]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
cfg2cmd: push vmstate to volid list to ensure it gets also deactivated
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5
6 use POSIX;
7 use IO::Handle;
8 use IO::Select;
9 use IO::File;
10 use IO::Dir;
11 use IO::Socket::UNIX;
12 use File::Basename;
13 use File::Path;
14 use File::stat;
15 use Getopt::Long;
16 use Digest::SHA;
17 use Fcntl ':flock';
18 use Cwd 'abs_path';
19 use IPC::Open3;
20 use JSON;
21 use Fcntl;
22 use PVE::SafeSyslog;
23 use Storable qw(dclone);
24 use MIME::Base64;
25 use PVE::Exception qw(raise raise_param_exc);
26 use PVE::Storage;
27 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach $IPV6RE);
28 use PVE::JSONSchema qw(get_standard_option);
29 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
30 use PVE::INotify;
31 use PVE::ProcFSTools;
32 use PVE::QemuConfig;
33 use PVE::QMPClient;
34 use PVE::RPCEnvironment;
35 use PVE::GuestHelpers;
36 use PVE::QemuServer::PCI qw(print_pci_addr print_pcie_addr print_pcie_root_port);
37 use PVE::QemuServer::Memory;
38 use PVE::QemuServer::USB qw(parse_usb_device);
39 use PVE::QemuServer::Cloudinit;
40 use PVE::SysFSTools;
41 use PVE::Systemd;
42 use Time::HiRes qw(gettimeofday);
43 use File::Copy qw(copy);
44 use URI::Escape;
45
46 my $EDK2_FW_BASE = '/usr/share/pve-edk2-firmware/';
47 my $OVMF = {
48 x86_64 => [
49 "$EDK2_FW_BASE/OVMF_CODE.fd",
50 "$EDK2_FW_BASE/OVMF_VARS.fd"
51 ],
52 aarch64 => [
53 "$EDK2_FW_BASE/AAVMF_CODE.fd",
54 "$EDK2_FW_BASE/AAVMF_VARS.fd"
55 ],
56 };
57
58 my $qemu_snap_storage = { rbd => 1 };
59
60 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
61
62 my $QEMU_FORMAT_RE = qr/raw|cow|qcow|qcow2|qed|vmdk|cloop/;
63
64 # Note about locking: we use flock on the config file protect
65 # against concurent actions.
66 # Aditionaly, we have a 'lock' setting in the config file. This
67 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
68 # allowed when such lock is set. But you can ignore this kind of
69 # lock with the --skiplock flag.
70
71 cfs_register_file('/qemu-server/',
72 \&parse_vm_config,
73 \&write_vm_config);
74
75 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
76 description => "Some command save/restore state from this location.",
77 type => 'string',
78 maxLength => 128,
79 optional => 1,
80 });
81
82 PVE::JSONSchema::register_standard_option('pve-qm-image-format', {
83 type => 'string',
84 enum => [qw(raw cow qcow qed qcow2 vmdk cloop)],
85 description => "The drive's backing file's data format.",
86 optional => 1,
87 });
88
89 PVE::JSONSchema::register_standard_option('pve-qemu-machine', {
90 description => "Specifies the Qemu machine type.",
91 type => 'string',
92 pattern => '(pc|pc(-i440fx)?-\d+\.\d+(\.pxe)?|q35|pc-q35-\d+\.\d+(\.pxe)?|virt(?:-\d+\.\d+)?)',
93 maxLength => 40,
94 optional => 1,
95 });
96
97 #no warnings 'redefine';
98
99 sub cgroups_write {
100 my ($controller, $vmid, $option, $value) = @_;
101
102 my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
103 PVE::ProcFSTools::write_proc_entry($path, $value);
104
105 }
106
107 my $nodename = PVE::INotify::nodename();
108
109 mkdir "/etc/pve/nodes/$nodename";
110 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
111 mkdir $confdir;
112
113 my $var_run_tmpdir = "/var/run/qemu-server";
114 mkdir $var_run_tmpdir;
115
116 my $lock_dir = "/var/lock/qemu-server";
117 mkdir $lock_dir;
118
119 my $cpu_vendor_list = {
120 # Intel CPUs
121 486 => 'GenuineIntel',
122 pentium => 'GenuineIntel',
123 pentium2 => 'GenuineIntel',
124 pentium3 => 'GenuineIntel',
125 coreduo => 'GenuineIntel',
126 core2duo => 'GenuineIntel',
127 Conroe => 'GenuineIntel',
128 Penryn => 'GenuineIntel',
129 Nehalem => 'GenuineIntel',
130 'Nehalem-IBRS' => 'GenuineIntel',
131 Westmere => 'GenuineIntel',
132 'Westmere-IBRS' => 'GenuineIntel',
133 SandyBridge => 'GenuineIntel',
134 'SandyBridge-IBRS' => 'GenuineIntel',
135 IvyBridge => 'GenuineIntel',
136 'IvyBridge-IBRS' => 'GenuineIntel',
137 Haswell => 'GenuineIntel',
138 'Haswell-IBRS' => 'GenuineIntel',
139 'Haswell-noTSX' => 'GenuineIntel',
140 'Haswell-noTSX-IBRS' => 'GenuineIntel',
141 Broadwell => 'GenuineIntel',
142 'Broadwell-IBRS' => 'GenuineIntel',
143 'Broadwell-noTSX' => 'GenuineIntel',
144 'Broadwell-noTSX-IBRS' => 'GenuineIntel',
145 'Skylake-Client' => 'GenuineIntel',
146 'Skylake-Client-IBRS' => 'GenuineIntel',
147 'Skylake-Server' => 'GenuineIntel',
148 'Skylake-Server-IBRS' => 'GenuineIntel',
149 'Cascadelake-Server' => 'GenuineIntel',
150 KnightsMill => 'GenuineIntel',
151
152
153 # AMD CPUs
154 athlon => 'AuthenticAMD',
155 phenom => 'AuthenticAMD',
156 Opteron_G1 => 'AuthenticAMD',
157 Opteron_G2 => 'AuthenticAMD',
158 Opteron_G3 => 'AuthenticAMD',
159 Opteron_G4 => 'AuthenticAMD',
160 Opteron_G5 => 'AuthenticAMD',
161 EPYC => 'AuthenticAMD',
162 'EPYC-IBPB' => 'AuthenticAMD',
163
164 # generic types, use vendor from host node
165 host => 'default',
166 kvm32 => 'default',
167 kvm64 => 'default',
168 qemu32 => 'default',
169 qemu64 => 'default',
170 max => 'default',
171 };
172
173 my @supported_cpu_flags = (
174 'pcid',
175 'spec-ctrl',
176 'ibpb',
177 'ssbd',
178 'virt-ssbd',
179 'amd-ssbd',
180 'amd-no-ssb',
181 'pdpe1gb',
182 'md-clear',
183 'hv-tlbflush',
184 'hv-evmcs',
185 'aes'
186 );
187 my $cpu_flag = qr/[+-](@{[join('|', @supported_cpu_flags)]})/;
188
189 my $cpu_fmt = {
190 cputype => {
191 description => "Emulated CPU type.",
192 type => 'string',
193 enum => [ sort { "\L$a" cmp "\L$b" } keys %$cpu_vendor_list ],
194 default => 'kvm64',
195 default_key => 1,
196 },
197 hidden => {
198 description => "Do not identify as a KVM virtual machine.",
199 type => 'boolean',
200 optional => 1,
201 default => 0
202 },
203 'hv-vendor-id' => {
204 type => 'string',
205 pattern => qr/[a-zA-Z0-9]{1,12}/,
206 format_description => 'vendor-id',
207 description => 'The Hyper-V vendor ID. Some drivers or programs inside Windows guests need a specific ID.',
208 optional => 1,
209 },
210 flags => {
211 description => "List of additional CPU flags separated by ';'."
212 . " Use '+FLAG' to enable, '-FLAG' to disable a flag."
213 . " Currently supported flags: @{[join(', ', @supported_cpu_flags)]}.",
214 format_description => '+FLAG[;-FLAG...]',
215 type => 'string',
216 pattern => qr/$cpu_flag(;$cpu_flag)*/,
217 optional => 1,
218 },
219 };
220
221 my $watchdog_fmt = {
222 model => {
223 default_key => 1,
224 type => 'string',
225 enum => [qw(i6300esb ib700)],
226 description => "Watchdog type to emulate.",
227 default => 'i6300esb',
228 optional => 1,
229 },
230 action => {
231 type => 'string',
232 enum => [qw(reset shutdown poweroff pause debug none)],
233 description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
234 optional => 1,
235 },
236 };
237 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
238
239 my $agent_fmt = {
240 enabled => {
241 description => "Enable/disable Qemu GuestAgent.",
242 type => 'boolean',
243 default => 0,
244 default_key => 1,
245 },
246 fstrim_cloned_disks => {
247 description => "Run fstrim after cloning/moving a disk.",
248 type => 'boolean',
249 optional => 1,
250 default => 0
251 },
252 };
253
254 my $vga_fmt = {
255 type => {
256 description => "Select the VGA type.",
257 type => 'string',
258 default => 'std',
259 optional => 1,
260 default_key => 1,
261 enum => [qw(cirrus qxl qxl2 qxl3 qxl4 none serial0 serial1 serial2 serial3 std virtio vmware)],
262 },
263 memory => {
264 description => "Sets the VGA memory (in MiB). Has no effect with serial display.",
265 type => 'integer',
266 optional => 1,
267 minimum => 4,
268 maximum => 512,
269 },
270 };
271
272 my $ivshmem_fmt = {
273 size => {
274 type => 'integer',
275 minimum => 1,
276 description => "The size of the file in MB.",
277 },
278 name => {
279 type => 'string',
280 pattern => '[a-zA-Z0-9\-]+',
281 optional => 1,
282 format_description => 'string',
283 description => "The name of the file. Will be prefixed with 'pve-shm-'. Default is the VMID. Will be deleted when the VM is stopped.",
284 },
285 };
286
287 my $audio_fmt = {
288 device => {
289 type => 'string',
290 enum => [qw(ich9-intel-hda intel-hda AC97)],
291 description => "Configure an audio device."
292 },
293 driver => {
294 type => 'string',
295 enum => ['spice'],
296 default => 'spice',
297 optional => 1,
298 description => "Driver backend for the audio device."
299 },
300 };
301
302 my $spice_enhancements_fmt = {
303 foldersharing => {
304 type => 'boolean',
305 optional => 1,
306 default => '0',
307 description => "Enable folder sharing via SPICE. Needs Spice-WebDAV daemon installed in the VM."
308 },
309 videostreaming => {
310 type => 'string',
311 enum => ['off', 'all', 'filter'],
312 default => 'off',
313 optional => 1,
314 description => "Enable video streaming. Uses compression for detected video streams."
315 },
316 };
317
318 my $confdesc = {
319 onboot => {
320 optional => 1,
321 type => 'boolean',
322 description => "Specifies whether a VM will be started during system bootup.",
323 default => 0,
324 },
325 autostart => {
326 optional => 1,
327 type => 'boolean',
328 description => "Automatic restart after crash (currently ignored).",
329 default => 0,
330 },
331 hotplug => {
332 optional => 1,
333 type => 'string', format => 'pve-hotplug-features',
334 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
335 default => 'network,disk,usb',
336 },
337 reboot => {
338 optional => 1,
339 type => 'boolean',
340 description => "Allow reboot. If set to '0' the VM exit on reboot.",
341 default => 1,
342 },
343 lock => {
344 optional => 1,
345 type => 'string',
346 description => "Lock/unlock the VM.",
347 enum => [qw(backup clone create migrate rollback snapshot snapshot-delete suspending suspended)],
348 },
349 cpulimit => {
350 optional => 1,
351 type => 'number',
352 description => "Limit of CPU usage.",
353 verbose_description => "Limit of CPU usage.\n\nNOTE: If the computer has 2 CPUs, it has total of '2' CPU time. Value '0' indicates no CPU limit.",
354 minimum => 0,
355 maximum => 128,
356 default => 0,
357 },
358 cpuunits => {
359 optional => 1,
360 type => 'integer',
361 description => "CPU weight for a VM.",
362 verbose_description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.",
363 minimum => 2,
364 maximum => 262144,
365 default => 1024,
366 },
367 memory => {
368 optional => 1,
369 type => 'integer',
370 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
371 minimum => 16,
372 default => 512,
373 },
374 balloon => {
375 optional => 1,
376 type => 'integer',
377 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
378 minimum => 0,
379 },
380 shares => {
381 optional => 1,
382 type => 'integer',
383 description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning. Auto-ballooning is done by pvestatd.",
384 minimum => 0,
385 maximum => 50000,
386 default => 1000,
387 },
388 keyboard => {
389 optional => 1,
390 type => 'string',
391 description => "Keybord layout for vnc server. Default is read from the '/etc/pve/datacenter.cfg' configuration file.".
392 "It should not be necessary to set it.",
393 enum => PVE::Tools::kvmkeymaplist(),
394 default => undef,
395 },
396 name => {
397 optional => 1,
398 type => 'string', format => 'dns-name',
399 description => "Set a name for the VM. Only used on the configuration web interface.",
400 },
401 scsihw => {
402 optional => 1,
403 type => 'string',
404 description => "SCSI controller model",
405 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
406 default => 'lsi',
407 },
408 description => {
409 optional => 1,
410 type => 'string',
411 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
412 },
413 ostype => {
414 optional => 1,
415 type => 'string',
416 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 win10 l24 l26 solaris)],
417 description => "Specify guest operating system.",
418 verbose_description => <<EODESC,
419 Specify guest operating system. This is used to enable special
420 optimization/features for specific operating systems:
421
422 [horizontal]
423 other;; unspecified OS
424 wxp;; Microsoft Windows XP
425 w2k;; Microsoft Windows 2000
426 w2k3;; Microsoft Windows 2003
427 w2k8;; Microsoft Windows 2008
428 wvista;; Microsoft Windows Vista
429 win7;; Microsoft Windows 7
430 win8;; Microsoft Windows 8/2012/2012r2
431 win10;; Microsoft Windows 10/2016
432 l24;; Linux 2.4 Kernel
433 l26;; Linux 2.6/3.X Kernel
434 solaris;; Solaris/OpenSolaris/OpenIndiania kernel
435 EODESC
436 },
437 boot => {
438 optional => 1,
439 type => 'string',
440 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
441 pattern => '[acdn]{1,4}',
442 default => 'cdn',
443 },
444 bootdisk => {
445 optional => 1,
446 type => 'string', format => 'pve-qm-bootdisk',
447 description => "Enable booting from specified disk.",
448 pattern => '(ide|sata|scsi|virtio)\d+',
449 },
450 smp => {
451 optional => 1,
452 type => 'integer',
453 description => "The number of CPUs. Please use option -sockets instead.",
454 minimum => 1,
455 default => 1,
456 },
457 sockets => {
458 optional => 1,
459 type => 'integer',
460 description => "The number of CPU sockets.",
461 minimum => 1,
462 default => 1,
463 },
464 cores => {
465 optional => 1,
466 type => 'integer',
467 description => "The number of cores per socket.",
468 minimum => 1,
469 default => 1,
470 },
471 numa => {
472 optional => 1,
473 type => 'boolean',
474 description => "Enable/disable NUMA.",
475 default => 0,
476 },
477 hugepages => {
478 optional => 1,
479 type => 'string',
480 description => "Enable/disable hugepages memory.",
481 enum => [qw(any 2 1024)],
482 },
483 vcpus => {
484 optional => 1,
485 type => 'integer',
486 description => "Number of hotplugged vcpus.",
487 minimum => 1,
488 default => 0,
489 },
490 acpi => {
491 optional => 1,
492 type => 'boolean',
493 description => "Enable/disable ACPI.",
494 default => 1,
495 },
496 agent => {
497 optional => 1,
498 description => "Enable/disable Qemu GuestAgent and its properties.",
499 type => 'string',
500 format => $agent_fmt,
501 },
502 kvm => {
503 optional => 1,
504 type => 'boolean',
505 description => "Enable/disable KVM hardware virtualization.",
506 default => 1,
507 },
508 tdf => {
509 optional => 1,
510 type => 'boolean',
511 description => "Enable/disable time drift fix.",
512 default => 0,
513 },
514 localtime => {
515 optional => 1,
516 type => 'boolean',
517 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
518 },
519 freeze => {
520 optional => 1,
521 type => 'boolean',
522 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
523 },
524 vga => {
525 optional => 1,
526 type => 'string', format => $vga_fmt,
527 description => "Configure the VGA hardware.",
528 verbose_description => "Configure the VGA Hardware. If you want to use ".
529 "high resolution modes (>= 1280x1024x16) you may need to increase " .
530 "the vga memory option. Since QEMU 2.9 the default VGA display type " .
531 "is 'std' for all OS types besides some Windows versions (XP and " .
532 "older) which use 'cirrus'. The 'qxl' option enables the SPICE " .
533 "display server. For win* OS you can select how many independent " .
534 "displays you want, Linux guests can add displays them self.\n".
535 "You can also run without any graphic card, using a serial device as terminal.",
536 },
537 watchdog => {
538 optional => 1,
539 type => 'string', format => 'pve-qm-watchdog',
540 description => "Create a virtual hardware watchdog device.",
541 verbose_description => "Create a virtual hardware watchdog device. Once enabled" .
542 " (by a guest action), the watchdog must be periodically polled " .
543 "by an agent inside the guest or else the watchdog will reset " .
544 "the guest (or execute the respective action specified)",
545 },
546 startdate => {
547 optional => 1,
548 type => 'string',
549 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
550 description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
551 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
552 default => 'now',
553 },
554 startup => get_standard_option('pve-startup-order'),
555 template => {
556 optional => 1,
557 type => 'boolean',
558 description => "Enable/disable Template.",
559 default => 0,
560 },
561 args => {
562 optional => 1,
563 type => 'string',
564 description => "Arbitrary arguments passed to kvm.",
565 verbose_description => <<EODESCR,
566 Arbitrary arguments passed to kvm, for example:
567
568 args: -no-reboot -no-hpet
569
570 NOTE: this option is for experts only.
571 EODESCR
572 },
573 tablet => {
574 optional => 1,
575 type => 'boolean',
576 default => 1,
577 description => "Enable/disable the USB tablet device.",
578 verbose_description => "Enable/disable the USB tablet device. This device is " .
579 "usually needed to allow absolute mouse positioning with VNC. " .
580 "Else the mouse runs out of sync with normal VNC clients. " .
581 "If you're running lots of console-only guests on one host, " .
582 "you may consider disabling this to save some context switches. " .
583 "This is turned off by default if you use spice (-vga=qxl).",
584 },
585 migrate_speed => {
586 optional => 1,
587 type => 'integer',
588 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
589 minimum => 0,
590 default => 0,
591 },
592 migrate_downtime => {
593 optional => 1,
594 type => 'number',
595 description => "Set maximum tolerated downtime (in seconds) for migrations.",
596 minimum => 0,
597 default => 0.1,
598 },
599 cdrom => {
600 optional => 1,
601 type => 'string', format => 'pve-qm-ide',
602 typetext => '<volume>',
603 description => "This is an alias for option -ide2",
604 },
605 cpu => {
606 optional => 1,
607 description => "Emulated CPU type.",
608 type => 'string',
609 format => $cpu_fmt,
610 },
611 parent => get_standard_option('pve-snapshot-name', {
612 optional => 1,
613 description => "Parent snapshot name. This is used internally, and should not be modified.",
614 }),
615 snaptime => {
616 optional => 1,
617 description => "Timestamp for snapshots.",
618 type => 'integer',
619 minimum => 0,
620 },
621 vmstate => {
622 optional => 1,
623 type => 'string', format => 'pve-volume-id',
624 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
625 },
626 vmstatestorage => get_standard_option('pve-storage-id', {
627 description => "Default storage for VM state volumes/files.",
628 optional => 1,
629 }),
630 runningmachine => get_standard_option('pve-qemu-machine', {
631 description => "Specifies the Qemu machine type of the running vm. This is used internally for snapshots.",
632 }),
633 machine => get_standard_option('pve-qemu-machine'),
634 arch => {
635 description => "Virtual processor architecture. Defaults to the host.",
636 optional => 1,
637 type => 'string',
638 enum => [qw(x86_64 aarch64)],
639 },
640 smbios1 => {
641 description => "Specify SMBIOS type 1 fields.",
642 type => 'string', format => 'pve-qm-smbios1',
643 maxLength => 512,
644 optional => 1,
645 },
646 protection => {
647 optional => 1,
648 type => 'boolean',
649 description => "Sets the protection flag of the VM. This will disable the remove VM and remove disk operations.",
650 default => 0,
651 },
652 bios => {
653 optional => 1,
654 type => 'string',
655 enum => [ qw(seabios ovmf) ],
656 description => "Select BIOS implementation.",
657 default => 'seabios',
658 },
659 vmgenid => {
660 type => 'string',
661 pattern => '(?:[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}|[01])',
662 format_description => 'UUID',
663 description => "Set VM Generation ID. Use '1' to autogenerate on create or update, pass '0' to disable explicitly.",
664 verbose_description => "The VM generation ID (vmgenid) device exposes a".
665 " 128-bit integer value identifier to the guest OS. This allows to".
666 " notify the guest operating system when the virtual machine is".
667 " executed with a different configuration (e.g. snapshot execution".
668 " or creation from a template). The guest operating system notices".
669 " the change, and is then able to react as appropriate by marking".
670 " its copies of distributed databases as dirty, re-initializing its".
671 " random number generator, etc.\n".
672 "Note that auto-creation only works when done throug API/CLI create".
673 " or update methods, but not when manually editing the config file.",
674 default => "1 (autogenerated)",
675 optional => 1,
676 },
677 hookscript => {
678 type => 'string',
679 format => 'pve-volume-id',
680 optional => 1,
681 description => "Script that will be executed during various steps in the vms lifetime.",
682 },
683 ivshmem => {
684 type => 'string',
685 format => $ivshmem_fmt,
686 description => "Inter-VM shared memory. Useful for direct communication between VMs, or to the host.",
687 optional => 1,
688 },
689 audio0 => {
690 type => 'string',
691 format => $audio_fmt,
692 description => "Configure a audio device, useful in combination with QXL/Spice.",
693 optional => 1
694 },
695 spice_enhancements => {
696 type => 'string',
697 format => $spice_enhancements_fmt,
698 description => "Configure additional enhancements for SPICE.",
699 optional => 1
700 },
701 };
702
703 my $cicustom_fmt = {
704 meta => {
705 type => 'string',
706 optional => 1,
707 description => 'Specify a custom file containing all meta data passed to the VM via cloud-init. This is provider specific meaning configdrive2 and nocloud differ.',
708 format => 'pve-volume-id',
709 format_description => 'volume',
710 },
711 network => {
712 type => 'string',
713 optional => 1,
714 description => 'Specify a custom file containing all network data passed to the VM via cloud-init.',
715 format => 'pve-volume-id',
716 format_description => 'volume',
717 },
718 user => {
719 type => 'string',
720 optional => 1,
721 description => 'Specify a custom file containing all user data passed to the VM via cloud-init.',
722 format => 'pve-volume-id',
723 format_description => 'volume',
724 },
725 };
726 PVE::JSONSchema::register_format('pve-qm-cicustom', $cicustom_fmt);
727
728 my $confdesc_cloudinit = {
729 citype => {
730 optional => 1,
731 type => 'string',
732 description => 'Specifies the cloud-init configuration format. The default depends on the configured operating system type (`ostype`. We use the `nocloud` format for Linux, and `configdrive2` for windows.',
733 enum => ['configdrive2', 'nocloud'],
734 },
735 ciuser => {
736 optional => 1,
737 type => 'string',
738 description => "cloud-init: User name to change ssh keys and password for instead of the image's configured default user.",
739 },
740 cipassword => {
741 optional => 1,
742 type => 'string',
743 description => 'cloud-init: Password to assign the user. Using this is generally not recommended. Use ssh keys instead. Also note that older cloud-init versions do not support hashed passwords.',
744 },
745 cicustom => {
746 optional => 1,
747 type => 'string',
748 description => 'cloud-init: Specify custom files to replace the automatically generated ones at start.',
749 format => 'pve-qm-cicustom',
750 },
751 searchdomain => {
752 optional => 1,
753 type => 'string',
754 description => "cloud-init: Sets DNS search domains for a container. Create will automatically use the setting from the host if neither searchdomain nor nameserver are set.",
755 },
756 nameserver => {
757 optional => 1,
758 type => 'string', format => 'address-list',
759 description => "cloud-init: Sets DNS server IP address for a container. Create will automatically use the setting from the host if neither searchdomain nor nameserver are set.",
760 },
761 sshkeys => {
762 optional => 1,
763 type => 'string',
764 format => 'urlencoded',
765 description => "cloud-init: Setup public SSH keys (one key per line, OpenSSH format).",
766 },
767 };
768
769 # what about other qemu settings ?
770 #cpu => 'string',
771 #machine => 'string',
772 #fda => 'file',
773 #fdb => 'file',
774 #mtdblock => 'file',
775 #sd => 'file',
776 #pflash => 'file',
777 #snapshot => 'bool',
778 #bootp => 'file',
779 ##tftp => 'dir',
780 ##smb => 'dir',
781 #kernel => 'file',
782 #append => 'string',
783 #initrd => 'file',
784 ##soundhw => 'string',
785
786 while (my ($k, $v) = each %$confdesc) {
787 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
788 }
789
790 my $MAX_IDE_DISKS = 4;
791 my $MAX_SCSI_DISKS = 14;
792 my $MAX_VIRTIO_DISKS = 16;
793 my $MAX_SATA_DISKS = 6;
794 my $MAX_USB_DEVICES = 5;
795 my $MAX_NETS = 32;
796 my $MAX_UNUSED_DISKS = 256;
797 my $MAX_HOSTPCI_DEVICES = 16;
798 my $MAX_SERIAL_PORTS = 4;
799 my $MAX_PARALLEL_PORTS = 3;
800 my $MAX_NUMA = 8;
801
802 my $numa_fmt = {
803 cpus => {
804 type => "string",
805 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
806 description => "CPUs accessing this NUMA node.",
807 format_description => "id[-id];...",
808 },
809 memory => {
810 type => "number",
811 description => "Amount of memory this NUMA node provides.",
812 optional => 1,
813 },
814 hostnodes => {
815 type => "string",
816 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
817 description => "Host NUMA nodes to use.",
818 format_description => "id[-id];...",
819 optional => 1,
820 },
821 policy => {
822 type => 'string',
823 enum => [qw(preferred bind interleave)],
824 description => "NUMA allocation policy.",
825 optional => 1,
826 },
827 };
828 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
829 my $numadesc = {
830 optional => 1,
831 type => 'string', format => $numa_fmt,
832 description => "NUMA topology.",
833 };
834 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
835
836 for (my $i = 0; $i < $MAX_NUMA; $i++) {
837 $confdesc->{"numa$i"} = $numadesc;
838 }
839
840 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
841 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
842 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
843 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
844
845 my $net_fmt_bridge_descr = <<__EOD__;
846 Bridge to attach the network device to. The Proxmox VE standard bridge
847 is called 'vmbr0'.
848
849 If you do not specify a bridge, we create a kvm user (NATed) network
850 device, which provides DHCP and DNS services. The following addresses
851 are used:
852
853 10.0.2.2 Gateway
854 10.0.2.3 DNS Server
855 10.0.2.4 SMB Server
856
857 The DHCP server assign addresses to the guest starting from 10.0.2.15.
858 __EOD__
859
860 my $net_fmt = {
861 macaddr => get_standard_option('mac-addr', {
862 description => "MAC address. That address must be unique withing your network. This is automatically generated if not specified.",
863 }),
864 model => {
865 type => 'string',
866 description => "Network Card Model. The 'virtio' model provides the best performance with very low CPU overhead. If your guest does not support this driver, it is usually best to use 'e1000'.",
867 enum => $nic_model_list,
868 default_key => 1,
869 },
870 (map { $_ => { keyAlias => 'model', alias => 'macaddr' }} @$nic_model_list),
871 bridge => {
872 type => 'string',
873 description => $net_fmt_bridge_descr,
874 format_description => 'bridge',
875 optional => 1,
876 },
877 queues => {
878 type => 'integer',
879 minimum => 0, maximum => 16,
880 description => 'Number of packet queues to be used on the device.',
881 optional => 1,
882 },
883 rate => {
884 type => 'number',
885 minimum => 0,
886 description => "Rate limit in mbps (megabytes per second) as floating point number.",
887 optional => 1,
888 },
889 tag => {
890 type => 'integer',
891 minimum => 1, maximum => 4094,
892 description => 'VLAN tag to apply to packets on this interface.',
893 optional => 1,
894 },
895 trunks => {
896 type => 'string',
897 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
898 description => 'VLAN trunks to pass through this interface.',
899 format_description => 'vlanid[;vlanid...]',
900 optional => 1,
901 },
902 firewall => {
903 type => 'boolean',
904 description => 'Whether this interface should be protected by the firewall.',
905 optional => 1,
906 },
907 link_down => {
908 type => 'boolean',
909 description => 'Whether this interface should be disconnected (like pulling the plug).',
910 optional => 1,
911 },
912 };
913
914 my $netdesc = {
915 optional => 1,
916 type => 'string', format => $net_fmt,
917 description => "Specify network devices.",
918 };
919
920 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
921
922 my $ipconfig_fmt = {
923 ip => {
924 type => 'string',
925 format => 'pve-ipv4-config',
926 format_description => 'IPv4Format/CIDR',
927 description => 'IPv4 address in CIDR format.',
928 optional => 1,
929 default => 'dhcp',
930 },
931 gw => {
932 type => 'string',
933 format => 'ipv4',
934 format_description => 'GatewayIPv4',
935 description => 'Default gateway for IPv4 traffic.',
936 optional => 1,
937 requires => 'ip',
938 },
939 ip6 => {
940 type => 'string',
941 format => 'pve-ipv6-config',
942 format_description => 'IPv6Format/CIDR',
943 description => 'IPv6 address in CIDR format.',
944 optional => 1,
945 default => 'dhcp',
946 },
947 gw6 => {
948 type => 'string',
949 format => 'ipv6',
950 format_description => 'GatewayIPv6',
951 description => 'Default gateway for IPv6 traffic.',
952 optional => 1,
953 requires => 'ip6',
954 },
955 };
956 PVE::JSONSchema::register_format('pve-qm-ipconfig', $ipconfig_fmt);
957 my $ipconfigdesc = {
958 optional => 1,
959 type => 'string', format => 'pve-qm-ipconfig',
960 description => <<'EODESCR',
961 cloud-init: Specify IP addresses and gateways for the corresponding interface.
