]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
replace remaining vm_destroy call-sites with destroy_vm
[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 get_host_arch $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, $drive->{file});
1899 }
1900
1901 # for compatibility only, we prefer scsi-hd (#2408, #2355, #2380)
1902 if ($path =~ m/^iscsi\:\/\// &&
1903 !qemu_machine_feature_enabled($machine_type, undef, 4, 1)) {
1904 $devicetype = 'generic';
1905 }
1906 }
1907
1908 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1909 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1910 } else {
1911 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1912 }
1913
1914 if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1915 $device .= ",rotation_rate=1";
1916 }
1917 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1918
1919 } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1920 my $maxdev = ($drive->{interface} eq 'sata') ? $MAX_SATA_DISKS : 2;
1921 my $controller = int($drive->{index} / $maxdev);
1922 my $unit = $drive->{index} % $maxdev;
1923 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1924
1925 $device = "ide-$devicetype";
1926 if ($drive->{interface} eq 'ide') {
1927 $device .= ",bus=ide.$controller,unit=$unit";
1928 } else {
1929 $device .= ",bus=ahci$controller.$unit";
1930 }
1931 $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1932
1933 if ($devicetype eq 'hd') {
1934 if (my $model = $drive->{model}) {
1935 $model = URI::Escape::uri_unescape($model);
1936 $device .= ",model=$model";
1937 }
1938 if ($drive->{ssd}) {
1939 $device .= ",rotation_rate=1";
1940 }
1941 }
1942 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1943 } elsif ($drive->{interface} eq 'usb') {
1944 die "implement me";
1945 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1946 } else {
1947 die "unsupported interface type";
1948 }
1949
1950 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1951
1952 if (my $serial = $drive->{serial}) {
1953 $serial = URI::Escape::uri_unescape($serial);
1954 $device .= ",serial=$serial";
1955 }
1956
1957
1958 return $device;
1959 }
1960
1961 sub get_initiator_name {
1962 my $initiator;
1963
1964 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1965 while (defined(my $line = <$fh>)) {
1966 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1967 $initiator = $1;
1968 last;
1969 }
1970 $fh->close();
1971
1972 return $initiator;
1973 }
1974
1975 sub print_drive_full {
1976 my ($storecfg, $vmid, $drive) = @_;
1977
1978 my $path;
1979 my $volid = $drive->{file};
1980 my $format;
1981
1982 if (drive_is_cdrom($drive)) {
1983 $path = get_iso_path($storecfg, $vmid, $volid);
1984 } else {
1985 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1986 if ($storeid) {
1987 $path = PVE::Storage::path($storecfg, $volid);
1988 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1989 $format = qemu_img_format($scfg, $volname);
1990 } else {
1991 $path = $volid;
1992 $format = "raw";
1993 }
1994 }
1995
1996 my $opts = '';
1997 my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1998 foreach my $o (@qemu_drive_options) {
1999 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
2000 }
2001
2002 # snapshot only accepts on|off
2003 if (defined($drive->{snapshot})) {
2004 my $v = $drive->{snapshot} ? 'on' : 'off';
2005 $opts .= ",snapshot=$v";
2006 }
2007
2008 foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
2009 my ($dir, $qmpname) = @$type;
2010 if (my $v = $drive->{"mbps$dir"}) {
2011 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
2012 }
2013 if (my $v = $drive->{"mbps${dir}_max"}) {
2014 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
2015 }
2016 if (my $v = $drive->{"bps${dir}_max_length"}) {
2017 $opts .= ",throttling.bps$qmpname-max-length=$v";
2018 }
2019 if (my $v = $drive->{"iops${dir}"}) {
2020 $opts .= ",throttling.iops$qmpname=$v";
2021 }
2022 if (my $v = $drive->{"iops${dir}_max"}) {
2023 $opts .= ",throttling.iops$qmpname-max=$v";
2024 }
2025 if (my $v = $drive->{"iops${dir}_max_length"}) {
2026 $opts .= ",throttling.iops$qmpname-max-length=$v";
2027 }
2028 }
2029
2030 $opts .= ",format=$format" if $format && !$drive->{format};
2031
2032 my $cache_direct = 0;
2033
2034 if (my $cache = $drive->{cache}) {
2035 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
2036 } elsif (!drive_is_cdrom($drive)) {
2037 $opts .= ",cache=none";
2038 $cache_direct = 1;
2039 }
2040
2041 # aio native works only with O_DIRECT
2042 if (!$drive->{aio}) {
2043 if($cache_direct) {
2044 $opts .= ",aio=native";
2045 } else {
2046 $opts .= ",aio=threads";
2047 }
2048 }
2049
2050 if (!drive_is_cdrom($drive)) {
2051 my $detectzeroes;
2052 if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
2053 $detectzeroes = 'off';
2054 } elsif ($drive->{discard}) {
2055 $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
2056 } else {
2057 # This used to be our default with discard not being specified:
2058 $detectzeroes = 'on';
2059 }
2060 $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
2061 }
2062
2063 my $pathinfo = $path ? "file=$path," : '';
2064
2065 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
2066 }
2067
2068 sub print_netdevice_full {
2069 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
2070
2071 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2072
2073 my $device = $net->{model};
2074 if ($net->{model} eq 'virtio') {
2075 $device = 'virtio-net-pci';
2076 };
2077
2078 my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
2079 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
2080 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
2081 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
2082 my $vectors = $net->{queues} * 2 + 2;
2083 $tmpstr .= ",vectors=$vectors,mq=on";
2084 }
2085 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
2086
2087 if ($use_old_bios_files) {
2088 my $romfile;
2089 if ($device eq 'virtio-net-pci') {
2090 $romfile = 'pxe-virtio.rom';
2091 } elsif ($device eq 'e1000') {
2092 $romfile = 'pxe-e1000.rom';
2093 } elsif ($device eq 'ne2k') {
2094 $romfile = 'pxe-ne2k_pci.rom';
2095 } elsif ($device eq 'pcnet') {
2096 $romfile = 'pxe-pcnet.rom';
2097 } elsif ($device eq 'rtl8139') {
2098 $romfile = 'pxe-rtl8139.rom';
2099 }
2100 $tmpstr .= ",romfile=$romfile" if $romfile;
2101 }
2102
2103 return $tmpstr;
2104 }
2105
2106 sub print_netdev_full {
2107 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
2108
2109 my $i = '';
2110 if ($netid =~ m/^net(\d+)$/) {
2111 $i = int($1);
2112 }
2113
2114 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
2115
2116 my $ifname = "tap${vmid}i$i";
2117
2118 # kvm uses TUNSETIFF ioctl, and that limits ifname length
2119 die "interface name '$ifname' is too long (max 15 character)\n"
2120 if length($ifname) >= 16;
2121
2122 my $vhostparam = '';
2123 if (is_native($arch)) {
2124 $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
2125 }
2126
2127 my $vmname = $conf->{name} || "vm$vmid";
2128
2129 my $netdev = "";
2130 my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
2131
2132 if ($net->{bridge}) {
2133 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
2134 } else {
2135 $netdev = "type=user,id=$netid,hostname=$vmname";
2136 }
2137
2138 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
2139
2140 return $netdev;
2141 }
2142
2143
2144 sub print_cpu_device {
2145 my ($conf, $id) = @_;
2146
2147 my $kvm = $conf->{kvm} // 1;
2148 my $cpu = $kvm ? "kvm64" : "qemu64";
2149 if (my $cputype = $conf->{cpu}) {
2150 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
2151 or die "Cannot parse cpu description: $cputype\n";
2152 $cpu = $cpuconf->{cputype};
2153 }
2154
2155 my $cores = $conf->{cores} || 1;
2156
2157 my $current_core = ($id - 1) % $cores;
2158 my $current_socket = int(($id - 1 - $current_core)/$cores);
2159
2160 return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
2161 }
2162
2163 my $vga_map = {
2164 'cirrus' => 'cirrus-vga',
2165 'std' => 'VGA',
2166 'vmware' => 'vmware-svga',
2167 'virtio' => 'virtio-vga',
2168 };
2169
2170 sub print_vga_device {
2171 my ($conf, $vga, $arch, $machine, $id, $qxlnum, $bridges) = @_;
2172
2173 my $type = $vga_map->{$vga->{type}};
2174 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
2175 $type = 'virtio-gpu';
2176 }
2177 my $vgamem_mb = $vga->{memory};
2178 if ($qxlnum) {
2179 $type = $id ? 'qxl' : 'qxl-vga';
2180 }
2181 die "no devicetype for $vga->{type}\n" if !$type;
2182
2183 my $memory = "";
2184 if ($vgamem_mb) {
2185 if ($vga->{type} eq 'virtio') {
2186 my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2187 $memory = ",max_hostmem=$bytes";
2188 } elsif ($qxlnum) {
2189 # from https://www.spice-space.org/multiple-monitors.html
2190 $memory = ",vgamem_mb=$vga->{memory}";
2191 my $ram = $vgamem_mb * 4;
2192 my $vram = $vgamem_mb * 2;
2193 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2194 } else {
2195 $memory = ",vgamem_mb=$vga->{memory}";
2196 }
2197 } elsif ($qxlnum && $id) {
2198 $memory = ",ram_size=67108864,vram_size=33554432";
2199 }
2200
2201 my $q35 = machine_type_is_q35($conf);
2202 my $vgaid = "vga" . ($id // '');
2203 my $pciaddr;
2204
2205 if ($q35 && $vgaid eq 'vga') {
2206 # the first display uses pcie.0 bus on q35 machines
2207 $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
2208 } else {
2209 $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
2210 }
2211
2212 return "$type,id=${vgaid}${memory}${pciaddr}";
2213 }
2214
2215 sub drive_is_cloudinit {
2216 my ($drive) = @_;
2217 return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2218 }
2219
2220 sub drive_is_cdrom {
2221 my ($drive, $exclude_cloudinit) = @_;
2222
2223 return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2224
2225 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2226
2227 }
2228
2229 sub parse_number_sets {
2230 my ($set) = @_;
2231 my $res = [];
2232 foreach my $part (split(/;/, $set)) {
2233 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2234 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2235 push @$res, [ $1, $2 ];
2236 } else {
2237 die "invalid range: $part\n";
2238 }
2239 }
2240 return $res;
2241 }
2242
2243 sub parse_numa {
2244 my ($data) = @_;
2245
2246 my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2247 $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2248 $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2249 return $res;
2250 }
2251
2252 sub parse_hostpci {
2253 my ($value) = @_;
2254
2255 return undef if !$value;
2256
2257 my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2258
2259 my @idlist = split(/;/, $res->{host});
2260 delete $res->{host};
2261 foreach my $id (@idlist) {
2262 if ($id =~ m/\./) { # full id 00:00.1
2263 push @{$res->{pciid}}, {
2264 id => $id,
2265 };
2266 } else { # partial id 00:00
2267 $res->{pciid} = PVE::SysFSTools::lspci($id);
2268 }
2269 }
2270 return $res;
2271 }
2272
2273 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2274 sub parse_net {
2275 my ($data) = @_;
2276
2277 my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2278 if ($@) {
2279 warn $@;
2280 return undef;
2281 }
2282 if (!defined($res->{macaddr})) {
2283 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2284 $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2285 }
2286 return $res;
2287 }
2288
2289 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2290 sub parse_ipconfig {
2291 my ($data) = @_;
2292
2293 my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2294 if ($@) {
2295 warn $@;
2296 return undef;
2297 }
2298
2299 if ($res->{gw} && !$res->{ip}) {
2300 warn 'gateway specified without specifying an IP address';
2301 return undef;
2302 }
2303 if ($res->{gw6} && !$res->{ip6}) {
2304 warn 'IPv6 gateway specified without specifying an IPv6 address';
2305 return undef;
2306 }
2307 if ($res->{gw} && $res->{ip} eq 'dhcp') {
2308 warn 'gateway specified together with DHCP';
2309 return undef;
2310 }
2311 if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2312 # gw6 + auto/dhcp
2313 warn "IPv6 gateway specified together with $res->{ip6} address";
2314 return undef;
2315 }
2316
2317 if (!$res->{ip} && !$res->{ip6}) {
2318 return { ip => 'dhcp', ip6 => 'dhcp' };
2319 }
2320
2321 return $res;
2322 }
2323
2324 sub print_net {
2325 my $net = shift;
2326
2327 return PVE::JSONSchema::print_property_string($net, $net_fmt);
2328 }
2329
2330 sub add_random_macs {
2331 my ($settings) = @_;
2332
2333 foreach my $opt (keys %$settings) {
2334 next if $opt !~ m/^net(\d+)$/;
2335 my $net = parse_net($settings->{$opt});
2336 next if !$net;
2337 $settings->{$opt} = print_net($net);
2338 }
2339 }
2340
2341 sub vm_is_volid_owner {
2342 my ($storecfg, $vmid, $volid) = @_;
2343
2344 if ($volid !~ m|^/|) {
2345 my ($path, $owner);
2346 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2347 if ($owner && ($owner == $vmid)) {
2348 return 1;
2349 }
2350 }
2351
2352 return undef;
2353 }
2354
2355 sub vmconfig_register_unused_drive {
2356 my ($storecfg, $vmid, $conf, $drive) = @_;
2357
2358 if (drive_is_cloudinit($drive)) {
2359 eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2360 warn $@ if $@;
2361 } elsif (!drive_is_cdrom($drive)) {
2362 my $volid = $drive->{file};
2363 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2364 PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2365 }
2366 }
2367 }
2368
2369 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2370 my $smbios1_fmt = {
2371 uuid => {
2372 type => 'string',
2373 pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2374 format_description => 'UUID',
2375 description => "Set SMBIOS1 UUID.",
2376 optional => 1,
2377 },
2378 version => {
2379 type => 'string',
2380 pattern => '[A-Za-z0-9+\/]+={0,2}',
2381 format_description => 'Base64 encoded string',
2382 description => "Set SMBIOS1 version.",
2383 optional => 1,
2384 },
2385 serial => {
2386 type => 'string',
2387 pattern => '[A-Za-z0-9+\/]+={0,2}',
2388 format_description => 'Base64 encoded string',
2389 description => "Set SMBIOS1 serial number.",
2390 optional => 1,
2391 },
2392 manufacturer => {
2393 type => 'string',
2394 pattern => '[A-Za-z0-9+\/]+={0,2}',
2395 format_description => 'Base64 encoded string',
2396 description => "Set SMBIOS1 manufacturer.",
2397 optional => 1,
2398 },
2399 product => {
2400 type => 'string',
2401 pattern => '[A-Za-z0-9+\/]+={0,2}',
2402 format_description => 'Base64 encoded string',
2403 description => "Set SMBIOS1 product ID.",
2404 optional => 1,
2405 },
2406 sku => {
2407 type => 'string',
2408 pattern => '[A-Za-z0-9+\/]+={0,2}',
2409 format_description => 'Base64 encoded string',
2410 description => "Set SMBIOS1 SKU string.",
2411 optional => 1,
2412 },
2413 family => {
2414 type => 'string',
2415 pattern => '[A-Za-z0-9+\/]+={0,2}',
2416 format_description => 'Base64 encoded string',
2417 description => "Set SMBIOS1 family string.",
2418 optional => 1,
2419 },
2420 base64 => {
2421 type => 'boolean',
2422 description => 'Flag to indicate that the SMBIOS values are base64 encoded',
2423 optional => 1,
2424 },
2425 };
2426
2427 sub parse_smbios1 {
2428 my ($data) = @_;
2429
2430 my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2431 warn $@ if $@;
2432 return $res;
2433 }
2434
2435 sub print_smbios1 {
2436 my ($smbios1) = @_;
2437 return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2438 }
2439
2440 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2441
2442 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2443 sub verify_bootdisk {
2444 my ($value, $noerr) = @_;
2445
2446 return $value if is_valid_drivename($value);
2447
2448 return undef if $noerr;
2449
2450 die "invalid boot disk '$value'\n";
2451 }
2452
2453 sub parse_watchdog {
2454 my ($value) = @_;
2455
2456 return undef if !$value;
2457
2458 my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2459 warn $@ if $@;
2460 return $res;
2461 }
2462
2463 sub parse_guest_agent {
2464 my ($value) = @_;
2465
2466 return {} if !defined($value->{agent});
2467
2468 my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2469 warn $@ if $@;
2470
2471 # if the agent is disabled ignore the other potentially set properties
2472 return {} if !$res->{enabled};
2473 return $res;
2474 }
2475
2476 sub parse_vga {
2477 my ($value) = @_;
2478
2479 return {} if !$value;
2480 my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2481 warn $@ if $@;
2482 return $res;
2483 }
2484
2485 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2486 sub verify_usb_device {
2487 my ($value, $noerr) = @_;
2488
2489 return $value if parse_usb_device($value);
2490
2491 return undef if $noerr;
2492
2493 die "unable to parse usb device\n";
2494 }
2495
2496 # add JSON properties for create and set function
2497 sub json_config_properties {
2498 my $prop = shift;
2499
2500 foreach my $opt (keys %$confdesc) {
2501 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2502 $prop->{$opt} = $confdesc->{$opt};
2503 }
2504
2505 return $prop;
2506 }
2507
2508 # return copy of $confdesc_cloudinit to generate documentation
2509 sub cloudinit_config_properties {
2510
2511 return dclone($confdesc_cloudinit);
2512 }
2513
2514 sub check_type {
2515 my ($key, $value) = @_;
2516
2517 die "unknown setting '$key'\n" if !$confdesc->{$key};
2518
2519 my $type = $confdesc->{$key}->{type};
2520
2521 if (!defined($value)) {
2522 die "got undefined value\n";
2523 }
2524
2525 if ($value =~ m/[\n\r]/) {
2526 die "property contains a line feed\n";
2527 }
2528
2529 if ($type eq 'boolean') {
2530 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2531 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2532 die "type check ('boolean') failed - got '$value'\n";
2533 } elsif ($type eq 'integer') {
2534 return int($1) if $value =~ m/^(\d+)$/;
2535 die "type check ('integer') failed - got '$value'\n";
2536 } elsif ($type eq 'number') {
2537 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2538 die "type check ('number') failed - got '$value'\n";
2539 } elsif ($type eq 'string') {
2540 if (my $fmt = $confdesc->{$key}->{format}) {
2541 PVE::JSONSchema::check_format($fmt, $value);
2542 return $value;
2543 }
2544 $value =~ s/^\"(.*)\"$/$1/;
2545 return $value;
2546 } else {
2547 die "internal error"
2548 }
2549 }
2550
2551 sub destroy_vm {
2552 my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2553
2554 my $conf = PVE::QemuConfig->load_config($vmid);
2555
2556 PVE::QemuConfig->check_lock($conf) if !$skiplock;
2557
2558 if ($conf->{template}) {
2559 # check if any base image is still used by a linked clone
2560 foreach_drive($conf, sub {
2561 my ($ds, $drive) = @_;
2562
2563 return if drive_is_cdrom($drive);
2564
2565 my $volid = $drive->{file};
2566
2567 return if !$volid || $volid =~ m|^/|;
2568
2569 die "base volume '$volid' is still in use by linked cloned\n"
2570 if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2571
2572 });
2573 }
2574
2575 # only remove disks owned by this VM
2576 foreach_drive($conf, sub {
2577 my ($ds, $drive) = @_;
2578
2579 return if drive_is_cdrom($drive, 1);
2580
2581 my $volid = $drive->{file};
2582
2583 return if !$volid || $volid =~ m|^/|;
2584
2585 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2586 return if !$path || !$owner || ($owner != $vmid);
2587
2588 eval {
2589 PVE::Storage::vdisk_free($storecfg, $volid);
2590 };
2591 warn "Could not remove disk '$volid', check manually: $@" if $@;
2592
2593 });
2594
2595 # also remove unused disk
2596 eval {
2597 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2598
2599 eval {
2600 PVE::Storage::foreach_volid($dl, sub {
2601 my ($volid, $sid, $volname, $d) = @_;
2602 PVE::Storage::vdisk_free($storecfg, $volid);
2603 });
2604 };
2605 warn $@ if $@;
2606
2607 };
2608 warn $@ if $@;
2609
2610 if ($keep_empty_config) {
2611 PVE::QemuConfig->write_config($vmid, { memory => 128 });
2612 } else {
2613 PVE::QemuConfig->destroy_config($vmid);
2614 }
2615 }
2616
2617 sub parse_vm_config {
2618 my ($filename, $raw) = @_;
2619
2620 return undef if !defined($raw);
2621
2622 my $res = {
2623 digest => Digest::SHA::sha1_hex($raw),
2624 snapshots => {},
2625 pending => {},
2626 };
2627
2628 $filename =~ m|/qemu-server/(\d+)\.conf$|
2629 || die "got strange filename '$filename'";
2630
2631 my $vmid = $1;
2632
2633 my $conf = $res;
2634 my $descr;
2635 my $section = '';
2636
2637 my @lines = split(/\n/, $raw);
2638 foreach my $line (@lines) {
2639 next if $line =~ m/^\s*$/;
2640
2641 if ($line =~ m/^\[PENDING\]\s*$/i) {
2642 $section = 'pending';
2643 if (defined($descr)) {
2644 $descr =~ s/\s+$//;
2645 $conf->{description} = $descr;
2646 }
2647 $descr = undef;
2648 $conf = $res->{$section} = {};
2649 next;
2650
2651 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2652 $section = $1;
2653 if (defined($descr)) {
2654 $descr =~ s/\s+$//;
2655 $conf->{description} = $descr;
2656 }
2657 $descr = undef;
2658 $conf = $res->{snapshots}->{$section} = {};
2659 next;
2660 }
2661
2662 if ($line =~ m/^\#(.*)\s*$/) {
2663 $descr = '' if !defined($descr);
2664 $descr .= PVE::Tools::decode_text($1) . "\n";
2665 next;
2666 }
2667
2668 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2669 $descr = '' if !defined($descr);
2670 $descr .= PVE::Tools::decode_text($2);
2671 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2672 $conf->{snapstate} = $1;
2673 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2674 my $key = $1;
2675 my $value = $2;
2676 $conf->{$key} = $value;
2677 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2678 my $value = $1;
2679 if ($section eq 'pending') {
2680 $conf->{delete} = $value; # we parse this later
2681 } else {
2682 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2683 }
2684 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2685 my $key = $1;
2686 my $value = $2;
2687 eval { $value = check_type($key, $value); };
2688 if ($@) {
2689 warn "vm $vmid - unable to parse value of '$key' - $@";
2690 } else {
2691 $key = 'ide2' if $key eq 'cdrom';
2692 my $fmt = $confdesc->{$key}->{format};
2693 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2694 my $v = parse_drive($key, $value);
2695 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2696 $v->{file} = $volid;
2697 $value = print_drive($vmid, $v);
2698 } else {
2699 warn "vm $vmid - unable to parse value of '$key'\n";
2700 next;
2701 }
2702 }
2703
2704 $conf->{$key} = $value;
2705 }
2706 }
2707 }
2708
2709 if (defined($descr)) {
2710 $descr =~ s/\s+$//;
2711 $conf->{description} = $descr;
2712 }
2713 delete $res->{snapstate}; # just to be sure
2714
2715 return $res;
2716 }
2717
2718 sub write_vm_config {
2719 my ($filename, $conf) = @_;
2720
2721 delete $conf->{snapstate}; # just to be sure
2722
2723 if ($conf->{cdrom}) {
2724 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2725 $conf->{ide2} = $conf->{cdrom};
2726 delete $conf->{cdrom};
2727 }
2728
2729 # we do not use 'smp' any longer
2730 if ($conf->{sockets}) {
2731 delete $conf->{smp};
2732 } elsif ($conf->{smp}) {
2733 $conf->{sockets} = $conf->{smp};
2734 delete $conf->{cores};
2735 delete $conf->{smp};
2736 }
2737
2738 my $used_volids = {};
2739
2740 my $cleanup_config = sub {
2741 my ($cref, $pending, $snapname) = @_;
2742
2743 foreach my $key (keys %$cref) {
2744 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2745 $key eq 'snapstate' || $key eq 'pending';
2746 my $value = $cref->{$key};
2747 if ($key eq 'delete') {
2748 die "propertry 'delete' is only allowed in [PENDING]\n"
2749 if !$pending;
2750 # fixme: check syntax?
