvm_api_version;
1140 }
1141
1142 my $kvm_user_version = {};
1143 my $kvm_mtime = {};
1144
1145 sub kvm_user_version {
1146     my ($binary) = @_;
1147
1148     $binary //= get_command_for_arch(get_host_arch()); # get the native arch by default
1149     my $st = stat($binary);
1150
1151     my $cachedmtime = $kvm_mtime->{$binary} // -1;
1152     return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1153         $cachedmtime == $st->mtime;
1154
1155     $kvm_user_version->{$binary} = 'unknown';
1156     $kvm_mtime->{$binary} = $st->mtime;
1157
1158     my $code = sub {
1159         my $line = shift;
1160         if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1161             $kvm_user_version->{$binary} = $2;
1162         }
1163     };
1164
1165     eval { run_command([$binary, '--version'], outfunc => $code); };
1166     warn $@ if $@;
1167
1168     return $kvm_user_version->{$binary};
1169
1170 }
1171
1172 sub kernel_has_vhost_net {
1173     return -c '/dev/vhost-net';
1174 }
1175
1176 sub option_exists {
1177     my $key = shift;
1178     return defined($confdesc->{$key});
1179 }
1180
1181 my $cdrom_path;
1182 sub get_cdrom_path {
1183
1184     return  $cdrom_path if $cdrom_path;
1185
1186     return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1187     return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1188     return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1189 }
1190
1191 sub get_iso_path {
1192     my ($storecfg, $vmid, $cdrom) = @_;
1193
1194     if ($cdrom eq 'cdrom') {
1195         return get_cdrom_path();
1196     } elsif ($cdrom eq 'none') {
1197         return '';
1198     } elsif ($cdrom =~ m|^/|) {
1199         return $cdrom;
1200     } else {
1201         return PVE::Storage::path($storecfg, $cdrom);
1202     }
1203 }
1204
1205 # try to convert old style file names to volume IDs
1206 sub filename_to_volume_id {
1207     my ($vmid, $file, $media) = @_;
1208
1209      if (!($file eq 'none' || $file eq 'cdrom' ||
1210           $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1211
1212         return undef if $file =~ m|/|;
1213
1214         if ($media && $media eq 'cdrom') {
1215             $file = "local:iso/$file";
1216         } else {
1217             $file = "local:$vmid/$file";
1218         }
1219     }
1220
1221     return $file;
1222 }
1223
1224 sub verify_media_type {
1225     my ($opt, $vtype, $media) = @_;
1226
1227     return if !$media;
1228
1229     my $etype;
1230     if ($media eq 'disk') {
1231         $etype = 'images';
1232     } elsif ($media eq 'cdrom') {
1233         $etype = 'iso';
1234     } else {
1235         die "internal error";
1236     }
1237
1238     return if ($vtype eq $etype);
1239
1240     raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1241 }
1242
1243 sub cleanup_drive_path {
1244     my ($opt, $storecfg, $drive) = @_;
1245
1246     # try to convert filesystem paths to volume IDs
1247
1248     if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1249         ($drive->{file} !~ m|^/dev/.+|) &&
1250         ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1251         ($drive->{file} !~ m/^\d+$/)) {
1252         my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1253         raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1254         $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1255         verify_media_type($opt, $vtype, $drive->{media});
1256         $drive->{file} = $volid;
1257     }
1258
1259     $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1260 }
1261
1262 sub parse_hotplug_features {
1263     my ($data) = @_;
1264
1265     my $res = {};
1266
1267     return $res if $data eq '0';
1268
1269     $data = $confdesc->{hotplug}->{default} if $data eq '1';
1270
1271     foreach my $feature (PVE::Tools::split_list($data)) {
1272         if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1273             $res->{$1} = 1;
1274         } else {
1275             die "invalid hotplug feature '$feature'\n";
1276         }
1277     }
1278     return $res;
1279 }
1280
1281 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1282 sub pve_verify_hotplug_features {
1283     my ($value, $noerr) = @_;
1284
1285     return $value if parse_hotplug_features($value);
1286
1287     return undef if $noerr;
1288
1289     die "unable to parse hotplug option\n";
1290 }
1291
1292 sub scsi_inquiry {
1293     my($fh, $noerr) = @_;
1294
1295     my $SG_IO = 0x2285;
1296     my $SG_GET_VERSION_NUM = 0x2282;
1297
1298     my $versionbuf = "\x00" x 8;
1299     my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1300     if (!$ret) {
1301         die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1302         return undef;
1303     }
1304     my $version = unpack("I", $versionbuf);
1305     if ($version < 30000) {
1306         die "scsi generic interface too old\n"  if !$noerr;
1307         return undef;
1308     }
1309
1310     my $buf = "\x00" x 36;
1311     my $sensebuf = "\x00" x 8;
1312     my $cmd = pack("C x3 C x1", 0x12, 36);
1313
1314     # see /usr/include/scsi/sg.h
1315     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";
1316
1317     my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1318                       length($sensebuf), 0, length($buf), $buf,
1319                       $cmd, $sensebuf, 6000);
1320
1321     $ret = ioctl($fh, $SG_IO, $packet);
1322     if (!$ret) {
1323         die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1324         return undef;
1325     }
1326
1327     my @res = unpack($sg_io_hdr_t, $packet);
1328     if ($res[17] || $res[18]) {
1329         die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1330         return undef;
1331     }
1332
1333     my $res = {};
1334     (my $byte0, my $byte1, $res->{vendor},
1335      $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1336
1337     $res->{removable} = $byte1 & 128 ? 1 : 0;
1338     $res->{type} = $byte0 & 31;
1339
1340     return $res;
1341 }
1342
1343 sub path_is_scsi {
1344     my ($path) = @_;
1345
1346     my $fh = IO::File->new("+<$path") || return undef;
1347     my $res = scsi_inquiry($fh, 1);
1348     close($fh);
1349
1350     return $res;
1351 }
1352
1353 sub print_tabletdevice_full {
1354     my ($conf, $arch) = @_;
1355
1356     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
1357
1358     # we use uhci for old VMs because tablet driver was buggy in older qemu
1359     my $usbbus;
1360     if (PVE::QemuServer::Machine::machine_type_is_q35($conf) || $arch eq 'aarch64') {
1361         $usbbus = 'ehci';
1362     } else {
1363         $usbbus = 'uhci';
1364     }
1365
1366     return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1367 }
1368
1369 sub print_keyboarddevice_full {
1370     my ($conf, $arch, $machine) = @_;
1371
1372     return undef if $arch ne 'aarch64';
1373
1374     return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1375 }
1376
1377 sub print_drivedevice_full {
1378     my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1379
1380     my $device = '';
1381     my $maxdev = 0;
1382
1383     if ($drive->{interface} eq 'virtio') {
1384         my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges, $arch, $machine_type);
1385         $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1386         $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1387     } elsif ($drive->{interface} eq 'scsi') {
1388
1389         my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1390         my $unit = $drive->{index} % $maxdev;
1391         my $devicetype = 'hd';
1392         my $path = '';
1393         if (drive_is_cdrom($drive)) {
1394             $devicetype = 'cd';
1395         } else {
1396             if ($drive->{file} =~ m|^/|) {
1397                 $path = $drive->{file};
1398                 if (my $info = path_is_scsi($path)) {
1399                     if ($info->{type} == 0 && $drive->{scsiblock}) {
1400                         $devicetype = 'block';
1401                     } elsif ($info->{type} == 1) { # tape
1402                         $devicetype = 'generic';
1403                     }
1404                 }
1405             } else {
1406                  $path = PVE::Storage::path($storecfg, $drive->{file});
1407             }
1408
1409             # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1410             my $version = PVE::QemuServer::Machine::extract_version($machine_type, kvm_user_version());
1411             if ($path =~ m/^iscsi\:\/\// &&
1412                !min_version($version, 4, 1)) {
1413                 $devicetype = 'generic';
1414             }
1415         }
1416
1417         if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1418             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1419         } else {
1420             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1421         }
1422
1423         if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1424             $device .= ",rotation_rate=1";
1425         }
1426         $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1427
1428     } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1429         my $maxdev = ($drive->{interface} eq 'sata') ? $PVE::QemuServer::Drive::MAX_SATA_DISKS : 2;
1430         my $controller = int($drive->{index} / $maxdev);
1431         my $unit = $drive->{index} % $maxdev;
1432         my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1433
1434         $device = "ide-$devicetype";
1435         if ($drive->{interface} eq 'ide') {
1436             $device .= ",bus=ide.$controller,unit=$unit";
1437         } else {
1438             $device .= ",bus=ahci$controller.$unit";
1439         }
1440         $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1441
1442         if ($devicetype eq 'hd') {
1443             if (my $model = $drive->{model}) {
1444                 $model = URI::Escape::uri_unescape($model);
1445                 $device .= ",model=$model";
1446             }
1447             if ($drive->{ssd}) {
1448                 $device .= ",rotation_rate=1";
1449             }
1450         }
1451         $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1452     } elsif ($drive->{interface} eq 'usb') {
1453         die "implement me";
1454         #  -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1455     } else {
1456         die "unsupported interface type";
1457     }
1458
1459     $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1460
1461     if (my $serial = $drive->{serial}) {
1462         $serial = URI::Escape::uri_unescape($serial);
1463         $device .= ",serial=$serial";
1464     }
1465
1466
1467     return $device;
1468 }
1469
1470 sub get_initiator_name {
1471     my $initiator;
1472
1473     my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1474     while (defined(my $line = <$fh>)) {
1475         next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1476         $initiator = $1;
1477         last;
1478     }
1479     $fh->close();
1480
1481     return $initiator;
1482 }
1483
1484 sub print_drive_commandline_full {
1485     my ($storecfg, $vmid, $drive) = @_;
1486
1487     my $path;
1488     my $volid = $drive->{file};
1489     my $format;
1490
1491     if (drive_is_cdrom($drive)) {
1492         $path = get_iso_path($storecfg, $vmid, $volid);
1493     } else {
1494         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1495         if ($storeid) {
1496             $path = PVE::Storage::path($storecfg, $volid);
1497             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1498             $format = qemu_img_format($scfg, $volname);
1499         } else {
1500             $path = $volid;
1501             $format = "raw";
1502         }
1503    }
1504
1505     my $opts = '';
1506     my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1507     foreach my $o (@qemu_drive_options) {
1508         $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1509     }
1510
1511     # snapshot only accepts on|off
1512     if (defined($drive->{snapshot})) {
1513         my $v = $drive->{snapshot} ? 'on' : 'off';
1514         $opts .= ",snapshot=$v";
1515     }
1516
1517     foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
1518         my ($dir, $qmpname) = @$type;
1519         if (my $v = $drive->{"mbps$dir"}) {
1520             $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1521         }
1522         if (my $v = $drive->{"mbps${dir}_max"}) {
1523             $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1524         }
1525         if (my $v = $drive->{"bps${dir}_max_length"}) {
1526             $opts .= ",throttling.bps$qmpname-max-length=$v";
1527         }
1528         if (my $v = $drive->{"iops${dir}"}) {
1529             $opts .= ",throttling.iops$qmpname=$v";
1530         }
1531         if (my $v = $drive->{"iops${dir}_max"}) {
1532             $opts .= ",throttling.iops$qmpname-max=$v";
1533         }
1534         if (my $v = $drive->{"iops${dir}_max_length"}) {
1535             $opts .= ",throttling.iops$qmpname-max-length=$v";
1536         }
1537     }
1538
1539     $opts .= ",format=$format" if $format && !$drive->{format};
1540
1541     my $cache_direct = 0;
1542
1543     if (my $cache = $drive->{cache}) {
1544         $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1545     } elsif (!drive_is_cdrom($drive)) {
1546         $opts .= ",cache=none";
1547         $cache_direct = 1;
1548     }
1549
1550     # aio native works only with O_DIRECT
1551     if (!$drive->{aio}) {
1552         if($cache_direct) {
1553             $opts .= ",aio=native";
1554         } else {
1555             $opts .= ",aio=threads";
1556         }
1557     }
1558
1559     if (!drive_is_cdrom($drive)) {
1560         my $detectzeroes;
1561         if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
1562             $detectzeroes = 'off';
1563         } elsif ($drive->{discard}) {
1564             $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
1565         } else {
1566             # This used to be our default with discard not being specified:
1567             $detectzeroes = 'on';
1568         }
1569         $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
1570     }
1571
1572     my $pathinfo = $path ? "file=$path," : '';
1573
1574     return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1575 }
1576
1577 sub print_netdevice_full {
1578     my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
1579
1580     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1581
1582     my $device = $net->{model};
1583     if ($net->{model} eq 'virtio') {
1584          $device = 'virtio-net-pci';
1585      };
1586
1587     my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
1588     my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1589     if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1590         #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1591         my $vectors = $net->{queues} * 2 + 2;
1592         $tmpstr .= ",vectors=$vectors,mq=on";
1593     }
1594     $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1595
1596     if ($use_old_bios_files) {
1597         my $romfile;
1598         if ($device eq 'virtio-net-pci') {
1599             $romfile = 'pxe-virtio.rom';
1600         } elsif ($device eq 'e1000') {
1601             $romfile = 'pxe-e1000.rom';
1602         } elsif ($device eq 'ne2k') {
1603             $romfile = 'pxe-ne2k_pci.rom';
1604         } elsif ($device eq 'pcnet') {
1605             $romfile = 'pxe-pcnet.rom';
1606         } elsif ($device eq 'rtl8139') {
1607             $romfile = 'pxe-rtl8139.rom';
1608         }
1609         $tmpstr .= ",romfile=$romfile" if $romfile;
1610     }
1611
1612     return $tmpstr;
1613 }
1614
1615 sub print_netdev_full {
1616     my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
1617
1618     my $i = '';
1619     if ($netid =~ m/^net(\d+)$/) {
1620         $i = int($1);
1621     }
1622
1623     die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1624
1625     my $ifname = "tap${vmid}i$i";
1626
1627     # kvm uses TUNSETIFF ioctl, and that limits ifname length
1628     die "interface name '$ifname' is too long (max 15 character)\n"
1629         if length($ifname) >= 16;
1630
1631     my $vhostparam = '';
1632     if (is_native($arch)) {
1633         $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
1634     }
1635
1636     my $vmname = $conf->{name} || "vm$vmid";
1637
1638     my $netdev = "";
1639     my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
1640
1641     if ($net->{bridge}) {
1642         $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1643     } else {
1644         $netdev = "type=user,id=$netid,hostname=$vmname";
1645     }
1646
1647     $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1648
1649     return $netdev;
1650 }
1651
1652 my $vga_map = {
1653     'cirrus' => 'cirrus-vga',
1654     'std' => 'VGA',
1655     'vmware' => 'vmware-svga',
1656     'virtio' => 'virtio-vga',
1657 };
1658
1659 sub print_vga_device {
1660     my ($conf, $vga, $arch, $machine_version, $machine, $id, $qxlnum, $bridges) = @_;
1661
1662     my $type = $vga_map->{$vga->{type}};
1663     if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
1664         $type = 'virtio-gpu';
1665     }
1666     my $vgamem_mb = $vga->{memory};
1667
1668     my $max_outputs = '';
1669     if ($qxlnum) {
1670         $type = $id ? 'qxl' : 'qxl-vga';
1671
1672         if (!$conf->{ostype} || $conf->{ostype} =~ m/^(?:l\d\d)|(?:other)$/) {
1673             # set max outputs so linux can have up to 4 qxl displays with one device
1674             if (min_version($machine_version, 4, 1)) {
1675                 $max_outputs = ",max_outputs=4";
1676             }
1677         }
1678     }
1679
1680     die "no devicetype for $vga->{type}\n" if !$type;
1681
1682     my $memory = "";
1683     if ($vgamem_mb) {
1684         if ($vga->{type} eq 'virtio') {
1685             my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
1686             $memory = ",max_hostmem=$bytes";
1687         } elsif ($qxlnum) {
1688             # from https://www.spice-space.org/multiple-monitors.html
1689             $memory = ",vgamem_mb=$vga->{memory}";
1690             my $ram = $vgamem_mb * 4;
1691             my $vram = $vgamem_mb * 2;
1692             $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
1693         } else {
1694             $memory = ",vgamem_mb=$vga->{memory}";
1695         }
1696     } elsif ($qxlnum && $id) {
1697         $memory = ",ram_size=67108864,vram_size=33554432";
1698     }
1699
1700     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
1701     my $vgaid = "vga" . ($id // '');
1702     my $pciaddr;
1703
1704     if ($q35 && $vgaid eq 'vga') {
1705         # the first display uses pcie.0 bus on q35 machines
1706         $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
1707     } else {
1708         $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
1709     }
1710
1711     return "$type,id=${vgaid}${memory}${max_outputs}${pciaddr}";
1712 }
1713
1714 sub parse_number_sets {
1715     my ($set) = @_;
1716     my $res = [];
1717     foreach my $part (split(/;/, $set)) {
1718         if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1719             die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1720             push @$res, [ $1, $2 ];
1721         } else {
1722             die "invalid range: $part\n";
1723         }
1724     }
1725     return $res;
1726 }
1727
1728 sub parse_numa {
1729     my ($data) = @_;
1730
1731     my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
1732     $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
1733     $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
1734     return $res;
1735 }
1736
1737 sub parse_hostpci {
1738     my ($value) = @_;
1739
1740     return undef if !$value;
1741
1742     my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
1743
1744     my @idlist = split(/;/, $res->{host});
1745     delete $res->{host};
1746     foreach my $id (@idlist) {
1747         my $devs = PVE::SysFSTools::lspci($id);
1748         die "no PCI device found for '$id'\n" if !scalar(@$devs);
1749         push @{$res->{pciid}}, @$devs;
1750     }
1751     return $res;
1752 }
1753
1754 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1755 sub parse_net {
1756     my ($data) = @_;
1757
1758     my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
1759     if ($@) {
1760         warn $@;
1761         return undef;
1762     }
1763     if (!defined($res->{macaddr})) {
1764         my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
1765         $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
1766     }
1767     return $res;
1768 }
1769
1770 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
1771 sub parse_ipconfig {
1772     my ($data) = @_;
1773
1774     my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
1775     if ($@) {
1776         warn $@;
1777         return undef;
1778     }
1779
1780     if ($res->{gw} && !$res->{ip}) {
1781         warn 'gateway specified without specifying an IP address';
1782         return undef;
1783     }
1784     if ($res->{gw6} && !$res->{ip6}) {
1785         warn 'IPv6 gateway specified without specifying an IPv6 address';
1786         return undef;
1787     }
1788     if ($res->{gw} && $res->{ip} eq 'dhcp') {
1789         warn 'gateway specified together with DHCP';
1790         return undef;
1791     }
1792     if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
1793         # gw6 + auto/dhcp
1794         warn "IPv6 gateway specified together with $res->{ip6} address";
1795         return undef;
1796     }
1797
1798     if (!$res->{ip} && !$res->{ip6}) {
1799         return { ip => 'dhcp', ip6 => 'dhcp' };
1800     }
1801
1802     return $res;
1803 }
1804
1805 sub print_net {
1806     my $net = shift;
1807
1808     return PVE::JSONSchema::print_property_string($net, $net_fmt);
1809 }
1810
1811 sub add_random_macs {
1812     my ($settings) = @_;
1813
1814     foreach my $opt (keys %$settings) {
1815         next if $opt !~ m/^net(\d+)$/;
1816         my $net = parse_net($settings->{$opt});
1817         next if !$net;
1818         $settings->{$opt} = print_net($net);
1819     }
1820 }
1821
1822 sub vm_is_volid_owner {
1823     my ($storecfg, $vmid, $volid) = @_;
1824
1825     if ($volid !~  m|^/|) {
1826         my ($path, $owner);
1827         eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1828         if ($owner && ($owner == $vmid)) {
1829             return 1;
1830         }
1831     }
1832
1833     return undef;
1834 }
1835
1836 sub vmconfig_register_unused_drive {
1837     my ($storecfg, $vmid, $conf, $drive) = @_;
1838
1839     if (drive_is_cloudinit($drive)) {
1840         eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
1841         warn $@ if $@;
1842     } elsif (!drive_is_cdrom($drive)) {
1843         my $volid = $drive->{file};
1844         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1845             PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
1846         }
1847     }
1848 }
1849
1850 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
1851 my $smbios1_fmt = {
1852     uuid => {
1853         type => 'string',
1854         pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1855         format_description => 'UUID',
1856         description => "Set SMBIOS1 UUID.",
1857         optional => 1,
1858     },
1859     version => {
1860         type => 'string',
1861         pattern => '[A-Za-z0-9+\/]+={0,2}',
1862         format_description => 'Base64 encoded string',
1863         description => "Set SMBIOS1 version.",
1864         optional => 1,
1865     },
1866     serial => {
1867         type => 'string',
1868         pattern => '[A-Za-z0-9+\/]+={0,2}',
1869         format_description => 'Base64 encoded string',
1870         description => "Set SMBIOS1 serial number.",
1871         optional => 1,
1872     },
1873     manufacturer => {
1874         type => 'string',
1875         pattern => '[A-Za-z0-9+\/]+={0,2}',
1876         format_description => 'Base64 encoded string',
1877         description => "Set SMBIOS1 manufacturer.",
1878         optional => 1,
1879     },
1880     product => {
1881         type => 'string',
1882         pattern => '[A-Za-z0-9+\/]+={0,2}',
1883         format_description => 'Base64 encoded string',
1884         description => "Set SMBIOS1 product ID.",
1885         optional => 1,
1886     },
1887     sku => {
1888         type => 'string',
1889         pattern => '[A-Za-z0-9+\/]+={0,2}',
1890         format_description => 'Base64 encoded string',
1891         description => "Set SMBIOS1 SKU string.",
1892         optional => 1,
1893     },
1894     family => {
1895         type => 'string',
1896         pattern => '[A-Za-z0-9+\/]+={0,2}',
1897         format_description => 'Base64 encoded string',
1898         description => "Set SMBIOS1 family string.",
1899         optional => 1,
1900     },
1901     base64 => {
1902         type => 'boolean',
1903         description => 'Flag to indicate that the SMBIOS values are base64 encoded',
1904         optional => 1,
1905     },
1906 };
1907
1908 sub parse_smbios1 {
1909     my ($data) = @_;
1910
1911     my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
1912     warn $@ if $@;
1913     return $res;
1914 }
1915
1916 sub print_smbios1 {
1917     my ($smbios1) = @_;
1918     return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
1919 }
1920
