PVE/QemuServer.pm

   1 package PVE::QemuServer;
   2
   3 use strict;
   4 use warnings;
   5
   6 use Cwd 'abs_path';
   7 use Digest::SHA;
   8 use Fcntl ':flock';
   9 use Fcntl;
  10 use File::Basename;
  11 use File::Copy qw(copy);
  12 use File::Path;
  13 use File::stat;
  14 use Getopt::Long;
  15 use IO::Dir;
  16 use IO::File;
  17 use IO::Handle;
  18 use IO::Select;
  19 use IO::Socket::UNIX;
  20 use IPC::Open3;
  21 use JSON;
  22 use MIME::Base64;
  23 use POSIX;
  24 use Storable qw(dclone);
  25 use Time::HiRes qw(gettimeofday);
  26 use URI::Escape;
  27 use UUID;
  28
  29 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file);
  30 use PVE::DataCenterConfig;
  31 use PVE::Exception qw(raise raise_param_exc);
  32 use PVE::GuestHelpers qw(safe_string_ne safe_num_ne safe_boolean_ne);
  33 use PVE::INotify;
  34 use PVE::JSONSchema qw(get_standard_option);
  35 use PVE::ProcFSTools;
  36 use PVE::RPCEnvironment;
  37 use PVE::Storage;
  38 use PVE::SysFSTools;
  39 use PVE::Systemd;
  40 use PVE::Tools qw(run_command file_read_firstline file_get_contents dir_glob_foreach get_host_arch $IPV6RE);
  41
  42 use PVE::QMPClient;
  43 use PVE::QemuConfig;
  44 use PVE::QemuServer::Helpers qw(min_version config_aware_timeout);
  45 use PVE::QemuServer::Cloudinit;
  46 use PVE::QemuServer::CPUConfig qw(print_cpu_device get_cpu_options);
  47 use PVE::QemuServer::Drive qw(is_valid_drivename drive_is_cloudinit drive_is_cdrom parse_drive print_drive);
  48 use PVE::QemuServer::Machine;
  49 use PVE::QemuServer::Memory;
  50 use PVE::QemuServer::Monitor qw(mon_cmd);
  51 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr print_pcie_root_port);
  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 ($conf->{template}) {
2040         # check if any base image is still used by a linked clone
2041         PVE::QemuConfig->foreach_volume($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, $volid);
2050
2051         });
2052     }
2053
2054     # only remove disks owned by this VM
2055     PVE::QemuConfig->foreach_volume($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, $node) = @_;
2343
2344     PVE::QemuConfig->foreach_volume($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 $nodename = nodename();
2366
2367     PVE::QemuConfig->foreach_volume($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     PVE::QemuConfig->foreach_volume($conf, sub {
2400         my ($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${id}",
2751     };
2752 }
2753
2754 sub audio_devs {
2755     my ($audio, $audiopciaddr, $machine_version) = @_;
2756
2757     my $devs = [];
2758
2759     my $id = $audio->{dev_id};
2760     my $audiodev = "";
2761     if (min_version($machine_version, 4, 2)) {
2762         $audiodev = ",audiodev=$audio->{backend_id}";
2763     }
2764
2765     if ($audio->{dev} eq 'AC97') {
2766         push @$devs, '-device', "AC97,id=${id}${audiopciaddr}$audiodev";
2767     } elsif ($audio->{dev} =~ /intel\-hda$/) {
2768         push @$devs, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
2769         push @$devs, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0$audiodev";
2770         push @$devs, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1$audiodev";
2771     } else {
2772         die "unkown audio device '$audio->{dev}', implement me!";
2773     }
2774
2775     push @$devs, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
2776
2777     return $devs;
2778 }
2779
2780 sub vga_conf_has_spice {
2781     my ($vga) = @_;
2782
2783     my $vgaconf = parse_vga($vga);
2784     my $vgatype = $vgaconf->{type};
2785     return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
2786
2787     return $1 || 1;
2788 }
2789
2790 sub is_native($) {
2791     my ($arch) = @_;
2792     return get_host_arch() eq $arch;
2793 }
2794
2795 sub get_vm_arch {
2796     my ($conf) = @_;
2797     return $conf->{arch} // get_host_arch();
2798 }
2799
2800 my $default_machines = {
2801     x86_64 => 'pc',
2802     aarch64 => 'virt',
2803 };
2804
2805 sub get_vm_machine {
2806     my ($conf, $forcemachine, $arch, $add_pve_version, $kvmversion) = @_;
2807
2808     my $machine = $forcemachine || $conf->{machine};
2809
2810     if (!$machine || $machine =~ m/^(?:pc|q35|virt)$/) {
2811         $arch //= 'x86_64';
2812         $machine ||= $default_machines->{$arch};
2813         if ($add_pve_version) {
2814             $kvmversion //= kvm_user_version();
2815             my $pvever = PVE::QemuServer::Machine::get_pve_version($kvmversion);
2816             $machine .= "+pve$pvever";
2817         }
2818     }
2819
2820     if ($add_pve_version && $machine !~ m/\+pve\d+$/) {
2821         # for version-pinned machines that do not include a pve-version (e.g.
2822         # pc-q35-4.1), we assume 0 to keep them stable in case we bump
2823         $machine .= '+pve0';
2824     }
2825
2826     return $machine;
2827 }
2828
2829 sub get_ovmf_files($) {
2830     my ($arch) = @_;
2831
2832     my $ovmf = $OVMF->{$arch}
2833         or die "no OVMF images known for architecture '$arch'\n";
2834
2835     return @$ovmf;
2836 }
2837
2838 my $Arch2Qemu = {
2839     aarch64 => '/usr/bin/qemu-system-aarch64',
2840     x86_64 => '/usr/bin/qemu-system-x86_64',
2841 };
2842 sub get_command_for_arch($) {
2843     my ($arch) = @_;
2844     return '/usr/bin/kvm' if is_native($arch);
2845
2846     my $cmd = $Arch2Qemu->{$arch}
2847         or die "don't know how to emulate architecture '$arch'\n";
2848     return $cmd;
2849 }
2850
2851 # To use query_supported_cpu_flags and query_understood_cpu_flags to get flags
2852 # to use in a QEMU command line (-cpu element), first array_intersect the result
2853 # of query_supported_ with query_understood_. This is necessary because:
2854 #
2855 # a) query_understood_ returns flags the host cannot use and
2856 # b) query_supported_ (rather the QMP call) doesn't actually return CPU
2857 #    flags, but CPU settings - with most of them being flags. Those settings
2858 #    (and some flags, curiously) cannot be specified as a "-cpu" argument.
2859 #
2860 # query_supported_ needs to start up to 2 temporary VMs and is therefore rather
2861 # expensive. If you need the value returned from this, you can get it much
2862 # cheaper from pmxcfs using PVE::Cluster::get_node_kv('cpuflags-$accel') with
2863 # $accel being 'kvm' or 'tcg'.
2864 #
2865 # pvestatd calls this function on startup and whenever the QEMU/KVM version
2866 # changes, automatically populating pmxcfs.
2867 #
2868 # Returns: { kvm => [ flagX, flagY, ... ], tcg => [ flag1, flag2, ... ] }
2869 # since kvm and tcg machines support different flags
2870 #
2871 sub query_supported_cpu_flags {
2872     my ($arch) = @_;
2873
2874     $arch //= get_host_arch();
2875     my $default_machine = $default_machines->{$arch};
2876
2877     my $flags = {};
2878
2879     # FIXME: Once this is merged, the code below should work for ARM as well:
2880     # https://lists.nongnu.org/archive/html/qemu-devel/2019-06/msg04947.html
2881     die "QEMU/KVM cannot detect CPU flags on ARM (aarch64)\n" if
2882         $arch eq "aarch64";
2883
2884     my $kvm_supported = defined(kvm_version());
2885     my $qemu_cmd = get_command_for_arch($arch);
2886     my $fakevmid = -1;
2887     my $pidfile = PVE::QemuServer::Helpers::pidfile_name($fakevmid);
2888
2889     # Start a temporary (frozen) VM with vmid -1 to allow sending a QMP command
2890     my $query_supported_run_qemu = sub {
2891         my ($kvm) = @_;
2892
2893         my $flags = {};
2894         my $cmd = [
2895             $qemu_cmd,
2896             '-machine', $default_machine,
2897             '-display', 'none',
2898             '-chardev', "socket,id=qmp,path=/var/run/qemu-server/$fakevmid.qmp,server,nowait",
2899             '-mon', 'chardev=qmp,mode=control',
2900             '-pidfile', $pidfile,
2901             '-S', '-daemonize'
2902         ];
2903
2904         if (!$kvm) {
2905             push @$cmd, '-accel', 'tcg';
2906         }
2907
2908         my $rc = run_command($cmd, noerr => 1, quiet => 0);
2909         die "QEMU flag querying VM exited with code " . $rc if $rc;
2910
2911         eval {
2912             my $cmd_result = mon_cmd(
2913                 $fakevmid,
2914                 'query-cpu-model-expansion',
2915                 type => 'full',
2916                 model => { name => 'host' }
2917             );
2918
2919             my $props = $cmd_result->{model}->{props};
2920             foreach my $prop (keys %$props) {
2921                 next if $props->{$prop} ne '1';
2922                 # QEMU returns some flags multiple times, with '_', '.' or '-'
2923                 # (e.g. lahf_lm and lahf-lm; sse4.2, sse4-2 and sse4_2; ...).
2924                 # We only keep those with underscores, to match /proc/cpuinfo
2925                 $prop =~ s/\.|-/_/g;
2926                 $flags->{$prop} = 1;
2927             }
2928         };
2929         my $err = $@;
2930
2931         # force stop with 10 sec timeout and 'nocheck'
2932         # always stop, even if QMP failed
2933         vm_stop(undef, $fakevmid, 1, 1, 10, 0, 1);
2934
2935         die $err if $err;
2936
2937         return [ sort keys %$flags ];
2938     };
2939
2940     # We need to query QEMU twice, since KVM and TCG have different supported flags
2941     PVE::QemuConfig->lock_config($fakevmid, sub {
2942         $flags->{tcg} = eval { $query_supported_run_qemu->(0) };
2943         warn "warning: failed querying supported tcg flags: $@\n" if $@;
2944
2945         if ($kvm_supported) {
2946             $flags->{kvm} = eval { $query_supported_run_qemu->(1) };
2947             warn "warning: failed querying supported kvm flags: $@\n" if $@;
2948         }
2949     });
2950
2951     return $flags;
2952 }
2953
2954 # Understood CPU flags are written to a file at 'pve-qemu' compile time
2955 my $understood_cpu_flag_dir = "/usr/share/kvm";
2956 sub query_understood_cpu_flags {
2957     my $arch = get_host_arch();
2958     my $filepath = "$understood_cpu_flag_dir/recognized-CPUID-flags-$arch";
2959
2960     die "Cannot query understood QEMU CPU flags for architecture: $arch (file not found)\n"
2961         if ! -e $filepath;
2962
2963     my $raw = file_get_contents($filepath);
2964     $raw =~ s/^\s+|\s+$//g;
2965     my @flags = split(/\s+/, $raw);
2966
2967     return \@flags;
2968 }
2969
2970 sub config_to_command {
2971     my ($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu) = @_;
2972
2973     my $cmd = [];
2974     my $globalFlags = [];
2975     my $machineFlags = [];
2976     my $rtcFlags = [];
2977     my $devices = [];
2978     my $pciaddr = '';
2979     my $bridges = {};
2980     my $ostype = $conf->{ostype};
2981     my $winversion = windows_version($ostype);
2982     my $kvm = $conf->{kvm};
2983     my $nodename = nodename();
2984
2985     my $arch = get_vm_arch($conf);
2986     my $kvm_binary = get_command_for_arch($arch);
2987     my $kvmver = kvm_user_version($kvm_binary);
2988
2989     if (!$kvmver || $kvmver !~ m/^(\d+)\.(\d+)/ || $1 < 3) {
2990         $kvmver //= "undefined";
2991         die "Detected old QEMU binary ('$kvmver', at least 3.0 is required)\n";
2992     }
2993
2994     my $add_pve_version = min_version($kvmver, 4, 1);
2995
2996     my $machine_type = get_vm_machine($conf, $forcemachine, $arch, $add_pve_version);
2997     my $machine_version = PVE::QemuServer::Machine::extract_version($machine_type, $kvmver);
2998     $kvm //= 1 if is_native($arch);
2999
3000     $machine_version =~ m/(\d+)\.(\d+)/;
3001     my ($machine_major, $machine_minor) = ($1, $2);
3002
3003     if ($kvmver =~ m/^\d+\.\d+\.(\d+)/ && $1 >= 90) {
3004         warn "warning: Installed QEMU version ($kvmver) is a release candidate, ignoring version checks\n";
3005     } elsif (!min_version($kvmver, $machine_major, $machine_minor)) {
3006         die "Installed QEMU version '$kvmver' is too old to run machine type '$machine_type', please upgrade node '$nodename'\n"
3007     } elsif (!PVE::QemuServer::Machine::can_run_pve_machine_version($machine_version, $kvmver)) {
3008         my $max_pve_version = PVE::QemuServer::Machine::get_pve_version($machine_version);
3009         die "Installed qemu-server (max feature level for $machine_major.$machine_minor is pve$max_pve_version)"
3010             ." is too old to run machine type '$machine_type', please upgrade node '$nodename'\n";
3011     }
3012
3013     # if a specific +pve version is required for a feature, use $version_guard
3014     # instead of min_version to allow machines to be run with the minimum
3015     # required version
3016     my $required_pve_version = 0;
3017     my $version_guard = sub {
3018         my ($major, $minor, $pve) = @_;
3019         return 0 if !min_version($machine_version, $major, $minor, $pve);
3020         my $max_pve = PVE::QemuServer::Machine::get_pve_version("$major.$minor");
3021         return 1 if min_version($machine_version, $major, $minor, $max_pve+1);
3022         $required_pve_version = $pve if $pve && $pve > $required_pve_version;
3023         return 1;
3024     };
3025
3026     if ($kvm) {
3027         die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3028             if !defined kvm_version();
3029     }
3030
3031     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3032     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3033     my $use_old_bios_files = undef;
3034     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3035
3036     my $cpuunits = defined($conf->{cpuunits}) ?
3037             $conf->{cpuunits} : $defaults->{cpuunits};
3038
3039     push @$cmd, $kvm_binary;
3040
3041     push @$cmd, '-id', $vmid;
3042
3043     my $vmname = $conf->{name} || "vm$vmid";
3044
3045     push @$cmd, '-name', $vmname;
3046
3047     my $use_virtio = 0;
3048
3049     my $qmpsocket = PVE::QemuServer::Helpers::qmp_socket($vmid);
3050     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3051     push @$cmd, '-mon', "chardev=qmp,mode=control";
3052
3053     if (min_version($machine_version, 2, 12)) {
3054         push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3055         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3056     }
3057
3058     push @$cmd, '-pidfile' , PVE::QemuServer::Helpers::pidfile_name($vmid);
3059
3060     push @$cmd, '-daemonize';
3061
3062     if ($conf->{smbios1}) {
3063         my $smbios_conf = parse_smbios1($conf->{smbios1});
3064         if ($smbios_conf->{base64}) {
3065             # Do not pass base64 flag to qemu
3066             delete $smbios_conf->{base64};
3067             my $smbios_string = "";
3068             foreach my $key (keys %$smbios_conf) {
3069                 my $value;
3070                 if ($key eq "uuid") {
3071                     $value = $smbios_conf->{uuid}
3072                 } else {
3073                     $value = decode_base64($smbios_conf->{$key});
3074                 }
3075                 # qemu accepts any binary data, only commas need escaping by double comma
3076                 $value =~ s/,/,,/g;
3077                 $smbios_string .= "," . $key . "=" . $value if $value;
3078             }
3079             push @$cmd, '-smbios', "type=1" . $smbios_string;
3080         } else {
3081             push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3082         }
3083     }
3084
3085     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3086         my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3087         die "uefi base image '$ovmf_code' not found\n" if ! -f $ovmf_code;
3088
3089         my ($path, $format);
3090         if (my $efidisk = $conf->{efidisk0}) {
3091             my $d = parse_drive('efidisk0', $efidisk);
3092             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3093             $format = $d->{format};
3094             if ($storeid) {
3095                 $path = PVE::Storage::path($storecfg, $d->{file});
3096                 if (!defined($format)) {
3097                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3098                     $format = qemu_img_format($scfg, $volname);
3099                 }
3100             } else {
3101                 $path = $d->{file};
3102                 die "efidisk format must be specified\n"
3103                     if !defined($format);
3104             }
3105         } else {
3106             warn "no efidisk configured! Using temporary efivars disk.\n";
3107             $path = "/tmp/$vmid-ovmf.fd";
3108             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3109             $format = 'raw';
3110         }
3111
3112         my $size_str = "";
3113
3114         if ($format eq 'raw' && $version_guard->(4, 1, 2)) {
3115             $size_str = ",size=" . (-s $ovmf_vars);
3116         }
3117
3118         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3119         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0$size_str,file=$path";
3120     }
3121
3122     # load q35 config
3123     if ($q35) {
3124         # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3125         if (min_version($machine_version, 4, 0)) {
3126             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3127         } else {
3128             push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3129         }
3130     }
3131
3132     if ($conf->{vmgenid}) {
3133         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3134     }
3135
3136     # add usb controllers
3137     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3138     push @$devices, @usbcontrollers if @usbcontrollers;
3139     my $vga = parse_vga($conf->{vga});
3140
3141     my $qxlnum = vga_conf_has_spice($conf->{vga});
3142     $vga->{type} = 'qxl' if $qxlnum;
3143
3144     if (!$vga->{type}) {
3145         if ($arch eq 'aarch64') {
3146             $vga->{type} = 'virtio';
3147         } elsif (min_version($machine_version, 2, 9)) {
3148             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3149         } else {
3150             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3151         }
3152     }
3153
3154     # enable absolute mouse coordinates (needed by vnc)
3155     my $tablet;
3156     if (defined($conf->{tablet})) {
3157         $tablet = $conf->{tablet};
3158     } else {
3159         $tablet = $defaults->{tablet};
3160         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3161         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3162     }
3163
3164     if ($tablet) {
3165         push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3166         my $kbd = print_keyboarddevice_full($conf, $arch);
3167         push @$devices, '-device', $kbd if defined($kbd);
3168     }
3169
3170     my $kvm_off = 0;
3171     my $gpu_passthrough;
3172
3173     # host pci devices
3174     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
3175         my $id = "hostpci$i";
3176         my $d = parse_hostpci($conf->{$id});
3177         next if !$d;
3178
3179         if (my $pcie = $d->{pcie}) {
3180             die "q35 machine model is not enabled" if !$q35;
3181             # win7 wants to have the pcie devices directly on the pcie bus
3182             # instead of in the root port
3183             if ($winversion == 7) {
3184                 $pciaddr = print_pcie_addr("${id}bus0");
3185             } else {
3186                 # add more root ports if needed, 4 are present by default
3187                 # by pve-q35 cfgs, rest added here on demand.