2751 next;
2752 }
2753 eval { $value = check_type($key, $value); };
2754 die "unable to parse value of '$key' - $@" if $@;
2755
2756 $cref->{$key} = $value;
2757
2758 if (!$snapname && is_valid_drivename($key)) {
2759 my $drive = parse_drive($key, $value);
2760 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2761 }
2762 }
2763 };
2764
2765 &$cleanup_config($conf);
2766
2767 &$cleanup_config($conf->{pending}, 1);
2768
2769 foreach my $snapname (keys %{$conf->{snapshots}}) {
2770 die "internal error" if $snapname eq 'pending';
2771 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2772 }
2773
2774 # remove 'unusedX' settings if we re-add a volume
2775 foreach my $key (keys %$conf) {
2776 my $value = $conf->{$key};
2777 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2778 delete $conf->{$key};
2779 }
2780 }
2781
2782 my $generate_raw_config = sub {
2783 my ($conf, $pending) = @_;
2784
2785 my $raw = '';
2786
2787 # add description as comment to top of file
2788 if (defined(my $descr = $conf->{description})) {
2789 if ($descr) {
2790 foreach my $cl (split(/\n/, $descr)) {
2791 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2792 }
2793 } else {
2794 $raw .= "#\n" if $pending;
2795 }
2796 }
2797
2798 foreach my $key (sort keys %$conf) {
2799 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2800 $raw .= "$key: $conf->{$key}\n";
2801 }
2802 return $raw;
2803 };
2804
2805 my $raw = &$generate_raw_config($conf);
2806
2807 if (scalar(keys %{$conf->{pending}})){
2808 $raw .= "\n[PENDING]\n";
2809 $raw .= &$generate_raw_config($conf->{pending}, 1);
2810 }
2811
2812 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2813 $raw .= "\n[$snapname]\n";
2814 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2815 }
2816
2817 return $raw;
2818 }
2819
2820 sub load_defaults {
2821
2822 my $res = {};
2823
2824 # we use static defaults from our JSON schema configuration
2825 foreach my $key (keys %$confdesc) {
2826 if (defined(my $default = $confdesc->{$key}->{default})) {
2827 $res->{$key} = $default;
2828 }
2829 }
2830
2831 return $res;
2832 }
2833
2834 sub config_list {
2835 my $vmlist = PVE::Cluster::get_vmlist();
2836 my $res = {};
2837 return $res if !$vmlist || !$vmlist->{ids};
2838 my $ids = $vmlist->{ids};
2839
2840 foreach my $vmid (keys %$ids) {
2841 my $d = $ids->{$vmid};
2842 next if !$d->{node} || $d->{node} ne $nodename;
2843 next if !$d->{type} || $d->{type} ne 'qemu';
2844 $res->{$vmid}->{exists} = 1;
2845 }
2846 return $res;
2847 }
2848
2849 # test if VM uses local resources (to prevent migration)
2850 sub check_local_resources {
2851 my ($conf, $noerr) = @_;
2852
2853 my @loc_res = ();
2854
2855 push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2856 push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2857
2858 push @loc_res, "ivshmem" if $conf->{ivshmem};
2859
2860 foreach my $k (keys %$conf) {
2861 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2862 # sockets are safe: they will recreated be on the target side post-migrate
2863 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2864 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2865 }
2866
2867 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2868
2869 return \@loc_res;
2870 }
2871
2872 # check if used storages are available on all nodes (use by migrate)
2873 sub check_storage_availability {
2874 my ($storecfg, $conf, $node) = @_;
2875
2876 foreach_drive($conf, sub {
2877 my ($ds, $drive) = @_;
2878
2879 my $volid = $drive->{file};
2880 return if !$volid;
2881
2882 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2883 return if !$sid;
2884
2885 # check if storage is available on both nodes
2886 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2887 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2888 });
2889 }
2890
2891 # list nodes where all VM images are available (used by has_feature API)
2892 sub shared_nodes {
2893 my ($conf, $storecfg) = @_;
2894
2895 my $nodelist = PVE::Cluster::get_nodelist();
2896 my $nodehash = { map { $_ => 1 } @$nodelist };
2897 my $nodename = PVE::INotify::nodename();
2898
2899 foreach_drive($conf, sub {
2900 my ($ds, $drive) = @_;
2901
2902 my $volid = $drive->{file};
2903 return if !$volid;
2904
2905 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2906 if ($storeid) {
2907 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2908 if ($scfg->{disable}) {
2909 $nodehash = {};
2910 } elsif (my $avail = $scfg->{nodes}) {
2911 foreach my $node (keys %$nodehash) {
2912 delete $nodehash->{$node} if !$avail->{$node};
2913 }
2914 } elsif (!$scfg->{shared}) {
2915 foreach my $node (keys %$nodehash) {
2916 delete $nodehash->{$node} if $node ne $nodename
2917 }
2918 }
2919 }
2920 });
2921
2922 return $nodehash
2923 }
2924
2925 sub check_local_storage_availability {
2926 my ($conf, $storecfg) = @_;
2927
2928 my $nodelist = PVE::Cluster::get_nodelist();
2929 my $nodehash = { map { $_ => {} } @$nodelist };
2930
2931 foreach_drive($conf, sub {
2932 my ($ds, $drive) = @_;
2933
2934 my $volid = $drive->{file};
2935 return if !$volid;
2936
2937 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2938 if ($storeid) {
2939 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2940
2941 if ($scfg->{disable}) {
2942 foreach my $node (keys %$nodehash) {
2943 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2944 }
2945 } elsif (my $avail = $scfg->{nodes}) {
2946 foreach my $node (keys %$nodehash) {
2947 if (!$avail->{$node}) {
2948 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
2949 }
2950 }
2951 }
2952 }
2953 });
2954
2955 foreach my $node (values %$nodehash) {
2956 if (my $unavail = $node->{unavailable_storages}) {
2957 $node->{unavailable_storages} = [ sort keys %$unavail ];
2958 }
2959 }
2960
2961 return $nodehash
2962 }
2963
2964 sub check_cmdline {
2965 my ($pidfile, $pid) = @_;
2966
2967 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2968 if (defined($fh)) {
2969 my $line = <$fh>;
2970 $fh->close;
2971 return undef if !$line;
2972 my @param = split(/\0/, $line);
2973
2974 my $cmd = $param[0];
2975 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
2976
2977 for (my $i = 0; $i < scalar (@param); $i++) {
2978 my $p = $param[$i];
2979 next if !$p;
2980 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2981 my $p = $param[$i+1];
2982 return 1 if $p && ($p eq $pidfile);
2983 return undef;
2984 }
2985 }
2986 }
2987 return undef;
2988 }
2989
2990 sub check_running {
2991 my ($vmid, $nocheck, $node) = @_;
2992
2993 my $filename = PVE::QemuConfig->config_file($vmid, $node);
2994
2995 die "unable to find configuration file for VM $vmid - no such machine\n"
2996 if !$nocheck && ! -f $filename;
2997
2998 my $pidfile = pidfile_name($vmid);
2999
3000 if (my $fd = IO::File->new("<$pidfile")) {
3001 my $st = stat($fd);
3002 my $line = <$fd>;
3003 close($fd);
3004
3005 my $mtime = $st->mtime;
3006 if ($mtime > time()) {
3007 warn "file '$filename' modified in future\n";
3008 }
3009
3010 if ($line =~ m/^(\d+)$/) {
3011 my $pid = $1;
3012 if (check_cmdline($pidfile, $pid)) {
3013 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
3014 return $pid;
3015 }
3016 }
3017 }
3018 }
3019
3020 return undef;
3021 }
3022
3023 sub vzlist {
3024
3025 my $vzlist = config_list();
3026
3027 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
3028
3029 while (defined(my $de = $fd->read)) {
3030 next if $de !~ m/^(\d+)\.pid$/;
3031 my $vmid = $1;
3032 next if !defined($vzlist->{$vmid});
3033 if (my $pid = check_running($vmid)) {
3034 $vzlist->{$vmid}->{pid} = $pid;
3035 }
3036 }
3037
3038 return $vzlist;
3039 }
3040
3041 sub disksize {
3042 my ($storecfg, $conf) = @_;
3043
3044 my $bootdisk = $conf->{bootdisk};
3045 return undef if !$bootdisk;
3046 return undef if !is_valid_drivename($bootdisk);
3047
3048 return undef if !$conf->{$bootdisk};
3049
3050 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
3051 return undef if !defined($drive);
3052
3053 return undef if drive_is_cdrom($drive);
3054
3055 my $volid = $drive->{file};
3056 return undef if !$volid;
3057
3058 return $drive->{size};
3059 }
3060
3061 our $vmstatus_return_properties = {
3062 vmid => get_standard_option('pve-vmid'),
3063 status => {
3064 description => "Qemu process status.",
3065 type => 'string',
3066 enum => ['stopped', 'running'],
3067 },
3068 maxmem => {
3069 description => "Maximum memory in bytes.",
3070 type => 'integer',
3071 optional => 1,
3072 renderer => 'bytes',
3073 },
3074 maxdisk => {
3075 description => "Root disk size in bytes.",
3076 type => 'integer',
3077 optional => 1,
3078 renderer => 'bytes',
3079 },
3080 name => {
3081 description => "VM name.",
3082 type => 'string',
3083 optional => 1,
3084 },
3085 qmpstatus => {
3086 description => "Qemu QMP agent status.",
3087 type => 'string',
3088 optional => 1,
3089 },
3090 pid => {
3091 description => "PID of running qemu process.",
3092 type => 'integer',
3093 optional => 1,
3094 },
3095 uptime => {
3096 description => "Uptime.",
3097 type => 'integer',
3098 optional => 1,
3099 renderer => 'duration',
3100 },
3101 cpus => {
3102 description => "Maximum usable CPUs.",
3103 type => 'number',
3104 optional => 1,
3105 },
3106 lock => {
3107 description => "The current config lock, if any.",
3108 type => 'string',
3109 optional => 1,
3110 }
3111 };
3112
3113 my $last_proc_pid_stat;
3114
3115 # get VM status information
3116 # This must be fast and should not block ($full == false)
3117 # We only query KVM using QMP if $full == true (this can be slow)
3118 sub vmstatus {
3119 my ($opt_vmid, $full) = @_;
3120
3121 my $res = {};
3122
3123 my $storecfg = PVE::Storage::config();
3124
3125 my $list = vzlist();
3126 my $defaults = load_defaults();
3127
3128 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
3129
3130 my $cpucount = $cpuinfo->{cpus} || 1;
3131
3132 foreach my $vmid (keys %$list) {
3133 next if $opt_vmid && ($vmid ne $opt_vmid);
3134
3135 my $conf = PVE::QemuConfig->load_config($vmid);
3136
3137 my $d = { vmid => $vmid };
3138 $d->{pid} = $list->{$vmid}->{pid};
3139
3140 # fixme: better status?
3141 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3142
3143 my $size = disksize($storecfg, $conf);
3144 if (defined($size)) {
3145 $d->{disk} = 0; # no info available
3146 $d->{maxdisk} = $size;
3147 } else {
3148 $d->{disk} = 0;
3149 $d->{maxdisk} = 0;
3150 }
3151
3152 $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3153 * ($conf->{cores} || $defaults->{cores});
3154 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3155 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3156
3157 $d->{name} = $conf->{name} || "VM $vmid";
3158 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3159 : $defaults->{memory}*(1024*1024);
3160
3161 if ($conf->{balloon}) {
3162 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3163 $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3164 : $defaults->{shares};
3165 }
3166
3167 $d->{uptime} = 0;
3168 $d->{cpu} = 0;
3169 $d->{mem} = 0;
3170
3171 $d->{netout} = 0;
3172 $d->{netin} = 0;
3173
3174 $d->{diskread} = 0;
3175 $d->{diskwrite} = 0;
3176
3177 $d->{template} = PVE::QemuConfig->is_template($conf);
3178
3179 $d->{serial} = 1 if conf_has_serial($conf);
3180 $d->{lock} = $conf->{lock} if $conf->{lock};
3181
3182 $res->{$vmid} = $d;
3183 }
3184
3185 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3186 foreach my $dev (keys %$netdev) {
3187 next if $dev !~ m/^tap([1-9]\d*)i/;
3188 my $vmid = $1;
3189 my $d = $res->{$vmid};
3190 next if !$d;
3191
3192 $d->{netout} += $netdev->{$dev}->{receive};
3193 $d->{netin} += $netdev->{$dev}->{transmit};
3194
3195 if ($full) {
3196 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3197 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3198 }
3199
3200 }
3201
3202 my $ctime = gettimeofday;
3203
3204 foreach my $vmid (keys %$list) {
3205
3206 my $d = $res->{$vmid};
3207 my $pid = $d->{pid};
3208 next if !$pid;
3209
3210 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3211 next if !$pstat; # not running
3212
3213 my $used = $pstat->{utime} + $pstat->{stime};
3214
3215 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3216
3217 if ($pstat->{vsize}) {
3218 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3219 }
3220
3221 my $old = $last_proc_pid_stat->{$pid};
3222 if (!$old) {
3223 $last_proc_pid_stat->{$pid} = {
3224 time => $ctime,
3225 used => $used,
3226 cpu => 0,
3227 };
3228 next;
3229 }
3230
3231 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3232
3233 if ($dtime > 1000) {
3234 my $dutime = $used - $old->{used};
3235
3236 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3237 $last_proc_pid_stat->{$pid} = {
3238 time => $ctime,
3239 used => $used,
3240 cpu => $d->{cpu},
3241 };
3242 } else {
3243 $d->{cpu} = $old->{cpu};
3244 }
3245 }
3246
3247 return $res if !$full;
3248
3249 my $qmpclient = PVE::QMPClient->new();
3250
3251 my $ballooncb = sub {
3252 my ($vmid, $resp) = @_;
3253
3254 my $info = $resp->{'return'};
3255 return if !$info->{max_mem};
3256
3257 my $d = $res->{$vmid};
3258
3259 # use memory assigned to VM
3260 $d->{maxmem} = $info->{max_mem};
3261 $d->{balloon} = $info->{actual};
3262
3263 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3264 $d->{mem} = $info->{total_mem} - $info->{free_mem};
3265 $d->{freemem} = $info->{free_mem};
3266 }
3267
3268 $d->{ballooninfo} = $info;
3269 };
3270
3271 my $blockstatscb = sub {
3272 my ($vmid, $resp) = @_;
3273 my $data = $resp->{'return'} || [];
3274 my $totalrdbytes = 0;
3275 my $totalwrbytes = 0;
3276
3277 for my $blockstat (@$data) {
3278 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3279 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3280
3281 $blockstat->{device} =~ s/drive-//;
3282 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3283 }
3284 $res->{$vmid}->{diskread} = $totalrdbytes;
3285 $res->{$vmid}->{diskwrite} = $totalwrbytes;
3286 };
3287
3288 my $statuscb = sub {
3289 my ($vmid, $resp) = @_;
3290
3291 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3292 # this fails if ballon driver is not loaded, so this must be
3293 # the last commnand (following command are aborted if this fails).
