]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
fix #947: reenable disk/cdrom passthrough
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QemuConfig;
31 use PVE::QMPClient;
32 use PVE::RPCEnvironment;
33 use Time::HiRes qw(gettimeofday);
34 use File::Copy qw(copy);
35 use URI::Escape;
36
37 my $qemu_snap_storage = {rbd => 1, sheepdog => 1};
38
39 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
40
41 # Note about locking: we use flock on the config file protect
42 # against concurent actions.
43 # Aditionaly, we have a 'lock' setting in the config file. This
44 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
45 # allowed when such lock is set. But you can ignore this kind of
46 # lock with the --skiplock flag.
47
48 cfs_register_file('/qemu-server/',
49 \&parse_vm_config,
50 \&write_vm_config);
51
52 PVE::JSONSchema::register_standard_option('skiplock', {
53 description => "Ignore locks - only root is allowed to use this option.",
54 type => 'boolean',
55 optional => 1,
56 });
57
58 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
59 description => "Some command save/restore state from this location.",
60 type => 'string',
61 maxLength => 128,
62 optional => 1,
63 });
64
65 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
66 description => "The name of the snapshot.",
67 type => 'string', format => 'pve-configid',
68 maxLength => 40,
69 });
70
71 #no warnings 'redefine';
72
73 sub cgroups_write {
74 my ($controller, $vmid, $option, $value) = @_;
75
76 my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
77 PVE::ProcFSTools::write_proc_entry($path, $value);
78
79 }
80
81 my $nodename = PVE::INotify::nodename();
82
83 mkdir "/etc/pve/nodes/$nodename";
84 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
85 mkdir $confdir;
86
87 my $var_run_tmpdir = "/var/run/qemu-server";
88 mkdir $var_run_tmpdir;
89
90 my $lock_dir = "/var/lock/qemu-server";
91 mkdir $lock_dir;
92
93 my $pcisysfs = "/sys/bus/pci";
94
95 my $cpu_fmt = {
96 cputype => {
97 description => "Emulated CPU type.",
98 type => 'string',
99 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge IvyBridge Haswell Haswell-noTSX Broadwell Broadwell-noTSX Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
100 format_description => 'cputype',
101 default => 'kvm64',
102 default_key => 1,
103 },
104 hidden => {
105 description => "Do not identify as a KVM virtual machine.",
106 type => 'boolean',
107 optional => 1,
108 default => 0
109 },
110 };
111
112 my $watchdog_fmt = {
113 model => {
114 default_key => 1,
115 type => 'string',
116 enum => [qw(i6300esb ib700)],
117 description => "Watchdog type to emulate.",
118 default => 'i6300esb',
119 optional => 1,
120 },
121 action => {
122 type => 'string',
123 enum => [qw(reset shutdown poweroff pause debug none)],
124 description => "The action to perform if after activation the guest fails to poll the watchdog in time.",
125 optional => 1,
126 },
127 };
128 PVE::JSONSchema::register_format('pve-qm-watchdog', $watchdog_fmt);
129
130 my $confdesc = {
131 onboot => {
132 optional => 1,
133 type => 'boolean',
134 description => "Specifies whether a VM will be started during system bootup.",
135 default => 0,
136 },
137 autostart => {
138 optional => 1,
139 type => 'boolean',
140 description => "Automatic restart after crash (currently ignored).",
141 default => 0,
142 },
143 hotplug => {
144 optional => 1,
145 type => 'string', format => 'pve-hotplug-features',
146 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'.",
147 default => 'network,disk,usb',
148 },
149 reboot => {
150 optional => 1,
151 type => 'boolean',
152 description => "Allow reboot. If set to '0' the VM exit on reboot.",
153 default => 1,
154 },
155 lock => {
156 optional => 1,
157 type => 'string',
158 description => "Lock/unlock the VM.",
159 enum => [qw(migrate backup snapshot rollback)],
160 },
161 cpulimit => {
162 optional => 1,
163 type => 'number',
164 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.",
165 minimum => 0,
166 maximum => 128,
167 default => 0,
168 },
169 cpuunits => {
170 optional => 1,
171 type => 'integer',
172 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.\n\nNOTE: You can disable fair-scheduler configuration by setting this to 0.",
173 minimum => 0,
174 maximum => 500000,
175 default => 1000,
176 },
177 memory => {
178 optional => 1,
179 type => 'integer',
180 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
181 minimum => 16,
182 default => 512,
183 },
184 balloon => {
185 optional => 1,
186 type => 'integer',
187 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
188 minimum => 0,
189 },
190 shares => {
191 optional => 1,
192 type => 'integer',
193 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",
194 minimum => 0,
195 maximum => 50000,
196 default => 1000,
197 },
198 keyboard => {
199 optional => 1,
200 type => 'string',
201 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
202 enum => PVE::Tools::kvmkeymaplist(),
203 default => 'en-us',
204 },
205 name => {
206 optional => 1,
207 type => 'string', format => 'dns-name',
208 description => "Set a name for the VM. Only used on the configuration web interface.",
209 },
210 scsihw => {
211 optional => 1,
212 type => 'string',
213 description => "scsi controller model",
214 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
215 default => 'lsi',
216 },
217 description => {
218 optional => 1,
219 type => 'string',
220 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
221 },
222 ostype => {
223 optional => 1,
224 type => 'string',
225 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
226 description => <<EODESC,
227 Used to enable special optimization/features for specific
228 operating systems:
229
230 other => unspecified OS
231 wxp => Microsoft Windows XP
232 w2k => Microsoft Windows 2000
233 w2k3 => Microsoft Windows 2003
234 w2k8 => Microsoft Windows 2008
235 wvista => Microsoft Windows Vista
236 win7 => Microsoft Windows 7
237 win8 => Microsoft Windows 8/2012
238 l24 => Linux 2.4 Kernel
239 l26 => Linux 2.6/3.X Kernel
240 solaris => solaris/opensolaris/openindiania kernel
241
242 other|l24|l26|solaris ... no special behaviour
243 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
244 EODESC
245 },
246 boot => {
247 optional => 1,
248 type => 'string',
249 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
250 pattern => '[acdn]{1,4}',
251 default => 'cdn',
252 },
253 bootdisk => {
254 optional => 1,
255 type => 'string', format => 'pve-qm-bootdisk',
256 description => "Enable booting from specified disk.",
257 pattern => '(ide|sata|scsi|virtio)\d+',
258 },
259 smp => {
260 optional => 1,
261 type => 'integer',
262 description => "The number of CPUs. Please use option -sockets instead.",
263 minimum => 1,
264 default => 1,
265 },
266 sockets => {
267 optional => 1,
268 type => 'integer',
269 description => "The number of CPU sockets.",
270 minimum => 1,
271 default => 1,
272 },
273 cores => {
274 optional => 1,
275 type => 'integer',
276 description => "The number of cores per socket.",
277 minimum => 1,
278 default => 1,
279 },
280 numa => {
281 optional => 1,
282 type => 'boolean',
283 description => "Enable/disable NUMA.",
284 default => 0,
285 },
286 vcpus => {
287 optional => 1,
288 type => 'integer',
289 description => "Number of hotplugged vcpus.",
290 minimum => 1,
291 default => 0,
292 },
293 acpi => {
294 optional => 1,
295 type => 'boolean',
296 description => "Enable/disable ACPI.",
297 default => 1,
298 },
299 agent => {
300 optional => 1,
301 type => 'boolean',
302 description => "Enable/disable Qemu GuestAgent.",
303 default => 0,
304 },
305 kvm => {
306 optional => 1,
307 type => 'boolean',
308 description => "Enable/disable KVM hardware virtualization.",
309 default => 1,
310 },
311 tdf => {
312 optional => 1,
313 type => 'boolean',
314 description => "Enable/disable time drift fix.",
315 default => 0,
316 },
317 localtime => {
318 optional => 1,
319 type => 'boolean',
320 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
321 },
322 freeze => {
323 optional => 1,
324 type => 'boolean',
325 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
326 },
327 vga => {
328 optional => 1,
329 type => 'string',
330 description => "Select the VGA type. If you want to use high resolution" .
331 " modes (>= 1280x1024x16) then you should use the options " .
332 "'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and " .
333 "'cirrus' for other OS types. The 'qxl' option enables the SPICE " .
334 "display sever. For win* OS you can select how many independent " .
335 "displays you want, Linux guests can add displays them self. " .
336 "You can also run without any graphic card, using a serial device" .
337 " as terminal.",
338 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
339 },
340 watchdog => {
341 optional => 1,
342 type => 'string', format => 'pve-qm-watchdog',
343 description => "Create a virtual hardware watchdog device. Once enabled" .
344 " (by a guest action), the watchdog must be periodically polled " .
345 "by an agent inside the guest or else the watchdog will reset " .