3188                 if ($i > 3) {
3189                     push @$devices, '-device', print_pcie_root_port($i);
3190                 }
3191                 $pciaddr = print_pcie_addr($id);
3192             }
3193         } else {
3194             $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3195         }
3196
3197         my $xvga = '';
3198         if ($d->{'x-vga'}) {
3199             $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3200             $kvm_off = 1;
3201             $vga->{type} = 'none' if !defined($conf->{vga});
3202             $gpu_passthrough = 1;
3203         }
3204
3205         my $pcidevices = $d->{pciid};
3206         my $multifunction = 1 if @$pcidevices > 1;
3207
3208         my $sysfspath;
3209         if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3210             my $pci_id = $pcidevices->[0]->{id};
3211             my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3212             $sysfspath = "/sys/bus/pci/devices/$pci_id/$uuid";
3213         } elsif ($d->{mdev}) {
3214             warn "ignoring mediated device '$id' with multifunction device\n";
3215         }
3216
3217         my $j=0;
3218         foreach my $pcidevice (@$pcidevices) {
3219             my $devicestr = "vfio-pci";
3220
3221             if ($sysfspath) {
3222                 $devicestr .= ",sysfsdev=$sysfspath";
3223             } else {
3224                 $devicestr .= ",host=$pcidevice->{id}";
3225             }
3226
3227             my $mf_addr = $multifunction ? ".$j" : '';
3228             $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3229
3230             if ($j == 0) {
3231                 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3232                 $devicestr .= "$xvga";
3233                 $devicestr .= ",multifunction=on" if $multifunction;
3234                 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3235             }
3236
3237             push @$devices, '-device', $devicestr;
3238             $j++;
3239         }
3240     }
3241
3242     # usb devices
3243     my $usb_dev_features = {};
3244     $usb_dev_features->{spice_usb3} = 1 if min_version($machine_version, 4, 0);
3245
3246     my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3247     push @$devices, @usbdevices if @usbdevices;
3248     # serial devices
3249     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3250         if (my $path = $conf->{"serial$i"}) {
3251             if ($path eq 'socket') {
3252                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3253                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3254                 # On aarch64, serial0 is the UART device. Qemu only allows
3255                 # connecting UART devices via the '-serial' command line, as
3256                 # the device has a fixed slot on the hardware...
3257                 if ($arch eq 'aarch64' && $i == 0) {
3258                     push @$devices, '-serial', "chardev:serial$i";
3259                 } else {
3260                     push @$devices, '-device', "isa-serial,chardev=serial$i";
3261                 }
3262             } else {
3263                 die "no such serial device\n" if ! -c $path;
3264                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3265                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3266             }
3267         }
3268     }
3269
3270     # parallel devices
3271     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3272         if (my $path = $conf->{"parallel$i"}) {
3273             die "no such parallel device\n" if ! -c $path;
3274             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3275             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3276             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3277         }
3278     }
3279
3280     if (min_version($machine_version, 4, 0) && (my $audio = conf_has_audio($conf))) {
3281         my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3282         my $audio_devs = audio_devs($audio, $audiopciaddr, $machine_version);
3283         push @$devices, @$audio_devs;
3284     }
3285
3286     my $sockets = 1;
3287     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3288     $sockets = $conf->{sockets} if  $conf->{sockets};
3289
3290     my $cores = $conf->{cores} || 1;
3291
3292     my $maxcpus = $sockets * $cores;
3293
3294     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3295
3296     my $allowed_vcpus = $cpuinfo->{cpus};
3297
3298     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3299         if ($allowed_vcpus < $maxcpus);
3300
3301     if($hotplug_features->{cpu} && min_version($machine_version, 2, 7)) {
3302
3303         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3304         for (my $i = 2; $i <= $vcpus; $i++)  {
3305             my $cpustr = print_cpu_device($conf,$i);
3306             push @$cmd, '-device', $cpustr;
3307         }
3308
3309     } else {
3310
3311         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3312     }
3313     push @$cmd, '-nodefaults';
3314
3315     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3316
3317     my $bootindex_hash = {};
3318     my $i = 1;
3319     foreach my $o (split(//, $bootorder)) {
3320         $bootindex_hash->{$o} = $i*100;
3321         $i++;
3322     }
3323
3324     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3325
3326     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3327
3328     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3329
3330     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3331         push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_version, $machine_type, undef, $qxlnum, $bridges);
3332         my $socket = PVE::QemuServer::Helpers::vnc_socket($vmid);
3333         push @$cmd,  '-vnc', "unix:$socket,password";
3334     } else {
3335         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3336         push @$cmd, '-nographic';
3337     }
3338
3339     # time drift fix
3340     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3341     my $useLocaltime = $conf->{localtime};
3342
3343     if ($winversion >= 5) { # windows
3344         $useLocaltime = 1 if !defined($conf->{localtime});
3345
3346         # use time drift fix when acpi is enabled
3347         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3348             $tdf = 1 if !defined($conf->{tdf});
3349         }
3350     }
3351
3352     if ($winversion >= 6) {
3353         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3354         push @$cmd, '-no-hpet';
3355     }
3356
3357     push @$rtcFlags, 'driftfix=slew' if $tdf;
3358
3359     if ($conf->{startdate} && $conf->{startdate} ne 'now') {
3360         push @$rtcFlags, "base=$conf->{startdate}";
3361     } elsif ($useLocaltime) {
3362         push @$rtcFlags, 'base=localtime';
3363     }
3364
3365     if ($forcecpu) {
3366         push @$cmd, '-cpu', $forcecpu;
3367     } else {
3368         push @$cmd, get_cpu_options($conf, $arch, $kvm, $kvm_off, $machine_version, $winversion, $gpu_passthrough);
3369     }
3370
3371     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3372
3373     push @$cmd, '-S' if $conf->{freeze};
3374
3375     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3376
3377     my $guest_agent = parse_guest_agent($conf);
3378
3379     if ($guest_agent->{enabled}) {
3380         my $qgasocket = PVE::QemuServer::Helpers::qmp_socket($vmid, 1);
3381         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3382
3383         if (!$guest_agent->{type} || $guest_agent->{type} eq 'virtio') {
3384             my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3385             push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3386             push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3387         } elsif ($guest_agent->{type} eq 'isa') {
3388             push @$devices, '-device', "isa-serial,chardev=qga0";
3389         }
3390     }
3391
3392     my $rng = parse_rng($conf->{rng0}) if $conf->{rng0};
3393     if ($rng && &$version_guard(4, 1, 2)) {
3394         check_rng_source($rng->{source});
3395
3396         my $max_bytes = $rng->{max_bytes} // $rng_fmt->{max_bytes}->{default};
3397         my $period = $rng->{period} // $rng_fmt->{period}->{default};
3398         my $limiter_str = "";
3399         if ($max_bytes) {
3400             $limiter_str = ",max-bytes=$max_bytes,period=$period";
3401         }
3402
3403         my $rng_addr = print_pci_addr("rng0", $bridges, $arch, $machine_type);
3404         push @$devices, '-object', "rng-random,filename=$rng->{source},id=rng0";
3405         push @$devices, '-device', "virtio-rng-pci,rng=rng0$limiter_str$rng_addr";
3406     }
3407
3408     my $spice_port;
3409
3410     if ($qxlnum) {
3411         if ($qxlnum > 1) {
3412             if ($winversion){
3413                 for (my $i = 1; $i < $qxlnum; $i++){
3414                     push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_version, $machine_type, $i, $qxlnum, $bridges);
3415                 }
3416             } else {
3417                 # assume other OS works like Linux
3418                 my ($ram, $vram) = ("134217728", "67108864");
3419                 if ($vga->{memory}) {
3420                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3421                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3422                 }
3423                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3424                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3425             }
3426         }
3427
3428         my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
3429
3430         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3431         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3432         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3433
3434         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3435         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3436         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3437
3438         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3439         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3440
3441         my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
3442         if ($spice_enhancement->{foldersharing}) {
3443             push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3444             push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3445         }
3446
3447         my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3448         $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
3449         push @$devices, '-spice', "$spice_opts";
3450     }
3451
3452     # enable balloon by default, unless explicitly disabled
3453     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3454         $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
3455         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3456     }
3457
3458     if ($conf->{watchdog}) {
3459         my $wdopts = parse_watchdog($conf->{watchdog});
3460         $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
3461         my $watchdog = $wdopts->{model} || 'i6300esb';
3462         push @$devices, '-device', "$watchdog$pciaddr";
3463         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3464     }
3465
3466     my $vollist = [];
3467     my $scsicontroller = {};
3468     my $ahcicontroller = {};
3469     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3470
3471     # Add iscsi initiator name if available
3472     if (my $initiator = get_initiator_name()) {
3473         push @$devices, '-iscsi', "initiator-name=$initiator";
3474     }
3475
3476     PVE::QemuConfig->foreach_volume($conf, sub {
3477         my ($ds, $drive) = @_;
3478
3479         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3480             push @$vollist, $drive->{file};
3481         }
3482
3483         # ignore efidisk here, already added in bios/fw handling code above
3484         return if $drive->{interface} eq 'efidisk';
3485
3486         $use_virtio = 1 if $ds =~ m/^virtio/;
3487
3488         if (drive_is_cdrom ($drive)) {
3489             if ($bootindex_hash->{d}) {
3490                 $drive->{bootindex} = $bootindex_hash->{d};
3491                 $bootindex_hash->{d} += 1;
3492             }
3493         } else {
3494             if ($bootindex_hash->{c}) {
3495                 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3496                 $bootindex_hash->{c} += 1;
3497             }
3498         }
3499
3500         if ($drive->{interface} eq 'virtio'){
3501            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3502         }
3503
3504         if ($drive->{interface} eq 'scsi') {
3505
3506             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3507
3508             die "scsi$drive->{index}: machine version 4.1~pve2 or higher is required to use more than 14 SCSI disks\n"
3509                 if $drive->{index} > 13 && !&$version_guard(4, 1, 2);
3510
3511             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
3512             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3513
3514             my $iothread = '';
3515             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3516                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3517                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3518             } elsif ($drive->{iothread}) {
3519                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3520             }
3521
3522             my $queues = '';
3523             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3524                 $queues = ",num_queues=$drive->{queues}";
3525             }
3526
3527             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3528             $scsicontroller->{$controller}=1;
3529         }
3530
3531         if ($drive->{interface} eq 'sata') {
3532             my $controller = int($drive->{index} / $PVE::QemuServer::Drive::MAX_SATA_DISKS);
3533             $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
3534             push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3535             $ahcicontroller->{$controller}=1;
3536         }
3537
3538         my $drive_cmd = print_drive_commandline_full($storecfg, $vmid, $drive);
3539         push @$devices, '-drive',$drive_cmd;
3540         push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
3541     });
3542
3543     for (my $i = 0; $i < $MAX_NETS; $i++) {
3544          next if !$conf->{"net$i"};
3545          my $d = parse_net($conf->{"net$i"});
3546          next if !$d;
3547
3548          $use_virtio = 1 if $d->{model} eq 'virtio';
3549
3550          if ($bootindex_hash->{n}) {
3551             $d->{bootindex} = $bootindex_hash->{n};
3552             $bootindex_hash->{n} += 1;
3553          }
3554
3555          my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
3556          push @$devices, '-netdev', $netdevfull;
3557
3558          my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
3559          push @$devices, '-device', $netdevicefull;
3560     }
3561
3562     if ($conf->{ivshmem}) {
3563         my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
3564
3565         my $bus;
3566         if ($q35) {
3567             $bus = print_pcie_addr("ivshmem");
3568         } else {
3569             $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
3570         }
3571
3572         my $ivshmem_name = $ivshmem->{name} // $vmid;
3573         my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
3574
3575         push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
3576         push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
3577     }
3578
3579     # pci.4 is nested in pci.1
3580     $bridges->{1} = 1 if $bridges->{4};
3581
3582     if (!$q35) {
3583         # add pci bridges
3584         if (min_version($machine_version, 2, 3)) {
3585            $bridges->{1} = 1;
3586            $bridges->{2} = 1;
3587         }
3588
3589         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3590
3591     }
3592
3593     for my $k (sort {$b cmp $a} keys %$bridges) {
3594         next if $q35 && $k < 4; # q35.cfg already includes bridges up to 3
3595         $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
3596         my $devstr = "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr";
3597         if ($q35) {
3598             # add after -readconfig pve-q35.cfg
3599             splice @$devices, 2, 0, '-device', $devstr;
3600         } else {
3601             unshift @$devices, '-device', $devstr if $k > 0;
3602         }
3603     }
3604
3605     if (!$kvm) {
3606         push @$machineFlags, 'accel=tcg';
3607     }
3608
3609     my $machine_type_min = $machine_type;
3610     if ($add_pve_version) {
3611         $machine_type_min =~ s/\+pve\d+$//;
3612         $machine_type_min .= "+pve$required_pve_version";
3613     }
3614     push @$machineFlags, "type=${machine_type_min}";
3615
3616     push @$cmd, @$devices;
3617     push @$cmd, '-rtc', join(',', @$rtcFlags) if scalar(@$rtcFlags);
3618     push @$cmd, '-machine', join(',', @$machineFlags) if scalar(@$machineFlags);
3619     push @$cmd, '-global', join(',', @$globalFlags) if scalar(@$globalFlags);
3620
3621     if (my $vmstate = $conf->{vmstate}) {
3622         my $statepath = PVE::Storage::path($storecfg, $vmstate);
3623         push @$vollist, $vmstate;
3624         push @$cmd, '-loadstate', $statepath;
3625         print "activating and using '$vmstate' as vmstate\n";
3626     }
3627
3628     # add custom args
3629     if ($conf->{args}) {
3630         my $aa = PVE::Tools::split_args($conf->{args});
3631         push @$cmd, @$aa;
3632     }
3633
3634     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3635 }
3636
3637 sub check_rng_source {
3638     my ($source) = @_;
3639
3640     # mostly relevant for /dev/hwrng, but doesn't hurt to check others too
3641     die "cannot create VirtIO RNG device: source file '$source' doesn't exist\n"
3642         if ! -e $source;
3643
3644     my $rng_current = '/sys/devices/virtual/misc/hw_random/rng_current';
3645     if ($source eq '/dev/hwrng' && file_read_firstline($rng_current) eq 'none') {
3646         # Needs to abort, otherwise QEMU crashes on first rng access.
3647         # Note that rng_current cannot be changed to 'none' manually, so
3648         # once the VM is past this point, it is no longer an issue.
3649         die "Cannot start VM with passed-through RNG device: '/dev/hwrng'"
3650             . " exists, but '$rng_current' is set to 'none'. Ensure that"
3651             . " a compatible hardware-RNG is attached to the host.\n";
3652     }
3653 }
3654
3655 sub spice_port {
3656     my ($vmid) = @_;
3657
3658     my $res = mon_cmd($vmid, 'query-spice');
3659
3660     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3661 }
3662
3663 sub vm_devices_list {
3664     my ($vmid) = @_;
3665
3666     my $res = mon_cmd($vmid, 'query-pci');
3667     my $devices_to_check = [];
3668     my $devices = {};
3669     foreach my $pcibus (@$res) {
3670         push @$devices_to_check, @{$pcibus->{devices}},
3671     }
3672
3673     while (@$devices_to_check) {
3674         my $to_check = [];
3675         for my $d (@$devices_to_check) {
3676             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3677             next if !$d->{'pci_bridge'};
3678
3679             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3680             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3681         }
3682         $devices_to_check = $to_check;
3683     }
3684
3685     my $resblock = mon_cmd($vmid, 'query-block');
3686     foreach my $block (@$resblock) {
3687         if($block->{device} =~ m/^drive-(\S+)/){
3688                 $devices->{$1} = 1;
3689         }
3690     }
3691
3692     my $resmice = mon_cmd($vmid, 'query-mice');
3693     foreach my $mice (@$resmice) {
3694         if ($mice->{name} eq 'QEMU HID Tablet') {
3695             $devices->{tablet} = 1;
3696             last;
3697         }
3698     }
3699
3700     # for usb devices there is no query-usb
3701     # but we can iterate over the entries in
3702     # qom-list path=/machine/peripheral
3703     my $resperipheral = mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3704     foreach my $per (@$resperipheral) {
3705         if ($per->{name} =~ m/^usb\d+$/) {
3706             $devices->{$per->{name}} = 1;
3707         }
3708     }
3709
3710     return $devices;
3711 }
3712
3713 sub vm_deviceplug {
3714     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
3715
3716     my $q35 = PVE::QemuServer::Machine::machine_type_is_q35($conf);
3717
3718     my $devices_list = vm_devices_list($vmid);
3719     return 1 if defined($devices_list->{$deviceid});
3720
3721     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
3722
3723     if ($deviceid eq 'tablet') {
3724
3725         qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
3726
3727     } elsif ($deviceid eq 'keyboard') {
3728
3729         qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
3730
3731     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3732
3733         die "usb hotplug currently not reliable\n";
3734         # since we can't reliably hot unplug all added usb devices
3735         # and usb passthrough disables live migration
3736         # we disable usb hotplugging for now
3737         qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3738
3739     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3740
3741         qemu_iothread_add($vmid, $deviceid, $device);
3742
3743         qemu_driveadd($storecfg, $vmid, $device);
3744         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
3745
3746         qemu_deviceadd($vmid, $devicefull);
3747         eval { qemu_deviceaddverify($vmid, $deviceid); };
3748         if (my $err = $@) {
3749             eval { qemu_drivedel($vmid, $deviceid); };
3750             warn $@ if $@;
3751             die $err;
3752         }
3753
3754     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3755
3756
3757         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3758         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3759         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3760
3761         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3762
3763         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3764             qemu_iothread_add($vmid, $deviceid, $device);
3765             $devicefull .= ",iothread=iothread-$deviceid";
3766         }
3767
3768         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3769             $devicefull .= ",num_queues=$device->{queues}";
3770         }
3771
3772         qemu_deviceadd($vmid, $devicefull);
3773         qemu_deviceaddverify($vmid, $deviceid);
3774
3775     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3776
3777         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
3778         qemu_driveadd($storecfg, $vmid, $device);
3779
3780         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
3781         eval { qemu_deviceadd($vmid, $devicefull); };
3782         if (my $err = $@) {
3783             eval { qemu_drivedel($vmid, $deviceid); };
3784             warn $@ if $@;
3785             die $err;
3786         }
3787
3788     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3789
3790         return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
3791
3792         my $machine_type = PVE::QemuServer::Machine::qemu_machine_pxe($vmid, $conf);
3793         my $use_old_bios_files = undef;
3794         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3795
3796         my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
3797         qemu_deviceadd($vmid, $netdevicefull);
3798         eval {
3799             qemu_deviceaddverify($vmid, $deviceid);
3800             qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
3801         };
3802         if (my $err = $@) {
3803             eval { qemu_netdevdel($vmid, $deviceid); };
3804             warn $@ if $@;
3805             die $err;
3806         }
3807
3808     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3809
3810         my $bridgeid = $2;
3811         my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
3812         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3813
3814         qemu_deviceadd($vmid, $devicefull);
3815         qemu_deviceaddverify($vmid, $deviceid);
3816
3817     } else {
3818         die "can't hotplug device '$deviceid'\n";
3819     }
3820
3821     return 1;
3822 }
3823
3824 # fixme: this should raise exceptions on error!