3294 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3295
3296 my $status = 'unknown';
3297 if (!defined($status = $resp->{'return'}->{status})) {
3298 warn "unable to get VM status\n";
3299 return;
3300 }
3301
3302 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3303 };
3304
3305 foreach my $vmid (keys %$list) {
3306 next if $opt_vmid && ($vmid ne $opt_vmid);
3307 next if !$res->{$vmid}->{pid}; # not running
3308 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3309 }
3310
3311 $qmpclient->queue_execute(undef, 2);
3312
3313 foreach my $vmid (keys %$list) {
3314 next if $opt_vmid && ($vmid ne $opt_vmid);
3315 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3316 }
3317
3318 return $res;
3319 }
3320
3321 sub foreach_drive {
3322 my ($conf, $func, @param) = @_;
3323
3324 foreach my $ds (valid_drive_names()) {
3325 next if !defined($conf->{$ds});
3326
3327 my $drive = parse_drive($ds, $conf->{$ds});
3328 next if !$drive;
3329
3330 &$func($ds, $drive, @param);
3331 }
3332 }
3333
3334 sub foreach_volid {
3335 my ($conf, $func, @param) = @_;
3336
3337 my $volhash = {};
3338
3339 my $test_volid = sub {
3340 my ($volid, $is_cdrom, $replicate, $shared, $snapname, $size) = @_;
3341
3342 return if !$volid;
3343
3344 $volhash->{$volid}->{cdrom} //= 1;
3345 $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3346
3347 $volhash->{$volid}->{replicate} //= 0;
3348 $volhash->{$volid}->{replicate} = 1 if $replicate;
3349
3350 $volhash->{$volid}->{shared} //= 0;
3351 $volhash->{$volid}->{shared} = 1 if $shared;
3352
3353 $volhash->{$volid}->{referenced_in_config} //= 0;
3354 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3355
3356 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3357 if defined($snapname);
3358 $volhash->{$volid}->{size} = $size if $size;
3359 };
3360
3361 foreach_drive($conf, sub {
3362 my ($ds, $drive) = @_;
3363 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef, $drive->{size});
3364 });
3365
3366 foreach my $snapname (keys %{$conf->{snapshots}}) {
3367 my $snap = $conf->{snapshots}->{$snapname};
3368 $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3369 foreach_drive($snap, sub {
3370 my ($ds, $drive) = @_;
3371 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3372 });
3373 }
3374
3375 foreach my $volid (keys %$volhash) {
3376 &$func($volid, $volhash->{$volid}, @param);
3377 }
3378 }
3379
3380 sub conf_has_serial {
3381 my ($conf) = @_;
3382
3383 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3384 if ($conf->{"serial$i"}) {
3385 return 1;
3386 }
3387 }
3388
3389 return 0;
3390 }
3391
3392 sub conf_has_audio {
3393 my ($conf, $id) = @_;
3394
3395 $id //= 0;
3396 my $audio = $conf->{"audio$id"};
3397 return undef if !defined($audio);
3398
3399 my $audioproperties = PVE::JSONSchema::parse_property_string($audio_fmt, $audio);
3400 my $audiodriver = $audioproperties->{driver} // 'spice';
3401
3402 return {
3403 dev => $audioproperties->{device},
3404 dev_id => "audiodev$id",
3405 backend => $audiodriver,
3406 backend_id => "$audiodriver-backend${id}",
3407 };
3408 }
3409
3410 sub vga_conf_has_spice {
3411 my ($vga) = @_;
3412
3413 my $vgaconf = parse_vga($vga);
3414 my $vgatype = $vgaconf->{type};
3415 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3416
3417 return $1 || 1;
3418 }
3419
3420 sub is_native($) {
3421 my ($arch) = @_;
3422 return get_host_arch() eq $arch;
3423 }
3424
3425 my $default_machines = {
3426 x86_64 => 'pc',
3427 aarch64 => 'virt',
3428 };
3429
3430 sub get_basic_machine_info {
3431 my ($conf, $forcemachine) = @_;
3432
3433 my $arch = $conf->{arch} // get_host_arch();
3434 my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3435 return ($arch, $machine);
3436 }
3437
3438 sub get_ovmf_files($) {
3439 my ($arch) = @_;
3440
3441 my $ovmf = $OVMF->{$arch}
3442 or die "no OVMF images known for architecture '$arch'\n";
3443
3444 return @$ovmf;
3445 }
3446
3447 my $Arch2Qemu = {
3448 aarch64 => '/usr/bin/qemu-system-aarch64',
3449 x86_64 => '/usr/bin/qemu-system-x86_64',
3450 };
3451 sub get_command_for_arch($) {
3452 my ($arch) = @_;
3453 return '/usr/bin/kvm' if is_native($arch);
3454
3455 my $cmd = $Arch2Qemu->{$arch}
3456 or die "don't know how to emulate architecture '$arch'\n";
3457 return $cmd;
3458 }
3459
3460 sub get_cpu_options {
3461 my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3462
3463 my $cpuFlags = [];
3464 my $ostype = $conf->{ostype};
3465
3466 my $cpu = $kvm ? "kvm64" : "qemu64";
3467 if ($arch eq 'aarch64') {
3468 $cpu = 'cortex-a57';
3469 }
3470 my $hv_vendor_id;
3471 if (my $cputype = $conf->{cpu}) {
3472 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3473 or die "Cannot parse cpu description: $cputype\n";
3474 $cpu = $cpuconf->{cputype};
3475 $kvm_off = 1 if $cpuconf->{hidden};
3476 $hv_vendor_id = $cpuconf->{'hv-vendor-id'};
3477
3478 if (defined(my $flags = $cpuconf->{flags})) {
3479 push @$cpuFlags, split(";", $flags);
3480 }
3481 }
3482
3483 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64' && $arch eq 'x86_64';
3484
3485 push @$cpuFlags , '-x2apic'
3486 if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3487
3488 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3489
3490 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3491
3492 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3) && $arch eq 'x86_64') {
3493
3494 push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3495 push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3496 }
3497
3498 add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough, $hv_vendor_id) if $kvm;
3499
3500 push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm && $arch eq 'x86_64';
3501
3502 push @$cpuFlags, 'kvm=off' if $kvm_off;
3503
3504 if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3505 push @$cpuFlags, "vendor=${cpu_vendor}"
3506 if $cpu_vendor ne 'default';
3507 } elsif ($arch ne 'aarch64') {
3508 die "internal error"; # should not happen
3509 }
3510
3511 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3512
3513 return ('-cpu', $cpu);
3514 }
3515
3516 sub config_to_command {
3517 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3518
3519 my $cmd = [];
3520 my $globalFlags = [];
3521 my $machineFlags = [];
3522 my $rtcFlags = [];
3523 my $devices = [];
3524 my $pciaddr = '';
3525 my $bridges = {};
3526 my $vernum = 0; # unknown
3527 my $ostype = $conf->{ostype};
3528 my $winversion = windows_version($ostype);
3529 my $kvm = $conf->{kvm};
3530
3531 my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3532 my $kvm_binary = get_command_for_arch($arch);
3533 my $kvmver = kvm_user_version($kvm_binary);
3534 $kvm //= 1 if is_native($arch);
3535
3536 if ($kvm) {
3537 die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3538 if !defined kvm_version();
3539 }
3540
3541 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3542 $vernum = $1*1000000+$2*1000;
3543 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3544 $vernum = $1*1000000+$2*1000+$3;
3545 }
3546
3547 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3548
3549 my $q35 = machine_type_is_q35($conf);
3550 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3551 my $use_old_bios_files = undef;
3552 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3553
3554 my $cpuunits = defined($conf->{cpuunits}) ?
3555 $conf->{cpuunits} : $defaults->{cpuunits};
3556
3557 push @$cmd, $kvm_binary;
3558
3559 push @$cmd, '-id', $vmid;
3560
3561 my $vmname = $conf->{name} || "vm$vmid";
3562
3563 push @$cmd, '-name', $vmname;
3564
3565 my $use_virtio = 0;
3566
3567 my $qmpsocket = qmp_socket($vmid);
3568 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3569 push @$cmd, '-mon', "chardev=qmp,mode=control";
3570
3571 if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3572 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3573 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3574 }
3575
3576 push @$cmd, '-pidfile' , pidfile_name($vmid);
3577
3578 push @$cmd, '-daemonize';
3579
3580 if ($conf->{smbios1}) {
3581 my $smbios_conf = parse_smbios1($conf->{smbios1});
3582 if ($smbios_conf->{base64}) {
3583 # Do not pass base64 flag to qemu
3584 delete $smbios_conf->{base64};
3585 my $smbios_string = "";
3586 foreach my $key (keys %$smbios_conf) {
3587 my $value;
3588 if ($key eq "uuid") {
3589 $value = $smbios_conf->{uuid}
3590 } else {
3591 $value = decode_base64($smbios_conf->{$key});
3592 }
3593 # qemu accepts any binary data, only commas need escaping by double comma
3594 $value =~ s/,/,,/g;
3595 $smbios_string .= "," . $key . "=" . $value if $value;
3596 }
3597 push @$cmd, '-smbios', "type=1" . $smbios_string;
3598 } else {
3599 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3600 }
3601 }
3602
3603 if ($conf->{vmgenid}) {
3604 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3605 }
3606
3607 my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3608 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3609 die "uefi base image not found\n" if ! -f $ovmf_code;
3610
3611 my $path;
3612 my $format;
3613 if (my $efidisk = $conf->{efidisk0}) {
3614 my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3615 my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3616 $format = $d->{format};
3617 if ($storeid) {
3618 $path = PVE::Storage::path($storecfg, $d->{file});
3619 if (!defined($format)) {
3620 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3621 $format = qemu_img_format($scfg, $volname);
3622 }
3623 } else {
3624 $path = $d->{file};
3625 die "efidisk format must be specified\n"
3626 if !defined($format);
3627 }
3628 } else {
3629 warn "no efidisk configured! Using temporary efivars disk.\n";
3630 $path = "/tmp/$vmid-ovmf.fd";
3631 PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3632 $format = 'raw';
3633 }
3634
3635 push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3636 push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3637 }
3638
3639 # load q35 config
3640 if ($q35) {
3641 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3642 if (qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0)) {
3643 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3644 } else {
3645 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3646 }
3647 }
3648
3649 # add usb controllers
3650 my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3651 push @$devices, @usbcontrollers if @usbcontrollers;
3652 my $vga = parse_vga($conf->{vga});
3653
3654 my $qxlnum = vga_conf_has_spice($conf->{vga});
3655 $vga->{type} = 'qxl' if $qxlnum;
3656
3657 if (!$vga->{type}) {
3658 if ($arch eq 'aarch64') {
3659 $vga->{type} = 'virtio';
3660 } elsif (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3661 $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3662 } else {
3663 $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3664 }
3665 }
3666
3667 # enable absolute mouse coordinates (needed by vnc)
3668 my $tablet;
3669 if (defined($conf->{tablet})) {
3670 $tablet = $conf->{tablet};
3671 } else {
3672 $tablet = $defaults->{tablet};
3673 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3674 $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3675 }
3676
3677 if ($tablet) {
3678 push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3679 my $kbd = print_keyboarddevice_full($conf, $arch);
3680 push @$devices, '-device', $kbd if defined($kbd);
3681 }
3682
3683 my $kvm_off = 0;
3684 my $gpu_passthrough;
3685
3686 # host pci devices
3687 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3688 my $id = "hostpci$i";
3689 my $d = parse_hostpci($conf->{$id});
3690 next if !$d;
3691
3692 if (my $pcie = $d->{pcie}) {
3693 die "q35 machine model is not enabled" if !$q35;
3694 # win7 wants to have the pcie devices directly on the pcie bus
3695 # instead of in the root port
3696 if ($winversion == 7) {
3697 $pciaddr = print_pcie_addr("${id}bus0");
3698 } else {
3699 # add more root ports if needed, 4 are present by default
3700 # by pve-q35 cfgs, rest added here on demand.
3701 if ($i > 3) {
3702 push @$devices, '-device', print_pcie_root_port($i);
3703 }
3704 $pciaddr = print_pcie_addr($id);
3705 }
3706 } else {
3707 $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3708 }
3709
3710 my $xvga = '';
3711 if ($d->{'x-vga'}) {
3712 $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3713 $kvm_off = 1;
3714 $vga->{type} = 'none' if !defined($conf->{vga});
3715 $gpu_passthrough = 1;
3716 }
3717
3718 my $pcidevices = $d->{pciid};
3719 my $multifunction = 1 if @$pcidevices > 1;
3720
3721 my $sysfspath;
3722 if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3723 my $pci_id = $pcidevices->[0]->{id};
3724 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3725 $sysfspath = "/sys/bus/pci/devices/0000:$pci_id/$uuid";
3726 } elsif ($d->{mdev}) {
3727 warn "ignoring mediated device '$id' with multifunction device\n";
3728 }
3729
3730 my $j=0;
3731 foreach my $pcidevice (@$pcidevices) {
3732 my $devicestr = "vfio-pci";
3733
3734 if ($sysfspath) {
3735 $devicestr .= ",sysfsdev=$sysfspath";
3736 } else {
3737 $devicestr .= ",host=$pcidevice->{id}";
3738 }
3739
3740 my $mf_addr = $multifunction ? ".$j" : '';
3741 $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3742
3743 if ($j == 0) {
3744 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3745 $devicestr .= "$xvga";
3746 $devicestr .= ",multifunction=on" if $multifunction;
3747 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3748 }
3749
3750 push @$devices, '-device', $devicestr;
3751 $j++;
3752 }
3753 }
3754
3755 # usb devices
3756 my $usb_dev_features = {};
3757 $usb_dev_features->{spice_usb3} = 1 if qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0);
3758
3759 my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3760 push @$devices, @usbdevices if @usbdevices;
3761 # serial devices
3762 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3763 if (my $path = $conf->{"serial$i"}) {
3764 if ($path eq 'socket') {
3765 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3766 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3767 # On aarch64, serial0 is the UART device. Qemu only allows
3768 # connecting UART devices via the '-serial' command line, as
3769 # the device has a fixed slot on the hardware...