1921 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
1922
1923 sub parse_watchdog {
1924     my ($value) = @_;
1925
1926     return undef if !$value;
1927
1928     my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
1929     warn $@ if $@;
1930     return $res;
1931 }
1932
1933 sub parse_guest_agent {
1934     my ($value) = @_;
1935
1936     return {} if !defined($value->{agent});
1937
1938     my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
1939     warn $@ if $@;
1940
1941     # if the agent is disabled ignore the other potentially set properties
1942     return {} if !$res->{enabled};
1943     return $res;
1944 }
1945
1946 sub parse_vga {
1947     my ($value) = @_;
1948
1949     return {} if !$value;
1950     my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
1951     warn $@ if $@;
1952     return $res;
1953 }
1954
1955 sub parse_rng {
1956     my ($value) = @_;
1957
1958     return undef if !$value;
1959
1960     my $res = eval { PVE::JSONSchema::parse_property_string($rng_fmt, $value) };
1961     warn $@ if $@;
1962     return $res;
1963 }
1964
1965 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1966 sub verify_usb_device {
1967     my ($value, $noerr) = @_;
1968
1969     return $value if parse_usb_device($value);
1970
1971     return undef if $noerr;
1972
1973     die "unable to parse usb device\n";
1974 }
1975
1976 # add JSON properties for create and set function
1977 sub json_config_properties {
1978     my $prop = shift;
1979
1980     foreach my $opt (keys %$confdesc) {
1981         next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' ||
1982             $opt eq 'runningmachine' || $opt eq 'runningcpu';
1983         $prop->{$opt} = $confdesc->{$opt};
1984     }
1985
1986     return $prop;
1987 }
1988
1989 # return copy of $confdesc_cloudinit to generate documentation
1990 sub cloudinit_config_properties {
1991
1992     return dclone($confdesc_cloudinit);
1993 }
1994
1995 sub check_type {
1996     my ($key, $value) = @_;
1997
1998     die "unknown setting '$key'\n" if !$confdesc->{$key};
1999
2000     my $type = $confdesc->{$key}->{type};
2001
2002     if (!defined($value)) {
2003         die "got undefined value\n";
2004     }
2005
2006     if ($value =~ m/[\n\r]/) {
2007         die "property contains a line feed\n";
2008     }
2009
2010     if ($type eq 'boolean') {
2011         return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2012         return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2013         die "type check ('boolean') failed - got '$value'\n";
2014     } elsif ($type eq 'integer') {
2015         return int($1) if $value =~ m/^(\d+)$/;
2016         die "type check ('integer') failed - got '$value'\n";
2017     } elsif ($type eq 'number') {
2018         return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2019         die "type check ('number') failed - got '$value'\n";
2020     } elsif ($type eq 'string') {
2021         if (my $fmt = $confdesc->{$key}->{format}) {
2022             PVE::JSONSchema::check_format($fmt, $value);
2023             return $value;
2024         }
2025         $value =~ s/^\"(.*)\"$/$1/;
2026         return $value;
2027     } else {
2028         die "internal error"
2029     }
2030 }
2031
2032 sub destroy_vm {
2033     my ($storecfg, $vmid, $skiplock, $replacement_conf) = @_;
2034
2035     my $conf = PVE::QemuConfig->load_config($vmid);
2036
2037     PVE::QemuConfig->check_lock($conf) if !$skiplock;
2038
2039     if ($conf->{template}) {
2040         # check if any base image is still used by a linked clone
2041         foreach_drive($conf, sub {
2042                 my ($ds, $drive) = @_;
2043                 return if drive_is_cdrom($drive);
2044
2045                 my $volid = $drive->{file};
2046                 return if !$volid || $volid =~ m|^/|;
2047
2048                 die "base volume '$volid' is still in use by linked cloned\n"
2049                     if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2050
2051         });
2052     }
2053
2054     # only remove disks owned by this VM
2055     foreach_drive($conf, sub {
2056         my ($ds, $drive) = @_;
2057         return if drive_is_cdrom($drive, 1);
2058
2059         my $volid = $drive->{file};
2060         return if !$volid || $volid =~ m|^/|;
2061
2062         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2063         return if !$path || !$owner || ($owner != $vmid);
2064
2065         eval { PVE::Storage::vdisk_free($storecfg, $volid) };
2066         warn "Could not remove disk '$volid', check manually: $@" if $@;
2067     });
2068
2069     # also remove unused disk
2070     my $vmdisks = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2071     PVE::Storage::foreach_volid($vmdisks, sub {
2072         my ($volid, $sid, $volname, $d) = @_;
2073         eval { PVE::Storage::vdisk_free($storecfg, $volid) };
2074         warn $@ if $@;
2075     });
2076
2077     if (defined $replacement_conf) {
2078         PVE::QemuConfig->write_config($vmid, $replacement_conf);
2079     } else {
2080         PVE::QemuConfig->destroy_config($vmid);
2081     }
2082 }
2083
2084 sub parse_vm_config {
2085     my ($filename, $raw) = @_;
2086
2087     return undef if !defined($raw);
2088
2089     my $res = {
2090         digest => Digest::SHA::sha1_hex($raw),
2091         snapshots => {},
2092         pending => {},
2093     };
2094
2095     $filename =~ m|/qemu-server/(\d+)\.conf$|
2096         || die "got strange filename '$filename'";
2097
2098     my $vmid = $1;
2099
2100     my $conf = $res;
2101     my $descr;
2102     my $section = '';
2103
2104     my @lines = split(/\n/, $raw);
2105     foreach my $line (@lines) {
2106         next if $line =~ m/^\s*$/;
2107
2108         if ($line =~ m/^\[PENDING\]\s*$/i) {
2109             $section = 'pending';
2110             if (defined($descr)) {
2111                 $descr =~ s/\s+$//;
2112                 $conf->{description} = $descr;
2113             }
2114             $descr = undef;
2115             $conf = $res->{$section} = {};
2116             next;
2117
2118         } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2119             $section = $1;
2120             if (defined($descr)) {
2121                 $descr =~ s/\s+$//;
2122                 $conf->{description} = $descr;
2123             }
2124             $descr = undef;
2125             $conf = $res->{snapshots}->{$section} = {};
2126             next;
2127         }
2128
2129         if ($line =~ m/^\#(.*)\s*$/) {
2130             $descr = '' if !defined($descr);
2131             $descr .= PVE::Tools::decode_text($1) . "\n";
2132             next;
2133         }
2134
2135         if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2136             $descr = '' if !defined($descr);
2137             $descr .= PVE::Tools::decode_text($2);
2138         } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2139             $conf->{snapstate} = $1;
2140         } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2141             my $key = $1;
2142             my $value = $2;
2143             $conf->{$key} = $value;
2144         } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2145             my $value = $1;
2146             if ($section eq 'pending') {
2147                 $conf->{delete} = $value; # we parse this later
2148             } else {
2149                 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2150             }
2151         } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2152             my $key = $1;
2153             my $value = $2;
2154             eval { $value = check_type($key, $value); };
2155             if ($@) {
2156                 warn "vm $vmid - unable to parse value of '$key' - $@";
2157             } else {
2158                 $key = 'ide2' if $key eq 'cdrom';
2159                 my $fmt = $confdesc->{$key}->{format};
2160                 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2161                     my $v = parse_drive($key, $value);
2162                     if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2163                         $v->{file} = $volid;
2164                         $value = print_drive($v);
2165                     } else {
2166                         warn "vm $vmid - unable to parse value of '$key'\n";
2167                         next;
2168                     }
2169                 }
2170
2171                 $conf->{$key} = $value;
2172             }
2173         }
2174     }
2175
2176     if (defined($descr)) {
2177         $descr =~ s/\s+$//;
2178         $conf->{description} = $descr;
2179     }
2180     delete $res->{snapstate}; # just to be sure
2181
2182     return $res;
2183 }
2184
2185 sub write_vm_config {
2186     my ($filename, $conf) = @_;
2187
2188     delete $conf->{snapstate}; # just to be sure
2189
2190     if ($conf->{cdrom}) {
2191         die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2192         $conf->{ide2} = $conf->{cdrom};
2193         delete $conf->{cdrom};
2194     }
2195
2196     # we do not use 'smp' any longer
2197     if ($conf->{sockets}) {
2198         delete $conf->{smp};
2199     } elsif ($conf->{smp}) {
2200         $conf->{sockets} = $conf->{smp};
2201         delete $conf->{cores};
2202         delete $conf->{smp};
2203     }
2204
2205     my $used_volids = {};
2206
2207     my $cleanup_config = sub {
2208         my ($cref, $pending, $snapname) = @_;
2209
2210         foreach my $key (keys %$cref) {
2211             next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2212                 $key eq 'snapstate' || $key eq 'pending';
2213             my $value = $cref->{$key};
2214             if ($key eq 'delete') {
2215                 die "propertry 'delete' is only allowed in [PENDING]\n"
2216                     if !$pending;
2217                 # fixme: check syntax?
2218                 next;
2219             }
2220             eval { $value = check_type($key, $value); };
2221             die "unable to parse value of '$key' - $@" if $@;
2222
2223             $cref->{$key} = $value;
2224
2225             if (!$snapname && is_valid_drivename($key)) {
2226                 my $drive = parse_drive($key, $value);
2227                 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2228             }
2229         }
2230     };
2231
2232     &$cleanup_config($conf);
2233
2234     &$cleanup_config($conf->{pending}, 1);
2235
2236     foreach my $snapname (keys %{$conf->{snapshots}}) {
2237         die "internal error: snapshot name '$snapname' is forbidden" if lc($snapname) eq 'pending';
2238         &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2239     }
2240
2241     # remove 'unusedX' settings if we re-add a volume
2242     foreach my $key (keys %$conf) {
2243         my $value = $conf->{$key};
2244         if ($key =~ m/^unused/ && $used_volids->{$value}) {
2245             delete $conf->{$key};
2246         }
2247     }
2248
2249     my $generate_raw_config = sub {
2250         my ($conf, $pending) = @_;
2251
2252         my $raw = '';
2253
2254         # add description as comment to top of file
2255         if (defined(my $descr = $conf->{description})) {
2256             if ($descr) {
2257                 foreach my $cl (split(/\n/, $descr)) {
2258                     $raw .= '#' .  PVE::Tools::encode_text($cl) . "\n";
2259                 }
2260             } else {
2261                 $raw .= "#\n" if $pending;
2262             }
2263         }
2264
2265         foreach my $key (sort keys %$conf) {
2266             next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2267             $raw .= "$key: $conf->{$key}\n";
2268         }
2269         return $raw;
2270     };
2271
2272     my $raw = &$generate_raw_config($conf);
2273
2274     if (scalar(keys %{$conf->{pending}})){
2275         $raw .= "\n[PENDING]\n";
2276         $raw .= &$generate_raw_config($conf->{pending}, 1);
2277     }
2278
2279     foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2280         $raw .= "\n[$snapname]\n";
2281         $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2282     }
2283
2284     return $raw;
2285 }
2286
2287 sub load_defaults {
2288
2289     my $res = {};
2290
2291     # we use static defaults from our JSON schema configuration
2292     foreach my $key (keys %$confdesc) {
2293         if (defined(my $default = $confdesc->{$key}->{default})) {
2294             $res->{$key} = $default;
2295         }
2296     }
2297
2298     return $res;
2299 }
2300
2301 sub config_list {
2302     my $vmlist = PVE::Cluster::get_vmlist();
2303     my $res = {};
2304     return $res if !$vmlist || !$vmlist->{ids};
2305     my $ids = $vmlist->{ids};
2306     my $nodename = nodename();
2307
2308     foreach my $vmid (keys %$ids) {
2309         my $d = $ids->{$vmid};
2310         next if !$d->{node} || $d->{node} ne $nodename;
2311         next if !$d->{type} || $d->{type} ne 'qemu';
2312         $res->{$vmid}->{exists} = 1;
2313     }
2314     return $res;
2315 }
2316
2317 # test if VM uses local resources (to prevent migration)
2318 sub check_local_resources {
2319     my ($conf, $noerr) = @_;
2320
2321     my @loc_res = ();
2322
2323     push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2324     push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2325
2326     push @loc_res, "ivshmem" if $conf->{ivshmem};
2327
2328     foreach my $k (keys %$conf) {
2329         next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2330         # sockets are safe: they will recreated be on the target side post-migrate
2331         next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2332         push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2333     }
2334
2335     die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2336
2337     return \@loc_res;
2338 }
2339
2340 # check if used storages are available on all nodes (use by migrate)
2341 sub check_storage_availability {
2342     my ($storecfg, $conf, $node) = @_;
2343
2344     foreach_drive($conf, sub {
2345         my ($ds, $drive) = @_;
2346
2347         my $volid = $drive->{file};
2348         return if !$volid;
2349
2350         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2351         return if !$sid;
2352
2353         # check if storage is available on both nodes
2354         my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2355         PVE::Storage::storage_check_node($storecfg, $sid, $node);
2356    });
2357 }
2358
2359 # list nodes where all VM images are available (used by has_feature API)
2360 sub shared_nodes {
2361     my ($conf, $storecfg) = @_;
2362
2363     my $nodelist = PVE::Cluster::get_nodelist();
2364     my $nodehash = { map { $_ => 1 } @$nodelist };
2365     my $nodename = nodename();
2366
2367     foreach_drive($conf, sub {
2368         my ($ds, $drive) = @_;
2369
2370         my $volid = $drive->{file};
2371         return if !$volid;
2372
2373         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2374         if ($storeid) {
2375             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2376             if ($scfg->{disable}) {
2377                 $nodehash = {};
2378             } elsif (my $avail = $scfg->{nodes}) {
2379                 foreach my $node (keys %$nodehash) {
2380                     delete $nodehash->{$node} if !$avail->{$node};
2381                 }
2382             } elsif (!$scfg->{shared}) {
2383                 foreach my $node (keys %$nodehash) {
2384                     delete $nodehash->{$node} if $node ne $nodename
2385                 }
2386             }
2387         }
2388     });
2389
2390     return $nodehash
2391 }
2392
2393 sub check_local_storage_availability {
2394     my ($conf, $storecfg) = @_;
2395
2396     my $nodelist = PVE::Cluster::get_nodelist();
2397     my $nodehash = { map { $_ => {} } @$nodelist };
2398
2399     foreach_drive($conf, sub {
2400         my ($ds, $drive) = @_;
2401
2402         my $volid = $drive->{file};
2403         return if !$volid;
2404
2405         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2406         if ($storeid) {
2407             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2408
2409             if ($scfg->{disable}) {
2410                 foreach my $node (keys %$nodehash) {
2411                     $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2412                 }
2413             } elsif (my $avail = $scfg->{nodes}) {
2414                 foreach my $node (keys %$nodehash) {
2415                     if (!$avail->{$node}) {
2416                         $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2417                     }
2418                 }
2419             }
2420         }
2421     });
2422
2423     foreach my $node (values %$nodehash) {
2424         if (my $unavail = $node->{unavailable_storages}) {
2425             $node->{unavailable_storages} = [ sort keys %$unavail ];
2426         }
2427     }
2428
2429     return $nodehash
2430 }
2431
2432 # Compat only, use assert_config_exists_on_node and vm_running_locally where possible
2433 sub check_running {
2434     my ($vmid, $nocheck, $node) = @_;
2435
2436     PVE::QemuConfig::assert_config_exists_on_node($vmid, $node) if !$nocheck;
2437     return PVE::QemuServer::Helpers::vm_running_locally($vmid);
2438 }
2439
2440 sub vzlist {
2441
2442     my $vzlist = config_list();
2443
2444     my $fd = IO::Dir->new($PVE::QemuServer::Helpers::var_run_tmpdir) || return $vzlist;
2445
2446     while (defined(my $de = $fd->read)) {
2447         next if $de !~ m/^(\d+)\.pid$/;
2448         my $vmid = $1;
2449         next if !defined($vzlist->{$vmid});
2450         if (my $pid = check_running($vmid)) {
2451             $vzlist->{$vmid}->{pid} = $pid;
2452         }
2453     }
2454
2455     return $vzlist;
2456 }
2457
2458 our $vmstatus_return_properties = {
2459     vmid => get_standard_option('pve-vmid'),
2460     status => {
2461         description => "Qemu process status.",
2462         type => 'string',
2463         enum => ['stopped', 'running'],
2464     },
2465     maxmem => {
2466         description => "Maximum memory in bytes.",
2467         type => 'integer',
2468         optional => 1,
2469         renderer => 'bytes',
2470     },
2471     maxdisk => {
2472         description => "Root disk size in bytes.",
2473         type => 'integer',
2474         optional => 1,
2475         renderer => 'bytes',
2476     },
2477     name => {
2478         description => "VM name.",
2479         type => 'string',
2480         optional => 1,
2481     },
2482     qmpstatus => {
2483         description => "Qemu QMP agent status.",
2484         type => 'string',
2485         optional => 1,
2486     },
2487     pid => {
2488         description => "PID of running qemu process.",
2489         type => 'integer',
2490         optional => 1,
2491     },
2492     uptime => {
2493         description => "Uptime.",
2494         type => 'integer',
2495         optional => 1,
2496         renderer => 'duration',
2497     },
2498     cpus => {
2499         description => "Maximum usable CPUs.",
2500         type => 'number',
2501         optional => 1,
2502     },
2503     lock => {
2504         description => "The current config lock, if any.",
2505         type => 'string',
2506         optional => 1,
2507     },
2508     tags => {
2509         description  => "The current configured tags, if any",
2510         type => 'string',
2511         optional => 1,
2512     },
2513 };
2514
2515 my $last_proc_pid_stat;
2516
2517 # get VM status information
2518 # This must be fast and should not block ($full == false)
2519 # We only query KVM using QMP if $full == true (this can be slow)
2520 sub vmstatus {
2521     my ($opt_vmid, $full) = @_;
2522
2523     my $res = {};
2524
2525     my $storecfg = PVE::Storage::config();
2526
2527     my $list = vzlist();
2528     my $defaults = load_defaults();
2529
2530     my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2531
2532     my $cpucount = $cpuinfo->{cpus} || 1;
2533
2534     foreach my $vmid (keys %$list) {
2535         next if $opt_vmid && ($vmid ne $opt_vmid);
2536
2537         my $conf = PVE::QemuConfig->load_config($vmid);
2538
2539         my $d = { vmid => $vmid };
2540         $d->{pid} = $list->{$vmid}->{pid};
2541
2542         # fixme: better status?
2543         $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2544
2545         my $size = PVE::QemuServer::Drive::bootdisk_size($storecfg, $conf);
2546         if (defined($size)) {
2547             $d->{disk} = 0; # no info available
2548             $d->{maxdisk} = $size;
2549         } else {
2550             $d->{disk} = 0;
2551             $d->{maxdisk} = 0;
2552         }
2553
2554         $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
2555             * ($conf->{cores} || $defaults->{cores});
2556         $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2557         $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2558
2559         $d->{name} = $conf->{name} || "VM $vmid";
2560         $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
2561             : $defaults->{memory}*(1024*1024);
2562
2563         if ($conf->{balloon}) {
2564             $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2565             $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
2566                 : $defaults->{shares};
2567         }
2568
2569         $d->{uptime} = 0;
2570         $d->{cpu} = 0;
2571         $d->{mem} = 0;
2572
2573         $d->{netout} = 0;
2574         $d->{netin} = 0;
2575
2576         $d->{diskread} = 0;
2577         $d->{diskwrite} = 0;
2578
2579         $d->{template} = PVE::QemuConfig->is_template($conf);
2580
2581         $d->{serial} = 1 if conf_has_serial($conf);
2582         $d->{lock} = $conf->{lock} if $conf->{lock};
2583         $d->{tags} = $conf->{tags} if defined($conf->{tags});
2584
2585         $res->{$vmid} = $d;
2586     }
2587
2588     my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2589     foreach my $dev (keys %$netdev) {
2590         next if $dev !~ m/^tap([1-9]\d*)i/;
2591         my $vmid = $1;
2592         my $d = $res->{$vmid};
2593         next if !$d;
2594
2595         $d->{netout} += $netdev->{$dev}->{receive};
2596         $d->{netin} += $netdev->{$dev}->{transmit};
2597
2598         if ($full) {
2599             $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
2600             $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
2601         }
2602
2603     }
2604
2605     my $ctime = gettimeofday;
2606
2607     foreach my $vmid (keys %$list) {
2608
2609         my $d = $res->{$vmid};
2610         my $pid = $d->{pid};
2611         next if !$pid;
2612
2613         my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2614         next if !$pstat; # not running
2615
2616         my $used = $pstat->{utime} + $pstat->{stime};
2617
2618         $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2619
2620         if ($pstat->{vsize}) {
2621             $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2622         }
2623
2624         my $old = $last_proc_pid_stat->{$pid};
2625         if (!$old) {
2626             $last_proc_pid_stat->{$pid} = {
2627                 time => $ctime,
2628                 used => $used,
2629                 cpu => 0,
2630             };
2631             next;
2632         }
2633
2634         my $dtime = ($ctime -  $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2635
2636         if ($dtime > 1000) {
2637             my $dutime = $used -  $old->{used};
2638
2639             $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2640             $last_proc_pid_stat->{$pid} = {
2641                 time => $ctime,
2642                 used => $used,
2643                 cpu => $d->{cpu},
2644             };
2645         } else {
2646             $d->{cpu} = $old->{cpu};
2647         }
2648     }
2649
2650     return $res if !$full;
2651
2652     my $qmpclient = PVE::QMPClient->new();
2653
2654     my $ballooncb = sub {
2655         my ($vmid, $resp) = @_;
2656
2657         my $info = $resp->{'return'};
2658         return if !$info->{max_mem};
2659
2660         my $d = $res->{$vmid};
2661
2662         # use memory assigned to VM
2663         $d->{maxmem} = $info->{max_mem};
2664         $d->{balloon} = $info->{actual};
2665
2666         if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2667             $d->{mem} = $info->{total_mem} - $info->{free_mem};
2668             $d->{freemem} = $info->{free_mem};
2669         }
2670
2671         $d->{ballooninfo} = $info;
2672     };
2673
2674     my $blockstatscb = sub {
2675         my ($vmid, $resp) = @_;
2676         my $data = $resp->{'return'} || [];
2677         my $totalrdbytes = 0;
2678         my $totalwrbytes = 0;
2679
2680         for my $blockstat (@$data) {
2681             $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2682             $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2683
2684             $blockstat->{device} =~ s/drive-//;
2685             $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
2686         }
2687         $res->{$vmid}->{diskread} = $totalrdbytes;
2688         $res->{$vmid}->{diskwrite} = $totalwrbytes;
2689     };
2690
2691     my $statuscb = sub {
2692         my ($vmid, $resp) = @_;
2693
2694         $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2695         # this fails if ballon driver is not loaded, so this must be
2696         # the last commnand (following command are aborted if this fails).
