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 || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
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 my $Arch2Qemu = {
3296     aarch64 => '/usr/bin/qemu-system-aarch64',
3297     x86_64 => '/usr/bin/qemu-system-x86_64',
3298 };
3299 sub get_command_for_arch($) {
3300     my ($arch) = @_;
3301     return '/usr/bin/kvm' if is_native($arch);
3302
3303     my $cmd = $Arch2Qemu->{$arch}
3304         or die "don't know how to emulate architecture '$arch'\n";
3305     return $cmd;
3306 }
3307
3308 sub get_cpu_options {
3309     my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3310
3311     my $cpuFlags = [];
3312     my $ostype = $conf->{ostype};
3313
3314     my $cpu = $kvm ? "kvm64" : "qemu64";
3315     if (my $cputype = $conf->{cpu}) {
3316         my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3317             or die "Cannot parse cpu description: $cputype\n";
3318         $cpu = $cpuconf->{cputype};
3319         $kvm_off = 1 if $cpuconf->{hidden};
3320
3321         if (defined(my $flags = $cpuconf->{flags})) {
3322             push @$cpuFlags, split(";", $flags);
3323         }
3324     }
3325
3326     push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
3327
3328     push @$cpuFlags , '-x2apic'
3329         if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3330
3331     push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3332
3333     push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3334
3335     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3336
3337         push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3338         push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3339     }
3340
3341     add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough) if $kvm;
3342
3343     push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm;
3344
3345     push @$cpuFlags, 'kvm=off' if $kvm_off;
3346
3347     if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3348         push @$cpuFlags, "vendor=${cpu_vendor}"
3349             if $cpu_vendor ne 'default';
3350     } elsif ($arch ne 'aarch64') {
3351         die "internal error"; # should not happen
3352     }
3353
3354     $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3355
3356     return ('-cpu', $cpu);
3357 }
3358
3359 sub config_to_command {
3360     my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3361
3362     my $cmd = [];
3363     my $globalFlags = [];
3364     my $machineFlags = [];
3365     my $rtcFlags = [];
3366     my $devices = [];
3367     my $pciaddr = '';
3368     my $bridges = {};
3369     my $kvmver = kvm_user_version();
3370     my $vernum = 0; # unknown
3371     my $ostype = $conf->{ostype};
3372     my $winversion = windows_version($ostype);
3373     my $kvm = $conf->{kvm};
3374
3375     my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3376     $kvm //= 1 if is_native($arch);
3377
3378     if ($kvm) {
3379         die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3380             if !defined kvm_version();
3381     }
3382
3383     if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3384         $vernum = $1*1000000+$2*1000;
3385     } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3386         $vernum = $1*1000000+$2*1000+$3;
3387     }
3388
3389     die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3390
3391     my $have_ovz = -f '/proc/vz/vestat';
3392
3393     my $q35 = machine_type_is_q35($conf);
3394     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3395     my $use_old_bios_files = undef;
3396     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3397
3398     my $cpuunits = defined($conf->{cpuunits}) ?
3399             $conf->{cpuunits} : $defaults->{cpuunits};
3400
3401     push @$cmd, get_command_for_arch($arch);
3402
3403     push @$cmd, '-id', $vmid;
3404
3405     my $vmname = $conf->{name} || "vm$vmid";
3406
3407     push @$cmd, '-name', $vmname;
3408
3409     my $use_virtio = 0;
3410
3411     my $qmpsocket = qmp_socket($vmid);
3412     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3413     push @$cmd, '-mon', "chardev=qmp,mode=control";
3414
3415     if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3416         my $eventsocket = qmp_socket($vmid, 0, 'event');
3417         push @$cmd, '-chardev', "socket,id=qmp-event,path=$eventsocket,server,nowait";
3418         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3419     }
3420
3421     push @$cmd, '-pidfile' , pidfile_name($vmid);
3422
3423     push @$cmd, '-daemonize';
3424
3425     if ($conf->{smbios1}) {
3426         push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3427     }
3428
3429     if ($conf->{vmgenid}) {
3430         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3431     }
3432
3433     my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3434     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3435         die "uefi base image not found\n" if ! -f $ovmf_code;
3436
3437         my $path;
3438         my $format;
3439         if (my $efidisk = $conf->{efidisk0}) {
3440             my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3441             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3442             $format = $d->{format};
3443             if ($storeid) {
3444                 $path = PVE::Storage::path($storecfg, $d->{file});
3445                 if (!defined($format)) {
3446                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3447                     $format = qemu_img_format($scfg, $volname);
3448                 }
3449             } else {
3450                 $path = $d->{file};
3451                 die "efidisk format must be specified\n"
3452                     if !defined($format);
3453             }
3454         } else {
3455             warn "no efidisk configured! Using temporary efivars disk.\n";
3456             $path = "/tmp/$vmid-ovmf.fd";
3457             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3458             $format = 'raw';
3459         }
3460
3461         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3462         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3463     }
3464
3465
3466     # add usb controllers
3467     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $q35, $usbdesc->{format}, $MAX_USB_DEVICES);
3468     push @$devices, @usbcontrollers if @usbcontrollers;
3469     my $vga = parse_vga($conf->{vga});
3470
3471     my $qxlnum = vga_conf_has_spice($conf->{vga});
3472     $vga->{type} = 'qxl' if $qxlnum;
3473
3474     if (!$vga->{type}) {
3475         if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3476             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3477         } else {
3478             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3479         }
3480     }
3481
3482     # enable absolute mouse coordinates (needed by vnc)
3483     my $tablet;
3484     if (defined($conf->{tablet})) {
3485         $tablet = $conf->{tablet};
3486     } else {
3487         $tablet = $defaults->{tablet};
3488         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3489         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3490     }
3491
3492     push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
3493
3494     my $kvm_off = 0;
3495     my $gpu_passthrough;
3496
3497     # host pci devices
3498     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
3499         my $d = parse_hostpci($conf->{"hostpci$i"});
3500         next if !$d;
3501
3502         my $pcie = $d->{pcie};
3503         if($pcie){
3504             die "q35 machine model is not enabled" if !$q35;
3505             $pciaddr = print_pcie_addr("hostpci$i");
3506         }else{
3507             $pciaddr = print_pci_addr("hostpci$i", $bridges);
3508         }
3509
3510         my $rombar = defined($d->{rombar}) && !$d->{rombar} ? ',rombar=0' : '';
3511         my $romfile = $d->{romfile};
3512
3513         my $xvga = '';
3514         if ($d->{'x-vga'}) {
3515             $xvga = ',x-vga=on';
3516             $kvm_off = 1;
3517             $vga->{type} = 'none';
3518             $gpu_passthrough = 1;
3519
3520             if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3521                 $xvga = "";
3522             }
3523         }
3524         my $pcidevices = $d->{pciid};
3525         my $multifunction = 1 if @$pcidevices > 1;
3526
3527         my $j=0;
3528         foreach my $pcidevice (@$pcidevices) {
3529
3530             my $id = "hostpci$i";
3531             $id .= ".$j" if $multifunction;
3532             my $addr = $pciaddr;
3533             $addr .= ".$j" if $multifunction;
3534             my $devicestr = "vfio-pci,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
3535
3536             if($j == 0){
3537                 $devicestr .= "$rombar$xvga";
3538                 $devicestr .= ",multifunction=on" if $multifunction;
3539                 $devicestr .= ",romfile=/usr/share/kvm/$romfile" if $romfile;
3540             }
3541
3542             push @$devices, '-device', $devicestr;
3543             $j++;
3544         }
3545     }
3546
3547     # usb devices
3548     my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES);
3549     push @$devices, @usbdevices if @usbdevices;
3550     # serial devices
3551     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3552         if (my $path = $conf->{"serial$i"}) {
3553             if ($path eq 'socket') {
3554                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3555                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3556                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3557             } else {
3558                 die "no such serial device\n" if ! -c $path;
3559                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3560                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3561             }
3562         }
3563     }
3564
3565     # parallel devices
3566     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3567         if (my $path = $conf->{"parallel$i"}) {
3568             die "no such parallel device\n" if ! -c $path;
3569             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3570             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3571             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3572         }
3573     }
3574
3575
3576     my $sockets = 1;
3577     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3578     $sockets = $conf->{sockets} if  $conf->{sockets};
3579
3580     my $cores = $conf->{cores} || 1;
3581
3582     my $maxcpus = $sockets * $cores;
3583
3584     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3585
3586     my $allowed_vcpus = $cpuinfo->{cpus};
3587
3588     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3589         if ($allowed_vcpus < $maxcpus);
3590
3591     if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3592
3593         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3594         for (my $i = 2; $i <= $vcpus; $i++)  {
3595             my $cpustr = print_cpu_device($conf,$i);
3596             push @$cmd, '-device', $cpustr;
3597         }
3598
3599     } else {
3600
3601         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3602     }
3603     push @$cmd, '-nodefaults';
3604
3605     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3606
3607     my $bootindex_hash = {};
3608     my $i = 1;
3609     foreach my $o (split(//, $bootorder)) {
3610         $bootindex_hash->{$o} = $i*100;
3611         $i++;
3612     }
3613
3614     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3615
3616     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3617
3618     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3619
3620     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3621         push @$devices, '-device', print_vga_device($conf, $vga, undef, $qxlnum, $bridges);
3622         my $socket = vnc_socket($vmid);
3623         push @$cmd,  '-vnc', "unix:$socket,x509,password";
3624     } else {
3625         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3626         push @$cmd, '-nographic';
3627     }
3628
3629     # time drift fix
3630     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3631
3632     my $useLocaltime = $conf->{localtime};
3633
3634     if ($winversion >= 5) { # windows
3635         $useLocaltime = 1 if !defined($conf->{localtime});
3636
3637         # use time drift fix when acpi is enabled
3638         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3639             $tdf = 1 if !defined($conf->{tdf});
3640         }
3641     }
3642
3643     if ($winversion >= 6) {
3644         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3645         push @$cmd, '-no-hpet';
3646     }
3647
3648     push @$rtcFlags, 'driftfix=slew' if $tdf;
3649
3650     if (!$kvm) {
3651         push @$machineFlags, 'accel=tcg';
3652     }
3653
3654     if ($machine_type) {
3655         push @$machineFlags, "type=${machine_type}";
3656     }
3657
3658     if ($conf->{startdate}) {
3659         push @$rtcFlags, "base=$conf->{startdate}";
3660     } elsif ($useLocaltime) {
3661         push @$rtcFlags, 'base=localtime';
3662     }
3663
3664     push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3665
3666     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3667
3668     push @$cmd, '-S' if $conf->{freeze};
3669
3670     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3671
3672     # enable sound
3673     #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3674     #push @$cmd, '-soundhw', 'es1370';
3675     #push @$cmd, '-soundhw', $soundhw if $soundhw;
3676
3677     if (parse_guest_agent($conf)->{enabled}) {
3678         my $qgasocket = qmp_socket($vmid, 1);
3679         my $pciaddr = print_pci_addr("qga0", $bridges);
3680         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3681         push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3682         push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3683     }
3684
3685     my $spice_port;
3686
3687     if ($qxlnum) {
3688         if ($qxlnum > 1) {
3689             if ($winversion){
3690                 for(my $i = 1; $i < $qxlnum; $i++){
3691                     push @$devices, '-device', print_vga_device($conf, $vga, $i, $qxlnum, $bridges);
3692                 }
3693             } else {
3694                 # assume other OS works like Linux
3695                 my ($ram, $vram) = ("134217728", "67108864");
3696                 if ($vga->{memory}) {
3697                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3698                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3699                 }
3700                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3701                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3702             }
3703         }
3704
3705         my $pciaddr = print_pci_addr("spice", $bridges);
3706
3707         my $nodename = PVE::INotify::nodename();
3708         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3709         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3710         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3711         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3712         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3713
3714         push @$devices, '-spice', "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3715
3716         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3717         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3718         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3719     }
3720
3721     # enable balloon by default, unless explicitly disabled
3722     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3723         $pciaddr = print_pci_addr("balloon0", $bridges);
3724         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3725     }
3726
3727     if ($conf->{watchdog}) {
3728         my $wdopts = parse_watchdog($conf->{watchdog});
3729         $pciaddr = print_pci_addr("watchdog", $bridges);
3730         my $watchdog = $wdopts->{model} || 'i6300esb';
3731         push @$devices, '-device', "$watchdog$pciaddr";
3732         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3733     }
3734
3735     my $vollist = [];
3736     my $scsicontroller = {};
3737     my $ahcicontroller = {};
3738     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3739
3740     # Add iscsi initiator name if available
3741     if (my $initiator = get_initiator_name()) {
3742         push @$devices, '-iscsi', "initiator-name=$initiator";
3743     }
3744
3745     foreach_drive($conf, sub {
3746         my ($ds, $drive) = @_;
3747
3748         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3749             push @$vollist, $drive->{file};