3770 if ($arch eq 'aarch64' && $i == 0) {
3771 push @$devices, '-serial', "chardev:serial$i";
3772 } else {
3773 push @$devices, '-device', "isa-serial,chardev=serial$i";
3774 }
3775 } else {
3776 die "no such serial device\n" if ! -c $path;
3777 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3778 push @$devices, '-device', "isa-serial,chardev=serial$i";
3779 }
3780 }
3781 }
3782
3783 # parallel devices
3784 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3785 if (my $path = $conf->{"parallel$i"}) {
3786 die "no such parallel device\n" if ! -c $path;
3787 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3788 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3789 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3790 }
3791 }
3792
3793 if (my $audio = conf_has_audio($conf)) {
3794
3795 my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3796
3797 my $id = $audio->{dev_id};
3798 if ($audio->{dev} eq 'AC97') {
3799 push @$devices, '-device', "AC97,id=${id}${audiopciaddr}";
3800 } elsif ($audio->{dev} =~ /intel\-hda$/) {
3801 push @$devices, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3802 push @$devices, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0";
3803 push @$devices, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1";
3804 } else {
3805 die "unkown audio device '$audio->{dev}', implement me!";
3806 }
3807
3808 push @$devices, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3809 }
3810
3811 my $sockets = 1;
3812 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3813 $sockets = $conf->{sockets} if $conf->{sockets};
3814
3815 my $cores = $conf->{cores} || 1;
3816
3817 my $maxcpus = $sockets * $cores;
3818
3819 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3820
3821 my $allowed_vcpus = $cpuinfo->{cpus};
3822
3823 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3824 if ($allowed_vcpus < $maxcpus);
3825
3826 if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3827
3828 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3829 for (my $i = 2; $i <= $vcpus; $i++) {
3830 my $cpustr = print_cpu_device($conf,$i);
3831 push @$cmd, '-device', $cpustr;
3832 }
3833
3834 } else {
3835
3836 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3837 }
3838 push @$cmd, '-nodefaults';
3839
3840 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3841
3842 my $bootindex_hash = {};
3843 my $i = 1;
3844 foreach my $o (split(//, $bootorder)) {
3845 $bootindex_hash->{$o} = $i*100;
3846 $i++;
3847 }
3848
3849 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3850
3851 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3852
3853 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
3854
3855 if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3856 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, undef, $qxlnum, $bridges);
3857 my $socket = vnc_socket($vmid);
3858 push @$cmd, '-vnc', "unix:$socket,password";
3859 } else {
3860 push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3861 push @$cmd, '-nographic';
3862 }
3863
3864 # time drift fix
3865 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3866
3867 my $useLocaltime = $conf->{localtime};
3868
3869 if ($winversion >= 5) { # windows
3870 $useLocaltime = 1 if !defined($conf->{localtime});
3871
3872 # use time drift fix when acpi is enabled
3873 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3874 $tdf = 1 if !defined($conf->{tdf});
3875 }
3876 }
3877
3878 if ($winversion >= 6) {
3879 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3880 push @$cmd, '-no-hpet';
3881 }
3882
3883 push @$rtcFlags, 'driftfix=slew' if $tdf;
3884
3885 if (!$kvm) {
3886 push @$machineFlags, 'accel=tcg';
3887 }
3888
3889 if ($machine_type) {
3890 push @$machineFlags, "type=${machine_type}";
3891 }
3892
3893 if (($conf->{startdate}) && ($conf->{startdate} ne 'now')) {
3894 push @$rtcFlags, "base=$conf->{startdate}";
3895 } elsif ($useLocaltime) {
3896 push @$rtcFlags, 'base=localtime';
3897 }
3898
3899 push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3900
3901 PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3902
3903 push @$cmd, '-S' if $conf->{freeze};
3904
3905 push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3906
3907 if (parse_guest_agent($conf)->{enabled}) {
3908 my $qgasocket = qmp_socket($vmid, 1);
3909 my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3910 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3911 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3912 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3913 }
3914
3915 my $spice_port;
3916
3917 if ($qxlnum) {
3918 if ($qxlnum > 1) {
3919 if ($winversion){
3920 for(my $i = 1; $i < $qxlnum; $i++){
3921 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, $i, $qxlnum, $bridges);
3922 }
3923 } else {
3924 # assume other OS works like Linux
3925 my ($ram, $vram) = ("134217728", "67108864");
3926 if ($vga->{memory}) {
3927 $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
3928 $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
3929 }
3930 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
3931 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
3932 }
3933 }
3934
3935 my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
3936
3937 my $nodename = PVE::INotify::nodename();
3938 my $pfamily = PVE::Tools::get_host_address_family($nodename);
3939 my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
3940 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
3941
3942 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3943 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3944 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3945
3946 my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
3947 $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
3948
3949 my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
3950 if ($spice_enhancement->{foldersharing}) {
3951 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
3952 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
3953 }
3954
3955 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
3956 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
3957 push @$devices, '-spice', "$spice_opts";
3958 }
3959
3960 # enable balloon by default, unless explicitly disabled
3961 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3962 $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
3963 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3964 }
3965
3966 if ($conf->{watchdog}) {
3967 my $wdopts = parse_watchdog($conf->{watchdog});
3968 $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
3969 my $watchdog = $wdopts->{model} || 'i6300esb';
3970 push @$devices, '-device', "$watchdog$pciaddr";
3971 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3972 }
3973
3974 my $vollist = [];
3975 my $scsicontroller = {};
3976 my $ahcicontroller = {};
3977 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3978
3979 # Add iscsi initiator name if available
3980 if (my $initiator = get_initiator_name()) {
3981 push @$devices, '-iscsi', "initiator-name=$initiator";
3982 }
3983
3984 foreach_drive($conf, sub {
3985 my ($ds, $drive) = @_;
3986
3987 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3988 push @$vollist, $drive->{file};
3989 }
3990
3991 # ignore efidisk here, already added in bios/fw handling code above
3992 return if $drive->{interface} eq 'efidisk';
3993
3994 $use_virtio = 1 if $ds =~ m/^virtio/;
3995
3996 if (drive_is_cdrom ($drive)) {
3997 if ($bootindex_hash->{d}) {
3998 $drive->{bootindex} = $bootindex_hash->{d};
3999 $bootindex_hash->{d} += 1;
4000 }
4001 } else {
4002 if ($bootindex_hash->{c}) {
4003 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
4004 $bootindex_hash->{c} += 1;
4005 }
4006 }
4007
4008 if($drive->{interface} eq 'virtio'){
4009 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
4010 }
4011
4012 if ($drive->{interface} eq 'scsi') {
4013
4014 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
4015
4016 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
4017 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
4018
4019 my $iothread = '';
4020 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
4021 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4022 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4023 } elsif ($drive->{iothread}) {
4024 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
4025 }
4026
4027 my $queues = '';
4028 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
4029 $queues = ",num_queues=$drive->{queues}";
4030 }
4031
4032 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
4033 $scsicontroller->{$controller}=1;
4034 }
4035
4036 if ($drive->{interface} eq 'sata') {
4037 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
4038 $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
4039 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
4040 $ahcicontroller->{$controller}=1;
4041 }
4042
4043 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
4044 push @$devices, '-drive',$drive_cmd;
4045 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4046 });
4047
4048 for (my $i = 0; $i < $MAX_NETS; $i++) {
4049 next if !$conf->{"net$i"};
4050 my $d = parse_net($conf->{"net$i"});
4051 next if !$d;
4052
4053 $use_virtio = 1 if $d->{model} eq 'virtio';
4054
4055 if ($bootindex_hash->{n}) {
4056 $d->{bootindex} = $bootindex_hash->{n};
4057 $bootindex_hash->{n} += 1;
4058 }
4059
4060 my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
4061 push @$devices, '-netdev', $netdevfull;
4062
4063 my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
4064 push @$devices, '-device', $netdevicefull;
4065 }
4066
4067 if ($conf->{ivshmem}) {
4068 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
4069
4070 my $bus;
4071 if ($q35) {
4072 $bus = print_pcie_addr("ivshmem");
4073 } else {
4074 $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
4075 }
4076
4077 my $ivshmem_name = $ivshmem->{name} // $vmid;
4078 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4079
4080 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4081 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
4082 }
4083
4084 if (!$q35) {
4085 # add pci bridges
4086 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
4087 $bridges->{1} = 1;
4088 $bridges->{2} = 1;
4089 }
4090
4091 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4092
4093 for my $k (sort {$b cmp $a} keys %$bridges) {
4094 $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
4095 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
4096 }
4097 }
4098
4099 push @$cmd, @$devices;
4100 push @$cmd, '-rtc', join(',', @$rtcFlags)
4101 if scalar(@$rtcFlags);
4102 push @$cmd, '-machine', join(',', @$machineFlags)
4103 if scalar(@$machineFlags);
4104 push @$cmd, '-global', join(',', @$globalFlags)
4105 if scalar(@$globalFlags);
4106
4107 if (my $vmstate = $conf->{vmstate}) {
4108 my $statepath = PVE::Storage::path($storecfg, $vmstate);
4109 push @$vollist, $vmstate;
4110 push @$cmd, '-loadstate', $statepath;
4111 }
4112
4113 # add custom args
4114 if ($conf->{args}) {
4115 my $aa = PVE::Tools::split_args($conf->{args});
4116 push @$cmd, @$aa;
4117 }
4118
4119 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
4120 }
4121
4122 sub vnc_socket {
4123 my ($vmid) = @_;
4124 return "${var_run_tmpdir}/$vmid.vnc";
4125 }
4126
4127 sub spice_port {
4128 my ($vmid) = @_;
4129
4130 my $res = vm_mon_cmd($vmid, 'query-spice');
4131
4132 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4133 }
4134
4135 sub qmp_socket {
4136 my ($vmid, $qga) = @_;
4137 my $sockettype = $qga ? 'qga' : 'qmp';
4138 return "${var_run_tmpdir}/$vmid.$sockettype";
4139 }
4140
4141 sub pidfile_name {
4142 my ($vmid) = @_;
4143 return "${var_run_tmpdir}/$vmid.pid";
4144 }
4145
4146 sub vm_devices_list {
4147 my ($vmid) = @_;
4148
4149 my $res = vm_mon_cmd($vmid, 'query-pci');
4150 my $devices_to_check = [];
4151 my $devices = {};
4152 foreach my $pcibus (@$res) {
4153 push @$devices_to_check, @{$pcibus->{devices}},
4154 }
4155
4156 while (@$devices_to_check) {
4157 my $to_check = [];
4158 for my $d (@$devices_to_check) {
4159 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4160 next if !$d->{'pci_bridge'};
4161
4162 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
4163 push @$to_check, @{$d->{'pci_bridge'}->{devices}};
4164 }
4165 $devices_to_check = $to_check;
4166 }
4167
4168 my $resblock = vm_mon_cmd($vmid, 'query-block');
4169 foreach my $block (@$resblock) {
4170 if($block->{device} =~ m/^drive-(\S+)/){
4171 $devices->{$1} = 1;
4172 }
4173 }
4174
4175 my $resmice = vm_mon_cmd($vmid, 'query-mice');
4176 foreach my $mice (@$resmice) {
4177 if ($mice->{name} eq 'QEMU HID Tablet') {
4178 $devices->{tablet} = 1;
4179 last;
4180 }
4181 }
4182
4183 # for usb devices there is no query-usb
4184 # but we can iterate over the entries in
4185 # qom-list path=/machine/peripheral
4186 my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
4187 foreach my $per (@$resperipheral) {
4188 if ($per->{name} =~ m/^usb\d+$/) {
4189 $devices->{$per->{name}} = 1;
4190 }
4191 }
4192
4193 return $devices;
4194 }
4195
4196 sub vm_deviceplug {
4197 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4198
4199 my $q35 = machine_type_is_q35($conf);
4200
4201 my $devices_list = vm_devices_list($vmid);
4202 return 1 if defined($devices_list->{$deviceid});
4203
4204 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
4205
4206 if ($deviceid eq 'tablet') {
4207
4208 qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
4209
4210 } elsif ($deviceid eq 'keyboard') {
4211
4212 qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
4213
4214 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4215
4216 die "usb hotplug currently not reliable\n";
4217 # since we can't reliably hot unplug all added usb devices
4218 # and usb passthrough disables live migration
4219 # we disable usb hotplugging for now
4220 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
4221
4222 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4223
4224 qemu_iothread_add($vmid, $deviceid, $device);
4225
4226 qemu_driveadd($storecfg, $vmid, $device);
4227 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4228
4229 qemu_deviceadd($vmid, $devicefull);
4230 eval { qemu_deviceaddverify($vmid, $deviceid); };
4231 if (my $err = $@) {
4232 eval { qemu_drivedel($vmid, $deviceid); };
4233 warn $@ if $@;
4234 die $err;
4235 }
4236
4237 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4238
4239
4240 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
4241 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4242 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
4243
4244 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4245
4246 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
4247 qemu_iothread_add($vmid, $deviceid, $device);
4248 $devicefull .= ",iothread=iothread-$deviceid";
4249 }
4250
4251 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
4252 $devicefull .= ",num_queues=$device->{queues}";
4253 }
4254
4255 qemu_deviceadd($vmid, $devicefull);
4256 qemu_deviceaddverify($vmid, $deviceid);
4257
4258 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4259
4260 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4261 qemu_driveadd($storecfg, $vmid, $device);
4262
4263 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4264 eval { qemu_deviceadd($vmid, $devicefull); };
4265 if (my $err = $@) {
4266 eval { qemu_drivedel($vmid, $deviceid); };
4267 warn $@ if $@;
4268 die $err;
4269 }
4270
4271 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4272
4273 return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
4274
4275 my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4276 my $use_old_bios_files = undef;
4277 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4278
4279 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
4280 qemu_deviceadd($vmid, $netdevicefull);
4281 eval {
4282 qemu_deviceaddverify($vmid, $deviceid);
4283 qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
4284 };
4285 if (my $err = $@) {
4286 eval { qemu_netdevdel($vmid, $deviceid); };
4287 warn $@ if $@;
4288 die $err;
4289 }
4290
4291 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4292
4293 my $bridgeid = $2;
4294 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4295 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4296
4297 qemu_deviceadd($vmid, $devicefull);
4298 qemu_deviceaddverify($vmid, $deviceid);
4299
4300 } else {
4301 die "can't hotplug device '$deviceid'\n";
4302 }
4303
4304 return 1;
4305 }
4306
4307 # fixme: this should raise exceptions on error!
4308 sub vm_deviceunplug {
4309 my ($vmid, $conf, $deviceid) = @_;
4310
4311 my $devices_list = vm_devices_list($vmid);
4312 return 1 if !defined($devices_list->{$deviceid});
4313
4314 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4315
4316 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
4317
4318 qemu_devicedel($vmid, $deviceid);
4319
4320 } elsif ($deviceid =~ m/^usb\d+$/) {
4321
4322 die "usb hotplug currently not reliable\n";
4323 # when unplugging usb devices this way,
4324 # there may be remaining usb controllers/hubs
4325 # so we disable it for now
4326 qemu_devicedel($vmid, $deviceid);
4327 qemu_devicedelverify($vmid, $deviceid);
4328
4329 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4330
4331 qemu_devicedel($vmid, $deviceid);
4332 qemu_devicedelverify($vmid, $deviceid);
4333 qemu_drivedel($vmid, $deviceid);
4334 qemu_iothread_del($conf, $vmid, $deviceid);
4335
4336 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4337
4338 qemu_devicedel($vmid, $deviceid);
4339 qemu_devicedelverify($vmid, $deviceid);
4340 qemu_iothread_del($conf, $vmid, $deviceid);
4341
4342 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4343
4344 qemu_devicedel($vmid, $deviceid);
4345 qemu_drivedel($vmid, $deviceid);
4346 qemu_deletescsihw($conf, $vmid, $deviceid);
4347
4348 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4349
4350 qemu_devicedel($vmid, $deviceid);
4351 qemu_devicedelverify($vmid, $deviceid);
4352 qemu_netdevdel($vmid, $deviceid);
4353
4354 } else {
4355 die "can't unplug device '$deviceid'\n";
4356 }
4357
4358 return 1;
4359 }
4360
4361 sub qemu_deviceadd {
4362 my ($vmid, $devicefull) = @_;
4363
4364 $devicefull = "driver=".$devicefull;
4365 my %options = split(/[=,]/, $devicefull);
4366
4367 vm_mon_cmd($vmid, "device_add" , %options);
4368 }
4369
4370 sub qemu_devicedel {
4371 my ($vmid, $deviceid) = @_;
4372
4373 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4374 }
4375
4376 sub qemu_iothread_add {
4377 my($vmid, $deviceid, $device) = @_;
4378
4379 if ($device->{iothread}) {
4380 my $iothreads = vm_iothreads_list($vmid);
4381 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4382 }
4383 }
4384
4385 sub qemu_iothread_del {
4386 my($conf, $vmid, $deviceid) = @_;
4387
4388 my $confid = $deviceid;
4389 if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
4390 $confid = 'scsi' . $1;
4391 }
4392 my $device = parse_drive($confid, $conf->{$confid});
4393 if ($device->{iothread}) {
4394 my $iothreads = vm_iothreads_list($vmid);
4395 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4396 }
4397 }
4398
4399 sub qemu_objectadd {
4400 my($vmid, $objectid, $qomtype) = @_;
4401
4402 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4403
4404 return 1;
4405 }
4406
4407 sub qemu_objectdel {
4408 my($vmid, $objectid) = @_;
4409
4410 vm_mon_cmd($vmid, "object-del", id => $objectid);
4411
4412 return 1;
4413 }
4414
4415 sub qemu_driveadd {
4416 my ($storecfg, $vmid, $device) = @_;
4417
4418 my $drive = print_drive_full($storecfg, $vmid, $device);
4419 $drive =~ s/\\/\\\\/g;
4420 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4421
4422 # If the command succeeds qemu prints: "OK"
4423 return 1 if $ret =~ m/OK/s;
4424
4425 die "adding drive failed: $ret\n";
4426 }
4427
4428 sub qemu_drivedel {
4429 my($vmid, $deviceid) = @_;
4430
4431 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4432 $ret =~ s/^\s+//;
4433
4434 return 1 if $ret eq "";
4435
4436 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4437 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4438
4439 die "deleting drive $deviceid failed : $ret\n";
4440 }
4441
4442 sub qemu_deviceaddverify {
4443 my ($vmid, $deviceid) = @_;
4444
4445 for (my $i = 0; $i <= 5; $i++) {
4446 my $devices_list = vm_devices_list($vmid);
4447 return 1 if defined($devices_list->{$deviceid});
4448 sleep 1;
4449 }
4450
4451 die "error on hotplug device '$deviceid'\n";
4452 }
4453
4454
4455 sub qemu_devicedelverify {
4456 my ($vmid, $deviceid) = @_;
4457
4458 # need to verify that the device is correctly removed as device_del
4459 # is async and empty return is not reliable
4460
4461 for (my $i = 0; $i <= 5; $i++) {
4462 my $devices_list = vm_devices_list($vmid);
4463 return 1 if !defined($devices_list->{$deviceid});
4464 sleep 1;
4465 }
4466
4467 die "error on hot-unplugging device '$deviceid'\n";
4468 }
4469
4470 sub qemu_findorcreatescsihw {
4471 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4472
4473 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4474
4475 my $scsihwid="$controller_prefix$controller";
4476 my $devices_list = vm_devices_list($vmid);
4477
4478 if(!defined($devices_list->{$scsihwid})) {
4479 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4480 }
4481
4482 return 1;
4483 }
4484
4485 sub qemu_deletescsihw {
4486 my ($conf, $vmid, $opt) = @_;
4487
4488 my $device = parse_drive($opt, $conf->{$opt});
4489
4490 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4491 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4492 return 1;
4493 }
4494
4495 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4496
4497 my $devices_list = vm_devices_list($vmid);
4498 foreach my $opt (keys %{$devices_list}) {
4499 if (PVE::QemuServer::is_valid_drivename($opt)) {
4500 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4501 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4502 return 1;
4503 }
4504 }
4505 }
4506
4507 my $scsihwid="scsihw$controller";
4508
4509 vm_deviceunplug($vmid, $conf, $scsihwid);
4510
4511 return 1;
4512 }
4513
4514 sub qemu_add_pci_bridge {
4515 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4516
4517 my $bridges = {};
4518
4519 my $bridgeid;
4520
4521 print_pci_addr($device, $bridges, $arch, $machine_type);
4522
4523 while (my ($k, $v) = each %$bridges) {
4524 $bridgeid = $k;
4525 }
4526 return 1 if !defined($bridgeid) || $bridgeid < 1;
4527
4528 my $bridge = "pci.$bridgeid";
4529 my $devices_list = vm_devices_list($vmid);
4530
4531 if (!defined($devices_list->{$bridge})) {
4532 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4533 }
4534
4535 return 1;
4536 }
4537
4538 sub qemu_set_link_status {
4539 my ($vmid, $device, $up) = @_;
4540
4541 vm_mon_cmd($vmid, "set_link", name => $device,
4542 up => $up ? JSON::true : JSON::false);
4543 }
4544
4545 sub qemu_netdevadd {
4546 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4547
4548 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4549 my %options = split(/[=,]/, $netdev);
4550
4551 vm_mon_cmd($vmid, "netdev_add", %options);
4552 return 1;
4553 }
4554
4555 sub qemu_netdevdel {
4556 my ($vmid, $deviceid) = @_;
4557
4558 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4559 }
4560
4561 sub qemu_usb_hotplug {
4562 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4563
4564 return if !$device;
4565
4566 # remove the old one first
4567 vm_deviceunplug($vmid, $conf, $deviceid);
4568
4569 # check if xhci controller is necessary and available
4570 if ($device->{usb3}) {
4571
4572 my $devicelist = vm_devices_list($vmid);
4573
4574 if (!$devicelist->{xhci}) {
4575 my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4576 qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4577 }
4578 }
4579 my $d = parse_usb_device($device->{host});
4580 $d->{usb3} = $device->{usb3};
4581
4582 # add the new one
4583 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4584 }
4585
4586 sub qemu_cpu_hotplug {
4587 my ($vmid, $conf, $vcpus) = @_;
4588
4589 my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4590
4591 my $sockets = 1;
4592 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4593 $sockets = $conf->{sockets} if $conf->{sockets};
4594 my $cores = $conf->{cores} || 1;
4595 my $maxcpus = $sockets * $cores;
4596
4597 $vcpus = $maxcpus if !$vcpus;
4598
4599 die "you can't add more vcpus than maxcpus\n"
4600 if $vcpus > $maxcpus;
4601
4602 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4603
4604 if ($vcpus < $currentvcpus) {
4605
4606 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4607
4608 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4609 qemu_devicedel($vmid, "cpu$i");
4610 my $retry = 0;
4611 my $currentrunningvcpus = undef;
4612 while (1) {
4613 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4614 last if scalar(@{$currentrunningvcpus}) == $i-1;
4615 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4616 $retry++;
4617 sleep 1;
4618 }
4619 #update conf after each succesfull cpu unplug
4620 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4621 PVE::QemuConfig->write_config($vmid, $conf);
4622 }
4623 } else {
4624 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4625 }
4626
4627 return;
4628 }
4629
4630 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4631 die "vcpus in running vm does not match its configuration\n"
4632 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4633
4634 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4635
4636 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4637 my $cpustr = print_cpu_device($conf, $i);
4638 qemu_deviceadd($vmid, $cpustr);
4639
4640 my $retry = 0;
4641 my $currentrunningvcpus = undef;
4642 while (1) {
4643 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4644 last if scalar(@{$currentrunningvcpus}) == $i;
4645 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4646 sleep 1;
4647 $retry++;
4648 }
4649 #update conf after each succesfull cpu hotplug
4650 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4651 PVE::QemuConfig->write_config($vmid, $conf);
4652 }
4653 } else {
4654
4655 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4656 vm_mon_cmd($vmid, "cpu-add", id => int($i));
4657 }
4658 }
4659 }
4660
4661 sub qemu_block_set_io_throttle {
4662 my ($vmid, $deviceid,
4663 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4664 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4665 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4666 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4667
4668 return if !check_running($vmid) ;
4669
4670 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4671 bps => int($bps),
4672 bps_rd => int($bps_rd),
4673 bps_wr => int($bps_wr),
4674 iops => int($iops),
4675 iops_rd => int($iops_rd),
4676 iops_wr => int($iops_wr),
4677 bps_max => int($bps_max),
4678 bps_rd_max => int($bps_rd_max),
4679 bps_wr_max => int($bps_wr_max),
4680 iops_max => int($iops_max),
4681 iops_rd_max => int($iops_rd_max),
4682 iops_wr_max => int($iops_wr_max),
4683 bps_max_length => int($bps_max_length),
4684 bps_rd_max_length => int($bps_rd_max_length),
4685 bps_wr_max_length => int($bps_wr_max_length),
4686 iops_max_length => int($iops_max_length),
4687 iops_rd_max_length => int($iops_rd_max_length),
4688 iops_wr_max_length => int($iops_wr_max_length),
4689 );
4690
4691 }
4692
4693 # old code, only used to shutdown old VM after update
4694 sub __read_avail {
4695 my ($fh, $timeout) = @_;
4696
4697 my $sel = new IO::Select;
4698 $sel->add($fh);
4699
4700 my $res = '';
4701 my $buf;
4702
4703 my @ready;
4704 while (scalar (@ready = $sel->can_read($timeout))) {
4705 my $count;
4706 if ($count = $fh->sysread($buf, 8192)) {
4707 if ($buf =~ /^(.*)\(qemu\) $/s) {
4708 $res .= $1;
4709 last;
4710 } else {
4711 $res .= $buf;
4712 }
4713 } else {
4714 if (!defined($count)) {
4715 die "$!\n";
4716 }
4717 last;
4718 }
4719 }
4720
4721 die "monitor read timeout\n" if !scalar(@ready);
4722
4723 return $res;
4724 }
4725
4726 sub qemu_block_resize {
4727 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4728
4729 my $running = check_running($vmid);
4730
4731 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4732
4733 return if !$running;
4734
4735 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4736
4737 }
4738
4739 sub qemu_volume_snapshot {
4740 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4741
4742 my $running = check_running($vmid);
4743
4744 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4745 vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4746 } else {
4747 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4748 }
4749 }
4750
4751 sub qemu_volume_snapshot_delete {
4752 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4753
4754 my $running = check_running($vmid);
4755
4756 if($running) {
4757
4758 $running = undef;
4759 my $conf = PVE::QemuConfig->load_config($vmid);
4760 foreach_drive($conf, sub {
4761 my ($ds, $drive) = @_;
4762 $running = 1 if $drive->{file} eq $volid;
4763 });
4764 }
4765
4766 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4767 vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4768 } else {
4769 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4770 }
4771 }
4772
4773 sub set_migration_caps {
4774 my ($vmid) = @_;
4775
4776 my $cap_ref = [];
4777
4778 my $enabled_cap = {
4779 "auto-converge" => 1,
4780 "xbzrle" => 1,
4781 "x-rdma-pin-all" => 0,
4782 "zero-blocks" => 0,
4783 "compress" => 0
4784 };
4785
4786 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4787
4788 for my $supported_capability (@$supported_capabilities) {
4789 push @$cap_ref, {
4790 capability => $supported_capability->{capability},
4791 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4792 };
4793 }
4794
4795 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4796 }
4797
4798 my $fast_plug_option = {
4799 'lock' => 1,
4800 'name' => 1,
4801 'onboot' => 1,
4802 'shares' => 1,
4803 'startup' => 1,
4804 'description' => 1,
4805 'protection' => 1,
4806 'vmstatestorage' => 1,
4807 'hookscript' => 1,
4808 };
4809
4810 # hotplug changes in [PENDING]
4811 # $selection hash can be used to only apply specified options, for
4812 # example: { cores => 1 } (only apply changed 'cores')
4813 # $errors ref is used to return error messages
4814 sub vmconfig_hotplug_pending {
4815 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4816
4817 my $defaults = load_defaults();
4818 my ($arch, $machine_type) = get_basic_machine_info($conf, undef);
4819
4820 # commit values which do not have any impact on running VM first
4821 # Note: those option cannot raise errors, we we do not care about
4822 # $selection and always apply them.