2697         $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2698
2699         my $status = 'unknown';
2700         if (!defined($status = $resp->{'return'}->{status})) {
2701             warn "unable to get VM status\n";
2702             return;
2703         }
2704
2705         $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2706     };
2707
2708     foreach my $vmid (keys %$list) {
2709         next if $opt_vmid && ($vmid ne $opt_vmid);
2710         next if !$res->{$vmid}->{pid}; # not running
2711         $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2712     }
2713
2714     $qmpclient->queue_execute(undef, 2);
2715
2716     foreach my $vmid (keys %$list) {
2717         next if $opt_vmid && ($vmid ne $opt_vmid);
2718         $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2719     }
2720
2721     return $res;
2722 }
2723
2724 sub conf_has_serial {
2725     my ($conf) = @_;
2726
2727     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
2728         if ($conf->{"serial$i"}) {
2729             return 1;
2730         }
2731     }
2732
2733     return 0;
2734 }
2735
2736 sub conf_has_audio {
2737     my ($conf, $id) = @_;
2738
2739     $id //= 0;
2740     my $audio = $conf->{"audio$id"};
2741     return undef if !defined($audio);
2742
2743     my $audioproperties = PVE::JSONSchema::parse_property_string($audio_fmt, $audio);
2744     my $audiodriver = $audioproperties->{driver} // 'spice';
2745
2746     return {
2747         dev => $audioproperties->{device},
2748         dev_id => "audiodev$id",
2749         backend => $audiodriver,
2750         backend_id => "$audiodriver-backend${id}",
2751     };
2752 }
2753
2754 sub vga_conf_has_spice {
2755     my ($vga) = @_;
2756
2757     my $vgaconf = parse_vga($vga);
2758     my $vgatype = $vgaconf->{type};
2759     return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
2760
2761     return $1 || 1;
2762 }
2763
2764 sub is_native($) {
2765     my ($arch) = @_;
2766     return get_host_arch() eq $arch;
2767 }
2768
2769 sub get_vm_arch {
2770     my ($conf) = @_;
2771     return $conf->{arch} // get_host_arch();
2772 }
2773
2774 my $default_machines = {
2775     x86_64 => 'pc',
2776     aarch64 => 'virt',
2777 };
2778
2779 sub get_vm_machine {
2780     my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
2781
2782     my $machine = $forcemachine || $conf->{machine};
2783
2784     if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
2785         $arch //= 'x86_64';
2786         $machine ||= $default_machines->{$arch};
2787         if ($add_pve_version) {
2788             $kvmversion //= kvm_user_version();
2789             my $pvever = PVE::QemuServer::Machine::get_pve_version($kvmversion);
2790             $machine .= "+pve$pvever";
2791         }
2792     }
2793
2794     if ($add_pve_version && $machine !~ m/\+pve\d+$/) {
2795         # for version-pinned machines that do not include a pve-version (e.g.
2796         # pc-q35-4.1), we assume 0 to keep them stable in case we bump
2797         $machine .= '+pve0';
2798     }
2799
2800     return $machine;
2801 }
2802
2803 sub get_ovmf_files($) {
2804     my ($arch) = @_;
2805
2806     my $ovmf = $OVMF->{$arch}
2807         or die "no OVMF images known for architecture '$arch'\n";
2808
2809     return @$ovmf;
2810 }
2811
2812 my $Arch2Qemu = {
2813     aarch64 => '/usr/bin/qemu-system-aarch64',
2814     x86_64 => '/usr/bin/qemu-system-x86_64',
2815 };
2816 sub get_command_for_arch($) {
2817     my ($arch) = @_;
2818     return '/usr/bin/kvm' if is_native($arch);
2819
2820     my $cmd = $Arch2Qemu->{$arch}
2821         or die "don't know how to emulate architecture '$arch'\n";
2822     return $cmd;
2823 }
2824
2825 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
2826 # to use in a QEMU command line (-cpu element), first array_intersect the result
2827 # of query_supported_ with query_understood_. This is necessary because:
2828 #
2829 # a) query_understood_ returns flags the host cannot use and
2830 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
2831 #    flags, but CPU settings - with most of them being flags. Those settings
2832 #    (and some flags, curiously) cannot be specified as a "-cpu" argument.
2833 #
2834 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
2835 # expensive. If you need the value returned from this, you can get it much
2836 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
2837 # $accel being 'kvm' or 'tcg'.
2838 #
2839 # pvestatd calls this function on startup and whenever the QEMU/KVM version
2840 # changes, automatically populating pmxcfs.
2841 #
2842 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
2843 # since kvm and tcg machines support different flags
2844 #
2845 sub query_supported_cpu_flags {
2846     my ($arch) = @_;
2847
2848     $arch //= get_host_arch();
2849     my $default_machine = $default_machines->{$arch};
2850
2851     my $flags = {};
2852
2853     # FIXME: Once this is merged, the code below should work for ARM as well:
2854     # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
2855     die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
2856         $arch eq "aarch64";
2857
2858     my $kvm_supported = defined(kvm_version());
2859     my $qemu_cmd = get_command_for_arch($arch);
2860     my $fakevmid = -1;
2861     my $pidfile = PVE::QemuServer::Helpers::pidfile_name($fakevmid);
2862
2863     # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
2864     my $query_supported_run_qemu = sub {
2865         my ($kvm) = @_;
2866
2867         my $flags = {};
2868         my $cmd = [
2869             $qemu_cmd,
2870             '-machine', $default_machine,
2871             '-display', 'none',
2872             '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server,nowait",
2873             '-mon', 'chardev=qmp,mode=control',
2874             '-pidfile', $pidfile,
2875             '-S', '-daemonize'
2876         ];
2877
2878         if (!$kvm) {
2879             push @$cmd, '-accel', 'tcg';
2880         }
2881
2882         my $rc = run_command($cmd, noerr => 1, quiet => 0);
2883         die "QEMU flag querying VM exited with code " . $rc if $rc;
2884
2885         eval {
2886             my $cmd_result = mon_cmd(
2887                 $fakevmid,
2888                 'query-cpu-model-expansion',
2889                 type => 'full',
2890                 model => { name => 'host' }
2891             );
2892
2893             my $props = $cmd_result->{model}->{props};
2894             foreach my $prop (keys %$props) {
2895                 next if $props->{$prop} ne '1';
2896                 # QEMU returns some flags multiple times, with '_', '.' or '-'
2897                 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
2898                 # We only keep those with underscores, to match /proc/cpuinfo
2899                 $prop =~ s/\.|-/_/g;
2900                 $flags->{$prop} = 1;
2901             }
2902         };
2903         my $err = $@;
2904
2905         # force stop with 10 sec timeout and 'nocheck'
2906         # always stop, even if QMP failed
2907         vm_stop(undef, $fakevmid, 1, 1, 10, 0, 1);
2908
2909         die $err if $err;
2910
2911         return [ sort keys %$flags ];
2912     };
2913
2914     # We need to query QEMU twice, since KVM and TCG have different supported flags
2915     PVE::QemuConfig->lock_config($fakevmid, sub {
2916         $flags->{tcg} = eval { $query_supported_run_qemu->(0) };
2917         warn "warning: failed querying supported tcg flags: $@\n" if $@;
2918
2919         if ($kvm_supported) {
2920             $flags->{kvm} = eval { $query_supported_run_qemu->(1) };
2921             warn "warning: failed querying supported kvm flags: $@\n" if $@;
2922         }
2923     });
2924
2925     return $flags;
2926 }
2927
2928 # Understood CPU flags are written to a file at 'pve-qemu' compile time
2929 my $understood_cpu_flag_dir = "/usr/share/kvm";
2930 sub query_understood_cpu_flags {
2931     my $arch = get_host_arch();
2932     my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
2933
2934     die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
2935         if ! -e $filepath;
2936
2937     my $raw = file_get_contents($filepath);
2938     $raw =~ s/^\s+|\s+$//g;
2939     my @flags = split(/\s+/, $raw);
2940
2941     return \@flags;
2942 }
2943
2944 sub config_to_command {
2945     my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu) = @_;
2946
2947     my $cmd = [];
2948     my $globalFlags = [];
2949     my $machineFlags = [];
2950     my $rtcFlags = [];
2951     my $devices = [];
2952     my $pciaddr = '';
2953     my $bridges = {};
2954     my $ostype = $conf->{ostype};
2955     my $winversion = windows_version($ostype);
2956     my $kvm = $conf->{kvm};
2957     my $nodename = nodename();
2958
2959     my $arch = get_vm_arch($conf);
2960     my $kvm_binary = get_command_for_arch($arch);
2961     my $kvmver = kvm_user_version($kvm_binary);
2962
2963     if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
2964         $kvmver //= "undefined";
2965         die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
2966     }
2967
2968     my $add_pve_version = min_version($kvmver, 4, 1);
2969
2970     my $machine_type = get_vm_machine($conf, $forcemachine, $arch, $add_pve_version);
2971     my $machine_version = PVE::QemuServer::Machine::extract_version($machine_type, $kvmver);
2972     $kvm //= 1 if is_native($arch);
2973
2974     $machine_version =~ m/(\d+)\.(\d+)/;
2975     my ($machine_major, $machine_minor) = ($1, $2);
2976
2977     if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
2978         warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
2979     } elsif (!min_version($kvmver, $machine_major, $machine_minor)) {
2980         die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type', please upgrade node '$nodename'\n"
2981     } elsif (!PVE::QemuServer::Machine::can_run_pve_machine_version($machine_version, $kvmver)) {
2982         my $max_pve_version = PVE::QemuServer::Machine::get_pve_version($machine_version);
2983         die "Installed qemu-server (max feature level for $machine_major.$machine_minor is pve$max_pve_version)"
2984             ." is too old to run machine type '$machine_type', please upgrade node '$nodename'\n";
2985     }
2986
2987     # if a specific +pve version is required for a feature, use $version_guard
2988     # instead of min_version to allow machines to be run with the minimum
2989     # required version
2990     my $required_pve_version = 0;
2991     my $version_guard = sub {
2992         my ($major, $minor, $pve) = @_;
2993         return 0 if !min_version($machine_version, $major, $minor, $pve);
2994         my $max_pve = PVE::QemuServer::Machine::get_pve_version("$major.$minor");
2995         return 1 if min_version($machine_version, $major, $minor, $max_pve+1);
2996         $required_pve_version = $pve if $pve && $pve > $required_pve_version;
2997         return 1;
2998     };
2999
3000     if ($kvm) {
3001         die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3002             if !defined kvm_version();
3003     }
3004
3005     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3006     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3007     my $use_old_bios_files = undef;
3008     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3009
3010     my $cpuunits = defined($conf->{cpuunits}) ?
3011             $conf->{cpuunits} : $defaults->{cpuunits};
3012
3013     push @$cmd, $kvm_binary;
3014
3015     push @$cmd, '-id', $vmid;
3016
3017     my $vmname = $conf->{name} || "vm$vmid";
3018
3019     push @$cmd, '-name', $vmname;
3020
3021     my $use_virtio = 0;
3022
3023     my $qmpsocket = PVE::QemuServer::Helpers::qmp_socket($vmid);
3024     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3025     push @$cmd, '-mon', "chardev=qmp,mode=control";
3026
3027     if (min_version($machine_version, 2, 12)) {
3028         push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3029         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3030     }
3031
3032     push @$cmd, '-pidfile' , PVE::QemuServer::Helpers::pidfile_name($vmid);
3033
3034     push @$cmd, '-daemonize';
3035
3036     if ($conf->{smbios1}) {
3037         my $smbios_conf = parse_smbios1($conf->{smbios1});
3038         if ($smbios_conf->{base64}) {
3039             # Do not pass base64 flag to qemu
3040             delete $smbios_conf->{base64};
3041             my $smbios_string = "";
3042             foreach my $key (keys %$smbios_conf) {
3043                 my $value;
3044                 if ($key eq "uuid") {
3045                     $value = $smbios_conf->{uuid}
3046                 } else {
3047                     $value = decode_base64($smbios_conf->{$key});
3048                 }
3049                 # qemu accepts any binary data, only commas need escaping by double comma
3050                 $value =~ s/,/,,/g;
3051                 $smbios_string .= "," . $key . "=" . $value if $value;
3052             }
3053             push @$cmd, '-smbios', "type=1" . $smbios_string;
3054         } else {
3055             push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3056         }
3057     }
3058
3059     my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3060     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3061         die "uefi base image not found\n" if ! -f $ovmf_code;
3062
3063         my $path;
3064         my $format;
3065         if (my $efidisk = $conf->{efidisk0}) {
3066             my $d = parse_drive('efidisk0', $efidisk);
3067             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3068             $format = $d->{format};
3069             if ($storeid) {
3070                 $path = PVE::Storage::path($storecfg, $d->{file});
3071                 if (!defined($format)) {
3072                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3073                     $format = qemu_img_format($scfg, $volname);
3074                 }
3075             } else {
3076                 $path = $d->{file};
3077                 die "efidisk format must be specified\n"
3078                     if !defined($format);
3079             }
3080         } else {
3081             warn "no efidisk configured! Using temporary efivars disk.\n";
3082             $path = "/tmp/$vmid-ovmf.fd";
3083             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3084             $format = 'raw';
3085         }
3086
3087         my $size_str = "";
3088
3089         if ($format eq 'raw' && $version_guard->(4, 1, 2)) {
3090             $size_str = ",size=" . (-s $ovmf_vars);
3091         }
3092
3093         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3094         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0$size_str,file=$path";
3095     }
3096
3097     # load q35 config
3098     if ($q35) {
3099         # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3100         if (min_version($machine_version, 4, 0)) {
3101             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3102         } else {
3103             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3104         }
3105     }
3106
3107     if ($conf->{vmgenid}) {
3108         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3109     }
3110
3111     # add usb controllers
3112     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3113     push @$devices, @usbcontrollers if @usbcontrollers;
3114     my $vga = parse_vga($conf->{vga});
3115
3116     my $qxlnum = vga_conf_has_spice($conf->{vga});
3117     $vga->{type} = 'qxl' if $qxlnum;
3118
3119     if (!$vga->{type}) {
3120         if ($arch eq 'aarch64') {
3121             $vga->{type} = 'virtio';
3122         } elsif (min_version($machine_version, 2, 9)) {
3123             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3124         } else {
3125             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3126         }
3127     }
3128
3129     # enable absolute mouse coordinates (needed by vnc)
3130     my $tablet;
3131     if (defined($conf->{tablet})) {
3132         $tablet = $conf->{tablet};
3133     } else {
3134         $tablet = $defaults->{tablet};
3135         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3136         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3137     }
3138
3139     if ($tablet) {
3140         push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3141         my $kbd = print_keyboarddevice_full($conf, $arch);
3142         push @$devices, '-device', $kbd if defined($kbd);
3143     }
3144
3145     my $kvm_off = 0;
3146     my $gpu_passthrough;
3147
3148     # host pci devices
3149     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
3150         my $id = "hostpci$i";
3151         my $d = parse_hostpci($conf->{$id});
3152         next if !$d;
3153
3154         if (my $pcie = $d->{pcie}) {
3155             die "q35 machine model is not enabled" if !$q35;
3156             # win7 wants to have the pcie devices directly on the pcie bus
3157             # instead of in the root port
3158             if ($winversion == 7) {
3159                 $pciaddr = print_pcie_addr("${id}bus0");
3160             } else {
3161                 # add more root ports if needed, 4 are present by default
3162                 # by pve-q35 cfgs, rest added here on demand.
3163                 if ($i > 3) {
3164                     push @$devices, '-device', print_pcie_root_port($i);
3165                 }
3166                 $pciaddr = print_pcie_addr($id);
3167             }
3168         } else {
3169             $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3170         }
3171
3172         my $xvga = '';
3173         if ($d->{'x-vga'}) {
3174             $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3175             $kvm_off = 1;
3176             $vga->{type} = 'none' if !defined($conf->{vga});
3177             $gpu_passthrough = 1;
3178         }
3179
3180         my $pcidevices = $d->{pciid};
3181         my $multifunction = 1 if @$pcidevices > 1;
3182
3183         my $sysfspath;
3184         if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3185             my $pci_id = $pcidevices->[0]->{id};
3186             my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3187             $sysfspath = "/sys/bus/pci/devices/$pci_id/$uuid";
3188         } elsif ($d->{mdev}) {
3189             warn "ignoring mediated device '$id' with multifunction device\n";
3190         }
3191
3192         my $j=0;
3193         foreach my $pcidevice (@$pcidevices) {
3194             my $devicestr = "vfio-pci";
3195
3196             if ($sysfspath) {
3197                 $devicestr .= ",sysfsdev=$sysfspath";
3198             } else {
3199                 $devicestr .= ",host=$pcidevice->{id}";
3200             }
3201
3202             my $mf_addr = $multifunction ? ".$j" : '';
3203             $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3204
3205             if ($j == 0) {
3206                 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3207                 $devicestr .= "$xvga";
3208                 $devicestr .= ",multifunction=on" if $multifunction;
3209                 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3210             }
3211
3212             push @$devices, '-device', $devicestr;
3213             $j++;
3214         }
3215     }
3216
3217     # usb devices
3218     my $usb_dev_features = {};
3219     $usb_dev_features->{spice_usb3} = 1 if min_version($machine_version, 4, 0);
3220
3221     my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3222     push @$devices, @usbdevices if @usbdevices;
3223     # serial devices
3224     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3225         if (my $path = $conf->{"serial$i"}) {
3226             if ($path eq 'socket') {
3227                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3228                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3229                 # On aarch64, serial0 is the UART device. Qemu only allows
3230                 # connecting UART devices via the '-serial' command line, as
3231                 # the device has a fixed slot on the hardware...
