PVE/QemuServer.pm

   1 package PVE::QemuServer;
   2
   3 use strict;
   4 use warnings;
   5 use POSIX;
   6 use IO::Handle;
   7 use IO::Select;
   8 use IO::File;
   9 use IO::Dir;
  10 use IO::Socket::UNIX;
  11 use File::Basename;
  12 use File::Path;
  13 use File::stat;
  14 use Getopt::Long;
  15 use Digest::SHA;
  16 use Fcntl ':flock';
  17 use Cwd 'abs_path';
  18 use IPC::Open3;
  19 use JSON;
  20 use Fcntl;
  21 use PVE::SafeSyslog;
  22 use Storable qw(dclone);
  23 use PVE::Exception qw(raise raise_param_exc);
  24 use PVE::Storage;
  25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach $IPV6RE);
  26 use PVE::JSONSchema qw(get_standard_option);
  27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
  28 use PVE::INotify;
  29 use PVE::ProcFSTools;
  30 use PVE::QemuConfig;
  31 use PVE::QMPClient;
  32 use PVE::RPCEnvironment;
  33 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr);
  34 use PVE::QemuServer::Memory;
  35 use PVE::QemuServer::USB qw(parse_usb_device);
  36 use PVE::QemuServer::Cloudinit;
  37 use PVE::Systemd;
  38 use Time::HiRes qw(gettimeofday);
  39 use File::Copy qw(copy);
  40 use URI::Escape;
  41
  42 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
  43 my $OVMF = {
  44     x86_64 => [
  45         "$EDK2_FW_BASE/OVMF_CODE.fd",
  46         "$EDK2_FW_BASE/OVMF_VARS.fd"
  47     ],
  48     aarch64 => [
  49         "$EDK2_FW_BASE/AAVMF_CODE.fd",
  50         "$EDK2_FW_BASE/AAVMF_VARS.fd"
  51     ],
  52 };
  53
  54 my $qemu_snap_storage = {rbd => 1, sheepdog => 1};
  55
  56 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
  57
  58 my $QEMU_FORMAT_RE = qr/raw|cow|qcow|qcow2|qed|vmdk|cloop/;
  59
  60 # Note about locking: we use flock on the config file protect
  61 # against concurent actions.
  62 # Aditionaly, we have a 'lock' setting in the config file. This
  63 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
  64 # allowed when such lock is set. But you can ignore this kind of
  65 # lock with the --skiplock flag.
  66
  67 cfs_register_file('/qemu-server/',
  68                   \&parse_vm_config,
  69                   \&write_vm_config);
  70
  71 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
  72     description => "Some command save/restore state from this location.",
  73     type => 'string',
  74     maxLength => 128,
  75     optional => 1,
  76 });
  77
  78 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
  79     description => "The name of the snapshot.",
  80     type => 'string', format => 'pve-configid',
  81     maxLength => 40,
  82 });
  83
  84 PVE::JSONSchema::register_standard_option('pve-qm-image-format', {
  85     type => 'string',
  86     enum => [qw(raw cow qcow qed qcow2 vmdk cloop)],
  87     description => "The drive's backing file's data format.",
  88     optional => 1,
  89 });
  90
  91 PVE::JSONSchema::register_standard_option('pve-qemu-machine', {
  92         description => "Specifies the Qemu machine type.",
  93         type => 'string',
  94         pattern => '(pc|pc(-i440fx)?-\d+\.\d+(\.pxe)?|q35|pc-q35-\d+\.\d+(\.pxe)?|virt(?:-\d+\.\d+)?)',
  95         maxLength => 40,
  96         optional => 1,
  97 });
  98
  99 #no warnings 'redefine';
 100
 101 sub cgroups_write {
 102    my ($controller, $vmid, $option, $value) = @_;
 103
 104    my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
 105    PVE::ProcFSTools::write_proc_entry($path, $value);
 106
 107 }
 108
 109 my $nodename = PVE::INotify::nodename();
 110
 111 mkdir "/etc/pve/nodes/$nodename";
 112 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
 113 mkdir $confdir;
 114
 115 my $var_run_tmpdir = "/var/run/qemu-server";
 116 mkdir $var_run_tmpdir;
 117
 118 my $lock_dir = "/var/lock/qemu-server";
 119 mkdir $lock_dir;
 120
 121 my $pcisysfs = "/sys/bus/pci";
 122
 123 my $cpu_vendor_list = {
 124     # Intel CPUs
 125     486 => 'GenuineIntel',
 126     pentium => 'GenuineIntel',
 127     pentium2  => 'GenuineIntel',
 128     pentium3  => 'GenuineIntel',
 129     coreduo => 'GenuineIntel',
 130     core2duo => 'GenuineIntel',
 131     Conroe  => 'GenuineIntel',
 132     Penryn  => 'GenuineIntel',
 133     Nehalem  => 'GenuineIntel',
 134     'Nehalem-IBRS'  => 'GenuineIntel',
 135     Westmere => 'GenuineIntel',
 136     'Westmere-IBRS' => 'GenuineIntel',
 137     SandyBridge => 'GenuineIntel',
 138     'SandyBridge-IBRS' => 'GenuineIntel',
 139     IvyBridge => 'GenuineIntel',
 140     'IvyBridge-IBRS' => 'GenuineIntel',
 141     Haswell => 'GenuineIntel',
 142     'Haswell-IBRS' => 'GenuineIntel',
 143     'Haswell-noTSX' => 'GenuineIntel',
 144     'Haswell-noTSX-IBRS' => 'GenuineIntel',
 145     Broadwell => 'GenuineIntel',
 146     'Broadwell-IBRS' => 'GenuineIntel',
 147     'Broadwell-noTSX' => 'GenuineIntel',
 148     'Broadwell-noTSX-IBRS' => 'GenuineIntel',
 149     'Skylake-Client' => 'GenuineIntel',
 150     'Skylake-Client-IBRS' => 'GenuineIntel',
 151     'Skylake-Server' => 'GenuineIntel',
 152     'Skylake-Server-IBRS' => 'GenuineIntel',
 153
 154     # AMD CPUs
 155     athlon => 'AuthenticAMD',
 156     phenom  => 'AuthenticAMD',
 157     Opteron_G1  => 'AuthenticAMD',
 158     Opteron_G2  => 'AuthenticAMD',
 159     Opteron_G3  => 'AuthenticAMD',
 160     Opteron_G4  => 'AuthenticAMD',
 161     Opteron_G5  => 'AuthenticAMD',
 162     EPYC => 'AuthenticAMD',
 163     'EPYC-IBPB' => 'AuthenticAMD',
 164
 165     # generic types, use vendor from host node
 166     host => 'default',
 167     kvm32 => 'default',
 168     kvm64 => 'default',
 169     qemu32 => 'default',
 170     qemu64 => 'default',
 171     max => 'default',
 172 };
 173
 174 my $cpu_flag = qr/[+-](pcid|spec-ctrl|ibpb|ssbd|virt-ssbd|amd-ssbd|amd-no-ssb|pdpe1gb)/;
 175
 176 my $cpu_fmt = {
 177     cputype => {
 178         description => "Emulated CPU type.",
 179         type => 'string',
 180         enum => [ sort { "\L$a" cmp "\L$b" } keys %$cpu_vendor_list ],
 181         default => 'kvm64',
 182         default_key => 1,
 183     },
 184     hidden => {
 185         description => "Do not identify as a KVM virtual machine.",
 186         type => 'boolean',
 187         optional => 1,
 188         default => 0
 189     },
 190     flags => {
 191         description => "List of additional CPU flags separated by ';'."
 192                      . " Use '+FLAG' to enable, '-FLAG' to disable a flag."
 193                      . " Currently supported flags: 'pcid', 'spec-ctrl', 'ibpb', 'ssbd', 'virt-ssbd', 'amd-ssbd', 'amd-no-ssb', 'pdpe1gb'.",
 194         format_description => '+FLAG[;-FLAG...]',
 195         type => 'string',
 196         pattern => qr/$cpu_flag(;$cpu_flag)*/,
 197         optional => 1,
 198     },
 199 };
 200
 201 my $watchdog_fmt = {
 202     model => {
 203         default_key => 1,
 204         type => 'string',
 205         enum => [qw(i6300esb ib700)],
 206         description => "Watchdog type to emulate.",
 207         default => 'i6300esb',
 208         optional => 1,
 209     },
 210     action => {
 211         type => 'string',
 212         enum => [qw(reset shutdown poweroff pause debug none)],
 213         description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
 214         optional => 1,
 215     },
 216 };
 217 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
 218
 219 my $agent_fmt = {
 220     enabled => {
 221         description => "Enable/disable Qemu GuestAgent.",
 222         type => 'boolean',
 223         default => 0,
 224         default_key => 1,
 225     },
 226     fstrim_cloned_disks => {
 227         description => "Run fstrim after cloning/moving a disk.",
 228         type => 'boolean',
 229         optional => 1,
 230         default => 0
 231     },
 232 };
 233
 234 my $vga_fmt = {
 235     type => {
 236         description => "Select the VGA type.",
 237         type => 'string',
 238         default => 'std',
 239         optional => 1,
 240         default_key => 1,
 241         enum => [qw(cirrus qxl qxl2 qxl3 qxl4 serial0 serial1 serial2 serial3 std virtio vmware)],
 242     },
 243     memory => {
 244         description => "Sets the VGA memory (in MiB). Has no effect with serial display.",
 245         type => 'integer',
 246         optional => 1,
 247         minimum => 4,
 248         maximum => 512,
 249     },
 250 };
 251
 252 my $confdesc = {
 253     onboot => {
 254         optional => 1,
 255         type => 'boolean',
 256         description => "Specifies whether a VM will be started during system bootup.",
 257         default => 0,
 258     },
 259     autostart => {
 260         optional => 1,
 261         type => 'boolean',
 262         description => "Automatic restart after crash (currently ignored).",
 263         default => 0,
 264     },
 265     hotplug => {
 266         optional => 1,
 267         type => 'string', format => 'pve-hotplug-features',
 268         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'.",
 269         default => 'network,disk,usb',
 270     },
 271     reboot => {
 272         optional => 1,
 273         type => 'boolean',
 274         description => "Allow reboot. If set to '0' the VM exit on reboot.",
 275         default => 1,
 276     },
 277     lock => {
 278         optional => 1,
 279         type => 'string',
 280         description => "Lock/unlock the VM.",
 281         enum => [qw(migrate backup snapshot rollback)],
 282     },
 283     cpulimit => {
 284         optional => 1,
 285         type => 'number',
 286         description => "Limit of CPU usage.",
 287         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.",
 288         minimum => 0,
 289         maximum => 128,
 290         default => 0,
 291     },
 292     cpuunits => {
 293         optional => 1,
 294         type => 'integer',
 295         description => "CPU weight for a VM.",
 296         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.",
 297         minimum => 2,
 298         maximum => 262144,
 299         default => 1024,
 300     },
 301     memory => {
 302         optional => 1,
 303         type => 'integer',
 304         description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
 305         minimum => 16,
 306         default => 512,
 307     },
 308     balloon => {
 309         optional => 1,
 310         type => 'integer',
 311         description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
 312         minimum => 0,
 313     },
 314     shares => {
 315         optional => 1,
 316         type => 'integer',
 317         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.",
 318         minimum => 0,
 319         maximum => 50000,
 320         default => 1000,
 321     },
 322     keyboard => {
 323         optional => 1,
 324         type => 'string',
 325         description => "Keybord layout for vnc server. Default is read from the '/etc/pve/datacenter.cfg' configuration file.".
 326                        "It should not be necessary to set it.",
 327         enum => PVE::Tools::kvmkeymaplist(),
 328         default => undef,
 329     },
 330     name => {
 331         optional => 1,
 332         type => 'string', format => 'dns-name',
 333         description => "Set a name for the VM. Only used on the configuration web interface.",
 334     },
 335     scsihw => {
 336         optional => 1,
 337         type => 'string',
 338         description => "SCSI controller model",
 339         enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
 340         default => 'lsi',
 341     },
 342     description => {
 343         optional => 1,
 344         type => 'string',
 345         description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
 346     },
 347     ostype => {
 348         optional => 1,
 349         type => 'string',
 350         enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 l24 l26 solaris)],
 351         description => "Specify guest operating system.",
 352         verbose_description => <<EODESC,
 353 Specify guest operating system. This is used to enable special
 354 optimization/features for specific operating systems:
 355
 356 [horizontal]
 357 other;; unspecified OS
 358 wxp;; Microsoft Windows XP
 359 w2k;; Microsoft Windows 2000
 360 w2k3;; Microsoft Windows 2003
 361 w2k8;; Microsoft Windows 2008
 362 wvista;; Microsoft Windows Vista
 363 win7;; Microsoft Windows 7
 364 win8;; Microsoft Windows 8/2012/2012r2
 365 win10;; Microsoft Windows 10/2016
 366 l24;; Linux 2.4 Kernel
 367 l26;; Linux 2.6/3.X Kernel
 368 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
 369 EODESC
 370     },
 371     boot => {
 372         optional => 1,
 373         type => 'string',
 374         description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
 375         pattern => '[acdn]{1,4}',
 376         default => 'cdn',
 377     },
 378     bootdisk => {
 379         optional => 1,
 380         type => 'string', format => 'pve-qm-bootdisk',
 381         description => "Enable booting from specified disk.",
 382         pattern => '(ide|sata|scsi|virtio)\d+',
 383     },
 384     smp => {
 385         optional => 1,
 386         type => 'integer',
 387         description => "The number of CPUs. Please use option -sockets instead.",
 388         minimum => 1,
 389         default => 1,
 390     },
 391     sockets => {
 392         optional => 1,
 393         type => 'integer',
 394         description => "The number of CPU sockets.",
 395         minimum => 1,
 396         default => 1,
 397     },
 398     cores => {
 399         optional => 1,
 400         type => 'integer',
 401         description => "The number of cores per socket.",
 402         minimum => 1,
 403         default => 1,
 404     },
 405     numa => {
 406         optional => 1,
 407         type => 'boolean',
 408         description => "Enable/disable NUMA.",
 409         default => 0,
 410     },
 411     hugepages => {
 412         optional => 1,
 413         type => 'string',
 414         description => "Enable/disable hugepages memory.",
 415         enum => [qw(any 2 1024)],
 416     },
 417     vcpus => {
 418         optional => 1,
 419         type => 'integer',
 420         description => "Number of hotplugged vcpus.",
 421         minimum => 1,
 422         default => 0,
 423     },
 424     acpi => {
 425         optional => 1,
 426         type => 'boolean',
 427         description => "Enable/disable ACPI.",
 428         default => 1,
 429     },
 430     agent => {
 431         optional => 1,
 432         description => "Enable/disable Qemu GuestAgent and its properties.",
 433         type => 'string',
 434         format => $agent_fmt,
 435     },
 436     kvm => {
 437         optional => 1,
 438         type => 'boolean',
 439         description => "Enable/disable KVM hardware virtualization.",
 440         default => 1,
 441     },
 442     tdf => {
 443         optional => 1,
 444         type => 'boolean',
 445         description => "Enable/disable time drift fix.",
 446         default => 0,
 447     },
 448     localtime => {
 449         optional => 1,
 450         type => 'boolean',
 451         description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
 452     },
 453     freeze => {
 454         optional => 1,
 455         type => 'boolean',
 456         description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
 457     },
 458     vga => {
 459         optional => 1,
 460         type => 'string', format => $vga_fmt,
 461         description => "Configure the VGA hardware.",
 462         verbose_description => "Configure the VGA Hardware. If you want to use ".
 463             "high resolution modes (>= 1280x1024x16) you may need to increase " .
 464             "the vga memory option. Since QEMU 2.9 the default VGA display type " .
 465             "is 'std' for all OS types besides some Windows versions (XP and " .
 466             "older) which use 'cirrus'. The 'qxl' option enables the SPICE " .
 467             "display server. For win* OS you can select how many independent " .
 468             "displays you want, Linux guests can add displays them self.\n".
 469             "You can also run without any graphic card, using a serial device as terminal.",
 470     },
 471     watchdog => {
 472         optional => 1,
 473         type => 'string', format => 'pve-qm-watchdog',
 474         description => "Create a virtual hardware watchdog device.",
 475         verbose_description => "Create a virtual hardware watchdog device. Once enabled" .
 476             " (by a guest action), the watchdog must be periodically polled " .
 477             "by an agent inside the guest or else the watchdog will reset " .
 478             "the guest (or execute the respective action specified)",
 479     },
 480     startdate => {
 481         optional => 1,
 482         type => 'string',
 483         typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
 484         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'.",
 485         pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
 486         default => 'now',
 487     },
 488     startup =>  get_standard_option('pve-startup-order'),
 489     template => {
 490         optional => 1,
 491         type => 'boolean',
 492         description => "Enable/disable Template.",
 493         default => 0,
 494     },
 495     args => {
 496         optional => 1,
 497         type => 'string',
 498         description => "Arbitrary arguments passed to kvm.",
 499         verbose_description => <<EODESCR,
 500 Arbitrary arguments passed to kvm, for example:
 501
 502 args: -no-reboot -no-hpet
 503
 504 NOTE: this option is for experts only.
 505 EODESCR
 506     },
 507     tablet => {
 508         optional => 1,
 509         type => 'boolean',
 510         default => 1,
 511         description => "Enable/disable the USB tablet device.",
 512         verbose_description => "Enable/disable the USB tablet device. This device is " .
 513             "usually needed to allow absolute mouse positioning with VNC. " .
 514             "Else the mouse runs out of sync with normal VNC clients. " .
 515             "If you're running lots of console-only guests on one host, " .
 516             "you may consider disabling this to save some context switches. " .
 517             "This is turned off by default if you use spice (-vga=qxl).",
 518     },
 519     migrate_speed => {
 520         optional => 1,
 521         type => 'integer',
 522         description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
 523         minimum => 0,
 524         default => 0,
 525     },
 526     migrate_downtime => {
 527         optional => 1,
 528         type => 'number',
 529         description => "Set maximum tolerated downtime (in seconds) for migrations.",
 530         minimum => 0,
 531         default => 0.1,
 532     },
 533     cdrom => {
 534         optional => 1,
 535         type => 'string', format => 'pve-qm-ide',
 536         typetext => '<volume>',
 537         description => "This is an alias for option -ide2",
 538     },
 539     cpu => {
 540         optional => 1,
 541         description => "Emulated CPU type.",
 542         type => 'string',
 543         format => $cpu_fmt,
 544     },
 545     parent => get_standard_option('pve-snapshot-name', {
 546         optional => 1,
 547         description => "Parent snapshot name. This is used internally, and should not be modified.",
 548     }),
 549     snaptime => {
 550         optional => 1,
 551         description => "Timestamp for snapshots.",
 552         type => 'integer',
 553         minimum => 0,
 554     },
 555     vmstate => {
 556         optional => 1,
 557         type => 'string', format => 'pve-volume-id',
 558         description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
 559     },
 560     vmstatestorage => get_standard_option('pve-storage-id', {
 561         description => "Default storage for VM state volumes/files.",
 562         optional => 1,
 563     }),
 564     runningmachine => get_standard_option('pve-qemu-machine', {
 565         description => "Specifies the Qemu machine type of the running vm. This is used internally for snapshots.",
 566     }),
 567     machine => get_standard_option('pve-qemu-machine'),
 568     arch => {
 569         description => "Virtual processor architecture. Defaults to the host.",
 570         optional => 1,
 571         type => 'string',
 572         enum => [qw(x86_64 aarch64)],
 573     },
 574     smbios1 => {
 575         description => "Specify SMBIOS type 1 fields.",
 576         type => 'string', format => 'pve-qm-smbios1',
 577         maxLength => 256,
 578         optional => 1,
 579     },
 580     protection => {
 581         optional => 1,
 582         type => 'boolean',
 583         description => "Sets the protection flag of the VM. This will disable the remove VM and remove disk operations.",
 584         default => 0,
 585     },
 586     bios => {
 587         optional => 1,
 588         type => 'string',
 589         enum => [ qw(seabios ovmf) ],
 590         description => "Select BIOS implementation.",
 591         default => 'seabios',
 592     },
 593     vmgenid => {
 594         type => 'string',
 595         pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
 596         format_description => 'UUID',
 597         description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0' to disable explicitly.",
 598         verbose_description => "The VM generation ID (vmgenid) device exposes a".
 599             " 128-bit integer value identifier to the guest OS. This allows to".
 600             " notify the guest operating system when the virtual machine is".
 601             " executed with a different configuration (e.g. snapshot execution".
 602             " or creation from a template). The guest operating system notices".
 603             " the change, and is then able to react as appropriate by marking".
 604             " its copies of distributed databases as dirty, re-initializing its".
 605             " random number generator, etc.\n".
 606             "Note that auto-creation only works when done throug API/CLI create".
 607             " or update methods, but not when manually editing the config file.",
 608         default => "1 (autogenerated)",
 609         optional => 1,
 610     },
 611 };
 612
 613 my $confdesc_cloudinit = {
 614     citype => {
 615         optional => 1,
 616         type => 'string',
 617         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.',
 618         enum => ['configdrive2', 'nocloud'],
 619     },
 620     ciuser => {
 621         optional => 1,
 622         type => 'string',
 623         description => "cloud-init: User name to change ssh keys and password for instead of the image's configured default user.",
 624     },
 625     cipassword => {
 626         optional => 1,
 627         type => 'string',
 628         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.',
 629     },
 630     searchdomain => {
 631         optional => 1,
 632         type => 'string',
 633         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.",
 634     },
 635     nameserver => {
 636         optional => 1,
 637         type => 'string', format => 'address-list',
 638         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.",
 639     },
 640     sshkeys => {
 641         optional => 1,
 642         type => 'string',
 643         format => 'urlencoded',
 644         description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
 645     },
 646 };
 647
 648 # what about other qemu settings ?
 649 #cpu => 'string',
 650 #machine => 'string',
 651 #fda => 'file',
 652 #fdb => 'file',
 653 #mtdblock => 'file',
 654 #sd => 'file',
 655 #pflash => 'file',
 656 #snapshot => 'bool',
 657 #bootp => 'file',
 658 ##tftp => 'dir',
 659 ##smb => 'dir',
 660 #kernel => 'file',
 661 #append => 'string',
 662 #initrd => 'file',
 663 ##soundhw => 'string',
 664
 665 while (my ($k, $v) = each %$confdesc) {
 666     PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
 667 }
 668
 669 my $MAX_IDE_DISKS = 4;
 670 my $MAX_SCSI_DISKS = 14;
 671 my $MAX_VIRTIO_DISKS = 16;
 672 my $MAX_SATA_DISKS = 6;
 673 my $MAX_USB_DEVICES = 5;
 674 my $MAX_NETS = 32;
 675 my $MAX_UNUSED_DISKS = 256;
 676 my $MAX_HOSTPCI_DEVICES = 4;
 677 my $MAX_SERIAL_PORTS = 4;
 678 my $MAX_PARALLEL_PORTS = 3;
 679 my $MAX_NUMA = 8;
 680
 681 my $numa_fmt = {
 682     cpus => {
 683         type => "string",
 684         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 685         description => "CPUs accessing this NUMA node.",
 686         format_description => "id[-id];...",
 687     },
 688     memory => {
 689         type => "number",
 690         description => "Amount of memory this NUMA node provides.",
 691         optional => 1,
 692     },
 693     hostnodes => {
 694         type => "string",
 695         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 696         description => "Host NUMA nodes to use.",
 697         format_description => "id[-id];...",
 698         optional => 1,
 699     },
 700     policy => {
 701         type => 'string',
 702         enum => [qw(preferred bind interleave)],
 703         description => "NUMA allocation policy.",
 704         optional => 1,
 705     },
 706 };
 707 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
 708 my $numadesc = {
 709     optional => 1,
 710     type => 'string', format => $numa_fmt,
 711     description => "NUMA topology.",
 712 };
 713 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
 714
 715 for (my $i = 0; $i < $MAX_NUMA; $i++)  {
 716     $confdesc->{"numa$i"} = $numadesc;
 717 }
 718
 719 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000',  'pcnet',  'virtio',
 720                       'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
 721                       'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
 722 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
 723
 724 my $net_fmt_bridge_descr = <<__EOD__;
 725 Bridge to attach the network device to. The Proxmox VE standard bridge
 726 is called 'vmbr0'.