3750         }
3751
3752         # ignore efidisk here, already added in bios/fw handling code above
3753         return if $drive->{interface} eq 'efidisk';
3754
3755         $use_virtio = 1 if $ds =~ m/^virtio/;
3756
3757         if (drive_is_cdrom ($drive)) {
3758             if ($bootindex_hash->{d}) {
3759                 $drive->{bootindex} = $bootindex_hash->{d};
3760                 $bootindex_hash->{d} += 1;
3761             }
3762         } else {
3763             if ($bootindex_hash->{c}) {
3764                 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3765                 $bootindex_hash->{c} += 1;
3766             }
3767         }
3768
3769         if($drive->{interface} eq 'virtio'){
3770            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3771         }
3772
3773         if ($drive->{interface} eq 'scsi') {
3774
3775             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3776
3777             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3778             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3779
3780             my $iothread = '';
3781             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3782                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3783                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3784             } elsif ($drive->{iothread}) {
3785                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3786             }
3787
3788             my $queues = '';
3789             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3790                 $queues = ",num_queues=$drive->{queues}";
3791             }
3792
3793             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3794             $scsicontroller->{$controller}=1;
3795         }
3796
3797         if ($drive->{interface} eq 'sata') {
3798            my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3799            $pciaddr = print_pci_addr("ahci$controller", $bridges);
3800            push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3801            $ahcicontroller->{$controller}=1;
3802         }
3803
3804         my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3805         push @$devices, '-drive',$drive_cmd;
3806         push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3807     });
3808
3809     for (my $i = 0; $i < $MAX_NETS; $i++) {
3810          next if !$conf->{"net$i"};
3811          my $d = parse_net($conf->{"net$i"});
3812          next if !$d;
3813
3814          $use_virtio = 1 if $d->{model} eq 'virtio';
3815
3816          if ($bootindex_hash->{n}) {
3817             $d->{bootindex} = $bootindex_hash->{n};
3818             $bootindex_hash->{n} += 1;
3819          }
3820
3821          my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3822          push @$devices, '-netdev', $netdevfull;
3823
3824          my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files);
3825          push @$devices, '-device', $netdevicefull;
3826     }
3827
3828     if (!$q35) {
3829         # add pci bridges
3830         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3831            $bridges->{1} = 1;
3832            $bridges->{2} = 1;
3833         }
3834
3835         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3836
3837         while (my ($k, $v) = each %$bridges) {
3838             $pciaddr = print_pci_addr("pci.$k");
3839             unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3840         }
3841     }
3842
3843     # add custom args
3844     if ($conf->{args}) {
3845         my $aa = PVE::Tools::split_args($conf->{args});
3846         push @$cmd, @$aa;
3847     }
3848
3849     push @$cmd, @$devices;
3850     push @$cmd, '-rtc', join(',', @$rtcFlags)
3851         if scalar(@$rtcFlags);
3852     push @$cmd, '-machine', join(',', @$machineFlags)
3853         if scalar(@$machineFlags);
3854     push @$cmd, '-global', join(',', @$globalFlags)
3855         if scalar(@$globalFlags);
3856
3857     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3858 }
3859
3860 sub vnc_socket {
3861     my ($vmid) = @_;
3862     return "${var_run_tmpdir}/$vmid.vnc";
3863 }
3864
3865 sub spice_port {
3866     my ($vmid) = @_;
3867
3868     my $res = vm_mon_cmd($vmid, 'query-spice');
3869
3870     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3871 }
3872
3873 sub qmp_socket {
3874     my ($vmid, $qga, $name) = @_;
3875     my $sockettype = $qga ? 'qga' : 'qmp';
3876     my $ext = $name ? '-'.$name : '';
3877     return "${var_run_tmpdir}/$vmid$ext.$sockettype";
3878 }
3879
3880 sub pidfile_name {
3881     my ($vmid) = @_;
3882     return "${var_run_tmpdir}/$vmid.pid";
3883 }
3884
3885 sub vm_devices_list {
3886     my ($vmid) = @_;
3887
3888     my $res = vm_mon_cmd($vmid, 'query-pci');
3889     my $devices_to_check = [];
3890     my $devices = {};
3891     foreach my $pcibus (@$res) {
3892         push @$devices_to_check, @{$pcibus->{devices}},
3893     }
3894
3895     while (@$devices_to_check) {
3896         my $to_check = [];
3897         for my $d (@$devices_to_check) {
3898             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3899             next if !$d->{'pci_bridge'};
3900
3901             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3902             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3903         }
3904         $devices_to_check = $to_check;
3905     }
3906
3907     my $resblock = vm_mon_cmd($vmid, 'query-block');
3908     foreach my $block (@$resblock) {
3909         if($block->{device} =~ m/^drive-(\S+)/){
3910                 $devices->{$1} = 1;
3911         }
3912     }
3913
3914     my $resmice = vm_mon_cmd($vmid, 'query-mice');
3915     foreach my $mice (@$resmice) {
3916         if ($mice->{name} eq 'QEMU HID Tablet') {
3917             $devices->{tablet} = 1;
3918             last;
3919         }
3920     }
3921
3922     # for usb devices there is no query-usb
3923     # but we can iterate over the entries in
3924     # qom-list path=/machine/peripheral
3925     my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3926     foreach my $per (@$resperipheral) {
3927         if ($per->{name} =~ m/^usb\d+$/) {
3928             $devices->{$per->{name}} = 1;
3929         }
3930     }
3931
3932     return $devices;
3933 }
3934
3935 sub vm_deviceplug {
3936     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3937
3938     my $q35 = machine_type_is_q35($conf);
3939
3940     my $devices_list = vm_devices_list($vmid);
3941     return 1 if defined($devices_list->{$deviceid});
3942
3943     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3944
3945     if ($deviceid eq 'tablet') {
3946
3947         qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3948
3949     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3950
3951         die "usb hotplug currently not reliable\n";
3952         # since we can't reliably hot unplug all added usb devices
3953         # and usb passthrough disables live migration
3954         # we disable usb hotplugging for now
3955         qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3956
3957     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3958
3959         qemu_iothread_add($vmid, $deviceid, $device);
3960
3961         qemu_driveadd($storecfg, $vmid, $device);
3962         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3963
3964         qemu_deviceadd($vmid, $devicefull);
3965         eval { qemu_deviceaddverify($vmid, $deviceid); };
3966         if (my $err = $@) {
3967             eval { qemu_drivedel($vmid, $deviceid); };
3968             warn $@ if $@;
3969             die $err;
3970         }
3971
3972     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3973
3974
3975         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3976         my $pciaddr = print_pci_addr($deviceid);
3977         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3978
3979         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3980
3981         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3982             qemu_iothread_add($vmid, $deviceid, $device);
3983             $devicefull .= ",iothread=iothread-$deviceid";
3984         }
3985
3986         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3987             $devicefull .= ",num_queues=$device->{queues}";
3988         }
3989
3990         qemu_deviceadd($vmid, $devicefull);
3991         qemu_deviceaddverify($vmid, $deviceid);
3992
3993     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3994
3995         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3996         qemu_driveadd($storecfg, $vmid, $device);
3997
3998         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3999         eval { qemu_deviceadd($vmid, $devicefull); };
4000         if (my $err = $@) {
4001             eval { qemu_drivedel($vmid, $deviceid); };
4002             warn $@ if $@;
4003             die $err;
4004         }
4005
4006     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4007
4008         return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
4009
4010         my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4011         my $use_old_bios_files = undef;
4012         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4013
4014         my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files);
4015         qemu_deviceadd($vmid, $netdevicefull);
4016         eval { qemu_deviceaddverify($vmid, $deviceid); };
4017         if (my $err = $@) {
4018             eval { qemu_netdevdel($vmid, $deviceid); };
4019             warn $@ if $@;
4020             die $err;
4021         }
4022
4023     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4024
4025         my $bridgeid = $2;
4026         my $pciaddr = print_pci_addr($deviceid);
4027         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4028
4029         qemu_deviceadd($vmid, $devicefull);
4030         qemu_deviceaddverify($vmid, $deviceid);
4031
4032     } else {
4033         die "can't hotplug device '$deviceid'\n";
4034     }
4035
4036     return 1;
4037 }
4038
4039 # fixme: this should raise exceptions on error!
4040 sub vm_deviceunplug {
4041     my ($vmid, $conf, $deviceid) = @_;
4042
4043     my $devices_list = vm_devices_list($vmid);
4044     return 1 if !defined($devices_list->{$deviceid});
4045
4046     die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4047
4048     if ($deviceid eq 'tablet') {
4049
4050         qemu_devicedel($vmid, $deviceid);
4051
4052     } elsif ($deviceid =~ m/^usb\d+$/) {
4053
4054         die "usb hotplug currently not reliable\n";
4055         # when unplugging usb devices this way,
4056         # there may be remaining usb controllers/hubs
4057         # so we disable it for now
4058         qemu_devicedel($vmid, $deviceid);
4059         qemu_devicedelverify($vmid, $deviceid);
4060
4061     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4062
4063         qemu_devicedel($vmid, $deviceid);
4064         qemu_devicedelverify($vmid, $deviceid);
4065         qemu_drivedel($vmid, $deviceid);
4066         qemu_iothread_del($conf, $vmid, $deviceid);
4067
4068     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4069
4070         qemu_devicedel($vmid, $deviceid);
4071         qemu_devicedelverify($vmid, $deviceid);
4072         qemu_iothread_del($conf, $vmid, $deviceid);
4073
4074     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4075
4076         qemu_devicedel($vmid, $deviceid);
4077         qemu_drivedel($vmid, $deviceid);
4078         qemu_deletescsihw($conf, $vmid, $deviceid);
4079
4080     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4081
4082         qemu_devicedel($vmid, $deviceid);
4083         qemu_devicedelverify($vmid, $deviceid);
4084         qemu_netdevdel($vmid, $deviceid);
4085
4086     } else {
4087         die "can't unplug device '$deviceid'\n";
4088     }
4089
4090     return 1;
4091 }
4092
4093 sub qemu_deviceadd {
4094     my ($vmid, $devicefull) = @_;
4095
4096     $devicefull = "driver=".$devicefull;
4097     my %options =  split(/[=,]/, $devicefull);
4098
4099     vm_mon_cmd($vmid, "device_add" , %options);
4100 }
4101
4102 sub qemu_devicedel {
4103     my ($vmid, $deviceid) = @_;
4104
4105     my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4106 }
4107
4108 sub qemu_iothread_add {
4109     my($vmid, $deviceid, $device) = @_;
4110
4111     if ($device->{iothread}) {
4112         my $iothreads = vm_iothreads_list($vmid);
4113         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4114     }
4115 }
4116
4117 sub qemu_iothread_del {
4118     my($conf, $vmid, $deviceid) = @_;
4119
4120     my $device = parse_drive($deviceid, $conf->{$deviceid});
4121     if ($device->{iothread}) {
4122         my $iothreads = vm_iothreads_list($vmid);
4123         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4124     }
4125 }
4126
4127 sub qemu_objectadd {
4128     my($vmid, $objectid, $qomtype) = @_;
4129
4130     vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4131
4132     return 1;
4133 }
4134
4135 sub qemu_objectdel {
4136     my($vmid, $objectid) = @_;
4137
4138     vm_mon_cmd($vmid, "object-del", id => $objectid);
4139
4140     return 1;
4141 }
4142
4143 sub qemu_driveadd {
4144     my ($storecfg, $vmid, $device) = @_;
4145
4146     my $drive = print_drive_full($storecfg, $vmid, $device);
4147     $drive =~ s/\\/\\\\/g;
4148     my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4149
4150     # If the command succeeds qemu prints: "OK"
4151     return 1 if $ret =~ m/OK/s;
4152
4153     die "adding drive failed: $ret\n";
4154 }
4155
4156 sub qemu_drivedel {
4157     my($vmid, $deviceid) = @_;
4158
4159     my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4160     $ret =~ s/^\s+//;
4161
4162     return 1 if $ret eq "";
4163
4164     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4165     return 1 if $ret =~ m/Device \'.*?\' not found/s;
4166
4167     die "deleting drive $deviceid failed : $ret\n";
4168 }
4169
4170 sub qemu_deviceaddverify {
4171     my ($vmid, $deviceid) = @_;
4172
4173     for (my $i = 0; $i <= 5; $i++) {
4174          my $devices_list = vm_devices_list($vmid);
4175          return 1 if defined($devices_list->{$deviceid});
4176          sleep 1;
4177     }
4178
4179     die "error on hotplug device '$deviceid'\n";
4180 }
4181
4182
4183 sub qemu_devicedelverify {
4184     my ($vmid, $deviceid) = @_;
4185
4186     # need to verify that the device is correctly removed as device_del
4187     # is async and empty return is not reliable
4188
4189     for (my $i = 0; $i <= 5; $i++) {
4190          my $devices_list = vm_devices_list($vmid);
4191          return 1 if !defined($devices_list->{$deviceid});
4192          sleep 1;
4193     }
4194
4195     die "error on hot-unplugging device '$deviceid'\n";
4196 }
4197
4198 sub qemu_findorcreatescsihw {
4199     my ($storecfg, $conf, $vmid, $device) = @_;
4200
4201     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4202
4203     my $scsihwid="$controller_prefix$controller";
4204     my $devices_list = vm_devices_list($vmid);
4205
4206     if(!defined($devices_list->{$scsihwid})) {
4207         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
4208     }
4209
4210     return 1;
4211 }
4212
4213 sub qemu_deletescsihw {
4214     my ($conf, $vmid, $opt) = @_;
4215
4216     my $device = parse_drive($opt, $conf->{$opt});
4217
4218     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4219         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4220         return 1;
4221     }
4222
4223     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4224
4225     my $devices_list = vm_devices_list($vmid);
4226     foreach my $opt (keys %{$devices_list}) {
4227         if (PVE::QemuServer::is_valid_drivename($opt)) {
4228             my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4229             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4230                 return 1;
4231             }
4232         }
4233     }
4234
4235     my $scsihwid="scsihw$controller";
4236
4237     vm_deviceunplug($vmid, $conf, $scsihwid);
4238
4239     return 1;
4240 }
4241
4242 sub qemu_add_pci_bridge {
4243     my ($storecfg, $conf, $vmid, $device) = @_;
4244
4245     my $bridges = {};
4246
4247     my $bridgeid;
4248
4249     print_pci_addr($device, $bridges);
4250
4251     while (my ($k, $v) = each %$bridges) {
4252         $bridgeid = $k;
4253     }
4254     return 1 if !defined($bridgeid) || $bridgeid < 1;
4255
4256     my $bridge = "pci.$bridgeid";
4257     my $devices_list = vm_devices_list($vmid);
4258
4259     if (!defined($devices_list->{$bridge})) {
4260         vm_deviceplug($storecfg, $conf, $vmid, $bridge);
4261     }
4262
4263     return 1;