962
963 IP addresses use CIDR notation, gateways are optional but need an IP of the same type specified.
964
965 The special string 'dhcp' can be used for IP addresses to use DHCP, in which case no explicit gateway should be provided.
966 For IPv6 the special string 'auto' can be used to use stateless autoconfiguration.
967
968 If cloud-init is enabled and neither an IPv4 nor an IPv6 address is specified, it defaults to using dhcp on IPv4.
969 EODESCR
970 };
971 PVE::JSONSchema::register_standard_option("pve-qm-ipconfig", $netdesc);
972
973 for (my $i = 0; $i < $MAX_NETS; $i++) {
974 $confdesc->{"net$i"} = $netdesc;
975 $confdesc_cloudinit->{"ipconfig$i"} = $ipconfigdesc;
976 }
977
978 foreach my $key (keys %$confdesc_cloudinit) {
979 $confdesc->{$key} = $confdesc_cloudinit->{$key};
980 }
981
982 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
983 sub verify_volume_id_or_qm_path {
984 my ($volid, $noerr) = @_;
985
986 if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
987 return $volid;
988 }
989
990 # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
991 $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
992 if ($@) {
993 return undef if $noerr;
994 die $@;
995 }
996 return $volid;
997 }
998
999 my $drivename_hash;
1000
1001 my %drivedesc_base = (
1002 volume => { alias => 'file' },
1003 file => {
1004 type => 'string',
1005 format => 'pve-volume-id-or-qm-path',
1006 default_key => 1,
1007 format_description => 'volume',
1008 description => "The drive's backing volume.",
1009 },
1010 media => {
1011 type => 'string',
1012 enum => [qw(cdrom disk)],
1013 description => "The drive's media type.",
1014 default => 'disk',
1015 optional => 1
1016 },
1017 cyls => {
1018 type => 'integer',
1019 description => "Force the drive's physical geometry to have a specific cylinder count.",
1020 optional => 1
1021 },
1022 heads => {
1023 type => 'integer',
1024 description => "Force the drive's physical geometry to have a specific head count.",
1025 optional => 1
1026 },
1027 secs => {
1028 type => 'integer',
1029 description => "Force the drive's physical geometry to have a specific sector count.",
1030 optional => 1
1031 },
1032 trans => {
1033 type => 'string',
1034 enum => [qw(none lba auto)],
1035 description => "Force disk geometry bios translation mode.",
1036 optional => 1,
1037 },
1038 snapshot => {
1039 type => 'boolean',
1040 description => "Controls qemu's snapshot mode feature."
1041 . " If activated, changes made to the disk are temporary and will"
1042 . " be discarded when the VM is shutdown.",
1043 optional => 1,
1044 },
1045 cache => {
1046 type => 'string',
1047 enum => [qw(none writethrough writeback unsafe directsync)],
1048 description => "The drive's cache mode",
1049 optional => 1,
1050 },
1051 format => get_standard_option('pve-qm-image-format'),
1052 size => {
1053 type => 'string',
1054 format => 'disk-size',
1055 format_description => 'DiskSize',
1056 description => "Disk size. This is purely informational and has no effect.",
1057 optional => 1,
1058 },
1059 backup => {
1060 type => 'boolean',
1061 description => "Whether the drive should be included when making backups.",
1062 optional => 1,
1063 },
1064 replicate => {
1065 type => 'boolean',
1066 description => 'Whether the drive should considered for replication jobs.',
1067 optional => 1,
1068 default => 1,
1069 },
1070 rerror => {
1071 type => 'string',
1072 enum => [qw(ignore report stop)],
1073 description => 'Read error action.',
1074 optional => 1,
1075 },
1076 werror => {
1077 type => 'string',
1078 enum => [qw(enospc ignore report stop)],
1079 description => 'Write error action.',
1080 optional => 1,
1081 },
1082 aio => {
1083 type => 'string',
1084 enum => [qw(native threads)],
1085 description => 'AIO type to use.',
1086 optional => 1,
1087 },
1088 discard => {
1089 type => 'string',
1090 enum => [qw(ignore on)],
1091 description => 'Controls whether to pass discard/trim requests to the underlying storage.',
1092 optional => 1,
1093 },
1094 detect_zeroes => {
1095 type => 'boolean',
1096 description => 'Controls whether to detect and try to optimize writes of zeroes.',
1097 optional => 1,
1098 },
1099 serial => {
1100 type => 'string',
1101 format => 'urlencoded',
1102 format_description => 'serial',
1103 maxLength => 20*3, # *3 since it's %xx url enoded
1104 description => "The drive's reported serial number, url-encoded, up to 20 bytes long.",
1105 optional => 1,
1106 },
1107 shared => {
1108 type => 'boolean',
1109 description => 'Mark this locally-managed volume as available on all nodes',
1110 verbose_description => "Mark this locally-managed volume as available on all nodes.\n\nWARNING: This option does not share the volume automatically, it assumes it is shared already!",
1111 optional => 1,
1112 default => 0,
1113 }
1114 );
1115
1116 my %iothread_fmt = ( iothread => {
1117 type => 'boolean',
1118 description => "Whether to use iothreads for this drive",
1119 optional => 1,
1120 });
1121
1122 my %model_fmt = (
1123 model => {
1124 type => 'string',
1125 format => 'urlencoded',
1126 format_description => 'model',
1127 maxLength => 40*3, # *3 since it's %xx url enoded
1128 description => "The drive's reported model name, url-encoded, up to 40 bytes long.",
1129 optional => 1,
1130 },
1131 );
1132
1133 my %queues_fmt = (
1134 queues => {
1135 type => 'integer',
1136 description => "Number of queues.",
1137 minimum => 2,
1138 optional => 1
1139 }
1140 );
1141
1142 my %scsiblock_fmt = (
1143 scsiblock => {
1144 type => 'boolean',
1145 description => "whether to use scsi-block for full passthrough of host block device\n\nWARNING: can lead to I/O errors in combination with low memory or high memory fragmentation on host",
1146 optional => 1,
1147 default => 0,
1148 },
1149 );
1150
1151 my %ssd_fmt = (
1152 ssd => {
1153 type => 'boolean',
1154 description => "Whether to expose this drive as an SSD, rather than a rotational hard disk.",
1155 optional => 1,
1156 },
1157 );
1158
1159 my %wwn_fmt = (
1160 wwn => {
1161 type => 'string',
1162 pattern => qr/^(0x)[0-9a-fA-F]{16}/,
1163 format_description => 'wwn',
1164 description => "The drive's worldwide name, encoded as 16 bytes hex string, prefixed by '0x'.",
1165 optional => 1,
1166 },
1167 );
1168
1169 my $add_throttle_desc = sub {
1170 my ($key, $type, $what, $unit, $longunit, $minimum) = @_;
1171 my $d = {
1172 type => $type,
1173 format_description => $unit,
1174 description => "Maximum $what in $longunit.",
1175 optional => 1,
1176 };
1177 $d->{minimum} = $minimum if defined($minimum);
1178 $drivedesc_base{$key} = $d;
1179 };
1180 # throughput: (leaky bucket)
1181 $add_throttle_desc->('bps', 'integer', 'r/w speed', 'bps', 'bytes per second');
1182 $add_throttle_desc->('bps_rd', 'integer', 'read speed', 'bps', 'bytes per second');
1183 $add_throttle_desc->('bps_wr', 'integer', 'write speed', 'bps', 'bytes per second');
1184 $add_throttle_desc->('mbps', 'number', 'r/w speed', 'mbps', 'megabytes per second');
1185 $add_throttle_desc->('mbps_rd', 'number', 'read speed', 'mbps', 'megabytes per second');
1186 $add_throttle_desc->('mbps_wr', 'number', 'write speed', 'mbps', 'megabytes per second');
1187 $add_throttle_desc->('iops', 'integer', 'r/w I/O', 'iops', 'operations per second');
1188 $add_throttle_desc->('iops_rd', 'integer', 'read I/O', 'iops', 'operations per second');
1189 $add_throttle_desc->('iops_wr', 'integer', 'write I/O', 'iops', 'operations per second');
1190
1191 # pools: (pool of IO before throttling starts taking effect)
1192 $add_throttle_desc->('mbps_max', 'number', 'unthrottled r/w pool', 'mbps', 'megabytes per second');
1193 $add_throttle_desc->('mbps_rd_max', 'number', 'unthrottled read pool', 'mbps', 'megabytes per second');
1194 $add_throttle_desc->('mbps_wr_max', 'number', 'unthrottled write pool', 'mbps', 'megabytes per second');
1195 $add_throttle_desc->('iops_max', 'integer', 'unthrottled r/w I/O pool', 'iops', 'operations per second');
1196 $add_throttle_desc->('iops_rd_max', 'integer', 'unthrottled read I/O pool', 'iops', 'operations per second');
1197 $add_throttle_desc->('iops_wr_max', 'integer', 'unthrottled write I/O pool', 'iops', 'operations per second');
1198
1199 # burst lengths
1200 $add_throttle_desc->('bps_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1201 $add_throttle_desc->('bps_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1202 $add_throttle_desc->('bps_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1203 $add_throttle_desc->('iops_max_length', 'integer', 'length of I/O bursts', 'seconds', 'seconds', 1);
1204 $add_throttle_desc->('iops_rd_max_length', 'integer', 'length of read I/O bursts', 'seconds', 'seconds', 1);
1205 $add_throttle_desc->('iops_wr_max_length', 'integer', 'length of write I/O bursts', 'seconds', 'seconds', 1);
1206
1207 # legacy support
1208 $drivedesc_base{'bps_rd_length'} = { alias => 'bps_rd_max_length' };
1209 $drivedesc_base{'bps_wr_length'} = { alias => 'bps_wr_max_length' };
1210 $drivedesc_base{'iops_rd_length'} = { alias => 'iops_rd_max_length' };
1211 $drivedesc_base{'iops_wr_length'} = { alias => 'iops_wr_max_length' };
1212
1213 my $ide_fmt = {
1214 %drivedesc_base,
1215 %model_fmt,
1216 %ssd_fmt,
1217 %wwn_fmt,
1218 };
1219 PVE::JSONSchema::register_format("pve-qm-ide", $ide_fmt);
1220
1221 my $idedesc = {
1222 optional => 1,
1223 type => 'string', format => $ide_fmt,
1224 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
1225 };
1226 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
1227
1228 my $scsi_fmt = {
1229 %drivedesc_base,
1230 %iothread_fmt,
1231 %queues_fmt,
1232 %scsiblock_fmt,
1233 %ssd_fmt,
1234 %wwn_fmt,
1235 };
1236 my $scsidesc = {
1237 optional => 1,
1238 type => 'string', format => $scsi_fmt,
1239 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
1240 };
1241 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
1242
1243 my $sata_fmt = {
1244 %drivedesc_base,
1245 %ssd_fmt,
1246 %wwn_fmt,
1247 };
1248 my $satadesc = {
1249 optional => 1,
1250 type => 'string', format => $sata_fmt,
1251 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
1252 };
1253 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
1254
1255 my $virtio_fmt = {
1256 %drivedesc_base,
1257 %iothread_fmt,
1258 };
1259 my $virtiodesc = {
1260 optional => 1,
1261 type => 'string', format => $virtio_fmt,
1262 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
1263 };
1264 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
1265
1266 my $alldrive_fmt = {
1267 %drivedesc_base,
1268 %iothread_fmt,
1269 %model_fmt,
1270 %queues_fmt,
1271 %scsiblock_fmt,
1272 %ssd_fmt,
1273 %wwn_fmt,
1274 };
1275
1276 my $efidisk_fmt = {
1277 volume => { alias => 'file' },
1278 file => {
1279 type => 'string',
1280 format => 'pve-volume-id-or-qm-path',
1281 default_key => 1,
1282 format_description => 'volume',
1283 description => "The drive's backing volume.",
1284 },
1285 format => get_standard_option('pve-qm-image-format'),
1286 size => {
1287 type => 'string',
1288 format => 'disk-size',
1289 format_description => 'DiskSize',
1290 description => "Disk size. This is purely informational and has no effect.",
1291 optional => 1,
1292 },
1293 };
1294
1295 my $efidisk_desc = {
1296 optional => 1,
1297 type => 'string', format => $efidisk_fmt,
1298 description => "Configure a Disk for storing EFI vars",
1299 };
1300
1301 PVE::JSONSchema::register_standard_option("pve-qm-efidisk", $efidisk_desc);
1302
1303 my $usb_fmt = {
1304 host => {
1305 default_key => 1,
1306 type => 'string', format => 'pve-qm-usb-device',
1307 format_description => 'HOSTUSBDEVICE|spice',
1308 description => <<EODESCR,
1309 The Host USB device or port or the value 'spice'. HOSTUSBDEVICE syntax is:
1310
1311 'bus-port(.port)*' (decimal numbers) or
1312 'vendor_id:product_id' (hexadeciaml numbers) or
1313 'spice'
1314
1315 You can use the 'lsusb -t' command to list existing usb devices.
1316
1317 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1318
1319 The value 'spice' can be used to add a usb redirection devices for spice.
1320 EODESCR
1321 },
1322 usb3 => {
1323 optional => 1,
1324 type => 'boolean',
1325 description => "Specifies whether if given host option is a USB3 device or port.",
1326 default => 0,
1327 },
1328 };
1329
1330 my $usbdesc = {
1331 optional => 1,
1332 type => 'string', format => $usb_fmt,
1333 description => "Configure an USB device (n is 0 to 4).",
1334 };
1335 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
1336
1337 my $PCIRE = qr/[a-f0-9]{2}:[a-f0-9]{2}(?:\.[a-f0-9])?/;
1338 my $hostpci_fmt = {
1339 host => {
1340 default_key => 1,
1341 type => 'string',
1342 pattern => qr/$PCIRE(;$PCIRE)*/,
1343 format_description => 'HOSTPCIID[;HOSTPCIID2...]',
1344 description => <<EODESCR,
1345 Host PCI device pass through. The PCI ID of a host's PCI device or a list
1346 of PCI virtual functions of the host. HOSTPCIID syntax is:
1347
1348 'bus:dev.func' (hexadecimal numbers)
1349
1350 You can us the 'lspci' command to list existing PCI devices.
1351 EODESCR
1352 },
1353 rombar => {
1354 type => 'boolean',
1355 description => "Specify whether or not the device's ROM will be visible in the guest's memory map.",
1356 optional => 1,
1357 default => 1,
1358 },
1359 romfile => {
1360 type => 'string',
1361 pattern => '[^,;]+',
1362 format_description => 'string',
1363 description => "Custom pci device rom filename (must be located in /usr/share/kvm/).",
1364 optional => 1,
1365 },
1366 pcie => {
1367 type => 'boolean',
1368 description => "Choose the PCI-express bus (needs the 'q35' machine model).",
1369 optional => 1,
1370 default => 0,
1371 },
1372 'x-vga' => {
1373 type => 'boolean',
1374 description => "Enable vfio-vga device support.",
1375 optional => 1,
1376 default => 0,
1377 },
1378 'mdev' => {
1379 type => 'string',
1380 format_description => 'string',
1381 pattern => '[^/\.:]+',
1382 optional => 1,
1383 description => <<EODESCR
1384 The type of mediated device to use.
1385 An instance of this type will be created on startup of the VM and
1386 will be cleaned up when the VM stops.
1387 EODESCR
1388 }
1389 };
1390 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
1391
1392 my $hostpcidesc = {
1393 optional => 1,
1394 type => 'string', format => 'pve-qm-hostpci',
1395 description => "Map host PCI devices into guest.",
1396 verbose_description => <<EODESCR,
1397 Map host PCI devices into guest.
1398
1399 NOTE: This option allows direct access to host hardware. So it is no longer
1400 possible to migrate such machines - use with special care.
1401
1402 CAUTION: Experimental! User reported problems with this option.
1403 EODESCR
1404 };
1405 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
1406
1407 my $serialdesc = {
1408 optional => 1,
1409 type => 'string',
1410 pattern => '(/dev/.+|socket)',
1411 description => "Create a serial device inside the VM (n is 0 to 3)",
1412 verbose_description => <<EODESCR,
1413 Create a serial device inside the VM (n is 0 to 3), and pass through a
1414 host serial device (i.e. /dev/ttyS0), or create a unix socket on the
1415 host side (use 'qm terminal' to open a terminal connection).
1416
1417 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
1418
1419 CAUTION: Experimental! User reported problems with this option.
1420 EODESCR
1421 };
1422
1423 my $paralleldesc= {
1424 optional => 1,
1425 type => 'string',
1426 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
1427 description => "Map host parallel devices (n is 0 to 2).",
1428 verbose_description => <<EODESCR,
1429 Map host parallel devices (n is 0 to 2).
1430
1431 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
1432
1433 CAUTION: Experimental! User reported problems with this option.
1434 EODESCR
1435 };
1436
1437 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
1438 $confdesc->{"parallel$i"} = $paralleldesc;
1439 }
1440
1441 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
1442 $confdesc->{"serial$i"} = $serialdesc;
1443 }
1444
1445 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
1446 $confdesc->{"hostpci$i"} = $hostpcidesc;
1447 }
1448
1449 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
1450 $drivename_hash->{"ide$i"} = 1;
1451 $confdesc->{"ide$i"} = $idedesc;
1452 }
1453
1454 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
1455 $drivename_hash->{"sata$i"} = 1;
1456 $confdesc->{"sata$i"} = $satadesc;
1457 }
1458
1459 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
1460 $drivename_hash->{"scsi$i"} = 1;
1461 $confdesc->{"scsi$i"} = $scsidesc ;
1462 }
1463
1464 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
1465 $drivename_hash->{"virtio$i"} = 1;
1466 $confdesc->{"virtio$i"} = $virtiodesc;
1467 }
1468
1469 $drivename_hash->{efidisk0} = 1;
1470 $confdesc->{efidisk0} = $efidisk_desc;
1471
1472 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1473 $confdesc->{"usb$i"} = $usbdesc;
1474 }
1475
1476 my $unuseddesc = {
1477 optional => 1,
1478 type => 'string', format => 'pve-volume-id',
1479 description => "Reference to unused volumes. This is used internally, and should not be modified manually.",
1480 };
1481
1482 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
1483 $confdesc->{"unused$i"} = $unuseddesc;
1484 }
1485
1486 my $kvm_api_version = 0;
1487
1488 sub kvm_version {
1489 return $kvm_api_version if $kvm_api_version;
1490
1491 open my $fh, '<', '/dev/kvm'
1492 or return undef;
1493
1494 # 0xae00 => KVM_GET_API_VERSION
1495 $kvm_api_version = ioctl($fh, 0xae00, 0);
1496
1497 return $kvm_api_version;
1498 }
1499
1500 my $kvm_user_version = {};
1501 my $kvm_mtime = {};
1502
1503 sub kvm_user_version {
1504 my ($binary) = @_;
1505
1506 $binary //= get_command_for_arch(get_host_arch()); # get the native arch by default
1507 my $st = stat($binary);
1508
1509 my $cachedmtime = $kvm_mtime->{$binary} // -1;
1510 return $kvm_user_version->{$binary} if $kvm_user_version->{$binary} &&
1511 $cachedmtime == $st->mtime;
1512
1513 $kvm_user_version->{$binary} = 'unknown';
1514 $kvm_mtime->{$binary} = $st->mtime;
1515
1516 my $code = sub {
1517 my $line = shift;
1518 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1519 $kvm_user_version->{$binary} = $2;
1520 }
1521 };
1522
1523 eval { run_command([$binary, '--version'], outfunc => $code); };
1524 warn $@ if $@;
1525
1526 return $kvm_user_version->{$binary};
1527
1528 }
1529
1530 sub kernel_has_vhost_net {
1531 return -c '/dev/vhost-net';
1532 }
1533
1534 sub valid_drive_names {
1535 # order is important - used to autoselect boot disk
1536 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
1537 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
1538 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
1539 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))),
1540 'efidisk0');
1541 }
1542
1543 sub is_valid_drivename {
1544 my $dev = shift;
1545
1546 return defined($drivename_hash->{$dev});
1547 }
1548
1549 sub option_exists {
1550 my $key = shift;
1551 return defined($confdesc->{$key});
1552 }
1553
1554 sub nic_models {
1555 return $nic_model_list;
1556 }
1557
1558 sub os_list_description {
1559
1560 return {
1561 other => 'Other',
1562 wxp => 'Windows XP',
1563 w2k => 'Windows 2000',
1564 w2k3 =>, 'Windows 2003',
1565 w2k8 => 'Windows 2008',
1566 wvista => 'Windows Vista',
1567 win7 => 'Windows 7',
1568 win8 => 'Windows 8/2012',
1569 win10 => 'Windows 10/2016',
1570 l24 => 'Linux 2.4',
1571 l26 => 'Linux 2.6',
1572 };
1573 }
1574
1575 my $cdrom_path;
1576
1577 sub get_cdrom_path {
1578
1579 return $cdrom_path if $cdrom_path;
1580
1581 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1582 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1583 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1584 }
1585
1586 sub get_iso_path {
1587 my ($storecfg, $vmid, $cdrom) = @_;
1588
1589 if ($cdrom eq 'cdrom') {
1590 return get_cdrom_path();
1591 } elsif ($cdrom eq 'none') {
1592 return '';
1593 } elsif ($cdrom =~ m|^/|) {
1594 return $cdrom;
1595 } else {
1596 return PVE::Storage::path($storecfg, $cdrom);
1597 }
1598 }
1599
1600 # try to convert old style file names to volume IDs
1601 sub filename_to_volume_id {
1602 my ($vmid, $file, $media) = @_;
1603
1604 if (!($file eq 'none' || $file eq 'cdrom' ||
1605 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1606
1607 return undef if $file =~ m|/|;
1608
1609 if ($media && $media eq 'cdrom') {
1610 $file = "local:iso/$file";
1611 } else {
1612 $file = "local:$vmid/$file";
1613 }
1614 }
1615
1616 return $file;
1617 }
1618
1619 sub verify_media_type {
1620 my ($opt, $vtype, $media) = @_;
1621
1622 return if !$media;
1623
1624 my $etype;
1625 if ($media eq 'disk') {
1626 $etype = 'images';
1627 } elsif ($media eq 'cdrom') {
1628 $etype = 'iso';
1629 } else {
1630 die "internal error";
1631 }
1632
1633 return if ($vtype eq $etype);
1634
1635 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1636 }
1637
1638 sub cleanup_drive_path {
1639 my ($opt, $storecfg, $drive) = @_;
1640
1641 # try to convert filesystem paths to volume IDs
1642
1643 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1644 ($drive->{file} !~ m|^/dev/.+|) &&
1645 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1646 ($drive->{file} !~ m/^\d+$/)) {
1647 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1648 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1649 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1650 verify_media_type($opt, $vtype, $drive->{media});
1651 $drive->{file} = $volid;
1652 }
1653
1654 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1655 }
1656
1657 sub parse_hotplug_features {
1658 my ($data) = @_;
1659
1660 my $res = {};
1661
1662 return $res if $data eq '0';
1663
1664 $data = $confdesc->{hotplug}->{default} if $data eq '1';
1665
1666 foreach my $feature (PVE::Tools::split_list($data)) {
1667 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1668 $res->{$1} = 1;
1669 } else {
1670 die "invalid hotplug feature '$feature'\n";
1671 }
1672 }
1673 return $res;
1674 }
1675
1676 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1677 sub pve_verify_hotplug_features {
1678 my ($value, $noerr) = @_;
1679
1680 return $value if parse_hotplug_features($value);
1681
1682 return undef if $noerr;
1683
1684 die "unable to parse hotplug option\n";
1685 }
1686
1687 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
1688 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
1689 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
1690 # [,aio=native|threads][,discard=ignore|on][,detect_zeroes=on|off]
1691 # [,iothread=on][,serial=serial][,model=model]
1692
1693 sub parse_drive {
1694 my ($key, $data) = @_;
1695
1696 my ($interface, $index);
1697
1698 if ($key =~ m/^([^\d]+)(\d+)$/) {
1699 $interface = $1;
1700 $index = $2;
1701 } else {
1702 return undef;
1703 }
1704
1705 my $desc = $key =~ /^unused\d+$/ ? $alldrive_fmt
1706 : $confdesc->{$key}->{format};
1707 if (!$desc) {
1708 warn "invalid drive key: $key\n";
1709 return undef;
1710 }
1711 my $res = eval { PVE::JSONSchema::parse_property_string($desc, $data) };
1712 return undef if !$res;
1713 $res->{interface} = $interface;
1714 $res->{index} = $index;
1715
1716 my $error = 0;
1717 foreach my $opt (qw(bps bps_rd bps_wr)) {
1718 if (my $bps = defined(delete $res->{$opt})) {
1719 if (defined($res->{"m$opt"})) {
1720 warn "both $opt and m$opt specified\n";
1721 ++$error;
1722 next;
1723 }
1724 $res->{"m$opt"} = sprintf("%.3f", $bps / (1024*1024.0));
1725 }
1726 }
1727
1728 # can't use the schema's 'requires' because of the mbps* => bps* "transforming aliases"
1729 for my $requirement (
1730 [mbps_max => 'mbps'],
1731 [mbps_rd_max => 'mbps_rd'],
1732 [mbps_wr_max => 'mbps_wr'],
1733 [miops_max => 'miops'],
1734 [miops_rd_max => 'miops_rd'],
1735 [miops_wr_max => 'miops_wr'],
1736 [bps_max_length => 'mbps_max'],
1737 [bps_rd_max_length => 'mbps_rd_max'],
1738 [bps_wr_max_length => 'mbps_wr_max'],
1739 [iops_max_length => 'iops_max'],
1740 [iops_rd_max_length => 'iops_rd_max'],
1741 [iops_wr_max_length => 'iops_wr_max']) {
1742 my ($option, $requires) = @$requirement;
1743 if ($res->{$option} && !$res->{$requires}) {
1744 warn "$option requires $requires\n";
1745 ++$error;
1746 }
1747 }
1748
1749 return undef if $error;
1750
1751 return undef if $res->{mbps_rd} && $res->{mbps};
1752 return undef if $res->{mbps_wr} && $res->{mbps};
1753 return undef if $res->{iops_rd} && $res->{iops};
1754 return undef if $res->{iops_wr} && $res->{iops};
1755
1756 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1757 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1758 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1759 return undef if $res->{interface} eq 'virtio';
1760 }
1761
1762 if (my $size = $res->{size}) {
1763 return undef if !defined($res->{size} = PVE::JSONSchema::parse_size($size));
1764 }
1765
1766 return $res;
1767 }
1768
1769 sub print_drive {
1770 my ($vmid, $drive) = @_;
1771 my $data = { %$drive };
1772 delete $data->{$_} for qw(index interface);
1773 return PVE::JSONSchema::print_property_string($data, $alldrive_fmt);
1774 }
1775
1776 sub scsi_inquiry {
1777 my($fh, $noerr) = @_;
1778
1779 my $SG_IO = 0x2285;
1780 my $SG_GET_VERSION_NUM = 0x2282;
1781
1782 my $versionbuf = "\x00" x 8;
1783 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1784 if (!$ret) {
1785 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1786 return undef;
1787 }
1788 my $version = unpack("I", $versionbuf);
1789 if ($version < 30000) {
1790 die "scsi generic interface too old\n" if !$noerr;
1791 return undef;
1792 }
1793
1794 my $buf = "\x00" x 36;
1795 my $sensebuf = "\x00" x 8;
1796 my $cmd = pack("C x3 C x1", 0x12, 36);
1797
1798 # see /usr/include/scsi/sg.h
1799 my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1800
1801 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1802 length($sensebuf), 0, length($buf), $buf,
1803 $cmd, $sensebuf, 6000);
1804
1805 $ret = ioctl($fh, $SG_IO, $packet);
1806 if (!$ret) {
1807 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1808 return undef;
1809 }
1810
1811 my @res = unpack($sg_io_hdr_t, $packet);
1812 if ($res[17] || $res[18]) {
1813 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1814 return undef;
1815 }
1816
1817 my $res = {};
1818 (my $byte0, my $byte1, $res->{vendor},
1819 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1820
1821 $res->{removable} = $byte1 & 128 ? 1 : 0;
1822 $res->{type} = $byte0 & 31;
1823
1824 return $res;
1825 }
1826
1827 sub path_is_scsi {
1828 my ($path) = @_;
1829
1830 my $fh = IO::File->new("+<$path") || return undef;
1831 my $res = scsi_inquiry($fh, 1);
1832 close($fh);
1833
1834 return $res;
1835 }
1836
1837 sub machine_type_is_q35 {
1838 my ($conf) = @_;
1839
1840 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1841 }
1842
1843 sub print_tabletdevice_full {
1844 my ($conf, $arch) = @_;
1845
1846 my $q35 = machine_type_is_q35($conf);
1847
1848 # we use uhci for old VMs because tablet driver was buggy in older qemu