 727
 728 If you do not specify a bridge, we create a kvm user (NATed) network
 729 device, which provides DHCP and DNS services. The following addresses
 730 are used:
 731
 732  10.0.2.2   Gateway
 733  10.0.2.3   DNS Server
 734  10.0.2.4   SMB Server
 735
 736 The DHCP server assign addresses to the guest starting from 10.0.2.15.
 737 __EOD__
 738
 739 my $net_fmt = {
 740     macaddr => {
 741         type => 'string',
 742         pattern => qr/[0-9a-f]{2}(?::[0-9a-f]{2}){5}/i,
 743         description => "MAC address. That address must be unique withing your network. This is automatically generated if not specified.",
 744         format_description => "XX:XX:XX:XX:XX:XX",
 745         optional => 1,
 746     },
 747     model => {
 748         type => 'string',
 749         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'.",
 750         enum => $nic_model_list,
 751         default_key => 1,
 752     },
 753     (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
 754     bridge => {
 755         type => 'string',
 756         description => $net_fmt_bridge_descr,
 757         format_description => 'bridge',
 758         optional => 1,
 759     },
 760     queues => {
 761         type => 'integer',
 762         minimum => 0, maximum => 16,
 763         description => 'Number of packet queues to be used on the device.',
 764         optional => 1,
 765     },
 766     rate => {
 767         type => 'number',
 768         minimum => 0,
 769         description => "Rate limit in mbps (megabytes per second) as floating point number.",
 770         optional => 1,
 771     },
 772     tag => {
 773         type => 'integer',
 774         minimum => 1, maximum => 4094,
 775         description => 'VLAN tag to apply to packets on this interface.',
 776         optional => 1,
 777     },
 778     trunks => {
 779         type => 'string',
 780         pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
 781         description => 'VLAN trunks to pass through this interface.',
 782         format_description => 'vlanid[;vlanid...]',
 783         optional => 1,
 784     },
 785     firewall => {
 786         type => 'boolean',
 787         description => 'Whether this interface should be protected by the firewall.',
 788         optional => 1,
 789     },
 790     link_down => {
 791         type => 'boolean',
 792         description => 'Whether this interface should be disconnected (like pulling the plug).',
 793         optional => 1,
 794     },
 795 };
 796
 797 my $netdesc = {
 798     optional => 1,
 799     type => 'string', format => $net_fmt,
 800     description => "Specify network devices.",
 801 };
 802
 803 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
 804
 805 my $ipconfig_fmt = {
 806     ip => {
 807         type => 'string',
 808         format => 'pve-ipv4-config',
 809         format_description => 'IPv4Format/CIDR',
 810         description => 'IPv4 address in CIDR format.',
 811         optional => 1,
 812         default => 'dhcp',
 813     },
 814     gw => {
 815         type => 'string',
 816         format => 'ipv4',
 817         format_description => 'GatewayIPv4',
 818         description => 'Default gateway for IPv4 traffic.',
 819         optional => 1,
 820         requires => 'ip',
 821     },
 822     ip6 => {
 823         type => 'string',
 824         format => 'pve-ipv6-config',
 825         format_description => 'IPv6Format/CIDR',
 826         description => 'IPv6 address in CIDR format.',
 827         optional => 1,
 828         default => 'dhcp',
 829     },
 830     gw6 => {
 831         type => 'string',
 832         format => 'ipv6',
 833         format_description => 'GatewayIPv6',
 834         description => 'Default gateway for IPv6 traffic.',
 835         optional => 1,
 836         requires => 'ip6',
 837     },
 838 };
 839 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
 840 my $ipconfigdesc = {
 841     optional => 1,
 842     type => 'string', format => 'pve-qm-ipconfig',
 843     description => <<'EODESCR',
 844 cloud-init: Specify IP addresses and gateways for the corresponding interface.
 845
 846 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
 847
 848 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit gateway should be provided.
 849 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
 850
 851 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using dhcp on IPv4.
 852 EODESCR
 853 };
 854 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
 855
 856 for (my $i = 0; $i < $MAX_NETS; $i++)  {
 857     $confdesc->{"net$i"} = $netdesc;
 858     $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
 859 }
 860
 861 foreach my $key (keys %$confdesc_cloudinit) {
 862     $confdesc->{$key} = $confdesc_cloudinit->{$key};
 863 }
 864
 865 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
 866 sub verify_volume_id_or_qm_path {
 867     my ($volid, $noerr) = @_;
 868
 869     if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
 870         return $volid;
 871     }
 872
 873     # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
 874     $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
 875     if ($@) {
 876         return undef if $noerr;
 877         die $@;
 878     }
 879     return $volid;
 880 }
 881
 882 my $drivename_hash;
 883
 884 my %drivedesc_base = (
 885     volume => { alias => 'file' },
 886     file => {
 887         type => 'string',
 888         format => 'pve-volume-id-or-qm-path',
 889         default_key => 1,
 890         format_description => 'volume',
 891         description => "The drive's backing volume.",
 892     },
 893     media => {
 894         type => 'string',
 895         enum => [qw(cdrom disk)],
 896         description => "The drive's media type.",
 897         default => 'disk',
 898         optional => 1
 899     },
 900     cyls => {
 901         type => 'integer',
 902         description => "Force the drive's physical geometry to have a specific cylinder count.",
 903         optional => 1
 904     },
 905     heads => {
 906         type => 'integer',
 907         description => "Force the drive's physical geometry to have a specific head count.",
 908         optional => 1
 909     },
 910     secs => {
 911         type => 'integer',
 912         description => "Force the drive's physical geometry to have a specific sector count.",
 913         optional => 1
 914     },
 915     trans => {
 916         type => 'string',
 917         enum => [qw(none lba auto)],
 918         description => "Force disk geometry bios translation mode.",
 919         optional => 1,
 920     },
 921     snapshot => {
 922         type => 'boolean',
 923         description => "Controls qemu's snapshot mode feature."
 924             . " If activated, changes made to the disk are temporary and will"
 925             . " be discarded when the VM is shutdown.",
 926         optional => 1,
 927     },
 928     cache => {
 929         type => 'string',
 930         enum => [qw(none writethrough writeback unsafe directsync)],
 931         description => "The drive's cache mode",
 932         optional => 1,
 933     },
 934     format => get_standard_option('pve-qm-image-format'),
 935     size => {
 936         type => 'string',
 937         format => 'disk-size',
 938         format_description => 'DiskSize',
 939         description => "Disk size. This is purely informational and has no effect.",
 940         optional => 1,
 941     },
 942     backup => {
 943         type => 'boolean',
 944         description => "Whether the drive should be included when making backups.",
 945         optional => 1,
 946     },
 947     replicate => {
 948         type => 'boolean',
 949         description => 'Whether the drive should considered for replication jobs.',
 950         optional => 1,
 951         default => 1,
 952     },
 953     rerror => {
 954         type => 'string',
 955         enum => [qw(ignore report stop)],
 956         description => 'Read error action.',
 957         optional => 1,
 958     },
 959     werror => {
 960         type => 'string',
 961         enum => [qw(enospc ignore report stop)],
 962         description => 'Write error action.',
 963         optional => 1,
 964     },
 965     aio => {
 966         type => 'string',
 967         enum => [qw(native threads)],
 968         description => 'AIO type to use.',
 969         optional => 1,
 970     },
 971     discard => {
 972         type => 'string',
 973         enum => [qw(ignore on)],
 974         description => 'Controls whether to pass discard/trim requests to the underlying storage.',
 975         optional => 1,
 976     },
 977     detect_zeroes => {
 978         type => 'boolean',
 979         description => 'Controls whether to detect and try to optimize writes of zeroes.',
 980         optional => 1,
 981     },
 982     serial => {
 983         type => 'string',
 984         format => 'urlencoded',
 985         format_description => 'serial',
 986         maxLength => 20*3, # *3 since it's %xx url enoded
 987         description => "The drive's reported serial number, url-encoded, up to 20 bytes long.",
 988         optional => 1,
 989     },
 990     shared => {
 991         type => 'boolean',
 992         description => 'Mark this locally-managed volume as available on all nodes',
 993         verbose_description => "Mark this locally-managed volume as available on all nodes.\n\nWARNING: This option does not share the volume automatically, it assumes it is shared already!",
 994         optional => 1,
 995         default => 0,
 996     }
 997 );
 998
 999 my %iothread_fmt = ( iothread => {
1000         type => 'boolean',
1001         description => "Whether to use iothreads for this drive",
1002         optional => 1,
1003 });
1004
1005 my %model_fmt = (
1006     model => {
1007         type => 'string',
1008         format => 'urlencoded',
1009         format_description => 'model',
1010         maxLength => 40*3, # *3 since it's %xx url enoded
1011         description => "The drive's reported model name, url-encoded, up to 40 bytes long.",
1012         optional => 1,
1013     },
1014 );
1015
1016 my %queues_fmt = (
1017     queues => {
1018         type => 'integer',
1019         description => "Number of queues.",
1020         minimum => 2,
1021         optional => 1
1022     }
1023 );
1024
1025 my %scsiblock_fmt = (
1026     scsiblock => {
1027         type => 'boolean',
1028         description => "whether to use scsi-block for full passthrough of host block device\n\nWARNING: can lead to I/O errors in combination with low memory or high memory fragmentation on host",
1029         optional => 1,
1030         default => 0,
1031     },
1032 );
1033
1034 my %ssd_fmt = (
1035     ssd => {
1036         type => 'boolean',
1037         description => "Whether to expose this drive as an SSD, rather than a rotational hard disk.",
1038         optional => 1,
1039     },
1040 );
1041
1042 my $add_throttle_desc = sub {
1043     my ($key, $type, $what, $unit, $longunit, $minimum) = @_;
1044     my $d = {
1045         type => $type,
1046         format_description => $unit,
1047         description => "Maximum $what in $longunit.",
1048         optional => 1,
1049     };
1050     $d->{minimum} = $minimum if defined($minimum);
1051     $drivedesc_base{$key} = $d;
1052 };
1053 # throughput: (leaky bucket)
1054 $add_throttle_desc->('bps',     'integer', 'r/w speed',   'bps',  'bytes per second');
1055 $add_throttle_desc->('bps_rd',  'integer', 'read speed',  'bps',  'bytes per second');
1056 $add_throttle_desc->('bps_wr',  'integer', 'write speed', 'bps',  'bytes per second');
1057 $add_throttle_desc->('mbps',    'number',  'r/w speed',   'mbps', 'megabytes per second');
1058 $add_throttle_desc->('mbps_rd', 'number',  'read speed',  'mbps', 'megabytes per second');
1059 $add_throttle_desc->('mbps_wr', 'number',  'write speed', 'mbps', 'megabytes per second');
1060 $add_throttle_desc->('iops',    'integer', 'r/w I/O',     'iops', 'operations per second');
1061 $add_throttle_desc->('iops_rd', 'integer', 'read I/O',    'iops', 'operations per second');
1062 $add_throttle_desc->('iops_wr', 'integer', 'write I/O',   'iops', 'operations per second');
1063
1064 # pools: (pool of IO before throttling starts taking effect)
1065 $add_throttle_desc->('mbps_max',    'number',  'unthrottled r/w pool',       'mbps', 'megabytes per second');
1066 $add_throttle_desc->('mbps_rd_max', 'number',  'unthrottled read pool',      'mbps', 'megabytes per second');
1067 $add_throttle_desc->('mbps_wr_max', 'number',  'unthrottled write pool',     'mbps', 'megabytes per second');
1068 $add_throttle_desc->('iops_max',    'integer', 'unthrottled r/w I/O pool',   'iops', 'operations per second');
1069 $add_throttle_desc->('iops_rd_max', 'integer', 'unthrottled read I/O pool',  'iops', 'operations per second');
1070 $add_throttle_desc->('iops_wr_max', 'integer', 'unthrottled write I/O pool', 'iops', 'operations per second');
1071
1072 # burst lengths
1073 $add_throttle_desc->('bps_max_length',     'integer', 'length of I/O bursts',       'seconds', 'seconds', 1);
1074 $add_throttle_desc->('bps_rd_max_length',  'integer', 'length of read I/O bursts',  'seconds', 'seconds', 1);
1075 $add_throttle_desc->('bps_wr_max_length',  'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1076 $add_throttle_desc->('iops_max_length',    'integer', 'length of I/O bursts',       'seconds', 'seconds', 1);
1077 $add_throttle_desc->('iops_rd_max_length', 'integer', 'length of read I/O bursts',  'seconds', 'seconds', 1);
1078 $add_throttle_desc->('iops_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1079
1080 # legacy support
1081 $drivedesc_base{'bps_rd_length'} = { alias => 'bps_rd_max_length' };
1082 $drivedesc_base{'bps_wr_length'} = { alias => 'bps_wr_max_length' };
1083 $drivedesc_base{'iops_rd_length'} = { alias => 'iops_rd_max_length' };
1084 $drivedesc_base{'iops_wr_length'} = { alias => 'iops_wr_max_length' };
1085
1086 my $ide_fmt = {
1087     %drivedesc_base,
1088     %model_fmt,
1089     %ssd_fmt,
1090 };
1091 PVE::JSONSchema::register_format("pve-qm-ide", $ide_fmt);
1092
1093 my $idedesc = {
1094     optional => 1,
1095     type => 'string', format => $ide_fmt,
1096     description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
1097 };
1098 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
1099
1100 my $scsi_fmt = {
1101     %drivedesc_base,
1102     %iothread_fmt,
1103     %queues_fmt,
1104     %scsiblock_fmt,
1105     %ssd_fmt,
1106 };
1107 my $scsidesc = {
1108     optional => 1,
1109     type => 'string', format => $scsi_fmt,
1110     description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
1111 };
1112 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
1113
1114 my $sata_fmt = {
1115     %drivedesc_base,
1116     %ssd_fmt,
1117 };
1118 my $satadesc = {
1119     optional => 1,
1120     type => 'string', format => $sata_fmt,
1121     description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
1122 };
1123 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
1124
1125 my $virtio_fmt = {
1126     %drivedesc_base,
1127     %iothread_fmt,
1128 };
1129 my $virtiodesc = {
1130     optional => 1,
1131     type => 'string', format => $virtio_fmt,
1132     description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
1133 };
1134 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
1135
1136 my $alldrive_fmt = {
1137     %drivedesc_base,
1138     %iothread_fmt,
1139     %model_fmt,
1140     %queues_fmt,
1141     %scsiblock_fmt,
1142     %ssd_fmt,
1143 };
1144
1145 my $efidisk_fmt = {
1146     volume => { alias => 'file' },
1147     file => {
1148         type => 'string',
1149         format => 'pve-volume-id-or-qm-path',
1150         default_key => 1,
1151         format_description => 'volume',
1152         description => "The drive's backing volume.",
1153     },
1154     format => get_standard_option('pve-qm-image-format'),
1155     size => {
1156         type => 'string',
1157         format => 'disk-size',
1158         format_description => 'DiskSize',
1159         description => "Disk size. This is purely informational and has no effect.",
1160         optional => 1,
1161     },
1162 };
1163
1164 my $efidisk_desc = {
1165     optional => 1,
1166     type => 'string', format => $efidisk_fmt,
1167     description => "Configure a Disk for storing EFI vars",
1168 };
1169
1170 PVE::JSONSchema::register_standard_option("pve-qm-efidisk", $efidisk_desc);
1171
1172 my $usb_fmt = {
1173     host => {
1174         default_key => 1,
1175         type => 'string', format => 'pve-qm-usb-device',
1176         format_description => 'HOSTUSBDEVICE|spice',
1177         description => <<EODESCR,
1178 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1179
1180  'bus-port(.port)*' (decimal numbers) or
1181  'vendor_id:product_id' (hexadeciaml numbers) or
1182  'spice'
1183
1184 You can use the 'lsusb -t' command to list existing usb devices.
1185
1186 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1187
1188 The value 'spice' can be used to add a usb redirection devices for spice.
1189 EODESCR
1190     },
1191     usb3 => {
1192         optional => 1,
1193         type => 'boolean',
1194         description => "Specifies whether if given host option is a USB3 device or port (this does currently not work reliably with spice redirection and is then ignored).",
1195         default => 0,
1196     },
1197 };
1198
1199 my $usbdesc = {
1200     optional => 1,
1201     type => 'string', format => $usb_fmt,
1202     description => "Configure an USB device (n is 0 to 4).",
1203 };
1204 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1205
1206 # NOTE: the match-groups of this regex are used in parse_hostpci
1207 my $PCIRE = qr/([a-f0-9]{2}:[a-f0-9]{2})(?:\.([a-f0-9]))?/;
1208 my $hostpci_fmt = {
1209     host => {
1210         default_key => 1,
1211         type => 'string',
1212         pattern => qr/$PCIRE(;$PCIRE)*/,
1213         format_description => 'HOSTPCIID[;HOSTPCIID2...]',
1214         description => <<EODESCR,
1215 Host PCI device pass through. The PCI ID of a host's PCI device or a list
1216 of PCI virtual functions of the host. HOSTPCIID syntax is:
1217
1218 'bus:dev.func' (hexadecimal numbers)
1219
1220 You can us the 'lspci' command to list existing PCI devices.
1221 EODESCR
1222     },
1223     rombar => {
1224         type => 'boolean',
1225         description =>  "Specify whether or not the device's ROM will be visible in the guest's memory map.",
1226         optional => 1,
1227         default => 1,
1228     },
1229     romfile => {
1230         type => 'string',
1231         pattern => '[^,;]+',
1232         format_description => 'string',
1233         description => "Custom pci device rom filename (must be located in /usr/share/kvm/).",
1234         optional => 1,
1235     },
1236     pcie => {
1237         type => 'boolean',
1238         description =>  "Choose the PCI-express bus (needs the 'q35' machine model).",
1239         optional => 1,
1240         default => 0,
1241     },
1242     'x-vga' => {
1243         type => 'boolean',
1244         description =>  "Enable vfio-vga device support.",
1245         optional => 1,
1246         default => 0,
1247     },
1248 };
1249 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
1250
1251 my $hostpcidesc = {
1252         optional => 1,
1253         type => 'string', format => 'pve-qm-hostpci',
1254         description => "Map host PCI devices into guest.",
1255         verbose_description =>  <<EODESCR,
1256 Map host PCI devices into guest.
1257
1258 NOTE: This option allows direct access to host hardware. So it is no longer
1259 possible to migrate such machines - use with special care.
1260
1261 CAUTION: Experimental! User reported problems with this option.
1262 EODESCR
1263 };
1264 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
1265
1266 my $serialdesc = {
1267         optional => 1,
1268         type => 'string',
1269         pattern => '(/dev/.+|socket)',
1270         description =>  "Create a serial device inside the VM (n is 0 to 3)",
1271         verbose_description =>  <<EODESCR,
1272 Create a serial device inside the VM (n is 0 to 3), and pass through a
1273 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1274 host side (use 'qm terminal' to open a terminal connection).
1275
1276 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
1277
1278 CAUTION: Experimental! User reported problems with this option.
1279 EODESCR
1280 };
1281
1282 my $paralleldesc= {
1283         optional => 1,
1284         type => 'string',
1285         pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1286         description =>  "Map host parallel devices (n is 0 to 2).",
1287         verbose_description =>  <<EODESCR,
1288 Map host parallel devices (n is 0 to 2).
1289
1290 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1291
1292 CAUTION: Experimental! User reported problems with this option.
1293 EODESCR
1294 };
1295
1296 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
1297     $confdesc->{"parallel$i"} = $paralleldesc;
1298 }
1299
1300 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
1301     $confdesc->{"serial$i"} = $serialdesc;
1302 }
1303
1304 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
1305     $confdesc->{"hostpci$i"} = $hostpcidesc;
1306 }
1307
1308 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++)  {
1309     $drivename_hash->{"ide$i"} = 1;
1310     $confdesc->{"ide$i"} = $idedesc;
1311 }
1312
1313 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++)  {
1314     $drivename_hash->{"sata$i"} = 1;
1315     $confdesc->{"sata$i"} = $satadesc;
1316 }
1317
1318 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++)  {
1319     $drivename_hash->{"scsi$i"} = 1;
1320     $confdesc->{"scsi$i"} = $scsidesc ;
1321 }
1322
1323 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++)  {
1324     $drivename_hash->{"virtio$i"} = 1;
1325     $confdesc->{"virtio$i"} = $virtiodesc;
1326 }
1327
1328 $drivename_hash->{efidisk0} = 1;
1329 $confdesc->{efidisk0} = $efidisk_desc;
1330
1331 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++)  {
1332     $confdesc->{"usb$i"} = $usbdesc;
1333 }
1334
1335 my $unuseddesc = {
1336     optional => 1,
1337     type => 'string', format => 'pve-volume-id',
1338     description => "Reference to unused volumes. This is used internally, and should not be modified manually.",
1339 };
1340
1341 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++)  {
1342     $confdesc->{"unused$i"} = $unuseddesc;
1343 }
1344
1345 my $kvm_api_version = 0;
1346
1347 sub kvm_version {
1348     return $kvm_api_version if $kvm_api_version;
1349
1350     open my $fh, '<', '/dev/kvm'
1351         or return undef;
1352
1353     # 0xae00 => KVM_GET_API_VERSION
1354     $kvm_api_version = ioctl($fh, 0xae00, 0);
1355
1356     return $kvm_api_version;
1357 }
1358
1359 my $kvm_user_version;
1360
1361 sub kvm_user_version {
1362
1363     return $kvm_user_version if $kvm_user_version;
1364
1365     $kvm_user_version = 'unknown';
1366
1367     my $code = sub {
1368         my $line = shift;
1369         if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1370             $kvm_user_version = $2;
1371         }
1372     };
1373
1374     eval { run_command("kvm -version", outfunc => $code); };
1375     warn $@ if $@;
1376
1377     return $kvm_user_version;
1378
1379 }
1380
1381 my $kernel_has_vhost_net = -c '/dev/vhost-net';
1382
1383 sub valid_drive_names {
1384     # order is important - used to autoselect boot disk
1385     return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
1386             (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
1387             (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
1388             (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))),
1389             'efidisk0');
1390 }
1391
1392 sub is_valid_drivename {
1393     my $dev = shift;
1394
1395     return defined($drivename_hash->{$dev});
1396 }
1397
1398 sub option_exists {
1399     my $key = shift;
1400     return defined($confdesc->{$key});
1401 }
1402
1403 sub nic_models {
1404     return $nic_model_list;
1405 }
1406
1407 sub os_list_description {
1408
1409     return {
1410         other => 'Other',
1411         wxp => 'Windows XP',
1412         w2k => 'Windows 2000',
1413         w2k3 =>, 'Windows 2003',
1414         w2k8 => 'Windows 2008',
1415         wvista => 'Windows Vista',
1416         win7 => 'Windows 7',
1417         win8 => 'Windows 8/2012',
1418         win10 => 'Windows 10/2016',
1419         l24 => 'Linux 2.4',
1420         l26 => 'Linux 2.6',
1421     };
1422 }
1423
1424 my $cdrom_path;
1425
1426 sub get_cdrom_path {
1427
1428     return  $cdrom_path if $cdrom_path;
1429
1430     return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1431     return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1432     return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1433 }
1434
1435 sub get_iso_path {
1436     my ($storecfg, $vmid, $cdrom) = @_;
1437
1438     if ($cdrom eq 'cdrom') {
1439         return get_cdrom_path();
1440     } elsif ($cdrom eq 'none') {
1441         return '';
1442     } elsif ($cdrom =~ m|^/|) {
1443         return $cdrom;
1444     } else {
1445         return PVE::Storage::path($storecfg, $cdrom);
1446     }
1447 }
1448
1449 # try to convert old style file names to volume IDs
1450 sub filename_to_volume_id {
1451     my ($vmid, $file, $media) = @_;
1452
1453      if (!($file eq 'none' || $file eq 'cdrom' ||
1454           $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1455
1456         return undef if $file =~ m|/|;
1457
1458         if ($media && $media eq 'cdrom') {
1459             $file = "local:iso/$file";
1460         } else {
1461             $file = "local:$vmid/$file";
1462         }
1463     }
1464
1465     return $file;
1466 }
1467
1468 sub verify_media_type {
1469     my ($opt, $vtype, $media) = @_;
1470
1471     return if !$media;
1472
1473     my $etype;
1474     if ($media eq 'disk') {
1475         $etype = 'images';
1476     } elsif ($media eq 'cdrom') {
1477         $etype = 'iso';
1478     } else {
1479         die "internal error";
1480     }
1481
1482     return if ($vtype eq $etype);
1483
1484     raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1485 }
1486
1487 sub cleanup_drive_path {
1488     my ($opt, $storecfg, $drive) = @_;
1489
1490     # try to convert filesystem paths to volume IDs
1491
1492     if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1493         ($drive->{file} !~ m|^/dev/.+|) &&
1494         ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1495         ($drive->{file} !~ m/^\d+$/)) {
1496         my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1497         raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1498         $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1499         verify_media_type($opt, $vtype, $drive->{media});
1500         $drive->{file} = $volid;
1501     }
1502
1503     $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1504 }
1505
1506 sub parse_hotplug_features {
1507     my ($data) = @_;
1508
1509     my $res = {};
1510
1511     return $res if $data eq '0';
1512
1513     $data = $confdesc->{hotplug}->{default} if $data eq '1';
1514
1515     foreach my $feature (PVE::Tools::split_list($data)) {
1516         if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1517             $res->{$1} = 1;
1518         } else {
1519             die "invalid hotplug feature '$feature'\n";
1520         }
1521     }
1522     return $res;
1523 }
1524
1525 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1526 sub pve_verify_hotplug_features {
1527     my ($value, $noerr) = @_;
1528
1529     return $value if parse_hotplug_features($value);
1530
1531     return undef if $noerr;
1532
1533     die "unable to parse hotplug option\n";
1534 }
1535
1536 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
1537 #        [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
1538 #        [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
1539 #        [,aio=native|threads][,discard=ignore|on][,detect_zeroes=on|off]
1540 #        [,iothread=on][,serial=serial][,model=model]
1541
1542 sub parse_drive {
1543     my ($key, $data) = @_;
1544
1545     my ($interface, $index);
1546
1547     if ($key =~ m/^([^\d]+)(\d+)$/) {
1548         $interface = $1;
1549         $index = $2;
1550     } else {
1551         return undef;
1552     }
1553
1554     my $desc = $key =~ /^unused\d+$/ ? $alldrive_fmt
1555                                      : $confdesc->{$key}->{format};
1556     if (!$desc) {
1557         warn "invalid drive key: $key\n";
1558         return undef;
1559     }
1560     my $res = eval { PVE::JSONSchema::parse_property_string($desc, $data) };
1561     return undef if !$res;
1562     $res->{interface} = $interface;
1563     $res->{index} = $index;
1564
1565     my $error = 0;
1566     foreach my $opt (qw(bps bps_rd bps_wr)) {
1567         if (my $bps = defined(delete $res->{$opt})) {
1568             if (defined($res->{"m$opt"})) {
1569                 warn "both $opt and m$opt specified\n";
1570                 ++$error;
1571                 next;
1572             }
1573             $res->{"m$opt"} = sprintf("%.3f", $bps / (1024*1024.0));
1574         }
1575     }
1576
1577     # can't use the schema's 'requires' because of the mbps* => bps* "transforming aliases"
1578     for my $requirement (
1579         [mbps_max => 'mbps'],
1580         [mbps_rd_max => 'mbps_rd'],
1581         [mbps_wr_max => 'mbps_wr'],
1582         [miops_max => 'miops'],
1583         [miops_rd_max => 'miops_rd'],
1584         [miops_wr_max => 'miops_wr'],
1585         [bps_max_length => 'mbps_max'],
1586         [bps_rd_max_length => 'mbps_rd_max'],
1587         [bps_wr_max_length => 'mbps_wr_max'],
1588         [iops_max_length => 'iops_max'],
1589         [iops_rd_max_length => 'iops_rd_max'],
1590         [iops_wr_max_length => 'iops_wr_max']) {
1591         my ($option, $requires) = @$requirement;
1592         if ($res->{$option} && !$res->{$requires}) {
1593             warn "$option requires $requires\n";
1594             ++$error;
1595         }
1596     }
1597
1598     return undef if $error;
1599
1600     return undef if $res->{mbps_rd} && $res->{mbps};
1601     return undef if $res->{mbps_wr} && $res->{mbps};
1602     return undef if $res->{iops_rd} && $res->{iops};
1603     return undef if $res->{iops_wr} && $res->{iops};
1604
1605     if ($res->{media} && ($res->{media} eq 'cdrom')) {
1606         return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1607         return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1608         return undef if $res->{interface} eq 'virtio';
1609     }
1610
1611     if (my $size = $res->{size}) {
1612         return undef if !defined($res->{size} = PVE::JSONSchema::parse_size($size));
1613     }
1614
1615     return $res;
1616 }
1617
1618 sub print_drive {
1619     my ($vmid, $drive) = @_;
1620     my $data = { %$drive };
1621     delete $data->{$_} for qw(index interface);
1622     return PVE::JSONSchema::print_property_string($data, $alldrive_fmt);
1623 }
1624
1625 sub scsi_inquiry {
1626     my($fh, $noerr) = @_;
1627
1628     my $SG_IO = 0x2285;
1629     my $SG_GET_VERSION_NUM = 0x2282;
1630
1631     my $versionbuf = "\x00" x 8;
1632     my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1633     if (!$ret) {
1634         die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1635         return undef;
1636     }
1637     my $version = unpack("I", $versionbuf);
1638     if ($version < 30000) {
1639         die "scsi generic interface too old\n"  if !$noerr;
1640         return undef;
1641     }
1642
1643     my $buf = "\x00" x 36;
1644     my $sensebuf = "\x00" x 8;
1645     my $cmd = pack("C x3 C x1", 0x12, 36);
1646
1647     # see /usr/include/scsi/sg.h
1648     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";
1649
1650     my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1651                       length($sensebuf), 0, length($buf), $buf,
1652                       $cmd, $sensebuf, 6000);
