]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
d69cd969a96ecd22cce0d4f2e4a47d614ce19932
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5
6 use POSIX;
7 use IO::Handle;
8 use IO::Select;
9 use IO::File;
10 use IO::Dir;
11 use IO::Socket::UNIX;
12 use File::Basename;
13 use File::Path;
14 use File::stat;
15 use Getopt::Long;
16 use Digest::SHA;
17 use Fcntl ':flock';
18 use Cwd 'abs_path';
19 use IPC::Open3;
20 use JSON;
21 use Fcntl;
22 use PVE::SafeSyslog;
23 use Storable qw(dclone);
24 use MIME::Base64;
25 use PVE::Exception qw(raise raise_param_exc);
26 use PVE::Storage;
27 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach $IPV6RE);
28 use PVE::JSONSchema qw(get_standard_option);
29 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
30 use PVE::INotify;
31 use PVE::ProcFSTools;
32 use PVE::QemuConfig;
33 use PVE::QMPClient;
34 use PVE::RPCEnvironment;
35 use PVE::GuestHelpers;
36 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr print_pcie_root_port);
37 use PVE::QemuServer::Memory;
38 use PVE::QemuServer::USB qw(parse_usb_device);
39 use PVE::QemuServer::Cloudinit;
40 use PVE::SysFSTools;
41 use PVE::Systemd;
42 use Time::HiRes qw(gettimeofday);
43 use File::Copy qw(copy);
44 use URI::Escape;
45
46 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
47 my $OVMF = {
48 x86_64 => [
49 "$EDK2_FW_BASE/OVMF_CODE.fd",
50 "$EDK2_FW_BASE/OVMF_VARS.fd"
51 ],
52 aarch64 => [
53 "$EDK2_FW_BASE/AAVMF_CODE.fd",
54 "$EDK2_FW_BASE/AAVMF_VARS.fd"
55 ],
56 };
57
58 my $qemu_snap_storage = { rbd => 1 };
59
60 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
61
62 my $QEMU_FORMAT_RE = qr/raw|cow|qcow|qcow2|qed|vmdk|cloop/;
63
64 # Note about locking: we use flock on the config file protect
65 # against concurent actions.
66 # Aditionaly, we have a 'lock' setting in the config file. This
67 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
68 # allowed when such lock is set. But you can ignore this kind of
69 # lock with the --skiplock flag.
70
71 cfs_register_file('/qemu-server/',
72 \&parse_vm_config,
73 \&write_vm_config);
74
75 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
76 description => "Some command save/restore state from this location.",
77 type => 'string',
78 maxLength => 128,
79 optional => 1,
80 });
81
82 PVE::JSONSchema::register_standard_option('pve-qm-image-format', {
83 type => 'string',
84 enum => [qw(raw cow qcow qed qcow2 vmdk cloop)],
85 description => "The drive's backing file's data format.",
86 optional => 1,
87 });
88
89 PVE::JSONSchema::register_standard_option('pve-qemu-machine', {
90 description => "Specifies the Qemu machine type.",
91 type => 'string',
92 pattern => '(pc|pc(-i440fx)?-\d+\.\d+(\.pxe)?|q35|pc-q35-\d+\.\d+(\.pxe)?|virt(?:-\d+\.\d+)?)',
93 maxLength => 40,
94 optional => 1,
95 });
96
97 #no warnings 'redefine';
98
99 sub cgroups_write {
100 my ($controller, $vmid, $option, $value) = @_;
101
102 my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
103 PVE::ProcFSTools::write_proc_entry($path, $value);
104
105 }
106
107 my $nodename = PVE::INotify::nodename();
108
109 mkdir "/etc/pve/nodes/$nodename";
110 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
111 mkdir $confdir;
112
113 my $var_run_tmpdir = "/var/run/qemu-server";
114 mkdir $var_run_tmpdir;
115
116 my $lock_dir = "/var/lock/qemu-server";
117 mkdir $lock_dir;
118
119 my $cpu_vendor_list = {
120 # Intel CPUs
121 486 => 'GenuineIntel',
122 pentium => 'GenuineIntel',
123 pentium2 => 'GenuineIntel',
124 pentium3 => 'GenuineIntel',
125 coreduo => 'GenuineIntel',
126 core2duo => 'GenuineIntel',
127 Conroe => 'GenuineIntel',
128 Penryn => 'GenuineIntel',
129 Nehalem => 'GenuineIntel',
130 'Nehalem-IBRS' => 'GenuineIntel',
131 Westmere => 'GenuineIntel',
132 'Westmere-IBRS' => 'GenuineIntel',
133 SandyBridge => 'GenuineIntel',
134 'SandyBridge-IBRS' => 'GenuineIntel',
135 IvyBridge => 'GenuineIntel',
136 'IvyBridge-IBRS' => 'GenuineIntel',
137 Haswell => 'GenuineIntel',
138 'Haswell-IBRS' => 'GenuineIntel',
139 'Haswell-noTSX' => 'GenuineIntel',
140 'Haswell-noTSX-IBRS' => 'GenuineIntel',
141 Broadwell => 'GenuineIntel',
142 'Broadwell-IBRS' => 'GenuineIntel',
143 'Broadwell-noTSX' => 'GenuineIntel',
144 'Broadwell-noTSX-IBRS' => 'GenuineIntel',
145 'Skylake-Client' => 'GenuineIntel',
146 'Skylake-Client-IBRS' => 'GenuineIntel',
147 'Skylake-Server' => 'GenuineIntel',
148 'Skylake-Server-IBRS' => 'GenuineIntel',
149 'Cascadelake-Server' => 'GenuineIntel',
150 KnightsMill => 'GenuineIntel',
151
152
153 # AMD CPUs
154 athlon => 'AuthenticAMD',
155 phenom => 'AuthenticAMD',
156 Opteron_G1 => 'AuthenticAMD',
157 Opteron_G2 => 'AuthenticAMD',
158 Opteron_G3 => 'AuthenticAMD',
159 Opteron_G4 => 'AuthenticAMD',
160 Opteron_G5 => 'AuthenticAMD',
161 EPYC => 'AuthenticAMD',
162 'EPYC-IBPB' => 'AuthenticAMD',
163
164 # generic types, use vendor from host node
165 host => 'default',
166 kvm32 => 'default',
167 kvm64 => 'default',
168 qemu32 => 'default',
169 qemu64 => 'default',
170 max => 'default',
171 };
172
173 my @supported_cpu_flags = (
174 'pcid',
175 'spec-ctrl',
176 'ibpb',
177 'ssbd',
178 'virt-ssbd',
179 'amd-ssbd',
180 'amd-no-ssb',
181 'pdpe1gb',
182 'md-clear',
183 'hv-tlbflush',
184 'hv-evmcs',
185 'aes'
186 );
187 my $cpu_flag = qr/[+-](@{[join('|', @supported_cpu_flags)]})/;
188
189 my $cpu_fmt = {
190 cputype => {
191 description => "Emulated CPU type.",
192 type => 'string',
193 enum => [ sort { "\L$a" cmp "\L$b" } keys %$cpu_vendor_list ],
194 default => 'kvm64',
195 default_key => 1,
196 },
197 hidden => {
198 description => "Do not identify as a KVM virtual machine.",
199 type => 'boolean',
200 optional => 1,
201 default => 0
202 },
203 'hv-vendor-id' => {
204 type => 'string',
205 pattern => qr/[a-zA-Z0-9]{1,12}/,
206 format_description => 'vendor-id',
207 description => 'The Hyper-V vendor ID. Some drivers or programs inside Windows guests need a specific ID.',
208 optional => 1,
209 },
210 flags => {
211 description => "List of additional CPU flags separated by ';'."
212 . " Use '+FLAG' to enable, '-FLAG' to disable a flag."
213 . " Currently supported flags: @{[join(', ', @supported_cpu_flags)]}.",
214 format_description => '+FLAG[;-FLAG...]',
215 type => 'string',
216 pattern => qr/$cpu_flag(;$cpu_flag)*/,
217 optional => 1,
218 },
219 };
220
221 my $watchdog_fmt = {
222 model => {
223 default_key => 1,
224 type => 'string',
225 enum => [qw(i6300esb ib700)],
226 description => "Watchdog type to emulate.",
227 default => 'i6300esb',
228 optional => 1,
229 },
230 action => {
231 type => 'string',
232 enum => [qw(reset shutdown poweroff pause debug none)],
233 description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
234 optional => 1,
235 },
236 };
237 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
238
239 my $agent_fmt = {
240 enabled => {
241 description => "Enable/disable Qemu GuestAgent.",
242 type => 'boolean',
243 default => 0,
244 default_key => 1,
245 },
246 fstrim_cloned_disks => {
247 description => "Run fstrim after cloning/moving a disk.",
248 type => 'boolean',
249 optional => 1,
250 default => 0
251 },
252 };
253
254 my $vga_fmt = {
255 type => {
256 description => "Select the VGA type.",
257 type => 'string',
258 default => 'std',
259 optional => 1,
260 default_key => 1,
261 enum => [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio vmware)],
262 },
263 memory => {
264 description => "Sets the VGA memory (in MiB). Has no effect with serial display.",
265 type => 'integer',
266 optional => 1,
267 minimum => 4,
268 maximum => 512,
269 },
270 };
271
272 my $ivshmem_fmt = {
273 size => {
274 type => 'integer',
275 minimum => 1,
276 description => "The size of the file in MB.",
277 },
278 name => {
279 type => 'string',
280 pattern => '[a-zA-Z0-9\-]+',
281 optional => 1,
282 format_description => 'string',
283 description => "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
284 },
285 };
286
287 my $audio_fmt = {
288 device => {
289 type => 'string',
290 enum => [qw(ich9-intel-hda intel-hda AC97)],
291 description => "Configure an audio device."
292 },
293 driver => {
294 type => 'string',
295 enum => ['spice'],
296 default => 'spice',
297 optional => 1,
298 description => "Driver backend for the audio device."
299 },
300 };
301
302 my $spice_enhancements_fmt = {
303 foldersharing => {
304 type => 'boolean',
305 optional => 1,
306 default => '0',
307 description => "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
308 },
309 videostreaming => {
310 type => 'string',
311 enum => ['off', 'all', 'filter'],
312 default => 'off',
313 optional => 1,
314 description => "Enable video streaming. Uses compression for detected video streams."
315 },
316 };
317
318 my $confdesc = {
319 onboot => {
320 optional => 1,
321 type => 'boolean',
322 description => "Specifies whether a VM will be started during system bootup.",
323 default => 0,
324 },
325 autostart => {
326 optional => 1,
327 type => 'boolean',
328 description => "Automatic restart after crash (currently ignored).",
329 default => 0,
330 },
331 hotplug => {
332 optional => 1,
333 type => 'string', format => 'pve-hotplug-features',
334 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'.",
335 default => 'network,disk,usb',
336 },
337 reboot => {
338 optional => 1,
339 type => 'boolean',
340 description => "Allow reboot. If set to '0' the VM exit on reboot.",
341 default => 1,
342 },
343 lock => {
344 optional => 1,
345 type => 'string',
346 description => "Lock/unlock the VM.",
347 enum => [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
348 },
349 cpulimit => {
350 optional => 1,
351 type => 'number',
352 description => "Limit of CPU usage.",
353 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.",
354 minimum => 0,
355 maximum => 128,
356 default => 0,
357 },
358 cpuunits => {
359 optional => 1,
360 type => 'integer',
361 description => "CPU weight for a VM.",
362 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.",
363 minimum => 2,
364 maximum => 262144,
365 default => 1024,
366 },
367 memory => {
368 optional => 1,
369 type => 'integer',
370 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
371 minimum => 16,
372 default => 512,
373 },
374 balloon => {
375 optional => 1,
376 type => 'integer',
377 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
378 minimum => 0,
379 },
380 shares => {
381 optional => 1,
382 type => 'integer',
383 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.",
384 minimum => 0,
385 maximum => 50000,
386 default => 1000,
387 },
388 keyboard => {
389 optional => 1,
390 type => 'string',
391 description => "Keybord layout for vnc server. Default is read from the '/etc/pve/datacenter.cfg' configuration file.".
392 "It should not be necessary to set it.",
393 enum => PVE::Tools::kvmkeymaplist(),
394 default => undef,
395 },
396 name => {
397 optional => 1,
398 type => 'string', format => 'dns-name',
399 description => "Set a name for the VM. Only used on the configuration web interface.",
400 },
401 scsihw => {
402 optional => 1,
403 type => 'string',
404 description => "SCSI controller model",
405 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
406 default => 'lsi',
407 },
408 description => {
409 optional => 1,
410 type => 'string',
411 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
412 },
413 ostype => {
414 optional => 1,
415 type => 'string',
416 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 l24 l26 solaris)],
417 description => "Specify guest operating system.",
418 verbose_description => <<EODESC,
419 Specify guest operating system. This is used to enable special
420 optimization/features for specific operating systems:
421
422 [horizontal]
423 other;; unspecified OS
424 wxp;; Microsoft Windows XP
425 w2k;; Microsoft Windows 2000
426 w2k3;; Microsoft Windows 2003
427 w2k8;; Microsoft Windows 2008
428 wvista;; Microsoft Windows Vista
429 win7;; Microsoft Windows 7
430 win8;; Microsoft Windows 8/2012/2012r2
431 win10;; Microsoft Windows 10/2016
432 l24;; Linux 2.4 Kernel
433 l26;; Linux 2.6/3.X Kernel
434 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
435 EODESC
436 },
437 boot => {
438 optional => 1,
439 type => 'string',
440 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
441 pattern => '[acdn]{1,4}',
442 default => 'cdn',
443 },
444 bootdisk => {
445 optional => 1,
446 type => 'string', format => 'pve-qm-bootdisk',
447 description => "Enable booting from specified disk.",
448 pattern => '(ide|sata|scsi|virtio)\d+',
449 },
450 smp => {
451 optional => 1,
452 type => 'integer',
453 description => "The number of CPUs. Please use option -sockets instead.",
454 minimum => 1,
455 default => 1,
456 },
457 sockets => {
458 optional => 1,
459 type => 'integer',
460 description => "The number of CPU sockets.",
461 minimum => 1,
462 default => 1,
463 },
464 cores => {
465 optional => 1,
466 type => 'integer',
467 description => "The number of cores per socket.",
468 minimum => 1,
469 default => 1,
470 },
471 numa => {
472 optional => 1,
473 type => 'boolean',
474 description => "Enable/disable NUMA.",
475 default => 0,
476 },
477 hugepages => {
478 optional => 1,
479 type => 'string',
480 description => "Enable/disable hugepages memory.",
481 enum => [qw(any 2 1024)],
482 },
483 vcpus => {
484 optional => 1,
485 type => 'integer',
486 description => "Number of hotplugged vcpus.",
487 minimum => 1,
488 default => 0,
489 },
490 acpi => {
491 optional => 1,
492 type => 'boolean',
493 description => "Enable/disable ACPI.",
494 default => 1,
495 },
496 agent => {
497 optional => 1,
498 description => "Enable/disable Qemu GuestAgent and its properties.",
499 type => 'string',
500 format => $agent_fmt,
501 },
502 kvm => {
503 optional => 1,
504 type => 'boolean',
505 description => "Enable/disable KVM hardware virtualization.",
506 default => 1,
507 },
508 tdf => {
509 optional => 1,
510 type => 'boolean',
511 description => "Enable/disable time drift fix.",
512 default => 0,
513 },
514 localtime => {
515 optional => 1,
516 type => 'boolean',
517 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
518 },
519 freeze => {
520 optional => 1,
521 type => 'boolean',
522 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
523 },
524 vga => {
525 optional => 1,
526 type => 'string', format => $vga_fmt,
527 description => "Configure the VGA hardware.",
528 verbose_description => "Configure the VGA Hardware. If you want to use ".
529 "high resolution modes (>= 1280x1024x16) you may need to increase " .
530 "the vga memory option. Since QEMU 2.9 the default VGA display type " .
531 "is 'std' for all OS types besides some Windows versions (XP and " .
532 "older) which use 'cirrus'. The 'qxl' option enables the SPICE " .
533 "display server. For win* OS you can select how many independent " .
534 "displays you want, Linux guests can add displays them self.\n".
535 "You can also run without any graphic card, using a serial device as terminal.",
536 },
537 watchdog => {
538 optional => 1,
539 type => 'string', format => 'pve-qm-watchdog',
540 description => "Create a virtual hardware watchdog device.",
541 verbose_description => "Create a virtual hardware watchdog device. Once enabled" .
542 " (by a guest action), the watchdog must be periodically polled " .
543 "by an agent inside the guest or else the watchdog will reset " .
544 "the guest (or execute the respective action specified)",
545 },
546 startdate => {
547 optional => 1,
548 type => 'string',
549 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
550 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'.",
551 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
552 default => 'now',
553 },
554 startup => get_standard_option('pve-startup-order'),
555 template => {
556 optional => 1,
557 type => 'boolean',
558 description => "Enable/disable Template.",
559 default => 0,
560 },
561 args => {
562 optional => 1,
563 type => 'string',
564 description => "Arbitrary arguments passed to kvm.",
565 verbose_description => <<EODESCR,
566 Arbitrary arguments passed to kvm, for example:
567
568 args: -no-reboot -no-hpet
569
570 NOTE: this option is for experts only.
571 EODESCR
572 },
573 tablet => {
574 optional => 1,
575 type => 'boolean',
576 default => 1,
577 description => "Enable/disable the USB tablet device.",
578 verbose_description => "Enable/disable the USB tablet device. This device is " .
579 "usually needed to allow absolute mouse positioning with VNC. " .
580 "Else the mouse runs out of sync with normal VNC clients. " .
581 "If you're running lots of console-only guests on one host, " .
582 "you may consider disabling this to save some context switches. " .
583 "This is turned off by default if you use spice (-vga=qxl).",
584 },
585 migrate_speed => {
586 optional => 1,
587 type => 'integer',
588 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
589 minimum => 0,
590 default => 0,
591 },
592 migrate_downtime => {
593 optional => 1,
594 type => 'number',
595 description => "Set maximum tolerated downtime (in seconds) for migrations.",
596 minimum => 0,
597 default => 0.1,
598 },
599 cdrom => {
600 optional => 1,
601 type => 'string', format => 'pve-qm-ide',
602 typetext => '<volume>',
603 description => "This is an alias for option -ide2",
604 },
605 cpu => {
606 optional => 1,
607 description => "Emulated CPU type.",
608 type => 'string',
609 format => $cpu_fmt,
610 },
611 parent => get_standard_option('pve-snapshot-name', {
612 optional => 1,
613 description => "Parent snapshot name. This is used internally, and should not be modified.",
614 }),
615 snaptime => {
616 optional => 1,
617 description => "Timestamp for snapshots.",
618 type => 'integer',
619 minimum => 0,
620 },
621 vmstate => {
622 optional => 1,
623 type => 'string', format => 'pve-volume-id',
624 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
625 },
626 vmstatestorage => get_standard_option('pve-storage-id', {
627 description => "Default storage for VM state volumes/files.",
628 optional => 1,
629 }),
630 runningmachine => get_standard_option('pve-qemu-machine', {
631 description => "Specifies the Qemu machine type of the running vm. This is used internally for snapshots.",
632 }),
633 machine => get_standard_option('pve-qemu-machine'),
634 arch => {
635 description => "Virtual processor architecture. Defaults to the host.",
636 optional => 1,
637 type => 'string',
638 enum => [qw(x86_64 aarch64)],
639 },
640 smbios1 => {
641 description => "Specify SMBIOS type 1 fields.",
642 type => 'string', format => 'pve-qm-smbios1',
643 maxLength => 512,
644 optional => 1,
645 },
646 protection => {
647 optional => 1,
648 type => 'boolean',
649 description => "Sets the protection flag of the VM. This will disable the remove VM and remove disk operations.",
650 default => 0,
651 },
652 bios => {
653 optional => 1,
654 type => 'string',
655 enum => [ qw(seabios ovmf) ],
656 description => "Select BIOS implementation.",
657 default => 'seabios',
658 },
659 vmgenid => {
660 type => 'string',
661 pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
662 format_description => 'UUID',
663 description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0' to disable explicitly.",
664 verbose_description => "The VM generation ID (vmgenid) device exposes a".
665 " 128-bit integer value identifier to the guest OS. This allows to".
666 " notify the guest operating system when the virtual machine is".
667 " executed with a different configuration (e.g. snapshot execution".
668 " or creation from a template). The guest operating system notices".
669 " the change, and is then able to react as appropriate by marking".
670 " its copies of distributed databases as dirty, re-initializing its".
671 " random number generator, etc.\n".
672 "Note that auto-creation only works when done throug API/CLI create".
673 " or update methods, but not when manually editing the config file.",
674 default => "1 (autogenerated)",
675 optional => 1,
676 },
677 hookscript => {
678 type => 'string',
679 format => 'pve-volume-id',
680 optional => 1,
681 description => "Script that will be executed during various steps in the vms lifetime.",
682 },
683 ivshmem => {
684 type => 'string',
685 format => $ivshmem_fmt,
686 description => "Inter-VM shared memory. Useful for direct communication between VMs, or to the host.",
687 optional => 1,
688 },
689 audio0 => {
690 type => 'string',
691 format => $audio_fmt,
692 description => "Configure a audio device, useful in combination with QXL/Spice.",
693 optional => 1
694 },
695 spice_enhancements => {
696 type => 'string',
697 format => $spice_enhancements_fmt,
698 description => "Configure additional enhancements for SPICE.",
699 optional => 1
700 },
701 };
702
703 my $cicustom_fmt = {
704 meta => {
705 type => 'string',
706 optional => 1,
707 description => 'Specify a custom file containing all meta data passed to the VM via cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
708 format => 'pve-volume-id',
709 format_description => 'volume',
710 },
711 network => {
712 type => 'string',
713 optional => 1,
714 description => 'Specify a custom file containing all network data passed to the VM via cloud-init.',
715 format => 'pve-volume-id',
716 format_description => 'volume',
717 },
718 user => {
719 type => 'string',
720 optional => 1,
721 description => 'Specify a custom file containing all user data passed to the VM via cloud-init.',
722 format => 'pve-volume-id',
723 format_description => 'volume',
724 },
725 };
726 PVE::JSONSchema::register_format('pve-qm-cicustom', $cicustom_fmt);
727
728 my $confdesc_cloudinit = {
729 citype => {
730 optional => 1,
731 type => 'string',
732 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.',
733 enum => ['configdrive2', 'nocloud'],
734 },
735 ciuser => {
736 optional => 1,
737 type => 'string',
738 description => "cloud-init: User name to change ssh keys and password for instead of the image's configured default user.",
739 },
740 cipassword => {
741 optional => 1,
742 type => 'string',
743 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.',
744 },
745 cicustom => {
746 optional => 1,
747 type => 'string',
748 description => 'cloud-init: Specify custom files to replace the automatically generated ones at start.',
749 format => 'pve-qm-cicustom',
750 },
751 searchdomain => {
752 optional => 1,
753 type => 'string',
754 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.",
755 },
756 nameserver => {
757 optional => 1,
758 type => 'string', format => 'address-list',
759 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.",
760 },
761 sshkeys => {
762 optional => 1,
763 type => 'string',
764 format => 'urlencoded',
765 description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
766 },
767 };
768
769 # what about other qemu settings ?
770 #cpu => 'string',
771 #machine => 'string',
772 #fda => 'file',
773 #fdb => 'file',
774 #mtdblock => 'file',
775 #sd => 'file',
776 #pflash => 'file',
777 #snapshot => 'bool',
778 #bootp => 'file',
779 ##tftp => 'dir',
780 ##smb => 'dir',
781 #kernel => 'file',
782 #append => 'string',
783 #initrd => 'file',
784 ##soundhw => 'string',
785
786 while (my ($k, $v) = each %$confdesc) {
787 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
788 }
789
790 my $MAX_IDE_DISKS = 4;
791 my $MAX_SCSI_DISKS = 14;
792 my $MAX_VIRTIO_DISKS = 16;
793 my $MAX_SATA_DISKS = 6;
794 my $MAX_USB_DEVICES = 5;
795 my $MAX_NETS = 32;
796 my $MAX_UNUSED_DISKS = 256;
797 my $MAX_HOSTPCI_DEVICES = 16;
798 my $MAX_SERIAL_PORTS = 4;
799 my $MAX_PARALLEL_PORTS = 3;
800 my $MAX_NUMA = 8;
801
802 my $numa_fmt = {
803 cpus => {
804 type => "string",
805 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
806 description => "CPUs accessing this NUMA node.",
807 format_description => "id[-id];...",
808 },
809 memory => {
810 type => "number",
811 description => "Amount of memory this NUMA node provides.",
812 optional => 1,
813 },
814 hostnodes => {
815 type => "string",
816 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
817 description => "Host NUMA nodes to use.",
818 format_description => "id[-id];...",
819 optional => 1,
820 },
821 policy => {
822 type => 'string',
823 enum => [qw(preferred bind interleave)],
824 description => "NUMA allocation policy.",
825 optional => 1,
826 },
827 };
828 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
829 my $numadesc = {
830 optional => 1,
831 type => 'string', format => $numa_fmt,
832 description => "NUMA topology.",
833 };
834 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
835
836 for (my $i = 0; $i < $MAX_NUMA; $i++) {
837 $confdesc->{"numa$i"} = $numadesc;
838 }
839
840 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
841 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
842 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
843 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
844
845 my $net_fmt_bridge_descr = <<__EOD__;
846 Bridge to attach the network device to. The Proxmox VE standard bridge
847 is called 'vmbr0'.
