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