4823
4824 my $add_error = sub {
4825 my ($opt, $msg) = @_;
4826 $errors->{$opt} = "hotplug problem - $msg";
4827 };
4828
4829 my $changes = 0;
4830 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4831 if ($fast_plug_option->{$opt}) {
4832 $conf->{$opt} = $conf->{pending}->{$opt};
4833 delete $conf->{pending}->{$opt};
4834 $changes = 1;
4835 }
4836 }
4837
4838 if ($changes) {
4839 PVE::QemuConfig->write_config($vmid, $conf);
4840 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4841 }
4842
4843 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4844
4845 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
4846 foreach my $opt (sort keys %$pending_delete_hash) {
4847 next if $selection && !$selection->{$opt};
4848 my $force = $pending_delete_hash->{$opt}->{force};
4849 eval {
4850 if ($opt eq 'hotplug') {
4851 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4852 } elsif ($opt eq 'tablet') {
4853 die "skip\n" if !$hotplug_features->{usb};
4854 if ($defaults->{tablet}) {
4855 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4856 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4857 if $arch eq 'aarch64';
4858 } else {
4859 vm_deviceunplug($vmid, $conf, 'tablet');
4860 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4861 }
4862 } elsif ($opt =~ m/^usb\d+/) {
4863 die "skip\n";
4864 # since we cannot reliably hot unplug usb devices
4865 # we are disabling it
4866 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4867 vm_deviceunplug($vmid, $conf, $opt);
4868 } elsif ($opt eq 'vcpus') {
4869 die "skip\n" if !$hotplug_features->{cpu};
4870 qemu_cpu_hotplug($vmid, $conf, undef);
4871 } elsif ($opt eq 'balloon') {
4872 # enable balloon device is not hotpluggable
4873 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4874 # here we reset the ballooning value to memory
4875 my $balloon = $conf->{memory} || $defaults->{memory};
4876 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4877 } elsif ($fast_plug_option->{$opt}) {
4878 # do nothing
4879 } elsif ($opt =~ m/^net(\d+)$/) {
4880 die "skip\n" if !$hotplug_features->{network};
4881 vm_deviceunplug($vmid, $conf, $opt);
4882 } elsif (is_valid_drivename($opt)) {
4883 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4884 vm_deviceunplug($vmid, $conf, $opt);
4885 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4886 } elsif ($opt =~ m/^memory$/) {
4887 die "skip\n" if !$hotplug_features->{memory};
4888 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4889 } elsif ($opt eq 'cpuunits') {
4890 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4891 } elsif ($opt eq 'cpulimit') {
4892 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4893 } else {
4894 die "skip\n";
4895 }
4896 };
4897 if (my $err = $@) {
4898 &$add_error($opt, $err) if $err ne "skip\n";
4899 } else {
4900 # save new config if hotplug was successful
4901 delete $conf->{$opt};
4902 PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
4903 PVE::QemuConfig->write_config($vmid, $conf);
4904 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4905 }
4906 }
4907
4908 my $apply_pending_cloudinit;
4909 $apply_pending_cloudinit = sub {
4910 my ($key, $value) = @_;
4911 $apply_pending_cloudinit = sub {}; # once is enough
4912
4913 my @cloudinit_opts = keys %$confdesc_cloudinit;
4914 foreach my $opt (keys %{$conf->{pending}}) {
4915 next if !grep { $_ eq $opt } @cloudinit_opts;
4916 $conf->{$opt} = delete $conf->{pending}->{$opt};
4917 }
4918
4919 my $new_conf = { %$conf };
4920 $new_conf->{$key} = $value;
4921 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4922 };
4923
4924 foreach my $opt (keys %{$conf->{pending}}) {
4925 next if $selection && !$selection->{$opt};
4926 my $value = $conf->{pending}->{$opt};
4927 eval {
4928 if ($opt eq 'hotplug') {
4929 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4930 } elsif ($opt eq 'tablet') {
4931 die "skip\n" if !$hotplug_features->{usb};
4932 if ($value == 1) {
4933 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4934 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4935 if $arch eq 'aarch64';
4936 } elsif ($value == 0) {
4937 vm_deviceunplug($vmid, $conf, 'tablet');
4938 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4939 }
4940 } elsif ($opt =~ m/^usb\d+$/) {
4941 die "skip\n";
4942 # since we cannot reliably hot unplug usb devices
4943 # we are disabling it
4944 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
4945 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
4946 die "skip\n" if !$d;
4947 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
4948 } elsif ($opt eq 'vcpus') {
4949 die "skip\n" if !$hotplug_features->{cpu};
4950 qemu_cpu_hotplug($vmid, $conf, $value);
4951 } elsif ($opt eq 'balloon') {
4952 # enable/disable balloning device is not hotpluggable
4953 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
4954 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4955 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4956
4957 # allow manual ballooning if shares is set to zero
4958 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4959 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4960 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4961 }
4962 } elsif ($opt =~ m/^net(\d+)$/) {
4963 # some changes can be done without hotplug
4964 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4965 $vmid, $opt, $value, $arch, $machine_type);
4966 } elsif (is_valid_drivename($opt)) {
4967 # some changes can be done without hotplug
4968 my $drive = parse_drive($opt, $value);
4969 if (drive_is_cloudinit($drive)) {
4970 &$apply_pending_cloudinit($opt, $value);
4971 }
4972 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4973 $vmid, $opt, $value, 1, $arch, $machine_type);
4974 } elsif ($opt =~ m/^memory$/) { #dimms
4975 die "skip\n" if !$hotplug_features->{memory};
4976 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4977 } elsif ($opt eq 'cpuunits') {
4978 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4979 } elsif ($opt eq 'cpulimit') {
4980 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4981 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4982 } else {
4983 die "skip\n"; # skip non-hot-pluggable options
4984 }
4985 };
4986 if (my $err = $@) {
4987 &$add_error($opt, $err) if $err ne "skip\n";
4988 } else {
4989 # save new config if hotplug was successful
4990 $conf->{$opt} = $value;
4991 delete $conf->{pending}->{$opt};
4992 PVE::QemuConfig->write_config($vmid, $conf);
4993 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4994 }
4995 }
4996 }
4997
4998 sub try_deallocate_drive {
4999 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5000
5001 if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
5002 my $volid = $drive->{file};
5003 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
5004 my $sid = PVE::Storage::parse_volume_id($volid);
5005 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5006
5007 # check if the disk is really unused
5008 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5009 if is_volume_in_use($storecfg, $conf, $key, $volid);
5010 PVE::Storage::vdisk_free($storecfg, $volid);
5011 return 1;
5012 } else {
5013 # If vm is not owner of this disk remove from config
5014 return 1;
5015 }
5016 }
5017
5018 return undef;
5019 }
5020
5021 sub vmconfig_delete_or_detach_drive {
5022 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5023
5024 my $drive = parse_drive($opt, $conf->{$opt});
5025
5026 my $rpcenv = PVE::RPCEnvironment::get();
5027 my $authuser = $rpcenv->get_user();
5028
5029 if ($force) {
5030 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5031 try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5032 } else {
5033 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
5034 }
5035 }
5036
5037
5038
5039 sub vmconfig_apply_pending {
5040 my ($vmid, $conf, $storecfg) = @_;
5041
5042 # cold plug
5043
5044 my $pending_delete_hash = PVE::QemuConfig->parse_pending_delete($conf->{pending}->{delete});
5045 foreach my $opt (sort keys %$pending_delete_hash) {
5046 die "internal error" if $opt =~ m/^unused/;
5047 my $force = $pending_delete_hash->{$opt}->{force};
5048 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5049 if (!defined($conf->{$opt})) {
5050 PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
5051 PVE::QemuConfig->write_config($vmid, $conf);
5052 } elsif (is_valid_drivename($opt)) {
5053 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
5054 PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
5055 delete $conf->{$opt};
5056 PVE::QemuConfig->write_config($vmid, $conf);
5057 } else {
5058 PVE::QemuConfig->remove_from_pending_delete($conf, $opt);
5059 delete $conf->{$opt};
5060 PVE::QemuConfig->write_config($vmid, $conf);
5061 }
5062 }
5063
5064 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5065
5066 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5067 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5068
5069 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
5070 # skip if nothing changed
5071 } elsif (is_valid_drivename($opt)) {
5072 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
5073 if defined($conf->{$opt});
5074 $conf->{$opt} = $conf->{pending}->{$opt};
5075 } else {
5076 $conf->{$opt} = $conf->{pending}->{$opt};
5077 }
5078
5079 delete $conf->{pending}->{$opt};
5080 PVE::QemuConfig->write_config($vmid, $conf);
5081 }
5082 }
5083
5084 my $safe_num_ne = sub {
5085 my ($a, $b) = @_;
5086
5087 return 0 if !defined($a) && !defined($b);
5088 return 1 if !defined($a);
5089 return 1 if !defined($b);
5090
5091 return $a != $b;
5092 };
5093
5094 my $safe_string_ne = sub {
5095 my ($a, $b) = @_;
5096
5097 return 0 if !defined($a) && !defined($b);
5098 return 1 if !defined($a);
5099 return 1 if !defined($b);
5100
5101 return $a ne $b;
5102 };
5103
5104 sub vmconfig_update_net {
5105 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5106
5107 my $newnet = parse_net($value);
5108
5109 if ($conf->{$opt}) {
5110 my $oldnet = parse_net($conf->{$opt});
5111
5112 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
5113 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
5114 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
5115 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
5116
5117 # for non online change, we try to hot-unplug
5118 die "skip\n" if !$hotplug;
5119 vm_deviceunplug($vmid, $conf, $opt);
5120 } else {
5121
5122 die "internal error" if $opt !~ m/net(\d+)/;
5123 my $iface = "tap${vmid}i$1";
5124
5125 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
5126 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
5127 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
5128 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
5129 PVE::Network::tap_unplug($iface);
5130 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
5131 } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
5132 # Rate can be applied on its own but any change above needs to
5133 # include the rate in tap_plug since OVS resets everything.
5134 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
5135 }
5136
5137 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
5138 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
5139 }
5140
5141 return 1;
5142 }
5143 }
5144
5145 if ($hotplug) {
5146 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5147 } else {
5148 die "skip\n";
5149 }
5150 }
5151
5152 sub vmconfig_update_disk {
5153 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force, $arch, $machine_type) = @_;
5154
5155 # fixme: do we need force?
5156
5157 my $drive = parse_drive($opt, $value);
5158
5159 if ($conf->{$opt}) {
5160
5161 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
5162
5163 my $media = $drive->{media} || 'disk';
5164 my $oldmedia = $old_drive->{media} || 'disk';
5165 die "unable to change media type\n" if $media ne $oldmedia;
5166
5167 if (!drive_is_cdrom($old_drive)) {
5168
5169 if ($drive->{file} ne $old_drive->{file}) {
5170
5171 die "skip\n" if !$hotplug;
5172
5173 # unplug and register as unused
5174 vm_deviceunplug($vmid, $conf, $opt);
5175 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
5176
5177 } else {
5178 # update existing disk
5179
5180 # skip non hotpluggable value
5181 if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
5182 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
5183 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
5184 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
5185 die "skip\n";
5186 }
5187
5188 # apply throttle
5189 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
5190 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
5191 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
5192 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
5193 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
5194 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
5195 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
5196 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
5197 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
5198 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
5199 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
5200 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
5201 &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
5202 &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
5203 &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
5204 &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
5205 &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
5206 &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
5207
5208 qemu_block_set_io_throttle($vmid,"drive-$opt",
5209 ($drive->{mbps} || 0)*1024*1024,
5210 ($drive->{mbps_rd} || 0)*1024*1024,
5211 ($drive->{mbps_wr} || 0)*1024*1024,
5212 $drive->{iops} || 0,
5213 $drive->{iops_rd} || 0,
5214 $drive->{iops_wr} || 0,
5215 ($drive->{mbps_max} || 0)*1024*1024,
5216 ($drive->{mbps_rd_max} || 0)*1024*1024,
5217 ($drive->{mbps_wr_max} || 0)*1024*1024,
5218 $drive->{iops_max} || 0,
5219 $drive->{iops_rd_max} || 0,
5220 $drive->{iops_wr_max} || 0,
5221 $drive->{bps_max_length} || 1,
5222 $drive->{bps_rd_max_length} || 1,
5223 $drive->{bps_wr_max_length} || 1,
5224 $drive->{iops_max_length} || 1,
5225 $drive->{iops_rd_max_length} || 1,
5226 $drive->{iops_wr_max_length} || 1);
5227
5228 }
5229
5230 return 1;
5231 }
5232
5233 } else { # cdrom
5234
5235 if ($drive->{file} eq 'none') {
5236 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
5237 if (drive_is_cloudinit($old_drive)) {
5238 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
5239 }
5240 } else {
5241 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
5242 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
5243 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
5244 }
5245
5246 return 1;
5247 }
5248 }
5249 }
5250
5251 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5252 # hotplug new disks
5253 PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
5254 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5255 }
5256
5257 sub vm_start {
5258 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
5259 $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
5260
5261 PVE::QemuConfig->lock_config($vmid, sub {
5262 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5263
5264 die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
5265
5266 my $is_suspended = PVE::QemuConfig->has_lock($conf, 'suspended');
5267
5268 PVE::QemuConfig->check_lock($conf)
5269 if !($skiplock || $is_suspended);
5270
5271 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
5272
5273 # clean up leftover reboot request files
5274 eval { clear_reboot_request($vmid); };
5275 warn $@ if $@;
5276
5277 if (!$statefile && scalar(keys %{$conf->{pending}})) {
5278 vmconfig_apply_pending($vmid, $conf, $storecfg);
5279 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5280 }
5281
5282 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
5283
5284 my $defaults = load_defaults();
5285
5286 # set environment variable useful inside network script
5287 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
5288
5289 my $local_volumes = {};
5290
5291 if ($targetstorage) {
5292 foreach_drive($conf, sub {
5293 my ($ds, $drive) = @_;
5294
5295 return if drive_is_cdrom($drive);
5296
5297 my $volid = $drive->{file};
5298
5299 return if !$volid;
5300
5301 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5302
5303 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5304 return if $scfg->{shared};
5305 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5306 });
5307
5308 my $format = undef;
5309
5310 foreach my $opt (sort keys %$local_volumes) {
5311
5312 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5313 my $drive = parse_drive($opt, $conf->{$opt});
5314
5315 #if remote storage is specified, use default format
5316 if ($targetstorage && $targetstorage ne "1") {
5317 $storeid = $targetstorage;
5318 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5319 $format = $defFormat;
5320 } else {
5321 #else we use same format than original
5322 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5323 $format = qemu_img_format($scfg, $volid);
5324 }
5325
5326 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5327 my $newdrive = $drive;
5328 $newdrive->{format} = $format;
5329 $newdrive->{file} = $newvolid;
5330 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5331 $local_volumes->{$opt} = $drivestr;
5332 #pass drive to conf for command line
5333 $conf->{$opt} = $drivestr;
5334 }
5335 }
5336
5337 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
5338
5339 if ($is_suspended) {
5340 # enforce machine type on suspended vm to ensure HW compatibility
5341 $forcemachine = $conf->{runningmachine};
5342 print "Resuming suspended VM\n";
5343 }
5344
5345 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5346
5347 my $migrate_uri;
5348 if ($statefile) {
5349 if ($statefile eq 'tcp') {
5350 my $localip = "localhost";
5351 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5352 my $nodename = PVE::INotify::nodename();
5353
5354 if (!defined($migration_type)) {
5355 if (defined($datacenterconf->{migration}->{type})) {
5356 $migration_type = $datacenterconf->{migration}->{type};
5357 } else {
5358 $migration_type = 'secure';
5359 }
5360 }
5361
5362 if ($migration_type eq 'insecure') {
5363 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5364 if ($migrate_network_addr) {
5365 $localip = $migrate_network_addr;
5366 } else {
5367 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5368 }
5369
5370 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5371 }
5372
5373 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5374 my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5375 $migrate_uri = "tcp:${localip}:${migrate_port}";
5376 push @$cmd, '-incoming', $migrate_uri;
5377 push @$cmd, '-S';
5378
5379 } elsif ($statefile eq 'unix') {
5380 # should be default for secure migrations as a ssh TCP forward
5381 # tunnel is not deterministic reliable ready and fails regurarly
5382 # to set up in time, so use UNIX socket forwards
5383 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5384 unlink $socket_addr;
5385
5386 $migrate_uri = "unix:$socket_addr";
5387
5388 push @$cmd, '-incoming', $migrate_uri;
5389 push @$cmd, '-S';
5390
5391 } elsif (-e $statefile) {
5392 push @$cmd, '-loadstate', $statefile;
5393 } else {
5394 my $statepath = PVE::Storage::path($storecfg, $statefile);
5395 push @$vollist, $statefile;
5396 push @$cmd, '-loadstate', $statepath;
5397 }
5398 } elsif ($paused) {
5399 push @$cmd, '-S';
5400 }
5401
5402 # host pci devices
5403 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
5404 my $d = parse_hostpci($conf->{"hostpci$i"});
5405 next if !$d;
5406 my $pcidevices = $d->{pciid};
5407 foreach my $pcidevice (@$pcidevices) {
5408 my $pciid = $pcidevice->{id};
5409
5410 my $info = PVE::SysFSTools::pci_device_info("0000:$pciid");
5411 die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5412 die "no pci device info for device '$pciid'\n" if !$info;
5413
5414 if ($d->{mdev}) {
5415 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5416 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5417 } else {
5418 die "can't unbind/bind pci group to vfio '$pciid'\n"
5419 if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5420 die "can't reset pci device '$pciid'\n"
5421 if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5422 }
5423 }
5424 }
5425
5426 PVE::Storage::activate_volumes($storecfg, $vollist);
5427
5428 eval {
5429 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5430 outfunc => sub {}, errfunc => sub {});
5431 };
5432 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5433 # timeout should be more than enough here...