1849 my $usbbus;
1850 if (machine_type_is_q35($conf) || $arch eq 'aarch64') {
1851 $usbbus = 'ehci';
1852 } else {
1853 $usbbus = 'uhci';
1854 }
1855
1856 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1857 }
1858
1859 sub print_keyboarddevice_full {
1860 my ($conf, $arch, $machine) = @_;
1861
1862 return undef if $arch ne 'aarch64';
1863
1864 return "usb-kbd,id=keyboard,bus=ehci.0,port=2";
1865 }
1866
1867 sub print_drivedevice_full {
1868 my ($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type) = @_;
1869
1870 my $device = '';
1871 my $maxdev = 0;
1872
1873 if ($drive->{interface} eq 'virtio') {
1874 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges, $arch, $machine_type);
1875 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1876 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1877 } elsif ($drive->{interface} eq 'scsi') {
1878
1879 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1880 my $unit = $drive->{index} % $maxdev;
1881 my $devicetype = 'hd';
1882 my $path = '';
1883 if (drive_is_cdrom($drive)) {
1884 $devicetype = 'cd';
1885 } else {
1886 if ($drive->{file} =~ m|^/|) {
1887 $path = $drive->{file};
1888 if (my $info = path_is_scsi($path)) {
1889 if ($info->{type} == 0 && $drive->{scsiblock}) {
1890 $devicetype = 'block';
1891 } elsif ($info->{type} == 1) { # tape
1892 $devicetype = 'generic';
1893 }
1894 }
1895 } else {
1896 $path = PVE::Storage::path($storecfg, $drive->{file});
1897 }
1898
1899 if($path =~ m/^iscsi\:\/\//){
1900 $devicetype = 'generic';
1901 }
1902 }
1903
1904 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1905 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1906 } else {
1907 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1908 }
1909
1910 if ($drive->{ssd} && ($devicetype eq 'block' || $devicetype eq 'hd')) {
1911 $device .= ",rotation_rate=1";
1912 }
1913 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1914
1915 } elsif ($drive->{interface} eq 'ide' || $drive->{interface} eq 'sata') {
1916 my $maxdev = ($drive->{interface} eq 'sata') ? $MAX_SATA_DISKS : 2;
1917 my $controller = int($drive->{index} / $maxdev);
1918 my $unit = $drive->{index} % $maxdev;
1919 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1920
1921 $device = "ide-$devicetype";
1922 if ($drive->{interface} eq 'ide') {
1923 $device .= ",bus=ide.$controller,unit=$unit";
1924 } else {
1925 $device .= ",bus=ahci$controller.$unit";
1926 }
1927 $device .= ",drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1928
1929 if ($devicetype eq 'hd') {
1930 if (my $model = $drive->{model}) {
1931 $model = URI::Escape::uri_unescape($model);
1932 $device .= ",model=$model";
1933 }
1934 if ($drive->{ssd}) {
1935 $device .= ",rotation_rate=1";
1936 }
1937 }
1938 $device .= ",wwn=$drive->{wwn}" if $drive->{wwn};
1939 } elsif ($drive->{interface} eq 'usb') {
1940 die "implement me";
1941 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1942 } else {
1943 die "unsupported interface type";
1944 }
1945
1946 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1947
1948 if (my $serial = $drive->{serial}) {
1949 $serial = URI::Escape::uri_unescape($serial);
1950 $device .= ",serial=$serial";
1951 }
1952
1953
1954 return $device;
1955 }
1956
1957 sub get_initiator_name {
1958 my $initiator;
1959
1960 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1961 while (defined(my $line = <$fh>)) {
1962 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1963 $initiator = $1;
1964 last;
1965 }
1966 $fh->close();
1967
1968 return $initiator;
1969 }
1970
1971 sub print_drive_full {
1972 my ($storecfg, $vmid, $drive) = @_;
1973
1974 my $path;
1975 my $volid = $drive->{file};
1976 my $format;
1977
1978 if (drive_is_cdrom($drive)) {
1979 $path = get_iso_path($storecfg, $vmid, $volid);
1980 } else {
1981 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1982 if ($storeid) {
1983 $path = PVE::Storage::path($storecfg, $volid);
1984 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1985 $format = qemu_img_format($scfg, $volname);
1986 } else {
1987 $path = $volid;
1988 $format = "raw";
1989 }
1990 }
1991
1992 my $opts = '';
1993 my @qemu_drive_options = qw(heads secs cyls trans media format cache rerror werror aio discard);
1994 foreach my $o (@qemu_drive_options) {
1995 $opts .= ",$o=$drive->{$o}" if defined($drive->{$o});
1996 }
1997
1998 # snapshot only accepts on|off
1999 if (defined($drive->{snapshot})) {
2000 my $v = $drive->{snapshot} ? 'on' : 'off';
2001 $opts .= ",snapshot=$v";
2002 }
2003
2004 foreach my $type (['', '-total'], [_rd => '-read'], [_wr => '-write']) {
2005 my ($dir, $qmpname) = @$type;
2006 if (my $v = $drive->{"mbps$dir"}) {
2007 $opts .= ",throttling.bps$qmpname=".int($v*1024*1024);
2008 }
2009 if (my $v = $drive->{"mbps${dir}_max"}) {
2010 $opts .= ",throttling.bps$qmpname-max=".int($v*1024*1024);
2011 }
2012 if (my $v = $drive->{"bps${dir}_max_length"}) {
2013 $opts .= ",throttling.bps$qmpname-max-length=$v";
2014 }
2015 if (my $v = $drive->{"iops${dir}"}) {
2016 $opts .= ",throttling.iops$qmpname=$v";
2017 }
2018 if (my $v = $drive->{"iops${dir}_max"}) {
2019 $opts .= ",throttling.iops$qmpname-max=$v";
2020 }
2021 if (my $v = $drive->{"iops${dir}_max_length"}) {
2022 $opts .= ",throttling.iops$qmpname-max-length=$v";
2023 }
2024 }
2025
2026 $opts .= ",format=$format" if $format && !$drive->{format};
2027
2028 my $cache_direct = 0;
2029
2030 if (my $cache = $drive->{cache}) {
2031 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
2032 } elsif (!drive_is_cdrom($drive)) {
2033 $opts .= ",cache=none";
2034 $cache_direct = 1;
2035 }
2036
2037 # aio native works only with O_DIRECT
2038 if (!$drive->{aio}) {
2039 if($cache_direct) {
2040 $opts .= ",aio=native";
2041 } else {
2042 $opts .= ",aio=threads";
2043 }
2044 }
2045
2046 if (!drive_is_cdrom($drive)) {
2047 my $detectzeroes;
2048 if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
2049 $detectzeroes = 'off';
2050 } elsif ($drive->{discard}) {
2051 $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
2052 } else {
2053 # This used to be our default with discard not being specified:
2054 $detectzeroes = 'on';
2055 }
2056 $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
2057 }
2058
2059 my $pathinfo = $path ? "file=$path," : '';
2060
2061 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
2062 }
2063
2064 sub print_netdevice_full {
2065 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files, $arch, $machine_type) = @_;
2066
2067 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2068
2069 my $device = $net->{model};
2070 if ($net->{model} eq 'virtio') {
2071 $device = 'virtio-net-pci';
2072 };
2073
2074 my $pciaddr = print_pci_addr("$netid", $bridges, $arch, $machine_type);
2075 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
2076 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
2077 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
2078 my $vectors = $net->{queues} * 2 + 2;
2079 $tmpstr .= ",vectors=$vectors,mq=on";
2080 }
2081 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
2082
2083 if ($use_old_bios_files) {
2084 my $romfile;
2085 if ($device eq 'virtio-net-pci') {
2086 $romfile = 'pxe-virtio.rom';
2087 } elsif ($device eq 'e1000') {
2088 $romfile = 'pxe-e1000.rom';
2089 } elsif ($device eq 'ne2k') {
2090 $romfile = 'pxe-ne2k_pci.rom';
2091 } elsif ($device eq 'pcnet') {
2092 $romfile = 'pxe-pcnet.rom';
2093 } elsif ($device eq 'rtl8139') {
2094 $romfile = 'pxe-rtl8139.rom';
2095 }
2096 $tmpstr .= ",romfile=$romfile" if $romfile;
2097 }
2098
2099 return $tmpstr;
2100 }
2101
2102 sub print_netdev_full {
2103 my ($vmid, $conf, $arch, $net, $netid, $hotplug) = @_;
2104
2105 my $i = '';
2106 if ($netid =~ m/^net(\d+)$/) {
2107 $i = int($1);
2108 }
2109
2110 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
2111
2112 my $ifname = "tap${vmid}i$i";
2113
2114 # kvm uses TUNSETIFF ioctl, and that limits ifname length
2115 die "interface name '$ifname' is too long (max 15 character)\n"
2116 if length($ifname) >= 16;
2117
2118 my $vhostparam = '';
2119 if (is_native($arch)) {
2120 $vhostparam = ',vhost=on' if kernel_has_vhost_net() && $net->{model} eq 'virtio';
2121 }
2122
2123 my $vmname = $conf->{name} || "vm$vmid";
2124
2125 my $netdev = "";
2126 my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
2127
2128 if ($net->{bridge}) {
2129 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
2130 } else {
2131 $netdev = "type=user,id=$netid,hostname=$vmname";
2132 }
2133
2134 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
2135
2136 return $netdev;
2137 }
2138
2139
2140 sub print_cpu_device {
2141 my ($conf, $id) = @_;
2142
2143 my $kvm = $conf->{kvm} // 1;
2144 my $cpu = $kvm ? "kvm64" : "qemu64";
2145 if (my $cputype = $conf->{cpu}) {
2146 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
2147 or die "Cannot parse cpu description: $cputype\n";
2148 $cpu = $cpuconf->{cputype};
2149 }
2150
2151 my $cores = $conf->{cores} || 1;
2152
2153 my $current_core = ($id - 1) % $cores;
2154 my $current_socket = int(($id - 1 - $current_core)/$cores);
2155
2156 return "$cpu-x86_64-cpu,id=cpu$id,socket-id=$current_socket,core-id=$current_core,thread-id=0";
2157 }
2158
2159 my $vga_map = {
2160 'cirrus' => 'cirrus-vga',
2161 'std' => 'VGA',
2162 'vmware' => 'vmware-svga',
2163 'virtio' => 'virtio-vga',
2164 };
2165
2166 sub print_vga_device {
2167 my ($conf, $vga, $arch, $machine, $id, $qxlnum, $bridges) = @_;
2168
2169 my $type = $vga_map->{$vga->{type}};
2170 if ($arch eq 'aarch64' && defined($type) && $type eq 'virtio-vga') {
2171 $type = 'virtio-gpu';
2172 }
2173 my $vgamem_mb = $vga->{memory};
2174 if ($qxlnum) {
2175 $type = $id ? 'qxl' : 'qxl-vga';
2176 }
2177 die "no devicetype for $vga->{type}\n" if !$type;
2178
2179 my $memory = "";
2180 if ($vgamem_mb) {
2181 if ($vga->{type} eq 'virtio') {
2182 my $bytes = PVE::Tools::convert_size($vgamem_mb, "mb" => "b");
2183 $memory = ",max_hostmem=$bytes";
2184 } elsif ($qxlnum) {
2185 # from https://www.spice-space.org/multiple-monitors.html
2186 $memory = ",vgamem_mb=$vga->{memory}";
2187 my $ram = $vgamem_mb * 4;
2188 my $vram = $vgamem_mb * 2;
2189 $memory .= ",ram_size_mb=$ram,vram_size_mb=$vram";
2190 } else {
2191 $memory = ",vgamem_mb=$vga->{memory}";
2192 }
2193 } elsif ($qxlnum && $id) {
2194 $memory = ",ram_size=67108864,vram_size=33554432";
2195 }
2196
2197 my $q35 = machine_type_is_q35($conf);
2198 my $vgaid = "vga" . ($id // '');
2199 my $pciaddr;
2200
2201 if ($q35 && $vgaid eq 'vga') {
2202 # the first display uses pcie.0 bus on q35 machines
2203 $pciaddr = print_pcie_addr($vgaid, $bridges, $arch, $machine);
2204 } else {
2205 $pciaddr = print_pci_addr($vgaid, $bridges, $arch, $machine);
2206 }
2207
2208 return "$type,id=${vgaid}${memory}${pciaddr}";
2209 }
2210
2211 sub drive_is_cloudinit {
2212 my ($drive) = @_;
2213 return $drive->{file} =~ m@[:/]vm-\d+-cloudinit(?:\.$QEMU_FORMAT_RE)?$@;
2214 }
2215
2216 sub drive_is_cdrom {
2217 my ($drive, $exclude_cloudinit) = @_;
2218
2219 return 0 if $exclude_cloudinit && drive_is_cloudinit($drive);
2220
2221 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
2222
2223 }
2224
2225 sub parse_number_sets {
2226 my ($set) = @_;
2227 my $res = [];
2228 foreach my $part (split(/;/, $set)) {
2229 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
2230 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
2231 push @$res, [ $1, $2 ];
2232 } else {
2233 die "invalid range: $part\n";
2234 }
2235 }
2236 return $res;
2237 }
2238
2239 sub parse_numa {
2240 my ($data) = @_;
2241
2242 my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
2243 $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
2244 $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
2245 return $res;
2246 }
2247
2248 sub parse_hostpci {
2249 my ($value) = @_;
2250
2251 return undef if !$value;
2252
2253 my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
2254
2255 my @idlist = split(/;/, $res->{host});
2256 delete $res->{host};
2257 foreach my $id (@idlist) {
2258 if ($id =~ m/\./) { # full id 00:00.1
2259 push @{$res->{pciid}}, {
2260 id => $id,
2261 };
2262 } else { # partial id 00:00
2263 $res->{pciid} = PVE::SysFSTools::lspci($id);
2264 }
2265 }
2266 return $res;
2267 }
2268
2269 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
2270 sub parse_net {
2271 my ($data) = @_;
2272
2273 my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
2274 if ($@) {
2275 warn $@;
2276 return undef;
2277 }
2278 if (!defined($res->{macaddr})) {
2279 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
2280 $res->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix});
2281 }
2282 return $res;
2283 }
2284
2285 # ipconfigX ip=cidr,gw=ip,ip6=cidr,gw6=ip
2286 sub parse_ipconfig {
2287 my ($data) = @_;
2288
2289 my $res = eval { PVE::JSONSchema::parse_property_string($ipconfig_fmt, $data) };
2290 if ($@) {
2291 warn $@;
2292 return undef;
2293 }
2294
2295 if ($res->{gw} && !$res->{ip}) {
2296 warn 'gateway specified without specifying an IP address';
2297 return undef;
2298 }
2299 if ($res->{gw6} && !$res->{ip6}) {
2300 warn 'IPv6 gateway specified without specifying an IPv6 address';
2301 return undef;
2302 }
2303 if ($res->{gw} && $res->{ip} eq 'dhcp') {
2304 warn 'gateway specified together with DHCP';
2305 return undef;
2306 }
2307 if ($res->{gw6} && $res->{ip6} !~ /^$IPV6RE/) {
2308 # gw6 + auto/dhcp
2309 warn "IPv6 gateway specified together with $res->{ip6} address";
2310 return undef;
2311 }
2312
2313 if (!$res->{ip} && !$res->{ip6}) {
2314 return { ip => 'dhcp', ip6 => 'dhcp' };
2315 }
2316
2317 return $res;
2318 }
2319
2320 sub print_net {
2321 my $net = shift;
2322
2323 return PVE::JSONSchema::print_property_string($net, $net_fmt);
2324 }
2325
2326 sub add_random_macs {
2327 my ($settings) = @_;
2328
2329 foreach my $opt (keys %$settings) {
2330 next if $opt !~ m/^net(\d+)$/;
2331 my $net = parse_net($settings->{$opt});
2332 next if !$net;
2333 $settings->{$opt} = print_net($net);
2334 }
2335 }
2336
2337 sub vm_is_volid_owner {
2338 my ($storecfg, $vmid, $volid) = @_;
2339
2340 if ($volid !~ m|^/|) {
2341 my ($path, $owner);
2342 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
2343 if ($owner && ($owner == $vmid)) {
2344 return 1;
2345 }
2346 }
2347
2348 return undef;
2349 }
2350
2351 sub split_flagged_list {
2352 my $text = shift || '';
2353 $text =~ s/[,;]/ /g;
2354 $text =~ s/^\s+//;
2355 return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
2356 }
2357
2358 sub join_flagged_list {
2359 my ($how, $lst) = @_;
2360 join $how, map { $lst->{$_} . $_ } keys %$lst;
2361 }
2362
2363 sub vmconfig_delete_pending_option {
2364 my ($conf, $key, $force) = @_;
2365
2366 delete $conf->{pending}->{$key};
2367 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2368 $pending_delete_hash->{$key} = $force ? '!' : '';
2369 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2370 }
2371
2372 sub vmconfig_undelete_pending_option {
2373 my ($conf, $key) = @_;
2374
2375 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
2376 delete $pending_delete_hash->{$key};
2377
2378 if (%$pending_delete_hash) {
2379 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2380 } else {
2381 delete $conf->{pending}->{delete};
2382 }
2383 }
2384
2385 sub vmconfig_register_unused_drive {
2386 my ($storecfg, $vmid, $conf, $drive) = @_;
2387
2388 if (drive_is_cloudinit($drive)) {
2389 eval { PVE::Storage::vdisk_free($storecfg, $drive->{file}) };
2390 warn $@ if $@;
2391 } elsif (!drive_is_cdrom($drive)) {
2392 my $volid = $drive->{file};
2393 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
2394 PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
2395 }
2396 }
2397 }
2398
2399 sub vmconfig_cleanup_pending {
2400 my ($conf) = @_;
2401
2402 # remove pending changes when nothing changed
2403 my $changes;
2404 foreach my $opt (keys %{$conf->{pending}}) {
2405 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
2406 $changes = 1;
2407 delete $conf->{pending}->{$opt};
2408 }
2409 }
2410
2411 my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
2412 my $pending_delete_hash = {};
2413 while (my ($opt, $force) = each %$current_delete_hash) {
2414 if (defined($conf->{$opt})) {
2415 $pending_delete_hash->{$opt} = $force;
2416 } else {
2417 $changes = 1;
2418 }
2419 }
2420
2421 if (%$pending_delete_hash) {
2422 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
2423 } else {
2424 delete $conf->{pending}->{delete};
2425 }
2426
2427 return $changes;
2428 }
2429
2430 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str][,base64=bool]
2431 my $smbios1_fmt = {
2432 uuid => {
2433 type => 'string',
2434 pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
2435 format_description => 'UUID',
2436 description => "Set SMBIOS1 UUID.",
2437 optional => 1,
2438 },
2439 version => {
2440 type => 'string',
2441 pattern => '[A-Za-z0-9+\/]+={0,2}',
2442 format_description => 'Base64 encoded string',
2443 description => "Set SMBIOS1 version.",
2444 optional => 1,
2445 },
2446 serial => {
2447 type => 'string',
2448 pattern => '[A-Za-z0-9+\/]+={0,2}',
2449 format_description => 'Base64 encoded string',
2450 description => "Set SMBIOS1 serial number.",
2451 optional => 1,
2452 },
2453 manufacturer => {
2454 type => 'string',
2455 pattern => '[A-Za-z0-9+\/]+={0,2}',
2456 format_description => 'Base64 encoded string',
2457 description => "Set SMBIOS1 manufacturer.",
2458 optional => 1,
2459 },
2460 product => {
2461 type => 'string',
2462 pattern => '[A-Za-z0-9+\/]+={0,2}',
2463 format_description => 'Base64 encoded string',
2464 description => "Set SMBIOS1 product ID.",
2465 optional => 1,
2466 },
2467 sku => {
2468 type => 'string',
2469 pattern => '[A-Za-z0-9+\/]+={0,2}',
2470 format_description => 'Base64 encoded string',
2471 description => "Set SMBIOS1 SKU string.",
2472 optional => 1,
2473 },
2474 family => {
2475 type => 'string',
2476 pattern => '[A-Za-z0-9+\/]+={0,2}',
2477 format_description => 'Base64 encoded string',
2478 description => "Set SMBIOS1 family string.",
2479 optional => 1,
2480 },
2481 base64 => {
2482 type => 'boolean',
2483 description => 'Flag to indicate that the SMBIOS values are base64 encoded',
2484 optional => 1,
2485 },
2486 };
2487
2488 sub parse_smbios1 {
2489 my ($data) = @_;
2490
2491 my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
2492 warn $@ if $@;
2493 return $res;
2494 }
2495
2496 sub print_smbios1 {
2497 my ($smbios1) = @_;
2498 return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
2499 }
2500
2501 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
2502
2503 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
2504 sub verify_bootdisk {
2505 my ($value, $noerr) = @_;
2506
2507 return $value if is_valid_drivename($value);
2508
2509 return undef if $noerr;
2510
2511 die "invalid boot disk '$value'\n";
2512 }
2513
2514 sub parse_watchdog {
2515 my ($value) = @_;
2516
2517 return undef if !$value;
2518
2519 my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
2520 warn $@ if $@;
2521 return $res;
2522 }
2523
2524 sub parse_guest_agent {
2525 my ($value) = @_;
2526
2527 return {} if !defined($value->{agent});
2528
2529 my $res = eval { PVE::JSONSchema::parse_property_string($agent_fmt, $value->{agent}) };
2530 warn $@ if $@;
2531
2532 # if the agent is disabled ignore the other potentially set properties
2533 return {} if !$res->{enabled};
2534 return $res;
2535 }
2536
2537 sub parse_vga {
2538 my ($value) = @_;
2539
2540 return {} if !$value;
2541 my $res = eval { PVE::JSONSchema::parse_property_string($vga_fmt, $value) };
2542 warn $@ if $@;
2543 return $res;
2544 }
2545
2546 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
2547 sub verify_usb_device {
2548 my ($value, $noerr) = @_;
2549
2550 return $value if parse_usb_device($value);
2551
2552 return undef if $noerr;
2553
2554 die "unable to parse usb device\n";
2555 }
2556
2557 # add JSON properties for create and set function
2558 sub json_config_properties {
2559 my $prop = shift;
2560
2561 foreach my $opt (keys %$confdesc) {
2562 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate' || $opt eq 'runningmachine';
2563 $prop->{$opt} = $confdesc->{$opt};
2564 }
2565
2566 return $prop;
2567 }
2568
2569 # return copy of $confdesc_cloudinit to generate documentation
2570 sub cloudinit_config_properties {
2571
2572 return dclone($confdesc_cloudinit);
2573 }
2574
2575 sub check_type {
2576 my ($key, $value) = @_;
2577
2578 die "unknown setting '$key'\n" if !$confdesc->{$key};
2579
2580 my $type = $confdesc->{$key}->{type};
2581
2582 if (!defined($value)) {
2583 die "got undefined value\n";
2584 }
2585
2586 if ($value =~ m/[\n\r]/) {
2587 die "property contains a line feed\n";
2588 }
2589
2590 if ($type eq 'boolean') {
2591 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
2592 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
2593 die "type check ('boolean') failed - got '$value'\n";
2594 } elsif ($type eq 'integer') {
2595 return int($1) if $value =~ m/^(\d+)$/;
2596 die "type check ('integer') failed - got '$value'\n";
2597 } elsif ($type eq 'number') {
2598 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
2599 die "type check ('number') failed - got '$value'\n";
2600 } elsif ($type eq 'string') {
2601 if (my $fmt = $confdesc->{$key}->{format}) {
2602 PVE::JSONSchema::check_format($fmt, $value);
2603 return $value;
2604 }
2605 $value =~ s/^\"(.*)\"$/$1/;
2606 return $value;
2607 } else {
2608 die "internal error"
2609 }
2610 }
2611
2612 sub touch_config {
2613 my ($vmid) = @_;
2614
2615 my $conf = PVE::QemuConfig->config_file($vmid);
2616 utime undef, undef, $conf;
2617 }
2618
2619 sub destroy_vm {
2620 my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
2621
2622 my $conffile = PVE::QemuConfig->config_file($vmid);
2623
2624 my $conf = PVE::QemuConfig->load_config($vmid);
2625
2626 PVE::QemuConfig->check_lock($conf) if !$skiplock;
2627
2628 if ($conf->{template}) {
2629 # check if any base image is still used by a linked clone
2630 foreach_drive($conf, sub {
2631 my ($ds, $drive) = @_;
2632
2633 return if drive_is_cdrom($drive);
2634
2635 my $volid = $drive->{file};
2636
2637 return if !$volid || $volid =~ m|^/|;
2638
2639 die "base volume '$volid' is still in use by linked cloned\n"
2640 if PVE::Storage::volume_is_base_and_used($storecfg, $volid);
2641
2642 });
2643 }
2644
2645 # only remove disks owned by this VM
2646 foreach_drive($conf, sub {
2647 my ($ds, $drive) = @_;
2648
2649 return if drive_is_cdrom($drive, 1);
2650
2651 my $volid = $drive->{file};
2652
2653 return if !$volid || $volid =~ m|^/|;
2654
2655 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
2656 return if !$path || !$owner || ($owner != $vmid);
2657
2658 eval {
2659 PVE::Storage::vdisk_free($storecfg, $volid);
2660 };
2661 warn "Could not remove disk '$volid', check manually: $@" if $@;
2662
2663 });
2664
2665 if ($keep_empty_config) {
2666 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
2667 } else {
2668 unlink $conffile;
2669 }
2670
2671 # also remove unused disk
2672 eval {
2673 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
2674
2675 eval {
2676 PVE::Storage::foreach_volid($dl, sub {
2677 my ($volid, $sid, $volname, $d) = @_;
2678 PVE::Storage::vdisk_free($storecfg, $volid);
2679 });
2680 };
2681 warn $@ if $@;
2682
2683 };
2684 warn $@ if $@;
2685 }
2686
2687 sub parse_vm_config {
2688 my ($filename, $raw) = @_;
2689
2690 return undef if !defined($raw);
2691
2692 my $res = {
2693 digest => Digest::SHA::sha1_hex($raw),
2694 snapshots => {},
2695 pending => {},
2696 };
2697
2698 $filename =~ m|/qemu-server/(\d+)\.conf$|
2699 || die "got strange filename '$filename'";
2700
2701 my $vmid = $1;
2702
2703 my $conf = $res;
2704 my $descr;
2705 my $section = '';
2706
2707 my @lines = split(/\n/, $raw);
2708 foreach my $line (@lines) {
2709 next if $line =~ m/^\s*$/;
2710
2711 if ($line =~ m/^\[PENDING\]\s*$/i) {
2712 $section = 'pending';
2713 if (defined($descr)) {
2714 $descr =~ s/\s+$//;
2715 $conf->{description} = $descr;
2716 }
2717 $descr = undef;
2718 $conf = $res->{$section} = {};
2719 next;
2720
2721 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2722 $section = $1;
2723 if (defined($descr)) {
2724 $descr =~ s/\s+$//;
2725 $conf->{description} = $descr;
2726 }
2727 $descr = undef;
2728 $conf = $res->{snapshots}->{$section} = {};
2729 next;
2730 }
2731
2732 if ($line =~ m/^\#(.*)\s*$/) {
2733 $descr = '' if !defined($descr);
2734 $descr .= PVE::Tools::decode_text($1) . "\n";
2735 next;
2736 }
2737
2738 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2739 $descr = '' if !defined($descr);
2740 $descr .= PVE::Tools::decode_text($2);
2741 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2742 $conf->{snapstate} = $1;
2743 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2744 my $key = $1;
2745 my $value = $2;
2746 $conf->{$key} = $value;
2747 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2748 my $value = $1;
2749 if ($section eq 'pending') {
2750 $conf->{delete} = $value; # we parse this later
2751 } else {
2752 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2753 }
2754 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(.+?)\s*$/) {
2755 my $key = $1;
2756 my $value = $2;
2757 eval { $value = check_type($key, $value); };
2758 if ($@) {
2759 warn "vm $vmid - unable to parse value of '$key' - $@";
2760 } else {
2761 $key = 'ide2' if $key eq 'cdrom';
2762 my $fmt = $confdesc->{$key}->{format};
2763 if ($fmt && $fmt =~ /^pve-qm-(?:ide|scsi|virtio|sata)$/) {
2764 my $v = parse_drive($key, $value);
2765 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2766 $v->{file} = $volid;
2767 $value = print_drive($vmid, $v);
2768 } else {
2769 warn "vm $vmid - unable to parse value of '$key'\n";
2770 next;
2771 }
2772 }
2773
2774 $conf->{$key} = $value;
2775 }
2776 }
2777 }
2778
2779 if (defined($descr)) {
2780 $descr =~ s/\s+$//;
2781 $conf->{description} = $descr;
2782 }
2783 delete $res->{snapstate}; # just to be sure
2784
2785 return $res;
2786 }
2787
2788 sub write_vm_config {
2789 my ($filename, $conf) = @_;
2790
2791 delete $conf->{snapstate}; # just to be sure
2792
2793 if ($conf->{cdrom}) {
2794 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2795 $conf->{ide2} = $conf->{cdrom};
2796 delete $conf->{cdrom};
2797 }
2798
2799 # we do not use 'smp' any longer
2800 if ($conf->{sockets}) {
2801 delete $conf->{smp};
2802 } elsif ($conf->{smp}) {
2803 $conf->{sockets} = $conf->{smp};
2804 delete $conf->{cores};
2805 delete $conf->{smp};
2806 }
2807
2808 my $used_volids = {};
2809
2810 my $cleanup_config = sub {
2811 my ($cref, $pending, $snapname) = @_;
2812
2813 foreach my $key (keys %$cref) {
2814 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2815 $key eq 'snapstate' || $key eq 'pending';
2816 my $value = $cref->{$key};
2817 if ($key eq 'delete') {
2818 die "propertry 'delete' is only allowed in [PENDING]\n"
2819 if !$pending;
2820 # fixme: check syntax?