848
849 If you do not specify a bridge, we create a kvm user (NATed) network
850 device, which provides DHCP and DNS services. The following addresses
851 are used:
852
853 10.0.2.2 Gateway
854 10.0.2.3 DNS Server
855 10.0.2.4 SMB Server
856
857 The DHCP server assign addresses to the guest starting from 10.0.2.15.
858 __EOD__
859
860 my $net_fmt = {
861 macaddr => get_standard_option('mac-addr', {
862 description => "MAC address. That address must be unique withing your network. This is automatically generated if not specified.",
863 }),
864 model => {
865 type => 'string',
866 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'.",
867 enum => $nic_model_list,
868 default_key => 1,
869 },
870 (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
871 bridge => {
872 type => 'string',
873 description => $net_fmt_bridge_descr,
874 format_description => 'bridge',
875 optional => 1,
876 },
877 queues => {
878 type => 'integer',
879 minimum => 0, maximum => 16,
880 description => 'Number of packet queues to be used on the device.',
881 optional => 1,
882 },
883 rate => {
884 type => 'number',
885 minimum => 0,
886 description => "Rate limit in mbps (megabytes per second) as floating point number.",
887 optional => 1,
888 },
889 tag => {
890 type => 'integer',
891 minimum => 1, maximum => 4094,
892 description => 'VLAN tag to apply to packets on this interface.',
893 optional => 1,
894 },
895 trunks => {
896 type => 'string',
897 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
898 description => 'VLAN trunks to pass through this interface.',
899 format_description => 'vlanid[;vlanid...]',
900 optional => 1,
901 },
902 firewall => {
903 type => 'boolean',
904 description => 'Whether this interface should be protected by the firewall.',
905 optional => 1,
906 },
907 link_down => {
908 type => 'boolean',
909 description => 'Whether this interface should be disconnected (like pulling the plug).',
910 optional => 1,
911 },
912 };
913
914 my $netdesc = {
915 optional => 1,
916 type => 'string', format => $net_fmt,
917 description => "Specify network devices.",
918 };
919
920 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
921
922 my $ipconfig_fmt = {
923 ip => {
924 type => 'string',
925 format => 'pve-ipv4-config',
926 format_description => 'IPv4Format/CIDR',
927 description => 'IPv4 address in CIDR format.',
928 optional => 1,
929 default => 'dhcp',
930 },
931 gw => {
932 type => 'string',
933 format => 'ipv4',
934 format_description => 'GatewayIPv4',
935 description => 'Default gateway for IPv4 traffic.',
936 optional => 1,
937 requires => 'ip',
938 },
939 ip6 => {
940 type => 'string',
941 format => 'pve-ipv6-config',
942 format_description => 'IPv6Format/CIDR',
943 description => 'IPv6 address in CIDR format.',
944 optional => 1,
945 default => 'dhcp',
946 },
947 gw6 => {
948 type => 'string',
949 format => 'ipv6',
950 format_description => 'GatewayIPv6',
951 description => 'Default gateway for IPv6 traffic.',
952 optional => 1,
953 requires => 'ip6',
954 },
955 };
956 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
957 my $ipconfigdesc = {
958 optional => 1,
959 type => 'string', format => 'pve-qm-ipconfig',
960 description => <<'EODESCR',
961 cloud-init: Specify IP addresses and gateways for the corresponding interface.
962
963 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
964
965 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit gateway should be provided.
966 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
967
968 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using dhcp on IPv4.
969 EODESCR
970 };
971 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
972
973 for (my $i = 0; $i < $MAX_NETS; $i++) {
974 $confdesc->{"net$i"} = $netdesc;
975 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
976 }
977
978 foreach my $key (keys %$confdesc_cloudinit) {
979 $confdesc->{$key} = $confdesc_cloudinit->{$key};
980 }
981
982 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
983 sub verify_volume_id_or_qm_path {
984 my ($volid, $noerr) = @_;
985
986 if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
987 return $volid;
988 }
989
990 # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
991 $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
992 if ($@) {
993 return undef if $noerr;
994 die $@;
995 }
996 return $volid;
997 }
998
999 my $drivename_hash;
1000
1001 my %drivedesc_base = (
1002 volume => { alias => 'file' },
1003 file => {
1004 type => 'string',
1005 format => 'pve-volume-id-or-qm-path',
1006 default_key => 1,
1007 format_description => 'volume',
1008 description => "The drive's backing volume.",
1009 },
1010 media => {
1011 type => 'string',
1012 enum => [qw(cdrom disk)],
1013 description => "The drive's media type.",
1014 default => 'disk',
1015 optional => 1
1016 },
1017 cyls => {
1018 type => 'integer',
1019 description => "Force the drive's physical geometry to have a specific cylinder count.",
1020 optional => 1
1021 },
1022 heads => {
1023 type => 'integer',
1024 description => "Force the drive's physical geometry to have a specific head count.",
1025 optional => 1
1026 },
1027 secs => {
1028 type => 'integer',
1029 description => "Force the drive's physical geometry to have a specific sector count.",
1030 optional => 1
1031 },
1032 trans => {
1033 type => 'string',
1034 enum => [qw(none lba auto)],
1035 description => "Force disk geometry bios translation mode.",
1036 optional => 1,
1037 },
1038 snapshot => {
1039 type => 'boolean',
1040 description => "Controls qemu's snapshot mode feature."
1041 . " If activated, changes made to the disk are temporary and will"
1042 . " be discarded when the VM is shutdown.",
1043 optional => 1,
1044 },
1045 cache => {
1046 type => 'string',
1047 enum => [qw(none writethrough writeback unsafe directsync)],
1048 description => "The drive's cache mode",
1049 optional => 1,
1050 },
1051 format => get_standard_option('pve-qm-image-format'),
1052 size => {
1053 type => 'string',
1054 format => 'disk-size',
1055 format_description => 'DiskSize',
1056 description => "Disk size. This is purely informational and has no effect.",
1057 optional => 1,
1058 },
1059 backup => {
1060 type => 'boolean',
1061 description => "Whether the drive should be included when making backups.",
1062 optional => 1,
1063 },
1064 replicate => {
1065 type => 'boolean',
1066 description => 'Whether the drive should considered for replication jobs.',
1067 optional => 1,
1068 default => 1,
1069 },
1070 rerror => {
1071 type => 'string',
1072 enum => [qw(ignore report stop)],
1073 description => 'Read error action.',
1074 optional => 1,
1075 },
1076 werror => {
1077 type => 'string',
1078 enum => [qw(enospc ignore report stop)],
1079 description => 'Write error action.',
1080 optional => 1,
1081 },
1082 aio => {
1083 type => 'string',
1084 enum => [qw(native threads)],
1085 description => 'AIO type to use.',
1086 optional => 1,
1087 },
1088 discard => {
1089 type => 'string',
1090 enum => [qw(ignore on)],
1091 description => 'Controls whether to pass discard/trim requests to the underlying storage.',
1092 optional => 1,
1093 },
1094 detect_zeroes => {
1095 type => 'boolean',
1096 description => 'Controls whether to detect and try to optimize writes of zeroes.',
1097 optional => 1,
1098 },
1099 serial => {
1100 type => 'string',
1101 format => 'urlencoded',
1102 format_description => 'serial',
1103 maxLength => 20*3, # *3 since it's %xx url enoded
1104 description => "The drive's reported serial number, url-encoded, up to 20 bytes long.",
1105 optional => 1,
1106 },
1107 shared => {
1108 type => 'boolean',
1109 description => 'Mark this locally-managed volume as available on all nodes',
1110 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!",
1111 optional => 1,
1112 default => 0,
1113 }
1114 );
1115
1116 my %iothread_fmt = ( iothread => {
1117 type => 'boolean',
1118 description => "Whether to use iothreads for this drive",
1119 optional => 1,
1120 });
1121
1122 my %model_fmt = (
1123 model => {
1124 type => 'string',
1125 format => 'urlencoded',
1126 format_description => 'model',
1127 maxLength => 40*3, # *3 since it's %xx url enoded
1128 description => "The drive's reported model name, url-encoded, up to 40 bytes long.",
1129 optional => 1,
1130 },
1131 );
1132
1133 my %queues_fmt = (
1134 queues => {
1135 type => 'integer',
1136 description => "Number of queues.",
1137 minimum => 2,
1138 optional => 1
1139 }
1140 );
1141
1142 my %scsiblock_fmt = (
1143 scsiblock => {
1144 type => 'boolean',
1145 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",
1146 optional => 1,
1147 default => 0,
1148 },
1149 );
1150
1151 my %ssd_fmt = (
1152 ssd => {
1153 type => 'boolean',
1154 description => "Whether to expose this drive as an SSD, rather than a rotational hard disk.",
1155 optional => 1,
1156 },
1157 );
1158
1159 my %wwn_fmt = (
1160 wwn => {
1161 type => 'string',
1162 pattern => qr/^(0x)[0-9a-fA-F]{16}/,
1163 format_description => 'wwn',
1164 description => "The drive's worldwide name, encoded as 16 bytes hex string, prefixed by '0x'.",
1165 optional => 1,
1166 },
1167 );
1168
1169 my $add_throttle_desc = sub {
1170 my ($key, $type, $what, $unit, $longunit, $minimum) = @_;
1171 my $d = {
1172 type => $type,
1173 format_description => $unit,
1174 description => "Maximum $what in $longunit.",
1175 optional => 1,
1176 };
1177 $d->{minimum} = $minimum if defined($minimum);
1178 $drivedesc_base{$key} = $d;
1179 };
1180 # throughput: (leaky bucket)
1181 $add_throttle_desc->('bps', 'integer', 'r/w speed', 'bps', 'bytes per second');
1182 $add_throttle_desc->('bps_rd', 'integer', 'read speed', 'bps', 'bytes per second');
1183 $add_throttle_desc->('bps_wr', 'integer', 'write speed', 'bps', 'bytes per second');
1184 $add_throttle_desc->('mbps', 'number', 'r/w speed', 'mbps', 'megabytes per second');
1185 $add_throttle_desc->('mbps_rd', 'number', 'read speed', 'mbps', 'megabytes per second');
1186 $add_throttle_desc->('mbps_wr', 'number', 'write speed', 'mbps', 'megabytes per second');
1187 $add_throttle_desc->('iops', 'integer', 'r/w I/O', 'iops', 'operations per second');
1188 $add_throttle_desc->('iops_rd', 'integer', 'read I/O', 'iops', 'operations per second');
1189 $add_throttle_desc->('iops_wr', 'integer', 'write I/O', 'iops', 'operations per second');
1190
1191 # pools: (pool of IO before throttling starts taking effect)
1192 $add_throttle_desc->('mbps_max', 'number', 'unthrottled r/w pool', 'mbps', 'megabytes per second');
1193 $add_throttle_desc->('mbps_rd_max', 'number', 'unthrottled read pool', 'mbps', 'megabytes per second');
1194 $add_throttle_desc->('mbps_wr_max', 'number', 'unthrottled write pool', 'mbps', 'megabytes per second');
1195 $add_throttle_desc->('iops_max', 'integer', 'unthrottled r/w I/O pool', 'iops', 'operations per second');
1196 $add_throttle_desc->('iops_rd_max', 'integer', 'unthrottled read I/O pool', 'iops', 'operations per second');
1197 $add_throttle_desc->('iops_wr_max', 'integer', 'unthrottled write I/O pool', 'iops', 'operations per second');
1198
1199 # burst lengths
1200 $add_throttle_desc->('bps_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1201 $add_throttle_desc->('bps_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1202 $add_throttle_desc->('bps_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1203 $add_throttle_desc->('iops_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1204 $add_throttle_desc->('iops_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1205 $add_throttle_desc->('iops_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1206
1207 # legacy support
1208 $drivedesc_base{'bps_rd_length'} = { alias => 'bps_rd_max_length' };
1209 $drivedesc_base{'bps_wr_length'} = { alias => 'bps_wr_max_length' };
1210 $drivedesc_base{'iops_rd_length'} = { alias => 'iops_rd_max_length' };
1211 $drivedesc_base{'iops_wr_length'} = { alias => 'iops_wr_max_length' };
1212
1213 my $ide_fmt = {
1214 %drivedesc_base,
1215 %model_fmt,
1216 %ssd_fmt,
1217 %wwn_fmt,
1218 };
1219 PVE::JSONSchema::register_format("pve-qm-ide", $ide_fmt);
1220
1221 my $idedesc = {
1222 optional => 1,
1223 type => 'string', format => $ide_fmt,
1224 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
1225 };
1226 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
1227
1228 my $scsi_fmt = {
1229 %drivedesc_base,
1230 %iothread_fmt,
1231 %queues_fmt,
1232 %scsiblock_fmt,
1233 %ssd_fmt,
1234 %wwn_fmt,
1235 };
1236 my $scsidesc = {
1237 optional => 1,
1238 type => 'string', format => $scsi_fmt,
1239 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
1240 };
1241 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
1242
1243 my $sata_fmt = {
1244 %drivedesc_base,
1245 %ssd_fmt,
1246 %wwn_fmt,
1247 };
1248 my $satadesc = {
1249 optional => 1,
1250 type => 'string', format => $sata_fmt,
1251 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
1252 };
1253 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
1254
1255 my $virtio_fmt = {
1256 %drivedesc_base,
1257 %iothread_fmt,
1258 };
1259 my $virtiodesc = {
1260 optional => 1,
1261 type => 'string', format => $virtio_fmt,
1262 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
1263 };
1264 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
1265
1266 my $alldrive_fmt = {
1267 %drivedesc_base,
1268 %iothread_fmt,
1269 %model_fmt,
1270 %queues_fmt,
1271 %scsiblock_fmt,
1272 %ssd_fmt,
1273 %wwn_fmt,
1274 };
1275
1276 my $efidisk_fmt = {
1277 volume => { alias => 'file' },
1278 file => {
1279 type => 'string',
1280 format => 'pve-volume-id-or-qm-path',
1281 default_key => 1,
1282 format_description => 'volume',
1283 description => "The drive's backing volume.",
1284 },
1285 format => get_standard_option('pve-qm-image-format'),
1286 size => {
1287 type => 'string',
1288 format => 'disk-size',
1289 format_description => 'DiskSize',
1290 description => "Disk size. This is purely informational and has no effect.",
1291 optional => 1,
1292 },
1293 };
1294
1295 my $efidisk_desc = {
1296 optional => 1,
1297 type => 'string', format => $efidisk_fmt,
1298 description => "Configure a Disk for storing EFI vars",
1299 };
1300
1301 PVE::JSONSchema::register_standard_option("pve-qm-efidisk", $efidisk_desc);
1302
1303 my $usb_fmt = {
1304 host => {
1305 default_key => 1,
1306 type => 'string', format => 'pve-qm-usb-device',
1307 format_description => 'HOSTUSBDEVICE|spice',
1308 description => <<EODESCR,
1309 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1310
1311 'bus-port(.port)*' (decimal numbers) or
1312 'vendor_id:product_id' (hexadeciaml numbers) or
1313 'spice'
1314
1315 You can use the 'lsusb -t' command to list existing usb devices.
1316
1317 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1318
1319 The value 'spice' can be used to add a usb redirection devices for spice.
1320 EODESCR
1321 },
1322 usb3 => {
1323 optional => 1,
1324 type => 'boolean',
1325 description => "Specifies whether if given host option is a USB3 device or port.",
1326 default => 0,
1327 },
1328 };
1329
1330 my $usbdesc = {
1331 optional => 1,
1332 type => 'string', format => $usb_fmt,
1333 description => "Configure an USB device (n is 0 to 4).",
1334 };
1335 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1336
1337 my $PCIRE = qr/[a-f0-9]{2}:[a-f0-9]{2}(?:\.[a-f0-9])?/;
1338 my $hostpci_fmt = {
1339 host => {
1340 default_key => 1,
1341 type => 'string',
1342 pattern => qr/$PCIRE(;$PCIRE)*/,
1343 format_description => 'HOSTPCIID[;HOSTPCIID2...]',
1344 description => <<EODESCR,
1345 Host PCI device pass through. The PCI ID of a host's PCI device or a list
1346 of PCI virtual functions of the host. HOSTPCIID syntax is:
1347
1348 'bus:dev.func' (hexadecimal numbers)
1349
1350 You can us the 'lspci' command to list existing PCI devices.
1351 EODESCR
1352 },
1353 rombar => {
1354 type => 'boolean',
1355 description => "Specify whether or not the device's ROM will be visible in the guest's memory map.",
1356 optional => 1,
1357 default => 1,
1358 },
1359 romfile => {
1360 type => 'string',
1361 pattern => '[^,;]+',
1362 format_description => 'string',
1363 description => "Custom pci device rom filename (must be located in /usr/share/kvm/).",
1364 optional => 1,
1365 },
1366 pcie => {
1367 type => 'boolean',
1368 description => "Choose the PCI-express bus (needs the 'q35' machine model).",
1369 optional => 1,
1370 default => 0,
1371 },
1372 'x-vga' => {
1373 type => 'boolean',
1374 description => "Enable vfio-vga device support.",
1375 optional => 1,
1376 default => 0,
1377 },
1378 'mdev' => {
1379 type => 'string',
1380 format_description => 'string',
1381 pattern => '[^/\.:]+',
1382 optional => 1,
1383 description => <<EODESCR
1384 The type of mediated device to use.
1385 An instance of this type will be created on startup of the VM and
1386 will be cleaned up when the VM stops.
1387 EODESCR
1388 }
1389 };
1390 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
1391
1392 my $hostpcidesc = {
1393 optional => 1,
1394 type => 'string', format => 'pve-qm-hostpci',
1395 description => "Map host PCI devices into guest.",
1396 verbose_description => <<EODESCR,
1397 Map host PCI devices into guest.
1398
1399 NOTE: This option allows direct access to host hardware. So it is no longer
1400 possible to migrate such machines - use with special care.
1401
1402 CAUTION: Experimental! User reported problems with this option.
1403 EODESCR
1404 };
1405 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
1406
1407 my $serialdesc = {
1408 optional => 1,
1409 type => 'string',
1410 pattern => '(/dev/.+|socket)',
1411 description => "Create a serial device inside the VM (n is 0 to 3)",
1412 verbose_description => <<EODESCR,
1413 Create a serial device inside the VM (n is 0 to 3), and pass through a
1414 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1415 host side (use 'qm terminal' to open a terminal connection).
1416
1417 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
1418
1419 CAUTION: Experimental! User reported problems with this option.
1420 EODESCR
1421 };
1422
1423 my $paralleldesc= {
1424 optional => 1,
1425 type => 'string',
1426 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1427 description => "Map host parallel devices (n is 0 to 2).",
1428 verbose_description => <<EODESCR,
1429 Map host parallel devices (n is 0 to 2).
1430
1431 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1432
1433 CAUTION: Experimental! User reported problems with this option.
1434 EODESCR
1435 };
1436
1437 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1438 $confdesc->{"parallel$i"} = $paralleldesc;
1439 }
1440
1441 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1442 $confdesc->{"serial$i"} = $serialdesc;
1443 }
1444
1445 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
1446 $confdesc->{"hostpci$i"} = $hostpcidesc;
1447 }
1448
1449 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
1450 $drivename_hash->{"ide$i"} = 1;
1451 $confdesc->{"ide$i"} = $idedesc;
1452 }
1453
1454 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
1455 $drivename_hash->{"sata$i"} = 1;
1456 $confdesc->{"sata$i"} = $satadesc;
1457 }
1458
1459 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
1460 $drivename_hash->{"scsi$i"} = 1;
1461 $confdesc->{"scsi$i"} = $scsidesc ;
1462 }
1463
1464 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
1465 $drivename_hash->{"virtio$i"} = 1;
1466 $confdesc->{"virtio$i"} = $virtiodesc;
1467 }
1468
1469 $drivename_hash->{efidisk0} = 1;
1470 $confdesc->{efidisk0} = $efidisk_desc;
1471
1472 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1473 $confdesc->{"usb$i"} = $usbdesc;
1474 }
1475
1476 my $unuseddesc = {
1477 optional => 1,
1478 type => 'string', format => 'pve-volume-id',
1479 description => "Reference to unused volumes. This is used internally, and should not be modified manually.",
1480 };
1481
1482 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
1483 $confdesc->{"unused$i"} = $unuseddesc;
1484 }
1485
1486 my $kvm_api_version = 0;
1487
1488 sub kvm_version {
1489 return $kvm_api_version if $kvm_api_version;
1490
1491 open my $fh, '<', '/dev/kvm'
1492 or return undef;
1493
1494 # 0xae00 => KVM_GET_API_VERSION
1495 $kvm_api_version = ioctl($fh, 0xae00, 0);
1496
1497 return $kvm_api_version;
1498 }
1499
1500 my $kvm_user_version = {};
1501 my $kvm_mtime = {};
1502
1503 sub kvm_user_version {
1504 my ($binary) = @_;
1505
1506 $binary //= get_command_for_arch(get_host_arch()); # get the native arch by default
1507 my $st = stat($binary);
1508
1509 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1510 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1511 $cachedmtime == $st->mtime;
1512
1513 $kvm_user_version->{$binary} = 'unknown';
1514 $kvm_mtime->{$binary} = $st->mtime;
1515
1516 my $code = sub {
1517 my $line = shift;
1518 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1519 $kvm_user_version->{$binary} = $2;
1520 }
1521 };
1522
1523 eval { run_command([$binary, '--version'], outfunc => $code); };
1524 warn $@ if $@;
1525
1526 return $kvm_user_version->{$binary};
1527
1528 }
1529
1530 sub kernel_has_vhost_net {
1531 return -c '/dev/vhost-net';
1532 }
1533
1534 sub valid_drive_names {
1535 # order is important - used to autoselect boot disk
1536 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
1537 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
1538 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
1539 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))),
1540 'efidisk0');
1541 }
1542
1543 sub is_valid_drivename {
1544 my $dev = shift;
1545
1546 return defined($drivename_hash->{$dev});
1547 }
1548
1549 sub option_exists {
1550 my $key = shift;
1551 return defined($confdesc->{$key});
1552 }
1553
1554 sub nic_models {
1555 return $nic_model_list;
1556 }
1557
1558 sub os_list_description {
1559
1560 return {
1561 other => 'Other',
1562 wxp => 'Windows XP',
1563 w2k => 'Windows 2000',
1564 w2k3 =>, 'Windows 2003',
1565 w2k8 => 'Windows 2008',
1566 wvista => 'Windows Vista',
1567 win7 => 'Windows 7',
1568 win8 => 'Windows 8/2012',
1569 win10 => 'Windows 10/2016',
1570 l24 => 'Linux 2.4',
1571 l26 => 'Linux 2.6',
1572 };
1573 }
1574
1575 my $cdrom_path;
1576
1577 sub get_cdrom_path {
1578
1579 return $cdrom_path if $cdrom_path;
1580
1581 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1582 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1583 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1584 }
1585
1586 sub get_iso_path {
1587 my ($storecfg, $vmid, $cdrom) = @_;
1588
1589 if ($cdrom eq 'cdrom') {
1590 return get_cdrom_path();
1591 } elsif ($cdrom eq 'none') {
1592 return '';
1593 } elsif ($cdrom =~ m|^/|) {
1594 return $cdrom;
1595 } else {
1596 return PVE::Storage::path($storecfg, $cdrom);
1597 }
1598 }
1599
1600 # try to convert old style file names to volume IDs
1601 sub filename_to_volume_id {
1602 my ($vmid, $file, $media) = @_;
1603
1604 if (!($file eq 'none' || $file eq 'cdrom' ||
1605 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1606
1607 return undef if $file =~ m|/|;
1608
1609 if ($media && $media eq 'cdrom') {
1610 $file = "local:iso/$file";
1611 } else {
1612 $file = "local:$vmid/$file";
1613 }
1614 }
1615
1616 return $file;
1617 }
1618
1619 sub verify_media_type {
1620 my ($opt, $vtype, $media) = @_;
1621
1622 return if !$media;
1623
1624 my $etype;
1625 if ($media eq 'disk') {
1626 $etype = 'images';
1627 } elsif ($media eq 'cdrom') {
1628 $etype = 'iso';
1629 } else {
1630 die "internal error";
1631 }
1632
1633 return if ($vtype eq $etype);
1634
1635 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1636 }
1637
1638 sub cleanup_drive_path {
1639 my ($opt, $storecfg, $drive) = @_;
1640
1641 # try to convert filesystem paths to volume IDs
1642
1643 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1644 ($drive->{file} !~ m|^/dev/.+|) &&
1645 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1646 ($drive->{file} !~ m/^\d+$/)) {
1647 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1648 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1649 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1650 verify_media_type($opt, $vtype, $drive->{media});
1651 $drive->{file} = $volid;
1652 }
1653
1654 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1655 }
1656
1657 sub parse_hotplug_features {
1658 my ($data) = @_;
1659
1660 my $res = {};
1661
1662 return $res if $data eq '0';
1663
1664 $data = $confdesc->{hotplug}->{default} if $data eq '1';
1665
1666 foreach my $feature (PVE::Tools::split_list($data)) {
1667 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1668 $res->{$1} = 1;
1669 } else {
1670 die "invalid hotplug feature '$feature'\n";
1671 }
1672 }
1673 return $res;
1674 }
1675
1676 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1677 sub pve_verify_hotplug_features {
1678 my ($value, $noerr) = @_;
1679
1680 return $value if parse_hotplug_features($value);
1681
1682 return undef if $noerr;
1683
1684 die "unable to parse hotplug option\n";
1685 }
1686
1687 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
1688 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
1689 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
1690 # [,aio=native|threads][,discard=ignore|on][,detect_zeroes=on|off]
1691 # [,iothread=on][,serial=serial][,model=model]
1692
1693 sub parse_drive {
1694 my ($key, $data) = @_;
1695
1696 my ($interface, $index);
1697
1698 if ($key =~ m/^([^\d]+)(\d+)$/) {
1699 $interface = $1;
1700 $index = $2;
1701 } else {
1702 return undef;
1703 }
1704
1705 my $desc = $key =~ /^unused\d+$/ ? $alldrive_fmt
1706 : $confdesc->{$key}->{format};
1707 if (!$desc) {
1708 warn "invalid drive key: $key\n";
1709 return undef;
1710 }
1711 my $res = eval { PVE::JSONSchema::parse_property_string($desc, $data) };
1712 return undef if !$res;
1713 $res->{interface} = $interface;
1714 $res->{index} = $index;
1715
1716 my $error = 0;
1717 foreach my $opt (qw(bps bps_rd bps_wr)) {
1718 if (my $bps = defined(delete $res->{$opt})) {
1719 if (defined($res->{"m$opt"})) {
1720 warn "both $opt and m$opt specified\n";
1721 ++$error;
1722 next;
1723 }
1724 $res->{"m$opt"} = sprintf("%.3f", $bps / (1024*1024.0));
1725 }
1726 }
1727
1728 # can't use the schema's 'requires' because of the mbps* => bps* "transforming aliases"
1729 for my $requirement (
1730 [mbps_max => 'mbps'],
1731 [mbps_rd_max => 'mbps_rd'],
1732 [mbps_wr_max => 'mbps_wr'],
1733 [miops_max => 'miops'],
1734 [miops_rd_max => 'miops_rd'],
1735 [miops_wr_max => 'miops_wr'],
1736 [bps_max_length => 'mbps_max'],
1737 [bps_rd_max_length => 'mbps_rd_max'],
1738 [bps_wr_max_length => 'mbps_wr_max'],
1739 [iops_max_length => 'iops_max'],
1740 [iops_rd_max_length => 'iops_rd_max'],
1741 [iops_wr_max_length => 'iops_wr_max']) {
1742 my ($option, $requires) = @$requirement;
1743 if ($res->{$option} && !$res->{$requires}) {
1744 warn "$option requires $requires\n";
1745 ++$error;
1746 }
1747 }
1748
1749 return undef if $error;
1750
1751 return undef if $res->{mbps_rd} && $res->{mbps};
1752 return undef if $res->{mbps_wr} && $res->{mbps};
1753 return undef if $res->{iops_rd} && $res->{iops};
1754 return undef if $res->{iops_wr} && $res->{iops};
1755
1756 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1757 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1758 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1759 return undef if $res->{interface} eq 'virtio';
1760 }
1761
1762 if (my $size = $res->{size}) {
1763 return undef if !defined($res->{size} = PVE::JSONSchema::parse_size($size));
1764 }
1765
1766 return $res;
1767 }
1768
1769 sub print_drive {
1770 my ($vmid, $drive) = @_;
1771 my $data = { %$drive };
1772 delete $data->{$_} for qw(index interface);
1773 return PVE::JSONSchema::print_property_string($data, $alldrive_fmt);
1774 }
1775
1776 sub scsi_inquiry {
1777 my($fh, $noerr) = @_;
1778
1779 my $SG_IO = 0x2285;
1780 my $SG_GET_VERSION_NUM = 0x2282;
1781
1782 my $versionbuf = "\x00" x 8;
1783 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1784 if (!$ret) {