5434 PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5435
5436 my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5437 : $defaults->{cpuunits};
5438
5439 my $start_timeout = ($conf->{hugepages} || $is_suspended) ? 300 : 30;
5440 my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5441
5442 my %properties = (
5443 Slice => 'qemu.slice',
5444 KillMode => 'none',
5445 CPUShares => $cpuunits
5446 );
5447
5448 if (my $cpulimit = $conf->{cpulimit}) {
5449 $properties{CPUQuota} = int($cpulimit * 100);
5450 }
5451 $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5452
5453 my $run_qemu = sub {
5454 PVE::Tools::run_fork sub {
5455 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5456 run_command($cmd, %run_params);
5457 };
5458 };
5459
5460 if ($conf->{hugepages}) {
5461
5462 my $code = sub {
5463 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5464 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5465
5466 PVE::QemuServer::Memory::hugepages_mount();
5467 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5468
5469 eval { $run_qemu->() };
5470 if (my $err = $@) {
5471 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5472 die $err;
5473 }
5474
5475 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5476 };
5477 eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5478
5479 } else {
5480 eval { $run_qemu->() };
5481 }
5482
5483 if (my $err = $@) {
5484 # deactivate volumes if start fails
5485 eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5486 die "start failed: $err";
5487 }
5488
5489 print "migration listens on $migrate_uri\n" if $migrate_uri;
5490
5491 if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix') {
5492 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5493 warn $@ if $@;
5494 }
5495
5496 #start nbd server for storage migration
5497 if ($targetstorage) {
5498 my $nodename = PVE::INotify::nodename();
5499 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5500 my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5501 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5502 my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5503
5504 vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5505
5506 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5507
5508 foreach my $opt (sort keys %$local_volumes) {
5509 my $volid = $local_volumes->{$opt};
5510 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5511 my $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}:exportname=drive-$opt";
5512 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5513 }
5514 }
5515
5516 if ($migratedfrom) {
5517 eval {
5518 set_migration_caps($vmid);
5519 };
5520 warn $@ if $@;
5521
5522 if ($spice_port) {
5523 print "spice listens on port $spice_port\n";
5524 if ($spice_ticket) {
5525 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5526 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5527 }
5528 }
5529
5530 } else {
5531 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5532 if !$statefile && $conf->{balloon};
5533
5534 foreach my $opt (keys %$conf) {
5535 next if $opt !~ m/^net\d+$/;
5536 my $nicconf = parse_net($conf->{$opt});
5537 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5538 }
5539 }
5540
5541 vm_mon_cmd_nocheck($vmid, 'qom-set',
5542 path => "machine/peripheral/balloon0",
5543 property => "guest-stats-polling-interval",
5544 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5545
5546 if ($is_suspended && (my $vmstate = $conf->{vmstate})) {
5547 print "Resumed VM, removing state\n";
5548 delete $conf->@{qw(lock vmstate runningmachine)};
5549 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5550 PVE::Storage::vdisk_free($storecfg, $vmstate);
5551 PVE::QemuConfig->write_config($vmid, $conf);
5552 }
5553
5554 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5555 });
5556 }
5557
5558 sub vm_mon_cmd {
5559 my ($vmid, $execute, %params) = @_;
5560
5561 my $cmd = { execute => $execute, arguments => \%params };
5562 vm_qmp_command($vmid, $cmd);
5563 }
5564
5565 sub vm_mon_cmd_nocheck {
5566 my ($vmid, $execute, %params) = @_;
5567
5568 my $cmd = { execute => $execute, arguments => \%params };
5569 vm_qmp_command($vmid, $cmd, 1);
5570 }
5571
5572 sub vm_qmp_command {
5573 my ($vmid, $cmd, $nocheck) = @_;
5574
5575 my $res;
5576
5577 my $timeout;
5578 if ($cmd->{arguments}) {
5579 $timeout = delete $cmd->{arguments}->{timeout};
5580 }
5581
5582 eval {
5583 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5584 my $sname = qmp_socket($vmid);
5585 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5586 my $qmpclient = PVE::QMPClient->new();
5587
5588 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5589 } else {
5590 die "unable to open monitor socket\n";
5591 }
5592 };
5593 if (my $err = $@) {
5594 syslog("err", "VM $vmid qmp command failed - $err");
5595 die $err;
5596 }
5597
5598 return $res;
5599 }
5600
5601 sub vm_human_monitor_command {
5602 my ($vmid, $cmdline) = @_;
5603
5604 my $cmd = {
5605 execute => 'human-monitor-command',
5606 arguments => { 'command-line' => $cmdline},
5607 };
5608
5609 return vm_qmp_command($vmid, $cmd);
5610 }
5611
5612 sub vm_commandline {
5613 my ($storecfg, $vmid, $snapname) = @_;
5614
5615 my $conf = PVE::QemuConfig->load_config($vmid);
5616
5617 if ($snapname) {
5618 my $snapshot = $conf->{snapshots}->{$snapname};
5619 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5620
5621 $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5622
5623 $conf = $snapshot;
5624 }
5625
5626 my $defaults = load_defaults();
5627
5628 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5629
5630 return PVE::Tools::cmd2string($cmd);
5631 }
5632
5633 sub vm_reset {
5634 my ($vmid, $skiplock) = @_;
5635
5636 PVE::QemuConfig->lock_config($vmid, sub {
5637
5638 my $conf = PVE::QemuConfig->load_config($vmid);
5639
5640 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5641
5642 vm_mon_cmd($vmid, "system_reset");
5643 });
5644 }
5645
5646 sub get_vm_volumes {
5647 my ($conf) = @_;
5648
5649 my $vollist = [];
5650 foreach_volid($conf, sub {
5651 my ($volid, $attr) = @_;
5652
5653 return if $volid =~ m|^/|;
5654
5655 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5656 return if !$sid;
5657
5658 push @$vollist, $volid;
5659 });
5660
5661 return $vollist;
5662 }
5663
5664 sub vm_stop_cleanup {
5665 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5666
5667 eval {
5668
5669 if (!$keepActive) {
5670 my $vollist = get_vm_volumes($conf);
5671 PVE::Storage::deactivate_volumes($storecfg, $vollist);
5672 }
5673
5674 foreach my $ext (qw(mon qmp pid vnc qga)) {
5675 unlink "/var/run/qemu-server/${vmid}.$ext";
5676 }
5677
5678 if ($conf->{ivshmem}) {
5679 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5680 # just delete it for now, VMs which have this already open do not
5681 # are affected, but new VMs will get a separated one. If this
5682 # becomes an issue we either add some sort of ref-counting or just
5683 # add a "don't delete on stop" flag to the ivshmem format.
5684 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5685 }
5686
5687 foreach my $key (keys %$conf) {
5688 next if $key !~ m/^hostpci(\d+)$/;
5689 my $hostpciindex = $1;
5690 my $d = parse_hostpci($conf->{$key});
5691 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5692
5693 foreach my $pci (@{$d->{pciid}}) {
5694 my $pciid = $pci->{id};
5695 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5696 }
5697 }
5698
5699 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5700 };
5701 warn $@ if $@; # avoid errors - just warn
5702 }
5703
5704 # call only in locked context
5705 sub _do_vm_stop {
5706 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5707
5708 my $pid = check_running($vmid, $nocheck);
5709 return if !$pid;
5710
5711 my $conf;
5712 if (!$nocheck) {
5713 $conf = PVE::QemuConfig->load_config($vmid);
5714 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5715 if (!defined($timeout) && $shutdown && $conf->{startup}) {
5716 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5717 $timeout = $opts->{down} if $opts->{down};
5718 }
5719 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5720 }
5721
5722 eval {
5723 if ($shutdown) {
5724 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5725 vm_qmp_command($vmid, {
5726 execute => "guest-shutdown",
5727 arguments => { timeout => $timeout }
5728 }, $nocheck);
5729 } else {
5730 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5731 }
5732 } else {
5733 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5734 }
5735 };
5736 my $err = $@;
5737
5738 if (!$err) {
5739 $timeout = 60 if !defined($timeout);
5740
5741 my $count = 0;
5742 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5743 $count++;
5744 sleep 1;
5745 }
5746
5747 if ($count >= $timeout) {
5748 if ($force) {
5749 warn "VM still running - terminating now with SIGTERM\n";
5750 kill 15, $pid;
5751 } else {
5752 die "VM quit/powerdown failed - got timeout\n";
5753 }
5754 } else {
5755 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5756 return;
5757 }
5758 } else {
5759 if ($force) {
5760 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5761 kill 15, $pid;
5762 } else {
5763 die "VM quit/powerdown failed\n";
5764 }
5765 }
5766
5767 # wait again
5768 $timeout = 10;
5769
5770 my $count = 0;
5771 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5772 $count++;
5773 sleep 1;
5774 }
5775
5776 if ($count >= $timeout) {
5777 warn "VM still running - terminating now with SIGKILL\n";
5778 kill 9, $pid;
5779 sleep 1;
5780 }
5781
5782 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5783 }
5784
5785 # Note: use $nocheck to skip tests if VM configuration file exists.
5786 # We need that when migration VMs to other nodes (files already moved)
5787 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5788 sub vm_stop {
5789 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5790
5791 $force = 1 if !defined($force) && !$shutdown;
5792
5793 if ($migratedfrom){
5794 my $pid = check_running($vmid, $nocheck, $migratedfrom);
5795 kill 15, $pid if $pid;
5796 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5797 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5798 return;
5799 }
5800
5801 PVE::QemuConfig->lock_config($vmid, sub {
5802 _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5803 });
5804 }
5805
5806 sub vm_reboot {
5807 my ($vmid, $timeout) = @_;
5808
5809 PVE::QemuConfig->lock_config($vmid, sub {
5810
5811 # only reboot if running, as qmeventd starts it again on a stop event
5812 return if !check_running($vmid);
5813
5814 create_reboot_request($vmid);
5815
5816 my $storecfg = PVE::Storage::config();
5817 _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5818
5819 });
5820 }
5821
5822 sub vm_suspend {
5823 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5824
5825 my $conf;
5826 my $path;
5827 my $storecfg;
5828 my $vmstate;
5829
5830 PVE::QemuConfig->lock_config($vmid, sub {
5831
5832 $conf = PVE::QemuConfig->load_config($vmid);
5833
5834 my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5835 PVE::QemuConfig->check_lock($conf)
5836 if !($skiplock || $is_backing_up);
5837
5838 die "cannot suspend to disk during backup\n"
5839 if $is_backing_up && $includestate;
5840
5841 if ($includestate) {
5842 $conf->{lock} = 'suspending';
5843 my $date = strftime("%Y-%m-%d", localtime(time()));
5844 $storecfg = PVE::Storage::config();
5845 $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5846 $path = PVE::Storage::path($storecfg, $vmstate);
5847 PVE::QemuConfig->write_config($vmid, $conf);
5848 } else {
5849 vm_mon_cmd($vmid, "stop");
5850 }
5851 });
5852
5853 if ($includestate) {
5854 # save vm state
5855 PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5856
5857 eval {
5858 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5859 for(;;) {
5860 my $state = vm_mon_cmd_nocheck($vmid, "query-savevm");
5861 if (!$state->{status}) {
5862 die "savevm not active\n";
5863 } elsif ($state->{status} eq 'active') {
5864 sleep(1);
5865 next;
5866 } elsif ($state->{status} eq 'completed') {
5867 print "State saved, quitting\n";
5868 last;
5869 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5870 die "query-savevm failed with error '$state->{error}'\n"
5871 } else {
5872 die "query-savevm returned status '$state->{status}'\n";
5873 }
5874 }
5875 };
5876 my $err = $@;
5877
5878 PVE::QemuConfig->lock_config($vmid, sub {
5879 $conf = PVE::QemuConfig->load_config($vmid);
5880 if ($err) {
5881 # cleanup, but leave suspending lock, to indicate something went wrong
5882 eval {
5883 vm_mon_cmd($vmid, "savevm-end");
5884 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5885 PVE::Storage::vdisk_free($storecfg, $vmstate);
5886 delete $conf->@{qw(vmstate runningmachine)};
5887 PVE::QemuConfig->write_config($vmid, $conf);
5888 };
5889 warn $@ if $@;
5890 die $err;
5891 }
5892
5893 die "lock changed unexpectedly\n"
5894 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5895
5896 vm_qmp_command($vmid, { execute => "quit" });
5897 $conf->{lock} = 'suspended';
5898 PVE::QemuConfig->write_config($vmid, $conf);
5899 });
5900 }
5901 }
5902
5903 sub vm_resume {
5904 my ($vmid, $skiplock, $nocheck) = @_;
5905
5906 PVE::QemuConfig->lock_config($vmid, sub {
5907 my $vm_mon_cmd = $nocheck ? \&vm_mon_cmd_nocheck : \&vm_mon_cmd;
5908 my $res = $vm_mon_cmd->($vmid, 'query-status');
5909 my $resume_cmd = 'cont';
5910
5911 if ($res->{status} && $res->{status} eq 'suspended') {
5912 $resume_cmd = 'system_wakeup';
5913 }
5914
5915 if (!$nocheck) {
5916
5917 my $conf = PVE::QemuConfig->load_config($vmid);
5918
5919 PVE::QemuConfig->check_lock($conf)
5920 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5921 }
5922
5923 $vm_mon_cmd->($vmid, $resume_cmd);
5924 });
5925 }
5926
5927 sub vm_sendkey {
5928 my ($vmid, $skiplock, $key) = @_;
5929
5930 PVE::QemuConfig->lock_config($vmid, sub {
5931
5932 my $conf = PVE::QemuConfig->load_config($vmid);
5933
5934 # there is no qmp command, so we use the human monitor command
5935 my $res = vm_human_monitor_command($vmid, "sendkey $key");
5936 die $res if $res ne '';
5937 });
5938 }
5939
5940 # vzdump restore implementaion
5941
5942 sub tar_archive_read_firstfile {
5943 my $archive = shift;
5944
5945 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5946
5947 # try to detect archive type first
5948 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
5949 die "unable to open file '$archive'\n";
5950 my $firstfile = <$fh>;
5951 kill 15, $pid;
5952 close $fh;
5953
5954 die "ERROR: archive contaions no data\n" if !$firstfile;
5955 chomp $firstfile;
5956
5957 return $firstfile;
5958 }
5959
5960 sub tar_restore_cleanup {
5961 my ($storecfg, $statfile) = @_;
5962
5963 print STDERR "starting cleanup\n";
5964
5965 if (my $fd = IO::File->new($statfile, "r")) {
5966 while (defined(my $line = <$fd>)) {
5967 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5968 my $volid = $2;
5969 eval {
5970 if ($volid =~ m|^/|) {
5971 unlink $volid || die 'unlink failed\n';
5972 } else {
5973 PVE::Storage::vdisk_free($storecfg, $volid);
5974 }
5975 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5976 };
5977 print STDERR "unable to cleanup '$volid' - $@" if $@;
5978 } else {
5979 print STDERR "unable to parse line in statfile - $line";
5980 }
5981 }
5982 $fd->close();
5983 }
5984 }
5985
5986 sub restore_archive {
5987 my ($archive, $vmid, $user, $opts) = @_;
5988
5989 my $format = $opts->{format};
5990 my $comp;
5991
5992 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5993 $format = 'tar' if !$format;
5994 $comp = 'gzip';
5995 } elsif ($archive =~ m/\.tar$/) {
5996 $format = 'tar' if !$format;
5997 } elsif ($archive =~ m/.tar.lzo$/) {
5998 $format = 'tar' if !$format;
5999 $comp = 'lzop';
6000 } elsif ($archive =~ m/\.vma$/) {
6001 $format = 'vma' if !$format;
6002 } elsif ($archive =~ m/\.vma\.gz$/) {
6003 $format = 'vma' if !$format;
6004 $comp = 'gzip';
6005 } elsif ($archive =~ m/\.vma\.lzo$/) {
6006 $format = 'vma' if !$format;
6007 $comp = 'lzop';
6008 } else {
6009 $format = 'vma' if !$format; # default
6010 }
6011
6012 # try to detect archive format
6013 if ($format eq 'tar') {
6014 return restore_tar_archive($archive, $vmid, $user, $opts);
6015 } else {
6016 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
6017 }
6018 }
6019
6020 sub restore_update_config_line {
6021 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
6022
6023 return if $line =~ m/^\#qmdump\#/;
6024 return if $line =~ m/^\#vzdump\#/;
6025 return if $line =~ m/^lock:/;
6026 return if $line =~ m/^unused\d+:/;
6027 return if $line =~ m/^parent:/;
6028
6029 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
6030 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6031 # try to convert old 1.X settings
6032 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6033 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
6034 my ($model, $macaddr) = split(/\=/, $devconfig);
6035 $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
6036 my $net = {
6037 model => $model,
6038 bridge => "vmbr$ind",
6039 macaddr => $macaddr,
6040 };
6041 my $netstr = print_net($net);
6042
6043 print $outfd "net$cookie->{netcount}: $netstr\n";
6044 $cookie->{netcount}++;
6045 }
6046 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6047 my ($id, $netstr) = ($1, $2);
6048 my $net = parse_net($netstr);
6049 $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
6050 $netstr = print_net($net);
6051 print $outfd "$id: $netstr\n";
6052 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
6053 my $virtdev = $1;
6054 my $value = $3;
6055 my $di = parse_drive($virtdev, $value);
6056 if (defined($di->{backup}) && !$di->{backup}) {
6057 print $outfd "#$line";
6058 } elsif ($map->{$virtdev}) {
6059 delete $di->{format}; # format can change on restore
6060 $di->{file} = $map->{$virtdev};
6061 $value = print_drive($vmid, $di);
6062 print $outfd "$virtdev: $value\n";
6063 } else {
6064 print $outfd $line;
6065 }
6066 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6067 my $vmgenid = $1;
6068 if ($vmgenid ne '0') {
6069 # always generate a new vmgenid if there was a valid one setup
6070 $vmgenid = generate_uuid();
6071 }
6072 print $outfd "vmgenid: $vmgenid\n";
6073 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6074 my ($uuid, $uuid_str);
6075 UUID::generate($uuid);
6076 UUID::unparse($uuid, $uuid_str);
6077 my $smbios1 = parse_smbios1($2);
6078 $smbios1->{uuid} = $uuid_str;
6079 print $outfd $1.print_smbios1($smbios1)."\n";
6080 } else {
6081 print $outfd $line;
6082 }
6083 }
6084
6085 sub scan_volids {
6086 my ($cfg, $vmid) = @_;
6087
6088 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
6089
6090 my $volid_hash = {};
6091 foreach my $storeid (keys %$info) {
6092 foreach my $item (@{$info->{$storeid}}) {
6093 next if !($item->{volid} && $item->{size});
6094 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
6095 $volid_hash->{$item->{volid}} = $item;
6096 }
6097 }
6098
6099 return $volid_hash;
6100 }
6101
6102 sub is_volume_in_use {
6103 my ($storecfg, $conf, $skip_drive, $volid) = @_;
6104
6105 my $path = PVE::Storage::path($storecfg, $volid);
6106
6107 my $scan_config = sub {
6108 my ($cref, $snapname) = @_;
6109
6110 foreach my $key (keys %$cref) {
6111 my $value = $cref->{$key};
6112 if (is_valid_drivename($key)) {
6113 next if $skip_drive && $key eq $skip_drive;
6114 my $drive = parse_drive($key, $value);
6115 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
6116 return 1 if $volid eq $drive->{file};
6117 if ($drive->{file} =~ m!^/!) {
6118 return 1 if $drive->{file} eq $path;
6119 } else {
6120 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
6121 next if !$storeid;
6122 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
6123 next if !$scfg;
6124 return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
6125 }
6126 }
6127 }
6128
6129 return 0;
6130 };
6131
6132 return 1 if &$scan_config($conf);
6133
6134 undef $skip_drive;
6135
6136 foreach my $snapname (keys %{$conf->{snapshots}}) {
6137 return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
6138 }
6139
6140 return 0;
6141 }
6142
6143 sub update_disksize {
6144 my ($vmid, $conf, $volid_hash) = @_;
6145
6146 my $changes;
6147 my $prefix = "VM $vmid:";
6148
6149 # used and unused disks
6150 my $referenced = {};
6151
6152 # Note: it is allowed to define multiple storages with same path (alias), so
6153 # we need to check both 'volid' and real 'path' (two different volid can point
6154 # to the same path).