2821 next;
2822 }
2823 eval { $value = check_type($key, $value); };
2824 die "unable to parse value of '$key' - $@" if $@;
2825
2826 $cref->{$key} = $value;
2827
2828 if (!$snapname && is_valid_drivename($key)) {
2829 my $drive = parse_drive($key, $value);
2830 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2831 }
2832 }
2833 };
2834
2835 &$cleanup_config($conf);
2836
2837 &$cleanup_config($conf->{pending}, 1);
2838
2839 foreach my $snapname (keys %{$conf->{snapshots}}) {
2840 die "internal error" if $snapname eq 'pending';
2841 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2842 }
2843
2844 # remove 'unusedX' settings if we re-add a volume
2845 foreach my $key (keys %$conf) {
2846 my $value = $conf->{$key};
2847 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2848 delete $conf->{$key};
2849 }
2850 }
2851
2852 my $generate_raw_config = sub {
2853 my ($conf, $pending) = @_;
2854
2855 my $raw = '';
2856
2857 # add description as comment to top of file
2858 if (defined(my $descr = $conf->{description})) {
2859 if ($descr) {
2860 foreach my $cl (split(/\n/, $descr)) {
2861 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2862 }
2863 } else {
2864 $raw .= "#\n" if $pending;
2865 }
2866 }
2867
2868 foreach my $key (sort keys %$conf) {
2869 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2870 $raw .= "$key: $conf->{$key}\n";
2871 }
2872 return $raw;
2873 };
2874
2875 my $raw = &$generate_raw_config($conf);
2876
2877 if (scalar(keys %{$conf->{pending}})){
2878 $raw .= "\n[PENDING]\n";
2879 $raw .= &$generate_raw_config($conf->{pending}, 1);
2880 }
2881
2882 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2883 $raw .= "\n[$snapname]\n";
2884 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2885 }
2886
2887 return $raw;
2888 }
2889
2890 sub load_defaults {
2891
2892 my $res = {};
2893
2894 # we use static defaults from our JSON schema configuration
2895 foreach my $key (keys %$confdesc) {
2896 if (defined(my $default = $confdesc->{$key}->{default})) {
2897 $res->{$key} = $default;
2898 }
2899 }
2900
2901 return $res;
2902 }
2903
2904 sub config_list {
2905 my $vmlist = PVE::Cluster::get_vmlist();
2906 my $res = {};
2907 return $res if !$vmlist || !$vmlist->{ids};
2908 my $ids = $vmlist->{ids};
2909
2910 foreach my $vmid (keys %$ids) {
2911 my $d = $ids->{$vmid};
2912 next if !$d->{node} || $d->{node} ne $nodename;
2913 next if !$d->{type} || $d->{type} ne 'qemu';
2914 $res->{$vmid}->{exists} = 1;
2915 }
2916 return $res;
2917 }
2918
2919 # test if VM uses local resources (to prevent migration)
2920 sub check_local_resources {
2921 my ($conf, $noerr) = @_;
2922
2923 my @loc_res = ();
2924
2925 push @loc_res, "hostusb" if $conf->{hostusb}; # old syntax
2926 push @loc_res, "hostpci" if $conf->{hostpci}; # old syntax
2927
2928 push @loc_res, "ivshmem" if $conf->{ivshmem};
2929
2930 foreach my $k (keys %$conf) {
2931 next if $k =~ m/^usb/ && ($conf->{$k} =~ m/^spice(?![^,])/);
2932 # sockets are safe: they will recreated be on the target side post-migrate
2933 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2934 push @loc_res, $k if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2935 }
2936
2937 die "VM uses local resources\n" if scalar @loc_res && !$noerr;
2938
2939 return \@loc_res;
2940 }
2941
2942 # check if used storages are available on all nodes (use by migrate)
2943 sub check_storage_availability {
2944 my ($storecfg, $conf, $node) = @_;
2945
2946 foreach_drive($conf, sub {
2947 my ($ds, $drive) = @_;
2948
2949 my $volid = $drive->{file};
2950 return if !$volid;
2951
2952 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2953 return if !$sid;
2954
2955 # check if storage is available on both nodes
2956 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2957 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2958 });
2959 }
2960
2961 # list nodes where all VM images are available (used by has_feature API)
2962 sub shared_nodes {
2963 my ($conf, $storecfg) = @_;
2964
2965 my $nodelist = PVE::Cluster::get_nodelist();
2966 my $nodehash = { map { $_ => 1 } @$nodelist };
2967 my $nodename = PVE::INotify::nodename();
2968
2969 foreach_drive($conf, sub {
2970 my ($ds, $drive) = @_;
2971
2972 my $volid = $drive->{file};
2973 return if !$volid;
2974
2975 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2976 if ($storeid) {
2977 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2978 if ($scfg->{disable}) {
2979 $nodehash = {};
2980 } elsif (my $avail = $scfg->{nodes}) {
2981 foreach my $node (keys %$nodehash) {
2982 delete $nodehash->{$node} if !$avail->{$node};
2983 }
2984 } elsif (!$scfg->{shared}) {
2985 foreach my $node (keys %$nodehash) {
2986 delete $nodehash->{$node} if $node ne $nodename
2987 }
2988 }
2989 }
2990 });
2991
2992 return $nodehash
2993 }
2994
2995 sub check_local_storage_availability {
2996 my ($conf, $storecfg) = @_;
2997
2998 my $nodelist = PVE::Cluster::get_nodelist();
2999 my $nodehash = { map { $_ => {} } @$nodelist };
3000
3001 foreach_drive($conf, sub {
3002 my ($ds, $drive) = @_;
3003
3004 my $volid = $drive->{file};
3005 return if !$volid;
3006
3007 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
3008 if ($storeid) {
3009 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3010
3011 if ($scfg->{disable}) {
3012 foreach my $node (keys %$nodehash) {
3013 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3014 }
3015 } elsif (my $avail = $scfg->{nodes}) {
3016 foreach my $node (keys %$nodehash) {
3017 if (!$avail->{$node}) {
3018 $nodehash->{$node}->{unavailable_storages}->{$storeid} = 1;
3019 }
3020 }
3021 }
3022 }
3023 });
3024
3025 foreach my $node (values %$nodehash) {
3026 if (my $unavail = $node->{unavailable_storages}) {
3027 $node->{unavailable_storages} = [ sort keys %$unavail ];
3028 }
3029 }
3030
3031 return $nodehash
3032 }
3033
3034 sub check_cmdline {
3035 my ($pidfile, $pid) = @_;
3036
3037 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
3038 if (defined($fh)) {
3039 my $line = <$fh>;
3040 $fh->close;
3041 return undef if !$line;
3042 my @param = split(/\0/, $line);
3043
3044 my $cmd = $param[0];
3045 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m@(?:^|/)qemu-system-[^/]+$@);
3046
3047 for (my $i = 0; $i < scalar (@param); $i++) {
3048 my $p = $param[$i];
3049 next if !$p;
3050 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
3051 my $p = $param[$i+1];
3052 return 1 if $p && ($p eq $pidfile);
3053 return undef;
3054 }
3055 }
3056 }
3057 return undef;
3058 }
3059
3060 sub check_running {
3061 my ($vmid, $nocheck, $node) = @_;
3062
3063 my $filename = PVE::QemuConfig->config_file($vmid, $node);
3064
3065 die "unable to find configuration file for VM $vmid - no such machine\n"
3066 if !$nocheck && ! -f $filename;
3067
3068 my $pidfile = pidfile_name($vmid);
3069
3070 if (my $fd = IO::File->new("<$pidfile")) {
3071 my $st = stat($fd);
3072 my $line = <$fd>;
3073 close($fd);
3074
3075 my $mtime = $st->mtime;
3076 if ($mtime > time()) {
3077 warn "file '$filename' modified in future\n";
3078 }
3079
3080 if ($line =~ m/^(\d+)$/) {
3081 my $pid = $1;
3082 if (check_cmdline($pidfile, $pid)) {
3083 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
3084 return $pid;
3085 }
3086 }
3087 }
3088 }
3089
3090 return undef;
3091 }
3092
3093 sub vzlist {
3094
3095 my $vzlist = config_list();
3096
3097 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
3098
3099 while (defined(my $de = $fd->read)) {
3100 next if $de !~ m/^(\d+)\.pid$/;
3101 my $vmid = $1;
3102 next if !defined($vzlist->{$vmid});
3103 if (my $pid = check_running($vmid)) {
3104 $vzlist->{$vmid}->{pid} = $pid;
3105 }
3106 }
3107
3108 return $vzlist;
3109 }
3110
3111 sub disksize {
3112 my ($storecfg, $conf) = @_;
3113
3114 my $bootdisk = $conf->{bootdisk};
3115 return undef if !$bootdisk;
3116 return undef if !is_valid_drivename($bootdisk);
3117
3118 return undef if !$conf->{$bootdisk};
3119
3120 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
3121 return undef if !defined($drive);
3122
3123 return undef if drive_is_cdrom($drive);
3124
3125 my $volid = $drive->{file};
3126 return undef if !$volid;
3127
3128 return $drive->{size};
3129 }
3130
3131 our $vmstatus_return_properties = {
3132 vmid => get_standard_option('pve-vmid'),
3133 status => {
3134 description => "Qemu process status.",
3135 type => 'string',
3136 enum => ['stopped', 'running'],
3137 },
3138 maxmem => {
3139 description => "Maximum memory in bytes.",
3140 type => 'integer',
3141 optional => 1,
3142 renderer => 'bytes',
3143 },
3144 maxdisk => {
3145 description => "Root disk size in bytes.",
3146 type => 'integer',
3147 optional => 1,
3148 renderer => 'bytes',
3149 },
3150 name => {
3151 description => "VM name.",
3152 type => 'string',
3153 optional => 1,
3154 },
3155 qmpstatus => {
3156 description => "Qemu QMP agent status.",
3157 type => 'string',
3158 optional => 1,
3159 },
3160 pid => {
3161 description => "PID of running qemu process.",
3162 type => 'integer',
3163 optional => 1,
3164 },
3165 uptime => {
3166 description => "Uptime.",
3167 type => 'integer',
3168 optional => 1,
3169 renderer => 'duration',
3170 },
3171 cpus => {
3172 description => "Maximum usable CPUs.",
3173 type => 'number',
3174 optional => 1,
3175 },
3176 lock => {
3177 description => "The current config lock, if any.",
3178 type => 'string',
3179 optional => 1,
3180 }
3181 };
3182
3183 my $last_proc_pid_stat;
3184
3185 # get VM status information
3186 # This must be fast and should not block ($full == false)
3187 # We only query KVM using QMP if $full == true (this can be slow)
3188 sub vmstatus {
3189 my ($opt_vmid, $full) = @_;
3190
3191 my $res = {};
3192
3193 my $storecfg = PVE::Storage::config();
3194
3195 my $list = vzlist();
3196 my $defaults = load_defaults();
3197
3198 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
3199
3200 my $cpucount = $cpuinfo->{cpus} || 1;
3201
3202 foreach my $vmid (keys %$list) {
3203 next if $opt_vmid && ($vmid ne $opt_vmid);
3204
3205 my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
3206 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
3207
3208 my $d = { vmid => $vmid };
3209 $d->{pid} = $list->{$vmid}->{pid};
3210
3211 # fixme: better status?
3212 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
3213
3214 my $size = disksize($storecfg, $conf);
3215 if (defined($size)) {
3216 $d->{disk} = 0; # no info available
3217 $d->{maxdisk} = $size;
3218 } else {
3219 $d->{disk} = 0;
3220 $d->{maxdisk} = 0;
3221 }
3222
3223 $d->{cpus} = ($conf->{sockets} || $defaults->{sockets})
3224 * ($conf->{cores} || $defaults->{cores});
3225 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
3226 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
3227
3228 $d->{name} = $conf->{name} || "VM $vmid";
3229 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024)
3230 : $defaults->{memory}*(1024*1024);
3231
3232 if ($conf->{balloon}) {
3233 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
3234 $d->{shares} = defined($conf->{shares}) ? $conf->{shares}
3235 : $defaults->{shares};
3236 }
3237
3238 $d->{uptime} = 0;
3239 $d->{cpu} = 0;
3240 $d->{mem} = 0;
3241
3242 $d->{netout} = 0;
3243 $d->{netin} = 0;
3244
3245 $d->{diskread} = 0;
3246 $d->{diskwrite} = 0;
3247
3248 $d->{template} = PVE::QemuConfig->is_template($conf);
3249
3250 $d->{serial} = 1 if conf_has_serial($conf);
3251 $d->{lock} = $conf->{lock} if $conf->{lock};
3252
3253 $res->{$vmid} = $d;
3254 }
3255
3256 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
3257 foreach my $dev (keys %$netdev) {
3258 next if $dev !~ m/^tap([1-9]\d*)i/;
3259 my $vmid = $1;
3260 my $d = $res->{$vmid};
3261 next if !$d;
3262
3263 $d->{netout} += $netdev->{$dev}->{receive};
3264 $d->{netin} += $netdev->{$dev}->{transmit};
3265
3266 if ($full) {
3267 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
3268 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
3269 }
3270
3271 }
3272
3273 my $ctime = gettimeofday;
3274
3275 foreach my $vmid (keys %$list) {
3276
3277 my $d = $res->{$vmid};
3278 my $pid = $d->{pid};
3279 next if !$pid;
3280
3281 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
3282 next if !$pstat; # not running
3283
3284 my $used = $pstat->{utime} + $pstat->{stime};
3285
3286 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
3287
3288 if ($pstat->{vsize}) {
3289 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
3290 }
3291
3292 my $old = $last_proc_pid_stat->{$pid};
3293 if (!$old) {
3294 $last_proc_pid_stat->{$pid} = {
3295 time => $ctime,
3296 used => $used,
3297 cpu => 0,
3298 };
3299 next;
3300 }
3301
3302 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
3303
3304 if ($dtime > 1000) {
3305 my $dutime = $used - $old->{used};
3306
3307 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
3308 $last_proc_pid_stat->{$pid} = {
3309 time => $ctime,
3310 used => $used,
3311 cpu => $d->{cpu},
3312 };
3313 } else {
3314 $d->{cpu} = $old->{cpu};
3315 }
3316 }
3317
3318 return $res if !$full;
3319
3320 my $qmpclient = PVE::QMPClient->new();
3321
3322 my $ballooncb = sub {
3323 my ($vmid, $resp) = @_;
3324
3325 my $info = $resp->{'return'};
3326 return if !$info->{max_mem};
3327
3328 my $d = $res->{$vmid};
3329
3330 # use memory assigned to VM
3331 $d->{maxmem} = $info->{max_mem};
3332 $d->{balloon} = $info->{actual};
3333
3334 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
3335 $d->{mem} = $info->{total_mem} - $info->{free_mem};
3336 $d->{freemem} = $info->{free_mem};
3337 }
3338
3339 $d->{ballooninfo} = $info;
3340 };
3341
3342 my $blockstatscb = sub {
3343 my ($vmid, $resp) = @_;
3344 my $data = $resp->{'return'} || [];
3345 my $totalrdbytes = 0;
3346 my $totalwrbytes = 0;
3347
3348 for my $blockstat (@$data) {
3349 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
3350 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
3351
3352 $blockstat->{device} =~ s/drive-//;
3353 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
3354 }
3355 $res->{$vmid}->{diskread} = $totalrdbytes;
3356 $res->{$vmid}->{diskwrite} = $totalwrbytes;
3357 };
3358
3359 my $statuscb = sub {
3360 my ($vmid, $resp) = @_;
3361
3362 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
3363 # this fails if ballon driver is not loaded, so this must be
3364 # the last commnand (following command are aborted if this fails).
3365 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
3366
3367 my $status = 'unknown';
3368 if (!defined($status = $resp->{'return'}->{status})) {
3369 warn "unable to get VM status\n";
3370 return;
3371 }
3372
3373 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
3374 };
3375
3376 foreach my $vmid (keys %$list) {
3377 next if $opt_vmid && ($vmid ne $opt_vmid);
3378 next if !$res->{$vmid}->{pid}; # not running
3379 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
3380 }
3381
3382 $qmpclient->queue_execute(undef, 2);
3383
3384 foreach my $vmid (keys %$list) {
3385 next if $opt_vmid && ($vmid ne $opt_vmid);
3386 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
3387 }
3388
3389 return $res;
3390 }
3391
3392 sub foreach_drive {
3393 my ($conf, $func, @param) = @_;
3394
3395 foreach my $ds (valid_drive_names()) {
3396 next if !defined($conf->{$ds});
3397
3398 my $drive = parse_drive($ds, $conf->{$ds});
3399 next if !$drive;
3400
3401 &$func($ds, $drive, @param);
3402 }
3403 }
3404
3405 sub foreach_volid {
3406 my ($conf, $func, @param) = @_;
3407
3408 my $volhash = {};
3409
3410 my $test_volid = sub {
3411 my ($volid, $is_cdrom, $replicate, $shared, $snapname, $size) = @_;
3412
3413 return if !$volid;
3414
3415 $volhash->{$volid}->{cdrom} //= 1;
3416 $volhash->{$volid}->{cdrom} = 0 if !$is_cdrom;
3417
3418 $volhash->{$volid}->{replicate} //= 0;
3419 $volhash->{$volid}->{replicate} = 1 if $replicate;
3420
3421 $volhash->{$volid}->{shared} //= 0;
3422 $volhash->{$volid}->{shared} = 1 if $shared;
3423
3424 $volhash->{$volid}->{referenced_in_config} //= 0;
3425 $volhash->{$volid}->{referenced_in_config} = 1 if !defined($snapname);
3426
3427 $volhash->{$volid}->{referenced_in_snapshot}->{$snapname} = 1
3428 if defined($snapname);
3429 $volhash->{$volid}->{size} = $size if $size;
3430 };
3431
3432 foreach_drive($conf, sub {
3433 my ($ds, $drive) = @_;
3434 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, undef, $drive->{size});
3435 });
3436
3437 foreach my $snapname (keys %{$conf->{snapshots}}) {
3438 my $snap = $conf->{snapshots}->{$snapname};
3439 $test_volid->($snap->{vmstate}, 0, 1, $snapname);
3440 foreach_drive($snap, sub {
3441 my ($ds, $drive) = @_;
3442 $test_volid->($drive->{file}, drive_is_cdrom($drive), $drive->{replicate} // 1, $drive->{shared}, $snapname);
3443 });
3444 }
3445
3446 foreach my $volid (keys %$volhash) {
3447 &$func($volid, $volhash->{$volid}, @param);
3448 }
3449 }
3450
3451 sub conf_has_serial {
3452 my ($conf) = @_;
3453
3454 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3455 if ($conf->{"serial$i"}) {
3456 return 1;
3457 }
3458 }
3459
3460 return 0;
3461 }
3462
3463 sub conf_has_audio {
3464 my ($conf, $id) = @_;
3465
3466 $id //= 0;
3467 my $audio = $conf->{"audio$id"};
3468 return undef if !defined($audio);
3469
3470 my $audioproperties = PVE::JSONSchema::parse_property_string($audio_fmt, $audio);
3471 my $audiodriver = $audioproperties->{driver} // 'spice';
3472
3473 return {
3474 dev => $audioproperties->{device},
3475 dev_id => "audiodev$id",
3476 backend => $audiodriver,
3477 backend_id => "$audiodriver-backend${id}",
3478 };
3479 }
3480
3481 sub vga_conf_has_spice {
3482 my ($vga) = @_;
3483
3484 my $vgaconf = parse_vga($vga);
3485 my $vgatype = $vgaconf->{type};
3486 return 0 if !$vgatype || $vgatype !~ m/^qxl([234])?$/;
3487
3488 return $1 || 1;
3489 }
3490
3491 my $host_arch; # FIXME: fix PVE::Tools::get_host_arch
3492 sub get_host_arch() {
3493 $host_arch = (POSIX::uname())[4] if !$host_arch;
3494 return $host_arch;
3495 }
3496
3497 sub is_native($) {
3498 my ($arch) = @_;
3499 return get_host_arch() eq $arch;
3500 }
3501
3502 my $default_machines = {
3503 x86_64 => 'pc',
3504 aarch64 => 'virt',
3505 };
3506
3507 sub get_basic_machine_info {
3508 my ($conf, $forcemachine) = @_;
3509
3510 my $arch = $conf->{arch} // get_host_arch();
3511 my $machine = $forcemachine || $conf->{machine} || $default_machines->{$arch};
3512 return ($arch, $machine);
3513 }
3514
3515 sub get_ovmf_files($) {
3516 my ($arch) = @_;
3517
3518 my $ovmf = $OVMF->{$arch}
3519 or die "no OVMF images known for architecture '$arch'\n";
3520
3521 return @$ovmf;
3522 }
3523
3524 my $Arch2Qemu = {
3525 aarch64 => '/usr/bin/qemu-system-aarch64',
3526 x86_64 => '/usr/bin/qemu-system-x86_64',
3527 };
3528 sub get_command_for_arch($) {
3529 my ($arch) = @_;
3530 return '/usr/bin/kvm' if is_native($arch);
3531
3532 my $cmd = $Arch2Qemu->{$arch}
3533 or die "don't know how to emulate architecture '$arch'\n";
3534 return $cmd;
3535 }
3536
3537 sub get_cpu_options {
3538 my ($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough) = @_;
3539
3540 my $cpuFlags = [];
3541 my $ostype = $conf->{ostype};
3542
3543 my $cpu = $kvm ? "kvm64" : "qemu64";
3544 if ($arch eq 'aarch64') {
3545 $cpu = 'cortex-a57';
3546 }
3547 my $hv_vendor_id;
3548 if (my $cputype = $conf->{cpu}) {
3549 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3550 or die "Cannot parse cpu description: $cputype\n";
3551 $cpu = $cpuconf->{cputype};
3552 $kvm_off = 1 if $cpuconf->{hidden};
3553 $hv_vendor_id = $cpuconf->{'hv-vendor-id'};
3554
3555 if (defined(my $flags = $cpuconf->{flags})) {
3556 push @$cpuFlags, split(";", $flags);
3557 }
3558 }
3559
3560 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64' && $arch eq 'x86_64';
3561
3562 push @$cpuFlags , '-x2apic'
3563 if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3564
3565 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3566
3567 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3568
3569 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3) && $arch eq 'x86_64') {
3570
3571 push @$cpuFlags , '+kvm_pv_unhalt' if $kvm;
3572 push @$cpuFlags , '+kvm_pv_eoi' if $kvm;
3573 }
3574
3575 add_hyperv_enlightenments($cpuFlags, $winversion, $machine_type, $kvmver, $conf->{bios}, $gpu_passthrough, $hv_vendor_id) if $kvm;
3576
3577 push @$cpuFlags, 'enforce' if $cpu ne 'host' && $kvm && $arch eq 'x86_64';
3578
3579 push @$cpuFlags, 'kvm=off' if $kvm_off;
3580
3581 if (my $cpu_vendor = $cpu_vendor_list->{$cpu}) {
3582 push @$cpuFlags, "vendor=${cpu_vendor}"
3583 if $cpu_vendor ne 'default';
3584 } elsif ($arch ne 'aarch64') {
3585 die "internal error"; # should not happen
3586 }
3587
3588 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3589
3590 return ('-cpu', $cpu);
3591 }
3592
3593 sub config_to_command {
3594 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
3595
3596 my $cmd = [];
3597 my $globalFlags = [];
3598 my $machineFlags = [];
3599 my $rtcFlags = [];
3600 my $devices = [];
3601 my $pciaddr = '';
3602 my $bridges = {};
3603 my $vernum = 0; # unknown
3604 my $ostype = $conf->{ostype};
3605 my $winversion = windows_version($ostype);
3606 my $kvm = $conf->{kvm};
3607
3608 my ($arch, $machine_type) = get_basic_machine_info($conf, $forcemachine);
3609 my $kvm_binary = get_command_for_arch($arch);
3610 my $kvmver = kvm_user_version($kvm_binary);
3611 $kvm //= 1 if is_native($arch);
3612
3613 if ($kvm) {
3614 die "KVM virtualisation configured, but not available. Either disable in VM configuration or enable in BIOS.\n"
3615 if !defined kvm_version();
3616 }
3617
3618 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
3619 $vernum = $1*1000000+$2*1000;
3620 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
3621 $vernum = $1*1000000+$2*1000+$3;
3622 }
3623
3624 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
3625
3626 my $q35 = machine_type_is_q35($conf);
3627 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3628 my $use_old_bios_files = undef;
3629 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3630
3631 my $cpuunits = defined($conf->{cpuunits}) ?
3632 $conf->{cpuunits} : $defaults->{cpuunits};
3633
3634 push @$cmd, $kvm_binary;
3635
3636 push @$cmd, '-id', $vmid;
3637
3638 my $vmname = $conf->{name} || "vm$vmid";
3639
3640 push @$cmd, '-name', $vmname;
3641
3642 my $use_virtio = 0;
3643
3644 my $qmpsocket = qmp_socket($vmid);
3645 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
3646 push @$cmd, '-mon', "chardev=qmp,mode=control";
3647
3648 if (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 12)) {
3649 push @$cmd, '-chardev', "socket,id=qmp-event,path=/var/run/qmeventd.sock,reconnect=5";
3650 push @$cmd, '-mon', "chardev=qmp-event,mode=control";
3651 }
3652
3653 push @$cmd, '-pidfile' , pidfile_name($vmid);
3654
3655 push @$cmd, '-daemonize';
3656
3657 if ($conf->{smbios1}) {
3658 my $smbios_conf = parse_smbios1($conf->{smbios1});
3659 if ($smbios_conf->{base64}) {
3660 # Do not pass base64 flag to qemu
3661 delete $smbios_conf->{base64};
3662 my $smbios_string = "";
3663 foreach my $key (keys %$smbios_conf) {
3664 my $value;
3665 if ($key eq "uuid") {
3666 $value = $smbios_conf->{uuid}
3667 } else {
3668 $value = decode_base64($smbios_conf->{$key});
3669 }
3670 # qemu accepts any binary data, only commas need escaping by double comma
3671 $value =~ s/,/,,/g;
3672 $smbios_string .= "," . $key . "=" . $value if $value;
3673 }
3674 push @$cmd, '-smbios', "type=1" . $smbios_string;
3675 } else {
3676 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
3677 }
3678 }
3679
3680 if ($conf->{vmgenid}) {
3681 push @$devices, '-device', 'vmgenid,guid='.$conf->{vmgenid};
3682 }
3683
3684 my ($ovmf_code, $ovmf_vars) = get_ovmf_files($arch);
3685 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
3686 die "uefi base image not found\n" if ! -f $ovmf_code;
3687
3688 my $path;
3689 my $format;
3690 if (my $efidisk = $conf->{efidisk0}) {
3691 my $d = PVE::JSONSchema::parse_property_string($efidisk_fmt, $efidisk);
3692 my ($storeid, $volname) = PVE::Storage::parse_volume_id($d->{file}, 1);
3693 $format = $d->{format};
3694 if ($storeid) {
3695 $path = PVE::Storage::path($storecfg, $d->{file});
3696 if (!defined($format)) {
3697 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
3698 $format = qemu_img_format($scfg, $volname);
3699 }
3700 } else {
3701 $path = $d->{file};
3702 die "efidisk format must be specified\n"
3703 if !defined($format);
3704 }
3705 } else {
3706 warn "no efidisk configured! Using temporary efivars disk.\n";
3707 $path = "/tmp/$vmid-ovmf.fd";
3708 PVE::Tools::file_copy($ovmf_vars, $path, -s $ovmf_vars);
3709 $format = 'raw';
3710 }
3711
3712 push @$cmd, '-drive', "if=pflash,unit=0,format=raw,readonly,file=$ovmf_code";
3713 push @$cmd, '-drive', "if=pflash,unit=1,format=$format,id=drive-efidisk0,file=$path";
3714 }
3715
3716 # load q35 config
3717 if ($q35) {
3718 # we use different pcie-port hardware for qemu >= 4.0 for passthrough
3719 if (qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0)) {
3720 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35-4.0.cfg';
3721 } else {
3722 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
3723 }
3724 }
3725
3726 # add usb controllers
3727 my @usbcontrollers = PVE::QemuServer::USB::get_usb_controllers($conf, $bridges, $arch, $machine_type, $usbdesc->{format}, $MAX_USB_DEVICES);
3728 push @$devices, @usbcontrollers if @usbcontrollers;
3729 my $vga = parse_vga($conf->{vga});
3730
3731 my $qxlnum = vga_conf_has_spice($conf->{vga});
3732 $vga->{type} = 'qxl' if $qxlnum;
3733
3734 if (!$vga->{type}) {
3735 if ($arch eq 'aarch64') {
3736 $vga->{type} = 'virtio';
3737 } elsif (qemu_machine_feature_enabled($machine_type, $kvmver, 2, 9)) {
3738 $vga->{type} = (!$winversion || $winversion >= 6) ? 'std' : 'cirrus';
3739 } else {
3740 $vga->{type} = ($winversion >= 6) ? 'std' : 'cirrus';
3741 }
3742 }
3743
3744 # enable absolute mouse coordinates (needed by vnc)
3745 my $tablet;
3746 if (defined($conf->{tablet})) {
3747 $tablet = $conf->{tablet};
3748 } else {
3749 $tablet = $defaults->{tablet};
3750 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
3751 $tablet = 0 if $vga->{type} =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
3752 }
3753
3754 if ($tablet) {
3755 push @$devices, '-device', print_tabletdevice_full($conf, $arch) if $tablet;
3756 my $kbd = print_keyboarddevice_full($conf, $arch);
3757 push @$devices, '-device', $kbd if defined($kbd);
3758 }
3759
3760 my $kvm_off = 0;
3761 my $gpu_passthrough;
3762
3763 # host pci devices
3764 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3765 my $id = "hostpci$i";
3766 my $d = parse_hostpci($conf->{$id});
3767 next if !$d;
3768
3769 if (my $pcie = $d->{pcie}) {
3770 die "q35 machine model is not enabled" if !$q35;
3771 # win7 wants to have the pcie devices directly on the pcie bus
3772 # instead of in the root port
3773 if ($winversion == 7) {
3774 $pciaddr = print_pcie_addr("${id}bus0");
3775 } else {
3776 # add more root ports if needed, 4 are present by default
3777 # by pve-q35 cfgs, rest added here on demand.
3778 if ($i > 3) {
3779 push @$devices, '-device', print_pcie_root_port($i);
3780 }
3781 $pciaddr = print_pcie_addr($id);
3782 }
3783 } else {
3784 $pciaddr = print_pci_addr($id, $bridges, $arch, $machine_type);
3785 }
3786
3787 my $xvga = '';
3788 if ($d->{'x-vga'}) {
3789 $xvga = ',x-vga=on' if !($conf->{bios} && $conf->{bios} eq 'ovmf');
3790 $kvm_off = 1;
3791 $vga->{type} = 'none' if !defined($conf->{vga});
3792 $gpu_passthrough = 1;
3793 }
3794
3795 my $pcidevices = $d->{pciid};
3796 my $multifunction = 1 if @$pcidevices > 1;
3797
3798 my $sysfspath;
3799 if ($d->{mdev} && scalar(@$pcidevices) == 1) {
3800 my $pci_id = $pcidevices->[0]->{id};
3801 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
3802 $sysfspath = "/sys/bus/pci/devices/0000:$pci_id/$uuid";
3803 } elsif ($d->{mdev}) {
3804 warn "ignoring mediated device '$id' with multifunction device\n";
3805 }
3806
3807 my $j=0;
3808 foreach my $pcidevice (@$pcidevices) {
3809 my $devicestr = "vfio-pci";
3810
3811 if ($sysfspath) {
3812 $devicestr .= ",sysfsdev=$sysfspath";
3813 } else {
3814 $devicestr .= ",host=$pcidevice->{id}";
3815 }
3816
3817 my $mf_addr = $multifunction ? ".$j" : '';
3818 $devicestr .= ",id=${id}${mf_addr}${pciaddr}${mf_addr}";
3819
3820 if ($j == 0) {
3821 $devicestr .= ',rombar=0' if defined($d->{rombar}) && !$d->{rombar};
3822 $devicestr .= "$xvga";
3823 $devicestr .= ",multifunction=on" if $multifunction;
3824 $devicestr .= ",romfile=/usr/share/kvm/$d->{romfile}" if $d->{romfile};
3825 }
3826
3827 push @$devices, '-device', $devicestr;
3828 $j++;
3829 }
3830 }
3831
3832 # usb devices
3833 my $usb_dev_features = {};
3834 $usb_dev_features->{spice_usb3} = 1 if qemu_machine_feature_enabled($machine_type, $kvmver, 4, 0);
3835
3836 my @usbdevices = PVE::QemuServer::USB::get_usb_devices($conf, $usbdesc->{format}, $MAX_USB_DEVICES, $usb_dev_features);
3837 push @$devices, @usbdevices if @usbdevices;
3838 # serial devices
3839 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
3840 if (my $path = $conf->{"serial$i"}) {
3841 if ($path eq 'socket') {
3842 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
3843 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
3844 # On aarch64, serial0 is the UART device. Qemu only allows
3845 # connecting UART devices via the '-serial' command line, as
3846 # the device has a fixed slot on the hardware...