1785 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1786 return undef;
1787 }
1788 my $version = unpack("I", $versionbuf);
1789 if ($version < 30000) {
1790 die "scsi generic interface too old\n" if !$noerr;
1791 return undef;
1792 }
1793
1794 my $buf = "\x00" x 36;
1795 my $sensebuf = "\x00" x 8;
1796 my $cmd = pack("C x3 C x1", 0x12, 36);
1797
1798 # see /usr/include/scsi/sg.h
1799 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";
1800
1801 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1802 length($sensebuf), 0, length($buf), $buf,
1803 $cmd, $sensebuf, 6000);
1804
1805 $ret = ioctl($fh, $SG_IO, $packet);
1806 if (!$ret) {
1807 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1808 return undef;
1809 }
1810
1811 my @res = unpack($sg_io_hdr_t, $packet);
1812 if ($res[17] || $res[18]) {
1813 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1814 return undef;
1815 }
1816
1817 my $res = {};
1818 (my $byte0, my $byte1, $res->{vendor},
1819 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1820
1821 $res->{removable} = $byte1 & 128 ? 1 : 0;
1822 $res->{type} = $byte0 & 31;
1823
1824 return $res;
1825 }
1826
1827 sub path_is_scsi {
1828 my ($path) = @_;
1829
1830 my $fh = IO::File->new("+<$path") || return undef;
1831 my $res = scsi_inquiry($fh, 1);
1832 close($fh);
1833
1834 return $res;
1835 }
1836
1837 sub machine_type_is_q35 {
1838 my ($conf) = @_;
1839
1840 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1841 }
1842
1843 sub print_tabletdevice_full {
1844 my ($conf, $arch) = @_;
1845
1846 my $q35 = machine_type_is_q35($conf);
1847
1848 # we use uhci for old VMs because tablet driver was buggy in older qemu
1849 my $usbbus;
1850 if (machine_type_is_q35($conf) || $arch eq 'aarch64') {
1851 $usbbus = 'ehci';
1852 } else {
1853 $usbbus = 'uhci';
1854 }
1855
1856 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1857 }
1858
1859 sub print_keyboarddevice_full {
1860 my ($conf, $arch, $machine) = @_;
1861
1862 return undef if $arch ne 'aarch64';
1863
1864 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1865 }
1866
1867 sub print_drivedevice_full {
1868 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1869
1870 my $device = '';
1871 my $maxdev = 0;
1872
1873 if ($drive->{interface} eq 'virtio') {
1874 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges, $arch, $machine_type);
1875 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1876 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1877 } elsif ($drive->{interface} eq 'scsi') {
1878
1879 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1880 my $unit = $drive->{index} % $maxdev;
1881 my $devicetype = 'hd';
1882 my $path = '';
1883 if (drive_is_cdrom($drive)) {
1884 $devicetype = 'cd';
1885 } else {
1886 if ($drive->{file} =~ m|^/|) {
1887 $path = $drive->{file};
1888 if (my $info = path_is_scsi($path)) {
1889 if ($info->{type} == 0 && $drive->{scsiblock}) {
1890 $devicetype = 'block';
1891 } elsif ($info->{type} == 1) { # tape
1892 $devicetype = 'generic';
1893 }
1894 }
1895 } else {
1896 $path = PVE::Storage::path($storecfg, $drive->{file});
1897 }
1898
1899 if($path =~ m/^iscsi\:\/\//){
1900 $devicetype = 'generic';
1901 }
1902 }
1903
1904 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1905 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1906 } else {
1907 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1908 }
1909
1910 if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1911 $device .= ",rotation_rate=1";
1912 }
1913 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1914
1915 } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1916 my $maxdev = ($drive->{interface} eq 'sata') ? $MAX_SATA_DISKS : 2;
1917 my $controller = int($drive->{index} / $maxdev);
1918 my $unit = $drive->{index} % $maxdev;
1919 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1920
1921 $device = "ide-$devicetype";
1922 if ($drive->{interface} eq 'ide') {
1923 $device .= ",bus=ide.$controller,unit=$unit";
1924 } else {
1925 $device .= ",bus=ahci$controller.$unit";
1926 }
1927 $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1928
1929 if ($devicetype eq 'hd') {
1930 if (my $model = $drive->{model}) {
1931 $model = URI::Escape::uri_unescape($model);
1932 $device .= ",model=$model";
1933 }
1934 if ($drive->{ssd}) {
1935 $device .= ",rotation_rate=1";
1936 }
1937 }
1938 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1939 } elsif ($drive->{interface} eq 'usb') {
1940 die "implement me";
1941 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1942 } else {
1943 die "unsupported interface type";
1944 }
1945
1946 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1947
1948 if (my $serial = $drive->{serial}) {
1949 $serial = URI::Escape::uri_unescape($serial);
1950 $device .= ",serial=$serial";
1951 }
1952
1953
1954 return $device;
1955 }
1956
1957 sub get_initiator_name {
1958 my $initiator;
1959
1960 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1961 while (defined(my $line = <$fh>)) {
1962 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1963 $initiator = $1;
1964 last;
1965 }
1966 $fh->close();
1967
1968 return $initiator;
1969 }
1970
1971 sub print_drive_full {
1972 my ($storecfg, $vmid, $drive) = @_;
1973
1974 my $path;
1975 my $volid = $drive->{file};
1976 my $format;
1977
1978 if (drive_is_cdrom($drive)) {
1979 $path = get_iso_path($storecfg, $vmid, $volid);
1980 } else {
1981 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1982 if ($storeid) {
1983 $path = PVE::Storage::path($storecfg, $volid);
1984 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1985 $format = qemu_img_format($scfg, $volname);
1986 } else {
1987 $path = $volid;
1988 $format = "raw";
1989 }
1990 }
1991
1992 my $opts = '';
1993 my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1994 foreach my $o (@qemu_drive_options) {
1995 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1996 }
1997
1998 # snapshot only accepts on|off
1999 if (defined($drive->{snapshot})) {
2000 my $v = $drive->{snapshot} ? 'on' : 'off';
2001 $opts .= ",snapshot=$v";
2002 }
2003
2004 foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
2005 my ($dir, $qmpname) = @$type;
2006 if (my $v = $drive->{"mbps$dir"}) {
2007 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
2008 }
2009 if (my $v = $drive->{"mbps${dir}_max"}) {
2010 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
2011 }
2012 if (my $v = $drive->{"bps${dir}_max_length"}) {
2013 $opts .= ",throttling.bps$qmpname-max-length=$v";
2014 }
2015 if (my $v = $drive->{"iops${dir}"}) {
2016 $opts .= ",throttling.iops$qmpname=$v";
2017 }
2018 if (my $v = $drive->{"iops${dir}_max"}) {
2019 $opts .= ",throttling.iops$qmpname-max=$v";
2020 }
2021 if (my $v = $drive->{"iops${dir}_max_length"}) {
2022 $opts .= ",throttling.iops$qmpname-max-length=$v";
2023 }
2024 }
2025
2026 $opts .= ",format=$format" if $format && !$drive->{format};
2027
2028 my $cache_direct = 0;
2029
2030 if (my $cache = $drive->{cache}) {
2031 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
2032 } elsif (!drive_is_cdrom($drive)) {
2033 $opts .= ",cache=none";
2034 $cache_direct = 1;
2035 }
2036
2037 # aio native works only with O_DIRECT
2038 if (!$drive->{aio}) {
2039 if($cache_direct) {
2040 $opts .= ",aio=native";
2041 } else {
2042 $opts .= ",aio=threads";
2043 }
2044 }
2045
2046 if (!drive_is_cdrom($drive)) {
2047 my $detectzeroes;
2048 if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
2049 $detectzeroes = 'off';
2050 } elsif ($drive->{discard}) {
2051 $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
2052 } else {
2053 # This used to be our default with discard not being specified:
2054 $detectzeroes = 'on';
2055 }
2056 $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
2057 }
2058
2059 my $pathinfo = $path ? "file=$path," : '';
2060
2061 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
2062 }
2063
2064 sub print_netdevice_full {
2065 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
2066
2067 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2068
2069 my $device = $net->{model};
2070 if ($net->{model} eq 'virtio') {
2071 $device = 'virtio-net-pci';
2072 };
2073
2074 my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
2075 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
2076 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
2077 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
2078 my $vectors = $net->{queues} * 2 + 2;
2079 $tmpstr .= ",vectors=$vectors,mq=on";
2080 }
2081 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
2082
2083 if ($use_old_bios_files) {
2084 my $romfile;
2085 if ($device eq 'virtio-net-pci') {
2086 $romfile = 'pxe-virtio.rom';
2087 } elsif ($device eq 'e1000') {
2088 $romfile = 'pxe-e1000.rom';
2089 } elsif ($device eq 'ne2k') {
2090 $romfile = 'pxe-ne2k_pci.rom';
2091 } elsif ($device eq 'pcnet') {
2092 $romfile = 'pxe-pcnet.rom';
2093 } elsif ($device eq 'rtl8139') {
2094 $romfile = 'pxe-rtl8139.rom';
2095 }
2096 $tmpstr .= ",romfile=$romfile" if $romfile;
2097 }
2098
2099 return $tmpstr;
2100 }
2101
2102 sub print_netdev_full {
2103 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
2104
2105 my $i = '';
2106 if ($netid =~ m/^net(\d+)$/) {
2107 $i = int($1);
2108 }
2109
2110 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
2111
2112 my $ifname = "tap${vmid}i$i";
2113
2114 # kvm uses TUNSETIFF ioctl, and that limits ifname length
2115 die "interface name '$ifname' is too long (max 15 character)\n"
2116 if length($ifname) >= 16;
2117
2118 my $vhostparam = '';
2119 if (is_native($arch)) {
2120 $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
2121 }
2122
2123 my $vmname = $conf->{name} || "vm$vmid";
2124
2125 my $netdev = "";
2126 my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
2127
2128 if ($net->{bridge}) {
2129 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
2130 } else {
2131 $netdev = "type=user,id=$netid,hostname=$vmname";
2132 }
2133
2134 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
2135
2136 return $netdev;
2137 }
2138
2139
2140 sub print_cpu_device {
2141 my ($conf, $id) = @_;
2142
2143 my $kvm = $conf->{kvm} // 1;
2144 my $cpu = $kvm ? "kvm64" : "qemu64";
2145 if (my $cputype = $conf->{cpu}) {
2146 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
2147 or die "Cannot parse cpu description: $cputype\n";
2148 $cpu = $cpuconf->{cputype};
2149 }
2150
2151 my $cores = $conf->{cores} || 1;
2152
2153 my $current_core = ($id - 1) % $cores;
2154 my $current_socket = int(($id - 1 - $current_core)/$cores);
2155
2156 return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
2157 }
2158
2159 my $vga_map = {
2160 'cirrus' => 'cirrus-vga',
2161 'std' => 'VGA',
2162 'vmware' => 'vmware-svga',
2163 'virtio' => 'virtio-vga',
2164 };
2165
2166 sub print_vga_device {
2167 my ($conf, $vga, $arch, $machine, $id, $qxlnum, $bridges) = @_;
2168
2169 my $type = $vga_map->{$vga->{type}};
2170 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
2171 $type = 'virtio-gpu';
2172 }
2173 my $vgamem_mb = $vga->{memory};
2174 if ($qxlnum) {
2175 $type = $id ? 'qxl' : 'qxl-vga';
2176 }
2177 die "no devicetype for $vga->{type}\n" if !$type;
2178
2179 my $memory = "";
2180 if ($vgamem_mb) {
2181 if ($vga->{type} eq 'virtio') {
2182 my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2183 $memory = ",max_hostmem=$bytes";
2184 } elsif ($qxlnum) {
2185 # from https://www.spice-space.org/multiple-monitors.html
2186 $memory = ",vgamem_mb=$vga->{memory}";
2187 my $ram = $vgamem_mb * 4;
2188 my $vram = $vgamem_mb * 2;
2189 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2190 } else {
2191 $memory = ",vgamem_mb=$vga->{memory}";
2192 }
2193 } elsif ($qxlnum && $id) {
2194 $memory = ",ram_size=67108864,vram_size=33554432";
2195 }
2196
2197 my $q35 = machine_type_is_q35($conf);
2198 my $vgaid = "vga" . ($id // '');
2199 my $pciaddr;
2200
2201 if ($q35 && $vgaid eq 'vga') {
2202 # the first display uses pcie.0 bus on q35 machines
2203 $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
2204 } else {
2205 $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
2206 }
2207
2208 return "$type,id=${vgaid}${memory}${pciaddr}";
2209 }
2210
2211 sub drive_is_cloudinit {
2212 my ($drive) = @_;
2213 return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2214 }
2215
2216 sub drive_is_cdrom {
2217 my ($drive, $exclude_cloudinit) = @_;
2218
2219 return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2220
2221 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2222
2223 }
2224
2225 sub parse_number_sets {
2226 my ($set) = @_;
2227 my $res = [];
2228 foreach my $part (split(/;/, $set)) {
2229 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2230 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2231 push @$res, [ $1, $2 ];
2232 } else {
2233 die "invalid range: $part\n";
2234 }
2235 }
2236 return $res;
2237 }
2238
2239 sub parse_numa {
2240 my ($data) = @_;
2241
2242 my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2243 $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2244 $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2245 return $res;
2246 }
2247
2248 sub parse_hostpci {
2249 my ($value) = @_;
2250
2251 return undef if !$value;
2252
2253 my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2254
2255 my @idlist = split(/;/, $res->{host});
2256 delete $res->{host};
2257 foreach my $id (@idlist) {
2258 if ($id =~ m/\./) { # full id 00:00.1
2259 push @{$res->{pciid}}, {
2260 id => $id,
2261 };
2262 } else { # partial id 00:00
2263 $res->{pciid} = PVE::SysFSTools::lspci($id);
2264 }
2265 }
2266 return $res;
2267 }
2268
2269 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2270 sub parse_net {
2271 my ($data) = @_;
2272
2273 my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2274 if ($@) {
2275 warn $@;
2276 return undef;
2277 }
2278 if (!defined($res->{macaddr})) {
2279 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2280 $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2281 }
2282 return $res;
2283 }
2284
2285 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2286 sub parse_ipconfig {
2287 my ($data) = @_;
2288
2289 my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2290 if ($@) {
2291 warn $@;
2292 return undef;
2293 }
2294
2295 if ($res->{gw} && !$res->{ip}) {
2296 warn 'gateway specified without specifying an IP address';
2297 return undef;
2298 }
2299 if ($res->{gw6} && !$res->{ip6}) {
2300 warn 'IPv6 gateway specified without specifying an IPv6 address';
2301 return undef;
2302 }
2303 if ($res->{gw} && $res->{ip} eq 'dhcp') {
2304 warn 'gateway specified together with DHCP';
2305 return undef;
2306 }
2307 if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2308 # gw6 + auto/dhcp
2309 warn "IPv6 gateway specified together with $res->{ip6} address";
2310 return undef;
2311 }
2312
2313 if (!$res->{ip} && !$res->{ip6}) {
2314 return { ip => 'dhcp', ip6 => 'dhcp' };
2315 }
2316
2317 return $res;
2318 }
2319
2320 sub print_net {
2321 my $net = shift;
2322
2323 return PVE::JSONSchema::print_property_string($net, $net_fmt);
2324 }
2325
2326 sub add_random_macs {
2327 my ($settings) = @_;
2328
2329 foreach my $opt (keys %$settings) {
2330 next if $opt !~ m/^net(\d+)$/;
2331 my $net = parse_net($settings->{$opt});
2332 next if !$net;
2333 $settings->{$opt} = print_net($net);
2334 }
2335 }
2336
2337 sub vm_is_volid_owner {
2338 my ($storecfg, $vmid, $volid) = @_;
2339
2340 if ($volid !~ m|^/|) {
2341 my ($path, $owner);
2342 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2343 if ($owner && ($owner == $vmid)) {
2344 return 1;
2345 }
2346 }
2347
2348 return undef;
2349 }
2350
2351 sub split_flagged_list {
2352 my $text = shift || '';
2353 $text =~ s/[,;]/ /g;
2354 $text =~ s/^\s+//;
2355 return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
2356 }
2357
2358 sub join_flagged_list {
2359 my ($how, $lst) = @_;
2360 join $how, map { $lst->{$_} . $_ } keys %$lst;
2361 }
2362
2363 sub vmconfig_delete_pending_option {
2364 my ($conf, $key, $force) = @_;
2365
2366 delete $conf->{pending}->{$key};
2367 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2368 $pending_delete_hash->{$key} = $force ? '!' : '';
2369 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2370 }
2371
2372 sub vmconfig_undelete_pending_option {
2373 my ($conf, $key) = @_;
2374
2375 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2376 delete $pending_delete_hash->{$key};
2377
2378 if (%$pending_delete_hash) {
2379 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2380 } else {
2381 delete $conf->{pending}->{delete};
2382 }
2383 }
2384
2385 sub vmconfig_register_unused_drive {
2386 my ($storecfg, $vmid, $conf, $drive) = @_;
2387
2388 if (drive_is_cloudinit($drive)) {
2389 eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2390 warn $@ if $@;
2391 } elsif (!drive_is_cdrom($drive)) {
2392 my $volid = $drive->{file};
2393 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2394 PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2395 }
2396 }
2397 }
2398
2399 sub vmconfig_cleanup_pending {
2400 my ($conf) = @_;
2401
2402 # remove pending changes when nothing changed
2403 my $changes;
2404 foreach my $opt (keys %{$conf->{pending}}) {
2405 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
2406 $changes = 1;
2407 delete $conf->{pending}->{$opt};
2408 }
2409 }
2410
2411 my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
2412 my $pending_delete_hash = {};
2413 while (my ($opt, $force) = each %$current_delete_hash) {
2414 if (defined($conf->{$opt})) {
2415 $pending_delete_hash->{$opt} = $force;
2416 } else {
2417 $changes = 1;
2418 }
2419 }
2420
2421 if (%$pending_delete_hash) {
2422 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2423 } else {
2424 delete $conf->{pending}->{delete};
2425 }
2426
2427 return $changes;
2428 }
2429
2430 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2431 my $smbios1_fmt = {
2432 uuid => {
2433 type => 'string',
2434 pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2435 format_description => 'UUID',
2436 description => "Set SMBIOS1 UUID.",
2437 optional => 1,
2438 },
2439 version => {
2440 type => 'string',
2441 pattern => '[A-Za-z0-9+\/]+={0,2}',
2442 format_description => 'Base64 encoded string',
2443 description => "Set SMBIOS1 version.",
2444 optional => 1,
2445 },
2446 serial => {
2447 type => 'string',
2448 pattern => '[A-Za-z0-9+\/]+={0,2}',
2449 format_description => 'Base64 encoded string',
2450 description => "Set SMBIOS1 serial number.",
2451 optional => 1,
2452 },
2453 manufacturer => {
2454 type => 'string',
2455 pattern => '[A-Za-z0-9+\/]+={0,2}',
2456 format_description => 'Base64 encoded string',
2457 description => "Set SMBIOS1 manufacturer.",
2458 optional => 1,
2459 },
2460 product => {
2461 type => 'string',
2462 pattern => '[A-Za-z0-9+\/]+={0,2}',
2463 format_description => 'Base64 encoded string',
2464 description => "Set SMBIOS1 product ID.",
2465 optional => 1,
2466 },
2467 sku => {
2468 type => 'string',
2469 pattern => '[A-Za-z0-9+\/]+={0,2}',
2470 format_description => 'Base64 encoded string',
2471 description => "Set SMBIOS1 SKU string.",
2472 optional => 1,
2473 },
2474 family => {
2475 type => 'string',
2476 pattern => '[A-Za-z0-9+\/]+={0,2}',
2477 format_description => 'Base64 encoded string',
2478 description => "Set SMBIOS1 family string.",
2479 optional => 1,
2480 },
2481 base64 => {
2482 type => 'boolean',
2483 description => 'Flag to indicate that the SMBIOS values are base64 encoded',
2484 optional => 1,
2485 },
2486 };
2487
2488 sub parse_smbios1 {
2489 my ($data) = @_;
2490
2491 my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2492 warn $@ if $@;
2493 return $res;
2494 }
2495
2496 sub print_smbios1 {
2497 my ($smbios1) = @_;
2498 return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2499 }
2500
2501 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2502
2503 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2504 sub verify_bootdisk {
2505 my ($value, $noerr) = @_;
2506
2507 return $value if is_valid_drivename($value);
2508
2509 return undef if $noerr;
2510
2511 die "invalid boot disk '$value'\n";
2512 }
2513
2514 sub parse_watchdog {
2515 my ($value) = @_;
2516
2517 return undef if !$value;
2518
2519 my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2520 warn $@ if $@;
2521 return $res;
2522 }
2523
2524 sub parse_guest_agent {
2525 my ($value) = @_;
2526
2527 return {} if !defined($value->{agent});
2528
2529 my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2530 warn $@ if $@;
2531
2532 # if the agent is disabled ignore the other potentially set properties
2533 return {} if !$res->{enabled};
2534 return $res;
2535 }
2536
2537 sub parse_vga {
2538 my ($value) = @_;
2539
2540 return {} if !$value;
2541 my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2542 warn $@ if $@;
2543 return $res;
2544 }
2545
2546 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2547 sub verify_usb_device {
2548 my ($value, $noerr) = @_;
2549
2550 return $value if parse_usb_device($value);
2551
2552 return undef if $noerr;
2553
2554 die "unable to parse usb device\n";
2555 }
2556
2557 # add JSON properties for create and set function
2558 sub json_config_properties {
2559 my $prop = shift;
2560
2561 foreach my $opt (keys %$confdesc) {
2562 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2563 $prop->{$opt} = $confdesc->{$opt};
2564 }
2565
2566 return $prop;
2567 }
2568
2569 # return copy of $confdesc_cloudinit to generate documentation
2570 sub cloudinit_config_properties {
2571
2572 return dclone($confdesc_cloudinit);
2573 }
2574
2575 sub check_type {
2576 my ($key, $value) = @_;
2577
2578 die "unknown setting '$key'\n" if !$confdesc->{$key};
2579
2580 my $type = $confdesc->{$key}->{type};
2581
2582 if (!defined($value)) {
2583 die "got undefined value\n";
2584 }
2585
2586 if ($value =~ m/[\n\r]/) {
2587 die "property contains a line feed\n";
2588 }
2589
2590 if ($type eq 'boolean') {
2591 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2592 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2593 die "type check ('boolean') failed - got '$value'\n";
2594 } elsif ($type eq 'integer') {
2595 return int($1) if $value =~ m/^(\d+)$/;
2596 die "type check ('integer') failed - got '$value'\n";
2597 } elsif ($type eq 'number') {
2598 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2599 die "type check ('number') failed - got '$value'\n";
2600 } elsif ($type eq 'string') {
2601 if (my $fmt = $confdesc->{$key}->{format}) {
2602 PVE::JSONSchema::check_format($fmt, $value);
2603 return $value;
2604 }
2605 $value =~ s/^\"(.*)\"$/$1/;
2606 return $value;
2607 } else {
2608 die "internal error"
2609 }
2610 }
2611
2612 sub touch_config {
2613 my ($vmid) = @_;
2614
2615 my $conf = PVE::QemuConfig->config_file($vmid);
2616 utime undef, undef, $conf;
2617 }
2618
2619 sub destroy_vm {
2620 my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2621
2622 my $conf = PVE::QemuConfig->load_config($vmid);
2623
2624 PVE::QemuConfig->check_lock($conf) if !$skiplock;
2625
2626 if ($conf->{template}) {
2627 # check if any base image is still used by a linked clone
2628 foreach_drive($conf, sub {
2629 my ($ds, $drive) = @_;
2630
2631 return if drive_is_cdrom($drive);
2632
2633 my $volid = $drive->{file};
2634
2635 return if !$volid || $volid =~ m|^/|;
2636
2637 die "base volume '$volid' is still in use by linked cloned\n"
2638 if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2639
2640 });
2641 }
2642
2643 # only remove disks owned by this VM
2644 foreach_drive($conf, sub {
2645 my ($ds, $drive) = @_;
2646
2647 return if drive_is_cdrom($drive, 1);
2648
2649 my $volid = $drive->{file};
2650
2651 return if !$volid || $volid =~ m|^/|;
2652
2653 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2654 return if !$path || !$owner || ($owner != $vmid);
2655
2656 eval {
2657 PVE::Storage::vdisk_free($storecfg, $volid);
2658 };
2659 warn "Could not remove disk '$volid', check manually: $@" if $@;
2660
2661 });
2662
2663 if ($keep_empty_config) {
2664 PVE::QemuConfig->write_config($vmid, "memory: 128\n");
2665 } else {
2666 PVE::QemuConfig->destroy_config($vmid);
2667 }
2668
2669 # also remove unused disk
2670 eval {
2671 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2672
2673 eval {
2674 PVE::Storage::foreach_volid($dl, sub {
2675 my ($volid, $sid, $volname, $d) = @_;
2676 PVE::Storage::vdisk_free($storecfg, $volid);
2677 });
2678 };
2679 warn $@ if $@;
2680
2681 };
2682 warn $@ if $@;
2683 }
2684
2685 sub parse_vm_config {
2686 my ($filename, $raw) = @_;
2687
2688 return undef if !defined($raw);
2689
2690 my $res = {
2691 digest => Digest::SHA::sha1_hex($raw),
2692 snapshots => {},
2693 pending => {},
2694 };
2695
2696 $filename =~ m|/qemu-server/(\d+)\.conf$|
2697 || die "got strange filename '$filename'";
2698
2699 my $vmid = $1;
2700
2701 my $conf = $res;
2702 my $descr;
2703 my $section = '';
2704
2705 my @lines = split(/\n/, $raw);
2706 foreach my $line (@lines) {
2707 next if $line =~ m/^\s*$/;
2708
2709 if ($line =~ m/^\[PENDING\]\s*$/i) {
2710 $section = 'pending';
2711 if (defined($descr)) {
2712 $descr =~ s/\s+$//;
2713 $conf->{description} = $descr;
2714 }
2715 $descr = undef;
2716 $conf = $res->{$section} = {};
2717 next;
2718
2719 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2720 $section = $1;
2721 if (defined($descr)) {
2722 $descr =~ s/\s+$//;
2723 $conf->{description} = $descr;
2724 }
2725 $descr = undef;
2726 $conf = $res->{snapshots}->{$section} = {};
2727 next;
2728 }
2729
2730 if ($line =~ m/^\#(.*)\s*$/) {
2731 $descr = '' if !defined($descr);
2732 $descr .= PVE::Tools::decode_text($1) . "\n";
2733 next;
2734 }
2735
2736 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2737 $descr = '' if !defined($descr);
2738 $descr .= PVE::Tools::decode_text($2);
2739 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2740 $conf->{snapstate} = $1;
2741 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2742 my $key = $1;
2743 my $value = $2;
2744 $conf->{$key} = $value;
2745 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2746 my $value = $1;
2747 if ($section eq 'pending') {
2748 $conf->{delete} = $value; # we parse this later
2749 } else {
2750 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2751 }
2752 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2753 my $key = $1;
2754 my $value = $2;
2755 eval { $value = check_type($key, $value); };
2756 if ($@) {
2757 warn "vm $vmid - unable to parse value of '$key' - $@";
2758 } else {
2759 $key = 'ide2' if $key eq 'cdrom';
2760 my $fmt = $confdesc->{$key}->{format};
2761 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2762 my $v = parse_drive($key, $value);
2763 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2764 $v->{file} = $volid;
2765 $value = print_drive($vmid, $v);
2766 } else {
2767 warn "vm $vmid - unable to parse value of '$key'\n";
2768 next;
2769 }
2770 }
2771
2772 $conf->{$key} = $value;
2773 }
2774 }
2775 }
2776
2777 if (defined($descr)) {
2778 $descr =~ s/\s+$//;
2779 $conf->{description} = $descr;
2780 }
2781 delete $res->{snapstate}; # just to be sure
2782
2783 return $res;
2784 }
2785
2786 sub write_vm_config {
2787 my ($filename, $conf) = @_;
2788
2789 delete $conf->{snapstate}; # just to be sure
2790
2791 if ($conf->{cdrom}) {
2792 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2793 $conf->{ide2} = $conf->{cdrom};
2794 delete $conf->{cdrom};
2795 }
2796
2797 # we do not use 'smp' any longer
2798 if ($conf->{sockets}) {
2799 delete $conf->{smp};
2800 } elsif ($conf->{smp}) {
2801 $conf->{sockets} = $conf->{smp};
2802 delete $conf->{cores};
2803 delete $conf->{smp};
2804 }
2805
2806 my $used_volids = {};
2807
2808 my $cleanup_config = sub {
2809 my ($cref, $pending, $snapname) = @_;
2810
2811 foreach my $key (keys %$cref) {
2812 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2813 $key eq 'snapstate' || $key eq 'pending';
2814 my $value = $cref->{$key};
2815 if ($key eq 'delete') {
2816 die "propertry 'delete' is only allowed in [PENDING]\n"
2817 if !$pending;
2818 # fixme: check syntax?