346 "the guest (or execute the respective action specified)",
347 },
348 startdate => {
349 optional => 1,
350 type => 'string',
351 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
352 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'.",
353 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
354 default => 'now',
355 },
356 startup => get_standard_option('pve-startup-order'),
357 template => {
358 optional => 1,
359 type => 'boolean',
360 description => "Enable/disable Template.",
361 default => 0,
362 },
363 args => {
364 optional => 1,
365 type => 'string',
366 description => <<EODESCR,
367 Arbitrary arguments passed to kvm, for example:
368
369 args: -no-reboot -no-hpet
370
371 NOTE: this option is for experts only.
372 EODESCR
373 },
374 tablet => {
375 optional => 1,
376 type => 'boolean',
377 default => 1,
378 description => "Enable/disable the USB tablet device. This device is " .
379 "usually needed to allow absolute mouse positioning with VNC. " .
380 "Else the mouse runs out of sync with normal VNC clients. " .
381 "If you're running lots of console-only guests on one host, " .
382 "you may consider disabling this to save some context switches. " .
383 "This is turned off by default if you use spice (-vga=qxl).",
384 },
385 migrate_speed => {
386 optional => 1,
387 type => 'integer',
388 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
389 minimum => 0,
390 default => 0,
391 },
392 migrate_downtime => {
393 optional => 1,
394 type => 'number',
395 description => "Set maximum tolerated downtime (in seconds) for migrations.",
396 minimum => 0,
397 default => 0.1,
398 },
399 cdrom => {
400 optional => 1,
401 type => 'string', format => 'pve-qm-drive',
402 typetext => 'volume',
403 description => "This is an alias for option -ide2",
404 },
405 cpu => {
406 optional => 1,
407 description => "Emulated CPU type.",
408 type => 'string',
409 format => $cpu_fmt,
410 },
411 parent => get_standard_option('pve-snapshot-name', {
412 optional => 1,
413 description => "Parent snapshot name. This is used internally, and should not be modified.",
414 }),
415 snaptime => {
416 optional => 1,
417 description => "Timestamp for snapshots.",
418 type => 'integer',
419 minimum => 0,
420 },
421 vmstate => {
422 optional => 1,
423 type => 'string', format => 'pve-volume-id',
424 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
425 },
426 machine => {
427 description => "Specific the Qemu machine type.",
428 type => 'string',
429 pattern => '(pc|pc(-i440fx)?-\d+\.\d+(\.pxe)?|q35|pc-q35-\d+\.\d+(\.pxe)?)',
430 maxLength => 40,
431 optional => 1,
432 },
433 smbios1 => {
434 description => "Specify SMBIOS type 1 fields.",
435 type => 'string', format => 'pve-qm-smbios1',
436 maxLength => 256,
437 optional => 1,
438 },
439 protection => {
440 optional => 1,
441 type => 'boolean',
442 description => "Sets the protection flag of the VM. This will prevent the remove operation.",
443 default => 0,
444 },
445 bios => {
446 optional => 1,
447 type => 'string',
448 enum => [ qw(seabios ovmf) ],
449 description => "Select BIOS implementation.",
450 default => 'seabios',
451 },
452 };
453
454 # what about other qemu settings ?
455 #cpu => 'string',
456 #machine => 'string',
457 #fda => 'file',
458 #fdb => 'file',
459 #mtdblock => 'file',
460 #sd => 'file',
461 #pflash => 'file',
462 #snapshot => 'bool',
463 #bootp => 'file',
464 ##tftp => 'dir',
465 ##smb => 'dir',
466 #kernel => 'file',
467 #append => 'string',
468 #initrd => 'file',
469 ##soundhw => 'string',
470
471 while (my ($k, $v) = each %$confdesc) {
472 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
473 }
474
475 my $MAX_IDE_DISKS = 4;
476 my $MAX_SCSI_DISKS = 14;
477 my $MAX_VIRTIO_DISKS = 16;
478 my $MAX_SATA_DISKS = 6;
479 my $MAX_USB_DEVICES = 5;
480 my $MAX_NETS = 32;
481 my $MAX_UNUSED_DISKS = 8;
482 my $MAX_HOSTPCI_DEVICES = 4;
483 my $MAX_SERIAL_PORTS = 4;
484 my $MAX_PARALLEL_PORTS = 3;
485 my $MAX_NUMA = 8;
486 my $MAX_MEM = 4194304;
487 my $STATICMEM = 1024;
488
489 my $numa_fmt = {
490 cpus => {
491 type => "string",
492 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
493 description => "CPUs accessing this numa node.",
494 format_description => "id[-id];...",
495 },
496 memory => {
497 type => "number",
498 description => "Amount of memory this numa node provides.",
499 format_description => "mb",
500 optional => 1,
501 },
502 hostnodes => {
503 type => "string",
504 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
505 description => "host numa nodes to use",
506 format_description => "id[-id];...",
507 optional => 1,
508 },
509 policy => {
510 type => 'string',
511 enum => [qw(preferred bind interleave)],
512 format_description => 'preferred|bind|interleave',
513 description => "numa allocation policy.",
514 optional => 1,
515 },
516 };
517 PVE::JSONSchema::register_format('pve-qm-numanode', $numa_fmt);
518 my $numadesc = {
519 optional => 1,
520 type => 'string', format => $numa_fmt,
521 description => "numa topology",
522 };
523 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
524
525 for (my $i = 0; $i < $MAX_NUMA; $i++) {
526 $confdesc->{"numa$i"} = $numadesc;
527 }
528
529 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
530 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
531 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
532 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
533
534 my $net_fmt = {
535 macaddr => {
536 type => 'string',
537 pattern => qr/[0-9a-f]{2}(?::[0-9a-f]{2}){5}/i,
538 description => "MAC address",
539 format_description => "XX:XX:XX:XX:XX:XX",
540 optional => 1,
541 },
542 model => { alias => 'macaddr', default_key => 1 },
543 (map { $_ => { group => 'model' } } @$nic_model_list),
544 bridge => {
545 type => 'string',
546 description => 'Bridge to attach the network device to.',
547 format_description => 'bridge',
548 optional => 1,
549 },
550 queues => {
551 type => 'integer',
552 minimum => 0, maximum => 16,
553 description => 'Number of packet queues to be used on the device.',
554 format_description => 'number',
555 optional => 1,
556 },
557 rate => {
558 type => 'number',
559 minimum => 0,
560 description => 'Rate limit in mbps as floating point number.',
561 format_description => 'mbps',
562 optional => 1,
563 },
564 tag => {
565 type => 'integer',
566 minimum => 2, maximum => 4094,
567 description => 'VLAN tag to apply to packets on this interface.',
568 format_description => 'vlanid',
569 optional => 1,
570 },
571 trunks => {
572 type => 'string',
573 pattern => qr/\d+(?:-\d+)?(?:;\d+(?:-\d+)?)*/,
574 description => 'VLAN trunks to pass through this interface.',
575 format_description => 'id;id...',
576 optional => 1,
577 },
578 firewall => {
579 type => 'boolean',
580 description => 'Whether this interface should be protected by the firewall.',
581 format_description => '0|1',
582 optional => 1,
583 },
584 link_down => {
585 type => 'boolean',
586 description => 'Whether this interface should be DISconnected (like pulling the plug).',
587 format_description => '0|1',
588 optional => 1,
589 },
590 };
591 my $netdesc = {
592 optional => 1,
593 type => 'string', format => 'pve-qm-net',
594 description => <<EODESCR,
595 Specify network devices.
596
597 MODEL is one of: $nic_model_list_txt
598
599 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
600 automatically generated if not specified.
601
602 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
603
604 Option 'rate' is used to limit traffic bandwidth from and to this interface. It is specified as floating point number, unit is 'Megabytes per second'.