1653
1654     $ret = ioctl($fh, $SG_IO, $packet);
1655     if (!$ret) {
1656         die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1657         return undef;
1658     }
1659
1660     my @res = unpack($sg_io_hdr_t, $packet);
1661     if ($res[17] || $res[18]) {
1662         die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1663         return undef;
1664     }
1665
1666     my $res = {};
1667     (my $byte0, my $byte1, $res->{vendor},
1668      $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1669
1670     $res->{removable} = $byte1 & 128 ? 1 : 0;
1671     $res->{type} = $byte0 & 31;
1672
1673     return $res;
1674 }
1675
1676 sub path_is_scsi {
1677     my ($path) = @_;
1678
1679     my $fh = IO::File->new("+<$path") || return undef;
1680     my $res = scsi_inquiry($fh, 1);
1681     close($fh);
1682
1683     return $res;
1684 }
1685
1686 sub machine_type_is_q35 {
1687     my ($conf) = @_;
1688
1689     return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1690 }
1691
1692 sub print_tabletdevice_full {
1693     my ($conf) = @_;
1694
1695     my $q35 = machine_type_is_q35($conf);
1696
1697     # we use uhci for old VMs because tablet driver was buggy in older qemu
1698     my $usbbus = $q35 ? "ehci" : "uhci";
1699
1700     return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1701 }
1702
1703 sub print_drivedevice_full {
1704     my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1705
1706     my $device = '';
1707     my $maxdev = 0;
1708
1709     if ($drive->{interface} eq 'virtio') {
1710         my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1711         $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1712         $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1713     } elsif ($drive->{interface} eq 'scsi') {
1714
1715         my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1716         my $unit = $drive->{index} % $maxdev;
1717         my $devicetype = 'hd';
1718         my $path = '';
1719         if (drive_is_cdrom($drive)) {
1720             $devicetype = 'cd';
1721         } else {
1722             if ($drive->{file} =~ m|^/|) {
1723                 $path = $drive->{file};
1724                 if (my $info = path_is_scsi($path)) {
1725                     if ($info->{type} == 0 && $drive->{scsiblock}) {
1726                         $devicetype = 'block';
1727                     } elsif ($info->{type} == 1) { # tape
1728                         $devicetype = 'generic';
1729                     }
1730                 }
1731             } else {
1732                  $path = PVE::Storage::path($storecfg, $drive->{file});
1733             }
1734
1735             if($path =~ m/^iscsi\:\/\//){
1736                 $devicetype = 'generic';
1737             }
1738         }
1739
1740         if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1741             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1742         } else {
1743             $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1744         }
1745
1746         if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1747             $device .= ",rotation_rate=1";
1748         }
1749
1750     } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1751         my $maxdev = ($drive->{interface} eq 'sata') ? $MAX_SATA_DISKS : 2;
1752         my $controller = int($drive->{index} / $maxdev);
1753         my $unit = $drive->{index} % $maxdev;
1754         my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1755
1756         $device = "ide-$devicetype";
1757         if ($drive->{interface} eq 'ide') {
1758             $device .= ",bus=ide.$controller,unit=$unit";
1759         } else {
1760             $device .= ",bus=ahci$controller.$unit";
1761         }
1762         $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1763
1764         if ($devicetype eq 'hd') {
1765             if (my $model = $drive->{model}) {
1766                 $model = URI::Escape::uri_unescape($model);
1767                 $device .= ",model=$model";
1768             }
1769             if ($drive->{ssd}) {
1770                 $device .= ",rotation_rate=1";
1771             }
1772         }
1773     } elsif ($drive->{interface} eq 'usb') {
1774         die "implement me";
1775         #  -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1776     } else {
1777         die "unsupported interface type";
1778     }
1779
1780     $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1781
1782     if (my $serial = $drive->{serial}) {
1783         $serial = URI::Escape::uri_unescape($serial);
1784         $device .= ",serial=$serial";
1785     }
1786
1787
1788     return $device;
1789 }
1790
1791 sub get_initiator_name {
1792     my $initiator;
1793
1794     my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1795     while (defined(my $line = <$fh>)) {
1796         next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1797         $initiator = $1;
1798         last;
1799     }
1800     $fh->close();
1801
1802     return $initiator;
1803 }
1804
1805 sub print_drive_full {
1806     my ($storecfg, $vmid, $drive) = @_;
1807
1808     my $path;
1809     my $volid = $drive->{file};
1810     my $format;
1811
1812     if (drive_is_cdrom($drive)) {
1813         $path = get_iso_path($storecfg, $vmid, $volid);
1814     } else {
1815         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1816         if ($storeid) {
1817             $path = PVE::Storage::path($storecfg, $volid);
1818             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1819             $format = qemu_img_format($scfg, $volname);
1820         } else {
1821             $path = $volid;
1822             $format = "raw";
1823         }
1824    }
1825
1826     my $opts = '';
1827     my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1828     foreach my $o (@qemu_drive_options) {
1829         $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1830     }
1831
1832     # snapshot only accepts on|off
1833     if (defined($drive->{snapshot})) {
1834         my $v = $drive->{snapshot} ? 'on' : 'off';
1835         $opts .= ",snapshot=$v";
1836     }
1837
1838     foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
1839         my ($dir, $qmpname) = @$type;
1840         if (my $v = $drive->{"mbps$dir"}) {
1841             $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
1842         }
1843         if (my $v = $drive->{"mbps${dir}_max"}) {
1844             $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
1845         }
1846         if (my $v = $drive->{"bps${dir}_max_length"}) {
1847             $opts .= ",throttling.bps$qmpname-max-length=$v";
1848         }
1849         if (my $v = $drive->{"iops${dir}"}) {
1850             $opts .= ",throttling.iops$qmpname=$v";
1851         }
1852         if (my $v = $drive->{"iops${dir}_max"}) {
1853             $opts .= ",throttling.iops$qmpname-max=$v";
1854         }
1855         if (my $v = $drive->{"iops${dir}_max_length"}) {
1856             $opts .= ",throttling.iops$qmpname-max-length=$v";
1857         }
1858     }
1859
1860     $opts .= ",format=$format" if $format && !$drive->{format};
1861
1862     my $cache_direct = 0;
1863
1864     if (my $cache = $drive->{cache}) {
1865         $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1866     } elsif (!drive_is_cdrom($drive)) {
1867         $opts .= ",cache=none";
1868         $cache_direct = 1;
1869     }
1870
1871     # aio native works only with O_DIRECT
1872     if (!$drive->{aio}) {
1873         if($cache_direct) {
1874             $opts .= ",aio=native";
1875         } else {
1876             $opts .= ",aio=threads";
1877         }
1878     }
1879
1880     if (!drive_is_cdrom($drive)) {
1881         my $detectzeroes;
1882         if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
1883             $detectzeroes = 'off';
1884         } elsif ($drive->{discard}) {
1885             $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
1886         } else {
1887             # This used to be our default with discard not being specified:
1888             $detectzeroes = 'on';
1889         }
1890         $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
1891     }
1892
1893     my $pathinfo = $path ? "file=$path," : '';
1894
1895     return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1896 }
1897
1898 sub print_netdevice_full {
1899     my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files) = @_;
1900
1901     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1902
1903     my $device = $net->{model};
1904     if ($net->{model} eq 'virtio') {
1905          $device = 'virtio-net-pci';
1906      };
1907
1908     my $pciaddr = print_pci_addr("$netid", $bridges);
1909     my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1910     if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1911         #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1912         my $vectors = $net->{queues} * 2 + 2;
1913         $tmpstr .= ",vectors=$vectors,mq=on";
1914     }
1915     $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1916
1917     if ($use_old_bios_files) {
1918         my $romfile;
1919         if ($device eq 'virtio-net-pci') {
1920             $romfile = 'pxe-virtio.rom';
1921         } elsif ($device eq 'e1000') {
1922             $romfile = 'pxe-e1000.rom';
1923         } elsif ($device eq 'ne2k') {
1924             $romfile = 'pxe-ne2k_pci.rom';
1925         } elsif ($device eq 'pcnet') {
1926             $romfile = 'pxe-pcnet.rom';
1927         } elsif ($device eq 'rtl8139') {
1928             $romfile = 'pxe-rtl8139.rom';
1929         }
1930         $tmpstr .= ",romfile=$romfile" if $romfile;
1931     }
1932
1933     return $tmpstr;
1934 }
1935
1936 sub print_netdev_full {
1937     my ($vmid, $conf, $net, $netid, $hotplug) = @_;
1938
1939     my $i = '';
1940     if ($netid =~ m/^net(\d+)$/) {
1941         $i = int($1);
1942     }
1943
1944     die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1945
1946     my $ifname = "tap${vmid}i$i";
1947
1948     # kvm uses TUNSETIFF ioctl, and that limits ifname length
1949     die "interface name '$ifname' is too long (max 15 character)\n"
1950         if length($ifname) >= 16;
1951
1952     my $vhostparam = '';
1953     $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1954
1955     my $vmname = $conf->{name} || "vm$vmid";
1956
1957     my $netdev = "";
1958     my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
1959
1960     if ($net->{bridge}) {
1961         $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1962     } else {
1963         $netdev = "type=user,id=$netid,hostname=$vmname";
1964     }
1965
1966     $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1967
1968     return $netdev;
1969 }
1970
1971
1972 sub print_cpu_device {
1973     my ($conf, $id) = @_;
1974
1975     my $kvm = $conf->{kvm} // 1;
1976     my $cpu = $kvm ? "kvm64" : "qemu64";
1977     if (my $cputype = $conf->{cpu}) {
1978         my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
1979             or die "Cannot parse cpu description: $cputype\n";
1980         $cpu = $cpuconf->{cputype};
1981     }
1982
1983     my $cores = $conf->{cores} || 1;
1984
1985     my $current_core = ($id - 1) % $cores;
1986     my $current_socket = int(($id - 1 - $current_core)/$cores);
1987
1988     return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
1989 }
1990
1991 my $vga_map = {
1992     'cirrus' => 'cirrus-vga',
1993     'std' => 'VGA',
1994     'vmware' => 'vmware-svga',
1995     'virtio' => 'virtio-vga',
1996 };
1997
1998 sub print_vga_device {
1999     my ($conf, $vga, $id, $qxlnum, $bridges) = @_;
2000
2001     my $type = $vga_map->{$vga->{type}};
2002     my $vgamem_mb = $vga->{memory};
2003     if ($qxlnum) {
2004         $type = $id ? 'qxl' : 'qxl-vga';
2005     }
2006     die "no devicetype for $vga->{type}\n" if !$type;
2007
2008     my $memory = "";
2009     if ($vgamem_mb) {
2010         if ($vga->{type} eq 'virtio') {
2011             my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2012             $memory = ",max_hostmem=$bytes";
2013         } elsif ($qxlnum) {
2014             # from https://www.spice-space.org/multiple-monitors.html
2015             $memory = ",vgamem_mb=$vga->{memory}";
2016             my $ram = $vgamem_mb * 4;
2017             my $vram = $vgamem_mb * 2;
2018             $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2019         } else {
2020             $memory = ",vgamem_mb=$vga->{memory}";
2021         }
2022     } elsif ($qxlnum && $id) {
2023         $memory = ",ram_size=67108864,vram_size=33554432";
2024     }
2025
2026     my $q35 = machine_type_is_q35($conf);
2027     my $vgaid = "vga" . ($id // '');
2028     my $pciaddr;
2029
2030     if ($q35 && $vgaid eq 'vga') {
2031         # the first display uses pcie.0 bus on q35 machines
2032         $pciaddr = print_pcie_addr($vgaid, $bridges);
2033     } else {
2034         $pciaddr = print_pci_addr($vgaid, $bridges);
2035     }
2036
2037     return "$type,id=${vgaid}${memory}${pciaddr}";
2038 }
2039
2040 sub drive_is_cloudinit {
2041     my ($drive) = @_;
2042     return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2043 }
2044
2045 sub drive_is_cdrom {
2046     my ($drive, $exclude_cloudinit) = @_;
2047
2048     return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2049
2050     return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2051
2052 }
2053
2054 sub parse_number_sets {
2055     my ($set) = @_;
2056     my $res = [];
2057     foreach my $part (split(/;/, $set)) {
2058         if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2059             die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2060             push @$res, [ $1, $2 ];
2061         } else {
2062             die "invalid range: $part\n";
2063         }
2064     }
2065     return $res;
2066 }
2067
2068 sub parse_numa {
2069     my ($data) = @_;
2070
2071     my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2072     $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2073     $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2074     return $res;
2075 }
2076
2077 sub parse_hostpci {
2078     my ($value) = @_;
2079
2080     return undef if !$value;
2081
2082     my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2083
2084     my @idlist = split(/;/, $res->{host});
2085     delete $res->{host};
2086     foreach my $id (@idlist) {
2087         if ($id =~ /^$PCIRE$/) {
2088             if (defined($2)) {
2089                 push @{$res->{pciid}}, { id => $1, function => $2 };
2090             } else {
2091                 my $pcidevices = lspci($1);
2092                 $res->{pciid} = $pcidevices->{$1};
2093             }
2094         } else {
2095             # should have been caught by parse_property_string already
2096             die "failed to parse PCI id: $id\n";
2097         }
2098     }
2099     return $res;
2100 }
2101
2102 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2103 sub parse_net {
2104     my ($data) = @_;
2105
2106     my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2107     if ($@) {
2108         warn $@;
2109         return undef;
2110     }
2111     if (!defined($res->{macaddr})) {
2112         my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2113         $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2114     }
2115     return $res;
2116 }
2117
2118 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2119 sub parse_ipconfig {
2120     my ($data) = @_;
2121
2122     my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2123     if ($@) {
2124         warn $@;
2125         return undef;
2126     }
2127
2128     if ($res->{gw} && !$res->{ip}) {
2129         warn 'gateway specified without specifying an IP address';
2130         return undef;
2131     }
2132     if ($res->{gw6} && !$res->{ip6}) {
2133         warn 'IPv6 gateway specified without specifying an IPv6 address';
2134         return undef;
2135     }
2136     if ($res->{gw} && $res->{ip} eq 'dhcp') {
2137         warn 'gateway specified together with DHCP';
2138         return undef;
2139     }
2140     if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2141         # gw6 + auto/dhcp
2142         warn "IPv6 gateway specified together with $res->{ip6} address";
2143         return undef;
2144     }
2145
2146     if (!$res->{ip} && !$res->{ip6}) {
2147         return { ip => 'dhcp', ip6 => 'dhcp' };
2148     }
2149
2150     return $res;
2151 }
2152
2153 sub print_net {
2154     my $net = shift;
2155
2156     return PVE::JSONSchema::print_property_string($net, $net_fmt);
2157 }
2158
2159 sub add_random_macs {
2160     my ($settings) = @_;
2161
2162     foreach my $opt (keys %$settings) {
2163         next if $opt !~ m/^net(\d+)$/;
2164         my $net = parse_net($settings->{$opt});
2165         next if !$net;
2166         $settings->{$opt} = print_net($net);
2167     }
2168 }
2169
2170 sub vm_is_volid_owner {
2171     my ($storecfg, $vmid, $volid) = @_;
2172
2173     if ($volid !~  m|^/|) {
2174         my ($path, $owner);
2175         eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2176         if ($owner && ($owner == $vmid)) {
2177             return 1;
2178         }
2179     }
2180
2181     return undef;
2182 }
2183
2184 sub split_flagged_list {
2185     my $text = shift || '';
2186     $text =~ s/[,;]/ /g;
2187     $text =~ s/^\s+//;
2188     return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
2189 }
2190
2191 sub join_flagged_list {
2192     my ($how, $lst) = @_;
2193     join $how, map { $lst->{$_} . $_ } keys %$lst;
2194 }
2195
2196 sub vmconfig_delete_pending_option {
2197     my ($conf, $key, $force) = @_;
2198
2199     delete $conf->{pending}->{$key};
2200     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2201     $pending_delete_hash->{$key} = $force ? '!' : '';
2202     $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2203 }
2204
2205 sub vmconfig_undelete_pending_option {
2206     my ($conf, $key) = @_;
2207
2208     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2209     delete $pending_delete_hash->{$key};
2210
2211     if (%$pending_delete_hash) {
2212         $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2213     } else {
2214         delete $conf->{pending}->{delete};
2215     }
2216 }
2217
2218 sub vmconfig_register_unused_drive {
2219     my ($storecfg, $vmid, $conf, $drive) = @_;
2220
2221     if (drive_is_cloudinit($drive)) {
2222         eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2223         warn $@ if $@;
2224     } elsif (!drive_is_cdrom($drive)) {
2225         my $volid = $drive->{file};
2226         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2227             PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2228         }
2229     }
2230 }
2231
2232 sub vmconfig_cleanup_pending {
2233     my ($conf) = @_;
2234
2235     # remove pending changes when nothing changed
2236     my $changes;
2237     foreach my $opt (keys %{$conf->{pending}}) {
2238         if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq  $conf->{$opt})) {
2239             $changes = 1;
2240             delete $conf->{pending}->{$opt};
2241         }
2242     }
2243
2244     my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
2245     my $pending_delete_hash = {};
2246     while (my ($opt, $force) = each %$current_delete_hash) {
2247         if (defined($conf->{$opt})) {
2248             $pending_delete_hash->{$opt} = $force;
2249         } else {
2250             $changes = 1;
2251         }
2252     }
2253
2254     if (%$pending_delete_hash) {
2255         $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2256     } else {
2257         delete $conf->{pending}->{delete};
2258     }
2259
2260     return $changes;
2261 }
2262
2263 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
2264 my $smbios1_fmt = {
2265     uuid => {
2266         type => 'string',
2267         pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2268         format_description => 'UUID',
2269         description => "Set SMBIOS1 UUID.",
2270         optional => 1,
2271     },
2272     version => {
2273         type => 'string',
2274         pattern => '\S+',
2275         format_description => 'string',
2276         description => "Set SMBIOS1 version.",
2277         optional => 1,
2278     },
2279     serial => {
2280         type => 'string',
2281         pattern => '\S+',
2282         format_description => 'string',
2283         description => "Set SMBIOS1 serial number.",
2284         optional => 1,
2285     },
2286     manufacturer => {
2287         type => 'string',
2288         pattern => '\S+',
2289         format_description => 'string',
2290         description => "Set SMBIOS1 manufacturer.",
2291         optional => 1,
2292     },
2293     product => {
2294         type => 'string',
2295         pattern => '\S+',
2296         format_description => 'string',
2297         description => "Set SMBIOS1 product ID.",
2298         optional => 1,
2299     },
2300     sku => {
2301         type => 'string',
2302         pattern => '\S+',
2303         format_description => 'string',
2304         description => "Set SMBIOS1 SKU string.",
2305         optional => 1,
2306     },
2307     family => {
2308         type => 'string',
2309         pattern => '\S+',
2310         format_description => 'string',
2311         description => "Set SMBIOS1 family string.",
2312         optional => 1,
2313     },
2314 };
2315
2316 sub parse_smbios1 {
2317     my ($data) = @_;
2318
2319     my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2320     warn $@ if $@;
2321     return $res;
2322 }
2323
2324 sub print_smbios1 {
2325     my ($smbios1) = @_;
2326     return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2327 }
2328
2329 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2330
2331 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2332 sub verify_bootdisk {
2333     my ($value, $noerr) = @_;
2334
2335     return $value if is_valid_drivename($value);
2336
2337     return undef if $noerr;
2338
2339     die "invalid boot disk '$value'\n";
2340 }
2341
2342 sub parse_watchdog {
2343     my ($value) = @_;
2344
2345     return undef if !$value;
2346
2347     my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2348     warn $@ if $@;
2349     return $res;
2350 }
2351
2352 sub parse_guest_agent {
2353     my ($value) = @_;
2354
2355     return {} if !defined($value->{agent});
2356
2357     my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2358     warn $@ if $@;
2359
2360     # if the agent is disabled ignore the other potentially set properties
2361     return {} if !$res->{enabled};
2362     return $res;
2363 }
2364
2365 sub parse_vga {
2366     my ($value) = @_;
2367
2368     return {} if !$value;
2369     my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2370     warn $@ if $@;
2371     return $res;
2372 }
2373
2374 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2375 sub verify_usb_device {
2376     my ($value, $noerr) = @_;
2377
2378     return $value if parse_usb_device($value);
2379
2380     return undef if $noerr;
2381
2382     die "unable to parse usb device\n";
2383 }
2384
2385 # add JSON properties for create and set function
2386 sub json_config_properties {
2387     my $prop = shift;
2388
2389     foreach my $opt (keys %$confdesc) {
2390         next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2391         $prop->{$opt} = $confdesc->{$opt};
2392     }
2393
2394     return $prop;
2395 }
2396
2397 # return copy of $confdesc_cloudinit to generate documentation
2398 sub cloudinit_config_properties {
2399
2400     return dclone($confdesc_cloudinit);
2401 }
2402
2403 sub check_type {
2404     my ($key, $value) = @_;
2405
2406     die "unknown setting '$key'\n" if !$confdesc->{$key};
2407
2408     my $type = $confdesc->{$key}->{type};
2409
2410     if (!defined($value)) {
2411         die "got undefined value\n";
2412     }
2413
2414     if ($value =~ m/[\n\r]/) {
2415         die "property contains a line feed\n";
2416     }
2417
2418     if ($type eq 'boolean') {
2419         return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2420         return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2421         die "type check ('boolean') failed - got '$value'\n";
2422     } elsif ($type eq 'integer') {
2423         return int($1) if $value =~ m/^(\d+)$/;
2424         die "type check ('integer') failed - got '$value'\n";
2425     } elsif ($type eq 'number') {
2426         return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2427         die "type check ('number') failed - got '$value'\n";
2428     } elsif ($type eq 'string') {
2429         if (my $fmt = $confdesc->{$key}->{format}) {
2430             PVE::JSONSchema::check_format($fmt, $value);
2431             return $value;
2432         }
2433         $value =~ s/^\"(.*)\"$/$1/;
2434         return $value;
2435     } else {
2436         die "internal error"
2437     }
2438 }
2439
2440 sub check_iommu_support{
2441     #fixme : need to check IOMMU support
2442     #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
2443
2444     my $iommu=1;
2445     return $iommu;
2446
2447 }
2448
2449 sub touch_config {
2450     my ($vmid) = @_;
2451
2452     my $conf = PVE::QemuConfig->config_file($vmid);
2453     utime undef, undef, $conf;
2454 }
2455
2456 sub destroy_vm {
2457     my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2458
2459     my $conffile = PVE::QemuConfig->config_file($vmid);
2460
2461     my $conf = PVE::QemuConfig->load_config($vmid);
2462
2463     PVE::QemuConfig->check_lock($conf) if !$skiplock;
2464
2465     if ($conf->{template}) {
2466         # check if any base image is still used by a linked clone
2467         foreach_drive($conf, sub {
2468                 my ($ds, $drive) = @_;
2469
2470                 return if drive_is_cdrom($drive);
2471
2472                 my $volid = $drive->{file};
2473
2474                 return if !$volid || $volid =~ m|^/|;
2475
2476                 die "base volume '$volid' is still in use by linked cloned\n"
2477                     if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2478
2479         });
2480     }
2481
2482     # only remove disks owned by this VM
2483     foreach_drive($conf, sub {
2484         my ($ds, $drive) = @_;
2485
2486         return if drive_is_cdrom($drive, 1);
2487
2488         my $volid = $drive->{file};
2489
2490         return if !$volid || $volid =~ m|^/|;
2491
2492         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2493         return if !$path || !$owner || ($owner != $vmid);
2494
2495         eval {
2496             PVE::Storage::vdisk_free($storecfg, $volid);
2497         };
2498         warn "Could not remove disk '$volid', check manually: $@" if $@;
2499
2500     });
2501
2502     if ($keep_empty_config) {
2503         PVE::Tools::file_set_contents($conffile, "memory: 128\n");
2504     } else {
2505         unlink $conffile;
2506     }
2507
2508     # also remove unused disk
2509     eval {
2510         my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2511
2512         eval {
2513             PVE::Storage::foreach_volid($dl, sub {
2514                 my ($volid, $sid, $volname, $d) = @_;
2515                 PVE::Storage::vdisk_free($storecfg, $volid);
2516             });
2517         };
2518         warn $@ if $@;
2519
2520     };
2521     warn $@ if $@;
2522 }
2523
2524 sub parse_vm_config {
2525     my ($filename, $raw) = @_;
2526
2527     return undef if !defined($raw);
2528
2529     my $res = {
2530         digest => Digest::SHA::sha1_hex($raw),
2531         snapshots => {},
2532         pending => {},
2533     };
2534
2535     $filename =~ m|/qemu-server/(\d+)\.conf$|
2536         || die "got strange filename '$filename'";
2537
2538     my $vmid = $1;
2539
2540     my $conf = $res;
2541     my $descr;
2542     my $section = '';
2543
2544     my @lines = split(/\n/, $raw);
2545     foreach my $line (@lines) {
2546         next if $line =~ m/^\s*$/;
2547
2548         if ($line =~ m/^\[PENDING\]\s*$/i) {
2549             $section = 'pending';
2550             if (defined($descr)) {
2551                 $descr =~ s/\s+$//;
2552                 $conf->{description} = $descr;
2553             }
2554             $descr = undef;
2555             $conf = $res->{$section} = {};
2556             next;
2557
2558         } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2559             $section = $1;
2560             if (defined($descr)) {
2561                 $descr =~ s/\s+$//;
2562                 $conf->{description} = $descr;
2563             }
2564             $descr = undef;
2565             $conf = $res->{snapshots}->{$section} = {};
2566             next;
2567         }
2568
2569         if ($line =~ m/^\#(.*)\s*$/) {
2570             $descr = '' if !defined($descr);
2571             $descr .= PVE::Tools::decode_text($1) . "\n";
2572             next;
2573         }
2574
2575         if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2576             $descr = '' if !defined($descr);
2577             $descr .= PVE::Tools::decode_text($2);
2578         } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2579             $conf->{snapstate} = $1;
2580         } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2581             my $key = $1;
2582             my $value = $2;
2583             $conf->{$key} = $value;
2584         } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2585             my $value = $1;
2586             if ($section eq 'pending') {
2587                 $conf->{delete} = $value; # we parse this later
2588             } else {
2589                 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2590             }
2591         } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2592             my $key = $1;
2593             my $value = $2;
2594             eval { $value = check_type($key, $value); };
2595             if ($@) {
2596                 warn "vm $vmid - unable to parse value of '$key' - $@";
2597             } else {
2598                 $key = 'ide2' if $key eq 'cdrom';
2599                 my $fmt = $confdesc->{$key}->{format};
2600                 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2601                     my $v = parse_drive($key, $value);
2602                     if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2603                         $v->{file} = $volid;
2604                         $value = print_drive($vmid, $v);
2605                     } else {
2606                         warn "vm $vmid - unable to parse value of '$key'\n";
2607                         next;
2608                     }
2609                 }
2610
2611                 $conf->{$key} = $value;
2612             }
2613         }
2614     }
2615
2616     if (defined($descr)) {
2617         $descr =~ s/\s+$//;
2618         $conf->{description} = $descr;
2619     }
2620     delete $res->{snapstate}; # just to be sure
2621
2622     return $res;
2623 }
2624
2625 sub write_vm_config {
2626     my ($filename, $conf) = @_;
2627
2628     delete $conf->{snapstate}; # just to be sure
2629
2630     if ($conf->{cdrom}) {
2631         die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2632         $conf->{ide2} = $conf->{cdrom};
2633         delete $conf->{cdrom};
2634     }
2635
2636     # we do not use 'smp' any longer
2637     if ($conf->{sockets}) {
2638         delete $conf->{smp};
2639     } elsif ($conf->{smp}) {
2640         $conf->{sockets} = $conf->{smp};
2641         delete $conf->{cores};
2642         delete $conf->{smp};
2643     }
2644
2645     my $used_volids = {};
2646
2647     my $cleanup_config = sub {
2648         my ($cref, $pending, $snapname) = @_;
2649
2650         foreach my $key (keys %$cref) {
2651             next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2652                 $key eq 'snapstate' || $key eq 'pending';
2653             my $value = $cref->{$key};
2654             if ($key eq 'delete') {
2655                 die "propertry 'delete' is only allowed in [PENDING]\n"
2656                     if !$pending;
2657                 # fixme: check syntax?
