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 $vhostparam = '';
1953     if (is_native($arch)) {
1954         $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1955     }
1956
1957     my $vmname = $conf->{name} || "vm$vmid";
1958
1959     my $netdev = "";
1960     my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
1961
1962     if ($net->{bridge}) {
1963         $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1964     } else {
1965         $netdev = "type=user,id=$netid,hostname=$vmname";
1966     }
1967
1968     $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1969
1970     return $netdev;
1971 }
1972
1973
1974 sub print_cpu_device {
1975     my ($conf, $id) = @_;
1976
1977     my $kvm = $conf->{kvm} // 1;
1978     my $cpu = $kvm ? "kvm64" : "qemu64";
1979     if (my $cputype = $conf->{cpu}) {
1980         my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
1981             or die "Cannot parse cpu description: $cputype\n";
1982         $cpu = $cpuconf->{cputype};
1983     }
1984
1985     my $cores = $conf->{cores} || 1;
1986
1987     my $current_core = ($id - 1) % $cores;
1988     my $current_socket = int(($id - 1 - $current_core)/$cores);
1989
1990     return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
1991 }
1992
1993 my $vga_map = {
1994     'cirrus' => 'cirrus-vga',
1995     'std' => 'VGA',
1996     'vmware' => 'vmware-svga',
1997     'virtio' => 'virtio-vga',
1998 };
1999
2000 sub print_vga_device {
2001     my ($conf, $vga, $id, $qxlnum, $bridges) = @_;
2002
2003     my $type = $vga_map->{$vga->{type}};
2004     my $vgamem_mb = $vga->{memory};
2005     if ($qxlnum) {
2006         $type = $id ? 'qxl' : 'qxl-vga';
2007     }
2008     die "no devicetype for $vga->{type}\n" if !$type;
2009
2010     my $memory = "";
2011     if ($vgamem_mb) {
2012         if ($vga->{type} eq 'virtio') {
2013             my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2014             $memory = ",max_hostmem=$bytes";
2015         } elsif ($qxlnum) {
2016             # from https://www.spice-space.org/multiple-monitors.html
2017             $memory = ",vgamem_mb=$vga->{memory}";
2018             my $ram = $vgamem_mb * 4;
2019             my $vram = $vgamem_mb * 2;
2020             $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2021         } else {
2022             $memory = ",vgamem_mb=$vga->{memory}";
2023         }
2024     } elsif ($qxlnum && $id) {
2025         $memory = ",ram_size=67108864,vram_size=33554432";
2026     }
2027
2028     my $q35 = machine_type_is_q35($conf);
2029     my $vgaid = "vga" . ($id // '');
2030     my $pciaddr;
2031
2032     if ($q35 && $vgaid eq 'vga') {
2033         # the first display uses pcie.0 bus on q35 machines
2034         $pciaddr = print_pcie_addr($vgaid, $bridges);
2035     } else {
2036         $pciaddr = print_pci_addr($vgaid, $bridges);
2037     }
2038
2039     return "$type,id=${vgaid}${memory}${pciaddr}";
2040 }
2041
2042 sub drive_is_cloudinit {
2043     my ($drive) = @_;
2044     return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2045 }
2046
2047 sub drive_is_cdrom {
2048     my ($drive, $exclude_cloudinit) = @_;
2049
2050     return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2051
2052     return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2053
2054 }
2055
2056 sub parse_number_sets {
2057     my ($set) = @_;
2058     my $res = [];
2059     foreach my $part (split(/;/, $set)) {
2060         if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2061             die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2062             push @$res, [ $1, $2 ];
2063         } else {
2064             die "invalid range: $part\n";
2065         }
2066     }
2067     return $res;
2068 }
2069
2070 sub parse_numa {
2071     my ($data) = @_;
2072
2073     my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2074     $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2075     $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2076     return $res;
2077 }
2078
2079 sub parse_hostpci {
2080     my ($value) = @_;
2081
2082     return undef if !$value;
2083
2084     my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2085
2086     my @idlist = split(/;/, $res->{host});
2087     delete $res->{host};
2088     foreach my $id (@idlist) {
2089         if ($id =~ /^$PCIRE$/) {
2090             if (defined($2)) {
2091                 push @{$res->{pciid}}, { id => $1, function => $2 };
2092             } else {
2093                 my $pcidevices = lspci($1);
2094                 $res->{pciid} = $pcidevices->{$1};
2095             }
2096         } else {
2097             # should have been caught by parse_property_string already
2098             die "failed to parse PCI id: $id\n";
2099         }
2100     }
2101     return $res;
2102 }
2103
2104 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2105 sub parse_net {
2106     my ($data) = @_;
2107
2108     my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2109     if ($@) {
2110         warn $@;
2111         return undef;
2112     }
2113     if (!defined($res->{macaddr})) {
2114         my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2115         $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2116     }
2117     return $res;
2118 }
2119
2120 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2121 sub parse_ipconfig {
2122     my ($data) = @_;
2123
2124     my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2125     if ($@) {
2126         warn $@;
2127         return undef;
2128     }
2129
2130     if ($res->{gw} && !$res->{ip}) {
2131         warn 'gateway specified without specifying an IP address';
2132         return undef;
2133     }
2134     if ($res->{gw6} && !$res->{ip6}) {
2135         warn 'IPv6 gateway specified without specifying an IPv6 address';
2136         return undef;
2137     }
2138     if ($res->{gw} && $res->{ip} eq 'dhcp') {
2139         warn 'gateway specified together with DHCP';
2140         return undef;
2141     }
2142     if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2143         # gw6 + auto/dhcp
2144         warn "IPv6 gateway specified together with $res->{ip6} address";
2145         return undef;
2146     }
2147
2148     if (!$res->{ip} && !$res->{ip6}) {
2149         return { ip => 'dhcp', ip6 => 'dhcp' };
2150     }
2151
2152     return $res;
2153 }
2154
2155 sub print_net {
2156     my $net = shift;
2157
2158     return PVE::JSONSchema::print_property_string($net, $net_fmt);
2159 }
2160
2161 sub add_random_macs {
2162     my ($settings) = @_;
2163
2164     foreach my $opt (keys %$settings) {
2165         next if $opt !~ m/^net(\d+)$/;
2166         my $net = parse_net($settings->{$opt});
2167         next if !$net;
2168         $settings->{$opt} = print_net($net);
2169     }
2170 }
2171
2172 sub vm_is_volid_owner {
2173     my ($storecfg, $vmid, $volid) = @_;
2174
2175     if ($volid !~  m|^/|) {
2176         my ($path, $owner);
2177         eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2178         if ($owner && ($owner == $vmid)) {
2179             return 1;
2180         }
2181     }
2182
2183     return undef;
2184 }
2185
2186 sub split_flagged_list {
2187     my $text = shift || '';
2188     $text =~ s/[,;]/ /g;
2189     $text =~ s/^\s+//;
2190     return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
2191 }
2192
2193 sub join_flagged_list {
2194     my ($how, $lst) = @_;
2195     join $how, map { $lst->{$_} . $_ } keys %$lst;
2196 }
2197
2198 sub vmconfig_delete_pending_option {
2199     my ($conf, $key, $force) = @_;
2200
2201     delete $conf->{pending}->{$key};
2202     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2203     $pending_delete_hash->{$key} = $force ? '!' : '';
2204     $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2205 }
2206
2207 sub vmconfig_undelete_pending_option {
2208     my ($conf, $key) = @_;
2209
2210     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2211     delete $pending_delete_hash->{$key};
2212
2213     if (%$pending_delete_hash) {
2214         $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2215     } else {
2216         delete $conf->{pending}->{delete};
2217     }
2218 }
2219
2220 sub vmconfig_register_unused_drive {
2221     my ($storecfg, $vmid, $conf, $drive) = @_;
2222
2223     if (drive_is_cloudinit($drive)) {
2224         eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2225         warn $@ if $@;
2226     } elsif (!drive_is_cdrom($drive)) {
2227         my $volid = $drive->{file};
2228         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2229             PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2230         }
2231     }
2232 }
2233
2234 sub vmconfig_cleanup_pending {
2235     my ($conf) = @_;
2236
2237     # remove pending changes when nothing changed
2238     my $changes;
2239     foreach my $opt (keys %{$conf->{pending}}) {
2240         if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq  $conf->{$opt})) {
2241             $changes = 1;
2242             delete $conf->{pending}->{$opt};
2243         }
2244     }
2245
2246     my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
2247     my $pending_delete_hash = {};
2248     while (my ($opt, $force) = each %$current_delete_hash) {
2249         if (defined($conf->{$opt})) {
2250             $pending_delete_hash->{$opt} = $force;
2251         } else {
2252             $changes = 1;
2253         }
2254     }
2255
2256     if (%$pending_delete_hash) {
2257         $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2258     } else {
2259         delete $conf->{pending}->{delete};
2260     }
2261
2262     return $changes;
2263 }
2264
2265 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
2266 my $smbios1_fmt = {
2267     uuid => {
2268         type => 'string',
2269         pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2270         format_description => 'UUID',
2271         description => "Set SMBIOS1 UUID.",
2272         optional => 1,
2273     },
2274     version => {
2275         type => 'string',
2276         pattern => '\S+',
2277         format_description => 'string',
2278         description => "Set SMBIOS1 version.",
2279         optional => 1,
2280     },
2281     serial => {
2282         type => 'string',
2283         pattern => '\S+',
2284         format_description => 'string',
2285         description => "Set SMBIOS1 serial number.",
2286         optional => 1,
2287     },
2288     manufacturer => {
2289         type => 'string',
2290         pattern => '\S+',
2291         format_description => 'string',
2292         description => "Set SMBIOS1 manufacturer.",
2293         optional => 1,
2294     },
2295     product => {
2296         type => 'string',
2297         pattern => '\S+',
2298         format_description => 'string',
2299         description => "Set SMBIOS1 product ID.",
2300         optional => 1,
2301     },
2302     sku => {
2303         type => 'string',
2304         pattern => '\S+',
2305         format_description => 'string',
2306         description => "Set SMBIOS1 SKU string.",
2307         optional => 1,
2308     },
2309     family => {
2310         type => 'string',
2311         pattern => '\S+',
2312         format_description => 'string',
2313         description => "Set SMBIOS1 family string.",
2314         optional => 1,
2315     },
2316 };
2317
2318 sub parse_smbios1 {
2319     my ($data) = @_;
2320
2321     my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2322     warn $@ if $@;
2323     return $res;
2324 }
2325
2326 sub print_smbios1 {
2327     my ($smbios1) = @_;
2328     return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2329 }
2330
2331 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2332
2333 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2334 sub verify_bootdisk {
2335     my ($value, $noerr) = @_;
2336
2337     return $value if is_valid_drivename($value);
2338
2339     return undef if $noerr;
2340
2341     die "invalid boot disk '$value'\n";
2342 }
2343
2344 sub parse_watchdog {
2345     my ($value) = @_;
2346
2347     return undef if !$value;
2348
2349     my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2350     warn $@ if $@;
2351     return $res;
2352 }
2353
2354 sub parse_guest_agent {
2355     my ($value) = @_;
2356
2357     return {} if !defined($value->{agent});
2358
2359     my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2360     warn $@ if $@;
2361
2362     # if the agent is disabled ignore the other potentially set properties
2363     return {} if !$res->{enabled};
2364     return $res;
2365 }
2366
2367 sub parse_vga {
2368     my ($value) = @_;
2369
2370     return {} if !$value;
2371     my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2372     warn $@ if $@;
2373     return $res;
2374 }
2375
2376 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2377 sub verify_usb_device {
2378     my ($value, $noerr) = @_;
2379
2380     return $value if parse_usb_device($value);
2381
2382     return undef if $noerr;
2383
2384     die "unable to parse usb device\n";
2385 }
2386
2387 # add JSON properties for create and set function
2388 sub json_config_properties {
2389     my $prop = shift;
2390
2391     foreach my $opt (keys %$confdesc) {
2392         next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2393         $prop->{$opt} = $confdesc->{$opt};
2394     }
2395
2396     return $prop;
2397 }
2398
2399 # return copy of $confdesc_cloudinit to generate documentation
2400 sub cloudinit_config_properties {
2401
2402     return dclone($confdesc_cloudinit);
2403 }
2404
2405 sub check_type {
2406     my ($key, $value) = @_;
2407
2408     die "unknown setting '$key'\n" if !$confdesc->{$key};
2409
2410     my $type = $confdesc->{$key}->{type};
2411
2412     if (!defined($value)) {
2413         die "got undefined value\n";
2414     }
2415
2416     if ($value =~ m/[\n\r]/) {
2417         die "property contains a line feed\n";
2418     }
2419
2420     if ($type eq 'boolean') {
2421         return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2422         return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2423         die "type check ('boolean') failed - got '$value'\n";
2424     } elsif ($type eq 'integer') {
2425         return int($1) if $value =~ m/^(\d+)$/;
2426         die "type check ('integer') failed - got '$value'\n";
2427     } elsif ($type eq 'number') {
2428         return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2429         die "type check ('number') failed - got '$value'\n";
2430     } elsif ($type eq 'string') {
2431         if (my $fmt = $confdesc->{$key}->{format}) {
2432             PVE::JSONSchema::check_format($fmt, $value);
2433             return $value;
2434         }
2435         $value =~ s/^\"(.*)\"$/$1/;
2436         return $value;
2437     } else {
2438         die "internal error"
2439     }
2440 }
2441
2442 sub check_iommu_support{
2443     #fixme : need to check IOMMU support
2444     #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
2445
2446     my $iommu=1;
2447     return $iommu;
2448
2449 }
2450
2451 sub touch_config {
2452     my ($vmid) = @_;
2453
2454     my $conf = PVE::QemuConfig->config_file($vmid);
2455     utime undef, undef, $conf;
2456 }
2457
2458 sub destroy_vm {
2459     my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2460
2461     my $conffile = PVE::QemuConfig->config_file($vmid);
2462
2463     my $conf = PVE::QemuConfig->load_config($vmid);
2464
2465     PVE::QemuConfig->check_lock($conf) if !$skiplock;
2466
2467     if ($conf->{template}) {
2468         # check if any base image is still used by a linked clone
2469         foreach_drive($conf, sub {
2470                 my ($ds, $drive) = @_;
2471
2472                 return if drive_is_cdrom($drive);
2473
2474                 my $volid = $drive->{file};
2475
2476                 return if !$volid || $volid =~ m|^/|;
2477
2478                 die "base volume '$volid' is still in use by linked cloned\n"
2479                     if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2480
2481         });
2482     }
2483
2484     # only remove disks owned by this VM
2485     foreach_drive($conf, sub {
2486         my ($ds, $drive) = @_;
2487
2488         return if drive_is_cdrom($drive, 1);
2489
2490         my $volid = $drive->{file};
2491
2492         return if !$volid || $volid =~ m|^/|;
2493
2494         my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2495         return if !$path || !$owner || ($owner != $vmid);
2496
2497         eval {
2498             PVE::Storage::vdisk_free($storecfg, $volid);
2499         };
2500         warn "Could not remove disk '$volid', check manually: $@" if $@;
2501
2502     });
2503
2504     if ($keep_empty_config) {
2505         PVE::Tools::file_set_contents($conffile, "memory: 128\n");
2506     } else {
2507         unlink $conffile;
2508     }
2509
2510     # also remove unused disk
2511     eval {
2512         my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2513
2514         eval {
2515             PVE::Storage::foreach_volid($dl, sub {
2516                 my ($volid, $sid, $volname, $d) = @_;
2517                 PVE::Storage::vdisk_free($storecfg, $volid);
2518             });
2519         };
2520         warn $@ if $@;
2521
2522     };
2523     warn $@ if $@;
2524 }
2525
2526 sub parse_vm_config {
2527     my ($filename, $raw) = @_;
2528
2529     return undef if !defined($raw);
2530
2531     my $res = {
2532         digest => Digest::SHA::sha1_hex($raw),
2533         snapshots => {},
2534         pending => {},
2535     };
2536
2537     $filename =~ m|/qemu-server/(\d+)\.conf$|
2538         || die "got strange filename '$filename'";
2539
2540     my $vmid = $1;
2541
2542     my $conf = $res;
2543     my $descr;
2544     my $section = '';
2545
2546     my @lines = split(/\n/, $raw);
2547     foreach my $line (@lines) {
2548         next if $line =~ m/^\s*$/;
2549
2550         if ($line =~ m/^\[PENDING\]\s*$/i) {
2551             $section = 'pending';
2552             if (defined($descr)) {
2553                 $descr =~ s/\s+$//;
2554                 $conf->{description} = $descr;
2555             }
2556             $descr = undef;
2557             $conf = $res->{$section} = {};
2558             next;
2559
2560         } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2561             $section = $1;
2562             if (defined($descr)) {
2563                 $descr =~ s/\s+$//;
2564                 $conf->{description} = $descr;
2565             }
2566             $descr = undef;
2567             $conf = $res->{snapshots}->{$section} = {};
2568             next;
2569         }
2570
2571         if ($line =~ m/^\#(.*)\s*$/) {
2572             $descr = '' if !defined($descr);
2573             $descr .= PVE::Tools::decode_text($1) . "\n";
2574             next;
2575         }
2576
2577         if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2578             $descr = '' if !defined($descr);
2579             $descr .= PVE::Tools::decode_text($2);
2580         } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2581             $conf->{snapstate} = $1;
2582         } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2583             my $key = $1;
2584             my $value = $2;
2585             $conf->{$key} = $value;
2586         } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2587             my $value = $1;
2588             if ($section eq 'pending') {
2589                 $conf->{delete} = $value; # we parse this later
2590             } else {
2591                 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2592             }
2593         } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2594             my $key = $1;
2595             my $value = $2;
2596             eval { $value = check_type($key, $value); };
2597             if ($@) {
2598                 warn "vm $vmid - unable to parse value of '$key' - $@";
2599             } else {
2600                 $key = 'ide2' if $key eq 'cdrom';
2601                 my $fmt = $confdesc->{$key}->{format};
2602                 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2603                     my $v = parse_drive($key, $value);
2604                     if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2605                         $v->{file} = $volid;
2606                         $value = print_drive($vmid, $v);
2607                     } else {
2608                         warn "vm $vmid - unable to parse value of '$key'\n";
2609                         next;
2610                     }
2611                 }
2612
2613                 $conf->{$key} = $value;
2614             }
2615         }
2616     }
2617
2618     if (defined($descr)) {
2619         $descr =~ s/\s+$//;
2620         $conf->{description} = $descr;
2621     }
2622     delete $res->{snapstate}; # just to be sure
2623
2624     return $res;
2625 }
2626
2627 sub write_vm_config {
2628     my ($filename, $conf) = @_;
2629
2630     delete $conf->{snapstate}; # just to be sure
2631
2632     if ($conf->{cdrom}) {
2633         die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2634         $conf->{ide2} = $conf->{cdrom};
2635         delete $conf->{cdrom};
2636     }
2637
2638     # we do not use 'smp' any longer
2639     if ($conf->{sockets}) {
2640         delete $conf->{smp};
2641     } elsif ($conf->{smp}) {
2642         $conf->{sockets} = $conf->{smp};
2643         delete $conf->{cores};
2644         delete $conf->{smp};
2645     }
2646
2647     my $used_volids = {};
2648
2649     my $cleanup_config = sub {
2650         my ($cref, $pending, $snapname) = @_;
2651
2652         foreach my $key (keys %$cref) {
2653             next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2654                 $key eq 'snapstate' || $key eq 'pending';
2655             my $value = $cref->{$key};
2656             if ($key eq 'delete') {
2657                 die "propertry 'delete' is only allowed in [PENDING]\n"
2658                     if !$pending;
2659                 # fixme: check syntax?