4264 }
4265
4266 sub qemu_set_link_status {
4267     my ($vmid, $device, $up) = @_;
4268
4269     vm_mon_cmd($vmid, "set_link", name => $device,
4270                up => $up ? JSON::true : JSON::false);
4271 }
4272
4273 sub qemu_netdevadd {
4274     my ($vmid, $conf, $device, $deviceid) = @_;
4275
4276     my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid, 1);
4277     my %options =  split(/[=,]/, $netdev);
4278
4279     vm_mon_cmd($vmid, "netdev_add",  %options);
4280     return 1;
4281 }
4282
4283 sub qemu_netdevdel {
4284     my ($vmid, $deviceid) = @_;
4285
4286     vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4287 }
4288
4289 sub qemu_usb_hotplug {
4290     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
4291
4292     return if !$device;
4293
4294     # remove the old one first
4295     vm_deviceunplug($vmid, $conf, $deviceid);
4296
4297     # check if xhci controller is necessary and available
4298     if ($device->{usb3}) {
4299
4300         my $devicelist = vm_devices_list($vmid);
4301
4302         if (!$devicelist->{xhci}) {
4303             my $pciaddr = print_pci_addr("xhci");
4304             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4305         }
4306     }
4307     my $d = parse_usb_device($device->{host});
4308     $d->{usb3} = $device->{usb3};
4309
4310     # add the new one
4311     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d);
4312 }
4313
4314 sub qemu_cpu_hotplug {
4315     my ($vmid, $conf, $vcpus) = @_;
4316
4317     my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4318
4319     my $sockets = 1;
4320     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4321     $sockets = $conf->{sockets} if  $conf->{sockets};
4322     my $cores = $conf->{cores} || 1;
4323     my $maxcpus = $sockets * $cores;
4324
4325     $vcpus = $maxcpus if !$vcpus;
4326
4327     die "you can't add more vcpus than maxcpus\n"
4328         if $vcpus > $maxcpus;
4329
4330     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4331
4332     if ($vcpus < $currentvcpus) {
4333
4334         if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4335
4336             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4337                 qemu_devicedel($vmid, "cpu$i");
4338                 my $retry = 0;
4339                 my $currentrunningvcpus = undef;
4340                 while (1) {
4341                     $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4342                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4343                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4344                     $retry++;
4345                     sleep 1;
4346                 }
4347                 #update conf after each succesfull cpu unplug
4348                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4349                 PVE::QemuConfig->write_config($vmid, $conf);
4350             }
4351         } else {
4352             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4353         }
4354
4355         return;
4356     }
4357
4358     my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4359     die "vcpus in running vm does not match its configuration\n"
4360         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4361
4362     if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4363
4364         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4365             my $cpustr = print_cpu_device($conf, $i);
4366             qemu_deviceadd($vmid, $cpustr);
4367
4368             my $retry = 0;
4369             my $currentrunningvcpus = undef;
4370             while (1) {
4371                 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4372                 last if scalar(@{$currentrunningvcpus}) == $i;
4373                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4374                 sleep 1;
4375                 $retry++;
4376             }
4377             #update conf after each succesfull cpu hotplug
4378             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4379             PVE::QemuConfig->write_config($vmid, $conf);
4380         }
4381     } else {
4382
4383         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4384             vm_mon_cmd($vmid, "cpu-add", id => int($i));
4385         }
4386     }
4387 }
4388
4389 sub qemu_block_set_io_throttle {
4390     my ($vmid, $deviceid,
4391         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4392         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4393         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4394         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4395
4396     return if !check_running($vmid) ;
4397
4398     vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4399         bps => int($bps),
4400         bps_rd => int($bps_rd),
4401         bps_wr => int($bps_wr),
4402         iops => int($iops),
4403         iops_rd => int($iops_rd),
4404         iops_wr => int($iops_wr),
4405         bps_max => int($bps_max),
4406         bps_rd_max => int($bps_rd_max),
4407         bps_wr_max => int($bps_wr_max),
4408         iops_max => int($iops_max),
4409         iops_rd_max => int($iops_rd_max),
4410         iops_wr_max => int($iops_wr_max),
4411         bps_max_length => int($bps_max_length),
4412         bps_rd_max_length => int($bps_rd_max_length),
4413         bps_wr_max_length => int($bps_wr_max_length),
4414         iops_max_length => int($iops_max_length),
4415         iops_rd_max_length => int($iops_rd_max_length),
4416         iops_wr_max_length => int($iops_wr_max_length),
4417     );
4418
4419 }
4420
4421 # old code, only used to shutdown old VM after update
4422 sub __read_avail {
4423     my ($fh, $timeout) = @_;
4424
4425     my $sel = new IO::Select;
4426     $sel->add($fh);
4427
4428     my $res = '';
4429     my $buf;
4430
4431     my @ready;
4432     while (scalar (@ready = $sel->can_read($timeout))) {
4433         my $count;
4434         if ($count = $fh->sysread($buf, 8192)) {
4435             if ($buf =~ /^(.*)\(qemu\) $/s) {
4436                 $res .= $1;
4437                 last;
4438             } else {
4439                 $res .= $buf;
4440             }
4441         } else {
4442             if (!defined($count)) {
4443                 die "$!\n";
4444             }
4445             last;
4446         }
4447     }
4448
4449     die "monitor read timeout\n" if !scalar(@ready);
4450
4451     return $res;
4452 }
4453
4454 sub qemu_block_resize {
4455     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4456
4457     my $running = check_running($vmid);
4458
4459     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4460
4461     return if !$running;
4462
4463     vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4464
4465 }
4466
4467 sub qemu_volume_snapshot {
4468     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4469
4470     my $running = check_running($vmid);
4471
4472     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4473         vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4474     } else {
4475         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4476     }
4477 }
4478
4479 sub qemu_volume_snapshot_delete {
4480     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4481
4482     my $running = check_running($vmid);
4483
4484     if($running) {
4485
4486         $running = undef;
4487         my $conf = PVE::QemuConfig->load_config($vmid);
4488         foreach_drive($conf, sub {
4489             my ($ds, $drive) = @_;
4490             $running = 1 if $drive->{file} eq $volid;
4491         });
4492     }
4493
4494     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4495         vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4496     } else {
4497         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4498     }
4499 }
4500
4501 sub set_migration_caps {
4502     my ($vmid) = @_;
4503
4504     my $cap_ref = [];
4505
4506     my $enabled_cap = {
4507         "auto-converge" => 1,
4508         "xbzrle" => 1,
4509         "x-rdma-pin-all" => 0,
4510         "zero-blocks" => 0,
4511         "compress" => 0
4512     };
4513
4514     my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4515
4516     for my $supported_capability (@$supported_capabilities) {
4517         push @$cap_ref, {
4518             capability => $supported_capability->{capability},
4519             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4520         };
4521     }
4522
4523     vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4524 }
4525
4526 my $fast_plug_option = {
4527     'lock' => 1,
4528     'name' => 1,
4529     'onboot' => 1,
4530     'shares' => 1,
4531     'startup' => 1,
4532     'description' => 1,
4533     'protection' => 1,
4534     'vmstatestorage' => 1,
4535 };
4536
4537 # hotplug changes in [PENDING]
4538 # $selection hash can be used to only apply specified options, for
4539 # example: { cores => 1 } (only apply changed 'cores')
4540 # $errors ref is used to return error messages
4541 sub vmconfig_hotplug_pending {
4542     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4543
4544     my $defaults = load_defaults();
4545
4546     # commit values which do not have any impact on running VM first
4547     # Note: those option cannot raise errors, we we do not care about
4548     # $selection and always apply them.
4549
4550     my $add_error = sub {
4551         my ($opt, $msg) = @_;
4552         $errors->{$opt} = "hotplug problem - $msg";
4553     };
4554
4555     my $changes = 0;
4556     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4557         if ($fast_plug_option->{$opt}) {
4558             $conf->{$opt} = $conf->{pending}->{$opt};
4559             delete $conf->{pending}->{$opt};
4560             $changes = 1;
4561         }
4562     }
4563
4564     if ($changes) {
4565         PVE::QemuConfig->write_config($vmid, $conf);
4566         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4567     }
4568
4569     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4570
4571     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4572     while (my ($opt, $force) = each %$pending_delete_hash) {
4573         next if $selection && !$selection->{$opt};
4574         eval {
4575             if ($opt eq 'hotplug') {
4576                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4577             } elsif ($opt eq 'tablet') {
4578                 die "skip\n" if !$hotplug_features->{usb};
4579                 if ($defaults->{tablet}) {
4580                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4581                 } else {
4582                     vm_deviceunplug($vmid, $conf, $opt);
4583                 }
4584             } elsif ($opt =~ m/^usb\d+/) {
4585                 die "skip\n";
4586                 # since we cannot reliably hot unplug usb devices
4587                 # we are disabling it
4588                 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4589                 vm_deviceunplug($vmid, $conf, $opt);
4590             } elsif ($opt eq 'vcpus') {
4591                 die "skip\n" if !$hotplug_features->{cpu};
4592                 qemu_cpu_hotplug($vmid, $conf, undef);
4593             } elsif ($opt eq 'balloon') {
4594                 # enable balloon device is not hotpluggable
4595                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4596                 # here we reset the ballooning value to memory
4597                 my $balloon = $conf->{memory} || $defaults->{memory};
4598                 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4599             } elsif ($fast_plug_option->{$opt}) {
4600                 # do nothing
4601             } elsif ($opt =~ m/^net(\d+)$/) {
4602                 die "skip\n" if !$hotplug_features->{network};
4603                 vm_deviceunplug($vmid, $conf, $opt);
4604             } elsif (is_valid_drivename($opt)) {
4605                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4606                 vm_deviceunplug($vmid, $conf, $opt);
4607                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4608             } elsif ($opt =~ m/^memory$/) {
4609                 die "skip\n" if !$hotplug_features->{memory};
4610                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4611             } elsif ($opt eq 'cpuunits') {
4612                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4613             } elsif ($opt eq 'cpulimit') {
4614                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4615             } else {
4616                 die "skip\n";
4617             }
4618         };
4619         if (my $err = $@) {
4620             &$add_error($opt, $err) if $err ne "skip\n";
4621         } else {
4622             # save new config if hotplug was successful
4623             delete $conf->{$opt};
4624             vmconfig_undelete_pending_option($conf, $opt);
4625             PVE::QemuConfig->write_config($vmid, $conf);
4626             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4627         }
4628     }
4629
4630     my $apply_pending_cloudinit;
4631     $apply_pending_cloudinit = sub {
4632         my ($key, $value) = @_;
4633         $apply_pending_cloudinit = sub {}; # once is enough
4634
4635         my @cloudinit_opts = keys %$confdesc_cloudinit;
4636         foreach my $opt (keys %{$conf->{pending}}) {
4637             next if !grep { $_ eq $opt } @cloudinit_opts;
4638             $conf->{$opt} = delete $conf->{pending}->{$opt};
4639         }
4640
4641         my $new_conf = { %$conf };
4642         $new_conf->{$key} = $value;
4643         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4644     };
4645
4646     foreach my $opt (keys %{$conf->{pending}}) {
4647         next if $selection && !$selection->{$opt};
4648         my $value = $conf->{pending}->{$opt};
4649         eval {
4650             if ($opt eq 'hotplug') {
4651                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4652             } elsif ($opt eq 'tablet') {
4653                 die "skip\n" if !$hotplug_features->{usb};
4654                 if ($value == 1) {
4655                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4656                 } elsif ($value == 0) {
4657                     vm_deviceunplug($vmid, $conf, $opt);
4658                 }
4659             } elsif ($opt =~ m/^usb\d+$/) {
4660                 die "skip\n";
4661                 # since we cannot reliably hot unplug usb devices
4662                 # we are disabling it
4663                 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4664                 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4665                 die "skip\n" if !$d;
4666                 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d);
4667             } elsif ($opt eq 'vcpus') {
4668                 die "skip\n" if !$hotplug_features->{cpu};
4669                 qemu_cpu_hotplug($vmid, $conf, $value);
4670             } elsif ($opt eq 'balloon') {
4671                 # enable/disable balloning device is not hotpluggable
4672                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4673                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4674                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4675
4676                 # allow manual ballooning if shares is set to zero
4677                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4678                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4679                     vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4680                 }
4681             } elsif ($opt =~ m/^net(\d+)$/) {
4682                 # some changes can be done without hotplug
4683                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4684                                     $vmid, $opt, $value);
4685             } elsif (is_valid_drivename($opt)) {
4686                 # some changes can be done without hotplug
4687                 my $drive = parse_drive($opt, $value);
4688                 if (drive_is_cloudinit($drive)) {
4689                     &$apply_pending_cloudinit($opt, $value);
4690                 }
4691                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4692                                      $vmid, $opt, $value, 1);
4693             } elsif ($opt =~ m/^memory$/) { #dimms
4694                 die "skip\n" if !$hotplug_features->{memory};
4695                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4696             } elsif ($opt eq 'cpuunits') {
4697                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4698             } elsif ($opt eq 'cpulimit') {