605
606 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
607
608 10.0.2.2 Gateway
609 10.0.2.3 DNS Server
610 10.0.2.4 SMB Server
611
612 The DHCP server assign addresses to the guest starting from 10.0.2.15.
613
614 EODESCR
615 };
616 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
617
618 for (my $i = 0; $i < $MAX_NETS; $i++) {
619 $confdesc->{"net$i"} = $netdesc;
620 }
621
622 PVE::JSONSchema::register_format('pve-volume-id-or-qm-path', \&verify_volume_id_or_qm_path);
623 sub verify_volume_id_or_qm_path {
624 my ($volid, $noerr) = @_;
625
626 if ($volid eq 'none' || $volid eq 'cdrom' || $volid =~ m|^/|) {
627 return $volid;
628 }
629
630 # if its neither 'none' nor 'cdrom' nor a path, check if its a volume-id
631 $volid = eval { PVE::JSONSchema::check_format('pve-volume-id', $volid, '') };
632 if ($@) {
633 return undef if $noerr;
634 die $@;
635 }
636 return $volid;
637 }
638
639 my $drivename_hash;
640
641 my %drivedesc_base = (
642 volume => { alias => 'file' },
643 file => {
644 type => 'string',
645 format => 'pve-volume-id-or-qm-path',
646 default_key => 1,
647 format_description => 'volume',
648 description => "The drive's backing volume.",
649 },
650 media => {
651 type => 'string',
652 format_description => 'cdrom|disk',
653 enum => [qw(cdrom disk)],
654 description => "The drive's media type.",
655 default => 'disk',
656 optional => 1
657 },
658 cyls => {
659 type => 'integer',
660 format_description => 'count',
661 description => "Force the drive's physical geometry to have a specific cylinder count.",
662 optional => 1
663 },
664 heads => {
665 type => 'integer',
666 format_description => 'count',
667 description => "Force the drive's physical geometry to have a specific head count.",
668 optional => 1
669 },
670 secs => {
671 type => 'integer',
672 format_description => 'count',
673 description => "Force the drive's physical geometry to have a specific sector count.",
674 optional => 1
675 },
676 trans => {
677 type => 'string',
678 format_description => 'none|lba|auto',
679 enum => [qw(none lba auto)],
680 description => "Force disk geometry bios translation mode.",
681 optional => 1,
682 },
683 snapshot => {
684 type => 'boolean',
685 format_description => 'on|off',
686 description => "Whether the drive should be included when making snapshots.",
687 optional => 1,
688 },
689 cache => {
690 type => 'string',
691 format_description => 'none|writethrough|writeback|unsafe|directsync',
692 enum => [qw(none writethrough writeback unsafe directsync)],
693 description => "The drive's cache mode",
694 optional => 1,
695 },
696 format => {
697 type => 'string',
698 format_description => 'drive format',
699 enum => [qw(raw cow qcow qed qcow2 vmdk cloop)],
700 description => "The drive's backing file's data format.",
701 optional => 1,
702 },
703 size => {
704 type => 'string',
705 format => 'disk-size',
706 description => "Disk size. This is purely informational and has no effect.",
707 optional => 1,
708 },
709 backup => {
710 type => 'boolean',
711 format_description => 'on|off',
712 description => "Whether the drive should be included when making backups.",
713 optional => 1,
714 },
715 werror => {
716 type => 'string',
717 format_description => 'enospc|ignore|report|stop',
718 enum => [qw(enospc ignore report stop)],
719 description => 'Write error action.',
720 optional => 1,
721 },
722 aio => {
723 type => 'string',
724 format_description => 'native|threads',
725 enum => [qw(native threads)],
726 description => 'AIO type to use.',
727 optional => 1,
728 },
729 discard => {
730 type => 'string',
731 format_description => 'ignore|on',
732 enum => [qw(ignore on)],
733 description => 'Controls whether to pass discard/trim requests to the underlying storage.',
734 optional => 1,
735 },
736 detect_zeroes => {
737 type => 'boolean',
738 description => 'Controls whether to detect and try to optimize writes of zeroes.',
739 optional => 1,
740 },
741 serial => {
742 type => 'string',
743 format => 'urlencoded',
744 format_description => 'serial',
745 maxLength => 20*3, # *3 since it's %xx url enoded
746 description => "The drive's reported serial number, url-encoded, up to 20 bytes long.",
747 optional => 1,
748 }
749 );
750
751 my %rerror_fmt = (
752 rerror => {
753 type => 'string',
754 format_description => 'ignore|report|stop',
755 enum => [qw(ignore report stop)],
756 description => 'Read error action.',
757 optional => 1,
758 },
759 );
760
761 my %iothread_fmt = ( iothread => {
762 type => 'boolean',
763 format_description => 'off|on',
764 description => "Whether to use iothreads for this drive",
765 optional => 1,
766 });
767
768 my %model_fmt = (
769 model => {
770 type => 'string',
771 format => 'urlencoded',
772 format_description => 'model',
773 maxLength => 40*3, # *3 since it's %xx url enoded
774 description => "The drive's reported model name, url-encoded, up to 40 bytes long.",
775 optional => 1,
776 },
777 );
778
779 my %queues_fmt = (
780 queues => {
781 type => 'integer',
782 format_description => 'nbqueues',
783 description => "Number of queues.",
784 minimum => 2,
785 optional => 1
786 }
787 );
788
789 my $add_throttle_desc = sub {
790 my ($key, $type, $what, $size, $longsize) = @_;
791 $drivedesc_base{$key} = {
792 type => $type,
793 format_description => $size,
794 description => "Maximum $what speed in $longsize per second.",
795 optional => 1,
796 };
797 };
798 # throughput: (leaky bucket)
799 $add_throttle_desc->('bps', 'integer', 'r/w speed', 'bps', 'bytes');
800 $add_throttle_desc->('bps_rd', 'integer', 'read speed', 'bps', 'bytes');
801 $add_throttle_desc->('bps_wr', 'integer', 'write speed', 'bps', 'bytes');
802 $add_throttle_desc->('mbps', 'number', 'r/w speed', 'mbps', 'megabytes');
803 $add_throttle_desc->('mbps_rd', 'number', 'read speed', 'mbps', 'megabytes');
804 $add_throttle_desc->('mbps_wr', 'number', 'write speed', 'mbps', 'megabytes');
805 $add_throttle_desc->('iops', 'integer', 'r/w I/O', 'iops', 'operations');
806 $add_throttle_desc->('iops_rd', 'integer', 'read I/O', 'iops', 'operations');
807 $add_throttle_desc->('iops_wr', 'integer', 'write I/O', 'iops', 'operations');
808
809 # pools: (pool of IO before throttling starts taking effect)
810 $add_throttle_desc->('mbps_max', 'number', 'unthrottled r/w pool', 'mbps', 'megabytes');
811 $add_throttle_desc->('mbps_rd_max', 'number', 'unthrottled read pool', 'mbps', 'megabytes');
812 $add_throttle_desc->('mbps_wr_max', 'number', 'unthrottled write pool', 'mbps', 'megabytes');
813 $add_throttle_desc->('iops_max', 'integer', 'unthrottled r/w I/O pool', 'iops', 'operations');
814 $add_throttle_desc->('iops_rd_max', 'integer', 'unthrottled read I/O pool', 'iops', 'operations');
815 $add_throttle_desc->('iops_wr_max', 'integer', 'unthrottled write I/O pool', 'iops', 'operations');
816
817 my $ide_fmt = {
818 %drivedesc_base,
819 %rerror_fmt,
820 %model_fmt,
821 };
822
823 my $idedesc = {
824 optional => 1,
825 type => 'string', format => $ide_fmt,
826 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
827 };
828 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
829
830 my $scsi_fmt = {
831 %drivedesc_base,
832 %iothread_fmt,
833 %queues_fmt,
834 };
835 my $scsidesc = {
836 optional => 1,
837 type => 'string', format => $scsi_fmt,
838 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
839 };
840 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
841
842 my $sata_fmt = {
843 %drivedesc_base,
844 %rerror_fmt,
845 };
846 my $satadesc = {
847 optional => 1,
848 type => 'string', format => $sata_fmt,
849 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
850 };
851 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
852
853 my $virtio_fmt = {
854 %drivedesc_base,
855 %iothread_fmt,
856 %rerror_fmt,
857 };
858 my $virtiodesc = {
859 optional => 1,
860 type => 'string', format => $virtio_fmt,
861 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
862 };
863 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
864
865 my $alldrive_fmt = {
866 %drivedesc_base,
867 %rerror_fmt,
868 %iothread_fmt,
869 %model_fmt,
870 %queues_fmt,
871 };
872
873 my $usb_fmt = {
874 host => {
875 default_key => 1,
876 type => 'string', format => 'pve-qm-usb-device',
877 format_description => 'HOSTUSBDEVICE|spice',
878 description => 'The Host USB device or port or the value spice',
879 },
880 usb3 => {
881 optional => 1,
882 type => 'boolean',
883 format_description => 'yes|no',
884 description => 'Specifies whether if given host option is a USB3 device or port',
885 },
886 };
887
888 my $usbdesc = {
889 optional => 1,
890 type => 'string', format => $usb_fmt,
891 description => <<EODESCR,
892 Configure an USB device (n is 0 to 4). This can be used to
893 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
894
895 'bus-port(.port)*' (decimal numbers) or
896 'vendor_id:product_id' (hexadeciaml numbers) or
897 'spice'
898
899 You can use the 'lsusb -t' command to list existing usb devices.
900
901 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
902
903 The value 'spice' can be used to add a usb redirection devices for spice.
904
905 The 'usb3' option determines whether the device is a USB3 device or not (this does currently not work reliably with spice redirection and is then ignored).