2658                 next;
2659             }
2660             eval { $value = check_type($key, $value); };
2661             die "unable to parse value of '$key' - $@" if $@;
2662
2663             $cref->{$key} = $value;
2664
2665             if (!$snapname && is_valid_drivename($key)) {
2666                 my $drive = parse_drive($key, $value);
2667                 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2668             }
2669         }
2670     };
2671
2672     &$cleanup_config($conf);
2673
2674     &$cleanup_config($conf->{pending}, 1);
2675
2676     foreach my $snapname (keys %{$conf->{snapshots}}) {
2677         die "internal error" if $snapname eq 'pending';
2678         &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2679     }
2680
2681     # remove 'unusedX' settings if we re-add a volume
2682     foreach my $key (keys %$conf) {
2683         my $value = $conf->{$key};
2684         if ($key =~ m/^unused/ && $used_volids->{$value}) {
2685             delete $conf->{$key};
2686         }
2687     }
2688
2689     my $generate_raw_config = sub {
2690         my ($conf, $pending) = @_;
2691
2692         my $raw = '';
2693
2694         # add description as comment to top of file
2695         if (defined(my $descr = $conf->{description})) {
2696             if ($descr) {
2697                 foreach my $cl (split(/\n/, $descr)) {
2698                     $raw .= '#' .  PVE::Tools::encode_text($cl) . "\n";
2699                 }
2700             } else {
2701                 $raw .= "#\n" if $pending;
2702             }
2703         }
2704
2705         foreach my $key (sort keys %$conf) {
2706             next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2707             $raw .= "$key: $conf->{$key}\n";
2708         }
2709         return $raw;
2710     };
2711
2712     my $raw = &$generate_raw_config($conf);
2713
2714     if (scalar(keys %{$conf->{pending}})){
2715         $raw .= "\n[PENDING]\n";
2716         $raw .= &$generate_raw_config($conf->{pending}, 1);
2717     }
2718
2719     foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2720         $raw .= "\n[$snapname]\n";
2721         $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2722     }
2723
2724     return $raw;
2725 }
2726
2727 sub load_defaults {
2728
2729     my $res = {};
2730
2731     # we use static defaults from our JSON schema configuration
2732     foreach my $key (keys %$confdesc) {
2733         if (defined(my $default = $confdesc->{$key}->{default})) {
2734             $res->{$key} = $default;
2735         }
2736     }
2737
2738     return $res;
2739 }
2740
2741 sub config_list {
2742     my $vmlist = PVE::Cluster::get_vmlist();
2743     my $res = {};
2744     return $res if !$vmlist || !$vmlist->{ids};
2745     my $ids = $vmlist->{ids};
2746
2747     foreach my $vmid (keys %$ids) {
2748         my $d = $ids->{$vmid};
2749         next if !$d->{node} || $d->{node} ne $nodename;
2750         next if !$d->{type} || $d->{type} ne 'qemu';
2751         $res->{$vmid}->{exists} = 1;
2752     }
2753     return $res;
2754 }
2755
2756 # test if VM uses local resources (to prevent migration)
2757 sub check_local_resources {
2758     my ($conf, $noerr) = @_;
2759
2760     my $loc_res = 0;
2761
2762     $loc_res = 1 if $conf->{hostusb}; # old syntax
2763     $loc_res = 1 if $conf->{hostpci}; # old syntax
2764
2765     foreach my $k (keys %$conf) {
2766         next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2767         # sockets are safe: they will recreated be on the target side post-migrate
2768         next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2769         $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2770     }
2771
2772     die "VM uses local resources\n" if $loc_res && !$noerr;
2773
2774     return $loc_res;
2775 }
2776
2777 # check if used storages are available on all nodes (use by migrate)
2778 sub check_storage_availability {
2779     my ($storecfg, $conf, $node) = @_;
2780
2781     foreach_drive($conf, sub {
2782         my ($ds, $drive) = @_;
2783
2784         my $volid = $drive->{file};
2785         return if !$volid;
2786
2787         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2788         return if !$sid;
2789
2790         # check if storage is available on both nodes
2791         my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2792         PVE::Storage::storage_check_node($storecfg, $sid, $node);
2793    });
2794 }
2795
2796 # list nodes where all VM images are available (used by has_feature API)
2797 sub shared_nodes {
2798     my ($conf, $storecfg) = @_;
2799
2800     my $nodelist = PVE::Cluster::get_nodelist();
2801     my $nodehash = { map { $_ => 1 } @$nodelist };
2802     my $nodename = PVE::INotify::nodename();
2803
2804     foreach_drive($conf, sub {
2805         my ($ds, $drive) = @_;
2806
2807         my $volid = $drive->{file};
2808         return if !$volid;
2809
2810         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2811         if ($storeid) {
2812             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2813             if ($scfg->{disable}) {
2814                 $nodehash = {};
2815             } elsif (my $avail = $scfg->{nodes}) {
2816                 foreach my $node (keys %$nodehash) {
2817                     delete $nodehash->{$node} if !$avail->{$node};
2818                 }
2819             } elsif (!$scfg->{shared}) {
2820                 foreach my $node (keys %$nodehash) {
2821                     delete $nodehash->{$node} if $node ne $nodename
2822                 }
2823             }
2824         }
2825     });
2826
2827     return $nodehash
2828 }
2829
2830 sub check_cmdline {
2831     my ($pidfile, $pid) = @_;
2832
2833     my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2834     if (defined($fh)) {
2835         my $line = <$fh>;
2836         $fh->close;
2837         return undef if !$line;
2838         my @param = split(/\0/, $line);
2839
2840         my $cmd = $param[0];
2841         return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2842
2843         for (my $i = 0; $i < scalar (@param); $i++) {
2844             my $p = $param[$i];
2845             next if !$p;
2846             if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2847                 my $p = $param[$i+1];
2848                 return 1 if $p && ($p eq $pidfile);
2849                 return undef;
2850             }
2851         }
2852     }
2853     return undef;
2854 }
2855
2856 sub check_running {
2857     my ($vmid, $nocheck, $node) = @_;
2858
2859     my $filename = PVE::QemuConfig->config_file($vmid, $node);
2860
2861     die "unable to find configuration file for VM $vmid - no such machine\n"
2862         if !$nocheck && ! -f $filename;
2863
2864     my $pidfile = pidfile_name($vmid);
2865
2866     if (my $fd = IO::File->new("<$pidfile")) {
2867         my $st = stat($fd);
2868         my $line = <$fd>;
2869         close($fd);
2870
2871         my $mtime = $st->mtime;
2872         if ($mtime > time()) {
2873             warn "file '$filename' modified in future\n";
2874         }
2875
2876         if ($line =~ m/^(\d+)$/) {
2877             my $pid = $1;
2878             if (check_cmdline($pidfile, $pid)) {
2879                 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2880                     return $pid;
2881                 }
2882             }
2883         }
2884     }
2885
2886     return undef;
2887 }
2888
2889 sub vzlist {
2890
2891     my $vzlist = config_list();
2892
2893     my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2894
2895     while (defined(my $de = $fd->read)) {
2896         next if $de !~ m/^(\d+)\.pid$/;
2897         my $vmid = $1;
2898         next if !defined($vzlist->{$vmid});
2899         if (my $pid = check_running($vmid)) {
2900             $vzlist->{$vmid}->{pid} = $pid;
2901         }
2902     }
2903
2904     return $vzlist;
2905 }
2906
2907 sub disksize {
2908     my ($storecfg, $conf) = @_;
2909
2910     my $bootdisk = $conf->{bootdisk};
2911     return undef if !$bootdisk;
2912     return undef if !is_valid_drivename($bootdisk);
2913
2914     return undef if !$conf->{$bootdisk};
2915
2916     my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2917     return undef if !defined($drive);
2918
2919     return undef if drive_is_cdrom($drive);
2920
2921     my $volid = $drive->{file};
2922     return undef if !$volid;
2923
2924     return $drive->{size};
2925 }
2926
2927 our $vmstatus_return_properties = {
2928     vmid => get_standard_option('pve-vmid'),
2929     status => {
2930         description => "Qemu process status.",
2931         type => 'string',
2932         enum => ['stopped', 'running'],
2933     },
2934     maxmem => {
2935         description => "Maximum memory in bytes.",
2936         type => 'integer',
2937         optional => 1,
2938         renderer => 'bytes',
2939     },
2940     maxdisk => {
2941         description => "Root disk size in bytes.",
2942         type => 'integer',
2943         optional => 1,
2944         renderer => 'bytes',
2945     },
2946     name => {
2947         description => "VM name.",
2948         type => 'string',
2949         optional => 1,
2950     },
2951     qmpstatus => {
2952         description => "Qemu QMP agent status.",
2953         type => 'string',
2954         optional => 1,
2955     },
2956     pid => {
2957         description => "PID of running qemu process.",
2958         type => 'integer',
2959         optional => 1,
2960     },
2961     uptime => {
2962         description => "Uptime.",
2963         type => 'integer',
2964         optional => 1,
2965         renderer => 'duration',
2966     },
2967     cpus => {
2968         description => "Maximum usable CPUs.",
2969         type => 'number',
2970         optional => 1,
2971     },
2972 };
2973
2974 my $last_proc_pid_stat;
2975
2976 # get VM status information
2977 # This must be fast and should not block ($full == false)
2978 # We only query KVM using QMP if $full == true (this can be slow)
2979 sub vmstatus {
2980     my ($opt_vmid, $full) = @_;
2981
2982     my $res = {};
2983
2984     my $storecfg = PVE::Storage::config();
2985
2986     my $list = vzlist();
2987     my $defaults = load_defaults();
2988
2989     my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2990
2991     my $cpucount = $cpuinfo->{cpus} || 1;
2992
2993     foreach my $vmid (keys %$list) {
2994         next if $opt_vmid && ($vmid ne $opt_vmid);
2995
2996         my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
2997         my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2998
2999         my $d = { vmid => $vmid };
3000         $d->{pid} = $list->{$vmid}->{pid};
3001
3002         # fixme: better status?
3003         $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3004
3005         my $size = disksize($storecfg, $conf);
3006         if (defined($size)) {
3007             $d->{disk} = 0; # no info available
3008             $d->{maxdisk} = $size;
3009         } else {
3010             $d->{disk} = 0;
3011             $d->{maxdisk} = 0;
3012         }
3013
3014         $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3015             * ($conf->{cores} || $defaults->{cores});
3016         $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3017         $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3018
3019         $d->{name} = $conf->{name} || "VM $vmid";
3020         $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3021             : $defaults->{memory}*(1024*1024);
3022
3023         if ($conf->{balloon}) {
3024             $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3025             $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3026                 : $defaults->{shares};
3027         }
3028
3029         $d->{uptime} = 0;
3030         $d->{cpu} = 0;
3031         $d->{mem} = 0;
3032
3033         $d->{netout} = 0;
3034         $d->{netin} = 0;
3035
3036         $d->{diskread} = 0;
3037         $d->{diskwrite} = 0;
3038
3039         $d->{template} = PVE::QemuConfig->is_template($conf);
3040
3041         $d->{serial} = 1 if conf_has_serial($conf);
3042
3043         $res->{$vmid} = $d;
3044     }
3045
3046     my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3047     foreach my $dev (keys %$netdev) {
3048         next if $dev !~ m/^tap([1-9]\d*)i/;
3049         my $vmid = $1;
3050         my $d = $res->{$vmid};
3051         next if !$d;
3052
3053         $d->{netout} += $netdev->{$dev}->{receive};
3054         $d->{netin} += $netdev->{$dev}->{transmit};
3055
3056         if ($full) {
3057             $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3058             $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3059         }
3060
3061     }
3062
3063     my $ctime = gettimeofday;
3064
3065     foreach my $vmid (keys %$list) {
3066
3067         my $d = $res->{$vmid};
3068         my $pid = $d->{pid};
3069         next if !$pid;
3070
3071         my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3072         next if !$pstat; # not running
3073
3074         my $used = $pstat->{utime} + $pstat->{stime};
3075
3076         $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3077
3078         if ($pstat->{vsize}) {
3079             $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3080         }
3081
3082         my $old = $last_proc_pid_stat->{$pid};
3083         if (!$old) {
3084             $last_proc_pid_stat->{$pid} = {
3085                 time => $ctime,
3086                 used => $used,
3087                 cpu => 0,
3088             };
3089             next;
3090         }
3091
3092         my $dtime = ($ctime -  $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3093
3094         if ($dtime > 1000) {
3095             my $dutime = $used -  $old->{used};
3096
3097             $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3098             $last_proc_pid_stat->{$pid} = {
3099                 time => $ctime,
3100                 used => $used,
3101                 cpu => $d->{cpu},
3102             };
3103         } else {
3104             $d->{cpu} = $old->{cpu};
3105         }
3106     }
3107
3108     return $res if !$full;
3109
3110     my $qmpclient = PVE::QMPClient->new();
3111
3112     my $ballooncb = sub {
3113         my ($vmid, $resp) = @_;
3114
3115         my $info = $resp->{'return'};
3116         return if !$info->{max_mem};
3117
3118         my $d = $res->{$vmid};
3119
3120         # use memory assigned to VM
3121         $d->{maxmem} = $info->{max_mem};
3122         $d->{balloon} = $info->{actual};
3123
3124         if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3125             $d->{mem} = $info->{total_mem} - $info->{free_mem};
3126             $d->{freemem} = $info->{free_mem};
3127         }
3128
3129         $d->{ballooninfo} = $info;
3130     };
3131
3132     my $blockstatscb = sub {
3133         my ($vmid, $resp) = @_;
3134         my $data = $resp->{'return'} || [];
3135         my $totalrdbytes = 0;
3136         my $totalwrbytes = 0;
3137
3138         for my $blockstat (@$data) {
3139             $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3140             $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3141
3142             $blockstat->{device} =~ s/drive-//;
3143             $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3144         }
3145         $res->{$vmid}->{diskread} = $totalrdbytes;
3146         $res->{$vmid}->{diskwrite} = $totalwrbytes;
3147     };
3148
3149     my $statuscb = sub {
3150         my ($vmid, $resp) = @_;
3151
3152         $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3153         # this fails if ballon driver is not loaded, so this must be
3154         # the last commnand (following command are aborted if this fails).
3155         $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3156
3157         my $status = 'unknown';
3158         if (!defined($status = $resp->{'return'}->{status})) {
3159             warn "unable to get VM status\n";
3160             return;
3161         }
3162
3163         $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3164     };
3165
3166     foreach my $vmid (keys %$list) {
3167         next if $opt_vmid && ($vmid ne $opt_vmid);
3168         next if !$res->{$vmid}->{pid}; # not running
3169         $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3170     }
3171
3172     $qmpclient->queue_execute(undef, 2);
3173
3174     foreach my $vmid (keys %$list) {
3175         next if $opt_vmid && ($vmid ne $opt_vmid);
3176         $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3177     }
3178
3179     return $res;
3180 }
3181
3182 sub foreach_drive {
3183     my ($conf, $func, @param) = @_;
3184
3185     foreach my $ds (valid_drive_names()) {
3186         next if !defined($conf->{$ds});
3187
3188         my $drive = parse_drive($ds, $conf->{$ds});
3189         next if !$drive;
3190
3191         &$func($ds, $drive, @param);
3192     }
3193 }
3194
3195 sub foreach_volid {
3196     my ($conf, $func, @param) = @_;
3197
3198     my $volhash = {};
3199
3200     my $test_volid = sub {
3201         my ($volid, $is_cdrom, $replicate, $shared, $snapname) = @_;
3202
3203         return if !$volid;
3204
3205         $volhash->{$volid}->{cdrom} //= 1;
3206         $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3207
3208         $volhash->{$volid}->{replicate} //= 0;
3209         $volhash->{$volid}->{replicate} = 1 if $replicate;
3210
3211         $volhash->{$volid}->{shared} //= 0;
3212         $volhash->{$volid}->{shared} = 1 if $shared;
3213
3214         $volhash->{$volid}->{referenced_in_config} //= 0;
3215         $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3216
3217         $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3218             if defined($snapname);
3219     };
3220
3221     foreach_drive($conf, sub {
3222         my ($ds, $drive) = @_;
3223         $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef);
3224     });
3225
3226     foreach my $snapname (keys %{$conf->{snapshots}}) {
3227         my $snap = $conf->{snapshots}->{$snapname};
3228         $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3229         foreach_drive($snap, sub {
3230             my ($ds, $drive) = @_;
3231             $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3232         });
3233     }
3234
3235     foreach my $volid (keys %$volhash) {
3236         &$func($volid, $volhash->{$volid}, @param);
3237     }
3238 }
3239
3240 sub conf_has_serial {
3241     my ($conf) = @_;
3242
3243     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3244         if ($conf->{"serial$i"}) {
3245             return 1;
3246         }
3247     }
3248
3249     return 0;
3250 }
3251
3252 sub vga_conf_has_spice {
3253     my ($vga) = @_;
3254
3255     my $vgaconf = parse_vga($vga);
3256     my $vgatype = $vgaconf->{type};
3257     return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3258
3259     return $1 || 1;
3260 }
3261
3262 my $host_arch; # FIXME: fix PVE::Tools::get_host_arch
3263 sub get_host_arch() {
3264     $host_arch = (POSIX::uname())[4] if !$host_arch;
3265     return $host_arch;
3266 }
3267
3268 sub is_native($) {
3269     my ($arch) = @_;
3270     return get_host_arch() eq $arch;
3271 }
3272
3273 my $default_machines = {
3274     x86_64 => 'pc',
3275     aarch64 => 'virt',
3276 };
3277
3278 sub get_basic_machine_info {
3279     my ($conf, $forcemachine) = @_;
3280
3281     my $arch = $conf->{arch} // get_host_arch();
3282     my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3283     return ($arch, $machine);
3284 }
3285
3286 sub get_ovmf_files($) {
3287     my ($arch) = @_;
3288
3289     my $ovmf = $OVMF->{$arch}
3290         or die "no OVMF images known for architecture '$arch'\n";
3291
3292     return @$ovmf;
3293 }
3294
3295 sub config_to_command {
3296     my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3297
3298     my $cmd = [];
3299     my $globalFlags = [];
3300     my $machineFlags = [];
3301     my $rtcFlags = [];
3302     my $cpuFlags = [];
3303     my $devices = [];
3304     my $pciaddr = '';
3305     my $bridges = {};
3306     my $kvmver = kvm_user_version();
3307     my $vernum = 0; # unknown
3308     my $ostype = $conf->{ostype};
3309     my $winversion = windows_version($ostype);
3310     my $kvm = $conf->{kvm};
3311
3312     my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3313     $kvm //= 1 if is_native($arch);
3314
3315     if ($kvm) {
3316         die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3317             if !defined kvm_version();
3318     }
3319
3320     if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3321         $vernum = $1*1000000+$2*1000;
3322     } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3323         $vernum = $1*1000000+$2*1000+$3;
3324     }
3325
3326     die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3327
3328     my $have_ovz = -f '/proc/vz/vestat';
3329
3330     my $q35 = machine_type_is_q35($conf);
3331     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3332     my $use_old_bios_files = undef;
3333     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3334
3335     my $cpuunits = defined($conf->{cpuunits}) ?