4699                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4700                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4701             } else {
4702                 die "skip\n";  # skip non-hot-pluggable options
4703             }
4704         };
4705         if (my $err = $@) {
4706             &$add_error($opt, $err) if $err ne "skip\n";
4707         } else {
4708             # save new config if hotplug was successful
4709             $conf->{$opt} = $value;
4710             delete $conf->{pending}->{$opt};
4711             PVE::QemuConfig->write_config($vmid, $conf);
4712             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4713         }
4714     }
4715 }
4716
4717 sub try_deallocate_drive {
4718     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4719
4720     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4721         my $volid = $drive->{file};
4722         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4723             my $sid = PVE::Storage::parse_volume_id($volid);
4724             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4725
4726             # check if the disk is really unused
4727             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4728                 if is_volume_in_use($storecfg, $conf, $key, $volid);
4729             PVE::Storage::vdisk_free($storecfg, $volid);
4730             return 1;
4731         } else {
4732             # If vm is not owner of this disk remove from config
4733             return 1;
4734         }
4735     }
4736
4737     return undef;
4738 }
4739
4740 sub vmconfig_delete_or_detach_drive {
4741     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4742
4743     my $drive = parse_drive($opt, $conf->{$opt});
4744
4745     my $rpcenv = PVE::RPCEnvironment::get();
4746     my $authuser = $rpcenv->get_user();
4747
4748     if ($force) {
4749         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4750         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4751     } else {
4752         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4753     }
4754 }
4755
4756 sub vmconfig_apply_pending {
4757     my ($vmid, $conf, $storecfg) = @_;
4758
4759     # cold plug
4760
4761     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4762     while (my ($opt, $force) = each %$pending_delete_hash) {
4763         die "internal error" if $opt =~ m/^unused/;
4764         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4765         if (!defined($conf->{$opt})) {
4766             vmconfig_undelete_pending_option($conf, $opt);
4767             PVE::QemuConfig->write_config($vmid, $conf);
4768         } elsif (is_valid_drivename($opt)) {
4769             vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4770             vmconfig_undelete_pending_option($conf, $opt);
4771             delete $conf->{$opt};
4772             PVE::QemuConfig->write_config($vmid, $conf);
4773         } else {
4774             vmconfig_undelete_pending_option($conf, $opt);
4775             delete $conf->{$opt};
4776             PVE::QemuConfig->write_config($vmid, $conf);
4777         }
4778     }
4779
4780     $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4781
4782     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4783         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4784
4785         if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
4786             # skip if nothing changed
4787         } elsif (is_valid_drivename($opt)) {
4788             vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4789                 if defined($conf->{$opt});
4790             $conf->{$opt} = $conf->{pending}->{$opt};
4791         } else {
4792             $conf->{$opt} = $conf->{pending}->{$opt};
4793         }
4794
4795         delete $conf->{pending}->{$opt};
4796         PVE::QemuConfig->write_config($vmid, $conf);
4797     }
4798 }
4799
4800 my $safe_num_ne = sub {
4801     my ($a, $b) = @_;
4802
4803     return 0 if !defined($a) && !defined($b);
4804     return 1 if !defined($a);
4805     return 1 if !defined($b);
4806
4807     return $a != $b;
4808 };
4809
4810 my $safe_string_ne = sub {
4811     my ($a, $b) = @_;
4812
4813     return 0 if !defined($a) && !defined($b);
4814     return 1 if !defined($a);
4815     return 1 if !defined($b);
4816
4817     return $a ne $b;
4818 };
4819
4820 sub vmconfig_update_net {
4821     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4822
4823     my $newnet = parse_net($value);
4824
4825     if ($conf->{$opt}) {
4826         my $oldnet = parse_net($conf->{$opt});
4827
4828         if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4829             &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4830             &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4831             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4832
4833             # for non online change, we try to hot-unplug
4834             die "skip\n" if !$hotplug;
4835             vm_deviceunplug($vmid, $conf, $opt);
4836         } else {
4837
4838             die "internal error" if $opt !~ m/net(\d+)/;
4839             my $iface = "tap${vmid}i$1";
4840
4841             if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4842                 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4843                 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4844                 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4845                 PVE::Network::tap_unplug($iface);
4846                 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4847             } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4848                 # Rate can be applied on its own but any change above needs to
4849                 # include the rate in tap_plug since OVS resets everything.
4850                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4851             }
4852
4853             if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4854                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4855             }
4856
4857             return 1;
4858         }
4859     }
4860
4861     if ($hotplug) {
4862         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4863     } else {
4864         die "skip\n";
4865     }
4866 }
4867
4868 sub vmconfig_update_disk {
4869     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4870
4871     # fixme: do we need force?
4872
4873     my $drive = parse_drive($opt, $value);
4874
4875     if ($conf->{$opt}) {
4876
4877         if (my $old_drive = parse_drive($opt, $conf->{$opt}))  {
4878
4879             my $media = $drive->{media} || 'disk';
4880             my $oldmedia = $old_drive->{media} || 'disk';
4881             die "unable to change media type\n" if $media ne $oldmedia;
4882
4883             if (!drive_is_cdrom($old_drive)) {
4884
4885                 if ($drive->{file} ne $old_drive->{file}) {
4886
4887                     die "skip\n" if !$hotplug;
4888
4889                     # unplug and register as unused
4890                     vm_deviceunplug($vmid, $conf, $opt);
4891                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4892
4893                 } else {
4894                     # update existing disk
4895
4896                     # skip non hotpluggable value
4897                     if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4898                         &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4899                         &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4900                         &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4901                         die "skip\n";
4902                     }
4903
4904                     # apply throttle
4905                     if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4906                         &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4907                         &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4908                         &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4909                         &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4910                         &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4911                         &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4912                         &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4913                         &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4914                         &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4915                         &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4916                         &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4917                         &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4918                         &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4919                         &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4920                         &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4921                         &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4922                         &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4923
4924                         qemu_block_set_io_throttle($vmid,"drive-$opt",
4925                                                    ($drive->{mbps} || 0)*1024*1024,
4926                                                    ($drive->{mbps_rd} || 0)*1024*1024,
4927                                                    ($drive->{mbps_wr} || 0)*1024*1024,
4928                                                    $drive->{iops} || 0,
4929                                                    $drive->{iops_rd} || 0,
4930                                                    $drive->{iops_wr} || 0,
4931                                                    ($drive->{mbps_max} || 0)*1024*1024,
4932                                                    ($drive->{mbps_rd_max} || 0)*1024*1024,
4933                                                    ($drive->{mbps_wr_max} || 0)*1024*1024,
4934                                                    $drive->{iops_max} || 0,
4935                                                    $drive->{iops_rd_max} || 0,
4936                                                    $drive->{iops_wr_max} || 0,
4937                                                    $drive->{bps_max_length} || 1,
4938                                                    $drive->{bps_rd_max_length} || 1,
4939                                                    $drive->{bps_wr_max_length} || 1,
4940                                                    $drive->{iops_max_length} || 1,
4941                                                    $drive->{iops_rd_max_length} || 1,
4942                                                    $drive->{iops_wr_max_length} || 1);
4943
4944                     }
4945
4946                     return 1;
4947                 }
4948
4949             } else { # cdrom
4950
4951                 if ($drive->{file} eq 'none') {
4952                     vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4953                     if (drive_is_cloudinit($old_drive)) {
4954                         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4955                     }
4956                 } else {
4957                     my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4958                     vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4959                     vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4960                 }
4961
4962                 return 1;
4963             }
4964         }
4965     }
4966
4967     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4968     # hotplug new disks
4969     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4970     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4971 }
4972
4973 sub vm_start {
4974     my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
4975         $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
4976
4977     PVE::QemuConfig->lock_config($vmid, sub {
4978         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
4979
4980         die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4981
4982         PVE::QemuConfig->check_lock($conf) if !$skiplock;
4983
4984         die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4985
4986         if (!$statefile && scalar(keys %{$conf->{pending}})) {
4987             vmconfig_apply_pending($vmid, $conf, $storecfg);
4988             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4989         }
4990
4991         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4992
4993         my $defaults = load_defaults();
4994
4995         # set environment variable useful inside network script
4996         $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4997
4998         my $local_volumes = {};
4999
5000         if ($targetstorage) {
5001             foreach_drive($conf, sub {
5002                 my ($ds, $drive) = @_;
5003
5004                 return if drive_is_cdrom($drive);
5005
5006                 my $volid = $drive->{file};
5007
5008                 return if !$volid;
5009
5010                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5011
5012                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5013                 return if $scfg->{shared};
5014                 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5015             });
5016
5017             my $format = undef;
5018
5019             foreach my $opt (sort keys %$local_volumes) {
5020
5021                 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5022                 my $drive = parse_drive($opt, $conf->{$opt});
5023
5024                 #if remote storage is specified, use default format
5025                 if ($targetstorage && $targetstorage ne "1") {
5026                     $storeid = $targetstorage;
5027                     my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5028                     $format = $defFormat;
5029                 } else {
5030                     #else we use same format than original
5031                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5032                     $format = qemu_img_format($scfg, $volid);
5033                 }
5034
5035                 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5036                 my $newdrive = $drive;
5037                 $newdrive->{format} = $format;
5038                 $newdrive->{file} = $newvolid;
5039                 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5040                 $local_volumes->{$opt} = $drivestr;
5041                 #pass drive to conf for command line
5042                 $conf->{$opt} = $drivestr;
5043             }
5044         }
5045
5046         my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5047
5048         my $migrate_port = 0;
5049         my $migrate_uri;
5050         if ($statefile) {
5051             if ($statefile eq 'tcp') {
5052                 my $localip = "localhost";
5053                 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5054                 my $nodename = PVE::INotify::nodename();
5055
5056                 if (!defined($migration_type)) {
5057                     if (defined($datacenterconf->{migration}->{type})) {
5058                         $migration_type = $datacenterconf->{migration}->{type};
5059                     } else {
5060                         $migration_type = 'secure';
5061                     }
5062                 }
5063
5064                 if ($migration_type eq 'insecure') {
5065                     my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5066                     if ($migrate_network_addr) {
5067                         $localip = $migrate_network_addr;
5068                     } else {
5069                         $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5070                     }
5071
5072                     $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5073                 }
5074
5075                 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5076                 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5077                 $migrate_uri = "tcp:${localip}:${migrate_port}";
5078                 push @$cmd, '-incoming', $migrate_uri;
5079                 push @$cmd, '-S';
5080
5081             } elsif ($statefile eq 'unix') {
5082                 # should be default for secure migrations as a ssh TCP forward
5083                 # tunnel is not deterministic reliable ready and fails regurarly
5084                 # to set up in time, so use UNIX socket forwards
5085                 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5086                 unlink $socket_addr;