3825 sub vm_deviceunplug {
3826     my ($vmid, $conf, $deviceid) = @_;
3827
3828     my $devices_list = vm_devices_list($vmid);
3829     return 1 if !defined($devices_list->{$deviceid});
3830
3831     die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3832
3833     if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
3834
3835         qemu_devicedel($vmid, $deviceid);
3836
3837     } elsif ($deviceid =~ m/^usb\d+$/) {
3838
3839         die "usb hotplug currently not reliable\n";
3840         # when unplugging usb devices this way,
3841         # there may be remaining usb controllers/hubs
3842         # so we disable it for now
3843         qemu_devicedel($vmid, $deviceid);
3844         qemu_devicedelverify($vmid, $deviceid);
3845
3846     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3847
3848         qemu_devicedel($vmid, $deviceid);
3849         qemu_devicedelverify($vmid, $deviceid);
3850         qemu_drivedel($vmid, $deviceid);
3851         qemu_iothread_del($conf, $vmid, $deviceid);
3852
3853     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3854
3855         qemu_devicedel($vmid, $deviceid);
3856         qemu_devicedelverify($vmid, $deviceid);
3857         qemu_iothread_del($conf, $vmid, $deviceid);
3858
3859     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3860
3861         qemu_devicedel($vmid, $deviceid);
3862         qemu_drivedel($vmid, $deviceid);
3863         qemu_deletescsihw($conf, $vmid, $deviceid);
3864
3865     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3866
3867         qemu_devicedel($vmid, $deviceid);
3868         qemu_devicedelverify($vmid, $deviceid);
3869         qemu_netdevdel($vmid, $deviceid);
3870
3871     } else {
3872         die "can't unplug device '$deviceid'\n";
3873     }
3874
3875     return 1;
3876 }
3877
3878 sub qemu_deviceadd {
3879     my ($vmid, $devicefull) = @_;
3880
3881     $devicefull = "driver=".$devicefull;
3882     my %options =  split(/[=,]/, $devicefull);
3883
3884     mon_cmd($vmid, "device_add" , %options);
3885 }
3886
3887 sub qemu_devicedel {
3888     my ($vmid, $deviceid) = @_;
3889
3890     my $ret = mon_cmd($vmid, "device_del", id => $deviceid);
3891 }
3892
3893 sub qemu_iothread_add {
3894     my($vmid, $deviceid, $device) = @_;
3895
3896     if ($device->{iothread}) {
3897         my $iothreads = vm_iothreads_list($vmid);
3898         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3899     }
3900 }
3901
3902 sub qemu_iothread_del {
3903     my($conf, $vmid, $deviceid) = @_;
3904
3905     my $confid = $deviceid;
3906     if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
3907         $confid = 'scsi' . $1;
3908     }
3909     my $device = parse_drive($confid, $conf->{$confid});
3910     if ($device->{iothread}) {
3911         my $iothreads = vm_iothreads_list($vmid);
3912         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3913     }
3914 }
3915
3916 sub qemu_objectadd {
3917     my($vmid, $objectid, $qomtype) = @_;
3918
3919     mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3920
3921     return 1;
3922 }
3923
3924 sub qemu_objectdel {
3925     my($vmid, $objectid) = @_;
3926
3927     mon_cmd($vmid, "object-del", id => $objectid);
3928
3929     return 1;
3930 }
3931
3932 sub qemu_driveadd {
3933     my ($storecfg, $vmid, $device) = @_;
3934
3935     my $drive = print_drive_commandline_full($storecfg, $vmid, $device);
3936     $drive =~ s/\\/\\\\/g;
3937     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_add auto \"$drive\"");
3938
3939     # If the command succeeds qemu prints: "OK"
3940     return 1 if $ret =~ m/OK/s;
3941
3942     die "adding drive failed: $ret\n";
3943 }
3944
3945 sub qemu_drivedel {
3946     my($vmid, $deviceid) = @_;
3947
3948     my $ret = PVE::QemuServer::Monitor::hmp_cmd($vmid, "drive_del drive-$deviceid");
3949     $ret =~ s/^\s+//;
3950
3951     return 1 if $ret eq "";
3952
3953     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3954     return 1 if $ret =~ m/Device \'.*?\' not found/s;
3955
3956     die "deleting drive $deviceid failed : $ret\n";
3957 }
3958
3959 sub qemu_deviceaddverify {
3960     my ($vmid, $deviceid) = @_;
3961
3962     for (my $i = 0; $i <= 5; $i++) {
3963          my $devices_list = vm_devices_list($vmid);
3964          return 1 if defined($devices_list->{$deviceid});
3965          sleep 1;
3966     }
3967
3968     die "error on hotplug device '$deviceid'\n";
3969 }
3970
3971
3972 sub qemu_devicedelverify {
3973     my ($vmid, $deviceid) = @_;
3974
3975     # need to verify that the device is correctly removed as device_del
3976     # is async and empty return is not reliable
3977
3978     for (my $i = 0; $i <= 5; $i++) {
3979          my $devices_list = vm_devices_list($vmid);
3980          return 1 if !defined($devices_list->{$deviceid});
3981          sleep 1;
3982     }
3983
3984     die "error on hot-unplugging device '$deviceid'\n";
3985 }
3986
3987 sub qemu_findorcreatescsihw {
3988     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
3989
3990     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3991
3992     my $scsihwid="$controller_prefix$controller";
3993     my $devices_list = vm_devices_list($vmid);
3994
3995     if(!defined($devices_list->{$scsihwid})) {
3996         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
3997     }
3998
3999     return 1;
4000 }
4001
4002 sub qemu_deletescsihw {
4003     my ($conf, $vmid, $opt) = @_;
4004
4005     my $device = parse_drive($opt, $conf->{$opt});
4006
4007     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4008         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4009         return 1;
4010     }
4011
4012     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4013
4014     my $devices_list = vm_devices_list($vmid);
4015     foreach my $opt (keys %{$devices_list}) {
4016         if (is_valid_drivename($opt)) {
4017             my $drive = parse_drive($opt, $conf->{$opt});
4018             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4019                 return 1;
4020             }
4021         }
4022     }
4023
4024     my $scsihwid="scsihw$controller";
4025
4026     vm_deviceunplug($vmid, $conf, $scsihwid);
4027
4028     return 1;
4029 }
4030
4031 sub qemu_add_pci_bridge {
4032     my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4033
4034     my $bridges = {};
4035
4036     my $bridgeid;
4037
4038     print_pci_addr($device, $bridges, $arch, $machine_type);
4039
4040     while (my ($k, $v) = each %$bridges) {
4041         $bridgeid = $k;
4042     }
4043     return 1 if !defined($bridgeid) || $bridgeid < 1;
4044
4045     my $bridge = "pci.$bridgeid";
4046     my $devices_list = vm_devices_list($vmid);
4047
4048     if (!defined($devices_list->{$bridge})) {
4049         vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4050     }
4051
4052     return 1;
4053 }
4054
4055 sub qemu_set_link_status {
4056     my ($vmid, $device, $up) = @_;
4057
4058     mon_cmd($vmid, "set_link", name => $device,
4059                up => $up ? JSON::true : JSON::false);
4060 }
4061
4062 sub qemu_netdevadd {
4063     my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4064
4065     my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4066     my %options =  split(/[=,]/, $netdev);
4067
4068     mon_cmd($vmid, "netdev_add",  %options);
4069     return 1;
4070 }
4071
4072 sub qemu_netdevdel {
4073     my ($vmid, $deviceid) = @_;
4074
4075     mon_cmd($vmid, "netdev_del", id => $deviceid);
4076 }
4077
4078 sub qemu_usb_hotplug {
4079     my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4080
4081     return if !$device;
4082
4083     # remove the old one first
4084     vm_deviceunplug($vmid, $conf, $deviceid);
4085
4086     # check if xhci controller is necessary and available
4087     if ($device->{usb3}) {
4088
4089         my $devicelist = vm_devices_list($vmid);
4090
4091         if (!$devicelist->{xhci}) {
4092             my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4093             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4094         }
4095     }
4096     my $d = parse_usb_device($device->{host});
4097     $d->{usb3} = $device->{usb3};
4098
4099     # add the new one
4100     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4101 }
4102
4103 sub qemu_cpu_hotplug {
4104     my ($vmid, $conf, $vcpus) = @_;
4105
4106     my $machine_type = PVE::QemuServer::Machine::get_current_qemu_machine($vmid);
4107
4108     my $sockets = 1;
4109     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4110     $sockets = $conf->{sockets} if  $conf->{sockets};
4111     my $cores = $conf->{cores} || 1;
4112     my $maxcpus = $sockets * $cores;
4113
4114     $vcpus = $maxcpus if !$vcpus;
4115
4116     die "you can't add more vcpus than maxcpus\n"
4117         if $vcpus > $maxcpus;
4118
4119     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4120
4121     if ($vcpus < $currentvcpus) {
4122
4123         if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4124
4125             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4126                 qemu_devicedel($vmid, "cpu$i");
4127                 my $retry = 0;
4128                 my $currentrunningvcpus = undef;
4129                 while (1) {
4130                     $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4131                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4132                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4133                     $retry++;
4134                     sleep 1;
4135                 }
4136                 #update conf after each succesfull cpu unplug
4137                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4138                 PVE::QemuConfig->write_config($vmid, $conf);
4139             }
4140         } else {
4141             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4142         }
4143
4144         return;
4145     }
4146
4147     my $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4148     die "vcpus in running vm does not match its configuration\n"
4149         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4150
4151     if (PVE::QemuServer::Machine::machine_version($machine_type, 2, 7)) {
4152
4153         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4154             my $cpustr = print_cpu_device($conf, $i);
4155             qemu_deviceadd($vmid, $cpustr);
4156
4157             my $retry = 0;
4158             my $currentrunningvcpus = undef;
4159             while (1) {
4160                 $currentrunningvcpus = mon_cmd($vmid, "query-cpus-fast");
4161                 last if scalar(@{$currentrunningvcpus}) == $i;
4162                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4163                 sleep 1;
4164                 $retry++;
4165             }
4166             #update conf after each succesfull cpu hotplug
4167             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4168             PVE::QemuConfig->write_config($vmid, $conf);
4169         }
4170     } else {
4171
4172         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4173             mon_cmd($vmid, "cpu-add", id => int($i));
4174         }
4175     }
4176 }
4177
4178 sub qemu_block_set_io_throttle {
4179     my ($vmid, $deviceid,
4180         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4181         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4182         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4183         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4184
4185     return if !check_running($vmid) ;
4186
4187     mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4188         bps => int($bps),
4189         bps_rd => int($bps_rd),
4190         bps_wr => int($bps_wr),
4191         iops => int($iops),
4192         iops_rd => int($iops_rd),
4193         iops_wr => int($iops_wr),
4194         bps_max => int($bps_max),
4195         bps_rd_max => int($bps_rd_max),
4196         bps_wr_max => int($bps_wr_max),
4197         iops_max => int($iops_max),
4198         iops_rd_max => int($iops_rd_max),
4199         iops_wr_max => int($iops_wr_max),
4200         bps_max_length => int($bps_max_length),
4201         bps_rd_max_length => int($bps_rd_max_length),
4202         bps_wr_max_length => int($bps_wr_max_length),
4203         iops_max_length => int($iops_max_length),
4204         iops_rd_max_length => int($iops_rd_max_length),
4205         iops_wr_max_length => int($iops_wr_max_length),
4206     );
4207
4208 }
4209
4210 # old code, only used to shutdown old VM after update
4211 sub __read_avail {
4212     my ($fh, $timeout) = @_;
4213
4214     my $sel = new IO::Select;
4215     $sel->add($fh);
4216
4217     my $res = '';
4218     my $buf;
4219
4220     my @ready;
4221     while (scalar (@ready = $sel->can_read($timeout))) {
4222         my $count;
4223         if ($count = $fh->sysread($buf, 8192)) {
4224             if ($buf =~ /^(.*)\(qemu\) $/s) {
4225                 $res .= $1;
4226                 last;
4227             } else {
4228                 $res .= $buf;
4229             }
4230         } else {
4231             if (!defined($count)) {
4232                 die "$!\n";
4233             }
4234             last;
4235         }
4236     }
4237
4238     die "monitor read timeout\n" if !scalar(@ready);
4239
4240     return $res;
4241 }
4242
4243 sub qemu_block_resize {
4244     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4245
4246     my $running = check_running($vmid);
4247
4248     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4249
4250     return if !$running;
4251
4252     my $padding = (1024 - $size % 1024) % 1024;
4253     $size = $size + $padding;
4254
4255     mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4256
4257 }
4258
4259 sub qemu_volume_snapshot {
4260     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4261
4262     my $running = check_running($vmid);
4263
4264     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4265         mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4266     } else {
4267         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4268     }
4269 }
4270
4271 sub qemu_volume_snapshot_delete {
4272     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4273
4274     my $running = check_running($vmid);
4275
4276     if($running) {
4277
4278         $running = undef;
4279         my $conf = PVE::QemuConfig->load_config($vmid);
4280         PVE::QemuConfig->foreach_volume($conf, sub {
4281             my ($ds, $drive) = @_;
4282             $running = 1 if $drive->{file} eq $volid;
4283         });
4284     }
4285
4286     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4287         mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4288     } else {
4289         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4290     }
4291 }
4292
4293 sub set_migration_caps {
4294     my ($vmid) = @_;
4295
4296     my $cap_ref = [];
4297
4298     my $enabled_cap = {
4299         "auto-converge" => 1,
4300         "xbzrle" => 1,
4301         "x-rdma-pin-all" => 0,
4302         "zero-blocks" => 0,
4303         "compress" => 0
4304     };
4305
4306     my $supported_capabilities = mon_cmd($vmid, "query-migrate-capabilities");
4307
4308     for my $supported_capability (@$supported_capabilities) {
4309         push @$cap_ref, {
4310             capability => $supported_capability->{capability},
4311             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4312         };
4313     }
4314
4315     mon_cmd($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4316 }
4317
4318 sub foreach_volid {
4319     my ($conf, $func, @param) = @_;
4320
4321     my $volhash = {};
4322
4323     my $test_volid = sub {
4324         my ($key, $drive, $snapname) = @_;
4325
4326         my $volid = $drive->{file};
4327         return if !$volid;
4328
4329         $volhash->{$volid}->{cdrom} //= 1;
4330         $volhash->{$volid}->{cdrom} = 0 if !drive_is_cdrom($drive);
4331
4332         my $replicate = $drive->{replicate} // 1;
4333         $volhash->{$volid}->{replicate} //= 0;
4334         $volhash->{$volid}->{replicate} = 1 if $replicate;
4335
4336         $volhash->{$volid}->{shared} //= 0;
4337         $volhash->{$volid}->{shared} = 1 if $drive->{shared};
4338
4339         $volhash->{$volid}->{referenced_in_config} //= 0;
4340         $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
4341
4342         $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
4343             if defined($snapname);
4344
4345         my $size = $drive->{size};
4346         $volhash->{$volid}->{size} //= $size if $size;
4347
4348         $volhash->{$volid}->{is_vmstate} //= 0;
4349         $volhash->{$volid}->{is_vmstate} = 1 if $key eq 'vmstate';
4350
4351         $volhash->{$volid}->{is_unused} //= 0;
4352         $volhash->{$volid}->{is_unused} = 1 if $key =~ /^unused\d+$/;
4353     };
4354
4355     my $include_opts = {
4356         extra_keys => ['vmstate'],
4357         include_unused => 1,
4358     };
4359
4360     PVE::QemuConfig->foreach_volume_full($conf, $include_opts, $test_volid);
4361     foreach my $snapname (keys %{$conf->{snapshots}}) {
4362         my $snap = $conf->{snapshots}->{$snapname};
4363         PVE::QemuConfig->foreach_volume_full($snap, $include_opts, $test_volid, $snapname);
4364     }
4365
4366     foreach my $volid (keys %$volhash) {
4367         &$func($volid, $volhash->{$volid}, @param);
4368     }
4369 }
4370
4371 my $fast_plug_option = {
4372     'lock' => 1,
4373     'name' => 1,
4374     'onboot' => 1,
4375     'shares' => 1,
4376     'startup' => 1,
4377     'description' => 1,
4378     'protection' => 1,
4379     'vmstatestorage' => 1,
4380     'hookscript' => 1,
4381     'tags' => 1,
4382 };
4383
4384 # hotplug changes in [PENDING]
4385 # $selection hash can be used to only apply specified options, for
4386 # example: { cores => 1 } (only apply changed 'cores')
4387 # $errors ref is used to return error messages
4388 sub vmconfig_hotplug_pending {
4389     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4390
4391     my $defaults = load_defaults();
4392     my $arch = get_vm_arch($conf);
4393     my $machine_type = get_vm_machine($conf, undef, $arch);
4394
4395     # commit values which do not have any impact on running VM first
4396     # Note: those option cannot raise errors, we we do not care about
4397     # $selection and always apply them.