6155
6156 my $referencedpath = {};
6157
6158 # update size info
6159 foreach my $opt (keys %$conf) {
6160 if (is_valid_drivename($opt)) {
6161 my $drive = parse_drive($opt, $conf->{$opt});
6162 my $volid = $drive->{file};
6163 next if !$volid;
6164
6165 $referenced->{$volid} = 1;
6166 if ($volid_hash->{$volid} &&
6167 (my $path = $volid_hash->{$volid}->{path})) {
6168 $referencedpath->{$path} = 1;
6169 }
6170
6171 next if drive_is_cdrom($drive);
6172 next if !$volid_hash->{$volid};
6173
6174 $drive->{size} = $volid_hash->{$volid}->{size};
6175 my $new = print_drive($vmid, $drive);
6176 if ($new ne $conf->{$opt}) {
6177 $changes = 1;
6178 $conf->{$opt} = $new;
6179 print "$prefix update disk '$opt' information.\n";
6180 }
6181 }
6182 }
6183
6184 # remove 'unusedX' entry if volume is used
6185 foreach my $opt (keys %$conf) {
6186 next if $opt !~ m/^unused\d+$/;
6187 my $volid = $conf->{$opt};
6188 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
6189 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6190 print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
6191 $changes = 1;
6192 delete $conf->{$opt};
6193 }
6194
6195 $referenced->{$volid} = 1;
6196 $referencedpath->{$path} = 1 if $path;
6197 }
6198
6199 foreach my $volid (sort keys %$volid_hash) {
6200 next if $volid =~ m/vm-$vmid-state-/;
6201 next if $referenced->{$volid};
6202 my $path = $volid_hash->{$volid}->{path};
6203 next if !$path; # just to be sure
6204 next if $referencedpath->{$path};
6205 $changes = 1;
6206 my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
6207 print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
6208 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6209 }
6210
6211 return $changes;
6212 }
6213
6214 sub rescan {
6215 my ($vmid, $nolock, $dryrun) = @_;
6216
6217 my $cfg = PVE::Storage::config();
6218
6219 # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
6220 # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
6221 foreach my $stor (keys %{$cfg->{ids}}) {
6222 delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
6223 }
6224
6225 print "rescan volumes...\n";
6226 my $volid_hash = scan_volids($cfg, $vmid);
6227
6228 my $updatefn = sub {
6229 my ($vmid) = @_;
6230
6231 my $conf = PVE::QemuConfig->load_config($vmid);
6232
6233 PVE::QemuConfig->check_lock($conf);
6234
6235 my $vm_volids = {};
6236 foreach my $volid (keys %$volid_hash) {
6237 my $info = $volid_hash->{$volid};
6238 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
6239 }
6240
6241 my $changes = update_disksize($vmid, $conf, $vm_volids);
6242
6243 PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
6244 };
6245
6246 if (defined($vmid)) {
6247 if ($nolock) {
6248 &$updatefn($vmid);
6249 } else {
6250 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6251 }
6252 } else {
6253 my $vmlist = config_list();
6254 foreach my $vmid (keys %$vmlist) {
6255 if ($nolock) {
6256 &$updatefn($vmid);
6257 } else {
6258 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6259 }
6260 }
6261 }
6262 }
6263
6264 sub restore_vma_archive {
6265 my ($archive, $vmid, $user, $opts, $comp) = @_;
6266
6267 my $readfrom = $archive;
6268
6269 my $cfg = PVE::Storage::config();
6270 my $commands = [];
6271 my $bwlimit = $opts->{bwlimit};
6272
6273 my $dbg_cmdstring = '';
6274 my $add_pipe = sub {
6275 my ($cmd) = @_;
6276 push @$commands, $cmd;
6277 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6278 $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6279 $readfrom = '-';
6280 };
6281
6282 my $input = undef;
6283 if ($archive eq '-') {
6284 $input = '<&STDIN';
6285 } else {
6286 # If we use a backup from a PVE defined storage we also consider that
6287 # storage's rate limit:
6288 my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6289 if (defined($volid)) {
6290 my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6291 my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6292 if ($readlimit) {
6293 print STDERR "applying read rate limit: $readlimit\n";
6294 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6295 $add_pipe->($cstream);
6296 }
6297 }
6298 }
6299
6300 if ($comp) {
6301 my $cmd;
6302 if ($comp eq 'gzip') {
6303 $cmd = ['zcat', $readfrom];
6304 } elsif ($comp eq 'lzop') {
6305 $cmd = ['lzop', '-d', '-c', $readfrom];
6306 } else {
6307 die "unknown compression method '$comp'\n";
6308 }
6309 $add_pipe->($cmd);
6310 }
6311
6312 my $tmpdir = "/var/tmp/vzdumptmp$$";
6313 rmtree $tmpdir;
6314
6315 # disable interrupts (always do cleanups)
6316 local $SIG{INT} =
6317 local $SIG{TERM} =
6318 local $SIG{QUIT} =
6319 local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6320
6321 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6322 POSIX::mkfifo($mapfifo, 0600);
6323 my $fifofh;
6324
6325 my $openfifo = sub {
6326 open($fifofh, '>', $mapfifo) || die $!;
6327 };
6328
6329 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6330
6331 my $oldtimeout;
6332 my $timeout = 5;
6333
6334 my $devinfo = {};
6335
6336 my $rpcenv = PVE::RPCEnvironment::get();
6337
6338 my $conffile = PVE::QemuConfig->config_file($vmid);
6339 my $tmpfn = "$conffile.$$.tmp";
6340
6341 # Note: $oldconf is undef if VM does not exists
6342 my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6343 my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6344
6345 my %storage_limits;
6346
6347 my $print_devmap = sub {
6348 my $virtdev_hash = {};
6349
6350 my $cfgfn = "$tmpdir/qemu-server.conf";
6351
6352 # we can read the config - that is already extracted
6353 my $fh = IO::File->new($cfgfn, "r") ||
6354 "unable to read qemu-server.conf - $!\n";
6355
6356 my $fwcfgfn = "$tmpdir/qemu-server.fw";
6357 if (-f $fwcfgfn) {
6358 my $pve_firewall_dir = '/etc/pve/firewall';
6359 mkdir $pve_firewall_dir; # make sure the dir exists
6360 PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6361 }
6362
6363 while (defined(my $line = <$fh>)) {
6364 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6365 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6366 die "archive does not contain data for drive '$virtdev'\n"
6367 if !$devinfo->{$devname};
6368 if (defined($opts->{storage})) {
6369 $storeid = $opts->{storage} || 'local';
6370 } elsif (!$storeid) {
6371 $storeid = 'local';
6372 }
6373 $format = 'raw' if !$format;
6374 $devinfo->{$devname}->{devname} = $devname;
6375 $devinfo->{$devname}->{virtdev} = $virtdev;
6376 $devinfo->{$devname}->{format} = $format;
6377 $devinfo->{$devname}->{storeid} = $storeid;
6378
6379 # check permission on storage
6380 my $pool = $opts->{pool}; # todo: do we need that?
6381 if ($user ne 'root@pam') {
6382 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6383 }
6384
6385 $storage_limits{$storeid} = $bwlimit;
6386
6387 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6388 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6389 my $virtdev = $1;
6390 my $drive = parse_drive($virtdev, $2);
6391 if (drive_is_cloudinit($drive)) {
6392 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6393 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6394 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
6395
6396 my $d = {
6397 format => $format,
6398 storeid => $opts->{storage} // $storeid,
6399 size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
6400 file => $drive->{file}, # to make drive_is_cloudinit check possible
6401 name => "vm-$vmid-cloudinit",
6402 is_cloudinit => 1,
6403 };
6404 $virtdev_hash->{$virtdev} = $d;
6405 }
6406 }
6407 }
6408
6409 foreach my $key (keys %storage_limits) {
6410 my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6411 next if !$limit;
6412 print STDERR "rate limit for storage $key: $limit KiB/s\n";
6413 $storage_limits{$key} = $limit * 1024;
6414 }
6415
6416 foreach my $devname (keys %$devinfo) {
6417 die "found no device mapping information for device '$devname'\n"
6418 if !$devinfo->{$devname}->{virtdev};
6419 }
6420
6421 # create empty/temp config
6422 if ($oldconf) {
6423 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6424 foreach_drive($oldconf, sub {
6425 my ($ds, $drive) = @_;
6426
6427 return if drive_is_cdrom($drive, 1);
6428
6429 my $volid = $drive->{file};
6430 return if !$volid || $volid =~ m|^/|;
6431
6432 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6433 return if !$path || !$owner || ($owner != $vmid);
6434
6435 # Note: only delete disk we want to restore
6436 # other volumes will become unused
6437 if ($virtdev_hash->{$ds}) {
6438 eval { PVE::Storage::vdisk_free($cfg, $volid); };
6439 if (my $err = $@) {
6440 warn $err;
6441 }
6442 }
6443 });
6444
6445 # delete vmstate files, after the restore we have no snapshots anymore
6446 foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6447 my $snap = $oldconf->{snapshots}->{$snapname};
6448 if ($snap->{vmstate}) {
6449 eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6450 if (my $err = $@) {
6451 warn $err;
6452 }
6453 }
6454 }
6455 }
6456
6457 my $map = {};
6458 foreach my $virtdev (sort keys %$virtdev_hash) {
6459 my $d = $virtdev_hash->{$virtdev};
6460 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6461 my $storeid = $d->{storeid};
6462 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6463
6464 my $map_opts = '';
6465 if (my $limit = $storage_limits{$storeid}) {
6466 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6467 }
6468
6469 # test if requested format is supported
6470 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6471 my $supported = grep { $_ eq $d->{format} } @$validFormats;
6472 $d->{format} = $defFormat if !$supported;
6473
6474 my $name;
6475 if ($d->{is_cloudinit}) {
6476 $name = $d->{name};
6477 $name .= ".$d->{format}" if $d->{format} ne 'raw';
6478 }
6479
6480 my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
6481 print STDERR "new volume ID is '$volid'\n";
6482 $d->{volid} = $volid;
6483
6484 PVE::Storage::activate_volumes($cfg, [$volid]);
6485
6486 my $write_zeros = 1;
6487 if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6488 $write_zeros = 0;
6489 }
6490
6491 if (!$d->{is_cloudinit}) {
6492 my $path = PVE::Storage::path($cfg, $volid);
6493
6494 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6495
6496 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6497 }
6498 $map->{$virtdev} = $volid;
6499 }
6500
6501 $fh->seek(0, 0) || die "seek failed - $!\n";
6502
6503 my $outfd = new IO::File ($tmpfn, "w") ||
6504 die "unable to write config for VM $vmid\n";
6505
6506 my $cookie = { netcount => 0 };
6507 while (defined(my $line = <$fh>)) {
6508 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6509 }
6510
6511 $fh->close();
6512 $outfd->close();
6513 };
6514
6515 eval {
6516 # enable interrupts
6517 local $SIG{INT} =
6518 local $SIG{TERM} =
6519 local $SIG{QUIT} =
6520 local $SIG{HUP} =
6521 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6522 local $SIG{ALRM} = sub { die "got timeout\n"; };
6523
6524 $oldtimeout = alarm($timeout);
6525
6526 my $parser = sub {
6527 my $line = shift;
6528
6529 print "$line\n";
6530
6531 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6532 my ($dev_id, $size, $devname) = ($1, $2, $3);
6533 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6534 } elsif ($line =~ m/^CTIME: /) {
6535 # we correctly received the vma config, so we can disable
6536 # the timeout now for disk allocation (set to 10 minutes, so
6537 # that we always timeout if something goes wrong)
6538 alarm(600);
6539 &$print_devmap();
6540 print $fifofh "done\n";
6541 my $tmp = $oldtimeout || 0;
6542 $oldtimeout = undef;
6543 alarm($tmp);
6544 close($fifofh);
6545 }
6546 };
6547
6548 print "restore vma archive: $dbg_cmdstring\n";
6549 run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6550 };
6551 my $err = $@;
6552
6553 alarm($oldtimeout) if $oldtimeout;
6554
6555 my $vollist = [];
6556 foreach my $devname (keys %$devinfo) {
6557 my $volid = $devinfo->{$devname}->{volid};
6558 push @$vollist, $volid if $volid;
6559 }
6560
6561 PVE::Storage::deactivate_volumes($cfg, $vollist);
6562
6563 unlink $mapfifo;
6564
6565 if ($err) {
6566 rmtree $tmpdir;
6567 unlink $tmpfn;
6568
6569 foreach my $devname (keys %$devinfo) {
6570 my $volid = $devinfo->{$devname}->{volid};
6571 next if !$volid;
6572 eval {
6573 if ($volid =~ m|^/|) {
6574 unlink $volid || die 'unlink failed\n';
6575 } else {
6576 PVE::Storage::vdisk_free($cfg, $volid);
6577 }
6578 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6579 };
6580 print STDERR "unable to cleanup '$volid' - $@" if $@;
6581 }
6582 die $err;
6583 }
6584
6585 rmtree $tmpdir;
6586
6587 rename($tmpfn, $conffile) ||
6588 die "unable to commit configuration file '$conffile'\n";
6589
6590 PVE::Cluster::cfs_update(); # make sure we read new file
6591
6592 eval { rescan($vmid, 1); };
6593 warn $@ if $@;
6594 }
6595
6596 sub restore_tar_archive {
6597 my ($archive, $vmid, $user, $opts) = @_;
6598
6599 if ($archive ne '-') {
6600 my $firstfile = tar_archive_read_firstfile($archive);
6601 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6602 if $firstfile ne 'qemu-server.conf';
6603 }
6604
6605 my $storecfg = PVE::Storage::config();
6606
6607 # destroy existing data - keep empty config
6608 my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6609 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6610
6611 my $tocmd = "/usr/lib/qemu-server/qmextract";
6612
6613 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6614 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6615 $tocmd .= ' --prealloc' if $opts->{prealloc};
6616 $tocmd .= ' --info' if $opts->{info};
6617
6618 # tar option "xf" does not autodetect compression when read from STDIN,
6619 # so we pipe to zcat
6620 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6621 PVE::Tools::shellquote("--to-command=$tocmd");
6622
6623 my $tmpdir = "/var/tmp/vzdumptmp$$";
6624 mkpath $tmpdir;
6625
6626 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6627 local $ENV{VZDUMP_VMID} = $vmid;
6628 local $ENV{VZDUMP_USER} = $user;
6629
6630 my $conffile = PVE::QemuConfig->config_file($vmid);
6631 my $tmpfn = "$conffile.$$.tmp";
6632
6633 # disable interrupts (always do cleanups)
6634 local $SIG{INT} =
6635 local $SIG{TERM} =
6636 local $SIG{QUIT} =
6637 local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6638
6639 eval {
6640 # enable interrupts
6641 local $SIG{INT} =
6642 local $SIG{TERM} =
6643 local $SIG{QUIT} =
6644 local $SIG{HUP} =
6645 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6646
6647 if ($archive eq '-') {
6648 print "extracting archive from STDIN\n";
6649 run_command($cmd, input => "<&STDIN");
6650 } else {
6651 print "extracting archive '$archive'\n";
6652 run_command($cmd);
6653 }
6654
6655 return if $opts->{info};
6656
6657 # read new mapping
6658 my $map = {};
6659 my $statfile = "$tmpdir/qmrestore.stat";
6660 if (my $fd = IO::File->new($statfile, "r")) {
6661 while (defined (my $line = <$fd>)) {
6662 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6663 $map->{$1} = $2 if $1;
6664 } else {
6665 print STDERR "unable to parse line in statfile - $line\n";
6666 }
6667 }
6668 $fd->close();
6669 }
6670
6671 my $confsrc = "$tmpdir/qemu-server.conf";
6672
6673 my $srcfd = new IO::File($confsrc, "r") ||
6674 die "unable to open file '$confsrc'\n";
6675
6676 my $outfd = new IO::File ($tmpfn, "w") ||
6677 die "unable to write config for VM $vmid\n";
6678
6679 my $cookie = { netcount => 0 };
6680 while (defined (my $line = <$srcfd>)) {
6681 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6682 }
6683
6684 $srcfd->close();
6685 $outfd->close();
6686 };
6687 my $err = $@;
6688
6689 if ($err) {
6690
6691 unlink $tmpfn;
6692
6693 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6694
6695 die $err;
6696 }
6697
6698 rmtree $tmpdir;
6699
6700 rename $tmpfn, $conffile ||
6701 die "unable to commit configuration file '$conffile'\n";
6702
6703 PVE::Cluster::cfs_update(); # make sure we read new file
6704
6705 eval { rescan($vmid, 1); };
6706 warn $@ if $@;
6707 };
6708
6709 sub foreach_storage_used_by_vm {
6710 my ($conf, $func) = @_;
6711
6712 my $sidhash = {};
6713
6714 foreach_drive($conf, sub {
6715 my ($ds, $drive) = @_;
6716 return if drive_is_cdrom($drive);
6717
6718 my $volid = $drive->{file};
6719
6720 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6721 $sidhash->{$sid} = $sid if $sid;
6722 });
6723
6724 foreach my $sid (sort keys %$sidhash) {
6725 &$func($sid);
6726 }
6727 }
6728
6729 sub do_snapshots_with_qemu {
6730 my ($storecfg, $volid) = @_;
6731
6732 my $storage_name = PVE::Storage::parse_volume_id($volid);
6733 my $scfg = $storecfg->{ids}->{$storage_name};
6734
6735 if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6736 return 1;
6737 }
6738
6739 if ($volid =~ m/\.(qcow2|qed)$/){
6740 return 1;
6741 }
6742
6743 return undef;
6744 }
6745
6746 sub qga_check_running {
6747 my ($vmid, $nowarn) = @_;
6748
6749 eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6750 if ($@) {
6751 warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6752 return 0;
6753 }
6754 return 1;
6755 }
6756
6757 sub template_create {
6758 my ($vmid, $conf, $disk) = @_;
6759
6760 my $storecfg = PVE::Storage::config();
6761
6762 foreach_drive($conf, sub {
6763 my ($ds, $drive) = @_;
6764
6765 return if drive_is_cdrom($drive);
6766 return if $disk && $ds ne $disk;
6767
6768 my $volid = $drive->{file};
6769 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6770
6771 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6772 $drive->{file} = $voliddst;
6773 $conf->{$ds} = print_drive($vmid, $drive);
6774 PVE::QemuConfig->write_config($vmid, $conf);
6775 });
6776 }
6777
6778 sub convert_iscsi_path {
6779 my ($path) = @_;
6780
6781 if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6782 my $portal = $1;
6783 my $target = $2;
6784 my $lun = $3;
6785
6786 my $initiator_name = get_initiator_name();
6787
6788 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6789 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6790 }
6791
6792 die "cannot convert iscsi path '$path', unkown format\n";
6793 }
6794
6795 sub qemu_img_convert {
6796 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6797
6798 my $storecfg = PVE::Storage::config();
6799 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6800 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6801
6802 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6803
6804 my $cachemode;
6805 my $src_path;
6806 my $src_is_iscsi = 0;