5087
5088                 $migrate_uri = "unix:$socket_addr";
5089
5090                 push @$cmd, '-incoming', $migrate_uri;
5091                 push @$cmd, '-S';
5092
5093             } else {
5094                 push @$cmd, '-loadstate', $statefile;
5095             }
5096         } elsif ($paused) {
5097             push @$cmd, '-S';
5098         }
5099
5100         # host pci devices
5101         for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
5102           my $d = parse_hostpci($conf->{"hostpci$i"});
5103           next if !$d;
5104           my $pcidevices = $d->{pciid};
5105           foreach my $pcidevice (@$pcidevices) {
5106                 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
5107
5108                 my $info = pci_device_info("0000:$pciid");
5109                 die "IOMMU not present\n" if !check_iommu_support();
5110                 die "no pci device info for device '$pciid'\n" if !$info;
5111                 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
5112                 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
5113           }
5114         }
5115
5116         PVE::Storage::activate_volumes($storecfg, $vollist);
5117
5118         if (!check_running($vmid, 1)) {
5119             eval {
5120                 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5121                     outfunc => sub {}, errfunc => sub {});
5122             };
5123         }
5124
5125         my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5126                                                   : $defaults->{cpuunits};
5127
5128         my $start_timeout = $conf->{hugepages} ? 300 : 30;
5129         my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5130
5131         my %properties = (
5132             Slice => 'qemu.slice',
5133             KillMode => 'none',
5134             CPUShares => $cpuunits
5135         );
5136
5137         if (my $cpulimit = $conf->{cpulimit}) {
5138             $properties{CPUQuota} = int($cpulimit * 100);
5139         }
5140         $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5141
5142         my $run_qemu = sub {
5143             PVE::Tools::run_fork sub {
5144                 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5145                 run_command($cmd, %run_params);
5146             };
5147         };
5148
5149         if ($conf->{hugepages}) {
5150
5151             my $code = sub {
5152                 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5153                 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5154
5155                 PVE::QemuServer::Memory::hugepages_mount();
5156                 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5157
5158                 eval { $run_qemu->() };
5159                 if (my $err = $@) {
5160                     PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5161                     die $err;
5162                 }
5163
5164                 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5165             };
5166             eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5167
5168         } else {
5169             eval { $run_qemu->() };
5170         }
5171
5172         if (my $err = $@) {
5173             # deactivate volumes if start fails
5174             eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5175             die "start failed: $err";
5176         }
5177
5178         print "migration listens on $migrate_uri\n" if $migrate_uri;
5179
5180         if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5181             eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5182             warn $@ if $@;
5183         }
5184
5185         #start nbd server for storage migration
5186         if ($targetstorage) {
5187             my $nodename = PVE::INotify::nodename();
5188             my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5189             my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5190             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5191             $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5192
5193             vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${migrate_port}" } } );
5194
5195             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5196
5197             foreach my $opt (sort keys %$local_volumes) {
5198                 my $volid = $local_volumes->{$opt};
5199                 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5200                 my $migrate_storage_uri = "nbd:${localip}:${migrate_port}:exportname=drive-$opt";
5201                 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5202             }
5203         }
5204
5205         if ($migratedfrom) {
5206             eval {
5207                 set_migration_caps($vmid);
5208             };
5209             warn $@ if $@;
5210
5211             if ($spice_port) {
5212                 print "spice listens on port $spice_port\n";
5213                 if ($spice_ticket) {
5214                     vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5215                     vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5216                 }
5217             }
5218
5219         } else {
5220             vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5221                 if !$statefile && $conf->{balloon};
5222
5223             foreach my $opt (keys %$conf) {
5224                 next if $opt !~  m/^net\d+$/;
5225                 my $nicconf = parse_net($conf->{$opt});
5226                 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5227             }
5228         }
5229
5230         vm_mon_cmd_nocheck($vmid, 'qom-set',
5231                     path => "machine/peripheral/balloon0",
5232                     property => "guest-stats-polling-interval",
5233                     value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5234
5235     });
5236 }
5237
5238 sub vm_mon_cmd {
5239     my ($vmid, $execute, %params) = @_;
5240
5241     my $cmd = { execute => $execute, arguments => \%params };
5242     vm_qmp_command($vmid, $cmd);
5243 }
5244
5245 sub vm_mon_cmd_nocheck {
5246     my ($vmid, $execute, %params) = @_;
5247
5248     my $cmd = { execute => $execute, arguments => \%params };
5249     vm_qmp_command($vmid, $cmd, 1);
5250 }
5251
5252 sub vm_qmp_command {
5253     my ($vmid, $cmd, $nocheck) = @_;
5254
5255     my $res;
5256
5257     my $timeout;
5258     if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
5259         $timeout = $cmd->{arguments}->{timeout};
5260         delete $cmd->{arguments}->{timeout};
5261     }
5262
5263     eval {
5264         die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5265         my $sname = qmp_socket($vmid);
5266         if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5267             my $qmpclient = PVE::QMPClient->new();
5268
5269             $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5270         } else {
5271             die "unable to open monitor socket\n";
5272         }
5273     };
5274     if (my $err = $@) {
5275         syslog("err", "VM $vmid qmp command failed - $err");
5276         die $err;
5277     }
5278
5279     return $res;
5280 }
5281
5282 sub vm_human_monitor_command {
5283     my ($vmid, $cmdline) = @_;
5284
5285     my $res;
5286
5287     my $cmd = {
5288         execute => 'human-monitor-command',
5289         arguments => { 'command-line' => $cmdline},
5290     };
5291
5292     return vm_qmp_command($vmid, $cmd);
5293 }
5294
5295 sub vm_commandline {
5296     my ($storecfg, $vmid) = @_;
5297
5298     my $conf = PVE::QemuConfig->load_config($vmid);
5299
5300     my $defaults = load_defaults();
5301
5302     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5303
5304     return PVE::Tools::cmd2string($cmd);
5305 }
5306
5307 sub vm_reset {
5308     my ($vmid, $skiplock) = @_;
5309
5310     PVE::QemuConfig->lock_config($vmid, sub {
5311
5312         my $conf = PVE::QemuConfig->load_config($vmid);
5313
5314         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5315
5316         vm_mon_cmd($vmid, "system_reset");
5317     });
5318 }
5319
5320 sub get_vm_volumes {
5321     my ($conf) = @_;
5322
5323     my $vollist = [];
5324     foreach_volid($conf, sub {
5325         my ($volid, $attr) = @_;
5326
5327         return if $volid =~ m|^/|;
5328
5329         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5330         return if !$sid;
5331
5332         push @$vollist, $volid;
5333     });
5334
5335     return $vollist;
5336 }
5337
5338 sub vm_stop_cleanup {
5339     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5340
5341     eval {
5342
5343         if (!$keepActive) {
5344             my $vollist = get_vm_volumes($conf);
5345             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5346         }
5347
5348         foreach my $ext (qw(mon qmp pid vnc qga)) {
5349             unlink "/var/run/qemu-server/${vmid}.$ext";
5350         }
5351
5352         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5353     };
5354     warn $@ if $@; # avoid errors - just warn
5355 }
5356
5357 # Note: use $nockeck to skip tests if VM configuration file exists.
5358 # We need that when migration VMs to other nodes (files already moved)
5359 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5360 sub vm_stop {
5361     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5362
5363     $force = 1 if !defined($force) && !$shutdown;
5364
5365     if ($migratedfrom){
5366         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5367         kill 15, $pid if $pid;
5368         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5369         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5370         return;
5371     }
5372
5373     PVE::QemuConfig->lock_config($vmid, sub {
5374
5375         my $pid = check_running($vmid, $nocheck);
5376         return if !$pid;
5377
5378         my $conf;
5379         if (!$nocheck) {
5380             $conf = PVE::QemuConfig->load_config($vmid);
5381             PVE::QemuConfig->check_lock($conf) if !$skiplock;
5382             if (!defined($timeout) && $shutdown && $conf->{startup}) {
5383                 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5384                 $timeout = $opts->{down} if $opts->{down};
5385             }
5386         }
5387
5388         $timeout = 60 if !defined($timeout);
5389
5390         eval {
5391             if ($shutdown) {
5392                 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5393                     vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
5394                 } else {
5395                     vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5396                 }
5397             } else {
5398                 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5399             }
5400         };
5401         my $err = $@;
5402
5403         if (!$err) {
5404             my $count = 0;
5405             while (($count < $timeout) && check_running($vmid, $nocheck)) {
5406                 $count++;
5407                 sleep 1;
5408             }
5409
5410             if ($count >= $timeout) {
5411                 if ($force) {
5412                     warn "VM still running - terminating now with SIGTERM\n";
5413                     kill 15, $pid;
5414                 } else {
5415                     die "VM quit/powerdown failed - got timeout\n";
5416                 }
5417             } else {
5418                 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5419                 return;
5420             }
5421         } else {
5422             if ($force) {
5423                 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5424                 kill 15, $pid;
5425             } else {
5426                 die "VM quit/powerdown failed\n";
5427             }
5428         }
5429
5430         # wait again
5431         $timeout = 10;
5432
5433         my $count = 0;
5434         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5435             $count++;
5436             sleep 1;
5437         }
5438
5439         if ($count >= $timeout) {
5440             warn "VM still running - terminating now with SIGKILL\n";
5441             kill 9, $pid;
5442             sleep 1;
5443         }
5444
5445         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5446    });
5447 }
5448
5449 sub vm_suspend {
5450     my ($vmid, $skiplock) = @_;
5451
5452     PVE::QemuConfig->lock_config($vmid, sub {
5453
5454         my $conf = PVE::QemuConfig->load_config($vmid);
5455
5456         PVE::QemuConfig->check_lock($conf)
5457             if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5458
5459         vm_mon_cmd($vmid, "stop");
5460     });
5461 }
5462
5463 sub vm_resume {
5464     my ($vmid, $skiplock, $nocheck) = @_;
5465
5466     PVE::QemuConfig->lock_config($vmid, sub {
5467
5468         my $res = vm_mon_cmd($vmid, 'query-status');
5469         my $resume_cmd = 'cont';
5470
5471         if ($res->{status} && $res->{status} eq 'suspended') {
5472             $resume_cmd = 'system_wakeup';
5473         }
5474
5475         if (!$nocheck) {
5476
5477             my $conf = PVE::QemuConfig->load_config($vmid);
5478
5479             PVE::QemuConfig->check_lock($conf)
5480                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5481
5482             vm_mon_cmd($vmid, $resume_cmd);
5483
5484         } else {
5485             vm_mon_cmd_nocheck($vmid, $resume_cmd);
5486         }
5487     });
5488 }
5489
5490 sub vm_sendkey {
5491     my ($vmid, $skiplock, $key) = @_;
5492
5493     PVE::QemuConfig->lock_config($vmid, sub {
5494
5495         my $conf = PVE::QemuConfig->load_config($vmid);
5496
5497         # there is no qmp command, so we use the human monitor command
5498         vm_human_monitor_command($vmid, "sendkey $key");
5499     });
5500 }
5501
5502 sub vm_destroy {
5503     my ($storecfg, $vmid, $skiplock) = @_;
5504
5505     PVE::QemuConfig->lock_config($vmid, sub {
5506
5507         my $conf = PVE::QemuConfig->load_config($vmid);
5508
5509         if (!check_running($vmid)) {
5510             destroy_vm($storecfg, $vmid, undef, $skiplock);
5511         } else {
5512             die "VM $vmid is running - destroy failed\n";
5513         }
5514     });
5515 }
5516
5517 # pci helpers
5518
5519 sub file_write {
5520     my ($filename, $buf) = @_;
5521
5522     my $fh = IO::File->new($filename, "w");
5523     return undef if !$fh;
5524
5525     my $res = print $fh $buf;
5526
5527     $fh->close();
5528
5529     return $res;
5530 }
5531
5532 sub pci_device_info {
5533     my ($name) = @_;
5534
5535     my $res;
5536
5537     return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
5538     my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
5539
5540     my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
5541     return undef if !defined($irq) || $irq !~ m/^\d+$/;
5542
5543     my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
5544     return undef if !defined($vendor) || $vendor !~ s/^0x//;
5545
5546     my $product = file_read_firstline("$pcisysfs/devices/$name/device");
5547     return undef if !defined($product) || $product !~ s/^0x//;
5548
5549     $res = {
5550         name => $name,
5551         vendor => $vendor,
5552         product => $product,
5553         domain => $domain,
5554         bus => $bus,
5555         slot => $slot,
5556         func => $func,
5557         irq => $irq,
5558         has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
5559     };
5560
5561     return $res;
5562 }
5563
5564 sub pci_dev_reset {
5565     my ($dev) = @_;
5566
5567     my $name = $dev->{name};
5568
5569     my $fn = "$pcisysfs/devices/$name/reset";
5570
5571     return file_write($fn, "1");
5572 }
5573
5574 sub pci_dev_bind_to_vfio {
5575     my ($dev) = @_;
5576
5577     my $name = $dev->{name};
5578
5579     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5580
5581     if (!-d $vfio_basedir) {
5582         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5583     }
5584     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5585
5586     my $testdir = "$vfio_basedir/$name";
5587     return 1 if -d $testdir;
5588
5589     my $data = "$dev->{vendor} $dev->{product}";
5590     return undef if !file_write("$vfio_basedir/new_id", $data);
5591
5592     my $fn = "$pcisysfs/devices/$name/driver/unbind";
5593     if (!file_write($fn, $name)) {
5594         return undef if -f $fn;
5595     }
5596
5597     $fn = "$vfio_basedir/bind";
5598     if (! -d $testdir) {
5599         return undef if !file_write($fn, $name);
5600     }
5601
5602     return -d $testdir;
5603 }
5604
5605 sub pci_dev_group_bind_to_vfio {
5606     my ($pciid) = @_;
5607