906
907 EODESCR
908 };
909 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
910
911 # NOTE: the match-groups of this regex are used in parse_hostpci
912 my $PCIRE = qr/([a-f0-9]{2}:[a-f0-9]{2})(?:\.([a-f0-9]))?/;
913 my $hostpci_fmt = {
914 host => {
915 default_key => 1,
916 type => 'string',
917 pattern => qr/$PCIRE(;$PCIRE)*/,
918 format_description => 'HOSTPCIID[;HOSTPCIID2...]',
919 description => "The PCI ID of a host's PCI device or a list of PCI virtual functions of the host.",
920 },
921 rombar => {
922 type => 'boolean',
923 optional => 1,
924 default => 1,
925 },
926 pcie => {
927 type => 'boolean',
928 optional => 1,
929 default => 0,
930 },
931 'x-vga' => {
932 type => 'boolean',
933 optional => 1,
934 default => 0,
935 },
936 };
937 PVE::JSONSchema::register_format('pve-qm-hostpci', $hostpci_fmt);
938
939 my $hostpcidesc = {
940 optional => 1,
941 type => 'string', format => 'pve-qm-hostpci',
942 description => <<EODESCR,
943 Map host pci devices. HOSTPCIDEVICE syntax is:
944
945 'bus:dev.func' (hexadecimal numbers)
946
947 You can us the 'lspci' command to list existing pci devices.
948
949 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
950
951 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
952
953 Experimental: user reported problems with this option.
954 EODESCR
955 };
956 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
957
958 my $serialdesc = {
959 optional => 1,
960 type => 'string',
961 pattern => '(/dev/.+|socket)',
962 description => <<EODESCR,
963 Create a serial device inside the VM (n is 0 to 3), and pass through a host serial device (i.e. /dev/ttyS0), or create a unix socket on the host side (use 'qm terminal' to open a terminal connection).
964
965 NOTE: If you pass through a host serial device, it is no longer possible to migrate such machines - use with special care.
966
967 Experimental: user reported problems with this option.
968 EODESCR
969 };
970
971 my $paralleldesc= {
972 optional => 1,
973 type => 'string',
974 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
975 description => <<EODESCR,
976 Map host parallel devices (n is 0 to 2).
977
978 NOTE: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
979
980 Experimental: user reported problems with this option.
981 EODESCR
982 };
983
984 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
985 $confdesc->{"parallel$i"} = $paralleldesc;
986 }
987
988 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
989 $confdesc->{"serial$i"} = $serialdesc;
990 }
991
992 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
993 $confdesc->{"hostpci$i"} = $hostpcidesc;
994 }
995
996 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
997 $drivename_hash->{"ide$i"} = 1;
998 $confdesc->{"ide$i"} = $idedesc;
999 }
1000
1001 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
1002 $drivename_hash->{"sata$i"} = 1;
1003 $confdesc->{"sata$i"} = $satadesc;
1004 }
1005
1006 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
1007 $drivename_hash->{"scsi$i"} = 1;
1008 $confdesc->{"scsi$i"} = $scsidesc ;
1009 }
1010
1011 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
1012 $drivename_hash->{"virtio$i"} = 1;
1013 $confdesc->{"virtio$i"} = $virtiodesc;
1014 }
1015
1016 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
1017 $confdesc->{"usb$i"} = $usbdesc;
1018 }
1019
1020 my $unuseddesc = {
1021 optional => 1,
1022 type => 'string', format => 'pve-volume-id',
1023 description => "Reference to unused volumes.",
1024 };
1025
1026 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
1027 $confdesc->{"unused$i"} = $unuseddesc;
1028 }
1029
1030 my $kvm_api_version = 0;
1031
1032 sub kvm_version {
1033
1034 return $kvm_api_version if $kvm_api_version;
1035
1036 my $fh = IO::File->new("</dev/kvm") ||
1037 return 0;
1038
1039 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
1040 $kvm_api_version = $v;
1041 }
1042
1043 $fh->close();
1044
1045 return $kvm_api_version;
1046 }
1047
1048 my $kvm_user_version;
1049
1050 sub kvm_user_version {
1051
1052 return $kvm_user_version if $kvm_user_version;
1053
1054 $kvm_user_version = 'unknown';
1055
1056 my $code = sub {
1057 my $line = shift;
1058 if ($line =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)(\.\d+)?[,\s]/) {
1059 $kvm_user_version = $2;
1060 }
1061 };
1062
1063 eval { run_command("kvm -version", outfunc => $code); };
1064 warn $@ if $@;
1065
1066 return $kvm_user_version;
1067
1068 }
1069
1070 my $kernel_has_vhost_net = -c '/dev/vhost-net';
1071
1072 sub valid_drive_names {
1073 # order is important - used to autoselect boot disk
1074 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
1075 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
1076 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
1077 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
1078 }
1079
1080 sub is_valid_drivename {
1081 my $dev = shift;
1082
1083 return defined($drivename_hash->{$dev});
1084 }
1085
1086 sub option_exists {
1087 my $key = shift;
1088 return defined($confdesc->{$key});
1089 }
1090
1091 sub nic_models {
1092 return $nic_model_list;
1093 }
1094
1095 sub os_list_description {
1096
1097 return {
1098 other => 'Other',
1099 wxp => 'Windows XP',
1100 w2k => 'Windows 2000',
1101 w2k3 =>, 'Windows 2003',
1102 w2k8 => 'Windows 2008',
1103 wvista => 'Windows Vista',
1104 win7 => 'Windows 7',
1105 win8 => 'Windows 8/2012',
1106 l24 => 'Linux 2.4',
1107 l26 => 'Linux 2.6',
1108 };
1109 }
1110
1111 my $cdrom_path;
1112
1113 sub get_cdrom_path {
1114
1115 return $cdrom_path if $cdrom_path;
1116
1117 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
1118 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
1119 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
1120 }
1121
1122 sub get_iso_path {
1123 my ($storecfg, $vmid, $cdrom) = @_;
1124
1125 if ($cdrom eq 'cdrom') {
1126 return get_cdrom_path();
1127 } elsif ($cdrom eq 'none') {
1128 return '';
1129 } elsif ($cdrom =~ m|^/|) {
1130 return $cdrom;
1131 } else {
1132 return PVE::Storage::path($storecfg, $cdrom);
1133 }
1134 }
1135
1136 # try to convert old style file names to volume IDs
1137 sub filename_to_volume_id {
1138 my ($vmid, $file, $media) = @_;
1139
1140 if (!($file eq 'none' || $file eq 'cdrom' ||
1141 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
1142
1143 return undef if $file =~ m|/|;
1144
1145 if ($media && $media eq 'cdrom') {
1146 $file = "local:iso/$file";
1147 } else {
1148 $file = "local:$vmid/$file";
1149 }
1150 }
1151
1152 return $file;
1153 }
1154
1155 sub verify_media_type {
1156 my ($opt, $vtype, $media) = @_;
1157
1158 return if !$media;
1159
1160 my $etype;
1161 if ($media eq 'disk') {
1162 $etype = 'images';
1163 } elsif ($media eq 'cdrom') {
1164 $etype = 'iso';
1165 } else {
1166 die "internal error";
1167 }
1168
1169 return if ($vtype eq $etype);
1170
1171 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
1172 }
1173
1174 sub cleanup_drive_path {
1175 my ($opt, $storecfg, $drive) = @_;
1176
1177 # try to convert filesystem paths to volume IDs
1178
1179 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
1180 ($drive->{file} !~ m|^/dev/.+|) &&
1181 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
1182 ($drive->{file} !~ m/^\d+$/)) {
1183 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
1184 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
1185 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
1186 verify_media_type($opt, $vtype, $drive->{media});
1187 $drive->{file} = $volid;
1188 }
1189
1190 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
1191 }
1192
1193 sub parse_hotplug_features {
1194 my ($data) = @_;
1195
1196 my $res = {};
1197
1198 return $res if $data eq '0';
1199
1200 $data = $confdesc->{hotplug}->{default} if $data eq '1';
1201
1202 foreach my $feature (PVE::Tools::split_list($data)) {
1203 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
1204 $res->{$1} = 1;
1205 } else {
1206 warn "ignoring unknown hotplug feature '$feature'\n";
1207 }
1208 }
1209 return $res;
1210 }
1211
1212 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
1213 sub pve_verify_hotplug_features {
1214 my ($value, $noerr) = @_;
1215
1216 return $value if parse_hotplug_features($value);
1217
1218 return undef if $noerr;
1219
1220 die "unable to parse hotplug option\n";
1221 }
1222
