]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
api: use shared methods in config GET
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5
6 use Cwd 'abs_path';
7 use Digest::SHA;
8 use Fcntl ':flock';
9 use Fcntl;
10 use File::Basename;
11 use File::Copy qw(copy);
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use IO::Dir;
16 use IO::File;
17 use IO::Handle;
18 use IO::Select;
19 use IO::Socket::UNIX;
20 use IPC::Open3;
21 use JSON;
22 use MIME::Base64;
23 use POSIX;
24 use Storable qw(dclone);
25 use Time::HiRes qw(gettimeofday);
26 use URI::Escape;
27
28 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
29 use PVE::Exception qw(raise raise_param_exc);
30 use PVE::GuestHelpers;
31 use PVE::INotify;
32 use PVE::JSONSchema qw(get_standard_option);
33 use PVE::ProcFSTools;
34 use PVE::RPCEnvironment;
35 use PVE::SafeSyslog;
36 use PVE::Storage;
37 use PVE::SysFSTools;
38 use PVE::Systemd;
39 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach $IPV6RE);
40
41 use PVE::QMPClient;
42 use PVE::QemuConfig;
43 use PVE::QemuServer::Cloudinit;
44 use PVE::QemuServer::Memory;
45 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr print_pcie_root_port);
46 use PVE::QemuServer::USB qw(parse_usb_device);
47
48 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
49 my $OVMF = {
50 x86_64 => [
51 "$EDK2_FW_BASE/OVMF_CODE.fd",
52 "$EDK2_FW_BASE/OVMF_VARS.fd"
53 ],
54 aarch64 => [
55 "$EDK2_FW_BASE/AAVMF_CODE.fd",
56 "$EDK2_FW_BASE/AAVMF_VARS.fd"
57 ],
58 };
59
60 my $qemu_snap_storage = { rbd => 1 };
61
62 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
63
64 my $QEMU_FORMAT_RE = qr/raw|cow|qcow|qcow2|qed|vmdk|cloop/;
65
66 # Note about locking: we use flock on the config file protect
67 # against concurent actions.
68 # Aditionaly, we have a 'lock' setting in the config file. This
69 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
70 # allowed when such lock is set. But you can ignore this kind of
71 # lock with the --skiplock flag.
72
73 cfs_register_file('/qemu-server/',
74 \&parse_vm_config,
75 \&write_vm_config);
76
77 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
78 description => "Some command save/restore state from this location.",
79 type => 'string',
80 maxLength => 128,
81 optional => 1,
82 });
83
84 PVE::JSONSchema::register_standard_option('pve-qm-image-format', {
85 type => 'string',
86 enum => [qw(raw cow qcow qed qcow2 vmdk cloop)],
87 description => "The drive's backing file's data format.",
88 optional => 1,
89 });
90
91 PVE::JSONSchema::register_standard_option('pve-qemu-machine', {
92 description => "Specifies the Qemu machine type.",
93 type => 'string',
94 pattern => '(pc|pc(-i440fx)?-\d+\.\d+(\.pxe)?|q35|pc-q35-\d+\.\d+(\.pxe)?|virt(?:-\d+\.\d+)?)',
95 maxLength => 40,
96 optional => 1,
97 });
98
99 #no warnings 'redefine';
100
101 sub cgroups_write {
102 my ($controller, $vmid, $option, $value) = @_;
103
104 my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
105 PVE::ProcFSTools::write_proc_entry($path, $value);
106
107 }
108
109 my $nodename = PVE::INotify::nodename();
110
111 mkdir "/etc/pve/nodes/$nodename";
112 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
113 mkdir $confdir;
114
115 my $var_run_tmpdir = "/var/run/qemu-server";
116 mkdir $var_run_tmpdir;
117
118 my $lock_dir = "/var/lock/qemu-server";
119 mkdir $lock_dir;
120
121 my $cpu_vendor_list = {
122 # Intel CPUs
123 486 => 'GenuineIntel',
124 pentium => 'GenuineIntel',
125 pentium2 => 'GenuineIntel',
126 pentium3 => 'GenuineIntel',
127 coreduo => 'GenuineIntel',
128 core2duo => 'GenuineIntel',
129 Conroe => 'GenuineIntel',
130 Penryn => 'GenuineIntel',
131 Nehalem => 'GenuineIntel',
132 'Nehalem-IBRS' => 'GenuineIntel',
133 Westmere => 'GenuineIntel',
134 'Westmere-IBRS' => 'GenuineIntel',
135 SandyBridge => 'GenuineIntel',
136 'SandyBridge-IBRS' => 'GenuineIntel',
137 IvyBridge => 'GenuineIntel',
138 'IvyBridge-IBRS' => 'GenuineIntel',
139 Haswell => 'GenuineIntel',
140 'Haswell-IBRS' => 'GenuineIntel',
141 'Haswell-noTSX' => 'GenuineIntel',
142 'Haswell-noTSX-IBRS' => 'GenuineIntel',
143 Broadwell => 'GenuineIntel',
144 'Broadwell-IBRS' => 'GenuineIntel',
145 'Broadwell-noTSX' => 'GenuineIntel',
146 'Broadwell-noTSX-IBRS' => 'GenuineIntel',
147 'Skylake-Client' => 'GenuineIntel',
148 'Skylake-Client-IBRS' => 'GenuineIntel',
149 'Skylake-Server' => 'GenuineIntel',
150 'Skylake-Server-IBRS' => 'GenuineIntel',
151 'Cascadelake-Server' => 'GenuineIntel',
152 KnightsMill => 'GenuineIntel',
153
154
155 # AMD CPUs
156 athlon => 'AuthenticAMD',
157 phenom => 'AuthenticAMD',
158 Opteron_G1 => 'AuthenticAMD',
159 Opteron_G2 => 'AuthenticAMD',
160 Opteron_G3 => 'AuthenticAMD',
161 Opteron_G4 => 'AuthenticAMD',
162 Opteron_G5 => 'AuthenticAMD',
163 EPYC => 'AuthenticAMD',
164 'EPYC-IBPB' => 'AuthenticAMD',
165
166 # generic types, use vendor from host node
167 host => 'default',
168 kvm32 => 'default',
169 kvm64 => 'default',
170 qemu32 => 'default',
171 qemu64 => 'default',
172 max => 'default',
173 };
174
175 my @supported_cpu_flags = (
176 'pcid',
177 'spec-ctrl',
178 'ibpb',
179 'ssbd',
180 'virt-ssbd',
181 'amd-ssbd',
182 'amd-no-ssb',
183 'pdpe1gb',
184 'md-clear',
185 'hv-tlbflush',
186 'hv-evmcs',
187 'aes'
188 );
189 my $cpu_flag = qr/[+-](@{[join('|', @supported_cpu_flags)]})/;
190
191 my $cpu_fmt = {
192 cputype => {
193 description => "Emulated CPU type.",
194 type => 'string',
195 enum => [ sort { "\L$a" cmp "\L$b" } keys %$cpu_vendor_list ],
196 default => 'kvm64',
197 default_key => 1,
198 },
199 hidden => {
200 description => "Do not identify as a KVM virtual machine.",
201 type => 'boolean',
202 optional => 1,
203 default => 0
204 },
205 'hv-vendor-id' => {
206 type => 'string',
207 pattern => qr/[a-zA-Z0-9]{1,12}/,
208 format_description => 'vendor-id',
209 description => 'The Hyper-V vendor ID. Some drivers or programs inside Windows guests need a specific ID.',
210 optional => 1,
211 },
212 flags => {
213 description => "List of additional CPU flags separated by ';'."
214 . " Use '+FLAG' to enable, '-FLAG' to disable a flag."
215 . " Currently supported flags: @{[join(', ', @supported_cpu_flags)]}.",
216 format_description => '+FLAG[;-FLAG...]',
217 type => 'string',
218 pattern => qr/$cpu_flag(;$cpu_flag)*/,
219 optional => 1,
220 },
221 };
222
223 my $watchdog_fmt = {
224 model => {
225 default_key => 1,
226 type => 'string',
227 enum => [qw(i6300esb ib700)],
228 description => "Watchdog type to emulate.",
229 default => 'i6300esb',
230 optional => 1,
231 },
232 action => {
233 type => 'string',
234 enum => [qw(reset shutdown poweroff pause debug none)],
235 description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
236 optional => 1,
237 },
238 };
239 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
240
241 my $agent_fmt = {
242 enabled => {
243 description => "Enable/disable Qemu GuestAgent.",
244 type => 'boolean',
245 default => 0,
246 default_key => 1,
247 },
248 fstrim_cloned_disks => {
249 description => "Run fstrim after cloning/moving a disk.",
250 type => 'boolean',
251 optional => 1,
252 default => 0
253 },
254 };
255
256 my $vga_fmt = {
257 type => {
258 description => "Select the VGA type.",
259 type => 'string',
260 default => 'std',
261 optional => 1,
262 default_key => 1,
263 enum => [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio vmware)],
264 },
265 memory => {
266 description => "Sets the VGA memory (in MiB). Has no effect with serial display.",
267 type => 'integer',
268 optional => 1,
269 minimum => 4,
270 maximum => 512,
271 },
272 };
273
274 my $ivshmem_fmt = {
275 size => {
276 type => 'integer',
277 minimum => 1,
278 description => "The size of the file in MB.",
279 },
280 name => {
281 type => 'string',
282 pattern => '[a-zA-Z0-9\-]+',
283 optional => 1,
284 format_description => 'string',
285 description => "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
286 },
287 };
288
289 my $audio_fmt = {
290 device => {
291 type => 'string',
292 enum => [qw(ich9-intel-hda intel-hda AC97)],
293 description => "Configure an audio device."
294 },
295 driver => {
296 type => 'string',
297 enum => ['spice'],
298 default => 'spice',
299 optional => 1,
300 description => "Driver backend for the audio device."
301 },
302 };
303
304 my $spice_enhancements_fmt = {
305 foldersharing => {
306 type => 'boolean',
307 optional => 1,
308 default => '0',
309 description => "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
310 },
311 videostreaming => {
312 type => 'string',
313 enum => ['off', 'all', 'filter'],
314 default => 'off',
315 optional => 1,
316 description => "Enable video streaming. Uses compression for detected video streams."
317 },
318 };
319
320 my $confdesc = {
321 onboot => {
322 optional => 1,
323 type => 'boolean',
324 description => "Specifies whether a VM will be started during system bootup.",
325 default => 0,
326 },
327 autostart => {
328 optional => 1,
329 type => 'boolean',
330 description => "Automatic restart after crash (currently ignored).",
331 default => 0,
332 },
333 hotplug => {
334 optional => 1,
335 type => 'string', format => 'pve-hotplug-features',
336 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'.",
337 default => 'network,disk,usb',
338 },
339 reboot => {
340 optional => 1,
341 type => 'boolean',
342 description => "Allow reboot. If set to '0' the VM exit on reboot.",
343 default => 1,
344 },
345 lock => {
346 optional => 1,
347 type => 'string',
348 description => "Lock/unlock the VM.",
349 enum => [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
350 },
351 cpulimit => {
352 optional => 1,
353 type => 'number',
354 description => "Limit of CPU usage.",
355 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.",
356 minimum => 0,
357 maximum => 128,
358 default => 0,
359 },
360 cpuunits => {
361 optional => 1,
362 type => 'integer',
363 description => "CPU weight for a VM.",
364 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.",
365 minimum => 2,
366 maximum => 262144,
367 default => 1024,
368 },
369 memory => {
370 optional => 1,
371 type => 'integer',
372 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
373 minimum => 16,
374 default => 512,
375 },
376 balloon => {
377 optional => 1,
378 type => 'integer',
379 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
380 minimum => 0,
381 },
382 shares => {
383 optional => 1,
384 type => 'integer',
385 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.",
386 minimum => 0,
387 maximum => 50000,
388 default => 1000,
389 },
390 keyboard => {
391 optional => 1,
392 type => 'string',
393 description => "Keybord layout for vnc server. Default is read from the '/etc/pve/datacenter.cfg' configuration file.".
394 "It should not be necessary to set it.",
395 enum => PVE::Tools::kvmkeymaplist(),
396 default => undef,
397 },
398 name => {
399 optional => 1,
400 type => 'string', format => 'dns-name',
401 description => "Set a name for the VM. Only used on the configuration web interface.",
402 },
403 scsihw => {
404 optional => 1,
405 type => 'string',
406 description => "SCSI controller model",
407 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
408 default => 'lsi',
409 },
410 description => {
411 optional => 1,
412 type => 'string',
413 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
414 },
415 ostype => {
416 optional => 1,
417 type => 'string',
418 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 l24 l26 solaris)],
419 description => "Specify guest operating system.",
420 verbose_description => <<EODESC,
421 Specify guest operating system. This is used to enable special
422 optimization/features for specific operating systems:
423
424 [horizontal]
425 other;; unspecified OS
426 wxp;; Microsoft Windows XP
427 w2k;; Microsoft Windows 2000
428 w2k3;; Microsoft Windows 2003
429 w2k8;; Microsoft Windows 2008
430 wvista;; Microsoft Windows Vista
431 win7;; Microsoft Windows 7
432 win8;; Microsoft Windows 8/2012/2012r2
433 win10;; Microsoft Windows 10/2016
434 l24;; Linux 2.4 Kernel
435 l26;; Linux 2.6/3.X Kernel
436 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
437 EODESC
438 },
439 boot => {
440 optional => 1,
441 type => 'string',
442 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
443 pattern => '[acdn]{1,4}',
444 default => 'cdn',
445 },
446 bootdisk => {
447 optional => 1,
448 type => 'string', format => 'pve-qm-bootdisk',
449 description => "Enable booting from specified disk.",
450 pattern => '(ide|sata|scsi|virtio)\d+',
451 },
452 smp => {
453 optional => 1,
454 type => 'integer',
455 description => "The number of CPUs. Please use option -sockets instead.",
456 minimum => 1,
457 default => 1,
458 },
459 sockets => {
460 optional => 1,
461 type => 'integer',
462 description => "The number of CPU sockets.",
463 minimum => 1,
464 default => 1,
465 },
466 cores => {
467 optional => 1,
468 type => 'integer',
469 description => "The number of cores per socket.",
470 minimum => 1,
471 default => 1,
472 },
473 numa => {
474 optional => 1,
475 type => 'boolean',
476 description => "Enable/disable NUMA.",
477 default => 0,
478 },
479 hugepages => {
480 optional => 1,
481 type => 'string',
482 description => "Enable/disable hugepages memory.",
483 enum => [qw(any 2 1024)],
484 },
485 vcpus => {
486 optional => 1,
487 type => 'integer',
488 description => "Number of hotplugged vcpus.",
489 minimum => 1,
490 default => 0,
491 },
492 acpi => {
493 optional => 1,
494 type => 'boolean',
495 description => "Enable/disable ACPI.",
496 default => 1,
497 },
498 agent => {
499 optional => 1,
500 description => "Enable/disable Qemu GuestAgent and its properties.",
501 type => 'string',
502 format => $agent_fmt,
503 },
504 kvm => {
505 optional => 1,
506 type => 'boolean',
507 description => "Enable/disable KVM hardware virtualization.",
508 default => 1,
509 },
510 tdf => {
511 optional => 1,
512 type => 'boolean',
513 description => "Enable/disable time drift fix.",
514 default => 0,
515 },
516 localtime => {
517 optional => 1,
518 type => 'boolean',
519 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
520 },
521 freeze => {
522 optional => 1,
523 type => 'boolean',
524 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
525 },
526 vga => {
527 optional => 1,
528 type => 'string', format => $vga_fmt,
529 description => "Configure the VGA hardware.",
530 verbose_description => "Configure the VGA Hardware. If you want to use ".
531 "high resolution modes (>= 1280x1024x16) you may need to increase " .
532 "the vga memory option. Since QEMU 2.9 the default VGA display type " .
533 "is 'std' for all OS types besides some Windows versions (XP and " .
534 "older) which use 'cirrus'. The 'qxl' option enables the SPICE " .
535 "display server. For win* OS you can select how many independent " .
536 "displays you want, Linux guests can add displays them self.\n".
537 "You can also run without any graphic card, using a serial device as terminal.",
538 },
539 watchdog => {
540 optional => 1,
541 type => 'string', format => 'pve-qm-watchdog',
542 description => "Create a virtual hardware watchdog device.",
543 verbose_description => "Create a virtual hardware watchdog device. Once enabled" .
544 " (by a guest action), the watchdog must be periodically polled " .
545 "by an agent inside the guest or else the watchdog will reset " .
546 "the guest (or execute the respective action specified)",
547 },
548 startdate => {
549 optional => 1,
550 type => 'string',
551 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
552 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'.",
553 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
554 default => 'now',
555 },
556 startup => get_standard_option('pve-startup-order'),
557 template => {
558 optional => 1,
559 type => 'boolean',
560 description => "Enable/disable Template.",
561 default => 0,
562 },
563 args => {
564 optional => 1,
565 type => 'string',
566 description => "Arbitrary arguments passed to kvm.",
567 verbose_description => <<EODESCR,
568 Arbitrary arguments passed to kvm, for example:
569
570 args: -no-reboot -no-hpet
571
572 NOTE: this option is for experts only.
573 EODESCR
574 },
575 tablet => {
576 optional => 1,
577 type => 'boolean',
578 default => 1,
579 description => "Enable/disable the USB tablet device.",
580 verbose_description => "Enable/disable the USB tablet device. This device is " .
581 "usually needed to allow absolute mouse positioning with VNC. " .
582 "Else the mouse runs out of sync with normal VNC clients. " .
583 "If you're running lots of console-only guests on one host, " .
584 "you may consider disabling this to save some context switches. " .
585 "This is turned off by default if you use spice (-vga=qxl).",
586 },
587 migrate_speed => {
588 optional => 1,
589 type => 'integer',
590 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
591 minimum => 0,
592 default => 0,
593 },
594 migrate_downtime => {
595 optional => 1,
596 type => 'number',
597 description => "Set maximum tolerated downtime (in seconds) for migrations.",
598 minimum => 0,
599 default => 0.1,
600 },
601 cdrom => {
602 optional => 1,
603 type => 'string', format => 'pve-qm-ide',
604 typetext => '<volume>',
605 description => "This is an alias for option -ide2",
606 },
607 cpu => {
608 optional => 1,
609 description => "Emulated CPU type.",
610 type => 'string',
611 format => $cpu_fmt,
612 },
613 parent => get_standard_option('pve-snapshot-name', {
614 optional => 1,
615 description => "Parent snapshot name. This is used internally, and should not be modified.",
616 }),
617 snaptime => {
618 optional => 1,
619 description => "Timestamp for snapshots.",
620 type => 'integer',
621 minimum => 0,
622 },
623 vmstate => {
624 optional => 1,
625 type => 'string', format => 'pve-volume-id',
626 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
627 },
628 vmstatestorage => get_standard_option('pve-storage-id', {
629 description => "Default storage for VM state volumes/files.",
630 optional => 1,
631 }),
632 runningmachine => get_standard_option('pve-qemu-machine', {
633 description => "Specifies the Qemu machine type of the running vm. This is used internally for snapshots.",
634 }),
635 machine => get_standard_option('pve-qemu-machine'),
636 arch => {
637 description => "Virtual processor architecture. Defaults to the host.",
638 optional => 1,
639 type => 'string',
640 enum => [qw(x86_64 aarch64)],
641 },
642 smbios1 => {
643 description => "Specify SMBIOS type 1 fields.",
644 type => 'string', format => 'pve-qm-smbios1',
645 maxLength => 512,
646 optional => 1,
647 },
648 protection => {
649 optional => 1,
650 type => 'boolean',
651 description => "Sets the protection flag of the VM. This will disable the remove VM and remove disk operations.",
652 default => 0,
653 },
654 bios => {
655 optional => 1,
656 type => 'string',
657 enum => [ qw(seabios ovmf) ],
658 description => "Select BIOS implementation.",
659 default => 'seabios',
660 },
661 vmgenid => {
662 type => 'string',
663 pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
664 format_description => 'UUID',
665 description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0' to disable explicitly.",
666 verbose_description => "The VM generation ID (vmgenid) device exposes a".
667 " 128-bit integer value identifier to the guest OS. This allows to".
668 " notify the guest operating system when the virtual machine is".
669 " executed with a different configuration (e.g. snapshot execution".
670 " or creation from a template). The guest operating system notices".
671 " the change, and is then able to react as appropriate by marking".
672 " its copies of distributed databases as dirty, re-initializing its".
673 " random number generator, etc.\n".
674 "Note that auto-creation only works when done throug API/CLI create".
675 " or update methods, but not when manually editing the config file.",
676 default => "1 (autogenerated)",
677 optional => 1,
678 },
679 hookscript => {
680 type => 'string',
681 format => 'pve-volume-id',
682 optional => 1,
683 description => "Script that will be executed during various steps in the vms lifetime.",
684 },
685 ivshmem => {
686 type => 'string',
687 format => $ivshmem_fmt,
688 description => "Inter-VM shared memory. Useful for direct communication between VMs, or to the host.",
689 optional => 1,
690 },
691 audio0 => {
692 type => 'string',
693 format => $audio_fmt,
694 description => "Configure a audio device, useful in combination with QXL/Spice.",
695 optional => 1
696 },
697 spice_enhancements => {
698 type => 'string',
699 format => $spice_enhancements_fmt,
700 description => "Configure additional enhancements for SPICE.",
701 optional => 1
702 },
703 };
704
705 my $cicustom_fmt = {
706 meta => {
707 type => 'string',
708 optional => 1,
709 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.',
710 format => 'pve-volume-id',
711 format_description => 'volume',
712 },
713 network => {
714 type => 'string',
715 optional => 1,
716 description => 'Specify a custom file containing all network data passed to the VM via cloud-init.',
717 format => 'pve-volume-id',
718 format_description => 'volume',
719 },
720 user => {
721 type => 'string',
722 optional => 1,
723 description => 'Specify a custom file containing all user data passed to the VM via cloud-init.',
724 format => 'pve-volume-id',
725 format_description => 'volume',
726 },
727 };
728 PVE::JSONSchema::register_format('pve-qm-cicustom', $cicustom_fmt);
729
730 my $confdesc_cloudinit = {
731 citype => {
732 optional => 1,
733 type => 'string',
734 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.',
735 enum => ['configdrive2', 'nocloud'],
736 },
737 ciuser => {
738 optional => 1,
739 type => 'string',
740 description => "cloud-init: User name to change ssh keys and password for instead of the image's configured default user.",
741 },
742 cipassword => {
743 optional => 1,
744 type => 'string',
745 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.',
746 },
747 cicustom => {
748 optional => 1,
749 type => 'string',
750 description => 'cloud-init: Specify custom files to replace the automatically generated ones at start.',
751 format => 'pve-qm-cicustom',
752 },
753 searchdomain => {
754 optional => 1,
755 type => 'string',
756 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.",
757 },
758 nameserver => {
759 optional => 1,
760 type => 'string', format => 'address-list',
761 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.",
762 },
763 sshkeys => {
764 optional => 1,
765 type => 'string',
766 format => 'urlencoded',
767 description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
768 },
769 };
770
771 # what about other qemu settings ?
772 #cpu => 'string',
773 #machine => 'string',
774 #fda => 'file',
775 #fdb => 'file',
776 #mtdblock => 'file',
777 #sd => 'file',
778 #pflash => 'file',
779 #snapshot => 'bool',
780 #bootp => 'file',
781 ##tftp => 'dir',
782 ##smb => 'dir',
783 #kernel => 'file',
784 #append => 'string',
785 #initrd => 'file',
786 ##soundhw => 'string',
787
788 while (my ($k, $v) = each %$confdesc) {
789 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
790 }
791
792 my $MAX_IDE_DISKS = 4;
793 my $MAX_SCSI_DISKS = 14;
794 my $MAX_VIRTIO_DISKS = 16;
795 my $MAX_SATA_DISKS = 6;
796 my $MAX_USB_DEVICES = 5;
797 my $MAX_NETS = 32;
798 my $MAX_UNUSED_DISKS = 256;
799 my $MAX_HOSTPCI_DEVICES = 16;
800 my $MAX_SERIAL_PORTS = 4;
801 my $MAX_PARALLEL_PORTS = 3;
802 my $MAX_NUMA = 8;
803
804 my $numa_fmt = {
805 cpus => {
806 type => "string",
807 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
808 description => "CPUs accessing this NUMA node.",
809 format_description => "id[-id];...",
810 },
811 memory => {
812 type => "number",
813 description => "Amount of memory this NUMA node provides.",
814 optional => 1,
815 },
816 hostnodes => {
817 type => "string",
818 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
819 description => "Host NUMA nodes to use.",
820 format_description => "id[-id];...",
821 optional => 1,
822 },
823 policy => {
824 type => 'string',
825 enum => [qw(preferred bind interleave)],
826 description => "NUMA allocation policy.",
827 optional => 1,
828 },
829 };
830 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
831 my $numadesc = {
832 optional => 1,
833 type => 'string', format => $numa_fmt,
834 description => "NUMA topology.",
835 };
836 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
837
838 for (my $i = 0; $i < $MAX_NUMA; $i++) {
839 $confdesc->{"numa$i"} = $numadesc;
840 }
841
842 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
843 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
844 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
845 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
846
847 my $net_fmt_bridge_descr = <<__EOD__;
848 Bridge to attach the network device to. The Proxmox VE standard bridge
849 is called 'vmbr0'.