2819 next;
2820 }
2821 eval { $value = check_type($key, $value); };
2822 die "unable to parse value of '$key' - $@" if $@;
2823
2824 $cref->{$key} = $value;
2825
2826 if (!$snapname && is_valid_drivename($key)) {
2827 my $drive = parse_drive($key, $value);
2828 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2829 }
2830 }
2831 };
2832
2833 &$cleanup_config($conf);
2834
2835 &$cleanup_config($conf->{pending}, 1);
2836
2837 foreach my $snapname (keys %{$conf->{snapshots}}) {
2838 die "internal error" if $snapname eq 'pending';
2839 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2840 }
2841
2842 # remove 'unusedX' settings if we re-add a volume
2843 foreach my $key (keys %$conf) {
2844 my $value = $conf->{$key};
2845 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2846 delete $conf->{$key};
2847 }
2848 }
2849
2850 my $generate_raw_config = sub {
2851 my ($conf, $pending) = @_;
2852
2853 my $raw = '';
2854
2855 # add description as comment to top of file
2856 if (defined(my $descr = $conf->{description})) {
2857 if ($descr) {
2858 foreach my $cl (split(/\n/, $descr)) {
2859 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2860 }
2861 } else {
2862 $raw .= "#\n" if $pending;
2863 }
2864 }
2865
2866 foreach my $key (sort keys %$conf) {
2867 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2868 $raw .= "$key: $conf->{$key}\n";
2869 }
2870 return $raw;
2871 };
2872
2873 my $raw = &$generate_raw_config($conf);
2874
2875 if (scalar(keys %{$conf->{pending}})){
2876 $raw .= "\n[PENDING]\n";
2877 $raw .= &$generate_raw_config($conf->{pending}, 1);
2878 }
2879
2880 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2881 $raw .= "\n[$snapname]\n";
2882 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2883 }
2884
2885 return $raw;
2886 }
2887
2888 sub load_defaults {
2889
2890 my $res = {};
2891
2892 # we use static defaults from our JSON schema configuration
2893 foreach my $key (keys %$confdesc) {
2894 if (defined(my $default = $confdesc->{$key}->{default})) {
2895 $res->{$key} = $default;
2896 }
2897 }
2898
2899 return $res;
2900 }
2901
2902 sub config_list {
2903 my $vmlist = PVE::Cluster::get_vmlist();
2904 my $res = {};
2905 return $res if !$vmlist || !$vmlist->{ids};
2906 my $ids = $vmlist->{ids};
2907
2908 foreach my $vmid (keys %$ids) {
2909 my $d = $ids->{$vmid};
2910 next if !$d->{node} || $d->{node} ne $nodename;
2911 next if !$d->{type} || $d->{type} ne 'qemu';
2912 $res->{$vmid}->{exists} = 1;
2913 }
2914 return $res;
2915 }
2916
2917 # test if VM uses local resources (to prevent migration)
2918 sub check_local_resources {
2919 my ($conf, $noerr) = @_;
2920
2921 my @loc_res = ();
2922
2923 push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2924 push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2925
2926 push @loc_res, "ivshmem" if $conf->{ivshmem};
2927
2928 foreach my $k (keys %$conf) {
2929 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2930 # sockets are safe: they will recreated be on the target side post-migrate
2931 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2932 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2933 }
2934
2935 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2936
2937 return \@loc_res;
2938 }
2939
2940 # check if used storages are available on all nodes (use by migrate)
2941 sub check_storage_availability {
2942 my ($storecfg, $conf, $node) = @_;
2943
2944 foreach_drive($conf, sub {
2945 my ($ds, $drive) = @_;
2946
2947 my $volid = $drive->{file};
2948 return if !$volid;
2949
2950 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2951 return if !$sid;
2952
2953 # check if storage is available on both nodes
2954 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2955 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2956 });
2957 }
2958
2959 # list nodes where all VM images are available (used by has_feature API)
2960 sub shared_nodes {
2961 my ($conf, $storecfg) = @_;
2962
2963 my $nodelist = PVE::Cluster::get_nodelist();
2964 my $nodehash = { map { $_ => 1 } @$nodelist };
2965 my $nodename = PVE::INotify::nodename();
2966
2967 foreach_drive($conf, sub {
2968 my ($ds, $drive) = @_;
2969
2970 my $volid = $drive->{file};
2971 return if !$volid;
2972
2973 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2974 if ($storeid) {
2975 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2976 if ($scfg->{disable}) {
2977 $nodehash = {};
2978 } elsif (my $avail = $scfg->{nodes}) {
2979 foreach my $node (keys %$nodehash) {
2980 delete $nodehash->{$node} if !$avail->{$node};
2981 }
2982 } elsif (!$scfg->{shared}) {
2983 foreach my $node (keys %$nodehash) {
2984 delete $nodehash->{$node} if $node ne $nodename
2985 }
2986 }
2987 }
2988 });
2989
2990 return $nodehash
2991 }
2992
2993 sub check_local_storage_availability {
2994 my ($conf, $storecfg) = @_;
2995
2996 my $nodelist = PVE::Cluster::get_nodelist();
2997 my $nodehash = { map { $_ => {} } @$nodelist };
2998
2999 foreach_drive($conf, sub {
3000 my ($ds, $drive) = @_;
3001
3002 my $volid = $drive->{file};
3003 return if !$volid;
3004
3005 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
3006 if ($storeid) {
3007 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3008
3009 if ($scfg->{disable}) {
3010 foreach my $node (keys %$nodehash) {
3011 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3012 }
3013 } elsif (my $avail = $scfg->{nodes}) {
3014 foreach my $node (keys %$nodehash) {
3015 if (!$avail->{$node}) {
3016 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3017 }
3018 }
3019 }
3020 }
3021 });
3022
3023 foreach my $node (values %$nodehash) {
3024 if (my $unavail = $node->{unavailable_storages}) {
3025 $node->{unavailable_storages} = [ sort keys %$unavail ];
3026 }
3027 }
3028
3029 return $nodehash
3030 }
3031
3032 sub check_cmdline {
3033 my ($pidfile, $pid) = @_;
3034
3035 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
3036 if (defined($fh)) {
3037 my $line = <$fh>;
3038 $fh->close;
3039 return undef if !$line;
3040 my @param = split(/\0/, $line);
3041
3042 my $cmd = $param[0];
3043 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
3044
3045 for (my $i = 0; $i < scalar (@param); $i++) {
3046 my $p = $param[$i];
3047 next if !$p;
3048 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
3049 my $p = $param[$i+1];
3050 return 1 if $p && ($p eq $pidfile);
3051 return undef;
3052 }
3053 }
3054 }
3055 return undef;
3056 }
3057
3058 sub check_running {
3059 my ($vmid, $nocheck, $node) = @_;
3060
3061 my $filename = PVE::QemuConfig->config_file($vmid, $node);
3062
3063 die "unable to find configuration file for VM $vmid - no such machine\n"
3064 if !$nocheck && ! -f $filename;
3065
3066 my $pidfile = pidfile_name($vmid);
3067
3068 if (my $fd = IO::File->new("<$pidfile")) {
3069 my $st = stat($fd);
3070 my $line = <$fd>;
3071 close($fd);
3072
3073 my $mtime = $st->mtime;
3074 if ($mtime > time()) {
3075 warn "file '$filename' modified in future\n";
3076 }
3077
3078 if ($line =~ m/^(\d+)$/) {
3079 my $pid = $1;
3080 if (check_cmdline($pidfile, $pid)) {
3081 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
3082 return $pid;
3083 }
3084 }
3085 }
3086 }
3087
3088 return undef;
3089 }
3090
3091 sub vzlist {
3092
3093 my $vzlist = config_list();
3094
3095 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
3096
3097 while (defined(my $de = $fd->read)) {
3098 next if $de !~ m/^(\d+)\.pid$/;
3099 my $vmid = $1;
3100 next if !defined($vzlist->{$vmid});
3101 if (my $pid = check_running($vmid)) {
3102 $vzlist->{$vmid}->{pid} = $pid;
3103 }
3104 }
3105
3106 return $vzlist;
3107 }
3108
3109 sub disksize {
3110 my ($storecfg, $conf) = @_;
3111
3112 my $bootdisk = $conf->{bootdisk};
3113 return undef if !$bootdisk;
3114 return undef if !is_valid_drivename($bootdisk);
3115
3116 return undef if !$conf->{$bootdisk};
3117
3118 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
3119 return undef if !defined($drive);
3120
3121 return undef if drive_is_cdrom($drive);
3122
3123 my $volid = $drive->{file};
3124 return undef if !$volid;
3125
3126 return $drive->{size};
3127 }
3128
3129 our $vmstatus_return_properties = {
3130 vmid => get_standard_option('pve-vmid'),
3131 status => {
3132 description => "Qemu process status.",
3133 type => 'string',
3134 enum => ['stopped', 'running'],
3135 },
3136 maxmem => {
3137 description => "Maximum memory in bytes.",
3138 type => 'integer',
3139 optional => 1,
3140 renderer => 'bytes',
3141 },
3142 maxdisk => {
3143 description => "Root disk size in bytes.",
3144 type => 'integer',
3145 optional => 1,
3146 renderer => 'bytes',
3147 },
3148 name => {
3149 description => "VM name.",
3150 type => 'string',
3151 optional => 1,
3152 },
3153 qmpstatus => {
3154 description => "Qemu QMP agent status.",
3155 type => 'string',
3156 optional => 1,
3157 },
3158 pid => {
3159 description => "PID of running qemu process.",
3160 type => 'integer',
3161 optional => 1,
3162 },
3163 uptime => {
3164 description => "Uptime.",
3165 type => 'integer',
3166 optional => 1,
3167 renderer => 'duration',
3168 },
3169 cpus => {
3170 description => "Maximum usable CPUs.",
3171 type => 'number',
3172 optional => 1,
3173 },
3174 lock => {
3175 description => "The current config lock, if any.",
3176 type => 'string',
3177 optional => 1,
3178 }
3179 };
3180
3181 my $last_proc_pid_stat;
3182
3183 # get VM status information
3184 # This must be fast and should not block ($full == false)
3185 # We only query KVM using QMP if $full == true (this can be slow)
3186 sub vmstatus {
3187 my ($opt_vmid, $full) = @_;
3188
3189 my $res = {};
3190
3191 my $storecfg = PVE::Storage::config();
3192
3193 my $list = vzlist();
3194 my $defaults = load_defaults();
3195
3196 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
3197
3198 my $cpucount = $cpuinfo->{cpus} || 1;
3199
3200 foreach my $vmid (keys %$list) {
3201 next if $opt_vmid && ($vmid ne $opt_vmid);
3202
3203 my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
3204 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
3205
3206 my $d = { vmid => $vmid };
3207 $d->{pid} = $list->{$vmid}->{pid};
3208
3209 # fixme: better status?
3210 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3211
3212 my $size = disksize($storecfg, $conf);
3213 if (defined($size)) {
3214 $d->{disk} = 0; # no info available
3215 $d->{maxdisk} = $size;
3216 } else {
3217 $d->{disk} = 0;
3218 $d->{maxdisk} = 0;
3219 }
3220
3221 $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3222 * ($conf->{cores} || $defaults->{cores});
3223 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3224 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3225
3226 $d->{name} = $conf->{name} || "VM $vmid";
3227 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3228 : $defaults->{memory}*(1024*1024);
3229
3230 if ($conf->{balloon}) {
3231 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3232 $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3233 : $defaults->{shares};
3234 }
3235
3236 $d->{uptime} = 0;
3237 $d->{cpu} = 0;
3238 $d->{mem} = 0;
3239
3240 $d->{netout} = 0;
3241 $d->{netin} = 0;
3242
3243 $d->{diskread} = 0;
3244 $d->{diskwrite} = 0;
3245
3246 $d->{template} = PVE::QemuConfig->is_template($conf);
3247
3248 $d->{serial} = 1 if conf_has_serial($conf);
3249 $d->{lock} = $conf->{lock} if $conf->{lock};
3250
3251 $res->{$vmid} = $d;
3252 }
3253
3254 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3255 foreach my $dev (keys %$netdev) {
3256 next if $dev !~ m/^tap([1-9]\d*)i/;
3257 my $vmid = $1;
3258 my $d = $res->{$vmid};
3259 next if !$d;
3260
3261 $d->{netout} += $netdev->{$dev}->{receive};
3262 $d->{netin} += $netdev->{$dev}->{transmit};
3263
3264 if ($full) {
3265 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3266 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3267 }
3268
3269 }
3270
3271 my $ctime = gettimeofday;
3272
3273 foreach my $vmid (keys %$list) {
3274
3275 my $d = $res->{$vmid};
3276 my $pid = $d->{pid};
3277 next if !$pid;
3278
3279 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3280 next if !$pstat; # not running
3281
3282 my $used = $pstat->{utime} + $pstat->{stime};
3283
3284 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3285
3286 if ($pstat->{vsize}) {
3287 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3288 }
3289
3290 my $old = $last_proc_pid_stat->{$pid};
3291 if (!$old) {
3292 $last_proc_pid_stat->{$pid} = {
3293 time => $ctime,
3294 used => $used,
3295 cpu => 0,
3296 };
3297 next;
3298 }
3299
3300 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3301
3302 if ($dtime > 1000) {
3303 my $dutime = $used - $old->{used};
3304
3305 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3306 $last_proc_pid_stat->{$pid} = {
3307 time => $ctime,
3308 used => $used,
3309 cpu => $d->{cpu},
3310 };
3311 } else {
3312 $d->{cpu} = $old->{cpu};
3313 }
3314 }
3315
3316 return $res if !$full;
3317
3318 my $qmpclient = PVE::QMPClient->new();
3319
3320 my $ballooncb = sub {
3321 my ($vmid, $resp) = @_;
3322
3323 my $info = $resp->{'return'};
3324 return if !$info->{max_mem};
3325
3326 my $d = $res->{$vmid};
3327
3328 # use memory assigned to VM
3329 $d->{maxmem} = $info->{max_mem};
3330 $d->{balloon} = $info->{actual};
3331
3332 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3333 $d->{mem} = $info->{total_mem} - $info->{free_mem};
3334 $d->{freemem} = $info->{free_mem};
3335 }
3336
3337 $d->{ballooninfo} = $info;
3338 };
3339
3340 my $blockstatscb = sub {
3341 my ($vmid, $resp) = @_;
3342 my $data = $resp->{'return'} || [];
3343 my $totalrdbytes = 0;
3344 my $totalwrbytes = 0;
3345
3346 for my $blockstat (@$data) {
3347 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3348 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3349
3350 $blockstat->{device} =~ s/drive-//;
3351 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3352 }
3353 $res->{$vmid}->{diskread} = $totalrdbytes;
3354 $res->{$vmid}->{diskwrite} = $totalwrbytes;
3355 };
3356
3357 my $statuscb = sub {
3358 my ($vmid, $resp) = @_;
3359
3360 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3361 # this fails if ballon driver is not loaded, so this must be
3362 # the last commnand (following command are aborted if this fails).
3363 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3364
3365 my $status = 'unknown';
3366 if (!defined($status = $resp->{'return'}->{status})) {
3367 warn "unable to get VM status\n";
3368 return;
3369 }
3370
3371 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3372 };
3373
3374 foreach my $vmid (keys %$list) {
3375 next if $opt_vmid && ($vmid ne $opt_vmid);
3376 next if !$res->{$vmid}->{pid}; # not running
3377 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3378 }
3379
3380 $qmpclient->queue_execute(undef, 2);
3381
3382 foreach my $vmid (keys %$list) {
3383 next if $opt_vmid && ($vmid ne $opt_vmid);
3384 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3385 }
3386
3387 return $res;
3388 }
3389
3390 sub foreach_drive {
3391 my ($conf, $func, @param) = @_;
3392
3393 foreach my $ds (valid_drive_names()) {
3394 next if !defined($conf->{$ds});
3395
3396 my $drive = parse_drive($ds, $conf->{$ds});
3397 next if !$drive;
3398
3399 &$func($ds, $drive, @param);
3400 }
3401 }
3402
3403 sub foreach_volid {
3404 my ($conf, $func, @param) = @_;
3405
3406 my $volhash = {};
3407
3408 my $test_volid = sub {
3409 my ($volid, $is_cdrom, $replicate, $shared, $snapname, $size) = @_;
3410
3411 return if !$volid;
3412
3413 $volhash->{$volid}->{cdrom} //= 1;
3414 $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3415
3416 $volhash->{$volid}->{replicate} //= 0;
3417 $volhash->{$volid}->{replicate} = 1 if $replicate;
3418
3419 $volhash->{$volid}->{shared} //= 0;
3420 $volhash->{$volid}->{shared} = 1 if $shared;
3421
3422 $volhash->{$volid}->{referenced_in_config} //= 0;
3423 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3424
3425 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3426 if defined($snapname);
3427 $volhash->{$volid}->{size} = $size if $size;
3428 };
3429
3430 foreach_drive($conf, sub {
3431 my ($ds, $drive) = @_;
3432 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef, $drive->{size});
3433 });
3434
3435 foreach my $snapname (keys %{$conf->{snapshots}}) {
3436 my $snap = $conf->{snapshots}->{$snapname};
3437 $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3438 foreach_drive($snap, sub {
3439 my ($ds, $drive) = @_;
3440 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3441 });
3442 }
3443
3444 foreach my $volid (keys %$volhash) {
3445 &$func($volid, $volhash->{$volid}, @param);
3446 }
3447 }
3448
3449 sub conf_has_serial {
3450 my ($conf) = @_;
3451
3452 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3453 if ($conf->{"serial$i"}) {
3454 return 1;
3455 }
3456 }
3457
3458 return 0;
3459 }
3460
3461 sub conf_has_audio {
3462 my ($conf, $id) = @_;
3463
3464 $id //= 0;
3465 my $audio = $conf->{"audio$id"};
3466 return undef if !defined($audio);
3467
3468 my $audioproperties = PVE::JSONSchema::parse_property_string($audio_fmt, $audio);
3469 my $audiodriver = $audioproperties->{driver} // 'spice';
3470
3471 return {
3472 dev => $audioproperties->{device},
3473 dev_id => "audiodev$id",
3474 backend => $audiodriver,
3475 backend_id => "$audiodriver-backend${id}",
3476 };
3477 }
3478
3479 sub vga_conf_has_spice {
3480 my ($vga) = @_;
3481
3482 my $vgaconf = parse_vga($vga);
3483 my $vgatype = $vgaconf->{type};
3484 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3485
3486 return $1 || 1;
3487 }
3488
3489 my $host_arch; # FIXME: fix PVE::Tools::get_host_arch
3490 sub get_host_arch() {
3491 $host_arch = (POSIX::uname())[4] if !$host_arch;
3492 return $host_arch;
3493 }
3494
3495 sub is_native($) {
3496 my ($arch) = @_;
3497 return get_host_arch() eq $arch;
3498 }
3499
3500 my $default_machines = {
3501 x86_64 => 'pc',
3502 aarch64 => 'virt',
3503 };
3504
3505 sub get_basic_machine_info {
3506 my ($conf, $forcemachine) = @_;
3507
3508 my $arch = $conf->{arch} // get_host_arch();
3509 my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3510 return ($arch, $machine);
3511 }
3512
3513 sub get_ovmf_files($) {
3514 my ($arch) = @_;
3515
3516 my $ovmf = $OVMF->{$arch}
3517 or die "no OVMF images known for architecture '$arch'\n";
3518
3519 return @$ovmf;
3520 }
3521
3522 my $Arch2Qemu = {
3523 aarch64 => '/usr/bin/qemu-system-aarch64',
3524 x86_64 => '/usr/bin/qemu-system-x86_64',
3525 };
3526 sub get_command_for_arch($) {
3527 my ($arch) = @_;
3528 return '/usr/bin/kvm' if is_native($arch);
3529
3530 my $cmd = $Arch2Qemu->{$arch}
3531 or die "don't know how to emulate architecture '$arch'\n";
3532 return $cmd;
3533 }
3534
3535 sub get_cpu_options {
3536 my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3537
3538 my $cpuFlags = [];
3539 my $ostype = $conf->{ostype};
3540
3541 my $cpu = $kvm ? "kvm64" : "qemu64";
3542 if ($arch eq 'aarch64') {
3543 $cpu = 'cortex-a57';
3544 }
3545 my $hv_vendor_id;
3546 if (my $cputype = $conf->{cpu}) {
3547 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3548 or die "Cannot parse cpu description: $cputype\n";
3549 $cpu = $cpuconf->{cputype};
3550 $kvm_off = 1 if $cpuconf->{hidden};
3551 $hv_vendor_id = $cpuconf->{'hv-vendor-id'};
3552
3553 if (defined(my $flags = $cpuconf->{flags})) {
3554 push @$cpuFlags, split(";", $flags);
3555 }
3556 }
3557
3558 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64' && $arch eq 'x86_64';
3559
3560 push @$cpuFlags , '-x2apic'
3561 if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3562
3563 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3564
3565 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3566
3567 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3) && $arch eq 'x86_64') {
3568
3569 push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3570 push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3571 }
3572
3573 add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough, $hv_vendor_id) if $kvm;
3574
3575 push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm && $arch eq 'x86_64';
3576
3577 push @$cpuFlags, 'kvm=off' if $kvm_off;
3578
3579 if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3580 push @$cpuFlags, "vendor=${cpu_vendor}"
3581 if $cpu_vendor ne 'default';
3582 } elsif ($arch ne 'aarch64') {
3583 die "internal error"; # should not happen
3584 }
3585
3586 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3587
3588 return ('-cpu', $cpu);
3589 }
3590
3591 sub config_to_command {
3592 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3593
3594 my $cmd = [];
3595 my $globalFlags = [];
3596 my $machineFlags = [];
3597 my $rtcFlags = [];
3598 my $devices = [];
3599 my $pciaddr = '';
3600 my $bridges = {};
3601 my $vernum = 0; # unknown
3602 my $ostype = $conf->{ostype};
3603 my $winversion = windows_version($ostype);
3604 my $kvm = $conf->{kvm};
3605
3606 my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3607 my $kvm_binary = get_command_for_arch($arch);
3608 my $kvmver = kvm_user_version($kvm_binary);
3609 $kvm //= 1 if is_native($arch);
3610
3611 if ($kvm) {
3612 die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3613 if !defined kvm_version();
3614 }
3615
3616 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3617 $vernum = $1*1000000+$2*1000;
3618 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3619 $vernum = $1*1000000+$2*1000+$3;
3620 }
3621
3622 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3623
3624 my $q35 = machine_type_is_q35($conf);
3625 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3626 my $use_old_bios_files = undef;
3627 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3628
3629 my $cpuunits = defined($conf->{cpuunits}) ?