4398
4399     my $add_error = sub {
4400         my ($opt, $msg) = @_;
4401         $errors->{$opt} = "hotplug problem - $msg";
4402     };
4403
4404     my $changes = 0;
4405     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4406         if ($fast_plug_option->{$opt}) {
4407             $conf->{$opt} = $conf->{pending}->{$opt};
4408             delete $conf->{pending}->{$opt};
4409             $changes = 1;
4410         }
4411     }
4412
4413     if ($changes) {
4414         PVE::QemuConfig->write_config($vmid, $conf);
4415     }
4416
4417     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4418
4419     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4420     foreach my $opt (sort keys %$pending_delete_hash) {
4421         next if $selection && !$selection->{$opt};
4422         my $force = $pending_delete_hash->{$opt}->{force};
4423         eval {
4424             if ($opt eq 'hotplug') {
4425                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4426             } elsif ($opt eq 'tablet') {
4427                 die "skip\n" if !$hotplug_features->{usb};
4428                 if ($defaults->{tablet}) {
4429                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4430                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4431                         if $arch eq 'aarch64';
4432                 } else {
4433                     vm_deviceunplug($vmid, $conf, 'tablet');
4434                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4435                 }
4436             } elsif ($opt =~ m/^usb\d+/) {
4437                 die "skip\n";
4438                 # since we cannot reliably hot unplug usb devices
4439                 # we are disabling it
4440                 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4441                 vm_deviceunplug($vmid, $conf, $opt);
4442             } elsif ($opt eq 'vcpus') {
4443                 die "skip\n" if !$hotplug_features->{cpu};
4444                 qemu_cpu_hotplug($vmid, $conf, undef);
4445             } elsif ($opt eq 'balloon') {
4446                 # enable balloon device is not hotpluggable
4447                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4448                 # here we reset the ballooning value to memory
4449                 my $balloon = $conf->{memory} || $defaults->{memory};
4450                 mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4451             } elsif ($fast_plug_option->{$opt}) {
4452                 # do nothing
4453             } elsif ($opt =~ m/^net(\d+)$/) {
4454                 die "skip\n" if !$hotplug_features->{network};
4455                 vm_deviceunplug($vmid, $conf, $opt);
4456             } elsif (is_valid_drivename($opt)) {
4457                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4458                 vm_deviceunplug($vmid, $conf, $opt);
4459                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4460             } elsif ($opt =~ m/^memory$/) {
4461                 die "skip\n" if !$hotplug_features->{memory};
4462                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4463             } elsif ($opt eq 'cpuunits') {
4464                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4465             } elsif ($opt eq 'cpulimit') {
4466                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4467             } else {
4468                 die "skip\n";
4469             }
4470         };
4471         if (my $err = $@) {
4472             &$add_error($opt, $err) if $err ne "skip\n";
4473         } else {
4474             delete $conf->{$opt};
4475             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4476         }
4477     }
4478
4479     my ($apply_pending_cloudinit, $apply_pending_cloudinit_done);
4480     $apply_pending_cloudinit = sub {
4481         return if $apply_pending_cloudinit_done; # once is enough
4482         $apply_pending_cloudinit_done = 1; # once is enough
4483
4484         my ($key, $value) = @_;
4485
4486         my @cloudinit_opts = keys %$confdesc_cloudinit;
4487         foreach my $opt (keys %{$conf->{pending}}) {
4488             next if !grep { $_ eq $opt } @cloudinit_opts;
4489             $conf->{$opt} = delete $conf->{pending}->{$opt};
4490         }
4491
4492         my $new_conf = { %$conf };
4493         $new_conf->{$key} = $value;
4494         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4495     };
4496
4497     foreach my $opt (keys %{$conf->{pending}}) {
4498         next if $selection && !$selection->{$opt};
4499         my $value = $conf->{pending}->{$opt};
4500         eval {
4501             if ($opt eq 'hotplug') {
4502                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4503             } elsif ($opt eq 'tablet') {
4504                 die "skip\n" if !$hotplug_features->{usb};
4505                 if ($value == 1) {
4506                     vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4507                     vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4508                         if $arch eq 'aarch64';
4509                 } elsif ($value == 0) {
4510                     vm_deviceunplug($vmid, $conf, 'tablet');
4511                     vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4512                 }
4513             } elsif ($opt =~ m/^usb\d+$/) {
4514                 die "skip\n";
4515                 # since we cannot reliably hot unplug usb devices
4516                 # we are disabling it
4517                 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4518                 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4519                 die "skip\n" if !$d;
4520                 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
4521             } elsif ($opt eq 'vcpus') {
4522                 die "skip\n" if !$hotplug_features->{cpu};
4523                 qemu_cpu_hotplug($vmid, $conf, $value);
4524             } elsif ($opt eq 'balloon') {
4525                 # enable/disable balloning device is not hotpluggable
4526                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4527                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4528                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4529
4530                 # allow manual ballooning if shares is set to zero
4531                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4532                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4533                     mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4534                 }
4535             } elsif ($opt =~ m/^net(\d+)$/) {
4536                 # some changes can be done without hotplug
4537                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4538                                     $vmid, $opt, $value, $arch, $machine_type);
4539             } elsif (is_valid_drivename($opt)) {
4540                 die "skip\n" if $opt eq 'efidisk0';
4541                 # some changes can be done without hotplug
4542                 my $drive = parse_drive($opt, $value);
4543                 if (drive_is_cloudinit($drive)) {
4544                     &$apply_pending_cloudinit($opt, $value);
4545                 }
4546                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4547                                      $vmid, $opt, $value, $arch, $machine_type);
4548             } elsif ($opt =~ m/^memory$/) { #dimms
4549                 die "skip\n" if !$hotplug_features->{memory};
4550                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4551             } elsif ($opt eq 'cpuunits') {
4552                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4553             } elsif ($opt eq 'cpulimit') {
4554                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4555                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4556             } else {
4557                 die "skip\n";  # skip non-hot-pluggable options
4558             }
4559         };
4560         if (my $err = $@) {
4561             &$add_error($opt, $err) if $err ne "skip\n";
4562         } else {
4563             $conf->{$opt} = $value;
4564             delete $conf->{pending}->{$opt};
4565         }
4566     }
4567
4568     PVE::QemuConfig->write_config($vmid, $conf);
4569 }
4570
4571 sub try_deallocate_drive {
4572     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4573
4574     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4575         my $volid = $drive->{file};
4576         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4577             my $sid = PVE::Storage::parse_volume_id($volid);
4578             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4579
4580             # check if the disk is really unused
4581             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4582                 if PVE::QemuServer::Drive::is_volume_in_use($storecfg, $conf, $key, $volid);
4583             PVE::Storage::vdisk_free($storecfg, $volid);
4584             return 1;
4585         } else {
4586             # If vm is not owner of this disk remove from config
4587             return 1;
4588         }
4589     }
4590
4591     return undef;
4592 }
4593
4594 sub vmconfig_delete_or_detach_drive {
4595     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4596
4597     my $drive = parse_drive($opt, $conf->{$opt});
4598
4599     my $rpcenv = PVE::RPCEnvironment::get();
4600     my $authuser = $rpcenv->get_user();
4601
4602     if ($force) {
4603         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4604         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4605     } else {
4606         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4607     }
4608 }
4609
4610
4611
4612 sub vmconfig_apply_pending {
4613     my ($vmid, $conf, $storecfg, $errors) = @_;
4614
4615     my $add_apply_error = sub {
4616         my ($opt, $msg) = @_;
4617         my $err_msg = "unable to apply pending change $opt : $msg";
4618         $errors->{$opt} = $err_msg;
4619         warn $err_msg;
4620     };
4621
4622     # cold plug
4623
4624     my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4625     foreach my $opt (sort keys %$pending_delete_hash) {
4626         my $force = $pending_delete_hash->{$opt}->{force};
4627         eval {
4628             if ($opt =~ m/^unused/) {
4629                 die "internal error";
4630             } elsif (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4631                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4632             }
4633         };
4634         if (my $err = $@) {
4635             $add_apply_error->($opt, $err);
4636         } else {
4637             PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4638             delete $conf->{$opt};
4639         }
4640     }
4641
4642     PVE::QemuConfig->cleanup_pending($conf);
4643
4644     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4645         next if $opt eq 'delete'; # just to be sure
4646         eval {
4647             if (defined($conf->{$opt}) && is_valid_drivename($opt)) {
4648                 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4649             }
4650         };
4651         if (my $err = $@) {
4652             $add_apply_error->($opt, $err);
4653         } else {
4654             $conf->{$opt} = delete $conf->{pending}->{$opt};
4655         }
4656     }
4657
4658     # write all changes at once to avoid unnecessary i/o
4659     PVE::QemuConfig->write_config($vmid, $conf);
4660 }
4661
4662 sub vmconfig_update_net {
4663     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4664
4665     my $newnet = parse_net($value);
4666
4667     if ($conf->{$opt}) {
4668         my $oldnet = parse_net($conf->{$opt});
4669
4670         if (safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4671             safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4672             safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4673             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4674
4675             # for non online change, we try to hot-unplug
4676             die "skip\n" if !$hotplug;
4677             vm_deviceunplug($vmid, $conf, $opt);
4678         } else {
4679
4680             die "internal error" if $opt !~ m/net(\d+)/;
4681             my $iface = "tap${vmid}i$1";
4682
4683             if (safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4684                 safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4685                 safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4686                 safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4687                 PVE::Network::tap_unplug($iface);
4688
4689                 if ($have_sdn) {
4690                     PVE::Network::SDN::Zones::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4691                 } else {
4692                     PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4693                 }
4694             } elsif (safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4695                 # Rate can be applied on its own but any change above needs to
4696                 # include the rate in tap_plug since OVS resets everything.
4697                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4698             }
4699
4700             if (safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4701                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4702             }
4703
4704             return 1;
4705         }
4706     }
4707
4708     if ($hotplug) {
4709         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
4710     } else {
4711         die "skip\n";
4712     }
4713 }
4714
4715 sub vmconfig_update_disk {
4716     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
4717
4718     my $drive = parse_drive($opt, $value);
4719
4720     if ($conf->{$opt}) {
4721
4722         if (my $old_drive = parse_drive($opt, $conf->{$opt}))  {
4723
4724             my $media = $drive->{media} || 'disk';
4725             my $oldmedia = $old_drive->{media} || 'disk';
4726             die "unable to change media type\n" if $media ne $oldmedia;
4727
4728             if (!drive_is_cdrom($old_drive)) {
4729
4730                 if ($drive->{file} ne $old_drive->{file}) {
4731
4732                     die "skip\n" if !$hotplug;
4733
4734                     # unplug and register as unused
4735                     vm_deviceunplug($vmid, $conf, $opt);
4736                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4737
4738                 } else {
4739                     # update existing disk
4740
4741                     # skip non hotpluggable value
4742                     if (safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4743                         safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4744                         safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4745                         safe_string_ne($drive->{cache}, $old_drive->{cache}) ||
4746                         safe_string_ne($drive->{ssd}, $old_drive->{ssd})) {
4747                         die "skip\n";
4748                     }
4749
4750                     # apply throttle
4751                     if (safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4752                         safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4753                         safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4754                         safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4755                         safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4756                         safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4757                         safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4758                         safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4759                         safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4760                         safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4761                         safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4762                         safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4763                         safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4764                         safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4765                         safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4766                         safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4767                         safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4768                         safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4769
4770                         qemu_block_set_io_throttle($vmid,"drive-$opt",
4771                                                    ($drive->{mbps} || 0)*1024*1024,
4772                                                    ($drive->{mbps_rd} || 0)*1024*1024,
4773                                                    ($drive->{mbps_wr} || 0)*1024*1024,
4774                                                    $drive->{iops} || 0,
4775                                                    $drive->{iops_rd} || 0,
4776                                                    $drive->{iops_wr} || 0,
4777                                                    ($drive->{mbps_max} || 0)*1024*1024,
4778                                                    ($drive->{mbps_rd_max} || 0)*1024*1024,
4779                                                    ($drive->{mbps_wr_max} || 0)*1024*1024,
4780                                                    $drive->{iops_max} || 0,
4781                                                    $drive->{iops_rd_max} || 0,
4782                                                    $drive->{iops_wr_max} || 0,
4783                                                    $drive->{bps_max_length} || 1,
4784                                                    $drive->{bps_rd_max_length} || 1,
4785                                                    $drive->{bps_wr_max_length} || 1,
4786                                                    $drive->{iops_max_length} || 1,
4787                                                    $drive->{iops_rd_max_length} || 1,
4788                                                    $drive->{iops_wr_max_length} || 1);
4789
4790                     }
4791
4792                     return 1;
4793                 }
4794
4795             } else { # cdrom
4796
4797                 if ($drive->{file} eq 'none') {
4798                     mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4799                     if (drive_is_cloudinit($old_drive)) {
4800                         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4801                     }
4802                 } else {
4803                     my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4804
4805                     # force eject if locked
4806                     mon_cmd($vmid, "eject", force => JSON::true, id => "$opt");
4807
4808                     if ($path) {
4809                         mon_cmd($vmid, "blockdev-change-medium",
4810                             id => "$opt", filename => "$path");
4811                     }
4812                 }
4813
4814                 return 1;
4815             }
4816         }
4817     }
4818
4819     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4820     # hotplug new disks
4821     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4822     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
4823 }
4824
4825 # called in locked context by incoming migration
4826 sub vm_migrate_get_nbd_disks {
4827     my ($storecfg, $conf, $replicated_volumes) = @_;
4828
4829     my $local_volumes = {};
4830     PVE::QemuConfig->foreach_volume($conf, sub {
4831         my ($ds, $drive) = @_;
4832
4833         return if drive_is_cdrom($drive);
4834
4835         my $volid = $drive->{file};
4836
4837         return if !$volid;
4838
4839         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
4840
4841         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4842         return if $scfg->{shared};
4843
4844         # replicated disks re-use existing state via bitmap
4845         my $use_existing = $replicated_volumes->{$volid} ? 1 : 0;
4846         $local_volumes->{$ds} = [$volid, $storeid, $volname, $drive, $use_existing];
4847     });
4848     return $local_volumes;
4849 }
4850
4851 # called in locked context by incoming migration
4852 sub vm_migrate_alloc_nbd_disks {
4853     my ($storecfg, $vmid, $source_volumes, $storagemap) = @_;
4854
4855     my $format = undef;
4856
4857     my $nbd = {};
4858     foreach my $opt (sort keys %$source_volumes) {
4859         my ($volid, $storeid, $volname, $drive, $use_existing) = @{$source_volumes->{$opt}};
4860
4861         if ($use_existing) {
4862             $nbd->{$opt}->{drivestr} = print_drive($drive);
4863             $nbd->{$opt}->{volid} = $volid;
4864             $nbd->{$opt}->{replicated} = 1;
4865             next;
4866         }
4867
4868         # If a remote storage is specified and the format of the original
4869         # volume is not available there, fall back to the default format.
4870         # Otherwise use the same format as the original.