1223 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
1224 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
1225 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
1226 # [,aio=native|threads][,discard=ignore|on][,detect_zeroes=on|off]
1227 # [,iothread=on][,serial=serial][,model=model]
1228
1229 sub parse_drive {
1230 my ($key, $data) = @_;
1231
1232 my ($interface, $index);
1233
1234 if ($key =~ m/^([^\d]+)(\d+)$/) {
1235 $interface = $1;
1236 $index = $2;
1237 } else {
1238 return undef;
1239 }
1240
1241 my $desc = $key =~ /^unused\d+$/ ? $alldrive_fmt
1242 : $confdesc->{$key}->{format};
1243 if (!$desc) {
1244 warn "invalid drive key: $key\n";
1245 return undef;
1246 }
1247 my $res = eval { PVE::JSONSchema::parse_property_string($desc, $data) };
1248 return undef if !$res;
1249 $res->{interface} = $interface;
1250 $res->{index} = $index;
1251
1252 my $error = 0;
1253 foreach my $opt (qw(bps bps_rd bps_wr)) {
1254 if (my $bps = defined(delete $res->{$opt})) {
1255 if (defined($res->{"m$opt"})) {
1256 warn "both $opt and m$opt specified\n";
1257 ++$error;
1258 next;
1259 }
1260 $res->{"m$opt"} = sprintf("%.3f", $bps / (1024*1024.0));
1261 }
1262 }
1263 return undef if $error;
1264
1265 return undef if $res->{mbps_rd} && $res->{mbps};
1266 return undef if $res->{mbps_wr} && $res->{mbps};
1267 return undef if $res->{iops_rd} && $res->{iops};
1268 return undef if $res->{iops_wr} && $res->{iops};
1269
1270 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1271 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1272 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1273 return undef if $res->{interface} eq 'virtio';
1274 }
1275
1276 if (my $size = $res->{size}) {
1277 return undef if !defined($res->{size} = PVE::JSONSchema::parse_size($size));
1278 }
1279
1280 return $res;
1281 }
1282
1283 sub print_drive {
1284 my ($vmid, $drive) = @_;
1285 my $data = { %$drive };
1286 delete $data->{$_} for qw(index interface);
1287 return PVE::JSONSchema::print_property_string($data, $alldrive_fmt);
1288 }
1289
1290 sub scsi_inquiry {
1291 my($fh, $noerr) = @_;
1292
1293 my $SG_IO = 0x2285;
1294 my $SG_GET_VERSION_NUM = 0x2282;
1295
1296 my $versionbuf = "\x00" x 8;
1297 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1298 if (!$ret) {
1299 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1300 return undef;
1301 }
1302 my $version = unpack("I", $versionbuf);
1303 if ($version < 30000) {
1304 die "scsi generic interface too old\n" if !$noerr;
1305 return undef;
1306 }
1307
1308 my $buf = "\x00" x 36;
1309 my $sensebuf = "\x00" x 8;
1310 my $cmd = pack("C x3 C x1", 0x12, 36);
1311
1312 # see /usr/include/scsi/sg.h
1313 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";
1314
1315 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1316 length($sensebuf), 0, length($buf), $buf,
1317 $cmd, $sensebuf, 6000);
1318
1319 $ret = ioctl($fh, $SG_IO, $packet);
1320 if (!$ret) {
1321 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1322 return undef;
1323 }
1324
1325 my @res = unpack($sg_io_hdr_t, $packet);
1326 if ($res[17] || $res[18]) {
1327 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1328 return undef;
1329 }
1330
1331 my $res = {};
1332 (my $byte0, my $byte1, $res->{vendor},
1333 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1334
1335 $res->{removable} = $byte1 & 128 ? 1 : 0;
1336 $res->{type} = $byte0 & 31;
1337
1338 return $res;
1339 }
1340
1341 sub path_is_scsi {
1342 my ($path) = @_;
1343
1344 my $fh = IO::File->new("+<$path") || return undef;
1345 my $res = scsi_inquiry($fh, 1);
1346 close($fh);
1347
1348 return $res;
1349 }
1350
1351 sub machine_type_is_q35 {
1352 my ($conf) = @_;
1353
1354 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1355 }
1356
1357 sub print_tabletdevice_full {
1358 my ($conf) = @_;
1359
1360 my $q35 = machine_type_is_q35($conf);
1361
1362 # we use uhci for old VMs because tablet driver was buggy in older qemu
1363 my $usbbus = $q35 ? "ehci" : "uhci";
1364
1365 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1366 }
1367
1368 sub print_drivedevice_full {
1369 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1370
1371 my $device = '';
1372 my $maxdev = 0;
1373
1374 if ($drive->{interface} eq 'virtio') {
1375 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1376 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1377 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1378 } elsif ($drive->{interface} eq 'scsi') {
1379
1380 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1381 my $unit = $drive->{index} % $maxdev;
1382 my $devicetype = 'hd';
1383 my $path = '';
1384 if (drive_is_cdrom($drive)) {
1385 $devicetype = 'cd';
1386 } else {
1387 if ($drive->{file} =~ m|^/|) {
1388 $path = $drive->{file};
1389 if (my $info = path_is_scsi($path)) {
1390 if ($info->{type} == 0) {
1391 $devicetype = 'block';
1392 } elsif ($info->{type} == 1) { # tape
1393 $devicetype = 'generic';
1394 }
1395 }
1396 } else {
1397 $path = PVE::Storage::path($storecfg, $drive->{file});
1398 }
1399
1400 if($path =~ m/^iscsi\:\/\//){
1401 $devicetype = 'generic';
1402 }
1403 }
1404
1405 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1406 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1407 } else {
1408 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1409 }
1410
1411 } elsif ($drive->{interface} eq 'ide'){
1412 $maxdev = 2;
1413 my $controller = int($drive->{index} / $maxdev);
1414 my $unit = $drive->{index} % $maxdev;
1415 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1416
1417 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1418 if ($devicetype eq 'hd' && (my $model = $drive->{model})) {
1419 $model = URI::Escape::uri_unescape($model);
1420 $device .= ",model=$model";
1421 }
1422 } elsif ($drive->{interface} eq 'sata'){
1423 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1424 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1425 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1426 } elsif ($drive->{interface} eq 'usb') {
1427 die "implement me";
1428 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1429 } else {
1430 die "unsupported interface type";
1431 }
1432
1433 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1434
1435 return $device;
1436 }
1437
1438 sub get_initiator_name {
1439 my $initiator;
1440
1441 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1442 while (defined(my $line = <$fh>)) {
1443 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1444 $initiator = $1;
1445 last;
1446 }
1447 $fh->close();
1448
1449 return $initiator;
1450 }
1451
1452 sub print_drive_full {
1453 my ($storecfg, $vmid, $drive) = @_;
1454
1455 my $path;
1456 my $volid = $drive->{file};
1457 my $format;
1458
1459 if (drive_is_cdrom($drive)) {
1460 $path = get_iso_path($storecfg, $vmid, $volid);
1461 } else {
1462 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1463 if ($storeid) {
1464 $path = PVE::Storage::path($storecfg, $volid);
1465 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1466 $format = qemu_img_format($scfg, $volname);
1467 } else {
1468 $path = $volid;
1469 $format = "raw";
1470 }
1471 }
1472
1473 my $opts = '';
1474 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1475 foreach my $o (@qemu_drive_options) {
1476 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1477 }
1478 if (my $serial = $drive->{serial}) {
1479 $serial = URI::Escape::uri_unescape($serial);
1480 $opts .= ",serial=$serial";
1481 }
1482
1483 $opts .= ",format=$format" if $format && !$drive->{format};
1484
1485 foreach my $o (qw(bps bps_rd bps_wr)) {
1486 my $v = $drive->{"m$o"};
1487 $opts .= ",$o=" . int($v*1024*1024) if $v;
1488 }
1489
1490 my $cache_direct = 0;
1491
1492 if (my $cache = $drive->{cache}) {
1493 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1494 } elsif (!drive_is_cdrom($drive)) {
1495 $opts .= ",cache=none";
1496 $cache_direct = 1;
1497 }
1498
1499 # aio native works only with O_DIRECT
1500 if (!$drive->{aio}) {
1501 if($cache_direct) {
1502 $opts .= ",aio=native";
1503 } else {
1504 $opts .= ",aio=threads";
1505 }
1506 }
1507
1508 if (!drive_is_cdrom($drive)) {
1509 my $detectzeroes;
1510 if (defined($drive->{detect_zeroes}) && !$drive->{detect_zeroes}) {
1511 $detectzeroes = 'off';
1512 } elsif ($drive->{discard}) {
1513 $detectzeroes = $drive->{discard} eq 'on' ? 'unmap' : 'on';