2660                 next;
2661             }
2662             eval { $value = check_type($key, $value); };
2663             die "unable to parse value of '$key' - $@" if $@;
2664
2665             $cref->{$key} = $value;
2666
2667             if (!$snapname && is_valid_drivename($key)) {
2668                 my $drive = parse_drive($key, $value);
2669                 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2670             }
2671         }
2672     };
2673
2674     &$cleanup_config($conf);
2675
2676     &$cleanup_config($conf->{pending}, 1);
2677
2678     foreach my $snapname (keys %{$conf->{snapshots}}) {
2679         die "internal error" if $snapname eq 'pending';
2680         &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2681     }
2682
2683     # remove 'unusedX' settings if we re-add a volume
2684     foreach my $key (keys %$conf) {
2685         my $value = $conf->{$key};
2686         if ($key =~ m/^unused/ && $used_volids->{$value}) {
2687             delete $conf->{$key};
2688         }
2689     }
2690
2691     my $generate_raw_config = sub {
2692         my ($conf, $pending) = @_;
2693
2694         my $raw = '';
2695
2696         # add description as comment to top of file
2697         if (defined(my $descr = $conf->{description})) {
2698             if ($descr) {
2699                 foreach my $cl (split(/\n/, $descr)) {
2700                     $raw .= '#' .  PVE::Tools::encode_text($cl) . "\n";
2701                 }
2702             } else {
2703                 $raw .= "#\n" if $pending;
2704             }
2705         }
2706
2707         foreach my $key (sort keys %$conf) {
2708             next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2709             $raw .= "$key: $conf->{$key}\n";
2710         }
2711         return $raw;
2712     };
2713
2714     my $raw = &$generate_raw_config($conf);
2715
2716     if (scalar(keys %{$conf->{pending}})){
2717         $raw .= "\n[PENDING]\n";
2718         $raw .= &$generate_raw_config($conf->{pending}, 1);
2719     }
2720
2721     foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2722         $raw .= "\n[$snapname]\n";
2723         $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2724     }
2725
2726     return $raw;
2727 }
2728
2729 sub load_defaults {
2730
2731     my $res = {};
2732
2733     # we use static defaults from our JSON schema configuration
2734     foreach my $key (keys %$confdesc) {
2735         if (defined(my $default = $confdesc->{$key}->{default})) {
2736             $res->{$key} = $default;
2737         }
2738     }
2739
2740     return $res;
2741 }
2742
2743 sub config_list {
2744     my $vmlist = PVE::Cluster::get_vmlist();
2745     my $res = {};
2746     return $res if !$vmlist || !$vmlist->{ids};
2747     my $ids = $vmlist->{ids};
2748
2749     foreach my $vmid (keys %$ids) {
2750         my $d = $ids->{$vmid};
2751         next if !$d->{node} || $d->{node} ne $nodename;
2752         next if !$d->{type} || $d->{type} ne 'qemu';
2753         $res->{$vmid}->{exists} = 1;
2754     }
2755     return $res;
2756 }
2757
2758 # test if VM uses local resources (to prevent migration)
2759 sub check_local_resources {
2760     my ($conf, $noerr) = @_;
2761
2762     my $loc_res = 0;
2763
2764     $loc_res = 1 if $conf->{hostusb}; # old syntax
2765     $loc_res = 1 if $conf->{hostpci}; # old syntax
2766
2767     foreach my $k (keys %$conf) {
2768         next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2769         # sockets are safe: they will recreated be on the target side post-migrate
2770         next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2771         $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2772     }
2773
2774     die "VM uses local resources\n" if $loc_res && !$noerr;
2775
2776     return $loc_res;
2777 }
2778
2779 # check if used storages are available on all nodes (use by migrate)
2780 sub check_storage_availability {
2781     my ($storecfg, $conf, $node) = @_;
2782
2783     foreach_drive($conf, sub {
2784         my ($ds, $drive) = @_;
2785
2786         my $volid = $drive->{file};
2787         return if !$volid;
2788
2789         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2790         return if !$sid;
2791
2792         # check if storage is available on both nodes
2793         my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2794         PVE::Storage::storage_check_node($storecfg, $sid, $node);
2795    });
2796 }
2797
2798 # list nodes where all VM images are available (used by has_feature API)
2799 sub shared_nodes {
2800     my ($conf, $storecfg) = @_;
2801
2802     my $nodelist = PVE::Cluster::get_nodelist();
2803     my $nodehash = { map { $_ => 1 } @$nodelist };
2804     my $nodename = PVE::INotify::nodename();
2805
2806     foreach_drive($conf, sub {
2807         my ($ds, $drive) = @_;
2808
2809         my $volid = $drive->{file};
2810         return if !$volid;
2811
2812         my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2813         if ($storeid) {
2814             my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2815             if ($scfg->{disable}) {
2816                 $nodehash = {};
2817             } elsif (my $avail = $scfg->{nodes}) {
2818                 foreach my $node (keys %$nodehash) {
2819                     delete $nodehash->{$node} if !$avail->{$node};
2820                 }
2821             } elsif (!$scfg->{shared}) {
2822                 foreach my $node (keys %$nodehash) {
2823                     delete $nodehash->{$node} if $node ne $nodename
2824                 }
2825             }
2826         }
2827     });
2828
2829     return $nodehash
2830 }
2831
2832 sub check_cmdline {
2833     my ($pidfile, $pid) = @_;
2834
2835     my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2836     if (defined($fh)) {
2837         my $line = <$fh>;
2838         $fh->close;
2839         return undef if !$line;
2840         my @param = split(/\0/, $line);
2841
2842         my $cmd = $param[0];
2843         return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
2844
2845         for (my $i = 0; $i < scalar (@param); $i++) {
2846             my $p = $param[$i];
2847             next if !$p;
2848             if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2849                 my $p = $param[$i+1];
2850                 return 1 if $p && ($p eq $pidfile);
2851                 return undef;
2852             }
2853         }
2854     }
2855     return undef;
2856 }
2857
2858 sub check_running {
2859     my ($vmid, $nocheck, $node) = @_;
2860
2861     my $filename = PVE::QemuConfig->config_file($vmid, $node);
2862
2863     die "unable to find configuration file for VM $vmid - no such machine\n"
2864         if !$nocheck && ! -f $filename;
2865
2866     my $pidfile = pidfile_name($vmid);
2867
2868     if (my $fd = IO::File->new("<$pidfile")) {
2869         my $st = stat($fd);
2870         my $line = <$fd>;
2871         close($fd);
2872
2873         my $mtime = $st->mtime;
2874         if ($mtime > time()) {
2875             warn "file '$filename' modified in future\n";
2876         }
2877
2878         if ($line =~ m/^(\d+)$/) {
2879             my $pid = $1;
2880             if (check_cmdline($pidfile, $pid)) {
2881                 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2882                     return $pid;
2883                 }
2884             }
2885         }
2886     }
2887
2888     return undef;
2889 }
2890
2891 sub vzlist {
2892
2893     my $vzlist = config_list();
2894
2895     my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2896
2897     while (defined(my $de = $fd->read)) {
2898         next if $de !~ m/^(\d+)\.pid$/;
2899         my $vmid = $1;
2900         next if !defined($vzlist->{$vmid});
2901         if (my $pid = check_running($vmid)) {
2902             $vzlist->{$vmid}->{pid} = $pid;
2903         }
2904     }
2905
2906     return $vzlist;
2907 }
2908
2909 sub disksize {
2910     my ($storecfg, $conf) = @_;
2911
2912     my $bootdisk = $conf->{bootdisk};
2913     return undef if !$bootdisk;
2914     return undef if !is_valid_drivename($bootdisk);
2915
2916     return undef if !$conf->{$bootdisk};
2917
2918     my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2919     return undef if !defined($drive);
2920
2921     return undef if drive_is_cdrom($drive);
2922
2923     my $volid = $drive->{file};
2924     return undef if !$volid;
2925
2926     return $drive->{size};
2927 }
2928
2929 our $vmstatus_return_properties = {
2930     vmid => get_standard_option('pve-vmid'),
2931     status => {
2932         description => "Qemu process status.",
2933         type => 'string',
2934         enum => ['stopped', 'running'],
2935     },
2936     maxmem => {
2937         description => "Maximum memory in bytes.",
2938         type => 'integer',
2939         optional => 1,
2940         renderer => 'bytes',
2941     },
2942     maxdisk => {
2943         description => "Root disk size in bytes.",
2944         type => 'integer',
2945         optional => 1,
2946         renderer => 'bytes',
2947     },
2948     name => {
2949         description => "VM name.",
2950         type => 'string',
2951         optional => 1,
2952     },
2953     qmpstatus => {
2954         description => "Qemu QMP agent status.",
2955         type => 'string',
2956         optional => 1,
2957     },
2958     pid => {
2959         description => "PID of running qemu process.",
2960         type => 'integer',
2961         optional => 1,
2962     },
2963     uptime => {
2964         description => "Uptime.",
2965         type => 'integer',
2966         optional => 1,
2967         renderer => 'duration',
2968     },
2969     cpus => {
2970         description => "Maximum usable CPUs.",
2971         type => 'number',
2972         optional => 1,
2973     },
2974 };
2975
2976 my $last_proc_pid_stat;
2977
2978 # get VM status information
2979 # This must be fast and should not block ($full == false)
2980 # We only query KVM using QMP if $full == true (this can be slow)
2981 sub vmstatus {
2982     my ($opt_vmid, $full) = @_;
2983
2984     my $res = {};
2985
2986     my $storecfg = PVE::Storage::config();
2987
2988     my $list = vzlist();
2989     my $defaults = load_defaults();
2990
2991     my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2992
2993     my $cpucount = $cpuinfo->{cpus} || 1;
2994
2995     foreach my $vmid (keys %$list) {
2996         next if $opt_vmid && ($vmid ne $opt_vmid);
2997
2998         my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
2999         my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
3000
3001         my $d = { vmid => $vmid };
3002         $d->{pid} = $list->{$vmid}->{pid};
3003
3004         # fixme: better status?
3005         $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3006
3007         my $size = disksize($storecfg, $conf);
3008         if (defined($size)) {
3009             $d->{disk} = 0; # no info available
3010             $d->{maxdisk} = $size;
3011         } else {
3012             $d->{disk} = 0;
3013             $d->{maxdisk} = 0;
3014         }
3015
3016         $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3017             * ($conf->{cores} || $defaults->{cores});
3018         $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3019         $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3020
3021         $d->{name} = $conf->{name} || "VM $vmid";
3022         $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3023             : $defaults->{memory}*(1024*1024);
3024
3025         if ($conf->{balloon}) {
3026             $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3027             $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3028                 : $defaults->{shares};
3029         }
3030
3031         $d->{uptime} = 0;
3032         $d->{cpu} = 0;
3033         $d->{mem} = 0;
3034
3035         $d->{netout} = 0;
3036         $d->{netin} = 0;
3037
3038         $d->{diskread} = 0;
3039         $d->{diskwrite} = 0;
3040
3041         $d->{template} = PVE::QemuConfig->is_template($conf);
3042
3043         $d->{serial} = 1 if conf_has_serial($conf);
3044
3045         $res->{$vmid} = $d;
3046     }
3047
3048     my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3049     foreach my $dev (keys %$netdev) {
3050         next if $dev !~ m/^tap([1-9]\d*)i/;
3051         my $vmid = $1;
3052         my $d = $res->{$vmid};
3053         next if !$d;
3054
3055         $d->{netout} += $netdev->{$dev}->{receive};
3056         $d->{netin} += $netdev->{$dev}->{transmit};
3057
3058         if ($full) {
3059             $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3060             $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3061         }
3062
3063     }
3064
3065     my $ctime = gettimeofday;
3066
3067     foreach my $vmid (keys %$list) {
3068
3069         my $d = $res->{$vmid};
3070         my $pid = $d->{pid};
3071         next if !$pid;
3072
3073         my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3074         next if !$pstat; # not running
3075
3076         my $used = $pstat->{utime} + $pstat->{stime};
3077
3078         $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3079
3080         if ($pstat->{vsize}) {
3081             $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3082         }
3083
3084         my $old = $last_proc_pid_stat->{$pid};
3085         if (!$old) {
3086             $last_proc_pid_stat->{$pid} = {
3087                 time => $ctime,
3088                 used => $used,
3089                 cpu => 0,
3090             };
3091             next;
3092         }
3093
3094         my $dtime = ($ctime -  $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3095
3096         if ($dtime > 1000) {
3097             my $dutime = $used -  $old->{used};
3098
3099             $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3100             $last_proc_pid_stat->{$pid} = {
3101                 time => $ctime,
3102                 used => $used,
3103                 cpu => $d->{cpu},
3104             };
3105         } else {
3106             $d->{cpu} = $old->{cpu};
3107         }
3108     }
3109
3110     return $res if !$full;
3111
3112     my $qmpclient = PVE::QMPClient->new();
3113
3114     my $ballooncb = sub {
3115         my ($vmid, $resp) = @_;
3116
3117         my $info = $resp->{'return'};
3118         return if !$info->{max_mem};
3119
3120         my $d = $res->{$vmid};
3121
3122         # use memory assigned to VM
3123         $d->{maxmem} = $info->{max_mem};
3124         $d->{balloon} = $info->{actual};
3125
3126         if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3127             $d->{mem} = $info->{total_mem} - $info->{free_mem};
3128             $d->{freemem} = $info->{free_mem};
3129         }
3130
3131         $d->{ballooninfo} = $info;
3132     };
3133
3134     my $blockstatscb = sub {
3135         my ($vmid, $resp) = @_;
3136         my $data = $resp->{'return'} || [];
3137         my $totalrdbytes = 0;
3138         my $totalwrbytes = 0;
3139
3140         for my $blockstat (@$data) {
3141             $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3142             $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3143
3144             $blockstat->{device} =~ s/drive-//;
3145             $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3146         }
3147         $res->{$vmid}->{diskread} = $totalrdbytes;
3148         $res->{$vmid}->{diskwrite} = $totalwrbytes;
3149     };
3150
3151     my $statuscb = sub {
3152         my ($vmid, $resp) = @_;
3153
3154         $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3155         # this fails if ballon driver is not loaded, so this must be
3156         # the last commnand (following command are aborted if this fails).