4871         if (!$storagemap->{identity}) {
4872             $storeid = map_storage($storagemap, $storeid);
4873             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
4874             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4875             my $fileFormat = qemu_img_format($scfg, $volname);
4876             $format = (grep {$fileFormat eq $_} @{$validFormats}) ? $fileFormat : $defFormat;
4877         } else {
4878             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4879             $format = qemu_img_format($scfg, $volname);
4880         }
4881
4882         my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
4883         my $newdrive = $drive;
4884         $newdrive->{format} = $format;
4885         $newdrive->{file} = $newvolid;
4886         my $drivestr = print_drive($newdrive);
4887         $nbd->{$opt}->{drivestr} = $drivestr;
4888         $nbd->{$opt}->{volid} = $newvolid;
4889     }
4890
4891     return $nbd;
4892 }
4893
4894 # see vm_start_nolock for parameters, additionally:
4895 # migrate_opts:
4896 #   storagemap = parsed storage map for allocating NBD disks
4897 sub vm_start {
4898     my ($storecfg, $vmid, $params, $migrate_opts) = @_;
4899
4900     return PVE::QemuConfig->lock_config($vmid, sub {
4901         my $conf = PVE::QemuConfig->load_config($vmid, $migrate_opts->{migratedfrom});
4902
4903         die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4904
4905         $params->{resume} = PVE::QemuConfig->has_lock($conf, 'suspended');
4906
4907         PVE::QemuConfig->check_lock($conf)
4908             if !($params->{skiplock} || $params->{resume});
4909
4910         die "VM $vmid already running\n" if check_running($vmid, undef, $migrate_opts->{migratedfrom});
4911
4912         if (my $storagemap = $migrate_opts->{storagemap}) {
4913             my $replicated = $migrate_opts->{replicated_volumes};
4914             my $disks = vm_migrate_get_nbd_disks($storecfg, $conf, $replicated);
4915             $migrate_opts->{nbd} = vm_migrate_alloc_nbd_disks($storecfg, $vmid, $disks, $storagemap);
4916
4917             foreach my $opt (keys %{$migrate_opts->{nbd}}) {
4918                 $conf->{$opt} = $migrate_opts->{nbd}->{$opt}->{drivestr};
4919             }
4920         }
4921
4922         return vm_start_nolock($storecfg, $vmid, $conf, $params, $migrate_opts);
4923     });
4924 }
4925
4926
4927 # params:
4928 #   statefile => 'tcp', 'unix' for migration or path/volid for RAM state
4929 #   skiplock => 0/1, skip checking for config lock
4930 #   forcemachine => to force Qemu machine (rollback/migration)
4931 #   forcecpu => a QEMU '-cpu' argument string to override get_cpu_options
4932 #   timeout => in seconds
4933 #   paused => start VM in paused state (backup)
4934 #   resume => resume from hibernation
4935 # migrate_opts:
4936 #   nbd => volumes for NBD exports (vm_migrate_alloc_nbd_disks)
4937 #   migratedfrom => source node
4938 #   spice_ticket => used for spice migration, passed via tunnel/stdin
4939 #   network => CIDR of migration network
4940 #   type => secure/insecure - tunnel over encrypted connection or plain-text
4941 #   nbd_proto_version => int, 0 for TCP, 1 for UNIX
4942 #   replicated_volumes = which volids should be re-used with bitmaps for nbd migration
4943 sub vm_start_nolock {
4944     my ($storecfg, $vmid, $conf, $params, $migrate_opts) = @_;
4945
4946     my $statefile = $params->{statefile};
4947     my $resume = $params->{resume};
4948
4949     my $migratedfrom = $migrate_opts->{migratedfrom};
4950     my $migration_type = $migrate_opts->{type};
4951
4952     my $res = {};
4953
4954     # clean up leftover reboot request files
4955     eval { clear_reboot_request($vmid); };
4956     warn $@ if $@;
4957
4958     if (!$statefile && scalar(keys %{$conf->{pending}})) {
4959         vmconfig_apply_pending($vmid, $conf, $storecfg);
4960         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4961     }
4962
4963     PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4964
4965     my $defaults = load_defaults();
4966
4967     # set environment variable useful inside network script
4968     $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4969
4970     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
4971
4972     my $forcemachine = $params->{forcemachine};
4973     my $forcecpu = $params->{forcecpu};
4974     if ($resume) {
4975         # enforce machine and CPU type on suspended vm to ensure HW compatibility
4976         $forcemachine = $conf->{runningmachine};
4977         $forcecpu = $conf->{runningcpu};
4978         print "Resuming suspended VM\n";
4979     }
4980
4981     my ($cmd, $vollist, $spice_port) =
4982         config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine, $forcecpu);
4983
4984     my $migration_ip;
4985     my $get_migration_ip = sub {
4986         my ($nodename) = @_;
4987
4988         return $migration_ip if defined($migration_ip);
4989
4990         my $cidr = $migrate_opts->{network};
4991
4992         if (!defined($cidr)) {
4993             my $dc_conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4994             $cidr = $dc_conf->{migration}->{network};
4995         }
4996
4997         if (defined($cidr)) {
4998             my $ips = PVE::Network::get_local_ip_from_cidr($cidr);
4999
5000             die "could not get IP: no address configured on local " .
5001                 "node for network '$cidr'\n" if scalar(@$ips) == 0;
5002
5003             die "could not get IP: multiple addresses configured on local " .
5004                 "node for network '$cidr'\n" if scalar(@$ips) > 1;
5005
5006             $migration_ip = @$ips[0];
5007         }
5008
5009         $migration_ip = PVE::Cluster::remote_node_ip($nodename, 1)
5010             if !defined($migration_ip);
5011
5012         return $migration_ip;
5013     };
5014
5015     my $migrate_uri;
5016     if ($statefile) {
5017         if ($statefile eq 'tcp') {
5018             my $localip = "localhost";
5019             my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5020             my $nodename = nodename();
5021
5022             if (!defined($migration_type)) {
5023                 if (defined($datacenterconf->{migration}->{type})) {
5024                     $migration_type = $datacenterconf->{migration}->{type};
5025                 } else {
5026                     $migration_type = 'secure';
5027                 }
5028             }
5029
5030             if ($migration_type eq 'insecure') {
5031                 $localip = $get_migration_ip->($nodename);
5032                 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5033             }
5034
5035             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5036             my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5037             $migrate_uri = "tcp:${localip}:${migrate_port}";
5038             push @$cmd, '-incoming', $migrate_uri;
5039             push @$cmd, '-S';
5040
5041         } elsif ($statefile eq 'unix') {
5042             # should be default for secure migrations as a ssh TCP forward
5043             # tunnel is not deterministic reliable ready and fails regurarly
5044             # to set up in time, so use UNIX socket forwards
5045             my $socket_addr = "/run/qemu-server/$vmid.migrate";
5046             unlink $socket_addr;
5047
5048             $migrate_uri = "unix:$socket_addr";
5049
5050             push @$cmd, '-incoming', $migrate_uri;
5051             push @$cmd, '-S';
5052
5053         } elsif (-e $statefile) {
5054             push @$cmd, '-loadstate', $statefile;
5055         } else {
5056             my $statepath = PVE::Storage::path($storecfg, $statefile);
5057             push @$vollist, $statefile;
5058             push @$cmd, '-loadstate', $statepath;
5059         }
5060     } elsif ($params->{paused}) {
5061         push @$cmd, '-S';
5062     }
5063
5064     # host pci devices
5065     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
5066       my $d = parse_hostpci($conf->{"hostpci$i"});
5067       next if !$d;
5068       my $pcidevices = $d->{pciid};
5069       foreach my $pcidevice (@$pcidevices) {
5070             my $pciid = $pcidevice->{id};
5071
5072             my $info = PVE::SysFSTools::pci_device_info("$pciid");
5073             die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5074             die "no pci device info for device '$pciid'\n" if !$info;
5075
5076             if ($d->{mdev}) {
5077                 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5078                 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5079             } else {
5080                 die "can't unbind/bind pci group to vfio '$pciid'\n"
5081                     if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5082                 die "can't reset pci device '$pciid'\n"
5083                     if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5084             }
5085       }
5086     }
5087
5088     PVE::Storage::activate_volumes($storecfg, $vollist);
5089
5090     eval {
5091         run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5092             outfunc => sub {}, errfunc => sub {});
5093     };
5094     # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5095     # timeout should be more than enough here...
5096     PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5097
5098     my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5099                                               : $defaults->{cpuunits};
5100
5101     my $start_timeout = $params->{timeout} // config_aware_timeout($conf, $resume);
5102     my %run_params = (
5103         timeout => $statefile ? undef : $start_timeout,
5104         umask => 0077,
5105         noerr => 1,
5106     );
5107
5108     # when migrating, prefix QEMU output so other side can pick up any
5109     # errors that might occur and show the user
5110     if ($migratedfrom) {
5111         $run_params{quiet} = 1;
5112         $run_params{logfunc} = sub { print "QEMU: $_[0]\n" };
5113     }
5114
5115     my %properties = (
5116         Slice => 'qemu.slice',
5117         KillMode => 'none',
5118         CPUShares => $cpuunits
5119     );
5120
5121     if (my $cpulimit = $conf->{cpulimit}) {
5122         $properties{CPUQuota} = int($cpulimit * 100);
5123     }
5124     $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5125
5126     my $run_qemu = sub {
5127         PVE::Tools::run_fork sub {
5128             PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5129
5130             my $exitcode = run_command($cmd, %run_params);
5131             die "QEMU exited with code $exitcode\n" if $exitcode;
5132         };
5133     };
5134
5135     if ($conf->{hugepages}) {
5136
5137         my $code = sub {
5138             my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5139             my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5140
5141             PVE::QemuServer::Memory::hugepages_mount();
5142             PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5143
5144             eval { $run_qemu->() };
5145             if (my $err = $@) {
5146                 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5147                 die $err;
5148             }
5149
5150             PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5151         };
5152         eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5153
5154     } else {
5155         eval { $run_qemu->() };
5156     }
5157
5158     if (my $err = $@) {
5159         # deactivate volumes if start fails
5160         eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5161         die "start failed: $err";
5162     }
5163
5164     print "migration listens on $migrate_uri\n" if $migrate_uri;
5165     $res->{migrate_uri} = $migrate_uri;
5166
5167     if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5168         eval { mon_cmd($vmid, "cont"); };
5169         warn $@ if $@;
5170     }
5171
5172     #start nbd server for storage migration
5173     if (my $nbd = $migrate_opts->{nbd}) {
5174         my $nbd_protocol_version = $migrate_opts->{nbd_proto_version} // 0;
5175
5176         my $migrate_storage_uri;
5177         # nbd_protocol_version > 0 for unix socket support
5178         if ($nbd_protocol_version > 0 && $migration_type eq 'secure') {
5179             my $socket_path = "/run/qemu-server/$vmid\_nbd.migrate";
5180             mon_cmd($vmid, "nbd-server-start", addr => { type => 'unix', data => { path => $socket_path } } );
5181             $migrate_storage_uri = "nbd:unix:$socket_path";
5182         } else {
5183             my $nodename = nodename();
5184             my $localip = $get_migration_ip->($nodename);
5185             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5186             my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5187
5188             mon_cmd($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5189             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5190             $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}";
5191         }
5192
5193         $res->{migrate_storage_uri} = $migrate_storage_uri;
5194
5195         foreach my $opt (sort keys %$nbd) {
5196             my $drivestr = $nbd->{$opt}->{drivestr};
5197             my $volid = $nbd->{$opt}->{volid};
5198             mon_cmd($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5199             my $nbd_uri = "$migrate_storage_uri:exportname=drive-$opt";
5200             print "storage migration listens on $nbd_uri volume:$drivestr\n";
5201             print "re-using replicated volume: $opt - $volid\n"
5202                 if $nbd->{$opt}->{replicated};
5203
5204             $res->{drives}->{$opt} = $nbd->{$opt};
5205             $res->{drives}->{$opt}->{nbd_uri} = $nbd_uri;
5206         }
5207     }
5208
5209     if ($migratedfrom) {
5210         eval {
5211             set_migration_caps($vmid);
5212         };
5213         warn $@ if $@;
5214
5215         if ($spice_port) {
5216             print "spice listens on port $spice_port\n";
5217             $res->{spice_port} = $spice_port;
5218             if ($migrate_opts->{spice_ticket}) {
5219                 mon_cmd($vmid, "set_password", protocol => 'spice', password => $migrate_opts->{spice_ticket});
5220                 mon_cmd($vmid, "expire_password", protocol => 'spice', time => "+30");
5221             }
5222         }
5223
5224     } else {
5225         mon_cmd($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5226             if !$statefile && $conf->{balloon};
5227
5228         foreach my $opt (keys %$conf) {
5229             next if $opt !~  m/^net\d+$/;
5230             my $nicconf = parse_net($conf->{$opt});
5231             qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5232         }
5233     }
5234
5235     mon_cmd($vmid, 'qom-set',
5236                 path => "machine/peripheral/balloon0",
5237                 property => "guest-stats-polling-interval",
5238                 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5239
5240     if ($resume) {
5241         print "Resumed VM, removing state\n";
5242         if (my $vmstate = $conf->{vmstate}) {
5243             PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5244             PVE::Storage::vdisk_free($storecfg, $vmstate);
5245         }
5246         delete $conf->@{qw(lock vmstate runningmachine runningcpu)};
5247         PVE::QemuConfig->write_config($vmid, $conf);
5248     }
5249
5250     PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5251
5252     return $res;
5253 }
5254
5255 sub vm_commandline {
5256     my ($storecfg, $vmid, $snapname) = @_;
5257
5258     my $conf = PVE::QemuConfig->load_config($vmid);
5259     my $forcemachine;
5260     my $forcecpu;
5261
5262     if ($snapname) {
5263         my $snapshot = $conf->{snapshots}->{$snapname};
5264         die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5265
5266         # check for machine or CPU overrides in snapshot
5267         $forcemachine = $snapshot->{runningmachine};
5268         $forcecpu = $snapshot->{runningcpu};
5269
5270         $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5271
5272         $conf = $snapshot;
5273     }
5274
5275     my $defaults = load_defaults();
5276
5277     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults,
5278         $forcemachine, $forcecpu);
5279
5280     return PVE::Tools::cmd2string($cmd);
5281 }
5282
5283 sub vm_reset {
5284     my ($vmid, $skiplock) = @_;
5285
5286     PVE::QemuConfig->lock_config($vmid, sub {
5287
5288         my $conf = PVE::QemuConfig->load_config($vmid);
5289
5290         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5291
5292         mon_cmd($vmid, "system_reset");
5293     });
5294 }
5295
5296 sub get_vm_volumes {
5297     my ($conf) = @_;
5298
5299     my $vollist = [];
5300     foreach_volid($conf, sub {
5301         my ($volid, $attr) = @_;
5302
5303         return if $volid =~ m|^/|;
5304
5305         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5306         return if !$sid;
5307
5308         push @$vollist, $volid;
5309     });
5310
5311     return $vollist;
5312 }
5313
5314 sub vm_stop_cleanup {
5315     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5316
5317     eval {
5318
5319         if (!$keepActive) {
5320             my $vollist = get_vm_volumes($conf);
5321             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5322         }
5323
5324         foreach my $ext (qw(mon qmp pid vnc qga)) {
5325             unlink "/var/run/qemu-server/${vmid}.$ext";
5326         }
5327
5328         if ($conf->{ivshmem}) {
5329             my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5330             # just delete it for now, VMs which have this already open do not
5331             # are affected, but new VMs will get a separated one. If this
5332             # becomes an issue we either add some sort of ref-counting or just
5333             # add a "don't delete on stop" flag to the ivshmem format.
5334             unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5335         }
5336
5337         foreach my $key (keys %$conf) {
5338             next if $key !~ m/^hostpci(\d+)$/;
5339             my $hostpciindex = $1;
5340             my $d = parse_hostpci($conf->{$key});
5341             my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5342
5343             foreach my $pci (@{$d->{pciid}}) {
5344                 my $pciid = $pci->{id};
5345                 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5346             }
5347         }
5348
5349         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5350     };
5351     warn $@ if $@; # avoid errors - just warn
5352 }
5353
5354 # call only in locked context
5355 sub _do_vm_stop {
5356     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5357
5358     my $pid = check_running($vmid, $nocheck);
5359     return if !$pid;
5360
5361     my $conf;
5362     if (!$nocheck) {
5363         $conf = PVE::QemuConfig->load_config($vmid);
5364         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5365         if (!defined($timeout) && $shutdown && $conf->{startup}) {
5366             my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5367             $timeout = $opts->{down} if $opts->{down};
5368         }
5369         PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5370     }
5371
5372     eval {
5373         if ($shutdown) {
5374             if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5375                 mon_cmd($vmid, "guest-shutdown", timeout => $timeout);
5376             } else {
5377                 mon_cmd($vmid, "system_powerdown");
5378             }
5379         } else {
5380             mon_cmd($vmid, "quit");
5381         }
5382     };
5383     my $err = $@;
5384
5385     if (!$err) {
5386         $timeout = 60 if !defined($timeout);
5387
5388         my $count = 0;
5389         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5390             $count++;
5391             sleep 1;
5392         }
5393
5394         if ($count >= $timeout) {
5395             if ($force) {
5396                 warn "VM still running - terminating now with SIGTERM\n";
5397                 kill 15, $pid;
5398             } else {
5399                 die "VM quit/powerdown failed - got timeout\n";
5400             }
5401         } else {
5402             vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5403             return;
5404         }
5405     } else {
5406        if (!check_running($vmid, $nocheck)) {
5407             warn "Unexpected: VM shutdown command failed, but VM not running anymore..\n";
5408             return;
5409        }
5410        if ($force) {
5411             warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5412             kill 15, $pid;
5413         } else {
5414             die "VM quit/powerdown failed\n";
5415         }
5416     }
5417
5418     # wait again
5419     $timeout = 10;
5420
5421     my $count = 0;
5422     while (($count < $timeout) && check_running($vmid, $nocheck)) {
5423         $count++;
5424         sleep 1;
5425     }
5426
5427     if ($count >= $timeout) {
5428         warn "VM still running - terminating now with SIGKILL\n";
5429         kill 9, $pid;
5430         sleep 1;
5431     }
5432
5433     vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5434 }
5435
5436 # Note: use $nocheck to skip tests if VM configuration file exists.