850
851 If you do not specify a bridge, we create a kvm user (NATed) network
852 device, which provides DHCP and DNS services. The following addresses
853 are used:
854
855 10.0.2.2 Gateway
856 10.0.2.3 DNS Server
857 10.0.2.4 SMB Server
858
859 The DHCP server assign addresses to the guest starting from 10.0.2.15.
860 __EOD__
861
862 my $net_fmt = {
863 macaddr => get_standard_option('mac-addr', {
864 description => "MAC address. That address must be unique withing your network. This is automatically generated if not specified.",
865 }),
866 model => {
867 type => 'string',
868 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'.",
869 enum => $nic_model_list,
870 default_key => 1,
871 },
872 (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
873 bridge => {
874 type => 'string',
875 description => $net_fmt_bridge_descr,
876 format_description => 'bridge',
877 optional => 1,
878 },
879 queues => {
880 type => 'integer',
881 minimum => 0, maximum => 16,
882 description => 'Number of packet queues to be used on the device.',
883 optional => 1,
884 },
885 rate => {
886 type => 'number',
887 minimum => 0,
888 description => "Rate limit in mbps (megabytes per second) as floating point number.",
889 optional => 1,
890 },
891 tag => {
892 type => 'integer',
893 minimum => 1, maximum => 4094,
894 description => 'VLAN tag to apply to packets on this interface.',
895 optional => 1,
896 },
897 trunks => {
898 type => 'string',
899 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
900 description => 'VLAN trunks to pass through this interface.',
901 format_description => 'vlanid[;vlanid...]',
902 optional => 1,
903 },
904 firewall => {
905 type => 'boolean',
906 description => 'Whether this interface should be protected by the firewall.',
907 optional => 1,
908 },
909 link_down => {
910 type => 'boolean',
911 description => 'Whether this interface should be disconnected (like pulling the plug).',
912 optional => 1,
913 },
914 };
915
916 my $netdesc = {
917 optional => 1,
918 type => 'string', format => $net_fmt,
919 description => "Specify network devices.",
920 };
921
922 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
923
924 my $ipconfig_fmt = {
925 ip => {
926 type => 'string',
927 format => 'pve-ipv4-config',
928 format_description => 'IPv4Format/CIDR',
929 description => 'IPv4 address in CIDR format.',
930 optional => 1,
931 default => 'dhcp',
932 },
933 gw => {
934 type => 'string',
935 format => 'ipv4',
936 format_description => 'GatewayIPv4',
937 description => 'Default gateway for IPv4 traffic.',
938 optional => 1,
939 requires => 'ip',
940 },
941 ip6 => {
942 type => 'string',
943 format => 'pve-ipv6-config',
944 format_description => 'IPv6Format/CIDR',
945 description => 'IPv6 address in CIDR format.',
946 optional => 1,
947 default => 'dhcp',
948 },
949 gw6 => {
950 type => 'string',
951 format => 'ipv6',
952 format_description => 'GatewayIPv6',
953 description => 'Default gateway for IPv6 traffic.',
954 optional => 1,
955 requires => 'ip6',
956 },
957 };
958 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
959 my $ipconfigdesc = {
960 optional => 1,
961 type => 'string', format => 'pve-qm-ipconfig',
962 description => <<'EODESCR',
963 cloud-init: Specify IP addresses and gateways for the corresponding interface.
964
965 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
966
967 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit gateway should be provided.
968 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
969
970 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using dhcp on IPv4.
971 EODESCR
972 };
973 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
974
975 for (my $i = 0; $i < $MAX_NETS; $i++) {
976 $confdesc->{"net$i"} = $netdesc;
977 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
978 }
979
980 foreach my $key (keys %$confdesc_cloudinit) {
981 $confdesc->{$key} = $confdesc_cloudinit->{$key};
982 }
983
984 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
985 sub verify_volume_id_or_qm_path {
986 my ($volid, $noerr) = @_;
987
988 if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
989 return $volid;
990 }
991
992 # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
993 $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
994 if ($@) {
995 return undef if $noerr;
996 die $@;
997 }
998 return $volid;
999 }
1000
1001 my $drivename_hash;
1002
1003 my %drivedesc_base = (
1004 volume => { alias => 'file' },
1005 file => {
1006 type => 'string',
1007 format => 'pve-volume-id-or-qm-path',
1008 default_key => 1,
1009 format_description => 'volume',
1010 description => "The drive's backing volume.",
1011 },
1012 media => {
1013 type => 'string',
1014 enum => [qw(cdrom disk)],
1015 description => "The drive's media type.",
1016 default => 'disk',
1017 optional => 1
1018 },
1019 cyls => {
1020 type => 'integer',
1021 description => "Force the drive's physical geometry to have a specific cylinder count.",
1022 optional => 1
1023 },
1024 heads => {
1025 type => 'integer',
1026 description => "Force the drive's physical geometry to have a specific head count.",
1027 optional => 1
1028 },
1029 secs => {
1030 type => 'integer',
1031 description => "Force the drive's physical geometry to have a specific sector count.",
1032 optional => 1
1033 },
1034 trans => {
1035 type => 'string',
1036 enum => [qw(none lba auto)],
1037 description => "Force disk geometry bios translation mode.",
1038 optional => 1,
1039 },
1040 snapshot => {
1041 type => 'boolean',
1042 description => "Controls qemu's snapshot mode feature."
1043 . " If activated, changes made to the disk are temporary and will"
1044 . " be discarded when the VM is shutdown.",
1045 optional => 1,
1046 },
1047 cache => {
1048 type => 'string',
1049 enum => [qw(none writethrough writeback unsafe directsync)],
1050 description => "The drive's cache mode",
1051 optional => 1,
1052 },
1053 format => get_standard_option('pve-qm-image-format'),
1054 size => {
1055 type => 'string',
1056 format => 'disk-size',
1057 format_description => 'DiskSize',
1058 description => "Disk size. This is purely informational and has no effect.",
1059 optional => 1,
1060 },
1061 backup => {
1062 type => 'boolean',
1063 description => "Whether the drive should be included when making backups.",
1064 optional => 1,
1065 },
1066 replicate => {
1067 type => 'boolean',
1068 description => 'Whether the drive should considered for replication jobs.',
1069 optional => 1,
1070 default => 1,
1071 },
1072 rerror => {
1073 type => 'string',
1074 enum => [qw(ignore report stop)],
1075 description => 'Read error action.',
1076 optional => 1,
1077 },
1078 werror => {
1079 type => 'string',
1080 enum => [qw(enospc ignore report stop)],
1081 description => 'Write error action.',
1082 optional => 1,
1083 },
1084 aio => {
1085 type => 'string',
1086 enum => [qw(native threads)],
1087 description => 'AIO type to use.',
1088 optional => 1,
1089 },
1090 discard => {
1091 type => 'string',
1092 enum => [qw(ignore on)],
1093 description => 'Controls whether to pass discard/trim requests to the underlying storage.',
1094 optional => 1,
1095 },
1096 detect_zeroes => {
1097 type => 'boolean',
1098 description => 'Controls whether to detect and try to optimize writes of zeroes.',
1099 optional => 1,
1100 },
1101 serial => {
1102 type => 'string',
1103 format => 'urlencoded',
1104 format_description => 'serial',
1105 maxLength => 20*3, # *3 since it's %xx url enoded
1106 description => "The drive's reported serial number, url-encoded, up to 20 bytes long.",
1107 optional => 1,
1108 },
1109 shared => {
1110 type => 'boolean',
1111 description => 'Mark this locally-managed volume as available on all nodes',
1112 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!",
1113 optional => 1,
1114 default => 0,
1115 }
1116 );
1117
1118 my %iothread_fmt = ( iothread => {
1119 type => 'boolean',
1120 description => "Whether to use iothreads for this drive",
1121 optional => 1,
1122 });
1123
1124 my %model_fmt = (
1125 model => {
1126 type => 'string',
1127 format => 'urlencoded',
1128 format_description => 'model',
1129 maxLength => 40*3, # *3 since it's %xx url enoded
1130 description => "The drive's reported model name, url-encoded, up to 40 bytes long.",
1131 optional => 1,
1132 },
1133 );
1134
1135 my %queues_fmt = (
1136 queues => {
1137 type => 'integer',
1138 description => "Number of queues.",
1139 minimum => 2,
1140 optional => 1
1141 }
1142 );
1143
1144 my %scsiblock_fmt = (
1145 scsiblock => {
1146 type => 'boolean',
1147 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",
1148 optional => 1,
1149 default => 0,
1150 },
1151 );
1152
1153 my %ssd_fmt = (
1154 ssd => {
1155 type => 'boolean',
1156 description => "Whether to expose this drive as an SSD, rather than a rotational hard disk.",
1157 optional => 1,
1158 },
1159 );
1160
1161 my %wwn_fmt = (
1162 wwn => {
1163 type => 'string',
1164 pattern => qr/^(0x)[0-9a-fA-F]{16}/,
1165 format_description => 'wwn',
1166 description => "The drive's worldwide name, encoded as 16 bytes hex string, prefixed by '0x'.",
1167 optional => 1,
1168 },
1169 );
1170
1171 my $add_throttle_desc = sub {
1172 my ($key, $type, $what, $unit, $longunit, $minimum) = @_;
1173 my $d = {
1174 type => $type,
1175 format_description => $unit,
1176 description => "Maximum $what in $longunit.",
1177 optional => 1,
1178 };
1179 $d->{minimum} = $minimum if defined($minimum);
1180 $drivedesc_base{$key} = $d;
1181 };
1182 # throughput: (leaky bucket)
1183 $add_throttle_desc->('bps', 'integer', 'r/w speed', 'bps', 'bytes per second');
1184 $add_throttle_desc->('bps_rd', 'integer', 'read speed', 'bps', 'bytes per second');
1185 $add_throttle_desc->('bps_wr', 'integer', 'write speed', 'bps', 'bytes per second');
1186 $add_throttle_desc->('mbps', 'number', 'r/w speed', 'mbps', 'megabytes per second');
1187 $add_throttle_desc->('mbps_rd', 'number', 'read speed', 'mbps', 'megabytes per second');
1188 $add_throttle_desc->('mbps_wr', 'number', 'write speed', 'mbps', 'megabytes per second');
1189 $add_throttle_desc->('iops', 'integer', 'r/w I/O', 'iops', 'operations per second');
1190 $add_throttle_desc->('iops_rd', 'integer', 'read I/O', 'iops', 'operations per second');
1191 $add_throttle_desc->('iops_wr', 'integer', 'write I/O', 'iops', 'operations per second');
1192
1193 # pools: (pool of IO before throttling starts taking effect)
1194 $add_throttle_desc->('mbps_max', 'number', 'unthrottled r/w pool', 'mbps', 'megabytes per second');
1195 $add_throttle_desc->('mbps_rd_max', 'number', 'unthrottled read pool', 'mbps', 'megabytes per second');
1196 $add_throttle_desc->('mbps_wr_max', 'number', 'unthrottled write pool', 'mbps', 'megabytes per second');
1197 $add_throttle_desc->('iops_max', 'integer', 'unthrottled r/w I/O pool', 'iops', 'operations per second');
1198 $add_throttle_desc->('iops_rd_max', 'integer', 'unthrottled read I/O pool', 'iops', 'operations per second');
1199 $add_throttle_desc->('iops_wr_max', 'integer', 'unthrottled write I/O pool', 'iops', 'operations per second');
1200
1201 # burst lengths
1202 $add_throttle_desc->('bps_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1203 $add_throttle_desc->('bps_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1204 $add_throttle_desc->('bps_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1205 $add_throttle_desc->('iops_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1206 $add_throttle_desc->('iops_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1207 $add_throttle_desc->('iops_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1208
1209 # legacy support
1210 $drivedesc_base{'bps_rd_length'} = { alias => 'bps_rd_max_length' };
1211 $drivedesc_base{'bps_wr_length'} = { alias => 'bps_wr_max_length' };
1212 $drivedesc_base{'iops_rd_length'} = { alias => 'iops_rd_max_length' };
1213 $drivedesc_base{'iops_wr_length'} = { alias => 'iops_wr_max_length' };
1214
1215 my $ide_fmt = {
1216 %drivedesc_base,
1217 %model_fmt,
1218 %ssd_fmt,
1219 %wwn_fmt,
1220 };
1221 PVE::JSONSchema::register_format("pve-qm-ide", $ide_fmt);
1222
1223 my $idedesc = {
1224 optional => 1,
1225 type => 'string', format => $ide_fmt,
1226 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
1227 };
1228 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
1229
1230 my $scsi_fmt = {
1231 %drivedesc_base,
1232 %iothread_fmt,
1233 %queues_fmt,
1234 %scsiblock_fmt,
1235 %ssd_fmt,
1236 %wwn_fmt,
1237 };
1238 my $scsidesc = {
1239 optional => 1,
1240 type => 'string', format => $scsi_fmt,
1241 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
1242 };
1243 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
1244
1245 my $sata_fmt = {
1246 %drivedesc_base,
1247 %ssd_fmt,
1248 %wwn_fmt,
1249 };
1250 my $satadesc = {
1251 optional => 1,
1252 type => 'string', format => $sata_fmt,
1253 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
1254 };
1255 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
1256
1257 my $virtio_fmt = {
1258 %drivedesc_base,
1259 %iothread_fmt,
1260 };
1261 my $virtiodesc = {
1262 optional => 1,
1263 type => 'string', format => $virtio_fmt,
1264 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
1265 };
1266 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
1267
1268 my $alldrive_fmt = {
1269 %drivedesc_base,
1270 %iothread_fmt,
1271 %model_fmt,
1272 %queues_fmt,
1273 %scsiblock_fmt,
1274 %ssd_fmt,
1275 %wwn_fmt,
1276 };
1277
1278 my $efidisk_fmt = {
1279 volume => { alias => 'file' },
1280 file => {
1281 type => 'string',
1282 format => 'pve-volume-id-or-qm-path',
1283 default_key => 1,
1284 format_description => 'volume',
1285 description => "The drive's backing volume.",
1286 },
1287 format => get_standard_option('pve-qm-image-format'),
1288 size => {
1289 type => 'string',
1290 format => 'disk-size',
1291 format_description => 'DiskSize',
1292 description => "Disk size. This is purely informational and has no effect.",
1293 optional => 1,
1294 },
1295 };
1296
1297 my $efidisk_desc = {
1298 optional => 1,
1299 type => 'string', format => $efidisk_fmt,
1300 description => "Configure a Disk for storing EFI vars",
1301 };
1302
1303 PVE::JSONSchema::register_standard_option("pve-qm-efidisk", $efidisk_desc);
1304
1305 my $usb_fmt = {
1306 host => {
1307 default_key => 1,
1308 type => 'string', format => 'pve-qm-usb-device',
1309 format_description => 'HOSTUSBDEVICE|spice',
1310 description => <<EODESCR,
1311 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1312
1313 'bus-port(.port)*' (decimal numbers) or
1314 'vendor_id:product_id' (hexadeciaml numbers) or
1315 'spice'
1316
1317 You can use the 'lsusb -t' command to list existing usb devices.
1318
1319 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1320
1321 The value 'spice' can be used to add a usb redirection devices for spice.
1322 EODESCR
1323 },
1324 usb3 => {
1325 optional => 1,
1326 type => 'boolean',
1327 description => "Specifies whether if given host option is a USB3 device or port.",
1328 default => 0,
1329 },
1330 };
1331
1332 my $usbdesc = {
1333 optional => 1,
1334 type => 'string', format => $usb_fmt,
1335 description => "Configure an USB device (n is 0 to 4).",
1336 };
1337 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1338
1339 my $PCIRE = qr/[a-f0-9]{2}:[a-f0-9]{2}(?:\.[a-f0-9])?/;
1340 my $hostpci_fmt = {
1341 host => {
1342 default_key => 1,
1343 type => 'string',
1344 pattern => qr/$PCIRE(;$PCIRE)*/,
1345 format_description => 'HOSTPCIID[;HOSTPCIID2...]',
1346 description => <<EODESCR,
1347 Host PCI device pass through. The PCI ID of a host's PCI device or a list
1348 of PCI virtual functions of the host. HOSTPCIID syntax is:
1349
1350 'bus:dev.func' (hexadecimal numbers)
1351
1352 You can us the 'lspci' command to list existing PCI devices.
1353 EODESCR
1354 },
1355 rombar => {
1356 type => 'boolean',
1357 description => "Specify whether or not the device's ROM will be visible in the guest's memory map.",
1358 optional => 1,
1359 default => 1,
1360 },
1361 romfile => {
1362 type => 'string',
1363 pattern => '[^,;]+',
1364 format_description => 'string',
1365 description => "Custom pci device rom filename (must be located in /usr/share/kvm/).",
1366 optional => 1,
1367 },
1368 pcie => {
1369 type => 'boolean',
1370 description => "Choose the PCI-express bus (needs the 'q35' machine model).",
1371 optional => 1,
1372 default => 0,
1373 },
1374 'x-vga' => {
1375 type => 'boolean',
1376 description => "Enable vfio-vga device support.",
1377 optional => 1,
1378 default => 0,
1379 },
1380 'mdev' => {
1381 type => 'string',
1382 format_description => 'string',
1383 pattern => '[^/\.:]+',
1384 optional => 1,
1385 description => <<EODESCR
1386 The type of mediated device to use.
1387 An instance of this type will be created on startup of the VM and
1388 will be cleaned up when the VM stops.
1389 EODESCR
1390 }
1391 };
1392 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
1393
1394 my $hostpcidesc = {
1395 optional => 1,
1396 type => 'string', format => 'pve-qm-hostpci',
1397 description => "Map host PCI devices into guest.",
1398 verbose_description => <<EODESCR,
1399 Map host PCI devices into guest.
1400
1401 NOTE: This option allows direct access to host hardware. So it is no longer
1402 possible to migrate such machines - use with special care.
1403
1404 CAUTION: Experimental! User reported problems with this option.
1405 EODESCR
1406 };
1407 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
1408
1409 my $serialdesc = {
1410 optional => 1,
1411 type => 'string',
1412 pattern => '(/dev/.+|socket)',
1413 description => "Create a serial device inside the VM (n is 0 to 3)",
1414 verbose_description => <<EODESCR,
1415 Create a serial device inside the VM (n is 0 to 3), and pass through a
1416 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1417 host side (use 'qm terminal' to open a terminal connection).
1418
1419 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
1420
1421 CAUTION: Experimental! User reported problems with this option.
1422 EODESCR
1423 };
1424
1425 my $paralleldesc= {
1426 optional => 1,
1427 type => 'string',
1428 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1429 description => "Map host parallel devices (n is 0 to 2).",
1430 verbose_description => <<EODESCR,
1431 Map host parallel devices (n is 0 to 2).
1432
1433 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1434
1435 CAUTION: Experimental! User reported problems with this option.
1436 EODESCR
1437 };
1438
1439 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1440 $confdesc->{"parallel$i"} = $paralleldesc;
1441 }
1442
1443 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1444 $confdesc->{"serial$i"} = $serialdesc;
1445 }
1446
1447 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
1448 $confdesc->{"hostpci$i"} = $hostpcidesc;
1449 }
1450
1451 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
1452 $drivename_hash->{"ide$i"} = 1;
1453 $confdesc->{"ide$i"} = $idedesc;
1454 }
1455
1456 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
1457 $drivename_hash->{"sata$i"} = 1;
1458 $confdesc->{"sata$i"} = $satadesc;
1459 }
1460
1461 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
1462 $drivename_hash->{"scsi$i"} = 1;
1463 $confdesc->{"scsi$i"} = $scsidesc ;
1464 }
1465
1466 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
1467 $drivename_hash->{"virtio$i"} = 1;
1468 $confdesc->{"virtio$i"} = $virtiodesc;
1469 }
1470
1471 $drivename_hash->{efidisk0} = 1;
1472 $confdesc->{efidisk0} = $efidisk_desc;
1473
1474 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1475 $confdesc->{"usb$i"} = $usbdesc;
1476 }
1477
1478 my $unuseddesc = {
1479 optional => 1,
1480 type => 'string', format => 'pve-volume-id',
1481 description => "Reference to unused volumes. This is used internally, and should not be modified manually.",
1482 };
1483
1484 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
1485 $confdesc->{"unused$i"} = $unuseddesc;
1486 }
1487
1488 my $kvm_api_version = 0;
1489
1490 sub kvm_version {
1491 return $kvm_api_version if $kvm_api_version;
1492
1493 open my $fh, '<', '/dev/kvm'
1494 or return undef;
1495
1496 # 0xae00 => KVM_GET_API_VERSION
1497 $kvm_api_version = ioctl($fh, 0xae00, 0);
1498
1499 return $kvm_api_version;
1500 }
1501
1502 my $kvm_user_version = {};
1503 my $kvm_mtime = {};
1504
1505 sub kvm_user_version {
1506 my ($binary) = @_;
1507
1508 $binary //= get_command_for_arch(get_host_arch()); # get the native arch by default
1509 my $st = stat($binary);
1510
1511 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1512 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1513 $cachedmtime == $st->mtime;
1514
1515 $kvm_user_version->{$binary} = 'unknown';
1516 $kvm_mtime->{$binary} = $st->mtime;
1517
1518 my $code = sub {
1519 my $line = shift;
1520 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1521 $kvm_user_version->{$binary} = $2;
1522 }
1523 };
1524
1525 eval { run_command([$binary, '--version'], outfunc => $code); };
1526 warn $@ if $@;
1527
1528 return $kvm_user_version->{$binary};
1529
1530 }
1531
1532 sub kernel_has_vhost_net {
1533 return -c '/dev/vhost-net';
1534 }
1535
1536 sub valid_drive_names {
1537 # order is important - used to autoselect boot disk
1538 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
1539 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
1540 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
1541 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))),
1542 'efidisk0');
1543 }
1544
1545 sub is_valid_drivename {
1546 my $dev = shift;
1547
1548 return defined($drivename_hash->{$dev});
1549 }
1550
1551 sub option_exists {
1552 my $key = shift;
1553 return defined($confdesc->{$key});
1554 }
1555
1556 sub nic_models {
1557 return $nic_model_list;
1558 }
1559
1560 sub os_list_description {
1561
1562 return {
1563 other => 'Other',
1564 wxp => 'Windows XP',
1565 w2k => 'Windows 2000',
1566 w2k3 =>, 'Windows 2003',
1567 w2k8 => 'Windows 2008',
1568 wvista => 'Windows Vista',
1569 win7 => 'Windows 7',
1570 win8 => 'Windows 8/2012',
1571 win10 => 'Windows 10/2016',
1572 l24 => 'Linux 2.4',
1573 l26 => 'Linux 2.6',
1574 };
1575 }
1576
1577 my $cdrom_path;
1578
1579 sub get_cdrom_path {
1580
1581 return $cdrom_path if $cdrom_path;
1582
1583 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1584 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1585 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1586 }
1587
1588 sub get_iso_path {
1589 my ($storecfg, $vmid, $cdrom) = @_;
1590
1591 if ($cdrom eq 'cdrom') {
1592 return get_cdrom_path();
1593 } elsif ($cdrom eq 'none') {
1594 return '';
1595 } elsif ($cdrom =~ m|^/|) {
1596 return $cdrom;
1597 } else {
1598 return PVE::Storage::path($storecfg, $cdrom);
1599 }
1600 }
1601
1602 # try to convert old style file names to volume IDs
1603 sub filename_to_volume_id {
1604 my ($vmid, $file, $media) = @_;
1605
1606 if (!($file eq 'none' || $file eq 'cdrom' ||
1607 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1608
1609 return undef if $file =~ m|/|;
1610
1611 if ($media && $media eq 'cdrom') {
1612 $file = "local:iso/$file";
1613 } else {
1614 $file = "local:$vmid/$file";
1615 }
1616 }
1617
1618 return $file;
1619 }
1620
1621 sub verify_media_type {
1622 my ($opt, $vtype, $media) = @_;
1623
1624 return if !$media;
1625
1626 my $etype;
1627 if ($media eq 'disk') {
1628 $etype = 'images';
1629 } elsif ($media eq 'cdrom') {
1630 $etype = 'iso';
1631 } else {
1632 die "internal error";
1633 }
1634
1635 return if ($vtype eq $etype);
1636
1637 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1638 }
1639
1640 sub cleanup_drive_path {
1641 my ($opt, $storecfg, $drive) = @_;
1642
1643 # try to convert filesystem paths to volume IDs
1644
1645 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1646 ($drive->{file} !~ m|^/dev/.+|) &&
1647 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1648 ($drive->{file} !~ m/^\d+$/)) {
1649 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1650 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1651 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1652 verify_media_type($opt, $vtype, $drive->{media});
1653 $drive->{file} = $volid;
1654 }
1655
1656 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1657 }
1658
1659 sub parse_hotplug_features {
1660 my ($data) = @_;
1661
1662 my $res = {};
1663
1664 return $res if $data eq '0';
1665
1666 $data = $confdesc->{hotplug}->{default} if $data eq '1';
1667
1668 foreach my $feature (PVE::Tools::split_list($data)) {
1669 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1670 $res->{$1} = 1;
1671 } else {
1672 die "invalid hotplug feature '$feature'\n";
1673 }
1674 }
1675 return $res;
1676 }
1677
1678 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1679 sub pve_verify_hotplug_features {
1680 my ($value, $noerr) = @_;
1681
1682 return $value if parse_hotplug_features($value);
1683
1684 return undef if $noerr;
1685
1686 die "unable to parse hotplug option\n";
1687 }
1688
1689 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
1690 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
1691 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
1692 # [,aio=native|threads][,discard=ignore|on][,detect_zeroes=on|off]
1693 # [,iothread=on][,serial=serial][,model=model]
1694
1695 sub parse_drive {
1696 my ($key, $data) = @_;
1697
1698 my ($interface, $index);
1699
1700 if ($key =~ m/^([^\d]+)(\d+)$/) {
1701 $interface = $1;
1702 $index = $2;
1703 } else {
1704 return undef;
1705 }
1706
1707 my $desc = $key =~ /^unused\d+$/ ? $alldrive_fmt
1708 : $confdesc->{$key}->{format};
1709 if (!$desc) {
1710 warn "invalid drive key: $key\n";
1711 return undef;
1712 }
1713 my $res = eval { PVE::JSONSchema::parse_property_string($desc, $data) };
1714 return undef if !$res;
1715 $res->{interface} = $interface;
1716 $res->{index} = $index;
1717
1718 my $error = 0;
1719 foreach my $opt (qw(bps bps_rd bps_wr)) {
1720 if (my $bps = defined(delete $res->{$opt})) {
1721 if (defined($res->{"m$opt"})) {
1722 warn "both $opt and m$opt specified\n";
1723 ++$error;
1724 next;
1725 }
1726 $res->{"m$opt"} = sprintf("%.3f", $bps / (1024*1024.0));
1727 }
1728 }
1729
1730 # can't use the schema's 'requires' because of the mbps* => bps* "transforming aliases"
1731 for my $requirement (
1732 [mbps_max => 'mbps'],
1733 [mbps_rd_max => 'mbps_rd'],
1734 [mbps_wr_max => 'mbps_wr'],
1735 [miops_max => 'miops'],
1736 [miops_rd_max => 'miops_rd'],
1737 [miops_wr_max => 'miops_wr'],
1738 [bps_max_length => 'mbps_max'],
1739 [bps_rd_max_length => 'mbps_rd_max'],
1740 [bps_wr_max_length => 'mbps_wr_max'],
1741 [iops_max_length => 'iops_max'],
1742 [iops_rd_max_length => 'iops_rd_max'],
1743 [iops_wr_max_length => 'iops_wr_max']) {
1744 my ($option, $requires) = @$requirement;
1745 if ($res->{$option} && !$res->{$requires}) {
1746 warn "$option requires $requires\n";
1747 ++$error;
1748 }
1749 }
1750
1751 return undef if $error;
1752
1753 return undef if $res->{mbps_rd} && $res->{mbps};
1754 return undef if $res->{mbps_wr} && $res->{mbps};
1755 return undef if $res->{iops_rd} && $res->{iops};
1756 return undef if $res->{iops_wr} && $res->{iops};
1757
1758 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1759 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1760 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1761 return undef if $res->{interface} eq 'virtio';
1762 }
1763
1764 if (my $size = $res->{size}) {
1765 return undef if !defined($res->{size} = PVE::JSONSchema::parse_size($size));
1766 }
1767
1768 return $res;
1769 }
1770
1771 sub print_drive {
1772 my ($vmid, $drive) = @_;
1773 my $data = { %$drive };
1774 delete $data->{$_} for qw(index interface);
1775 return PVE::JSONSchema::print_property_string($data, $alldrive_fmt);
1776 }
1777
1778 sub scsi_inquiry {
1779 my($fh, $noerr) = @_;
1780
1781 my $SG_IO = 0x2285;
1782 my $SG_GET_VERSION_NUM = 0x2282;
1783
1784 my $versionbuf = "\x00" x 8;
1785 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1786 if (!$ret) {