3157         $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3158
3159         my $status = 'unknown';
3160         if (!defined($status = $resp->{'return'}->{status})) {
3161             warn "unable to get VM status\n";
3162             return;
3163         }
3164
3165         $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3166     };
3167
3168     foreach my $vmid (keys %$list) {
3169         next if $opt_vmid && ($vmid ne $opt_vmid);
3170         next if !$res->{$vmid}->{pid}; # not running
3171         $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3172     }
3173
3174     $qmpclient->queue_execute(undef, 2);
3175
3176     foreach my $vmid (keys %$list) {
3177         next if $opt_vmid && ($vmid ne $opt_vmid);
3178         $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3179     }
3180
3181     return $res;
3182 }
3183
3184 sub foreach_drive {
3185     my ($conf, $func, @param) = @_;
3186
3187     foreach my $ds (valid_drive_names()) {
3188         next if !defined($conf->{$ds});
3189
3190         my $drive = parse_drive($ds, $conf->{$ds});
3191         next if !$drive;
3192
3193         &$func($ds, $drive, @param);
3194     }
3195 }
3196
3197 sub foreach_volid {
3198     my ($conf, $func, @param) = @_;
3199
3200     my $volhash = {};
3201
3202     my $test_volid = sub {
3203         my ($volid, $is_cdrom, $replicate, $shared, $snapname) = @_;
3204
3205         return if !$volid;
3206
3207         $volhash->{$volid}->{cdrom} //= 1;
3208         $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3209
3210         $volhash->{$volid}->{replicate} //= 0;
3211         $volhash->{$volid}->{replicate} = 1 if $replicate;
3212
3213         $volhash->{$volid}->{shared} //= 0;
3214         $volhash->{$volid}->{shared} = 1 if $shared;
3215
3216         $volhash->{$volid}->{referenced_in_config} //= 0;
3217         $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3218
3219         $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3220             if defined($snapname);
3221     };
3222
3223     foreach_drive($conf, sub {
3224         my ($ds, $drive) = @_;
3225         $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef);
3226     });
3227
3228     foreach my $snapname (keys %{$conf->{snapshots}}) {
3229         my $snap = $conf->{snapshots}->{$snapname};
3230         $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3231         foreach_drive($snap, sub {
3232             my ($ds, $drive) = @_;
3233             $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3234         });
3235     }
3236
3237     foreach my $volid (keys %$volhash) {
3238         &$func($volid, $volhash->{$volid}, @param);
3239     }
3240 }
3241
3242 sub conf_has_serial {
3243     my ($conf) = @_;
3244
3245     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3246         if ($conf->{"serial$i"}) {
3247             return 1;
3248         }
3249     }
3250
3251     return 0;
3252 }
3253
3254 sub vga_conf_has_spice {
3255     my ($vga) = @_;
3256
3257     my $vgaconf = parse_vga($vga);
3258     my $vgatype = $vgaconf->{type};
3259     return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3260
3261     return $1 || 1;
3262 }
3263
3264 my $host_arch; # FIXME: fix PVE::Tools::get_host_arch
3265 sub get_host_arch() {
3266     $host_arch = (POSIX::uname())[4] if !$host_arch;
3267     return $host_arch;
3268 }
3269
3270 sub is_native($) {
3271     my ($arch) = @_;
3272     return get_host_arch() eq $arch;
3273 }
3274
3275 my $default_machines = {
3276     x86_64 => 'pc',
3277     aarch64 => 'virt',
3278 };
3279
3280 sub get_basic_machine_info {
3281     my ($conf, $forcemachine) = @_;
3282
3283     my $arch = $conf->{arch} // get_host_arch();
3284     my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3285     return ($arch, $machine);
3286 }
3287
3288 sub get_ovmf_files($) {
3289     my ($arch) = @_;
3290
3291     my $ovmf = $OVMF->{$arch}
3292         or die "no OVMF images known for architecture '$arch'\n";
3293
3294     return @$ovmf;
3295 }
3296
3297 my $Arch2Qemu = {
3298     aarch64 => '/usr/bin/qemu-system-aarch64',
3299     x86_64 => '/usr/bin/qemu-system-x86_64',
3300 };
3301 sub get_command_for_arch($) {
3302     my ($arch) = @_;
3303     return '/usr/bin/kvm' if is_native($arch);
3304
3305     my $cmd = $Arch2Qemu->{$arch}
3306         or die "don't know how to emulate architecture '$arch'\n";
3307     return $cmd;
3308 }
3309
3310 sub get_cpu_options {
3311     my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3312
3313     my $cpuFlags = [];
3314     my $ostype = $conf->{ostype};
3315
3316     my $cpu = $kvm ? "kvm64" : "qemu64";
3317     if ($arch eq 'aarch64') {
3318         $cpu = 'cortex-a57';
3319     }
3320     if (my $cputype = $conf->{cpu}) {
3321         my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3322             or die "Cannot parse cpu description: $cputype\n";
3323         $cpu = $cpuconf->{cputype};
3324         $kvm_off = 1 if $cpuconf->{hidden};
3325
3326         if (defined(my $flags = $cpuconf->{flags})) {
3327             push @$cpuFlags, split(";", $flags);
3328         }
3329     }
3330
3331     push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
3332
3333     push @$cpuFlags , '-x2apic'
3334         if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3335
3336     push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3337
3338     push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3339
3340     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3341
3342         push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3343         push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3344     }
3345
3346     add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough) if $kvm;
3347
3348     push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm;
3349
3350     push @$cpuFlags, 'kvm=off' if $kvm_off;
3351
3352     if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3353         push @$cpuFlags, "vendor=${cpu_vendor}"
3354             if $cpu_vendor ne 'default';
3355     } elsif ($arch ne 'aarch64') {
3356         die "internal error"; # should not happen
3357     }
3358
3359     $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3360
3361     return ('-cpu', $cpu);
3362 }
3363
3364 sub config_to_command {
3365     my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3366
3367     my $cmd = [];
3368     my $globalFlags = [];
3369     my $machineFlags = [];
3370     my $rtcFlags = [];
3371     my $devices = [];
3372     my $pciaddr = '';
3373     my $bridges = {};
3374     my $kvmver = kvm_user_version();
3375     my $vernum = 0; # unknown
3376     my $ostype = $conf->{ostype};
3377     my $winversion = windows_version($ostype);
3378     my $kvm = $conf->{kvm};
3379
3380     my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3381     $kvm //= 1 if is_native($arch);
3382
3383     if ($kvm) {
3384         die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3385             if !defined kvm_version();
3386     }
3387
3388     if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3389         $vernum = $1*1000000+$2*1000;
3390     } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3391         $vernum = $1*1000000+$2*1000+$3;
3392     }
3393
3394     die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3395
3396     my $have_ovz = -f '/proc/vz/vestat';
3397
3398     my $q35 = machine_type_is_q35($conf);
3399     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3400     my $use_old_bios_files = undef;
3401     ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3402
3403     my $cpuunits = defined($conf->{cpuunits}) ?
3404             $conf->{cpuunits} : $defaults->{cpuunits};
3405
3406     push @$cmd, get_command_for_arch($arch);
3407
3408     push @$cmd, '-id', $vmid;
3409
3410     my $vmname = $conf->{name} || "vm$vmid";
3411
3412     push @$cmd, '-name', $vmname;
3413
3414     my $use_virtio = 0;
3415
3416     my $qmpsocket = qmp_socket($vmid);
3417     push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3418     push @$cmd, '-mon', "chardev=qmp,mode=control";
3419
3420     if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3421         my $eventsocket = qmp_socket($vmid, 0, 'event');
3422         push @$cmd, '-chardev', "socket,id=qmp-event,path=$eventsocket,server,nowait";
3423         push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3424     }
3425
3426     push @$cmd, '-pidfile' , pidfile_name($vmid);
3427
3428     push @$cmd, '-daemonize';
3429
3430     if ($conf->{smbios1}) {
3431         push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3432     }
3433
3434     if ($conf->{vmgenid}) {
3435         push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3436     }
3437
3438     my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3439     if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3440         die "uefi base image not found\n" if ! -f $ovmf_code;
3441
3442         my $path;
3443         my $format;
3444         if (my $efidisk = $conf->{efidisk0}) {
3445             my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3446             my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3447             $format = $d->{format};
3448             if ($storeid) {
3449                 $path = PVE::Storage::path($storecfg, $d->{file});
3450                 if (!defined($format)) {
3451                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3452                     $format = qemu_img_format($scfg, $volname);
3453                 }
3454             } else {
3455                 $path = $d->{file};
3456                 die "efidisk format must be specified\n"
3457                     if !defined($format);
3458             }
3459         } else {
3460             warn "no efidisk configured! Using temporary efivars disk.\n";
3461             $path = "/tmp/$vmid-ovmf.fd";
3462             PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3463             $format = 'raw';
3464         }
3465
3466         push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3467         push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3468     }
3469
3470
3471     # add usb controllers
3472     my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $q35, $usbdesc->{format}, $MAX_USB_DEVICES);
3473     push @$devices, @usbcontrollers if @usbcontrollers;
3474     my $vga = parse_vga($conf->{vga});
3475
3476     my $qxlnum = vga_conf_has_spice($conf->{vga});
3477     $vga->{type} = 'qxl' if $qxlnum;
3478
3479     if (!$vga->{type}) {
3480         if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3481             $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3482         } else {
3483             $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3484         }
3485     }
3486
3487     # enable absolute mouse coordinates (needed by vnc)
3488     my $tablet;
3489     if (defined($conf->{tablet})) {
3490         $tablet = $conf->{tablet};
3491     } else {
3492         $tablet = $defaults->{tablet};
3493         $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3494         $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3495     }
3496
3497     push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
3498
3499     my $kvm_off = 0;
3500     my $gpu_passthrough;
3501
3502     # host pci devices
3503     for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
3504         my $d = parse_hostpci($conf->{"hostpci$i"});
3505         next if !$d;
3506
3507         my $pcie = $d->{pcie};
3508         if($pcie){
3509             die "q35 machine model is not enabled" if !$q35;
3510             $pciaddr = print_pcie_addr("hostpci$i");
3511         }else{
3512             $pciaddr = print_pci_addr("hostpci$i", $bridges);
3513         }
3514
3515         my $rombar = defined($d->{rombar}) && !$d->{rombar} ? ',rombar=0' : '';
3516         my $romfile = $d->{romfile};
3517
3518         my $xvga = '';
3519         if ($d->{'x-vga'}) {
3520             $xvga = ',x-vga=on';
3521             $kvm_off = 1;
3522             $vga->{type} = 'none';
3523             $gpu_passthrough = 1;
3524
3525             if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3526                 $xvga = "";
3527             }
3528         }
3529         my $pcidevices = $d->{pciid};
3530         my $multifunction = 1 if @$pcidevices > 1;
3531
3532         my $j=0;
3533         foreach my $pcidevice (@$pcidevices) {
3534
3535             my $id = "hostpci$i";
3536             $id .= ".$j" if $multifunction;
3537             my $addr = $pciaddr;
3538             $addr .= ".$j" if $multifunction;
3539             my $devicestr = "vfio-pci,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
3540
3541             if($j == 0){
3542                 $devicestr .= "$rombar$xvga";
3543                 $devicestr .= ",multifunction=on" if $multifunction;
3544                 $devicestr .= ",romfile=/usr/share/kvm/$romfile" if $romfile;
3545             }
3546
3547             push @$devices, '-device', $devicestr;
3548             $j++;
3549         }
3550     }
3551
3552     # usb devices
3553     my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES);
3554     push @$devices, @usbdevices if @usbdevices;
3555     # serial devices
3556     for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++)  {
3557         if (my $path = $conf->{"serial$i"}) {
3558             if ($path eq 'socket') {
3559                 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3560                 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3561                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3562             } else {
3563                 die "no such serial device\n" if ! -c $path;
3564                 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3565                 push @$devices, '-device', "isa-serial,chardev=serial$i";
3566             }
3567         }
3568     }
3569
3570     # parallel devices
3571     for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++)  {
3572         if (my $path = $conf->{"parallel$i"}) {
3573             die "no such parallel device\n" if ! -c $path;
3574             my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3575             push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3576             push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3577         }
3578     }
3579
3580
3581     my $sockets = 1;
3582     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3583     $sockets = $conf->{sockets} if  $conf->{sockets};
3584
3585     my $cores = $conf->{cores} || 1;
3586
3587     my $maxcpus = $sockets * $cores;
3588
3589     my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3590
3591     my $allowed_vcpus = $cpuinfo->{cpus};
3592
3593     die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3594         if ($allowed_vcpus < $maxcpus);
3595
3596     if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3597
3598         push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3599         for (my $i = 2; $i <= $vcpus; $i++)  {
3600             my $cpustr = print_cpu_device($conf,$i);
3601             push @$cmd, '-device', $cpustr;
3602         }
3603
3604     } else {
3605
3606         push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3607     }
3608     push @$cmd, '-nodefaults';
3609
3610     my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3611
3612     my $bootindex_hash = {};
3613     my $i = 1;
3614     foreach my $o (split(//, $bootorder)) {
3615         $bootindex_hash->{$o} = $i*100;
3616         $i++;
3617     }
3618
3619     push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3620
3621     push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3622
3623     push @$cmd, '-no-reboot' if  defined($conf->{reboot}) && $conf->{reboot} == 0;
3624
3625     if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3626         push @$devices, '-device', print_vga_device($conf, $vga, undef, $qxlnum, $bridges);
3627         my $socket = vnc_socket($vmid);
3628         push @$cmd,  '-vnc', "unix:$socket,x509,password";
3629     } else {
3630         push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3631         push @$cmd, '-nographic';
3632     }
3633
3634     # time drift fix
3635     my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3636
3637     my $useLocaltime = $conf->{localtime};
3638
3639     if ($winversion >= 5) { # windows
3640         $useLocaltime = 1 if !defined($conf->{localtime});
3641
3642         # use time drift fix when acpi is enabled
3643         if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3644             $tdf = 1 if !defined($conf->{tdf});
3645         }
3646     }
3647
3648     if ($winversion >= 6) {
3649         push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3650         push @$cmd, '-no-hpet';
3651     }
3652
3653     push @$rtcFlags, 'driftfix=slew' if $tdf;
3654
3655     if (!$kvm) {
3656         push @$machineFlags, 'accel=tcg';
3657     }
3658
3659     if ($machine_type) {
3660         push @$machineFlags, "type=${machine_type}";
3661     }
3662
3663     if ($conf->{startdate}) {
3664         push @$rtcFlags, "base=$conf->{startdate}";
3665     } elsif ($useLocaltime) {
3666         push @$rtcFlags, 'base=localtime';
3667     }
3668
3669     push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3670
3671     PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3672
3673     push @$cmd, '-S' if $conf->{freeze};
3674
3675     push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3676
3677     # enable sound
3678     #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3679     #push @$cmd, '-soundhw', 'es1370';
3680     #push @$cmd, '-soundhw', $soundhw if $soundhw;
3681
3682     if (parse_guest_agent($conf)->{enabled}) {
3683         my $qgasocket = qmp_socket($vmid, 1);
3684         my $pciaddr = print_pci_addr("qga0", $bridges);
3685         push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3686         push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3687         push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3688     }
3689
3690     my $spice_port;
3691
3692     if ($qxlnum) {
3693         if ($qxlnum > 1) {
3694             if ($winversion){
3695                 for(my $i = 1; $i < $qxlnum; $i++){
3696                     push @$devices, '-device', print_vga_device($conf, $vga, $i, $qxlnum, $bridges);
3697                 }
3698             } else {
3699                 # assume other OS works like Linux
3700                 my ($ram, $vram) = ("134217728", "67108864");
3701                 if ($vga->{memory}) {
3702                     $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3703                     $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3704                 }
3705                 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3706                 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3707             }
3708         }
3709
3710         my $pciaddr = print_pci_addr("spice", $bridges);
3711
3712         my $nodename = PVE::INotify::nodename();
3713         my $pfamily = PVE::Tools::get_host_address_family($nodename);
3714         my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3715         die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3716         my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3717         $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3718
3719         push @$devices, '-spice', "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3720
3721         push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3722         push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3723         push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3724     }
3725
3726     # enable balloon by default, unless explicitly disabled
3727     if (!defined($conf->{balloon}) || $conf->{balloon}) {
3728         $pciaddr = print_pci_addr("balloon0", $bridges);
3729         push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3730     }
3731
3732     if ($conf->{watchdog}) {
3733         my $wdopts = parse_watchdog($conf->{watchdog});
3734         $pciaddr = print_pci_addr("watchdog", $bridges);
3735         my $watchdog = $wdopts->{model} || 'i6300esb';
3736         push @$devices, '-device', "$watchdog$pciaddr";
3737         push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3738     }
3739
3740     my $vollist = [];
3741     my $scsicontroller = {};
3742     my $ahcicontroller = {};
3743     my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3744
3745     # Add iscsi initiator name if available
3746     if (my $initiator = get_initiator_name()) {
3747         push @$devices, '-iscsi', "initiator-name=$initiator";
3748     }
3749
3750     foreach_drive($conf, sub {
3751         my ($ds, $drive) = @_;
3752
3753         if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3754             push @$vollist, $drive->{file};
3755         }
3756
3757         # ignore efidisk here, already added in bios/fw handling code above
3758         return if $drive->{interface} eq 'efidisk';
3759
3760         $use_virtio = 1 if $ds =~ m/^virtio/;
3761
3762         if (drive_is_cdrom ($drive)) {
3763             if ($bootindex_hash->{d}) {
3764                 $drive->{bootindex} = $bootindex_hash->{d};
3765                 $bootindex_hash->{d} += 1;
3766             }
3767         } else {
3768             if ($bootindex_hash->{c}) {
3769                 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3770                 $bootindex_hash->{c} += 1;
3771             }
3772         }
3773
3774         if($drive->{interface} eq 'virtio'){
3775            push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3776         }
3777
3778         if ($drive->{interface} eq 'scsi') {
3779
3780             my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3781
3782             $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3783             my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3784
3785             my $iothread = '';
3786             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3787                 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3788                 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3789             } elsif ($drive->{iothread}) {
3790                 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
3791             }
3792
3793             my $queues = '';
3794             if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3795                 $queues = ",num_queues=$drive->{queues}";
3796             }
3797
3798             push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3799             $scsicontroller->{$controller}=1;
3800         }
3801
3802         if ($drive->{interface} eq 'sata') {
3803            my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3804            $pciaddr = print_pci_addr("ahci$controller", $bridges);
3805            push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3806            $ahcicontroller->{$controller}=1;
3807         }
3808
3809         my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3810         push @$devices, '-drive',$drive_cmd;
3811         push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3812     });
3813
3814     for (my $i = 0; $i < $MAX_NETS; $i++) {
3815          next if !$conf->{"net$i"};
3816          my $d = parse_net($conf->{"net$i"});
3817          next if !$d;
3818
3819          $use_virtio = 1 if $d->{model} eq 'virtio';
3820
3821          if ($bootindex_hash->{n}) {
3822             $d->{bootindex} = $bootindex_hash->{n};
3823             $bootindex_hash->{n} += 1;
3824          }
3825
3826          my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3827          push @$devices, '-netdev', $netdevfull;
3828
3829          my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files);
3830          push @$devices, '-device', $netdevicefull;
3831     }
3832
3833     if (!$q35) {
3834         # add pci bridges
3835         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3836            $bridges->{1} = 1;
3837            $bridges->{2} = 1;
3838         }
3839
3840         $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3841
3842         while (my ($k, $v) = each %$bridges) {
3843             $pciaddr = print_pci_addr("pci.$k");
3844             unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3845         }
3846     }
3847
3848     # add custom args
3849     if ($conf->{args}) {
3850         my $aa = PVE::Tools::split_args($conf->{args});
3851         push @$cmd, @$aa;
3852     }
3853
3854     push @$cmd, @$devices;
3855     push @$cmd, '-rtc', join(',', @$rtcFlags)
3856         if scalar(@$rtcFlags);
3857     push @$cmd, '-machine', join(',', @$machineFlags)
3858         if scalar(@$machineFlags);
3859     push @$cmd, '-global', join(',', @$globalFlags)
3860         if scalar(@$globalFlags);
3861
3862     return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3863 }
3864
3865 sub vnc_socket {
3866     my ($vmid) = @_;
3867     return "${var_run_tmpdir}/$vmid.vnc";
3868 }
3869
3870 sub spice_port {
3871     my ($vmid) = @_;
3872
3873     my $res = vm_mon_cmd($vmid, 'query-spice');
3874
3875     return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3876 }
3877
3878 sub qmp_socket {
3879     my ($vmid, $qga, $name) = @_;
3880     my $sockettype = $qga ? 'qga' : 'qmp';
3881     my $ext = $name ? '-'.$name : '';
3882     return "${var_run_tmpdir}/$vmid$ext.$sockettype";
3883 }
3884
3885 sub pidfile_name {
3886     my ($vmid) = @_;
3887     return "${var_run_tmpdir}/$vmid.pid";
3888 }
3889
3890 sub vm_devices_list {
3891     my ($vmid) = @_;
3892
3893     my $res = vm_mon_cmd($vmid, 'query-pci');
3894     my $devices_to_check = [];
3895     my $devices = {};
3896     foreach my $pcibus (@$res) {
3897         push @$devices_to_check, @{$pcibus->{devices}},
3898     }
3899
3900     while (@$devices_to_check) {
3901         my $to_check = [];
3902         for my $d (@$devices_to_check) {
3903             $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
3904             next if !$d->{'pci_bridge'};
3905
3906             $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
3907             push @$to_check, @{$d->{'pci_bridge'}->{devices}};
3908         }
3909         $devices_to_check = $to_check;
3910     }
3911
3912     my $resblock = vm_mon_cmd($vmid, 'query-block');
3913     foreach my $block (@$resblock) {
3914         if($block->{device} =~ m/^drive-(\S+)/){
3915                 $devices->{$1} = 1;
3916         }
3917     }
3918
3919     my $resmice = vm_mon_cmd($vmid, 'query-mice');
3920     foreach my $mice (@$resmice) {
3921         if ($mice->{name} eq 'QEMU HID Tablet') {
3922             $devices->{tablet} = 1;
3923             last;
3924         }
3925     }
3926
3927     # for usb devices there is no query-usb
3928     # but we can iterate over the entries in
3929     # qom-list path=/machine/peripheral
3930     my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
3931     foreach my $per (@$resperipheral) {
3932         if ($per->{name} =~ m/^usb\d+$/) {
3933             $devices->{$per->{name}} = 1;
3934         }
3935     }
3936
3937     return $devices;
3938 }
3939
3940 sub vm_deviceplug {
3941     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3942
3943     my $q35 = machine_type_is_q35($conf);
3944
3945     my $devices_list = vm_devices_list($vmid);
3946     return 1 if defined($devices_list->{$deviceid});
3947
3948     qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3949
3950     if ($deviceid eq 'tablet') {
3951
3952         qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3953
3954     } elsif ($deviceid =~ m/^usb(\d+)$/) {
3955
3956         die "usb hotplug currently not reliable\n";
3957         # since we can't reliably hot unplug all added usb devices
3958         # and usb passthrough disables live migration
3959         # we disable usb hotplugging for now
3960         qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
3961
3962     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3963
3964         qemu_iothread_add($vmid, $deviceid, $device);
3965
3966         qemu_driveadd($storecfg, $vmid, $device);
3967         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3968
3969         qemu_deviceadd($vmid, $devicefull);
3970         eval { qemu_deviceaddverify($vmid, $deviceid); };
3971         if (my $err = $@) {
3972             eval { qemu_drivedel($vmid, $deviceid); };
3973             warn $@ if $@;
3974             die $err;
3975         }
3976
3977     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3978
3979
3980         my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3981         my $pciaddr = print_pci_addr($deviceid);
3982         my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3983
3984         my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3985
3986         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3987             qemu_iothread_add($vmid, $deviceid, $device);
3988             $devicefull .= ",iothread=iothread-$deviceid";
3989         }
3990
3991         if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3992             $devicefull .= ",num_queues=$device->{queues}";
3993         }
3994
3995         qemu_deviceadd($vmid, $devicefull);
3996         qemu_deviceaddverify($vmid, $deviceid);
3997
3998     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3999
4000         qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
4001         qemu_driveadd($storecfg, $vmid, $device);
4002
4003         my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
4004         eval { qemu_deviceadd($vmid, $devicefull); };
4005         if (my $err = $@) {
4006             eval { qemu_drivedel($vmid, $deviceid); };
4007             warn $@ if $@;
4008             die $err;
4009         }
4010
4011     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4012
4013         return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
4014
4015         my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4016         my $use_old_bios_files = undef;
4017         ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4018
4019         my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files);
4020         qemu_deviceadd($vmid, $netdevicefull);
4021         eval { qemu_deviceaddverify($vmid, $deviceid); };
4022         if (my $err = $@) {
4023             eval { qemu_netdevdel($vmid, $deviceid); };
4024             warn $@ if $@;
4025             die $err;
4026         }
4027
4028     } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4029
4030         my $bridgeid = $2;
4031         my $pciaddr = print_pci_addr($deviceid);
4032         my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4033
4034         qemu_deviceadd($vmid, $devicefull);
4035         qemu_deviceaddverify($vmid, $deviceid);
4036
4037     } else {
4038         die "can't hotplug device '$deviceid'\n";
4039     }
4040
4041     return 1;
4042 }
4043
4044 # fixme: this should raise exceptions on error!