3847 if ($arch eq 'aarch64' && $i == 0) {
3848 push @$devices, '-serial', "chardev:serial$i";
3849 } else {
3850 push @$devices, '-device', "isa-serial,chardev=serial$i";
3851 }
3852 } else {
3853 die "no such serial device\n" if ! -c $path;
3854 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
3855 push @$devices, '-device', "isa-serial,chardev=serial$i";
3856 }
3857 }
3858 }
3859
3860 # parallel devices
3861 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
3862 if (my $path = $conf->{"parallel$i"}) {
3863 die "no such parallel device\n" if ! -c $path;
3864 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
3865 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
3866 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
3867 }
3868 }
3869
3870 if (my $audio = conf_has_audio($conf)) {
3871
3872 my $audiopciaddr = print_pci_addr("audio0", $bridges, $arch, $machine_type);
3873
3874 my $id = $audio->{dev_id};
3875 if ($audio->{dev} eq 'AC97') {
3876 push @$devices, '-device', "AC97,id=${id}${audiopciaddr}";
3877 } elsif ($audio->{dev} =~ /intel\-hda$/) {
3878 push @$devices, '-device', "$audio->{dev},id=${id}${audiopciaddr}";
3879 push @$devices, '-device', "hda-micro,id=${id}-codec0,bus=${id}.0,cad=0";
3880 push @$devices, '-device', "hda-duplex,id=${id}-codec1,bus=${id}.0,cad=1";
3881 } else {
3882 die "unkown audio device '$audio->{dev}', implement me!";
3883 }
3884
3885 push @$devices, '-audiodev', "$audio->{backend},id=$audio->{backend_id}";
3886 }
3887
3888 my $sockets = 1;
3889 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3890 $sockets = $conf->{sockets} if $conf->{sockets};
3891
3892 my $cores = $conf->{cores} || 1;
3893
3894 my $maxcpus = $sockets * $cores;
3895
3896 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
3897
3898 my $allowed_vcpus = $cpuinfo->{cpus};
3899
3900 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
3901 if ($allowed_vcpus < $maxcpus);
3902
3903 if($hotplug_features->{cpu} && qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 7)) {
3904
3905 push @$cmd, '-smp', "1,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3906 for (my $i = 2; $i <= $vcpus; $i++) {
3907 my $cpustr = print_cpu_device($conf,$i);
3908 push @$cmd, '-device', $cpustr;
3909 }
3910
3911 } else {
3912
3913 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
3914 }
3915 push @$cmd, '-nodefaults';
3916
3917 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
3918
3919 my $bootindex_hash = {};
3920 my $i = 1;
3921 foreach my $o (split(//, $bootorder)) {
3922 $bootindex_hash->{$o} = $i*100;
3923 $i++;
3924 }
3925
3926 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000,splash=/usr/share/qemu-server/bootsplash.jpg";
3927
3928 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
3929
3930 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
3931
3932 if ($vga->{type} && $vga->{type} !~ m/^serial\d+$/ && $vga->{type} ne 'none'){
3933 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, undef, $qxlnum, $bridges);
3934 my $socket = vnc_socket($vmid);
3935 push @$cmd, '-vnc', "unix:$socket,password";
3936 } else {
3937 push @$cmd, '-vga', 'none' if $vga->{type} eq 'none';
3938 push @$cmd, '-nographic';
3939 }
3940
3941 # time drift fix
3942 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
3943
3944 my $useLocaltime = $conf->{localtime};
3945
3946 if ($winversion >= 5) { # windows
3947 $useLocaltime = 1 if !defined($conf->{localtime});
3948
3949 # use time drift fix when acpi is enabled
3950 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3951 $tdf = 1 if !defined($conf->{tdf});
3952 }
3953 }
3954
3955 if ($winversion >= 6) {
3956 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3957 push @$cmd, '-no-hpet';
3958 }
3959
3960 push @$rtcFlags, 'driftfix=slew' if $tdf;
3961
3962 if (!$kvm) {
3963 push @$machineFlags, 'accel=tcg';
3964 }
3965
3966 if ($machine_type) {
3967 push @$machineFlags, "type=${machine_type}";
3968 }
3969
3970 if (($conf->{startdate}) && ($conf->{startdate} ne 'now')) {
3971 push @$rtcFlags, "base=$conf->{startdate}";
3972 } elsif ($useLocaltime) {
3973 push @$rtcFlags, 'base=localtime';
3974 }
3975
3976 push @$cmd, get_cpu_options($conf, $arch, $kvm, $machine_type, $kvm_off, $kvmver, $winversion, $gpu_passthrough);
3977
3978 PVE::QemuServer::Memory::config($conf, $vmid, $sockets, $cores, $defaults, $hotplug_features, $cmd);
3979
3980 push @$cmd, '-S' if $conf->{freeze};
3981
3982 push @$cmd, '-k', $conf->{keyboard} if defined($conf->{keyboard});
3983
3984 if (parse_guest_agent($conf)->{enabled}) {
3985 my $qgasocket = qmp_socket($vmid, 1);
3986 my $pciaddr = print_pci_addr("qga0", $bridges, $arch, $machine_type);
3987 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3988 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3989 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3990 }
3991
3992 my $spice_port;
3993
3994 if ($qxlnum) {
3995 if ($qxlnum > 1) {
3996 if ($winversion){
3997 for(my $i = 1; $i < $qxlnum; $i++){
3998 push @$devices, '-device', print_vga_device($conf, $vga, $arch, $machine_type, $i, $qxlnum, $bridges);
3999 }
4000 } else {
4001 # assume other OS works like Linux
4002 my ($ram, $vram) = ("134217728", "67108864");
4003 if ($vga->{memory}) {
4004 $ram = PVE::Tools::convert_size($qxlnum*4*$vga->{memory}, 'mb' => 'b');
4005 $vram = PVE::Tools::convert_size($qxlnum*2*$vga->{memory}, 'mb' => 'b');
4006 }
4007 push @$cmd, '-global', "qxl-vga.ram_size=$ram";
4008 push @$cmd, '-global', "qxl-vga.vram_size=$vram";
4009 }
4010 }
4011
4012 my $pciaddr = print_pci_addr("spice", $bridges, $arch, $machine_type);
4013
4014 my $nodename = PVE::INotify::nodename();
4015 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4016 my @nodeaddrs = PVE::Tools::getaddrinfo_all('localhost', family => $pfamily);
4017 die "failed to get an ip address of type $pfamily for 'localhost'\n" if !@nodeaddrs;
4018
4019 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
4020 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
4021 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
4022
4023 my $localhost = PVE::Network::addr_to_ip($nodeaddrs[0]->{addr});
4024 $spice_port = PVE::Tools::next_spice_port($pfamily, $localhost);
4025
4026 my $spice_enhancement = PVE::JSONSchema::parse_property_string($spice_enhancements_fmt, $conf->{spice_enhancements} // '');
4027 if ($spice_enhancement->{foldersharing}) {
4028 push @$devices, '-chardev', "spiceport,id=foldershare,name=org.spice-space.webdav.0";
4029 push @$devices, '-device', "virtserialport,chardev=foldershare,name=org.spice-space.webdav.0";
4030 }
4031
4032 my $spice_opts = "tls-port=${spice_port},addr=$localhost,tls-ciphers=HIGH,seamless-migration=on";
4033 $spice_opts .= ",streaming-video=$spice_enhancement->{videostreaming}" if $spice_enhancement->{videostreaming};
4034 push @$devices, '-spice', "$spice_opts";
4035 }
4036
4037 # enable balloon by default, unless explicitly disabled
4038 if (!defined($conf->{balloon}) || $conf->{balloon}) {
4039 $pciaddr = print_pci_addr("balloon0", $bridges, $arch, $machine_type);
4040 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
4041 }
4042
4043 if ($conf->{watchdog}) {
4044 my $wdopts = parse_watchdog($conf->{watchdog});
4045 $pciaddr = print_pci_addr("watchdog", $bridges, $arch, $machine_type);
4046 my $watchdog = $wdopts->{model} || 'i6300esb';
4047 push @$devices, '-device', "$watchdog$pciaddr";
4048 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
4049 }
4050
4051 my $vollist = [];
4052 my $scsicontroller = {};
4053 my $ahcicontroller = {};
4054 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
4055
4056 # Add iscsi initiator name if available
4057 if (my $initiator = get_initiator_name()) {
4058 push @$devices, '-iscsi', "initiator-name=$initiator";
4059 }
4060
4061 foreach_drive($conf, sub {
4062 my ($ds, $drive) = @_;
4063
4064 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
4065 push @$vollist, $drive->{file};
4066 }
4067
4068 # ignore efidisk here, already added in bios/fw handling code above
4069 return if $drive->{interface} eq 'efidisk';
4070
4071 $use_virtio = 1 if $ds =~ m/^virtio/;
4072
4073 if (drive_is_cdrom ($drive)) {
4074 if ($bootindex_hash->{d}) {
4075 $drive->{bootindex} = $bootindex_hash->{d};
4076 $bootindex_hash->{d} += 1;
4077 }
4078 } else {
4079 if ($bootindex_hash->{c}) {
4080 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
4081 $bootindex_hash->{c} += 1;
4082 }
4083 }
4084
4085 if($drive->{interface} eq 'virtio'){
4086 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
4087 }
4088
4089 if ($drive->{interface} eq 'scsi') {
4090
4091 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
4092
4093 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges, $arch, $machine_type);
4094 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
4095
4096 my $iothread = '';
4097 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
4098 $iothread .= ",iothread=iothread-$controller_prefix$controller";
4099 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
4100 } elsif ($drive->{iothread}) {
4101 warn "iothread is only valid with virtio disk or virtio-scsi-single controller, ignoring\n";
4102 }
4103
4104 my $queues = '';
4105 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
4106 $queues = ",num_queues=$drive->{queues}";
4107 }
4108
4109 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
4110 $scsicontroller->{$controller}=1;
4111 }
4112
4113 if ($drive->{interface} eq 'sata') {
4114 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
4115 $pciaddr = print_pci_addr("ahci$controller", $bridges, $arch, $machine_type);
4116 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
4117 $ahcicontroller->{$controller}=1;
4118 }
4119
4120 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
4121 push @$devices, '-drive',$drive_cmd;
4122 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges, $arch, $machine_type);
4123 });
4124
4125 for (my $i = 0; $i < $MAX_NETS; $i++) {
4126 next if !$conf->{"net$i"};
4127 my $d = parse_net($conf->{"net$i"});
4128 next if !$d;
4129
4130 $use_virtio = 1 if $d->{model} eq 'virtio';
4131
4132 if ($bootindex_hash->{n}) {
4133 $d->{bootindex} = $bootindex_hash->{n};
4134 $bootindex_hash->{n} += 1;
4135 }
4136
4137 my $netdevfull = print_netdev_full($vmid, $conf, $arch, $d, "net$i");
4138 push @$devices, '-netdev', $netdevfull;
4139
4140 my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files, $arch, $machine_type);
4141 push @$devices, '-device', $netdevicefull;
4142 }
4143
4144 if ($conf->{ivshmem}) {
4145 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
4146
4147 my $bus;
4148 if ($q35) {
4149 $bus = print_pcie_addr("ivshmem");
4150 } else {
4151 $bus = print_pci_addr("ivshmem", $bridges, $arch, $machine_type);
4152 }
4153
4154 my $ivshmem_name = $ivshmem->{name} // $vmid;
4155 my $path = '/dev/shm/pve-shm-' . $ivshmem_name;
4156
4157 push @$devices, '-device', "ivshmem-plain,memdev=ivshmem$bus,";
4158 push @$devices, '-object', "memory-backend-file,id=ivshmem,share=on,mem-path=$path,size=$ivshmem->{size}M";
4159 }
4160
4161 if (!$q35) {
4162 # add pci bridges
4163 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
4164 $bridges->{1} = 1;
4165 $bridges->{2} = 1;
4166 }
4167
4168 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
4169
4170 for my $k (sort {$b cmp $a} keys %$bridges) {
4171 $pciaddr = print_pci_addr("pci.$k", undef, $arch, $machine_type);
4172 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
4173 }
4174 }
4175
4176 push @$cmd, @$devices;
4177 push @$cmd, '-rtc', join(',', @$rtcFlags)
4178 if scalar(@$rtcFlags);
4179 push @$cmd, '-machine', join(',', @$machineFlags)
4180 if scalar(@$machineFlags);
4181 push @$cmd, '-global', join(',', @$globalFlags)
4182 if scalar(@$globalFlags);
4183
4184 if (my $vmstate = $conf->{vmstate}) {
4185 my $statepath = PVE::Storage::path($storecfg, $vmstate);
4186 push @$vollist, $statepath;
4187 push @$cmd, '-loadstate', $statepath;
4188 }
4189
4190 # add custom args
4191 if ($conf->{args}) {
4192 my $aa = PVE::Tools::split_args($conf->{args});
4193 push @$cmd, @$aa;
4194 }
4195
4196 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
4197 }
4198
4199 sub vnc_socket {
4200 my ($vmid) = @_;
4201 return "${var_run_tmpdir}/$vmid.vnc";
4202 }
4203
4204 sub spice_port {
4205 my ($vmid) = @_;
4206
4207 my $res = vm_mon_cmd($vmid, 'query-spice');
4208
4209 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
4210 }
4211
4212 sub qmp_socket {
4213 my ($vmid, $qga, $name) = @_;
4214 my $sockettype = $qga ? 'qga' : 'qmp';
4215 my $ext = $name ? '-'.$name : '';
4216 return "${var_run_tmpdir}/$vmid$ext.$sockettype";
4217 }
4218
4219 sub pidfile_name {
4220 my ($vmid) = @_;
4221 return "${var_run_tmpdir}/$vmid.pid";
4222 }
4223
4224 sub vm_devices_list {
4225 my ($vmid) = @_;
4226
4227 my $res = vm_mon_cmd($vmid, 'query-pci');
4228 my $devices_to_check = [];
4229 my $devices = {};
4230 foreach my $pcibus (@$res) {
4231 push @$devices_to_check, @{$pcibus->{devices}},
4232 }
4233
4234 while (@$devices_to_check) {
4235 my $to_check = [];
4236 for my $d (@$devices_to_check) {
4237 $devices->{$d->{'qdev_id'}} = 1 if $d->{'qdev_id'};
4238 next if !$d->{'pci_bridge'};
4239
4240 $devices->{$d->{'qdev_id'}} += scalar(@{$d->{'pci_bridge'}->{devices}});
4241 push @$to_check, @{$d->{'pci_bridge'}->{devices}};
4242 }
4243 $devices_to_check = $to_check;
4244 }
4245
4246 my $resblock = vm_mon_cmd($vmid, 'query-block');
4247 foreach my $block (@$resblock) {
4248 if($block->{device} =~ m/^drive-(\S+)/){
4249 $devices->{$1} = 1;
4250 }
4251 }
4252
4253 my $resmice = vm_mon_cmd($vmid, 'query-mice');
4254 foreach my $mice (@$resmice) {
4255 if ($mice->{name} eq 'QEMU HID Tablet') {
4256 $devices->{tablet} = 1;
4257 last;
4258 }
4259 }
4260
4261 # for usb devices there is no query-usb
4262 # but we can iterate over the entries in
4263 # qom-list path=/machine/peripheral
4264 my $resperipheral = vm_mon_cmd($vmid, 'qom-list', path => '/machine/peripheral');
4265 foreach my $per (@$resperipheral) {
4266 if ($per->{name} =~ m/^usb\d+$/) {
4267 $devices->{$per->{name}} = 1;
4268 }
4269 }
4270
4271 return $devices;
4272 }
4273
4274 sub vm_deviceplug {
4275 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4276
4277 my $q35 = machine_type_is_q35($conf);
4278
4279 my $devices_list = vm_devices_list($vmid);
4280 return 1 if defined($devices_list->{$deviceid});
4281
4282 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid, $arch, $machine_type); # add PCI bridge if we need it for the device
4283
4284 if ($deviceid eq 'tablet') {
4285
4286 qemu_deviceadd($vmid, print_tabletdevice_full($conf, $arch));
4287
4288 } elsif ($deviceid eq 'keyboard') {
4289
4290 qemu_deviceadd($vmid, print_keyboarddevice_full($conf, $arch));
4291
4292 } elsif ($deviceid =~ m/^usb(\d+)$/) {
4293
4294 die "usb hotplug currently not reliable\n";
4295 # since we can't reliably hot unplug all added usb devices
4296 # and usb passthrough disables live migration
4297 # we disable usb hotplugging for now
4298 qemu_deviceadd($vmid, PVE::QemuServer::USB::print_usbdevice_full($conf, $deviceid, $device));
4299
4300 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4301
4302 qemu_iothread_add($vmid, $deviceid, $device);
4303
4304 qemu_driveadd($storecfg, $vmid, $device);
4305 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4306
4307 qemu_deviceadd($vmid, $devicefull);
4308 eval { qemu_deviceaddverify($vmid, $deviceid); };
4309 if (my $err = $@) {
4310 eval { qemu_drivedel($vmid, $deviceid); };
4311 warn $@ if $@;
4312 die $err;
4313 }
4314
4315 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4316
4317
4318 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
4319 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4320 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
4321
4322 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
4323
4324 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
4325 qemu_iothread_add($vmid, $deviceid, $device);
4326 $devicefull .= ",iothread=iothread-$deviceid";
4327 }
4328
4329 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
4330 $devicefull .= ",num_queues=$device->{queues}";
4331 }
4332
4333 qemu_deviceadd($vmid, $devicefull);
4334 qemu_deviceaddverify($vmid, $deviceid);
4335
4336 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4337
4338 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device, $arch, $machine_type);
4339 qemu_driveadd($storecfg, $vmid, $device);
4340
4341 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device, $arch, $machine_type);
4342 eval { qemu_deviceadd($vmid, $devicefull); };
4343 if (my $err = $@) {
4344 eval { qemu_drivedel($vmid, $deviceid); };
4345 warn $@ if $@;
4346 die $err;
4347 }
4348
4349 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4350
4351 return undef if !qemu_netdevadd($vmid, $conf, $arch, $device, $deviceid);
4352
4353 my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
4354 my $use_old_bios_files = undef;
4355 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
4356
4357 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files, $arch, $machine_type);
4358 qemu_deviceadd($vmid, $netdevicefull);
4359 eval {
4360 qemu_deviceaddverify($vmid, $deviceid);
4361 qemu_set_link_status($vmid, $deviceid, !$device->{link_down});
4362 };
4363 if (my $err = $@) {
4364 eval { qemu_netdevdel($vmid, $deviceid); };
4365 warn $@ if $@;
4366 die $err;
4367 }
4368
4369 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
4370
4371 my $bridgeid = $2;
4372 my $pciaddr = print_pci_addr($deviceid, undef, $arch, $machine_type);
4373 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
4374
4375 qemu_deviceadd($vmid, $devicefull);
4376 qemu_deviceaddverify($vmid, $deviceid);
4377
4378 } else {
4379 die "can't hotplug device '$deviceid'\n";
4380 }
4381
4382 return 1;
4383 }
4384
4385 # fixme: this should raise exceptions on error!
4386 sub vm_deviceunplug {
4387 my ($vmid, $conf, $deviceid) = @_;
4388
4389 my $devices_list = vm_devices_list($vmid);
4390 return 1 if !defined($devices_list->{$deviceid});
4391
4392 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
4393
4394 if ($deviceid eq 'tablet' || $deviceid eq 'keyboard') {
4395
4396 qemu_devicedel($vmid, $deviceid);
4397
4398 } elsif ($deviceid =~ m/^usb\d+$/) {
4399
4400 die "usb hotplug currently not reliable\n";
4401 # when unplugging usb devices this way,
4402 # there may be remaining usb controllers/hubs
4403 # so we disable it for now
4404 qemu_devicedel($vmid, $deviceid);
4405 qemu_devicedelverify($vmid, $deviceid);
4406
4407 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
4408
4409 qemu_devicedel($vmid, $deviceid);
4410 qemu_devicedelverify($vmid, $deviceid);
4411 qemu_drivedel($vmid, $deviceid);
4412 qemu_iothread_del($conf, $vmid, $deviceid);
4413
4414 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
4415
4416 qemu_devicedel($vmid, $deviceid);
4417 qemu_devicedelverify($vmid, $deviceid);
4418 qemu_iothread_del($conf, $vmid, $deviceid);
4419
4420 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
4421
4422 qemu_devicedel($vmid, $deviceid);
4423 qemu_drivedel($vmid, $deviceid);
4424 qemu_deletescsihw($conf, $vmid, $deviceid);
4425
4426 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
4427
4428 qemu_devicedel($vmid, $deviceid);
4429 qemu_devicedelverify($vmid, $deviceid);
4430 qemu_netdevdel($vmid, $deviceid);
4431
4432 } else {
4433 die "can't unplug device '$deviceid'\n";
4434 }
4435
4436 return 1;
4437 }
4438
4439 sub qemu_deviceadd {
4440 my ($vmid, $devicefull) = @_;
4441
4442 $devicefull = "driver=".$devicefull;
4443 my %options = split(/[=,]/, $devicefull);
4444
4445 vm_mon_cmd($vmid, "device_add" , %options);
4446 }
4447
4448 sub qemu_devicedel {
4449 my ($vmid, $deviceid) = @_;
4450
4451 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
4452 }
4453
4454 sub qemu_iothread_add {
4455 my($vmid, $deviceid, $device) = @_;
4456
4457 if ($device->{iothread}) {
4458 my $iothreads = vm_iothreads_list($vmid);
4459 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
4460 }
4461 }
4462
4463 sub qemu_iothread_del {
4464 my($conf, $vmid, $deviceid) = @_;
4465
4466 my $confid = $deviceid;
4467 if ($deviceid =~ m/^(?:virtioscsi|scsihw)(\d+)$/) {
4468 $confid = 'scsi' . $1;
4469 }
4470 my $device = parse_drive($confid, $conf->{$confid});
4471 if ($device->{iothread}) {
4472 my $iothreads = vm_iothreads_list($vmid);
4473 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
4474 }
4475 }
4476
4477 sub qemu_objectadd {
4478 my($vmid, $objectid, $qomtype) = @_;
4479
4480 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
4481
4482 return 1;
4483 }
4484
4485 sub qemu_objectdel {
4486 my($vmid, $objectid) = @_;
4487
4488 vm_mon_cmd($vmid, "object-del", id => $objectid);
4489
4490 return 1;
4491 }
4492
4493 sub qemu_driveadd {
4494 my ($storecfg, $vmid, $device) = @_;
4495
4496 my $drive = print_drive_full($storecfg, $vmid, $device);
4497 $drive =~ s/\\/\\\\/g;
4498 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
4499
4500 # If the command succeeds qemu prints: "OK"
4501 return 1 if $ret =~ m/OK/s;
4502
4503 die "adding drive failed: $ret\n";
4504 }
4505
4506 sub qemu_drivedel {
4507 my($vmid, $deviceid) = @_;
4508
4509 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
4510 $ret =~ s/^\s+//;
4511
4512 return 1 if $ret eq "";
4513
4514 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
4515 return 1 if $ret =~ m/Device \'.*?\' not found/s;
4516
4517 die "deleting drive $deviceid failed : $ret\n";
4518 }
4519
4520 sub qemu_deviceaddverify {
4521 my ($vmid, $deviceid) = @_;
4522
4523 for (my $i = 0; $i <= 5; $i++) {
4524 my $devices_list = vm_devices_list($vmid);
4525 return 1 if defined($devices_list->{$deviceid});
4526 sleep 1;
4527 }
4528
4529 die "error on hotplug device '$deviceid'\n";
4530 }
4531
4532
4533 sub qemu_devicedelverify {
4534 my ($vmid, $deviceid) = @_;
4535
4536 # need to verify that the device is correctly removed as device_del
4537 # is async and empty return is not reliable
4538
4539 for (my $i = 0; $i <= 5; $i++) {
4540 my $devices_list = vm_devices_list($vmid);
4541 return 1 if !defined($devices_list->{$deviceid});
4542 sleep 1;
4543 }
4544
4545 die "error on hot-unplugging device '$deviceid'\n";
4546 }
4547
4548 sub qemu_findorcreatescsihw {
4549 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4550
4551 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4552
4553 my $scsihwid="$controller_prefix$controller";
4554 my $devices_list = vm_devices_list($vmid);
4555
4556 if(!defined($devices_list->{$scsihwid})) {
4557 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device, $arch, $machine_type);
4558 }
4559
4560 return 1;
4561 }
4562
4563 sub qemu_deletescsihw {
4564 my ($conf, $vmid, $opt) = @_;
4565
4566 my $device = parse_drive($opt, $conf->{$opt});
4567
4568 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
4569 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
4570 return 1;
4571 }
4572
4573 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
4574
4575 my $devices_list = vm_devices_list($vmid);
4576 foreach my $opt (keys %{$devices_list}) {
4577 if (PVE::QemuServer::is_valid_drivename($opt)) {
4578 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
4579 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
4580 return 1;
4581 }
4582 }
4583 }
4584
4585 my $scsihwid="scsihw$controller";
4586
4587 vm_deviceunplug($vmid, $conf, $scsihwid);
4588
4589 return 1;
4590 }
4591
4592 sub qemu_add_pci_bridge {
4593 my ($storecfg, $conf, $vmid, $device, $arch, $machine_type) = @_;
4594
4595 my $bridges = {};
4596
4597 my $bridgeid;
4598
4599 print_pci_addr($device, $bridges, $arch, $machine_type);
4600
4601 while (my ($k, $v) = each %$bridges) {
4602 $bridgeid = $k;
4603 }
4604 return 1 if !defined($bridgeid) || $bridgeid < 1;
4605
4606 my $bridge = "pci.$bridgeid";
4607 my $devices_list = vm_devices_list($vmid);
4608
4609 if (!defined($devices_list->{$bridge})) {
4610 vm_deviceplug($storecfg, $conf, $vmid, $bridge, $arch, $machine_type);
4611 }
4612
4613 return 1;
4614 }
4615
4616 sub qemu_set_link_status {
4617 my ($vmid, $device, $up) = @_;
4618
4619 vm_mon_cmd($vmid, "set_link", name => $device,
4620 up => $up ? JSON::true : JSON::false);
4621 }
4622
4623 sub qemu_netdevadd {
4624 my ($vmid, $conf, $arch, $device, $deviceid) = @_;
4625
4626 my $netdev = print_netdev_full($vmid, $conf, $arch, $device, $deviceid, 1);
4627 my %options = split(/[=,]/, $netdev);
4628
4629 vm_mon_cmd($vmid, "netdev_add", %options);
4630 return 1;
4631 }
4632
4633 sub qemu_netdevdel {
4634 my ($vmid, $deviceid) = @_;
4635
4636 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
4637 }
4638
4639 sub qemu_usb_hotplug {
4640 my ($storecfg, $conf, $vmid, $deviceid, $device, $arch, $machine_type) = @_;
4641
4642 return if !$device;
4643
4644 # remove the old one first
4645 vm_deviceunplug($vmid, $conf, $deviceid);
4646
4647 # check if xhci controller is necessary and available
4648 if ($device->{usb3}) {
4649
4650 my $devicelist = vm_devices_list($vmid);
4651
4652 if (!$devicelist->{xhci}) {
4653 my $pciaddr = print_pci_addr("xhci", undef, $arch, $machine_type);
4654 qemu_deviceadd($vmid, "nec-usb-xhci,id=xhci$pciaddr");
4655 }
4656 }
4657 my $d = parse_usb_device($device->{host});
4658 $d->{usb3} = $device->{usb3};
4659
4660 # add the new one
4661 vm_deviceplug($storecfg, $conf, $vmid, $deviceid, $d, $arch, $machine_type);
4662 }
4663
4664 sub qemu_cpu_hotplug {
4665 my ($vmid, $conf, $vcpus) = @_;
4666
4667 my $machine_type = PVE::QemuServer::get_current_qemu_machine($vmid);
4668
4669 my $sockets = 1;
4670 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
4671 $sockets = $conf->{sockets} if $conf->{sockets};
4672 my $cores = $conf->{cores} || 1;
4673 my $maxcpus = $sockets * $cores;
4674
4675 $vcpus = $maxcpus if !$vcpus;
4676
4677 die "you can't add more vcpus than maxcpus\n"
4678 if $vcpus > $maxcpus;
4679
4680 my $currentvcpus = $conf->{vcpus} || $maxcpus;
4681
4682 if ($vcpus < $currentvcpus) {
4683
4684 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4685
4686 for (my $i = $currentvcpus; $i > $vcpus; $i--) {
4687 qemu_devicedel($vmid, "cpu$i");
4688 my $retry = 0;
4689 my $currentrunningvcpus = undef;
4690 while (1) {
4691 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4692 last if scalar(@{$currentrunningvcpus}) == $i-1;
4693 raise_param_exc({ vcpus => "error unplugging cpu$i" }) if $retry > 5;
4694 $retry++;
4695 sleep 1;
4696 }
4697 #update conf after each succesfull cpu unplug
4698 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4699 PVE::QemuConfig->write_config($vmid, $conf);
4700 }
4701 } else {
4702 die "cpu hot-unplugging requires qemu version 2.7 or higher\n";
4703 }
4704
4705 return;
4706 }
4707
4708 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4709 die "vcpus in running vm does not match its configuration\n"
4710 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
4711
4712 if (qemu_machine_feature_enabled ($machine_type, undef, 2, 7)) {
4713
4714 for (my $i = $currentvcpus+1; $i <= $vcpus; $i++) {
4715 my $cpustr = print_cpu_device($conf, $i);
4716 qemu_deviceadd($vmid, $cpustr);
4717
4718 my $retry = 0;
4719 my $currentrunningvcpus = undef;
4720 while (1) {
4721 $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
4722 last if scalar(@{$currentrunningvcpus}) == $i;
4723 raise_param_exc({ vcpus => "error hotplugging cpu$i" }) if $retry > 10;
4724 sleep 1;
4725 $retry++;
4726 }
4727 #update conf after each succesfull cpu hotplug
4728 $conf->{vcpus} = scalar(@{$currentrunningvcpus});
4729 PVE::QemuConfig->write_config($vmid, $conf);
4730 }
4731 } else {
4732
4733 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
4734 vm_mon_cmd($vmid, "cpu-add", id => int($i));
4735 }
4736 }
4737 }
4738
4739 sub qemu_block_set_io_throttle {
4740 my ($vmid, $deviceid,
4741 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
4742 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max,
4743 $bps_max_length, $bps_rd_max_length, $bps_wr_max_length,
4744 $iops_max_length, $iops_rd_max_length, $iops_wr_max_length) = @_;
4745
4746 return if !check_running($vmid) ;
4747
4748 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
4749 bps => int($bps),
4750 bps_rd => int($bps_rd),
4751 bps_wr => int($bps_wr),
4752 iops => int($iops),
4753 iops_rd => int($iops_rd),
4754 iops_wr => int($iops_wr),
4755 bps_max => int($bps_max),
4756 bps_rd_max => int($bps_rd_max),
4757 bps_wr_max => int($bps_wr_max),
4758 iops_max => int($iops_max),
4759 iops_rd_max => int($iops_rd_max),
4760 iops_wr_max => int($iops_wr_max),
4761 bps_max_length => int($bps_max_length),
4762 bps_rd_max_length => int($bps_rd_max_length),
4763 bps_wr_max_length => int($bps_wr_max_length),
4764 iops_max_length => int($iops_max_length),
4765 iops_rd_max_length => int($iops_rd_max_length),
4766 iops_wr_max_length => int($iops_wr_max_length),
4767 );
4768
4769 }
4770
4771 # old code, only used to shutdown old VM after update
4772 sub __read_avail {
4773 my ($fh, $timeout) = @_;
4774
4775 my $sel = new IO::Select;
4776 $sel->add($fh);
4777
4778 my $res = '';
4779 my $buf;
4780
4781 my @ready;
4782 while (scalar (@ready = $sel->can_read($timeout))) {
4783 my $count;
4784 if ($count = $fh->sysread($buf, 8192)) {
4785 if ($buf =~ /^(.*)\(qemu\) $/s) {
4786 $res .= $1;
4787 last;
4788 } else {
4789 $res .= $buf;
4790 }
4791 } else {
4792 if (!defined($count)) {
4793 die "$!\n";
4794 }
4795 last;
4796 }
4797 }
4798
4799 die "monitor read timeout\n" if !scalar(@ready);
4800
4801 return $res;
4802 }
4803
4804 sub qemu_block_resize {
4805 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4806
4807 my $running = check_running($vmid);
4808
4809 $size = 0 if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4810
4811 return if !$running;
4812
4813 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4814
4815 }
4816
4817 sub qemu_volume_snapshot {
4818 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4819
4820 my $running = check_running($vmid);
4821
4822 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4823 vm_mon_cmd($vmid, 'blockdev-snapshot-internal-sync', device => $deviceid, name => $snap);
4824 } else {
4825 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4826 }
4827 }
4828
4829 sub qemu_volume_snapshot_delete {
4830 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4831
4832 my $running = check_running($vmid);
4833
4834 if($running) {
4835
4836 $running = undef;
4837 my $conf = PVE::QemuConfig->load_config($vmid);
4838 foreach_drive($conf, sub {
4839 my ($ds, $drive) = @_;
4840 $running = 1 if $drive->{file} eq $volid;
4841 });
4842 }
4843
4844 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4845 vm_mon_cmd($vmid, 'blockdev-snapshot-delete-internal-sync', device => $deviceid, name => $snap);
4846 } else {
4847 PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4848 }
4849 }
4850
4851 sub set_migration_caps {
4852 my ($vmid) = @_;
4853
4854 my $cap_ref = [];
4855
4856 my $enabled_cap = {
4857 "auto-converge" => 1,
4858 "xbzrle" => 1,
4859 "x-rdma-pin-all" => 0,
4860 "zero-blocks" => 0,
4861 "compress" => 0
4862 };
4863
4864 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4865
4866 for my $supported_capability (@$supported_capabilities) {
4867 push @$cap_ref, {
4868 capability => $supported_capability->{capability},
4869 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4870 };
4871 }
4872
4873 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4874 }
4875
4876 my $fast_plug_option = {
4877 'lock' => 1,
4878 'name' => 1,
4879 'onboot' => 1,
4880 'shares' => 1,
4881 'startup' => 1,
4882 'description' => 1,
4883 'protection' => 1,
4884 'vmstatestorage' => 1,
4885 'hookscript' => 1,
4886 };
4887
4888 # hotplug changes in [PENDING]
4889 # $selection hash can be used to only apply specified options, for
4890 # example: { cores => 1 } (only apply changed 'cores')
4891 # $errors ref is used to return error messages
4892 sub vmconfig_hotplug_pending {
4893 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4894
4895 my $defaults = load_defaults();
4896 my ($arch, $machine_type) = get_basic_machine_info($conf, undef);
4897
4898 # commit values which do not have any impact on running VM first
4899 # Note: those option cannot raise errors, we we do not care about
4900 # $selection and always apply them.