1787 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1788 return undef;
1789 }
1790 my $version = unpack("I", $versionbuf);
1791 if ($version < 30000) {
1792 die "scsi generic interface too old\n" if !$noerr;
1793 return undef;
1794 }
1795
1796 my $buf = "\x00" x 36;
1797 my $sensebuf = "\x00" x 8;
1798 my $cmd = pack("C x3 C x1", 0x12, 36);
1799
1800 # see /usr/include/scsi/sg.h
1801 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";
1802
1803 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1804 length($sensebuf), 0, length($buf), $buf,
1805 $cmd, $sensebuf, 6000);
1806
1807 $ret = ioctl($fh, $SG_IO, $packet);
1808 if (!$ret) {
1809 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1810 return undef;
1811 }
1812
1813 my @res = unpack($sg_io_hdr_t, $packet);
1814 if ($res[17] || $res[18]) {
1815 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1816 return undef;
1817 }
1818
1819 my $res = {};
1820 (my $byte0, my $byte1, $res->{vendor},
1821 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1822
1823 $res->{removable} = $byte1 & 128 ? 1 : 0;
1824 $res->{type} = $byte0 & 31;
1825
1826 return $res;
1827 }
1828
1829 sub path_is_scsi {
1830 my ($path) = @_;
1831
1832 my $fh = IO::File->new("+<$path") || return undef;
1833 my $res = scsi_inquiry($fh, 1);
1834 close($fh);
1835
1836 return $res;
1837 }
1838
1839 sub machine_type_is_q35 {
1840 my ($conf) = @_;
1841
1842 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1843 }
1844
1845 sub print_tabletdevice_full {
1846 my ($conf, $arch) = @_;
1847
1848 my $q35 = machine_type_is_q35($conf);
1849
1850 # we use uhci for old VMs because tablet driver was buggy in older qemu
1851 my $usbbus;
1852 if (machine_type_is_q35($conf) || $arch eq 'aarch64') {
1853 $usbbus = 'ehci';
1854 } else {
1855 $usbbus = 'uhci';
1856 }
1857
1858 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1859 }
1860
1861 sub print_keyboarddevice_full {
1862 my ($conf, $arch, $machine) = @_;
1863
1864 return undef if $arch ne 'aarch64';
1865
1866 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1867 }
1868
1869 sub print_drivedevice_full {
1870 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1871
1872 my $device = '';
1873 my $maxdev = 0;
1874
1875 if ($drive->{interface} eq 'virtio') {
1876 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges, $arch, $machine_type);
1877 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1878 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1879 } elsif ($drive->{interface} eq 'scsi') {
1880
1881 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1882 my $unit = $drive->{index} % $maxdev;
1883 my $devicetype = 'hd';
1884 my $path = '';
1885 if (drive_is_cdrom($drive)) {
1886 $devicetype = 'cd';
1887 } else {
1888 if ($drive->{file} =~ m|^/|) {
1889 $path = $drive->{file};
1890 if (my $info = path_is_scsi($path)) {
1891 if ($info->{type} == 0 && $drive->{scsiblock}) {
1892 $devicetype = 'block';
1893 } elsif ($info->{type} == 1) { # tape
1894 $devicetype = 'generic';
1895 }
1896 }
1897 } else {
1898 $path = PVE::Storage::path($storecfg, $drive->{file});
1899 }
1900
1901 if($path =~ m/^iscsi\:\/\//){
1902 $devicetype = 'generic';
1903 }
1904 }
1905
1906 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1907 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1908 } else {
1909 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1910 }
1911
1912 if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1913 $device .= ",rotation_rate=1";
1914 }
1915 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1916
1917 } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1918 my $maxdev = ($drive->{interface} eq 'sata') ? $MAX_SATA_DISKS : 2;
1919 my $controller = int($drive->{index} / $maxdev);
1920 my $unit = $drive->{index} % $maxdev;
1921 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1922
1923 $device = "ide-$devicetype";
1924 if ($drive->{interface} eq 'ide') {
1925 $device .= ",bus=ide.$controller,unit=$unit";
1926 } else {
1927 $device .= ",bus=ahci$controller.$unit";
1928 }
1929 $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1930
1931 if ($devicetype eq 'hd') {
1932 if (my $model = $drive->{model}) {
1933 $model = URI::Escape::uri_unescape($model);
1934 $device .= ",model=$model";
1935 }
1936 if ($drive->{ssd}) {
1937 $device .= ",rotation_rate=1";
1938 }
1939 }
1940 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1941 } elsif ($drive->{interface} eq 'usb') {
1942 die "implement me";
1943 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1944 } else {
1945 die "unsupported interface type";
1946 }
1947
1948 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1949
1950 if (my $serial = $drive->{serial}) {
1951 $serial = URI::Escape::uri_unescape($serial);
1952 $device .= ",serial=$serial";
1953 }
1954
1955
1956 return $device;
1957 }
1958
1959 sub get_initiator_name {
1960 my $initiator;
1961
1962 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1963 while (defined(my $line = <$fh>)) {
1964 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1965 $initiator = $1;
1966 last;
1967 }
1968 $fh->close();
1969
1970 return $initiator;
1971 }
1972
1973 sub print_drive_full {
1974 my ($storecfg, $vmid, $drive) = @_;
1975
1976 my $path;
1977 my $volid = $drive->{file};
1978 my $format;
1979
1980 if (drive_is_cdrom($drive)) {
1981 $path = get_iso_path($storecfg, $vmid, $volid);
1982 } else {
1983 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1984 if ($storeid) {
1985 $path = PVE::Storage::path($storecfg, $volid);
1986 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1987 $format = qemu_img_format($scfg, $volname);
1988 } else {
1989 $path = $volid;
1990 $format = "raw";
1991 }
1992 }
1993
1994 my $opts = '';
1995 my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1996 foreach my $o (@qemu_drive_options) {
1997 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1998 }
1999
2000 # snapshot only accepts on|off
2001 if (defined($drive->{snapshot})) {
2002 my $v = $drive->{snapshot} ? 'on' : 'off';
2003 $opts .= ",snapshot=$v";
2004 }
2005
2006 foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
2007 my ($dir, $qmpname) = @$type;
2008 if (my $v = $drive->{"mbps$dir"}) {
2009 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
2010 }
2011 if (my $v = $drive->{"mbps${dir}_max"}) {
2012 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
2013 }
2014 if (my $v = $drive->{"bps${dir}_max_length"}) {
2015 $opts .= ",throttling.bps$qmpname-max-length=$v";
2016 }
2017 if (my $v = $drive->{"iops${dir}"}) {
2018 $opts .= ",throttling.iops$qmpname=$v";
2019 }
2020 if (my $v = $drive->{"iops${dir}_max"}) {
2021 $opts .= ",throttling.iops$qmpname-max=$v";
2022 }
2023 if (my $v = $drive->{"iops${dir}_max_length"}) {
2024 $opts .= ",throttling.iops$qmpname-max-length=$v";
2025 }
2026 }
2027
2028 $opts .= ",format=$format" if $format && !$drive->{format};
2029
2030 my $cache_direct = 0;
2031
2032 if (my $cache = $drive->{cache}) {
2033 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
2034 } elsif (!drive_is_cdrom($drive)) {
2035 $opts .= ",cache=none";
2036 $cache_direct = 1;
2037 }
2038
2039 # aio native works only with O_DIRECT
2040 if (!$drive->{aio}) {
2041 if($cache_direct) {
2042 $opts .= ",aio=native";
2043 } else {
2044 $opts .= ",aio=threads";
2045 }
2046 }
2047
2048 if (!drive_is_cdrom($drive)) {
2049 my $detectzeroes;
2050 if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
2051 $detectzeroes = 'off';
2052 } elsif ($drive->{discard}) {
2053 $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
2054 } else {
2055 # This used to be our default with discard not being specified:
2056 $detectzeroes = 'on';
2057 }
2058 $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
2059 }
2060
2061 my $pathinfo = $path ? "file=$path," : '';
2062
2063 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
2064 }
2065
2066 sub print_netdevice_full {
2067 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
2068
2069 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2070
2071 my $device = $net->{model};
2072 if ($net->{model} eq 'virtio') {
2073 $device = 'virtio-net-pci';
2074 };
2075
2076 my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
2077 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
2078 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
2079 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
2080 my $vectors = $net->{queues} * 2 + 2;
2081 $tmpstr .= ",vectors=$vectors,mq=on";
2082 }
2083 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
2084
2085 if ($use_old_bios_files) {
2086 my $romfile;
2087 if ($device eq 'virtio-net-pci') {
2088 $romfile = 'pxe-virtio.rom';
2089 } elsif ($device eq 'e1000') {
2090 $romfile = 'pxe-e1000.rom';
2091 } elsif ($device eq 'ne2k') {
2092 $romfile = 'pxe-ne2k_pci.rom';
2093 } elsif ($device eq 'pcnet') {
2094 $romfile = 'pxe-pcnet.rom';
2095 } elsif ($device eq 'rtl8139') {
2096 $romfile = 'pxe-rtl8139.rom';
2097 }
2098 $tmpstr .= ",romfile=$romfile" if $romfile;
2099 }
2100
2101 return $tmpstr;
2102 }
2103
2104 sub print_netdev_full {
2105 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
2106
2107 my $i = '';
2108 if ($netid =~ m/^net(\d+)$/) {
2109 $i = int($1);
2110 }
2111
2112 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
2113
2114 my $ifname = "tap${vmid}i$i";
2115
2116 # kvm uses TUNSETIFF ioctl, and that limits ifname length
2117 die "interface name '$ifname' is too long (max 15 character)\n"
2118 if length($ifname) >= 16;
2119
2120 my $vhostparam = '';
2121 if (is_native($arch)) {
2122 $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
2123 }
2124
2125 my $vmname = $conf->{name} || "vm$vmid";
2126
2127 my $netdev = "";
2128 my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
2129
2130 if ($net->{bridge}) {
2131 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
2132 } else {
2133 $netdev = "type=user,id=$netid,hostname=$vmname";
2134 }
2135
2136 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
2137
2138 return $netdev;
2139 }
2140
2141
2142 sub print_cpu_device {
2143 my ($conf, $id) = @_;
2144
2145 my $kvm = $conf->{kvm} // 1;
2146 my $cpu = $kvm ? "kvm64" : "qemu64";
2147 if (my $cputype = $conf->{cpu}) {
2148 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
2149 or die "Cannot parse cpu description: $cputype\n";
2150 $cpu = $cpuconf->{cputype};
2151 }
2152
2153 my $cores = $conf->{cores} || 1;
2154
2155 my $current_core = ($id - 1) % $cores;
2156 my $current_socket = int(($id - 1 - $current_core)/$cores);
2157
2158 return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
2159 }
2160
2161 my $vga_map = {
2162 'cirrus' => 'cirrus-vga',
2163 'std' => 'VGA',
2164 'vmware' => 'vmware-svga',
2165 'virtio' => 'virtio-vga',
2166 };
2167
2168 sub print_vga_device {
2169 my ($conf, $vga, $arch, $machine, $id, $qxlnum, $bridges) = @_;
2170
2171 my $type = $vga_map->{$vga->{type}};
2172 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
2173 $type = 'virtio-gpu';
2174 }
2175 my $vgamem_mb = $vga->{memory};
2176 if ($qxlnum) {
2177 $type = $id ? 'qxl' : 'qxl-vga';
2178 }
2179 die "no devicetype for $vga->{type}\n" if !$type;
2180
2181 my $memory = "";
2182 if ($vgamem_mb) {
2183 if ($vga->{type} eq 'virtio') {
2184 my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2185 $memory = ",max_hostmem=$bytes";
2186 } elsif ($qxlnum) {
2187 # from https://www.spice-space.org/multiple-monitors.html
2188 $memory = ",vgamem_mb=$vga->{memory}";
2189 my $ram = $vgamem_mb * 4;
2190 my $vram = $vgamem_mb * 2;
2191 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2192 } else {
2193 $memory = ",vgamem_mb=$vga->{memory}";
2194 }
2195 } elsif ($qxlnum && $id) {
2196 $memory = ",ram_size=67108864,vram_size=33554432";
2197 }
2198
2199 my $q35 = machine_type_is_q35($conf);
2200 my $vgaid = "vga" . ($id // '');
2201 my $pciaddr;
2202
2203 if ($q35 && $vgaid eq 'vga') {
2204 # the first display uses pcie.0 bus on q35 machines
2205 $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
2206 } else {
2207 $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
2208 }
2209
2210 return "$type,id=${vgaid}${memory}${pciaddr}";
2211 }
2212
2213 sub drive_is_cloudinit {
2214 my ($drive) = @_;
2215 return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2216 }
2217
2218 sub drive_is_cdrom {
2219 my ($drive, $exclude_cloudinit) = @_;
2220
2221 return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2222
2223 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2224
2225 }
2226
2227 sub parse_number_sets {
2228 my ($set) = @_;
2229 my $res = [];
2230 foreach my $part (split(/;/, $set)) {
2231 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2232 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2233 push @$res, [ $1, $2 ];
2234 } else {
2235 die "invalid range: $part\n";
2236 }
2237 }
2238 return $res;
2239 }
2240
2241 sub parse_numa {
2242 my ($data) = @_;
2243
2244 my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2245 $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2246 $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2247 return $res;
2248 }
2249
2250 sub parse_hostpci {
2251 my ($value) = @_;
2252
2253 return undef if !$value;
2254
2255 my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2256
2257 my @idlist = split(/;/, $res->{host});
2258 delete $res->{host};
2259 foreach my $id (@idlist) {
2260 if ($id =~ m/\./) { # full id 00:00.1
2261 push @{$res->{pciid}}, {
2262 id => $id,
2263 };
2264 } else { # partial id 00:00
2265 $res->{pciid} = PVE::SysFSTools::lspci($id);
2266 }
2267 }
2268 return $res;
2269 }
2270
2271 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2272 sub parse_net {
2273 my ($data) = @_;
2274
2275 my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2276 if ($@) {
2277 warn $@;
2278 return undef;
2279 }
2280 if (!defined($res->{macaddr})) {
2281 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2282 $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2283 }
2284 return $res;
2285 }
2286
2287 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2288 sub parse_ipconfig {
2289 my ($data) = @_;
2290
2291 my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2292 if ($@) {
2293 warn $@;
2294 return undef;
2295 }
2296
2297 if ($res->{gw} && !$res->{ip}) {
2298 warn 'gateway specified without specifying an IP address';
2299 return undef;
2300 }
2301 if ($res->{gw6} && !$res->{ip6}) {
2302 warn 'IPv6 gateway specified without specifying an IPv6 address';
2303 return undef;
2304 }
2305 if ($res->{gw} && $res->{ip} eq 'dhcp') {
2306 warn 'gateway specified together with DHCP';
2307 return undef;
2308 }
2309 if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2310 # gw6 + auto/dhcp
2311 warn "IPv6 gateway specified together with $res->{ip6} address";
2312 return undef;
2313 }
2314
2315 if (!$res->{ip} && !$res->{ip6}) {
2316 return { ip => 'dhcp', ip6 => 'dhcp' };
2317 }
2318
2319 return $res;
2320 }
2321
2322 sub print_net {
2323 my $net = shift;
2324
2325 return PVE::JSONSchema::print_property_string($net, $net_fmt);
2326 }
2327
2328 sub add_random_macs {
2329 my ($settings) = @_;
2330
2331 foreach my $opt (keys %$settings) {
2332 next if $opt !~ m/^net(\d+)$/;
2333 my $net = parse_net($settings->{$opt});
2334 next if !$net;
2335 $settings->{$opt} = print_net($net);
2336 }
2337 }
2338
2339 sub vm_is_volid_owner {
2340 my ($storecfg, $vmid, $volid) = @_;
2341
2342 if ($volid !~ m|^/|) {
2343 my ($path, $owner);
2344 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2345 if ($owner && ($owner == $vmid)) {
2346 return 1;
2347 }
2348 }
2349
2350 return undef;
2351 }
2352
2353 sub split_flagged_list {
2354 my $text = shift || '';
2355 $text =~ s/[,;]/ /g;
2356 $text =~ s/^\s+//;
2357 return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
2358 }
2359
2360 sub join_flagged_list {
2361 my ($how, $lst) = @_;
2362 join $how, map { $lst->{$_} . $_ } keys %$lst;
2363 }
2364
2365 sub vmconfig_delete_pending_option {
2366 my ($conf, $key, $force) = @_;
2367
2368 delete $conf->{pending}->{$key};
2369 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2370 $pending_delete_hash->{$key} = $force ? '!' : '';
2371 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2372 }
2373
2374 sub vmconfig_undelete_pending_option {
2375 my ($conf, $key) = @_;
2376
2377 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2378 delete $pending_delete_hash->{$key};
2379
2380 if (%$pending_delete_hash) {
2381 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2382 } else {
2383 delete $conf->{pending}->{delete};
2384 }
2385 }
2386
2387 sub vmconfig_register_unused_drive {
2388 my ($storecfg, $vmid, $conf, $drive) = @_;
2389
2390 if (drive_is_cloudinit($drive)) {
2391 eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2392 warn $@ if $@;
2393 } elsif (!drive_is_cdrom($drive)) {
2394 my $volid = $drive->{file};
2395 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2396 PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2397 }
2398 }
2399 }
2400
2401 sub vmconfig_cleanup_pending {
2402 my ($conf) = @_;
2403
2404 # remove pending changes when nothing changed
2405 my $changes;
2406 foreach my $opt (keys %{$conf->{pending}}) {
2407 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
2408 $changes = 1;
2409 delete $conf->{pending}->{$opt};
2410 }
2411 }
2412
2413 my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
2414 my $pending_delete_hash = {};
2415 while (my ($opt, $force) = each %$current_delete_hash) {
2416 if (defined($conf->{$opt})) {
2417 $pending_delete_hash->{$opt} = $force;
2418 } else {
2419 $changes = 1;
2420 }
2421 }
2422
2423 if (%$pending_delete_hash) {
2424 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2425 } else {
2426 delete $conf->{pending}->{delete};
2427 }
2428
2429 return $changes;
2430 }
2431
2432 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2433 my $smbios1_fmt = {
2434 uuid => {
2435 type => 'string',
2436 pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2437 format_description => 'UUID',
2438 description => "Set SMBIOS1 UUID.",
2439 optional => 1,
2440 },
2441 version => {
2442 type => 'string',
2443 pattern => '[A-Za-z0-9+\/]+={0,2}',
2444 format_description => 'Base64 encoded string',
2445 description => "Set SMBIOS1 version.",
2446 optional => 1,
2447 },
2448 serial => {
2449 type => 'string',
2450 pattern => '[A-Za-z0-9+\/]+={0,2}',
2451 format_description => 'Base64 encoded string',
2452 description => "Set SMBIOS1 serial number.",
2453 optional => 1,
2454 },
2455 manufacturer => {
2456 type => 'string',
2457 pattern => '[A-Za-z0-9+\/]+={0,2}',
2458 format_description => 'Base64 encoded string',
2459 description => "Set SMBIOS1 manufacturer.",
2460 optional => 1,
2461 },
2462 product => {
2463 type => 'string',
2464 pattern => '[A-Za-z0-9+\/]+={0,2}',
2465 format_description => 'Base64 encoded string',
2466 description => "Set SMBIOS1 product ID.",
2467 optional => 1,
2468 },
2469 sku => {
2470 type => 'string',
2471 pattern => '[A-Za-z0-9+\/]+={0,2}',
2472 format_description => 'Base64 encoded string',
2473 description => "Set SMBIOS1 SKU string.",
2474 optional => 1,
2475 },
2476 family => {
2477 type => 'string',
2478 pattern => '[A-Za-z0-9+\/]+={0,2}',
2479 format_description => 'Base64 encoded string',
2480 description => "Set SMBIOS1 family string.",
2481 optional => 1,
2482 },
2483 base64 => {
2484 type => 'boolean',
2485 description => 'Flag to indicate that the SMBIOS values are base64 encoded',
2486 optional => 1,
2487 },
2488 };
2489
2490 sub parse_smbios1 {
2491 my ($data) = @_;
2492
2493 my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2494 warn $@ if $@;
2495 return $res;
2496 }
2497
2498 sub print_smbios1 {
2499 my ($smbios1) = @_;
2500 return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2501 }
2502
2503 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2504
2505 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2506 sub verify_bootdisk {
2507 my ($value, $noerr) = @_;
2508
2509 return $value if is_valid_drivename($value);
2510
2511 return undef if $noerr;
2512
2513 die "invalid boot disk '$value'\n";
2514 }
2515
2516 sub parse_watchdog {
2517 my ($value) = @_;
2518
2519 return undef if !$value;
2520
2521 my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2522 warn $@ if $@;
2523 return $res;
2524 }
2525
2526 sub parse_guest_agent {
2527 my ($value) = @_;
2528
2529 return {} if !defined($value->{agent});
2530
2531 my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2532 warn $@ if $@;
2533
2534 # if the agent is disabled ignore the other potentially set properties
2535 return {} if !$res->{enabled};
2536 return $res;
2537 }
2538
2539 sub parse_vga {
2540 my ($value) = @_;
2541
2542 return {} if !$value;
2543 my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2544 warn $@ if $@;
2545 return $res;
2546 }
2547
2548 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2549 sub verify_usb_device {
2550 my ($value, $noerr) = @_;
2551
2552 return $value if parse_usb_device($value);
2553
2554 return undef if $noerr;
2555
2556 die "unable to parse usb device\n";
2557 }
2558
2559 # add JSON properties for create and set function
2560 sub json_config_properties {
2561 my $prop = shift;
2562
2563 foreach my $opt (keys %$confdesc) {
2564 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2565 $prop->{$opt} = $confdesc->{$opt};
2566 }
2567
2568 return $prop;
2569 }
2570
2571 # return copy of $confdesc_cloudinit to generate documentation
2572 sub cloudinit_config_properties {
2573
2574 return dclone($confdesc_cloudinit);
2575 }
2576
2577 sub check_type {
2578 my ($key, $value) = @_;
2579
2580 die "unknown setting '$key'\n" if !$confdesc->{$key};
2581
2582 my $type = $confdesc->{$key}->{type};
2583
2584 if (!defined($value)) {
2585 die "got undefined value\n";
2586 }
2587
2588 if ($value =~ m/[\n\r]/) {
2589 die "property contains a line feed\n";
2590 }
2591
2592 if ($type eq 'boolean') {
2593 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2594 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2595 die "type check ('boolean') failed - got '$value'\n";
2596 } elsif ($type eq 'integer') {
2597 return int($1) if $value =~ m/^(\d+)$/;
2598 die "type check ('integer') failed - got '$value'\n";
2599 } elsif ($type eq 'number') {
2600 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2601 die "type check ('number') failed - got '$value'\n";
2602 } elsif ($type eq 'string') {
2603 if (my $fmt = $confdesc->{$key}->{format}) {
2604 PVE::JSONSchema::check_format($fmt, $value);
2605 return $value;
2606 }
2607 $value =~ s/^\"(.*)\"$/$1/;
2608 return $value;
2609 } else {
2610 die "internal error"
2611 }
2612 }
2613
2614 sub touch_config {
2615 my ($vmid) = @_;
2616
2617 my $conf = PVE::QemuConfig->config_file($vmid);
2618 utime undef, undef, $conf;
2619 }
2620
2621 sub destroy_vm {
2622 my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2623
2624 my $conf = PVE::QemuConfig->load_config($vmid);
2625
2626 PVE::QemuConfig->check_lock($conf) if !$skiplock;
2627
2628 if ($conf->{template}) {
2629 # check if any base image is still used by a linked clone
2630 foreach_drive($conf, sub {
2631 my ($ds, $drive) = @_;
2632
2633 return if drive_is_cdrom($drive);
2634
2635 my $volid = $drive->{file};
2636
2637 return if !$volid || $volid =~ m|^/|;
2638
2639 die "base volume '$volid' is still in use by linked cloned\n"
2640 if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2641
2642 });
2643 }
2644
2645 # only remove disks owned by this VM
2646 foreach_drive($conf, sub {
2647 my ($ds, $drive) = @_;
2648
2649 return if drive_is_cdrom($drive, 1);
2650
2651 my $volid = $drive->{file};
2652
2653 return if !$volid || $volid =~ m|^/|;
2654
2655 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2656 return if !$path || !$owner || ($owner != $vmid);
2657
2658 eval {
2659 PVE::Storage::vdisk_free($storecfg, $volid);
2660 };
2661 warn "Could not remove disk '$volid', check manually: $@" if $@;
2662
2663 });
2664
2665 if ($keep_empty_config) {
2666 PVE::QemuConfig->write_config($vmid, { memory => 128 });
2667 } else {
2668 PVE::QemuConfig->destroy_config($vmid);
2669 }
2670
2671 # also remove unused disk
2672 eval {
2673 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2674
2675 eval {
2676 PVE::Storage::foreach_volid($dl, sub {
2677 my ($volid, $sid, $volname, $d) = @_;
2678 PVE::Storage::vdisk_free($storecfg, $volid);
2679 });
2680 };
2681 warn $@ if $@;
2682
2683 };
2684 warn $@ if $@;
2685 }
2686
2687 sub parse_vm_config {
2688 my ($filename, $raw) = @_;
2689
2690 return undef if !defined($raw);
2691
2692 my $res = {
2693 digest => Digest::SHA::sha1_hex($raw),
2694 snapshots => {},
2695 pending => {},
2696 };
2697
2698 $filename =~ m|/qemu-server/(\d+)\.conf$|
2699 || die "got strange filename '$filename'";
2700
2701 my $vmid = $1;
2702
2703 my $conf = $res;
2704 my $descr;
2705 my $section = '';
2706
2707 my @lines = split(/\n/, $raw);
2708 foreach my $line (@lines) {
2709 next if $line =~ m/^\s*$/;
2710
2711 if ($line =~ m/^\[PENDING\]\s*$/i) {
2712 $section = 'pending';
2713 if (defined($descr)) {
2714 $descr =~ s/\s+$//;
2715 $conf->{description} = $descr;
2716 }
2717 $descr = undef;
2718 $conf = $res->{$section} = {};
2719 next;
2720
2721 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2722 $section = $1;
2723 if (defined($descr)) {
2724 $descr =~ s/\s+$//;
2725 $conf->{description} = $descr;
2726 }
2727 $descr = undef;
2728 $conf = $res->{snapshots}->{$section} = {};
2729 next;
2730 }
2731
2732 if ($line =~ m/^\#(.*)\s*$/) {
2733 $descr = '' if !defined($descr);
2734 $descr .= PVE::Tools::decode_text($1) . "\n";
2735 next;
2736 }
2737
2738 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2739 $descr = '' if !defined($descr);
2740 $descr .= PVE::Tools::decode_text($2);
2741 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2742 $conf->{snapstate} = $1;
2743 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2744 my $key = $1;
2745 my $value = $2;
2746 $conf->{$key} = $value;
2747 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2748 my $value = $1;
2749 if ($section eq 'pending') {
2750 $conf->{delete} = $value; # we parse this later
2751 } else {
2752 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2753 }
2754 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2755 my $key = $1;
2756 my $value = $2;
2757 eval { $value = check_type($key, $value); };
2758 if ($@) {
2759 warn "vm $vmid - unable to parse value of '$key' - $@";
2760 } else {
2761 $key = 'ide2' if $key eq 'cdrom';
2762 my $fmt = $confdesc->{$key}->{format};
2763 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2764 my $v = parse_drive($key, $value);
2765 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2766 $v->{file} = $volid;
2767 $value = print_drive($vmid, $v);
2768 } else {
2769 warn "vm $vmid - unable to parse value of '$key'\n";
2770 next;
2771 }
2772 }
2773
2774 $conf->{$key} = $value;
2775 }
2776 }
2777 }
2778
2779 if (defined($descr)) {
2780 $descr =~ s/\s+$//;
2781 $conf->{description} = $descr;
2782 }
2783 delete $res->{snapstate}; # just to be sure
2784
2785 return $res;
2786 }
2787
2788 sub write_vm_config {
2789 my ($filename, $conf) = @_;
2790
2791 delete $conf->{snapstate}; # just to be sure
2792
2793 if ($conf->{cdrom}) {
2794 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2795 $conf->{ide2} = $conf->{cdrom};
2796 delete $conf->{cdrom};
2797 }
2798
2799 # we do not use 'smp' any longer
2800 if ($conf->{sockets}) {
2801 delete $conf->{smp};
2802 } elsif ($conf->{smp}) {
2803 $conf->{sockets} = $conf->{smp};
2804 delete $conf->{cores};
2805 delete $conf->{smp};
2806 }
2807
2808 my $used_volids = {};
2809
2810 my $cleanup_config = sub {
2811 my ($cref, $pending, $snapname) = @_;
2812
2813 foreach my $key (keys %$cref) {
2814 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2815 $key eq 'snapstate' || $key eq 'pending';
2816 my $value = $cref->{$key};
2817 if ($key eq 'delete') {
2818 die "propertry 'delete' is only allowed in [PENDING]\n"
2819 if !$pending;
2820 # fixme: check syntax?