6807 my $src_format = 'raw';
6808
6809 if ($src_storeid) {
6810 PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6811 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6812 $src_format = qemu_img_format($src_scfg, $src_volname);
6813 $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6814 $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6815 $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6816 } elsif (-f $src_volid) {
6817 $src_path = $src_volid;
6818 if ($src_path =~ m/\.($QEMU_FORMAT_RE)$/) {
6819 $src_format = $1;
6820 }
6821 }
6822
6823 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6824
6825 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6826 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6827 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6828 my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6829
6830 my $cmd = [];
6831 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6832 push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6833 push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6834 push @$cmd, '-T', $cachemode if defined($cachemode);
6835
6836 if ($src_is_iscsi) {
6837 push @$cmd, '--image-opts';
6838 $src_path = convert_iscsi_path($src_path);
6839 } else {
6840 push @$cmd, '-f', $src_format;
6841 }
6842
6843 if ($dst_is_iscsi) {
6844 push @$cmd, '--target-image-opts';
6845 $dst_path = convert_iscsi_path($dst_path);
6846 } else {
6847 push @$cmd, '-O', $dst_format;
6848 }
6849
6850 push @$cmd, $src_path;
6851
6852 if (!$dst_is_iscsi && $is_zero_initialized) {
6853 push @$cmd, "zeroinit:$dst_path";
6854 } else {
6855 push @$cmd, $dst_path;
6856 }
6857
6858 my $parser = sub {
6859 my $line = shift;
6860 if($line =~ m/\((\S+)\/100\%\)/){
6861 my $percent = $1;
6862 my $transferred = int($size * $percent / 100);
6863 my $remaining = $size - $transferred;
6864
6865 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6866 }
6867
6868 };
6869
6870 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6871 my $err = $@;
6872 die "copy failed: $err" if $err;
6873 }
6874
6875 sub qemu_img_format {
6876 my ($scfg, $volname) = @_;
6877
6878 if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6879 return $1;
6880 } else {
6881 return "raw";
6882 }
6883 }
6884
6885 sub qemu_drive_mirror {
6886 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
6887
6888 $jobs = {} if !$jobs;
6889
6890 my $qemu_target;
6891 my $format;
6892 $jobs->{"drive-$drive"} = {};
6893
6894 if ($dst_volid =~ /^nbd:/) {
6895 $qemu_target = $dst_volid;
6896 $format = "nbd";
6897 } else {
6898 my $storecfg = PVE::Storage::config();
6899 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6900
6901 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6902
6903 $format = qemu_img_format($dst_scfg, $dst_volname);
6904
6905 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6906
6907 $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6908 }
6909
6910 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6911 $opts->{format} = $format if $format;
6912
6913 if (defined($bwlimit)) {
6914 $opts->{speed} = $bwlimit * 1024;
6915 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
6916 } else {
6917 print "drive mirror is starting for drive-$drive\n";
6918 }
6919
6920 # if a job already runs for this device we get an error, catch it for cleanup
6921 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
6922 if (my $err = $@) {
6923 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
6924 warn "$@\n" if $@;
6925 die "mirroring error: $err\n";
6926 }
6927
6928 qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
6929 }
6930
6931 sub qemu_drive_mirror_monitor {
6932 my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
6933
6934 eval {
6935 my $err_complete = 0;
6936
6937 while (1) {
6938 die "storage migration timed out\n" if $err_complete > 300;
6939
6940 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6941
6942 my $running_mirror_jobs = {};
6943 foreach my $stat (@$stats) {
6944 next if $stat->{type} ne 'mirror';
6945 $running_mirror_jobs->{$stat->{device}} = $stat;
6946 }
6947
6948 my $readycounter = 0;
6949
6950 foreach my $job (keys %$jobs) {
6951
6952 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
6953 print "$job : finished\n";
6954 delete $jobs->{$job};
6955 next;
6956 }
6957
6958 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
6959
6960 my $busy = $running_mirror_jobs->{$job}->{busy};
6961 my $ready = $running_mirror_jobs->{$job}->{ready};
6962 if (my $total = $running_mirror_jobs->{$job}->{len}) {
6963 my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
6964 my $remaining = $total - $transferred;
6965 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6966
6967 print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6968 }
6969
6970 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
6971 }
6972
6973 last if scalar(keys %$jobs) == 0;
6974
6975 if ($readycounter == scalar(keys %$jobs)) {
6976 print "all mirroring jobs are ready \n";
6977 last if $skipcomplete; #do the complete later
6978
6979 if ($vmiddst && $vmiddst != $vmid) {
6980 my $agent_running = $qga && qga_check_running($vmid);
6981 if ($agent_running) {
6982 print "freeze filesystem\n";
6983 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
6984 } else {
6985 print "suspend vm\n";
6986 eval { PVE::QemuServer::vm_suspend($vmid, 1); };
6987 }
6988
6989 # if we clone a disk for a new target vm, we don't switch the disk
6990 PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
6991
6992 if ($agent_running) {
6993 print "unfreeze filesystem\n";
6994 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
6995 } else {
6996 print "resume vm\n";
6997 eval { PVE::QemuServer::vm_resume($vmid, 1, 1); };
6998 }
6999
7000 last;
7001 } else {
7002
7003 foreach my $job (keys %$jobs) {
7004 # try to switch the disk if source and destination are on the same guest
7005 print "$job: Completing block job...\n";
7006
7007 eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
7008 if ($@ =~ m/cannot be completed/) {
7009 print "$job: Block job cannot be completed, try again.\n";
7010 $err_complete++;
7011 }else {
7012 print "$job: Completed successfully.\n";
7013 $jobs->{$job}->{complete} = 1;
7014 }
7015 }
7016 }
7017 }
7018 sleep 1;
7019 }
7020 };
7021 my $err = $@;
7022
7023 if ($err) {
7024 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7025 die "mirroring error: $err";
7026 }
7027
7028 }
7029
7030 sub qemu_blockjobs_cancel {
7031 my ($vmid, $jobs) = @_;
7032
7033 foreach my $job (keys %$jobs) {
7034 print "$job: Cancelling block job\n";
7035 eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
7036 $jobs->{$job}->{cancel} = 1;
7037 }
7038
7039 while (1) {
7040 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7041
7042 my $running_jobs = {};
7043 foreach my $stat (@$stats) {
7044 $running_jobs->{$stat->{device}} = $stat;
7045 }
7046
7047 foreach my $job (keys %$jobs) {
7048
7049 if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
7050 print "$job: Done.\n";
7051 delete $jobs->{$job};
7052 }
7053 }
7054
7055 last if scalar(keys %$jobs) == 0;
7056
7057 sleep 1;
7058 }
7059 }
7060
7061 sub clone_disk {
7062 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
7063 $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
7064
7065 my $newvolid;
7066
7067 if (!$full) {
7068 print "create linked clone of drive $drivename ($drive->{file})\n";
7069 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
7070 push @$newvollist, $newvolid;
7071 } else {
7072
7073 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
7074 $storeid = $storage if $storage;
7075
7076 my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
7077 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
7078
7079 print "create full clone of drive $drivename ($drive->{file})\n";
7080 my $name = undef;
7081 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
7082 push @$newvollist, $newvolid;
7083
7084 PVE::Storage::activate_volumes($storecfg, [$newvolid]);
7085
7086 my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
7087 if (!$running || $snapname) {
7088 # TODO: handle bwlimits
7089 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
7090 } else {
7091
7092 my $kvmver = get_running_qemu_version ($vmid);
7093 if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
7094 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
7095 if $drive->{iothread};
7096 }
7097
7098 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga, $bwlimit);
7099 }
7100 }
7101
7102 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
7103
7104 my $disk = $drive;
7105 $disk->{format} = undef;
7106 $disk->{file} = $newvolid;
7107 $disk->{size} = $size;
7108
7109 return $disk;
7110 }
7111
7112 # this only works if VM is running
7113 sub get_current_qemu_machine {
7114 my ($vmid) = @_;
7115
7116 my $cmd = { execute => 'query-machines', arguments => {} };
7117 my $res = vm_qmp_command($vmid, $cmd);
7118
7119 my ($current, $default);
7120 foreach my $e (@$res) {
7121 $default = $e->{name} if $e->{'is-default'};
7122 $current = $e->{name} if $e->{'is-current'};
7123 }
7124
7125 # fallback to the default machine if current is not supported by qemu
7126 return $current || $default || 'pc';
7127 }
7128
7129 sub get_running_qemu_version {
7130 my ($vmid) = @_;
7131 my $cmd = { execute => 'query-version', arguments => {} };
7132 my $res = vm_qmp_command($vmid, $cmd);
7133 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
7134 }
7135
7136 sub qemu_machine_feature_enabled {
7137 my ($machine, $kvmver, $version_major, $version_minor) = @_;
7138
7139 my $current_major;
7140 my $current_minor;
7141
7142 if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
7143
7144 $current_major = $3;
7145 $current_minor = $4;
7146
7147 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
7148
7149 $current_major = $1;
7150 $current_minor = $2;
7151 }
7152
7153 return 1 if version_cmp($current_major, $version_major, $current_minor, $version_minor) >= 0;
7154 }
7155
7156 # gets in pairs the versions you want to compares, i.e.:
7157 # ($a-major, $b-major, $a-minor, $b-minor, $a-extra, $b-extra, ...)
7158 # returns 0 if same, -1 if $a is older than $b, +1 if $a is newer than $b
7159 sub version_cmp {
7160 my @versions = @_;
7161
7162 my $size = scalar(@versions);
7163
7164 return 0 if $size == 0;
7165 die "cannot compare odd count of versions" if $size & 1;
7166
7167 for (my $i = 0; $i < $size; $i += 2) {
7168 my ($a, $b) = splice(@versions, 0, 2);
7169 $a //= 0;
7170 $b //= 0;
7171
7172 return 1 if $a > $b;
7173 return -1 if $a < $b;
7174 }
7175 return 0;
7176 }
7177
7178 # dies if a) VM not running or not exisiting b) Version query failed
7179 # So, any defined return value is valid, any invalid state can be caught by eval
7180 sub runs_at_least_qemu_version {
7181 my ($vmid, $major, $minor, $extra) = @_;
7182
7183 my $v = vm_qmp_command($vmid, { execute => 'query-version' });
7184 die "could not query currently running version for VM $vmid\n" if !defined($v);
7185 $v = $v->{qemu};
7186
7187 return version_cmp($v->{major}, $major, $v->{minor}, $minor, $v->{micro}, $extra) >= 0;
7188 }
7189
7190 sub qemu_machine_pxe {
7191 my ($vmid, $conf) = @_;
7192
7193 my $machine = PVE::QemuServer::get_current_qemu_machine($vmid);
7194
7195 if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
7196 $machine .= '.pxe';
7197 }
7198
7199 return $machine;
7200 }
7201
7202 sub qemu_use_old_bios_files {
7203 my ($machine_type) = @_;
7204
7205 return if !$machine_type;
7206
7207 my $use_old_bios_files = undef;
7208
7209 if ($machine_type =~ m/^(\S+)\.pxe$/) {
7210 $machine_type = $1;
7211 $use_old_bios_files = 1;
7212 } else {
7213 my $kvmver = kvm_user_version();
7214 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
7215 # load new efi bios files on migration. So this hack is required to allow
7216 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
7217 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
7218 $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
7219 }
7220
7221 return ($use_old_bios_files, $machine_type);
7222 }
7223
7224 sub create_efidisk($$$$$) {
7225 my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7226
7227 my (undef, $ovmf_vars) = get_ovmf_files($arch);
7228 die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7229
7230 my $vars_size_b = -s $ovmf_vars;
7231 my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7232 my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7233 PVE::Storage::activate_volumes($storecfg, [$volid]);
7234
7235 qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7236
7237 return ($volid, $vars_size);
7238 }
7239
7240 sub vm_iothreads_list {
7241 my ($vmid) = @_;
7242
7243 my $res = vm_mon_cmd($vmid, 'query-iothreads');
7244
7245 my $iothreads = {};
7246 foreach my $iothread (@$res) {
7247 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7248 }
7249
7250 return $iothreads;
7251 }
7252
7253 sub scsihw_infos {
7254 my ($conf, $drive) = @_;
7255
7256 my $maxdev = 0;
7257
7258 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7259 $maxdev = 7;
7260 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7261 $maxdev = 1;
7262 } else {
7263 $maxdev = 256;
7264 }
7265
7266 my $controller = int($drive->{index} / $maxdev);
7267 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7268
7269 return ($maxdev, $controller, $controller_prefix);
7270 }
7271
7272 sub add_hyperv_enlightenments {
7273 my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough, $hv_vendor_id) = @_;
7274
7275 return if $winversion < 6;
7276 return if $bios && $bios eq 'ovmf' && $winversion < 8;
7277
7278 if ($gpu_passthrough || defined($hv_vendor_id)) {
7279 $hv_vendor_id //= 'proxmox';
7280 push @$cpuFlags , "hv_vendor_id=$hv_vendor_id";
7281 }
7282
7283 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
7284 push @$cpuFlags , 'hv_spinlocks=0x1fff';
7285 push @$cpuFlags , 'hv_vapic';
7286 push @$cpuFlags , 'hv_time';
7287 } else {
7288 push @$cpuFlags , 'hv_spinlocks=0xffff';
7289 }
7290
7291 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
7292 push @$cpuFlags , 'hv_reset';
7293 push @$cpuFlags , 'hv_vpindex';
7294 push @$cpuFlags , 'hv_runtime';
7295 }
7296
7297 if ($winversion >= 7) {
7298 push @$cpuFlags , 'hv_relaxed';
7299
7300 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
7301 push @$cpuFlags , 'hv_synic';
7302 push @$cpuFlags , 'hv_stimer';
7303 }
7304
7305 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 3, 1)) {
7306 push @$cpuFlags , 'hv_ipi';
7307 }
7308 }
7309 }
7310
7311 sub windows_version {
7312 my ($ostype) = @_;
7313
7314 return 0 if !$ostype;
7315
7316 my $winversion = 0;
7317
7318 if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7319 $winversion = 5;
7320 } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7321 $winversion = 6;
7322 } elsif ($ostype =~ m/^win(\d+)$/) {
7323 $winversion = $1;
7324 }
7325
7326 return $winversion;
7327 }
7328
7329 sub resolve_dst_disk_format {
7330 my ($storecfg, $storeid, $src_volname, $format) = @_;
7331 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7332
7333 if (!$format) {
7334 # if no target format is specified, use the source disk format as hint
7335 if ($src_volname) {
7336 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7337 $format = qemu_img_format($scfg, $src_volname);
7338 } else {
7339 return $defFormat;
7340 }
7341 }
7342
7343 # test if requested format is supported - else use default
7344 my $supported = grep { $_ eq $format } @$validFormats;
7345 $format = $defFormat if !$supported;
7346 return $format;
7347 }
7348
7349 sub resolve_first_disk {
7350 my $conf = shift;
7351 my @disks = PVE::QemuServer::valid_drive_names();
7352 my $firstdisk;
7353 foreach my $ds (reverse @disks) {
7354 next if !$conf->{$ds};
7355 my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
7356 next if PVE::QemuServer::drive_is_cdrom($disk);
7357 $firstdisk = $ds;
7358 }
7359 return $firstdisk;
7360 }
7361
7362 sub generate_uuid {
7363 my ($uuid, $uuid_str);
7364 UUID::generate($uuid);
7365 UUID::unparse($uuid, $uuid_str);
7366 return $uuid_str;
7367 }
7368
7369 sub generate_smbios1_uuid {
7370 return "uuid=".generate_uuid();
7371 }
7372
7373 sub nbd_stop {
7374 my ($vmid) = @_;
7375
7376 vm_mon_cmd($vmid, 'nbd-server-stop');
7377 }
7378
7379 sub create_reboot_request {
7380 my ($vmid) = @_;
7381 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7382 or die "failed to create reboot trigger file: $!\n";
7383 close($fh);
7384 }
7385
7386 sub clear_reboot_request {
7387 my ($vmid) = @_;
7388 my $path = "/run/qemu-server/$vmid.reboot";
7389 my $res = 0;
7390
7391 $res = unlink($path);
7392 die "could not remove reboot request for $vmid: $!"
7393 if !$res && $! != POSIX::ENOENT;
7394
7395 return $res;
7396 }
7397
7398 # bash completion helper
7399
7400 sub complete_backup_archives {
7401 my ($cmdname, $pname, $cvalue) = @_;
7402
7403 my $cfg = PVE::Storage::config();
7404
7405 my $storeid;
7406
7407 if ($cvalue =~ m/^([^:]+):/) {
7408 $storeid = $1;
7409 }
7410
7411 my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7412
7413 my $res = [];
7414 foreach my $id (keys %$data) {
7415 foreach my $item (@{$data->{$id}}) {
7416 next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7417 push @$res, $item->{volid} if defined($item->{volid});
7418 }
7419 }
7420
7421 return $res;
7422 }
7423
7424 my $complete_vmid_full = sub {
7425 my ($running) = @_;
7426
7427 my $idlist = vmstatus();
7428
7429 my $res = [];
7430
7431 foreach my $id (keys %$idlist) {
7432 my $d = $idlist->{$id};
7433 if (defined($running)) {
7434 next if $d->{template};
7435 next if $running && $d->{status} ne 'running';
7436 next if !$running && $d->{status} eq 'running';
7437 }
7438 push @$res, $id;
7439
7440 }
7441 return $res;
7442 };
7443
7444 sub complete_vmid {
7445 return &$complete_vmid_full();
7446 }
7447
7448 sub complete_vmid_stopped {
7449 return &$complete_vmid_full(0);
7450 }
7451
7452 sub complete_vmid_running {
7453 return &$complete_vmid_full(1);
7454 }
7455
7456 sub complete_storage {
7457
7458 my $cfg = PVE::Storage::config();
7459 my $ids = $cfg->{ids};
7460
7461 my $res = [];
7462 foreach my $sid (keys %$ids) {
7463 next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7464 next if !$ids->{$sid}->{content}->{images};
7465 push @$res, $sid;
7466 }
7467
7468 return $res;
7469 }
7470
7471 1;