5437 # We need that when migration VMs to other nodes (files already moved)
5438 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5439 sub vm_stop {
5440     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5441
5442     $force = 1 if !defined($force) && !$shutdown;
5443
5444     if ($migratedfrom){
5445         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5446         kill 15, $pid if $pid;
5447         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5448         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5449         return;
5450     }
5451
5452     PVE::QemuConfig->lock_config($vmid, sub {
5453         _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5454    });
5455 }
5456
5457 sub vm_reboot {
5458     my ($vmid, $timeout) = @_;
5459
5460     PVE::QemuConfig->lock_config($vmid, sub {
5461         eval {
5462
5463             # only reboot if running, as qmeventd starts it again on a stop event
5464             return if !check_running($vmid);
5465
5466             create_reboot_request($vmid);
5467
5468             my $storecfg = PVE::Storage::config();
5469             _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5470
5471         };
5472         if (my $err = $@) {
5473             # avoid that the next normal shutdown will be confused for a reboot
5474             clear_reboot_request($vmid);
5475             die $err;
5476         }
5477    });
5478 }
5479
5480 # note: if using the statestorage parameter, the caller has to check privileges
5481 sub vm_suspend {
5482     my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5483
5484     my $conf;
5485     my $path;
5486     my $storecfg;
5487     my $vmstate;
5488
5489     PVE::QemuConfig->lock_config($vmid, sub {
5490
5491         $conf = PVE::QemuConfig->load_config($vmid);
5492
5493         my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5494         PVE::QemuConfig->check_lock($conf)
5495             if !($skiplock || $is_backing_up);
5496
5497         die "cannot suspend to disk during backup\n"
5498             if $is_backing_up && $includestate;
5499
5500         if ($includestate) {
5501             $conf->{lock} = 'suspending';
5502             my $date = strftime("%Y-%m-%d", localtime(time()));
5503             $storecfg = PVE::Storage::config();
5504             if (!$statestorage) {
5505                 $statestorage = find_vmstate_storage($conf, $storecfg);
5506                 # check permissions for the storage
5507                 my $rpcenv = PVE::RPCEnvironment::get();
5508                 if ($rpcenv->{type} ne 'cli') {
5509                     my $authuser = $rpcenv->get_user();
5510                     $rpcenv->check($authuser, "/storage/$statestorage", ['Datastore.AllocateSpace']);
5511                 }
5512             }
5513
5514
5515             $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5516             $path = PVE::Storage::path($storecfg, $vmstate);
5517             PVE::QemuConfig->write_config($vmid, $conf);
5518         } else {
5519             mon_cmd($vmid, "stop");
5520         }
5521     });
5522
5523     if ($includestate) {
5524         # save vm state
5525         PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5526
5527         eval {
5528             mon_cmd($vmid, "savevm-start", statefile => $path);
5529             for(;;) {
5530                 my $state = mon_cmd($vmid, "query-savevm");
5531                 if (!$state->{status}) {
5532                     die "savevm not active\n";
5533                 } elsif ($state->{status} eq 'active') {
5534                     sleep(1);
5535                     next;
5536                 } elsif ($state->{status} eq 'completed') {
5537                     print "State saved, quitting\n";
5538                     last;
5539                 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5540                     die "query-savevm failed with error '$state->{error}'\n"
5541                 } else {
5542                     die "query-savevm returned status '$state->{status}'\n";
5543                 }
5544             }
5545         };
5546         my $err = $@;
5547
5548         PVE::QemuConfig->lock_config($vmid, sub {
5549             $conf = PVE::QemuConfig->load_config($vmid);
5550             if ($err) {
5551                 # cleanup, but leave suspending lock, to indicate something went wrong
5552                 eval {
5553                     mon_cmd($vmid, "savevm-end");
5554                     PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5555                     PVE::Storage::vdisk_free($storecfg, $vmstate);
5556                     delete $conf->@{qw(vmstate runningmachine runningcpu)};
5557                     PVE::QemuConfig->write_config($vmid, $conf);
5558                 };
5559                 warn $@ if $@;
5560                 die $err;
5561             }
5562
5563             die "lock changed unexpectedly\n"
5564                 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5565
5566             mon_cmd($vmid, "quit");
5567             $conf->{lock} = 'suspended';
5568             PVE::QemuConfig->write_config($vmid, $conf);
5569         });
5570     }
5571 }
5572
5573 sub vm_resume {
5574     my ($vmid, $skiplock, $nocheck) = @_;
5575
5576     PVE::QemuConfig->lock_config($vmid, sub {
5577         my $res = mon_cmd($vmid, 'query-status');
5578         my $resume_cmd = 'cont';
5579
5580         if ($res->{status} && $res->{status} eq 'suspended') {
5581             $resume_cmd = 'system_wakeup';
5582         }
5583
5584         if (!$nocheck) {
5585
5586             my $conf = PVE::QemuConfig->load_config($vmid);
5587
5588             PVE::QemuConfig->check_lock($conf)
5589                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5590         }
5591
5592         mon_cmd($vmid, $resume_cmd);
5593     });
5594 }
5595
5596 sub vm_sendkey {
5597     my ($vmid, $skiplock, $key) = @_;
5598
5599     PVE::QemuConfig->lock_config($vmid, sub {
5600
5601         my $conf = PVE::QemuConfig->load_config($vmid);
5602
5603         # there is no qmp command, so we use the human monitor command
5604         my $res = PVE::QemuServer::Monitor::hmp_cmd($vmid, "sendkey $key");
5605         die $res if $res ne '';
5606     });
5607 }
5608
5609 # vzdump restore implementaion
5610
5611 sub tar_archive_read_firstfile {
5612     my $archive = shift;
5613
5614     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5615
5616     # try to detect archive type first
5617     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5618         die "unable to open file '$archive'\n";
5619     my $firstfile = <$fh>;
5620     kill 15, $pid;
5621     close $fh;
5622
5623     die "ERROR: archive contaions no data\n" if !$firstfile;
5624     chomp $firstfile;
5625
5626     return $firstfile;
5627 }
5628
5629 sub tar_restore_cleanup {
5630     my ($storecfg, $statfile) = @_;
5631
5632     print STDERR "starting cleanup\n";
5633
5634     if (my $fd = IO::File->new($statfile, "r")) {
5635         while (defined(my $line = <$fd>)) {
5636             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5637                 my $volid = $2;
5638                 eval {
5639                     if ($volid =~ m|^/|) {
5640                         unlink $volid || die 'unlink failed\n';
5641                     } else {
5642                         PVE::Storage::vdisk_free($storecfg, $volid);
5643                     }
5644                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5645                 };
5646                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5647             } else {
5648                 print STDERR "unable to parse line in statfile - $line";
5649             }
5650         }
5651         $fd->close();
5652     }
5653 }
5654
5655 sub restore_file_archive {
5656     my ($archive, $vmid, $user, $opts) = @_;
5657
5658     my $info = PVE::Storage::archive_info($archive);
5659     my $format = $opts->{format} // $info->{format};
5660     my $comp = $info->{compression};
5661
5662     # try to detect archive format
5663     if ($format eq 'tar') {
5664         return restore_tar_archive($archive, $vmid, $user, $opts);
5665     } else {
5666         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5667     }
5668 }
5669
5670 # hepler to remove disks that will not be used after restore
5671 my $restore_cleanup_oldconf = sub {
5672     my ($storecfg, $vmid, $oldconf, $virtdev_hash) = @_;
5673
5674     PVE::QemuConfig->foreach_volume($oldconf, sub {
5675         my ($ds, $drive) = @_;
5676
5677         return if drive_is_cdrom($drive, 1);
5678
5679         my $volid = $drive->{file};
5680         return if !$volid || $volid =~ m|^/|;
5681
5682         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
5683         return if !$path || !$owner || ($owner != $vmid);
5684
5685         # Note: only delete disk we want to restore
5686         # other volumes will become unused
5687         if ($virtdev_hash->{$ds}) {
5688             eval { PVE::Storage::vdisk_free($storecfg, $volid); };
5689             if (my $err = $@) {
5690                 warn $err;
5691             }
5692         }
5693     });
5694
5695     # delete vmstate files, after the restore we have no snapshots anymore
5696     foreach my $snapname (keys %{$oldconf->{snapshots}}) {
5697         my $snap = $oldconf->{snapshots}->{$snapname};
5698         if ($snap->{vmstate}) {
5699             eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5700             if (my $err = $@) {
5701                 warn $err;
5702             }
5703         }
5704     }
5705 };
5706
5707 # Helper to parse vzdump backup device hints
5708 #
5709 # $rpcenv: Environment, used to ckeck storage permissions
5710 # $user: User ID, to check storage permissions
5711 # $storecfg: Storage configuration
5712 # $fh: the file handle for reading the configuration
5713 # $devinfo: should contain device sizes for all backu-up'ed devices
5714 # $options: backup options (pool, default storage)
5715 #
5716 # Return: $virtdev_hash, updates $devinfo (add devname, virtdev, format, storeid)
5717 my $parse_backup_hints = sub {
5718     my ($rpcenv, $user, $storecfg, $fh, $devinfo, $options) = @_;
5719
5720     my $virtdev_hash = {};
5721
5722     while (defined(my $line = <$fh>)) {
5723         if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5724             my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5725             die "archive does not contain data for drive '$virtdev'\n"
5726                 if !$devinfo->{$devname};
5727
5728             if (defined($options->{storage})) {
5729                 $storeid = $options->{storage} || 'local';
5730             } elsif (!$storeid) {
5731                 $storeid = 'local';
5732             }
5733             $format = 'raw' if !$format;
5734             $devinfo->{$devname}->{devname} = $devname;
5735             $devinfo->{$devname}->{virtdev} = $virtdev;
5736             $devinfo->{$devname}->{format} = $format;
5737             $devinfo->{$devname}->{storeid} = $storeid;
5738
5739             # check permission on storage
5740             my $pool = $options->{pool}; # todo: do we need that?
5741             if ($user ne 'root@pam') {
5742                 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5743             }
5744
5745             $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5746         } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
5747             my $virtdev = $1;
5748             my $drive = parse_drive($virtdev, $2);
5749             if (drive_is_cloudinit($drive)) {
5750                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5751                 $storeid = $options->{storage} if defined ($options->{storage});
5752                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5753                 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
5754
5755                 $virtdev_hash->{$virtdev} = {
5756                     format => $format,
5757                     storeid => $storeid,
5758                     size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
5759                     is_cloudinit => 1,
5760                 };
5761             }
5762         }
5763     }
5764
5765     return $virtdev_hash;
5766 };
5767
5768 # Helper to allocate and activate all volumes required for a restore
5769 #
5770 # $storecfg: Storage configuration
5771 # $virtdev_hash: as returned by parse_backup_hints()
5772 #
5773 # Returns: { $virtdev => $volid }
5774 my $restore_allocate_devices = sub {
5775     my ($storecfg, $virtdev_hash, $vmid) = @_;
5776
5777     my $map = {};
5778     foreach my $virtdev (sort keys %$virtdev_hash) {
5779         my $d = $virtdev_hash->{$virtdev};
5780         my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5781         my $storeid = $d->{storeid};
5782         my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5783
5784         # test if requested format is supported
5785         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5786         my $supported = grep { $_ eq $d->{format} } @$validFormats;
5787         $d->{format} = $defFormat if !$supported;
5788
5789         my $name;
5790         if ($d->{is_cloudinit}) {
5791             $name = "vm-$vmid-cloudinit";
5792             $name .= ".$d->{format}" if $d->{format} ne 'raw';
5793         }
5794
5795         my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
5796
5797         print STDERR "new volume ID is '$volid'\n";
5798         $d->{volid} = $volid;
5799
5800         PVE::Storage::activate_volumes($storecfg, [$volid]);
5801
5802         $map->{$virtdev} = $volid;
5803     }
5804
5805     return $map;
5806 };
5807
5808 my $restore_update_config_line = sub {
5809     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5810
5811     return if $line =~ m/^\#qmdump\#/;
5812     return if $line =~ m/^\#vzdump\#/;
5813     return if $line =~ m/^lock:/;
5814     return if $line =~ m/^unused\d+:/;
5815     return if $line =~ m/^parent:/;
5816
5817     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5818     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5819         # try to convert old 1.X settings
5820         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5821         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5822             my ($model, $macaddr) = split(/\=/, $devconfig);
5823             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5824             my $net = {
5825                 model => $model,
5826                 bridge => "vmbr$ind",
5827                 macaddr => $macaddr,
5828             };
5829             my $netstr = print_net($net);
5830
5831             print $outfd "net$cookie->{netcount}: $netstr\n";
5832             $cookie->{netcount}++;
5833         }
5834     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5835         my ($id, $netstr) = ($1, $2);
5836         my $net = parse_net($netstr);
5837         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5838         $netstr = print_net($net);
5839         print $outfd "$id: $netstr\n";
5840     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5841         my $virtdev = $1;
5842         my $value = $3;
5843         my $di = parse_drive($virtdev, $value);
5844         if (defined($di->{backup}) && !$di->{backup}) {
5845             print $outfd "#$line";
5846         } elsif ($map->{$virtdev}) {
5847             delete $di->{format}; # format can change on restore
5848             $di->{file} = $map->{$virtdev};
5849             $value = print_drive($di);
5850             print $outfd "$virtdev: $value\n";
5851         } else {
5852             print $outfd $line;
5853         }
5854     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5855         my $vmgenid = $1;
5856         if ($vmgenid ne '0') {
5857             # always generate a new vmgenid if there was a valid one setup
5858             $vmgenid = generate_uuid();
5859         }
5860         print $outfd "vmgenid: $vmgenid\n";
5861     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5862         my ($uuid, $uuid_str);
5863         UUID::generate($uuid);
5864         UUID::unparse($uuid, $uuid_str);
5865         my $smbios1 = parse_smbios1($2);
5866         $smbios1->{uuid} = $uuid_str;
5867         print $outfd $1.print_smbios1($smbios1)."\n";
5868     } else {
5869         print $outfd $line;
5870     }
5871 };
5872
5873 my $restore_deactivate_volumes = sub {
5874     my ($storecfg, $devinfo) = @_;
5875
5876     my $vollist = [];
5877     foreach my $devname (keys %$devinfo) {
5878         my $volid = $devinfo->{$devname}->{volid};
5879         push @$vollist, $volid if $volid;
5880     }
5881
5882     PVE::Storage::deactivate_volumes($storecfg, $vollist);
5883 };
5884
5885 my $restore_destroy_volumes = sub {
5886     my ($storecfg, $devinfo) = @_;
5887
5888     foreach my $devname (keys %$devinfo) {
5889         my $volid = $devinfo->{$devname}->{volid};
5890         next if !$volid;
5891         eval {
5892             if ($volid =~ m|^/|) {
5893                 unlink $volid || die 'unlink failed\n';
5894             } else {
5895                 PVE::Storage::vdisk_free($storecfg, $volid);
5896             }
5897             print STDERR "temporary volume '$volid' sucessfuly removed\n";
5898         };
5899         print STDERR "unable to cleanup '$volid' - $@" if $@;
5900     }
5901 };
5902
5903 sub scan_volids {
5904     my ($cfg, $vmid) = @_;
5905
5906     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5907
5908     my $volid_hash = {};
5909     foreach my $storeid (keys %$info) {
5910         foreach my $item (@{$info->{$storeid}}) {
5911             next if !($item->{volid} && $item->{size});
5912             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5913             $volid_hash->{$item->{volid}} = $item;
5914         }
5915     }
5916
5917     return $volid_hash;
5918 }
5919
5920 sub update_disk_config {
5921     my ($vmid, $conf, $volid_hash) = @_;
5922
5923     my $changes;
5924     my $prefix = "VM $vmid:";
5925
5926     # used and unused disks
5927     my $referenced = {};
5928
5929     # Note: it is allowed to define multiple storages with same path (alias), so
5930     # we need to check both 'volid' and real 'path' (two different volid can point
5931     # to the same path).
5932
5933     my $referencedpath = {};
5934
5935     # update size info
5936     PVE::QemuConfig->foreach_volume($conf, sub {
5937         my ($opt, $drive) = @_;
5938
5939         my $volid = $drive->{file};
5940         return if !$volid;
5941
5942         # mark volid as "in-use" for next step
5943         $referenced->{$volid} = 1;
5944         if ($volid_hash->{$volid} &&
5945             (my $path = $volid_hash->{$volid}->{path})) {
5946             $referencedpath->{$path} = 1;
5947         }
5948
5949         return if drive_is_cdrom($drive);
5950         return if !$volid_hash->{$volid};
5951
5952         my ($updated, $old_size, $new_size) = PVE::QemuServer::Drive::update_disksize($drive, $volid_hash);
5953         if (defined($updated)) {
5954             $changes = 1;
5955             $conf->{$opt} = print_drive($updated);
5956             print "$prefix size of disk '$volid' ($opt) updated from $old_size to $new_size\n";
5957         }
5958     });
5959
5960     # remove 'unusedX' entry if volume is used
5961     PVE::QemuConfig->foreach_unused_volume($conf, sub {
5962         my ($opt, $drive) = @_;
5963
5964         my $volid = $drive->{file};
5965         return if !$volid;
5966
5967         my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5968         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5969             print "$prefix remove entry '$opt', its volume '$volid' is in use\n";
5970             $changes = 1;
5971             delete $conf->{$opt};
5972         }
5973
5974         $referenced->{$volid} = 1;
5975         $referencedpath->{$path} = 1 if $path;
5976     });
5977
5978     foreach my $volid (sort keys %$volid_hash) {
5979         next if $volid =~ m/vm-$vmid-state-/;
5980         next if $referenced->{$volid};
5981         my $path = $volid_hash->{$volid}->{path};
5982         next if !$path; # just to be sure
5983         next if $referencedpath->{$path};
5984         $changes = 1;
5985         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5986         print "$prefix add unreferenced volume '$volid' as '$key' to config\n";
5987         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5988     }
5989
5990     return $changes;
5991 }
5992
5993 sub rescan {
5994     my ($vmid, $nolock, $dryrun) = @_;
5995
5996     my $cfg = PVE::Storage::config();
5997
5998     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5999     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
6000     foreach my $stor (keys %{$cfg->{ids}}) {
6001         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
6002     }
6003
6004     print "rescan volumes...\n";
6005     my $volid_hash = scan_volids($cfg, $vmid);
6006
6007     my $updatefn =  sub {
6008         my ($vmid) = @_;
6009
6010         my $conf = PVE::QemuConfig->load_config($vmid);
6011
6012         PVE::QemuConfig->check_lock($conf);
6013
6014         my $vm_volids = {};
6015         foreach my $volid (keys %$volid_hash) {
6016             my $info = $volid_hash->{$volid};
6017             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
6018         }
6019
6020         my $changes = update_disk_config($vmid, $conf, $vm_volids);
6021
6022         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
6023     };
6024
6025     if (defined($vmid)) {
6026         if ($nolock) {
6027             &$updatefn($vmid);
6028         } else {
6029             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6030         }
6031     } else {
6032         my $vmlist = config_list();
6033         foreach my $vmid (keys %$vmlist) {
6034             if ($nolock) {
6035                 &$updatefn($vmid);
6036             } else {
6037                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6038             }
6039         }
6040     }
6041 }
6042
6043 sub restore_proxmox_backup_archive {
6044     my ($archive, $vmid, $user, $options) = @_;
6045
6046     my $storecfg = PVE::Storage::config();
6047
6048     my ($storeid, $volname) = PVE::Storage::parse_volume_id($archive);
6049     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6050
6051     my $server = $scfg->{server};
6052     my $datastore = $scfg->{datastore};
6053     my $username = $scfg->{username} // 'root@pam';
6054     my $fingerprint = $scfg->{fingerprint};
6055
6056     my $repo = "$username\@$server:$datastore";
6057     my $password = PVE::Storage::PBSPlugin::pbs_get_password($scfg, $storeid);
6058     local $ENV{PBS_PASSWORD} = $password;
6059     local $ENV{PBS_FINGERPRINT} = $fingerprint if defined($fingerprint);
6060
6061     my ($vtype, $pbs_backup_name, undef, undef, undef, undef, $format) =
6062         PVE::Storage::parse_volname($storecfg, $archive);
6063
6064     die "got unexpected vtype '$vtype'\n" if $vtype ne 'backup';
6065
6066     die "got unexpected backup format '$format'\n" if $format ne 'pbs-vm';
6067
6068     my $tmpdir = "/var/tmp/vzdumptmp$$";
6069     rmtree $tmpdir;
6070     mkpath $tmpdir;
6071
6072     my $conffile = PVE::QemuConfig->config_file($vmid);
6073     my $tmpfn = "$conffile.$$.tmp";
6074      # disable interrupts (always do cleanups)
6075     local $SIG{INT} =
6076         local $SIG{TERM} =
6077         local $SIG{QUIT} =
6078         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6079
6080     # Note: $oldconf is undef if VM does not exists
6081     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6082     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6083
6084     my $rpcenv = PVE::RPCEnvironment::get();
6085     my $devinfo = {};
6086
6087     eval {
6088         # enable interrupts
6089         local $SIG{INT} =
6090             local $SIG{TERM} =
6091             local $SIG{QUIT} =
6092             local $SIG{HUP} =
6093             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6094
6095         my $cfgfn = "$tmpdir/qemu-server.conf";
6096         my $firewall_config_fn = "$tmpdir/fw.conf";
6097         my $index_fn = "$tmpdir/index.json";
6098
6099         my $cmd = "restore";
6100
6101         my $param = [$pbs_backup_name, "index.json", $index_fn];
6102         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6103         my $index = PVE::Tools::file_get_contents($index_fn);
6104         $index = decode_json($index);
6105
6106         # print Dumper($index);
6107         foreach my $info (@{$index->{files}}) {
6108             if ($info->{filename} =~ m/^(drive-\S+).img.fidx$/) {
6109                 my $devname = $1;
6110                 if ($info->{size} =~ m/^(\d+)$/) { # untaint size
6111                     $devinfo->{$devname}->{size} = $1;
6112                 } else {
6113                     die "unable to parse file size in 'index.json' - got '$info->{size}'\n";
6114                 }
6115             }
6116         }
6117
6118         my $is_qemu_server_backup = scalar(grep { $_->{filename} eq 'qemu-server.conf.blob' } @{$index->{files}});
6119         if (!$is_qemu_server_backup) {
6120             die "backup does not look like a qemu-server backup (missing 'qemu-server.conf' file)\n";
6121         }
6122         my $has_firewall_config = scalar(grep { $_->{filename} eq 'fw.conf.blob' } @{$index->{files}});
6123
6124         $param = [$pbs_backup_name, "qemu-server.conf", $cfgfn];
6125         PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6126
6127         if ($has_firewall_config) {
6128             $param = [$pbs_backup_name, "fw.conf", $firewall_config_fn];
6129             PVE::Storage::PBSPlugin::run_raw_client_cmd($scfg, $storeid, $cmd, $param);
6130
6131             my $pve_firewall_dir = '/etc/pve/firewall';
6132             mkdir $pve_firewall_dir; # make sure the dir exists
6133             PVE::Tools::file_copy($firewall_config_fn, "${pve_firewall_dir}/$vmid.fw");
6134         }
6135
6136         my $fh = IO::File->new($cfgfn, "r") ||
6137             "unable to read qemu-server.conf - $!\n";
6138
6139         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $storecfg, $fh, $devinfo, $options);
6140
6141         # fixme: rate limit?