2821 next;
2822 }
2823 eval { $value = check_type($key, $value); };
2824 die "unable to parse value of '$key' - $@" if $@;
2825
2826 $cref->{$key} = $value;
2827
2828 if (!$snapname && is_valid_drivename($key)) {
2829 my $drive = parse_drive($key, $value);
2830 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2831 }
2832 }
2833 };
2834
2835 &$cleanup_config($conf);
2836
2837 &$cleanup_config($conf->{pending}, 1);
2838
2839 foreach my $snapname (keys %{$conf->{snapshots}}) {
2840 die "internal error" if $snapname eq 'pending';
2841 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2842 }
2843
2844 # remove 'unusedX' settings if we re-add a volume
2845 foreach my $key (keys %$conf) {
2846 my $value = $conf->{$key};
2847 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2848 delete $conf->{$key};
2849 }
2850 }
2851
2852 my $generate_raw_config = sub {
2853 my ($conf, $pending) = @_;
2854
2855 my $raw = '';
2856
2857 # add description as comment to top of file
2858 if (defined(my $descr = $conf->{description})) {
2859 if ($descr) {
2860 foreach my $cl (split(/\n/, $descr)) {
2861 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2862 }
2863 } else {
2864 $raw .= "#\n" if $pending;
2865 }
2866 }
2867
2868 foreach my $key (sort keys %$conf) {
2869 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2870 $raw .= "$key: $conf->{$key}\n";
2871 }
2872 return $raw;
2873 };
2874
2875 my $raw = &$generate_raw_config($conf);
2876
2877 if (scalar(keys %{$conf->{pending}})){
2878 $raw .= "\n[PENDING]\n";
2879 $raw .= &$generate_raw_config($conf->{pending}, 1);
2880 }
2881
2882 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2883 $raw .= "\n[$snapname]\n";
2884 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2885 }
2886
2887 return $raw;
2888 }
2889
2890 sub load_defaults {
2891
2892 my $res = {};
2893
2894 # we use static defaults from our JSON schema configuration
2895 foreach my $key (keys %$confdesc) {
2896 if (defined(my $default = $confdesc->{$key}->{default})) {
2897 $res->{$key} = $default;
2898 }
2899 }
2900
2901 return $res;
2902 }
2903
2904 sub config_list {
2905 my $vmlist = PVE::Cluster::get_vmlist();
2906 my $res = {};
2907 return $res if !$vmlist || !$vmlist->{ids};
2908 my $ids = $vmlist->{ids};
2909
2910 foreach my $vmid (keys %$ids) {
2911 my $d = $ids->{$vmid};
2912 next if !$d->{node} || $d->{node} ne $nodename;
2913 next if !$d->{type} || $d->{type} ne 'qemu';
2914 $res->{$vmid}->{exists} = 1;
2915 }
2916 return $res;
2917 }
2918
2919 # test if VM uses local resources (to prevent migration)
2920 sub check_local_resources {
2921 my ($conf, $noerr) = @_;
2922
2923 my @loc_res = ();
2924
2925 push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2926 push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2927
2928 push @loc_res, "ivshmem" if $conf->{ivshmem};
2929
2930 foreach my $k (keys %$conf) {
2931 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2932 # sockets are safe: they will recreated be on the target side post-migrate
2933 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2934 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2935 }
2936
2937 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2938
2939 return \@loc_res;
2940 }
2941
2942 # check if used storages are available on all nodes (use by migrate)
2943 sub check_storage_availability {
2944 my ($storecfg, $conf, $node) = @_;
2945
2946 foreach_drive($conf, sub {
2947 my ($ds, $drive) = @_;
2948
2949 my $volid = $drive->{file};
2950 return if !$volid;
2951
2952 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2953 return if !$sid;
2954
2955 # check if storage is available on both nodes
2956 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2957 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2958 });
2959 }
2960
2961 # list nodes where all VM images are available (used by has_feature API)
2962 sub shared_nodes {
2963 my ($conf, $storecfg) = @_;
2964
2965 my $nodelist = PVE::Cluster::get_nodelist();
2966 my $nodehash = { map { $_ => 1 } @$nodelist };
2967 my $nodename = PVE::INotify::nodename();
2968
2969 foreach_drive($conf, sub {
2970 my ($ds, $drive) = @_;
2971
2972 my $volid = $drive->{file};
2973 return if !$volid;
2974
2975 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2976 if ($storeid) {
2977 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2978 if ($scfg->{disable}) {
2979 $nodehash = {};
2980 } elsif (my $avail = $scfg->{nodes}) {
2981 foreach my $node (keys %$nodehash) {
2982 delete $nodehash->{$node} if !$avail->{$node};
2983 }
2984 } elsif (!$scfg->{shared}) {
2985 foreach my $node (keys %$nodehash) {
2986 delete $nodehash->{$node} if $node ne $nodename
2987 }
2988 }
2989 }
2990 });
2991
2992 return $nodehash
2993 }
2994
2995 sub check_local_storage_availability {
2996 my ($conf, $storecfg) = @_;
2997
2998 my $nodelist = PVE::Cluster::get_nodelist();
2999 my $nodehash = { map { $_ => {} } @$nodelist };
3000
3001 foreach_drive($conf, sub {
3002 my ($ds, $drive) = @_;
3003
3004 my $volid = $drive->{file};
3005 return if !$volid;
3006
3007 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
3008 if ($storeid) {
3009 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3010
3011 if ($scfg->{disable}) {
3012 foreach my $node (keys %$nodehash) {
3013 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3014 }
3015 } elsif (my $avail = $scfg->{nodes}) {
3016 foreach my $node (keys %$nodehash) {
3017 if (!$avail->{$node}) {
3018 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3019 }
3020 }
3021 }
3022 }
3023 });
3024
3025 foreach my $node (values %$nodehash) {
3026 if (my $unavail = $node->{unavailable_storages}) {
3027 $node->{unavailable_storages} = [ sort keys %$unavail ];
3028 }
3029 }
3030
3031 return $nodehash
3032 }
3033
3034 sub check_cmdline {
3035 my ($pidfile, $pid) = @_;
3036
3037 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
3038 if (defined($fh)) {
3039 my $line = <$fh>;
3040 $fh->close;
3041 return undef if !$line;
3042 my @param = split(/\0/, $line);
3043
3044 my $cmd = $param[0];
3045 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
3046
3047 for (my $i = 0; $i < scalar (@param); $i++) {
3048 my $p = $param[$i];
3049 next if !$p;
3050 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
3051 my $p = $param[$i+1];
3052 return 1 if $p && ($p eq $pidfile);
3053 return undef;
3054 }
3055 }
3056 }
3057 return undef;
3058 }
3059
3060 sub check_running {
3061 my ($vmid, $nocheck, $node) = @_;
3062
3063 my $filename = PVE::QemuConfig->config_file($vmid, $node);
3064
3065 die "unable to find configuration file for VM $vmid - no such machine\n"
3066 if !$nocheck && ! -f $filename;
3067
3068 my $pidfile = pidfile_name($vmid);
3069
3070 if (my $fd = IO::File->new("<$pidfile")) {
3071 my $st = stat($fd);
3072 my $line = <$fd>;
3073 close($fd);
3074
3075 my $mtime = $st->mtime;
3076 if ($mtime > time()) {
3077 warn "file '$filename' modified in future\n";
3078 }
3079
3080 if ($line =~ m/^(\d+)$/) {
3081 my $pid = $1;
3082 if (check_cmdline($pidfile, $pid)) {
3083 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
3084 return $pid;
3085 }
3086 }
3087 }
3088 }
3089
3090 return undef;
3091 }
3092
3093 sub vzlist {
3094
3095 my $vzlist = config_list();
3096
3097 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
3098
3099 while (defined(my $de = $fd->read)) {
3100 next if $de !~ m/^(\d+)\.pid$/;
3101 my $vmid = $1;
3102 next if !defined($vzlist->{$vmid});
3103 if (my $pid = check_running($vmid)) {
3104 $vzlist->{$vmid}->{pid} = $pid;
3105 }
3106 }
3107
3108 return $vzlist;
3109 }
3110
3111 sub disksize {
3112 my ($storecfg, $conf) = @_;
3113
3114 my $bootdisk = $conf->{bootdisk};
3115 return undef if !$bootdisk;
3116 return undef if !is_valid_drivename($bootdisk);
3117
3118 return undef if !$conf->{$bootdisk};
3119
3120 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
3121 return undef if !defined($drive);
3122
3123 return undef if drive_is_cdrom($drive);
3124
3125 my $volid = $drive->{file};
3126 return undef if !$volid;
3127
3128 return $drive->{size};
3129 }
3130
3131 our $vmstatus_return_properties = {
3132 vmid => get_standard_option('pve-vmid'),
3133 status => {
3134 description => "Qemu process status.",
3135 type => 'string',
3136 enum => ['stopped', 'running'],
3137 },
3138 maxmem => {
3139 description => "Maximum memory in bytes.",
3140 type => 'integer',
3141 optional => 1,
3142 renderer => 'bytes',
3143 },
3144 maxdisk => {
3145 description => "Root disk size in bytes.",
3146 type => 'integer',
3147 optional => 1,
3148 renderer => 'bytes',
3149 },
3150 name => {
3151 description => "VM name.",
3152 type => 'string',
3153 optional => 1,
3154 },
3155 qmpstatus => {
3156 description => "Qemu QMP agent status.",
3157 type => 'string',
3158 optional => 1,
3159 },
3160 pid => {
3161 description => "PID of running qemu process.",
3162 type => 'integer',
3163 optional => 1,
3164 },
3165 uptime => {
3166 description => "Uptime.",
3167 type => 'integer',
3168 optional => 1,
3169 renderer => 'duration',
3170 },
3171 cpus => {
3172 description => "Maximum usable CPUs.",
3173 type => 'number',
3174 optional => 1,
3175 },
3176 lock => {
3177 description => "The current config lock, if any.",
3178 type => 'string',
3179 optional => 1,
3180 }
3181 };
3182
3183 my $last_proc_pid_stat;
3184
3185 # get VM status information
3186 # This must be fast and should not block ($full == false)
3187 # We only query KVM using QMP if $full == true (this can be slow)
3188 sub vmstatus {
3189 my ($opt_vmid, $full) = @_;
3190
3191 my $res = {};
3192
3193 my $storecfg = PVE::Storage::config();
3194
3195 my $list = vzlist();
3196 my $defaults = load_defaults();
3197
3198 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
3199
3200 my $cpucount = $cpuinfo->{cpus} || 1;
3201
3202 foreach my $vmid (keys %$list) {
3203 next if $opt_vmid && ($vmid ne $opt_vmid);
3204
3205 my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
3206 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
3207
3208 my $d = { vmid => $vmid };
3209 $d->{pid} = $list->{$vmid}->{pid};
3210
3211 # fixme: better status?
3212 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3213
3214 my $size = disksize($storecfg, $conf);
3215 if (defined($size)) {
3216 $d->{disk} = 0; # no info available
3217 $d->{maxdisk} = $size;
3218 } else {
3219 $d->{disk} = 0;
3220 $d->{maxdisk} = 0;
3221 }
3222
3223 $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3224 * ($conf->{cores} || $defaults->{cores});
3225 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3226 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3227
3228 $d->{name} = $conf->{name} || "VM $vmid";
3229 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3230 : $defaults->{memory}*(1024*1024);
3231
3232 if ($conf->{balloon}) {
3233 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3234 $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3235 : $defaults->{shares};
3236 }
3237
3238 $d->{uptime} = 0;
3239 $d->{cpu} = 0;
3240 $d->{mem} = 0;
3241
3242 $d->{netout} = 0;
3243 $d->{netin} = 0;
3244
3245 $d->{diskread} = 0;
3246 $d->{diskwrite} = 0;
3247
3248 $d->{template} = PVE::QemuConfig->is_template($conf);
3249
3250 $d->{serial} = 1 if conf_has_serial($conf);
3251 $d->{lock} = $conf->{lock} if $conf->{lock};
3252
3253 $res->{$vmid} = $d;
3254 }
3255
3256 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3257 foreach my $dev (keys %$netdev) {
3258 next if $dev !~ m/^tap([1-9]\d*)i/;
3259 my $vmid = $1;
3260 my $d = $res->{$vmid};
3261 next if !$d;
3262
3263 $d->{netout} += $netdev->{$dev}->{receive};
3264 $d->{netin} += $netdev->{$dev}->{transmit};
3265
3266 if ($full) {
3267 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3268 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3269 }
3270
3271 }
3272
3273 my $ctime = gettimeofday;
3274
3275 foreach my $vmid (keys %$list) {
3276
3277 my $d = $res->{$vmid};
3278 my $pid = $d->{pid};
3279 next if !$pid;
3280
3281 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3282 next if !$pstat; # not running
3283
3284 my $used = $pstat->{utime} + $pstat->{stime};
3285
3286 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3287
3288 if ($pstat->{vsize}) {
3289 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3290 }
3291
3292 my $old = $last_proc_pid_stat->{$pid};
3293 if (!$old) {
3294 $last_proc_pid_stat->{$pid} = {
3295 time => $ctime,
3296 used => $used,
3297 cpu => 0,
3298 };
3299 next;
3300 }
3301
3302 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3303
3304 if ($dtime > 1000) {
3305 my $dutime = $used - $old->{used};
3306
3307 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3308 $last_proc_pid_stat->{$pid} = {
3309 time => $ctime,
3310 used => $used,
3311 cpu => $d->{cpu},
3312 };
3313 } else {
3314 $d->{cpu} = $old->{cpu};
3315 }
3316 }
3317
3318 return $res if !$full;
3319
3320 my $qmpclient = PVE::QMPClient->new();
3321
3322 my $ballooncb = sub {
3323 my ($vmid, $resp) = @_;
3324
3325 my $info = $resp->{'return'};
3326 return if !$info->{max_mem};
3327
3328 my $d = $res->{$vmid};
3329
3330 # use memory assigned to VM
3331 $d->{maxmem} = $info->{max_mem};
3332 $d->{balloon} = $info->{actual};
3333
3334 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3335 $d->{mem} = $info->{total_mem} - $info->{free_mem};
3336 $d->{freemem} = $info->{free_mem};
3337 }
3338
3339 $d->{ballooninfo} = $info;
3340 };
3341
3342 my $blockstatscb = sub {
3343 my ($vmid, $resp) = @_;
3344 my $data = $resp->{'return'} || [];
3345 my $totalrdbytes = 0;
3346 my $totalwrbytes = 0;
3347
3348 for my $blockstat (@$data) {
3349 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3350 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3351
3352 $blockstat->{device} =~ s/drive-//;
3353 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3354 }
3355 $res->{$vmid}->{diskread} = $totalrdbytes;
3356 $res->{$vmid}->{diskwrite} = $totalwrbytes;
3357 };
3358
3359 my $statuscb = sub {
3360 my ($vmid, $resp) = @_;
3361
3362 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3363 # this fails if ballon driver is not loaded, so this must be
3364 # the last commnand (following command are aborted if this fails).
3365 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3366
3367 my $status = 'unknown';
3368 if (!defined($status = $resp->{'return'}->{status})) {
3369 warn "unable to get VM status\n";
3370 return;
3371 }
3372
3373 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3374 };
3375
3376 foreach my $vmid (keys %$list) {
3377 next if $opt_vmid && ($vmid ne $opt_vmid);
3378 next if !$res->{$vmid}->{pid}; # not running
3379 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3380 }
3381
3382 $qmpclient->queue_execute(undef, 2);
3383
3384 foreach my $vmid (keys %$list) {
3385 next if $opt_vmid && ($vmid ne $opt_vmid);
3386 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3387 }
3388
3389 return $res;
3390 }
3391
3392 sub foreach_drive {
3393 my ($conf, $func, @param) = @_;
3394
3395 foreach my $ds (valid_drive_names()) {
3396 next if !defined($conf->{$ds});
3397
3398 my $drive = parse_drive($ds, $conf->{$ds});
3399 next if !$drive;
3400
3401 &$func($ds, $drive, @param);
3402 }
3403 }
3404
3405 sub foreach_volid {
3406 my ($conf, $func, @param) = @_;
3407
3408 my $volhash = {};
3409
3410 my $test_volid = sub {
3411 my ($volid, $is_cdrom, $replicate, $shared, $snapname, $size) = @_;
3412
3413 return if !$volid;
3414
3415 $volhash->{$volid}->{cdrom} //= 1;
3416 $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3417
3418 $volhash->{$volid}->{replicate} //= 0;
3419 $volhash->{$volid}->{replicate} = 1 if $replicate;
3420
3421 $volhash->{$volid}->{shared} //= 0;
3422 $volhash->{$volid}->{shared} = 1 if $shared;
3423
3424 $volhash->{$volid}->{referenced_in_config} //= 0;
3425 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3426
3427 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3428 if defined($snapname);
3429 $volhash->{$volid}->{size} = $size if $size;
3430 };
3431
3432 foreach_drive($conf, sub {
3433 my ($ds, $drive) = @_;
3434 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef, $drive->{size});
3435 });
3436
3437 foreach my $snapname (keys %{$conf->{snapshots}}) {
3438 my $snap = $conf->{snapshots}->{$snapname};
3439 $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3440 foreach_drive($snap, sub {
3441 my ($ds, $drive) = @_;
3442 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3443 });
3444 }
3445
3446 foreach my $volid (keys %$volhash) {
3447 &$func($volid, $volhash->{$volid}, @param);
3448 }
3449 }
3450
3451 sub conf_has_serial {
3452 my ($conf) = @_;
3453
3454 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3455 if ($conf->{"serial$i"}) {
3456 return 1;
3457 }
3458 }
3459
3460 return 0;
3461 }
3462
3463 sub conf_has_audio {
3464 my ($conf, $id) = @_;
3465
3466 $id //= 0;
3467 my $audio = $conf->{"audio$id"};
3468 return undef if !defined($audio);
3469
3470 my $audioproperties = PVE::JSONSchema::parse_property_string($audio_fmt, $audio);
3471 my $audiodriver = $audioproperties->{driver} // 'spice';
3472
3473 return {
3474 dev => $audioproperties->{device},
3475 dev_id => "audiodev$id",
3476 backend => $audiodriver,
3477 backend_id => "$audiodriver-backend${id}",
3478 };
3479 }
3480
3481 sub vga_conf_has_spice {
3482 my ($vga) = @_;
3483
3484 my $vgaconf = parse_vga($vga);
3485 my $vgatype = $vgaconf->{type};
3486 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3487
3488 return $1 || 1;
3489 }
3490
3491 my $host_arch; # FIXME: fix PVE::Tools::get_host_arch
3492 sub get_host_arch() {
3493 $host_arch = (POSIX::uname())[4] if !$host_arch;
3494 return $host_arch;
3495 }
3496
3497 sub is_native($) {
3498 my ($arch) = @_;
3499 return get_host_arch() eq $arch;
3500 }
3501
3502 my $default_machines = {
3503 x86_64 => 'pc',
3504 aarch64 => 'virt',
3505 };
3506
3507 sub get_basic_machine_info {
3508 my ($conf, $forcemachine) = @_;
3509
3510 my $arch = $conf->{arch} // get_host_arch();
3511 my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3512 return ($arch, $machine);
3513 }
3514
3515 sub get_ovmf_files($) {
3516 my ($arch) = @_;
3517
3518 my $ovmf = $OVMF->{$arch}
3519 or die "no OVMF images known for architecture '$arch'\n";
3520
3521 return @$ovmf;
3522 }
3523
3524 my $Arch2Qemu = {
3525 aarch64 => '/usr/bin/qemu-system-aarch64',
3526 x86_64 => '/usr/bin/qemu-system-x86_64',
3527 };
3528 sub get_command_for_arch($) {
3529 my ($arch) = @_;
3530 return '/usr/bin/kvm' if is_native($arch);
3531
3532 my $cmd = $Arch2Qemu->{$arch}
3533 or die "don't know how to emulate architecture '$arch'\n";
3534 return $cmd;
3535 }
3536
3537 sub get_cpu_options {
3538 my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3539
3540 my $cpuFlags = [];
3541 my $ostype = $conf->{ostype};
3542
3543 my $cpu = $kvm ? "kvm64" : "qemu64";
3544 if ($arch eq 'aarch64') {
3545 $cpu = 'cortex-a57';
3546 }
3547 my $hv_vendor_id;
3548 if (my $cputype = $conf->{cpu}) {
3549 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3550 or die "Cannot parse cpu description: $cputype\n";
3551 $cpu = $cpuconf->{cputype};
3552 $kvm_off = 1 if $cpuconf->{hidden};
3553 $hv_vendor_id = $cpuconf->{'hv-vendor-id'};
3554
3555 if (defined(my $flags = $cpuconf->{flags})) {
3556 push @$cpuFlags, split(";", $flags);
3557 }
3558 }
3559
3560 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64' && $arch eq 'x86_64';
3561
3562 push @$cpuFlags , '-x2apic'
3563 if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3564
3565 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3566
3567 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3568
3569 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3) && $arch eq 'x86_64') {
3570
3571 push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3572 push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3573 }
3574
3575 add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough, $hv_vendor_id) if $kvm;
3576
3577 push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm && $arch eq 'x86_64';
3578
3579 push @$cpuFlags, 'kvm=off' if $kvm_off;
3580
3581 if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3582 push @$cpuFlags, "vendor=${cpu_vendor}"
3583 if $cpu_vendor ne 'default';
3584 } elsif ($arch ne 'aarch64') {
3585 die "internal error"; # should not happen
3586 }
3587
3588 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3589
3590 return ('-cpu', $cpu);
3591 }
3592
3593 sub config_to_command {
3594 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3595
3596 my $cmd = [];
3597 my $globalFlags = [];
3598 my $machineFlags = [];
3599 my $rtcFlags = [];
3600 my $devices = [];
3601 my $pciaddr = '';
3602 my $bridges = {};
3603 my $vernum = 0; # unknown
3604 my $ostype = $conf->{ostype};
3605 my $winversion = windows_version($ostype);
3606 my $kvm = $conf->{kvm};
3607
3608 my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3609 my $kvm_binary = get_command_for_arch($arch);
3610 my $kvmver = kvm_user_version($kvm_binary);
3611 $kvm //= 1 if is_native($arch);
3612
3613 if ($kvm) {
3614 die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3615 if !defined kvm_version();
3616 }
3617
3618 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3619 $vernum = $1*1000000+$2*1000;
3620 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3621 $vernum = $1*1000000+$2*1000+$3;
3622 }
3623
3624 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3625
3626 my $q35 = machine_type_is_q35($conf);
3627 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3628 my $use_old_bios_files = undef;
3629 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3630
3631 my $cpuunits = defined($conf->{cpuunits}) ?