3232                 if ($arch eq 'aarch64' && $i == 0) {
3233                     push @$devices, '-serial', "chardev:serial$i";
3234                 } else {
3235                     push @$devices, '-device', "isa-serial,chardev=serial$i";
3236                 }
3237             } else {
3238                 die "no such serial device\n" if ! -c $path;
3239                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3240                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3241             }
3242         }
3243     }
3244
3245     # parallel devices
3246     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3247         if (my $path = $conf->{"parallel$i"}) {
3248             die "no such parallel device\n" if ! -c $path;
3249             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3250             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3251             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3252         }
3253     }
3254
3255     if (my $audio = conf_has_audio($conf)) {
3256
3257         my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3258
3259         my $id = $audio->{dev_id};
3260         if ($audio->{dev} eq 'AC97') {
3261             push @$devices, '-device', "AC97,id=${id}${audiopciaddr}";
3262         } elsif ($audio->{dev} =~ /intel\-hda$/) {
3263             push @$devices, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3264             push @$devices, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0";
3265             push @$devices, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1";
3266         } else {
3267             die "unkown audio device '$audio->{dev}', implement me!";
3268         }
3269
3270         push @$devices, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3271     }
3272
3273     my $sockets = 1;
3274     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3275     $sockets = $conf->{sockets} if  $conf->{sockets};
3276
3277     my $cores = $conf->{cores} || 1;
3278
3279     my $maxcpus = $sockets * $cores;
3280
3281     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3282
3283     my $allowed_vcpus = $cpuinfo->{cpus};
3284
3285     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3286         if ($allowed_vcpus < $maxcpus);
3287
3288     if($hotplug_features->{cpu} && min_version($machine_version, 2, 7)) {
3289
3290         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3291         for (my $i = 2; $i <= $vcpus; $i++)  {
3292             my $cpustr = print_cpu_device($conf,$i);
3293             push @$cmd, '-device', $cpustr;
3294         }
3295
3296     } else {
3297
3298         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3299     }
3300     push @$cmd, '-nodefaults';
3301
3302     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3303
3304     my $bootindex_hash = {};
3305     my $i = 1;
3306     foreach my $o (split(//, $bootorder)) {
3307         $bootindex_hash->{$o} = $i*100;
3308         $i++;
3309     }
3310
3311     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3312
3313     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3314
3315     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3316
3317     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3318         push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3319         my $socket = PVE::QemuServer::Helpers::vnc_socket($vmid);
3320         push @$cmd,  '-vnc', "unix:$socket,password";
3321     } else {
3322         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3323         push @$cmd, '-nographic';
3324     }
3325
3326     # time drift fix
3327     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3328
3329     my $useLocaltime = $conf->{localtime};
3330
3331     if ($winversion >= 5) { # windows
3332         $useLocaltime = 1 if !defined($conf->{localtime});
3333
3334         # use time drift fix when acpi is enabled
3335         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3336             $tdf = 1 if !defined($conf->{tdf});
3337         }
3338     }
3339
3340     if ($winversion >= 6) {
3341         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3342         push @$cmd, '-no-hpet';
3343     }
3344
3345     push @$rtcFlags, 'driftfix=slew' if $tdf;
3346
3347     if (($conf->{startdate}) && ($conf->{startdate} ne 'now')) {
3348         push @$rtcFlags, "base=$conf->{startdate}";
3349     } elsif ($useLocaltime) {
3350         push @$rtcFlags, 'base=localtime';
3351     }
3352
3353     if ($forcecpu) {
3354         push @$cmd, '-cpu', $forcecpu;
3355     } else {
3356         push @$cmd, get_cpu_options($conf, $arch, $kvm, $kvm_off,
3357             $machine_version, $winversion, $gpu_passthrough);
3358     }
3359
3360     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3361
3362     push @$cmd, '-S' if $conf->{freeze};
3363
3364     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3365
3366     my $guest_agent = parse_guest_agent($conf);
3367
3368     if ($guest_agent->{enabled}) {
3369         my $qgasocket = PVE::QemuServer::Helpers::qmp_socket($vmid, 1);
3370         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3371
3372         if (!$guest_agent->{type} || $guest_agent->{type} eq 'virtio') {
3373             my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3374             push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3375             push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3376         } elsif ($guest_agent->{type} eq 'isa') {
3377             push @$devices, '-device', "isa-serial,chardev=qga0";
3378         }
3379     }
3380
3381     my $rng = parse_rng($conf->{rng0}) if $conf->{rng0};
3382     if ($rng && &$version_guard(4, 1, 2)) {
3383         my $max_bytes = $rng->{max_bytes} // $rng_fmt->{max_bytes}->{default};
3384         my $period = $rng->{period} // $rng_fmt->{period}->{default};
3385
3386         my $limiter_str = "";
3387         if ($max_bytes) {
3388             $limiter_str = ",max-bytes=$max_bytes,period=$period";
3389         }
3390
3391         # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
3392         die "cannot create VirtIO RNG device: source file '$rng->{source}' doesn't exist\n"
3393             if ! -e $rng->{source};
3394
3395         my $rng_addr = print_pci_addr("rng0", $bridges, $arch, $machine_type);
3396
3397         push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3398         push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3399     }
3400
3401     my $spice_port;
3402
3403     if ($qxlnum) {
3404         if ($qxlnum > 1) {
3405             if ($winversion){
3406                 for(my $i = 1; $i < $qxlnum; $i++){
3407                     push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3408                 }
3409             } else {
3410                 # assume other OS works like Linux
3411                 my ($ram, $vram) = ("134217728", "67108864");
3412                 if ($vga->{memory}) {
3413                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3414                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3415                 }
3416                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3417                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3418             }
3419         }
3420
3421         my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
3422
3423         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3424         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3425         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3426
3427         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3428         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3429         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3430
3431         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3432         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3433
3434         my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
3435         if ($spice_enhancement->{foldersharing}) {
3436             push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3437             push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3438         }
3439
3440         my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3441         $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
3442         push @$devices, '-spice', "$spice_opts";
3443     }
3444
3445     # enable balloon by default, unless explicitly disabled
3446     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3447         $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
3448         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3449     }
3450
3451     if ($conf->{watchdog}) {
3452         my $wdopts = parse_watchdog($conf->{watchdog});
3453         $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
3454         my $watchdog = $wdopts->{model} || 'i6300esb';
3455         push @$devices, '-device', "$watchdog$pciaddr";
3456         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3457     }
3458
3459     my $vollist = [];
3460     my $scsicontroller = {};
3461     my $ahcicontroller = {};
3462     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3463
3464     # Add iscsi initiator name if available
3465     if (my $initiator = get_initiator_name()) {
3466         push @$devices, '-iscsi', "initiator-name=$initiator";
3467     }
3468
3469     foreach_drive($conf, sub {
3470         my ($ds, $drive) = @_;
3471
3472         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3473             push @$vollist, $drive->{file};
3474         }
3475
3476         # ignore efidisk here, already added in bios/fw handling code above
3477         return if $drive->{interface} eq 'efidisk';
3478
3479         $use_virtio = 1 if $ds =~ m/^virtio/;
3480
3481         if (drive_is_cdrom ($drive)) {
3482             if ($bootindex_hash->{d}) {
3483                 $drive->{bootindex} = $bootindex_hash->{d};
3484                 $bootindex_hash->{d} += 1;
3485             }
3486         } else {
3487             if ($bootindex_hash->{c}) {
3488                 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3489                 $bootindex_hash->{c} += 1;
3490             }
3491         }
3492
3493         if($drive->{interface} eq 'virtio'){
3494            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3495         }
3496
3497         if ($drive->{interface} eq 'scsi') {
3498
3499             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3500
3501             die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
3502                 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
3503
3504             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
3505             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3506
3507             my $iothread = '';
3508             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3509                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3510                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3511             } elsif ($drive->{iothread}) {
3512                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3513             }
3514
3515             my $queues = '';
3516             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3517                 $queues = ",num_queues=$drive->{queues}";
3518             }
3519
3520             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3521             $scsicontroller->{$controller}=1;
3522         }
3523
3524         if ($drive->{interface} eq 'sata') {
3525            my $controller = int($drive->{index} / $PVE::QemuServer::Drive::MAX_SATA_DISKS);
3526            $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
3527            push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3528            $ahcicontroller->{$controller}=1;
3529         }
3530
3531         my $drive_cmd = print_drive_commandline_full($storecfg, $vmid, $drive);
3532         push @$devices, '-drive',$drive_cmd;
3533         push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
3534     });
3535
3536     for (my $i = 0; $i < $MAX_NETS; $i++) {
3537          next if !$conf->{"net$i"};
3538          my $d = parse_net($conf->{"net$i"});
3539          next if !$d;
3540
3541          $use_virtio = 1 if $d->{model} eq 'virtio';
3542
3543          if ($bootindex_hash->{n}) {
3544             $d->{bootindex} = $bootindex_hash->{n};
3545             $bootindex_hash->{n} += 1;
3546          }
3547
3548          my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
3549          push @$devices, '-netdev', $netdevfull;
3550
3551          my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
3552          push @$devices, '-device', $netdevicefull;
3553     }
3554
3555     if ($conf->{ivshmem}) {
3556         my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
3557
3558         my $bus;
3559         if ($q35) {
3560             $bus = print_pcie_addr("ivshmem");
3561         } else {
3562             $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
3563         }
3564
3565         my $ivshmem_name = $ivshmem->{name} // $vmid;
3566         my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
3567
3568         push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
3569         push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
3570     }
3571
3572     # pci.4 is nested in pci.1
3573     $bridges->{1} = 1 if $bridges->{4};
3574
3575     if (!$q35) {
3576         # add pci bridges
3577         if (min_version($machine_version, 2, 3)) {
3578            $bridges->{1} = 1;
3579            $bridges->{2} = 1;
3580         }
3581
3582         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3583
3584     }
3585
3586     for my $k (sort {$b cmp $a} keys %$bridges) {
3587         next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
3588         $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
3589         my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
3590         if ($q35) {
3591             # add after -readconfig pve-q35.cfg
3592             splice @$devices, 2, 0, '-device', $devstr;
3593         } else {
3594             unshift @$devices, '-device', $devstr if $k > 0;
3595         }
3596     }
3597
3598     if (!$kvm) {
3599         push @$machineFlags, 'accel=tcg';
3600     }
3601
3602     my $machine_type_min = $machine_type;
3603     if ($add_pve_version) {
3604         $machine_type_min =~ s/\+pve\d+$//;
3605         $machine_type_min .= "+pve$required_pve_version";
3606     }
3607     push @$machineFlags, "type=${machine_type_min}";
3608
3609     push @$cmd, @$devices;
3610     push @$cmd, '-rtc', join(',', @$rtcFlags)
3611         if scalar(@$rtcFlags);
3612     push @$cmd, '-machine', join(',', @$machineFlags)
3613         if scalar(@$machineFlags);
3614     push @$cmd, '-global', join(',', @$globalFlags)
3615         if scalar(@$globalFlags);
3616
3617     if (my $vmstate = $conf->{vmstate}) {
3618         my $statepath = PVE::Storage::path($storecfg, $vmstate);
3619         push @$vollist, $vmstate;
3620         push @$cmd, '-loadstate', $statepath;
3621         print "activating and using '$vmstate' as vmstate\n";
3622     }
3623
3624     # add custom args
3625     if ($conf->{args}) {
3626         my $aa = PVE::Tools::split_args($conf->{args});
3627         push @$cmd, @$aa;
3628     }
3629
3630     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3631 }
3632
3633 sub spice_port {
3634     my ($vmid) = @_;
3635
3636     my $res = mon_cmd($vmid, 'query-spice');
3637
3638     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3639 }
3640
3641 sub vm_devices_list {
3642     my ($vmid) = @_;
3643
3644     my $res = mon_cmd($vmid, 'query-pci');
3645     my $devices_to_check = [];
3646     my $devices = {};
3647     foreach my $pcibus (@$res) {
3648         push @$devices_to_check, @{$pcibus->{devices}},
3649     }
3650
3651     while (@$devices_to_check) {
3652         my $to_check = [];
3653         for my $d (@$devices_to_check) {
3654             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3655             next if !$d->{'pci_bridge'};
3656
3657             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3658             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3659         }
3660         $devices_to_check = $to_check;
3661     }
3662
3663     my $resblock = mon_cmd($vmid, 'query-block');
3664     foreach my $block (@$resblock) {
3665         if($block->{device} =~ m/^drive-(\S+)/){
3666                 $devices->{$1} = 1;
3667         }
3668     }
3669
3670     my $resmice = mon_cmd($vmid, 'query-mice');
3671     foreach my $mice (@$resmice) {
3672         if ($mice->{name} eq 'QEMU HID Tablet') {
3673             $devices->{tablet} = 1;
3674             last;
3675         }
3676     }
3677
3678     # for usb devices there is no query-usb
3679     # but we can iterate over the entries in
3680     # qom-list path=/machine/peripheral
3681     my $resperipheral = mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3682     foreach my $per (@$resperipheral) {
3683         if ($per->{name} =~ m/^usb\d+$/) {
3684             $devices->{$per->{name}} = 1;
3685         }
3686     }
3687
3688     return $devices;
3689 }
3690
3691 sub vm_deviceplug {
3692     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
3693
3694     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3695
3696     my $devices_list = vm_devices_list($vmid);
3697     return 1 if defined($devices_list->{$deviceid});
3698
3699     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
3700
3701     if ($deviceid eq 'tablet') {
3702
3703         qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
3704
3705     } elsif ($deviceid eq 'keyboard') {
3706
3707         qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
3708
3709     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3710
3711         die "usb hotplug currently not reliable\n";
3712         # since we can't reliably hot unplug all added usb devices
3713         # and usb passthrough disables live migration
3714         # we disable usb hotplugging for now
3715         qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3716
3717     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3718
3719         qemu_iothread_add($vmid, $deviceid, $device);
3720
3721         qemu_driveadd($storecfg, $vmid, $device);
3722         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
3723
3724         qemu_deviceadd($vmid, $devicefull);
3725         eval { qemu_deviceaddverify($vmid, $deviceid); };
3726         if (my $err = $@) {
3727             eval { qemu_drivedel($vmid, $deviceid); };
3728             warn $@ if $@;
3729             die $err;
3730         }
3731
3732     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3733
3734
3735         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3736         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3737         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3738
3739         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3740
3741         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3742             qemu_iothread_add($vmid, $deviceid, $device);
3743             $devicefull .= ",iothread=iothread-$deviceid";
3744         }
3745
3746         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3747             $devicefull .= ",num_queues=$device->{queues}";
3748         }
3749
3750         qemu_deviceadd($vmid, $devicefull);
3751         qemu_deviceaddverify($vmid, $deviceid);
3752
3753     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3754
3755         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
3756         qemu_driveadd($storecfg, $vmid, $device);
3757
3758         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
3759         eval { qemu_deviceadd($vmid, $devicefull); };
3760         if (my $err = $@) {
3761             eval { qemu_drivedel($vmid, $deviceid); };
3762             warn $@ if $@;
3763             die $err;
3764         }
3765
3766     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3767
3768         return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
3769
3770         my $machine_type = PVE::QemuServer::Machine::qemu_machine_pxe($vmid, $conf);
3771         my $use_old_bios_files = undef;
3772         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3773
3774         my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
3775         qemu_deviceadd($vmid, $netdevicefull);
3776         eval {
3777             qemu_deviceaddverify($vmid, $deviceid);
3778             qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
3779         };
3780         if (my $err = $@) {
3781             eval { qemu_netdevdel($vmid, $deviceid); };
3782             warn $@ if $@;
3783             die $err;
3784         }
3785
3786     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3787
3788         my $bridgeid = $2;
3789         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3790         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3791
3792         qemu_deviceadd($vmid, $devicefull);
3793         qemu_deviceaddverify($vmid, $deviceid);
3794
3795     } else {
3796         die "can't hotplug device '$deviceid'\n";
3797     }
3798
3799     return 1;
3800 }
3801
3802 # fixme: this should raise exceptions on error!
3803 sub vm_deviceunplug {
3804     my ($vmid, $conf, $deviceid) = @_;
3805
3806     my $devices_list = vm_devices_list($vmid);
3807     return 1 if !defined($devices_list->{$deviceid});
3808
3809     die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3810
3811     if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
3812
3813         qemu_devicedel($vmid, $deviceid);
3814
3815     } elsif ($deviceid =~ m/^usb\d+$/) {
3816
3817         die "usb hotplug currently not reliable\n";
3818         # when unplugging usb devices this way,
3819         # there may be remaining usb controllers/hubs
3820         # so we disable it for now
3821         qemu_devicedel($vmid, $deviceid);
3822         qemu_devicedelverify($vmid, $deviceid);
3823
3824     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3825
3826         qemu_devicedel($vmid, $deviceid);
3827         qemu_devicedelverify($vmid, $deviceid);
3828         qemu_drivedel($vmid, $deviceid);
3829         qemu_iothread_del($conf, $vmid, $deviceid);
3830
3831     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3832
3833         qemu_devicedel($vmid, $deviceid);
3834         qemu_devicedelverify($vmid, $deviceid);
3835         qemu_iothread_del($conf, $vmid, $deviceid);
3836
3837     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3838
3839         qemu_devicedel($vmid, $deviceid);
3840         qemu_drivedel($vmid, $deviceid);
3841         qemu_deletescsihw($conf, $vmid, $deviceid);
3842
3843     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3844
3845         qemu_devicedel($vmid, $deviceid);
3846         qemu_devicedelverify($vmid, $deviceid);
3847         qemu_netdevdel($vmid, $deviceid);
3848
3849     } else {
3850         die "can't unplug device '$deviceid'\n";
3851     }
3852
3853     return 1;
3854 }
3855
3856 sub qemu_deviceadd {
3857     my ($vmid, $devicefull) = @_;
3858
3859     $devicefull = "driver=".$devicefull;
3860     my %options =  split(/[=,]/, $devicefull);
3861
3862     mon_cmd($vmid, "device_add" , %options);
3863 }
3864
3865 sub qemu_devicedel {
3866     my ($vmid, $deviceid) = @_;
3867
3868     my $ret = mon_cmd($vmid, "device_del", id => $deviceid);
3869 }
3870
3871 sub qemu_iothread_add {
3872     my($vmid, $deviceid, $device) = @_;
3873
3874     if ($device->{iothread}) {
3875         my $iothreads = vm_iothreads_list($vmid);
3876         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3877     }
3878 }
3879
3880 sub qemu_iothread_del {
3881     my($conf, $vmid, $deviceid) = @_;
3882
3883     my $confid = $deviceid;
3884     if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
3885         $confid = 'scsi' . $1;
3886     }
3887     my $device = parse_drive($confid, $conf->{$confid});
3888     if ($device->{iothread}) {
3889         my $iothreads = vm_iothreads_list($vmid);
3890         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3891     }
3892 }
3893
3894 sub qemu_objectadd {
3895     my($vmid, $objectid, $qomtype) = @_;
3896
3897     mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3898
3899     return 1;
3900 }
3901
3902 sub qemu_objectdel {
3903     my($vmid, $objectid) = @_;
3904
3905     mon_cmd($vmid, "object-del", id => $objectid);
3906
3907     return 1;
3908 }
3909
3910 sub qemu_driveadd {
3911     my ($storecfg, $vmid, $device) = @_;
3912
3913     my $drive = print_drive_commandline_full($storecfg, $vmid, $device);
3914     $drive =~ s/\\/\\\\/g;
3915     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_add auto \"$drive\"");
3916
3917     # If the command succeeds qemu prints: "OK"
3918     return 1 if $ret =~ m/OK/s;
3919
3920     die "adding drive failed: $ret\n";
3921 }
3922
3923 sub qemu_drivedel {
3924     my($vmid, $deviceid) = @_;
3925
3926     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-$deviceid");
3927     $ret =~ s/^\s+//;
3928
3929     return 1 if $ret eq "";
3930
3931     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3932     return 1 if $ret =~ m/Device \'.*?\' not found/s;
3933
3934     die "deleting drive $deviceid failed : $ret\n";
3935 }
3936
3937 sub qemu_deviceaddverify {
3938     my ($vmid, $deviceid) = @_;
3939
3940     for (my $i = 0; $i <= 5; $i++) {
3941          my $devices_list = vm_devices_list($vmid);
3942          return 1 if defined($devices_list->{$deviceid});
3943          sleep 1;
3944     }
3945
3946     die "error on hotplug device '$deviceid'\n";
3947 }
3948
3949
3950 sub qemu_devicedelverify {
3951     my ($vmid, $deviceid) = @_;
3952
3953     # need to verify that the device is correctly removed as device_del
3954     # is async and empty return is not reliable
3955
3956     for (my $i = 0; $i <= 5; $i++) {
3957          my $devices_list = vm_devices_list($vmid);
3958          return 1 if !defined($devices_list->{$deviceid});
3959          sleep 1;
3960     }
3961
3962     die "error on hot-unplugging device '$deviceid'\n";
3963 }
3964
3965 sub qemu_findorcreatescsihw {
3966     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
3967
3968     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3969
3970     my $scsihwid="$controller_prefix$controller";
3971     my $devices_list = vm_devices_list($vmid);
3972
3973     if(!defined($devices_list->{$scsihwid})) {
3974         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
3975     }
3976
3977     return 1;
3978 }
3979
3980 sub qemu_deletescsihw {
3981     my ($conf, $vmid, $opt) = @_;
3982
3983     my $device = parse_drive($opt, $conf->{$opt});
3984
3985     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
3986         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3987         return 1;
3988     }
3989
3990     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3991
3992     my $devices_list = vm_devices_list($vmid);
3993     foreach my $opt (keys %{$devices_list}) {
3994         if (is_valid_drivename($opt)) {
3995             my $drive = parse_drive($opt, $conf->{$opt});
3996             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
3997                 return 1;
3998             }
3999         }
4000     }
4001
4002     my $scsihwid="scsihw$controller";
4003
4004     vm_deviceunplug($vmid, $conf, $scsihwid);
4005
4006     return 1;
4007 }
4008
4009 sub qemu_add_pci_bridge {
4010     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4011
4012     my $bridges = {};
4013
4014     my $bridgeid;
4015
4016     print_pci_addr($device, $bridges, $arch, $machine_type);
4017
4018     while (my ($k, $v) = each %$bridges) {
4019         $bridgeid = $k;
4020     }
4021     return 1 if !defined($bridgeid) || $bridgeid < 1;
4022
4023     my $bridge = "pci.$bridgeid";
4024     my $devices_list = vm_devices_list($vmid);
4025
4026     if (!defined($devices_list->{$bridge})) {
4027         vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4028     }
4029
4030     return 1;
4031 }
4032
4033 sub qemu_set_link_status {
4034     my ($vmid, $device, $up) = @_;
4035
4036     mon_cmd($vmid, "set_link", name => $device,
4037                up => $up ? JSON::true : JSON::false);
4038 }
4039
4040 sub qemu_netdevadd {
4041     my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4042
4043     my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4044     my %options =  split(/[=,]/, $netdev);
4045
4046     mon_cmd($vmid, "netdev_add",  %options);
4047     return 1;
4048 }
4049
4050 sub qemu_netdevdel {
4051     my ($vmid, $deviceid) = @_;
4052
4053     mon_cmd($vmid, "netdev_del", id => $deviceid);
4054 }
4055
4056 sub qemu_usb_hotplug {
4057     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4058
4059     return if !$device;
4060
4061     # remove the old one first
4062     vm_deviceunplug($vmid, $conf, $deviceid);
4063
4064     # check if xhci controller is necessary and available
4065     if ($device->{usb3}) {
4066
4067         my $devicelist = vm_devices_list($vmid);
4068
4069         if (!$devicelist->{xhci}) {
4070             my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4071             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4072         }
4073     }
4074     my $d = parse_usb_device($device->{host});
4075     $d->{usb3} = $device->{usb3};
4076
4077     # add the new one
4078     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4079 }
4080
4081 sub qemu_cpu_hotplug {
4082     my ($vmid, $conf, $vcpus) = @_;
4083
4084     my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4085
4086     my $sockets = 1;
4087     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4088     $sockets = $conf->{sockets} if  $conf->{sockets};
4089     my $cores = $conf->{cores} || 1;
4090     my $maxcpus = $sockets * $cores;
4091
4092     $vcpus = $maxcpus if !$vcpus;
4093
4094     die "you can't add more vcpus than maxcpus\n"
4095         if $vcpus > $maxcpus;
4096
4097     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4098
4099     if ($vcpus < $currentvcpus) {
4100
4101         if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4102
4103             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4104                 qemu_devicedel($vmid, "cpu$i");
4105                 my $retry = 0;
4106                 my $currentrunningvcpus = undef;
4107                 while (1) {
4108                     $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4109                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4110                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4111                     $retry++;
4112                     sleep 1;
4113                 }
4114                 #update conf after each succesfull cpu unplug
4115                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4116                 PVE::QemuConfig->write_config($vmid, $conf);
4117             }
4118         } else {
4119             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4120         }
4121
4122         return;
4123     }
4124
4125     my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4126     die "vcpus in running vm does not match its configuration\n"
4127         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4128
4129     if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4130
4131         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4132             my $cpustr = print_cpu_device($conf, $i);
4133             qemu_deviceadd($vmid, $cpustr);
4134
4135             my $retry = 0;
4136             my $currentrunningvcpus = undef;
4137             while (1) {
4138                 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4139                 last if scalar(@{$currentrunningvcpus}) == $i;
4140                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4141                 sleep 1;
4142                 $retry++;
4143             }
4144             #update conf after each succesfull cpu hotplug
4145             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4146             PVE::QemuConfig->write_config($vmid, $conf);
4147         }
4148     } else {
4149
4150         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4151             mon_cmd($vmid, "cpu-add", id => int($i));
4152         }
4153     }
4154 }
4155
4156 sub qemu_block_set_io_throttle {
4157     my ($vmid, $deviceid,
4158         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4159         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4160         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4161         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4162
4163     return if !check_running($vmid) ;
4164
4165     mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4166         bps => int($bps),
4167         bps_rd => int($bps_rd),
4168         bps_wr => int($bps_wr),
4169         iops => int($iops),
4170         iops_rd => int($iops_rd),
4171         iops_wr => int($iops_wr),
4172         bps_max => int($bps_max),
4173         bps_rd_max => int($bps_rd_max),
4174         bps_wr_max => int($bps_wr_max),
4175         iops_max => int($iops_max),
4176         iops_rd_max => int($iops_rd_max),
4177         iops_wr_max => int($iops_wr_max),
4178         bps_max_length => int($bps_max_length),
4179         bps_rd_max_length => int($bps_rd_max_length),
4180         bps_wr_max_length => int($bps_wr_max_length),
4181         iops_max_length => int($iops_max_length),
4182         iops_rd_max_length => int($iops_rd_max_length),
4183         iops_wr_max_length => int($iops_wr_max_length),
4184     );
4185
4186 }
4187
4188 # old code, only used to shutdown old VM after update
4189 sub __read_avail {
4190     my ($fh, $timeout) = @_;
4191
4192     my $sel = new IO::Select;
4193     $sel->add($fh);
4194
4195     my $res = '';
4196     my $buf;
4197
4198     my @ready;
4199     while (scalar (@ready = $sel->can_read($timeout))) {
4200         my $count;
4201         if ($count = $fh->sysread($buf, 8192)) {
4202             if ($buf =~ /^(.*)\(qemu\) $/s) {
4203                 $res .= $1;
4204                 last;
4205             } else {
4206                 $res .= $buf;
4207             }
4208         } else {
4209             if (!defined($count)) {
4210                 die "$!\n";
4211             }
4212             last;
4213         }
4214     }
4215
4216     die "monitor read timeout\n" if !scalar(@ready);
4217
4218     return $res;
4219 }
4220
4221 sub qemu_block_resize {
4222     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4223
4224     my $running = check_running($vmid);
4225
4226     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4227
4228     return if !$running;
4229
4230     my $padding = (1024 - $size % 1024) % 1024;
4231     $size = $size + $padding;
4232
4233     mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4234
4235 }
4236
4237 sub qemu_volume_snapshot {
4238     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4239
4240     my $running = check_running($vmid);
4241
4242     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4243         mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4244     } else {
4245         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4246     }
4247 }
4248
4249 sub qemu_volume_snapshot_delete {
4250     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4251
4252     my $running = check_running($vmid);
4253
4254     if($running) {
4255
4256         $running = undef;
4257         my $conf = PVE::QemuConfig->load_config($vmid);
4258         foreach_drive($conf, sub {
4259             my ($ds, $drive) = @_;
4260             $running = 1 if $drive->{file} eq $volid;
4261         });
4262     }
4263
4264     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4265         mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4266     } else {
4267         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4268     }
4269 }
4270
4271 sub set_migration_caps {
4272     my ($vmid) = @_;
4273
4274     my $cap_ref = [];
4275
4276     my $enabled_cap = {
4277         "auto-converge" => 1,
4278         "xbzrle" => 1,
4279         "x-rdma-pin-all" => 0,
4280         "zero-blocks" => 0,
4281         "compress" => 0
4282     };
4283
4284     my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4285
4286     for my $supported_capability (@$supported_capabilities) {
4287         push @$cap_ref, {
4288             capability => $supported_capability->{capability},
4289             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4290         };
4291     }
4292
4293     mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4294 }
4295
4296 my $fast_plug_option = {
4297     'lock' => 1,
4298     'name' => 1,
4299     'onboot' => 1,
4300     'shares' => 1,
4301     'startup' => 1,
4302     'description' => 1,
4303     'protection' => 1,
4304     'vmstatestorage' => 1,
4305     'hookscript' => 1,
4306     'tags' => 1,
4307 };
4308
4309 # hotplug changes in [PENDING]
4310 # $selection hash can be used to only apply specified options, for
4311 # example: { cores => 1 } (only apply changed 'cores')
4312 # $errors ref is used to return error messages
4313 sub vmconfig_hotplug_pending {
4314     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4315
4316     my $defaults = load_defaults();
4317     my $arch = get_vm_arch($conf);
4318     my $machine_type = get_vm_machine($conf, undef, $arch);
4319
4320     # commit values which do not have any impact on running VM first
4321     # Note: those option cannot raise errors, we we do not care about
4322     # $selection and always apply them.