6142
6143         # create empty/temp config
6144         PVE::Tools::file_set_contents($conffile, "memory: 128\nlock: create");
6145
6146         $restore_cleanup_oldconf->($storecfg, $vmid, $oldconf, $virtdev_hash) if $oldconf;
6147
6148         # allocate volumes
6149         my $map = $restore_allocate_devices->($storecfg, $virtdev_hash, $vmid);
6150
6151         foreach my $virtdev (sort keys %$virtdev_hash) {
6152             my $d = $virtdev_hash->{$virtdev};
6153             next if $d->{is_cloudinit}; # no need to restore cloudinit
6154
6155             my $volid = $d->{volid};
6156
6157             my $path = PVE::Storage::path($storecfg, $volid);
6158
6159             my $pbs_restore_cmd = [
6160                 '/usr/bin/pbs-restore',
6161                 '--repository', $repo,
6162                 $pbs_backup_name,
6163                 "$d->{devname}.img.fidx",
6164                 $path,
6165                 '--verbose',
6166                 ];
6167
6168             if (PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $volid)) {
6169                 push @$pbs_restore_cmd, '--skip-zero';
6170             }
6171
6172             my $dbg_cmdstring = PVE::Tools::cmd2string($pbs_restore_cmd);
6173             print "restore proxmox backup image: $dbg_cmdstring\n";
6174             run_command($pbs_restore_cmd);
6175         }
6176
6177         $fh->seek(0, 0) || die "seek failed - $!\n";
6178
6179         my $outfd = new IO::File ($tmpfn, "w") ||
6180             die "unable to write config for VM $vmid\n";
6181
6182         my $cookie = { netcount => 0 };
6183         while (defined(my $line = <$fh>)) {
6184             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $options->{unique});
6185         }
6186
6187         $fh->close();
6188         $outfd->close();
6189     };
6190     my $err = $@;
6191
6192     $restore_deactivate_volumes->($storecfg, $devinfo);
6193
6194     rmtree $tmpdir;
6195
6196     if ($err) {
6197         unlink $tmpfn;
6198         $restore_destroy_volumes->($storecfg, $devinfo);
6199         die $err;
6200     }
6201
6202     rename($tmpfn, $conffile) ||
6203         die "unable to commit configuration file '$conffile'\n";
6204
6205     PVE::Cluster::cfs_update(); # make sure we read new file
6206
6207     eval { rescan($vmid, 1); };
6208     warn $@ if $@;
6209 }
6210
6211 sub restore_vma_archive {
6212     my ($archive, $vmid, $user, $opts, $comp) = @_;
6213
6214     my $readfrom = $archive;
6215
6216     my $cfg = PVE::Storage::config();
6217     my $commands = [];
6218     my $bwlimit = $opts->{bwlimit};
6219
6220     my $dbg_cmdstring = '';
6221     my $add_pipe = sub {
6222         my ($cmd) = @_;
6223         push @$commands, $cmd;
6224         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6225         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6226         $readfrom = '-';
6227     };
6228
6229     my $input = undef;
6230     if ($archive eq '-') {
6231         $input = '<&STDIN';
6232     } else {
6233         # If we use a backup from a PVE defined storage we also consider that
6234         # storage's rate limit:
6235         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6236         if (defined($volid)) {
6237             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6238             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6239             if ($readlimit) {
6240                 print STDERR "applying read rate limit: $readlimit\n";
6241                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6242                 $add_pipe->($cstream);
6243             }
6244         }
6245     }
6246
6247     if ($comp) {
6248         my $info = PVE::Storage::decompressor_info('vma', $comp);
6249         my $cmd = $info->{decompressor};
6250         push @$cmd, $readfrom;
6251         $add_pipe->($cmd);
6252     }
6253
6254     my $tmpdir = "/var/tmp/vzdumptmp$$";
6255     rmtree $tmpdir;
6256
6257     # disable interrupts (always do cleanups)
6258     local $SIG{INT} =
6259         local $SIG{TERM} =
6260         local $SIG{QUIT} =
6261         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6262
6263     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6264     POSIX::mkfifo($mapfifo, 0600);
6265     my $fifofh;
6266
6267     my $openfifo = sub {
6268         open($fifofh, '>', $mapfifo) || die $!;
6269     };
6270
6271     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6272
6273     my $oldtimeout;
6274     my $timeout = 5;
6275
6276     my $devinfo = {};
6277
6278     my $rpcenv = PVE::RPCEnvironment::get();
6279
6280     my $conffile = PVE::QemuConfig->config_file($vmid);
6281     my $tmpfn = "$conffile.$$.tmp";
6282
6283     # Note: $oldconf is undef if VM does not exist
6284     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6285     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6286
6287     my %storage_limits;
6288
6289     my $print_devmap = sub {
6290         my $cfgfn = "$tmpdir/qemu-server.conf";
6291
6292         # we can read the config - that is already extracted
6293         my $fh = IO::File->new($cfgfn, "r") ||
6294             "unable to read qemu-server.conf - $!\n";
6295
6296         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6297         if (-f $fwcfgfn) {
6298             my $pve_firewall_dir = '/etc/pve/firewall';
6299             mkdir $pve_firewall_dir; # make sure the dir exists
6300             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6301         }
6302
6303         my $virtdev_hash = $parse_backup_hints->($rpcenv, $user, $cfg, $fh, $devinfo, $opts);
6304
6305         foreach my $key (keys %storage_limits) {
6306             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6307             next if !$limit;
6308             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6309             $storage_limits{$key} = $limit * 1024;
6310         }
6311
6312         foreach my $devname (keys %$devinfo) {
6313             die "found no device mapping information for device '$devname'\n"
6314                 if !$devinfo->{$devname}->{virtdev};
6315         }
6316
6317         # create empty/temp config
6318         if ($oldconf) {
6319             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6320             $restore_cleanup_oldconf->($cfg, $vmid, $oldconf, $virtdev_hash);
6321         }
6322
6323         # allocate volumes
6324         my $map = $restore_allocate_devices->($cfg, $virtdev_hash, $vmid);
6325
6326         # print restore information to $fifofh
6327         foreach my $virtdev (sort keys %$virtdev_hash) {
6328             my $d = $virtdev_hash->{$virtdev};
6329             next if $d->{is_cloudinit}; # no need to restore cloudinit
6330
6331             my $storeid = $d->{storeid};
6332             my $volid = $d->{volid};
6333
6334             my $map_opts = '';
6335             if (my $limit = $storage_limits{$storeid}) {
6336                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6337             }
6338
6339             my $write_zeros = 1;
6340             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6341                 $write_zeros = 0;
6342             }
6343
6344             my $path = PVE::Storage::path($cfg, $volid);
6345
6346             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6347
6348             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6349         }
6350
6351         $fh->seek(0, 0) || die "seek failed - $!\n";
6352
6353         my $outfd = new IO::File ($tmpfn, "w") ||
6354             die "unable to write config for VM $vmid\n";
6355
6356         my $cookie = { netcount => 0 };
6357         while (defined(my $line = <$fh>)) {
6358             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6359         }
6360
6361         $fh->close();
6362         $outfd->close();
6363     };
6364
6365     eval {
6366         # enable interrupts
6367         local $SIG{INT} =
6368             local $SIG{TERM} =
6369             local $SIG{QUIT} =
6370             local $SIG{HUP} =
6371             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6372         local $SIG{ALRM} = sub { die "got timeout\n"; };
6373
6374         $oldtimeout = alarm($timeout);
6375
6376         my $parser = sub {
6377             my $line = shift;
6378
6379             print "$line\n";
6380
6381             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6382                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6383                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6384             } elsif ($line =~ m/^CTIME: /) {
6385                 # we correctly received the vma config, so we can disable
6386                 # the timeout now for disk allocation (set to 10 minutes, so
6387                 # that we always timeout if something goes wrong)
6388                 alarm(600);
6389                 &$print_devmap();
6390                 print $fifofh "done\n";
6391                 my $tmp = $oldtimeout || 0;
6392                 $oldtimeout = undef;
6393                 alarm($tmp);
6394                 close($fifofh);
6395             }
6396         };
6397
6398         print "restore vma archive: $dbg_cmdstring\n";
6399         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6400     };
6401     my $err = $@;
6402
6403     alarm($oldtimeout) if $oldtimeout;
6404
6405     $restore_deactivate_volumes->($cfg, $devinfo);
6406
6407     unlink $mapfifo;
6408     rmtree $tmpdir;
6409
6410     if ($err) {
6411         unlink $tmpfn;
6412         $restore_destroy_volumes->($cfg, $devinfo);
6413         die $err;
6414     }
6415
6416     rename($tmpfn, $conffile) ||
6417         die "unable to commit configuration file '$conffile'\n";
6418
6419     PVE::Cluster::cfs_update(); # make sure we read new file
6420
6421     eval { rescan($vmid, 1); };
6422     warn $@ if $@;
6423 }
6424
6425 sub restore_tar_archive {
6426     my ($archive, $vmid, $user, $opts) = @_;
6427
6428     if ($archive ne '-') {
6429         my $firstfile = tar_archive_read_firstfile($archive);
6430         die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6431             if $firstfile ne 'qemu-server.conf';
6432     }
6433
6434     my $storecfg = PVE::Storage::config();
6435
6436     # avoid zombie disks when restoring over an existing VM -> cleanup first
6437     # pass keep_empty_config=1 to keep the config (thus VMID) reserved for us
6438     # skiplock=1 because qmrestore has set the 'create' lock itself already
6439     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6440     destroy_vm($storecfg, $vmid, 1, { lock => 'restore' }) if -f $vmcfgfn;
6441
6442     my $tocmd = "/usr/lib/qemu-server/qmextract";
6443
6444     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6445     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6446     $tocmd .= ' --prealloc' if $opts->{prealloc};
6447     $tocmd .= ' --info' if $opts->{info};
6448
6449     # tar option "xf" does not autodetect compression when read from STDIN,
6450     # so we pipe to zcat
6451     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6452         PVE::Tools::shellquote("--to-command=$tocmd");
6453
6454     my $tmpdir = "/var/tmp/vzdumptmp$$";
6455     mkpath $tmpdir;
6456
6457     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6458     local $ENV{VZDUMP_VMID} = $vmid;
6459     local $ENV{VZDUMP_USER} = $user;
6460
6461     my $conffile = PVE::QemuConfig->config_file($vmid);
6462     my $tmpfn = "$conffile.$$.tmp";
6463
6464     # disable interrupts (always do cleanups)
6465     local $SIG{INT} =
6466         local $SIG{TERM} =
6467         local $SIG{QUIT} =
6468         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6469
6470     eval {
6471         # enable interrupts
6472         local $SIG{INT} =
6473             local $SIG{TERM} =
6474             local $SIG{QUIT} =
6475             local $SIG{HUP} =
6476             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6477
6478         if ($archive eq '-') {
6479             print "extracting archive from STDIN\n";
6480             run_command($cmd, input => "<&STDIN");
6481         } else {
6482             print "extracting archive '$archive'\n";
6483             run_command($cmd);
6484         }
6485
6486         return if $opts->{info};
6487
6488         # read new mapping
6489         my $map = {};
6490         my $statfile = "$tmpdir/qmrestore.stat";
6491         if (my $fd = IO::File->new($statfile, "r")) {
6492             while (defined (my $line = <$fd>)) {
6493                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6494                     $map->{$1} = $2 if $1;
6495                 } else {
6496                     print STDERR "unable to parse line in statfile - $line\n";
6497                 }
6498             }
6499             $fd->close();
6500         }
6501
6502         my $confsrc = "$tmpdir/qemu-server.conf";
6503
6504         my $srcfd = new IO::File($confsrc, "r") ||
6505             die "unable to open file '$confsrc'\n";
6506
6507         my $outfd = new IO::File ($tmpfn, "w") ||
6508             die "unable to write config for VM $vmid\n";
6509
6510         my $cookie = { netcount => 0 };
6511         while (defined (my $line = <$srcfd>)) {
6512             $restore_update_config_line->($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6513         }
6514
6515         $srcfd->close();
6516         $outfd->close();
6517     };
6518     if (my $err = $@) {
6519         unlink $tmpfn;
6520         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6521         die $err;
6522     }
6523
6524     rmtree $tmpdir;
6525
6526     rename $tmpfn, $conffile ||
6527         die "unable to commit configuration file '$conffile'\n";
6528
6529     PVE::Cluster::cfs_update(); # make sure we read new file
6530
6531     eval { rescan($vmid, 1); };
6532     warn $@ if $@;
6533 };
6534
6535 sub foreach_storage_used_by_vm {
6536     my ($conf, $func) = @_;
6537
6538     my $sidhash = {};
6539
6540     PVE::QemuConfig->foreach_volume($conf, sub {
6541         my ($ds, $drive) = @_;
6542         return if drive_is_cdrom($drive);
6543
6544         my $volid = $drive->{file};
6545
6546         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6547         $sidhash->{$sid} = $sid if $sid;
6548     });
6549
6550     foreach my $sid (sort keys %$sidhash) {
6551         &$func($sid);
6552     }
6553 }
6554
6555 my $qemu_snap_storage = {
6556     rbd => 1,
6557 };
6558 sub do_snapshots_with_qemu {
6559     my ($storecfg, $volid) = @_;
6560
6561     my $storage_name = PVE::Storage::parse_volume_id($volid);
6562     my $scfg = $storecfg->{ids}->{$storage_name};
6563
6564     if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6565         return 1;
6566     }
6567
6568     if ($volid =~ m/\.(qcow2|qed)$/){
6569         return 1;
6570     }
6571
6572     return undef;
6573 }
6574
6575 sub qga_check_running {
6576     my ($vmid, $nowarn) = @_;
6577
6578     eval { mon_cmd($vmid, "guest-ping", timeout => 3); };
6579     if ($@) {
6580         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6581         return 0;
6582     }
6583     return 1;
6584 }
6585
6586 sub template_create {
6587     my ($vmid, $conf, $disk) = @_;
6588
6589     my $storecfg = PVE::Storage::config();
6590
6591     PVE::QemuConfig->foreach_volume($conf, sub {
6592         my ($ds, $drive) = @_;
6593
6594         return if drive_is_cdrom($drive);
6595         return if $disk && $ds ne $disk;
6596
6597         my $volid = $drive->{file};
6598         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6599
6600         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6601         $drive->{file} = $voliddst;
6602         $conf->{$ds} = print_drive($drive);
6603         PVE::QemuConfig->write_config($vmid, $conf);
6604     });
6605 }
6606
6607 sub convert_iscsi_path {
6608     my ($path) = @_;
6609
6610     if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6611         my $portal = $1;
6612         my $target = $2;
6613         my $lun = $3;
6614
6615         my $initiator_name = get_initiator_name();
6616
6617         return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6618                "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6619     }
6620
6621     die "cannot convert iscsi path '$path', unkown format\n";
6622 }
6623
6624 sub qemu_img_convert {
6625     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6626
6627     my $storecfg = PVE::Storage::config();
6628     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6629     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6630
6631     die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6632
6633     my $cachemode;
6634     my $src_path;
6635     my $src_is_iscsi = 0;
6636     my $src_format;
6637
6638     if ($src_storeid) {
6639         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6640         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6641         $src_format = qemu_img_format($src_scfg, $src_volname);
6642         $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6643         $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6644         $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6645     } elsif (-f $src_volid) {
6646         $src_path = $src_volid;
6647         if ($src_path =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6648             $src_format = $1;
6649         }
6650     }
6651
6652     die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6653
6654     my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6655     my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6656     my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6657     my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6658
6659     my $cmd = [];
6660     push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6661     push @$cmd, '-l', "snapshot.name=$snapname"
6662         if $snapname && $src_format && $src_format eq "qcow2";
6663     push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6664     push @$cmd, '-T', $cachemode if defined($cachemode);
6665
6666     if ($src_is_iscsi) {
6667         push @$cmd, '--image-opts';
6668         $src_path = convert_iscsi_path($src_path);
6669     } elsif ($src_format) {
6670         push @$cmd, '-f', $src_format;
6671     }
6672
6673     if ($dst_is_iscsi) {
6674         push @$cmd, '--target-image-opts';
6675         $dst_path = convert_iscsi_path($dst_path);
6676     } else {
6677         push @$cmd, '-O', $dst_format;
6678     }
6679
6680     push @$cmd, $src_path;
6681
6682     if (!$dst_is_iscsi && $is_zero_initialized) {
6683         push @$cmd, "zeroinit:$dst_path";
6684     } else {
6685         push @$cmd, $dst_path;
6686     }
6687
6688     my $parser = sub {
6689         my $line = shift;
6690         if($line =~ m/\((\S+)\/100\%\)/){
6691             my $percent = $1;
6692             my $transferred = int($size * $percent / 100);
6693             my $remaining = $size - $transferred;
6694
6695             print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6696         }
6697
6698     };
6699
6700     eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6701     my $err = $@;
6702     die "copy failed: $err" if $err;