4901
4902 my $add_error = sub {
4903 my ($opt, $msg) = @_;
4904 $errors->{$opt} = "hotplug problem - $msg";
4905 };
4906
4907 my $changes = 0;
4908 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4909 if ($fast_plug_option->{$opt}) {
4910 $conf->{$opt} = $conf->{pending}->{$opt};
4911 delete $conf->{pending}->{$opt};
4912 $changes = 1;
4913 }
4914 }
4915
4916 if ($changes) {
4917 PVE::QemuConfig->write_config($vmid, $conf);
4918 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4919 }
4920
4921 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4922
4923 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4924 while (my ($opt, $force) = each %$pending_delete_hash) {
4925 next if $selection && !$selection->{$opt};
4926 eval {
4927 if ($opt eq 'hotplug') {
4928 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4929 } elsif ($opt eq 'tablet') {
4930 die "skip\n" if !$hotplug_features->{usb};
4931 if ($defaults->{tablet}) {
4932 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
4933 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
4934 if $arch eq 'aarch64';
4935 } else {
4936 vm_deviceunplug($vmid, $conf, 'tablet');
4937 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
4938 }
4939 } elsif ($opt =~ m/^usb\d+/) {
4940 die "skip\n";
4941 # since we cannot reliably hot unplug usb devices
4942 # we are disabling it
4943 die "skip\n" if !$hotplug_features->{usb} || $conf->{$opt} =~ m/spice/i;
4944 vm_deviceunplug($vmid, $conf, $opt);
4945 } elsif ($opt eq 'vcpus') {
4946 die "skip\n" if !$hotplug_features->{cpu};
4947 qemu_cpu_hotplug($vmid, $conf, undef);
4948 } elsif ($opt eq 'balloon') {
4949 # enable balloon device is not hotpluggable
4950 die "skip\n" if defined($conf->{balloon}) && $conf->{balloon} == 0;
4951 # here we reset the ballooning value to memory
4952 my $balloon = $conf->{memory} || $defaults->{memory};
4953 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4954 } elsif ($fast_plug_option->{$opt}) {
4955 # do nothing
4956 } elsif ($opt =~ m/^net(\d+)$/) {
4957 die "skip\n" if !$hotplug_features->{network};
4958 vm_deviceunplug($vmid, $conf, $opt);
4959 } elsif (is_valid_drivename($opt)) {
4960 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4961 vm_deviceunplug($vmid, $conf, $opt);
4962 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4963 } elsif ($opt =~ m/^memory$/) {
4964 die "skip\n" if !$hotplug_features->{memory};
4965 PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4966 } elsif ($opt eq 'cpuunits') {
4967 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4968 } elsif ($opt eq 'cpulimit') {
4969 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4970 } else {
4971 die "skip\n";
4972 }
4973 };
4974 if (my $err = $@) {
4975 &$add_error($opt, $err) if $err ne "skip\n";
4976 } else {
4977 # save new config if hotplug was successful
4978 delete $conf->{$opt};
4979 vmconfig_undelete_pending_option($conf, $opt);
4980 PVE::QemuConfig->write_config($vmid, $conf);
4981 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4982 }
4983 }
4984
4985 my $apply_pending_cloudinit;
4986 $apply_pending_cloudinit = sub {
4987 my ($key, $value) = @_;
4988 $apply_pending_cloudinit = sub {}; # once is enough
4989
4990 my @cloudinit_opts = keys %$confdesc_cloudinit;
4991 foreach my $opt (keys %{$conf->{pending}}) {
4992 next if !grep { $_ eq $opt } @cloudinit_opts;
4993 $conf->{$opt} = delete $conf->{pending}->{$opt};
4994 }
4995
4996 my $new_conf = { %$conf };
4997 $new_conf->{$key} = $value;
4998 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($new_conf, $vmid);
4999 };
5000
5001 foreach my $opt (keys %{$conf->{pending}}) {
5002 next if $selection && !$selection->{$opt};
5003 my $value = $conf->{pending}->{$opt};
5004 eval {
5005 if ($opt eq 'hotplug') {
5006 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
5007 } elsif ($opt eq 'tablet') {
5008 die "skip\n" if !$hotplug_features->{usb};
5009 if ($value == 1) {
5010 vm_deviceplug($storecfg, $conf, $vmid, 'tablet', $arch, $machine_type);
5011 vm_deviceplug($storecfg, $conf, $vmid, 'keyboard', $arch, $machine_type)
5012 if $arch eq 'aarch64';
5013 } elsif ($value == 0) {
5014 vm_deviceunplug($vmid, $conf, 'tablet');
5015 vm_deviceunplug($vmid, $conf, 'keyboard') if $arch eq 'aarch64';
5016 }
5017 } elsif ($opt =~ m/^usb\d+$/) {
5018 die "skip\n";
5019 # since we cannot reliably hot unplug usb devices
5020 # we are disabling it
5021 die "skip\n" if !$hotplug_features->{usb} || $value =~ m/spice/i;
5022 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format}, $value) };
5023 die "skip\n" if !$d;
5024 qemu_usb_hotplug($storecfg, $conf, $vmid, $opt, $d, $arch, $machine_type);
5025 } elsif ($opt eq 'vcpus') {
5026 die "skip\n" if !$hotplug_features->{cpu};
5027 qemu_cpu_hotplug($vmid, $conf, $value);
5028 } elsif ($opt eq 'balloon') {
5029 # enable/disable balloning device is not hotpluggable
5030 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
5031 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
5032 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
5033
5034 # allow manual ballooning if shares is set to zero
5035 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
5036 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
5037 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
5038 }
5039 } elsif ($opt =~ m/^net(\d+)$/) {
5040 # some changes can be done without hotplug
5041 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
5042 $vmid, $opt, $value, $arch, $machine_type);
5043 } elsif (is_valid_drivename($opt)) {
5044 # some changes can be done without hotplug
5045 my $drive = parse_drive($opt, $value);
5046 if (drive_is_cloudinit($drive)) {
5047 &$apply_pending_cloudinit($opt, $value);
5048 }
5049 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
5050 $vmid, $opt, $value, 1, $arch, $machine_type);
5051 } elsif ($opt =~ m/^memory$/) { #dimms
5052 die "skip\n" if !$hotplug_features->{memory};
5053 $value = PVE::QemuServer::Memory::qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
5054 } elsif ($opt eq 'cpuunits') {
5055 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
5056 } elsif ($opt eq 'cpulimit') {
5057 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
5058 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
5059 } else {
5060 die "skip\n"; # skip non-hot-pluggable options
5061 }
5062 };
5063 if (my $err = $@) {
5064 &$add_error($opt, $err) if $err ne "skip\n";
5065 } else {
5066 # save new config if hotplug was successful
5067 $conf->{$opt} = $value;
5068 delete $conf->{pending}->{$opt};
5069 PVE::QemuConfig->write_config($vmid, $conf);
5070 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5071 }
5072 }
5073 }
5074
5075 sub try_deallocate_drive {
5076 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
5077
5078 if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
5079 my $volid = $drive->{file};
5080 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
5081 my $sid = PVE::Storage::parse_volume_id($volid);
5082 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
5083
5084 # check if the disk is really unused
5085 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
5086 if is_volume_in_use($storecfg, $conf, $key, $volid);
5087 PVE::Storage::vdisk_free($storecfg, $volid);
5088 return 1;
5089 } else {
5090 # If vm is not owner of this disk remove from config
5091 return 1;
5092 }
5093 }
5094
5095 return undef;
5096 }
5097
5098 sub vmconfig_delete_or_detach_drive {
5099 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
5100
5101 my $drive = parse_drive($opt, $conf->{$opt});
5102
5103 my $rpcenv = PVE::RPCEnvironment::get();
5104 my $authuser = $rpcenv->get_user();
5105
5106 if ($force) {
5107 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
5108 try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
5109 } else {
5110 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
5111 }
5112 }
5113
5114 sub vmconfig_apply_pending {
5115 my ($vmid, $conf, $storecfg) = @_;
5116
5117 # cold plug
5118
5119 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
5120 while (my ($opt, $force) = each %$pending_delete_hash) {
5121 die "internal error" if $opt =~ m/^unused/;
5122 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5123 if (!defined($conf->{$opt})) {
5124 vmconfig_undelete_pending_option($conf, $opt);
5125 PVE::QemuConfig->write_config($vmid, $conf);
5126 } elsif (is_valid_drivename($opt)) {
5127 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
5128 vmconfig_undelete_pending_option($conf, $opt);
5129 delete $conf->{$opt};
5130 PVE::QemuConfig->write_config($vmid, $conf);
5131 } else {
5132 vmconfig_undelete_pending_option($conf, $opt);
5133 delete $conf->{$opt};
5134 PVE::QemuConfig->write_config($vmid, $conf);
5135 }
5136 }
5137
5138 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5139
5140 foreach my $opt (keys %{$conf->{pending}}) { # add/change
5141 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5142
5143 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
5144 # skip if nothing changed
5145 } elsif (is_valid_drivename($opt)) {
5146 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
5147 if defined($conf->{$opt});
5148 $conf->{$opt} = $conf->{pending}->{$opt};
5149 } else {
5150 $conf->{$opt} = $conf->{pending}->{$opt};
5151 }
5152
5153 delete $conf->{pending}->{$opt};
5154 PVE::QemuConfig->write_config($vmid, $conf);
5155 }
5156 }
5157
5158 my $safe_num_ne = sub {
5159 my ($a, $b) = @_;
5160
5161 return 0 if !defined($a) && !defined($b);
5162 return 1 if !defined($a);
5163 return 1 if !defined($b);
5164
5165 return $a != $b;
5166 };
5167
5168 my $safe_string_ne = sub {
5169 my ($a, $b) = @_;
5170
5171 return 0 if !defined($a) && !defined($b);
5172 return 1 if !defined($a);
5173 return 1 if !defined($b);
5174
5175 return $a ne $b;
5176 };
5177
5178 sub vmconfig_update_net {
5179 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $arch, $machine_type) = @_;
5180
5181 my $newnet = parse_net($value);
5182
5183 if ($conf->{$opt}) {
5184 my $oldnet = parse_net($conf->{$opt});
5185
5186 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
5187 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
5188 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
5189 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
5190
5191 # for non online change, we try to hot-unplug
5192 die "skip\n" if !$hotplug;
5193 vm_deviceunplug($vmid, $conf, $opt);
5194 } else {
5195
5196 die "internal error" if $opt !~ m/net(\d+)/;
5197 my $iface = "tap${vmid}i$1";
5198
5199 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
5200 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
5201 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
5202 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
5203 PVE::Network::tap_unplug($iface);
5204 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
5205 } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
5206 # Rate can be applied on its own but any change above needs to
5207 # include the rate in tap_plug since OVS resets everything.
5208 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
5209 }
5210
5211 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
5212 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
5213 }
5214
5215 return 1;
5216 }
5217 }
5218
5219 if ($hotplug) {
5220 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet, $arch, $machine_type);
5221 } else {
5222 die "skip\n";
5223 }
5224 }
5225
5226 sub vmconfig_update_disk {
5227 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force, $arch, $machine_type) = @_;
5228
5229 # fixme: do we need force?
5230
5231 my $drive = parse_drive($opt, $value);
5232
5233 if ($conf->{$opt}) {
5234
5235 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
5236
5237 my $media = $drive->{media} || 'disk';
5238 my $oldmedia = $old_drive->{media} || 'disk';
5239 die "unable to change media type\n" if $media ne $oldmedia;
5240
5241 if (!drive_is_cdrom($old_drive)) {
5242
5243 if ($drive->{file} ne $old_drive->{file}) {
5244
5245 die "skip\n" if !$hotplug;
5246
5247 # unplug and register as unused
5248 vm_deviceunplug($vmid, $conf, $opt);
5249 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
5250
5251 } else {
5252 # update existing disk
5253
5254 # skip non hotpluggable value
5255 if (&$safe_string_ne($drive->{discard}, $old_drive->{discard}) ||
5256 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
5257 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
5258 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
5259 die "skip\n";
5260 }
5261
5262 # apply throttle
5263 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
5264 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
5265 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
5266 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
5267 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
5268 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
5269 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
5270 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
5271 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
5272 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
5273 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
5274 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max}) ||
5275 &$safe_num_ne($drive->{bps_max_length}, $old_drive->{bps_max_length}) ||
5276 &$safe_num_ne($drive->{bps_rd_max_length}, $old_drive->{bps_rd_max_length}) ||
5277 &$safe_num_ne($drive->{bps_wr_max_length}, $old_drive->{bps_wr_max_length}) ||
5278 &$safe_num_ne($drive->{iops_max_length}, $old_drive->{iops_max_length}) ||
5279 &$safe_num_ne($drive->{iops_rd_max_length}, $old_drive->{iops_rd_max_length}) ||
5280 &$safe_num_ne($drive->{iops_wr_max_length}, $old_drive->{iops_wr_max_length})) {
5281
5282 qemu_block_set_io_throttle($vmid,"drive-$opt",
5283 ($drive->{mbps} || 0)*1024*1024,
5284 ($drive->{mbps_rd} || 0)*1024*1024,
5285 ($drive->{mbps_wr} || 0)*1024*1024,
5286 $drive->{iops} || 0,
5287 $drive->{iops_rd} || 0,
5288 $drive->{iops_wr} || 0,
5289 ($drive->{mbps_max} || 0)*1024*1024,
5290 ($drive->{mbps_rd_max} || 0)*1024*1024,
5291 ($drive->{mbps_wr_max} || 0)*1024*1024,
5292 $drive->{iops_max} || 0,
5293 $drive->{iops_rd_max} || 0,
5294 $drive->{iops_wr_max} || 0,
5295 $drive->{bps_max_length} || 1,
5296 $drive->{bps_rd_max_length} || 1,
5297 $drive->{bps_wr_max_length} || 1,
5298 $drive->{iops_max_length} || 1,
5299 $drive->{iops_rd_max_length} || 1,
5300 $drive->{iops_wr_max_length} || 1);
5301
5302 }
5303
5304 return 1;
5305 }
5306
5307 } else { # cdrom
5308
5309 if ($drive->{file} eq 'none') {
5310 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
5311 if (drive_is_cloudinit($old_drive)) {
5312 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive);
5313 }
5314 } else {
5315 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
5316 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
5317 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
5318 }
5319
5320 return 1;
5321 }
5322 }
5323 }
5324
5325 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
5326 # hotplug new disks
5327 PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
5328 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive, $arch, $machine_type);
5329 }
5330
5331 sub vm_start {
5332 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
5333 $forcemachine, $spice_ticket, $migration_network, $migration_type, $targetstorage) = @_;
5334
5335 PVE::QemuConfig->lock_config($vmid, sub {
5336 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5337
5338 die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
5339
5340 my $is_suspended = PVE::QemuConfig->has_lock($conf, 'suspended');
5341
5342 PVE::QemuConfig->check_lock($conf)
5343 if !($skiplock || $is_suspended);
5344
5345 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
5346
5347 # clean up leftover reboot request files
5348 eval { clear_reboot_request($vmid); };
5349 warn $@ if $@;
5350
5351 if (!$statefile && scalar(keys %{$conf->{pending}})) {
5352 vmconfig_apply_pending($vmid, $conf, $storecfg);
5353 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
5354 }
5355
5356 PVE::QemuServer::Cloudinit::generate_cloudinitconfig($conf, $vmid);
5357
5358 my $defaults = load_defaults();
5359
5360 # set environment variable useful inside network script
5361 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
5362
5363 my $local_volumes = {};
5364
5365 if ($targetstorage) {
5366 foreach_drive($conf, sub {
5367 my ($ds, $drive) = @_;
5368
5369 return if drive_is_cdrom($drive);
5370
5371 my $volid = $drive->{file};
5372
5373 return if !$volid;
5374
5375 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid);
5376
5377 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5378 return if $scfg->{shared};
5379 $local_volumes->{$ds} = [$volid, $storeid, $volname];
5380 });
5381
5382 my $format = undef;
5383
5384 foreach my $opt (sort keys %$local_volumes) {
5385
5386 my ($volid, $storeid, $volname) = @{$local_volumes->{$opt}};
5387 my $drive = parse_drive($opt, $conf->{$opt});
5388
5389 #if remote storage is specified, use default format
5390 if ($targetstorage && $targetstorage ne "1") {
5391 $storeid = $targetstorage;
5392 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5393 $format = $defFormat;
5394 } else {
5395 #else we use same format than original
5396 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
5397 $format = qemu_img_format($scfg, $volid);
5398 }
5399
5400 my $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $format, undef, ($drive->{size}/1024));
5401 my $newdrive = $drive;
5402 $newdrive->{format} = $format;
5403 $newdrive->{file} = $newvolid;
5404 my $drivestr = PVE::QemuServer::print_drive($vmid, $newdrive);
5405 $local_volumes->{$opt} = $drivestr;
5406 #pass drive to conf for command line
5407 $conf->{$opt} = $drivestr;
5408 }
5409 }
5410
5411 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
5412
5413 if ($is_suspended) {
5414 # enforce machine type on suspended vm to ensure HW compatibility
5415 $forcemachine = $conf->{runningmachine};
5416 print "Resuming suspended VM\n";
5417 }
5418
5419 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
5420
5421 my $migrate_uri;
5422 if ($statefile) {
5423 if ($statefile eq 'tcp') {
5424 my $localip = "localhost";
5425 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
5426 my $nodename = PVE::INotify::nodename();
5427
5428 if (!defined($migration_type)) {
5429 if (defined($datacenterconf->{migration}->{type})) {
5430 $migration_type = $datacenterconf->{migration}->{type};
5431 } else {
5432 $migration_type = 'secure';
5433 }
5434 }
5435
5436 if ($migration_type eq 'insecure') {
5437 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5438 if ($migrate_network_addr) {
5439 $localip = $migrate_network_addr;
5440 } else {
5441 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
5442 }
5443
5444 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5445 }
5446
5447 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5448 my $migrate_port = PVE::Tools::next_migrate_port($pfamily);
5449 $migrate_uri = "tcp:${localip}:${migrate_port}";
5450 push @$cmd, '-incoming', $migrate_uri;
5451 push @$cmd, '-S';
5452
5453 } elsif ($statefile eq 'unix') {
5454 # should be default for secure migrations as a ssh TCP forward
5455 # tunnel is not deterministic reliable ready and fails regurarly
5456 # to set up in time, so use UNIX socket forwards
5457 my $socket_addr = "/run/qemu-server/$vmid.migrate";
5458 unlink $socket_addr;
5459
5460 $migrate_uri = "unix:$socket_addr";
5461
5462 push @$cmd, '-incoming', $migrate_uri;
5463 push @$cmd, '-S';
5464
5465 } else {
5466 push @$cmd, '-loadstate', $statefile;
5467 }
5468 } elsif ($paused) {
5469 push @$cmd, '-S';
5470 }
5471
5472 # host pci devices
5473 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
5474 my $d = parse_hostpci($conf->{"hostpci$i"});
5475 next if !$d;
5476 my $pcidevices = $d->{pciid};
5477 foreach my $pcidevice (@$pcidevices) {
5478 my $pciid = $pcidevice->{id};
5479
5480 my $info = PVE::SysFSTools::pci_device_info("0000:$pciid");
5481 die "IOMMU not present\n" if !PVE::SysFSTools::check_iommu_support();
5482 die "no pci device info for device '$pciid'\n" if !$info;
5483
5484 if ($d->{mdev}) {
5485 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $i);
5486 PVE::SysFSTools::pci_create_mdev_device($pciid, $uuid, $d->{mdev});
5487 } else {
5488 die "can't unbind/bind pci group to vfio '$pciid'\n"
5489 if !PVE::SysFSTools::pci_dev_group_bind_to_vfio($pciid);
5490 die "can't reset pci device '$pciid'\n"
5491 if $info->{has_fl_reset} and !PVE::SysFSTools::pci_dev_reset($info);
5492 }
5493 }
5494 }
5495
5496 PVE::Storage::activate_volumes($storecfg, $vollist);
5497
5498 eval {
5499 run_command(['/bin/systemctl', 'stop', "$vmid.scope"],
5500 outfunc => sub {}, errfunc => sub {});
5501 };
5502 # Issues with the above 'stop' not being fully completed are extremely rare, a very low
5503 # timeout should be more than enough here...
5504 PVE::Systemd::wait_for_unit_removed("$vmid.scope", 5);
5505
5506 my $cpuunits = defined($conf->{cpuunits}) ? $conf->{cpuunits}
5507 : $defaults->{cpuunits};
5508
5509 my $start_timeout = ($conf->{hugepages} || $is_suspended) ? 300 : 30;
5510 my %run_params = (timeout => $statefile ? undef : $start_timeout, umask => 0077);
5511
5512 my %properties = (
5513 Slice => 'qemu.slice',
5514 KillMode => 'none',
5515 CPUShares => $cpuunits
5516 );
5517
5518 if (my $cpulimit = $conf->{cpulimit}) {
5519 $properties{CPUQuota} = int($cpulimit * 100);
5520 }
5521 $properties{timeout} = 10 if $statefile; # setting up the scope shoul be quick
5522
5523 my $run_qemu = sub {
5524 PVE::Tools::run_fork sub {
5525 PVE::Systemd::enter_systemd_scope($vmid, "Proxmox VE VM $vmid", %properties);
5526 run_command($cmd, %run_params);
5527 };
5528 };
5529
5530 if ($conf->{hugepages}) {
5531
5532 my $code = sub {
5533 my $hugepages_topology = PVE::QemuServer::Memory::hugepages_topology($conf);
5534 my $hugepages_host_topology = PVE::QemuServer::Memory::hugepages_host_topology();
5535
5536 PVE::QemuServer::Memory::hugepages_mount();
5537 PVE::QemuServer::Memory::hugepages_allocate($hugepages_topology, $hugepages_host_topology);
5538
5539 eval { $run_qemu->() };
5540 if (my $err = $@) {
5541 PVE::QemuServer::Memory::hugepages_reset($hugepages_host_topology);
5542 die $err;
5543 }
5544
5545 PVE::QemuServer::Memory::hugepages_pre_deallocate($hugepages_topology);
5546 };
5547 eval { PVE::QemuServer::Memory::hugepages_update_locked($code); };
5548
5549 } else {
5550 eval { $run_qemu->() };
5551 }
5552
5553 if (my $err = $@) {
5554 # deactivate volumes if start fails
5555 eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
5556 die "start failed: $err";
5557 }
5558
5559 print "migration listens on $migrate_uri\n" if $migrate_uri;
5560
5561 if ($statefile && $statefile ne 'tcp' && $statefile ne 'unix') {
5562 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
5563 warn $@ if $@;
5564 }
5565
5566 #start nbd server for storage migration
5567 if ($targetstorage) {
5568 my $nodename = PVE::INotify::nodename();
5569 my $migrate_network_addr = PVE::Cluster::get_local_migration_ip($migration_network);
5570 my $localip = $migrate_network_addr ? $migrate_network_addr : PVE::Cluster::remote_node_ip($nodename, 1);
5571 my $pfamily = PVE::Tools::get_host_address_family($nodename);
5572 my $storage_migrate_port = PVE::Tools::next_migrate_port($pfamily);
5573
5574 vm_mon_cmd_nocheck($vmid, "nbd-server-start", addr => { type => 'inet', data => { host => "${localip}", port => "${storage_migrate_port}" } } );
5575
5576 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
5577
5578 foreach my $opt (sort keys %$local_volumes) {
5579 my $volid = $local_volumes->{$opt};
5580 vm_mon_cmd_nocheck($vmid, "nbd-server-add", device => "drive-$opt", writable => JSON::true );
5581 my $migrate_storage_uri = "nbd:${localip}:${storage_migrate_port}:exportname=drive-$opt";
5582 print "storage migration listens on $migrate_storage_uri volume:$volid\n";
5583 }
5584 }
5585
5586 if ($migratedfrom) {
5587 eval {
5588 set_migration_caps($vmid);
5589 };
5590 warn $@ if $@;
5591
5592 if ($spice_port) {
5593 print "spice listens on port $spice_port\n";
5594 if ($spice_ticket) {
5595 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
5596 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
5597 }
5598 }
5599
5600 } else {
5601 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
5602 if !$statefile && $conf->{balloon};
5603
5604 foreach my $opt (keys %$conf) {
5605 next if $opt !~ m/^net\d+$/;
5606 my $nicconf = parse_net($conf->{$opt});
5607 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
5608 }
5609 }
5610
5611 vm_mon_cmd_nocheck($vmid, 'qom-set',
5612 path => "machine/peripheral/balloon0",
5613 property => "guest-stats-polling-interval",
5614 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
5615
5616 if ($is_suspended && (my $vmstate = $conf->{vmstate})) {
5617 print "Resumed VM, removing state\n";
5618 delete $conf->@{qw(lock vmstate runningmachine)};
5619 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5620 PVE::Storage::vdisk_free($storecfg, $vmstate);
5621 PVE::QemuConfig->write_config($vmid, $conf);
5622 }
5623
5624 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
5625 });
5626 }
5627
5628 sub vm_mon_cmd {
5629 my ($vmid, $execute, %params) = @_;
5630
5631 my $cmd = { execute => $execute, arguments => \%params };
5632 vm_qmp_command($vmid, $cmd);
5633 }
5634
5635 sub vm_mon_cmd_nocheck {
5636 my ($vmid, $execute, %params) = @_;
5637
5638 my $cmd = { execute => $execute, arguments => \%params };
5639 vm_qmp_command($vmid, $cmd, 1);
5640 }
5641
5642 sub vm_qmp_command {
5643 my ($vmid, $cmd, $nocheck) = @_;
5644
5645 my $res;
5646
5647 my $timeout;
5648 if ($cmd->{arguments}) {
5649 $timeout = delete $cmd->{arguments}->{timeout};
5650 }
5651
5652 eval {
5653 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
5654 my $sname = qmp_socket($vmid);
5655 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
5656 my $qmpclient = PVE::QMPClient->new();
5657
5658 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
5659 } else {
5660 die "unable to open monitor socket\n";
5661 }
5662 };
5663 if (my $err = $@) {
5664 syslog("err", "VM $vmid qmp command failed - $err");
5665 die $err;
5666 }
5667
5668 return $res;
5669 }
5670
5671 sub vm_human_monitor_command {
5672 my ($vmid, $cmdline) = @_;
5673
5674 my $cmd = {
5675 execute => 'human-monitor-command',
5676 arguments => { 'command-line' => $cmdline},
5677 };
5678
5679 return vm_qmp_command($vmid, $cmd);
5680 }
5681
5682 sub vm_commandline {
5683 my ($storecfg, $vmid, $snapname) = @_;
5684
5685 my $conf = PVE::QemuConfig->load_config($vmid);
5686
5687 if ($snapname) {
5688 my $snapshot = $conf->{snapshots}->{$snapname};
5689 die "snapshot '$snapname' does not exist\n" if !defined($snapshot);
5690
5691 $snapshot->{digest} = $conf->{digest}; # keep file digest for API
5692
5693 $conf = $snapshot;
5694 }
5695
5696 my $defaults = load_defaults();
5697
5698 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
5699
5700 return PVE::Tools::cmd2string($cmd);
5701 }
5702
5703 sub vm_reset {
5704 my ($vmid, $skiplock) = @_;
5705
5706 PVE::QemuConfig->lock_config($vmid, sub {
5707
5708 my $conf = PVE::QemuConfig->load_config($vmid);
5709
5710 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5711
5712 vm_mon_cmd($vmid, "system_reset");
5713 });
5714 }
5715
5716 sub get_vm_volumes {
5717 my ($conf) = @_;
5718
5719 my $vollist = [];
5720 foreach_volid($conf, sub {
5721 my ($volid, $attr) = @_;
5722
5723 return if $volid =~ m|^/|;
5724
5725 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5726 return if !$sid;
5727
5728 push @$vollist, $volid;
5729 });
5730
5731 return $vollist;
5732 }
5733
5734 sub vm_stop_cleanup {
5735 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
5736
5737 eval {
5738
5739 if (!$keepActive) {
5740 my $vollist = get_vm_volumes($conf);
5741 PVE::Storage::deactivate_volumes($storecfg, $vollist);
5742 }
5743
5744 foreach my $ext (qw(mon qmp pid vnc qga)) {
5745 unlink "/var/run/qemu-server/${vmid}.$ext";
5746 }
5747
5748 if ($conf->{ivshmem}) {
5749 my $ivshmem = PVE::JSONSchema::parse_property_string($ivshmem_fmt, $conf->{ivshmem});
5750 # just delete it for now, VMs which have this already open do not
5751 # are affected, but new VMs will get a separated one. If this
5752 # becomes an issue we either add some sort of ref-counting or just
5753 # add a "don't delete on stop" flag to the ivshmem format.