4323
4324     my $add_error = sub {
4325         my ($opt, $msg) = @_;
4326         $errors->{$opt} = "hotplug problem - $msg";
4327     };
4328
4329     my $changes = 0;
4330     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4331         if ($fast_plug_option->{$opt}) {
4332             $conf->{$opt} = $conf->{pending}->{$opt};
4333             delete $conf->{pending}->{$opt};
4334             $changes = 1;
4335         }
4336     }
4337
4338     if ($changes) {
4339         PVE::QemuConfig->write_config($vmid, $conf);
4340     }
4341
4342     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4343
4344     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4345     foreach my $opt (sort keys %$pending_delete_hash) {
4346         next if $selection && !$selection->{$opt};
4347         my $force = $pending_delete_hash->{$opt}->{force};
4348         eval {
4349             if ($opt eq 'hotplug') {
4350                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4351             } elsif ($opt eq 'tablet') {
4352                 die "skip\n" if !$hotplug_features->{usb};
4353                 if ($defaults->{tablet}) {
4354                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4355                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4356                         if $arch eq 'aarch64';
4357                 } else {
4358                     vm_deviceunplug($vmid, $conf, 'tablet');
4359                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4360                 }
4361             } elsif ($opt =~ m/^usb\d+/) {
4362                 die "skip\n";
4363                 # since we cannot reliably hot unplug usb devices
4364                 # we are disabling it
4365                 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4366                 vm_deviceunplug($vmid, $conf, $opt);
4367             } elsif ($opt eq 'vcpus') {
4368                 die "skip\n" if !$hotplug_features->{cpu};
4369                 qemu_cpu_hotplug($vmid, $conf, undef);
4370             } elsif ($opt eq 'balloon') {
4371                 # enable balloon device is not hotpluggable
4372                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4373                 # here we reset the ballooning value to memory
4374                 my $balloon = $conf->{memory} || $defaults->{memory};
4375                 mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4376             } elsif ($fast_plug_option->{$opt}) {
4377                 # do nothing
4378             } elsif ($opt =~ m/^net(\d+)$/) {
4379                 die "skip\n" if !$hotplug_features->{network};
4380                 vm_deviceunplug($vmid, $conf, $opt);
4381             } elsif (is_valid_drivename($opt)) {
4382                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4383                 vm_deviceunplug($vmid, $conf, $opt);
4384                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4385             } elsif ($opt =~ m/^memory$/) {
4386                 die "skip\n" if !$hotplug_features->{memory};
4387                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4388             } elsif ($opt eq 'cpuunits') {
4389                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4390             } elsif ($opt eq 'cpulimit') {
4391                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4392             } else {
4393                 die "skip\n";
4394             }
4395         };
4396         if (my $err = $@) {
4397             &$add_error($opt, $err) if $err ne "skip\n";
4398         } else {
4399             delete $conf->{$opt};
4400             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4401         }
4402     }
4403
4404     my ($apply_pending_cloudinit, $apply_pending_cloudinit_done);
4405     $apply_pending_cloudinit = sub {
4406         return if $apply_pending_cloudinit_done; # once is enough
4407         $apply_pending_cloudinit_done = 1; # once is enough
4408
4409         my ($key, $value) = @_;
4410
4411         my @cloudinit_opts = keys %$confdesc_cloudinit;
4412         foreach my $opt (keys %{$conf->{pending}}) {
4413             next if !grep { $_ eq $opt } @cloudinit_opts;
4414             $conf->{$opt} = delete $conf->{pending}->{$opt};
4415         }
4416
4417         my $new_conf = { %$conf };
4418         $new_conf->{$key} = $value;
4419         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4420     };
4421
4422     foreach my $opt (keys %{$conf->{pending}}) {
4423         next if $selection && !$selection->{$opt};
4424         my $value = $conf->{pending}->{$opt};
4425         eval {
4426             if ($opt eq 'hotplug') {
4427                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4428             } elsif ($opt eq 'tablet') {
4429                 die "skip\n" if !$hotplug_features->{usb};
4430                 if ($value == 1) {
4431                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4432                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4433                         if $arch eq 'aarch64';
4434                 } elsif ($value == 0) {
4435                     vm_deviceunplug($vmid, $conf, 'tablet');
4436                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4437                 }
4438             } elsif ($opt =~ m/^usb\d+$/) {
4439                 die "skip\n";
4440                 # since we cannot reliably hot unplug usb devices
4441                 # we are disabling it
4442                 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4443                 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4444                 die "skip\n" if !$d;
4445                 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
4446             } elsif ($opt eq 'vcpus') {
4447                 die "skip\n" if !$hotplug_features->{cpu};
4448                 qemu_cpu_hotplug($vmid, $conf, $value);
4449             } elsif ($opt eq 'balloon') {
4450                 # enable/disable balloning device is not hotpluggable
4451                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4452                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4453                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4454
4455                 # allow manual ballooning if shares is set to zero
4456                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4457                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4458                     mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4459                 }
4460             } elsif ($opt =~ m/^net(\d+)$/) {
4461                 # some changes can be done without hotplug
4462                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4463                                     $vmid, $opt, $value, $arch, $machine_type);
4464             } elsif (is_valid_drivename($opt)) {
4465                 die "skip\n" if $opt eq 'efidisk0';
4466                 # some changes can be done without hotplug
4467                 my $drive = parse_drive($opt, $value);
4468                 if (drive_is_cloudinit($drive)) {
4469                     &$apply_pending_cloudinit($opt, $value);
4470                 }
4471                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4472                                      $vmid, $opt, $value, $arch, $machine_type);
4473             } elsif ($opt =~ m/^memory$/) { #dimms
4474                 die "skip\n" if !$hotplug_features->{memory};
4475                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4476             } elsif ($opt eq 'cpuunits') {
4477                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4478             } elsif ($opt eq 'cpulimit') {
4479                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4480                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4481             } else {
4482                 die "skip\n";  # skip non-hot-pluggable options
4483             }
4484         };
4485         if (my $err = $@) {
4486             &$add_error($opt, $err) if $err ne "skip\n";
4487         } else {
4488             $conf->{$opt} = $value;
4489             delete $conf->{pending}->{$opt};
4490         }
4491     }
4492
4493     PVE::QemuConfig->write_config($vmid, $conf);
4494 }
4495
4496 sub try_deallocate_drive {
4497     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4498
4499     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4500         my $volid = $drive->{file};
4501         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4502             my $sid = PVE::Storage::parse_volume_id($volid);
4503             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4504
4505             # check if the disk is really unused
4506             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4507                 if PVE::QemuServer::Drive::is_volume_in_use($storecfg, $conf, $key, $volid);
4508             PVE::Storage::vdisk_free($storecfg, $volid);
4509             return 1;
4510         } else {
4511             # If vm is not owner of this disk remove from config
4512             return 1;
4513         }
4514     }
4515
4516     return undef;
4517 }
4518
4519 sub vmconfig_delete_or_detach_drive {
4520     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4521
4522     my $drive = parse_drive($opt, $conf->{$opt});
4523
4524     my $rpcenv = PVE::RPCEnvironment::get();
4525     my $authuser = $rpcenv->get_user();
4526
4527     if ($force) {
4528         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4529         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4530     } else {
4531         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4532     }
4533 }
4534
4535
4536
4537 sub vmconfig_apply_pending {
4538     my ($vmid, $conf, $storecfg, $errors) = @_;
4539
4540     my $add_apply_error = sub {
4541         my ($opt, $msg) = @_;
4542         my $err_msg = "unable to apply pending change $opt : $msg";
4543         $errors->{$opt} = $err_msg;
4544         warn $err_msg;
4545     };
4546
4547     # cold plug
4548
4549     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4550     foreach my $opt (sort keys %$pending_delete_hash) {
4551         my $force = $pending_delete_hash->{$opt}->{force};
4552         eval {
4553             if ($opt =~ m/^unused/) {
4554                 die "internal error";
4555             } elsif (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4556                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4557             }
4558         };
4559         if (my $err = $@) {
4560             $add_apply_error->($opt, $err);
4561         } else {
4562             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4563             delete $conf->{$opt};
4564         }
4565     }
4566
4567     PVE::QemuConfig->cleanup_pending($conf);
4568
4569     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4570         next if $opt eq 'delete'; # just to be sure
4571         eval {
4572             if (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4573                 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4574             }
4575         };
4576         if (my $err = $@) {
4577             $add_apply_error->($opt, $err);
4578         } else {
4579             $conf->{$opt} = delete $conf->{pending}->{$opt};
4580         }
4581     }
4582
4583     # write all changes at once to avoid unnecessary i/o
4584     PVE::QemuConfig->write_config($vmid, $conf);
4585 }
4586
4587 sub vmconfig_update_net {
4588     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4589
4590     my $newnet = parse_net($value);
4591
4592     if ($conf->{$opt}) {
4593         my $oldnet = parse_net($conf->{$opt});
4594
4595         if (safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4596             safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4597             safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4598             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4599
4600             # for non online change, we try to hot-unplug
4601             die "skip\n" if !$hotplug;
4602             vm_deviceunplug($vmid, $conf, $opt);
4603         } else {
4604
4605             die "internal error" if $opt !~ m/net(\d+)/;
4606             my $iface = "tap${vmid}i$1";
4607
4608             if (safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4609                 safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4610                 safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4611                 safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4612                 PVE::Network::tap_unplug($iface);
4613
4614                 if ($have_sdn) {
4615                     PVE::Network::SDN::Zones::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4616                 } else {
4617                     PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4618                 }
4619             } elsif (safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4620                 # Rate can be applied on its own but any change above needs to
4621                 # include the rate in tap_plug since OVS resets everything.
4622                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4623             }
4624
4625             if (safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4626                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4627             }
4628
4629             return 1;
4630         }
4631     }
4632
4633     if ($hotplug) {
4634         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
4635     } else {
4636         die "skip\n";
4637     }
4638 }
4639
4640 sub vmconfig_update_disk {
4641     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4642
4643     my $drive = parse_drive($opt, $value);
4644
4645     if ($conf->{$opt}) {
4646
4647         if (my $old_drive = parse_drive($opt, $conf->{$opt}))  {
4648
4649             my $media = $drive->{media} || 'disk';
4650             my $oldmedia = $old_drive->{media} || 'disk';
4651             die "unable to change media type\n" if $media ne $oldmedia;
4652
4653             if (!drive_is_cdrom($old_drive)) {
4654
4655                 if ($drive->{file} ne $old_drive->{file}) {
4656
4657                     die "skip\n" if !$hotplug;
4658
4659                     # unplug and register as unused
4660                     vm_deviceunplug($vmid, $conf, $opt);
4661                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4662
4663                 } else {
4664                     # update existing disk
4665
4666                     # skip non hotpluggable value
4667                     if (safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4668                         safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4669                         safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4670                         safe_string_ne($drive->{cache}, $old_drive->{cache}) ||
4671                         safe_string_ne($drive->{ssd}, $old_drive->{ssd})) {
4672                         die "skip\n";
4673                     }
4674
4675                     # apply throttle
4676                     if (safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4677                         safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4678                         safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4679                         safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4680                         safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4681                         safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4682                         safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4683                         safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4684                         safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4685                         safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4686                         safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4687                         safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4688                         safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4689                         safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4690                         safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4691                         safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4692                         safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4693                         safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4694
4695                         qemu_block_set_io_throttle($vmid,"drive-$opt",
4696                                                    ($drive->{mbps} || 0)*1024*1024,
4697                                                    ($drive->{mbps_rd} || 0)*1024*1024,
4698                                                    ($drive->{mbps_wr} || 0)*1024*1024,
4699                                                    $drive->{iops} || 0,
4700                                                    $drive->{iops_rd} || 0,
4701                                                    $drive->{iops_wr} || 0,
4702                                                    ($drive->{mbps_max} || 0)*1024*1024,
4703                                                    ($drive->{mbps_rd_max} || 0)*1024*1024,
4704                                                    ($drive->{mbps_wr_max} || 0)*1024*1024,
4705                                                    $drive->{iops_max} || 0,
4706                                                    $drive->{iops_rd_max} || 0,
4707                                                    $drive->{iops_wr_max} || 0,
4708                                                    $drive->{bps_max_length} || 1,
4709                                                    $drive->{bps_rd_max_length} || 1,
4710                                                    $drive->{bps_wr_max_length} || 1,
4711                                                    $drive->{iops_max_length} || 1,
4712                                                    $drive->{iops_rd_max_length} || 1,
4713                                                    $drive->{iops_wr_max_length} || 1);
4714
4715                     }
4716
4717                     return 1;
4718                 }
4719
4720             } else { # cdrom
4721
4722                 if ($drive->{file} eq 'none') {
4723                     mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4724                     if (drive_is_cloudinit($old_drive)) {
4725                         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4726                     }
4727                 } else {
4728                     my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4729
4730                     # force eject if locked
4731                     mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4732
4733                     if ($path) {
4734                         mon_cmd($vmid, "blockdev-change-medium",
4735                             id => "$opt", filename => "$path");
4736                     }
4737                 }
4738
4739                 return 1;
4740             }
4741         }
4742     }
4743
4744     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4745     # hotplug new disks
4746     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4747     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
4748 }
4749
4750 # called in locked context by incoming migration
4751 sub vm_migrate_get_nbd_disks {
4752     my ($storecfg, $conf, $replicated_volumes) = @_;
4753
4754     my $local_volumes = {};
4755     foreach_drive($conf, sub {
4756         my ($ds, $drive) = @_;
4757
4758         return if drive_is_cdrom($drive);
4759
4760         my $volid = $drive->{file};
4761
4762         return if !$volid;
4763
4764         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
4765
4766         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4767         return if $scfg->{shared};
4768
4769         # replicated disks re-use existing state via bitmap
4770         my $use_existing = $replicated_volumes->{$volid} ? 1 : 0;
4771         $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
4772     });
4773     return $local_volumes;
4774 }
4775
4776 # called in locked context by incoming migration
4777 sub vm_migrate_alloc_nbd_disks {
4778     my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
4779
4780     my $format = undef;
4781
4782     my $nbd = {};
4783     foreach my $opt (sort keys %$source_volumes) {
4784         my ($volid, $storeid, $volname, $drive, $use_existing) = @{$source_volumes->{$opt}};
4785
4786         if ($use_existing) {
4787             $nbd->{$opt}->{drivestr} = print_drive($drive);
4788             $nbd->{$opt}->{volid} = $volid;
4789             $nbd->{$opt}->{replicated} = 1;
4790             next;
4791         }
4792
4793         # If a remote storage is specified and the format of the original
4794         # volume is not available there, fall back to the default format.
4795         # Otherwise use the same format as the original.
4796         if (!$storagemap->{identity}) {
4797             $storeid = map_storage($storagemap, $storeid);
4798             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
4799             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4800             my $fileFormat = qemu_img_format($scfg, $volname);
4801             $format = (grep {$fileFormat eq $_} @{$validFormats}) ? $fileFormat : $defFormat;
4802         } else {
4803             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4804             $format = qemu_img_format($scfg, $volname);
4805         }
4806
4807         my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
4808         my $newdrive = $drive;
4809         $newdrive->{format} = $format;
4810         $newdrive->{file} = $newvolid;
4811         my $drivestr = print_drive($newdrive);
4812         $nbd->{$opt}->{drivestr} = $drivestr;
4813         $nbd->{$opt}->{volid} = $newvolid;
4814     }
4815
4816     return $nbd;
4817 }
4818
4819 # see vm_start_nolock for parameters, additionally:
4820 # migrate_opts:
4821 #   storagemap = parsed storage map for allocating NBD disks
4822 sub vm_start {
4823     my ($storecfg, $vmid, $params, $migrate_opts) = @_;
4824
4825     return PVE::QemuConfig->lock_config($vmid, sub {
4826         my $conf = PVE::QemuConfig->load_config($vmid, $migrate_opts->{migratedfrom});
4827
4828         die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4829
4830         $params->{resume} = PVE::QemuConfig->has_lock($conf, 'suspended');
4831
4832         PVE::QemuConfig->check_lock($conf)
4833             if !($params->{skiplock} || $params->{resume});
4834
4835         die "VM $vmid already running\n" if check_running($vmid, undef, $migrate_opts->{migratedfrom});
4836
4837         if (my $storagemap = $migrate_opts->{storagemap}) {
4838             my $replicated = $migrate_opts->{replicated_volumes};
4839             my $disks = vm_migrate_get_nbd_disks($storecfg, $conf, $replicated);
4840             $migrate_opts->{nbd} = vm_migrate_alloc_nbd_disks($storecfg, $vmid, $disks, $storagemap);
4841
4842             foreach my $opt (keys %{$migrate_opts->{nbd}}) {
4843                 $conf->{$opt} = $migrate_opts->{nbd}->{$opt}->{drivestr};
4844             }
4845         }
4846
4847         return vm_start_nolock($storecfg, $vmid, $conf, $params, $migrate_opts);
4848     });
4849 }
4850
4851
4852 # params:
4853 #   statefile => 'tcp', 'unix' for migration or path/volid for RAM state
4854 #   skiplock => 0/1, skip checking for config lock
4855 #   forcemachine => to force Qemu machine (rollback/migration)
4856 #   forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
4857 #   timeout => in seconds
4858 #   paused => start VM in paused state (backup)
4859 #   resume => resume from hibernation
4860 # migrate_opts:
4861 #   nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
4862 #   migratedfrom => source node
4863 #   spice_ticket => used for spice migration, passed via tunnel/stdin
4864 #   network => CIDR of migration network
4865 #   type => secure/insecure - tunnel over encrypted connection or plain-text
4866 #   nbd_proto_version => int, 0 for TCP, 1 for UNIX
4867 #   replicated_volumes = which volids should be re-used with bitmaps for nbd migration
4868 sub vm_start_nolock {
4869     my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
4870
4871     my $statefile = $params->{statefile};
4872     my $resume = $params->{resume};
4873
4874     my $migratedfrom = $migrate_opts->{migratedfrom};
4875     my $migration_type = $migrate_opts->{type};
4876
4877     my $res = {};
4878
4879     # clean up leftover reboot request files
4880     eval { clear_reboot_request($vmid); };
4881     warn $@ if $@;
4882
4883     if (!$statefile && scalar(keys %{$conf->{pending}})) {
4884         vmconfig_apply_pending($vmid, $conf, $storecfg);
4885         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4886     }
4887
4888     PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4889
4890     my $defaults = load_defaults();
4891
4892     # set environment variable useful inside network script
4893     $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4894
4895     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
4896
4897     my $forcemachine = $params->{forcemachine};
4898     my $forcecpu = $params->{forcecpu};
4899     if ($resume) {
4900         # enforce machine and CPU type on suspended vm to ensure HW compatibility
4901         $forcemachine = $conf->{runningmachine};
4902         $forcecpu = $conf->{runningcpu};
4903         print "Resuming suspended VM\n";
4904     }
4905
4906     my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf,
4907         $defaults, $forcemachine, $forcecpu);
4908
4909     my $migration_ip;
4910     my $get_migration_ip = sub {
4911         my ($nodename) = @_;
4912
4913         return $migration_ip if defined($migration_ip);
4914
4915         my $cidr = $migrate_opts->{network};
4916
4917         if (!defined($cidr)) {
4918             my $dc_conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4919             $cidr = $dc_conf->{migration}->{network};
4920         }
4921
4922         if (defined($cidr)) {
4923             my $ips = PVE::Network::get_local_ip_from_cidr($cidr);
4924
4925             die "could not get IP: no address configured on local " .
4926                 "node for network '$cidr'\n" if scalar(@$ips) == 0;
4927
4928             die "could not get IP: multiple addresses configured on local " .