1514 } else {
1515 # This used to be our default with discard not being specified:
1516 $detectzeroes = 'on';
1517 }
1518 $opts .= ",detect-zeroes=$detectzeroes" if $detectzeroes;
1519 }
1520
1521 my $pathinfo = $path ? "file=$path," : '';
1522
1523 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1524 }
1525
1526 sub print_netdevice_full {
1527 my ($vmid, $conf, $net, $netid, $bridges, $use_old_bios_files) = @_;
1528
1529 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1530
1531 my $device = $net->{model};
1532 if ($net->{model} eq 'virtio') {
1533 $device = 'virtio-net-pci';
1534 };
1535
1536 my $pciaddr = print_pci_addr("$netid", $bridges);
1537 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1538 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1539 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1540 my $vectors = $net->{queues} * 2 + 2;
1541 $tmpstr .= ",vectors=$vectors,mq=on";
1542 }
1543 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1544
1545 if ($use_old_bios_files) {
1546 my $romfile;
1547 if ($device eq 'virtio-net-pci') {
1548 $romfile = 'pxe-virtio.rom';
1549 } elsif ($device eq 'e1000') {
1550 $romfile = 'pxe-e1000.rom';
1551 } elsif ($device eq 'ne2k') {
1552 $romfile = 'pxe-ne2k_pci.rom';
1553 } elsif ($device eq 'pcnet') {
1554 $romfile = 'pxe-pcnet.rom';
1555 } elsif ($device eq 'rtl8139') {
1556 $romfile = 'pxe-rtl8139.rom';
1557 }
1558 $tmpstr .= ",romfile=$romfile" if $romfile;
1559 }
1560
1561 return $tmpstr;
1562 }
1563
1564 sub print_netdev_full {
1565 my ($vmid, $conf, $net, $netid, $hotplug) = @_;
1566
1567 my $i = '';
1568 if ($netid =~ m/^net(\d+)$/) {
1569 $i = int($1);
1570 }
1571
1572 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1573
1574 my $ifname = "tap${vmid}i$i";
1575
1576 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1577 die "interface name '$ifname' is too long (max 15 character)\n"
1578 if length($ifname) >= 16;
1579
1580 my $vhostparam = '';
1581 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1582
1583 my $vmname = $conf->{name} || "vm$vmid";
1584
1585 my $netdev = "";
1586 my $script = $hotplug ? "pve-bridge-hotplug" : "pve-bridge";
1587
1588 if ($net->{bridge}) {
1589 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/$script,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1590 } else {
1591 $netdev = "type=user,id=$netid,hostname=$vmname";
1592 }
1593
1594 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1595
1596 return $netdev;
1597 }
1598
1599 sub drive_is_cdrom {
1600 my ($drive) = @_;
1601
1602 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1603
1604 }
1605
1606 sub parse_number_sets {
1607 my ($set) = @_;
1608 my $res = [];
1609 foreach my $part (split(/;/, $set)) {
1610 if ($part =~ /^\s*(\d+)(?:-(\d+))?\s*$/) {
1611 die "invalid range: $part ($2 < $1)\n" if defined($2) && $2 < $1;
1612 push @$res, [ $1, $2 ];
1613 } else {
1614 die "invalid range: $part\n";
1615 }
1616 }
1617 return $res;
1618 }
1619
1620 sub parse_numa {
1621 my ($data) = @_;
1622
1623 my $res = PVE::JSONSchema::parse_property_string($numa_fmt, $data);
1624 $res->{cpus} = parse_number_sets($res->{cpus}) if defined($res->{cpus});
1625 $res->{hostnodes} = parse_number_sets($res->{hostnodes}) if defined($res->{hostnodes});
1626 return $res;
1627 }
1628
1629 sub parse_hostpci {
1630 my ($value) = @_;
1631
1632 return undef if !$value;
1633
1634 my $res = PVE::JSONSchema::parse_property_string($hostpci_fmt, $value);
1635
1636 my @idlist = split(/;/, $res->{host});
1637 delete $res->{host};
1638 foreach my $id (@idlist) {
1639 if ($id =~ /^$PCIRE$/) {
1640 push @{$res->{pciid}}, { id => $1, function => ($2//'0') };
1641 } else {
1642 # should have been caught by parse_property_string already
1643 die "failed to parse PCI id: $id\n";
1644 }
1645 }
1646 return $res;
1647 }
1648
1649 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1650 sub parse_net {
1651 my ($data) = @_;
1652
1653 my $res = eval { PVE::JSONSchema::parse_property_string($net_fmt, $data) };
1654 if ($@) {
1655 warn $@;
1656 return undef;
1657 }
1658 $res->{macaddr} = PVE::Tools::random_ether_addr() if !defined($res->{macaddr});
1659 return $res;
1660 }
1661
1662 sub print_net {
1663 my $net = shift;
1664
1665 return PVE::JSONSchema::print_property_string($net, $net_fmt);
1666 }
1667
1668 sub add_random_macs {
1669 my ($settings) = @_;
1670
1671 foreach my $opt (keys %$settings) {
1672 next if $opt !~ m/^net(\d+)$/;
1673 my $net = parse_net($settings->{$opt});
1674 next if !$net;
1675 $settings->{$opt} = print_net($net);
1676 }
1677 }
1678
1679 sub vm_is_volid_owner {
1680 my ($storecfg, $vmid, $volid) = @_;
1681
1682 if ($volid !~ m|^/|) {
1683 my ($path, $owner);
1684 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1685 if ($owner && ($owner == $vmid)) {
1686 return 1;
1687 }
1688 }
1689
1690 return undef;
1691 }
1692
1693 sub split_flagged_list {
1694 my $text = shift || '';
1695 $text =~ s/[,;]/ /g;
1696 $text =~ s/^\s+//;
1697 return { map { /^(!?)(.*)$/ && ($2, $1) } ($text =~ /\S+/g) };
1698 }
1699
1700 sub join_flagged_list {
1701 my ($how, $lst) = @_;
1702 join $how, map { $lst->{$_} . $_ } keys %$lst;
1703 }
1704
1705 sub vmconfig_delete_pending_option {
1706 my ($conf, $key, $force) = @_;
1707
1708 delete $conf->{pending}->{$key};
1709 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
1710 $pending_delete_hash->{$key} = $force ? '!' : '';
1711 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
1712 }
1713
1714 sub vmconfig_undelete_pending_option {
1715 my ($conf, $key) = @_;
1716
1717 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
1718 delete $pending_delete_hash->{$key};
1719
1720 if (%$pending_delete_hash) {
1721 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
1722 } else {
1723 delete $conf->{pending}->{delete};
1724 }
1725 }
1726
1727 sub vmconfig_register_unused_drive {
1728 my ($storecfg, $vmid, $conf, $drive) = @_;
1729
1730 if (!drive_is_cdrom($drive)) {
1731 my $volid = $drive->{file};
1732 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1733 PVE::QemuConfig->add_unused_volume($conf, $volid, $vmid);
1734 }
1735 }
1736 }
1737
1738 sub vmconfig_cleanup_pending {
1739 my ($conf) = @_;
1740
1741 # remove pending changes when nothing changed
1742 my $changes;
1743 foreach my $opt (keys %{$conf->{pending}}) {
1744 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1745 $changes = 1;
1746 delete $conf->{pending}->{$opt};
1747 }
1748 }
1749
1750 my $current_delete_hash = split_flagged_list($conf->{pending}->{delete});
1751 my $pending_delete_hash = {};
1752 while (my ($opt, $force) = each %$current_delete_hash) {
1753 if (defined($conf->{$opt})) {
1754 $pending_delete_hash->{$opt} = $force;
1755 } else {
1756 $changes = 1;
1757 }
1758 }
1759
1760 if (%$pending_delete_hash) {
1761 $conf->{pending}->{delete} = join_flagged_list(',', $pending_delete_hash);
1762 } else {
1763 delete $conf->{pending}->{delete};
1764 }
1765
1766 return $changes;
1767 }
1768
1769 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1770 my $smbios1_fmt = {
1771 uuid => {
1772 type => 'string',
1773 pattern => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1774 format_description => 'UUID',
1775 optional => 1,
1776 },
1777 version => {
1778 type => 'string',
1779 pattern => '\S+',
1780 format_description => 'str',
1781 optional => 1,
1782 },
1783 serial => {
1784 type => 'string',
1785 pattern => '\S+',
1786 format_description => 'str',
1787 optional => 1,
1788 },
1789 manufacturer => {
1790 type => 'string',
1791 pattern => '\S+',
1792 format_description => 'name',
1793 optional => 1,
1794 },
1795 product => {
1796 type => 'string',
1797 pattern => '\S+',
1798 format_description => 'name',
1799 optional => 1,
1800 },
1801 sku => {
1802 type => 'string',
1803 pattern => '\S+',
1804 format_description => 'str',
1805 optional => 1,
1806 },
1807 family => {
1808 type => 'string',
1809 pattern => '\S+',
1810 format_description => 'str',
1811 optional => 1,
1812 },
1813 };
1814
1815 sub parse_smbios1 {
1816 my ($data) = @_;
1817
1818 my $res = eval { PVE::JSONSchema::parse_property_string($smbios1_fmt, $data) };
1819 warn $@ if $@;
1820 return $res;
1821 }
1822
1823 sub print_smbios1 {
1824 my ($smbios1) = @_;
1825 return PVE::JSONSchema::print_property_string($smbios1, $smbios1_fmt);