3630 $conf->{cpuunits} : $defaults->{cpuunits};
3631
3632 push @$cmd, $kvm_binary;
3633
3634 push @$cmd, '-id', $vmid;
3635
3636 my $vmname = $conf->{name} || "vm$vmid";
3637
3638 push @$cmd, '-name', $vmname;
3639
3640 my $use_virtio = 0;
3641
3642 my $qmpsocket = qmp_socket($vmid);
3643 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3644 push @$cmd, '-mon', "chardev=qmp,mode=control";
3645
3646 if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3647 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3648 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3649 }
3650
3651 push @$cmd, '-pidfile' , pidfile_name($vmid);
3652
3653 push @$cmd, '-daemonize';
3654
3655 if ($conf->{smbios1}) {
3656 my $smbios_conf = parse_smbios1($conf->{smbios1});
3657 if ($smbios_conf->{base64}) {
3658 # Do not pass base64 flag to qemu
3659 delete $smbios_conf->{base64};
3660 my $smbios_string = "";
3661 foreach my $key (keys %$smbios_conf) {
3662 my $value;
3663 if ($key eq "uuid") {
3664 $value = $smbios_conf->{uuid}
3665 } else {
3666 $value = decode_base64($smbios_conf->{$key});
3667 }
3668 # qemu accepts any binary data, only commas need escaping by double comma
3669 $value =~ s/,/,,/g;
3670 $smbios_string .= "," . $key . "=" . $value if $value;
3671 }
3672 push @$cmd, '-smbios', "type=1" . $smbios_string;
3673 } else {
3674 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3675 }
3676 }
3677
3678 if ($conf->{vmgenid}) {
3679 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3680 }
3681
3682 my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3683 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3684 die "uefi base image not found\n" if ! -f $ovmf_code;
3685
3686 my $path;
3687 my $format;
3688 if (my $efidisk = $conf->{efidisk0}) {
3689 my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3690 my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3691 $format = $d->{format};
3692 if ($storeid) {
3693 $path = PVE::Storage::path($storecfg, $d->{file});
3694 if (!defined($format)) {
3695 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3696 $format = qemu_img_format($scfg, $volname);
3697 }
3698 } else {
3699 $path = $d->{file};
3700 die "efidisk format must be specified\n"
3701 if !defined($format);
3702 }
3703 } else {
3704 warn "no efidisk configured! Using temporary efivars disk.\n";
3705 $path = "/tmp/$vmid-ovmf.fd";
3706 PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3707 $format = 'raw';
3708 }
3709
3710 push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3711 push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3712 }
3713
3714 # load q35 config
3715 if ($q35) {
3716 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3717 if (qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0)) {
3718 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3719 } else {
3720 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3721 }
3722 }
3723
3724 # add usb controllers
3725 my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3726 push @$devices, @usbcontrollers if @usbcontrollers;
3727 my $vga = parse_vga($conf->{vga});
3728
3729 my $qxlnum = vga_conf_has_spice($conf->{vga});
3730 $vga->{type} = 'qxl' if $qxlnum;
3731
3732 if (!$vga->{type}) {
3733 if ($arch eq 'aarch64') {
3734 $vga->{type} = 'virtio';
3735 } elsif (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3736 $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3737 } else {
3738 $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3739 }
3740 }
3741
3742 # enable absolute mouse coordinates (needed by vnc)
3743 my $tablet;
3744 if (defined($conf->{tablet})) {
3745 $tablet = $conf->{tablet};
3746 } else {
3747 $tablet = $defaults->{tablet};
3748 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3749 $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3750 }
3751
3752 if ($tablet) {
3753 push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3754 my $kbd = print_keyboarddevice_full($conf, $arch);
3755 push @$devices, '-device', $kbd if defined($kbd);
3756 }
3757
3758 my $kvm_off = 0;
3759 my $gpu_passthrough;
3760
3761 # host pci devices
3762 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3763 my $id = "hostpci$i";
3764 my $d = parse_hostpci($conf->{$id});
3765 next if !$d;
3766
3767 if (my $pcie = $d->{pcie}) {
3768 die "q35 machine model is not enabled" if !$q35;
3769 # win7 wants to have the pcie devices directly on the pcie bus
3770 # instead of in the root port
3771 if ($winversion == 7) {
3772 $pciaddr = print_pcie_addr("${id}bus0");
3773 } else {
3774 # add more root ports if needed, 4 are present by default
3775 # by pve-q35 cfgs, rest added here on demand.
3776 if ($i > 3) {
3777 push @$devices, '-device', print_pcie_root_port($i);
3778 }
3779 $pciaddr = print_pcie_addr($id);
3780 }
3781 } else {
3782 $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3783 }
3784
3785 my $xvga = '';
3786 if ($d->{'x-vga'}) {
3787 $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3788 $kvm_off = 1;
3789 $vga->{type} = 'none' if !defined($conf->{vga});
3790 $gpu_passthrough = 1;
3791 }
3792
3793 my $pcidevices = $d->{pciid};
3794 my $multifunction = 1 if @$pcidevices > 1;
3795
3796 my $sysfspath;
3797 if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3798 my $pci_id = $pcidevices->[0]->{id};
3799 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3800 $sysfspath = "/sys/bus/pci/devices/0000:$pci_id/$uuid";
3801 } elsif ($d->{mdev}) {
3802 warn "ignoring mediated device '$id' with multifunction device\n";
3803 }
3804
3805 my $j=0;
3806 foreach my $pcidevice (@$pcidevices) {
3807 my $devicestr = "vfio-pci";
3808
3809 if ($sysfspath) {
3810 $devicestr .= ",sysfsdev=$sysfspath";
3811 } else {
3812 $devicestr .= ",host=$pcidevice->{id}";
3813 }
3814
3815 my $mf_addr = $multifunction ? ".$j" : '';
3816 $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3817
3818 if ($j == 0) {
3819 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3820 $devicestr .= "$xvga";
3821 $devicestr .= ",multifunction=on" if $multifunction;
3822 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3823 }
3824
3825 push @$devices, '-device', $devicestr;
3826 $j++;
3827 }
3828 }
3829
3830 # usb devices
3831 my $usb_dev_features = {};
3832 $usb_dev_features->{spice_usb3} = 1 if qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0);
3833
3834 my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3835 push @$devices, @usbdevices if @usbdevices;
3836 # serial devices
3837 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3838 if (my $path = $conf->{"serial$i"}) {
3839 if ($path eq 'socket') {
3840 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3841 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3842 # On aarch64, serial0 is the UART device. Qemu only allows
3843 # connecting UART devices via the '-serial' command line, as
3844 # the device has a fixed slot on the hardware...
3845 if ($arch eq 'aarch64' && $i == 0) {
3846 push @$devices, '-serial', "chardev:serial$i";
3847 } else {
3848 push @$devices, '-device', "isa-serial,chardev=serial$i";
3849 }
3850 } else {
3851 die "no such serial device\n" if ! -c $path;
3852 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3853 push @$devices, '-device', "isa-serial,chardev=serial$i";
3854 }
3855 }
3856 }
3857
3858 # parallel devices
3859 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3860 if (my $path = $conf->{"parallel$i"}) {
3861 die "no such parallel device\n" if ! -c $path;
3862 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3863 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3864 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3865 }
3866 }
3867
3868 if (my $audio = conf_has_audio($conf)) {
3869
3870 my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3871
3872 my $id = $audio->{dev_id};
3873 if ($audio->{dev} eq 'AC97') {
3874 push @$devices, '-device', "AC97,id=${id}${audiopciaddr}";
3875 } elsif ($audio->{dev} =~ /intel\-hda$/) {
3876 push @$devices, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3877 push @$devices, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0";
3878 push @$devices, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1";
3879 } else {
3880 die "unkown audio device '$audio->{dev}', implement me!";
3881 }
3882
3883 push @$devices, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3884 }
3885
3886 my $sockets = 1;
3887 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3888 $sockets = $conf->{sockets} if $conf->{sockets};
3889
3890 my $cores = $conf->{cores} || 1;
3891
3892 my $maxcpus = $sockets * $cores;
3893
3894 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3895
3896 my $allowed_vcpus = $cpuinfo->{cpus};
3897
3898 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3899 if ($allowed_vcpus < $maxcpus);
3900
3901 if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3902
3903 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3904 for (my $i = 2; $i <= $vcpus; $i++) {
3905 my $cpustr = print_cpu_device($conf,$i);
3906 push @$cmd, '-device', $cpustr;
3907 }
3908
3909 } else {
3910
3911 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3912 }
3913 push @$cmd, '-nodefaults';
3914
3915 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3916
3917 my $bootindex_hash = {};
3918 my $i = 1;
3919 foreach my $o (split(//, $bootorder)) {
3920 $bootindex_hash->{$o} = $i*100;
3921 $i++;
3922 }
3923
3924 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3925
3926 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3927
3928 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
3929
3930 if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3931 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, undef, $qxlnum, $bridges);
3932 my $socket = vnc_socket($vmid);
3933 push @$cmd, '-vnc', "unix:$socket,password";
3934 } else {
3935 push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3936 push @$cmd, '-nographic';
3937 }
3938
3939 # time drift fix
3940 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3941
3942 my $useLocaltime = $conf->{localtime};
3943
3944 if ($winversion >= 5) { # windows
3945 $useLocaltime = 1 if !defined($conf->{localtime});
3946
3947 # use time drift fix when acpi is enabled
3948 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3949 $tdf = 1 if !defined($conf->{tdf});
3950 }
3951 }
3952
3953 if ($winversion >= 6) {
3954 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3955 push @$cmd, '-no-hpet';
3956 }
3957
3958 push @$rtcFlags, 'driftfix=slew' if $tdf;
3959
3960 if (!$kvm) {
3961 push @$machineFlags, 'accel=tcg';
3962 }
3963
3964 if ($machine_type) {
3965 push @$machineFlags, "type=${machine_type}";
3966 }
3967
3968 if (($conf->{startdate}) && ($conf->{startdate} ne 'now')) {
3969 push @$rtcFlags, "base=$conf->{startdate}";
3970 } elsif ($useLocaltime) {
3971 push @$rtcFlags, 'base=localtime';
3972 }
3973
3974 push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3975
3976 PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3977
3978 push @$cmd, '-S' if $conf->{freeze};
3979
3980 push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3981
3982 if (parse_guest_agent($conf)->{enabled}) {
3983 my $qgasocket = qmp_socket($vmid, 1);
3984 my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3985 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3986 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3987 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3988 }
3989
3990 my $spice_port;
3991
3992 if ($qxlnum) {
3993 if ($qxlnum > 1) {
3994 if ($winversion){
3995 for(my $i = 1; $i < $qxlnum; $i++){
3996 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, $i, $qxlnum, $bridges);
3997 }
3998 } else {
3999 # assume other OS works like Linux
4000 my ($ram, $vram) = ("134217728", "67108864");
4001 if ($vga->{memory}) {
4002 $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
4003 $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
4004 }
4005 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
4006 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
4007 }
4008 }
4009
4010 my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
4011
4012 my $nodename = PVE::INotify::nodename();
4013 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4014 my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
4015 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
4016
4017 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
4018 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
4019 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
4020
4021 my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
4022 $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
4023
4024 my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
4025 if ($spice_enhancement->{foldersharing}) {
4026 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
4027 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
4028 }
4029
4030 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
4031 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
4032 push @$devices, '-spice', "$spice_opts";
4033 }
4034
4035 # enable balloon by default, unless explicitly disabled
4036 if (!defined($conf->{balloon}) || $conf->{balloon}) {
4037 $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
4038 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
4039 }
4040
4041 if ($conf->{watchdog}) {
4042 my $wdopts = parse_watchdog($conf->{watchdog});
4043 $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
4044 my $watchdog = $wdopts->{model} || 'i6300esb';
4045 push @$devices, '-device', "$watchdog$pciaddr";
4046 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
4047 }
4048
4049 my $vollist = [];
4050 my $scsicontroller = {};
4051 my $ahcicontroller = {};
4052 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
4053
4054 # Add iscsi initiator name if available
4055 if (my $initiator = get_initiator_name()) {
4056 push @$devices, '-iscsi', "initiator-name=$initiator";
4057 }
4058
4059 foreach_drive($conf, sub {
4060 my ($ds, $drive) = @_;
4061
4062 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
4063 push @$vollist, $drive->{file};
4064 }
4065
4066 # ignore efidisk here, already added in bios/fw handling code above
4067 return if $drive->{interface} eq 'efidisk';
4068
4069 $use_virtio = 1 if $ds =~ m/^virtio/;
4070
4071 if (drive_is_cdrom ($drive)) {
4072 if ($bootindex_hash->{d}) {
4073 $drive->{bootindex} = $bootindex_hash->{d};
4074 $bootindex_hash->{d} += 1;
4075 }
4076 } else {
4077 if ($bootindex_hash->{c}) {
4078 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
4079 $bootindex_hash->{c} += 1;
4080 }
4081 }
4082
4083 if($drive->{interface} eq 'virtio'){
4084 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
4085 }
4086
4087 if ($drive->{interface} eq 'scsi') {
4088
4089 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
4090
4091 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
4092 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
4093
4094 my $iothread = '';
4095 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
4096 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4097 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4098 } elsif ($drive->{iothread}) {
4099 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
4100 }
4101
4102 my $queues = '';
4103 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
4104 $queues = ",num_queues=$drive->{queues}";
4105 }
4106
4107 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
4108 $scsicontroller->{$controller}=1;
4109 }
4110
4111 if ($drive->{interface} eq 'sata') {
4112 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
4113 $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
4114 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
4115 $ahcicontroller->{$controller}=1;
4116 }
4117
4118 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
4119 push @$devices, '-drive',$drive_cmd;
4120 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4121 });
4122
4123 for (my $i = 0; $i < $MAX_NETS; $i++) {
4124 next if !$conf->{"net$i"};
4125 my $d = parse_net($conf->{"net$i"});
4126 next if !$d;
4127
4128 $use_virtio = 1 if $d->{model} eq 'virtio';
4129
4130 if ($bootindex_hash->{n}) {
4131 $d->{bootindex} = $bootindex_hash->{n};
4132 $bootindex_hash->{n} += 1;
4133 }
4134
4135 my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
4136 push @$devices, '-netdev', $netdevfull;
4137
4138 my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
4139 push @$devices, '-device', $netdevicefull;
4140 }
4141
4142 if ($conf->{ivshmem}) {
4143 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
4144
4145 my $bus;
4146 if ($q35) {
4147 $bus = print_pcie_addr("ivshmem");
4148 } else {
4149 $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
4150 }
4151
4152 my $ivshmem_name = $ivshmem->{name} // $vmid;
4153 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4154
4155 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4156 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
4157 }
4158
4159 if (!$q35) {
4160 # add pci bridges
4161 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
4162 $bridges->{1} = 1;
4163 $bridges->{2} = 1;
4164 }
4165
4166 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4167
4168 for my $k (sort {$b cmp $a} keys %$bridges) {
4169 $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
4170 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
4171 }
4172 }
4173
4174 push @$cmd, @$devices;
4175 push @$cmd, '-rtc', join(',', @$rtcFlags)
4176 if scalar(@$rtcFlags);
4177 push @$cmd, '-machine', join(',', @$machineFlags)
4178 if scalar(@$machineFlags);
4179 push @$cmd, '-global', join(',', @$globalFlags)
4180 if scalar(@$globalFlags);
4181
4182 if (my $vmstate = $conf->{vmstate}) {
4183 my $statepath = PVE::Storage::path($storecfg, $vmstate);
4184 push @$vollist, $statepath;
4185 push @$cmd, '-loadstate', $statepath;
4186 }
4187
4188 # add custom args
4189 if ($conf->{args}) {
4190 my $aa = PVE::Tools::split_args($conf->{args});
4191 push @$cmd, @$aa;
4192 }
4193
4194 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
4195 }
4196
4197 sub vnc_socket {
4198 my ($vmid) = @_;
4199 return "${var_run_tmpdir}/$vmid.vnc";
4200 }
4201
4202 sub spice_port {
4203 my ($vmid) = @_;
4204
4205 my $res = vm_mon_cmd($vmid, 'query-spice');
4206
4207 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4208 }
4209
4210 sub qmp_socket {
4211 my ($vmid, $qga, $name) = @_;
4212 my $sockettype = $qga ? 'qga' : 'qmp';
4213 my $ext = $name ? '-'.$name : '';
4214 return "${var_run_tmpdir}/$vmid$ext.$sockettype";
4215 }
4216
4217 sub pidfile_name {
4218 my ($vmid) = @_;
4219 return "${var_run_tmpdir}/$vmid.pid";
4220 }
4221
4222 sub vm_devices_list {
4223 my ($vmid) = @_;
4224
4225 my $res = vm_mon_cmd($vmid, 'query-pci');
4226 my $devices_to_check = [];
4227 my $devices = {};
4228 foreach my $pcibus (@$res) {
4229 push @$devices_to_check, @{$pcibus->{devices}},
4230 }
4231
4232 while (@$devices_to_check) {
4233 my $to_check = [];
4234 for my $d (@$devices_to_check) {
4235 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4236 next if !$d->{'pci_bridge'};
4237
4238 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
4239 push @$to_check, @{$d->{'pci_bridge'}->{devices}};
4240 }
4241 $devices_to_check = $to_check;
4242 }
4243
4244 my $resblock = vm_mon_cmd($vmid, 'query-block');
4245 foreach my $block (@$resblock) {
4246 if($block->{device} =~ m/^drive-(\S+)/){
4247 $devices->{$1} = 1;
4248 }
4249 }
4250
4251 my $resmice = vm_mon_cmd($vmid, 'query-mice');
4252 foreach my $mice (@$resmice) {
4253 if ($mice->{name} eq 'QEMU HID Tablet') {
4254 $devices->{tablet} = 1;
4255 last;
4256 }
4257 }
4258
4259 # for usb devices there is no query-usb
4260 # but we can iterate over the entries in
4261 # qom-list path=/machine/peripheral
4262 my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
4263 foreach my $per (@$resperipheral) {
4264 if ($per->{name} =~ m/^usb\d+$/) {
4265 $devices->{$per->{name}} = 1;
4266 }
4267 }
4268
4269 return $devices;
4270 }
4271
4272 sub vm_deviceplug {
4273 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4274
4275 my $q35 = machine_type_is_q35($conf);
4276
4277 my $devices_list = vm_devices_list($vmid);
4278 return 1 if defined($devices_list->{$deviceid});
4279
4280 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
4281
4282 if ($deviceid eq 'tablet') {
4283
4284 qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
4285
4286 } elsif ($deviceid eq 'keyboard') {
4287
4288 qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
4289
4290 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4291
4292 die "usb hotplug currently not reliable\n";
4293 # since we can't reliably hot unplug all added usb devices
4294 # and usb passthrough disables live migration
4295 # we disable usb hotplugging for now
4296 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
4297
4298 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4299
4300 qemu_iothread_add($vmid, $deviceid, $device);
4301
4302 qemu_driveadd($storecfg, $vmid, $device);
4303 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4304
4305 qemu_deviceadd($vmid, $devicefull);
4306 eval { qemu_deviceaddverify($vmid, $deviceid); };
4307 if (my $err = $@) {
4308 eval { qemu_drivedel($vmid, $deviceid); };
4309 warn $@ if $@;
4310 die $err;
4311 }
4312
4313 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4314
4315
4316 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
4317 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4318 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
4319
4320 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4321
4322 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
4323 qemu_iothread_add($vmid, $deviceid, $device);
4324 $devicefull .= ",iothread=iothread-$deviceid";
4325 }
4326
4327 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
4328 $devicefull .= ",num_queues=$device->{queues}";
4329 }
4330
4331 qemu_deviceadd($vmid, $devicefull);
4332 qemu_deviceaddverify($vmid, $deviceid);
4333
4334 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4335
4336 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4337 qemu_driveadd($storecfg, $vmid, $device);
4338
4339 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4340 eval { qemu_deviceadd($vmid, $devicefull); };
4341 if (my $err = $@) {
4342 eval { qemu_drivedel($vmid, $deviceid); };
4343 warn $@ if $@;
4344 die $err;
4345 }
4346
4347 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4348
4349 return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
4350
4351 my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4352 my $use_old_bios_files = undef;
4353 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4354
4355 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
4356 qemu_deviceadd($vmid, $netdevicefull);
4357 eval {
4358 qemu_deviceaddverify($vmid, $deviceid);
4359 qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
4360 };
4361 if (my $err = $@) {
4362 eval { qemu_netdevdel($vmid, $deviceid); };
4363 warn $@ if $@;
4364 die $err;
4365 }
4366
4367 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4368
4369 my $bridgeid = $2;
4370 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4371 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4372
4373 qemu_deviceadd($vmid, $devicefull);
4374 qemu_deviceaddverify($vmid, $deviceid);
4375
4376 } else {
4377 die "can't hotplug device '$deviceid'\n";
4378 }
4379
4380 return 1;
4381 }
4382
4383 # fixme: this should raise exceptions on error!
4384 sub vm_deviceunplug {
4385 my ($vmid, $conf, $deviceid) = @_;
4386
4387 my $devices_list = vm_devices_list($vmid);
4388 return 1 if !defined($devices_list->{$deviceid});
4389
4390 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4391
4392 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
4393
4394 qemu_devicedel($vmid, $deviceid);
4395
4396 } elsif ($deviceid =~ m/^usb\d+$/) {
4397
4398 die "usb hotplug currently not reliable\n";
4399 # when unplugging usb devices this way,
4400 # there may be remaining usb controllers/hubs
4401 # so we disable it for now
4402 qemu_devicedel($vmid, $deviceid);
4403 qemu_devicedelverify($vmid, $deviceid);
4404
4405 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4406
4407 qemu_devicedel($vmid, $deviceid);
4408 qemu_devicedelverify($vmid, $deviceid);
4409 qemu_drivedel($vmid, $deviceid);
4410 qemu_iothread_del($conf, $vmid, $deviceid);
4411
4412 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4413
4414 qemu_devicedel($vmid, $deviceid);
4415 qemu_devicedelverify($vmid, $deviceid);
4416 qemu_iothread_del($conf, $vmid, $deviceid);
4417
4418 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4419
4420 qemu_devicedel($vmid, $deviceid);
4421 qemu_drivedel($vmid, $deviceid);
4422 qemu_deletescsihw($conf, $vmid, $deviceid);
4423
4424 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4425
4426 qemu_devicedel($vmid, $deviceid);
4427 qemu_devicedelverify($vmid, $deviceid);
4428 qemu_netdevdel($vmid, $deviceid);
4429
4430 } else {
4431 die "can't unplug device '$deviceid'\n";
4432 }
4433
4434 return 1;
4435 }
4436
4437 sub qemu_deviceadd {
4438 my ($vmid, $devicefull) = @_;
4439
4440 $devicefull = "driver=".$devicefull;
4441 my %options = split(/[=,]/, $devicefull);
4442
4443 vm_mon_cmd($vmid, "device_add" , %options);
4444 }
4445
4446 sub qemu_devicedel {
4447 my ($vmid, $deviceid) = @_;
4448
4449 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4450 }
4451
4452 sub qemu_iothread_add {
4453 my($vmid, $deviceid, $device) = @_;
4454
4455 if ($device->{iothread}) {
4456 my $iothreads = vm_iothreads_list($vmid);
4457 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4458 }
4459 }
4460
4461 sub qemu_iothread_del {
4462 my($conf, $vmid, $deviceid) = @_;
4463
4464 my $confid = $deviceid;
4465 if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
4466 $confid = 'scsi' . $1;
4467 }
4468 my $device = parse_drive($confid, $conf->{$confid});
4469 if ($device->{iothread}) {
4470 my $iothreads = vm_iothreads_list($vmid);
4471 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4472 }
4473 }
4474
4475 sub qemu_objectadd {
4476 my($vmid, $objectid, $qomtype) = @_;
4477
4478 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4479
4480 return 1;
4481 }
4482
4483 sub qemu_objectdel {
4484 my($vmid, $objectid) = @_;
4485
4486 vm_mon_cmd($vmid, "object-del", id => $objectid);
4487
4488 return 1;
4489 }
4490
4491 sub qemu_driveadd {
4492 my ($storecfg, $vmid, $device) = @_;
4493
4494 my $drive = print_drive_full($storecfg, $vmid, $device);
4495 $drive =~ s/\\/\\\\/g;
4496 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4497
4498 # If the command succeeds qemu prints: "OK"
4499 return 1 if $ret =~ m/OK/s;
4500
4501 die "adding drive failed: $ret\n";
4502 }
4503
4504 sub qemu_drivedel {
4505 my($vmid, $deviceid) = @_;
4506
4507 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4508 $ret =~ s/^\s+//;
4509
4510 return 1 if $ret eq "";
4511
4512 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4513 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4514
4515 die "deleting drive $deviceid failed : $ret\n";
4516 }
4517
4518 sub qemu_deviceaddverify {
4519 my ($vmid, $deviceid) = @_;
4520
4521 for (my $i = 0; $i <= 5; $i++) {
4522 my $devices_list = vm_devices_list($vmid);
4523 return 1 if defined($devices_list->{$deviceid});
4524 sleep 1;
4525 }
4526
4527 die "error on hotplug device '$deviceid'\n";
4528 }
4529
4530
4531 sub qemu_devicedelverify {
4532 my ($vmid, $deviceid) = @_;
4533
4534 # need to verify that the device is correctly removed as device_del
4535 # is async and empty return is not reliable
4536
4537 for (my $i = 0; $i <= 5; $i++) {
4538 my $devices_list = vm_devices_list($vmid);
4539 return 1 if !defined($devices_list->{$deviceid});
4540 sleep 1;
4541 }
4542
4543 die "error on hot-unplugging device '$deviceid'\n";
4544 }
4545
4546 sub qemu_findorcreatescsihw {
4547 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4548
4549 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4550
4551 my $scsihwid="$controller_prefix$controller";
4552 my $devices_list = vm_devices_list($vmid);
4553
4554 if(!defined($devices_list->{$scsihwid})) {
4555 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4556 }
4557
4558 return 1;
4559 }
4560
4561 sub qemu_deletescsihw {
4562 my ($conf, $vmid, $opt) = @_;
4563
4564 my $device = parse_drive($opt, $conf->{$opt});
4565
4566 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4567 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4568 return 1;
4569 }
4570
4571 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4572
4573 my $devices_list = vm_devices_list($vmid);
4574 foreach my $opt (keys %{$devices_list}) {
4575 if (PVE::QemuServer::is_valid_drivename($opt)) {
4576 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4577 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4578 return 1;
4579 }
4580 }
4581 }
4582
4583 my $scsihwid="scsihw$controller";
4584
4585 vm_deviceunplug($vmid, $conf, $scsihwid);
4586
4587 return 1;
4588 }
4589
4590 sub qemu_add_pci_bridge {
4591 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4592
4593 my $bridges = {};
4594
4595 my $bridgeid;
4596
4597 print_pci_addr($device, $bridges, $arch, $machine_type);
4598
4599 while (my ($k, $v) = each %$bridges) {
4600 $bridgeid = $k;
4601 }
4602 return 1 if !defined($bridgeid) || $bridgeid < 1;
4603
4604 my $bridge = "pci.$bridgeid";
4605 my $devices_list = vm_devices_list($vmid);
4606
4607 if (!defined($devices_list->{$bridge})) {
4608 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4609 }
4610
4611 return 1;
4612 }
4613
4614 sub qemu_set_link_status {
4615 my ($vmid, $device, $up) = @_;
4616
4617 vm_mon_cmd($vmid, "set_link", name => $device,
4618 up => $up ? JSON::true : JSON::false);
4619 }
4620
4621 sub qemu_netdevadd {
4622 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4623
4624 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4625 my %options = split(/[=,]/, $netdev);
4626
4627 vm_mon_cmd($vmid, "netdev_add", %options);
4628 return 1;
4629 }
4630
4631 sub qemu_netdevdel {
4632 my ($vmid, $deviceid) = @_;
4633
4634 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4635 }
4636
4637 sub qemu_usb_hotplug {
4638 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4639
4640 return if !$device;
4641
4642 # remove the old one first
4643 vm_deviceunplug($vmid, $conf, $deviceid);
4644
4645 # check if xhci controller is necessary and available
4646 if ($device->{usb3}) {
4647
4648 my $devicelist = vm_devices_list($vmid);
4649
4650 if (!$devicelist->{xhci}) {
4651 my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4652 qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4653 }
4654 }
4655 my $d = parse_usb_device($device->{host});
4656 $d->{usb3} = $device->{usb3};
4657
4658 # add the new one
4659 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4660 }
4661
4662 sub qemu_cpu_hotplug {
4663 my ($vmid, $conf, $vcpus) = @_;
4664
4665 my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4666
4667 my $sockets = 1;
4668 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4669 $sockets = $conf->{sockets} if $conf->{sockets};
4670 my $cores = $conf->{cores} || 1;
4671 my $maxcpus = $sockets * $cores;
4672
4673 $vcpus = $maxcpus if !$vcpus;
4674
4675 die "you can't add more vcpus than maxcpus\n"
4676 if $vcpus > $maxcpus;
4677
4678 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4679
4680 if ($vcpus < $currentvcpus) {
4681
4682 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4683
4684 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4685 qemu_devicedel($vmid, "cpu$i");
4686 my $retry = 0;
4687 my $currentrunningvcpus = undef;
4688 while (1) {
4689 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4690 last if scalar(@{$currentrunningvcpus}) == $i-1;
4691 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4692 $retry++;
4693 sleep 1;
4694 }
4695 #update conf after each succesfull cpu unplug
4696 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4697 PVE::QemuConfig->write_config($vmid, $conf);
4698 }
4699 } else {
4700 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4701 }
4702
4703 return;
4704 }
4705
4706 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4707 die "vcpus in running vm does not match its configuration\n"
4708 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4709
4710 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4711
4712 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4713 my $cpustr = print_cpu_device($conf, $i);
4714 qemu_deviceadd($vmid, $cpustr);
4715
4716 my $retry = 0;
4717 my $currentrunningvcpus = undef;
4718 while (1) {
4719 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4720 last if scalar(@{$currentrunningvcpus}) == $i;
4721 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4722 sleep 1;
4723 $retry++;
4724 }
4725 #update conf after each succesfull cpu hotplug
4726 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4727 PVE::QemuConfig->write_config($vmid, $conf);
4728 }
4729 } else {
4730
4731 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4732 vm_mon_cmd($vmid, "cpu-add", id => int($i));
4733 }
4734 }
4735 }
4736
4737 sub qemu_block_set_io_throttle {
4738 my ($vmid, $deviceid,
4739 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4740 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4741 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4742 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4743
4744 return if !check_running($vmid) ;
4745
4746 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4747 bps => int($bps),
4748 bps_rd => int($bps_rd),
4749 bps_wr => int($bps_wr),
4750 iops => int($iops),
4751 iops_rd => int($iops_rd),
4752 iops_wr => int($iops_wr),
4753 bps_max => int($bps_max),
4754 bps_rd_max => int($bps_rd_max),
4755 bps_wr_max => int($bps_wr_max),
4756 iops_max => int($iops_max),
4757 iops_rd_max => int($iops_rd_max),
4758 iops_wr_max => int($iops_wr_max),
4759 bps_max_length => int($bps_max_length),
4760 bps_rd_max_length => int($bps_rd_max_length),
4761 bps_wr_max_length => int($bps_wr_max_length),
4762 iops_max_length => int($iops_max_length),
4763 iops_rd_max_length => int($iops_rd_max_length),
4764 iops_wr_max_length => int($iops_wr_max_length),
4765 );
4766
4767 }
4768
4769 # old code, only used to shutdown old VM after update
4770 sub __read_avail {
4771 my ($fh, $timeout) = @_;
4772
4773 my $sel = new IO::Select;
4774 $sel->add($fh);
4775
4776 my $res = '';
4777 my $buf;
4778
4779 my @ready;
4780 while (scalar (@ready = $sel->can_read($timeout))) {
4781 my $count;
4782 if ($count = $fh->sysread($buf, 8192)) {
4783 if ($buf =~ /^(.*)\(qemu\) $/s) {
4784 $res .= $1;
4785 last;
4786 } else {
4787 $res .= $buf;
4788 }
4789 } else {
4790 if (!defined($count)) {
4791 die "$!\n";
4792 }
4793 last;
4794 }
4795 }
4796
4797 die "monitor read timeout\n" if !scalar(@ready);
4798
4799 return $res;
4800 }
4801
4802 sub qemu_block_resize {
4803 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4804
4805 my $running = check_running($vmid);
4806
4807 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4808
4809 return if !$running;
4810
4811 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4812
4813 }
4814
4815 sub qemu_volume_snapshot {
4816 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4817
4818 my $running = check_running($vmid);
4819
4820 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4821 vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4822 } else {
4823 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4824 }
4825 }
4826
4827 sub qemu_volume_snapshot_delete {
4828 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4829
4830 my $running = check_running($vmid);
4831
4832 if($running) {
4833
4834 $running = undef;
4835 my $conf = PVE::QemuConfig->load_config($vmid);
4836 foreach_drive($conf, sub {
4837 my ($ds, $drive) = @_;
4838 $running = 1 if $drive->{file} eq $volid;
4839 });
4840 }
4841
4842 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4843 vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4844 } else {
4845 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4846 }
4847 }
4848
4849 sub set_migration_caps {
4850 my ($vmid) = @_;
4851
4852 my $cap_ref = [];
4853
4854 my $enabled_cap = {
4855 "auto-converge" => 1,
4856 "xbzrle" => 1,
4857 "x-rdma-pin-all" => 0,
4858 "zero-blocks" => 0,
4859 "compress" => 0
4860 };
4861
4862 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4863
4864 for my $supported_capability (@$supported_capabilities) {
4865 push @$cap_ref, {
4866 capability => $supported_capability->{capability},
4867 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4868 };
4869 }
4870
4871 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4872 }
4873
4874 my $fast_plug_option = {
4875 'lock' => 1,
4876 'name' => 1,
4877 'onboot' => 1,
4878 'shares' => 1,
4879 'startup' => 1,
4880 'description' => 1,
4881 'protection' => 1,
4882 'vmstatestorage' => 1,
4883 'hookscript' => 1,
4884 };
4885
4886 # hotplug changes in [PENDING]
4887 # $selection hash can be used to only apply specified options, for
4888 # example: { cores => 1 } (only apply changed 'cores')
4889 # $errors ref is used to return error messages
4890 sub vmconfig_hotplug_pending {
4891 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4892
4893 my $defaults = load_defaults();
4894 my ($arch, $machine_type) = get_basic_machine_info($conf, undef);
4895
4896 # commit values which do not have any impact on running VM first
4897 # Note: those option cannot raise errors, we we do not care about
4898 # $selection and always apply them.