4045 sub vm_deviceunplug {
4046     my ($vmid, $conf, $deviceid) = @_;
4047
4048     my $devices_list = vm_devices_list($vmid);
4049     return 1 if !defined($devices_list->{$deviceid});
4050
4051     die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4052
4053     if ($deviceid eq 'tablet') {
4054
4055         qemu_devicedel($vmid, $deviceid);
4056
4057     } elsif ($deviceid =~ m/^usb\d+$/) {
4058
4059         die "usb hotplug currently not reliable\n";
4060         # when unplugging usb devices this way,
4061         # there may be remaining usb controllers/hubs
4062         # so we disable it for now
4063         qemu_devicedel($vmid, $deviceid);
4064         qemu_devicedelverify($vmid, $deviceid);
4065
4066     } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4067
4068         qemu_devicedel($vmid, $deviceid);
4069         qemu_devicedelverify($vmid, $deviceid);
4070         qemu_drivedel($vmid, $deviceid);
4071         qemu_iothread_del($conf, $vmid, $deviceid);
4072
4073     } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4074
4075         qemu_devicedel($vmid, $deviceid);
4076         qemu_devicedelverify($vmid, $deviceid);
4077         qemu_iothread_del($conf, $vmid, $deviceid);
4078
4079     } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4080
4081         qemu_devicedel($vmid, $deviceid);
4082         qemu_drivedel($vmid, $deviceid);
4083         qemu_deletescsihw($conf, $vmid, $deviceid);
4084
4085     } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4086
4087         qemu_devicedel($vmid, $deviceid);
4088         qemu_devicedelverify($vmid, $deviceid);
4089         qemu_netdevdel($vmid, $deviceid);
4090
4091     } else {
4092         die "can't unplug device '$deviceid'\n";
4093     }
4094
4095     return 1;
4096 }
4097
4098 sub qemu_deviceadd {
4099     my ($vmid, $devicefull) = @_;
4100
4101     $devicefull = "driver=".$devicefull;
4102     my %options =  split(/[=,]/, $devicefull);
4103
4104     vm_mon_cmd($vmid, "device_add" , %options);
4105 }
4106
4107 sub qemu_devicedel {
4108     my ($vmid, $deviceid) = @_;
4109
4110     my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4111 }
4112
4113 sub qemu_iothread_add {
4114     my($vmid, $deviceid, $device) = @_;
4115
4116     if ($device->{iothread}) {
4117         my $iothreads = vm_iothreads_list($vmid);
4118         qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4119     }
4120 }
4121
4122 sub qemu_iothread_del {
4123     my($conf, $vmid, $deviceid) = @_;
4124
4125     my $device = parse_drive($deviceid, $conf->{$deviceid});
4126     if ($device->{iothread}) {
4127         my $iothreads = vm_iothreads_list($vmid);
4128         qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4129     }
4130 }
4131
4132 sub qemu_objectadd {
4133     my($vmid, $objectid, $qomtype) = @_;
4134
4135     vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4136
4137     return 1;
4138 }
4139
4140 sub qemu_objectdel {
4141     my($vmid, $objectid) = @_;
4142
4143     vm_mon_cmd($vmid, "object-del", id => $objectid);
4144
4145     return 1;
4146 }
4147
4148 sub qemu_driveadd {
4149     my ($storecfg, $vmid, $device) = @_;
4150
4151     my $drive = print_drive_full($storecfg, $vmid, $device);
4152     $drive =~ s/\\/\\\\/g;
4153     my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4154
4155     # If the command succeeds qemu prints: "OK"
4156     return 1 if $ret =~ m/OK/s;
4157
4158     die "adding drive failed: $ret\n";
4159 }
4160
4161 sub qemu_drivedel {
4162     my($vmid, $deviceid) = @_;
4163
4164     my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4165     $ret =~ s/^\s+//;
4166
4167     return 1 if $ret eq "";
4168
4169     # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4170     return 1 if $ret =~ m/Device \'.*?\' not found/s;
4171
4172     die "deleting drive $deviceid failed : $ret\n";
4173 }
4174
4175 sub qemu_deviceaddverify {
4176     my ($vmid, $deviceid) = @_;
4177
4178     for (my $i = 0; $i <= 5; $i++) {
4179          my $devices_list = vm_devices_list($vmid);
4180          return 1 if defined($devices_list->{$deviceid});
4181          sleep 1;
4182     }
4183
4184     die "error on hotplug device '$deviceid'\n";
4185 }
4186
4187
4188 sub qemu_devicedelverify {
4189     my ($vmid, $deviceid) = @_;
4190
4191     # need to verify that the device is correctly removed as device_del
4192     # is async and empty return is not reliable
4193
4194     for (my $i = 0; $i <= 5; $i++) {
4195          my $devices_list = vm_devices_list($vmid);
4196          return 1 if !defined($devices_list->{$deviceid});
4197          sleep 1;
4198     }
4199
4200     die "error on hot-unplugging device '$deviceid'\n";
4201 }
4202
4203 sub qemu_findorcreatescsihw {
4204     my ($storecfg, $conf, $vmid, $device) = @_;
4205
4206     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4207
4208     my $scsihwid="$controller_prefix$controller";
4209     my $devices_list = vm_devices_list($vmid);
4210
4211     if(!defined($devices_list->{$scsihwid})) {
4212         vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
4213     }
4214
4215     return 1;
4216 }
4217
4218 sub qemu_deletescsihw {
4219     my ($conf, $vmid, $opt) = @_;
4220
4221     my $device = parse_drive($opt, $conf->{$opt});
4222
4223     if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4224         vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4225         return 1;
4226     }
4227
4228     my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4229
4230     my $devices_list = vm_devices_list($vmid);
4231     foreach my $opt (keys %{$devices_list}) {
4232         if (PVE::QemuServer::is_valid_drivename($opt)) {
4233             my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4234             if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4235                 return 1;
4236             }
4237         }
4238     }
4239
4240     my $scsihwid="scsihw$controller";
4241
4242     vm_deviceunplug($vmid, $conf, $scsihwid);
4243
4244     return 1;
4245 }
4246
4247 sub qemu_add_pci_bridge {
4248     my ($storecfg, $conf, $vmid, $device) = @_;
4249
4250     my $bridges = {};
4251
4252     my $bridgeid;
4253
4254     print_pci_addr($device, $bridges);
4255
4256     while (my ($k, $v) = each %$bridges) {
4257         $bridgeid = $k;
4258     }
4259     return 1 if !defined($bridgeid) || $bridgeid < 1;
4260
4261     my $bridge = "pci.$bridgeid";
4262     my $devices_list = vm_devices_list($vmid);
4263
4264     if (!defined($devices_list->{$bridge})) {
4265         vm_deviceplug($storecfg, $conf, $vmid, $bridge);
4266     }
4267
4268     return 1;
4269 }
4270
4271 sub qemu_set_link_status {
4272     my ($vmid, $device, $up) = @_;
4273
4274     vm_mon_cmd($vmid, "set_link", name => $device,
4275                up => $up ? JSON::true : JSON::false);
4276 }
4277
4278 sub qemu_netdevadd {
4279     my ($vmid, $conf, $device, $deviceid) = @_;
4280
4281     my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid, 1);
4282     my %options =  split(/[=,]/, $netdev);
4283
4284     vm_mon_cmd($vmid, "netdev_add",  %options);
4285     return 1;
4286 }
4287
4288 sub qemu_netdevdel {
4289     my ($vmid, $deviceid) = @_;
4290
4291     vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4292 }
4293
4294 sub qemu_usb_hotplug {
4295     my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
4296
4297     return if !$device;
4298
4299     # remove the old one first
4300     vm_deviceunplug($vmid, $conf, $deviceid);
4301
4302     # check if xhci controller is necessary and available
4303     if ($device->{usb3}) {
4304
4305         my $devicelist = vm_devices_list($vmid);
4306
4307         if (!$devicelist->{xhci}) {
4308             my $pciaddr = print_pci_addr("xhci");
4309             qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4310         }
4311     }
4312     my $d = parse_usb_device($device->{host});
4313     $d->{usb3} = $device->{usb3};
4314
4315     # add the new one
4316     vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d);
4317 }
4318
4319 sub qemu_cpu_hotplug {
4320     my ($vmid, $conf, $vcpus) = @_;
4321
4322     my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4323
4324     my $sockets = 1;
4325     $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4326     $sockets = $conf->{sockets} if  $conf->{sockets};
4327     my $cores = $conf->{cores} || 1;
4328     my $maxcpus = $sockets * $cores;
4329
4330     $vcpus = $maxcpus if !$vcpus;
4331
4332     die "you can't add more vcpus than maxcpus\n"
4333         if $vcpus > $maxcpus;
4334
4335     my $currentvcpus = $conf->{vcpus} || $maxcpus;
4336
4337     if ($vcpus < $currentvcpus) {
4338
4339         if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4340
4341             for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4342                 qemu_devicedel($vmid, "cpu$i");
4343                 my $retry = 0;
4344                 my $currentrunningvcpus = undef;
4345                 while (1) {
4346                     $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4347                     last if scalar(@{$currentrunningvcpus}) == $i-1;
4348                     raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4349                     $retry++;
4350                     sleep 1;
4351                 }
4352                 #update conf after each succesfull cpu unplug
4353                 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4354                 PVE::QemuConfig->write_config($vmid, $conf);
4355             }
4356         } else {
4357             die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4358         }
4359
4360         return;
4361     }
4362
4363     my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4364     die "vcpus in running vm does not match its configuration\n"
4365         if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4366
4367     if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4368
4369         for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4370             my $cpustr = print_cpu_device($conf, $i);
4371             qemu_deviceadd($vmid, $cpustr);
4372
4373             my $retry = 0;
4374             my $currentrunningvcpus = undef;
4375             while (1) {
4376                 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4377                 last if scalar(@{$currentrunningvcpus}) == $i;
4378                 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4379                 sleep 1;
4380                 $retry++;
4381             }
4382             #update conf after each succesfull cpu hotplug
4383             $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4384             PVE::QemuConfig->write_config($vmid, $conf);
4385         }
4386     } else {
4387
4388         for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4389             vm_mon_cmd($vmid, "cpu-add", id => int($i));
4390         }
4391     }
4392 }
4393
4394 sub qemu_block_set_io_throttle {
4395     my ($vmid, $deviceid,
4396         $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4397         $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4398         $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4399         $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4400
4401     return if !check_running($vmid) ;
4402
4403     vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4404         bps => int($bps),
4405         bps_rd => int($bps_rd),
4406         bps_wr => int($bps_wr),
4407         iops => int($iops),
4408         iops_rd => int($iops_rd),
4409         iops_wr => int($iops_wr),
4410         bps_max => int($bps_max),
4411         bps_rd_max => int($bps_rd_max),
4412         bps_wr_max => int($bps_wr_max),
4413         iops_max => int($iops_max),
4414         iops_rd_max => int($iops_rd_max),
4415         iops_wr_max => int($iops_wr_max),
4416         bps_max_length => int($bps_max_length),
4417         bps_rd_max_length => int($bps_rd_max_length),
4418         bps_wr_max_length => int($bps_wr_max_length),
4419         iops_max_length => int($iops_max_length),
4420         iops_rd_max_length => int($iops_rd_max_length),
4421         iops_wr_max_length => int($iops_wr_max_length),
4422     );
4423
4424 }
4425
4426 # old code, only used to shutdown old VM after update
4427 sub __read_avail {
4428     my ($fh, $timeout) = @_;
4429
4430     my $sel = new IO::Select;
4431     $sel->add($fh);
4432
4433     my $res = '';
4434     my $buf;
4435
4436     my @ready;
4437     while (scalar (@ready = $sel->can_read($timeout))) {
4438         my $count;
4439         if ($count = $fh->sysread($buf, 8192)) {
4440             if ($buf =~ /^(.*)\(qemu\) $/s) {
4441                 $res .= $1;
4442                 last;
4443             } else {
4444                 $res .= $buf;
4445             }
4446         } else {
4447             if (!defined($count)) {
4448                 die "$!\n";
4449             }
4450             last;
4451         }
4452     }
4453
4454     die "monitor read timeout\n" if !scalar(@ready);
4455
4456     return $res;
4457 }
4458
4459 sub qemu_block_resize {
4460     my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4461
4462     my $running = check_running($vmid);
4463
4464     $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4465
4466     return if !$running;
4467
4468     vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4469
4470 }
4471
4472 sub qemu_volume_snapshot {
4473     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4474
4475     my $running = check_running($vmid);
4476
4477     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4478         vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4479     } else {
4480         PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4481     }
4482 }
4483
4484 sub qemu_volume_snapshot_delete {
4485     my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4486
4487     my $running = check_running($vmid);
4488
4489     if($running) {
4490
4491         $running = undef;
4492         my $conf = PVE::QemuConfig->load_config($vmid);
4493         foreach_drive($conf, sub {
4494             my ($ds, $drive) = @_;
4495             $running = 1 if $drive->{file} eq $volid;
4496         });
4497     }
4498
4499     if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4500         vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4501     } else {
4502         PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4503     }
4504 }
4505
4506 sub set_migration_caps {
4507     my ($vmid) = @_;
4508
4509     my $cap_ref = [];
4510
4511     my $enabled_cap = {
4512         "auto-converge" => 1,
4513         "xbzrle" => 1,
4514         "x-rdma-pin-all" => 0,
4515         "zero-blocks" => 0,
4516         "compress" => 0
4517     };
4518
4519     my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4520
4521     for my $supported_capability (@$supported_capabilities) {
4522         push @$cap_ref, {
4523             capability => $supported_capability->{capability},
4524             state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4525         };
4526     }
4527
4528     vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4529 }
4530
4531 my $fast_plug_option = {
4532     'lock' => 1,
4533     'name' => 1,
4534     'onboot' => 1,
4535     'shares' => 1,
4536     'startup' => 1,
4537     'description' => 1,
4538     'protection' => 1,
4539     'vmstatestorage' => 1,
4540 };
4541
4542 # hotplug changes in [PENDING]
4543 # $selection hash can be used to only apply specified options, for
4544 # example: { cores => 1 } (only apply changed 'cores')
4545 # $errors ref is used to return error messages
4546 sub vmconfig_hotplug_pending {
4547     my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4548
4549     my $defaults = load_defaults();
4550
4551     # commit values which do not have any impact on running VM first
4552     # Note: those option cannot raise errors, we we do not care about
4553     # $selection and always apply them.