3336             $conf->{cpuunits} : $defaults->{cpuunits};
3337
3338     push @$cmd, '/usr/bin/kvm';
3339
3340     push @$cmd, '-id', $vmid;
3341
3342     my $vmname = $conf->{name} || "vm$vmid";
3343
3344     push @$cmd, '-name', $vmname;
3345
3346     my $use_virtio = 0;
3347
3348     my $qmpsocket = qmp_socket($vmid);
3349     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3350     push @$cmd, '-mon', "chardev=qmp,mode=control";
3351
3352     if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3353         my $eventsocket = qmp_socket($vmid, 0, 'event');
3354         push @$cmd, '-chardev', "socket,id=qmp-event,path=$eventsocket,server,nowait";
3355         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3356     }
3357
3358     push @$cmd, '-pidfile' , pidfile_name($vmid);
3359
3360     push @$cmd, '-daemonize';
3361
3362     if ($conf->{smbios1}) {
3363         push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3364     }
3365
3366     if ($conf->{vmgenid}) {
3367         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3368     }
3369
3370     my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3371     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3372         die "uefi base image not found\n" if ! -f $ovmf_code;
3373
3374         my $path;
3375         my $format;
3376         if (my $efidisk = $conf->{efidisk0}) {
3377             my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3378             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3379             $format = $d->{format};
3380             if ($storeid) {
3381                 $path = PVE::Storage::path($storecfg, $d->{file});
3382                 if (!defined($format)) {
3383                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3384                     $format = qemu_img_format($scfg, $volname);
3385                 }
3386             } else {
3387                 $path = $d->{file};
3388                 die "efidisk format must be specified\n"
3389                     if !defined($format);
3390             }
3391         } else {
3392             warn "no efidisk configured! Using temporary efivars disk.\n";
3393             $path = "/tmp/$vmid-ovmf.fd";
3394             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3395             $format = 'raw';
3396         }
3397
3398         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3399         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3400     }
3401
3402
3403     # add usb controllers
3404     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $q35, $usbdesc->{format}, $MAX_USB_DEVICES);
3405     push @$devices, @usbcontrollers if @usbcontrollers;
3406     my $vga = parse_vga($conf->{vga});
3407
3408     my $qxlnum = vga_conf_has_spice($conf->{vga});
3409     $vga->{type} = 'qxl' if $qxlnum;
3410
3411     if (!$vga->{type}) {
3412         if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3413             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3414         } else {
3415             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3416         }
3417     }
3418
3419     # enable absolute mouse coordinates (needed by vnc)
3420     my $tablet;
3421     if (defined($conf->{tablet})) {
3422         $tablet = $conf->{tablet};
3423     } else {
3424         $tablet = $defaults->{tablet};
3425         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3426         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3427     }
3428
3429     push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
3430
3431     my $kvm_off = 0;
3432     my $gpu_passthrough;
3433
3434     # host pci devices
3435     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
3436         my $d = parse_hostpci($conf->{"hostpci$i"});
3437         next if !$d;
3438
3439         my $pcie = $d->{pcie};
3440         if($pcie){
3441             die "q35 machine model is not enabled" if !$q35;
3442             $pciaddr = print_pcie_addr("hostpci$i");
3443         }else{
3444             $pciaddr = print_pci_addr("hostpci$i", $bridges);
3445         }
3446
3447         my $rombar = defined($d->{rombar}) && !$d->{rombar} ? ',rombar=0' : '';
3448         my $romfile = $d->{romfile};
3449
3450         my $xvga = '';
3451         if ($d->{'x-vga'}) {
3452             $xvga = ',x-vga=on';
3453             $kvm_off = 1;
3454             $vga->{type} = 'none';
3455             $gpu_passthrough = 1;
3456
3457             if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3458                 $xvga = "";
3459             }
3460         }
3461         my $pcidevices = $d->{pciid};
3462         my $multifunction = 1 if @$pcidevices > 1;
3463
3464         my $j=0;
3465         foreach my $pcidevice (@$pcidevices) {
3466
3467             my $id = "hostpci$i";
3468             $id .= ".$j" if $multifunction;
3469             my $addr = $pciaddr;
3470             $addr .= ".$j" if $multifunction;
3471             my $devicestr = "vfio-pci,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
3472
3473             if($j == 0){
3474                 $devicestr .= "$rombar$xvga";
3475                 $devicestr .= ",multifunction=on" if $multifunction;
3476                 $devicestr .= ",romfile=/usr/share/kvm/$romfile" if $romfile;
3477             }
3478
3479             push @$devices, '-device', $devicestr;
3480             $j++;
3481         }
3482     }
3483
3484     # usb devices
3485     my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES);
3486     push @$devices, @usbdevices if @usbdevices;
3487     # serial devices
3488     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3489         if (my $path = $conf->{"serial$i"}) {
3490             if ($path eq 'socket') {
3491                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3492                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3493                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3494             } else {
3495                 die "no such serial device\n" if ! -c $path;
3496                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3497                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3498             }
3499         }
3500     }
3501
3502     # parallel devices
3503     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3504         if (my $path = $conf->{"parallel$i"}) {
3505             die "no such parallel device\n" if ! -c $path;
3506             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3507             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3508             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3509         }
3510     }
3511
3512
3513     my $sockets = 1;
3514     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3515     $sockets = $conf->{sockets} if  $conf->{sockets};
3516
3517     my $cores = $conf->{cores} || 1;
3518
3519     my $maxcpus = $sockets * $cores;
3520
3521     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3522
3523     my $allowed_vcpus = $cpuinfo->{cpus};
3524
3525     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3526         if ($allowed_vcpus < $maxcpus);
3527
3528     if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3529
3530         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3531         for (my $i = 2; $i <= $vcpus; $i++)  {
3532             my $cpustr = print_cpu_device($conf,$i);
3533             push @$cmd, '-device', $cpustr;
3534         }
3535
3536     } else {
3537
3538         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3539     }
3540     push @$cmd, '-nodefaults';
3541
3542     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3543
3544     my $bootindex_hash = {};
3545     my $i = 1;
3546     foreach my $o (split(//, $bootorder)) {
3547         $bootindex_hash->{$o} = $i*100;
3548         $i++;
3549     }
3550
3551     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3552
3553     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3554
3555     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3556
3557     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3558         push @$devices, '-device', print_vga_device($conf, $vga, undef, $qxlnum, $bridges);
3559         my $socket = vnc_socket($vmid);
3560         push @$cmd,  '-vnc', "unix:$socket,x509,password";
3561     } else {
3562         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3563         push @$cmd, '-nographic';
3564     }
3565
3566     # time drift fix
3567     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3568
3569     my $useLocaltime = $conf->{localtime};
3570
3571     if ($winversion >= 5) { # windows
3572         $useLocaltime = 1 if !defined($conf->{localtime});
3573
3574         # use time drift fix when acpi is enabled
3575         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3576             $tdf = 1 if !defined($conf->{tdf});
3577         }
3578     }
3579
3580     if ($winversion >= 6) {
3581         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3582         push @$cmd, '-no-hpet';
3583     }
3584
3585     push @$rtcFlags, 'driftfix=slew' if $tdf;
3586
3587     if (!$kvm) {
3588         push @$machineFlags, 'accel=tcg';
3589     }
3590
3591     if ($machine_type) {
3592         push @$machineFlags, "type=${machine_type}";
3593     }
3594
3595     if ($conf->{startdate}) {
3596         push @$rtcFlags, "base=$conf->{startdate}";
3597     } elsif ($useLocaltime) {
3598         push @$rtcFlags, 'base=localtime';
3599     }
3600
3601     my $cpu = $kvm ? "kvm64" : "qemu64";
3602     if (my $cputype = $conf->{cpu}) {
3603         my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3604             or die "Cannot parse cpu description: $cputype\n";
3605         $cpu = $cpuconf->{cputype};
3606         $kvm_off = 1 if $cpuconf->{hidden};
3607
3608         if (defined(my $flags = $cpuconf->{flags})) {
3609             push @$cpuFlags, split(";", $flags);
3610         }
3611     }
3612
3613     push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
3614
3615     push @$cpuFlags , '-x2apic'
3616         if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3617
3618     push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3619
3620     push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3621
3622     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3623
3624         push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3625         push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3626     }
3627
3628     add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough) if $kvm;
3629
3630     push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm;
3631
3632     push @$cpuFlags, 'kvm=off' if $kvm_off;
3633
3634     my $cpu_vendor = $cpu_vendor_list->{$cpu} ||
3635         die "internal error"; # should not happen
3636
3637     push @$cpuFlags, "vendor=${cpu_vendor}"
3638         if $cpu_vendor ne 'default';
3639
3640     $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3641
3642     push @$cmd, '-cpu', $cpu;
3643
3644     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3645
3646     push @$cmd, '-S' if $conf->{freeze};
3647
3648     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3649
3650     # enable sound
3651     #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3652     #push @$cmd, '-soundhw', 'es1370';
3653     #push @$cmd, '-soundhw', $soundhw if $soundhw;
3654
3655     if (parse_guest_agent($conf)->{enabled}) {
3656         my $qgasocket = qmp_socket($vmid, 1);
3657         my $pciaddr = print_pci_addr("qga0", $bridges);
3658         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3659         push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3660         push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3661     }
3662
3663     my $spice_port;
3664
3665     if ($qxlnum) {
3666         if ($qxlnum > 1) {
3667             if ($winversion){
3668                 for(my $i = 1; $i < $qxlnum; $i++){
3669                     push @$devices, '-device', print_vga_device($conf, $vga, $i, $qxlnum, $bridges);
3670                 }
3671             } else {
3672                 # assume other OS works like Linux
3673                 my ($ram, $vram) = ("134217728", "67108864");
3674                 if ($vga->{memory}) {
3675                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3676                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3677                 }
3678                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3679                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3680             }
3681         }
3682
3683         my $pciaddr = print_pci_addr("spice", $bridges);
3684
3685         my $nodename = PVE::INotify::nodename();
3686         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3687         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3688         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3689         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3690         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3691
3692         push @$devices, '-spice', "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3693
3694         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3695         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3696         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3697     }
3698
3699     # enable balloon by default, unless explicitly disabled
3700     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3701         $pciaddr = print_pci_addr("balloon0", $bridges);
3702         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3703     }
3704
3705     if ($conf->{watchdog}) {
3706         my $wdopts = parse_watchdog($conf->{watchdog});
3707         $pciaddr = print_pci_addr("watchdog", $bridges);
3708         my $watchdog = $wdopts->{model} || 'i6300esb';
3709         push @$devices, '-device', "$watchdog$pciaddr";
3710         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3711     }
3712
3713     my $vollist = [];
3714     my $scsicontroller = {};
3715     my $ahcicontroller = {};
3716     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3717
3718     # Add iscsi initiator name if available
3719     if (my $initiator = get_initiator_name()) {
3720         push @$devices, '-iscsi', "initiator-name=$initiator";
3721     }
3722
3723     foreach_drive($conf, sub {
3724         my ($ds, $drive) = @_;
3725
3726         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3727             push @$vollist, $drive->{file};
3728         }
3729
3730         # ignore efidisk here, already added in bios/fw handling code above
3731         return if $drive->{interface} eq 'efidisk';
3732
3733         $use_virtio = 1 if $ds =~ m/^virtio/;
3734
3735         if (drive_is_cdrom ($drive)) {
3736             if ($bootindex_hash->{d}) {
3737                 $drive->{bootindex} = $bootindex_hash->{d};
3738                 $bootindex_hash->{d} += 1;
3739             }
3740         } else {
3741             if ($bootindex_hash->{c}) {
3742                 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3743                 $bootindex_hash->{c} += 1;
3744             }
3745         }
3746
3747         if($drive->{interface} eq 'virtio'){
3748            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3749         }
3750
3751         if ($drive->{interface} eq 'scsi') {
3752
3753             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3754
3755             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3756             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3757
3758             my $iothread = '';
3759             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3760                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3761                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3762             } elsif ($drive->{iothread}) {
3763                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3764             }
3765
3766             my $queues = '';
3767             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3768                 $queues = ",num_queues=$drive->{queues}";
3769             }
3770
3771             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3772             $scsicontroller->{$controller}=1;
3773         }
3774
3775         if ($drive->{interface} eq 'sata') {
3776            my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3777            $pciaddr = print_pci_addr("ahci$controller", $bridges);
3778            push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3779            $ahcicontroller->{$controller}=1;
3780         }
3781
3782         my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3783         push @$devices, '-drive',$drive_cmd;
3784         push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3785     });
3786
3787     for (my $i = 0; $i < $MAX_NETS; $i++) {
3788          next if !$conf->{"net$i"};
3789          my $d = parse_net($conf->{"net$i"});
3790          next if !$d;
3791
3792          $use_virtio = 1 if $d->{model} eq 'virtio';
3793
3794          if ($bootindex_hash->{n}) {
3795             $d->{bootindex} = $bootindex_hash->{n};
3796             $bootindex_hash->{n} += 1;
3797          }
3798
3799          my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3800          push @$devices, '-netdev', $netdevfull;
3801
3802          my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files);
3803          push @$devices, '-device', $netdevicefull;
3804     }
3805
3806     if (!$q35) {
3807         # add pci bridges
3808         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3809            $bridges->{1} = 1;
3810            $bridges->{2} = 1;
3811         }
3812
3813         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3814
3815         while (my ($k, $v) = each %$bridges) {
3816             $pciaddr = print_pci_addr("pci.$k");
3817             unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3818         }
3819     }
3820
3821     # add custom args
3822     if ($conf->{args}) {
3823         my $aa = PVE::Tools::split_args($conf->{args});
3824         push @$cmd, @$aa;
3825     }
3826
3827     push @$cmd, @$devices;
3828     push @$cmd, '-rtc', join(',', @$rtcFlags)
3829         if scalar(@$rtcFlags);
3830     push @$cmd, '-machine', join(',', @$machineFlags)
3831         if scalar(@$machineFlags);
3832     push @$cmd, '-global', join(',', @$globalFlags)
3833         if scalar(@$globalFlags);
3834
3835     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3836 }
3837
3838 sub vnc_socket {
3839     my ($vmid) = @_;
3840     return "${var_run_tmpdir}/$vmid.vnc";
3841 }
3842
3843 sub spice_port {
3844     my ($vmid) = @_;
3845
3846     my $res = vm_mon_cmd($vmid, 'query-spice');
3847
3848     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3849 }
3850
3851 sub qmp_socket {
3852     my ($vmid, $qga, $name) = @_;
3853     my $sockettype = $qga ? 'qga' : 'qmp';
3854     my $ext = $name ? '-'.$name : '';
3855     return "${var_run_tmpdir}/$vmid$ext.$sockettype";
3856 }
3857
3858 sub pidfile_name {
3859     my ($vmid) = @_;
3860     return "${var_run_tmpdir}/$vmid.pid";
3861 }
3862
3863 sub vm_devices_list {
3864     my ($vmid) = @_;
3865
3866     my $res = vm_mon_cmd($vmid, 'query-pci');
3867     my $devices_to_check = [];
3868     my $devices = {};
3869     foreach my $pcibus (@$res) {
3870         push @$devices_to_check, @{$pcibus->{devices}},
3871     }
3872
3873     while (@$devices_to_check) {
3874         my $to_check = [];
3875         for my $d (@$devices_to_check) {
3876             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3877             next if !$d->{'pci_bridge'};
3878
3879             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3880             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3881         }
3882         $devices_to_check = $to_check;
3883     }
3884
3885     my $resblock = vm_mon_cmd($vmid, 'query-block');
3886     foreach my $block (@$resblock) {
3887         if($block->{device} =~ m/^drive-(\S+)/){
3888                 $devices->{$1} = 1;
3889         }
3890     }
3891
3892     my $resmice = vm_mon_cmd($vmid, 'query-mice');
3893     foreach my $mice (@$resmice) {
3894         if ($mice->{name} eq 'QEMU HID Tablet') {
3895             $devices->{tablet} = 1;
3896             last;
3897         }
3898     }
3899
3900     # for usb devices there is no query-usb
3901     # but we can iterate over the entries in
3902     # qom-list path=/machine/peripheral
3903     my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3904     foreach my $per (@$resperipheral) {
3905         if ($per->{name} =~ m/^usb\d+$/) {
3906             $devices->{$per->{name}} = 1;
3907         }
3908     }
3909
3910     return $devices;
3911 }
3912
3913 sub vm_deviceplug {
3914     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3915
3916     my $q35 = machine_type_is_q35($conf);
3917
3918     my $devices_list = vm_devices_list($vmid);
3919     return 1 if defined($devices_list->{$deviceid});
3920
3921     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3922
3923     if ($deviceid eq 'tablet') {
3924
3925         qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3926
3927     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3928
3929         die "usb hotplug currently not reliable\n";
3930         # since we can't reliably hot unplug all added usb devices
3931         # and usb passthrough disables live migration
3932         # we disable usb hotplugging for now
3933         qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3934
3935     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3936
3937         qemu_iothread_add($vmid, $deviceid, $device);
3938
3939         qemu_driveadd($storecfg, $vmid, $device);
3940         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3941
3942         qemu_deviceadd($vmid, $devicefull);
3943         eval { qemu_deviceaddverify($vmid, $deviceid); };
3944         if (my $err = $@) {
3945             eval { qemu_drivedel($vmid, $deviceid); };
3946             warn $@ if $@;
3947             die $err;
3948         }
3949
3950     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3951
3952
3953         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3954         my $pciaddr = print_pci_addr($deviceid);
3955         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3956
3957         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3958
3959         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3960             qemu_iothread_add($vmid, $deviceid, $device);
3961             $devicefull .= ",iothread=iothread-$deviceid";
3962         }
3963
3964         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3965             $devicefull .= ",num_queues=$device->{queues}";
3966         }
3967
3968         qemu_deviceadd($vmid, $devicefull);
3969         qemu_deviceaddverify($vmid, $deviceid);
3970
3971     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3972
3973         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3974         qemu_driveadd($storecfg, $vmid, $device);
3975
3976         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3977         eval { qemu_deviceadd($vmid, $devicefull); };
3978         if (my $err = $@) {
3979             eval { qemu_drivedel($vmid, $deviceid); };
3980             warn $@ if $@;
3981             die $err;
3982         }
3983
3984     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3985
3986         return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3987
3988         my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
3989         my $use_old_bios_files = undef;
3990         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3991
3992         my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files);
3993         qemu_deviceadd($vmid, $netdevicefull);
3994         eval { qemu_deviceaddverify($vmid, $deviceid); };
3995         if (my $err = $@) {
3996             eval { qemu_netdevdel($vmid, $deviceid); };
3997             warn $@ if $@;
3998             die $err;
3999         }
4000
4001     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4002
4003         my $bridgeid = $2;
4004         my $pciaddr = print_pci_addr($deviceid);
4005         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4006
4007         qemu_deviceadd($vmid, $devicefull);
4008         qemu_deviceaddverify($vmid, $deviceid);
4009
4010     } else {
4011         die "can't hotplug device '$deviceid'\n";
4012     }
4013
4014     return 1;
4015 }
4016
4017 # fixme: this should raise exceptions on error!
4018 sub vm_deviceunplug {
4019     my ($vmid, $conf, $deviceid) = @_;
4020
4021     my $devices_list = vm_devices_list($vmid);
4022     return 1 if !defined($devices_list->{$deviceid});
4023
4024     die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4025
4026     if ($deviceid eq 'tablet') {
4027
4028         qemu_devicedel($vmid, $deviceid);
4029
4030     } elsif ($deviceid =~ m/^usb\d+$/) {
4031
4032         die "usb hotplug currently not reliable\n";
4033         # when unplugging usb devices this way,
4034         # there may be remaining usb controllers/hubs
4035         # so we disable it for now
4036         qemu_devicedel($vmid, $deviceid);
4037         qemu_devicedelverify($vmid, $deviceid);
4038
4039     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4040
4041         qemu_devicedel($vmid, $deviceid);
4042         qemu_devicedelverify($vmid, $deviceid);
4043         qemu_drivedel($vmid, $deviceid);
4044         qemu_iothread_del($conf, $vmid, $deviceid);
4045
4046     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4047
4048         qemu_devicedel($vmid, $deviceid);
4049         qemu_devicedelverify($vmid, $deviceid);
4050         qemu_iothread_del($conf, $vmid, $deviceid);
4051
4052     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4053
4054         qemu_devicedel($vmid, $deviceid);
4055         qemu_drivedel($vmid, $deviceid);
4056         qemu_deletescsihw($conf, $vmid, $deviceid);
4057
4058     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4059
4060         qemu_devicedel($vmid, $deviceid);
4061         qemu_devicedelverify($vmid, $deviceid);
4062         qemu_netdevdel($vmid, $deviceid);
4063
4064     } else {
4065         die "can't unplug device '$deviceid'\n";
4066     }
4067
4068     return 1;
4069 }
4070
4071 sub qemu_deviceadd {
4072     my ($vmid, $devicefull) = @_;
4073
4074     $devicefull = "driver=".$devicefull;
4075     my %options =  split(/[=,]/, $devicefull);
4076
4077     vm_mon_cmd($vmid, "device_add" , %options);
4078 }
4079
4080 sub qemu_devicedel {
4081     my ($vmid, $deviceid) = @_;
4082
4083     my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4084 }
4085
4086 sub qemu_iothread_add {
4087     my($vmid, $deviceid, $device) = @_;
4088
4089     if ($device->{iothread}) {
4090         my $iothreads = vm_iothreads_list($vmid);
4091         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4092     }
4093 }
4094
4095 sub qemu_iothread_del {
4096     my($conf, $vmid, $deviceid) = @_;
4097
4098     my $device = parse_drive($deviceid, $conf->{$deviceid});
4099     if ($device->{iothread}) {
4100         my $iothreads = vm_iothreads_list($vmid);
4101         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4102     }
4103 }
4104
4105 sub qemu_objectadd {
4106     my($vmid, $objectid, $qomtype) = @_;
4107
4108     vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4109
4110     return 1;
4111 }
4112
4113 sub qemu_objectdel {
4114     my($vmid, $objectid) = @_;
4115
4116     vm_mon_cmd($vmid, "object-del", id => $objectid);
4117
4118     return 1;
4119 }
4120
4121 sub qemu_driveadd {
4122     my ($storecfg, $vmid, $device) = @_;
4123
4124     my $drive = print_drive_full($storecfg, $vmid, $device);
4125     $drive =~ s/\\/\\\\/g;
4126     my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4127
4128     # If the command succeeds qemu prints: "OK"
4129     return 1 if $ret =~ m/OK/s;
4130
4131     die "adding drive failed: $ret\n";
4132 }
4133
4134 sub qemu_drivedel {
4135     my($vmid, $deviceid) = @_;
4136
4137     my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4138     $ret =~ s/^\s+//;
4139
4140     return 1 if $ret eq "";
4141
4142     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4143     return 1 if $ret =~ m/Device \'.*?\' not found/s;
4144
4145     die "deleting drive $deviceid failed : $ret\n";
4146 }
4147
4148 sub qemu_deviceaddverify {
4149     my ($vmid, $deviceid) = @_;
4150
4151     for (my $i = 0; $i <= 5; $i++) {
4152          my $devices_list = vm_devices_list($vmid);
4153          return 1 if defined($devices_list->{$deviceid});
4154          sleep 1;
4155     }
4156
4157     die "error on hotplug device '$deviceid'\n";
4158 }
4159
4160
4161 sub qemu_devicedelverify {
4162     my ($vmid, $deviceid) = @_;
4163
4164     # need to verify that the device is correctly removed as device_del
4165     # is async and empty return is not reliable
4166
4167     for (my $i = 0; $i <= 5; $i++) {
4168          my $devices_list = vm_devices_list($vmid);
4169          return 1 if !defined($devices_list->{$deviceid});
4170          sleep 1;
4171     }
4172
4173     die "error on hot-unplugging device '$deviceid'\n";
4174 }
4175
4176 sub qemu_findorcreatescsihw {
4177     my ($storecfg, $conf, $vmid, $device) = @_;
4178
4179     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4180
4181     my $scsihwid="$controller_prefix$controller";
4182     my $devices_list = vm_devices_list($vmid);
4183
4184     if(!defined($devices_list->{$scsihwid})) {
4185         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
4186     }
4187
4188     return 1;
4189 }
4190
4191 sub qemu_deletescsihw {
4192     my ($conf, $vmid, $opt) = @_;
4193
4194     my $device = parse_drive($opt, $conf->{$opt});
4195
4196     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4197         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4198         return 1;
4199     }
4200
4201     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4202
4203     my $devices_list = vm_devices_list($vmid);
4204     foreach my $opt (keys %{$devices_list}) {
4205         if (PVE::QemuServer::is_valid_drivename($opt)) {
4206             my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4207             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4208                 return 1;
4209             }
4210         }
4211     }
4212
4213     my $scsihwid="scsihw$controller";
4214
4215     vm_deviceunplug($vmid, $conf, $scsihwid);
4216
4217     return 1;
4218 }
4219
4220 sub qemu_add_pci_bridge {
4221     my ($storecfg, $conf, $vmid, $device) = @_;
4222
4223     my $bridges = {};
4224
4225     my $bridgeid;
4226
4227     print_pci_addr($device, $bridges);
4228
4229     while (my ($k, $v) = each %$bridges) {
4230         $bridgeid = $k;
4231     }
4232     return 1 if !defined($bridgeid) || $bridgeid < 1;
4233
4234     my $bridge = "pci.$bridgeid";
4235     my $devices_list = vm_devices_list($vmid);
4236
4237     if (!defined($devices_list->{$bridge})) {
4238         vm_deviceplug($storecfg, $conf, $vmid, $bridge);
4239     }
4240
4241     return 1;
4242 }
4243
4244 sub qemu_set_link_status {
4245     my ($vmid, $device, $up) = @_;
4246
4247     vm_mon_cmd($vmid, "set_link", name => $device,
4248                up => $up ? JSON::true : JSON::false);
4249 }
4250
4251 sub qemu_netdevadd {
4252     my ($vmid, $conf, $device, $deviceid) = @_;
4253
4254     my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid, 1);
4255     my %options =  split(/[=,]/, $netdev);
4256
4257     vm_mon_cmd($vmid, "netdev_add",  %options);
4258     return 1;
4259 }
4260
4261 sub qemu_netdevdel {
4262     my ($vmid, $deviceid) = @_;
4263
4264     vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4265 }
4266
4267 sub qemu_usb_hotplug {
4268     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
4269
4270     return if !$device;
4271
4272     # remove the old one first
4273     vm_deviceunplug($vmid, $conf, $deviceid);
4274
4275     # check if xhci controller is necessary and available
4276     if ($device->{usb3}) {
4277
4278         my $devicelist = vm_devices_list($vmid);
4279
4280         if (!$devicelist->{xhci}) {
4281             my $pciaddr = print_pci_addr("xhci");
4282             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4283         }
4284     }
4285     my $d = parse_usb_device($device->{host});
4286     $d->{usb3} = $device->{usb3};
4287
4288     # add the new one
4289     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d);
4290 }
4291
4292 sub qemu_cpu_hotplug {
4293     my ($vmid, $conf, $vcpus) = @_;
4294
4295     my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4296
4297     my $sockets = 1;
4298     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4299     $sockets = $conf->{sockets} if  $conf->{sockets};
4300     my $cores = $conf->{cores} || 1;
4301     my $maxcpus = $sockets * $cores;
4302
4303     $vcpus = $maxcpus if !$vcpus;
4304
4305     die "you can't add more vcpus than maxcpus\n"
4306         if $vcpus > $maxcpus;
4307
4308     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4309
4310     if ($vcpus < $currentvcpus) {
4311
4312         if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4313
4314             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4315                 qemu_devicedel($vmid, "cpu$i");
4316                 my $retry = 0;
4317                 my $currentrunningvcpus = undef;
4318                 while (1) {
4319                     $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4320                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4321                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4322                     $retry++;
4323                     sleep 1;
4324                 }
4325                 #update conf after each succesfull cpu unplug
4326                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4327                 PVE::QemuConfig->write_config($vmid, $conf);
4328             }
4329         } else {
4330             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4331         }
4332
4333         return;
4334     }
4335
4336     my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4337     die "vcpus in running vm does not match its configuration\n"
4338         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4339
4340     if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4341
4342         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4343             my $cpustr = print_cpu_device($conf, $i);
4344             qemu_deviceadd($vmid, $cpustr);
4345
4346             my $retry = 0;
4347             my $currentrunningvcpus = undef;
4348             while (1) {
4349                 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4350                 last if scalar(@{$currentrunningvcpus}) == $i;
4351                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4352                 sleep 1;
4353                 $retry++;
4354             }
4355             #update conf after each succesfull cpu hotplug
4356             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4357             PVE::QemuConfig->write_config($vmid, $conf);
4358         }
4359     } else {
4360
4361         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4362             vm_mon_cmd($vmid, "cpu-add", id => int($i));
4363         }
4364     }
4365 }
4366
4367 sub qemu_block_set_io_throttle {
4368     my ($vmid, $deviceid,
4369         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4370         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4371         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4372         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4373
4374     return if !check_running($vmid) ;
4375
4376     vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4377         bps => int($bps),
4378         bps_rd => int($bps_rd),
4379         bps_wr => int($bps_wr),
4380         iops => int($iops),
4381         iops_rd => int($iops_rd),
4382         iops_wr => int($iops_wr),
4383         bps_max => int($bps_max),
4384         bps_rd_max => int($bps_rd_max),
4385         bps_wr_max => int($bps_wr_max),
4386         iops_max => int($iops_max),
4387         iops_rd_max => int($iops_rd_max),
4388         iops_wr_max => int($iops_wr_max),
4389         bps_max_length => int($bps_max_length),
4390         bps_rd_max_length => int($bps_rd_max_length),
4391         bps_wr_max_length => int($bps_wr_max_length),
4392         iops_max_length => int($iops_max_length),
4393         iops_rd_max_length => int($iops_rd_max_length),
4394         iops_wr_max_length => int($iops_wr_max_length),
4395     );
4396
4397 }
4398
4399 # old code, only used to shutdown old VM after update
4400 sub __read_avail {
4401     my ($fh, $timeout) = @_;
4402
4403     my $sel = new IO::Select;
4404     $sel->add($fh);
4405
4406     my $res = '';
4407     my $buf;
4408
4409     my @ready;
4410     while (scalar (@ready = $sel->can_read($timeout))) {
4411         my $count;
4412         if ($count = $fh->sysread($buf, 8192)) {
4413             if ($buf =~ /^(.*)\(qemu\) $/s) {
4414                 $res .= $1;
4415                 last;
4416             } else {
4417                 $res .= $buf;
4418             }
4419         } else {
4420             if (!defined($count)) {
4421                 die "$!\n";
4422             }
4423             last;
4424         }
4425     }
4426
4427     die "monitor read timeout\n" if !scalar(@ready);
4428
4429     return $res;
4430 }
4431
4432 sub qemu_block_resize {
4433     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4434
4435     my $running = check_running($vmid);
4436
4437     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4438
4439     return if !$running;
4440
4441     vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4442
4443 }
4444
4445 sub qemu_volume_snapshot {
4446     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4447
4448     my $running = check_running($vmid);
4449
4450     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4451         vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4452     } else {
4453         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4454     }
4455 }
4456
4457 sub qemu_volume_snapshot_delete {
4458     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4459
4460     my $running = check_running($vmid);
4461
4462     if($running) {
4463
4464         $running = undef;
4465         my $conf = PVE::QemuConfig->load_config($vmid);
4466         foreach_drive($conf, sub {
4467             my ($ds, $drive) = @_;
4468             $running = 1 if $drive->{file} eq $volid;
4469         });
4470     }
4471
4472     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4473         vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4474     } else {
4475         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4476     }
4477 }
4478
4479 sub set_migration_caps {
4480     my ($vmid) = @_;
4481
4482     my $cap_ref = [];
4483
4484     my $enabled_cap = {
4485         "auto-converge" => 1,
4486         "xbzrle" => 1,
4487         "x-rdma-pin-all" => 0,
4488         "zero-blocks" => 0,
4489         "compress" => 0
4490     };
4491
4492     my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4493
4494     for my $supported_capability (@$supported_capabilities) {
4495         push @$cap_ref, {
4496             capability => $supported_capability->{capability},
4497             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4498         };
4499     }
4500
4501     vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4502 }
4503
4504 my $fast_plug_option = {
4505     'lock' => 1,
4506     'name' => 1,
4507     'onboot' => 1,
4508     'shares' => 1,
4509     'startup' => 1,
4510     'description' => 1,
4511     'protection' => 1,
4512     'vmstatestorage' => 1,
4513 };
4514
4515 # hotplug changes in [PENDING]
4516 # $selection hash can be used to only apply specified options, for
4517 # example: { cores => 1 } (only apply changed 'cores')
4518 # $errors ref is used to return error messages
4519 sub vmconfig_hotplug_pending {
4520     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4521
4522     my $defaults = load_defaults();
4523
4524     # commit values which do not have any impact on running VM first
4525     # Note: those option cannot raise errors, we we do not care about
4526     # $selection and always apply them.