4899
4900 my $add_error = sub {
4901 my ($opt, $msg) = @_;
4902 $errors->{$opt} = "hotplug problem - $msg";
4903 };
4904
4905 my $changes = 0;
4906 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4907 if ($fast_plug_option->{$opt}) {
4908 $conf->{$opt} = $conf->{pending}->{$opt};
4909 delete $conf->{pending}->{$opt};
4910 $changes = 1;
4911 }
4912 }
4913
4914 if ($changes) {
4915 PVE::QemuConfig->write_config($vmid, $conf);
4916 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4917 }
4918
4919 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4920
4921 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4922 while (my ($opt, $force) = each %$pending_delete_hash) {
4923 next if $selection && !$selection->{$opt};
4924 eval {
4925 if ($opt eq 'hotplug') {
4926 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4927 } elsif ($opt eq 'tablet') {
4928 die "skip\n" if !$hotplug_features->{usb};
4929 if ($defaults->{tablet}) {
4930 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4931 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4932 if $arch eq 'aarch64';
4933 } else {
4934 vm_deviceunplug($vmid, $conf, 'tablet');
4935 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4936 }
4937 } elsif ($opt =~ m/^usb\d+/) {
4938 die "skip\n";
4939 # since we cannot reliably hot unplug usb devices
4940 # we are disabling it
4941 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4942 vm_deviceunplug($vmid, $conf, $opt);
4943 } elsif ($opt eq 'vcpus') {
4944 die "skip\n" if !$hotplug_features->{cpu};
4945 qemu_cpu_hotplug($vmid, $conf, undef);
4946 } elsif ($opt eq 'balloon') {
4947 # enable balloon device is not hotpluggable
4948 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4949 # here we reset the ballooning value to memory
4950 my $balloon = $conf->{memory} || $defaults->{memory};
4951 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4952 } elsif ($fast_plug_option->{$opt}) {
4953 # do nothing
4954 } elsif ($opt =~ m/^net(\d+)$/) {
4955 die "skip\n" if !$hotplug_features->{network};
4956 vm_deviceunplug($vmid, $conf, $opt);
4957 } elsif (is_valid_drivename($opt)) {
4958 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4959 vm_deviceunplug($vmid, $conf, $opt);
4960 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4961 } elsif ($opt =~ m/^memory$/) {
4962 die "skip\n" if !$hotplug_features->{memory};
4963 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4964 } elsif ($opt eq 'cpuunits') {
4965 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4966 } elsif ($opt eq 'cpulimit') {
4967 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4968 } else {
4969 die "skip\n";
4970 }
4971 };
4972 if (my $err = $@) {
4973 &$add_error($opt, $err) if $err ne "skip\n";
4974 } else {
4975 # save new config if hotplug was successful
4976 delete $conf->{$opt};
4977 vmconfig_undelete_pending_option($conf, $opt);
4978 PVE::QemuConfig->write_config($vmid, $conf);
4979 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4980 }
4981 }
4982
4983 my $apply_pending_cloudinit;
4984 $apply_pending_cloudinit = sub {
4985 my ($key, $value) = @_;
4986 $apply_pending_cloudinit = sub {}; # once is enough
4987
4988 my @cloudinit_opts = keys %$confdesc_cloudinit;
4989 foreach my $opt (keys %{$conf->{pending}}) {
4990 next if !grep { $_ eq $opt } @cloudinit_opts;
4991 $conf->{$opt} = delete $conf->{pending}->{$opt};
4992 }
4993
4994 my $new_conf = { %$conf };
4995 $new_conf->{$key} = $value;
4996 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4997 };
4998
4999 foreach my $opt (keys %{$conf->{pending}}) {
5000 next if $selection && !$selection->{$opt};
5001 my $value = $conf->{pending}->{$opt};
5002 eval {
5003 if ($opt eq 'hotplug') {
5004 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
5005 } elsif ($opt eq 'tablet') {
5006 die "skip\n" if !$hotplug_features->{usb};
5007 if ($value == 1) {
5008 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5009 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5010 if $arch eq 'aarch64';
5011 } elsif ($value == 0) {
5012 vm_deviceunplug($vmid, $conf, 'tablet');
5013 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5014 }
5015 } elsif ($opt =~ m/^usb\d+$/) {
5016 die "skip\n";
5017 # since we cannot reliably hot unplug usb devices
5018 # we are disabling it
5019 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
5020 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
5021 die "skip\n" if !$d;
5022 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
5023 } elsif ($opt eq 'vcpus') {
5024 die "skip\n" if !$hotplug_features->{cpu};
5025 qemu_cpu_hotplug($vmid, $conf, $value);
5026 } elsif ($opt eq 'balloon') {
5027 # enable/disable balloning device is not hotpluggable
5028 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
5029 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
5030 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5031
5032 # allow manual ballooning if shares is set to zero
5033 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
5034 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
5035 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
5036 }
5037 } elsif ($opt =~ m/^net(\d+)$/) {
5038 # some changes can be done without hotplug
5039 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
5040 $vmid, $opt, $value, $arch, $machine_type);
5041 } elsif (is_valid_drivename($opt)) {
5042 # some changes can be done without hotplug
5043 my $drive = parse_drive($opt, $value);
5044 if (drive_is_cloudinit($drive)) {
5045 &$apply_pending_cloudinit($opt, $value);
5046 }
5047 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
5048 $vmid, $opt, $value, 1, $arch, $machine_type);
5049 } elsif ($opt =~ m/^memory$/) { #dimms
5050 die "skip\n" if !$hotplug_features->{memory};
5051 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
5052 } elsif ($opt eq 'cpuunits') {
5053 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
5054 } elsif ($opt eq 'cpulimit') {
5055 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
5056 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
5057 } else {
5058 die "skip\n"; # skip non-hot-pluggable options
5059 }
5060 };
5061 if (my $err = $@) {
5062 &$add_error($opt, $err) if $err ne "skip\n";
5063 } else {
5064 # save new config if hotplug was successful
5065 $conf->{$opt} = $value;
5066 delete $conf->{pending}->{$opt};
5067 PVE::QemuConfig->write_config($vmid, $conf);
5068 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5069 }
5070 }
5071 }
5072
5073 sub try_deallocate_drive {
5074 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5075
5076 if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
5077 my $volid = $drive->{file};
5078 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
5079 my $sid = PVE::Storage::parse_volume_id($volid);
5080 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5081
5082 # check if the disk is really unused
5083 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5084 if is_volume_in_use($storecfg, $conf, $key, $volid);
5085 PVE::Storage::vdisk_free($storecfg, $volid);
5086 return 1;
5087 } else {
5088 # If vm is not owner of this disk remove from config
5089 return 1;
5090 }
5091 }
5092
5093 return undef;
5094 }
5095
5096 sub vmconfig_delete_or_detach_drive {
5097 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5098
5099 my $drive = parse_drive($opt, $conf->{$opt});
5100
5101 my $rpcenv = PVE::RPCEnvironment::get();
5102 my $authuser = $rpcenv->get_user();
5103
5104 if ($force) {
5105 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5106 try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5107 } else {
5108 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
5109 }
5110 }
5111
5112 sub vmconfig_apply_pending {
5113 my ($vmid, $conf, $storecfg) = @_;
5114
5115 # cold plug
5116
5117 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
5118 while (my ($opt, $force) = each %$pending_delete_hash) {
5119 die "internal error" if $opt =~ m/^unused/;
5120 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5121 if (!defined($conf->{$opt})) {
5122 vmconfig_undelete_pending_option($conf, $opt);
5123 PVE::QemuConfig->write_config($vmid, $conf);
5124 } elsif (is_valid_drivename($opt)) {
5125 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
5126 vmconfig_undelete_pending_option($conf, $opt);
5127 delete $conf->{$opt};
5128 PVE::QemuConfig->write_config($vmid, $conf);
5129 } else {
5130 vmconfig_undelete_pending_option($conf, $opt);
5131 delete $conf->{$opt};
5132 PVE::QemuConfig->write_config($vmid, $conf);
5133 }
5134 }
5135
5136 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5137
5138 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5139 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5140
5141 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
5142 # skip if nothing changed
5143 } elsif (is_valid_drivename($opt)) {
5144 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
5145 if defined($conf->{$opt});
5146 $conf->{$opt} = $conf->{pending}->{$opt};
5147 } else {
5148 $conf->{$opt} = $conf->{pending}->{$opt};
5149 }
5150
5151 delete $conf->{pending}->{$opt};
5152 PVE::QemuConfig->write_config($vmid, $conf);
5153 }
5154 }
5155
5156 my $safe_num_ne = sub {
5157 my ($a, $b) = @_;
5158
5159 return 0 if !defined($a) && !defined($b);
5160 return 1 if !defined($a);
5161 return 1 if !defined($b);
5162
5163 return $a != $b;
5164 };
5165
5166 my $safe_string_ne = sub {
5167 my ($a, $b) = @_;
5168
5169 return 0 if !defined($a) && !defined($b);
5170 return 1 if !defined($a);
5171 return 1 if !defined($b);
5172
5173 return $a ne $b;
5174 };
5175
5176 sub vmconfig_update_net {
5177 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5178
5179 my $newnet = parse_net($value);
5180
5181 if ($conf->{$opt}) {
5182 my $oldnet = parse_net($conf->{$opt});
5183
5184 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
5185 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
5186 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
5187 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
5188
5189 # for non online change, we try to hot-unplug
5190 die "skip\n" if !$hotplug;
5191 vm_deviceunplug($vmid, $conf, $opt);
5192 } else {
5193
5194 die "internal error" if $opt !~ m/net(\d+)/;
5195 my $iface = "tap${vmid}i$1";
5196
5197 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
5198 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
5199 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
5200 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
5201 PVE::Network::tap_unplug($iface);
5202 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
5203 } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
5204 # Rate can be applied on its own but any change above needs to
5205 # include the rate in tap_plug since OVS resets everything.
5206 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
5207 }
5208
5209 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
5210 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
5211 }
5212
5213 return 1;
5214 }
5215 }
5216
5217 if ($hotplug) {
5218 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5219 } else {
5220 die "skip\n";
5221 }
5222 }
5223
5224 sub vmconfig_update_disk {
5225 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force, $arch, $machine_type) = @_;
5226
5227 # fixme: do we need force?
5228
5229 my $drive = parse_drive($opt, $value);
5230
5231 if ($conf->{$opt}) {
5232
5233 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
5234
5235 my $media = $drive->{media} || 'disk';
5236 my $oldmedia = $old_drive->{media} || 'disk';
5237 die "unable to change media type\n" if $media ne $oldmedia;
5238
5239 if (!drive_is_cdrom($old_drive)) {
5240
5241 if ($drive->{file} ne $old_drive->{file}) {
5242
5243 die "skip\n" if !$hotplug;
5244
5245 # unplug and register as unused
5246 vm_deviceunplug($vmid, $conf, $opt);
5247 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
5248
5249 } else {
5250 # update existing disk
5251
5252 # skip non hotpluggable value
5253 if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
5254 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
5255 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
5256 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
5257 die "skip\n";
5258 }
5259
5260 # apply throttle
5261 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
5262 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
5263 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
5264 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
5265 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
5266 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
5267 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
5268 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
5269 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
5270 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
5271 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
5272 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
5273 &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
5274 &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
5275 &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
5276 &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
5277 &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
5278 &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
5279
5280 qemu_block_set_io_throttle($vmid,"drive-$opt",
5281 ($drive->{mbps} || 0)*1024*1024,
5282 ($drive->{mbps_rd} || 0)*1024*1024,
5283 ($drive->{mbps_wr} || 0)*1024*1024,
5284 $drive->{iops} || 0,
5285 $drive->{iops_rd} || 0,
5286 $drive->{iops_wr} || 0,
5287 ($drive->{mbps_max} || 0)*1024*1024,
5288 ($drive->{mbps_rd_max} || 0)*1024*1024,
5289 ($drive->{mbps_wr_max} || 0)*1024*1024,
5290 $drive->{iops_max} || 0,
5291 $drive->{iops_rd_max} || 0,
5292 $drive->{iops_wr_max} || 0,
5293 $drive->{bps_max_length} || 1,
5294 $drive->{bps_rd_max_length} || 1,
5295 $drive->{bps_wr_max_length} || 1,
5296 $drive->{iops_max_length} || 1,
5297 $drive->{iops_rd_max_length} || 1,
5298 $drive->{iops_wr_max_length} || 1);
5299
5300 }
5301
5302 return 1;
5303 }
5304
5305 } else { # cdrom
5306
5307 if ($drive->{file} eq 'none') {
5308 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
5309 if (drive_is_cloudinit($old_drive)) {
5310 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
5311 }
5312 } else {
5313 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
5314 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
5315 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
5316 }
5317
5318 return 1;
5319 }
5320 }
5321 }
5322
5323 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5324 # hotplug new disks
5325 PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
5326 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5327 }
5328
5329 sub vm_start {
5330 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
5331 $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
5332
5333 PVE::QemuConfig->lock_config($vmid, sub {
5334 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5335
5336 die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
5337
5338 my $is_suspended = PVE::QemuConfig->has_lock($conf, 'suspended');
5339
5340 PVE::QemuConfig->check_lock($conf)
5341 if !($skiplock || $is_suspended);
5342
5343 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
5344
5345 # clean up leftover reboot request files
5346 eval { clear_reboot_request($vmid); };
5347 warn $@ if $@;
5348
5349 if (!$statefile && scalar(keys %{$conf->{pending}})) {
5350 vmconfig_apply_pending($vmid, $conf, $storecfg);
5351 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5352 }
5353
5354 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
5355
5356 my $defaults = load_defaults();
5357
5358 # set environment variable useful inside network script
5359 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
5360
5361 my $local_volumes = {};
5362
5363 if ($targetstorage) {
5364 foreach_drive($conf, sub {
5365 my ($ds, $drive) = @_;
5366
5367 return if drive_is_cdrom($drive);
5368
5369 my $volid = $drive->{file};
5370
5371 return if !$volid;
5372
5373 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5374
5375 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5376 return if $scfg->{shared};
5377 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5378 });
5379
5380 my $format = undef;
5381
5382 foreach my $opt (sort keys %$local_volumes) {
5383
5384 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5385 my $drive = parse_drive($opt, $conf->{$opt});
5386
5387 #if remote storage is specified, use default format
5388 if ($targetstorage && $targetstorage ne "1") {
5389 $storeid = $targetstorage;
5390 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5391 $format = $defFormat;
5392 } else {
5393 #else we use same format than original
5394 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5395 $format = qemu_img_format($scfg, $volid);
5396 }
5397
5398 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5399 my $newdrive = $drive;
5400 $newdrive->{format} = $format;
5401 $newdrive->{file} = $newvolid;
5402 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5403 $local_volumes->{$opt} = $drivestr;
5404 #pass drive to conf for command line
5405 $conf->{$opt} = $drivestr;
5406 }
5407 }
5408
5409 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
5410
5411 if ($is_suspended) {
5412 # enforce machine type on suspended vm to ensure HW compatibility
5413 $forcemachine = $conf->{runningmachine};
5414 print "Resuming suspended VM\n";
5415 }
5416
5417 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5418
5419 my $migrate_uri;
5420 if ($statefile) {
5421 if ($statefile eq 'tcp') {
5422 my $localip = "localhost";
5423 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5424 my $nodename = PVE::INotify::nodename();
5425
5426 if (!defined($migration_type)) {
5427 if (defined($datacenterconf->{migration}->{type})) {
5428 $migration_type = $datacenterconf->{migration}->{type};
5429 } else {
5430 $migration_type = 'secure';
5431 }
5432 }
5433
5434 if ($migration_type eq 'insecure') {
5435 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5436 if ($migrate_network_addr) {
5437 $localip = $migrate_network_addr;
5438 } else {
5439 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5440 }
5441
5442 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5443 }
5444
5445 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5446 my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5447 $migrate_uri = "tcp:${localip}:${migrate_port}";
5448 push @$cmd, '-incoming', $migrate_uri;
5449 push @$cmd, '-S';
5450
5451 } elsif ($statefile eq 'unix') {
5452 # should be default for secure migrations as a ssh TCP forward
5453 # tunnel is not deterministic reliable ready and fails regurarly
5454 # to set up in time, so use UNIX socket forwards
5455 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5456 unlink $socket_addr;
5457
5458 $migrate_uri = "unix:$socket_addr";
5459
5460 push @$cmd, '-incoming', $migrate_uri;
5461 push @$cmd, '-S';
5462
5463 } elsif (-e $statefile) {
5464 push @$cmd, '-loadstate', $statefile;
5465 } else {
5466 my $statepath = PVE::Storage::path($storecfg, $statefile);
5467 push @$vollist, $statepath;
5468 push @$cmd, '-loadstate', $statepath;
5469 }
5470 } elsif ($paused) {
5471 push @$cmd, '-S';
5472 }
5473
5474 # host pci devices
5475 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
5476 my $d = parse_hostpci($conf->{"hostpci$i"});
5477 next if !$d;
5478 my $pcidevices = $d->{pciid};
5479 foreach my $pcidevice (@$pcidevices) {
5480 my $pciid = $pcidevice->{id};
5481
5482 my $info = PVE::SysFSTools::pci_device_info("0000:$pciid");
5483 die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5484 die "no pci device info for device '$pciid'\n" if !$info;
5485
5486 if ($d->{mdev}) {
5487 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5488 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5489 } else {
5490 die "can't unbind/bind pci group to vfio '$pciid'\n"
5491 if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5492 die "can't reset pci device '$pciid'\n"
5493 if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5494 }
5495 }
5496 }
5497
5498 PVE::Storage::activate_volumes($storecfg, $vollist);
5499
5500 eval {
5501 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5502 outfunc => sub {}, errfunc => sub {});
5503 };
5504 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5505 # timeout should be more than enough here...