4929                 "node for network '$cidr'\n" if scalar(@$ips) > 1;
4930
4931             $migration_ip = @$ips[0];
4932         }
4933
4934         $migration_ip = PVE::Cluster::remote_node_ip($nodename, 1)
4935             if !defined($migration_ip);
4936
4937         return $migration_ip;
4938     };
4939
4940     my $migrate_uri;
4941     if ($statefile) {
4942         if ($statefile eq 'tcp') {
4943             my $localip = "localhost";
4944             my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4945             my $nodename = nodename();
4946
4947             if (!defined($migration_type)) {
4948                 if (defined($datacenterconf->{migration}->{type})) {
4949                     $migration_type = $datacenterconf->{migration}->{type};
4950                 } else {
4951                     $migration_type = 'secure';
4952                 }
4953             }
4954
4955             if ($migration_type eq 'insecure') {
4956                 $localip = $get_migration_ip->($nodename);
4957                 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
4958             }
4959
4960             my $pfamily = PVE::Tools::get_host_address_family($nodename);
4961             my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
4962             $migrate_uri = "tcp:${localip}:${migrate_port}";
4963             push @$cmd, '-incoming', $migrate_uri;
4964             push @$cmd, '-S';
4965
4966         } elsif ($statefile eq 'unix') {
4967             # should be default for secure migrations as a ssh TCP forward
4968             # tunnel is not deterministic reliable ready and fails regurarly
4969             # to set up in time, so use UNIX socket forwards
4970             my $socket_addr = "/run/qemu-server/$vmid.migrate";
4971             unlink $socket_addr;
4972
4973             $migrate_uri = "unix:$socket_addr";
4974
4975             push @$cmd, '-incoming', $migrate_uri;
4976             push @$cmd, '-S';
4977
4978         } elsif (-e $statefile) {
4979             push @$cmd, '-loadstate', $statefile;
4980         } else {
4981             my $statepath = PVE::Storage::path($storecfg, $statefile);
4982             push @$vollist, $statefile;
4983             push @$cmd, '-loadstate', $statepath;
4984         }
4985     } elsif ($params->{paused}) {
4986         push @$cmd, '-S';
4987     }
4988
4989     # host pci devices
4990     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
4991       my $d = parse_hostpci($conf->{"hostpci$i"});
4992       next if !$d;
4993       my $pcidevices = $d->{pciid};
4994       foreach my $pcidevice (@$pcidevices) {
4995             my $pciid = $pcidevice->{id};
4996
4997             my $info = PVE::SysFSTools::pci_device_info("$pciid");
4998             die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
4999             die "no pci device info for device '$pciid'\n" if !$info;
5000
5001             if ($d->{mdev}) {
5002                 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5003                 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5004             } else {
5005                 die "can't unbind/bind pci group to vfio '$pciid'\n"
5006                     if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5007                 die "can't reset pci device '$pciid'\n"
5008                     if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5009             }
5010       }
5011     }
5012
5013     PVE::Storage::activate_volumes($storecfg, $vollist);
5014
5015     eval {
5016         run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5017             outfunc => sub {}, errfunc => sub {});
5018     };
5019     # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5020     # timeout should be more than enough here...
5021     PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5022
5023     my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5024                                               : $defaults->{cpuunits};
5025
5026     my $start_timeout = $params->{timeout} // config_aware_timeout($conf, $resume);
5027     my %run_params = (
5028         timeout => $statefile ? undef : $start_timeout,
5029         umask => 0077,
5030         noerr => 1,
5031     );
5032
5033     # when migrating, prefix QEMU output so other side can pick up any
5034     # errors that might occur and show the user
5035     if ($migratedfrom) {
5036         $run_params{quiet} = 1;
5037         $run_params{logfunc} = sub { print "QEMU: $_[0]\n" };
5038     }
5039
5040     my %properties = (
5041         Slice => 'qemu.slice',
5042         KillMode => 'none',
5043         CPUShares => $cpuunits
5044     );
5045
5046     if (my $cpulimit = $conf->{cpulimit}) {
5047         $properties{CPUQuota} = int($cpulimit * 100);
5048     }
5049     $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5050
5051     my $run_qemu = sub {
5052         PVE::Tools::run_fork sub {
5053             PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5054
5055             my $exitcode = run_command($cmd, %run_params);
5056             die "QEMU exited with code $exitcode\n" if $exitcode;
5057         };
5058     };
5059
5060     if ($conf->{hugepages}) {
5061
5062         my $code = sub {
5063             my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5064             my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5065
5066             PVE::QemuServer::Memory::hugepages_mount();
5067             PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5068
5069             eval { $run_qemu->() };
5070             if (my $err = $@) {
5071                 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5072                 die $err;
5073             }
5074
5075             PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5076         };
5077         eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5078
5079     } else {
5080         eval { $run_qemu->() };
5081     }
5082
5083     if (my $err = $@) {
5084         # deactivate volumes if start fails
5085         eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5086         die "start failed: $err";
5087     }
5088
5089     print "migration listens on $migrate_uri\n" if $migrate_uri;
5090     $res->{migrate_uri} = $migrate_uri;
5091
5092     if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5093         eval { mon_cmd($vmid, "cont"); };
5094         warn $@ if $@;
5095     }
5096
5097     #start nbd server for storage migration
5098     if (my $nbd = $migrate_opts->{nbd}) {
5099         my $nbd_protocol_version = $migrate_opts->{nbd_proto_version} // 0;
5100
5101         my $migrate_storage_uri;
5102         # nbd_protocol_version > 0 for unix socket support
5103         if ($nbd_protocol_version > 0 && $migration_type eq 'secure') {
5104             my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5105             mon_cmd($vmid, "nbd-server-start", addr => { type => 'unix', data => { path => $socket_path } } );
5106             $migrate_storage_uri = "nbd:unix:$socket_path";
5107         } else {
5108             my $nodename = nodename();
5109             my $localip = $get_migration_ip->($nodename);
5110             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5111             my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5112
5113             mon_cmd($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5114             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5115             $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5116         }
5117
5118         $res->{migrate_storage_uri} = $migrate_storage_uri;
5119
5120         foreach my $opt (sort keys %$nbd) {
5121             my $drivestr = $nbd->{$opt}->{drivestr};
5122             my $volid = $nbd->{$opt}->{volid};
5123             mon_cmd($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5124             my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
5125             print "storage migration listens on $nbd_uri volume:$drivestr\n";
5126             print "re-using replicated volume: $opt - $volid\n"
5127                 if $nbd->{$opt}->{replicated};
5128
5129             $res->{drives}->{$opt} = $nbd->{$opt};
5130             $res->{drives}->{$opt}->{nbd_uri} = $nbd_uri;
5131         }
5132     }
5133
5134     if ($migratedfrom) {
5135         eval {
5136             set_migration_caps($vmid);
5137         };
5138         warn $@ if $@;
5139
5140         if ($spice_port) {
5141             print "spice listens on port $spice_port\n";
5142             $res->{spice_port} = $spice_port;
5143             if ($migrate_opts->{spice_ticket}) {
5144                 mon_cmd($vmid, "set_password", protocol => 'spice', password => $migrate_opts->{spice_ticket});
5145                 mon_cmd($vmid, "expire_password", protocol => 'spice', time => "+30");
5146             }
5147         }
5148
5149     } else {
5150         mon_cmd($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5151             if !$statefile && $conf->{balloon};
5152
5153         foreach my $opt (keys %$conf) {
5154             next if $opt !~  m/^net\d+$/;
5155             my $nicconf = parse_net($conf->{$opt});
5156             qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5157         }
5158     }
5159
5160     mon_cmd($vmid, 'qom-set',
5161                 path => "machine/peripheral/balloon0",
5162                 property => "guest-stats-polling-interval",
5163                 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5164
5165     if ($resume) {
5166         print "Resumed VM, removing state\n";
5167         if (my $vmstate = $conf->{vmstate}) {
5168             PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5169             PVE::Storage::vdisk_free($storecfg, $vmstate);
5170         }
5171         delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
5172         PVE::QemuConfig->write_config($vmid, $conf);
5173     }
5174
5175     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5176
5177     return $res;
5178 }
5179
5180 sub vm_commandline {
5181     my ($storecfg, $vmid, $snapname) = @_;
5182
5183     my $conf = PVE::QemuConfig->load_config($vmid);
5184     my $forcemachine;
5185     my $forcecpu;
5186
5187     if ($snapname) {
5188         my $snapshot = $conf->{snapshots}->{$snapname};
5189         die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5190
5191         # check for machine or CPU overrides in snapshot
5192         $forcemachine = $snapshot->{runningmachine};
5193         $forcecpu = $snapshot->{runningcpu};
5194
5195         $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5196
5197         $conf = $snapshot;
5198     }
5199
5200     my $defaults = load_defaults();
5201
5202     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults,
5203         $forcemachine, $forcecpu);
5204
5205     return PVE::Tools::cmd2string($cmd);
5206 }
5207
5208 sub vm_reset {
5209     my ($vmid, $skiplock) = @_;
5210
5211     PVE::QemuConfig->lock_config($vmid, sub {
5212
5213         my $conf = PVE::QemuConfig->load_config($vmid);
5214
5215         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5216
5217         mon_cmd($vmid, "system_reset");
5218     });
5219 }
5220
5221 sub get_vm_volumes {
5222     my ($conf) = @_;
5223
5224     my $vollist = [];
5225     foreach_volid($conf, sub {
5226         my ($volid, $attr) = @_;
5227
5228         return if $volid =~ m|^/|;
5229
5230         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5231         return if !$sid;
5232
5233         push @$vollist, $volid;
5234     });
5235
5236     return $vollist;
5237 }
5238
5239 sub vm_stop_cleanup {
5240     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5241
5242     eval {
5243
5244         if (!$keepActive) {
5245             my $vollist = get_vm_volumes($conf);
5246             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5247         }
5248
5249         foreach my $ext (qw(mon qmp pid vnc qga)) {
5250             unlink "/var/run/qemu-server/${vmid}.$ext";
5251         }
5252
5253         if ($conf->{ivshmem}) {
5254             my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5255             # just delete it for now, VMs which have this already open do not
5256             # are affected, but new VMs will get a separated one. If this
5257             # becomes an issue we either add some sort of ref-counting or just
5258             # add a "don't delete on stop" flag to the ivshmem format.
5259             unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5260         }
5261
5262         foreach my $key (keys %$conf) {
5263             next if $key !~ m/^hostpci(\d+)$/;
5264             my $hostpciindex = $1;
5265             my $d = parse_hostpci($conf->{$key});
5266             my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5267
5268             foreach my $pci (@{$d->{pciid}}) {
5269                 my $pciid = $pci->{id};
5270                 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5271             }
5272         }
5273
5274         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5275     };
5276     warn $@ if $@; # avoid errors - just warn
5277 }
5278
5279 # call only in locked context
5280 sub _do_vm_stop {
5281     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5282
5283     my $pid = check_running($vmid, $nocheck);
5284     return if !$pid;
5285
5286     my $conf;
5287     if (!$nocheck) {
5288         $conf = PVE::QemuConfig->load_config($vmid);
5289         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5290         if (!defined($timeout) && $shutdown && $conf->{startup}) {
5291             my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5292             $timeout = $opts->{down} if $opts->{down};
5293         }
5294         PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5295     }
5296
5297     eval {
5298         if ($shutdown) {
5299             if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5300                 mon_cmd($vmid, "guest-shutdown", timeout => $timeout);
5301             } else {
5302                 mon_cmd($vmid, "system_powerdown");
5303             }
5304         } else {
5305             mon_cmd($vmid, "quit");
5306         }
5307     };
5308     my $err = $@;
5309
5310     if (!$err) {
5311         $timeout = 60 if !defined($timeout);
5312
5313         my $count = 0;
5314         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5315             $count++;
5316             sleep 1;
5317         }
5318
5319         if ($count >= $timeout) {
5320             if ($force) {
5321                 warn "VM still running - terminating now with SIGTERM\n";
5322                 kill 15, $pid;
5323             } else {
5324                 die "VM quit/powerdown failed - got timeout\n";
5325             }
5326         } else {
5327             vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5328             return;
5329         }
5330     } else {
5331         if ($force) {
5332             warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5333             kill 15, $pid;
5334         } else {
5335             die "VM quit/powerdown failed\n";
5336         }
5337     }
5338
5339     # wait again
5340     $timeout = 10;
5341
5342     my $count = 0;
5343     while (($count < $timeout) && check_running($vmid, $nocheck)) {
5344         $count++;
5345         sleep 1;
5346     }
5347
5348     if ($count >= $timeout) {
5349         warn "VM still running - terminating now with SIGKILL\n";
5350         kill 9, $pid;
5351         sleep 1;
5352     }
5353
5354     vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5355 }
5356
5357 # Note: use $nocheck to skip tests if VM configuration file exists.
5358 # We need that when migration VMs to other nodes (files already moved)
5359 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5360 sub vm_stop {
5361     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5362
5363     $force = 1 if !defined($force) && !$shutdown;
5364
5365     if ($migratedfrom){
5366         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5367         kill 15, $pid if $pid;
5368         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5369         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5370         return;
5371     }
5372
5373     PVE::QemuConfig->lock_config($vmid, sub {
5374         _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5375    });
5376 }
5377
5378 sub vm_reboot {
5379     my ($vmid, $timeout) = @_;
5380
5381     PVE::QemuConfig->lock_config($vmid, sub {
5382         eval {
5383
5384             # only reboot if running, as qmeventd starts it again on a stop event
5385             return if !check_running($vmid);
5386
5387             create_reboot_request($vmid);
5388
5389             my $storecfg = PVE::Storage::config();
5390             _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5391
5392         };
5393         if (my $err = $@) {
5394             # avoid that the next normal shutdown will be confused for a reboot
5395             clear_reboot_request($vmid);
5396             die $err;
5397         }
5398    });
5399 }
5400
5401 # note: if using the statestorage parameter, the caller has to check privileges
5402 sub vm_suspend {
5403     my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5404
5405     my $conf;
5406     my $path;
5407     my $storecfg;
5408     my $vmstate;
5409
5410     PVE::QemuConfig->lock_config($vmid, sub {
5411
5412         $conf = PVE::QemuConfig->load_config($vmid);
5413
5414         my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5415         PVE::QemuConfig->check_lock($conf)
5416             if !($skiplock || $is_backing_up);
5417
5418         die "cannot suspend to disk during backup\n"
5419             if $is_backing_up && $includestate;
5420
5421         if ($includestate) {
5422             $conf->{lock} = 'suspending';
5423             my $date = strftime("%Y-%m-%d", localtime(time()));
5424             $storecfg = PVE::Storage::config();
5425             if (!$statestorage) {
5426                 $statestorage = find_vmstate_storage($conf, $storecfg);
5427                 # check permissions for the storage
5428                 my $rpcenv = PVE::RPCEnvironment::get();
5429                 if ($rpcenv->{type} ne 'cli') {
5430                     my $authuser = $rpcenv->get_user();
5431                     $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
5432                 }
5433             }
5434
5435
5436             $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5437             $path = PVE::Storage::path($storecfg, $vmstate);
5438             PVE::QemuConfig->write_config($vmid, $conf);
5439         } else {
5440             mon_cmd($vmid, "stop");
5441         }
5442     });
5443
5444     if ($includestate) {
5445         # save vm state
5446         PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5447
5448         eval {
5449             mon_cmd($vmid, "savevm-start", statefile => $path);
5450             for(;;) {
5451                 my $state = mon_cmd($vmid, "query-savevm");
5452                 if (!$state->{status}) {
5453                     die "savevm not active\n";
5454                 } elsif ($state->{status} eq 'active') {
5455                     sleep(1);
5456                     next;
5457                 } elsif ($state->{status} eq 'completed') {
5458                     print "State saved, quitting\n";
5459                     last;
5460                 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5461                     die "query-savevm failed with error '$state->{error}'\n"
5462                 } else {
5463                     die "query-savevm returned status '$state->{status}'\n";
5464                 }
5465             }
5466         };
5467         my $err = $@;
5468
5469         PVE::QemuConfig->lock_config($vmid, sub {
5470             $conf = PVE::QemuConfig->load_config($vmid);
5471             if ($err) {
5472                 # cleanup, but leave suspending lock, to indicate something went wrong
5473                 eval {
5474                     mon_cmd($vmid, "savevm-end");
5475                     PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5476                     PVE::Storage::vdisk_free($storecfg, $vmstate);
5477                     delete $conf->@{qw(vmstate runningmachine runningcpu)};
5478                     PVE::QemuConfig->write_config($vmid, $conf);
5479                 };
5480                 warn $@ if $@;
5481                 die $err;
5482             }
5483
5484             die "lock changed unexpectedly\n"
5485                 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5486
5487             mon_cmd($vmid, "quit");
5488             $conf->{lock} = 'suspended';
5489             PVE::QemuConfig->write_config($vmid, $conf);
5490         });
5491     }
5492 }
5493
5494 sub vm_resume {
5495     my ($vmid, $skiplock, $nocheck) = @_;
5496
5497     PVE::QemuConfig->lock_config($vmid, sub {
5498         my $res = mon_cmd($vmid, 'query-status');
5499         my $resume_cmd = 'cont';
5500
5501         if ($res->{status} && $res->{status} eq 'suspended') {
5502             $resume_cmd = 'system_wakeup';
5503         }
5504
5505         if (!$nocheck) {
5506
5507             my $conf = PVE::QemuConfig->load_config($vmid);
5508
5509             PVE::QemuConfig->check_lock($conf)
5510                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5511         }
5512
5513         mon_cmd($vmid, $resume_cmd);
5514     });
5515 }
5516
5517 sub vm_sendkey {
5518     my ($vmid, $skiplock, $key) = @_;
5519
5520     PVE::QemuConfig->lock_config($vmid, sub {
5521
5522         my $conf = PVE::QemuConfig->load_config($vmid);
5523
5524         # there is no qmp command, so we use the human monitor command
5525         my $res = PVE::QemuServer::Monitor::hmp_cmd($vmid, "sendkey $key");
5526         die $res if $res ne '';
5527     });
5528 }
5529
5530 # vzdump restore implementaion
5531
5532 sub tar_archive_read_firstfile {
5533     my $archive = shift;
5534
5535     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5536
5537     # try to detect archive type first
5538     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5539         die "unable to open file '$archive'\n";
5540     my $firstfile = <$fh>;
5541     kill 15, $pid;
5542     close $fh;
5543
5544     die "ERROR: archive contaions no data\n" if !$firstfile;
5545     chomp $firstfile;
5546
5547     return $firstfile;
5548 }
5549
5550 sub tar_restore_cleanup {
5551     my ($storecfg, $statfile) = @_;
5552
5553     print STDERR "starting cleanup\n";
5554
5555     if (my $fd = IO::File->new($statfile, "r")) {
5556         while (defined(my $line = <$fd>)) {
5557             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5558                 my $volid = $2;
5559                 eval {
5560                     if ($volid =~ m|^/|) {
5561                         unlink $volid || die 'unlink failed\n';
5562                     } else {
5563                         PVE::Storage::vdisk_free($storecfg, $volid);
5564                     }
5565                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5566                 };
5567                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5568             } else {
5569                 print STDERR "unable to parse line in statfile - $line";
5570             }
5571         }
5572         $fd->close();
5573     }
5574 }
5575
5576 sub restore_file_archive {
5577     my ($archive, $vmid, $user, $opts) = @_;
5578
5579     my $format = $opts->{format};
5580     my $comp;
5581
5582     if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5583         $format = 'tar' if !$format;
5584         $comp = 'gzip';
5585     } elsif ($archive =~ m/\.tar$/) {
5586         $format = 'tar' if !$format;
5587     } elsif ($archive =~ m/.tar.lzo$/) {
5588         $format = 'tar' if !$format;
5589         $comp = 'lzop';
5590     } elsif ($archive =~ m/\.vma$/) {
5591         $format = 'vma' if !$format;
5592     } elsif ($archive =~ m/\.vma\.gz$/) {
5593         $format = 'vma' if !$format;
5594         $comp = 'gzip';
5595     } elsif ($archive =~ m/\.vma\.lzo$/) {
5596         $format = 'vma' if !$format;
5597         $comp = 'lzop';
5598     } else {
5599         $format = 'vma' if !$format; # default
5600     }
5601
5602     # try to detect archive format
5603     if ($format eq 'tar') {
5604         return restore_tar_archive($archive, $vmid, $user, $opts);
5605     } else {
5606         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5607     }
5608 }
5609
5610 # hepler to remove disks that will not be used after restore
5611 my $restore_cleanup_oldconf = sub {
5612     my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
5613
5614     foreach_drive($oldconf, sub {
5615         my ($ds, $drive) = @_;
5616
5617         return if drive_is_cdrom($drive, 1);
5618
5619         my $volid = $drive->{file};
5620         return if !$volid || $volid =~ m|^/|;
5621
5622         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
5623         return if !$path || !$owner || ($owner != $vmid);
5624
5625         # Note: only delete disk we want to restore
5626         # other volumes will become unused
5627         if ($virtdev_hash->{$ds}) {
5628             eval { PVE::Storage::vdisk_free($storecfg, $volid); };
5629             if (my $err = $@) {
5630                 warn $err;
5631             }
5632         }
5633     });
5634
5635     # delete vmstate files, after the restore we have no snapshots anymore
5636     foreach my $snapname (keys %{$oldconf->{snapshots}}) {
5637         my $snap = $oldconf->{snapshots}->{$snapname};
5638         if ($snap->{vmstate}) {
5639             eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5640             if (my $err = $@) {
5641                 warn $err;
5642             }
5643         }
5644     }
5645 };
5646
5647 # Helper to parse vzdump backup device hints
5648 #
5649 # $rpcenv: Environment, used to ckeck storage permissions
5650 # $user: User ID, to check storage permissions
5651 # $storecfg: Storage configuration
5652 # $fh: the file handle for reading the configuration
5653 # $devinfo: should contain device sizes for all backu-up'ed devices
5654 # $options: backup options (pool, default storage)
5655 #
5656 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
5657 my $parse_backup_hints = sub {
5658     my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
5659
5660     my $virtdev_hash = {};
5661
5662     while (defined(my $line = <$fh>)) {
5663         if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5664             my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5665             die "archive does not contain data for drive '$virtdev'\n"
5666                 if !$devinfo->{$devname};
5667
5668             if (defined($options->{storage})) {
5669                 $storeid = $options->{storage} || 'local';
5670             } elsif (!$storeid) {
5671                 $storeid = 'local';
5672             }
5673             $format = 'raw' if !$format;
5674             $devinfo->{$devname}->{devname} = $devname;
5675             $devinfo->{$devname}->{virtdev} = $virtdev;
5676             $devinfo->{$devname}->{format} = $format;
5677             $devinfo->{$devname}->{storeid} = $storeid;
5678
5679             # check permission on storage
5680             my $pool = $options->{pool}; # todo: do we need that?