4554
4555     my $add_error = sub {
4556         my ($opt, $msg) = @_;
4557         $errors->{$opt} = "hotplug problem - $msg";
4558     };
4559
4560     my $changes = 0;
4561     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4562         if ($fast_plug_option->{$opt}) {
4563             $conf->{$opt} = $conf->{pending}->{$opt};
4564             delete $conf->{pending}->{$opt};
4565             $changes = 1;
4566         }
4567     }
4568
4569     if ($changes) {
4570         PVE::QemuConfig->write_config($vmid, $conf);
4571         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4572     }
4573
4574     my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4575
4576     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4577     while (my ($opt, $force) = each %$pending_delete_hash) {
4578         next if $selection && !$selection->{$opt};
4579         eval {
4580             if ($opt eq 'hotplug') {
4581                 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4582             } elsif ($opt eq 'tablet') {
4583                 die "skip\n" if !$hotplug_features->{usb};
4584                 if ($defaults->{tablet}) {
4585                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4586                 } else {
4587                     vm_deviceunplug($vmid, $conf, $opt);
4588                 }
4589             } elsif ($opt =~ m/^usb\d+/) {
4590                 die "skip\n";
4591                 # since we cannot reliably hot unplug usb devices
4592                 # we are disabling it
4593                 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4594                 vm_deviceunplug($vmid, $conf, $opt);
4595             } elsif ($opt eq 'vcpus') {
4596                 die "skip\n" if !$hotplug_features->{cpu};
4597                 qemu_cpu_hotplug($vmid, $conf, undef);
4598             } elsif ($opt eq 'balloon') {
4599                 # enable balloon device is not hotpluggable
4600                 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4601                 # here we reset the ballooning value to memory
4602                 my $balloon = $conf->{memory} || $defaults->{memory};
4603                 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4604             } elsif ($fast_plug_option->{$opt}) {
4605                 # do nothing
4606             } elsif ($opt =~ m/^net(\d+)$/) {
4607                 die "skip\n" if !$hotplug_features->{network};
4608                 vm_deviceunplug($vmid, $conf, $opt);
4609             } elsif (is_valid_drivename($opt)) {
4610                 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4611                 vm_deviceunplug($vmid, $conf, $opt);
4612                 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4613             } elsif ($opt =~ m/^memory$/) {
4614                 die "skip\n" if !$hotplug_features->{memory};
4615                 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4616             } elsif ($opt eq 'cpuunits') {
4617                 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4618             } elsif ($opt eq 'cpulimit') {
4619                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4620             } else {
4621                 die "skip\n";
4622             }
4623         };
4624         if (my $err = $@) {
4625             &$add_error($opt, $err) if $err ne "skip\n";
4626         } else {
4627             # save new config if hotplug was successful
4628             delete $conf->{$opt};
4629             vmconfig_undelete_pending_option($conf, $opt);
4630             PVE::QemuConfig->write_config($vmid, $conf);
4631             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4632         }
4633     }
4634
4635     my $apply_pending_cloudinit;
4636     $apply_pending_cloudinit = sub {
4637         my ($key, $value) = @_;
4638         $apply_pending_cloudinit = sub {}; # once is enough
4639
4640         my @cloudinit_opts = keys %$confdesc_cloudinit;
4641         foreach my $opt (keys %{$conf->{pending}}) {
4642             next if !grep { $_ eq $opt } @cloudinit_opts;
4643             $conf->{$opt} = delete $conf->{pending}->{$opt};
4644         }
4645
4646         my $new_conf = { %$conf };
4647         $new_conf->{$key} = $value;
4648         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4649     };
4650
4651     foreach my $opt (keys %{$conf->{pending}}) {
4652         next if $selection && !$selection->{$opt};
4653         my $value = $conf->{pending}->{$opt};
4654         eval {
4655             if ($opt eq 'hotplug') {
4656                 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4657             } elsif ($opt eq 'tablet') {
4658                 die "skip\n" if !$hotplug_features->{usb};
4659                 if ($value == 1) {
4660                     vm_deviceplug($storecfg, $conf, $vmid, $opt);
4661                 } elsif ($value == 0) {
4662                     vm_deviceunplug($vmid, $conf, $opt);
4663                 }
4664             } elsif ($opt =~ m/^usb\d+$/) {
4665                 die "skip\n";
4666                 # since we cannot reliably hot unplug usb devices
4667                 # we are disabling it
4668                 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4669                 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4670                 die "skip\n" if !$d;
4671                 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d);
4672             } elsif ($opt eq 'vcpus') {
4673                 die "skip\n" if !$hotplug_features->{cpu};
4674                 qemu_cpu_hotplug($vmid, $conf, $value);
4675             } elsif ($opt eq 'balloon') {
4676                 # enable/disable balloning device is not hotpluggable
4677                 my $old_balloon_enabled =  !!(!defined($conf->{balloon}) || $conf->{balloon});
4678                 my $new_balloon_enabled =  !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4679                 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4680
4681                 # allow manual ballooning if shares is set to zero
4682                 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4683                     my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4684                     vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4685                 }
4686             } elsif ($opt =~ m/^net(\d+)$/) {
4687                 # some changes can be done without hotplug
4688                 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4689                                     $vmid, $opt, $value);
4690             } elsif (is_valid_drivename($opt)) {
4691                 # some changes can be done without hotplug
4692                 my $drive = parse_drive($opt, $value);
4693                 if (drive_is_cloudinit($drive)) {
4694                     &$apply_pending_cloudinit($opt, $value);
4695                 }
4696                 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4697                                      $vmid, $opt, $value, 1);
4698             } elsif ($opt =~ m/^memory$/) { #dimms
4699                 die "skip\n" if !$hotplug_features->{memory};
4700                 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4701             } elsif ($opt eq 'cpuunits') {
4702                 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4703             } elsif ($opt eq 'cpulimit') {
4704                 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4705                 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4706             } else {
4707                 die "skip\n";  # skip non-hot-pluggable options
4708             }
4709         };
4710         if (my $err = $@) {
4711             &$add_error($opt, $err) if $err ne "skip\n";
4712         } else {
4713             # save new config if hotplug was successful
4714             $conf->{$opt} = $value;
4715             delete $conf->{pending}->{$opt};
4716             PVE::QemuConfig->write_config($vmid, $conf);
4717             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4718         }
4719     }
4720 }
4721
4722 sub try_deallocate_drive {
4723     my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4724
4725     if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4726         my $volid = $drive->{file};
4727         if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4728             my $sid = PVE::Storage::parse_volume_id($volid);
4729             $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4730
4731             # check if the disk is really unused
4732             die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4733                 if is_volume_in_use($storecfg, $conf, $key, $volid);
4734             PVE::Storage::vdisk_free($storecfg, $volid);
4735             return 1;
4736         } else {
4737             # If vm is not owner of this disk remove from config
4738             return 1;
4739         }
4740     }
4741
4742     return undef;
4743 }
4744
4745 sub vmconfig_delete_or_detach_drive {
4746     my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4747
4748     my $drive = parse_drive($opt, $conf->{$opt});
4749
4750     my $rpcenv = PVE::RPCEnvironment::get();
4751     my $authuser = $rpcenv->get_user();
4752
4753     if ($force) {
4754         $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4755         try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4756     } else {
4757         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4758     }
4759 }
4760
4761 sub vmconfig_apply_pending {
4762     my ($vmid, $conf, $storecfg) = @_;
4763
4764     # cold plug
4765
4766     my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4767     while (my ($opt, $force) = each %$pending_delete_hash) {
4768         die "internal error" if $opt =~ m/^unused/;
4769         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4770         if (!defined($conf->{$opt})) {
4771             vmconfig_undelete_pending_option($conf, $opt);
4772             PVE::QemuConfig->write_config($vmid, $conf);
4773         } elsif (is_valid_drivename($opt)) {
4774             vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4775             vmconfig_undelete_pending_option($conf, $opt);
4776             delete $conf->{$opt};
4777             PVE::QemuConfig->write_config($vmid, $conf);
4778         } else {
4779             vmconfig_undelete_pending_option($conf, $opt);
4780             delete $conf->{$opt};
4781             PVE::QemuConfig->write_config($vmid, $conf);
4782         }
4783     }
4784
4785     $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4786
4787     foreach my $opt (keys %{$conf->{pending}}) { # add/change
4788         $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4789
4790         if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
4791             # skip if nothing changed
4792         } elsif (is_valid_drivename($opt)) {
4793             vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4794                 if defined($conf->{$opt});
4795             $conf->{$opt} = $conf->{pending}->{$opt};
4796         } else {
4797             $conf->{$opt} = $conf->{pending}->{$opt};
4798         }
4799
4800         delete $conf->{pending}->{$opt};
4801         PVE::QemuConfig->write_config($vmid, $conf);
4802     }
4803 }
4804
4805 my $safe_num_ne = sub {
4806     my ($a, $b) = @_;
4807
4808     return 0 if !defined($a) && !defined($b);
4809     return 1 if !defined($a);
4810     return 1 if !defined($b);
4811
4812     return $a != $b;
4813 };
4814
4815 my $safe_string_ne = sub {
4816     my ($a, $b) = @_;
4817
4818     return 0 if !defined($a) && !defined($b);
4819     return 1 if !defined($a);
4820     return 1 if !defined($b);
4821
4822     return $a ne $b;
4823 };
4824
4825 sub vmconfig_update_net {
4826     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4827
4828     my $newnet = parse_net($value);
4829
4830     if ($conf->{$opt}) {
4831         my $oldnet = parse_net($conf->{$opt});
4832
4833         if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4834             &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4835             &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4836             !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4837
4838             # for non online change, we try to hot-unplug
4839             die "skip\n" if !$hotplug;
4840             vm_deviceunplug($vmid, $conf, $opt);
4841         } else {
4842
4843             die "internal error" if $opt !~ m/net(\d+)/;
4844             my $iface = "tap${vmid}i$1";
4845
4846             if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4847                 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4848                 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4849                 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4850                 PVE::Network::tap_unplug($iface);
4851                 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4852             } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4853                 # Rate can be applied on its own but any change above needs to
4854                 # include the rate in tap_plug since OVS resets everything.
4855                 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4856             }
4857
4858             if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4859                 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4860             }
4861
4862             return 1;
4863         }
4864     }
4865
4866     if ($hotplug) {
4867         vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4868     } else {
4869         die "skip\n";
4870     }
4871 }
4872
4873 sub vmconfig_update_disk {
4874     my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4875
4876     # fixme: do we need force?
4877
4878     my $drive = parse_drive($opt, $value);
4879
4880     if ($conf->{$opt}) {
4881
4882         if (my $old_drive = parse_drive($opt, $conf->{$opt}))  {
4883
4884             my $media = $drive->{media} || 'disk';
4885             my $oldmedia = $old_drive->{media} || 'disk';
4886             die "unable to change media type\n" if $media ne $oldmedia;
4887
4888             if (!drive_is_cdrom($old_drive)) {
4889
4890                 if ($drive->{file} ne $old_drive->{file}) {
4891
4892                     die "skip\n" if !$hotplug;
4893
4894                     # unplug and register as unused
4895                     vm_deviceunplug($vmid, $conf, $opt);
4896                     vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4897
4898                 } else {
4899                     # update existing disk
4900
4901                     # skip non hotpluggable value
4902                     if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
4903                         &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4904                         &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4905                         &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4906                         die "skip\n";
4907                     }
4908
4909                     # apply throttle
4910                     if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4911                         &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4912                         &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4913                         &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4914                         &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4915                         &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4916                         &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4917                         &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4918                         &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4919                         &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4920                         &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4921                         &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
4922                         &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
4923                         &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
4924                         &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
4925                         &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
4926                         &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
4927                         &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
4928
4929                         qemu_block_set_io_throttle($vmid,"drive-$opt",
4930                                                    ($drive->{mbps} || 0)*1024*1024,
4931                                                    ($drive->{mbps_rd} || 0)*1024*1024,
4932                                                    ($drive->{mbps_wr} || 0)*1024*1024,
4933                                                    $drive->{iops} || 0,
4934                                                    $drive->{iops_rd} || 0,
4935                                                    $drive->{iops_wr} || 0,
4936                                                    ($drive->{mbps_max} || 0)*1024*1024,
4937                                                    ($drive->{mbps_rd_max} || 0)*1024*1024,
4938                                                    ($drive->{mbps_wr_max} || 0)*1024*1024,
4939                                                    $drive->{iops_max} || 0,
4940                                                    $drive->{iops_rd_max} || 0,
4941                                                    $drive->{iops_wr_max} || 0,
4942                                                    $drive->{bps_max_length} || 1,
4943                                                    $drive->{bps_rd_max_length} || 1,
4944                                                    $drive->{bps_wr_max_length} || 1,
4945                                                    $drive->{iops_max_length} || 1,
4946                                                    $drive->{iops_rd_max_length} || 1,
4947                                                    $drive->{iops_wr_max_length} || 1);
4948
4949                     }
4950
4951                     return 1;
4952                 }
4953
4954             } else { # cdrom
4955
4956                 if ($drive->{file} eq 'none') {
4957                     vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4958                     if (drive_is_cloudinit($old_drive)) {
4959                         vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
4960                     }
4961                 } else {
4962                     my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4963                     vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4964                     vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4965                 }
4966
4967                 return 1;
4968             }
4969         }
4970     }
4971
4972     die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4973     # hotplug new disks
4974     PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4975     vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4976 }
4977
4978 sub vm_start {
4979     my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
4980         $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
4981
4982     PVE::QemuConfig->lock_config($vmid, sub {
4983         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
4984
4985         die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4986
4987         PVE::QemuConfig->check_lock($conf) if !$skiplock;
4988
4989         die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4990
4991         if (!$statefile && scalar(keys %{$conf->{pending}})) {
4992             vmconfig_apply_pending($vmid, $conf, $storecfg);
4993             $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4994         }
4995
4996         PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
4997
4998         my $defaults = load_defaults();
4999
5000         # set environment variable useful inside network script
5001         $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
5002
5003         my $local_volumes = {};
5004
5005         if ($targetstorage) {
5006             foreach_drive($conf, sub {
5007                 my ($ds, $drive) = @_;
5008
5009                 return if drive_is_cdrom($drive);
5010
5011                 my $volid = $drive->{file};
5012
5013                 return if !$volid;
5014
5015                 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5016
5017                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5018                 return if $scfg->{shared};
5019                 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5020             });
5021
5022             my $format = undef;
5023
5024             foreach my $opt (sort keys %$local_volumes) {
5025
5026                 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5027                 my $drive = parse_drive($opt, $conf->{$opt});
5028
5029                 #if remote storage is specified, use default format
5030                 if ($targetstorage && $targetstorage ne "1") {
5031                     $storeid = $targetstorage;
5032                     my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5033                     $format = $defFormat;
5034                 } else {
5035                     #else we use same format than original
5036                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5037                     $format = qemu_img_format($scfg, $volid);
5038                 }
5039
5040                 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5041                 my $newdrive = $drive;
5042                 $newdrive->{format} = $format;
5043                 $newdrive->{file} = $newvolid;
5044                 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5045                 $local_volumes->{$opt} = $drivestr;
5046                 #pass drive to conf for command line
5047                 $conf->{$opt} = $drivestr;
5048             }
5049         }
5050
5051         my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5052
5053         my $migrate_port = 0;
5054         my $migrate_uri;
5055         if ($statefile) {
5056             if ($statefile eq 'tcp') {
5057                 my $localip = "localhost";
5058                 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5059                 my $nodename = PVE::INotify::nodename();
5060
5061                 if (!defined($migration_type)) {
5062                     if (defined($datacenterconf->{migration}->{type})) {
5063                         $migration_type = $datacenterconf->{migration}->{type};
5064                     } else {
5065                         $migration_type = 'secure';
5066                     }
5067                 }
5068
5069                 if ($migration_type eq 'insecure') {
5070                     my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5071                     if ($migrate_network_addr) {
5072                         $localip = $migrate_network_addr;
5073                     } else {
5074                         $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5075                     }
5076
5077                     $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5078                 }
5079
5080                 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5081                 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5082                 $migrate_uri = "tcp:${localip}:${migrate_port}";
5083                 push @$cmd, '-incoming', $migrate_uri;
5084                 push @$cmd, '-S';
5085
5086             } elsif ($statefile eq 'unix') {
5087                 # should be default for secure migrations as a ssh TCP forward
5088                 # tunnel is not deterministic reliable ready and fails regurarly
5089                 # to set up in time, so use UNIX socket forwards
5090                 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5091                 unlink $socket_addr;
5092
5093                 $migrate_uri = "unix:$socket_addr";
5094
5095                 push @$cmd, '-incoming', $migrate_uri;
5096                 push @$cmd, '-S';
5097
5098             } else {
5099                 push @$cmd, '-loadstate', $statefile;
5100             }
5101         } elsif ($paused) {
5102             push @$cmd, '-S';
5103         }
5104
5105         # host pci devices
5106         for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++)  {
5107           my $d = parse_hostpci($conf->{"hostpci$i"});
5108           next if !$d;
5109           my $pcidevices = $d->{pciid};