3632 $conf->{cpuunits} : $defaults->{cpuunits};
3633
3634 push @$cmd, $kvm_binary;
3635
3636 push @$cmd, '-id', $vmid;
3637
3638 my $vmname = $conf->{name} || "vm$vmid";
3639
3640 push @$cmd, '-name', $vmname;
3641
3642 my $use_virtio = 0;
3643
3644 my $qmpsocket = qmp_socket($vmid);
3645 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3646 push @$cmd, '-mon', "chardev=qmp,mode=control";
3647
3648 if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3649 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3650 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3651 }
3652
3653 push @$cmd, '-pidfile' , pidfile_name($vmid);
3654
3655 push @$cmd, '-daemonize';
3656
3657 if ($conf->{smbios1}) {
3658 my $smbios_conf = parse_smbios1($conf->{smbios1});
3659 if ($smbios_conf->{base64}) {
3660 # Do not pass base64 flag to qemu
3661 delete $smbios_conf->{base64};
3662 my $smbios_string = "";
3663 foreach my $key (keys %$smbios_conf) {
3664 my $value;
3665 if ($key eq "uuid") {
3666 $value = $smbios_conf->{uuid}
3667 } else {
3668 $value = decode_base64($smbios_conf->{$key});
3669 }
3670 # qemu accepts any binary data, only commas need escaping by double comma
3671 $value =~ s/,/,,/g;
3672 $smbios_string .= "," . $key . "=" . $value if $value;
3673 }
3674 push @$cmd, '-smbios', "type=1" . $smbios_string;
3675 } else {
3676 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3677 }
3678 }
3679
3680 if ($conf->{vmgenid}) {
3681 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3682 }
3683
3684 my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3685 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3686 die "uefi base image not found\n" if ! -f $ovmf_code;
3687
3688 my $path;
3689 my $format;
3690 if (my $efidisk = $conf->{efidisk0}) {
3691 my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3692 my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3693 $format = $d->{format};
3694 if ($storeid) {
3695 $path = PVE::Storage::path($storecfg, $d->{file});
3696 if (!defined($format)) {
3697 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3698 $format = qemu_img_format($scfg, $volname);
3699 }
3700 } else {
3701 $path = $d->{file};
3702 die "efidisk format must be specified\n"
3703 if !defined($format);
3704 }
3705 } else {
3706 warn "no efidisk configured! Using temporary efivars disk.\n";
3707 $path = "/tmp/$vmid-ovmf.fd";
3708 PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3709 $format = 'raw';
3710 }
3711
3712 push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3713 push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3714 }
3715
3716 # load q35 config
3717 if ($q35) {
3718 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3719 if (qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0)) {
3720 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3721 } else {
3722 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3723 }
3724 }
3725
3726 # add usb controllers
3727 my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3728 push @$devices, @usbcontrollers if @usbcontrollers;
3729 my $vga = parse_vga($conf->{vga});
3730
3731 my $qxlnum = vga_conf_has_spice($conf->{vga});
3732 $vga->{type} = 'qxl' if $qxlnum;
3733
3734 if (!$vga->{type}) {
3735 if ($arch eq 'aarch64') {
3736 $vga->{type} = 'virtio';
3737 } elsif (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3738 $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3739 } else {
3740 $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3741 }
3742 }
3743
3744 # enable absolute mouse coordinates (needed by vnc)
3745 my $tablet;
3746 if (defined($conf->{tablet})) {
3747 $tablet = $conf->{tablet};
3748 } else {
3749 $tablet = $defaults->{tablet};
3750 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3751 $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3752 }
3753
3754 if ($tablet) {
3755 push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3756 my $kbd = print_keyboarddevice_full($conf, $arch);
3757 push @$devices, '-device', $kbd if defined($kbd);
3758 }
3759
3760 my $kvm_off = 0;
3761 my $gpu_passthrough;
3762
3763 # host pci devices
3764 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3765 my $id = "hostpci$i";
3766 my $d = parse_hostpci($conf->{$id});
3767 next if !$d;
3768
3769 if (my $pcie = $d->{pcie}) {
3770 die "q35 machine model is not enabled" if !$q35;
3771 # win7 wants to have the pcie devices directly on the pcie bus
3772 # instead of in the root port
3773 if ($winversion == 7) {
3774 $pciaddr = print_pcie_addr("${id}bus0");
3775 } else {
3776 # add more root ports if needed, 4 are present by default
3777 # by pve-q35 cfgs, rest added here on demand.
3778 if ($i > 3) {
3779 push @$devices, '-device', print_pcie_root_port($i);
3780 }
3781 $pciaddr = print_pcie_addr($id);
3782 }
3783 } else {
3784 $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3785 }
3786
3787 my $xvga = '';
3788 if ($d->{'x-vga'}) {
3789 $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3790 $kvm_off = 1;
3791 $vga->{type} = 'none' if !defined($conf->{vga});
3792 $gpu_passthrough = 1;
3793 }
3794
3795 my $pcidevices = $d->{pciid};
3796 my $multifunction = 1 if @$pcidevices > 1;
3797
3798 my $sysfspath;
3799 if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3800 my $pci_id = $pcidevices->[0]->{id};
3801 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3802 $sysfspath = "/sys/bus/pci/devices/0000:$pci_id/$uuid";
3803 } elsif ($d->{mdev}) {
3804 warn "ignoring mediated device '$id' with multifunction device\n";
3805 }
3806
3807 my $j=0;
3808 foreach my $pcidevice (@$pcidevices) {
3809 my $devicestr = "vfio-pci";
3810
3811 if ($sysfspath) {
3812 $devicestr .= ",sysfsdev=$sysfspath";
3813 } else {
3814 $devicestr .= ",host=$pcidevice->{id}";
3815 }
3816
3817 my $mf_addr = $multifunction ? ".$j" : '';
3818 $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3819
3820 if ($j == 0) {
3821 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3822 $devicestr .= "$xvga";
3823 $devicestr .= ",multifunction=on" if $multifunction;
3824 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3825 }
3826
3827 push @$devices, '-device', $devicestr;
3828 $j++;
3829 }
3830 }
3831
3832 # usb devices
3833 my $usb_dev_features = {};
3834 $usb_dev_features->{spice_usb3} = 1 if qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0);
3835
3836 my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3837 push @$devices, @usbdevices if @usbdevices;
3838 # serial devices
3839 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3840 if (my $path = $conf->{"serial$i"}) {
3841 if ($path eq 'socket') {
3842 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3843 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3844 # On aarch64, serial0 is the UART device. Qemu only allows
3845 # connecting UART devices via the '-serial' command line, as
3846 # the device has a fixed slot on the hardware...
3847 if ($arch eq 'aarch64' && $i == 0) {
3848 push @$devices, '-serial', "chardev:serial$i";
3849 } else {
3850 push @$devices, '-device', "isa-serial,chardev=serial$i";
3851 }
3852 } else {
3853 die "no such serial device\n" if ! -c $path;
3854 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3855 push @$devices, '-device', "isa-serial,chardev=serial$i";
3856 }
3857 }
3858 }
3859
3860 # parallel devices
3861 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3862 if (my $path = $conf->{"parallel$i"}) {
3863 die "no such parallel device\n" if ! -c $path;
3864 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3865 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3866 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3867 }
3868 }
3869
3870 if (my $audio = conf_has_audio($conf)) {
3871
3872 my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3873
3874 my $id = $audio->{dev_id};
3875 if ($audio->{dev} eq 'AC97') {
3876 push @$devices, '-device', "AC97,id=${id}${audiopciaddr}";
3877 } elsif ($audio->{dev} =~ /intel\-hda$/) {
3878 push @$devices, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3879 push @$devices, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0";
3880 push @$devices, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1";
3881 } else {
3882 die "unkown audio device '$audio->{dev}', implement me!";
3883 }
3884
3885 push @$devices, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3886 }
3887
3888 my $sockets = 1;
3889 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3890 $sockets = $conf->{sockets} if $conf->{sockets};
3891
3892 my $cores = $conf->{cores} || 1;
3893
3894 my $maxcpus = $sockets * $cores;
3895
3896 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3897
3898 my $allowed_vcpus = $cpuinfo->{cpus};
3899
3900 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3901 if ($allowed_vcpus < $maxcpus);
3902
3903 if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3904
3905 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3906 for (my $i = 2; $i <= $vcpus; $i++) {
3907 my $cpustr = print_cpu_device($conf,$i);
3908 push @$cmd, '-device', $cpustr;
3909 }
3910
3911 } else {
3912
3913 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3914 }
3915 push @$cmd, '-nodefaults';
3916
3917 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3918
3919 my $bootindex_hash = {};
3920 my $i = 1;
3921 foreach my $o (split(//, $bootorder)) {
3922 $bootindex_hash->{$o} = $i*100;
3923 $i++;
3924 }
3925
3926 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3927
3928 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3929
3930 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
3931
3932 if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3933 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, undef, $qxlnum, $bridges);
3934 my $socket = vnc_socket($vmid);
3935 push @$cmd, '-vnc', "unix:$socket,password";
3936 } else {
3937 push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3938 push @$cmd, '-nographic';
3939 }
3940
3941 # time drift fix
3942 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3943
3944 my $useLocaltime = $conf->{localtime};
3945
3946 if ($winversion >= 5) { # windows
3947 $useLocaltime = 1 if !defined($conf->{localtime});
3948
3949 # use time drift fix when acpi is enabled
3950 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3951 $tdf = 1 if !defined($conf->{tdf});
3952 }
3953 }
3954
3955 if ($winversion >= 6) {
3956 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3957 push @$cmd, '-no-hpet';
3958 }
3959
3960 push @$rtcFlags, 'driftfix=slew' if $tdf;
3961
3962 if (!$kvm) {
3963 push @$machineFlags, 'accel=tcg';
3964 }
3965
3966 if ($machine_type) {
3967 push @$machineFlags, "type=${machine_type}";
3968 }
3969
3970 if (($conf->{startdate}) && ($conf->{startdate} ne 'now')) {
3971 push @$rtcFlags, "base=$conf->{startdate}";
3972 } elsif ($useLocaltime) {
3973 push @$rtcFlags, 'base=localtime';
3974 }
3975
3976 push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3977
3978 PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3979
3980 push @$cmd, '-S' if $conf->{freeze};
3981
3982 push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3983
3984 if (parse_guest_agent($conf)->{enabled}) {
3985 my $qgasocket = qmp_socket($vmid, 1);
3986 my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3987 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3988 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3989 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3990 }
3991
3992 my $spice_port;
3993
3994 if ($qxlnum) {
3995 if ($qxlnum > 1) {
3996 if ($winversion){
3997 for(my $i = 1; $i < $qxlnum; $i++){
3998 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, $i, $qxlnum, $bridges);
3999 }
4000 } else {
4001 # assume other OS works like Linux
4002 my ($ram, $vram) = ("134217728", "67108864");
4003 if ($vga->{memory}) {
4004 $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
4005 $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
4006 }
4007 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
4008 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
4009 }
4010 }
4011
4012 my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
4013
4014 my $nodename = PVE::INotify::nodename();
4015 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4016 my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
4017 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
4018
4019 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
4020 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
4021 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
4022
4023 my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
4024 $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
4025
4026 my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
4027 if ($spice_enhancement->{foldersharing}) {
4028 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
4029 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
4030 }
4031
4032 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
4033 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
4034 push @$devices, '-spice', "$spice_opts";
4035 }
4036
4037 # enable balloon by default, unless explicitly disabled
4038 if (!defined($conf->{balloon}) || $conf->{balloon}) {
4039 $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
4040 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
4041 }
4042
4043 if ($conf->{watchdog}) {
4044 my $wdopts = parse_watchdog($conf->{watchdog});
4045 $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
4046 my $watchdog = $wdopts->{model} || 'i6300esb';
4047 push @$devices, '-device', "$watchdog$pciaddr";
4048 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
4049 }
4050
4051 my $vollist = [];
4052 my $scsicontroller = {};
4053 my $ahcicontroller = {};
4054 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
4055
4056 # Add iscsi initiator name if available
4057 if (my $initiator = get_initiator_name()) {
4058 push @$devices, '-iscsi', "initiator-name=$initiator";
4059 }
4060
4061 foreach_drive($conf, sub {
4062 my ($ds, $drive) = @_;
4063
4064 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
4065 push @$vollist, $drive->{file};
4066 }
4067
4068 # ignore efidisk here, already added in bios/fw handling code above
4069 return if $drive->{interface} eq 'efidisk';
4070
4071 $use_virtio = 1 if $ds =~ m/^virtio/;
4072
4073 if (drive_is_cdrom ($drive)) {
4074 if ($bootindex_hash->{d}) {
4075 $drive->{bootindex} = $bootindex_hash->{d};
4076 $bootindex_hash->{d} += 1;
4077 }
4078 } else {
4079 if ($bootindex_hash->{c}) {
4080 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
4081 $bootindex_hash->{c} += 1;
4082 }
4083 }
4084
4085 if($drive->{interface} eq 'virtio'){
4086 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
4087 }
4088
4089 if ($drive->{interface} eq 'scsi') {
4090
4091 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
4092
4093 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
4094 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
4095
4096 my $iothread = '';
4097 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
4098 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4099 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4100 } elsif ($drive->{iothread}) {
4101 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
4102 }
4103
4104 my $queues = '';
4105 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
4106 $queues = ",num_queues=$drive->{queues}";
4107 }
4108
4109 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
4110 $scsicontroller->{$controller}=1;
4111 }
4112
4113 if ($drive->{interface} eq 'sata') {
4114 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
4115 $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
4116 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
4117 $ahcicontroller->{$controller}=1;
4118 }
4119
4120 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
4121 push @$devices, '-drive',$drive_cmd;
4122 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4123 });
4124
4125 for (my $i = 0; $i < $MAX_NETS; $i++) {
4126 next if !$conf->{"net$i"};
4127 my $d = parse_net($conf->{"net$i"});
4128 next if !$d;
4129
4130 $use_virtio = 1 if $d->{model} eq 'virtio';
4131
4132 if ($bootindex_hash->{n}) {
4133 $d->{bootindex} = $bootindex_hash->{n};
4134 $bootindex_hash->{n} += 1;
4135 }
4136
4137 my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
4138 push @$devices, '-netdev', $netdevfull;
4139
4140 my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
4141 push @$devices, '-device', $netdevicefull;
4142 }
4143
4144 if ($conf->{ivshmem}) {
4145 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
4146
4147 my $bus;
4148 if ($q35) {
4149 $bus = print_pcie_addr("ivshmem");
4150 } else {
4151 $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
4152 }
4153
4154 my $ivshmem_name = $ivshmem->{name} // $vmid;
4155 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4156
4157 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4158 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
4159 }
4160
4161 if (!$q35) {
4162 # add pci bridges
4163 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
4164 $bridges->{1} = 1;
4165 $bridges->{2} = 1;
4166 }
4167
4168 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4169
4170 for my $k (sort {$b cmp $a} keys %$bridges) {
4171 $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
4172 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
4173 }
4174 }
4175
4176 push @$cmd, @$devices;
4177 push @$cmd, '-rtc', join(',', @$rtcFlags)
4178 if scalar(@$rtcFlags);
4179 push @$cmd, '-machine', join(',', @$machineFlags)
4180 if scalar(@$machineFlags);
4181 push @$cmd, '-global', join(',', @$globalFlags)
4182 if scalar(@$globalFlags);
4183
4184 if (my $vmstate = $conf->{vmstate}) {
4185 my $statepath = PVE::Storage::path($storecfg, $vmstate);
4186 push @$vollist, $statepath;
4187 push @$cmd, '-loadstate', $statepath;
4188 }
4189
4190 # add custom args
4191 if ($conf->{args}) {
4192 my $aa = PVE::Tools::split_args($conf->{args});
4193 push @$cmd, @$aa;
4194 }
4195
4196 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
4197 }
4198
4199 sub vnc_socket {
4200 my ($vmid) = @_;
4201 return "${var_run_tmpdir}/$vmid.vnc";
4202 }
4203
4204 sub spice_port {
4205 my ($vmid) = @_;
4206
4207 my $res = vm_mon_cmd($vmid, 'query-spice');
4208
4209 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4210 }
4211
4212 sub qmp_socket {
4213 my ($vmid, $qga, $name) = @_;
4214 my $sockettype = $qga ? 'qga' : 'qmp';
4215 my $ext = $name ? '-'.$name : '';
4216 return "${var_run_tmpdir}/$vmid$ext.$sockettype";
4217 }
4218
4219 sub pidfile_name {
4220 my ($vmid) = @_;
4221 return "${var_run_tmpdir}/$vmid.pid";
4222 }
4223
4224 sub vm_devices_list {
4225 my ($vmid) = @_;
4226
4227 my $res = vm_mon_cmd($vmid, 'query-pci');
4228 my $devices_to_check = [];
4229 my $devices = {};
4230 foreach my $pcibus (@$res) {
4231 push @$devices_to_check, @{$pcibus->{devices}},
4232 }
4233
4234 while (@$devices_to_check) {
4235 my $to_check = [];
4236 for my $d (@$devices_to_check) {
4237 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4238 next if !$d->{'pci_bridge'};
4239
4240 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
4241 push @$to_check, @{$d->{'pci_bridge'}->{devices}};
4242 }
4243 $devices_to_check = $to_check;
4244 }
4245
4246 my $resblock = vm_mon_cmd($vmid, 'query-block');
4247 foreach my $block (@$resblock) {
4248 if($block->{device} =~ m/^drive-(\S+)/){
4249 $devices->{$1} = 1;
4250 }
4251 }
4252
4253 my $resmice = vm_mon_cmd($vmid, 'query-mice');
4254 foreach my $mice (@$resmice) {
4255 if ($mice->{name} eq 'QEMU HID Tablet') {
4256 $devices->{tablet} = 1;
4257 last;
4258 }
4259 }
4260
4261 # for usb devices there is no query-usb
4262 # but we can iterate over the entries in
4263 # qom-list path=/machine/peripheral
4264 my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
4265 foreach my $per (@$resperipheral) {
4266 if ($per->{name} =~ m/^usb\d+$/) {
4267 $devices->{$per->{name}} = 1;
4268 }
4269 }
4270
4271 return $devices;
4272 }
4273
4274 sub vm_deviceplug {
4275 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4276
4277 my $q35 = machine_type_is_q35($conf);
4278
4279 my $devices_list = vm_devices_list($vmid);
4280 return 1 if defined($devices_list->{$deviceid});
4281
4282 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
4283
4284 if ($deviceid eq 'tablet') {
4285
4286 qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
4287
4288 } elsif ($deviceid eq 'keyboard') {
4289
4290 qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
4291
4292 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4293
4294 die "usb hotplug currently not reliable\n";
4295 # since we can't reliably hot unplug all added usb devices
4296 # and usb passthrough disables live migration
4297 # we disable usb hotplugging for now
4298 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
4299
4300 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4301
4302 qemu_iothread_add($vmid, $deviceid, $device);
4303
4304 qemu_driveadd($storecfg, $vmid, $device);
4305 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4306
4307 qemu_deviceadd($vmid, $devicefull);
4308 eval { qemu_deviceaddverify($vmid, $deviceid); };
4309 if (my $err = $@) {
4310 eval { qemu_drivedel($vmid, $deviceid); };
4311 warn $@ if $@;
4312 die $err;
4313 }
4314
4315 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4316
4317
4318 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
4319 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4320 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
4321
4322 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4323
4324 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
4325 qemu_iothread_add($vmid, $deviceid, $device);
4326 $devicefull .= ",iothread=iothread-$deviceid";
4327 }
4328
4329 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
4330 $devicefull .= ",num_queues=$device->{queues}";
4331 }
4332
4333 qemu_deviceadd($vmid, $devicefull);
4334 qemu_deviceaddverify($vmid, $deviceid);
4335
4336 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4337
4338 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4339 qemu_driveadd($storecfg, $vmid, $device);
4340
4341 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4342 eval { qemu_deviceadd($vmid, $devicefull); };
4343 if (my $err = $@) {
4344 eval { qemu_drivedel($vmid, $deviceid); };
4345 warn $@ if $@;
4346 die $err;
4347 }
4348
4349 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4350
4351 return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
4352
4353 my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4354 my $use_old_bios_files = undef;
4355 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4356
4357 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
4358 qemu_deviceadd($vmid, $netdevicefull);
4359 eval {
4360 qemu_deviceaddverify($vmid, $deviceid);
4361 qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
4362 };
4363 if (my $err = $@) {
4364 eval { qemu_netdevdel($vmid, $deviceid); };
4365 warn $@ if $@;
4366 die $err;
4367 }
4368
4369 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4370
4371 my $bridgeid = $2;
4372 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4373 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4374
4375 qemu_deviceadd($vmid, $devicefull);
4376 qemu_deviceaddverify($vmid, $deviceid);
4377
4378 } else {
4379 die "can't hotplug device '$deviceid'\n";
4380 }
4381
4382 return 1;
4383 }
4384
4385 # fixme: this should raise exceptions on error!
4386 sub vm_deviceunplug {
4387 my ($vmid, $conf, $deviceid) = @_;
4388
4389 my $devices_list = vm_devices_list($vmid);
4390 return 1 if !defined($devices_list->{$deviceid});
4391
4392 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4393
4394 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
4395
4396 qemu_devicedel($vmid, $deviceid);
4397
4398 } elsif ($deviceid =~ m/^usb\d+$/) {
4399
4400 die "usb hotplug currently not reliable\n";
4401 # when unplugging usb devices this way,
4402 # there may be remaining usb controllers/hubs
4403 # so we disable it for now
4404 qemu_devicedel($vmid, $deviceid);
4405 qemu_devicedelverify($vmid, $deviceid);
4406
4407 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4408
4409 qemu_devicedel($vmid, $deviceid);
4410 qemu_devicedelverify($vmid, $deviceid);
4411 qemu_drivedel($vmid, $deviceid);
4412 qemu_iothread_del($conf, $vmid, $deviceid);
4413
4414 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4415
4416 qemu_devicedel($vmid, $deviceid);
4417 qemu_devicedelverify($vmid, $deviceid);
4418 qemu_iothread_del($conf, $vmid, $deviceid);
4419
4420 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4421
4422 qemu_devicedel($vmid, $deviceid);
4423 qemu_drivedel($vmid, $deviceid);
4424 qemu_deletescsihw($conf, $vmid, $deviceid);
4425
4426 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4427
4428 qemu_devicedel($vmid, $deviceid);
4429 qemu_devicedelverify($vmid, $deviceid);
4430 qemu_netdevdel($vmid, $deviceid);
4431
4432 } else {
4433 die "can't unplug device '$deviceid'\n";
4434 }
4435
4436 return 1;
4437 }
4438
4439 sub qemu_deviceadd {
4440 my ($vmid, $devicefull) = @_;
4441
4442 $devicefull = "driver=".$devicefull;
4443 my %options = split(/[=,]/, $devicefull);
4444
4445 vm_mon_cmd($vmid, "device_add" , %options);
4446 }
4447
4448 sub qemu_devicedel {
4449 my ($vmid, $deviceid) = @_;
4450
4451 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4452 }
4453
4454 sub qemu_iothread_add {
4455 my($vmid, $deviceid, $device) = @_;
4456
4457 if ($device->{iothread}) {
4458 my $iothreads = vm_iothreads_list($vmid);
4459 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4460 }
4461 }
4462
4463 sub qemu_iothread_del {
4464 my($conf, $vmid, $deviceid) = @_;
4465
4466 my $confid = $deviceid;
4467 if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
4468 $confid = 'scsi' . $1;
4469 }
4470 my $device = parse_drive($confid, $conf->{$confid});
4471 if ($device->{iothread}) {
4472 my $iothreads = vm_iothreads_list($vmid);
4473 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4474 }
4475 }
4476
4477 sub qemu_objectadd {
4478 my($vmid, $objectid, $qomtype) = @_;
4479
4480 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4481
4482 return 1;
4483 }
4484
4485 sub qemu_objectdel {
4486 my($vmid, $objectid) = @_;
4487
4488 vm_mon_cmd($vmid, "object-del", id => $objectid);
4489
4490 return 1;
4491 }
4492
4493 sub qemu_driveadd {
4494 my ($storecfg, $vmid, $device) = @_;
4495
4496 my $drive = print_drive_full($storecfg, $vmid, $device);
4497 $drive =~ s/\\/\\\\/g;
4498 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4499
4500 # If the command succeeds qemu prints: "OK"
4501 return 1 if $ret =~ m/OK/s;
4502
4503 die "adding drive failed: $ret\n";
4504 }
4505
4506 sub qemu_drivedel {
4507 my($vmid, $deviceid) = @_;
4508
4509 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4510 $ret =~ s/^\s+//;
4511
4512 return 1 if $ret eq "";
4513
4514 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4515 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4516
4517 die "deleting drive $deviceid failed : $ret\n";
4518 }
4519
4520 sub qemu_deviceaddverify {
4521 my ($vmid, $deviceid) = @_;
4522
4523 for (my $i = 0; $i <= 5; $i++) {
4524 my $devices_list = vm_devices_list($vmid);
4525 return 1 if defined($devices_list->{$deviceid});
4526 sleep 1;
4527 }
4528
4529 die "error on hotplug device '$deviceid'\n";
4530 }
4531
4532
4533 sub qemu_devicedelverify {
4534 my ($vmid, $deviceid) = @_;
4535
4536 # need to verify that the device is correctly removed as device_del
4537 # is async and empty return is not reliable
4538
4539 for (my $i = 0; $i <= 5; $i++) {
4540 my $devices_list = vm_devices_list($vmid);
4541 return 1 if !defined($devices_list->{$deviceid});
4542 sleep 1;
4543 }
4544
4545 die "error on hot-unplugging device '$deviceid'\n";
4546 }
4547
4548 sub qemu_findorcreatescsihw {
4549 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4550
4551 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4552
4553 my $scsihwid="$controller_prefix$controller";
4554 my $devices_list = vm_devices_list($vmid);
4555
4556 if(!defined($devices_list->{$scsihwid})) {
4557 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4558 }
4559
4560 return 1;
4561 }
4562
4563 sub qemu_deletescsihw {
4564 my ($conf, $vmid, $opt) = @_;
4565
4566 my $device = parse_drive($opt, $conf->{$opt});
4567
4568 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4569 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4570 return 1;
4571 }
4572
4573 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4574
4575 my $devices_list = vm_devices_list($vmid);
4576 foreach my $opt (keys %{$devices_list}) {
4577 if (PVE::QemuServer::is_valid_drivename($opt)) {
4578 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4579 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4580 return 1;
4581 }
4582 }
4583 }
4584
4585 my $scsihwid="scsihw$controller";
4586
4587 vm_deviceunplug($vmid, $conf, $scsihwid);
4588
4589 return 1;
4590 }
4591
4592 sub qemu_add_pci_bridge {
4593 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4594
4595 my $bridges = {};
4596
4597 my $bridgeid;
4598
4599 print_pci_addr($device, $bridges, $arch, $machine_type);
4600
4601 while (my ($k, $v) = each %$bridges) {
4602 $bridgeid = $k;
4603 }
4604 return 1 if !defined($bridgeid) || $bridgeid < 1;
4605
4606 my $bridge = "pci.$bridgeid";
4607 my $devices_list = vm_devices_list($vmid);
4608
4609 if (!defined($devices_list->{$bridge})) {
4610 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4611 }
4612
4613 return 1;
4614 }
4615
4616 sub qemu_set_link_status {
4617 my ($vmid, $device, $up) = @_;
4618
4619 vm_mon_cmd($vmid, "set_link", name => $device,
4620 up => $up ? JSON::true : JSON::false);
4621 }
4622
4623 sub qemu_netdevadd {
4624 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4625
4626 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4627 my %options = split(/[=,]/, $netdev);
4628
4629 vm_mon_cmd($vmid, "netdev_add", %options);
4630 return 1;
4631 }
4632
4633 sub qemu_netdevdel {
4634 my ($vmid, $deviceid) = @_;
4635
4636 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4637 }
4638
4639 sub qemu_usb_hotplug {
4640 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4641
4642 return if !$device;
4643
4644 # remove the old one first
4645 vm_deviceunplug($vmid, $conf, $deviceid);
4646
4647 # check if xhci controller is necessary and available
4648 if ($device->{usb3}) {
4649
4650 my $devicelist = vm_devices_list($vmid);
4651
4652 if (!$devicelist->{xhci}) {
4653 my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4654 qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4655 }
4656 }
4657 my $d = parse_usb_device($device->{host});
4658 $d->{usb3} = $device->{usb3};
4659
4660 # add the new one
4661 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4662 }
4663
4664 sub qemu_cpu_hotplug {
4665 my ($vmid, $conf, $vcpus) = @_;
4666
4667 my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4668
4669 my $sockets = 1;
4670 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4671 $sockets = $conf->{sockets} if $conf->{sockets};
4672 my $cores = $conf->{cores} || 1;
4673 my $maxcpus = $sockets * $cores;
4674
4675 $vcpus = $maxcpus if !$vcpus;
4676
4677 die "you can't add more vcpus than maxcpus\n"
4678 if $vcpus > $maxcpus;
4679
4680 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4681
4682 if ($vcpus < $currentvcpus) {
4683
4684 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4685
4686 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4687 qemu_devicedel($vmid, "cpu$i");
4688 my $retry = 0;
4689 my $currentrunningvcpus = undef;
4690 while (1) {
4691 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4692 last if scalar(@{$currentrunningvcpus}) == $i-1;
4693 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4694 $retry++;
4695 sleep 1;
4696 }
4697 #update conf after each succesfull cpu unplug
4698 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4699 PVE::QemuConfig->write_config($vmid, $conf);
4700 }
4701 } else {
4702 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4703 }
4704
4705 return;
4706 }
4707
4708 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4709 die "vcpus in running vm does not match its configuration\n"
4710 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4711
4712 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4713
4714 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4715 my $cpustr = print_cpu_device($conf, $i);
4716 qemu_deviceadd($vmid, $cpustr);
4717
4718 my $retry = 0;
4719 my $currentrunningvcpus = undef;
4720 while (1) {
4721 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4722 last if scalar(@{$currentrunningvcpus}) == $i;
4723 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4724 sleep 1;
4725 $retry++;
4726 }
4727 #update conf after each succesfull cpu hotplug
4728 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4729 PVE::QemuConfig->write_config($vmid, $conf);
4730 }
4731 } else {
4732
4733 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4734 vm_mon_cmd($vmid, "cpu-add", id => int($i));
4735 }
4736 }
4737 }
4738
4739 sub qemu_block_set_io_throttle {
4740 my ($vmid, $deviceid,
4741 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4742 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4743 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4744 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4745
4746 return if !check_running($vmid) ;
4747
4748 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4749 bps => int($bps),
4750 bps_rd => int($bps_rd),
4751 bps_wr => int($bps_wr),
4752 iops => int($iops),
4753 iops_rd => int($iops_rd),
4754 iops_wr => int($iops_wr),
4755 bps_max => int($bps_max),
4756 bps_rd_max => int($bps_rd_max),
4757 bps_wr_max => int($bps_wr_max),
4758 iops_max => int($iops_max),
4759 iops_rd_max => int($iops_rd_max),
4760 iops_wr_max => int($iops_wr_max),
4761 bps_max_length => int($bps_max_length),
4762 bps_rd_max_length => int($bps_rd_max_length),
4763 bps_wr_max_length => int($bps_wr_max_length),
4764 iops_max_length => int($iops_max_length),
4765 iops_rd_max_length => int($iops_rd_max_length),
4766 iops_wr_max_length => int($iops_wr_max_length),
4767 );
4768
4769 }
4770
4771 # old code, only used to shutdown old VM after update
4772 sub __read_avail {
4773 my ($fh, $timeout) = @_;
4774
4775 my $sel = new IO::Select;
4776 $sel->add($fh);
4777
4778 my $res = '';
4779 my $buf;
4780
4781 my @ready;
4782 while (scalar (@ready = $sel->can_read($timeout))) {
4783 my $count;
4784 if ($count = $fh->sysread($buf, 8192)) {
4785 if ($buf =~ /^(.*)\(qemu\) $/s) {
4786 $res .= $1;
4787 last;
4788 } else {
4789 $res .= $buf;
4790 }
4791 } else {
4792 if (!defined($count)) {
4793 die "$!\n";
4794 }
4795 last;
4796 }
4797 }
4798
4799 die "monitor read timeout\n" if !scalar(@ready);
4800
4801 return $res;
4802 }
4803
4804 sub qemu_block_resize {
4805 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4806
4807 my $running = check_running($vmid);
4808
4809 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4810
4811 return if !$running;
4812
4813 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4814
4815 }
4816
4817 sub qemu_volume_snapshot {
4818 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4819
4820 my $running = check_running($vmid);
4821
4822 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4823 vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4824 } else {
4825 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4826 }
4827 }
4828
4829 sub qemu_volume_snapshot_delete {
4830 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4831
4832 my $running = check_running($vmid);
4833
4834 if($running) {
4835
4836 $running = undef;
4837 my $conf = PVE::QemuConfig->load_config($vmid);
4838 foreach_drive($conf, sub {
4839 my ($ds, $drive) = @_;
4840 $running = 1 if $drive->{file} eq $volid;
4841 });
4842 }
4843
4844 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4845 vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4846 } else {
4847 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4848 }
4849 }
4850
4851 sub set_migration_caps {
4852 my ($vmid) = @_;
4853
4854 my $cap_ref = [];
4855
4856 my $enabled_cap = {
4857 "auto-converge" => 1,
4858 "xbzrle" => 1,
4859 "x-rdma-pin-all" => 0,
4860 "zero-blocks" => 0,
4861 "compress" => 0
4862 };
4863
4864 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4865
4866 for my $supported_capability (@$supported_capabilities) {
4867 push @$cap_ref, {
4868 capability => $supported_capability->{capability},
4869 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4870 };
4871 }
4872
4873 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4874 }
4875
4876 my $fast_plug_option = {
4877 'lock' => 1,
4878 'name' => 1,
4879 'onboot' => 1,
4880 'shares' => 1,
4881 'startup' => 1,
4882 'description' => 1,
4883 'protection' => 1,
4884 'vmstatestorage' => 1,
4885 'hookscript' => 1,
4886 };
4887
4888 # hotplug changes in [PENDING]
4889 # $selection hash can be used to only apply specified options, for
4890 # example: { cores => 1 } (only apply changed 'cores')
4891 # $errors ref is used to return error messages
4892 sub vmconfig_hotplug_pending {
4893 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4894
4895 my $defaults = load_defaults();
4896 my ($arch, $machine_type) = get_basic_machine_info($conf, undef);
4897
4898 # commit values which do not have any impact on running VM first
4899 # Note: those option cannot raise errors, we we do not care about
4900 # $selection and always apply them.