4527
4528     my $add_error = sub {
4529         my ($opt, $msg) = @_;
4530         $errors->{$opt} = "hotplug problem - $msg";
4531     };
4532
4533     my $changes = 0;
4534     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4535         if ($fast_plug_option->{$opt}) {
4536             $conf->{$opt} = $conf->{pending}->{$opt};
4537             delete $conf->{pending}->{$opt};
4538             $changes = 1;
4539         }
4540     }
4541
4542     if ($changes) {
4543         PVE::QemuConfig->write_config($vmid, $conf);
4544         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4545     }
4546
4547     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4548
4549     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4550     while (my ($opt, $force) = each %$pending_delete_hash) {
4551         next if $selection && !$selection->{$opt};
4552         eval {
4553             if ($opt eq 'hotplug') {
4554                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4555             } elsif ($opt eq 'tablet') {
4556                 die "skip\n" if !$hotplug_features->{usb};
4557                 if ($defaults->{tablet}) {
4558                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4559                 } else {
4560                     vm_deviceunplug($vmid, $conf, $opt);
4561                 }
4562             } elsif ($opt =~ m/^usb\d+/) {
4563                 die "skip\n";
4564                 # since we cannot reliably hot unplug usb devices
4565                 # we are disabling it
4566                 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4567                 vm_deviceunplug($vmid, $conf, $opt);
4568             } elsif ($opt eq 'vcpus') {
4569                 die "skip\n" if !$hotplug_features->{cpu};
4570                 qemu_cpu_hotplug($vmid, $conf, undef);
4571             } elsif ($opt eq 'balloon') {
4572                 # enable balloon device is not hotpluggable
4573                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4574                 # here we reset the ballooning value to memory
4575                 my $balloon = $conf->{memory} || $defaults->{memory};
4576                 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4577             } elsif ($fast_plug_option->{$opt}) {
4578                 # do nothing
4579             } elsif ($opt =~ m/^net(\d+)$/) {
4580                 die "skip\n" if !$hotplug_features->{network};
4581                 vm_deviceunplug($vmid, $conf, $opt);
4582             } elsif (is_valid_drivename($opt)) {
4583                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4584                 vm_deviceunplug($vmid, $conf, $opt);
4585                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4586             } elsif ($opt =~ m/^memory$/) {
4587                 die "skip\n" if !$hotplug_features->{memory};
4588                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4589             } elsif ($opt eq 'cpuunits') {
4590                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4591             } elsif ($opt eq 'cpulimit') {
4592                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4593             } else {
4594                 die "skip\n";
4595             }
4596         };
4597         if (my $err = $@) {
4598             &$add_error($opt, $err) if $err ne "skip\n";
4599         } else {
4600             # save new config if hotplug was successful
4601             delete $conf->{$opt};
4602             vmconfig_undelete_pending_option($conf, $opt);
4603             PVE::QemuConfig->write_config($vmid, $conf);
4604             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4605         }
4606     }
4607
4608     my $apply_pending_cloudinit;
4609     $apply_pending_cloudinit = sub {
4610         my ($key, $value) = @_;
4611         $apply_pending_cloudinit = sub {}; # once is enough
4612
4613         my @cloudinit_opts = keys %$confdesc_cloudinit;
4614         foreach my $opt (keys %{$conf->{pending}}) {
4615             next if !grep { $_ eq $opt } @cloudinit_opts;
4616             $conf->{$opt} = delete $conf->{pending}->{$opt};
4617         }
4618
4619         my $new_conf = { %$conf };
4620         $new_conf->{$key} = $value;
4621         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4622     };
4623
4624     foreach my $opt (keys %{$conf->{pending}}) {
4625         next if $selection && !$selection->{$opt};
4626         my $value = $conf->{pending}->{$opt};
4627         eval {
4628             if ($opt eq 'hotplug') {
4629                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4630             } elsif ($opt eq 'tablet') {
4631                 die "skip\n" if !$hotplug_features->{usb};
4632                 if ($value == 1) {
4633                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4634                 } elsif ($value == 0) {
4635                     vm_deviceunplug($vmid, $conf, $opt);
4636                 }
4637             } elsif ($opt =~ m/^usb\d+$/) {
4638                 die "skip\n";
4639                 # since we cannot reliably hot unplug usb devices
4640                 # we are disabling it
4641                 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4642                 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4643                 die "skip\n" if !$d;
4644                 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d);
4645             } elsif ($opt eq 'vcpus') {
4646                 die "skip\n" if !$hotplug_features->{cpu};
4647                 qemu_cpu_hotplug($vmid, $conf, $value);
4648             } elsif ($opt eq 'balloon') {
4649                 # enable/disable balloning device is not hotpluggable
4650                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4651                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4652                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4653
4654                 # allow manual ballooning if shares is set to zero
4655                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4656                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4657                     vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4658                 }
4659             } elsif ($opt =~ m/^net(\d+)$/) {
4660                 # some changes can be done without hotplug
4661                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4662                                     $vmid, $opt, $value);
4663             } elsif (is_valid_drivename($opt)) {
4664                 # some changes can be done without hotplug
4665                 my $drive = parse_drive($opt, $value);
4666                 if (drive_is_cloudinit($drive)) {
4667                     &$apply_pending_cloudinit($opt, $value);
4668                 }
4669                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4670                                      $vmid, $opt, $value, 1);
4671             } elsif ($opt =~ m/^memory$/) { #dimms
4672                 die "skip\n" if !$hotplug_features->{memory};
4673                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4674             } elsif ($opt eq 'cpuunits') {
4675                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4676             } elsif ($opt eq 'cpulimit') {
4677                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4678                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4679             } else {
4680                 die "skip\n";  # skip non-hot-pluggable options
4681             }
4682         };
4683         if (my $err = $@) {
4684             &$add_error($opt, $err) if $err ne "skip\n";
4685         } else {
4686             # save new config if hotplug was successful
4687             $conf->{$opt} = $value;
4688             delete $conf->{pending}->{$opt};
4689             PVE::QemuConfig->write_config($vmid, $conf);
4690             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4691         }
4692     }
4693 }
4694
4695 sub try_deallocate_drive {
4696     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4697
4698     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4699         my $volid = $drive->{file};
4700         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4701             my $sid = PVE::Storage::parse_volume_id($volid);
4702             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4703
4704             # check if the disk is really unused
4705             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4706                 if is_volume_in_use($storecfg, $conf, $key, $volid);
4707             PVE::Storage::vdisk_free($storecfg, $volid);
4708             return 1;
4709         } else {
4710             # If vm is not owner of this disk remove from config
4711             return 1;
4712         }
4713     }
4714
4715     return undef;
4716 }
4717
4718 sub vmconfig_delete_or_detach_drive {
4719     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4720
4721     my $drive = parse_drive($opt, $conf->{$opt});
4722
4723     my $rpcenv = PVE::RPCEnvironment::get();
4724     my $authuser = $rpcenv->get_user();
4725
4726     if ($force) {
4727         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4728         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4729     } else {
4730         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4731     }
4732 }
4733
4734 sub vmconfig_apply_pending {
4735     my ($vmid, $conf, $storecfg) = @_;
4736
4737     # cold plug
4738
4739     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4740     while (my ($opt, $force) = each %$pending_delete_hash) {
4741         die "internal error" if $opt =~ m/^unused/;
4742         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4743         if (!defined($conf->{$opt})) {
4744             vmconfig_undelete_pending_option($conf, $opt);
4745             PVE::QemuConfig->write_config($vmid, $conf);
4746         } elsif (is_valid_drivename($opt)) {
4747             vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4748             vmconfig_undelete_pending_option($conf, $opt);
4749             delete $conf->{$opt};
4750             PVE::QemuConfig->write_config($vmid, $conf);
4751         } else {
4752             vmconfig_undelete_pending_option($conf, $opt);
4753             delete $conf->{$opt};
4754             PVE::QemuConfig->write_config($vmid, $conf);
4755         }
4756     }
4757
4758     $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4759
4760     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4761         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4762
4763         if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
4764             # skip if nothing changed
4765         } elsif (is_valid_drivename($opt)) {
4766             vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4767                 if defined($conf->{$opt});
4768             $conf->{$opt} = $conf->{pending}->{$opt};
4769         } else {
4770             $conf->{$opt} = $conf->{pending}->{$opt};
4771         }
4772
4773         delete $conf->{pending}->{$opt};
4774         PVE::QemuConfig->write_config($vmid, $conf);
4775     }
4776 }
4777
4778 my $safe_num_ne = sub {
4779     my ($a, $b) = @_;
4780
4781     return 0 if !defined($a) && !defined($b);
4782     return 1 if !defined($a);
4783     return 1 if !defined($b);
4784
4785     return $a != $b;
4786 };
4787
4788 my $safe_string_ne = sub {
4789     my ($a, $b) = @_;
4790
4791     return 0 if !defined($a) && !defined($b);
4792     return 1 if !defined($a);
4793     return 1 if !defined($b);
4794
4795     return $a ne $b;
4796 };
4797
4798 sub vmconfig_update_net {
4799     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4800
4801     my $newnet = parse_net($value);
4802
4803     if ($conf->{$opt}) {
4804         my $oldnet = parse_net($conf->{$opt});
4805
4806         if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4807             &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4808             &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4809             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4810
4811             # for non online change, we try to hot-unplug
4812             die "skip\n" if !$hotplug;
4813             vm_deviceunplug($vmid, $conf, $opt);
4814         } else {
4815
4816             die "internal error" if $opt !~ m/net(\d+)/;
4817             my $iface = "tap${vmid}i$1";
4818
4819             if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4820                 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4821                 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4822                 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4823                 PVE::Network::tap_unplug($iface);
4824                 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4825             } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4826                 # Rate can be applied on its own but any change above needs to
4827                 # include the rate in tap_plug since OVS resets everything.
4828                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4829             }
4830
4831             if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4832                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4833             }
4834
4835             return 1;
4836         }
4837     }
4838
4839     if ($hotplug) {
4840         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4841     } else {
4842         die "skip\n";
4843     }
4844 }
4845
4846 sub vmconfig_update_disk {
4847     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4848
4849     # fixme: do we need force?
4850
4851     my $drive = parse_drive($opt, $value);
4852
4853     if ($conf->{$opt}) {
4854
4855         if (my $old_drive = parse_drive($opt, $conf->{$opt}))  {
4856
4857             my $media = $drive->{media} || 'disk';
4858             my $oldmedia = $old_drive->{media} || 'disk';
4859             die "unable to change media type\n" if $media ne $oldmedia;
4860
4861             if (!drive_is_cdrom($old_drive)) {
4862
4863                 if ($drive->{file} ne $old_drive->{file}) {
4864
4865                     die "skip\n" if !$hotplug;
4866
4867                     # unplug and register as unused
4868                     vm_deviceunplug($vmid, $conf, $opt);
4869                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4870
4871                 } else {
4872                     # update existing disk
4873
4874                     # skip non hotpluggable value
4875                     if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4876                         &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4877                         &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4878                         &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4879                         die "skip\n";
4880                     }
4881
4882                     # apply throttle
4883                     if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4884                         &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4885                         &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4886                         &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4887                         &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4888                         &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4889                         &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4890                         &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4891                         &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4892                         &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4893                         &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4894                         &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4895                         &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4896                         &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4897                         &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4898                         &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4899                         &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4900                         &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4901
4902                         qemu_block_set_io_throttle($vmid,"drive-$opt",
4903                                                    ($drive->{mbps} || 0)*1024*1024,
4904                                                    ($drive->{mbps_rd} || 0)*1024*1024,
4905                                                    ($drive->{mbps_wr} || 0)*1024*1024,
4906                                                    $drive->{iops} || 0,
4907                                                    $drive->{iops_rd} || 0,
4908                                                    $drive->{iops_wr} || 0,
4909                                                    ($drive->{mbps_max} || 0)*1024*1024,
4910                                                    ($drive->{mbps_rd_max} || 0)*1024*1024,
4911                                                    ($drive->{mbps_wr_max} || 0)*1024*1024,
4912                                                    $drive->{iops_max} || 0,
4913                                                    $drive->{iops_rd_max} || 0,
4914                                                    $drive->{iops_wr_max} || 0,
4915                                                    $drive->{bps_max_length} || 1,
4916                                                    $drive->{bps_rd_max_length} || 1,
4917                                                    $drive->{bps_wr_max_length} || 1,
4918                                                    $drive->{iops_max_length} || 1,
4919                                                    $drive->{iops_rd_max_length} || 1,
4920                                                    $drive->{iops_wr_max_length} || 1);
4921
4922                     }
4923
4924                     return 1;
4925                 }
4926
4927             } else { # cdrom
4928
4929                 if ($drive->{file} eq 'none') {
4930                     vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4931                     if (drive_is_cloudinit($old_drive)) {
4932                         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4933                     }
4934                 } else {
4935                     my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4936                     vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4937                     vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4938                 }
4939
4940                 return 1;
4941             }
4942         }
4943     }
4944
4945     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4946     # hotplug new disks
4947     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4948     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4949 }
4950
4951 sub vm_start {
4952     my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
4953         $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
4954
4955     PVE::QemuConfig->lock_config($vmid, sub {
4956         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
4957
4958         die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4959
4960         PVE::QemuConfig->check_lock($conf) if !$skiplock;
4961
4962         die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4963
4964         if (!$statefile && scalar(keys %{$conf->{pending}})) {
4965             vmconfig_apply_pending($vmid, $conf, $storecfg);
4966             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4967         }
4968
4969         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4970
4971         my $defaults = load_defaults();
4972
4973         # set environment variable useful inside network script
4974         $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4975
4976         my $local_volumes = {};
4977
4978         if ($targetstorage) {
4979             foreach_drive($conf, sub {
4980                 my ($ds, $drive) = @_;
4981
4982                 return if drive_is_cdrom($drive);
4983
4984                 my $volid = $drive->{file};
4985
4986                 return if !$volid;
4987
4988                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
4989
4990                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
4991                 return if $scfg->{shared};
4992                 $local_volumes->{$ds} = [$volid, $storeid, $volname];
4993             });
4994
4995             my $format = undef;
4996
4997             foreach my $opt (sort keys %$local_volumes) {
4998
4999                 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5000                 my $drive = parse_drive($opt, $conf->{$opt});
5001
5002                 #if remote storage is specified, use default format
5003                 if ($targetstorage && $targetstorage ne "1") {
5004                     $storeid = $targetstorage;
5005                     my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5006                     $format = $defFormat;
5007                 } else {
5008                     #else we use same format than original
5009                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5010                     $format = qemu_img_format($scfg, $volid);
5011                 }
5012
5013                 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5014                 my $newdrive = $drive;
5015                 $newdrive->{format} = $format;
5016                 $newdrive->{file} = $newvolid;
5017                 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5018                 $local_volumes->{$opt} = $drivestr;
5019                 #pass drive to conf for command line
5020                 $conf->{$opt} = $drivestr;
5021             }
5022         }
5023
5024         my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5025
5026         my $migrate_port = 0;
5027         my $migrate_uri;
5028         if ($statefile) {
5029             if ($statefile eq 'tcp') {
5030                 my $localip = "localhost";
5031                 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5032                 my $nodename = PVE::INotify::nodename();
5033
5034                 if (!defined($migration_type)) {
5035                     if (defined($datacenterconf->{migration}->{type})) {
5036                         $migration_type = $datacenterconf->{migration}->{type};
5037                     } else {
5038                         $migration_type = 'secure';
5039                     }
5040                 }
5041
5042                 if ($migration_type eq 'insecure') {
5043                     my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5044                     if ($migrate_network_addr) {
5045                         $localip = $migrate_network_addr;
5046                     } else {
5047                         $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5048                     }
5049
5050                     $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5051                 }
5052
5053                 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5054                 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5055                 $migrate_uri = "tcp:${localip}:${migrate_port}";
5056                 push @$cmd, '-incoming', $migrate_uri;
5057                 push @$cmd, '-S';
5058
5059             } elsif ($statefile eq 'unix') {
5060                 # should be default for secure migrations as a ssh TCP forward
5061                 # tunnel is not deterministic reliable ready and fails regurarly
5062                 # to set up in time, so use UNIX socket forwards
5063                 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5064                 unlink $socket_addr;
5065
5066                 $migrate_uri = "unix:$socket_addr";
5067
5068                 push @$cmd, '-incoming', $migrate_uri;
5069                 push @$cmd, '-S';
5070
5071             } else {
5072                 push @$cmd, '-loadstate', $statefile;
5073             }
5074         } elsif ($paused) {
5075             push @$cmd, '-S';
5076         }
5077
5078         # host pci devices
5079         for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
5080           my $d = parse_hostpci($conf->{"hostpci$i"});
5081           next if !$d;
5082           my $pcidevices = $d->{pciid};
5083           foreach my $pcidevice (@$pcidevices) {
5084                 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
5085
5086                 my $info = pci_device_info("0000:$pciid");
5087                 die "IOMMU not present\n" if !check_iommu_support();
5088                 die "no pci device info for device '$pciid'\n" if !$info;
5089                 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
5090                 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
5091           }
5092         }
5093
5094         PVE::Storage::activate_volumes($storecfg, $vollist);
5095
5096         if (!check_running($vmid, 1)) {
5097             eval {
5098                 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5099                     outfunc => sub {}, errfunc => sub {});
5100             };
5101         }
5102
5103         my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5104                                                   : $defaults->{cpuunits};
5105
5106         my $start_timeout = $conf->{hugepages} ? 300 : 30;
5107         my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5108
5109         my %properties = (
5110             Slice => 'qemu.slice',
5111             KillMode => 'none',
5112             CPUShares => $cpuunits
5113         );
5114
5115         if (my $cpulimit = $conf->{cpulimit}) {
5116             $properties{CPUQuota} = int($cpulimit * 100);
5117         }
5118         $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5119
5120         my $run_qemu = sub {
5121             PVE::Tools::run_fork sub {
5122                 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5123                 run_command($cmd, %run_params);
5124             };
5125         };
5126
5127         if ($conf->{hugepages}) {
5128
5129             my $code = sub {
5130                 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5131                 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5132
5133                 PVE::QemuServer::Memory::hugepages_mount();
5134                 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5135
5136                 eval { $run_qemu->() };
5137                 if (my $err = $@) {
5138                     PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5139                     die $err;
5140                 }
5141
5142                 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5143             };
5144             eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5145
5146         } else {
5147             eval { $run_qemu->() };
5148         }
5149
5150         if (my $err = $@) {
5151             # deactivate volumes if start fails
5152             eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5153             die "start failed: $err";
5154         }
5155
5156         print "migration listens on $migrate_uri\n" if $migrate_uri;
5157
5158         if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5159             eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5160             warn $@ if $@;
5161         }
5162
5163         #start nbd server for storage migration
5164         if ($targetstorage) {
5165             my $nodename = PVE::INotify::nodename();
5166             my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5167             my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5168             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5169             $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5170
5171             vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${migrate_port}" } } );
5172
5173             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5174
5175             foreach my $opt (sort keys %$local_volumes) {
5176                 my $volid = $local_volumes->{$opt};
5177                 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5178                 my $migrate_storage_uri = "nbd:${localip}:${migrate_port}:exportname=drive-$opt";
5179                 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5180             }
5181         }
5182
5183         if ($migratedfrom) {
5184             eval {
5185                 set_migration_caps($vmid);
5186             };
5187             warn $@ if $@;
5188
5189             if ($spice_port) {
5190                 print "spice listens on port $spice_port\n";
5191                 if ($spice_ticket) {
5192                     vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5193                     vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5194                 }
5195             }
5196
5197         } else {
5198             vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5199                 if !$statefile && $conf->{balloon};
5200
5201             foreach my $opt (keys %$conf) {
5202                 next if $opt !~  m/^net\d+$/;
5203                 my $nicconf = parse_net($conf->{$opt});
5204                 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5205             }
5206         }
5207
5208         vm_mon_cmd_nocheck($vmid, 'qom-set',
5209                     path => "machine/peripheral/balloon0",
5210                     property => "guest-stats-polling-interval",
5211                     value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5212
5213     });
5214 }
5215
5216 sub vm_mon_cmd {
5217     my ($vmid, $execute, %params) = @_;
5218
5219     my $cmd = { execute => $execute, arguments => \%params };
5220     vm_qmp_command($vmid, $cmd);
5221 }
5222
5223 sub vm_mon_cmd_nocheck {
5224     my ($vmid, $execute, %params) = @_;
5225
5226     my $cmd = { execute => $execute, arguments => \%params };
5227     vm_qmp_command($vmid, $cmd, 1);
5228 }
5229
5230 sub vm_qmp_command {
5231     my ($vmid, $cmd, $nocheck) = @_;
5232
5233     my $res;
5234
5235     my $timeout;
5236     if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
5237         $timeout = $cmd->{arguments}->{timeout};
5238         delete $cmd->{arguments}->{timeout};
5239     }
5240
5241     eval {
5242         die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5243         my $sname = qmp_socket($vmid);
5244         if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5245             my $qmpclient = PVE::QMPClient->new();
5246
5247             $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5248         } else {
5249             die "unable to open monitor socket\n";
5250         }
5251     };
5252     if (my $err = $@) {
5253         syslog("err", "VM $vmid qmp command failed - $err");
5254         die $err;
5255     }
5256
5257     return $res;
5258 }
5259
5260 sub vm_human_monitor_command {
5261     my ($vmid, $cmdline) = @_;
5262
5263     my $res;
5264
5265     my $cmd = {
5266         execute => 'human-monitor-command',
5267         arguments => { 'command-line' => $cmdline},
5268     };
5269
5270     return vm_qmp_command($vmid, $cmd);
5271 }
5272
5273 sub vm_commandline {
5274     my ($storecfg, $vmid) = @_;
5275
5276     my $conf = PVE::QemuConfig->load_config($vmid);
5277
5278     my $defaults = load_defaults();
5279
5280     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5281
5282     return PVE::Tools::cmd2string($cmd);
5283 }
5284
5285 sub vm_reset {
5286     my ($vmid, $skiplock) = @_;
5287
5288     PVE::QemuConfig->lock_config($vmid, sub {
5289
5290         my $conf = PVE::QemuConfig->load_config($vmid);
5291
5292         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5293
5294         vm_mon_cmd($vmid, "system_reset");
5295     });
5296 }
5297
5298 sub get_vm_volumes {
5299     my ($conf) = @_;
5300
5301     my $vollist = [];
5302     foreach_volid($conf, sub {
5303         my ($volid, $attr) = @_;
5304
5305         return if $volid =~ m|^/|;
5306
5307         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5308         return if !$sid;
5309
5310         push @$vollist, $volid;
5311     });
5312
5313     return $vollist;
5314 }
5315
5316 sub vm_stop_cleanup {
5317     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5318
5319     eval {
5320
5321         if (!$keepActive) {
5322             my $vollist = get_vm_volumes($conf);
5323             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5324         }
5325
5326         foreach my $ext (qw(mon qmp pid vnc qga)) {
5327             unlink "/var/run/qemu-server/${vmid}.$ext";
5328         }
5329
5330         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5331     };
5332     warn $@ if $@; # avoid errors - just warn
5333 }
5334
5335 # Note: use $nockeck to skip tests if VM configuration file exists.