5110           foreach my $pcidevice (@$pcidevices) {
5111                 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
5112
5113                 my $info = pci_device_info("0000:$pciid");
5114                 die "IOMMU not present\n" if !check_iommu_support();
5115                 die "no pci device info for device '$pciid'\n" if !$info;
5116                 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
5117                 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
5118           }
5119         }
5120
5121         PVE::Storage::activate_volumes($storecfg, $vollist);
5122
5123         if (!check_running($vmid, 1)) {
5124             eval {
5125                 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5126                     outfunc => sub {}, errfunc => sub {});
5127             };
5128         }
5129
5130         my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5131                                                   : $defaults->{cpuunits};
5132
5133         my $start_timeout = $conf->{hugepages} ? 300 : 30;
5134         my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5135
5136         my %properties = (
5137             Slice => 'qemu.slice',
5138             KillMode => 'none',
5139             CPUShares => $cpuunits
5140         );
5141
5142         if (my $cpulimit = $conf->{cpulimit}) {
5143             $properties{CPUQuota} = int($cpulimit * 100);
5144         }
5145         $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5146
5147         my $run_qemu = sub {
5148             PVE::Tools::run_fork sub {
5149                 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5150                 run_command($cmd, %run_params);
5151             };
5152         };
5153
5154         if ($conf->{hugepages}) {
5155
5156             my $code = sub {
5157                 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5158                 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5159
5160                 PVE::QemuServer::Memory::hugepages_mount();
5161                 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5162
5163                 eval { $run_qemu->() };
5164                 if (my $err = $@) {
5165                     PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5166                     die $err;
5167                 }
5168
5169                 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5170             };
5171             eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5172
5173         } else {
5174             eval { $run_qemu->() };
5175         }
5176
5177         if (my $err = $@) {
5178             # deactivate volumes if start fails
5179             eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5180             die "start failed: $err";
5181         }
5182
5183         print "migration listens on $migrate_uri\n" if $migrate_uri;
5184
5185         if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix')  {
5186             eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5187             warn $@ if $@;
5188         }
5189
5190         #start nbd server for storage migration
5191         if ($targetstorage) {
5192             my $nodename = PVE::INotify::nodename();
5193             my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5194             my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5195             my $pfamily = PVE::Tools::get_host_address_family($nodename);
5196             $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5197
5198             vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${migrate_port}" } } );
5199
5200             $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5201
5202             foreach my $opt (sort keys %$local_volumes) {
5203                 my $volid = $local_volumes->{$opt};
5204                 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5205                 my $migrate_storage_uri = "nbd:${localip}:${migrate_port}:exportname=drive-$opt";
5206                 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5207             }
5208         }
5209
5210         if ($migratedfrom) {
5211             eval {
5212                 set_migration_caps($vmid);
5213             };
5214             warn $@ if $@;
5215
5216             if ($spice_port) {
5217                 print "spice listens on port $spice_port\n";
5218                 if ($spice_ticket) {
5219                     vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5220                     vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5221                 }
5222             }
5223
5224         } else {
5225             vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5226                 if !$statefile && $conf->{balloon};
5227
5228             foreach my $opt (keys %$conf) {
5229                 next if $opt !~  m/^net\d+$/;
5230                 my $nicconf = parse_net($conf->{$opt});
5231                 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5232             }
5233         }
5234
5235         vm_mon_cmd_nocheck($vmid, 'qom-set',
5236                     path => "machine/peripheral/balloon0",
5237                     property => "guest-stats-polling-interval",
5238                     value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5239
5240     });
5241 }
5242
5243 sub vm_mon_cmd {
5244     my ($vmid, $execute, %params) = @_;
5245
5246     my $cmd = { execute => $execute, arguments => \%params };
5247     vm_qmp_command($vmid, $cmd);
5248 }
5249
5250 sub vm_mon_cmd_nocheck {
5251     my ($vmid, $execute, %params) = @_;
5252
5253     my $cmd = { execute => $execute, arguments => \%params };
5254     vm_qmp_command($vmid, $cmd, 1);
5255 }
5256
5257 sub vm_qmp_command {
5258     my ($vmid, $cmd, $nocheck) = @_;
5259
5260     my $res;
5261
5262     my $timeout;
5263     if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
5264         $timeout = $cmd->{arguments}->{timeout};
5265         delete $cmd->{arguments}->{timeout};
5266     }
5267
5268     eval {
5269         die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5270         my $sname = qmp_socket($vmid);
5271         if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5272             my $qmpclient = PVE::QMPClient->new();
5273
5274             $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5275         } else {
5276             die "unable to open monitor socket\n";
5277         }
5278     };
5279     if (my $err = $@) {
5280         syslog("err", "VM $vmid qmp command failed - $err");
5281         die $err;
5282     }
5283
5284     return $res;
5285 }
5286
5287 sub vm_human_monitor_command {
5288     my ($vmid, $cmdline) = @_;
5289
5290     my $res;
5291
5292     my $cmd = {
5293         execute => 'human-monitor-command',
5294         arguments => { 'command-line' => $cmdline},
5295     };
5296
5297     return vm_qmp_command($vmid, $cmd);
5298 }
5299
5300 sub vm_commandline {
5301     my ($storecfg, $vmid) = @_;
5302
5303     my $conf = PVE::QemuConfig->load_config($vmid);
5304
5305     my $defaults = load_defaults();
5306
5307     my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5308
5309     return PVE::Tools::cmd2string($cmd);
5310 }
5311
5312 sub vm_reset {
5313     my ($vmid, $skiplock) = @_;
5314
5315     PVE::QemuConfig->lock_config($vmid, sub {
5316
5317         my $conf = PVE::QemuConfig->load_config($vmid);
5318
5319         PVE::QemuConfig->check_lock($conf) if !$skiplock;
5320
5321         vm_mon_cmd($vmid, "system_reset");
5322     });
5323 }
5324
5325 sub get_vm_volumes {
5326     my ($conf) = @_;
5327
5328     my $vollist = [];
5329     foreach_volid($conf, sub {
5330         my ($volid, $attr) = @_;
5331
5332         return if $volid =~ m|^/|;
5333
5334         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5335         return if !$sid;
5336
5337         push @$vollist, $volid;
5338     });
5339
5340     return $vollist;
5341 }
5342
5343 sub vm_stop_cleanup {
5344     my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5345
5346     eval {
5347
5348         if (!$keepActive) {
5349             my $vollist = get_vm_volumes($conf);
5350             PVE::Storage::deactivate_volumes($storecfg, $vollist);
5351         }
5352
5353         foreach my $ext (qw(mon qmp pid vnc qga)) {
5354             unlink "/var/run/qemu-server/${vmid}.$ext";
5355         }
5356
5357         vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5358     };
5359     warn $@ if $@; # avoid errors - just warn
5360 }
5361
5362 # Note: use $nockeck to skip tests if VM configuration file exists.
5363 # We need that when migration VMs to other nodes (files already moved)
5364 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5365 sub vm_stop {
5366     my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5367
5368     $force = 1 if !defined($force) && !$shutdown;
5369
5370     if ($migratedfrom){
5371         my $pid = check_running($vmid, $nocheck, $migratedfrom);
5372         kill 15, $pid if $pid;
5373         my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5374         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5375         return;
5376     }
5377
5378     PVE::QemuConfig->lock_config($vmid, sub {
5379
5380         my $pid = check_running($vmid, $nocheck);
5381         return if !$pid;
5382
5383         my $conf;
5384         if (!$nocheck) {
5385             $conf = PVE::QemuConfig->load_config($vmid);
5386             PVE::QemuConfig->check_lock($conf) if !$skiplock;
5387             if (!defined($timeout) && $shutdown && $conf->{startup}) {
5388                 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5389                 $timeout = $opts->{down} if $opts->{down};
5390             }
5391         }
5392
5393         $timeout = 60 if !defined($timeout);
5394
5395         eval {
5396             if ($shutdown) {
5397                 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5398                     vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
5399                 } else {
5400                     vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5401                 }
5402             } else {
5403                 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5404             }
5405         };
5406         my $err = $@;
5407
5408         if (!$err) {
5409             my $count = 0;
5410             while (($count < $timeout) && check_running($vmid, $nocheck)) {
5411                 $count++;
5412                 sleep 1;
5413             }
5414
5415             if ($count >= $timeout) {
5416                 if ($force) {
5417                     warn "VM still running - terminating now with SIGTERM\n";
5418                     kill 15, $pid;
5419                 } else {
5420                     die "VM quit/powerdown failed - got timeout\n";
5421                 }
5422             } else {
5423                 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5424                 return;
5425             }
5426         } else {
5427             if ($force) {
5428                 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5429                 kill 15, $pid;
5430             } else {
5431                 die "VM quit/powerdown failed\n";
5432             }
5433         }
5434
5435         # wait again
5436         $timeout = 10;
5437
5438         my $count = 0;
5439         while (($count < $timeout) && check_running($vmid, $nocheck)) {
5440             $count++;
5441             sleep 1;
5442         }
5443
5444         if ($count >= $timeout) {
5445             warn "VM still running - terminating now with SIGKILL\n";
5446             kill 9, $pid;
5447             sleep 1;
5448         }
5449
5450         vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5451    });
5452 }
5453
5454 sub vm_suspend {
5455     my ($vmid, $skiplock) = @_;
5456
5457     PVE::QemuConfig->lock_config($vmid, sub {
5458
5459         my $conf = PVE::QemuConfig->load_config($vmid);
5460
5461         PVE::QemuConfig->check_lock($conf)
5462             if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5463
5464         vm_mon_cmd($vmid, "stop");
5465     });
5466 }
5467
5468 sub vm_resume {
5469     my ($vmid, $skiplock, $nocheck) = @_;
5470
5471     PVE::QemuConfig->lock_config($vmid, sub {
5472
5473         my $res = vm_mon_cmd($vmid, 'query-status');
5474         my $resume_cmd = 'cont';
5475
5476         if ($res->{status} && $res->{status} eq 'suspended') {
5477             $resume_cmd = 'system_wakeup';
5478         }
5479
5480         if (!$nocheck) {
5481
5482             my $conf = PVE::QemuConfig->load_config($vmid);
5483
5484             PVE::QemuConfig->check_lock($conf)
5485                 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5486
5487             vm_mon_cmd($vmid, $resume_cmd);
5488
5489         } else {
5490             vm_mon_cmd_nocheck($vmid, $resume_cmd);
5491         }
5492     });
5493 }
5494
5495 sub vm_sendkey {
5496     my ($vmid, $skiplock, $key) = @_;
5497
5498     PVE::QemuConfig->lock_config($vmid, sub {
5499
5500         my $conf = PVE::QemuConfig->load_config($vmid);
5501
5502         # there is no qmp command, so we use the human monitor command
5503         vm_human_monitor_command($vmid, "sendkey $key");
5504     });
5505 }
5506
5507 sub vm_destroy {
5508     my ($storecfg, $vmid, $skiplock) = @_;
5509
5510     PVE::QemuConfig->lock_config($vmid, sub {
5511
5512         my $conf = PVE::QemuConfig->load_config($vmid);
5513
5514         if (!check_running($vmid)) {
5515             destroy_vm($storecfg, $vmid, undef, $skiplock);
5516         } else {
5517             die "VM $vmid is running - destroy failed\n";
5518         }
5519     });
5520 }
5521
5522 # pci helpers
5523
5524 sub file_write {
5525     my ($filename, $buf) = @_;
5526
5527     my $fh = IO::File->new($filename, "w");
5528     return undef if !$fh;
5529
5530     my $res = print $fh $buf;
5531
5532     $fh->close();
5533
5534     return $res;
5535 }
5536
5537 sub pci_device_info {
5538     my ($name) = @_;
5539
5540     my $res;
5541
5542     return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
5543     my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
5544
5545     my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
5546     return undef if !defined($irq) || $irq !~ m/^\d+$/;
5547
5548     my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
5549     return undef if !defined($vendor) || $vendor !~ s/^0x//;
5550
5551     my $product = file_read_firstline("$pcisysfs/devices/$name/device");
5552     return undef if !defined($product) || $product !~ s/^0x//;
5553
5554     $res = {
5555         name => $name,
5556         vendor => $vendor,
5557         product => $product,
5558         domain => $domain,
5559         bus => $bus,
5560         slot => $slot,
5561         func => $func,
5562         irq => $irq,
5563         has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
5564     };
5565
5566     return $res;
5567 }
5568
5569 sub pci_dev_reset {
5570     my ($dev) = @_;
5571
5572     my $name = $dev->{name};
5573
5574     my $fn = "$pcisysfs/devices/$name/reset";
5575
5576     return file_write($fn, "1");
5577 }
5578
5579 sub pci_dev_bind_to_vfio {
5580     my ($dev) = @_;
5581
5582     my $name = $dev->{name};
5583
5584     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5585
5586     if (!-d $vfio_basedir) {
5587         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5588     }
5589     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5590
5591     my $testdir = "$vfio_basedir/$name";
5592     return 1 if -d $testdir;
5593
5594     my $data = "$dev->{vendor} $dev->{product}";
5595     return undef if !file_write("$vfio_basedir/new_id", $data);
5596
5597     my $fn = "$pcisysfs/devices/$name/driver/unbind";
5598     if (!file_write($fn, $name)) {
5599         return undef if -f $fn;
5600     }
5601
5602     $fn = "$vfio_basedir/bind";
5603     if (! -d $testdir) {
5604         return undef if !file_write($fn, $name);
5605     }
5606
5607     return -d $testdir;
5608 }
5609
5610 sub pci_dev_group_bind_to_vfio {
5611     my ($pciid) = @_;
5612
5613     my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
5614
5615     if (!-d $vfio_basedir) {
5616         system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
5617     }
5618     die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
5619
5620     # get IOMMU group devices
5621     opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
5622       my @devs = grep /^0000:/, readdir($D);
5623     closedir($D);
5624
5625     foreach my $pciid (@devs) {
5626         $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
5627
5628         # pci bridges, switches or root ports are not supported
5629         # they have a pci_bus subdirectory so skip them
5630         next if (-e "$pcisysfs/devices/$pciid/pci_bus");
5631
5632         my $info = pci_device_info($1);
5633         pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
5634     }
5635
5636     return 1;
5637 }
5638
5639 # vzdump restore implementaion
5640
5641 sub tar_archive_read_firstfile {
5642     my $archive = shift;
5643
5644     die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5645
5646     # try to detect archive type first
5647     my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5648         die "unable to open file '$archive'\n";
5649     my $firstfile = <$fh>;
5650     kill 15, $pid;
5651     close $fh;
5652
5653     die "ERROR: archive contaions no data\n" if !$firstfile;
5654     chomp $firstfile;
5655
5656     return $firstfile;
5657 }
5658
5659 sub tar_restore_cleanup {
5660     my ($storecfg, $statfile) = @_;
5661
5662     print STDERR "starting cleanup\n";
5663
5664     if (my $fd = IO::File->new($statfile, "r")) {
5665         while (defined(my $line = <$fd>)) {
5666             if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5667                 my $volid = $2;
5668                 eval {
5669                     if ($volid =~ m|^/|) {
5670                         unlink $volid || die 'unlink failed\n';
5671                     } else {
5672                         PVE::Storage::vdisk_free($storecfg, $volid);
5673                     }
5674                     print STDERR "temporary volume '$volid' sucessfuly removed\n";
5675                 };
5676                 print STDERR "unable to cleanup '$volid' - $@" if $@;
5677             } else {
5678                 print STDERR "unable to parse line in statfile - $line";
5679             }
5680         }
5681         $fd->close();
5682     }
5683 }
5684
5685 sub restore_archive {
5686     my ($archive, $vmid, $user, $opts) = @_;
5687
5688     my $format = $opts->{format};
5689     my $comp;
5690
5691     if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5692         $format = 'tar' if !$format;
5693         $comp = 'gzip';
5694     } elsif ($archive =~ m/\.tar$/) {
5695         $format = 'tar' if !$format;
5696     } elsif ($archive =~ m/.tar.lzo$/) {
5697         $format = 'tar' if !$format;
5698         $comp = 'lzop';
5699     } elsif ($archive =~ m/\.vma$/) {
5700         $format = 'vma' if !$format;
5701     } elsif ($archive =~ m/\.vma\.gz$/) {
5702         $format = 'vma' if !$format;
5703         $comp = 'gzip';
5704     } elsif ($archive =~ m/\.vma\.lzo$/) {
5705         $format = 'vma' if !$format;
5706         $comp = 'lzop';
5707     } else {
5708         $format = 'vma' if !$format; # default
5709     }
5710
5711     # try to detect archive format
5712     if ($format eq 'tar') {
5713         return restore_tar_archive($archive, $vmid, $user, $opts);
5714     } else {
5715         return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5716     }
5717 }
5718
5719 sub restore_update_config_line {
5720     my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5721
5722     return if $line =~ m/^\#qmdump\#/;
5723     return if $line =~ m/^\#vzdump\#/;
5724     return if $line =~ m/^lock:/;
5725     return if $line =~ m/^unused\d+:/;
5726     return if $line =~ m/^parent:/;
5727     return if $line =~ m/^template:/; # restored VM is never a template
5728
5729     my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
5730     if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5731         # try to convert old 1.X settings
5732         my ($id, $ind, $ethcfg) = ($1, $2, $3);
5733         foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5734             my ($model, $macaddr) = split(/\=/, $devconfig);
5735             $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
5736             my $net = {
5737                 model => $model,
5738                 bridge => "vmbr$ind",
5739                 macaddr => $macaddr,
5740             };
5741             my $netstr = print_net($net);
5742
5743             print $outfd "net$cookie->{netcount}: $netstr\n";
5744             $cookie->{netcount}++;
5745         }
5746     } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5747         my ($id, $netstr) = ($1, $2);
5748         my $net = parse_net($netstr);
5749         $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
5750         $netstr = print_net($net);
5751         print $outfd "$id: $netstr\n";
5752     } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
5753         my $virtdev = $1;
5754         my $value = $3;
5755         my $di = parse_drive($virtdev, $value);
5756         if (defined($di->{backup}) && !$di->{backup}) {
5757             print $outfd "#$line";
5758         } elsif ($map->{$virtdev}) {
5759             delete $di->{format}; # format can change on restore
5760             $di->{file} = $map->{$virtdev};
5761             $value = print_drive($vmid, $di);
5762             print $outfd "$virtdev: $value\n";
5763         } else {
5764             print $outfd $line;
5765         }
5766     } elsif (($line =~ m/^vmgenid: (.*)/)) {
5767         my $vmgenid = $1;
5768         if ($vmgenid ne '0') {
5769             # always generate a new vmgenid if there was a valid one setup
5770             $vmgenid = generate_uuid();
5771         }
5772         print $outfd "vmgenid: $vmgenid\n";
5773     } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
5774         my ($uuid, $uuid_str);
5775         UUID::generate($uuid);
5776         UUID::unparse($uuid, $uuid_str);
5777         my $smbios1 = parse_smbios1($2);
5778         $smbios1->{uuid} = $uuid_str;
5779         print $outfd $1.print_smbios1($smbios1)."\n";
5780     } else {
5781         print $outfd $line;
5782     }
5783 }
5784
5785 sub scan_volids {
5786     my ($cfg, $vmid) = @_;
5787
5788     my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5789
5790     my $volid_hash = {};
5791     foreach my $storeid (keys %$info) {
5792         foreach my $item (@{$info->{$storeid}}) {
5793             next if !($item->{volid} && $item->{size});
5794             $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5795             $volid_hash->{$item->{volid}} = $item;
5796         }
5797     }
5798
5799     return $volid_hash;
5800 }
5801
5802 sub is_volume_in_use {
5803     my ($storecfg, $conf, $skip_drive, $volid) = @_;
5804
5805     my $path = PVE::Storage::path($storecfg, $volid);
5806
5807     my $scan_config = sub {
5808         my ($cref, $snapname) = @_;
5809
5810         foreach my $key (keys %$cref) {
5811             my $value = $cref->{$key};
5812             if (is_valid_drivename($key)) {
5813                 next if $skip_drive && $key eq $skip_drive;
5814                 my $drive = parse_drive($key, $value);
5815                 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
5816                 return 1 if $volid eq $drive->{file};
5817                 if ($drive->{file} =~ m!^/!) {
5818                     return 1 if $drive->{file} eq $path;
5819                 } else {
5820                     my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
5821                     next if !$storeid;
5822                     my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
5823                     next if !$scfg;
5824                     return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
5825                 }
5826             }
5827         }
5828
5829         return 0;
5830     };
5831
5832     return 1 if &$scan_config($conf);
5833
5834     undef $skip_drive;
5835
5836     foreach my $snapname (keys %{$conf->{snapshots}}) {
5837         return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
5838     }
5839
5840     return 0;
5841 }
5842
5843 sub update_disksize {
5844     my ($vmid, $conf, $volid_hash) = @_;
5845
5846     my $changes;
5847     my $prefix = "VM $vmid:";
5848
5849     # used and unused disks
5850     my $referenced = {};
5851
5852     # Note: it is allowed to define multiple storages with same path (alias), so
5853     # we need to check both 'volid' and real 'path' (two different volid can point
5854     # to the same path).