1826 }
1827
1828 PVE::JSONSchema::register_format('pve-qm-smbios1', $smbios1_fmt);
1829
1830 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1831 sub verify_bootdisk {
1832 my ($value, $noerr) = @_;
1833
1834 return $value if is_valid_drivename($value);
1835
1836 return undef if $noerr;
1837
1838 die "invalid boot disk '$value'\n";
1839 }
1840
1841 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1842 sub verify_net {
1843 my ($value, $noerr) = @_;
1844
1845 return $value if parse_net($value);
1846
1847 return undef if $noerr;
1848
1849 die "unable to parse network options\n";
1850 }
1851
1852 sub parse_watchdog {
1853 my ($value) = @_;
1854
1855 return undef if !$value;
1856
1857 my $res = eval { PVE::JSONSchema::parse_property_string($watchdog_fmt, $value) };
1858 warn $@ if $@;
1859 return $res;
1860 }
1861
1862 sub parse_usb_device {
1863 my ($value) = @_;
1864
1865 return undef if !$value;
1866
1867 my $res = {};
1868 if ($value =~ m/^(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1869 $res->{vendorid} = $2;
1870 $res->{productid} = $4;
1871 } elsif ($value =~ m/^(\d+)\-(\d+(\.\d+)*)$/) {
1872 $res->{hostbus} = $1;
1873 $res->{hostport} = $2;
1874 } elsif ($value =~ m/^spice$/i) {
1875 $res->{spice} = 1;
1876 } else {
1877 return undef;
1878 }
1879
1880 return $res;
1881 }
1882
1883 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1884 sub verify_usb_device {
1885 my ($value, $noerr) = @_;
1886
1887 return $value if parse_usb_device($value);
1888
1889 return undef if $noerr;
1890
1891 die "unable to parse usb device\n";
1892 }
1893
1894 # add JSON properties for create and set function
1895 sub json_config_properties {
1896 my $prop = shift;
1897
1898 foreach my $opt (keys %$confdesc) {
1899 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1900 $prop->{$opt} = $confdesc->{$opt};
1901 }
1902
1903 return $prop;
1904 }
1905
1906 sub check_type {
1907 my ($key, $value) = @_;
1908
1909 die "unknown setting '$key'\n" if !$confdesc->{$key};
1910
1911 my $type = $confdesc->{$key}->{type};
1912
1913 if (!defined($value)) {
1914 die "got undefined value\n";
1915 }
1916
1917 if ($value =~ m/[\n\r]/) {
1918 die "property contains a line feed\n";
1919 }
1920
1921 if ($type eq 'boolean') {
1922 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1923 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1924 die "type check ('boolean') failed - got '$value'\n";
1925 } elsif ($type eq 'integer') {
1926 return int($1) if $value =~ m/^(\d+)$/;
1927 die "type check ('integer') failed - got '$value'\n";
1928 } elsif ($type eq 'number') {
1929 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1930 die "type check ('number') failed - got '$value'\n";
1931 } elsif ($type eq 'string') {
1932 if (my $fmt = $confdesc->{$key}->{format}) {
1933 if ($fmt eq 'pve-qm-drive') {
1934 # special case - we need to pass $key to parse_drive()
1935 my $drive = parse_drive($key, $value);
1936 return $value if $drive;
1937 die "unable to parse drive options\n";
1938 }
1939 PVE::JSONSchema::check_format($fmt, $value);
1940 return $value;
1941 }
1942 $value =~ s/^\"(.*)\"$/$1/;
1943 return $value;
1944 } else {
1945 die "internal error"
1946 }
1947 }
1948
1949 sub check_iommu_support{
1950 #fixme : need to check IOMMU support
1951 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1952
1953 my $iommu=1;
1954 return $iommu;
1955
1956 }
1957
1958 sub touch_config {
1959 my ($vmid) = @_;
1960
1961 my $conf = PVE::QemuConfig->config_file($vmid);
1962 utime undef, undef, $conf;
1963 }
1964
1965 sub destroy_vm {
1966 my ($storecfg, $vmid, $keep_empty_config, $skiplock) = @_;
1967
1968 my $conffile = PVE::QemuConfig->config_file($vmid);
1969
1970 my $conf = PVE::QemuConfig->load_config($vmid);
1971
1972 PVE::QemuConfig->check_lock($conf) if !$skiplock;
1973
1974 # only remove disks owned by this VM
1975 foreach_drive($conf, sub {
1976 my ($ds, $drive) = @_;
1977
1978 return if drive_is_cdrom($drive);
1979
1980 my $volid = $drive->{file};
1981
1982 return if !$volid || $volid =~ m|^/|;
1983
1984 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1985 return if !$path || !$owner || ($owner != $vmid);
1986
1987 PVE::Storage::vdisk_free($storecfg, $volid);
1988 });
1989
1990 if ($keep_empty_config) {
1991 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1992 } else {
1993 unlink $conffile;
1994 }
1995
1996 # also remove unused disk
1997 eval {
1998 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1999
2000 eval {
2001 PVE::Storage::foreach_volid($dl, sub {
2002 my ($volid, $sid, $volname, $d) = @_;
2003 PVE::Storage::vdisk_free($storecfg, $volid);
2004 });
2005 };
2006 warn $@ if $@;
2007
2008 };
2009 warn $@ if $@;
2010 }
2011
2012 sub parse_vm_config {
2013 my ($filename, $raw) = @_;
2014
2015 return undef if !defined($raw);
2016
2017 my $res = {
2018 digest => Digest::SHA::sha1_hex($raw),
2019 snapshots => {},
2020 pending => {},
2021 };
2022
2023 $filename =~ m|/qemu-server/(\d+)\.conf$|
2024 || die "got strange filename '$filename'";
2025
2026 my $vmid = $1;
2027
2028 my $conf = $res;
2029 my $descr;
2030 my $section = '';
2031
2032 my @lines = split(/\n/, $raw);
2033 foreach my $line (@lines) {
2034 next if $line =~ m/^\s*$/;
2035
2036 if ($line =~ m/^\[PENDING\]\s*$/i) {
2037 $section = 'pending';
2038 if (defined($descr)) {
2039 $descr =~ s/\s+$//;
2040 $conf->{description} = $descr;
2041 }
2042 $descr = undef;
2043 $conf = $res->{$section} = {};
2044 next;
2045
2046 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
2047 $section = $1;
2048 if (defined($descr)) {
2049 $descr =~ s/\s+$//;
2050 $conf->{description} = $descr;
2051 }
2052 $descr = undef;
2053 $conf = $res->{snapshots}->{$section} = {};
2054 next;
2055 }
2056
2057 if ($line =~ m/^\#(.*)\s*$/) {
2058 $descr = '' if !defined($descr);
2059 $descr .= PVE::Tools::decode_text($1) . "\n";
2060 next;
2061 }
2062
2063 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
2064 $descr = '' if !defined($descr);
2065 $descr .= PVE::Tools::decode_text($2);
2066 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
2067 $conf->{snapstate} = $1;
2068 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2069 my $key = $1;
2070 my $value = $2;
2071 $conf->{$key} = $value;
2072 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2073 my $value = $1;
2074 if ($section eq 'pending') {
2075 $conf->{delete} = $value; # we parse this later
2076 } else {
2077 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2078 }
2079 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2080 my $key = $1;
2081 my $value = $2;
2082 eval { $value = check_type($key, $value); };
2083 if ($@) {
2084 warn "vm $vmid - unable to parse value of '$key' - $@";
2085 } else {
2086 my $fmt = $confdesc->{$key}->{format};
2087 if ($fmt && $fmt eq 'pve-qm-drive') {
2088 my $v = parse_drive($key, $value);
2089 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2090 $v->{file} = $volid;
2091 $value = print_drive($vmid, $v);
2092 } else {
2093 warn "vm $vmid - unable to parse value of '$key'\n";
2094 next;
2095 }
2096 }
2097
2098 if ($key eq 'cdrom') {
2099 $conf->{ide2} = $value;
2100 } else {
2101 $conf->{$key} = $value;
2102 }
2103 }
2104 }
2105 }
2106
2107 if (defined($descr)) {
2108 $descr =~ s/\s+$//;
2109 $conf->{description} = $descr;
2110 }
2111 delete $res->{snapstate}; # just to be sure
2112
2113 return $res;
2114 }
2115
2116 sub write_vm_config {
2117 my ($filename, $conf) = @_;
2118
2119 delete $conf->{snapstate}; # just to be sure
2120
2121 if ($conf->{cdrom}) {
2122 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2123 $conf->{ide2} = $conf->{cdrom};
2124 delete $conf->{cdrom};
2125 }
2126
2127 # we do not use 'smp' any longer
2128 if ($conf->{sockets}) {
2129 delete $conf->{smp};
2130 } elsif ($conf->{smp}) {
2131 $conf->{sockets} = $conf->{smp};
2132 delete $conf->{cores};
2133 delete $conf->{smp};
2134 }
2135
2136 my $used_volids = {};
2137
2138 my $cleanup_config = sub {
2139 my ($cref, $pending, $snapname) = @_;
2140
2141 foreach my $key (keys %$cref) {
2142 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2143 $key eq 'snapstate' || $key eq 'pending';
2144 my $value = $cref->{$key};
2145 if ($key eq 'delete') {
2146 die "propertry 'delete' is only allowed in [PENDING]\n"
2147 if !$pending;
2148 # fixme: check syntax?