5681             if ($user ne 'root@pam') {
5682                 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5683             }
5684
5685             $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5686         } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
5687             my $virtdev = $1;
5688             my $drive = parse_drive($virtdev, $2);
5689             if (drive_is_cloudinit($drive)) {
5690                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5691                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5692                 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
5693
5694                 $virtdev_hash->{$virtdev} = {
5695                     format => $format,
5696                     storeid => $options->{storage} // $storeid,
5697                     size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
5698                     is_cloudinit => 1,
5699                 };
5700             }
5701         }
5702     }
5703
5704     return $virtdev_hash;
5705 };
5706
5707 # Helper to allocate and activate all volumes required for a restore
5708 #
5709 # $storecfg: Storage configuration
5710 # $virtdev_hash: as returned by parse_backup_hints()
5711 #
5712 # Returns: { $virtdev => $volid }
5713 my $restore_allocate_devices = sub {
5714     my ($storecfg, $virtdev_hash, $vmid) = @_;
5715
5716     my $map = {};
5717     foreach my $virtdev (sort keys %$virtdev_hash) {
5718         my $d = $virtdev_hash->{$virtdev};
5719         my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5720         my $storeid = $d->{storeid};
5721         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5722
5723         # test if requested format is supported
5724         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5725         my $supported = grep { $_ eq $d->{format} } @$validFormats;
5726         $d->{format} = $defFormat if !$supported;
5727
5728         my $name;
5729         if ($d->{is_cloudinit}) {
5730             $name = "vm-$vmid-cloudinit";
5731             $name .= ".$d->{format}" if $d->{format} ne 'raw';
5732         }
5733
5734         my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
5735
5736         print STDERR "new volume ID is '$volid'\n";
5737         $d->{volid} = $volid;
5738
5739         PVE::Storage::activate_volumes($storecfg, [$volid]);
5740
5741         $map->{$virtdev} = $volid;
5742     }
5743
5744     return $map;
5745 };
5746
5747 my $restore_update_config_line = sub {
5748     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5749
5750     return if $line =~ m/^\#qmdump\#/;
5751     return if $line =~ m/^\#vzdump\#/;
5752     return if $line =~ m/^lock:/;
5753     return if $line =~ m/^unused\d+:/;
5754     return if $line =~ m/^parent:/;
5755
5756     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5757     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5758         # try to convert old 1.X settings
5759         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5760         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5761             my ($model, $macaddr) = split(/\=/, $devconfig);
5762             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5763             my $net = {
5764                 model => $model,
5765                 bridge => "vmbr$ind",
5766                 macaddr => $macaddr,
5767             };
5768             my $netstr = print_net($net);
5769
5770             print $outfd "net$cookie->{netcount}: $netstr\n";
5771             $cookie->{netcount}++;
5772         }
5773     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5774         my ($id, $netstr) = ($1, $2);
5775         my $net = parse_net($netstr);
5776         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5777         $netstr = print_net($net);
5778         print $outfd "$id: $netstr\n";
5779     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5780         my $virtdev = $1;
5781         my $value = $3;
5782         my $di = parse_drive($virtdev, $value);
5783         if (defined($di->{backup}) && !$di->{backup}) {
5784             print $outfd "#$line";
5785         } elsif ($map->{$virtdev}) {
5786             delete $di->{format}; # format can change on restore
5787             $di->{file} = $map->{$virtdev};
5788             $value = print_drive($di);
5789             print $outfd "$virtdev: $value\n";
5790         } else {
5791             print $outfd $line;
5792         }
5793     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5794         my $vmgenid = $1;
5795         if ($vmgenid ne '0') {
5796             # always generate a new vmgenid if there was a valid one setup
5797             $vmgenid = generate_uuid();
5798         }
5799         print $outfd "vmgenid: $vmgenid\n";
5800     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5801         my ($uuid, $uuid_str);
5802         UUID::generate($uuid);
5803         UUID::unparse($uuid, $uuid_str);
5804         my $smbios1 = parse_smbios1($2);
5805         $smbios1->{uuid} = $uuid_str;
5806         print $outfd $1.print_smbios1($smbios1)."\n";
5807     } else {
5808         print $outfd $line;
5809     }
5810 };
5811
5812 my $restore_deactivate_volumes = sub {
5813     my ($storecfg, $devinfo) = @_;
5814
5815     my $vollist = [];
5816     foreach my $devname (keys %$devinfo) {
5817         my $volid = $devinfo->{$devname}->{volid};
5818         push @$vollist, $volid if $volid;
5819     }
5820
5821     PVE::Storage::deactivate_volumes($storecfg, $vollist);
5822 };
5823
5824 my $restore_destroy_volumes = sub {
5825     my ($storecfg, $devinfo) = @_;
5826
5827     foreach my $devname (keys %$devinfo) {
5828         my $volid = $devinfo->{$devname}->{volid};
5829         next if !$volid;
5830         eval {
5831             if ($volid =~ m|^/|) {
5832                 unlink $volid || die 'unlink failed\n';
5833             } else {
5834                 PVE::Storage::vdisk_free($storecfg, $volid);
5835             }
5836             print STDERR "temporary volume '$volid' sucessfuly removed\n";
5837         };
5838         print STDERR "unable to cleanup '$volid' - $@" if $@;
5839     }
5840 };
5841
5842 sub scan_volids {
5843     my ($cfg, $vmid) = @_;
5844
5845     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5846
5847     my $volid_hash = {};
5848     foreach my $storeid (keys %$info) {
5849         foreach my $item (@{$info->{$storeid}}) {
5850             next if !($item->{volid} && $item->{size});
5851             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5852             $volid_hash->{$item->{volid}} = $item;
5853         }
5854     }
5855
5856     return $volid_hash;
5857 }
5858
5859 sub update_disk_config {
5860     my ($vmid, $conf, $volid_hash) = @_;
5861
5862     my $changes;
5863     my $prefix = "VM $vmid:";
5864
5865     # used and unused disks
5866     my $referenced = {};
5867
5868     # Note: it is allowed to define multiple storages with same path (alias), so
5869     # we need to check both 'volid' and real 'path' (two different volid can point
5870     # to the same path).
5871
5872     my $referencedpath = {};
5873
5874     # update size info
5875     PVE::QemuConfig->foreach_volume($conf, sub {
5876         my ($opt, $drive) = @_;
5877
5878         my $volid = $drive->{file};
5879         return if !$volid;
5880
5881         # mark volid as "in-use" for next step
5882         $referenced->{$volid} = 1;
5883         if ($volid_hash->{$volid} &&
5884             (my $path = $volid_hash->{$volid}->{path})) {
5885             $referencedpath->{$path} = 1;
5886         }
5887
5888         return if drive_is_cdrom($drive);
5889         return if !$volid_hash->{$volid};
5890
5891         my ($updated, $old_size, $new_size) = PVE::QemuServer::Drive::update_disksize($drive, $volid_hash);
5892         if (defined($updated)) {
5893             $changes = 1;
5894             $conf->{$opt} = print_drive($updated);
5895             print "$prefix size of disk '$volid' ($opt) updated from $old_size to $new_size\n";
5896         }
5897     });
5898
5899     # remove 'unusedX' entry if volume is used
5900     PVE::QemuConfig->foreach_unused_volume($conf, sub {
5901         my ($opt, $drive) = @_;
5902
5903         my $volid = $drive->{file};
5904         return if !$volid;
5905
5906         my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5907         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5908             print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
5909             $changes = 1;
5910             delete $conf->{$opt};
5911         }
5912
5913         $referenced->{$volid} = 1;
5914         $referencedpath->{$path} = 1 if $path;
5915     });
5916
5917     foreach my $volid (sort keys %$volid_hash) {
5918         next if $volid =~ m/vm-$vmid-state-/;
5919         next if $referenced->{$volid};
5920         my $path = $volid_hash->{$volid}->{path};
5921         next if !$path; # just to be sure
5922         next if $referencedpath->{$path};
5923         $changes = 1;
5924         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5925         print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
5926         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5927     }
5928
5929     return $changes;
5930 }
5931
5932 sub rescan {
5933     my ($vmid, $nolock, $dryrun) = @_;
5934
5935     my $cfg = PVE::Storage::config();
5936
5937     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5938     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
5939     foreach my $stor (keys %{$cfg->{ids}}) {
5940         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
5941     }
5942
5943     print "rescan volumes...\n";
5944     my $volid_hash = scan_volids($cfg, $vmid);
5945
5946     my $updatefn =  sub {
5947         my ($vmid) = @_;
5948
5949         my $conf = PVE::QemuConfig->load_config($vmid);
5950
5951         PVE::QemuConfig->check_lock($conf);
5952
5953         my $vm_volids = {};
5954         foreach my $volid (keys %$volid_hash) {
5955             my $info = $volid_hash->{$volid};
5956             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5957         }
5958
5959         my $changes = update_disk_config($vmid, $conf, $vm_volids);
5960
5961         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
5962     };
5963
5964     if (defined($vmid)) {
5965         if ($nolock) {
5966             &$updatefn($vmid);
5967         } else {
5968             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5969         }
5970     } else {
5971         my $vmlist = config_list();
5972         foreach my $vmid (keys %$vmlist) {
5973             if ($nolock) {
5974                 &$updatefn($vmid);
5975             } else {
5976                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5977             }
5978         }
5979     }
5980 }
5981
5982 sub restore_proxmox_backup_archive {
5983     my ($archive, $vmid, $user, $options) = @_;
5984
5985     my $storecfg = PVE::Storage::config();
5986
5987     my ($storeid, $volname) = PVE::Storage::parse_volume_id($archive);
5988     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5989
5990     my $server = $scfg->{server};
5991     my $datastore = $scfg->{datastore};
5992     my $username = $scfg->{username} // 'root@pam';
5993     my $fingerprint = $scfg->{fingerprint};
5994
5995     my $repo = "$username\@$server:$datastore";
5996     my $password = PVE::Storage::PBSPlugin::pbs_get_password($scfg, $storeid);
5997     local $ENV{PBS_PASSWORD} = $password;
5998     local $ENV{PBS_FINGERPRINT} = $fingerprint if defined($fingerprint);
5999
6000     my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6001         PVE::Storage::parse_volname($storecfg, $archive);
6002
6003     die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6004
6005     die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6006
6007     my $tmpdir = "/var/tmp/vzdumptmp$$";
6008     rmtree $tmpdir;
6009     mkpath $tmpdir;
6010
6011     my $conffile = PVE::QemuConfig->config_file($vmid);
6012     my $tmpfn = "$conffile.$$.tmp";
6013      # disable interrupts (always do cleanups)
6014     local $SIG{INT} =
6015         local $SIG{TERM} =
6016         local $SIG{QUIT} =
6017         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6018
6019     # Note: $oldconf is undef if VM does not exists
6020     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6021     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6022
6023     my $rpcenv = PVE::RPCEnvironment::get();
6024     my $devinfo = {};
6025
6026     eval {
6027         # enable interrupts
6028         local $SIG{INT} =
6029             local $SIG{TERM} =
6030             local $SIG{QUIT} =
6031             local $SIG{HUP} =
6032             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6033
6034         my $cfgfn = "$tmpdir/qemu-server.conf";
6035         my $firewall_config_fn = "$tmpdir/fw.conf";
6036         my $index_fn = "$tmpdir/index.json";
6037
6038         my $cmd = "restore";
6039
6040         my $param = [$pbs_backup_name, "index.json", $index_fn];
6041         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6042         my $index = PVE::Tools::file_get_contents($index_fn);
6043         $index = decode_json($index);
6044
6045         # print Dumper($index);
6046         foreach my $info (@{$index->{files}}) {
6047             if ($info->{filename} =~ m/^(drive-\S+).img.fidx$/) {
6048                 my $devname = $1;
6049                 if ($info->{size} =~ m/^(\d+)$/) { # untaint size
6050                     $devinfo->{$devname}->{size} = $1;
6051                 } else {
6052                     die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
6053                 }
6054             }
6055         }
6056
6057         my $is_qemu_server_backup = scalar(grep { $_->{filename} eq 'qemu-server.conf.blob' } @{$index->{files}});
6058         if (!$is_qemu_server_backup) {
6059             die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
6060         }
6061         my $has_firewall_config = scalar(grep { $_->{filename} eq 'fw.conf.blob' } @{$index->{files}});
6062
6063         $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
6064         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6065
6066         if ($has_firewall_config) {
6067             $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
6068             PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6069
6070             my $pve_firewall_dir = '/etc/pve/firewall';
6071             mkdir $pve_firewall_dir; # make sure the dir exists
6072             PVE::Tools::file_copy($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
6073         }
6074
6075         my $fh = IO::File->new($cfgfn, "r") ||
6076             "unable to read qemu-server.conf - $!\n";
6077
6078         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
6079
6080         # fixme: rate limit?
6081
6082         # create empty/temp config
6083         PVE::Tools::file_set_contents($conffile, "memory: 128\nlock: create");
6084
6085         $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
6086
6087         # allocate volumes
6088         my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
6089
6090         foreach my $virtdev (sort keys %$virtdev_hash) {
6091             my $d = $virtdev_hash->{$virtdev};
6092             next if $d->{is_cloudinit}; # no need to restore cloudinit
6093
6094             my $volid = $d->{volid};
6095
6096             my $path = PVE::Storage::path($storecfg, $volid);
6097
6098             my $pbs_restore_cmd = [
6099                 '/usr/bin/pbs-restore',
6100                 '--repository', $repo,
6101                 $pbs_backup_name,
6102                 "$d->{devname}.img.fidx",
6103                 $path,
6104                 '--verbose',
6105                 ];
6106
6107             if (PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $volid)) {
6108                 push @$pbs_restore_cmd, '--skip-zero';
6109             }
6110
6111             my $dbg_cmdstring = PVE::Tools::cmd2string($pbs_restore_cmd);
6112             print "restore proxmox backup image: $dbg_cmdstring\n";
6113             run_command($pbs_restore_cmd);
6114         }
6115
6116         $fh->seek(0, 0) || die "seek failed - $!\n";
6117
6118         my $outfd = new IO::File ($tmpfn, "w") ||
6119             die "unable to write config for VM $vmid\n";
6120
6121         my $cookie = { netcount => 0 };
6122         while (defined(my $line = <$fh>)) {
6123             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $options->{unique});
6124         }
6125
6126         $fh->close();
6127         $outfd->close();
6128     };
6129     my $err = $@;
6130
6131     $restore_deactivate_volumes->($storecfg, $devinfo);
6132
6133     rmtree $tmpdir;
6134
6135     if ($err) {
6136         unlink $tmpfn;
6137         $restore_destroy_volumes->($storecfg, $devinfo);
6138         die $err;
6139     }
6140
6141     rename($tmpfn, $conffile) ||
6142         die "unable to commit configuration file '$conffile'\n";
6143
6144     PVE::Cluster::cfs_update(); # make sure we read new file
6145
6146     eval { rescan($vmid, 1); };
6147     warn $@ if $@;
6148 }
6149
6150 sub restore_vma_archive {
6151     my ($archive, $vmid, $user, $opts, $comp) = @_;
6152
6153     my $readfrom = $archive;
6154
6155     my $cfg = PVE::Storage::config();
6156     my $commands = [];
6157     my $bwlimit = $opts->{bwlimit};
6158
6159     my $dbg_cmdstring = '';
6160     my $add_pipe = sub {
6161         my ($cmd) = @_;
6162         push @$commands, $cmd;
6163         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6164         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6165         $readfrom = '-';
6166     };
6167
6168     my $input = undef;
6169     if ($archive eq '-') {
6170         $input = '<&STDIN';
6171     } else {
6172         # If we use a backup from a PVE defined storage we also consider that
6173         # storage's rate limit:
6174         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6175         if (defined($volid)) {
6176             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6177             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6178             if ($readlimit) {
6179                 print STDERR "applying read rate limit: $readlimit\n";
6180                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6181                 $add_pipe->($cstream);
6182             }
6183         }
6184     }
6185
6186     if ($comp) {
6187         my $cmd;
6188         if ($comp eq 'gzip') {
6189             $cmd = ['zcat', $readfrom];
6190         } elsif ($comp eq 'lzop') {
6191             $cmd = ['lzop', '-d', '-c', $readfrom];
6192         } else {
6193             die "unknown compression method '$comp'\n";
6194         }
6195         $add_pipe->($cmd);
6196     }
6197
6198     my $tmpdir = "/var/tmp/vzdumptmp$$";
6199     rmtree $tmpdir;
6200
6201     # disable interrupts (always do cleanups)
6202     local $SIG{INT} =
6203         local $SIG{TERM} =
6204         local $SIG{QUIT} =
6205         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6206
6207     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6208     POSIX::mkfifo($mapfifo, 0600);
6209     my $fifofh;
6210
6211     my $openfifo = sub {
6212         open($fifofh, '>', $mapfifo) || die $!;
6213     };
6214
6215     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6216
6217     my $oldtimeout;
6218     my $timeout = 5;
6219
6220     my $devinfo = {};
6221
6222     my $rpcenv = PVE::RPCEnvironment::get();
6223
6224     my $conffile = PVE::QemuConfig->config_file($vmid);
6225     my $tmpfn = "$conffile.$$.tmp";
6226
6227     # Note: $oldconf is undef if VM does not exist
6228     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6229     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6230
6231     my %storage_limits;
6232
6233     my $print_devmap = sub {
6234         my $cfgfn = "$tmpdir/qemu-server.conf";
6235
6236         # we can read the config - that is already extracted
6237         my $fh = IO::File->new($cfgfn, "r") ||
6238             "unable to read qemu-server.conf - $!\n";
6239
6240         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6241         if (-f $fwcfgfn) {
6242             my $pve_firewall_dir = '/etc/pve/firewall';
6243             mkdir $pve_firewall_dir; # make sure the dir exists
6244             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6245         }
6246
6247         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
6248
6249         foreach my $key (keys %storage_limits) {
6250             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6251             next if !$limit;
6252             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6253             $storage_limits{$key} = $limit * 1024;
6254         }
6255
6256         foreach my $devname (keys %$devinfo) {
6257             die "found no device mapping information for device '$devname'\n"
6258                 if !$devinfo->{$devname}->{virtdev};
6259         }
6260
6261         # create empty/temp config
6262         if ($oldconf) {
6263             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6264             $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
6265         }
6266
6267         # allocate volumes
6268         my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
6269
6270         # print restore information to $fifofh
6271         foreach my $virtdev (sort keys %$virtdev_hash) {
6272             my $d = $virtdev_hash->{$virtdev};
6273             next if $d->{is_cloudinit}; # no need to restore cloudinit
6274
6275             my $storeid = $d->{storeid};
6276             my $volid = $d->{volid};
6277
6278             my $map_opts = '';
6279             if (my $limit = $storage_limits{$storeid}) {
6280                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6281             }
6282
6283             my $write_zeros = 1;
6284             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6285                 $write_zeros = 0;
6286             }
6287
6288             my $path = PVE::Storage::path($cfg, $volid);
6289
6290             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6291
6292             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6293         }
6294
6295         $fh->seek(0, 0) || die "seek failed - $!\n";
6296
6297         my $outfd = new IO::File ($tmpfn, "w") ||
6298             die "unable to write config for VM $vmid\n";
6299
6300         my $cookie = { netcount => 0 };
6301         while (defined(my $line = <$fh>)) {
6302             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6303         }
6304
6305         $fh->close();
6306         $outfd->close();
6307     };
6308
6309     eval {
6310         # enable interrupts
6311         local $SIG{INT} =
6312             local $SIG{TERM} =
6313             local $SIG{QUIT} =
6314             local $SIG{HUP} =
6315             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6316         local $SIG{ALRM} = sub { die "got timeout\n"; };
6317
6318         $oldtimeout = alarm($timeout);
6319
6320         my $parser = sub {
6321             my $line = shift;
6322
6323             print "$line\n";
6324
6325             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6326                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6327                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6328             } elsif ($line =~ m/^CTIME: /) {
6329                 # we correctly received the vma config, so we can disable
6330                 # the timeout now for disk allocation (set to 10 minutes, so
6331                 # that we always timeout if something goes wrong)
6332                 alarm(600);
6333                 &$print_devmap();
6334                 print $fifofh "done\n";
6335                 my $tmp = $oldtimeout || 0;
6336                 $oldtimeout = undef;
6337                 alarm($tmp);
6338                 close($fifofh);
6339             }
6340         };
6341
6342         print "restore vma archive: $dbg_cmdstring\n";
6343         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6344     };
6345     my $err = $@;
6346
6347     alarm($oldtimeout) if $oldtimeout;
6348
6349     $restore_deactivate_volumes->($cfg, $devinfo);
6350
6351     unlink $mapfifo;
6352     rmtree $tmpdir;
6353
6354     if ($err) {
6355         unlink $tmpfn;
6356         $restore_destroy_volumes->($cfg, $devinfo);
6357         die $err;
6358     }
6359
6360     rename($tmpfn, $conffile) ||
6361         die "unable to commit configuration file '$conffile'\n";
6362
6363     PVE::Cluster::cfs_update(); # make sure we read new file
6364
6365     eval { rescan($vmid, 1); };
6366     warn $@ if $@;
6367 }
6368
6369 sub restore_tar_archive {
6370     my ($archive, $vmid, $user, $opts) = @_;
6371
6372     if ($archive ne '-') {
6373         my $firstfile = tar_archive_read_firstfile($archive);
6374         die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6375             if $firstfile ne 'qemu-server.conf';
6376     }
6377
6378     my $storecfg = PVE::Storage::config();
6379
6380     # avoid zombie disks when restoring over an existing VM -> cleanup first
6381     # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
6382     # skiplock=1 because qmrestore has set the 'create' lock itself already
6383     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6384     destroy_vm($storecfg, $vmid, 1, { lock => 'restore' }) if -f $vmcfgfn;
6385
6386     my $tocmd = "/usr/lib/qemu-server/qmextract";
6387
6388     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6389     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6390     $tocmd .= ' --prealloc' if $opts->{prealloc};
6391     $tocmd .= ' --info' if $opts->{info};
6392
6393     # tar option "xf" does not autodetect compression when read from STDIN,
6394     # so we pipe to zcat
6395     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6396         PVE::Tools::shellquote("--to-command=$tocmd");
6397
6398     my $tmpdir = "/var/tmp/vzdumptmp$$";
6399     mkpath $tmpdir;
6400
6401     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6402     local $ENV{VZDUMP_VMID} = $vmid;
6403     local $ENV{VZDUMP_USER} = $user;
6404
6405     my $conffile = PVE::QemuConfig->config_file($vmid);
6406     my $tmpfn = "$conffile.$$.tmp";
6407
6408     # disable interrupts (always do cleanups)
6409     local $SIG{INT} =
6410         local $SIG{TERM} =
6411         local $SIG{QUIT} =