5506 PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5507
5508 my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5509 : $defaults->{cpuunits};
5510
5511 my $start_timeout = ($conf->{hugepages} || $is_suspended) ? 300 : 30;
5512 my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5513
5514 my %properties = (
5515 Slice => 'qemu.slice',
5516 KillMode => 'none',
5517 CPUShares => $cpuunits
5518 );
5519
5520 if (my $cpulimit = $conf->{cpulimit}) {
5521 $properties{CPUQuota} = int($cpulimit * 100);
5522 }
5523 $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5524
5525 my $run_qemu = sub {
5526 PVE::Tools::run_fork sub {
5527 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5528 run_command($cmd, %run_params);
5529 };
5530 };
5531
5532 if ($conf->{hugepages}) {
5533
5534 my $code = sub {
5535 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5536 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5537
5538 PVE::QemuServer::Memory::hugepages_mount();
5539 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5540
5541 eval { $run_qemu->() };
5542 if (my $err = $@) {
5543 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5544 die $err;
5545 }
5546
5547 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5548 };
5549 eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5550
5551 } else {
5552 eval { $run_qemu->() };
5553 }
5554
5555 if (my $err = $@) {
5556 # deactivate volumes if start fails
5557 eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5558 die "start failed: $err";
5559 }
5560
5561 print "migration listens on $migrate_uri\n" if $migrate_uri;
5562
5563 if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix') {
5564 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5565 warn $@ if $@;
5566 }
5567
5568 #start nbd server for storage migration
5569 if ($targetstorage) {
5570 my $nodename = PVE::INotify::nodename();
5571 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5572 my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5573 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5574 my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5575
5576 vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5577
5578 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5579
5580 foreach my $opt (sort keys %$local_volumes) {
5581 my $volid = $local_volumes->{$opt};
5582 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5583 my $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}:exportname=drive-$opt";
5584 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5585 }
5586 }
5587
5588 if ($migratedfrom) {
5589 eval {
5590 set_migration_caps($vmid);
5591 };
5592 warn $@ if $@;
5593
5594 if ($spice_port) {
5595 print "spice listens on port $spice_port\n";
5596 if ($spice_ticket) {
5597 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5598 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5599 }
5600 }
5601
5602 } else {
5603 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5604 if !$statefile && $conf->{balloon};
5605
5606 foreach my $opt (keys %$conf) {
5607 next if $opt !~ m/^net\d+$/;
5608 my $nicconf = parse_net($conf->{$opt});
5609 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5610 }
5611 }
5612
5613 vm_mon_cmd_nocheck($vmid, 'qom-set',
5614 path => "machine/peripheral/balloon0",
5615 property => "guest-stats-polling-interval",
5616 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5617
5618 if ($is_suspended && (my $vmstate = $conf->{vmstate})) {
5619 print "Resumed VM, removing state\n";
5620 delete $conf->@{qw(lock vmstate runningmachine)};
5621 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5622 PVE::Storage::vdisk_free($storecfg, $vmstate);
5623 PVE::QemuConfig->write_config($vmid, $conf);
5624 }
5625
5626 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5627 });
5628 }
5629
5630 sub vm_mon_cmd {
5631 my ($vmid, $execute, %params) = @_;
5632
5633 my $cmd = { execute => $execute, arguments => \%params };
5634 vm_qmp_command($vmid, $cmd);
5635 }
5636
5637 sub vm_mon_cmd_nocheck {
5638 my ($vmid, $execute, %params) = @_;
5639
5640 my $cmd = { execute => $execute, arguments => \%params };
5641 vm_qmp_command($vmid, $cmd, 1);
5642 }
5643
5644 sub vm_qmp_command {
5645 my ($vmid, $cmd, $nocheck) = @_;
5646
5647 my $res;
5648
5649 my $timeout;
5650 if ($cmd->{arguments}) {
5651 $timeout = delete $cmd->{arguments}->{timeout};
5652 }
5653
5654 eval {
5655 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5656 my $sname = qmp_socket($vmid);
5657 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5658 my $qmpclient = PVE::QMPClient->new();
5659
5660 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5661 } else {
5662 die "unable to open monitor socket\n";
5663 }
5664 };
5665 if (my $err = $@) {
5666 syslog("err", "VM $vmid qmp command failed - $err");
5667 die $err;
5668 }
5669
5670 return $res;
5671 }
5672
5673 sub vm_human_monitor_command {
5674 my ($vmid, $cmdline) = @_;
5675
5676 my $cmd = {
5677 execute => 'human-monitor-command',
5678 arguments => { 'command-line' => $cmdline},
5679 };
5680
5681 return vm_qmp_command($vmid, $cmd);
5682 }
5683
5684 sub vm_commandline {
5685 my ($storecfg, $vmid, $snapname) = @_;
5686
5687 my $conf = PVE::QemuConfig->load_config($vmid);
5688
5689 if ($snapname) {
5690 my $snapshot = $conf->{snapshots}->{$snapname};
5691 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5692
5693 $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5694
5695 $conf = $snapshot;
5696 }
5697
5698 my $defaults = load_defaults();
5699
5700 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5701
5702 return PVE::Tools::cmd2string($cmd);
5703 }
5704
5705 sub vm_reset {
5706 my ($vmid, $skiplock) = @_;
5707
5708 PVE::QemuConfig->lock_config($vmid, sub {
5709
5710 my $conf = PVE::QemuConfig->load_config($vmid);
5711
5712 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5713
5714 vm_mon_cmd($vmid, "system_reset");
5715 });
5716 }
5717
5718 sub get_vm_volumes {
5719 my ($conf) = @_;
5720
5721 my $vollist = [];
5722 foreach_volid($conf, sub {
5723 my ($volid, $attr) = @_;
5724
5725 return if $volid =~ m|^/|;
5726
5727 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5728 return if !$sid;
5729
5730 push @$vollist, $volid;
5731 });
5732
5733 return $vollist;
5734 }
5735
5736 sub vm_stop_cleanup {
5737 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5738
5739 eval {
5740
5741 if (!$keepActive) {
5742 my $vollist = get_vm_volumes($conf);
5743 PVE::Storage::deactivate_volumes($storecfg, $vollist);
5744 }
5745
5746 foreach my $ext (qw(mon qmp pid vnc qga)) {
5747 unlink "/var/run/qemu-server/${vmid}.$ext";
5748 }
5749
5750 if ($conf->{ivshmem}) {
5751 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5752 # just delete it for now, VMs which have this already open do not
5753 # are affected, but new VMs will get a separated one. If this
5754 # becomes an issue we either add some sort of ref-counting or just
5755 # add a "don't delete on stop" flag to the ivshmem format.
5756 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5757 }
5758
5759 foreach my $key (keys %$conf) {
5760 next if $key !~ m/^hostpci(\d+)$/;
5761 my $hostpciindex = $1;
5762 my $d = parse_hostpci($conf->{$key});
5763 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5764
5765 foreach my $pci (@{$d->{pciid}}) {
5766 my $pciid = $pci->{id};
5767 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5768 }
5769 }
5770
5771 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5772 };
5773 warn $@ if $@; # avoid errors - just warn
5774 }
5775
5776 # call only in locked context
5777 sub _do_vm_stop {
5778 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5779
5780 my $pid = check_running($vmid, $nocheck);
5781 return if !$pid;
5782
5783 my $conf;
5784 if (!$nocheck) {
5785 $conf = PVE::QemuConfig->load_config($vmid);
5786 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5787 if (!defined($timeout) && $shutdown && $conf->{startup}) {
5788 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5789 $timeout = $opts->{down} if $opts->{down};
5790 }
5791 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5792 }
5793
5794 eval {
5795 if ($shutdown) {
5796 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5797 vm_qmp_command($vmid, {
5798 execute => "guest-shutdown",
5799 arguments => { timeout => $timeout }
5800 }, $nocheck);
5801 } else {
5802 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5803 }
5804 } else {
5805 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5806 }
5807 };
5808 my $err = $@;
5809
5810 if (!$err) {
5811 $timeout = 60 if !defined($timeout);
5812
5813 my $count = 0;
5814 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5815 $count++;
5816 sleep 1;
5817 }
5818
5819 if ($count >= $timeout) {
5820 if ($force) {
5821 warn "VM still running - terminating now with SIGTERM\n";
5822 kill 15, $pid;
5823 } else {
5824 die "VM quit/powerdown failed - got timeout\n";
5825 }
5826 } else {
5827 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5828 return;
5829 }
5830 } else {
5831 if ($force) {
5832 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5833 kill 15, $pid;
5834 } else {
5835 die "VM quit/powerdown failed\n";
5836 }
5837 }
5838
5839 # wait again
5840 $timeout = 10;
5841
5842 my $count = 0;
5843 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5844 $count++;
5845 sleep 1;
5846 }
5847
5848 if ($count >= $timeout) {
5849 warn "VM still running - terminating now with SIGKILL\n";
5850 kill 9, $pid;
5851 sleep 1;
5852 }
5853
5854 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5855 }
5856
5857 # Note: use $nocheck to skip tests if VM configuration file exists.
5858 # We need that when migration VMs to other nodes (files already moved)
5859 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5860 sub vm_stop {
5861 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5862
5863 $force = 1 if !defined($force) && !$shutdown;
5864
5865 if ($migratedfrom){
5866 my $pid = check_running($vmid, $nocheck, $migratedfrom);
5867 kill 15, $pid if $pid;
5868 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5869 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5870 return;
5871 }
5872
5873 PVE::QemuConfig->lock_config($vmid, sub {
5874 _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5875 });
5876 }
5877
5878 sub vm_reboot {
5879 my ($vmid, $timeout) = @_;
5880
5881 PVE::QemuConfig->lock_config($vmid, sub {
5882
5883 # only reboot if running, as qmeventd starts it again on a stop event
5884 return if !check_running($vmid);
5885
5886 create_reboot_request($vmid);
5887
5888 my $storecfg = PVE::Storage::config();
5889 _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5890
5891 });
5892 }
5893
5894 sub vm_suspend {
5895 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5896
5897 my $conf;
5898 my $path;
5899 my $storecfg;
5900 my $vmstate;
5901
5902 PVE::QemuConfig->lock_config($vmid, sub {
5903
5904 $conf = PVE::QemuConfig->load_config($vmid);
5905
5906 my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5907 PVE::QemuConfig->check_lock($conf)
5908 if !($skiplock || $is_backing_up);
5909
5910 die "cannot suspend to disk during backup\n"
5911 if $is_backing_up && $includestate;
5912
5913 if ($includestate) {
5914 $conf->{lock} = 'suspending';
5915 my $date = strftime("%Y-%m-%d", localtime(time()));
5916 $storecfg = PVE::Storage::config();
5917 $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5918 $path = PVE::Storage::path($storecfg, $vmstate);
5919 PVE::QemuConfig->write_config($vmid, $conf);
5920 } else {
5921 vm_mon_cmd($vmid, "stop");
5922 }
5923 });
5924
5925 if ($includestate) {
5926 # save vm state
5927 PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5928
5929 eval {
5930 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5931 for(;;) {
5932 my $state = vm_mon_cmd_nocheck($vmid, "query-savevm");
5933 if (!$state->{status}) {
5934 die "savevm not active\n";
5935 } elsif ($state->{status} eq 'active') {
5936 sleep(1);
5937 next;
5938 } elsif ($state->{status} eq 'completed') {
5939 print "State saved, quitting\n";
5940 last;
5941 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5942 die "query-savevm failed with error '$state->{error}'\n"
5943 } else {
5944 die "query-savevm returned status '$state->{status}'\n";
5945 }
5946 }
5947 };
5948 my $err = $@;
5949
5950 PVE::QemuConfig->lock_config($vmid, sub {
5951 $conf = PVE::QemuConfig->load_config($vmid);
5952 if ($err) {
5953 # cleanup, but leave suspending lock, to indicate something went wrong
5954 eval {
5955 vm_mon_cmd($vmid, "savevm-end");
5956 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5957 PVE::Storage::vdisk_free($storecfg, $vmstate);
5958 delete $conf->@{qw(vmstate runningmachine)};
5959 PVE::QemuConfig->write_config($vmid, $conf);
5960 };
5961 warn $@ if $@;
5962 die $err;
5963 }
5964
5965 die "lock changed unexpectedly\n"
5966 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5967
5968 vm_qmp_command($vmid, { execute => "quit" });
5969 $conf->{lock} = 'suspended';
5970 PVE::QemuConfig->write_config($vmid, $conf);
5971 });
5972 }
5973 }
5974
5975 sub vm_resume {
5976 my ($vmid, $skiplock, $nocheck) = @_;
5977
5978 PVE::QemuConfig->lock_config($vmid, sub {
5979 my $vm_mon_cmd = $nocheck ? \&vm_mon_cmd_nocheck : \&vm_mon_cmd;
5980 my $res = $vm_mon_cmd->($vmid, 'query-status');
5981 my $resume_cmd = 'cont';
5982
5983 if ($res->{status} && $res->{status} eq 'suspended') {
5984 $resume_cmd = 'system_wakeup';
5985 }
5986
5987 if (!$nocheck) {
5988
5989 my $conf = PVE::QemuConfig->load_config($vmid);
5990
5991 PVE::QemuConfig->check_lock($conf)
5992 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5993 }
5994
5995 $vm_mon_cmd->($vmid, $resume_cmd);
5996 });
5997 }
5998
5999 sub vm_sendkey {
6000 my ($vmid, $skiplock, $key) = @_;
6001
6002 PVE::QemuConfig->lock_config($vmid, sub {
6003
6004 my $conf = PVE::QemuConfig->load_config($vmid);
6005
6006 # there is no qmp command, so we use the human monitor command
6007 my $res = vm_human_monitor_command($vmid, "sendkey $key");
6008 die $res if $res ne '';
6009 });
6010 }
6011
6012 sub vm_destroy {
6013 my ($storecfg, $vmid, $skiplock) = @_;
6014
6015 PVE::QemuConfig->lock_config($vmid, sub {
6016
6017 my $conf = PVE::QemuConfig->load_config($vmid);
6018
6019 if (!check_running($vmid)) {
6020 destroy_vm($storecfg, $vmid, undef, $skiplock);
6021 } else {
6022 die "VM $vmid is running - destroy failed\n";
6023 }
6024 });
6025 }
6026
6027 # vzdump restore implementaion
6028
6029 sub tar_archive_read_firstfile {
6030 my $archive = shift;
6031
6032 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
6033
6034 # try to detect archive type first
6035 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6036 die "unable to open file '$archive'\n";
6037 my $firstfile = <$fh>;
6038 kill 15, $pid;
6039 close $fh;
6040
6041 die "ERROR: archive contaions no data\n" if !$firstfile;
6042 chomp $firstfile;
6043
6044 return $firstfile;
6045 }
6046
6047 sub tar_restore_cleanup {
6048 my ($storecfg, $statfile) = @_;
6049
6050 print STDERR "starting cleanup\n";
6051
6052 if (my $fd = IO::File->new($statfile, "r")) {
6053 while (defined(my $line = <$fd>)) {
6054 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6055 my $volid = $2;
6056 eval {
6057 if ($volid =~ m|^/|) {
6058 unlink $volid || die 'unlink failed\n';
6059 } else {
6060 PVE::Storage::vdisk_free($storecfg, $volid);
6061 }
6062 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6063 };
6064 print STDERR "unable to cleanup '$volid' - $@" if $@;
6065 } else {
6066 print STDERR "unable to parse line in statfile - $line";
6067 }
6068 }
6069 $fd->close();
6070 }
6071 }
6072
6073 sub restore_archive {
6074 my ($archive, $vmid, $user, $opts) = @_;
6075
6076 my $format = $opts->{format};
6077 my $comp;
6078
6079 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
6080 $format = 'tar' if !$format;
6081 $comp = 'gzip';
6082 } elsif ($archive =~ m/\.tar$/) {
6083 $format = 'tar' if !$format;
6084 } elsif ($archive =~ m/.tar.lzo$/) {
6085 $format = 'tar' if !$format;
6086 $comp = 'lzop';
6087 } elsif ($archive =~ m/\.vma$/) {
6088 $format = 'vma' if !$format;
6089 } elsif ($archive =~ m/\.vma\.gz$/) {
6090 $format = 'vma' if !$format;
6091 $comp = 'gzip';
6092 } elsif ($archive =~ m/\.vma\.lzo$/) {
6093 $format = 'vma' if !$format;
6094 $comp = 'lzop';
6095 } else {
6096 $format = 'vma' if !$format; # default
6097 }
6098
6099 # try to detect archive format
6100 if ($format eq 'tar') {
6101 return restore_tar_archive($archive, $vmid, $user, $opts);
6102 } else {
6103 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
6104 }
6105 }
6106
6107 sub restore_update_config_line {
6108 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
6109
6110 return if $line =~ m/^\#qmdump\#/;
6111 return if $line =~ m/^\#vzdump\#/;
6112 return if $line =~ m/^lock:/;
6113 return if $line =~ m/^unused\d+:/;
6114 return if $line =~ m/^parent:/;
6115
6116 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
6117 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6118 # try to convert old 1.X settings
6119 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6120 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
6121 my ($model, $macaddr) = split(/\=/, $devconfig);
6122 $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
6123 my $net = {
6124 model => $model,
6125 bridge => "vmbr$ind",
6126 macaddr => $macaddr,
6127 };
6128 my $netstr = print_net($net);
6129
6130 print $outfd "net$cookie->{netcount}: $netstr\n";
6131 $cookie->{netcount}++;
6132 }
6133 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6134 my ($id, $netstr) = ($1, $2);
6135 my $net = parse_net($netstr);
6136 $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
6137 $netstr = print_net($net);
6138 print $outfd "$id: $netstr\n";
6139 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
6140 my $virtdev = $1;
6141 my $value = $3;
6142 my $di = parse_drive($virtdev, $value);
6143 if (defined($di->{backup}) && !$di->{backup}) {
6144 print $outfd "#$line";
6145 } elsif ($map->{$virtdev}) {
6146 delete $di->{format}; # format can change on restore
6147 $di->{file} = $map->{$virtdev};
6148 $value = print_drive($vmid, $di);
6149 print $outfd "$virtdev: $value\n";
6150 } else {
6151 print $outfd $line;
6152 }
6153 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6154 my $vmgenid = $1;
6155 if ($vmgenid ne '0') {
6156 # always generate a new vmgenid if there was a valid one setup
6157 $vmgenid = generate_uuid();
6158 }
6159 print $outfd "vmgenid: $vmgenid\n";
6160 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6161 my ($uuid, $uuid_str);
6162 UUID::generate($uuid);
6163 UUID::unparse($uuid, $uuid_str);
6164 my $smbios1 = parse_smbios1($2);
6165 $smbios1->{uuid} = $uuid_str;
6166 print $outfd $1.print_smbios1($smbios1)."\n";
6167 } else {
6168 print $outfd $line;
6169 }
6170 }
6171
6172 sub scan_volids {
6173 my ($cfg, $vmid) = @_;
6174
6175 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
6176
6177 my $volid_hash = {};
6178 foreach my $storeid (keys %$info) {
6179 foreach my $item (@{$info->{$storeid}}) {
6180 next if !($item->{volid} && $item->{size});
6181 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
6182 $volid_hash->{$item->{volid}} = $item;
6183 }
6184 }
6185
6186 return $volid_hash;
6187 }
6188
6189 sub is_volume_in_use {
6190 my ($storecfg, $conf, $skip_drive, $volid) = @_;
6191
6192 my $path = PVE::Storage::path($storecfg, $volid);
6193
6194 my $scan_config = sub {
6195 my ($cref, $snapname) = @_;
6196
6197 foreach my $key (keys %$cref) {
6198 my $value = $cref->{$key};
6199 if (is_valid_drivename($key)) {
6200 next if $skip_drive && $key eq $skip_drive;
6201 my $drive = parse_drive($key, $value);
6202 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
6203 return 1 if $volid eq $drive->{file};
6204 if ($drive->{file} =~ m!^/!) {
6205 return 1 if $drive->{file} eq $path;
6206 } else {
6207 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
6208 next if !$storeid;
6209 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
6210 next if !$scfg;
6211 return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
6212 }
6213 }
6214 }
6215
6216 return 0;
6217 };
6218
6219 return 1 if &$scan_config($conf);
6220
6221 undef $skip_drive;
6222
6223 foreach my $snapname (keys %{$conf->{snapshots}}) {
6224 return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
6225 }
6226
6227 return 0;
6228 }
6229
6230 sub update_disksize {
6231 my ($vmid, $conf, $volid_hash) = @_;
6232
6233 my $changes;
6234 my $prefix = "VM $vmid:";
6235
6236 # used and unused disks
6237 my $referenced = {};
6238
6239 # Note: it is allowed to define multiple storages with same path (alias), so
6240 # we need to check both 'volid' and real 'path' (two different volid can point
6241 # to the same path).
6242
6243 my $referencedpath = {};
6244
6245 # update size info
6246 foreach my $opt (keys %$conf) {
6247 if (is_valid_drivename($opt)) {
6248 my $drive = parse_drive($opt, $conf->{$opt});
6249 my $volid = $drive->{file};
6250 next if !$volid;
6251
6252 $referenced->{$volid} = 1;
6253 if ($volid_hash->{$volid} &&
6254 (my $path = $volid_hash->{$volid}->{path})) {
6255 $referencedpath->{$path} = 1;
6256 }
6257
6258 next if drive_is_cdrom($drive);
6259 next if !$volid_hash->{$volid};
6260
6261 $drive->{size} = $volid_hash->{$volid}->{size};
6262 my $new = print_drive($vmid, $drive);
6263 if ($new ne $conf->{$opt}) {
6264 $changes = 1;
6265 $conf->{$opt} = $new;
6266 print "$prefix update disk '$opt' information.\n";
6267 }
6268 }
6269 }
6270
6271 # remove 'unusedX' entry if volume is used
6272 foreach my $opt (keys %$conf) {
6273 next if $opt !~ m/^unused\d+$/;
6274 my $volid = $conf->{$opt};
6275 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
6276 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6277 print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
6278 $changes = 1;
6279 delete $conf->{$opt};
6280 }
6281
6282 $referenced->{$volid} = 1;
6283 $referencedpath->{$path} = 1 if $path;
6284 }
6285
6286 foreach my $volid (sort keys %$volid_hash) {
6287 next if $volid =~ m/vm-$vmid-state-/;
6288 next if $referenced->{$volid};
6289 my $path = $volid_hash->{$volid}->{path};
6290 next if !$path; # just to be sure
6291 next if $referencedpath->{$path};
6292 $changes = 1;
6293 my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
6294 print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
6295 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6296 }
6297
6298 return $changes;
6299 }
6300
6301 sub rescan {
6302 my ($vmid, $nolock, $dryrun) = @_;
6303
6304 my $cfg = PVE::Storage::config();
6305
6306 # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
6307 # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
6308 foreach my $stor (keys %{$cfg->{ids}}) {
6309 delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
6310 }
6311
6312 print "rescan volumes...\n";
6313 my $volid_hash = scan_volids($cfg, $vmid);
6314
6315 my $updatefn = sub {
6316 my ($vmid) = @_;
6317
6318 my $conf = PVE::QemuConfig->load_config($vmid);
6319
6320 PVE::QemuConfig->check_lock($conf);
6321
6322 my $vm_volids = {};
6323 foreach my $volid (keys %$volid_hash) {
6324 my $info = $volid_hash->{$volid};
6325 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
6326 }
6327
6328 my $changes = update_disksize($vmid, $conf, $vm_volids);
6329
6330 PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
6331 };
6332
6333 if (defined($vmid)) {
6334 if ($nolock) {
6335 &$updatefn($vmid);
6336 } else {
6337 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6338 }
6339 } else {
6340 my $vmlist = config_list();
6341 foreach my $vmid (keys %$vmlist) {
6342 if ($nolock) {
6343 &$updatefn($vmid);
6344 } else {
6345 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6346 }
6347 }
6348 }
6349 }
6350
6351 sub restore_vma_archive {
6352 my ($archive, $vmid, $user, $opts, $comp) = @_;
6353
6354 my $readfrom = $archive;
6355
6356 my $cfg = PVE::Storage::config();
6357 my $commands = [];
6358 my $bwlimit = $opts->{bwlimit};
6359
6360 my $dbg_cmdstring = '';
6361 my $add_pipe = sub {
6362 my ($cmd) = @_;
6363 push @$commands, $cmd;
6364 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6365 $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6366 $readfrom = '-';
6367 };
6368
6369 my $input = undef;
6370 if ($archive eq '-') {
6371 $input = '<&STDIN';
6372 } else {
6373 # If we use a backup from a PVE defined storage we also consider that
6374 # storage's rate limit:
6375 my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6376 if (defined($volid)) {
6377 my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6378 my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6379 if ($readlimit) {
6380 print STDERR "applying read rate limit: $readlimit\n";
6381 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6382 $add_pipe->($cstream);
6383 }
6384 }
6385 }
6386
6387 if ($comp) {
6388 my $cmd;
6389 if ($comp eq 'gzip') {
6390 $cmd = ['zcat', $readfrom];
6391 } elsif ($comp eq 'lzop') {
6392 $cmd = ['lzop', '-d', '-c', $readfrom];
6393 } else {
6394 die "unknown compression method '$comp'\n";
6395 }
6396 $add_pipe->($cmd);
6397 }
6398
6399 my $tmpdir = "/var/tmp/vzdumptmp$$";
6400 rmtree $tmpdir;
6401
6402 # disable interrupts (always do cleanups)
6403 local $SIG{INT} =
6404 local $SIG{TERM} =
6405 local $SIG{QUIT} =
6406 local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6407
6408 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6409 POSIX::mkfifo($mapfifo, 0600);
6410 my $fifofh;
6411
6412 my $openfifo = sub {
6413 open($fifofh, '>', $mapfifo) || die $!;
6414 };
6415
6416 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6417
6418 my $oldtimeout;
6419 my $timeout = 5;
6420
6421 my $devinfo = {};
6422
6423 my $rpcenv = PVE::RPCEnvironment::get();
6424
6425 my $conffile = PVE::QemuConfig->config_file($vmid);
6426 my $tmpfn = "$conffile.$$.tmp";
6427
6428 # Note: $oldconf is undef if VM does not exists
6429 my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6430 my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6431
6432 my %storage_limits;
6433
6434 my $print_devmap = sub {
6435 my $virtdev_hash = {};
6436
6437 my $cfgfn = "$tmpdir/qemu-server.conf";
6438
6439 # we can read the config - that is already extracted
6440 my $fh = IO::File->new($cfgfn, "r") ||
6441 "unable to read qemu-server.conf - $!\n";
6442
6443 my $fwcfgfn = "$tmpdir/qemu-server.fw";
6444 if (-f $fwcfgfn) {
6445 my $pve_firewall_dir = '/etc/pve/firewall';
6446 mkdir $pve_firewall_dir; # make sure the dir exists
6447 PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6448 }
6449
6450 while (defined(my $line = <$fh>)) {
6451 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6452 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6453 die "archive does not contain data for drive '$virtdev'\n"
6454 if !$devinfo->{$devname};
6455 if (defined($opts->{storage})) {
6456 $storeid = $opts->{storage} || 'local';
6457 } elsif (!$storeid) {
6458 $storeid = 'local';
6459 }
6460 $format = 'raw' if !$format;
6461 $devinfo->{$devname}->{devname} = $devname;
6462 $devinfo->{$devname}->{virtdev} = $virtdev;
6463 $devinfo->{$devname}->{format} = $format;
6464 $devinfo->{$devname}->{storeid} = $storeid;
6465
6466 # check permission on storage
6467 my $pool = $opts->{pool}; # todo: do we need that?