5608     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5609
5610     if (!-d $vfio_basedir) {
5611         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5612     }
5613     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5614
5615     # get IOMMU group devices
5616     opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
5617       my @devs = grep /^0000:/, readdir($D);
5618     closedir($D);
5619
5620     foreach my $pciid (@devs) {
5621         $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
5622
5623         # pci bridges, switches or root ports are not supported
5624         # they have a pci_bus subdirectory so skip them
5625         next if (-e "$pcisysfs/devices/$pciid/pci_bus");
5626
5627         my $info = pci_device_info($1);
5628         pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
5629     }
5630
5631     return 1;
5632 }
5633
5634 # vzdump restore implementaion
5635
5636 sub tar_archive_read_firstfile {
5637     my $archive = shift;
5638
5639     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5640
5641     # try to detect archive type first
5642     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5643         die "unable to open file '$archive'\n";
5644     my $firstfile = <$fh>;
5645     kill 15, $pid;
5646     close $fh;
5647
5648     die "ERROR: archive contaions no data\n" if !$firstfile;
5649     chomp $firstfile;
5650
5651     return $firstfile;
5652 }
5653
5654 sub tar_restore_cleanup {
5655     my ($storecfg, $statfile) = @_;
5656
5657     print STDERR "starting cleanup\n";
5658
5659     if (my $fd = IO::File->new($statfile, "r")) {
5660         while (defined(my $line = <$fd>)) {
5661             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5662                 my $volid = $2;
5663                 eval {
5664                     if ($volid =~ m|^/|) {
5665                         unlink $volid || die 'unlink failed\n';
5666                     } else {
5667                         PVE::Storage::vdisk_free($storecfg, $volid);
5668                     }
5669                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5670                 };
5671                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5672             } else {
5673                 print STDERR "unable to parse line in statfile - $line";
5674             }
5675         }
5676         $fd->close();
5677     }
5678 }
5679
5680 sub restore_archive {
5681     my ($archive, $vmid, $user, $opts) = @_;
5682
5683     my $format = $opts->{format};
5684     my $comp;
5685
5686     if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5687         $format = 'tar' if !$format;
5688         $comp = 'gzip';
5689     } elsif ($archive =~ m/\.tar$/) {
5690         $format = 'tar' if !$format;
5691     } elsif ($archive =~ m/.tar.lzo$/) {
5692         $format = 'tar' if !$format;
5693         $comp = 'lzop';
5694     } elsif ($archive =~ m/\.vma$/) {
5695         $format = 'vma' if !$format;
5696     } elsif ($archive =~ m/\.vma\.gz$/) {
5697         $format = 'vma' if !$format;
5698         $comp = 'gzip';
5699     } elsif ($archive =~ m/\.vma\.lzo$/) {
5700         $format = 'vma' if !$format;
5701         $comp = 'lzop';
5702     } else {
5703         $format = 'vma' if !$format; # default
5704     }
5705
5706     # try to detect archive format
5707     if ($format eq 'tar') {
5708         return restore_tar_archive($archive, $vmid, $user, $opts);
5709     } else {
5710         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5711     }
5712 }
5713
5714 sub restore_update_config_line {
5715     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5716
5717     return if $line =~ m/^\#qmdump\#/;
5718     return if $line =~ m/^\#vzdump\#/;
5719     return if $line =~ m/^lock:/;
5720     return if $line =~ m/^unused\d+:/;
5721     return if $line =~ m/^parent:/;
5722     return if $line =~ m/^template:/; # restored VM is never a template
5723
5724     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5725     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5726         # try to convert old 1.X settings
5727         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5728         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5729             my ($model, $macaddr) = split(/\=/, $devconfig);
5730             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5731             my $net = {
5732                 model => $model,
5733                 bridge => "vmbr$ind",
5734                 macaddr => $macaddr,
5735             };
5736             my $netstr = print_net($net);
5737
5738             print $outfd "net$cookie->{netcount}: $netstr\n";
5739             $cookie->{netcount}++;
5740         }
5741     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5742         my ($id, $netstr) = ($1, $2);
5743         my $net = parse_net($netstr);
5744         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5745         $netstr = print_net($net);
5746         print $outfd "$id: $netstr\n";
5747     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5748         my $virtdev = $1;
5749         my $value = $3;
5750         my $di = parse_drive($virtdev, $value);
5751         if (defined($di->{backup}) && !$di->{backup}) {
5752             print $outfd "#$line";
5753         } elsif ($map->{$virtdev}) {
5754             delete $di->{format}; # format can change on restore
5755             $di->{file} = $map->{$virtdev};
5756             $value = print_drive($vmid, $di);
5757             print $outfd "$virtdev: $value\n";
5758         } else {
5759             print $outfd $line;
5760         }
5761     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5762         my $vmgenid = $1;
5763         if ($vmgenid ne '0') {
5764             # always generate a new vmgenid if there was a valid one setup
5765             $vmgenid = generate_uuid();
5766         }
5767         print $outfd "vmgenid: $vmgenid\n";
5768     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5769         my ($uuid, $uuid_str);
5770         UUID::generate($uuid);
5771         UUID::unparse($uuid, $uuid_str);
5772         my $smbios1 = parse_smbios1($2);
5773         $smbios1->{uuid} = $uuid_str;
5774         print $outfd $1.print_smbios1($smbios1)."\n";
5775     } else {
5776         print $outfd $line;
5777     }
5778 }
5779
5780 sub scan_volids {
5781     my ($cfg, $vmid) = @_;
5782
5783     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5784
5785     my $volid_hash = {};
5786     foreach my $storeid (keys %$info) {
5787         foreach my $item (@{$info->{$storeid}}) {
5788             next if !($item->{volid} && $item->{size});
5789             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5790             $volid_hash->{$item->{volid}} = $item;
5791         }
5792     }
5793
5794     return $volid_hash;
5795 }
5796
5797 sub is_volume_in_use {
5798     my ($storecfg, $conf, $skip_drive, $volid) = @_;
5799
5800     my $path = PVE::Storage::path($storecfg, $volid);
5801
5802     my $scan_config = sub {
5803         my ($cref, $snapname) = @_;
5804
5805         foreach my $key (keys %$cref) {
5806             my $value = $cref->{$key};
5807             if (is_valid_drivename($key)) {
5808                 next if $skip_drive && $key eq $skip_drive;
5809                 my $drive = parse_drive($key, $value);
5810                 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
5811                 return 1 if $volid eq $drive->{file};
5812                 if ($drive->{file} =~ m!^/!) {
5813                     return 1 if $drive->{file} eq $path;
5814                 } else {
5815                     my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
5816                     next if !$storeid;
5817                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
5818                     next if !$scfg;
5819                     return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
5820                 }
5821             }
5822         }
5823
5824         return 0;
5825     };
5826
5827     return 1 if &$scan_config($conf);
5828
5829     undef $skip_drive;
5830
5831     foreach my $snapname (keys %{$conf->{snapshots}}) {
5832         return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
5833     }
5834
5835     return 0;
5836 }
5837
5838 sub update_disksize {
5839     my ($vmid, $conf, $volid_hash) = @_;
5840
5841     my $changes;
5842     my $prefix = "VM $vmid:";
5843
5844     # used and unused disks
5845     my $referenced = {};
5846
5847     # Note: it is allowed to define multiple storages with same path (alias), so
5848     # we need to check both 'volid' and real 'path' (two different volid can point
5849     # to the same path).
5850
5851     my $referencedpath = {};
5852
5853     # update size info
5854     foreach my $opt (keys %$conf) {
5855         if (is_valid_drivename($opt)) {
5856             my $drive = parse_drive($opt, $conf->{$opt});
5857             my $volid = $drive->{file};
5858             next if !$volid;
5859
5860             $referenced->{$volid} = 1;
5861             if ($volid_hash->{$volid} &&
5862                 (my $path = $volid_hash->{$volid}->{path})) {
5863                 $referencedpath->{$path} = 1;
5864             }
5865
5866             next if drive_is_cdrom($drive);
5867             next if !$volid_hash->{$volid};
5868
5869             $drive->{size} = $volid_hash->{$volid}->{size};
5870             my $new = print_drive($vmid, $drive);
5871             if ($new ne $conf->{$opt}) {
5872                 $changes = 1;
5873                 $conf->{$opt} = $new;
5874                 print "$prefix update disk '$opt' information.\n";
5875             }
5876         }
5877     }
5878
5879     # remove 'unusedX' entry if volume is used
5880     foreach my $opt (keys %$conf) {
5881         next if $opt !~ m/^unused\d+$/;
5882         my $volid = $conf->{$opt};
5883         my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5884         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5885             print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
5886             $changes = 1;
5887             delete $conf->{$opt};
5888         }
5889
5890         $referenced->{$volid} = 1;
5891         $referencedpath->{$path} = 1 if $path;
5892     }
5893
5894     foreach my $volid (sort keys %$volid_hash) {
5895         next if $volid =~ m/vm-$vmid-state-/;
5896         next if $referenced->{$volid};
5897         my $path = $volid_hash->{$volid}->{path};
5898         next if !$path; # just to be sure
5899         next if $referencedpath->{$path};
5900         $changes = 1;
5901         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5902         print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
5903         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5904     }
5905
5906     return $changes;
5907 }
5908
5909 sub rescan {
5910     my ($vmid, $nolock, $dryrun) = @_;
5911
5912     my $cfg = PVE::Storage::config();
5913
5914     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5915     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
5916     foreach my $stor (keys %{$cfg->{ids}}) {
5917         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
5918     }
5919
5920     print "rescan volumes...\n";
5921     my $volid_hash = scan_volids($cfg, $vmid);
5922
5923     my $updatefn =  sub {
5924         my ($vmid) = @_;
5925
5926         my $conf = PVE::QemuConfig->load_config($vmid);
5927
5928         PVE::QemuConfig->check_lock($conf);
5929
5930         my $vm_volids = {};
5931         foreach my $volid (keys %$volid_hash) {
5932             my $info = $volid_hash->{$volid};
5933             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5934         }
5935
5936         my $changes = update_disksize($vmid, $conf, $vm_volids);
5937
5938         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
5939     };
5940
5941     if (defined($vmid)) {
5942         if ($nolock) {
5943             &$updatefn($vmid);
5944         } else {
5945             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5946         }
5947     } else {
5948         my $vmlist = config_list();
5949         foreach my $vmid (keys %$vmlist) {
5950             if ($nolock) {
5951                 &$updatefn($vmid);
5952             } else {
5953                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5954             }
5955         }
5956     }
5957 }
5958
5959 sub restore_vma_archive {
5960     my ($archive, $vmid, $user, $opts, $comp) = @_;
5961
5962     my $readfrom = $archive;
5963
5964     my $cfg = PVE::Storage::config();
5965     my $commands = [];
5966     my $bwlimit = $opts->{bwlimit};
5967
5968     my $dbg_cmdstring = '';
5969     my $add_pipe = sub {
5970         my ($cmd) = @_;
5971         push @$commands, $cmd;
5972         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
5973         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
5974         $readfrom = '-';
5975     };
5976
5977     my $input = undef;
5978     if ($archive eq '-') {
5979         $input = '<&STDIN';
5980     } else {
5981         # If we use a backup from a PVE defined storage we also consider that
5982         # storage's rate limit:
5983         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
5984         if (defined($volid)) {
5985             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
5986             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
5987             if ($readlimit) {
5988                 print STDERR "applying read rate limit: $readlimit\n";
5989                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
5990                 $add_pipe->($cstream);
5991             }
5992         }
5993     }
5994
5995     if ($comp) {
5996         my $cmd;
5997         if ($comp eq 'gzip') {
5998             $cmd = ['zcat', $readfrom];
5999         } elsif ($comp eq 'lzop') {
6000             $cmd = ['lzop', '-d', '-c', $readfrom];
6001         } else {
6002             die "unknown compression method '$comp'\n";
6003         }
6004         $add_pipe->($cmd);
6005     }
6006
6007     my $tmpdir = "/var/tmp/vzdumptmp$$";
6008     rmtree $tmpdir;
6009
6010     # disable interrupts (always do cleanups)
6011     local $SIG{INT} =
6012         local $SIG{TERM} =
6013         local $SIG{QUIT} =
6014         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6015
6016     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6017     POSIX::mkfifo($mapfifo, 0600);
6018     my $fifofh;
6019
6020     my $openfifo = sub {
6021         open($fifofh, '>', $mapfifo) || die $!;
6022     };
6023
6024     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6025
6026     my $oldtimeout;
6027     my $timeout = 5;
6028
6029     my $devinfo = {};
6030
6031     my $rpcenv = PVE::RPCEnvironment::get();
6032
6033     my $conffile = PVE::QemuConfig->config_file($vmid);
6034     my $tmpfn = "$conffile.$$.tmp";
6035
6036     # Note: $oldconf is undef if VM does not exists
6037     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6038     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6039
6040     my %storage_limits;
6041
6042     my $print_devmap = sub {
6043         my $virtdev_hash = {};
6044
6045         my $cfgfn = "$tmpdir/qemu-server.conf";
6046
6047         # we can read the config - that is already extracted
6048         my $fh = IO::File->new($cfgfn, "r") ||
6049             "unable to read qemu-server.conf - $!\n";
6050
6051         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6052         if (-f $fwcfgfn) {
6053             my $pve_firewall_dir = '/etc/pve/firewall';
6054             mkdir $pve_firewall_dir; # make sure the dir exists
6055             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6056         }
6057
6058         while (defined(my $line = <$fh>)) {
6059             if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6060                 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6061                 die "archive does not contain data for drive '$virtdev'\n"
6062                     if !$devinfo->{$devname};
6063                 if (defined($opts->{storage})) {
6064                     $storeid = $opts->{storage} || 'local';
6065                 } elsif (!$storeid) {
6066                     $storeid = 'local';
6067                 }
6068                 $format = 'raw' if !$format;
6069                 $devinfo->{$devname}->{devname} = $devname;
6070                 $devinfo->{$devname}->{virtdev} = $virtdev;
6071                 $devinfo->{$devname}->{format} = $format;
6072                 $devinfo->{$devname}->{storeid} = $storeid;
6073
6074                 # check permission on storage
6075                 my $pool = $opts->{pool}; # todo: do we need that?