5855
5856     my $referencedpath = {};
5857
5858     # update size info
5859     foreach my $opt (keys %$conf) {
5860         if (is_valid_drivename($opt)) {
5861             my $drive = parse_drive($opt, $conf->{$opt});
5862             my $volid = $drive->{file};
5863             next if !$volid;
5864
5865             $referenced->{$volid} = 1;
5866             if ($volid_hash->{$volid} &&
5867                 (my $path = $volid_hash->{$volid}->{path})) {
5868                 $referencedpath->{$path} = 1;
5869             }
5870
5871             next if drive_is_cdrom($drive);
5872             next if !$volid_hash->{$volid};
5873
5874             $drive->{size} = $volid_hash->{$volid}->{size};
5875             my $new = print_drive($vmid, $drive);
5876             if ($new ne $conf->{$opt}) {
5877                 $changes = 1;
5878                 $conf->{$opt} = $new;
5879                 print "$prefix update disk '$opt' information.\n";
5880             }
5881         }
5882     }
5883
5884     # remove 'unusedX' entry if volume is used
5885     foreach my $opt (keys %$conf) {
5886         next if $opt !~ m/^unused\d+$/;
5887         my $volid = $conf->{$opt};
5888         my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5889         if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
5890             print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
5891             $changes = 1;
5892             delete $conf->{$opt};
5893         }
5894
5895         $referenced->{$volid} = 1;
5896         $referencedpath->{$path} = 1 if $path;
5897     }
5898
5899     foreach my $volid (sort keys %$volid_hash) {
5900         next if $volid =~ m/vm-$vmid-state-/;
5901         next if $referenced->{$volid};
5902         my $path = $volid_hash->{$volid}->{path};
5903         next if !$path; # just to be sure
5904         next if $referencedpath->{$path};
5905         $changes = 1;
5906         my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
5907         print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
5908         $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
5909     }
5910
5911     return $changes;
5912 }
5913
5914 sub rescan {
5915     my ($vmid, $nolock, $dryrun) = @_;
5916
5917     my $cfg = PVE::Storage::config();
5918
5919     # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
5920     # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
5921     foreach my $stor (keys %{$cfg->{ids}}) {
5922         delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
5923     }
5924
5925     print "rescan volumes...\n";
5926     my $volid_hash = scan_volids($cfg, $vmid);
5927
5928     my $updatefn =  sub {
5929         my ($vmid) = @_;
5930
5931         my $conf = PVE::QemuConfig->load_config($vmid);
5932
5933         PVE::QemuConfig->check_lock($conf);
5934
5935         my $vm_volids = {};
5936         foreach my $volid (keys %$volid_hash) {
5937             my $info = $volid_hash->{$volid};
5938             $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5939         }
5940
5941         my $changes = update_disksize($vmid, $conf, $vm_volids);
5942
5943         PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
5944     };
5945
5946     if (defined($vmid)) {
5947         if ($nolock) {
5948             &$updatefn($vmid);
5949         } else {
5950             PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5951         }
5952     } else {
5953         my $vmlist = config_list();
5954         foreach my $vmid (keys %$vmlist) {
5955             if ($nolock) {
5956                 &$updatefn($vmid);
5957             } else {
5958                 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5959             }
5960         }
5961     }
5962 }
5963
5964 sub restore_vma_archive {
5965     my ($archive, $vmid, $user, $opts, $comp) = @_;
5966
5967     my $readfrom = $archive;
5968
5969     my $cfg = PVE::Storage::config();
5970     my $commands = [];
5971     my $bwlimit = $opts->{bwlimit};
5972
5973     my $dbg_cmdstring = '';
5974     my $add_pipe = sub {
5975         my ($cmd) = @_;
5976         push @$commands, $cmd;
5977         $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
5978         $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
5979         $readfrom = '-';
5980     };
5981
5982     my $input = undef;
5983     if ($archive eq '-') {
5984         $input = '<&STDIN';
5985     } else {
5986         # If we use a backup from a PVE defined storage we also consider that
5987         # storage's rate limit:
5988         my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
5989         if (defined($volid)) {
5990             my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
5991             my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
5992             if ($readlimit) {
5993                 print STDERR "applying read rate limit: $readlimit\n";
5994                 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
5995                 $add_pipe->($cstream);
5996             }
5997         }
5998     }
5999
6000     if ($comp) {
6001         my $cmd;
6002         if ($comp eq 'gzip') {
6003             $cmd = ['zcat', $readfrom];
6004         } elsif ($comp eq 'lzop') {
6005             $cmd = ['lzop', '-d', '-c', $readfrom];
6006         } else {
6007             die "unknown compression method '$comp'\n";
6008         }
6009         $add_pipe->($cmd);
6010     }
6011
6012     my $tmpdir = "/var/tmp/vzdumptmp$$";
6013     rmtree $tmpdir;
6014
6015     # disable interrupts (always do cleanups)
6016     local $SIG{INT} =
6017         local $SIG{TERM} =
6018         local $SIG{QUIT} =
6019         local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6020
6021     my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6022     POSIX::mkfifo($mapfifo, 0600);
6023     my $fifofh;
6024
6025     my $openfifo = sub {
6026         open($fifofh, '>', $mapfifo) || die $!;
6027     };
6028
6029     $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6030
6031     my $oldtimeout;
6032     my $timeout = 5;
6033
6034     my $devinfo = {};
6035
6036     my $rpcenv = PVE::RPCEnvironment::get();
6037
6038     my $conffile = PVE::QemuConfig->config_file($vmid);
6039     my $tmpfn = "$conffile.$$.tmp";
6040
6041     # Note: $oldconf is undef if VM does not exists
6042     my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6043     my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6044
6045     my %storage_limits;
6046
6047     my $print_devmap = sub {
6048         my $virtdev_hash = {};
6049
6050         my $cfgfn = "$tmpdir/qemu-server.conf";
6051
6052         # we can read the config - that is already extracted
6053         my $fh = IO::File->new($cfgfn, "r") ||
6054             "unable to read qemu-server.conf - $!\n";
6055
6056         my $fwcfgfn = "$tmpdir/qemu-server.fw";
6057         if (-f $fwcfgfn) {
6058             my $pve_firewall_dir = '/etc/pve/firewall';
6059             mkdir $pve_firewall_dir; # make sure the dir exists
6060             PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6061         }
6062
6063         while (defined(my $line = <$fh>)) {
6064             if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6065                 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6066                 die "archive does not contain data for drive '$virtdev'\n"
6067                     if !$devinfo->{$devname};
6068                 if (defined($opts->{storage})) {
6069                     $storeid = $opts->{storage} || 'local';
6070                 } elsif (!$storeid) {
6071                     $storeid = 'local';
6072                 }
6073                 $format = 'raw' if !$format;
6074                 $devinfo->{$devname}->{devname} = $devname;
6075                 $devinfo->{$devname}->{virtdev} = $virtdev;
6076                 $devinfo->{$devname}->{format} = $format;
6077                 $devinfo->{$devname}->{storeid} = $storeid;
6078
6079                 # check permission on storage
6080                 my $pool = $opts->{pool}; # todo: do we need that?
6081                 if ($user ne 'root@pam') {
6082                     $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6083                 }
6084
6085                 $storage_limits{$storeid} = $bwlimit;
6086
6087                 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6088             }
6089         }
6090
6091         foreach my $key (keys %storage_limits) {
6092             my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6093             next if !$limit;
6094             print STDERR "rate limit for storage $key: $limit KiB/s\n";
6095             $storage_limits{$key} = $limit * 1024;
6096         }
6097
6098         foreach my $devname (keys %$devinfo) {
6099             die "found no device mapping information for device '$devname'\n"
6100                 if !$devinfo->{$devname}->{virtdev};
6101         }
6102
6103         # create empty/temp config
6104         if ($oldconf) {
6105             PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6106             foreach_drive($oldconf, sub {
6107                 my ($ds, $drive) = @_;
6108
6109                 return if drive_is_cdrom($drive);
6110
6111                 my $volid = $drive->{file};
6112
6113                 return if !$volid || $volid =~ m|^/|;
6114
6115                 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6116                 return if !$path || !$owner || ($owner != $vmid);
6117
6118                 # Note: only delete disk we want to restore
6119                 # other volumes will become unused
6120                 if ($virtdev_hash->{$ds}) {
6121                     eval { PVE::Storage::vdisk_free($cfg, $volid); };
6122                     if (my $err = $@) {
6123                         warn $err;
6124                     }
6125                 }
6126             });
6127
6128             # delete vmstate files
6129             # since after the restore we have no snapshots anymore
6130             foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6131                 my $snap = $oldconf->{snapshots}->{$snapname};
6132                 if ($snap->{vmstate}) {
6133                     eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6134                     if (my $err = $@) {
6135                         warn $err;
6136                     }
6137                 }
6138             }
6139         }
6140
6141         my $map = {};
6142         foreach my $virtdev (sort keys %$virtdev_hash) {
6143             my $d = $virtdev_hash->{$virtdev};
6144             my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6145             my $storeid = $d->{storeid};
6146             my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6147
6148             my $map_opts = '';
6149             if (my $limit = $storage_limits{$storeid}) {
6150                 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6151             }
6152
6153             # test if requested format is supported
6154             my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6155             my $supported = grep { $_ eq $d->{format} } @$validFormats;
6156             $d->{format} = $defFormat if !$supported;
6157
6158             my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid,
6159                                                   $d->{format}, undef, $alloc_size);
6160             print STDERR "new volume ID is '$volid'\n";
6161             $d->{volid} = $volid;
6162             my $path = PVE::Storage::path($cfg, $volid);
6163
6164             PVE::Storage::activate_volumes($cfg,[$volid]);
6165
6166             my $write_zeros = 1;
6167             if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6168                 $write_zeros = 0;
6169             }
6170
6171             print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6172
6173             print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6174             $map->{$virtdev} = $volid;
6175         }
6176
6177         $fh->seek(0, 0) || die "seek failed - $!\n";
6178
6179         my $outfd = new IO::File ($tmpfn, "w") ||
6180             die "unable to write config for VM $vmid\n";
6181
6182         my $cookie = { netcount => 0 };
6183         while (defined(my $line = <$fh>)) {
6184             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6185         }
6186
6187         $fh->close();
6188         $outfd->close();
6189     };
6190
6191     eval {
6192         # enable interrupts
6193         local $SIG{INT} =
6194             local $SIG{TERM} =
6195             local $SIG{QUIT} =
6196             local $SIG{HUP} =
6197             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6198         local $SIG{ALRM} = sub { die "got timeout\n"; };
6199
6200         $oldtimeout = alarm($timeout);
6201
6202         my $parser = sub {
6203             my $line = shift;
6204
6205             print "$line\n";
6206
6207             if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6208                 my ($dev_id, $size, $devname) = ($1, $2, $3);
6209                 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6210             } elsif ($line =~ m/^CTIME: /) {
6211                 # we correctly received the vma config, so we can disable
6212                 # the timeout now for disk allocation (set to 10 minutes, so
6213                 # that we always timeout if something goes wrong)
6214                 alarm(600);
6215                 &$print_devmap();
6216                 print $fifofh "done\n";
6217                 my $tmp = $oldtimeout || 0;
6218                 $oldtimeout = undef;
6219                 alarm($tmp);
6220                 close($fifofh);
6221             }
6222         };
6223
6224         print "restore vma archive: $dbg_cmdstring\n";
6225         run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6226     };
6227     my $err = $@;
6228
6229     alarm($oldtimeout) if $oldtimeout;
6230
6231     my $vollist = [];
6232     foreach my $devname (keys %$devinfo) {
6233         my $volid = $devinfo->{$devname}->{volid};
6234         push @$vollist, $volid if $volid;
6235     }
6236
6237     PVE::Storage::deactivate_volumes($cfg, $vollist);
6238
6239     unlink $mapfifo;
6240
6241     if ($err) {
6242         rmtree $tmpdir;
6243         unlink $tmpfn;
6244
6245         foreach my $devname (keys %$devinfo) {
6246             my $volid = $devinfo->{$devname}->{volid};
6247             next if !$volid;
6248             eval {
6249                 if ($volid =~ m|^/|) {
6250                     unlink $volid || die 'unlink failed\n';
6251                 } else {
6252                     PVE::Storage::vdisk_free($cfg, $volid);
6253                 }
6254                 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6255             };
6256             print STDERR "unable to cleanup '$volid' - $@" if $@;
6257         }
6258         die $err;
6259     }
6260
6261     rmtree $tmpdir;
6262
6263     rename($tmpfn, $conffile) ||
6264         die "unable to commit configuration file '$conffile'\n";
6265
6266     PVE::Cluster::cfs_update(); # make sure we read new file
6267
6268     eval { rescan($vmid, 1); };
6269     warn $@ if $@;
6270 }
6271
6272 sub restore_tar_archive {
6273     my ($archive, $vmid, $user, $opts) = @_;
6274
6275     if ($archive ne '-') {
6276         my $firstfile = tar_archive_read_firstfile($archive);
6277         die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6278             if $firstfile ne 'qemu-server.conf';
6279     }
6280
6281     my $storecfg = PVE::Storage::config();
6282
6283     # destroy existing data - keep empty config
6284     my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6285     destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6286
6287     my $tocmd = "/usr/lib/qemu-server/qmextract";
6288
6289     $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6290     $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6291     $tocmd .= ' --prealloc' if $opts->{prealloc};
6292     $tocmd .= ' --info' if $opts->{info};
6293
6294     # tar option "xf" does not autodetect compression when read from STDIN,
6295     # so we pipe to zcat
6296     my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6297         PVE::Tools::shellquote("--to-command=$tocmd");
6298
6299     my $tmpdir = "/var/tmp/vzdumptmp$$";
6300     mkpath $tmpdir;
6301
6302     local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6303     local $ENV{VZDUMP_VMID} = $vmid;
6304     local $ENV{VZDUMP_USER} = $user;
6305
6306     my $conffile = PVE::QemuConfig->config_file($vmid);
6307     my $tmpfn = "$conffile.$$.tmp";
6308
6309     # disable interrupts (always do cleanups)
6310     local $SIG{INT} =
6311         local $SIG{TERM} =
6312         local $SIG{QUIT} =
6313         local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6314
6315     eval {
6316         # enable interrupts
6317         local $SIG{INT} =
6318             local $SIG{TERM} =
6319             local $SIG{QUIT} =
6320             local $SIG{HUP} =
6321             local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6322
6323         if ($archive eq '-') {
6324             print "extracting archive from STDIN\n";
6325             run_command($cmd, input => "<&STDIN");
6326         } else {
6327             print "extracting archive '$archive'\n";
6328             run_command($cmd);
6329         }
6330
6331         return if $opts->{info};
6332
6333         # read new mapping
6334         my $map = {};
6335         my $statfile = "$tmpdir/qmrestore.stat";
6336         if (my $fd = IO::File->new($statfile, "r")) {
6337             while (defined (my $line = <$fd>)) {
6338                 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6339                     $map->{$1} = $2 if $1;
6340                 } else {
6341                     print STDERR "unable to parse line in statfile - $line\n";
6342                 }
6343             }
6344             $fd->close();
6345         }
6346
6347         my $confsrc = "$tmpdir/qemu-server.conf";
6348
6349         my $srcfd = new IO::File($confsrc, "r") ||
6350             die "unable to open file '$confsrc'\n";
6351
6352         my $outfd = new IO::File ($tmpfn, "w") ||
6353             die "unable to write config for VM $vmid\n";
6354
6355         my $cookie = { netcount => 0 };
6356         while (defined (my $line = <$srcfd>)) {
6357             restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6358         }
6359
6360         $srcfd->close();
6361         $outfd->close();
6362     };
6363     my $err = $@;
6364
6365     if ($err) {
6366
6367         unlink $tmpfn;
6368
6369         tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6370
6371         die $err;
6372     }
6373
6374     rmtree $tmpdir;
6375
6376     rename $tmpfn, $conffile ||
6377         die "unable to commit configuration file '$conffile'\n";
6378
6379     PVE::Cluster::cfs_update(); # make sure we read new file
6380
6381     eval { rescan($vmid, 1); };
6382     warn $@ if $@;
6383 };
6384
6385 sub foreach_storage_used_by_vm {
6386     my ($conf, $func) = @_;
6387
6388     my $sidhash = {};
6389
6390     foreach_drive($conf, sub {
6391         my ($ds, $drive) = @_;
6392         return if drive_is_cdrom($drive);
6393
6394         my $volid = $drive->{file};
6395
6396         my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6397         $sidhash->{$sid} = $sid if $sid;
6398     });
6399
6400     foreach my $sid (sort keys %$sidhash) {
6401         &$func($sid);
6402     }
6403 }
6404
6405 sub do_snapshots_with_qemu {
6406     my ($storecfg, $volid) = @_;
6407
6408     my $storage_name = PVE::Storage::parse_volume_id($volid);
6409
6410     if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}}
6411         && !$storecfg->{ids}->{$storage_name}->{krbd}){
6412         return 1;
6413     }
6414
6415     if ($volid =~ m/\.(qcow2|qed)$/){
6416         return 1;
6417     }
6418
6419     return undef;
6420 }
6421
6422 sub qga_check_running {
6423     my ($vmid, $nowarn) = @_;
6424
6425     eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6426     if ($@) {
6427         warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6428         return 0;
6429     }
6430     return 1;
6431 }
6432
6433 sub template_create {
6434     my ($vmid, $conf, $disk) = @_;
6435
6436     my $storecfg = PVE::Storage::config();
6437
6438     foreach_drive($conf, sub {
6439         my ($ds, $drive) = @_;
6440
6441         return if drive_is_cdrom($drive);
6442         return if $disk && $ds ne $disk;
6443
6444         my $volid = $drive->{file};