4901
4902 my $add_error = sub {
4903 my ($opt, $msg) = @_;
4904 $errors->{$opt} = "hotplug problem - $msg";
4905 };
4906
4907 my $changes = 0;
4908 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4909 if ($fast_plug_option->{$opt}) {
4910 $conf->{$opt} = $conf->{pending}->{$opt};
4911 delete $conf->{pending}->{$opt};
4912 $changes = 1;
4913 }
4914 }
4915
4916 if ($changes) {
4917 PVE::QemuConfig->write_config($vmid, $conf);
4918 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4919 }
4920
4921 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4922
4923 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4924 while (my ($opt, $force) = each %$pending_delete_hash) {
4925 next if $selection && !$selection->{$opt};
4926 eval {
4927 if ($opt eq 'hotplug') {
4928 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4929 } elsif ($opt eq 'tablet') {
4930 die "skip\n" if !$hotplug_features->{usb};
4931 if ($defaults->{tablet}) {
4932 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4933 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4934 if $arch eq 'aarch64';
4935 } else {
4936 vm_deviceunplug($vmid, $conf, 'tablet');
4937 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4938 }
4939 } elsif ($opt =~ m/^usb\d+/) {
4940 die "skip\n";
4941 # since we cannot reliably hot unplug usb devices
4942 # we are disabling it
4943 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4944 vm_deviceunplug($vmid, $conf, $opt);
4945 } elsif ($opt eq 'vcpus') {
4946 die "skip\n" if !$hotplug_features->{cpu};
4947 qemu_cpu_hotplug($vmid, $conf, undef);
4948 } elsif ($opt eq 'balloon') {
4949 # enable balloon device is not hotpluggable
4950 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4951 # here we reset the ballooning value to memory
4952 my $balloon = $conf->{memory} || $defaults->{memory};
4953 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4954 } elsif ($fast_plug_option->{$opt}) {
4955 # do nothing
4956 } elsif ($opt =~ m/^net(\d+)$/) {
4957 die "skip\n" if !$hotplug_features->{network};
4958 vm_deviceunplug($vmid, $conf, $opt);
4959 } elsif (is_valid_drivename($opt)) {
4960 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4961 vm_deviceunplug($vmid, $conf, $opt);
4962 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4963 } elsif ($opt =~ m/^memory$/) {
4964 die "skip\n" if !$hotplug_features->{memory};
4965 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4966 } elsif ($opt eq 'cpuunits') {
4967 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4968 } elsif ($opt eq 'cpulimit') {
4969 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4970 } else {
4971 die "skip\n";
4972 }
4973 };
4974 if (my $err = $@) {
4975 &$add_error($opt, $err) if $err ne "skip\n";
4976 } else {
4977 # save new config if hotplug was successful
4978 delete $conf->{$opt};
4979 vmconfig_undelete_pending_option($conf, $opt);
4980 PVE::QemuConfig->write_config($vmid, $conf);
4981 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4982 }
4983 }
4984
4985 my $apply_pending_cloudinit;
4986 $apply_pending_cloudinit = sub {
4987 my ($key, $value) = @_;
4988 $apply_pending_cloudinit = sub {}; # once is enough
4989
4990 my @cloudinit_opts = keys %$confdesc_cloudinit;
4991 foreach my $opt (keys %{$conf->{pending}}) {
4992 next if !grep { $_ eq $opt } @cloudinit_opts;
4993 $conf->{$opt} = delete $conf->{pending}->{$opt};
4994 }
4995
4996 my $new_conf = { %$conf };
4997 $new_conf->{$key} = $value;
4998 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4999 };
5000
5001 foreach my $opt (keys %{$conf->{pending}}) {
5002 next if $selection && !$selection->{$opt};
5003 my $value = $conf->{pending}->{$opt};
5004 eval {
5005 if ($opt eq 'hotplug') {
5006 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
5007 } elsif ($opt eq 'tablet') {
5008 die "skip\n" if !$hotplug_features->{usb};
5009 if ($value == 1) {
5010 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5011 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5012 if $arch eq 'aarch64';
5013 } elsif ($value == 0) {
5014 vm_deviceunplug($vmid, $conf, 'tablet');
5015 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5016 }
5017 } elsif ($opt =~ m/^usb\d+$/) {
5018 die "skip\n";
5019 # since we cannot reliably hot unplug usb devices
5020 # we are disabling it
5021 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
5022 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
5023 die "skip\n" if !$d;
5024 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
5025 } elsif ($opt eq 'vcpus') {
5026 die "skip\n" if !$hotplug_features->{cpu};
5027 qemu_cpu_hotplug($vmid, $conf, $value);
5028 } elsif ($opt eq 'balloon') {
5029 # enable/disable balloning device is not hotpluggable
5030 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
5031 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
5032 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5033
5034 # allow manual ballooning if shares is set to zero
5035 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
5036 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
5037 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
5038 }
5039 } elsif ($opt =~ m/^net(\d+)$/) {
5040 # some changes can be done without hotplug
5041 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
5042 $vmid, $opt, $value, $arch, $machine_type);
5043 } elsif (is_valid_drivename($opt)) {
5044 # some changes can be done without hotplug
5045 my $drive = parse_drive($opt, $value);
5046 if (drive_is_cloudinit($drive)) {
5047 &$apply_pending_cloudinit($opt, $value);
5048 }
5049 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
5050 $vmid, $opt, $value, 1, $arch, $machine_type);
5051 } elsif ($opt =~ m/^memory$/) { #dimms
5052 die "skip\n" if !$hotplug_features->{memory};
5053 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
5054 } elsif ($opt eq 'cpuunits') {
5055 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
5056 } elsif ($opt eq 'cpulimit') {
5057 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
5058 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
5059 } else {
5060 die "skip\n"; # skip non-hot-pluggable options
5061 }
5062 };
5063 if (my $err = $@) {
5064 &$add_error($opt, $err) if $err ne "skip\n";
5065 } else {
5066 # save new config if hotplug was successful
5067 $conf->{$opt} = $value;
5068 delete $conf->{pending}->{$opt};
5069 PVE::QemuConfig->write_config($vmid, $conf);
5070 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5071 }
5072 }
5073 }
5074
5075 sub try_deallocate_drive {
5076 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5077
5078 if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
5079 my $volid = $drive->{file};
5080 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
5081 my $sid = PVE::Storage::parse_volume_id($volid);
5082 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5083
5084 # check if the disk is really unused
5085 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5086 if is_volume_in_use($storecfg, $conf, $key, $volid);
5087 PVE::Storage::vdisk_free($storecfg, $volid);
5088 return 1;
5089 } else {
5090 # If vm is not owner of this disk remove from config
5091 return 1;
5092 }
5093 }
5094
5095 return undef;
5096 }
5097
5098 sub vmconfig_delete_or_detach_drive {
5099 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5100
5101 my $drive = parse_drive($opt, $conf->{$opt});
5102
5103 my $rpcenv = PVE::RPCEnvironment::get();
5104 my $authuser = $rpcenv->get_user();
5105
5106 if ($force) {
5107 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5108 try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5109 } else {
5110 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
5111 }
5112 }
5113
5114 sub vmconfig_apply_pending {
5115 my ($vmid, $conf, $storecfg) = @_;
5116
5117 # cold plug
5118
5119 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
5120 while (my ($opt, $force) = each %$pending_delete_hash) {
5121 die "internal error" if $opt =~ m/^unused/;
5122 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5123 if (!defined($conf->{$opt})) {
5124 vmconfig_undelete_pending_option($conf, $opt);
5125 PVE::QemuConfig->write_config($vmid, $conf);
5126 } elsif (is_valid_drivename($opt)) {
5127 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
5128 vmconfig_undelete_pending_option($conf, $opt);
5129 delete $conf->{$opt};
5130 PVE::QemuConfig->write_config($vmid, $conf);
5131 } else {
5132 vmconfig_undelete_pending_option($conf, $opt);
5133 delete $conf->{$opt};
5134 PVE::QemuConfig->write_config($vmid, $conf);
5135 }
5136 }
5137
5138 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5139
5140 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5141 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5142
5143 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
5144 # skip if nothing changed
5145 } elsif (is_valid_drivename($opt)) {
5146 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
5147 if defined($conf->{$opt});
5148 $conf->{$opt} = $conf->{pending}->{$opt};
5149 } else {
5150 $conf->{$opt} = $conf->{pending}->{$opt};
5151 }
5152
5153 delete $conf->{pending}->{$opt};
5154 PVE::QemuConfig->write_config($vmid, $conf);
5155 }
5156 }
5157
5158 my $safe_num_ne = sub {
5159 my ($a, $b) = @_;
5160
5161 return 0 if !defined($a) && !defined($b);
5162 return 1 if !defined($a);
5163 return 1 if !defined($b);
5164
5165 return $a != $b;
5166 };
5167
5168 my $safe_string_ne = sub {
5169 my ($a, $b) = @_;
5170
5171 return 0 if !defined($a) && !defined($b);
5172 return 1 if !defined($a);
5173 return 1 if !defined($b);
5174
5175 return $a ne $b;
5176 };
5177
5178 sub vmconfig_update_net {
5179 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5180
5181 my $newnet = parse_net($value);
5182
5183 if ($conf->{$opt}) {
5184 my $oldnet = parse_net($conf->{$opt});
5185
5186 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
5187 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
5188 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
5189 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
5190
5191 # for non online change, we try to hot-unplug
5192 die "skip\n" if !$hotplug;
5193 vm_deviceunplug($vmid, $conf, $opt);
5194 } else {
5195
5196 die "internal error" if $opt !~ m/net(\d+)/;
5197 my $iface = "tap${vmid}i$1";
5198
5199 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
5200 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
5201 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
5202 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
5203 PVE::Network::tap_unplug($iface);
5204 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
5205 } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
5206 # Rate can be applied on its own but any change above needs to
5207 # include the rate in tap_plug since OVS resets everything.
5208 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
5209 }
5210
5211 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
5212 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
5213 }
5214
5215 return 1;
5216 }
5217 }
5218
5219 if ($hotplug) {
5220 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5221 } else {
5222 die "skip\n";
5223 }
5224 }
5225
5226 sub vmconfig_update_disk {
5227 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force, $arch, $machine_type) = @_;
5228
5229 # fixme: do we need force?
5230
5231 my $drive = parse_drive($opt, $value);
5232
5233 if ($conf->{$opt}) {
5234
5235 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
5236
5237 my $media = $drive->{media} || 'disk';
5238 my $oldmedia = $old_drive->{media} || 'disk';
5239 die "unable to change media type\n" if $media ne $oldmedia;
5240
5241 if (!drive_is_cdrom($old_drive)) {
5242
5243 if ($drive->{file} ne $old_drive->{file}) {
5244
5245 die "skip\n" if !$hotplug;
5246
5247 # unplug and register as unused
5248 vm_deviceunplug($vmid, $conf, $opt);
5249 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
5250
5251 } else {
5252 # update existing disk
5253
5254 # skip non hotpluggable value
5255 if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
5256 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
5257 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
5258 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
5259 die "skip\n";
5260 }
5261
5262 # apply throttle
5263 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
5264 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
5265 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
5266 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
5267 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
5268 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
5269 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
5270 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
5271 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
5272 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
5273 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
5274 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
5275 &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
5276 &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
5277 &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
5278 &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
5279 &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
5280 &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
5281
5282 qemu_block_set_io_throttle($vmid,"drive-$opt",
5283 ($drive->{mbps} || 0)*1024*1024,
5284 ($drive->{mbps_rd} || 0)*1024*1024,
5285 ($drive->{mbps_wr} || 0)*1024*1024,
5286 $drive->{iops} || 0,
5287 $drive->{iops_rd} || 0,
5288 $drive->{iops_wr} || 0,
5289 ($drive->{mbps_max} || 0)*1024*1024,
5290 ($drive->{mbps_rd_max} || 0)*1024*1024,
5291 ($drive->{mbps_wr_max} || 0)*1024*1024,
5292 $drive->{iops_max} || 0,
5293 $drive->{iops_rd_max} || 0,
5294 $drive->{iops_wr_max} || 0,
5295 $drive->{bps_max_length} || 1,
5296 $drive->{bps_rd_max_length} || 1,
5297 $drive->{bps_wr_max_length} || 1,
5298 $drive->{iops_max_length} || 1,
5299 $drive->{iops_rd_max_length} || 1,
5300 $drive->{iops_wr_max_length} || 1);
5301
5302 }
5303
5304 return 1;
5305 }
5306
5307 } else { # cdrom
5308
5309 if ($drive->{file} eq 'none') {
5310 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
5311 if (drive_is_cloudinit($old_drive)) {
5312 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
5313 }
5314 } else {
5315 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
5316 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
5317 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
5318 }
5319
5320 return 1;
5321 }
5322 }
5323 }
5324
5325 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5326 # hotplug new disks
5327 PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
5328 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5329 }
5330
5331 sub vm_start {
5332 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
5333 $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
5334
5335 PVE::QemuConfig->lock_config($vmid, sub {
5336 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5337
5338 die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
5339
5340 my $is_suspended = PVE::QemuConfig->has_lock($conf, 'suspended');
5341
5342 PVE::QemuConfig->check_lock($conf)
5343 if !($skiplock || $is_suspended);
5344
5345 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
5346
5347 # clean up leftover reboot request files
5348 eval { clear_reboot_request($vmid); };
5349 warn $@ if $@;
5350
5351 if (!$statefile && scalar(keys %{$conf->{pending}})) {
5352 vmconfig_apply_pending($vmid, $conf, $storecfg);
5353 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5354 }
5355
5356 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
5357
5358 my $defaults = load_defaults();
5359
5360 # set environment variable useful inside network script
5361 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
5362
5363 my $local_volumes = {};
5364
5365 if ($targetstorage) {
5366 foreach_drive($conf, sub {
5367 my ($ds, $drive) = @_;
5368
5369 return if drive_is_cdrom($drive);
5370
5371 my $volid = $drive->{file};
5372
5373 return if !$volid;
5374
5375 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5376
5377 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5378 return if $scfg->{shared};
5379 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5380 });
5381
5382 my $format = undef;
5383
5384 foreach my $opt (sort keys %$local_volumes) {
5385
5386 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5387 my $drive = parse_drive($opt, $conf->{$opt});
5388
5389 #if remote storage is specified, use default format
5390 if ($targetstorage && $targetstorage ne "1") {
5391 $storeid = $targetstorage;
5392 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5393 $format = $defFormat;
5394 } else {
5395 #else we use same format than original
5396 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5397 $format = qemu_img_format($scfg, $volid);
5398 }
5399
5400 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5401 my $newdrive = $drive;
5402 $newdrive->{format} = $format;
5403 $newdrive->{file} = $newvolid;
5404 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5405 $local_volumes->{$opt} = $drivestr;
5406 #pass drive to conf for command line
5407 $conf->{$opt} = $drivestr;
5408 }
5409 }
5410
5411 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
5412
5413 if ($is_suspended) {
5414 # enforce machine type on suspended vm to ensure HW compatibility
5415 $forcemachine = $conf->{runningmachine};
5416 print "Resuming suspended VM\n";
5417 }
5418
5419 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5420
5421 my $migrate_uri;
5422 if ($statefile) {
5423 if ($statefile eq 'tcp') {
5424 my $localip = "localhost";
5425 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5426 my $nodename = PVE::INotify::nodename();
5427
5428 if (!defined($migration_type)) {
5429 if (defined($datacenterconf->{migration}->{type})) {
5430 $migration_type = $datacenterconf->{migration}->{type};
5431 } else {
5432 $migration_type = 'secure';
5433 }
5434 }
5435
5436 if ($migration_type eq 'insecure') {
5437 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5438 if ($migrate_network_addr) {
5439 $localip = $migrate_network_addr;
5440 } else {
5441 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5442 }
5443
5444 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5445 }
5446
5447 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5448 my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5449 $migrate_uri = "tcp:${localip}:${migrate_port}";
5450 push @$cmd, '-incoming', $migrate_uri;
5451 push @$cmd, '-S';
5452
5453 } elsif ($statefile eq 'unix') {
5454 # should be default for secure migrations as a ssh TCP forward
5455 # tunnel is not deterministic reliable ready and fails regurarly
5456 # to set up in time, so use UNIX socket forwards
5457 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5458 unlink $socket_addr;
5459
5460 $migrate_uri = "unix:$socket_addr";
5461
5462 push @$cmd, '-incoming', $migrate_uri;
5463 push @$cmd, '-S';
5464
5465 } elsif (-e $statefile) {
5466 push @$cmd, '-loadstate', $statefile;
5467 } else {
5468 my $statepath = PVE::Storage::path($storecfg, $statefile);
5469 push @$vollist, $statepath;
5470 push @$cmd, '-loadstate', $statepath;
5471 }
5472 } elsif ($paused) {
5473 push @$cmd, '-S';
5474 }
5475
5476 # host pci devices
5477 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
5478 my $d = parse_hostpci($conf->{"hostpci$i"});
5479 next if !$d;
5480 my $pcidevices = $d->{pciid};
5481 foreach my $pcidevice (@$pcidevices) {
5482 my $pciid = $pcidevice->{id};
5483
5484 my $info = PVE::SysFSTools::pci_device_info("0000:$pciid");
5485 die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5486 die "no pci device info for device '$pciid'\n" if !$info;
5487
5488 if ($d->{mdev}) {
5489 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5490 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5491 } else {
5492 die "can't unbind/bind pci group to vfio '$pciid'\n"
5493 if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5494 die "can't reset pci device '$pciid'\n"
5495 if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5496 }
5497 }
5498 }
5499
5500 PVE::Storage::activate_volumes($storecfg, $vollist);
5501
5502 eval {
5503 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5504 outfunc => sub {}, errfunc => sub {});
5505 };
5506 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5507 # timeout should be more than enough here...