6076                 if ($user ne 'root@pam') {
6077                     $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6078                 }
6079
6080                 $storage_limits{$storeid} = $bwlimit;
6081
6082                 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6083             }
6084         }
6085
6086         foreach my $key (keys %storage_limits) {
6087             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6088             next if !$limit;
6089             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6090             $storage_limits{$key} = $limit * 1024;
6091         }
6092
6093         foreach my $devname (keys %$devinfo) {
6094             die "found no device mapping information for device '$devname'\n"
6095                 if !$devinfo->{$devname}->{virtdev};
6096         }
6097
6098         # create empty/temp config
6099         if ($oldconf) {
6100             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6101             foreach_drive($oldconf, sub {
6102                 my ($ds, $drive) = @_;
6103
6104                 return if drive_is_cdrom($drive);
6105
6106                 my $volid = $drive->{file};
6107
6108                 return if !$volid || $volid =~ m|^/|;
6109
6110                 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6111                 return if !$path || !$owner || ($owner != $vmid);
6112
6113                 # Note: only delete disk we want to restore
6114                 # other volumes will become unused
6115                 if ($virtdev_hash->{$ds}) {
6116                     eval { PVE::Storage::vdisk_free($cfg, $volid); };
6117                     if (my $err = $@) {
6118                         warn $err;
6119                     }
6120                 }
6121             });
6122
6123             # delete vmstate files
6124             # since after the restore we have no snapshots anymore
6125             foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6126                 my $snap = $oldconf->{snapshots}->{$snapname};
6127                 if ($snap->{vmstate}) {
6128                     eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6129                     if (my $err = $@) {
6130                         warn $err;
6131                     }
6132                 }
6133             }
6134         }
6135
6136         my $map = {};
6137         foreach my $virtdev (sort keys %$virtdev_hash) {
6138             my $d = $virtdev_hash->{$virtdev};
6139             my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6140             my $storeid = $d->{storeid};
6141             my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6142
6143             my $map_opts = '';
6144             if (my $limit = $storage_limits{$storeid}) {
6145                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6146             }
6147
6148             # test if requested format is supported
6149             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6150             my $supported = grep { $_ eq $d->{format} } @$validFormats;
6151             $d->{format} = $defFormat if !$supported;
6152
6153             my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid,
6154                                                   $d->{format}, undef, $alloc_size);
6155             print STDERR "new volume ID is '$volid'\n";
6156             $d->{volid} = $volid;
6157             my $path = PVE::Storage::path($cfg, $volid);
6158
6159             PVE::Storage::activate_volumes($cfg,[$volid]);
6160
6161             my $write_zeros = 1;
6162             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6163                 $write_zeros = 0;
6164             }
6165
6166             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6167
6168             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6169             $map->{$virtdev} = $volid;
6170         }
6171
6172         $fh->seek(0, 0) || die "seek failed - $!\n";
6173
6174         my $outfd = new IO::File ($tmpfn, "w") ||
6175             die "unable to write config for VM $vmid\n";
6176
6177         my $cookie = { netcount => 0 };
6178         while (defined(my $line = <$fh>)) {
6179             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6180         }
6181
6182         $fh->close();
6183         $outfd->close();
6184     };
6185
6186     eval {
6187         # enable interrupts
6188         local $SIG{INT} =
6189             local $SIG{TERM} =
6190             local $SIG{QUIT} =
6191             local $SIG{HUP} =
6192             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6193         local $SIG{ALRM} = sub { die "got timeout\n"; };
6194
6195         $oldtimeout = alarm($timeout);
6196
6197         my $parser = sub {
6198             my $line = shift;
6199
6200             print "$line\n";
6201
6202             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6203                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6204                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6205             } elsif ($line =~ m/^CTIME: /) {
6206                 # we correctly received the vma config, so we can disable
6207                 # the timeout now for disk allocation (set to 10 minutes, so
6208                 # that we always timeout if something goes wrong)
6209                 alarm(600);
6210                 &$print_devmap();
6211                 print $fifofh "done\n";
6212                 my $tmp = $oldtimeout || 0;
6213                 $oldtimeout = undef;
6214                 alarm($tmp);
6215                 close($fifofh);
6216             }
6217         };
6218
6219         print "restore vma archive: $dbg_cmdstring\n";
6220         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6221     };
6222     my $err = $@;
6223
6224     alarm($oldtimeout) if $oldtimeout;
6225
6226     my $vollist = [];
6227     foreach my $devname (keys %$devinfo) {
6228         my $volid = $devinfo->{$devname}->{volid};
6229         push @$vollist, $volid if $volid;
6230     }
6231
6232     PVE::Storage::deactivate_volumes($cfg, $vollist);
6233
6234     unlink $mapfifo;
6235
6236     if ($err) {
6237         rmtree $tmpdir;
6238         unlink $tmpfn;
6239
6240         foreach my $devname (keys %$devinfo) {
6241             my $volid = $devinfo->{$devname}->{volid};
6242             next if !$volid;
6243             eval {
6244                 if ($volid =~ m|^/|) {
6245                     unlink $volid || die 'unlink failed\n';
6246                 } else {
6247                     PVE::Storage::vdisk_free($cfg, $volid);
6248                 }
6249                 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6250             };
6251             print STDERR "unable to cleanup '$volid' - $@" if $@;
6252         }
6253         die $err;
6254     }
6255
6256     rmtree $tmpdir;
6257
6258     rename($tmpfn, $conffile) ||
6259         die "unable to commit configuration file '$conffile'\n";
6260
6261     PVE::Cluster::cfs_update(); # make sure we read new file
6262
6263     eval { rescan($vmid, 1); };
6264     warn $@ if $@;
6265 }
6266
6267 sub restore_tar_archive {
6268     my ($archive, $vmid, $user, $opts) = @_;
6269
6270     if ($archive ne '-') {
6271         my $firstfile = tar_archive_read_firstfile($archive);
6272         die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6273             if $firstfile ne 'qemu-server.conf';
6274     }
6275
6276     my $storecfg = PVE::Storage::config();
6277
6278     # destroy existing data - keep empty config
6279     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6280     destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6281
6282     my $tocmd = "/usr/lib/qemu-server/qmextract";
6283
6284     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6285     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6286     $tocmd .= ' --prealloc' if $opts->{prealloc};
6287     $tocmd .= ' --info' if $opts->{info};
6288
6289     # tar option "xf" does not autodetect compression when read from STDIN,
6290     # so we pipe to zcat
6291     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6292         PVE::Tools::shellquote("--to-command=$tocmd");
6293
6294     my $tmpdir = "/var/tmp/vzdumptmp$$";
6295     mkpath $tmpdir;
6296
6297     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6298     local $ENV{VZDUMP_VMID} = $vmid;
6299     local $ENV{VZDUMP_USER} = $user;
6300
6301     my $conffile = PVE::QemuConfig->config_file($vmid);
6302     my $tmpfn = "$conffile.$$.tmp";
6303
6304     # disable interrupts (always do cleanups)
6305     local $SIG{INT} =
6306         local $SIG{TERM} =
6307         local $SIG{QUIT} =
6308         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6309
6310     eval {
6311         # enable interrupts
6312         local $SIG{INT} =
6313             local $SIG{TERM} =
6314             local $SIG{QUIT} =
6315             local $SIG{HUP} =
6316             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6317
6318         if ($archive eq '-') {
6319             print "extracting archive from STDIN\n";
6320             run_command($cmd, input => "<&STDIN");
6321         } else {
6322             print "extracting archive '$archive'\n";
6323             run_command($cmd);
6324         }
6325
6326         return if $opts->{info};
6327
6328         # read new mapping
6329         my $map = {};
6330         my $statfile = "$tmpdir/qmrestore.stat";
6331         if (my $fd = IO::File->new($statfile, "r")) {
6332             while (defined (my $line = <$fd>)) {
6333                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6334                     $map->{$1} = $2 if $1;
6335                 } else {
6336                     print STDERR "unable to parse line in statfile - $line\n";
6337                 }
6338             }
6339             $fd->close();
6340         }
6341
6342         my $confsrc = "$tmpdir/qemu-server.conf";
6343
6344         my $srcfd = new IO::File($confsrc, "r") ||
6345             die "unable to open file '$confsrc'\n";
6346
6347         my $outfd = new IO::File ($tmpfn, "w") ||
6348             die "unable to write config for VM $vmid\n";
6349
6350         my $cookie = { netcount => 0 };
6351         while (defined (my $line = <$srcfd>)) {
6352             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6353         }
6354
6355         $srcfd->close();
6356         $outfd->close();
6357     };
6358     my $err = $@;
6359
6360     if ($err) {
6361
6362         unlink $tmpfn;
6363
6364         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6365
6366         die $err;
6367     }
6368
6369     rmtree $tmpdir;
6370
6371     rename $tmpfn, $conffile ||
6372         die "unable to commit configuration file '$conffile'\n";
6373
6374     PVE::Cluster::cfs_update(); # make sure we read new file
6375
6376     eval { rescan($vmid, 1); };
6377     warn $@ if $@;
6378 };
6379
6380 sub foreach_storage_used_by_vm {
6381     my ($conf, $func) = @_;
6382
6383     my $sidhash = {};
6384
6385     foreach_drive($conf, sub {
6386         my ($ds, $drive) = @_;
6387         return if drive_is_cdrom($drive);
6388
6389         my $volid = $drive->{file};
6390
6391         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6392         $sidhash->{$sid} = $sid if $sid;
6393     });
6394
6395     foreach my $sid (sort keys %$sidhash) {
6396         &$func($sid);
6397     }
6398 }
6399
6400 sub do_snapshots_with_qemu {
6401     my ($storecfg, $volid) = @_;
6402
6403     my $storage_name = PVE::Storage::parse_volume_id($volid);
6404
6405     if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}}
6406         && !$storecfg->{ids}->{$storage_name}->{krbd}){
6407         return 1;
6408     }
6409
6410     if ($volid =~ m/\.(qcow2|qed)$/){
6411         return 1;
6412     }
6413
6414     return undef;
6415 }
6416
6417 sub qga_check_running {
6418     my ($vmid, $nowarn) = @_;
6419
6420     eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6421     if ($@) {
6422         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6423         return 0;
6424     }
6425     return 1;
6426 }
6427
6428 sub template_create {
6429     my ($vmid, $conf, $disk) = @_;
6430
6431     my $storecfg = PVE::Storage::config();
6432
6433     foreach_drive($conf, sub {
6434         my ($ds, $drive) = @_;
6435
6436         return if drive_is_cdrom($drive);
6437         return if $disk && $ds ne $disk;
6438
6439         my $volid = $drive->{file};
6440         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6441
6442         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6443         $drive->{file} = $voliddst;
6444         $conf->{$ds} = print_drive($vmid, $drive);
6445         PVE::QemuConfig->write_config($vmid, $conf);
6446     });
6447 }
6448
6449 sub qemu_img_convert {
6450     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6451
6452     my $storecfg = PVE::Storage::config();
6453     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6454     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6455
6456     if ($src_storeid && $dst_storeid) {
6457
6458         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6459
6460         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6461         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6462
6463         my $src_format = qemu_img_format($src_scfg, $src_volname);
6464         my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6465
6466         my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6467         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6468
6469         my $cmd = [];
6470         push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6471         push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6472         push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6473         push @$cmd, '-T', 'none' if $src_scfg->{type} eq 'zfspool';
6474         push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path;
6475         if ($is_zero_initialized) {
6476             push @$cmd, "zeroinit:$dst_path";
6477         } else {
6478             push @$cmd, $dst_path;
6479         }
6480
6481         my $parser = sub {
6482             my $line = shift;
6483             if($line =~ m/\((\S+)\/100\%\)/){
6484                 my $percent = $1;
6485                 my $transferred = int($size * $percent / 100);
6486                 my $remaining = $size - $transferred;
6487
6488                 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6489             }
6490
6491         };
6492
6493         eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6494         my $err = $@;
6495         die "copy failed: $err" if $err;
6496     }
6497 }
6498
6499 sub qemu_img_format {
6500     my ($scfg, $volname) = @_;
6501
6502     if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6503         return $1;
6504     } else {
6505         return "raw";
6506     }
6507 }
6508
6509 sub qemu_drive_mirror {
6510     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga) = @_;
6511
6512     $jobs = {} if !$jobs;
6513
6514     my $qemu_target;
6515     my $format;
6516     $jobs->{"drive-$drive"} = {};
6517
6518     if ($dst_volid =~ /^nbd:/) {
6519         $qemu_target = $dst_volid;
6520         $format = "nbd";
6521     } else {
6522         my $storecfg = PVE::Storage::config();
6523         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6524
6525         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6526
6527         $format = qemu_img_format($dst_scfg, $dst_volname);
6528
6529         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6530
6531         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6532     }
6533
6534     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6535     $opts->{format} = $format if $format;
6536
6537     print "drive mirror is starting for drive-$drive\n";
6538
6539     eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); }; #if a job already run for this device,it's throw an error
6540
6541     if (my $err = $@) {
6542         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6543         die "mirroring error: $err";
6544     }
6545
6546     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
6547 }
6548
6549 sub qemu_drive_mirror_monitor {
6550     my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
6551
6552     eval {
6553         my $err_complete = 0;
6554
6555         while (1) {