5754 unlink '/dev/shm/pve-shm-' . ($ivshmem->{name} // $vmid);
5755 }
5756
5757 foreach my $key (keys %$conf) {
5758 next if $key !~ m/^hostpci(\d+)$/;
5759 my $hostpciindex = $1;
5760 my $d = parse_hostpci($conf->{$key});
5761 my $uuid = PVE::SysFSTools::generate_mdev_uuid($vmid, $hostpciindex);
5762
5763 foreach my $pci (@{$d->{pciid}}) {
5764 my $pciid = $pci->{id};
5765 PVE::SysFSTools::pci_cleanup_mdev_device($pciid, $uuid);
5766 }
5767 }
5768
5769 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
5770 };
5771 warn $@ if $@; # avoid errors - just warn
5772 }
5773
5774 # call only in locked context
5775 sub _do_vm_stop {
5776 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive) = @_;
5777
5778 my $pid = check_running($vmid, $nocheck);
5779 return if !$pid;
5780
5781 my $conf;
5782 if (!$nocheck) {
5783 $conf = PVE::QemuConfig->load_config($vmid);
5784 PVE::QemuConfig->check_lock($conf) if !$skiplock;
5785 if (!defined($timeout) && $shutdown && $conf->{startup}) {
5786 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
5787 $timeout = $opts->{down} if $opts->{down};
5788 }
5789 PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
5790 }
5791
5792 eval {
5793 if ($shutdown) {
5794 if (defined($conf) && parse_guest_agent($conf)->{enabled}) {
5795 vm_qmp_command($vmid, {
5796 execute => "guest-shutdown",
5797 arguments => { timeout => $timeout }
5798 }, $nocheck);
5799 } else {
5800 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
5801 }
5802 } else {
5803 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
5804 }
5805 };
5806 my $err = $@;
5807
5808 if (!$err) {
5809 $timeout = 60 if !defined($timeout);
5810
5811 my $count = 0;
5812 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5813 $count++;
5814 sleep 1;
5815 }
5816
5817 if ($count >= $timeout) {
5818 if ($force) {
5819 warn "VM still running - terminating now with SIGTERM\n";
5820 kill 15, $pid;
5821 } else {
5822 die "VM quit/powerdown failed - got timeout\n";
5823 }
5824 } else {
5825 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5826 return;
5827 }
5828 } else {
5829 if ($force) {
5830 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
5831 kill 15, $pid;
5832 } else {
5833 die "VM quit/powerdown failed\n";
5834 }
5835 }
5836
5837 # wait again
5838 $timeout = 10;
5839
5840 my $count = 0;
5841 while (($count < $timeout) && check_running($vmid, $nocheck)) {
5842 $count++;
5843 sleep 1;
5844 }
5845
5846 if ($count >= $timeout) {
5847 warn "VM still running - terminating now with SIGKILL\n";
5848 kill 9, $pid;
5849 sleep 1;
5850 }
5851
5852 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
5853 }
5854
5855 # Note: use $nocheck to skip tests if VM configuration file exists.
5856 # We need that when migration VMs to other nodes (files already moved)
5857 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
5858 sub vm_stop {
5859 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
5860
5861 $force = 1 if !defined($force) && !$shutdown;
5862
5863 if ($migratedfrom){
5864 my $pid = check_running($vmid, $nocheck, $migratedfrom);
5865 kill 15, $pid if $pid;
5866 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
5867 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
5868 return;
5869 }
5870
5871 PVE::QemuConfig->lock_config($vmid, sub {
5872 _do_vm_stop($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive);
5873 });
5874 }
5875
5876 sub vm_reboot {
5877 my ($vmid, $timeout) = @_;
5878
5879 PVE::QemuConfig->lock_config($vmid, sub {
5880
5881 # only reboot if running, as qmeventd starts it again on a stop event
5882 return if !check_running($vmid);
5883
5884 create_reboot_request($vmid);
5885
5886 my $storecfg = PVE::Storage::config();
5887 _do_vm_stop($storecfg, $vmid, undef, undef, $timeout, 1);
5888
5889 });
5890 }
5891
5892 sub vm_suspend {
5893 my ($vmid, $skiplock, $includestate, $statestorage) = @_;
5894
5895 my $conf;
5896 my $path;
5897 my $storecfg;
5898 my $vmstate;
5899
5900 PVE::QemuConfig->lock_config($vmid, sub {
5901
5902 $conf = PVE::QemuConfig->load_config($vmid);
5903
5904 my $is_backing_up = PVE::QemuConfig->has_lock($conf, 'backup');
5905 PVE::QemuConfig->check_lock($conf)
5906 if !($skiplock || $is_backing_up);
5907
5908 die "cannot suspend to disk during backup\n"
5909 if $is_backing_up && $includestate;
5910
5911 if ($includestate) {
5912 $conf->{lock} = 'suspending';
5913 my $date = strftime("%Y-%m-%d", localtime(time()));
5914 $storecfg = PVE::Storage::config();
5915 $vmstate = PVE::QemuConfig->__snapshot_save_vmstate($vmid, $conf, "suspend-$date", $storecfg, $statestorage, 1);
5916 $path = PVE::Storage::path($storecfg, $vmstate);
5917 PVE::QemuConfig->write_config($vmid, $conf);
5918 } else {
5919 vm_mon_cmd($vmid, "stop");
5920 }
5921 });
5922
5923 if ($includestate) {
5924 # save vm state
5925 PVE::Storage::activate_volumes($storecfg, [$vmstate]);
5926
5927 eval {
5928 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5929 for(;;) {
5930 my $state = vm_mon_cmd_nocheck($vmid, "query-savevm");
5931 if (!$state->{status}) {
5932 die "savevm not active\n";
5933 } elsif ($state->{status} eq 'active') {
5934 sleep(1);
5935 next;
5936 } elsif ($state->{status} eq 'completed') {
5937 print "State saved, quitting\n";
5938 last;
5939 } elsif ($state->{status} eq 'failed' && $state->{error}) {
5940 die "query-savevm failed with error '$state->{error}'\n"
5941 } else {
5942 die "query-savevm returned status '$state->{status}'\n";
5943 }
5944 }
5945 };
5946 my $err = $@;
5947
5948 PVE::QemuConfig->lock_config($vmid, sub {
5949 $conf = PVE::QemuConfig->load_config($vmid);
5950 if ($err) {
5951 # cleanup, but leave suspending lock, to indicate something went wrong
5952 eval {
5953 vm_mon_cmd($vmid, "savevm-end");
5954 PVE::Storage::deactivate_volumes($storecfg, [$vmstate]);
5955 PVE::Storage::vdisk_free($storecfg, $vmstate);
5956 delete $conf->@{qw(vmstate runningmachine)};
5957 PVE::QemuConfig->write_config($vmid, $conf);
5958 };
5959 warn $@ if $@;
5960 die $err;
5961 }
5962
5963 die "lock changed unexpectedly\n"
5964 if !PVE::QemuConfig->has_lock($conf, 'suspending');
5965
5966 vm_qmp_command($vmid, { execute => "quit" });
5967 $conf->{lock} = 'suspended';
5968 PVE::QemuConfig->write_config($vmid, $conf);
5969 });
5970 }
5971 }
5972
5973 sub vm_resume {
5974 my ($vmid, $skiplock, $nocheck) = @_;
5975
5976 PVE::QemuConfig->lock_config($vmid, sub {
5977 my $vm_mon_cmd = $nocheck ? \&vm_mon_cmd_nocheck : \&vm_mon_cmd;
5978 my $res = $vm_mon_cmd->($vmid, 'query-status');
5979 my $resume_cmd = 'cont';
5980
5981 if ($res->{status} && $res->{status} eq 'suspended') {
5982 $resume_cmd = 'system_wakeup';
5983 }
5984
5985 if (!$nocheck) {
5986
5987 my $conf = PVE::QemuConfig->load_config($vmid);
5988
5989 PVE::QemuConfig->check_lock($conf)
5990 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
5991 }
5992
5993 $vm_mon_cmd->($vmid, $resume_cmd);
5994 });
5995 }
5996
5997 sub vm_sendkey {
5998 my ($vmid, $skiplock, $key) = @_;
5999
6000 PVE::QemuConfig->lock_config($vmid, sub {
6001
6002 my $conf = PVE::QemuConfig->load_config($vmid);
6003
6004 # there is no qmp command, so we use the human monitor command
6005 my $res = vm_human_monitor_command($vmid, "sendkey $key");
6006 die $res if $res ne '';
6007 });
6008 }
6009
6010 sub vm_destroy {
6011 my ($storecfg, $vmid, $skiplock) = @_;
6012
6013 PVE::QemuConfig->lock_config($vmid, sub {
6014
6015 my $conf = PVE::QemuConfig->load_config($vmid);
6016
6017 if (!check_running($vmid)) {
6018 destroy_vm($storecfg, $vmid, undef, $skiplock);
6019 } else {
6020 die "VM $vmid is running - destroy failed\n";
6021 }
6022 });
6023 }
6024
6025 # vzdump restore implementaion
6026
6027 sub tar_archive_read_firstfile {
6028 my $archive = shift;
6029
6030 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
6031
6032 # try to detect archive type first
6033 my $pid = open (my $fh, '-|', 'tar', 'tf', $archive) ||
6034 die "unable to open file '$archive'\n";
6035 my $firstfile = <$fh>;
6036 kill 15, $pid;
6037 close $fh;
6038
6039 die "ERROR: archive contaions no data\n" if !$firstfile;
6040 chomp $firstfile;
6041
6042 return $firstfile;
6043 }
6044
6045 sub tar_restore_cleanup {
6046 my ($storecfg, $statfile) = @_;
6047
6048 print STDERR "starting cleanup\n";
6049
6050 if (my $fd = IO::File->new($statfile, "r")) {
6051 while (defined(my $line = <$fd>)) {
6052 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6053 my $volid = $2;
6054 eval {
6055 if ($volid =~ m|^/|) {
6056 unlink $volid || die 'unlink failed\n';
6057 } else {
6058 PVE::Storage::vdisk_free($storecfg, $volid);
6059 }
6060 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6061 };
6062 print STDERR "unable to cleanup '$volid' - $@" if $@;
6063 } else {
6064 print STDERR "unable to parse line in statfile - $line";
6065 }
6066 }
6067 $fd->close();
6068 }
6069 }
6070
6071 sub restore_archive {
6072 my ($archive, $vmid, $user, $opts) = @_;
6073
6074 my $format = $opts->{format};
6075 my $comp;
6076
6077 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
6078 $format = 'tar' if !$format;
6079 $comp = 'gzip';
6080 } elsif ($archive =~ m/\.tar$/) {
6081 $format = 'tar' if !$format;
6082 } elsif ($archive =~ m/.tar.lzo$/) {
6083 $format = 'tar' if !$format;
6084 $comp = 'lzop';
6085 } elsif ($archive =~ m/\.vma$/) {
6086 $format = 'vma' if !$format;
6087 } elsif ($archive =~ m/\.vma\.gz$/) {
6088 $format = 'vma' if !$format;
6089 $comp = 'gzip';
6090 } elsif ($archive =~ m/\.vma\.lzo$/) {
6091 $format = 'vma' if !$format;
6092 $comp = 'lzop';
6093 } else {
6094 $format = 'vma' if !$format; # default
6095 }
6096
6097 # try to detect archive format
6098 if ($format eq 'tar') {
6099 return restore_tar_archive($archive, $vmid, $user, $opts);
6100 } else {
6101 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
6102 }
6103 }
6104
6105 sub restore_update_config_line {
6106 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
6107
6108 return if $line =~ m/^\#qmdump\#/;
6109 return if $line =~ m/^\#vzdump\#/;
6110 return if $line =~ m/^lock:/;
6111 return if $line =~ m/^unused\d+:/;
6112 return if $line =~ m/^parent:/;
6113
6114 my $dc = PVE::Cluster::cfs_read_file('datacenter.cfg');
6115 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
6116 # try to convert old 1.X settings
6117 my ($id, $ind, $ethcfg) = ($1, $2, $3);
6118 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
6119 my ($model, $macaddr) = split(/\=/, $devconfig);
6120 $macaddr = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if !$macaddr || $unique;
6121 my $net = {
6122 model => $model,
6123 bridge => "vmbr$ind",
6124 macaddr => $macaddr,
6125 };
6126 my $netstr = print_net($net);
6127
6128 print $outfd "net$cookie->{netcount}: $netstr\n";
6129 $cookie->{netcount}++;
6130 }
6131 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
6132 my ($id, $netstr) = ($1, $2);
6133 my $net = parse_net($netstr);
6134 $net->{macaddr} = PVE::Tools::random_ether_addr($dc->{mac_prefix}) if $net->{macaddr};
6135 $netstr = print_net($net);
6136 print $outfd "$id: $netstr\n";
6137 } elsif ($line =~ m/^((ide|scsi|virtio|sata|efidisk)\d+):\s*(\S+)\s*$/) {
6138 my $virtdev = $1;
6139 my $value = $3;
6140 my $di = parse_drive($virtdev, $value);
6141 if (defined($di->{backup}) && !$di->{backup}) {
6142 print $outfd "#$line";
6143 } elsif ($map->{$virtdev}) {
6144 delete $di->{format}; # format can change on restore
6145 $di->{file} = $map->{$virtdev};
6146 $value = print_drive($vmid, $di);
6147 print $outfd "$virtdev: $value\n";
6148 } else {
6149 print $outfd $line;
6150 }
6151 } elsif (($line =~ m/^vmgenid: (.*)/)) {
6152 my $vmgenid = $1;
6153 if ($vmgenid ne '0') {
6154 # always generate a new vmgenid if there was a valid one setup
6155 $vmgenid = generate_uuid();
6156 }
6157 print $outfd "vmgenid: $vmgenid\n";
6158 } elsif (($line =~ m/^(smbios1: )(.*)/) && $unique) {
6159 my ($uuid, $uuid_str);
6160 UUID::generate($uuid);
6161 UUID::unparse($uuid, $uuid_str);
6162 my $smbios1 = parse_smbios1($2);
6163 $smbios1->{uuid} = $uuid_str;
6164 print $outfd $1.print_smbios1($smbios1)."\n";
6165 } else {
6166 print $outfd $line;
6167 }
6168 }
6169
6170 sub scan_volids {
6171 my ($cfg, $vmid) = @_;
6172
6173 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
6174
6175 my $volid_hash = {};
6176 foreach my $storeid (keys %$info) {
6177 foreach my $item (@{$info->{$storeid}}) {
6178 next if !($item->{volid} && $item->{size});
6179 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
6180 $volid_hash->{$item->{volid}} = $item;
6181 }
6182 }
6183
6184 return $volid_hash;
6185 }
6186
6187 sub is_volume_in_use {
6188 my ($storecfg, $conf, $skip_drive, $volid) = @_;
6189
6190 my $path = PVE::Storage::path($storecfg, $volid);
6191
6192 my $scan_config = sub {
6193 my ($cref, $snapname) = @_;
6194
6195 foreach my $key (keys %$cref) {
6196 my $value = $cref->{$key};
6197 if (is_valid_drivename($key)) {
6198 next if $skip_drive && $key eq $skip_drive;
6199 my $drive = parse_drive($key, $value);
6200 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
6201 return 1 if $volid eq $drive->{file};
6202 if ($drive->{file} =~ m!^/!) {
6203 return 1 if $drive->{file} eq $path;
6204 } else {
6205 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
6206 next if !$storeid;
6207 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
6208 next if !$scfg;
6209 return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
6210 }
6211 }
6212 }
6213
6214 return 0;
6215 };
6216
6217 return 1 if &$scan_config($conf);
6218
6219 undef $skip_drive;
6220
6221 foreach my $snapname (keys %{$conf->{snapshots}}) {
6222 return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
6223 }
6224
6225 return 0;
6226 }
6227
6228 sub update_disksize {
6229 my ($vmid, $conf, $volid_hash) = @_;
6230
6231 my $changes;
6232 my $prefix = "VM $vmid:";
6233
6234 # used and unused disks
6235 my $referenced = {};
6236
6237 # Note: it is allowed to define multiple storages with same path (alias), so
6238 # we need to check both 'volid' and real 'path' (two different volid can point
6239 # to the same path).
6240
6241 my $referencedpath = {};
6242
6243 # update size info
6244 foreach my $opt (keys %$conf) {
6245 if (is_valid_drivename($opt)) {
6246 my $drive = parse_drive($opt, $conf->{$opt});
6247 my $volid = $drive->{file};
6248 next if !$volid;
6249
6250 $referenced->{$volid} = 1;
6251 if ($volid_hash->{$volid} &&
6252 (my $path = $volid_hash->{$volid}->{path})) {
6253 $referencedpath->{$path} = 1;
6254 }
6255
6256 next if drive_is_cdrom($drive);
6257 next if !$volid_hash->{$volid};
6258
6259 $drive->{size} = $volid_hash->{$volid}->{size};
6260 my $new = print_drive($vmid, $drive);
6261 if ($new ne $conf->{$opt}) {
6262 $changes = 1;
6263 $conf->{$opt} = $new;
6264 print "$prefix update disk '$opt' information.\n";
6265 }
6266 }
6267 }
6268
6269 # remove 'unusedX' entry if volume is used
6270 foreach my $opt (keys %$conf) {
6271 next if $opt !~ m/^unused\d+$/;
6272 my $volid = $conf->{$opt};
6273 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
6274 if ($referenced->{$volid} || ($path && $referencedpath->{$path})) {
6275 print "$prefix remove entry '$opt', its volume '$volid' is in use.\n";
6276 $changes = 1;
6277 delete $conf->{$opt};
6278 }
6279
6280 $referenced->{$volid} = 1;
6281 $referencedpath->{$path} = 1 if $path;
6282 }
6283
6284 foreach my $volid (sort keys %$volid_hash) {
6285 next if $volid =~ m/vm-$vmid-state-/;
6286 next if $referenced->{$volid};
6287 my $path = $volid_hash->{$volid}->{path};
6288 next if !$path; # just to be sure
6289 next if $referencedpath->{$path};
6290 $changes = 1;
6291 my $key = PVE::QemuConfig->add_unused_volume($conf, $volid);
6292 print "$prefix add unreferenced volume '$volid' as '$key' to config.\n";
6293 $referencedpath->{$path} = 1; # avoid to add more than once (aliases)
6294 }
6295
6296 return $changes;
6297 }
6298
6299 sub rescan {
6300 my ($vmid, $nolock, $dryrun) = @_;
6301
6302 my $cfg = PVE::Storage::config();
6303
6304 # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
6305 # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
6306 foreach my $stor (keys %{$cfg->{ids}}) {
6307 delete($cfg->{ids}->{$stor}) if ! $cfg->{ids}->{$stor}->{content}->{images};
6308 }
6309
6310 print "rescan volumes...\n";
6311 my $volid_hash = scan_volids($cfg, $vmid);
6312
6313 my $updatefn = sub {
6314 my ($vmid) = @_;
6315
6316 my $conf = PVE::QemuConfig->load_config($vmid);
6317
6318 PVE::QemuConfig->check_lock($conf);
6319
6320 my $vm_volids = {};
6321 foreach my $volid (keys %$volid_hash) {
6322 my $info = $volid_hash->{$volid};
6323 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
6324 }
6325
6326 my $changes = update_disksize($vmid, $conf, $vm_volids);
6327
6328 PVE::QemuConfig->write_config($vmid, $conf) if $changes && !$dryrun;
6329 };
6330
6331 if (defined($vmid)) {
6332 if ($nolock) {
6333 &$updatefn($vmid);
6334 } else {
6335 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6336 }
6337 } else {
6338 my $vmlist = config_list();
6339 foreach my $vmid (keys %$vmlist) {
6340 if ($nolock) {
6341 &$updatefn($vmid);
6342 } else {
6343 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
6344 }
6345 }
6346 }
6347 }
6348
6349 sub restore_vma_archive {
6350 my ($archive, $vmid, $user, $opts, $comp) = @_;
6351
6352 my $readfrom = $archive;
6353
6354 my $cfg = PVE::Storage::config();
6355 my $commands = [];
6356 my $bwlimit = $opts->{bwlimit};
6357
6358 my $dbg_cmdstring = '';
6359 my $add_pipe = sub {
6360 my ($cmd) = @_;
6361 push @$commands, $cmd;
6362 $dbg_cmdstring .= ' | ' if length($dbg_cmdstring);
6363 $dbg_cmdstring .= PVE::Tools::cmd2string($cmd);
6364 $readfrom = '-';
6365 };
6366
6367 my $input = undef;
6368 if ($archive eq '-') {
6369 $input = '<&STDIN';
6370 } else {
6371 # If we use a backup from a PVE defined storage we also consider that
6372 # storage's rate limit:
6373 my (undef, $volid) = PVE::Storage::path_to_volume_id($cfg, $archive);
6374 if (defined($volid)) {
6375 my ($sid, undef) = PVE::Storage::parse_volume_id($volid);
6376 my $readlimit = PVE::Storage::get_bandwidth_limit('restore', [$sid], $bwlimit);
6377 if ($readlimit) {
6378 print STDERR "applying read rate limit: $readlimit\n";
6379 my $cstream = ['cstream', '-t', $readlimit*1024, '--', $readfrom];
6380 $add_pipe->($cstream);
6381 }
6382 }
6383 }
6384
6385 if ($comp) {
6386 my $cmd;
6387 if ($comp eq 'gzip') {
6388 $cmd = ['zcat', $readfrom];
6389 } elsif ($comp eq 'lzop') {
6390 $cmd = ['lzop', '-d', '-c', $readfrom];
6391 } else {
6392 die "unknown compression method '$comp'\n";
6393 }
6394 $add_pipe->($cmd);
6395 }
6396
6397 my $tmpdir = "/var/tmp/vzdumptmp$$";
6398 rmtree $tmpdir;
6399
6400 # disable interrupts (always do cleanups)
6401 local $SIG{INT} =
6402 local $SIG{TERM} =
6403 local $SIG{QUIT} =
6404 local $SIG{HUP} = sub { warn "got interrupt - ignored\n"; };
6405
6406 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
6407 POSIX::mkfifo($mapfifo, 0600);
6408 my $fifofh;
6409
6410 my $openfifo = sub {
6411 open($fifofh, '>', $mapfifo) || die $!;
6412 };
6413
6414 $add_pipe->(['vma', 'extract', '-v', '-r', $mapfifo, $readfrom, $tmpdir]);
6415
6416 my $oldtimeout;
6417 my $timeout = 5;
6418
6419 my $devinfo = {};
6420
6421 my $rpcenv = PVE::RPCEnvironment::get();
6422
6423 my $conffile = PVE::QemuConfig->config_file($vmid);
6424 my $tmpfn = "$conffile.$$.tmp";
6425
6426 # Note: $oldconf is undef if VM does not exists
6427 my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
6428 my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
6429
6430 my %storage_limits;
6431
6432 my $print_devmap = sub {
6433 my $virtdev_hash = {};
6434
6435 my $cfgfn = "$tmpdir/qemu-server.conf";
6436
6437 # we can read the config - that is already extracted
6438 my $fh = IO::File->new($cfgfn, "r") ||
6439 "unable to read qemu-server.conf - $!\n";
6440
6441 my $fwcfgfn = "$tmpdir/qemu-server.fw";
6442 if (-f $fwcfgfn) {
6443 my $pve_firewall_dir = '/etc/pve/firewall';
6444 mkdir $pve_firewall_dir; # make sure the dir exists
6445 PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
6446 }
6447
6448 while (defined(my $line = <$fh>)) {
6449 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
6450 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
6451 die "archive does not contain data for drive '$virtdev'\n"
6452 if !$devinfo->{$devname};
6453 if (defined($opts->{storage})) {
6454 $storeid = $opts->{storage} || 'local';
6455 } elsif (!$storeid) {
6456 $storeid = 'local';
6457 }
6458 $format = 'raw' if !$format;
6459 $devinfo->{$devname}->{devname} = $devname;
6460 $devinfo->{$devname}->{virtdev} = $virtdev;
6461 $devinfo->{$devname}->{format} = $format;
6462 $devinfo->{$devname}->{storeid} = $storeid;
6463
6464 # check permission on storage
6465 my $pool = $opts->{pool}; # todo: do we need that?