5508 PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5509
5510 my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5511 : $defaults->{cpuunits};
5512
5513 my $start_timeout = ($conf->{hugepages} || $is_suspended) ? 300 : 30;
5514 my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5515
5516 my %properties = (
5517 Slice => 'qemu.slice',
5518 KillMode => 'none',
5519 CPUShares => $cpuunits
5520 );
5521
5522 if (my $cpulimit = $conf->{cpulimit}) {
5523 $properties{CPUQuota} = int($cpulimit * 100);
5524 }
5525 $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5526
5527 my $run_qemu = sub {
5528 PVE::Tools::run_fork sub {
5529 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5530 run_command($cmd, %run_params);
5531 };
5532 };
5533
5534 if ($conf->{hugepages}) {
5535
5536 my $code = sub {
5537 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5538 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5539
5540 PVE::QemuServer::Memory::hugepages_mount();
5541 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5542
5543 eval { $run_qemu->() };
5544 if (my $err = $@) {
5545 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5546 die $err;
5547 }
5548
5549 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5550 };
5551 eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5552
5553 } else {
5554 eval { $run_qemu->() };
5555 }
5556
5557 if (my $err = $@) {
5558 # deactivate volumes if start fails
5559 eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5560 die "start failed: $err";
5561 }
5562
5563 print "migration listens on $migrate_uri\n" if $migrate_uri;
5564
5565 if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix') {
5566 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5567 warn $@ if $@;
5568 }
5569
5570 #start nbd server for storage migration
5571 if ($targetstorage) {
5572 my $nodename = PVE::INotify::nodename();
5573 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5574 my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5575 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5576 my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5577
5578 vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5579
5580 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5581
5582 foreach my $opt (sort keys %$local_volumes) {
5583 my $volid = $local_volumes->{$opt};
5584 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5585 my $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}:exportname=drive-$opt";
5586 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5587 }
5588 }
5589
5590 if ($migratedfrom) {
5591 eval {
5592 set_migration_caps($vmid);
5593 };
5594 warn $@ if $@;
5595
5596 if ($spice_port) {
5597 print "spice listens on port $spice_port\n";
5598 if ($spice_ticket) {
5599 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5600 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5601 }
5602 }
5603
5604 } else {
5605 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5606 if !$statefile && $conf->{balloon};
5607
5608 foreach my $opt (keys %$conf) {
5609 next if $opt !~ m/^net\d+$/;
5610 my $nicconf = parse_net($conf->{$opt});
5611 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5612 }
5613 }
5614
5615 vm_mon_cmd_nocheck($vmid, 'qom-set',
5616 path => "machine/peripheral/balloon0",
5617 property => "guest-stats-polling-interval",
5618 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5619
5620 if ($is_suspended && (my $vmstate = $conf->{vmstate})) {
5621 print "Resumed VM, removing state\n";
5622 delete $conf->@{qw(lock vmstate runningmachine)};
5623 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5624 PVE::Storage::vdisk_free($storecfg, $vmstate);
5625 PVE::QemuConfig->write_config($vmid, $conf);
5626 }
5627
5628 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5629 });
5630 }
5631
5632 sub vm_mon_cmd {
5633 my ($vmid, $execute, %params) = @_;
5634
5635 my $cmd = { execute => $execute, arguments => \%params };
5636 vm_qmp_command($vmid, $cmd);
5637 }
5638
5639 sub vm_mon_cmd_nocheck {
5640 my ($vmid, $execute, %params) = @_;
5641
5642 my $cmd = { execute => $execute, arguments => \%params };
5643 vm_qmp_command($vmid, $cmd, 1);
5644 }
5645
5646 sub vm_qmp_command {
5647 my ($vmid, $cmd, $nocheck) = @_;
5648
5649 my $res;
5650
5651 my $timeout;
5652 if ($cmd->{arguments}) {
5653 $timeout = delete $cmd->{arguments}->{timeout};
5654 }
5655
5656 eval {
5657 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5658 my $sname = qmp_socket($vmid);
5659 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5660 my $qmpclient = PVE::QMPClient->new();
5661
5662 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5663 } else {
5664 die "unable to open monitor socket\n";
5665 }
5666 };
5667 if (my $err = $@) {
5668 syslog("err", "VM $vmid qmp command failed - $err");
5669 die $err;
5670 }
5671
5672 return $res;
5673 }
5674
5675 sub vm_human_monitor_command {
5676 my ($vmid, $cmdline) = @_;
5677
5678 my $cmd = {
5679 execute => 'human-monitor-command',
5680 arguments => { 'command-line' => $cmdline},
5681 };
5682
5683 return vm_qmp_command($vmid, $cmd);
5684 }
5685
5686 sub vm_commandline {
5687 my ($storecfg, $vmid, $snapname) = @_;
5688
5689 my $conf = PVE::QemuConfig->load_config($vmid);
5690
5691 if ($snapname) {
5692 my $snapshot = $conf->{snapshots}->{$snapname};
5693 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5694
5695 $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5696
5697 $conf = $snapshot;
5698 }
5699
5700 my $defaults = load_defaults();
5701
5702 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5703
5704 return PVE::Tools::cmd2string($cmd);
5705 }
5706
5707 sub vm_reset {
5708 my ($vmid, $skiplock) = @_;
5709
5710 PVE::QemuConfig->lock_config($vmid, sub {
5711
5712 my $conf = PVE::QemuConfig->load_config($vmid);
5713
5714 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5715
5716 vm_mon_cmd($vmid, "system_reset");
5717 });
5718 }
5719
5720 sub get_vm_volumes {
5721 my ($conf) = @_;
5722
5723 my $vollist = [];
5724 foreach_volid($conf, sub {
5725 my ($volid, $attr) = @_;
5726
5727 return if $volid =~ m|^/|;
5728
5729 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5730 return if !$sid;
5731
5732 push @$vollist, $volid;
5733 });
5734
5735 return $vollist;
5736 }
5737
5738 sub vm_stop_cleanup {
5739 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5740
5741 eval {
5742
5743 if (!$keepActive) {
5744 my $vollist = get_vm_volumes($conf);
5745 PVE::Storage::deactivate_volumes($storecfg, $vollist);
5746 }
5747
5748 foreach my $ext (qw(mon qmp pid vnc qga)) {
5749 unlink "/var/run/qemu-server/${vmid}.$ext";
5750 }
5751
5752 if ($conf->{ivshmem}) {
5753 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5754 # just delete it for now, VMs which have this already open do not
5755 # are affected, but new VMs will get a separated one. If this
5756 # becomes an issue we either add some sort of ref-counting or just
5757 # add a "don't delete on stop" flag to the ivshmem format.
5758 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5759 }
5760
5761 foreach my $key (keys %$conf) {
5762 next if $key !~ m/^hostpci(\d+)$/;
5763 my $hostpciindex = $1;
5764 my $d = parse_hostpci($conf->{$key});
5765 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5766
5767 foreach my $pci (@{$d->{pciid}}) {
5768 my $pciid = $pci->{id};
5769 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5770 }
5771 }
5772
5773 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5774 };
5775 warn $@ if $@; # avoid errors - just warn
5776 }
5777
5778 # call only in locked context
5779 sub _do_vm_stop {
5780 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5781
5782 my $pid = check_running($vmid, $nocheck);
5783 return if !$pid;
5784
5785 my $conf;
5786 if (!$nocheck) {
5787 $conf = PVE::QemuConfig->load_config($vmid);
5788 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5789 if (!defined($timeout) && $shutdown && $conf->{startup}) {
5790 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5791 $timeout = $opts->{down} if $opts->{down};
5792 }
5793 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5794 }
5795
5796 eval {
5797 if ($shutdown) {
5798 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5799 vm_qmp_command($vmid, {
5800 execute => "guest-shutdown",
5801 arguments => { timeout => $timeout }
5802 }, $nocheck);
5803 } else {
5804 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5805 }
5806 } else {
5807 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5808 }
5809 };
5810 my $err = $@;
5811
5812 if (!$err) {
5813 $timeout = 60 if !defined($timeout);
5814
5815 my $count = 0;
5816 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5817 $count++;
5818 sleep 1;
5819 }
5820
5821 if ($count >= $timeout) {
5822 if ($force) {
5823 warn "VM still running - terminating now with SIGTERM\n";
5824 kill 15, $pid;
5825 } else {
5826 die "VM quit/powerdown failed - got timeout\n";
5827 }
5828 } else {
5829 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5830 return;
5831 }
5832 } else {
5833 if ($force) {
5834 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5835 kill 15, $pid;
5836 } else {
5837 die "VM quit/powerdown failed\n";
5838 }
5839 }
5840
5841 # wait again
5842 $timeout = 10;
5843
5844 my $count = 0;
5845 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5846 $count++;
5847 sleep 1;
5848 }
5849
5850 if ($count >= $timeout) {
5851 warn "VM still running - terminating now with SIGKILL\n";
5852 kill 9, $pid;
5853 sleep 1;
5854 }
5855
5856 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5857 }
5858
5859 # Note: use $nocheck to skip tests if VM configuration file exists.
5860 # We need that when migration VMs to other nodes (files already moved)
5861 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5862 sub vm_stop {
5863 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5864
5865 $force = 1 if !defined($force) && !$shutdown;
5866
5867 if ($migratedfrom){
5868 my $pid = check_running($vmid, $nocheck, $migratedfrom);
5869 kill 15, $pid if $pid;
5870 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5871 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5872 return;
5873 }
5874
5875 PVE::QemuConfig->lock_config($vmid, sub {
5876 _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5877 });
5878 }
5879
5880 sub vm_reboot {
5881 my ($vmid, $timeout) = @_;
5882
5883 PVE::QemuConfig->lock_config($vmid, sub {
5884
5885 # only reboot if running, as qmeventd starts it again on a stop event
5886 return if !check_running($vmid);
5887
5888 create_reboot_request($vmid);
5889
5890 my $storecfg = PVE::Storage::config();
5891 _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5892
5893 });
5894 }
5895
5896 sub vm_suspend {
5897 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5898
5899 my $conf;
5900 my $path;
5901 my $storecfg;
5902 my $vmstate;
5903
5904 PVE::QemuConfig->lock_config($vmid, sub {
5905
5906 $conf = PVE::QemuConfig->load_config($vmid);
5907
5908 my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5909 PVE::QemuConfig->check_lock($conf)
5910 if !($skiplock || $is_backing_up);
5911
5912 die "cannot suspend to disk during backup\n"
5913 if $is_backing_up && $includestate;
5914
5915 if ($includestate) {
5916 $conf->{lock} = 'suspending';
5917 my $date = strftime("%Y-%m-%d", localtime(time()));
5918 $storecfg = PVE::Storage::config();
5919 $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5920 $path = PVE::Storage::path($storecfg, $vmstate);
5921 PVE::QemuConfig->write_config($vmid, $conf);
5922 } else {
5923 vm_mon_cmd($vmid, "stop");
5924 }
5925 });
5926
5927 if ($includestate) {
5928 # save vm state
5929 PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5930
5931 eval {
5932 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5933 for(;;) {
5934 my $state = vm_mon_cmd_nocheck($vmid, "query-savevm");
5935 if (!$state->{status}) {
5936 die "savevm not active\n";
5937 } elsif ($state->{status} eq 'active') {
5938 sleep(1);
5939 next;
5940 } elsif ($state->{status} eq 'completed') {
5941 print "State saved, quitting\n";
5942 last;
5943 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5944 die "query-savevm failed with error '$state->{error}'\n"
5945 } else {
5946 die "query-savevm returned status '$state->{status}'\n";
5947 }
5948 }
5949 };
5950 my $err = $@;
5951
5952 PVE::QemuConfig->lock_config($vmid, sub {
5953 $conf = PVE::QemuConfig->load_config($vmid);
5954 if ($err) {
5955 # cleanup, but leave suspending lock, to indicate something went wrong
5956 eval {
5957 vm_mon_cmd($vmid, "savevm-end");
5958 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5959 PVE::Storage::vdisk_free($storecfg, $vmstate);
5960 delete $conf->@{qw(vmstate runningmachine)};
5961 PVE::QemuConfig->write_config($vmid, $conf);
5962 };
5963 warn $@ if $@;
5964 die $err;
5965 }
5966
5967 die "lock changed unexpectedly\n"
5968 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5969
5970 vm_qmp_command($vmid, { execute => "quit" });
5971 $conf->{lock} = 'suspended';
5972 PVE::QemuConfig->write_config($vmid, $conf);
5973 });
5974 }
5975 }
5976
5977 sub vm_resume {
5978 my ($vmid, $skiplock, $nocheck) = @_;
5979
5980 PVE::QemuConfig->lock_config($vmid, sub {
5981 my $vm_mon_cmd = $nocheck ? \&vm_mon_cmd_nocheck : \&vm_mon_cmd;
5982 my $res = $vm_mon_cmd->($vmid, 'query-status');
5983 my $resume_cmd = 'cont';
5984
5985 if ($res->{status} && $res->{status} eq 'suspended') {
5986 $resume_cmd = 'system_wakeup';
5987 }
5988
5989 if (!$nocheck) {
5990
5991 my $conf = PVE::QemuConfig->load_config($vmid);
5992
5993 PVE::QemuConfig->check_lock($conf)
5994 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5995 }
5996
5997 $vm_mon_cmd->($vmid, $resume_cmd);
5998 });
5999 }
6000
6001 sub vm_sendkey {
6002 my ($vmid, $skiplock, $key) = @_;
6003
6004 PVE::QemuConfig->lock_config($vmid, sub {
6005
6006 my $conf = PVE::QemuConfig->load_config($vmid);
6007
6008 # there is no qmp command, so we use the human monitor command
6009 my $res = vm_human_monitor_command($vmid, "sendkey $key");
6010 die $res if $res ne '';
6011 });
6012 }
6013
6014 sub vm_destroy {
6015 my ($storecfg, $vmid, $skiplock) = @_;
6016
6017 PVE::QemuConfig->lock_config($vmid, sub {
6018
6019 my $conf = PVE::QemuConfig->load_config($vmid);
6020
6021 if (!check_running($vmid)) {
6022 destroy_vm($storecfg, $vmid, undef, $skiplock);
6023 } else {
6024 die "VM $vmid is running - destroy failed\n";
6025 }
6026 });
6027 }
6028
6029 # vzdump restore implementaion
6030
6031 sub tar_archive_read_firstfile {
6032 my $archive = shift;
6033
6034 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
6035
6036 # try to detect archive type first
6037 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6038 die "unable to open file '$archive'\n";
6039 my $firstfile = <$fh>;
6040 kill 15, $pid;
6041 close $fh;
6042
6043 die "ERROR: archive contaions no data\n" if !$firstfile;
6044 chomp $firstfile;
6045
6046 return $firstfile;
6047 }
6048
6049 sub tar_restore_cleanup {
6050 my ($storecfg, $statfile) = @_;
6051
6052 print STDERR "starting cleanup\n";
6053
6054 if (my $fd = IO::File->new($statfile, "r")) {
6055 while (defined(my $line = <$fd>)) {
6056 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6057 my $volid = $2;
6058 eval {
6059 if ($volid =~ m|^/|) {
6060 unlink $volid || die 'unlink failed\n';
6061 } else {
6062 PVE::Storage::vdisk_free($storecfg, $volid);
6063 }
6064 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6065 };
6066 print STDERR "unable to cleanup '$volid' - $@" if $@;
6067 } else {
6068 print STDERR "unable to parse line in statfile - $line";
6069 }
6070 }
6071 $fd->close();
6072 }
6073 }
6074
6075 sub restore_archive {
6076 my ($archive, $vmid, $user, $opts) = @_;
6077
6078 my $format = $opts->{format};
6079 my $comp;
6080
6081 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
6082 $format = 'tar' if !$format;
6083 $comp = 'gzip';
6084 } elsif ($archive =~ m/\.tar$/) {
6085 $format = 'tar' if !$format;
6086 } elsif ($archive =~ m/.tar.lzo$/) {
6087 $format = 'tar' if !$format;
6088 $comp = 'lzop';
6089 } elsif ($archive =~ m/\.vma$/) {
6090 $format = 'vma' if !$format;
6091 } elsif ($archive =~ m/\.vma\.gz$/) {
6092 $format = 'vma' if !$format;
6093 $comp = 'gzip';
6094 } elsif ($archive =~ m/\.vma\.lzo$/) {
6095 $format = 'vma' if !$format;
6096 $comp = 'lzop';
6097 } else {
6098 $format = 'vma' if !$format; # default
6099 }
6100
6101 # try to detect archive format
6102 if ($format eq 'tar') {
6103 return restore_tar_archive($archive, $vmid, $user, $opts);
6104 } else {
6105 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
6106 }
6107 }
6108
6109 sub restore_update_config_line {
6110 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
6111
6112 return if $line =~ m/^\#qmdump\#/;
6113 return if $line =~ m/^\#vzdump\#/;
6114 return if $line =~ m/^lock:/;
6115 return if $line =~ m/^unused\d+:/;
6116 return if $line =~ m/^parent:/;
6117
6118 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
6119 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6120 # try to convert old 1.X settings
6121 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6122 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
6123 my ($model, $macaddr) = split(/\=/, $devconfig);
6124 $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
6125 my $net = {
6126 model => $model,
6127 bridge => "vmbr$ind",
6128 macaddr => $macaddr,
6129 };
6130 my $netstr = print_net($net);
6131
6132 print $outfd "net$cookie->{netcount}: $netstr\n";
6133 $cookie->{netcount}++;
6134 }
6135 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6136 my ($id, $netstr) = ($1, $2);
6137 my $net = parse_net($netstr);
6138 $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
6139 $netstr = print_net($net);
6140 print $outfd "$id: $netstr\n";
6141 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
6142 my $virtdev = $1;
6143 my $value = $3;
6144 my $di = parse_drive($virtdev, $value);
6145 if (defined($di->{backup}) && !$di->{backup}) {
6146 print $outfd "#$line";
6147 } elsif ($map->{$virtdev}) {
6148 delete $di->{format}; # format can change on restore
6149 $di->{file} = $map->{$virtdev};
6150 $value = print_drive($vmid, $di);
6151 print $outfd "$virtdev: $value\n";
6152 } else {
6153 print $outfd $line;
6154 }
6155 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6156 my $vmgenid = $1;
6157 if ($vmgenid ne '0') {
6158 # always generate a new vmgenid if there was a valid one setup
6159 $vmgenid = generate_uuid();
6160 }
6161 print $outfd "vmgenid: $vmgenid\n";
6162 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6163 my ($uuid, $uuid_str);
6164 UUID::generate($uuid);
6165 UUID::unparse($uuid, $uuid_str);
6166 my $smbios1 = parse_smbios1($2);
6167 $smbios1->{uuid} = $uuid_str;
6168 print $outfd $1.print_smbios1($smbios1)."\n";
6169 } else {
6170 print $outfd $line;
6171 }
6172 }
6173
6174 sub scan_volids {
6175 my ($cfg, $vmid) = @_;
6176
6177 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
6178
6179 my $volid_hash = {};
6180 foreach my $storeid (keys %$info) {
6181 foreach my $item (@{$info->{$storeid}}) {
6182 next if !($item->{volid} && $item->{size});
6183 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
6184 $volid_hash->{$item->{volid}} = $item;
6185 }
6186 }
6187
6188 return $volid_hash;
6189 }
6190
6191 sub is_volume_in_use {
6192 my ($storecfg, $conf, $skip_drive, $volid) = @_;
6193
6194 my $path = PVE::Storage::path($storecfg, $volid);
6195
6196 my $scan_config = sub {
6197 my ($cref, $snapname) = @_;
6198
6199 foreach my $key (keys %$cref) {
6200 my $value = $cref->{$key};
6201 if (is_valid_drivename($key)) {
6202 next if $skip_drive && $key eq $skip_drive;
6203 my $drive = parse_drive($key, $value);
6204 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
6205 return 1 if $volid eq $drive->{file};
6206 if ($drive->{file} =~ m!^/!) {
6207 return 1 if $drive->{file} eq $path;
6208 } else {
6209 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
6210 next if !$storeid;
6211 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
6212 next if !$scfg;
6213 return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
6214 }
6215 }
6216 }
6217
6218 return 0;
6219 };
6220
6221 return 1 if &$scan_config($conf);
6222
6223 undef $skip_drive;
6224
6225 foreach my $snapname (keys %{$conf->{snapshots}}) {
6226 return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
6227 }
6228
6229 return 0;
6230 }
6231
6232 sub update_disksize {
6233 my ($vmid, $conf, $volid_hash) = @_;
6234
6235 my $changes;
6236 my $prefix = "VM $vmid:";
6237
6238 # used and unused disks
6239 my $referenced = {};
6240
6241 # Note: it is allowed to define multiple storages with same path (alias), so
6242 # we need to check both 'volid' and real 'path' (two different volid can point
6243 # to the same path).
6244
6245 my $referencedpath = {};
6246
6247 # update size info
6248 foreach my $opt (keys %$conf) {
6249 if (is_valid_drivename($opt)) {
6250 my $drive = parse_drive($opt, $conf->{$opt});
6251 my $volid = $drive->{file};
6252 next if !$volid;
6253
6254 $referenced->{$volid} = 1;
6255 if ($volid_hash->{$volid} &&
6256 (my $path = $volid_hash->{$volid}->{path})) {
6257 $referencedpath->{$path} = 1;
6258 }
6259
6260 next if drive_is_cdrom($drive);
6261 next if !$volid_hash->{$volid};
6262
6263 $drive->{size} = $volid_hash->{$volid}->{size};
6264 my $new = print_drive($vmid, $drive);
6265 if ($new ne $conf->{$opt}) {
6266 $changes = 1;
6267 $conf->{$opt} = $new;
6268 print "$prefix update disk '$opt' information.\n";
6269 }
6270 }
6271 }
6272
6273 # remove 'unusedX' entry if volume is used
6274 foreach my $opt (keys %$conf) {
6275 next if $opt !~ m/^unused\d+$/;
6276 my $volid = $conf->{$opt};
6277 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
6278 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6279 print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
6280 $changes = 1;
6281 delete $conf->{$opt};
6282 }
6283
6284 $referenced->{$volid} = 1;
6285 $referencedpath->{$path} = 1 if $path;
6286 }
6287
6288 foreach my $volid (sort keys %$volid_hash) {
6289 next if $volid =~ m/vm-$vmid-state-/;
6290 next if $referenced->{$volid};
6291 my $path = $volid_hash->{$volid}->{path};
6292 next if !$path; # just to be sure
6293 next if $referencedpath->{$path};
6294 $changes = 1;
6295 my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
6296 print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
6297 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6298 }
6299
6300 return $changes;
6301 }
6302
6303 sub rescan {
6304 my ($vmid, $nolock, $dryrun) = @_;
6305
6306 my $cfg = PVE::Storage::config();
6307
6308 # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
6309 # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
6310 foreach my $stor (keys %{$cfg->{ids}}) {
6311 delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
6312 }
6313
6314 print "rescan volumes...\n";
6315 my $volid_hash = scan_volids($cfg, $vmid);
6316
6317 my $updatefn = sub {
6318 my ($vmid) = @_;
6319
6320 my $conf = PVE::QemuConfig->load_config($vmid);
6321
6322 PVE::QemuConfig->check_lock($conf);
6323
6324 my $vm_volids = {};
6325 foreach my $volid (keys %$volid_hash) {
6326 my $info = $volid_hash->{$volid};
6327 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
6328 }
6329
6330 my $changes = update_disksize($vmid, $conf, $vm_volids);
6331
6332 PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
6333 };
6334
6335 if (defined($vmid)) {
6336 if ($nolock) {
6337 &$updatefn($vmid);
6338 } else {
6339 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6340 }
6341 } else {
6342 my $vmlist = config_list();
6343 foreach my $vmid (keys %$vmlist) {
6344 if ($nolock) {
6345 &$updatefn($vmid);
6346 } else {
6347 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6348 }
6349 }
6350 }
6351 }
6352
6353 sub restore_vma_archive {
6354 my ($archive, $vmid, $user, $opts, $comp) = @_;
6355
6356 my $readfrom = $archive;
6357
6358 my $cfg = PVE::Storage::config();
6359 my $commands = [];
6360 my $bwlimit = $opts->{bwlimit};
6361
6362 my $dbg_cmdstring = '';
6363 my $add_pipe = sub {
6364 my ($cmd) = @_;
6365 push @$commands, $cmd;
6366 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6367 $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6368 $readfrom = '-';
6369 };
6370
6371 my $input = undef;
6372 if ($archive eq '-') {
6373 $input = '<&STDIN';
6374 } else {
6375 # If we use a backup from a PVE defined storage we also consider that
6376 # storage's rate limit:
6377 my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6378 if (defined($volid)) {
6379 my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6380 my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6381 if ($readlimit) {
6382 print STDERR "applying read rate limit: $readlimit\n";
6383 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6384 $add_pipe->($cstream);
6385 }
6386 }
6387 }
6388
6389 if ($comp) {
6390 my $cmd;
6391 if ($comp eq 'gzip') {
6392 $cmd = ['zcat', $readfrom];
6393 } elsif ($comp eq 'lzop') {
6394 $cmd = ['lzop', '-d', '-c', $readfrom];
6395 } else {
6396 die "unknown compression method '$comp'\n";
6397 }
6398 $add_pipe->($cmd);
6399 }
6400
6401 my $tmpdir = "/var/tmp/vzdumptmp$$";
6402 rmtree $tmpdir;
6403
6404 # disable interrupts (always do cleanups)
6405 local $SIG{INT} =
6406 local $SIG{TERM} =
6407 local $SIG{QUIT} =
6408 local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6409
6410 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6411 POSIX::mkfifo($mapfifo, 0600);
6412 my $fifofh;
6413
6414 my $openfifo = sub {
6415 open($fifofh, '>', $mapfifo) || die $!;
6416 };
6417
6418 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6419
6420 my $oldtimeout;
6421 my $timeout = 5;
6422
6423 my $devinfo = {};
6424
6425 my $rpcenv = PVE::RPCEnvironment::get();
6426
6427 my $conffile = PVE::QemuConfig->config_file($vmid);
6428 my $tmpfn = "$conffile.$$.tmp";
6429
6430 # Note: $oldconf is undef if VM does not exists
6431 my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6432 my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6433
6434 my %storage_limits;
6435
6436 my $print_devmap = sub {
6437 my $virtdev_hash = {};
6438
6439 my $cfgfn = "$tmpdir/qemu-server.conf";
6440
6441 # we can read the config - that is already extracted
6442 my $fh = IO::File->new($cfgfn, "r") ||
6443 "unable to read qemu-server.conf - $!\n";
6444
6445 my $fwcfgfn = "$tmpdir/qemu-server.fw";
6446 if (-f $fwcfgfn) {
6447 my $pve_firewall_dir = '/etc/pve/firewall';
6448 mkdir $pve_firewall_dir; # make sure the dir exists
6449 PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6450 }
6451
6452 while (defined(my $line = <$fh>)) {
6453 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6454 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6455 die "archive does not contain data for drive '$virtdev'\n"
6456 if !$devinfo->{$devname};
6457 if (defined($opts->{storage})) {
6458 $storeid = $opts->{storage} || 'local';
6459 } elsif (!$storeid) {
6460 $storeid = 'local';
6461 }
6462 $format = 'raw' if !$format;
6463 $devinfo->{$devname}->{devname} = $devname;
6464 $devinfo->{$devname}->{virtdev} = $virtdev;
6465 $devinfo->{$devname}->{format} = $format;
6466 $devinfo->{$devname}->{storeid} = $storeid;
6467
6468 # check permission on storage
6469 my $pool = $opts->{pool}; # todo: do we need that?