6556             die "storage migration timed out\n" if $err_complete > 300;
6557
6558             my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6559
6560             my $running_mirror_jobs = {};
6561             foreach my $stat (@$stats) {
6562                 next if $stat->{type} ne 'mirror';
6563                 $running_mirror_jobs->{$stat->{device}} = $stat;
6564             }
6565
6566             my $readycounter = 0;
6567
6568             foreach my $job (keys %$jobs) {
6569
6570                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6571                     print "$job : finished\n";
6572                     delete $jobs->{$job};
6573                     next;
6574                 }
6575
6576                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6577
6578                 my $busy = $running_mirror_jobs->{$job}->{busy};
6579                 my $ready = $running_mirror_jobs->{$job}->{ready};
6580                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6581                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6582                     my $remaining = $total - $transferred;
6583                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6584
6585                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6586                 }
6587
6588                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6589             }
6590
6591             last if scalar(keys %$jobs) == 0;
6592
6593             if ($readycounter == scalar(keys %$jobs)) {
6594                 print "all mirroring jobs are ready \n";
6595                 last if $skipcomplete; #do the complete later
6596
6597                 if ($vmiddst && $vmiddst != $vmid) {
6598                     my $agent_running = $qga && qga_check_running($vmid);
6599                     if ($agent_running) {
6600                         print "freeze filesystem\n";
6601                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6602                     } else {
6603                         print "suspend vm\n";
6604                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6605                     }
6606
6607                     # if we clone a disk for a new target vm, we don't switch the disk
6608                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6609
6610                     if ($agent_running) {
6611                         print "unfreeze filesystem\n";
6612                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6613                     } else {
6614                         print "resume vm\n";
6615                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6616                     }
6617
6618                     last;
6619                 } else {
6620
6621                     foreach my $job (keys %$jobs) {
6622                         # try to switch the disk if source and destination are on the same guest
6623                         print "$job: Completing block job...\n";
6624
6625                         eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
6626                         if ($@ =~ m/cannot be completed/) {
6627                             print "$job: Block job cannot be completed, try again.\n";
6628                             $err_complete++;
6629                         }else {
6630                             print "$job: Completed successfully.\n";
6631                             $jobs->{$job}->{complete} = 1;
6632                         }
6633                     }
6634                 }
6635             }
6636             sleep 1;
6637         }
6638     };
6639     my $err = $@;
6640
6641     if ($err) {
6642         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6643         die "mirroring error: $err";
6644     }
6645
6646 }
6647
6648 sub qemu_blockjobs_cancel {
6649     my ($vmid, $jobs) = @_;
6650
6651     foreach my $job (keys %$jobs) {
6652         print "$job: Cancelling block job\n";
6653         eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
6654         $jobs->{$job}->{cancel} = 1;
6655     }
6656
6657     while (1) {
6658         my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6659
6660         my $running_jobs = {};
6661         foreach my $stat (@$stats) {
6662             $running_jobs->{$stat->{device}} = $stat;
6663         }
6664
6665         foreach my $job (keys %$jobs) {
6666
6667             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6668                 print "$job: Done.\n";
6669                 delete $jobs->{$job};
6670             }
6671         }
6672
6673         last if scalar(keys %$jobs) == 0;
6674
6675         sleep 1;
6676     }
6677 }
6678
6679 sub clone_disk {
6680     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6681         $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga) = @_;
6682
6683     my $newvolid;
6684
6685     if (!$full) {
6686         print "create linked clone of drive $drivename ($drive->{file})\n";
6687         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6688         push @$newvollist, $newvolid;
6689     } else {
6690
6691         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6692         $storeid = $storage if $storage;
6693
6694         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6695         my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6696
6697         print "create full clone of drive $drivename ($drive->{file})\n";
6698         my $name = undef;
6699         if (drive_is_cloudinit($drive)) {
6700             $name = "vm-$newvmid-cloudinit";
6701             # cloudinit only supports raw and qcow2 atm:
6702             if ($dst_format eq 'qcow2') {
6703                 $name .= '.qcow2';
6704             } elsif ($dst_format ne 'raw') {
6705                 die "clone: unhandled format for cloudinit image\n";
6706             }
6707         }
6708         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
6709         push @$newvollist, $newvolid;
6710
6711         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6712
6713         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6714         if (!$running || $snapname) {
6715             qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6716         } else {
6717
6718             my $kvmver = get_running_qemu_version ($vmid);
6719             if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
6720                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6721                     if $drive->{iothread};
6722             }
6723
6724             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga);
6725         }
6726     }
6727
6728     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6729
6730     my $disk = $drive;
6731     $disk->{format} = undef;
6732     $disk->{file} = $newvolid;
6733     $disk->{size} = $size;
6734
6735     return $disk;
6736 }
6737
6738 # this only works if VM is running
6739 sub get_current_qemu_machine {
6740     my ($vmid) = @_;
6741
6742     my $cmd = { execute => 'query-machines', arguments => {} };
6743     my $res = vm_qmp_command($vmid, $cmd);
6744
6745     my ($current, $default);
6746     foreach my $e (@$res) {
6747         $default = $e->{name} if $e->{'is-default'};
6748         $current = $e->{name} if $e->{'is-current'};
6749     }
6750
6751     # fallback to the default machine if current is not supported by qemu
6752     return $current || $default || 'pc';
6753 }
6754
6755 sub get_running_qemu_version {
6756     my ($vmid) = @_;
6757     my $cmd = { execute => 'query-version', arguments => {} };
6758     my $res = vm_qmp_command($vmid, $cmd);
6759     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6760 }
6761
6762 sub qemu_machine_feature_enabled {
6763     my ($machine, $kvmver, $version_major, $version_minor) = @_;
6764
6765     my $current_major;
6766     my $current_minor;
6767
6768     if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
6769
6770         $current_major = $3;
6771         $current_minor = $4;
6772
6773     } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6774
6775         $current_major = $1;
6776         $current_minor = $2;
6777     }
6778
6779     return 1 if $current_major > $version_major ||
6780                 ($current_major == $version_major &&
6781                  $current_minor >= $version_minor);
6782 }
6783
6784 sub qemu_machine_pxe {
6785     my ($vmid, $conf, $machine) = @_;
6786
6787     $machine =  PVE::QemuServer::get_current_qemu_machine($vmid) if !$machine;
6788
6789     if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
6790         $machine .= '.pxe';
6791     }
6792
6793     return $machine;
6794 }
6795
6796 sub qemu_use_old_bios_files {
6797     my ($machine_type) = @_;
6798
6799     return if !$machine_type;
6800
6801     my $use_old_bios_files = undef;
6802
6803     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6804         $machine_type = $1;
6805         $use_old_bios_files = 1;
6806     } else {
6807         my $kvmver = kvm_user_version();
6808         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6809         # load new efi bios files on migration. So this hack is required to allow
6810         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6811         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6812         $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
6813     }
6814
6815     return ($use_old_bios_files, $machine_type);
6816 }
6817
6818 sub create_efidisk($$$$$) {
6819     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
6820
6821     my (undef, $ovmf_vars) = get_ovmf_files($arch);
6822     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
6823
6824     my $vars_size = PVE::Tools::convert_size(-s $ovmf_vars, 'b' => 'kb');
6825     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
6826     PVE::Storage::activate_volumes($storecfg, [$volid]);
6827
6828     my $path = PVE::Storage::path($storecfg, $volid);
6829     eval {
6830         run_command(['/usr/bin/qemu-img', 'convert', '-n', '-f', 'raw', '-O', $fmt, $ovmf_vars, $path]);
6831     };
6832     die "Copying EFI vars image failed: $@" if $@;
6833
6834     return ($volid, $vars_size);
6835 }
6836
6837 sub lspci {
6838
6839     my $devices = {};
6840
6841     dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6842             my (undef, $id, $function) = @_;
6843             my $res = { id => $id, function => $function};
6844             push @{$devices->{$id}}, $res;
6845     });
6846
6847     # Entries should be sorted by functions.
6848     foreach my $id (keys %$devices) {
6849         my $dev = $devices->{$id};
6850         $devices->{$id} = [ sort { $a->{function} <=> $b->{function} } @$dev ];
6851     }
6852
6853     return $devices;
6854 }
6855
6856 sub vm_iothreads_list {
6857     my ($vmid) = @_;
6858
6859     my $res = vm_mon_cmd($vmid, 'query-iothreads');
6860
6861     my $iothreads = {};
6862     foreach my $iothread (@$res) {
6863         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6864     }
6865
6866     return $iothreads;
6867 }
6868
6869 sub scsihw_infos {
6870     my ($conf, $drive) = @_;
6871
6872     my $maxdev = 0;
6873
6874     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
6875         $maxdev = 7;
6876     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
6877         $maxdev = 1;
6878     } else {
6879         $maxdev = 256;
6880     }
6881
6882     my $controller = int($drive->{index} / $maxdev);
6883     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
6884
6885     return ($maxdev, $controller, $controller_prefix);
6886 }
6887
6888 sub add_hyperv_enlightenments {
6889     my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough) = @_;
6890
6891     return if $winversion < 6;
6892     return if $bios && $bios eq 'ovmf' && $winversion < 8;
6893
6894     push @$cpuFlags , 'hv_vendor_id=proxmox' if $gpu_passthrough;
6895
6896     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
6897         push @$cpuFlags , 'hv_spinlocks=0x1fff';
6898         push @$cpuFlags , 'hv_vapic';
6899         push @$cpuFlags , 'hv_time';
6900     } else {
6901         push @$cpuFlags , 'hv_spinlocks=0xffff';
6902     }
6903
6904     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
6905         push @$cpuFlags , 'hv_reset';
6906         push @$cpuFlags , 'hv_vpindex';
6907         push @$cpuFlags , 'hv_runtime';
6908     }
6909
6910     if ($winversion >= 7) {
6911         push @$cpuFlags , 'hv_relaxed';
6912
6913         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
6914             push @$cpuFlags , 'hv_synic';
6915             push @$cpuFlags , 'hv_stimer';
6916         }
6917     }
6918 }
6919
6920 sub windows_version {
6921     my ($ostype) = @_;
6922
6923     return 0 if !$ostype;
6924
6925     my $winversion = 0;
6926
6927     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
6928         $winversion = 5;
6929     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
6930         $winversion = 6;
6931     } elsif ($ostype =~ m/^win(\d+)$/) {
6932         $winversion = $1;
6933     }
6934
6935     return $winversion;
6936 }
6937
6938 sub resolve_dst_disk_format {
6939         my ($storecfg, $storeid, $src_volname, $format) = @_;
6940         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6941
6942         if (!$format) {
6943             # if no target format is specified, use the source disk format as hint
6944             if ($src_volname) {
6945                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6946                 $format = qemu_img_format($scfg, $src_volname);
6947             } else {
6948                 return $defFormat;
6949             }
6950         }
6951
6952         # test if requested format is supported - else use default
6953         my $supported = grep { $_ eq $format } @$validFormats;
6954         $format = $defFormat if !$supported;
6955         return $format;
6956 }
6957
6958 sub resolve_first_disk {
6959     my $conf = shift;
6960     my @disks = PVE::QemuServer::valid_drive_names();
6961     my $firstdisk;
6962     foreach my $ds (reverse @disks) {
6963         next if !$conf->{$ds};
6964         my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
6965         next if PVE::QemuServer::drive_is_cdrom($disk);
6966         $firstdisk = $ds;
6967     }
6968     return $firstdisk;
6969 }
6970
6971 sub generate_uuid {
6972     my ($uuid, $uuid_str);
6973     UUID::generate($uuid);
6974     UUID::unparse($uuid, $uuid_str);
6975     return $uuid_str;
6976 }
6977
6978 sub generate_smbios1_uuid {
6979     return "uuid=".generate_uuid();
6980 }
6981
6982 sub nbd_stop {
6983     my ($vmid) = @_;
6984
6985     vm_mon_cmd($vmid, 'nbd-server-stop');
6986 }
6987
6988 # bash completion helper
6989
6990 sub complete_backup_archives {
6991     my ($cmdname, $pname, $cvalue) = @_;
6992
6993     my $cfg = PVE::Storage::config();
6994
6995     my $storeid;
6996
6997     if ($cvalue =~ m/^([^:]+):/) {
6998         $storeid = $1;
6999     }
7000
7001     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7002
7003     my $res = [];
7004     foreach my $id (keys %$data) {
7005         foreach my $item (@{$data->{$id}}) {
7006             next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7007             push @$res, $item->{volid} if defined($item->{volid});
7008         }
7009     }
7010
7011     return $res;
7012 }
7013
7014 my $complete_vmid_full = sub {
7015     my ($running) = @_;
7016
7017     my $idlist = vmstatus();
7018
7019     my $res = [];
7020
7021     foreach my $id (keys %$idlist) {
7022         my $d = $idlist->{$id};
7023         if (defined($running)) {
7024             next if $d->{template};
7025             next if $running && $d->{status} ne 'running';
7026             next if !$running && $d->{status} eq 'running';
7027         }
7028         push @$res, $id;
7029
7030     }
7031     return $res;
7032 };
7033
7034 sub complete_vmid {
7035     return &$complete_vmid_full();
7036 }
7037
7038 sub complete_vmid_stopped {
7039     return &$complete_vmid_full(0);
7040 }
7041
7042 sub complete_vmid_running {
7043     return &$complete_vmid_full(1);
7044 }
7045
7046 sub complete_storage {
7047
7048     my $cfg = PVE::Storage::config();
7049     my $ids = $cfg->{ids};
7050
7051     my $res = [];
7052     foreach my $sid (keys %$ids) {
7053         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7054         next if !$ids->{$sid}->{content}->{images};
7055         push @$res, $sid;
7056     }
7057
7058     return $res;
7059 }
7060
7061 1;