6412         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6413
6414     eval {
6415         # enable interrupts
6416         local $SIG{INT} =
6417             local $SIG{TERM} =
6418             local $SIG{QUIT} =
6419             local $SIG{HUP} =
6420             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6421
6422         if ($archive eq '-') {
6423             print "extracting archive from STDIN\n";
6424             run_command($cmd, input => "<&STDIN");
6425         } else {
6426             print "extracting archive '$archive'\n";
6427             run_command($cmd);
6428         }
6429
6430         return if $opts->{info};
6431
6432         # read new mapping
6433         my $map = {};
6434         my $statfile = "$tmpdir/qmrestore.stat";
6435         if (my $fd = IO::File->new($statfile, "r")) {
6436             while (defined (my $line = <$fd>)) {
6437                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6438                     $map->{$1} = $2 if $1;
6439                 } else {
6440                     print STDERR "unable to parse line in statfile - $line\n";
6441                 }
6442             }
6443             $fd->close();
6444         }
6445
6446         my $confsrc = "$tmpdir/qemu-server.conf";
6447
6448         my $srcfd = new IO::File($confsrc, "r") ||
6449             die "unable to open file '$confsrc'\n";
6450
6451         my $outfd = new IO::File ($tmpfn, "w") ||
6452             die "unable to write config for VM $vmid\n";
6453
6454         my $cookie = { netcount => 0 };
6455         while (defined (my $line = <$srcfd>)) {
6456             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6457         }
6458
6459         $srcfd->close();
6460         $outfd->close();
6461     };
6462     if (my $err = $@) {
6463         unlink $tmpfn;
6464         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6465         die $err;
6466     }
6467
6468     rmtree $tmpdir;
6469
6470     rename $tmpfn, $conffile ||
6471         die "unable to commit configuration file '$conffile'\n";
6472
6473     PVE::Cluster::cfs_update(); # make sure we read new file
6474
6475     eval { rescan($vmid, 1); };
6476     warn $@ if $@;
6477 };
6478
6479 sub foreach_storage_used_by_vm {
6480     my ($conf, $func) = @_;
6481
6482     my $sidhash = {};
6483
6484     foreach_drive($conf, sub {
6485         my ($ds, $drive) = @_;
6486         return if drive_is_cdrom($drive);
6487
6488         my $volid = $drive->{file};
6489
6490         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6491         $sidhash->{$sid} = $sid if $sid;
6492     });
6493
6494     foreach my $sid (sort keys %$sidhash) {
6495         &$func($sid);
6496     }
6497 }
6498
6499 my $qemu_snap_storage = {
6500     rbd => 1,
6501 };
6502 sub do_snapshots_with_qemu {
6503     my ($storecfg, $volid) = @_;
6504
6505     my $storage_name = PVE::Storage::parse_volume_id($volid);
6506     my $scfg = $storecfg->{ids}->{$storage_name};
6507
6508     if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6509         return 1;
6510     }
6511
6512     if ($volid =~ m/\.(qcow2|qed)$/){
6513         return 1;
6514     }
6515
6516     return undef;
6517 }
6518
6519 sub qga_check_running {
6520     my ($vmid, $nowarn) = @_;
6521
6522     eval { mon_cmd($vmid, "guest-ping", timeout => 3); };
6523     if ($@) {
6524         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6525         return 0;
6526     }
6527     return 1;
6528 }
6529
6530 sub template_create {
6531     my ($vmid, $conf, $disk) = @_;
6532
6533     my $storecfg = PVE::Storage::config();
6534
6535     foreach_drive($conf, sub {
6536         my ($ds, $drive) = @_;
6537
6538         return if drive_is_cdrom($drive);
6539         return if $disk && $ds ne $disk;
6540
6541         my $volid = $drive->{file};
6542         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6543
6544         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6545         $drive->{file} = $voliddst;
6546         $conf->{$ds} = print_drive($drive);
6547         PVE::QemuConfig->write_config($vmid, $conf);
6548     });
6549 }
6550
6551 sub convert_iscsi_path {
6552     my ($path) = @_;
6553
6554     if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6555         my $portal = $1;
6556         my $target = $2;
6557         my $lun = $3;
6558
6559         my $initiator_name = get_initiator_name();
6560
6561         return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6562                "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6563     }
6564
6565     die "cannot convert iscsi path '$path', unkown format\n";
6566 }
6567
6568 sub qemu_img_convert {
6569     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6570
6571     my $storecfg = PVE::Storage::config();
6572     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6573     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6574
6575     die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6576
6577     my $cachemode;
6578     my $src_path;
6579     my $src_is_iscsi = 0;
6580     my $src_format;
6581
6582     if ($src_storeid) {
6583         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6584         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6585         $src_format = qemu_img_format($src_scfg, $src_volname);
6586         $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6587         $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6588         $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6589     } elsif (-f $src_volid) {
6590         $src_path = $src_volid;
6591         if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6592             $src_format = $1;
6593         }
6594     }
6595
6596     die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6597
6598     my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6599     my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6600     my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6601     my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6602
6603     my $cmd = [];
6604     push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6605     push @$cmd, '-l', "snapshot.name=$snapname"
6606         if $snapname && $src_format && $src_format eq "qcow2";
6607     push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6608     push @$cmd, '-T', $cachemode if defined($cachemode);
6609
6610     if ($src_is_iscsi) {
6611         push @$cmd, '--image-opts';
6612         $src_path = convert_iscsi_path($src_path);
6613     } elsif ($src_format) {
6614         push @$cmd, '-f', $src_format;
6615     }
6616
6617     if ($dst_is_iscsi) {
6618         push @$cmd, '--target-image-opts';
6619         $dst_path = convert_iscsi_path($dst_path);
6620     } else {
6621         push @$cmd, '-O', $dst_format;
6622     }
6623
6624     push @$cmd, $src_path;
6625
6626     if (!$dst_is_iscsi && $is_zero_initialized) {
6627         push @$cmd, "zeroinit:$dst_path";
6628     } else {
6629         push @$cmd, $dst_path;
6630     }
6631
6632     my $parser = sub {
6633         my $line = shift;
6634         if($line =~ m/\((\S+)\/100\%\)/){
6635             my $percent = $1;
6636             my $transferred = int($size * $percent / 100);
6637             my $remaining = $size - $transferred;
6638
6639             print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6640         }
6641
6642     };
6643
6644     eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6645     my $err = $@;
6646     die "copy failed: $err" if $err;
6647 }
6648
6649 sub qemu_img_format {
6650     my ($scfg, $volname) = @_;
6651
6652     if ($scfg->{path} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6653         return $1;
6654     } else {
6655         return "raw";
6656     }
6657 }
6658
6659 sub qemu_drive_mirror {
6660     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
6661
6662     $jobs = {} if !$jobs;
6663
6664     my $qemu_target;
6665     my $format;
6666     $jobs->{"drive-$drive"} = {};
6667
6668     if ($dst_volid =~ /^nbd:/) {
6669         $qemu_target = $dst_volid;
6670         $format = "nbd";
6671     } else {
6672         my $storecfg = PVE::Storage::config();
6673         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6674
6675         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6676
6677         $format = qemu_img_format($dst_scfg, $dst_volname);
6678
6679         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6680
6681         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6682     }
6683
6684     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6685     $opts->{format} = $format if $format;
6686
6687     if (defined($src_bitmap)) {
6688         $opts->{sync} = 'incremental';
6689         $opts->{bitmap} = $src_bitmap;
6690         print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
6691     }
6692
6693     if (defined($bwlimit)) {
6694         $opts->{speed} = $bwlimit * 1024;
6695         print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
6696     } else {
6697         print "drive mirror is starting for drive-$drive\n";
6698     }
6699
6700     # if a job already runs for this device we get an error, catch it for cleanup
6701     eval { mon_cmd($vmid, "drive-mirror", %$opts); };
6702     if (my $err = $@) {
6703         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6704         warn "$@\n" if $@;
6705         die "mirroring error: $err\n";
6706     }
6707
6708     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $completion, $qga);
6709 }
6710
6711 # $completion can be either
6712 # 'complete': wait until all jobs are ready, block-job-complete them (default)
6713 # 'cancel': wait until all jobs are ready, block-job-cancel them
6714 # 'skip': wait until all jobs are ready, return with block jobs in ready state
6715 sub qemu_drive_mirror_monitor {
6716     my ($vmid, $vmiddst, $jobs, $completion, $qga) = @_;
6717
6718     $completion //= 'complete';
6719
6720     eval {
6721         my $err_complete = 0;
6722
6723         while (1) {
6724             die "storage migration timed out\n" if $err_complete > 300;
6725
6726             my $stats = mon_cmd($vmid, "query-block-jobs");
6727
6728             my $running_mirror_jobs = {};
6729             foreach my $stat (@$stats) {
6730                 next if $stat->{type} ne 'mirror';
6731                 $running_mirror_jobs->{$stat->{device}} = $stat;
6732             }
6733
6734             my $readycounter = 0;
6735
6736             foreach my $job (keys %$jobs) {
6737
6738                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6739                     print "$job : finished\n";
6740                     delete $jobs->{$job};
6741                     next;
6742                 }
6743
6744                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6745
6746                 my $busy = $running_mirror_jobs->{$job}->{busy};
6747                 my $ready = $running_mirror_jobs->{$job}->{ready};
6748                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6749                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6750                     my $remaining = $total - $transferred;
6751                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6752
6753                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6754                 }
6755
6756                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6757             }
6758
6759             last if scalar(keys %$jobs) == 0;
6760
6761             if ($readycounter == scalar(keys %$jobs)) {
6762                 print "all mirroring jobs are ready \n";
6763                 last if $completion eq 'skip'; #do the complete later
6764
6765                 if ($vmiddst && $vmiddst != $vmid) {
6766                     my $agent_running = $qga && qga_check_running($vmid);
6767                     if ($agent_running) {
6768                         print "freeze filesystem\n";
6769                         eval { mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6770                     } else {
6771                         print "suspend vm\n";
6772                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6773                     }
6774
6775                     # if we clone a disk for a new target vm, we don't switch the disk
6776                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6777
6778                     if ($agent_running) {
6779                         print "unfreeze filesystem\n";
6780                         eval { mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6781                     } else {
6782                         print "resume vm\n";
6783                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6784                     }
6785
6786                     last;
6787                 } else {
6788
6789                     foreach my $job (keys %$jobs) {
6790                         # try to switch the disk if source and destination are on the same guest
6791                         print "$job: Completing block job...\n";
6792
6793                         my $op;
6794                         if ($completion eq 'complete') {
6795                             $op = 'block-job-complete';
6796                         } elsif ($completion eq 'cancel') {
6797                             $op = 'block-job-cancel';
6798                         } else {
6799                             die "invalid completion value: $completion\n";
6800                         }
6801                         eval { mon_cmd($vmid, $op, device => $job) };
6802                         if ($@ =~ m/cannot be completed/) {
6803                             print "$job: Block job cannot be completed, try again.\n";
6804                             $err_complete++;
6805                         }else {
6806                             print "$job: Completed successfully.\n";
6807                             $jobs->{$job}->{complete} = 1;
6808                         }
6809                     }
6810                 }
6811             }
6812             sleep 1;
6813         }
6814     };
6815     my $err = $@;
6816
6817     if ($err) {
6818         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6819         die "mirroring error: $err";
6820     }
6821
6822 }
6823
6824 sub qemu_blockjobs_cancel {
6825     my ($vmid, $jobs) = @_;
6826
6827     foreach my $job (keys %$jobs) {
6828         print "$job: Cancelling block job\n";
6829         eval { mon_cmd($vmid, "block-job-cancel", device => $job); };
6830         $jobs->{$job}->{cancel} = 1;
6831     }
6832
6833     while (1) {
6834         my $stats = mon_cmd($vmid, "query-block-jobs");
6835
6836         my $running_jobs = {};
6837         foreach my $stat (@$stats) {
6838             $running_jobs->{$stat->{device}} = $stat;
6839         }
6840
6841         foreach my $job (keys %$jobs) {
6842
6843             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6844                 print "$job: Done.\n";
6845                 delete $jobs->{$job};
6846             }
6847         }
6848
6849         last if scalar(keys %$jobs) == 0;
6850
6851         sleep 1;
6852     }
6853 }
6854
6855 sub clone_disk {
6856     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6857         $newvmid, $storage, $format, $full, $newvollist, $jobs, $completion, $qga, $bwlimit, $conf) = @_;
6858
6859     my $newvolid;
6860
6861     if (!$full) {
6862         print "create linked clone of drive $drivename ($drive->{file})\n";
6863         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6864         push @$newvollist, $newvolid;
6865     } else {
6866
6867         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6868         $storeid = $storage if $storage;
6869
6870         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6871         my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6872
6873         print "create full clone of drive $drivename ($drive->{file})\n";
6874         my $name = undef;
6875         if (drive_is_cloudinit($drive)) {
6876             $name = "vm-$newvmid-cloudinit";
6877             $name .= ".$dst_format" if $dst_format ne 'raw';
6878             $snapname = undef;
6879             $size = PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE;
6880         } elsif ($drivename eq 'efidisk0') {
6881             $size = get_efivars_size($conf);
6882         }
6883         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
6884         push @$newvollist, $newvolid;
6885
6886         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6887
6888         if (drive_is_cloudinit($drive)) {
6889             goto no_data_clone;
6890         }
6891
6892         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6893         if (!$running || $snapname) {
6894             # TODO: handle bwlimits
6895             if ($drivename eq 'efidisk0') {
6896                 # the relevant data on the efidisk may be smaller than the source
6897                 # e.g. on RBD/ZFS, so we use dd to copy only the amount
6898                 # that is given by the OVMF_VARS.fd
6899                 my $src_path = PVE::Storage::path($storecfg, $drive->{file});
6900                 my $dst_path = PVE::Storage::path($storecfg, $newvolid);
6901                 run_command(['qemu-img', 'dd', '-n', '-O', $dst_format, "bs=1", "count=$size", "if=$src_path", "of=$dst_path"]);
6902             } else {
6903                 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6904             }
6905         } else {
6906
6907             my $kvmver = get_running_qemu_version ($vmid);
6908             if (!min_version($kvmver, 2, 7)) {
6909                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6910                     if $drive->{iothread};
6911             }
6912
6913             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $completion, $qga, $bwlimit);
6914         }
6915     }
6916
6917 no_data_clone:
6918     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6919
6920     my $disk = $drive;
6921     $disk->{format} = undef;
6922     $disk->{file} = $newvolid;
6923     $disk->{size} = $size;
6924
6925     return $disk;
6926 }
6927
6928 sub get_running_qemu_version {
6929     my ($vmid) = @_;
6930     my $res = mon_cmd($vmid, "query-version");
6931     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6932 }
6933
6934 sub qemu_use_old_bios_files {
6935     my ($machine_type) = @_;
6936
6937     return if !$machine_type;
6938
6939     my $use_old_bios_files = undef;
6940
6941     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6942         $machine_type = $1;
6943         $use_old_bios_files = 1;
6944     } else {
6945         my $version = PVE::QemuServer::Machine::extract_version($machine_type, kvm_user_version());
6946         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6947         # load new efi bios files on migration. So this hack is required to allow
6948         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6949         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6950         $use_old_bios_files = !min_version($version, 2, 4);
6951     }
6952
6953     return ($use_old_bios_files, $machine_type);
6954 }
6955
6956 sub get_efivars_size {
6957     my ($conf) = @_;
6958     my $arch = get_vm_arch($conf);
6959     my (undef, $ovmf_vars) = get_ovmf_files($arch);
6960     die "uefi vars image '$ovmf_vars' not found\n" if ! -f $ovmf_vars;
6961     return -s $ovmf_vars;
6962 }
6963
6964 sub update_efidisk_size {
6965     my ($conf) = @_;
6966
6967     return if !defined($conf->{efidisk0});
6968
6969     my $disk = PVE::QemuServer::parse_drive('efidisk0', $conf->{efidisk0});
6970     $disk->{size} = get_efivars_size($conf);
6971     $conf->{efidisk0} = print_drive($disk);
6972
6973     return;
6974 }
6975
6976 sub create_efidisk($$$$$) {
6977     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
6978
6979     my (undef, $ovmf_vars) = get_ovmf_files($arch);
6980     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
6981
6982     my $vars_size_b = -s $ovmf_vars;
6983     my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
6984     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
6985     PVE::Storage::activate_volumes($storecfg, [$volid]);
6986
6987     qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
6988     my ($size) = PVE::Storage::volume_size_info($storecfg, $volid, 3);
6989
6990     return ($volid, $size/1024);
6991 }
6992
6993 sub vm_iothreads_list {
6994     my ($vmid) = @_;
6995
6996     my $res = mon_cmd($vmid, 'query-iothreads');
6997
6998     my $iothreads = {};
6999     foreach my $iothread (@$res) {
7000         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7001     }
7002
7003     return $iothreads;
7004 }
7005
7006 sub scsihw_infos {
7007     my ($conf, $drive) = @_;
7008
7009     my $maxdev = 0;
7010
7011     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7012         $maxdev = 7;
7013     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7014         $maxdev = 1;
7015     } else {
7016         $maxdev = 256;
7017     }
7018
7019     my $controller = int($drive->{index} / $maxdev);
7020     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7021
7022     return ($maxdev, $controller, $controller_prefix);
7023 }
7024
7025 sub windows_version {
7026     my ($ostype) = @_;
7027
7028     return 0 if !$ostype;
7029
7030     my $winversion = 0;
7031
7032     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7033         $winversion = 5;
7034     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7035         $winversion = 6;
7036     } elsif ($ostype =~ m/^win(\d+)$/) {
7037         $winversion = $1;
7038     }
7039
7040     return $winversion;
7041 }
7042
7043 sub resolve_dst_disk_format {
7044         my ($storecfg, $storeid, $src_volname, $format) = @_;
7045         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7046
7047         if (!$format) {
7048             # if no target format is specified, use the source disk format as hint
7049             if ($src_volname) {
7050                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7051                 $format = qemu_img_format($scfg, $src_volname);
7052             } else {
7053                 return $defFormat;
7054             }
7055         }
7056
7057         # test if requested format is supported - else use default
7058         my $supported = grep { $_ eq $format } @$validFormats;
7059         $format = $defFormat if !$supported;
7060         return $format;
7061 }
7062
7063 # NOTE: if this logic changes, please update docs & possibly gui logic
7064 sub find_vmstate_storage {
7065     my ($conf, $storecfg) = @_;
7066
7067     # first, return storage from conf if set
7068     return $conf->{vmstatestorage} if $conf->{vmstatestorage};
7069
7070     my ($target, $shared, $local);
7071
7072     foreach_storage_used_by_vm($conf, sub {
7073         my ($sid) = @_;
7074         my $scfg = PVE::Storage::storage_config($storecfg, $sid);
7075         my $dst = $scfg->{shared} ? \$shared : \$local;
7076         $$dst = $sid if !$$dst || $scfg->{path}; # prefer file based storage
7077     });
7078
7079     # second, use shared storage where VM has at least one disk
7080     # third, use local storage where VM has at least one disk
7081     # fall back to local storage
7082     $target = $shared // $local // 'local';
7083
7084     return $target;
7085 }
7086
7087 sub generate_uuid {
7088     my ($uuid, $uuid_str);
7089     UUID::generate($uuid);
7090     UUID::unparse($uuid, $uuid_str);
7091     return $uuid_str;
7092 }
7093
7094 sub generate_smbios1_uuid {
7095     return "uuid=".generate_uuid();
7096 }
7097
7098 sub nbd_stop {
7099     my ($vmid) = @_;
7100
7101     mon_cmd($vmid, 'nbd-server-stop');
7102 }
7103
7104 sub create_reboot_request {
7105     my ($vmid) = @_;
7106     open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7107         or die "failed to create reboot trigger file: $!\n";
7108     close($fh);
7109 }
7110
7111 sub clear_reboot_request {
7112     my ($vmid) = @_;
7113     my $path = "/run/qemu-server/$vmid.reboot";
7114     my $res = 0;
7115
7116     $res = unlink($path);
7117     die "could not remove reboot request for $vmid: $!"
7118         if !$res && $! != POSIX::ENOENT;
7119
7120     return $res;
7121 }
7122
7123 # bash completion helper
7124
7125 sub complete_backup_archives {
7126     my ($cmdname, $pname, $cvalue) = @_;
7127
7128     my $cfg = PVE::Storage::config();
7129
7130     my $storeid;
7131
7132     if ($cvalue =~ m/^([^:]+):/) {
7133         $storeid = $1;
7134     }
7135
7136     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7137
7138     my $res = [];
7139     foreach my $id (keys %$data) {
7140         foreach my $item (@{$data->{$id}}) {
7141             next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7142             push @$res, $item->{volid} if defined($item->{volid});
7143         }
7144     }
7145
7146     return $res;
7147 }
7148
7149 my $complete_vmid_full = sub {
7150     my ($running) = @_;
7151
7152     my $idlist = vmstatus();
7153
7154     my $res = [];
7155
7156     foreach my $id (keys %$idlist) {
7157         my $d = $idlist->{$id};
7158         if (defined($running)) {
7159             next if $d->{template};
7160             next if $running && $d->{status} ne 'running';
7161             next if !$running && $d->{status} eq 'running';
7162         }
7163         push @$res, $id;
7164
7165     }
7166     return $res;
7167 };
7168
7169 sub complete_vmid {
7170     return &$complete_vmid_full();
7171 }
7172
7173 sub complete_vmid_stopped {
7174     return &$complete_vmid_full(0);
7175 }
7176
7177 sub complete_vmid_running {
7178     return &$complete_vmid_full(1);
7179 }
7180
7181 sub complete_storage {
7182
7183     my $cfg = PVE::Storage::config();
7184     my $ids = $cfg->{ids};
7185
7186     my $res = [];
7187     foreach my $sid (keys %$ids) {
7188         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7189         next if !$ids->{$sid}->{content}->{images};
7190         push @$res, $sid;
7191     }
7192
7193     return $res;
7194 }
7195
7196 sub complete_migration_storage {
7197     my ($cmd, $param, $current_value, $all_args) = @_;
7198
7199     my $targetnode = @$all_args[1];
7200
7201     my $cfg = PVE::Storage::config();
7202     my $ids = $cfg->{ids};
7203
7204     my $res = [];
7205     foreach my $sid (keys %$ids) {
7206         next if !PVE::Storage::storage_check_enabled($cfg, $sid, $targetnode, 1);
7207         next if !$ids->{$sid}->{content}->{images};
7208         push @$res, $sid;
7209     }
7210
7211     return $res;
7212 }
7213
7214 1;