6470 if ($user ne 'root@pam') {
6471 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6472 }
6473
6474 $storage_limits{$storeid} = $bwlimit;
6475
6476 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6477 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6478 my $virtdev = $1;
6479 my $drive = parse_drive($virtdev, $2);
6480 if (drive_is_cloudinit($drive)) {
6481 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6482 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6483 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
6484
6485 my $d = {
6486 format => $format,
6487 storeid => $opts->{storage} // $storeid,
6488 size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
6489 file => $drive->{file}, # to make drive_is_cloudinit check possible
6490 name => "vm-$vmid-cloudinit",
6491 is_cloudinit => 1,
6492 };
6493 $virtdev_hash->{$virtdev} = $d;
6494 }
6495 }
6496 }
6497
6498 foreach my $key (keys %storage_limits) {
6499 my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6500 next if !$limit;
6501 print STDERR "rate limit for storage $key: $limit KiB/s\n";
6502 $storage_limits{$key} = $limit * 1024;
6503 }
6504
6505 foreach my $devname (keys %$devinfo) {
6506 die "found no device mapping information for device '$devname'\n"
6507 if !$devinfo->{$devname}->{virtdev};
6508 }
6509
6510 # create empty/temp config
6511 if ($oldconf) {
6512 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6513 foreach_drive($oldconf, sub {
6514 my ($ds, $drive) = @_;
6515
6516 return if drive_is_cdrom($drive, 1);
6517
6518 my $volid = $drive->{file};
6519 return if !$volid || $volid =~ m|^/|;
6520
6521 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6522 return if !$path || !$owner || ($owner != $vmid);
6523
6524 # Note: only delete disk we want to restore
6525 # other volumes will become unused
6526 if ($virtdev_hash->{$ds}) {
6527 eval { PVE::Storage::vdisk_free($cfg, $volid); };
6528 if (my $err = $@) {
6529 warn $err;
6530 }
6531 }
6532 });
6533
6534 # delete vmstate files, after the restore we have no snapshots anymore
6535 foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6536 my $snap = $oldconf->{snapshots}->{$snapname};
6537 if ($snap->{vmstate}) {
6538 eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6539 if (my $err = $@) {
6540 warn $err;
6541 }
6542 }
6543 }
6544 }
6545
6546 my $map = {};
6547 foreach my $virtdev (sort keys %$virtdev_hash) {
6548 my $d = $virtdev_hash->{$virtdev};
6549 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6550 my $storeid = $d->{storeid};
6551 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6552
6553 my $map_opts = '';
6554 if (my $limit = $storage_limits{$storeid}) {
6555 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6556 }
6557
6558 # test if requested format is supported
6559 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6560 my $supported = grep { $_ eq $d->{format} } @$validFormats;
6561 $d->{format} = $defFormat if !$supported;
6562
6563 my $name;
6564 if ($d->{is_cloudinit}) {
6565 $name = $d->{name};
6566 $name .= ".$d->{format}" if $d->{format} ne 'raw';
6567 }
6568
6569 my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
6570 print STDERR "new volume ID is '$volid'\n";
6571 $d->{volid} = $volid;
6572
6573 PVE::Storage::activate_volumes($cfg, [$volid]);
6574
6575 my $write_zeros = 1;
6576 if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6577 $write_zeros = 0;
6578 }
6579
6580 if (!$d->{is_cloudinit}) {
6581 my $path = PVE::Storage::path($cfg, $volid);
6582
6583 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6584
6585 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6586 }
6587 $map->{$virtdev} = $volid;
6588 }
6589
6590 $fh->seek(0, 0) || die "seek failed - $!\n";
6591
6592 my $outfd = new IO::File ($tmpfn, "w") ||
6593 die "unable to write config for VM $vmid\n";
6594
6595 my $cookie = { netcount => 0 };
6596 while (defined(my $line = <$fh>)) {
6597 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6598 }
6599
6600 $fh->close();
6601 $outfd->close();
6602 };
6603
6604 eval {
6605 # enable interrupts
6606 local $SIG{INT} =
6607 local $SIG{TERM} =
6608 local $SIG{QUIT} =
6609 local $SIG{HUP} =
6610 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6611 local $SIG{ALRM} = sub { die "got timeout\n"; };
6612
6613 $oldtimeout = alarm($timeout);
6614
6615 my $parser = sub {
6616 my $line = shift;
6617
6618 print "$line\n";
6619
6620 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6621 my ($dev_id, $size, $devname) = ($1, $2, $3);
6622 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6623 } elsif ($line =~ m/^CTIME: /) {
6624 # we correctly received the vma config, so we can disable
6625 # the timeout now for disk allocation (set to 10 minutes, so
6626 # that we always timeout if something goes wrong)
6627 alarm(600);
6628 &$print_devmap();
6629 print $fifofh "done\n";
6630 my $tmp = $oldtimeout || 0;
6631 $oldtimeout = undef;
6632 alarm($tmp);
6633 close($fifofh);
6634 }
6635 };
6636
6637 print "restore vma archive: $dbg_cmdstring\n";
6638 run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6639 };
6640 my $err = $@;
6641
6642 alarm($oldtimeout) if $oldtimeout;
6643
6644 my $vollist = [];
6645 foreach my $devname (keys %$devinfo) {
6646 my $volid = $devinfo->{$devname}->{volid};
6647 push @$vollist, $volid if $volid;
6648 }
6649
6650 PVE::Storage::deactivate_volumes($cfg, $vollist);
6651
6652 unlink $mapfifo;
6653
6654 if ($err) {
6655 rmtree $tmpdir;
6656 unlink $tmpfn;
6657
6658 foreach my $devname (keys %$devinfo) {
6659 my $volid = $devinfo->{$devname}->{volid};
6660 next if !$volid;
6661 eval {
6662 if ($volid =~ m|^/|) {
6663 unlink $volid || die 'unlink failed\n';
6664 } else {
6665 PVE::Storage::vdisk_free($cfg, $volid);
6666 }
6667 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6668 };
6669 print STDERR "unable to cleanup '$volid' - $@" if $@;
6670 }
6671 die $err;
6672 }
6673
6674 rmtree $tmpdir;
6675
6676 rename($tmpfn, $conffile) ||
6677 die "unable to commit configuration file '$conffile'\n";
6678
6679 PVE::Cluster::cfs_update(); # make sure we read new file
6680
6681 eval { rescan($vmid, 1); };
6682 warn $@ if $@;
6683 }
6684
6685 sub restore_tar_archive {
6686 my ($archive, $vmid, $user, $opts) = @_;
6687
6688 if ($archive ne '-') {
6689 my $firstfile = tar_archive_read_firstfile($archive);
6690 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6691 if $firstfile ne 'qemu-server.conf';
6692 }
6693
6694 my $storecfg = PVE::Storage::config();
6695
6696 # destroy existing data - keep empty config
6697 my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6698 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6699
6700 my $tocmd = "/usr/lib/qemu-server/qmextract";
6701
6702 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6703 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6704 $tocmd .= ' --prealloc' if $opts->{prealloc};
6705 $tocmd .= ' --info' if $opts->{info};
6706
6707 # tar option "xf" does not autodetect compression when read from STDIN,
6708 # so we pipe to zcat
6709 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6710 PVE::Tools::shellquote("--to-command=$tocmd");
6711
6712 my $tmpdir = "/var/tmp/vzdumptmp$$";
6713 mkpath $tmpdir;
6714
6715 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6716 local $ENV{VZDUMP_VMID} = $vmid;
6717 local $ENV{VZDUMP_USER} = $user;
6718
6719 my $conffile = PVE::QemuConfig->config_file($vmid);
6720 my $tmpfn = "$conffile.$$.tmp";
6721
6722 # disable interrupts (always do cleanups)
6723 local $SIG{INT} =
6724 local $SIG{TERM} =
6725 local $SIG{QUIT} =
6726 local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6727
6728 eval {
6729 # enable interrupts
6730 local $SIG{INT} =
6731 local $SIG{TERM} =
6732 local $SIG{QUIT} =
6733 local $SIG{HUP} =
6734 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6735
6736 if ($archive eq '-') {
6737 print "extracting archive from STDIN\n";
6738 run_command($cmd, input => "<&STDIN");
6739 } else {
6740 print "extracting archive '$archive'\n";
6741 run_command($cmd);
6742 }
6743
6744 return if $opts->{info};
6745
6746 # read new mapping
6747 my $map = {};
6748 my $statfile = "$tmpdir/qmrestore.stat";
6749 if (my $fd = IO::File->new($statfile, "r")) {
6750 while (defined (my $line = <$fd>)) {
6751 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6752 $map->{$1} = $2 if $1;
6753 } else {
6754 print STDERR "unable to parse line in statfile - $line\n";
6755 }
6756 }
6757 $fd->close();
6758 }
6759
6760 my $confsrc = "$tmpdir/qemu-server.conf";
6761
6762 my $srcfd = new IO::File($confsrc, "r") ||
6763 die "unable to open file '$confsrc'\n";
6764
6765 my $outfd = new IO::File ($tmpfn, "w") ||
6766 die "unable to write config for VM $vmid\n";
6767
6768 my $cookie = { netcount => 0 };
6769 while (defined (my $line = <$srcfd>)) {
6770 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6771 }
6772
6773 $srcfd->close();
6774 $outfd->close();
6775 };
6776 my $err = $@;
6777
6778 if ($err) {
6779
6780 unlink $tmpfn;
6781
6782 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6783
6784 die $err;
6785 }
6786
6787 rmtree $tmpdir;
6788
6789 rename $tmpfn, $conffile ||
6790 die "unable to commit configuration file '$conffile'\n";
6791
6792 PVE::Cluster::cfs_update(); # make sure we read new file
6793
6794 eval { rescan($vmid, 1); };
6795 warn $@ if $@;
6796 };
6797
6798 sub foreach_storage_used_by_vm {
6799 my ($conf, $func) = @_;
6800
6801 my $sidhash = {};
6802
6803 foreach_drive($conf, sub {
6804 my ($ds, $drive) = @_;
6805 return if drive_is_cdrom($drive);
6806
6807 my $volid = $drive->{file};
6808
6809 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6810 $sidhash->{$sid} = $sid if $sid;
6811 });
6812
6813 foreach my $sid (sort keys %$sidhash) {
6814 &$func($sid);
6815 }
6816 }
6817
6818 sub do_snapshots_with_qemu {
6819 my ($storecfg, $volid) = @_;
6820
6821 my $storage_name = PVE::Storage::parse_volume_id($volid);
6822 my $scfg = $storecfg->{ids}->{$storage_name};
6823
6824 if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6825 return 1;
6826 }
6827
6828 if ($volid =~ m/\.(qcow2|qed)$/){
6829 return 1;
6830 }
6831
6832 return undef;
6833 }
6834
6835 sub qga_check_running {
6836 my ($vmid, $nowarn) = @_;
6837
6838 eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6839 if ($@) {
6840 warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6841 return 0;
6842 }
6843 return 1;
6844 }
6845
6846 sub template_create {
6847 my ($vmid, $conf, $disk) = @_;
6848
6849 my $storecfg = PVE::Storage::config();
6850
6851 foreach_drive($conf, sub {
6852 my ($ds, $drive) = @_;
6853
6854 return if drive_is_cdrom($drive);
6855 return if $disk && $ds ne $disk;
6856
6857 my $volid = $drive->{file};
6858 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6859
6860 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6861 $drive->{file} = $voliddst;
6862 $conf->{$ds} = print_drive($vmid, $drive);
6863 PVE::QemuConfig->write_config($vmid, $conf);
6864 });
6865 }
6866
6867 sub convert_iscsi_path {
6868 my ($path) = @_;
6869
6870 if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6871 my $portal = $1;
6872 my $target = $2;
6873 my $lun = $3;
6874
6875 my $initiator_name = get_initiator_name();
6876
6877 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6878 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6879 }
6880
6881 die "cannot convert iscsi path '$path', unkown format\n";
6882 }
6883
6884 sub qemu_img_convert {
6885 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6886
6887 my $storecfg = PVE::Storage::config();
6888 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6889 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6890
6891 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6892
6893 my $cachemode;
6894 my $src_path;
6895 my $src_is_iscsi = 0;
6896 my $src_format = 'raw';
6897
6898 if ($src_storeid) {
6899 PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6900 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6901 $src_format = qemu_img_format($src_scfg, $src_volname);
6902 $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6903 $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6904 $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6905 } elsif (-f $src_volid) {
6906 $src_path = $src_volid;
6907 if ($src_path =~ m/\.($QEMU_FORMAT_RE)$/) {
6908 $src_format = $1;
6909 }
6910 }
6911
6912 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6913
6914 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6915 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6916 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6917 my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6918
6919 my $cmd = [];
6920 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6921 push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6922 push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6923 push @$cmd, '-T', $cachemode if defined($cachemode);
6924
6925 if ($src_is_iscsi) {
6926 push @$cmd, '--image-opts';
6927 $src_path = convert_iscsi_path($src_path);
6928 } else {
6929 push @$cmd, '-f', $src_format;
6930 }
6931
6932 if ($dst_is_iscsi) {
6933 push @$cmd, '--target-image-opts';
6934 $dst_path = convert_iscsi_path($dst_path);
6935 } else {
6936 push @$cmd, '-O', $dst_format;
6937 }
6938
6939 push @$cmd, $src_path;
6940
6941 if (!$dst_is_iscsi && $is_zero_initialized) {
6942 push @$cmd, "zeroinit:$dst_path";
6943 } else {
6944 push @$cmd, $dst_path;
6945 }
6946
6947 my $parser = sub {
6948 my $line = shift;
6949 if($line =~ m/\((\S+)\/100\%\)/){
6950 my $percent = $1;
6951 my $transferred = int($size * $percent / 100);
6952 my $remaining = $size - $transferred;
6953
6954 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6955 }
6956
6957 };
6958
6959 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6960 my $err = $@;
6961 die "copy failed: $err" if $err;
6962 }
6963
6964 sub qemu_img_format {
6965 my ($scfg, $volname) = @_;
6966
6967 if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6968 return $1;
6969 } else {
6970 return "raw";
6971 }
6972 }
6973
6974 sub qemu_drive_mirror {
6975 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
6976
6977 $jobs = {} if !$jobs;
6978
6979 my $qemu_target;
6980 my $format;
6981 $jobs->{"drive-$drive"} = {};
6982
6983 if ($dst_volid =~ /^nbd:/) {
6984 $qemu_target = $dst_volid;
6985 $format = "nbd";
6986 } else {
6987 my $storecfg = PVE::Storage::config();
6988 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6989
6990 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6991
6992 $format = qemu_img_format($dst_scfg, $dst_volname);
6993
6994 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6995
6996 $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6997 }
6998
6999 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
7000 $opts->{format} = $format if $format;
7001
7002 if (defined($bwlimit)) {
7003 $opts->{speed} = $bwlimit * 1024;
7004 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7005 } else {
7006 print "drive mirror is starting for drive-$drive\n";
7007 }
7008
7009 # if a job already runs for this device we get an error, catch it for cleanup
7010 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
7011 if (my $err = $@) {
7012 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7013 warn "$@\n" if $@;
7014 die "mirroring error: $err\n";
7015 }
7016
7017 qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
7018 }
7019
7020 sub qemu_drive_mirror_monitor {
7021 my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
7022
7023 eval {
7024 my $err_complete = 0;
7025
7026 while (1) {
7027 die "storage migration timed out\n" if $err_complete > 300;
7028
7029 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7030
7031 my $running_mirror_jobs = {};
7032 foreach my $stat (@$stats) {
7033 next if $stat->{type} ne 'mirror';
7034 $running_mirror_jobs->{$stat->{device}} = $stat;
7035 }
7036
7037 my $readycounter = 0;
7038
7039 foreach my $job (keys %$jobs) {
7040
7041 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
7042 print "$job : finished\n";
7043 delete $jobs->{$job};
7044 next;
7045 }
7046
7047 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
7048
7049 my $busy = $running_mirror_jobs->{$job}->{busy};
7050 my $ready = $running_mirror_jobs->{$job}->{ready};
7051 if (my $total = $running_mirror_jobs->{$job}->{len}) {
7052 my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
7053 my $remaining = $total - $transferred;
7054 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7055
7056 print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
7057 }
7058
7059 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
7060 }
7061
7062 last if scalar(keys %$jobs) == 0;
7063
7064 if ($readycounter == scalar(keys %$jobs)) {
7065 print "all mirroring jobs are ready \n";
7066 last if $skipcomplete; #do the complete later
7067
7068 if ($vmiddst && $vmiddst != $vmid) {
7069 my $agent_running = $qga && qga_check_running($vmid);
7070 if ($agent_running) {
7071 print "freeze filesystem\n";
7072 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
7073 } else {
7074 print "suspend vm\n";
7075 eval { PVE::QemuServer::vm_suspend($vmid, 1); };
7076 }
7077
7078 # if we clone a disk for a new target vm, we don't switch the disk
7079 PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
7080
7081 if ($agent_running) {
7082 print "unfreeze filesystem\n";
7083 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
7084 } else {
7085 print "resume vm\n";
7086 eval { PVE::QemuServer::vm_resume($vmid, 1, 1); };
7087 }
7088
7089 last;
7090 } else {
7091
7092 foreach my $job (keys %$jobs) {
7093 # try to switch the disk if source and destination are on the same guest
7094 print "$job: Completing block job...\n";
7095
7096 eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
7097 if ($@ =~ m/cannot be completed/) {
7098 print "$job: Block job cannot be completed, try again.\n";
7099 $err_complete++;
7100 }else {
7101 print "$job: Completed successfully.\n";
7102 $jobs->{$job}->{complete} = 1;
7103 }
7104 }
7105 }
7106 }
7107 sleep 1;
7108 }
7109 };
7110 my $err = $@;
7111
7112 if ($err) {
7113 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7114 die "mirroring error: $err";
7115 }
7116
7117 }
7118
7119 sub qemu_blockjobs_cancel {
7120 my ($vmid, $jobs) = @_;
7121
7122 foreach my $job (keys %$jobs) {
7123 print "$job: Cancelling block job\n";
7124 eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
7125 $jobs->{$job}->{cancel} = 1;
7126 }
7127
7128 while (1) {
7129 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7130
7131 my $running_jobs = {};
7132 foreach my $stat (@$stats) {
7133 $running_jobs->{$stat->{device}} = $stat;
7134 }
7135
7136 foreach my $job (keys %$jobs) {
7137
7138 if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
7139 print "$job: Done.\n";
7140 delete $jobs->{$job};
7141 }
7142 }
7143
7144 last if scalar(keys %$jobs) == 0;
7145
7146 sleep 1;
7147 }
7148 }
7149
7150 sub clone_disk {
7151 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
7152 $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
7153
7154 my $newvolid;
7155
7156 if (!$full) {
7157 print "create linked clone of drive $drivename ($drive->{file})\n";
7158 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
7159 push @$newvollist, $newvolid;
7160 } else {
7161
7162 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
7163 $storeid = $storage if $storage;
7164
7165 my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
7166 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
7167
7168 print "create full clone of drive $drivename ($drive->{file})\n";
7169 my $name = undef;
7170 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
7171 push @$newvollist, $newvolid;
7172
7173 PVE::Storage::activate_volumes($storecfg, [$newvolid]);
7174
7175 my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
7176 if (!$running || $snapname) {
7177 # TODO: handle bwlimits
7178 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
7179 } else {
7180
7181 my $kvmver = get_running_qemu_version ($vmid);
7182 if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
7183 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
7184 if $drive->{iothread};
7185 }
7186
7187 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga, $bwlimit);
7188 }
7189 }
7190
7191 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
7192
7193 my $disk = $drive;
7194 $disk->{format} = undef;
7195 $disk->{file} = $newvolid;
7196 $disk->{size} = $size;
7197
7198 return $disk;
7199 }
7200
7201 # this only works if VM is running
7202 sub get_current_qemu_machine {
7203 my ($vmid) = @_;
7204
7205 my $cmd = { execute => 'query-machines', arguments => {} };
7206 my $res = vm_qmp_command($vmid, $cmd);
7207
7208 my ($current, $default);
7209 foreach my $e (@$res) {
7210 $default = $e->{name} if $e->{'is-default'};
7211 $current = $e->{name} if $e->{'is-current'};
7212 }
7213
7214 # fallback to the default machine if current is not supported by qemu
7215 return $current || $default || 'pc';
7216 }
7217
7218 sub get_running_qemu_version {
7219 my ($vmid) = @_;
7220 my $cmd = { execute => 'query-version', arguments => {} };
7221 my $res = vm_qmp_command($vmid, $cmd);
7222 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
7223 }
7224
7225 sub qemu_machine_feature_enabled {
7226 my ($machine, $kvmver, $version_major, $version_minor) = @_;
7227
7228 my $current_major;
7229 my $current_minor;
7230
7231 if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
7232
7233 $current_major = $3;
7234 $current_minor = $4;
7235
7236 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
7237
7238 $current_major = $1;
7239 $current_minor = $2;
7240 }
7241
7242 return 1 if $current_major > $version_major ||
7243 ($current_major == $version_major &&
7244 $current_minor >= $version_minor);
7245 }
7246
7247 sub qemu_machine_pxe {
7248 my ($vmid, $conf) = @_;
7249
7250 my $machine = PVE::QemuServer::get_current_qemu_machine($vmid);
7251
7252 if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
7253 $machine .= '.pxe';
7254 }
7255
7256 return $machine;
7257 }
7258
7259 sub qemu_use_old_bios_files {
7260 my ($machine_type) = @_;
7261
7262 return if !$machine_type;
7263
7264 my $use_old_bios_files = undef;
7265
7266 if ($machine_type =~ m/^(\S+)\.pxe$/) {
7267 $machine_type = $1;
7268 $use_old_bios_files = 1;
7269 } else {
7270 my $kvmver = kvm_user_version();
7271 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
7272 # load new efi bios files on migration. So this hack is required to allow
7273 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
7274 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
7275 $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
7276 }
7277
7278 return ($use_old_bios_files, $machine_type);
7279 }
7280
7281 sub create_efidisk($$$$$) {
7282 my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7283
7284 my (undef, $ovmf_vars) = get_ovmf_files($arch);
7285 die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7286
7287 my $vars_size_b = -s $ovmf_vars;
7288 my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7289 my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7290 PVE::Storage::activate_volumes($storecfg, [$volid]);
7291
7292 qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7293
7294 return ($volid, $vars_size);
7295 }
7296
7297 sub vm_iothreads_list {
7298 my ($vmid) = @_;
7299
7300 my $res = vm_mon_cmd($vmid, 'query-iothreads');
7301
7302 my $iothreads = {};
7303 foreach my $iothread (@$res) {
7304 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7305 }
7306
7307 return $iothreads;
7308 }
7309
7310 sub scsihw_infos {
7311 my ($conf, $drive) = @_;
7312
7313 my $maxdev = 0;
7314
7315 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7316 $maxdev = 7;
7317 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7318 $maxdev = 1;
7319 } else {
7320 $maxdev = 256;
7321 }
7322
7323 my $controller = int($drive->{index} / $maxdev);
7324 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7325
7326 return ($maxdev, $controller, $controller_prefix);
7327 }
7328
7329 sub add_hyperv_enlightenments {
7330 my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough, $hv_vendor_id) = @_;
7331
7332 return if $winversion < 6;
7333 return if $bios && $bios eq 'ovmf' && $winversion < 8;
7334
7335 if ($gpu_passthrough || defined($hv_vendor_id)) {
7336 $hv_vendor_id //= 'proxmox';
7337 push @$cpuFlags , "hv_vendor_id=$hv_vendor_id";
7338 }
7339
7340 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
7341 push @$cpuFlags , 'hv_spinlocks=0x1fff';
7342 push @$cpuFlags , 'hv_vapic';
7343 push @$cpuFlags , 'hv_time';
7344 } else {
7345 push @$cpuFlags , 'hv_spinlocks=0xffff';
7346 }
7347
7348 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
7349 push @$cpuFlags , 'hv_reset';
7350 push @$cpuFlags , 'hv_vpindex';
7351 push @$cpuFlags , 'hv_runtime';
7352 }
7353
7354 if ($winversion >= 7) {
7355 push @$cpuFlags , 'hv_relaxed';
7356
7357 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
7358 push @$cpuFlags , 'hv_synic';
7359 push @$cpuFlags , 'hv_stimer';
7360 }
7361
7362 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 3, 1)) {
7363 push @$cpuFlags , 'hv_ipi';
7364 }
7365 }
7366 }
7367
7368 sub windows_version {
7369 my ($ostype) = @_;
7370
7371 return 0 if !$ostype;
7372
7373 my $winversion = 0;
7374
7375 if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7376 $winversion = 5;
7377 } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7378 $winversion = 6;
7379 } elsif ($ostype =~ m/^win(\d+)$/) {
7380 $winversion = $1;
7381 }
7382
7383 return $winversion;
7384 }
7385
7386 sub resolve_dst_disk_format {
7387 my ($storecfg, $storeid, $src_volname, $format) = @_;
7388 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7389
7390 if (!$format) {
7391 # if no target format is specified, use the source disk format as hint
7392 if ($src_volname) {
7393 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7394 $format = qemu_img_format($scfg, $src_volname);
7395 } else {
7396 return $defFormat;
7397 }
7398 }
7399
7400 # test if requested format is supported - else use default
7401 my $supported = grep { $_ eq $format } @$validFormats;
7402 $format = $defFormat if !$supported;
7403 return $format;
7404 }
7405
7406 sub resolve_first_disk {
7407 my $conf = shift;
7408 my @disks = PVE::QemuServer::valid_drive_names();
7409 my $firstdisk;
7410 foreach my $ds (reverse @disks) {
7411 next if !$conf->{$ds};
7412 my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
7413 next if PVE::QemuServer::drive_is_cdrom($disk);
7414 $firstdisk = $ds;
7415 }
7416 return $firstdisk;
7417 }
7418
7419 sub generate_uuid {
7420 my ($uuid, $uuid_str);
7421 UUID::generate($uuid);
7422 UUID::unparse($uuid, $uuid_str);
7423 return $uuid_str;
7424 }
7425
7426 sub generate_smbios1_uuid {
7427 return "uuid=".generate_uuid();
7428 }
7429
7430 sub nbd_stop {
7431 my ($vmid) = @_;
7432
7433 vm_mon_cmd($vmid, 'nbd-server-stop');
7434 }
7435
7436 sub create_reboot_request {
7437 my ($vmid) = @_;
7438 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7439 or die "failed to create reboot trigger file: $!\n";
7440 close($fh);
7441 }
7442
7443 sub clear_reboot_request {
7444 my ($vmid) = @_;
7445 my $path = "/run/qemu-server/$vmid.reboot";
7446 my $res = 0;
7447
7448 $res = unlink($path);
7449 die "could not remove reboot request for $vmid: $!"
7450 if !$res && $! != POSIX::ENOENT;
7451
7452 return $res;
7453 }
7454
7455 # bash completion helper
7456
7457 sub complete_backup_archives {
7458 my ($cmdname, $pname, $cvalue) = @_;
7459
7460 my $cfg = PVE::Storage::config();
7461
7462 my $storeid;
7463
7464 if ($cvalue =~ m/^([^:]+):/) {
7465 $storeid = $1;
7466 }
7467
7468 my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7469
7470 my $res = [];
7471 foreach my $id (keys %$data) {
7472 foreach my $item (@{$data->{$id}}) {
7473 next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7474 push @$res, $item->{volid} if defined($item->{volid});
7475 }
7476 }
7477
7478 return $res;
7479 }
7480
7481 my $complete_vmid_full = sub {
7482 my ($running) = @_;
7483
7484 my $idlist = vmstatus();
7485
7486 my $res = [];
7487
7488 foreach my $id (keys %$idlist) {
7489 my $d = $idlist->{$id};
7490 if (defined($running)) {
7491 next if $d->{template};
7492 next if $running && $d->{status} ne 'running';
7493 next if !$running && $d->{status} eq 'running';
7494 }
7495 push @$res, $id;
7496
7497 }
7498 return $res;
7499 };
7500
7501 sub complete_vmid {
7502 return &$complete_vmid_full();
7503 }
7504
7505 sub complete_vmid_stopped {
7506 return &$complete_vmid_full(0);
7507 }
7508
7509 sub complete_vmid_running {
7510 return &$complete_vmid_full(1);
7511 }
7512
7513 sub complete_storage {
7514
7515 my $cfg = PVE::Storage::config();
7516 my $ids = $cfg->{ids};
7517
7518 my $res = [];
7519 foreach my $sid (keys %$ids) {
7520 next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7521 next if !$ids->{$sid}->{content}->{images};
7522 push @$res, $sid;
7523 }
7524
7525 return $res;
7526 }
7527
7528 1;