5336 # We need that when migration VMs to other nodes (files already moved)
5337 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5338 sub vm_stop {
5339     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5340
5341     $force = 1 if !defined($force) && !$shutdown;
5342
5343     if ($migratedfrom){
5344         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5345         kill 15, $pid if $pid;
5346         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5347         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5348         return;
5349     }
5350
5351     PVE::QemuConfig->lock_config($vmid, sub {
5352
5353         my $pid = check_running($vmid, $nocheck);
5354         return if !$pid;
5355
5356         my $conf;
5357         if (!$nocheck) {
5358             $conf = PVE::QemuConfig->load_config($vmid);
5359             PVE::QemuConfig->check_lock($conf) if !$skiplock;
5360             if (!defined($timeout) && $shutdown && $conf->{startup}) {
5361                 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5362                 $timeout = $opts->{down} if $opts->{down};
5363             }
5364         }
5365
5366         $timeout = 60 if !defined($timeout);
5367
5368         eval {
5369             if ($shutdown) {
5370                 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5371                     vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
5372                 } else {
5373                     vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5374                 }
5375             } else {
5376                 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5377             }
5378         };
5379         my $err = $@;
5380
5381         if (!$err) {
5382             my $count = 0;
5383             while (($count < $timeout) && check_running($vmid, $nocheck)) {
5384                 $count++;
5385                 sleep 1;
5386             }
5387
5388             if ($count >= $timeout) {
5389                 if ($force) {
5390                     warn "VM still running - terminating now with SIGTERM\n";
5391                     kill 15, $pid;
5392                 } else {
5393                     die "VM quit/powerdown failed - got timeout\n";
5394                 }
5395             } else {
5396                 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5397                 return;
5398             }
5399         } else {
5400             if ($force) {
5401                 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5402                 kill 15, $pid;
5403             } else {
5404                 die "VM quit/powerdown failed\n";
5405             }
5406         }
5407
5408         # wait again
5409         $timeout = 10;
5410
5411         my $count = 0;
5412         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5413             $count++;
5414             sleep 1;
5415         }
5416
5417         if ($count >= $timeout) {
5418             warn "VM still running - terminating now with SIGKILL\n";
5419             kill 9, $pid;
5420             sleep 1;
5421         }
5422
5423         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5424    });
5425 }
5426
5427 sub vm_suspend {
5428     my ($vmid, $skiplock) = @_;
5429
5430     PVE::QemuConfig->lock_config($vmid, sub {
5431
5432         my $conf = PVE::QemuConfig->load_config($vmid);
5433
5434         PVE::QemuConfig->check_lock($conf)
5435             if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5436
5437         vm_mon_cmd($vmid, "stop");
5438     });
5439 }
5440
5441 sub vm_resume {
5442     my ($vmid, $skiplock, $nocheck) = @_;
5443
5444     PVE::QemuConfig->lock_config($vmid, sub {
5445
5446         my $res = vm_mon_cmd($vmid, 'query-status');
5447         my $resume_cmd = 'cont';
5448
5449         if ($res->{status} && $res->{status} eq 'suspended') {
5450             $resume_cmd = 'system_wakeup';
5451         }
5452
5453         if (!$nocheck) {
5454
5455             my $conf = PVE::QemuConfig->load_config($vmid);
5456
5457             PVE::QemuConfig->check_lock($conf)
5458                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5459
5460             vm_mon_cmd($vmid, $resume_cmd);
5461
5462         } else {
5463             vm_mon_cmd_nocheck($vmid, $resume_cmd);
5464         }
5465     });
5466 }
5467
5468 sub vm_sendkey {
5469     my ($vmid, $skiplock, $key) = @_;
5470
5471     PVE::QemuConfig->lock_config($vmid, sub {
5472
5473         my $conf = PVE::QemuConfig->load_config($vmid);
5474
5475         # there is no qmp command, so we use the human monitor command
5476         vm_human_monitor_command($vmid, "sendkey $key");
5477     });
5478 }
5479
5480 sub vm_destroy {
5481     my ($storecfg, $vmid, $skiplock) = @_;
5482
5483     PVE::QemuConfig->lock_config($vmid, sub {
5484
5485         my $conf = PVE::QemuConfig->load_config($vmid);
5486
5487         if (!check_running($vmid)) {
5488             destroy_vm($storecfg, $vmid, undef, $skiplock);
5489         } else {
5490             die "VM $vmid is running - destroy failed\n";
5491         }
5492     });
5493 }
5494
5495 # pci helpers
5496
5497 sub file_write {
5498     my ($filename, $buf) = @_;
5499
5500     my $fh = IO::File->new($filename, "w");
5501     return undef if !$fh;
5502
5503     my $res = print $fh $buf;
5504
5505     $fh->close();
5506
5507     return $res;
5508 }
5509
5510 sub pci_device_info {
5511     my ($name) = @_;
5512
5513     my $res;
5514
5515     return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
5516     my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
5517
5518     my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
5519     return undef if !defined($irq) || $irq !~ m/^\d+$/;
5520
5521     my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
5522     return undef if !defined($vendor) || $vendor !~ s/^0x//;
5523
5524     my $product = file_read_firstline("$pcisysfs/devices/$name/device");
5525     return undef if !defined($product) || $product !~ s/^0x//;
5526
5527     $res = {
5528         name => $name,
5529         vendor => $vendor,
5530         product => $product,
5531         domain => $domain,
5532         bus => $bus,
5533         slot => $slot,
5534         func => $func,
5535         irq => $irq,
5536         has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
5537     };
5538
5539     return $res;
5540 }
5541
5542 sub pci_dev_reset {
5543     my ($dev) = @_;
5544
5545     my $name = $dev->{name};
5546
5547     my $fn = "$pcisysfs/devices/$name/reset";
5548
5549     return file_write($fn, "1");
5550 }
5551
5552 sub pci_dev_bind_to_vfio {
5553     my ($dev) = @_;
5554
5555     my $name = $dev->{name};
5556
5557     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5558
5559     if (!-d $vfio_basedir) {
5560         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5561     }
5562     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5563
5564     my $testdir = "$vfio_basedir/$name";
5565     return 1 if -d $testdir;
5566
5567     my $data = "$dev->{vendor} $dev->{product}";
5568     return undef if !file_write("$vfio_basedir/new_id", $data);
5569
5570     my $fn = "$pcisysfs/devices/$name/driver/unbind";
5571     if (!file_write($fn, $name)) {
5572         return undef if -f $fn;
5573     }
5574
5575     $fn = "$vfio_basedir/bind";
5576     if (! -d $testdir) {
5577         return undef if !file_write($fn, $name);
5578     }
5579
5580     return -d $testdir;
5581 }
5582
5583 sub pci_dev_group_bind_to_vfio {
5584     my ($pciid) = @_;
5585
5586     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5587
5588     if (!-d $vfio_basedir) {
5589         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5590     }
5591     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5592
5593     # get IOMMU group devices
5594     opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
5595       my @devs = grep /^0000:/, readdir($D);
5596     closedir($D);
5597
5598     foreach my $pciid (@devs) {
5599         $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
5600
5601         # pci bridges, switches or root ports are not supported
5602         # they have a pci_bus subdirectory so skip them
5603         next if (-e "$pcisysfs/devices/$pciid/pci_bus");
5604
5605         my $info = pci_device_info($1);
5606         pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
5607     }
5608
5609     return 1;
5610 }
5611
5612 # vzdump restore implementaion
5613
5614 sub tar_archive_read_firstfile {
5615     my $archive = shift;
5616
5617     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5618
5619     # try to detect archive type first
5620     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5621         die "unable to open file '$archive'\n";
5622     my $firstfile = <$fh>;
5623     kill 15, $pid;
5624     close $fh;
5625
5626     die "ERROR: archive contaions no data\n" if !$firstfile;
5627     chomp $firstfile;
5628
5629     return $firstfile;
5630 }
5631
5632 sub tar_restore_cleanup {
5633     my ($storecfg, $statfile) = @_;
5634
5635     print STDERR "starting cleanup\n";
5636
5637     if (my $fd = IO::File->new($statfile, "r")) {
5638         while (defined(my $line = <$fd>)) {
5639             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5640                 my $volid = $2;
5641                 eval {
5642                     if ($volid =~ m|^/|) {
5643                         unlink $volid || die 'unlink failed\n';
5644                     } else {
5645                         PVE::Storage::vdisk_free($storecfg, $volid);
5646                     }
5647                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5648                 };
5649                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5650             } else {
5651                 print STDERR "unable to parse line in statfile - $line";
5652             }
5653         }
5654         $fd->close();
5655     }
5656 }
5657
5658 sub restore_archive {
5659     my ($archive, $vmid, $user, $opts) = @_;
5660
5661     my $format = $opts->{format};
5662     my $comp;
5663
5664     if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5665         $format = 'tar' if !$format;
5666         $comp = 'gzip';
5667     } elsif ($archive =~ m/\.tar$/) {
5668         $format = 'tar' if !$format;
5669     } elsif ($archive =~ m/.tar.lzo$/) {
5670         $format = 'tar' if !$format;
5671         $comp = 'lzop';
5672     } elsif ($archive =~ m/\.vma$/) {
5673         $format = 'vma' if !$format;
5674     } elsif ($archive =~ m/\.vma\.gz$/) {
5675         $format = 'vma' if !$format;
5676         $comp = 'gzip';
5677     } elsif ($archive =~ m/\.vma\.lzo$/) {
5678         $format = 'vma' if !$format;
5679         $comp = 'lzop';
5680     } else {
5681         $format = 'vma' if !$format; # default
5682     }
5683
5684     # try to detect archive format
5685     if ($format eq 'tar') {
5686         return restore_tar_archive($archive, $vmid, $user, $opts);
5687     } else {
5688         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5689     }
5690 }
5691
5692 sub restore_update_config_line {
5693     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5694
5695     return if $line =~ m/^\#qmdump\#/;
5696     return if $line =~ m/^\#vzdump\#/;
5697     return if $line =~ m/^lock:/;
5698     return if $line =~ m/^unused\d+:/;
5699     return if $line =~ m/^parent:/;
5700     return if $line =~ m/^template:/; # restored VM is never a template
5701
5702     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5703     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5704         # try to convert old 1.X settings
5705         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5706         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5707             my ($model, $macaddr) = split(/\=/, $devconfig);
5708             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5709             my $net = {
5710                 model => $model,
5711                 bridge => "vmbr$ind",
5712                 macaddr => $macaddr,
5713             };
5714             my $netstr = print_net($net);
5715
5716             print $outfd "net$cookie->{netcount}: $netstr\n";
5717             $cookie->{netcount}++;
5718         }
5719     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5720         my ($id, $netstr) = ($1, $2);
5721         my $net = parse_net($netstr);
5722         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5723         $netstr = print_net($net);
5724         print $outfd "$id: $netstr\n";
5725     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5726         my $virtdev = $1;
5727         my $value = $3;
5728         my $di = parse_drive($virtdev, $value);
5729         if (defined($di->{backup}) && !$di->{backup}) {
5730             print $outfd "#$line";
5731         } elsif ($map->{$virtdev}) {
5732             delete $di->{format}; # format can change on restore
5733             $di->{file} = $map->{$virtdev};
5734             $value = print_drive($vmid, $di);
5735             print $outfd "$virtdev: $value\n";
5736         } else {
5737             print $outfd $line;
5738         }
5739     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5740         my $vmgenid = $1;
5741         if ($vmgenid ne '0') {
5742             # always generate a new vmgenid if there was a valid one setup
5743             $vmgenid = generate_uuid();
5744         }
5745         print $outfd "vmgenid: $vmgenid\n";
5746     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5747         my ($uuid, $uuid_str);
5748         UUID::generate($uuid);
5749         UUID::unparse($uuid, $uuid_str);
5750         my $smbios1 = parse_smbios1($2);
5751         $smbios1->{uuid} = $uuid_str;
5752         print $outfd $1.print_smbios1($smbios1)."\n";
5753     } else {
5754         print $outfd $line;
5755     }
5756 }
5757
5758 sub scan_volids {
5759     my ($cfg, $vmid) = @_;
5760
5761     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5762
5763     my $volid_hash = {};
5764     foreach my $storeid (keys %$info) {
5765         foreach my $item (@{$info->{$storeid}}) {
5766             next if !($item->{volid} && $item->{size});
5767             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5768             $volid_hash->{$item->{volid}} = $item;
5769         }
5770     }
5771
5772     return $volid_hash;
5773 }
5774
5775 sub is_volume_in_use {
5776     my ($storecfg, $conf, $skip_drive, $volid) = @_;
5777
5778     my $path = PVE::Storage::path($storecfg, $volid);
5779
5780     my $scan_config = sub {
5781         my ($cref, $snapname) = @_;
5782
5783         foreach my $key (keys %$cref) {
5784             my $value = $cref->{$key};
5785             if (is_valid_drivename($key)) {
5786                 next if $skip_drive && $key eq $skip_drive;
5787                 my $drive = parse_drive($key, $value);
5788                 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
5789                 return 1 if $volid eq $drive->{file};
5790                 if ($drive->{file} =~ m!^/!) {
5791                     return 1 if $drive->{file} eq $path;
5792                 } else {
5793                     my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
5794                     next if !$storeid;
5795                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
5796                     next if !$scfg;
5797                     return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
5798                 }
5799             }
5800         }
5801
5802         return 0;
5803     };
5804
5805     return 1 if &$scan_config($conf);
5806
5807     undef $skip_drive;
5808
5809     foreach my $snapname (keys %{$conf->{snapshots}}) {
5810         return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
5811     }
5812
5813     return 0;
5814 }
5815
5816 sub update_disksize {
5817     my ($vmid, $conf, $volid_hash) = @_;
5818
5819     my $changes;
5820     my $prefix = "VM $vmid:";
5821
5822     # used and unused disks
5823     my $referenced = {};
5824
5825     # Note: it is allowed to define multiple storages with same path (alias), so
5826     # we need to check both 'volid' and real 'path' (two different volid can point
5827     # to the same path).
5828
5829     my $referencedpath = {};
5830
5831     # update size info
5832     foreach my $opt (keys %$conf) {
5833         if (is_valid_drivename($opt)) {
5834             my $drive = parse_drive($opt, $conf->{$opt});
5835             my $volid = $drive->{file};
5836             next if !$volid;
5837
5838             $referenced->{$volid} = 1;
5839             if ($volid_hash->{$volid} &&
5840                 (my $path = $volid_hash->{$volid}->{path})) {
5841                 $referencedpath->{$path} = 1;
5842             }
5843
5844             next if drive_is_cdrom($drive);
5845             next if !$volid_hash->{$volid};
5846
5847             $drive->{size} = $volid_hash->{$volid}->{size};
5848             my $new = print_drive($vmid, $drive);
5849             if ($new ne $conf->{$opt}) {
5850                 $changes = 1;
5851                 $conf->{$opt} = $new;
5852                 print "$prefix update disk '$opt' information.\n";
5853             }
5854         }
5855     }
5856
5857     # remove 'unusedX' entry if volume is used
5858     foreach my $opt (keys %$conf) {
5859         next if $opt !~ m/^unused\d+$/;
5860         my $volid = $conf->{$opt};
5861         my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5862         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5863             print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
5864             $changes = 1;
5865             delete $conf->{$opt};
5866         }
5867
5868         $referenced->{$volid} = 1;
5869         $referencedpath->{$path} = 1 if $path;
5870     }
5871
5872     foreach my $volid (sort keys %$volid_hash) {
5873         next if $volid =~ m/vm-$vmid-state-/;
5874         next if $referenced->{$volid};
5875         my $path = $volid_hash->{$volid}->{path};
5876         next if !$path; # just to be sure
5877         next if $referencedpath->{$path};
5878         $changes = 1;
5879         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5880         print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
5881         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5882     }
5883
5884     return $changes;
5885 }
5886
5887 sub rescan {
5888     my ($vmid, $nolock, $dryrun) = @_;
5889
5890     my $cfg = PVE::Storage::config();
5891
5892     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5893     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
5894     foreach my $stor (keys %{$cfg->{ids}}) {
5895         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
5896     }
5897
5898     print "rescan volumes...\n";
5899     my $volid_hash = scan_volids($cfg, $vmid);
5900
5901     my $updatefn =  sub {
5902         my ($vmid) = @_;
5903
5904         my $conf = PVE::QemuConfig->load_config($vmid);
5905
5906         PVE::QemuConfig->check_lock($conf);
5907
5908         my $vm_volids = {};
5909         foreach my $volid (keys %$volid_hash) {
5910             my $info = $volid_hash->{$volid};
5911             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5912         }
5913
5914         my $changes = update_disksize($vmid, $conf, $vm_volids);
5915
5916         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
5917     };
5918
5919     if (defined($vmid)) {
5920         if ($nolock) {
5921             &$updatefn($vmid);
5922         } else {
5923             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5924         }
5925     } else {
5926         my $vmlist = config_list();
5927         foreach my $vmid (keys %$vmlist) {
5928             if ($nolock) {
5929                 &$updatefn($vmid);
5930             } else {
5931                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5932             }
5933         }
5934     }
5935 }
5936
5937 sub restore_vma_archive {
5938     my ($archive, $vmid, $user, $opts, $comp) = @_;
5939
5940     my $readfrom = $archive;
5941
5942     my $cfg = PVE::Storage::config();
5943     my $commands = [];
5944     my $bwlimit = $opts->{bwlimit};
5945
5946     my $dbg_cmdstring = '';
5947     my $add_pipe = sub {
5948         my ($cmd) = @_;
5949         push @$commands, $cmd;
5950         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
5951         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
5952         $readfrom = '-';
5953     };
5954
5955     my $input = undef;
5956     if ($archive eq '-') {
5957         $input = '<&STDIN';
5958     } else {
5959         # If we use a backup from a PVE defined storage we also consider that
5960         # storage's rate limit:
5961         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
5962         if (defined($volid)) {
5963             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
5964             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
5965             if ($readlimit) {
5966                 print STDERR "applying read rate limit: $readlimit\n";
5967                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
5968                 $add_pipe->($cstream);
5969             }
5970         }
5971     }
5972
5973     if ($comp) {
5974         my $cmd;
5975         if ($comp eq 'gzip') {
5976             $cmd = ['zcat', $readfrom];
5977         } elsif ($comp eq 'lzop') {
5978             $cmd = ['lzop', '-d', '-c', $readfrom];
5979         } else {
5980             die "unknown compression method '$comp'\n";
5981         }
5982         $add_pipe->($cmd);
5983     }
5984
5985     my $tmpdir = "/var/tmp/vzdumptmp$$";
5986     rmtree $tmpdir;
5987
5988     # disable interrupts (always do cleanups)
5989     local $SIG{INT} =
5990         local $SIG{TERM} =
5991         local $SIG{QUIT} =
5992         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
5993
5994     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5995     POSIX::mkfifo($mapfifo, 0600);
5996     my $fifofh;
5997
5998     my $openfifo = sub {
5999         open($fifofh, '>', $mapfifo) || die $!;
6000     };
6001
6002     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6003
6004     my $oldtimeout;
6005     my $timeout = 5;
6006
6007     my $devinfo = {};
6008
6009     my $rpcenv = PVE::RPCEnvironment::get();
6010
6011     my $conffile = PVE::QemuConfig->config_file($vmid);
6012     my $tmpfn = "$conffile.$$.tmp";
6013
6014     # Note: $oldconf is undef if VM does not exists
6015     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6016     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6017
6018     my %storage_limits;
6019
6020     my $print_devmap = sub {
6021         my $virtdev_hash = {};
6022
6023         my $cfgfn = "$tmpdir/qemu-server.conf";
6024
6025         # we can read the config - that is already extracted
6026         my $fh = IO::File->new($cfgfn, "r") ||
6027             "unable to read qemu-server.conf - $!\n";
6028
6029         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6030         if (-f $fwcfgfn) {
6031             my $pve_firewall_dir = '/etc/pve/firewall';
6032             mkdir $pve_firewall_dir; # make sure the dir exists
6033             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6034         }
6035
6036         while (defined(my $line = <$fh>)) {
6037             if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6038                 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6039                 die "archive does not contain data for drive '$virtdev'\n"
6040                     if !$devinfo->{$devname};
6041                 if (defined($opts->{storage})) {
6042                     $storeid = $opts->{storage} || 'local';
6043                 } elsif (!$storeid) {
6044                     $storeid = 'local';
6045                 }
6046                 $format = 'raw' if !$format;
6047                 $devinfo->{$devname}->{devname} = $devname;
6048                 $devinfo->{$devname}->{virtdev} = $virtdev;
6049                 $devinfo->{$devname}->{format} = $format;
6050                 $devinfo->{$devname}->{storeid} = $storeid;
6051
6052                 # check permission on storage
6053                 my $pool = $opts->{pool}; # todo: do we need that?