6445         return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6446
6447         my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6448         $drive->{file} = $voliddst;
6449         $conf->{$ds} = print_drive($vmid, $drive);
6450         PVE::QemuConfig->write_config($vmid, $conf);
6451     });
6452 }
6453
6454 sub qemu_img_convert {
6455     my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6456
6457     my $storecfg = PVE::Storage::config();
6458     my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6459     my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6460
6461     if ($src_storeid && $dst_storeid) {
6462
6463         PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6464
6465         my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6466         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6467
6468         my $src_format = qemu_img_format($src_scfg, $src_volname);
6469         my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6470
6471         my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6472         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6473
6474         my $cmd = [];
6475         push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6476         push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6477         push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6478         push @$cmd, '-T', 'none' if $src_scfg->{type} eq 'zfspool';
6479         push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path;
6480         if ($is_zero_initialized) {
6481             push @$cmd, "zeroinit:$dst_path";
6482         } else {
6483             push @$cmd, $dst_path;
6484         }
6485
6486         my $parser = sub {
6487             my $line = shift;
6488             if($line =~ m/\((\S+)\/100\%\)/){
6489                 my $percent = $1;
6490                 my $transferred = int($size * $percent / 100);
6491                 my $remaining = $size - $transferred;
6492
6493                 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6494             }
6495
6496         };
6497
6498         eval  { run_command($cmd, timeout => undef, outfunc => $parser); };
6499         my $err = $@;
6500         die "copy failed: $err" if $err;
6501     }
6502 }
6503
6504 sub qemu_img_format {
6505     my ($scfg, $volname) = @_;
6506
6507     if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6508         return $1;
6509     } else {
6510         return "raw";
6511     }
6512 }
6513
6514 sub qemu_drive_mirror {
6515     my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga) = @_;
6516
6517     $jobs = {} if !$jobs;
6518
6519     my $qemu_target;
6520     my $format;
6521     $jobs->{"drive-$drive"} = {};
6522
6523     if ($dst_volid =~ /^nbd:/) {
6524         $qemu_target = $dst_volid;
6525         $format = "nbd";
6526     } else {
6527         my $storecfg = PVE::Storage::config();
6528         my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6529
6530         my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6531
6532         $format = qemu_img_format($dst_scfg, $dst_volname);
6533
6534         my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6535
6536         $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6537     }
6538
6539     my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6540     $opts->{format} = $format if $format;
6541
6542     print "drive mirror is starting for drive-$drive\n";
6543
6544     eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); }; #if a job already run for this device,it's throw an error
6545
6546     if (my $err = $@) {
6547         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6548         die "mirroring error: $err";
6549     }
6550
6551     qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
6552 }
6553
6554 sub qemu_drive_mirror_monitor {
6555     my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
6556
6557     eval {
6558         my $err_complete = 0;
6559
6560         while (1) {
6561             die "storage migration timed out\n" if $err_complete > 300;
6562
6563             my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6564
6565             my $running_mirror_jobs = {};
6566             foreach my $stat (@$stats) {
6567                 next if $stat->{type} ne 'mirror';
6568                 $running_mirror_jobs->{$stat->{device}} = $stat;
6569             }
6570
6571             my $readycounter = 0;
6572
6573             foreach my $job (keys %$jobs) {
6574
6575                 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6576                     print "$job : finished\n";
6577                     delete $jobs->{$job};
6578                     next;
6579                 }
6580
6581                 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6582
6583                 my $busy = $running_mirror_jobs->{$job}->{busy};
6584                 my $ready = $running_mirror_jobs->{$job}->{ready};
6585                 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6586                     my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6587                     my $remaining = $total - $transferred;
6588                     my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6589
6590                     print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6591                 }
6592
6593                 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6594             }
6595
6596             last if scalar(keys %$jobs) == 0;
6597
6598             if ($readycounter == scalar(keys %$jobs)) {
6599                 print "all mirroring jobs are ready \n";
6600                 last if $skipcomplete; #do the complete later
6601
6602                 if ($vmiddst && $vmiddst != $vmid) {
6603                     my $agent_running = $qga && qga_check_running($vmid);
6604                     if ($agent_running) {
6605                         print "freeze filesystem\n";
6606                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6607                     } else {
6608                         print "suspend vm\n";
6609                         eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6610                     }
6611
6612                     # if we clone a disk for a new target vm, we don't switch the disk
6613                     PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6614
6615                     if ($agent_running) {
6616                         print "unfreeze filesystem\n";
6617                         eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6618                     } else {
6619                         print "resume vm\n";
6620                         eval {  PVE::QemuServer::vm_resume($vmid, 1, 1); };
6621                     }
6622
6623                     last;
6624                 } else {
6625
6626                     foreach my $job (keys %$jobs) {
6627                         # try to switch the disk if source and destination are on the same guest
6628                         print "$job: Completing block job...\n";
6629
6630                         eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
6631                         if ($@ =~ m/cannot be completed/) {
6632                             print "$job: Block job cannot be completed, try again.\n";
6633                             $err_complete++;
6634                         }else {
6635                             print "$job: Completed successfully.\n";
6636                             $jobs->{$job}->{complete} = 1;
6637                         }
6638                     }
6639                 }
6640             }
6641             sleep 1;
6642         }
6643     };
6644     my $err = $@;
6645
6646     if ($err) {
6647         eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6648         die "mirroring error: $err";
6649     }
6650
6651 }
6652
6653 sub qemu_blockjobs_cancel {
6654     my ($vmid, $jobs) = @_;
6655
6656     foreach my $job (keys %$jobs) {
6657         print "$job: Cancelling block job\n";
6658         eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
6659         $jobs->{$job}->{cancel} = 1;
6660     }
6661
6662     while (1) {
6663         my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6664
6665         my $running_jobs = {};
6666         foreach my $stat (@$stats) {
6667             $running_jobs->{$stat->{device}} = $stat;
6668         }
6669
6670         foreach my $job (keys %$jobs) {
6671
6672             if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
6673                 print "$job: Done.\n";
6674                 delete $jobs->{$job};
6675             }
6676         }
6677
6678         last if scalar(keys %$jobs) == 0;
6679
6680         sleep 1;
6681     }
6682 }
6683
6684 sub clone_disk {
6685     my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6686         $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga) = @_;
6687
6688     my $newvolid;
6689
6690     if (!$full) {
6691         print "create linked clone of drive $drivename ($drive->{file})\n";
6692         $newvolid = PVE::Storage::vdisk_clone($storecfg,  $drive->{file}, $newvmid, $snapname);
6693         push @$newvollist, $newvolid;
6694     } else {
6695
6696         my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6697         $storeid = $storage if $storage;
6698
6699         my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
6700         my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6701
6702         print "create full clone of drive $drivename ($drive->{file})\n";
6703         my $name = undef;
6704         if (drive_is_cloudinit($drive)) {
6705             $name = "vm-$newvmid-cloudinit";
6706             # cloudinit only supports raw and qcow2 atm:
6707             if ($dst_format eq 'qcow2') {
6708                 $name .= '.qcow2';
6709             } elsif ($dst_format ne 'raw') {
6710                 die "clone: unhandled format for cloudinit image\n";
6711             }
6712         }
6713         $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
6714         push @$newvollist, $newvolid;
6715
6716         PVE::Storage::activate_volumes($storecfg, [$newvolid]);
6717
6718         my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6719         if (!$running || $snapname) {
6720             qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6721         } else {
6722
6723             my $kvmver = get_running_qemu_version ($vmid);
6724             if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
6725                 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
6726                     if $drive->{iothread};
6727             }
6728
6729             qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga);
6730         }
6731     }
6732
6733     my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6734
6735     my $disk = $drive;
6736     $disk->{format} = undef;
6737     $disk->{file} = $newvolid;
6738     $disk->{size} = $size;
6739
6740     return $disk;
6741 }
6742
6743 # this only works if VM is running
6744 sub get_current_qemu_machine {
6745     my ($vmid) = @_;
6746
6747     my $cmd = { execute => 'query-machines', arguments => {} };
6748     my $res = vm_qmp_command($vmid, $cmd);
6749
6750     my ($current, $default);
6751     foreach my $e (@$res) {
6752         $default = $e->{name} if $e->{'is-default'};
6753         $current = $e->{name} if $e->{'is-current'};
6754     }
6755
6756     # fallback to the default machine if current is not supported by qemu
6757     return $current || $default || 'pc';
6758 }
6759
6760 sub get_running_qemu_version {
6761     my ($vmid) = @_;
6762     my $cmd = { execute => 'query-version', arguments => {} };
6763     my $res = vm_qmp_command($vmid, $cmd);
6764     return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
6765 }
6766
6767 sub qemu_machine_feature_enabled {
6768     my ($machine, $kvmver, $version_major, $version_minor) = @_;
6769
6770     my $current_major;
6771     my $current_minor;
6772
6773     if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
6774
6775         $current_major = $3;
6776         $current_minor = $4;
6777
6778     } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6779
6780         $current_major = $1;
6781         $current_minor = $2;
6782     }
6783
6784     return 1 if $current_major > $version_major ||
6785                 ($current_major == $version_major &&
6786                  $current_minor >= $version_minor);
6787 }
6788
6789 sub qemu_machine_pxe {
6790     my ($vmid, $conf, $machine) = @_;
6791
6792     $machine =  PVE::QemuServer::get_current_qemu_machine($vmid) if !$machine;
6793
6794     if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
6795         $machine .= '.pxe';
6796     }
6797
6798     return $machine;
6799 }
6800
6801 sub qemu_use_old_bios_files {
6802     my ($machine_type) = @_;
6803
6804     return if !$machine_type;
6805
6806     my $use_old_bios_files = undef;
6807
6808     if ($machine_type =~ m/^(\S+)\.pxe$/) {
6809         $machine_type = $1;
6810         $use_old_bios_files = 1;
6811     } else {
6812         my $kvmver = kvm_user_version();
6813         # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6814         # load new efi bios files on migration. So this hack is required to allow
6815         # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6816         # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6817         $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
6818     }
6819
6820     return ($use_old_bios_files, $machine_type);
6821 }
6822
6823 sub create_efidisk($$$$$) {
6824     my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
6825
6826     my (undef, $ovmf_vars) = get_ovmf_files($arch);
6827     die "EFI vars default image not found\n" if ! -f $ovmf_vars;
6828
6829     my $vars_size = PVE::Tools::convert_size(-s $ovmf_vars, 'b' => 'kb');
6830     my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
6831     PVE::Storage::activate_volumes($storecfg, [$volid]);
6832
6833     my $path = PVE::Storage::path($storecfg, $volid);
6834     eval {
6835         run_command(['/usr/bin/qemu-img', 'convert', '-n', '-f', 'raw', '-O', $fmt, $ovmf_vars, $path]);
6836     };
6837     die "Copying EFI vars image failed: $@" if $@;
6838
6839     return ($volid, $vars_size);
6840 }
6841
6842 sub lspci {
6843
6844     my $devices = {};
6845
6846     dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6847             my (undef, $id, $function) = @_;
6848             my $res = { id => $id, function => $function};
6849             push @{$devices->{$id}}, $res;
6850     });
6851
6852     # Entries should be sorted by functions.
6853     foreach my $id (keys %$devices) {
6854         my $dev = $devices->{$id};
6855         $devices->{$id} = [ sort { $a->{function} <=> $b->{function} } @$dev ];
6856     }
6857
6858     return $devices;
6859 }
6860
6861 sub vm_iothreads_list {
6862     my ($vmid) = @_;
6863
6864     my $res = vm_mon_cmd($vmid, 'query-iothreads');
6865
6866     my $iothreads = {};
6867     foreach my $iothread (@$res) {
6868         $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6869     }
6870
6871     return $iothreads;
6872 }
6873
6874 sub scsihw_infos {
6875     my ($conf, $drive) = @_;
6876
6877     my $maxdev = 0;
6878
6879     if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
6880         $maxdev = 7;
6881     } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
6882         $maxdev = 1;
6883     } else {
6884         $maxdev = 256;
6885     }
6886
6887     my $controller = int($drive->{index} / $maxdev);
6888     my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
6889
6890     return ($maxdev, $controller, $controller_prefix);
6891 }
6892
6893 sub add_hyperv_enlightenments {
6894     my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough) = @_;
6895
6896     return if $winversion < 6;
6897     return if $bios && $bios eq 'ovmf' && $winversion < 8;
6898
6899     push @$cpuFlags , 'hv_vendor_id=proxmox' if $gpu_passthrough;
6900
6901     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
6902         push @$cpuFlags , 'hv_spinlocks=0x1fff';
6903         push @$cpuFlags , 'hv_vapic';
6904         push @$cpuFlags , 'hv_time';
6905     } else {
6906         push @$cpuFlags , 'hv_spinlocks=0xffff';
6907     }
6908
6909     if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
6910         push @$cpuFlags , 'hv_reset';
6911         push @$cpuFlags , 'hv_vpindex';
6912         push @$cpuFlags , 'hv_runtime';
6913     }
6914
6915     if ($winversion >= 7) {
6916         push @$cpuFlags , 'hv_relaxed';
6917
6918         if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
6919             push @$cpuFlags , 'hv_synic';
6920             push @$cpuFlags , 'hv_stimer';
6921         }
6922     }
6923 }
6924
6925 sub windows_version {
6926     my ($ostype) = @_;
6927
6928     return 0 if !$ostype;
6929
6930     my $winversion = 0;
6931
6932     if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
6933         $winversion = 5;
6934     } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
6935         $winversion = 6;
6936     } elsif ($ostype =~ m/^win(\d+)$/) {
6937         $winversion = $1;
6938     }
6939
6940     return $winversion;
6941 }
6942
6943 sub resolve_dst_disk_format {
6944         my ($storecfg, $storeid, $src_volname, $format) = @_;
6945         my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6946
6947         if (!$format) {
6948             # if no target format is specified, use the source disk format as hint
6949             if ($src_volname) {
6950                 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6951                 $format = qemu_img_format($scfg, $src_volname);
6952             } else {
6953                 return $defFormat;
6954             }
6955         }
6956
6957         # test if requested format is supported - else use default
6958         my $supported = grep { $_ eq $format } @$validFormats;
6959         $format = $defFormat if !$supported;
6960         return $format;
6961 }
6962
6963 sub resolve_first_disk {
6964     my $conf = shift;
6965     my @disks = PVE::QemuServer::valid_drive_names();
6966     my $firstdisk;
6967     foreach my $ds (reverse @disks) {
6968         next if !$conf->{$ds};
6969         my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
6970         next if PVE::QemuServer::drive_is_cdrom($disk);
6971         $firstdisk = $ds;
6972     }
6973     return $firstdisk;
6974 }
6975
6976 sub generate_uuid {
6977     my ($uuid, $uuid_str);
6978     UUID::generate($uuid);
6979     UUID::unparse($uuid, $uuid_str);
6980     return $uuid_str;
6981 }
6982
6983 sub generate_smbios1_uuid {
6984     return "uuid=".generate_uuid();
6985 }
6986
6987 sub nbd_stop {
6988     my ($vmid) = @_;
6989
6990     vm_mon_cmd($vmid, 'nbd-server-stop');
6991 }
6992
6993 # bash completion helper
6994
6995 sub complete_backup_archives {
6996     my ($cmdname, $pname, $cvalue) = @_;
6997
6998     my $cfg = PVE::Storage::config();
6999
7000     my $storeid;
7001
7002     if ($cvalue =~ m/^([^:]+):/) {
7003         $storeid = $1;
7004     }
7005
7006     my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7007
7008     my $res = [];
7009     foreach my $id (keys %$data) {
7010         foreach my $item (@{$data->{$id}}) {
7011             next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7012             push @$res, $item->{volid} if defined($item->{volid});
7013         }
7014     }
7015
7016     return $res;
7017 }
7018
7019 my $complete_vmid_full = sub {
7020     my ($running) = @_;
7021
7022     my $idlist = vmstatus();
7023
7024     my $res = [];
7025
7026     foreach my $id (keys %$idlist) {
7027         my $d = $idlist->{$id};
7028         if (defined($running)) {
7029             next if $d->{template};
7030             next if $running && $d->{status} ne 'running';
7031             next if !$running && $d->{status} eq 'running';
7032         }
7033         push @$res, $id;
7034
7035     }
7036     return $res;
7037 };
7038
7039 sub complete_vmid {
7040     return &$complete_vmid_full();
7041 }
7042
7043 sub complete_vmid_stopped {
7044     return &$complete_vmid_full(0);
7045 }
7046
7047 sub complete_vmid_running {
7048     return &$complete_vmid_full(1);
7049 }
7050
7051 sub complete_storage {
7052
7053     my $cfg = PVE::Storage::config();
7054     my $ids = $cfg->{ids};
7055
7056     my $res = [];
7057     foreach my $sid (keys %$ids) {
7058         next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7059         next if !$ids->{$sid}->{content}->{images};
7060         push @$res, $sid;
7061     }
7062
7063     return $res;
7064 }
7065
7066 1;