6703 }
6704
6705 sub qemu_img_format {
6706     my ($scfg, $volname) = @_;
6707
6708     if ($scfg->{path} && $volname =~ m/\.($PVE::QemuServer::Drive::QEMU_FORMAT_RE)$/) {
6709         return $1;
6710     } else {
6711         return "raw";
6712     }
6713 }
6714
6715 sub qemu_drive_mirror {
6716     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $completion, $qga, $bwlimit, $src_bitmap) = @_;
6717
6718     $jobs = {} if !$jobs;
6719
6720     my $qemu_target;
6721     my $format;
6722     $jobs->{"drive-$drive"} = {};
6723
6724     if ($dst_volid =~ /^nbd:/) {
6725         $qemu_target = $dst_volid;
6726         $format = "nbd";
6727     } else {
6728         my $storecfg = PVE::Storage::config();
6729         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6730
6731         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6732
6733         $format = qemu_img_format($dst_scfg, $dst_volname);
6734
6735         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6736
6737         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6738     }
6739
6740     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6741     $opts->{format} = $format if $format;
6742
6743     if (defined($src_bitmap)) {
6744         $opts->{sync} = 'incremental';
6745         $opts->{bitmap} = $src_bitmap;
6746         print "drive mirror re-using dirty bitmap '$src_bitmap'\n";
6747     }
6748
6749     if (defined($bwlimit)) {
6750         $opts->{speed} = $bwlimit * 1024;
6751         print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
6752     } else {
6753         print "drive mirror is starting for drive-$drive\n";
6754     }
6755
6756     # if a job already runs for this device we get an error, catch it for cleanup
6757     eval { mon_cmd($vmid, "drive-mirror", %$opts); };
6758     if (my $err = $@) {
6759         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6760         warn "$@\n" if $@;
6761         die "mirroring error: $err\n";
6762     }
6763
6764     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $completion, $qga);
6765 }
6766
6767 # $completion can be either
6768 # 'complete': wait until all jobs are ready, block-job-complete them (default)
6769 # 'cancel': wait until all jobs are ready, block-job-cancel them
6770 # 'skip': wait until all jobs are ready, return with block jobs in ready state
6771 sub qemu_drive_mirror_monitor {
6772     my ($vmid, $vmiddst, $jobs, $completion, $qga) = @_;
6773
6774     $completion //= 'complete';
6775
6776     eval {
6777         my $err_complete = 0;
6778
6779         while (1) {
6780             die "storage migration timed out\n" if $err_complete > 300;
6781
6782             my $stats = mon_cmd($vmid, "query-block-jobs");
6783
6784             my $running_mirror_jobs = {};
6785             foreach my $stat (@$stats) {
6786                 next if $stat->{type} ne 'mirror';
6787                 $running_mirror_jobs->{$stat->{device}} = $stat;
6788             }
6789
6790             my $readycounter = 0;
6791
6792             foreach my $job (keys %$jobs) {
6793
6794                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6795                     print "$job : finished\n";
6796                     delete $jobs->{$job};
6797                     next;
6798                 }
6799
6800                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6801
6802                 my $busy = $running_mirror_jobs->{$job}->{busy};
6803                 my $ready = $running_mirror_jobs->{$job}->{ready};
6804                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6805                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6806                     my $remaining = $total - $transferred;
6807                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6808
6809                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6810                 }
6811
6812                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6813             }
6814
6815             last if scalar(keys %$jobs) == 0;
6816
6817             if ($readycounter == scalar(keys %$jobs)) {
6818                 print "all mirroring jobs are ready \n";
6819                 last if $completion eq 'skip'; #do the complete later
6820
6821                 if ($vmiddst && $vmiddst != $vmid) {
6822                     my $agent_running = $qga && qga_check_running($vmid);
6823                     if ($agent_running) {
6824                         print "freeze filesystem\n";
6825                         eval { mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6826                     } else {
6827                         print "suspend vm\n";
6828                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6829                     }
6830
6831                     # if we clone a disk for a new target vm, we don't switch the disk
6832                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6833
6834                     if ($agent_running) {
6835                         print "unfreeze filesystem\n";
6836                         eval { mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6837                     } else {
6838                         print "resume vm\n";
6839                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6840                     }
6841
6842                     last;
6843                 } else {
6844
6845                     foreach my $job (keys %$jobs) {
6846                         # try to switch the disk if source and destination are on the same guest
6847                         print "$job: Completing block job...\n";
6848
6849                         my $op;
6850                         if ($completion eq 'complete') {
6851                             $op = 'block-job-complete';
6852                         } elsif ($completion eq 'cancel') {
6853                             $op = 'block-job-cancel';
6854                         } else {
6855                             die "invalid completion value: $completion\n";
6856                         }
6857                         eval { mon_cmd($vmid, $op, device => $job) };
6858                         if ($@ =~ m/cannot be completed/) {
6859                             print "$job: Block job cannot be completed, try again.\n";
6860                             $err_complete++;
6861                         }else {
6862                             print "$job: Completed successfully.\n";
6863                             $jobs->{$job}->{complete} = 1;
6864                         }
6865                     }
6866                 }
6867             }
6868             sleep 1;
6869         }
6870     };
6871     my $err = $@;
6872
6873     if ($err) {
6874         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6875         die "mirroring error: $err";
6876     }
6877
6878 }
6879
6880 sub qemu_blockjobs_cancel {
6881     my ($vmid, $jobs) = @_;
6882
6883     foreach my $job (keys %$jobs) {
6884         print "$job: Cancelling block job\n";
6885         eval { mon_cmd($vmid, "block-job-cancel", device => $job); };
6886         $jobs->{$job}->{cancel} = 1;
6887     }
6888
6889     while (1) {
6890         my $stats = mon_cmd($vmid, "query-block-jobs");
6891
6892         my $running_jobs = {};
6893         foreach my $stat (@$stats) {
6894             $running_jobs->{$stat->{device}} = $stat;
6895         }
6896
6897         foreach my $job (keys %$jobs) {
6898
6899             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6900                 print "$job: Done.\n";
6901                 delete $jobs->{$job};
6902             }
6903         }
6904
6905         last if scalar(keys %$jobs) == 0;
6906
6907         sleep 1;
6908     }
6909 }
6910
6911 sub clone_disk {
6912     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6913         $newvmid, $storage, $format, $full, $newvollist, $jobs, $completion, $qga, $bwlimit, $conf) = @_;
6914
6915     my $newvolid;
6916
6917     if (!$full) {
6918         print "create linked clone of drive $drivename ($drive->{file})\n";
6919         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6920         push @$newvollist, $newvolid;
6921     } else {
6922
6923         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6924         $storeid = $storage if $storage;
6925
6926         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6927         my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6928
6929         print "create full clone of drive $drivename ($drive->{file})\n";
6930         my $name = undef;
6931         if (drive_is_cloudinit($drive)) {
6932             $name = "vm-$newvmid-cloudinit";
6933             $name .= ".$dst_format" if $dst_format ne 'raw';
6934             $snapname = undef;
6935             $size = PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE;
6936         } elsif ($drivename eq 'efidisk0') {
6937             $size = get_efivars_size($conf);
6938         }
6939         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
6940         push @$newvollist, $newvolid;
6941
6942         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6943
6944         if (drive_is_cloudinit($drive)) {
6945             goto no_data_clone;
6946         }
6947
6948         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6949         if (!$running || $snapname) {
6950             # TODO: handle bwlimits
6951             if ($drivename eq 'efidisk0') {
6952                 # the relevant data on the efidisk may be smaller than the source
6953                 # e.g. on RBD/ZFS, so we use dd to copy only the amount
6954                 # that is given by the OVMF_VARS.fd
6955                 my $src_path = PVE::Storage::path($storecfg, $drive->{file});
6956                 my $dst_path = PVE::Storage::path($storecfg, $newvolid);
6957                 run_command(['qemu-img', 'dd', '-n', '-O', $dst_format, "bs=1", "count=$size", "if=$src_path", "of=$dst_path"]);
6958             } else {
6959                 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6960             }
6961         } else {
6962
6963             my $kvmver = get_running_qemu_version ($vmid);
6964             if (!min_version($kvmver, 2, 7)) {
6965                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6966                     if $drive->{iothread};
6967             }
6968
6969             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $completion, $qga, $bwlimit);
6970         }
6971     }
6972
6973 no_data_clone:
6974     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6975
6976     my $disk = $drive;
6977     $disk->{format} = undef;
6978     $disk->{file} = $newvolid;
6979     $disk->{size} = $size;
6980
6981     return $disk;
6982 }
6983
6984 sub get_running_qemu_version {
6985     my ($vmid) = @_;
6986     my $res = mon_cmd($vmid, "query-version");
6987     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6988 }
6989
6990 sub qemu_use_old_bios_files {
6991     my ($machine_type) = @_;
6992
6993     return if !$machine_type;
6994
6995     my $use_old_bios_files = undef;
6996
6997     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6998         $machine_type = $1;
6999         $use_old_bios_files = 1;
7000     } else {
7001         my $version = PVE::QemuServer::Machine::extract_version($machine_type, kvm_user_version());
7002         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
7003         # load new efi bios files on migration. So this hack is required to allow
7004         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
7005         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
7006         $use_old_bios_files = !min_version($version, 2, 4);
7007     }
7008
7009     return ($use_old_bios_files, $machine_type);
7010 }
7011
7012 sub get_efivars_size {
7013     my ($conf) = @_;
7014     my $arch = get_vm_arch($conf);
7015     my (undef, $ovmf_vars) = get_ovmf_files($arch);
7016     die "uefi vars image '$ovmf_vars' not found\n" if ! -f $ovmf_vars;
7017     return -s $ovmf_vars;
7018 }
7019
7020 sub update_efidisk_size {
7021     my ($conf) = @_;
7022
7023     return if !defined($conf->{efidisk0});
7024
7025     my $disk = PVE::QemuServer::parse_drive('efidisk0', $conf->{efidisk0});
7026     $disk->{size} = get_efivars_size($conf);
7027     $conf->{efidisk0} = print_drive($disk);
7028
7029     return;
7030 }
7031
7032 sub create_efidisk($$$$$) {
7033     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7034
7035     my (undef, $ovmf_vars) = get_ovmf_files($arch);
7036     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7037
7038     my $vars_size_b = -s $ovmf_vars;
7039     my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7040     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7041     PVE::Storage::activate_volumes($storecfg, [$volid]);
7042
7043     qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7044     my ($size) = PVE::Storage::volume_size_info($storecfg, $volid, 3);
7045
7046     return ($volid, $size/1024);
7047 }
7048
7049 sub vm_iothreads_list {
7050     my ($vmid) = @_;
7051
7052     my $res = mon_cmd($vmid, 'query-iothreads');
7053
7054     my $iothreads = {};
7055     foreach my $iothread (@$res) {
7056         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7057     }
7058
7059     return $iothreads;
7060 }
7061
7062 sub scsihw_infos {
7063     my ($conf, $drive) = @_;
7064
7065     my $maxdev = 0;
7066
7067     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7068         $maxdev = 7;
7069     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7070         $maxdev = 1;
7071     } else {
7072         $maxdev = 256;
7073     }
7074
7075     my $controller = int($drive->{index} / $maxdev);
7076     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7077
7078     return ($maxdev, $controller, $controller_prefix);
7079 }
7080
7081 sub windows_version {
7082     my ($ostype) = @_;
7083
7084     return 0 if !$ostype;
7085
7086     my $winversion = 0;
7087
7088     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7089         $winversion = 5;
7090     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7091         $winversion = 6;
7092     } elsif ($ostype =~ m/^win(\d+)$/) {
7093         $winversion = $1;
7094     }
7095
7096     return $winversion;
7097 }
7098
7099 sub resolve_dst_disk_format {
7100         my ($storecfg, $storeid, $src_volname, $format) = @_;
7101         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7102
7103         if (!$format) {
7104             # if no target format is specified, use the source disk format as hint
7105             if ($src_volname) {
7106                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7107                 $format = qemu_img_format($scfg, $src_volname);
7108             } else {
7109                 return $defFormat;
7110             }
7111         }
7112
7113         # test if requested format is supported - else use default
7114         my $supported = grep { $_ eq $format } @$validFormats;
7115         $format = $defFormat if !$supported;
7116         return $format;
7117 }
7118
7119 # NOTE: if this logic changes, please update docs & possibly gui logic
7120 sub find_vmstate_storage {
7121     my ($conf, $storecfg) = @_;
7122
7123     # first, return storage from conf if set
7124     return $conf->{vmstatestorage} if $conf->{vmstatestorage};
7125
7126     my ($target, $shared, $local);
7127
7128     foreach_storage_used_by_vm($conf, sub {
7129         my ($sid) = @_;
7130         my $scfg = PVE::Storage::storage_config($storecfg, $sid);
7131         my $dst = $scfg->{shared} ? \$shared : \$local;
7132         $$dst = $sid if !$$dst || $scfg->{path}; # prefer file based storage
7133     });
7134
7135     # second, use shared storage where VM has at least one disk
7136     # third, use local storage where VM has at least one disk
7137     # fall back to local storage
7138     $target = $shared // $local // 'local';
7139
7140     return $target;
7141 }
7142
7143 sub generate_uuid {
7144     my ($uuid, $uuid_str);
7145     UUID::generate($uuid);
7146     UUID::unparse($uuid, $uuid_str);
7147     return $uuid_str;
7148 }
7149
7150 sub generate_smbios1_uuid {
7151     return "uuid=".generate_uuid();
7152 }
7153
7154 sub nbd_stop {
7155     my ($vmid) = @_;
7156
7157     mon_cmd($vmid, 'nbd-server-stop');
7158 }
7159
7160 sub create_reboot_request {
7161     my ($vmid) = @_;
7162     open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7163         or die "failed to create reboot trigger file: $!\n";
7164     close($fh);
7165 }
7166
7167 sub clear_reboot_request {
7168     my ($vmid) = @_;
7169     my $path = "/run/qemu-server/$vmid.reboot";
7170     my $res = 0;
7171
7172     $res = unlink($path);
7173     die "could not remove reboot request for $vmid: $!"
7174         if !$res && $! != POSIX::ENOENT;
7175
7176     return $res;
7177 }
7178
7179 # bash completion helper
7180
7181 sub complete_backup_archives {
7182     my ($cmdname, $pname, $cvalue) = @_;
7183
7184     my $cfg = PVE::Storage::config();
7185
7186     my $storeid;
7187
7188     if ($cvalue =~ m/^([^:]+):/) {
7189         $storeid = $1;
7190     }
7191
7192     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7193
7194     my $res = [];
7195     foreach my $id (keys %$data) {
7196         foreach my $item (@{$data->{$id}}) {
7197             next if $item->{format} !~ m/^vma\.(${\PVE::Storage::Plugin::COMPRESSOR_RE})$/;
7198             push @$res, $item->{volid} if defined($item->{volid});
7199         }
7200     }
7201
7202     return $res;
7203 }
7204
7205 my $complete_vmid_full = sub {
7206     my ($running) = @_;
7207
7208     my $idlist = vmstatus();
7209
7210     my $res = [];
7211
7212     foreach my $id (keys %$idlist) {
7213         my $d = $idlist->{$id};
7214         if (defined($running)) {
7215             next if $d->{template};
7216             next if $running && $d->{status} ne 'running';
7217             next if !$running && $d->{status} eq 'running';
7218         }
7219         push @$res, $id;
7220
7221     }
7222     return $res;
7223 };
7224
7225 sub complete_vmid {
7226     return &$complete_vmid_full();
7227 }
7228
7229 sub complete_vmid_stopped {
7230     return &$complete_vmid_full(0);
7231 }
7232
7233 sub complete_vmid_running {
7234     return &$complete_vmid_full(1);
7235 }
7236
7237 sub complete_storage {
7238
7239     my $cfg = PVE::Storage::config();
7240     my $ids = $cfg->{ids};
7241
7242     my $res = [];
7243     foreach my $sid (keys %$ids) {
7244         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7245         next if !$ids->{$sid}->{content}->{images};
7246         push @$res, $sid;
7247     }
7248
7249     return $res;
7250 }
7251
7252 sub complete_migration_storage {
7253     my ($cmd, $param, $current_value, $all_args) = @_;
7254
7255     my $targetnode = @$all_args[1];
7256
7257     my $cfg = PVE::Storage::config();
7258     my $ids = $cfg->{ids};
7259
7260     my $res = [];
7261     foreach my $sid (keys %$ids) {
7262         next if !PVE::Storage::storage_check_enabled($cfg, $sid, $targetnode, 1);
7263         next if !$ids->{$sid}->{content}->{images};
7264         push @$res, $sid;
7265     }
7266
7267     return $res;
7268 }
7269
7270 1;