6466 if ($user ne 'root@pam') {
6467 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
6468 }
6469
6470 $storage_limits{$storeid} = $bwlimit;
6471
6472 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
6473 } elsif ($line =~ m/^((?:ide|sata|scsi)\d+):\s*(.*)\s*$/) {
6474 my $virtdev = $1;
6475 my $drive = parse_drive($virtdev, $2);
6476 if (drive_is_cloudinit($drive)) {
6477 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6478 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6479 my $format = qemu_img_format($scfg, $volname); # has 'raw' fallback
6480
6481 my $d = {
6482 format => $format,
6483 storeid => $opts->{storage} // $storeid,
6484 size => PVE::QemuServer::Cloudinit::CLOUDINIT_DISK_SIZE,
6485 file => $drive->{file}, # to make drive_is_cloudinit check possible
6486 name => "vm-$vmid-cloudinit",
6487 is_cloudinit => 1,
6488 };
6489 $virtdev_hash->{$virtdev} = $d;
6490 }
6491 }
6492 }
6493
6494 foreach my $key (keys %storage_limits) {
6495 my $limit = PVE::Storage::get_bandwidth_limit('restore', [$key], $bwlimit);
6496 next if !$limit;
6497 print STDERR "rate limit for storage $key: $limit KiB/s\n";
6498 $storage_limits{$key} = $limit * 1024;
6499 }
6500
6501 foreach my $devname (keys %$devinfo) {
6502 die "found no device mapping information for device '$devname'\n"
6503 if !$devinfo->{$devname}->{virtdev};
6504 }
6505
6506 # create empty/temp config
6507 if ($oldconf) {
6508 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
6509 foreach_drive($oldconf, sub {
6510 my ($ds, $drive) = @_;
6511
6512 return if drive_is_cdrom($drive, 1);
6513
6514 my $volid = $drive->{file};
6515 return if !$volid || $volid =~ m|^/|;
6516
6517 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
6518 return if !$path || !$owner || ($owner != $vmid);
6519
6520 # Note: only delete disk we want to restore
6521 # other volumes will become unused
6522 if ($virtdev_hash->{$ds}) {
6523 eval { PVE::Storage::vdisk_free($cfg, $volid); };
6524 if (my $err = $@) {
6525 warn $err;
6526 }
6527 }
6528 });
6529
6530 # delete vmstate files, after the restore we have no snapshots anymore
6531 foreach my $snapname (keys %{$oldconf->{snapshots}}) {
6532 my $snap = $oldconf->{snapshots}->{$snapname};
6533 if ($snap->{vmstate}) {
6534 eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
6535 if (my $err = $@) {
6536 warn $err;
6537 }
6538 }
6539 }
6540 }
6541
6542 my $map = {};
6543 foreach my $virtdev (sort keys %$virtdev_hash) {
6544 my $d = $virtdev_hash->{$virtdev};
6545 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
6546 my $storeid = $d->{storeid};
6547 my $scfg = PVE::Storage::storage_config($cfg, $storeid);
6548
6549 my $map_opts = '';
6550 if (my $limit = $storage_limits{$storeid}) {
6551 $map_opts .= "throttling.bps=$limit:throttling.group=$storeid:";
6552 }
6553
6554 # test if requested format is supported
6555 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $storeid);
6556 my $supported = grep { $_ eq $d->{format} } @$validFormats;
6557 $d->{format} = $defFormat if !$supported;
6558
6559 my $name;
6560 if ($d->{is_cloudinit}) {
6561 $name = $d->{name};
6562 $name .= ".$d->{format}" if $d->{format} ne 'raw';
6563 }
6564
6565 my $volid = PVE::Storage::vdisk_alloc($cfg, $storeid, $vmid, $d->{format}, $name, $alloc_size);
6566 print STDERR "new volume ID is '$volid'\n";
6567 $d->{volid} = $volid;
6568
6569 PVE::Storage::activate_volumes($cfg, [$volid]);
6570
6571 my $write_zeros = 1;
6572 if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
6573 $write_zeros = 0;
6574 }
6575
6576 if (!$d->{is_cloudinit}) {
6577 my $path = PVE::Storage::path($cfg, $volid);
6578
6579 print $fifofh "${map_opts}format=$d->{format}:${write_zeros}:$d->{devname}=$path\n";
6580
6581 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
6582 }
6583 $map->{$virtdev} = $volid;
6584 }
6585
6586 $fh->seek(0, 0) || die "seek failed - $!\n";
6587
6588 my $outfd = new IO::File ($tmpfn, "w") ||
6589 die "unable to write config for VM $vmid\n";
6590
6591 my $cookie = { netcount => 0 };
6592 while (defined(my $line = <$fh>)) {
6593 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6594 }
6595
6596 $fh->close();
6597 $outfd->close();
6598 };
6599
6600 eval {
6601 # enable interrupts
6602 local $SIG{INT} =
6603 local $SIG{TERM} =
6604 local $SIG{QUIT} =
6605 local $SIG{HUP} =
6606 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6607 local $SIG{ALRM} = sub { die "got timeout\n"; };
6608
6609 $oldtimeout = alarm($timeout);
6610
6611 my $parser = sub {
6612 my $line = shift;
6613
6614 print "$line\n";
6615
6616 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
6617 my ($dev_id, $size, $devname) = ($1, $2, $3);
6618 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
6619 } elsif ($line =~ m/^CTIME: /) {
6620 # we correctly received the vma config, so we can disable
6621 # the timeout now for disk allocation (set to 10 minutes, so
6622 # that we always timeout if something goes wrong)
6623 alarm(600);
6624 &$print_devmap();
6625 print $fifofh "done\n";
6626 my $tmp = $oldtimeout || 0;
6627 $oldtimeout = undef;
6628 alarm($tmp);
6629 close($fifofh);
6630 }
6631 };
6632
6633 print "restore vma archive: $dbg_cmdstring\n";
6634 run_command($commands, input => $input, outfunc => $parser, afterfork => $openfifo);
6635 };
6636 my $err = $@;
6637
6638 alarm($oldtimeout) if $oldtimeout;
6639
6640 my $vollist = [];
6641 foreach my $devname (keys %$devinfo) {
6642 my $volid = $devinfo->{$devname}->{volid};
6643 push @$vollist, $volid if $volid;
6644 }
6645
6646 PVE::Storage::deactivate_volumes($cfg, $vollist);
6647
6648 unlink $mapfifo;
6649
6650 if ($err) {
6651 rmtree $tmpdir;
6652 unlink $tmpfn;
6653
6654 foreach my $devname (keys %$devinfo) {
6655 my $volid = $devinfo->{$devname}->{volid};
6656 next if !$volid;
6657 eval {
6658 if ($volid =~ m|^/|) {
6659 unlink $volid || die 'unlink failed\n';
6660 } else {
6661 PVE::Storage::vdisk_free($cfg, $volid);
6662 }
6663 print STDERR "temporary volume '$volid' sucessfuly removed\n";
6664 };
6665 print STDERR "unable to cleanup '$volid' - $@" if $@;
6666 }
6667 die $err;
6668 }
6669
6670 rmtree $tmpdir;
6671
6672 rename($tmpfn, $conffile) ||
6673 die "unable to commit configuration file '$conffile'\n";
6674
6675 PVE::Cluster::cfs_update(); # make sure we read new file
6676
6677 eval { rescan($vmid, 1); };
6678 warn $@ if $@;
6679 }
6680
6681 sub restore_tar_archive {
6682 my ($archive, $vmid, $user, $opts) = @_;
6683
6684 if ($archive ne '-') {
6685 my $firstfile = tar_archive_read_firstfile($archive);
6686 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
6687 if $firstfile ne 'qemu-server.conf';
6688 }
6689
6690 my $storecfg = PVE::Storage::config();
6691
6692 # destroy existing data - keep empty config
6693 my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
6694 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
6695
6696 my $tocmd = "/usr/lib/qemu-server/qmextract";
6697
6698 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
6699 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
6700 $tocmd .= ' --prealloc' if $opts->{prealloc};
6701 $tocmd .= ' --info' if $opts->{info};
6702
6703 # tar option "xf" does not autodetect compression when read from STDIN,
6704 # so we pipe to zcat
6705 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
6706 PVE::Tools::shellquote("--to-command=$tocmd");
6707
6708 my $tmpdir = "/var/tmp/vzdumptmp$$";
6709 mkpath $tmpdir;
6710
6711 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
6712 local $ENV{VZDUMP_VMID} = $vmid;
6713 local $ENV{VZDUMP_USER} = $user;
6714
6715 my $conffile = PVE::QemuConfig->config_file($vmid);
6716 my $tmpfn = "$conffile.$$.tmp";
6717
6718 # disable interrupts (always do cleanups)
6719 local $SIG{INT} =
6720 local $SIG{TERM} =
6721 local $SIG{QUIT} =
6722 local $SIG{HUP} = sub { print STDERR "got interrupt - ignored\n"; };
6723
6724 eval {
6725 # enable interrupts
6726 local $SIG{INT} =
6727 local $SIG{TERM} =
6728 local $SIG{QUIT} =
6729 local $SIG{HUP} =
6730 local $SIG{PIPE} = sub { die "interrupted by signal\n"; };
6731
6732 if ($archive eq '-') {
6733 print "extracting archive from STDIN\n";
6734 run_command($cmd, input => "<&STDIN");
6735 } else {
6736 print "extracting archive '$archive'\n";
6737 run_command($cmd);
6738 }
6739
6740 return if $opts->{info};
6741
6742 # read new mapping
6743 my $map = {};
6744 my $statfile = "$tmpdir/qmrestore.stat";
6745 if (my $fd = IO::File->new($statfile, "r")) {
6746 while (defined (my $line = <$fd>)) {
6747 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
6748 $map->{$1} = $2 if $1;
6749 } else {
6750 print STDERR "unable to parse line in statfile - $line\n";
6751 }
6752 }
6753 $fd->close();
6754 }
6755
6756 my $confsrc = "$tmpdir/qemu-server.conf";
6757
6758 my $srcfd = new IO::File($confsrc, "r") ||
6759 die "unable to open file '$confsrc'\n";
6760
6761 my $outfd = new IO::File ($tmpfn, "w") ||
6762 die "unable to write config for VM $vmid\n";
6763
6764 my $cookie = { netcount => 0 };
6765 while (defined (my $line = <$srcfd>)) {
6766 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
6767 }
6768
6769 $srcfd->close();
6770 $outfd->close();
6771 };
6772 my $err = $@;
6773
6774 if ($err) {
6775
6776 unlink $tmpfn;
6777
6778 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
6779
6780 die $err;
6781 }
6782
6783 rmtree $tmpdir;
6784
6785 rename $tmpfn, $conffile ||
6786 die "unable to commit configuration file '$conffile'\n";
6787
6788 PVE::Cluster::cfs_update(); # make sure we read new file
6789
6790 eval { rescan($vmid, 1); };
6791 warn $@ if $@;
6792 };
6793
6794 sub foreach_storage_used_by_vm {
6795 my ($conf, $func) = @_;
6796
6797 my $sidhash = {};
6798
6799 foreach_drive($conf, sub {
6800 my ($ds, $drive) = @_;
6801 return if drive_is_cdrom($drive);
6802
6803 my $volid = $drive->{file};
6804
6805 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
6806 $sidhash->{$sid} = $sid if $sid;
6807 });
6808
6809 foreach my $sid (sort keys %$sidhash) {
6810 &$func($sid);
6811 }
6812 }
6813
6814 sub do_snapshots_with_qemu {
6815 my ($storecfg, $volid) = @_;
6816
6817 my $storage_name = PVE::Storage::parse_volume_id($volid);
6818 my $scfg = $storecfg->{ids}->{$storage_name};
6819
6820 if ($qemu_snap_storage->{$scfg->{type}} && !$scfg->{krbd}){
6821 return 1;
6822 }
6823
6824 if ($volid =~ m/\.(qcow2|qed)$/){
6825 return 1;
6826 }
6827
6828 return undef;
6829 }
6830
6831 sub qga_check_running {
6832 my ($vmid, $nowarn) = @_;
6833
6834 eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
6835 if ($@) {
6836 warn "Qemu Guest Agent is not running - $@" if !$nowarn;
6837 return 0;
6838 }
6839 return 1;
6840 }
6841
6842 sub template_create {
6843 my ($vmid, $conf, $disk) = @_;
6844
6845 my $storecfg = PVE::Storage::config();
6846
6847 foreach_drive($conf, sub {
6848 my ($ds, $drive) = @_;
6849
6850 return if drive_is_cdrom($drive);
6851 return if $disk && $ds ne $disk;
6852
6853 my $volid = $drive->{file};
6854 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6855
6856 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6857 $drive->{file} = $voliddst;
6858 $conf->{$ds} = print_drive($vmid, $drive);
6859 PVE::QemuConfig->write_config($vmid, $conf);
6860 });
6861 }
6862
6863 sub convert_iscsi_path {
6864 my ($path) = @_;
6865
6866 if ($path =~ m|^iscsi://([^/]+)/([^/]+)/(.+)$|) {
6867 my $portal = $1;
6868 my $target = $2;
6869 my $lun = $3;
6870
6871 my $initiator_name = get_initiator_name();
6872
6873 return "file.driver=iscsi,file.transport=tcp,file.initiator-name=$initiator_name,".
6874 "file.portal=$portal,file.target=$target,file.lun=$lun,driver=raw";
6875 }
6876
6877 die "cannot convert iscsi path '$path', unkown format\n";
6878 }
6879
6880 sub qemu_img_convert {
6881 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
6882
6883 my $storecfg = PVE::Storage::config();
6884 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6885 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6886
6887 die "destination '$dst_volid' is not a valid volid form qemu-img convert\n" if !$dst_storeid;
6888
6889 my $cachemode;
6890 my $src_path;
6891 my $src_is_iscsi = 0;
6892 my $src_format = 'raw';
6893
6894 if ($src_storeid) {
6895 PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
6896 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6897 $src_format = qemu_img_format($src_scfg, $src_volname);
6898 $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6899 $src_is_iscsi = ($src_path =~ m|^iscsi://|);
6900 $cachemode = 'none' if $src_scfg->{type} eq 'zfspool';
6901 } elsif (-f $src_volid) {
6902 $src_path = $src_volid;
6903 if ($src_path =~ m/\.($QEMU_FORMAT_RE)$/) {
6904 $src_format = $1;
6905 }
6906 }
6907
6908 die "source '$src_volid' is not a valid volid nor path for qemu-img convert\n" if !$src_path;
6909
6910 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6911 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6912 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6913 my $dst_is_iscsi = ($dst_path =~ m|^iscsi://|);
6914
6915 my $cmd = [];
6916 push @$cmd, '/usr/bin/qemu-img', 'convert', '-p', '-n';
6917 push @$cmd, '-l', "snapshot.name=$snapname" if($snapname && $src_format eq "qcow2");
6918 push @$cmd, '-t', 'none' if $dst_scfg->{type} eq 'zfspool';
6919 push @$cmd, '-T', $cachemode if defined($cachemode);
6920
6921 if ($src_is_iscsi) {
6922 push @$cmd, '--image-opts';
6923 $src_path = convert_iscsi_path($src_path);
6924 } else {
6925 push @$cmd, '-f', $src_format;
6926 }
6927
6928 if ($dst_is_iscsi) {
6929 push @$cmd, '--target-image-opts';
6930 $dst_path = convert_iscsi_path($dst_path);
6931 } else {
6932 push @$cmd, '-O', $dst_format;
6933 }
6934
6935 push @$cmd, $src_path;
6936
6937 if (!$dst_is_iscsi && $is_zero_initialized) {
6938 push @$cmd, "zeroinit:$dst_path";
6939 } else {
6940 push @$cmd, $dst_path;
6941 }
6942
6943 my $parser = sub {
6944 my $line = shift;
6945 if($line =~ m/\((\S+)\/100\%\)/){
6946 my $percent = $1;
6947 my $transferred = int($size * $percent / 100);
6948 my $remaining = $size - $transferred;
6949
6950 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6951 }
6952
6953 };
6954
6955 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6956 my $err = $@;
6957 die "copy failed: $err" if $err;
6958 }
6959
6960 sub qemu_img_format {
6961 my ($scfg, $volname) = @_;
6962
6963 if ($scfg->{path} && $volname =~ m/\.($QEMU_FORMAT_RE)$/) {
6964 return $1;
6965 } else {
6966 return "raw";
6967 }
6968 }
6969
6970 sub qemu_drive_mirror {
6971 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
6972
6973 $jobs = {} if !$jobs;
6974
6975 my $qemu_target;
6976 my $format;
6977 $jobs->{"drive-$drive"} = {};
6978
6979 if ($dst_volid =~ /^nbd:/) {
6980 $qemu_target = $dst_volid;
6981 $format = "nbd";
6982 } else {
6983 my $storecfg = PVE::Storage::config();
6984 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6985
6986 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6987
6988 $format = qemu_img_format($dst_scfg, $dst_volname);
6989
6990 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6991
6992 $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
6993 }
6994
6995 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
6996 $opts->{format} = $format if $format;
6997
6998 if (defined($bwlimit)) {
6999 $opts->{speed} = $bwlimit * 1024;
7000 print "drive mirror is starting for drive-$drive with bandwidth limit: ${bwlimit} KB/s\n";
7001 } else {
7002 print "drive mirror is starting for drive-$drive\n";
7003 }
7004
7005 # if a job already runs for this device we get an error, catch it for cleanup
7006 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
7007 if (my $err = $@) {
7008 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7009 warn "$@\n" if $@;
7010 die "mirroring error: $err\n";
7011 }
7012
7013 qemu_drive_mirror_monitor ($vmid, $vmiddst, $jobs, $skipcomplete, $qga);
7014 }
7015
7016 sub qemu_drive_mirror_monitor {
7017 my ($vmid, $vmiddst, $jobs, $skipcomplete, $qga) = @_;
7018
7019 eval {
7020 my $err_complete = 0;
7021
7022 while (1) {
7023 die "storage migration timed out\n" if $err_complete > 300;
7024
7025 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7026
7027 my $running_mirror_jobs = {};
7028 foreach my $stat (@$stats) {
7029 next if $stat->{type} ne 'mirror';
7030 $running_mirror_jobs->{$stat->{device}} = $stat;
7031 }
7032
7033 my $readycounter = 0;
7034
7035 foreach my $job (keys %$jobs) {
7036
7037 if(defined($jobs->{$job}->{complete}) && !defined($running_mirror_jobs->{$job})) {
7038 print "$job : finished\n";
7039 delete $jobs->{$job};
7040 next;
7041 }
7042
7043 die "$job: mirroring has been cancelled\n" if !defined($running_mirror_jobs->{$job});
7044
7045 my $busy = $running_mirror_jobs->{$job}->{busy};
7046 my $ready = $running_mirror_jobs->{$job}->{ready};
7047 if (my $total = $running_mirror_jobs->{$job}->{len}) {
7048 my $transferred = $running_mirror_jobs->{$job}->{offset} || 0;
7049 my $remaining = $total - $transferred;
7050 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
7051
7052 print "$job: transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
7053 }
7054
7055 $readycounter++ if $running_mirror_jobs->{$job}->{ready};
7056 }
7057
7058 last if scalar(keys %$jobs) == 0;
7059
7060 if ($readycounter == scalar(keys %$jobs)) {
7061 print "all mirroring jobs are ready \n";
7062 last if $skipcomplete; #do the complete later
7063
7064 if ($vmiddst && $vmiddst != $vmid) {
7065 my $agent_running = $qga && qga_check_running($vmid);
7066 if ($agent_running) {
7067 print "freeze filesystem\n";
7068 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
7069 } else {
7070 print "suspend vm\n";
7071 eval { PVE::QemuServer::vm_suspend($vmid, 1); };
7072 }
7073
7074 # if we clone a disk for a new target vm, we don't switch the disk
7075 PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs);
7076
7077 if ($agent_running) {
7078 print "unfreeze filesystem\n";
7079 eval { PVE::QemuServer::vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
7080 } else {
7081 print "resume vm\n";
7082 eval { PVE::QemuServer::vm_resume($vmid, 1, 1); };
7083 }
7084
7085 last;
7086 } else {
7087
7088 foreach my $job (keys %$jobs) {
7089 # try to switch the disk if source and destination are on the same guest
7090 print "$job: Completing block job...\n";
7091
7092 eval { vm_mon_cmd($vmid, "block-job-complete", device => $job) };
7093 if ($@ =~ m/cannot be completed/) {
7094 print "$job: Block job cannot be completed, try again.\n";
7095 $err_complete++;
7096 }else {
7097 print "$job: Completed successfully.\n";
7098 $jobs->{$job}->{complete} = 1;
7099 }
7100 }
7101 }
7102 }
7103 sleep 1;
7104 }
7105 };
7106 my $err = $@;
7107
7108 if ($err) {
7109 eval { PVE::QemuServer::qemu_blockjobs_cancel($vmid, $jobs) };
7110 die "mirroring error: $err";
7111 }
7112
7113 }
7114
7115 sub qemu_blockjobs_cancel {
7116 my ($vmid, $jobs) = @_;
7117
7118 foreach my $job (keys %$jobs) {
7119 print "$job: Cancelling block job\n";
7120 eval { vm_mon_cmd($vmid, "block-job-cancel", device => $job); };
7121 $jobs->{$job}->{cancel} = 1;
7122 }
7123
7124 while (1) {
7125 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
7126
7127 my $running_jobs = {};
7128 foreach my $stat (@$stats) {
7129 $running_jobs->{$stat->{device}} = $stat;
7130 }
7131
7132 foreach my $job (keys %$jobs) {
7133
7134 if (defined($jobs->{$job}->{cancel}) && !defined($running_jobs->{$job})) {
7135 print "$job: Done.\n";
7136 delete $jobs->{$job};
7137 }
7138 }
7139
7140 last if scalar(keys %$jobs) == 0;
7141
7142 sleep 1;
7143 }
7144 }
7145
7146 sub clone_disk {
7147 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
7148 $newvmid, $storage, $format, $full, $newvollist, $jobs, $skipcomplete, $qga, $bwlimit) = @_;
7149
7150 my $newvolid;
7151
7152 if (!$full) {
7153 print "create linked clone of drive $drivename ($drive->{file})\n";
7154 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
7155 push @$newvollist, $newvolid;
7156 } else {
7157
7158 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
7159 $storeid = $storage if $storage;
7160
7161 my $dst_format = resolve_dst_disk_format($storecfg, $storeid, $volname, $format);
7162 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
7163
7164 print "create full clone of drive $drivename ($drive->{file})\n";
7165 my $name = undef;
7166 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $dst_format, $name, ($size/1024));
7167 push @$newvollist, $newvolid;
7168
7169 PVE::Storage::activate_volumes($storecfg, [$newvolid]);
7170
7171 my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
7172 if (!$running || $snapname) {
7173 # TODO: handle bwlimits
7174 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
7175 } else {
7176
7177 my $kvmver = get_running_qemu_version ($vmid);
7178 if (!qemu_machine_feature_enabled (undef, $kvmver, 2, 7)) {
7179 die "drive-mirror with iothread requires qemu version 2.7 or higher\n"
7180 if $drive->{iothread};
7181 }
7182
7183 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit, $jobs, $skipcomplete, $qga, $bwlimit);
7184 }
7185 }
7186
7187 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
7188
7189 my $disk = $drive;
7190 $disk->{format} = undef;
7191 $disk->{file} = $newvolid;
7192 $disk->{size} = $size;
7193
7194 return $disk;
7195 }
7196
7197 # this only works if VM is running
7198 sub get_current_qemu_machine {
7199 my ($vmid) = @_;
7200
7201 my $cmd = { execute => 'query-machines', arguments => {} };
7202 my $res = vm_qmp_command($vmid, $cmd);
7203
7204 my ($current, $default);
7205 foreach my $e (@$res) {
7206 $default = $e->{name} if $e->{'is-default'};
7207 $current = $e->{name} if $e->{'is-current'};
7208 }
7209
7210 # fallback to the default machine if current is not supported by qemu
7211 return $current || $default || 'pc';
7212 }
7213
7214 sub get_running_qemu_version {
7215 my ($vmid) = @_;
7216 my $cmd = { execute => 'query-version', arguments => {} };
7217 my $res = vm_qmp_command($vmid, $cmd);
7218 return "$res->{qemu}->{major}.$res->{qemu}->{minor}";
7219 }
7220
7221 sub qemu_machine_feature_enabled {
7222 my ($machine, $kvmver, $version_major, $version_minor) = @_;
7223
7224 my $current_major;
7225 my $current_minor;
7226
7227 if ($machine && $machine =~ m/^((?:pc(-i440fx|-q35)?|virt)-(\d+)\.(\d+))/) {
7228
7229 $current_major = $3;
7230 $current_minor = $4;
7231
7232 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
7233
7234 $current_major = $1;
7235 $current_minor = $2;
7236 }
7237
7238 return 1 if $current_major > $version_major ||
7239 ($current_major == $version_major &&
7240 $current_minor >= $version_minor);
7241 }
7242
7243 sub qemu_machine_pxe {
7244 my ($vmid, $conf) = @_;
7245
7246 my $machine = PVE::QemuServer::get_current_qemu_machine($vmid);
7247
7248 if ($conf->{machine} && $conf->{machine} =~ m/\.pxe$/) {
7249 $machine .= '.pxe';
7250 }
7251
7252 return $machine;
7253 }
7254
7255 sub qemu_use_old_bios_files {
7256 my ($machine_type) = @_;
7257
7258 return if !$machine_type;
7259
7260 my $use_old_bios_files = undef;
7261
7262 if ($machine_type =~ m/^(\S+)\.pxe$/) {
7263 $machine_type = $1;
7264 $use_old_bios_files = 1;
7265 } else {
7266 my $kvmver = kvm_user_version();
7267 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
7268 # load new efi bios files on migration. So this hack is required to allow
7269 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
7270 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
7271 $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
7272 }
7273
7274 return ($use_old_bios_files, $machine_type);
7275 }
7276
7277 sub create_efidisk($$$$$) {
7278 my ($storecfg, $storeid, $vmid, $fmt, $arch) = @_;
7279
7280 my (undef, $ovmf_vars) = get_ovmf_files($arch);
7281 die "EFI vars default image not found\n" if ! -f $ovmf_vars;
7282
7283 my $vars_size_b = -s $ovmf_vars;
7284 my $vars_size = PVE::Tools::convert_size($vars_size_b, 'b' => 'kb');
7285 my $volid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $vmid, $fmt, undef, $vars_size);
7286 PVE::Storage::activate_volumes($storecfg, [$volid]);
7287
7288 qemu_img_convert($ovmf_vars, $volid, $vars_size_b, undef, 0);
7289
7290 return ($volid, $vars_size);
7291 }
7292
7293 sub vm_iothreads_list {
7294 my ($vmid) = @_;
7295
7296 my $res = vm_mon_cmd($vmid, 'query-iothreads');
7297
7298 my $iothreads = {};
7299 foreach my $iothread (@$res) {
7300 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
7301 }
7302
7303 return $iothreads;
7304 }
7305
7306 sub scsihw_infos {
7307 my ($conf, $drive) = @_;
7308
7309 my $maxdev = 0;
7310
7311 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)) {
7312 $maxdev = 7;
7313 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
7314 $maxdev = 1;
7315 } else {
7316 $maxdev = 256;
7317 }
7318
7319 my $controller = int($drive->{index} / $maxdev);
7320 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
7321
7322 return ($maxdev, $controller, $controller_prefix);
7323 }
7324
7325 sub add_hyperv_enlightenments {
7326 my ($cpuFlags, $winversion, $machine_type, $kvmver, $bios, $gpu_passthrough, $hv_vendor_id) = @_;
7327
7328 return if $winversion < 6;
7329 return if $bios && $bios eq 'ovmf' && $winversion < 8;
7330
7331 if ($gpu_passthrough || defined($hv_vendor_id)) {
7332 $hv_vendor_id //= 'proxmox';
7333 push @$cpuFlags , "hv_vendor_id=$hv_vendor_id";
7334 }
7335
7336 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
7337 push @$cpuFlags , 'hv_spinlocks=0x1fff';
7338 push @$cpuFlags , 'hv_vapic';
7339 push @$cpuFlags , 'hv_time';
7340 } else {
7341 push @$cpuFlags , 'hv_spinlocks=0xffff';
7342 }
7343
7344 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 6)) {
7345 push @$cpuFlags , 'hv_reset';
7346 push @$cpuFlags , 'hv_vpindex';
7347 push @$cpuFlags , 'hv_runtime';
7348 }
7349
7350 if ($winversion >= 7) {
7351 push @$cpuFlags , 'hv_relaxed';
7352
7353 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 12)) {
7354 push @$cpuFlags , 'hv_synic';
7355 push @$cpuFlags , 'hv_stimer';
7356 }
7357
7358 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 3, 1)) {
7359 push @$cpuFlags , 'hv_ipi';
7360 }
7361 }
7362 }
7363
7364 sub windows_version {
7365 my ($ostype) = @_;
7366
7367 return 0 if !$ostype;
7368
7369 my $winversion = 0;
7370
7371 if($ostype eq 'wxp' || $ostype eq 'w2k3' || $ostype eq 'w2k') {
7372 $winversion = 5;
7373 } elsif($ostype eq 'w2k8' || $ostype eq 'wvista') {
7374 $winversion = 6;
7375 } elsif ($ostype =~ m/^win(\d+)$/) {
7376 $winversion = $1;
7377 }
7378
7379 return $winversion;
7380 }
7381
7382 sub resolve_dst_disk_format {
7383 my ($storecfg, $storeid, $src_volname, $format) = @_;
7384 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
7385
7386 if (!$format) {
7387 # if no target format is specified, use the source disk format as hint
7388 if ($src_volname) {
7389 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
7390 $format = qemu_img_format($scfg, $src_volname);
7391 } else {
7392 return $defFormat;
7393 }
7394 }
7395
7396 # test if requested format is supported - else use default
7397 my $supported = grep { $_ eq $format } @$validFormats;
7398 $format = $defFormat if !$supported;
7399 return $format;
7400 }
7401
7402 sub resolve_first_disk {
7403 my $conf = shift;
7404 my @disks = PVE::QemuServer::valid_drive_names();
7405 my $firstdisk;
7406 foreach my $ds (reverse @disks) {
7407 next if !$conf->{$ds};
7408 my $disk = PVE::QemuServer::parse_drive($ds, $conf->{$ds});
7409 next if PVE::QemuServer::drive_is_cdrom($disk);
7410 $firstdisk = $ds;
7411 }
7412 return $firstdisk;
7413 }
7414
7415 sub generate_uuid {
7416 my ($uuid, $uuid_str);
7417 UUID::generate($uuid);
7418 UUID::unparse($uuid, $uuid_str);
7419 return $uuid_str;
7420 }
7421
7422 sub generate_smbios1_uuid {
7423 return "uuid=".generate_uuid();
7424 }
7425
7426 sub nbd_stop {
7427 my ($vmid) = @_;
7428
7429 vm_mon_cmd($vmid, 'nbd-server-stop');
7430 }
7431
7432 sub create_reboot_request {
7433 my ($vmid) = @_;
7434 open(my $fh, '>', "/run/qemu-server/$vmid.reboot")
7435 or die "failed to create reboot trigger file: $!\n";
7436 close($fh);
7437 }
7438
7439 sub clear_reboot_request {
7440 my ($vmid) = @_;
7441 my $path = "/run/qemu-server/$vmid.reboot";
7442 my $res = 0;
7443
7444 $res = unlink($path);
7445 die "could not remove reboot request for $vmid: $!"
7446 if !$res && $! != POSIX::ENOENT;
7447
7448 return $res;
7449 }
7450
7451 # bash completion helper
7452
7453 sub complete_backup_archives {
7454 my ($cmdname, $pname, $cvalue) = @_;
7455
7456 my $cfg = PVE::Storage::config();
7457
7458 my $storeid;
7459
7460 if ($cvalue =~ m/^([^:]+):/) {
7461 $storeid = $1;
7462 }
7463
7464 my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
7465
7466 my $res = [];
7467 foreach my $id (keys %$data) {
7468 foreach my $item (@{$data->{$id}}) {
7469 next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
7470 push @$res, $item->{volid} if defined($item->{volid});
7471 }
7472 }
7473
7474 return $res;
7475 }
7476
7477 my $complete_vmid_full = sub {
7478 my ($running) = @_;
7479
7480 my $idlist = vmstatus();
7481
7482 my $res = [];
7483
7484 foreach my $id (keys %$idlist) {
7485 my $d = $idlist->{$id};
7486 if (defined($running)) {
7487 next if $d->{template};
7488 next if $running && $d->{status} ne 'running';
7489 next if !$running && $d->{status} eq 'running';
7490 }
7491 push @$res, $id;
7492
7493 }
7494 return $res;
7495 };
7496
7497 sub complete_vmid {
7498 return &$complete_vmid_full();
7499 }
7500
7501 sub complete_vmid_stopped {
7502 return &$complete_vmid_full(0);
7503 }
7504
7505 sub complete_vmid_running {
7506 return &$complete_vmid_full(1);
7507 }
7508
7509 sub complete_storage {
7510
7511 my $cfg = PVE::Storage::config();
7512 my $ids = $cfg->{ids};
7513
7514 my $res = [];
7515 foreach my $sid (keys %$ids) {
7516 next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
7517 next if !$ids->{$sid}->{content}->{images};
7518 push @$res, $sid;
7519 }
7520
7521 return $res;
7522 }
7523
7524 1;