6468 if ($user ne 'root@pam') {
6469 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6470 }
6471
6472 $storage_limits{$storeid} = $bwlimit;
6473
6474 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6475 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6476 my $virtdev = $1;
6477 my $drive = parse_drive($virtdev, $2);
6478 if (drive_is_cloudinit($drive)) {
6479 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6480 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6481 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
6482
6483 my $d = {
6484 format => $format,
6485 storeid => $opts->{storage} // $storeid,
6486 size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
6487 file => $drive->{file}, # to make drive_is_cloudinit check possible
6488 name => "vm-$vmid-cloudinit",
6489 is_cloudinit => 1,
6490 };
6491 $virtdev_hash->{$virtdev} = $d;
6492 }
6493 }
6494 }
6495
6496 foreach my $key (keys %storage_limits) {
6497 my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6498 next if !$limit;
6499 print STDERR "rate limit for storage $key: $limit KiB/s\n";
6500 $storage_limits{$key} = $limit * 1024;
6501 }
6502
6503 foreach my $devname (keys %$devinfo) {
6504 die "found no device mapping information for device '$devname'\n"
6505 if !$devinfo->{$devname}->{virtdev};
6506 }
6507
6508 # create empty/temp config
6509 if ($oldconf) {
6510 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6511 foreach_drive($oldconf, sub {
6512 my ($ds, $drive) = @_;
6513
6514 return if drive_is_cdrom($drive, 1);
6515
6516 my $volid = $drive->{file};
6517 return if !$volid || $volid =~ m|^/|;
6518
6519 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6520 return if !$path || !$owner || ($owner != $vmid);
6521
6522 # Note: only delete disk we want to restore
6523 # other volumes will become unused
6524 if ($virtdev_hash->{$ds}) {
6525 eval { PVE::Storage::vdisk_free($cfg, $volid); };
6526 if (my $err = $@) {
6527 warn $err;
6528 }
6529 }
6530 });
6531
6532 # delete vmstate files, after the restore we have no snapshots anymore
6533 foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6534 my $snap = $oldconf->{snapshots}->{$snapname};
6535 if ($snap->{vmstate}) {
6536 eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6537 if (my $err = $@) {
6538 warn $err;
6539 }
6540 }
6541 }
6542 }
6543
6544 my $map = {};
6545 foreach my $virtdev (sort keys %$virtdev_hash) {
6546 my $d = $virtdev_hash->{$virtdev};
6547 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6548 my $storeid = $d->{storeid};
6549 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6550
6551 my $map_opts = '';
6552 if (my $limit = $storage_limits{$storeid}) {
6553 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6554 }
6555
6556 # test if requested format is supported
6557 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6558 my $supported = grep { $_ eq $d->{format} } @$validFormats;
6559 $d->{format} = $defFormat if !$supported;
6560
6561 my $name;
6562 if ($d->{is_cloudinit}) {
6563 $name = $d->{name};
6564 $name .= ".$d->{format}" if $d->{format} ne 'raw';
6565 }
6566
6567 my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
6568 print STDERR "new volume ID is '$volid'\n";
6569 $d->{volid} = $volid;
6570
6571 PVE::Storage::activate_volumes($cfg, [$volid]);
6572
6573 my $write_zeros = 1;
6574 if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6575 $write_zeros = 0;
6576 }
6577
6578 if (!$d->{is_cloudinit}) {
6579 my $path = PVE::Storage::path($cfg, $volid);
6580
6581 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6582
6583 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6584 }
6585 $map->{$virtdev} = $volid;
6586 }
6587
6588 $fh->seek(0, 0) || die "seek failed - $!\n";
6589
6590 my $outfd = new IO::File ($tmpfn, "w") ||
6591 die "unable to write config for VM $vmid\n";
6592
6593 my $cookie = { netcount => 0 };
6594 while (defined(my $line = <$fh>)) {
6595 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6596 }
6597
6598 $fh->close();
6599 $outfd->close();
6600 };
6601
6602 eval {
6603 # enable interrupts
6604 local $SIG{INT} =
6605 local $SIG{TERM} =
6606 local $SIG{QUIT} =
6607 local $SIG{HUP} =
6608 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6609 local $SIG{ALRM} = sub { die "got timeout\n"; };
6610
6611 $oldtimeout = alarm($timeout);
6612
6613 my $parser = sub {
6614 my $line = shift;
6615
6616 print "$line\n";
6617
6618 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6619 my ($dev_id, $size, $devname) = ($1, $2, $3);
6620 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6621 } elsif ($line =~ m/^CTIME: /) {
6622 # we correctly received the vma config, so we can disable
6623 # the timeout now for disk allocation (set to 10 minutes, so
6624 # that we always timeout if something goes wrong)
6625 alarm(600);
6626 &$print_devmap();
6627 print $fifofh "done\n";
6628 my $tmp = $oldtimeout || 0;
6629 $oldtimeout = undef;
6630 alarm($tmp);
6631 close($fifofh);
6632 }
6633 };
6634
6635 print "restore vma archive: $dbg_cmdstring\n";
6636 run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6637 };
6638 my $err = $@;
6639
6640 alarm($oldtimeout) if $oldtimeout;
6641
6642 my $vollist = [];
6643 foreach my $devname (keys %$devinfo) {
6644 my $volid = $devinfo->{$devname}->{volid};
6645 push @$vollist, $volid if $volid;
6646 }
6647
6648 PVE::Storage::deactivate_volumes($cfg, $vollist);
6649
6650 unlink $mapfifo;
6651
6652 if ($err) {
6653 rmtree $tmpdir;
6654 unlink $tmpfn;
6655
6656 foreach my $devname (keys %$devinfo) {
6657 my $volid = $devinfo->{$devname}->{volid};
6658 next if !$volid;
6659 eval {
6660 if ($volid =~ m|^/|) {
6661 unlink $volid || die 'unlink failed\n';
6662 } else {
6663 PVE::Storage::vdisk_free($cfg, $volid);
6664 }
6665 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6666 };
6667 print STDERR "unable to cleanup '$volid' - $@" if $@;
6668 }
6669 die $err;
6670 }
6671
6672 rmtree $tmpdir;
6673
6674 rename($tmpfn, $conffile) ||
6675 die "unable to commit configuration file '$conffile'\n";
6676
6677 PVE::Cluster::cfs_update(); # make sure we read new file
6678
6679 eval { rescan($vmid, 1); };
6680 warn $@ if $@;
6681 }
6682
6683 sub restore_tar_archive {
6684 my ($archive, $vmid, $user, $opts) = @_;
6685
6686 if ($archive ne '-') {
6687 my $firstfile = tar_archive_read_firstfile($archive);
6688 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6689 if $firstfile ne 'qemu-server.conf';
6690 }
6691
6692 my $storecfg = PVE::Storage::config();
6693
6694 # destroy existing data - keep empty config
6695 my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6696 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6697
6698 my $tocmd = "/usr/lib/qemu-server/qmextract";
6699
6700 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6701 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6702 $tocmd .= ' --prealloc' if $opts->{prealloc};
6703 $tocmd .= ' --info' if $opts->{info};
6704
6705 # tar option "xf" does not autodetect compression when read from STDIN,
6706 # so we pipe to zcat
6707 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6708 PVE::Tools::shellquote("--to-command=$tocmd");
6709
6710 my $tmpdir = "/var/tmp/vzdumptmp$$";
6711 mkpath $tmpdir;
6712
6713 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6714 local $ENV{VZDUMP_VMID} = $vmid;
6715 local $ENV{VZDUMP_USER} = $user;
6716
6717 my $conffile = PVE::QemuConfig->config_file($vmid);
6718 my $tmpfn = "$conffile.$$.tmp";
6719
6720 # disable interrupts (always do cleanups)
6721 local $SIG{INT} =
6722 local $SIG{TERM} =
6723 local $SIG{QUIT} =
6724 local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6725
6726 eval {
6727 # enable interrupts
6728 local $SIG{INT} =
6729 local $SIG{TERM} =
6730 local $SIG{QUIT} =
6731 local $SIG{HUP} =
6732 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6733
6734 if ($archive eq '-') {
6735 print "extracting archive from STDIN\n";
6736 run_command($cmd, input => "<&STDIN");
6737 } else {
6738 print "extracting archive '$archive'\n";
6739 run_command($cmd);
6740 }
6741
6742 return if $opts->{info};
6743
6744 # read new mapping
6745 my $map = {};
6746 my $statfile = "$tmpdir/qmrestore.stat";
6747 if (my $fd = IO::File->new($statfile, "r")) {
6748 while (defined (my $line = <$fd>)) {
6749 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6750 $map->{$1} = $2 if $1;
6751 } else {
6752 print STDERR "unable to parse line in statfile - $line\n";
6753 }
6754 }
6755 $fd->close();
6756 }
6757
6758 my $confsrc = "$tmpdir/qemu-server.conf";
6759
6760 my $srcfd = new IO::File($confsrc, "r") ||
6761 die "unable to open file '$confsrc'\n";
6762
6763 my $outfd = new IO::File ($tmpfn, "w") ||
6764 die "unable to write config for VM $vmid\n";
6765
6766 my $cookie = { netcount => 0 };
6767 while (defined (my $line = <$srcfd>)) {
6768 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6769 }
6770
6771 $srcfd->close();
6772 $outfd->close();
6773 };
6774 my $err = $@;
6775
6776 if ($err) {
6777
6778 unlink $tmpfn;
6779
6780 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6781
6782 die $err;
6783 }
6784
6785 rmtree $tmpdir;
6786
6787 rename $tmpfn, $conffile ||
6788 die "unable to commit configuration file '$conffile'\n";
6789
6790 PVE::Cluster::cfs_update(); # make sure we read new file
6791
6792 eval { rescan($vmid, 1); };
6793 warn $@ if $@;
6794 };
6795
6796 sub foreach_storage_used_by_vm {
6797 my ($conf, $func) = @_;
6798
6799 my $sidhash = {};
6800
6801 foreach_drive($conf, sub {
6802 my ($ds, $drive) = @_;
6803 return if drive_is_cdrom($drive);
6804
6805 my $volid = $drive->{file};
6806
6807 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6808 $sidhash->{$sid} = $sid if $sid;
6809 });
6810
6811 foreach my $sid (sort keys %$sidhash) {
6812 &$func($sid);
6813 }
6814 }
6815
6816 sub do_snapshots_with_qemu {
6817 my ($storecfg, $volid) = @_;
6818
6819 my $storage_name = PVE::Storage::parse_volume_id($volid);
6820 my $scfg = $storecfg->{ids}->{$storage_name};
6821
6822 if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6823 return 1;
6824 }
6825
6826 if ($volid =~ m/\.(qcow2|qed)$/){
6827 return 1;
6828 }
6829
6830 return undef;
6831 }
6832
6833 sub qga_check_running {
6834 my ($vmid, $nowarn) = @_;
6835
6836 eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6837 if ($@) {
6838 warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6839 return 0;
6840 }
6841 return 1;
6842 }
6843
6844 sub template_create {
6845 my ($vmid, $conf, $disk) = @_;
6846
6847 my $storecfg = PVE::Storage::config();
6848
6849 foreach_drive($conf, sub {
6850 my ($ds, $drive) = @_;
6851
6852 return if drive_is_cdrom($drive);
6853 return if $disk && $ds ne $disk;
6854
6855 my $volid = $drive->{file};
6856 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6857
6858 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6859 $drive->{file} = $voliddst;
6860 $conf->{$ds} = print_drive($vmid, $drive);
6861 PVE::QemuConfig->write_config($vmid, $conf);
6862 });
6863 }
6864
6865 sub convert_iscsi_path {
6866 my ($path) = @_;
6867
6868 if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6869 my $portal = $1;
6870 my $target = $2;
6871 my $lun = $3;
6872
6873 my $initiator_name = get_initiator_name();
6874
6875 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6876 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6877 }
6878
6879 die "cannot convert iscsi path '$path', unkown format\n";
6880 }
6881
6882 sub qemu_img_convert {
6883 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6884
6885 my $storecfg = PVE::Storage::config();
6886 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6887 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6888
6889 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6890
6891 my $cachemode;
6892 my $src_path;
6893 my $src_is_iscsi = 0;
6894 my $src_format = 'raw';
6895
6896 if ($src_storeid) {
6897 PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6898 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6899 $src_format = qemu_img_format($src_scfg, $src_volname);
6900 $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6901 $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6902 $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6903 } elsif (-f $src_volid) {
6904 $src_path = $src_volid;
6905 if ($src_path =~ m/\.($QEMU_FORMAT_RE)$/) {
6906 $src_format = $1;
6907 }
6908 }
6909
6910 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6911
6912 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6913 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6914 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6915 my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6916
6917 my $cmd = [];
6918 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6919 push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6920 push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6921 push @$cmd, '-T', $cachemode if defined($cachemode);
6922
6923 if ($src_is_iscsi) {
6924 push @$cmd, '--image-opts';
6925 $src_path = convert_iscsi_path($src_path);
6926 } else {
6927 push @$cmd, '-f', $src_format;
6928 }
6929
6930 if ($dst_is_iscsi) {
6931 push @$cmd, '--target-image-opts';
6932 $dst_path = convert_iscsi_path($dst_path);
6933 } else {
6934 push @$cmd, '-O', $dst_format;
6935 }
6936
6937 push @$cmd, $src_path;
6938
6939 if (!$dst_is_iscsi && $is_zero_initialized) {
6940 push @$cmd, "zeroinit:$dst_path";
6941 } else {
6942 push @$cmd, $dst_path;
6943 }
6944
6945 my $parser = sub {
6946 my $line = shift;
6947 if($line =~ m/\((\S+)\/100\%\)/){
6948 my $percent = $1;
6949 my $transferred = int($size * $percent / 100);
6950 my $remaining = $size - $transferred;
6951
6952 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6953 }
6954
6955 };
6956
6957 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6958 my $err = $@;
6959 die "copy failed: $err" if $err;
6960 }
6961
6962 sub qemu_img_format {
6963 my ($scfg, $volname) = @_;
6964
6965 if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6966 return $1;
6967 } else {
6968 return "raw";
6969 }
6970 }
6971
6972 sub qemu_drive_mirror {
6973 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
6974
6975 $jobs = {} if !$jobs;
6976
6977 my $qemu_target;
6978 my $format;
6979 $jobs->{"drive-$drive"} = {};
6980
6981 if ($dst_volid =~ /^nbd:/) {
6982 $qemu_target = $dst_volid;
6983 $format = "nbd";
6984 } else {
6985 my $storecfg = PVE::Storage::config();
6986 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6987
6988 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6989
6990 $format = qemu_img_format($dst_scfg, $dst_volname);
6991
6992 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6993
6994 $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6995 }
6996
6997 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6998 $opts->{format} = $format if $format;
6999
7000 if (defined($bwlimit)) {
7001 $opts->{speed} = $bwlimit * 1024;
7002 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7003 } else {
7004 print "drive mirror is starting for drive-$drive\n";
7005 }
7006
7007 # if a job already runs for this device we get an error, catch it for cleanup
7008 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
7009 if (my $err = $@) {
7010 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7011 warn "$@\n" if $@;
7012 die "mirroring error: $err\n";
7013 }
7014
7015 qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
7016 }
7017
7018 sub qemu_drive_mirror_monitor {
7019 my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
7020
7021 eval {
7022 my $err_complete = 0;
7023
7024 while (1) {
7025 die "storage migration timed out\n" if $err_complete > 300;
7026
7027 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7028
7029 my $running_mirror_jobs = {};
7030 foreach my $stat (@$stats) {
7031 next if $stat->{type} ne 'mirror';
7032 $running_mirror_jobs->{$stat->{device}} = $stat;
7033 }
7034
7035 my $readycounter = 0;
7036
7037 foreach my $job (keys %$jobs) {
7038
7039 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
7040 print "$job : finished\n";
7041 delete $jobs->{$job};
7042 next;
7043 }
7044
7045 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
7046
7047 my $busy = $running_mirror_jobs->{$job}->{busy};
7048 my $ready = $running_mirror_jobs->{$job}->{ready};
7049 if (my $total = $running_mirror_jobs->{$job}->{len}) {
7050 my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
7051 my $remaining = $total - $transferred;
7052 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7053
7054 print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
7055 }
7056
7057 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
7058 }
7059
7060 last if scalar(keys %$jobs) == 0;
7061
7062 if ($readycounter == scalar(keys %$jobs)) {
7063 print "all mirroring jobs are ready \n";
7064 last if $skipcomplete; #do the complete later
7065
7066 if ($vmiddst && $vmiddst != $vmid) {
7067 my $agent_running = $qga && qga_check_running($vmid);
7068 if ($agent_running) {
7069 print "freeze filesystem\n";
7070 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
7071 } else {
7072 print "suspend vm\n";
7073 eval { PVE::QemuServer::vm_suspend($vmid, 1); };
7074 }
7075
7076 # if we clone a disk for a new target vm, we don't switch the disk
7077 PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
7078
7079 if ($agent_running) {
7080 print "unfreeze filesystem\n";
7081 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
7082 } else {
7083 print "resume vm\n";
7084 eval { PVE::QemuServer::vm_resume($vmid, 1, 1); };
7085 }
7086
7087 last;
7088 } else {
7089
7090 foreach my $job (keys %$jobs) {
7091 # try to switch the disk if source and destination are on the same guest
7092 print "$job: Completing block job...\n";
7093
7094 eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
7095 if ($@ =~ m/cannot be completed/) {
7096 print "$job: Block job cannot be completed, try again.\n";
7097 $err_complete++;
7098 }else {
7099 print "$job: Completed successfully.\n";
7100 $jobs->{$job}->{complete} = 1;
7101 }
7102 }
7103 }
7104 }
7105 sleep 1;
7106 }
7107 };
7108 my $err = $@;
7109
7110 if ($err) {
7111 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7112 die "mirroring error: $err";
7113 }
7114
7115 }
7116
7117 sub qemu_blockjobs_cancel {
7118 my ($vmid, $jobs) = @_;
7119
7120 foreach my $job (keys %$jobs) {
7121 print "$job: Cancelling block job\n";
7122 eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
7123 $jobs->{$job}->{cancel} = 1;
7124 }
7125
7126 while (1) {
7127 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7128
7129 my $running_jobs = {};
7130 foreach my $stat (@$stats) {
7131 $running_jobs->{$stat->{device}} = $stat;
7132 }
7133
7134 foreach my $job (keys %$jobs) {
7135
7136 if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
7137 print "$job: Done.\n";
7138 delete $jobs->{$job};
7139 }
7140 }
7141
7142 last if scalar(keys %$jobs) == 0;
7143
7144 sleep 1;
7145 }
7146 }
7147
7148 sub clone_disk {
7149 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
7150 $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
7151
7152 my $newvolid;
7153
7154 if (!$full) {
7155 print "create linked clone of drive $drivename ($drive->{file})\n";
7156 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
7157 push @$newvollist, $newvolid;
7158 } else {
7159
7160 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
7161 $storeid = $storage if $storage;
7162
7163 my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
7164 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
7165
7166 print "create full clone of drive $drivename ($drive->{file})\n";
7167 my $name = undef;
7168 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
7169 push @$newvollist, $newvolid;
7170
7171 PVE::Storage::activate_volumes($storecfg, [$newvolid]);
7172
7173 my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
7174 if (!$running || $snapname) {
7175 # TODO: handle bwlimits
7176 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
7177 } else {
7178
7179 my $kvmver = get_running_qemu_version ($vmid);
7180 if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
7181 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
7182 if $drive->{iothread};
7183 }
7184
7185 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga, $bwlimit);
7186 }
7187 }
7188
7189 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
7190
7191 my $disk = $drive;
7192 $disk->{format} = undef;
7193 $disk->{file} = $newvolid;
7194 $disk->{size} = $size;
7195
7196 return $disk;
7197 }
7198
7199 # this only works if VM is running
7200 sub get_current_qemu_machine {
7201 my ($vmid) = @_;
7202
7203 my $cmd = { execute => 'query-machines', arguments => {} };
7204 my $res = vm_qmp_command($vmid, $cmd);
7205
7206 my ($current, $default);
7207 foreach my $e (@$res) {
7208 $default = $e->{name} if $e->{'is-default'};
7209 $current = $e->{name} if $e->{'is-current'};
7210 }
7211
7212 # fallback to the default machine if current is not supported by qemu
7213 return $current || $default || 'pc';
7214 }
7215
7216 sub get_running_qemu_version {
7217 my ($vmid) = @_;
7218 my $cmd = { execute => 'query-version', arguments => {} };
7219 my $res = vm_qmp_command($vmid, $cmd);
7220 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
7221 }
7222
7223 sub qemu_machine_feature_enabled {
7224 my ($machine, $kvmver, $version_major, $version_minor) = @_;
7225
7226 my $current_major;
7227 my $current_minor;
7228
7229 if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
7230
7231 $current_major = $3;
7232 $current_minor = $4;
7233
7234 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
7235
7236 $current_major = $1;
7237 $current_minor = $2;
7238 }
7239
7240 return 1 if $current_major > $version_major ||
7241 ($current_major == $version_major &&
7242 $current_minor >= $version_minor);
7243 }
7244
7245 sub qemu_machine_pxe {
7246 my ($vmid, $conf) = @_;
7247
7248 my $machine = PVE::QemuServer::get_current_qemu_machine($vmid);
7249
7250 if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
7251 $machine .= '.pxe';
7252 }
7253
7254 return $machine;
7255 }
7256
7257 sub qemu_use_old_bios_files {
7258 my ($machine_type) = @_;
7259
7260 return if !$machine_type;
7261
7262 my $use_old_bios_files = undef;
7263
7264 if ($machine_type =~ m/^(\S+)\.pxe$/) {
7265 $machine_type = $1;
7266 $use_old_bios_files = 1;
7267 } else {
7268 my $kvmver = kvm_user_version();
7269 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
7270 # load new efi bios files on migration. So this hack is required to allow
7271 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
7272 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
7273 $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
7274 }
7275
7276 return ($use_old_bios_files, $machine_type);
7277 }
7278
7279 sub create_efidisk($$$$$) {
7280 my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7281
7282 my (undef, $ovmf_vars) = get_ovmf_files($arch);
7283 die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7284
7285 my $vars_size_b = -s $ovmf_vars;
7286 my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7287 my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7288 PVE::Storage::activate_volumes($storecfg, [$volid]);
7289
7290 qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7291
7292 return ($volid, $vars_size);
7293 }
7294
7295 sub vm_iothreads_list {
7296 my ($vmid) = @_;
7297
7298 my $res = vm_mon_cmd($vmid, 'query-iothreads');
7299
7300 my $iothreads = {};
7301 foreach my $iothread (@$res) {
7302 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7303 }
7304
7305 return $iothreads;
7306 }
7307
7308 sub scsihw_infos {
7309 my ($conf, $drive) = @_;
7310
7311 my $maxdev = 0;
7312
7313 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7314 $maxdev = 7;
7315 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7316 $maxdev = 1;
7317 } else {
7318 $maxdev = 256;
7319 }
7320
7321 my $controller = int($drive->{index} / $maxdev);
7322 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7323
7324 return ($maxdev, $controller, $controller_prefix);
7325 }
7326
7327 sub add_hyperv_enlightenments {
7328 my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough, $hv_vendor_id) = @_;
7329
7330 return if $winversion < 6;
7331 return if $bios && $bios eq 'ovmf' && $winversion < 8;
7332
7333 if ($gpu_passthrough || defined($hv_vendor_id)) {
7334 $hv_vendor_id //= 'proxmox';
7335 push @$cpuFlags , "hv_vendor_id=$hv_vendor_id";
7336 }
7337
7338 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
7339 push @$cpuFlags , 'hv_spinlocks=0x1fff';
7340 push @$cpuFlags , 'hv_vapic';
7341 push @$cpuFlags , 'hv_time';
7342 } else {
7343 push @$cpuFlags , 'hv_spinlocks=0xffff';
7344 }
7345
7346 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
7347 push @$cpuFlags , 'hv_reset';
7348 push @$cpuFlags , 'hv_vpindex';
7349 push @$cpuFlags , 'hv_runtime';
7350 }
7351
7352 if ($winversion >= 7) {
7353 push @$cpuFlags , 'hv_relaxed';
7354
7355 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
7356 push @$cpuFlags , 'hv_synic';
7357 push @$cpuFlags , 'hv_stimer';
7358 }
7359
7360 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 3, 1)) {
7361 push @$cpuFlags , 'hv_ipi';
7362 }
7363 }
7364 }
7365
7366 sub windows_version {
7367 my ($ostype) = @_;
7368
7369 return 0 if !$ostype;
7370
7371 my $winversion = 0;
7372
7373 if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7374 $winversion = 5;
7375 } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7376 $winversion = 6;
7377 } elsif ($ostype =~ m/^win(\d+)$/) {
7378 $winversion = $1;
7379 }
7380
7381 return $winversion;
7382 }
7383
7384 sub resolve_dst_disk_format {
7385 my ($storecfg, $storeid, $src_volname, $format) = @_;
7386 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7387
7388 if (!$format) {
7389 # if no target format is specified, use the source disk format as hint
7390 if ($src_volname) {
7391 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7392 $format = qemu_img_format($scfg, $src_volname);
7393 } else {
7394 return $defFormat;
7395 }
7396 }
7397
7398 # test if requested format is supported - else use default
7399 my $supported = grep { $_ eq $format } @$validFormats;
7400 $format = $defFormat if !$supported;
7401 return $format;
7402 }
7403
7404 sub resolve_first_disk {
7405 my $conf = shift;
7406 my @disks = PVE::QemuServer::valid_drive_names();
7407 my $firstdisk;
7408 foreach my $ds (reverse @disks) {
7409 next if !$conf->{$ds};
7410 my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
7411 next if PVE::QemuServer::drive_is_cdrom($disk);
7412 $firstdisk = $ds;
7413 }
7414 return $firstdisk;
7415 }
7416
7417 sub generate_uuid {
7418 my ($uuid, $uuid_str);
7419 UUID::generate($uuid);
7420 UUID::unparse($uuid, $uuid_str);
7421 return $uuid_str;
7422 }
7423
7424 sub generate_smbios1_uuid {
7425 return "uuid=".generate_uuid();
7426 }
7427
7428 sub nbd_stop {
7429 my ($vmid) = @_;
7430
7431 vm_mon_cmd($vmid, 'nbd-server-stop');
7432 }
7433
7434 sub create_reboot_request {
7435 my ($vmid) = @_;
7436 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7437 or die "failed to create reboot trigger file: $!\n";
7438 close($fh);
7439 }
7440
7441 sub clear_reboot_request {
7442 my ($vmid) = @_;
7443 my $path = "/run/qemu-server/$vmid.reboot";
7444 my $res = 0;
7445
7446 $res = unlink($path);
7447 die "could not remove reboot request for $vmid: $!"
7448 if !$res && $! != POSIX::ENOENT;
7449
7450 return $res;
7451 }
7452
7453 # bash completion helper
7454
7455 sub complete_backup_archives {
7456 my ($cmdname, $pname, $cvalue) = @_;
7457
7458 my $cfg = PVE::Storage::config();
7459
7460 my $storeid;
7461
7462 if ($cvalue =~ m/^([^:]+):/) {
7463 $storeid = $1;
7464 }
7465
7466 my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7467
7468 my $res = [];
7469 foreach my $id (keys %$data) {
7470 foreach my $item (@{$data->{$id}}) {
7471 next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7472 push @$res, $item->{volid} if defined($item->{volid});
7473 }
7474 }
7475
7476 return $res;
7477 }
7478
7479 my $complete_vmid_full = sub {
7480 my ($running) = @_;
7481
7482 my $idlist = vmstatus();
7483
7484 my $res = [];
7485
7486 foreach my $id (keys %$idlist) {
7487 my $d = $idlist->{$id};
7488 if (defined($running)) {
7489 next if $d->{template};
7490 next if $running && $d->{status} ne 'running';
7491 next if !$running && $d->{status} eq 'running';
7492 }
7493 push @$res, $id;
7494
7495 }
7496 return $res;
7497 };
7498
7499 sub complete_vmid {
7500 return &$complete_vmid_full();
7501 }
7502
7503 sub complete_vmid_stopped {
7504 return &$complete_vmid_full(0);
7505 }
7506
7507 sub complete_vmid_running {
7508 return &$complete_vmid_full(1);
7509 }
7510
7511 sub complete_storage {
7512
7513 my $cfg = PVE::Storage::config();
7514 my $ids = $cfg->{ids};
7515
7516 my $res = [];
7517 foreach my $sid (keys %$ids) {
7518 next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7519 next if !$ids->{$sid}->{content}->{images};
7520 push @$res, $sid;
7521 }
7522
7523 return $res;
7524 }
7525
7526 1;