6054                 if ($user ne 'root@pam') {
6055                     $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6056                 }
6057
6058                 $storage_limits{$storeid} = $bwlimit;
6059
6060                 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6061             }
6062         }
6063
6064         foreach my $key (keys %storage_limits) {
6065             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6066             next if !$limit;
6067             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6068             $storage_limits{$key} = $limit * 1024;
6069         }
6070
6071         foreach my $devname (keys %$devinfo) {
6072             die "found no device mapping information for device '$devname'\n"
6073                 if !$devinfo->{$devname}->{virtdev};
6074         }
6075
6076         # create empty/temp config
6077         if ($oldconf) {
6078             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6079             foreach_drive($oldconf, sub {
6080                 my ($ds, $drive) = @_;
6081
6082                 return if drive_is_cdrom($drive);
6083
6084                 my $volid = $drive->{file};
6085
6086                 return if !$volid || $volid =~ m|^/|;
6087
6088                 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6089                 return if !$path || !$owner || ($owner != $vmid);
6090
6091                 # Note: only delete disk we want to restore
6092                 # other volumes will become unused
6093                 if ($virtdev_hash->{$ds}) {
6094                     eval { PVE::Storage::vdisk_free($cfg, $volid); };
6095                     if (my $err = $@) {
6096                         warn $err;
6097                     }
6098                 }
6099             });
6100
6101             # delete vmstate files
6102             # since after the restore we have no snapshots anymore
6103             foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6104                 my $snap = $oldconf->{snapshots}->{$snapname};
6105                 if ($snap->{vmstate}) {
6106                     eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6107                     if (my $err = $@) {
6108                         warn $err;
6109                     }
6110                 }
6111             }
6112         }
6113
6114         my $map = {};
6115         foreach my $virtdev (sort keys %$virtdev_hash) {
6116             my $d = $virtdev_hash->{$virtdev};
6117             my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6118             my $storeid = $d->{storeid};
6119             my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6120
6121             my $map_opts = '';
6122             if (my $limit = $storage_limits{$storeid}) {
6123                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6124             }
6125
6126             # test if requested format is supported
6127             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6128             my $supported = grep { $_ eq $d->{format} } @$validFormats;
6129             $d->{format} = $defFormat if !$supported;
6130
6131             my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid,
6132                                                   $d->{format}, undef, $alloc_size);
6133             print STDERR "new volume ID is '$volid'\n";
6134             $d->{volid} = $volid;
6135             my $path = PVE::Storage::path($cfg, $volid);
6136
6137             PVE::Storage::activate_volumes($cfg,[$volid]);
6138
6139             my $write_zeros = 1;
6140             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6141                 $write_zeros = 0;
6142             }
6143
6144             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6145
6146             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6147             $map->{$virtdev} = $volid;
6148         }
6149
6150         $fh->seek(0, 0) || die "seek failed - $!\n";
6151
6152         my $outfd = new IO::File ($tmpfn, "w") ||
6153             die "unable to write config for VM $vmid\n";
6154
6155         my $cookie = { netcount => 0 };
6156         while (defined(my $line = <$fh>)) {
6157             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6158         }
6159
6160         $fh->close();
6161         $outfd->close();
6162     };
6163
6164     eval {
6165         # enable interrupts
6166         local $SIG{INT} =
6167             local $SIG{TERM} =
6168             local $SIG{QUIT} =
6169             local $SIG{HUP} =
6170             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6171         local $SIG{ALRM} = sub { die "got timeout\n"; };
6172
6173         $oldtimeout = alarm($timeout);
6174
6175         my $parser = sub {
6176             my $line = shift;
6177
6178             print "$line\n";
6179
6180             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6181                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6182                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6183             } elsif ($line =~ m/^CTIME: /) {
6184                 # we correctly received the vma config, so we can disable
6185                 # the timeout now for disk allocation (set to 10 minutes, so
6186                 # that we always timeout if something goes wrong)
6187                 alarm(600);
6188                 &$print_devmap();
6189                 print $fifofh "done\n";
6190                 my $tmp = $oldtimeout || 0;
6191                 $oldtimeout = undef;
6192                 alarm($tmp);
6193                 close($fifofh);
6194             }
6195         };
6196
6197         print "restore vma archive: $dbg_cmdstring\n";
6198         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6199     };
6200     my $err = $@;
6201
6202     alarm($oldtimeout) if $oldtimeout;
6203
6204     my $vollist = [];
6205     foreach my $devname (keys %$devinfo) {
6206         my $volid = $devinfo->{$devname}->{volid};
6207         push @$vollist, $volid if $volid;
6208     }
6209
6210     PVE::Storage::deactivate_volumes($cfg, $vollist);
6211
6212     unlink $mapfifo;
6213
6214     if ($err) {
6215         rmtree $tmpdir;
6216         unlink $tmpfn;
6217
6218         foreach my $devname (keys %$devinfo) {
6219             my $volid = $devinfo->{$devname}->{volid};
6220             next if !$volid;
6221             eval {
6222                 if ($volid =~ m|^/|) {
6223                     unlink $volid || die 'unlink failed\n';
6224                 } else {
6225                     PVE::Storage::vdisk_free($cfg, $volid);
6226                 }
6227                 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6228             };
6229             print STDERR "unable to cleanup '$volid' - $@" if $@;
6230         }
6231         die $err;
6232     }
6233
6234     rmtree $tmpdir;
6235
6236     rename($tmpfn, $conffile) ||
6237         die "unable to commit configuration file '$conffile'\n";
6238
6239     PVE::Cluster::cfs_update(); # make sure we read new file
6240
6241     eval { rescan($vmid, 1); };
6242     warn $@ if $@;
6243 }
6244
6245 sub restore_tar_archive {
6246     my ($archive, $vmid, $user, $opts) = @_;
6247
6248     if ($archive ne '-') {
6249         my $firstfile = tar_archive_read_firstfile($archive);
6250         die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6251             if $firstfile ne 'qemu-server.conf';
6252     }
6253
6254     my $storecfg = PVE::Storage::config();
6255
6256     # destroy existing data - keep empty config
6257     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6258     destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6259
6260     my $tocmd = "/usr/lib/qemu-server/qmextract";
6261
6262     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6263     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6264     $tocmd .= ' --prealloc' if $opts->{prealloc};
6265     $tocmd .= ' --info' if $opts->{info};
6266
6267     # tar option "xf" does not autodetect compression when read from STDIN,
6268     # so we pipe to zcat
6269     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6270         PVE::Tools::shellquote("--to-command=$tocmd");
6271
6272     my $tmpdir = "/var/tmp/vzdumptmp$$";
6273     mkpath $tmpdir;
6274
6275     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6276     local $ENV{VZDUMP_VMID} = $vmid;
6277     local $ENV{VZDUMP_USER} = $user;
6278
6279     my $conffile = PVE::QemuConfig->config_file($vmid);
6280     my $tmpfn = "$conffile.$$.tmp";
6281
6282     # disable interrupts (always do cleanups)
6283     local $SIG{INT} =
6284         local $SIG{TERM} =
6285         local $SIG{QUIT} =
6286         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6287
6288     eval {
6289         # enable interrupts
6290         local $SIG{INT} =
6291             local $SIG{TERM} =
6292             local $SIG{QUIT} =
6293             local $SIG{HUP} =
6294             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6295
6296         if ($archive eq '-') {
6297             print "extracting archive from STDIN\n";
6298             run_command($cmd, input => "<&STDIN");
6299         } else {
6300             print "extracting archive '$archive'\n";
6301             run_command($cmd);
6302         }
6303
6304         return if $opts->{info};
6305
6306         # read new mapping
6307         my $map = {};
6308         my $statfile = "$tmpdir/qmrestore.stat";
6309         if (my $fd = IO::File->new($statfile, "r")) {
6310             while (defined (my $line = <$fd>)) {
6311                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6312                     $map->{$1} = $2 if $1;
6313                 } else {
6314                     print STDERR "unable to parse line in statfile - $line\n";
6315                 }
6316             }
6317             $fd->close();
6318         }
6319
6320         my $confsrc = "$tmpdir/qemu-server.conf";
6321
6322         my $srcfd = new IO::File($confsrc, "r") ||
6323             die "unable to open file '$confsrc'\n";
6324
6325         my $outfd = new IO::File ($tmpfn, "w") ||
6326             die "unable to write config for VM $vmid\n";
6327
6328         my $cookie = { netcount => 0 };
6329         while (defined (my $line = <$srcfd>)) {
6330             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6331         }
6332
6333         $srcfd->close();
6334         $outfd->close();
6335     };
6336     my $err = $@;
6337
6338     if ($err) {
6339
6340         unlink $tmpfn;
6341
6342         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6343
6344         die $err;
6345     }
6346
6347     rmtree $tmpdir;
6348
6349     rename $tmpfn, $conffile ||
6350         die "unable to commit configuration file '$conffile'\n";
6351
6352     PVE::Cluster::cfs_update(); # make sure we read new file
6353
6354     eval { rescan($vmid, 1); };
6355     warn $@ if $@;
6356 };
6357
6358 sub foreach_storage_used_by_vm {
6359     my ($conf, $func) = @_;
6360
6361     my $sidhash = {};
6362
6363     foreach_drive($conf, sub {
6364         my ($ds, $drive) = @_;
6365         return if drive_is_cdrom($drive);
6366
6367         my $volid = $drive->{file};
6368
6369         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6370         $sidhash->{$sid} = $sid if $sid;
6371     });
6372
6373     foreach my $sid (sort keys %$sidhash) {
6374         &$func($sid);
6375     }
6376 }
6377
6378 sub do_snapshots_with_qemu {
6379     my ($storecfg, $volid) = @_;
6380
6381     my $storage_name = PVE::Storage::parse_volume_id($volid);
6382
6383     if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}}
6384         && !$storecfg->{ids}->{$storage_name}->{krbd}){
6385         return 1;
6386     }
6387
6388     if ($volid =~ m/\.(qcow2|qed)$/){
6389         return 1;
6390     }
6391
6392     return undef;
6393 }
6394
6395 sub qga_check_running {
6396     my ($vmid, $nowarn) = @_;
6397
6398     eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6399     if ($@) {
6400         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6401         return 0;
6402     }
6403     return 1;
6404 }
6405
6406 sub template_create {
6407     my ($vmid, $conf, $disk) = @_;
6408
6409     my $storecfg = PVE::Storage::config();
6410
6411     foreach_drive($conf, sub {
6412         my ($ds, $drive) = @_;
6413
6414         return if drive_is_cdrom($drive);
6415         return if $disk && $ds ne $disk;
6416
6417         my $volid = $drive->{file};
6418         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6419
6420         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6421         $drive->{file} = $voliddst;
6422         $conf->{$ds} = print_drive($vmid, $drive);
6423         PVE::QemuConfig->write_config($vmid, $conf);
6424     });
6425 }
6426
6427 sub qemu_img_convert {
6428     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6429
6430     my $storecfg = PVE::Storage::config();
6431     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6432     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6433
6434     if ($src_storeid && $dst_storeid) {
6435
6436         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6437
6438         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6439         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6440
6441         my $src_format = qemu_img_format($src_scfg, $src_volname);
6442         my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6443
6444         my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6445         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6446
6447         my $cmd = [];
6448         push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6449         push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6450         push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6451         push @$cmd, '-T', 'none' if $src_scfg->{type} eq 'zfspool';
6452         push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path;
6453         if ($is_zero_initialized) {
6454             push @$cmd, "zeroinit:$dst_path";
6455         } else {
6456             push @$cmd, $dst_path;
6457         }
6458
6459         my $parser = sub {
6460             my $line = shift;
6461             if($line =~ m/\((\S+)\/100\%\)/){
6462                 my $percent = $1;
6463                 my $transferred = int($size * $percent / 100);
6464                 my $remaining = $size - $transferred;
6465
6466                 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6467             }
6468
6469         };
6470
6471         eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6472         my $err = $@;
6473         die "copy failed: $err" if $err;
6474     }
6475 }
6476
6477 sub qemu_img_format {
6478     my ($scfg, $volname) = @_;
6479
6480     if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6481         return $1;
6482     } else {
6483         return "raw";
6484     }
6485 }
6486
6487 sub qemu_drive_mirror {
6488     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga) = @_;
6489
6490     $jobs = {} if !$jobs;
6491
6492     my $qemu_target;
6493     my $format;
6494     $jobs->{"drive-$drive"} = {};
6495
6496     if ($dst_volid =~ /^nbd:/) {
6497         $qemu_target = $dst_volid;
6498         $format = "nbd";
6499     } else {
6500         my $storecfg = PVE::Storage::config();
6501         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6502
6503         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6504
6505         $format = qemu_img_format($dst_scfg, $dst_volname);
6506
6507         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6508
6509         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6510     }
6511
6512     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6513     $opts->{format} = $format if $format;
6514
6515     print "drive mirror is starting for drive-$drive\n";
6516
6517     eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); }; #if a job already run for this device,it's throw an error
6518
6519     if (my $err = $@) {
6520         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6521         die "mirroring error: $err";
6522     }
6523
6524     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
6525 }
6526
6527 sub qemu_drive_mirror_monitor {
6528     my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
6529
6530     eval {
6531         my $err_complete = 0;
6532
6533         while (1) {
6534             die "storage migration timed out\n" if $err_complete > 300;
6535
6536             my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6537
6538             my $running_mirror_jobs = {};
6539             foreach my $stat (@$stats) {
6540                 next if $stat->{type} ne 'mirror';
6541                 $running_mirror_jobs->{$stat->{device}} = $stat;
6542             }
6543
6544             my $readycounter = 0;
6545
6546             foreach my $job (keys %$jobs) {
6547
6548                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6549                     print "$job : finished\n";
6550                     delete $jobs->{$job};
6551                     next;
6552                 }
6553
6554                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6555
6556                 my $busy = $running_mirror_jobs->{$job}->{busy};
6557                 my $ready = $running_mirror_jobs->{$job}->{ready};
6558                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6559                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6560                     my $remaining = $total - $transferred;
6561                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6562
6563                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6564                 }
6565
6566                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6567             }
6568
6569             last if scalar(keys %$jobs) == 0;
6570
6571             if ($readycounter == scalar(keys %$jobs)) {
6572                 print "all mirroring jobs are ready \n";
6573                 last if $skipcomplete; #do the complete later
6574
6575                 if ($vmiddst && $vmiddst != $vmid) {
6576                     my $agent_running = $qga && qga_check_running($vmid);
6577                     if ($agent_running) {
6578                         print "freeze filesystem\n";
6579                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6580                     } else {
6581                         print "suspend vm\n";
6582                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6583                     }
6584
6585                     # if we clone a disk for a new target vm, we don't switch the disk
6586                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6587
6588                     if ($agent_running) {
6589                         print "unfreeze filesystem\n";
6590                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6591                     } else {
6592                         print "resume vm\n";
6593                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6594                     }
6595
6596                     last;
6597                 } else {
6598
6599                     foreach my $job (keys %$jobs) {
6600                         # try to switch the disk if source and destination are on the same guest
6601                         print "$job: Completing block job...\n";
6602
6603                         eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
6604                         if ($@ =~ m/cannot be completed/) {
6605                             print "$job: Block job cannot be completed, try again.\n";
6606                             $err_complete++;
6607                         }else {
6608                             print "$job: Completed successfully.\n";
6609                             $jobs->{$job}->{complete} = 1;
6610                         }
6611                     }
6612                 }
6613             }
6614             sleep 1;
6615         }
6616     };
6617     my $err = $@;
6618
6619     if ($err) {
6620         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6621         die "mirroring error: $err";
6622     }
6623
6624 }
6625
6626 sub qemu_blockjobs_cancel {
6627     my ($vmid, $jobs) = @_;
6628
6629     foreach my $job (keys %$jobs) {
6630         print "$job: Cancelling block job\n";
6631         eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
6632         $jobs->{$job}->{cancel} = 1;
6633     }
6634
6635     while (1) {
6636         my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6637
6638         my $running_jobs = {};
6639         foreach my $stat (@$stats) {
6640             $running_jobs->{$stat->{device}} = $stat;
6641         }
6642
6643         foreach my $job (keys %$jobs) {
6644
6645             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6646                 print "$job: Done.\n";
6647                 delete $jobs->{$job};
6648             }
6649         }
6650
6651         last if scalar(keys %$jobs) == 0;
6652
6653         sleep 1;
6654     }
6655 }
6656
6657 sub clone_disk {
6658     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6659         $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga) = @_;
6660
6661     my $newvolid;
6662
6663     if (!$full) {
6664         print "create linked clone of drive $drivename ($drive->{file})\n";
6665         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6666         push @$newvollist, $newvolid;
6667     } else {
6668
6669         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6670         $storeid = $storage if $storage;
6671
6672         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6673         my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6674
6675         print "create full clone of drive $drivename ($drive->{file})\n";
6676         my $name = undef;
6677         if (drive_is_cloudinit($drive)) {
6678             $name = "vm-$newvmid-cloudinit";
6679             # cloudinit only supports raw and qcow2 atm:
6680             if ($dst_format eq 'qcow2') {
6681                 $name .= '.qcow2';
6682             } elsif ($dst_format ne 'raw') {
6683                 die "clone: unhandled format for cloudinit image\n";
6684             }
6685         }
6686         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
6687         push @$newvollist, $newvolid;
6688
6689         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6690
6691         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6692         if (!$running || $snapname) {
6693             qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6694         } else {
6695
6696             my $kvmver = get_running_qemu_version ($vmid);
6697             if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
6698                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6699                     if $drive->{iothread};
6700             }
6701
6702             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga);
6703         }
6704     }
6705
6706     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6707
6708     my $disk = $drive;
6709     $disk->{format} = undef;
6710     $disk->{file} = $newvolid;
6711     $disk->{size} = $size;
6712
6713     return $disk;
6714 }
6715
6716 # this only works if VM is running
6717 sub get_current_qemu_machine {
6718     my ($vmid) = @_;
6719
6720     my $cmd = { execute => 'query-machines', arguments => {} };
6721     my $res = vm_qmp_command($vmid, $cmd);
6722
6723     my ($current, $default);
6724     foreach my $e (@$res) {
6725         $default = $e->{name} if $e->{'is-default'};
6726         $current = $e->{name} if $e->{'is-current'};
6727     }
6728
6729     # fallback to the default machine if current is not supported by qemu
6730     return $current || $default || 'pc';
6731 }
6732
6733 sub get_running_qemu_version {
6734     my ($vmid) = @_;
6735     my $cmd = { execute => 'query-version', arguments => {} };
6736     my $res = vm_qmp_command($vmid, $cmd);
6737     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6738 }
6739
6740 sub qemu_machine_feature_enabled {
6741     my ($machine, $kvmver, $version_major, $version_minor) = @_;
6742
6743     my $current_major;
6744     my $current_minor;
6745
6746     if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
6747
6748         $current_major = $3;
6749         $current_minor = $4;
6750
6751     } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6752
6753         $current_major = $1;
6754         $current_minor = $2;
6755     }
6756
6757     return 1 if $current_major > $version_major ||
6758                 ($current_major == $version_major &&
6759                  $current_minor >= $version_minor);
6760 }
6761
6762 sub qemu_machine_pxe {
6763     my ($vmid, $conf, $machine) = @_;
6764
6765     $machine =  PVE::QemuServer::get_current_qemu_machine($vmid) if !$machine;
6766
6767     if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
6768         $machine .= '.pxe';
6769     }
6770
6771     return $machine;
6772 }
6773
6774 sub qemu_use_old_bios_files {
6775     my ($machine_type) = @_;
6776
6777     return if !$machine_type;
6778
6779     my $use_old_bios_files = undef;
6780
6781     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6782         $machine_type = $1;
6783         $use_old_bios_files = 1;
6784     } else {
6785         my $kvmver = kvm_user_version();
6786         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6787         # load new efi bios files on migration. So this hack is required to allow
6788         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6789         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6790         $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
6791     }
6792
6793     return ($use_old_bios_files, $machine_type);
6794 }
6795
6796 sub create_efidisk($$$$$) {
6797     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
6798
6799     my (undef, $ovmf_vars) = get_ovmf_files($arch);
6800     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
6801
6802     my $vars_size = PVE::Tools::convert_size(-s $ovmf_vars, 'b' => 'kb');
6803     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
6804     PVE::Storage::activate_volumes($storecfg, [$volid]);
6805
6806     my $path = PVE::Storage::path($storecfg, $volid);
6807     eval {
6808         run_command(['/usr/bin/qemu-img', 'convert', '-n', '-f', 'raw', '-O', $fmt, $ovmf_vars, $path]);
6809     };
6810     die "Copying EFI vars image failed: $@" if $@;
6811
6812     return ($volid, $vars_size);
6813 }
6814
6815 sub lspci {
6816
6817     my $devices = {};
6818
6819     dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6820             my (undef, $id, $function) = @_;
6821             my $res = { id => $id, function => $function};
6822             push @{$devices->{$id}}, $res;
6823     });
6824
6825     # Entries should be sorted by functions.
6826     foreach my $id (keys %$devices) {
6827         my $dev = $devices->{$id};
6828         $devices->{$id} = [ sort { $a->{function} <=> $b->{function} } @$dev ];
6829     }
6830
6831     return $devices;
6832 }
6833
6834 sub vm_iothreads_list {
6835     my ($vmid) = @_;
6836
6837     my $res = vm_mon_cmd($vmid, 'query-iothreads');
6838
6839     my $iothreads = {};
6840     foreach my $iothread (@$res) {
6841         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6842     }
6843
6844     return $iothreads;
6845 }
6846
6847 sub scsihw_infos {
6848     my ($conf, $drive) = @_;
6849
6850     my $maxdev = 0;
6851
6852     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
6853         $maxdev = 7;
6854     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
6855         $maxdev = 1;
6856     } else {
6857         $maxdev = 256;
6858     }
6859
6860     my $controller = int($drive->{index} / $maxdev);
6861     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
6862
6863     return ($maxdev, $controller, $controller_prefix);
6864 }
6865
6866 sub add_hyperv_enlightenments {
6867     my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough) = @_;
6868
6869     return if $winversion < 6;
6870     return if $bios && $bios eq 'ovmf' && $winversion < 8;
6871
6872     push @$cpuFlags , 'hv_vendor_id=proxmox' if $gpu_passthrough;
6873
6874     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
6875         push @$cpuFlags , 'hv_spinlocks=0x1fff';
6876         push @$cpuFlags , 'hv_vapic';
6877         push @$cpuFlags , 'hv_time';
6878     } else {
6879         push @$cpuFlags , 'hv_spinlocks=0xffff';
6880     }
6881
6882     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
6883         push @$cpuFlags , 'hv_reset';
6884         push @$cpuFlags , 'hv_vpindex';
6885         push @$cpuFlags , 'hv_runtime';
6886     }
6887
6888     if ($winversion >= 7) {
6889         push @$cpuFlags , 'hv_relaxed';
6890
6891         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
6892             push @$cpuFlags , 'hv_synic';
6893             push @$cpuFlags , 'hv_stimer';
6894         }
6895     }
6896 }
6897
6898 sub windows_version {
6899     my ($ostype) = @_;
6900
6901     return 0 if !$ostype;
6902
6903     my $winversion = 0;
6904
6905     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
6906         $winversion = 5;
6907     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
6908         $winversion = 6;
6909     } elsif ($ostype =~ m/^win(\d+)$/) {
6910         $winversion = $1;
6911     }
6912
6913     return $winversion;
6914 }
6915
6916 sub resolve_dst_disk_format {
6917         my ($storecfg, $storeid, $src_volname, $format) = @_;
6918         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6919
6920         if (!$format) {
6921             # if no target format is specified, use the source disk format as hint
6922             if ($src_volname) {
6923                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6924                 $format = qemu_img_format($scfg, $src_volname);
6925             } else {
6926                 return $defFormat;
6927             }
6928         }
6929
6930         # test if requested format is supported - else use default
6931         my $supported = grep { $_ eq $format } @$validFormats;
6932         $format = $defFormat if !$supported;
6933         return $format;
6934 }
6935
6936 sub resolve_first_disk {
6937     my $conf = shift;
6938     my @disks = PVE::QemuServer::valid_drive_names();
6939     my $firstdisk;
6940     foreach my $ds (reverse @disks) {
6941         next if !$conf->{$ds};
6942         my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
6943         next if PVE::QemuServer::drive_is_cdrom($disk);
6944         $firstdisk = $ds;
6945     }
6946     return $firstdisk;
6947 }
6948
6949 sub generate_uuid {
6950     my ($uuid, $uuid_str);
6951     UUID::generate($uuid);
6952     UUID::unparse($uuid, $uuid_str);
6953     return $uuid_str;
6954 }
6955
6956 sub generate_smbios1_uuid {
6957     return "uuid=".generate_uuid();
6958 }
6959
6960 sub nbd_stop {
6961     my ($vmid) = @_;
6962
6963     vm_mon_cmd($vmid, 'nbd-server-stop');
6964 }
6965
6966 # bash completion helper
6967
6968 sub complete_backup_archives {
6969     my ($cmdname, $pname, $cvalue) = @_;
6970
6971     my $cfg = PVE::Storage::config();
6972
6973     my $storeid;
6974
6975     if ($cvalue =~ m/^([^:]+):/) {
6976         $storeid = $1;
6977     }
6978
6979     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
6980
6981     my $res = [];
6982     foreach my $id (keys %$data) {
6983         foreach my $item (@{$data->{$id}}) {
6984             next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
6985             push @$res, $item->{volid} if defined($item->{volid});
6986         }
6987     }
6988
6989     return $res;
6990 }
6991
6992 my $complete_vmid_full = sub {
6993     my ($running) = @_;
6994
6995     my $idlist = vmstatus();
6996
6997     my $res = [];
6998
6999     foreach my $id (keys %$idlist) {
7000         my $d = $idlist->{$id};
7001         if (defined($running)) {
7002             next if $d->{template};
7003             next if $running && $d->{status} ne 'running';
7004             next if !$running && $d->{status} eq 'running';
7005         }
7006         push @$res, $id;
7007
7008     }
7009     return $res;
7010 };
7011
7012 sub complete_vmid {
7013     return &$complete_vmid_full();
7014 }
7015
7016 sub complete_vmid_stopped {
7017     return &$complete_vmid_full(0);
7018 }
7019
7020 sub complete_vmid_running {
7021     return &$complete_vmid_full(1);
7022 }
7023
7024 sub complete_storage {
7025
7026     my $cfg = PVE::Storage::config();
7027     my $ids = $cfg->{ids};
7028
7029     my $res = [];
7030     foreach my $sid (keys %$ids) {
7031         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7032         next if !$ids->{$sid}->{content}->{images};
7033         push @$res, $sid;
7034     }
7035
7036     return $res;
7037 }
7038
7039 1;