2149 next;
2150 }
2151 eval { $value = check_type($key, $value); };
2152 die "unable to parse value of '$key' - $@" if $@;
2153
2154 $cref->{$key} = $value;
2155
2156 if (!$snapname && is_valid_drivename($key)) {
2157 my $drive = parse_drive($key, $value);
2158 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2159 }
2160 }
2161 };
2162
2163 &$cleanup_config($conf);
2164
2165 &$cleanup_config($conf->{pending}, 1);
2166
2167 foreach my $snapname (keys %{$conf->{snapshots}}) {
2168 die "internal error" if $snapname eq 'pending';
2169 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2170 }
2171
2172 # remove 'unusedX' settings if we re-add a volume
2173 foreach my $key (keys %$conf) {
2174 my $value = $conf->{$key};
2175 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2176 delete $conf->{$key};
2177 }
2178 }
2179
2180 my $generate_raw_config = sub {
2181 my ($conf, $pending) = @_;
2182
2183 my $raw = '';
2184
2185 # add description as comment to top of file
2186 if (defined(my $descr = $conf->{description})) {
2187 if ($descr) {
2188 foreach my $cl (split(/\n/, $descr)) {
2189 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2190 }
2191 } else {
2192 $raw .= "#\n" if $pending;
2193 }
2194 }
2195
2196 foreach my $key (sort keys %$conf) {
2197 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2198 $raw .= "$key: $conf->{$key}\n";
2199 }
2200 return $raw;
2201 };
2202
2203 my $raw = &$generate_raw_config($conf);
2204
2205 if (scalar(keys %{$conf->{pending}})){
2206 $raw .= "\n[PENDING]\n";
2207 $raw .= &$generate_raw_config($conf->{pending}, 1);
2208 }
2209
2210 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2211 $raw .= "\n[$snapname]\n";
2212 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2213 }
2214
2215 return $raw;
2216 }
2217
2218 sub load_defaults {
2219
2220 my $res = {};
2221
2222 # we use static defaults from our JSON schema configuration
2223 foreach my $key (keys %$confdesc) {
2224 if (defined(my $default = $confdesc->{$key}->{default})) {
2225 $res->{$key} = $default;
2226 }
2227 }
2228
2229 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2230 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2231
2232 return $res;
2233 }
2234
2235 sub config_list {
2236 my $vmlist = PVE::Cluster::get_vmlist();
2237 my $res = {};
2238 return $res if !$vmlist || !$vmlist->{ids};
2239 my $ids = $vmlist->{ids};
2240
2241 foreach my $vmid (keys %$ids) {
2242 my $d = $ids->{$vmid};
2243 next if !$d->{node} || $d->{node} ne $nodename;
2244 next if !$d->{type} || $d->{type} ne 'qemu';
2245 $res->{$vmid}->{exists} = 1;
2246 }
2247 return $res;
2248 }
2249
2250 # test if VM uses local resources (to prevent migration)
2251 sub check_local_resources {
2252 my ($conf, $noerr) = @_;
2253
2254 my $loc_res = 0;
2255
2256 $loc_res = 1 if $conf->{hostusb}; # old syntax
2257 $loc_res = 1 if $conf->{hostpci}; # old syntax
2258
2259 foreach my $k (keys %$conf) {
2260 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2261 # sockets are safe: they will recreated be on the target side post-migrate
2262 next if $k =~ m/^serial/ && ($conf->{$k} eq 'socket');
2263 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2264 }
2265
2266 die "VM uses local resources\n" if $loc_res && !$noerr;
2267
2268 return $loc_res;
2269 }
2270
2271 # check if used storages are available on all nodes (use by migrate)
2272 sub check_storage_availability {
2273 my ($storecfg, $conf, $node) = @_;
2274
2275 foreach_drive($conf, sub {
2276 my ($ds, $drive) = @_;
2277
2278 my $volid = $drive->{file};
2279 return if !$volid;
2280
2281 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2282 return if !$sid;
2283
2284 # check if storage is available on both nodes
2285 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2286 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2287 });
2288 }
2289
2290 # list nodes where all VM images are available (used by has_feature API)
2291 sub shared_nodes {
2292 my ($conf, $storecfg) = @_;
2293
2294 my $nodelist = PVE::Cluster::get_nodelist();
2295 my $nodehash = { map { $_ => 1 } @$nodelist };
2296 my $nodename = PVE::INotify::nodename();
2297
2298 foreach_drive($conf, sub {
2299 my ($ds, $drive) = @_;
2300
2301 my $volid = $drive->{file};
2302 return if !$volid;
2303
2304 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2305 if ($storeid) {
2306 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2307 if ($scfg->{disable}) {
2308 $nodehash = {};
2309 } elsif (my $avail = $scfg->{nodes}) {
2310 foreach my $node (keys %$nodehash) {
2311 delete $nodehash->{$node} if !$avail->{$node};
2312 }
2313 } elsif (!$scfg->{shared}) {
2314 foreach my $node (keys %$nodehash) {
2315 delete $nodehash->{$node} if $node ne $nodename
2316 }
2317 }
2318 }
2319 });
2320
2321 return $nodehash
2322 }
2323
2324 sub check_cmdline {
2325 my ($pidfile, $pid) = @_;
2326
2327 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2328 if (defined($fh)) {
2329 my $line = <$fh>;
2330 $fh->close;
2331 return undef if !$line;
2332 my @param = split(/\0/, $line);
2333
2334 my $cmd = $param[0];
2335 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2336
2337 for (my $i = 0; $i < scalar (@param); $i++) {
2338 my $p = $param[$i];
2339 next if !$p;
2340 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2341 my $p = $param[$i+1];
2342 return 1 if $p && ($p eq $pidfile);
2343 return undef;
2344 }
2345 }
2346 }
2347 return undef;
2348 }
2349
2350 sub check_running {
2351 my ($vmid, $nocheck, $node) = @_;
2352
2353 my $filename = PVE::QemuConfig->config_file($vmid, $node);
2354
2355 die "unable to find configuration file for VM $vmid - no such machine\n"
2356 if !$nocheck && ! -f $filename;
2357
2358 my $pidfile = pidfile_name($vmid);
2359
2360 if (my $fd = IO::File->new("<$pidfile")) {
2361 my $st = stat($fd);
2362 my $line = <$fd>;
2363 close($fd);
2364
2365 my $mtime = $st->mtime;
2366 if ($mtime > time()) {
2367 warn "file '$filename' modified in future\n";
2368 }
2369
2370 if ($line =~ m/^(\d+)$/) {
2371 my $pid = $1;
2372 if (check_cmdline($pidfile, $pid)) {
2373 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2374 return $pid;
2375 }
2376 }
2377 }
2378 }
2379
2380 return undef;
2381 }
2382
2383 sub vzlist {
2384
2385 my $vzlist = config_list();
2386
2387 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2388
2389 while (defined(my $de = $fd->read)) {
2390 next if $de !~ m/^(\d+)\.pid$/;
2391 my $vmid = $1;
2392 next if !defined($vzlist->{$vmid});
2393 if (my $pid = check_running($vmid)) {
2394 $vzlist->{$vmid}->{pid} = $pid;
2395 }
2396 }
2397
2398 return $vzlist;
2399 }
2400
2401 sub disksize {
2402 my ($storecfg, $conf) = @_;
2403
2404 my $bootdisk = $conf->{bootdisk};
2405 return undef if !$bootdisk;
2406 return undef if !is_valid_drivename($bootdisk);
2407
2408 return undef if !$conf->{$bootdisk};
2409
2410 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2411 return undef if !defined($drive);
2412
2413 return undef if drive_is_cdrom($drive);
2414
2415 my $volid = $drive->{file};
2416 return undef if !$volid;
2417
2418 return $drive->{size};
2419 }
2420
2421 my $last_proc_pid_stat;
2422
2423 # get VM status information
2424 # This must be fast and should not block ($full == false)
2425 # We only query KVM using QMP if $full == true (this can be slow)
2426 sub vmstatus {
2427 my ($opt_vmid, $full) = @_;
2428
2429 my $res = {};
2430
2431 my $storecfg = PVE::Storage::config();
2432
2433 my $list = vzlist();
2434 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2435
2436 my $cpucount = $cpuinfo->{cpus} || 1;
2437
2438 foreach my $vmid (keys %$list) {
2439 next if $opt_vmid && ($vmid ne $opt_vmid);
2440
2441 my $cfspath = PVE::QemuConfig->cfs_config_path($vmid);
2442 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2443
2444 my $d = {};
2445 $d->{pid} = $list->{$vmid}->{pid};
2446
2447 # fixme: better status?
2448 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2449
2450 my $size = disksize($storecfg, $conf);
2451 if (defined($size)) {
2452 $d->{disk} = 0; # no info available
2453 $d->{maxdisk} = $size;
2454 } else {
2455 $d->{disk} = 0;
2456 $d->{maxdisk} = 0;
2457 }
2458
2459 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2460 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2461 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2462
2463 $d->{name} = $conf->{name} || "VM $vmid";
2464 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2465
2466 if ($conf->{balloon}) {
2467 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2468 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2469 }
2470
2471 $d->{uptime} = 0;
2472 $d->{cpu} = 0;
2473 $d->{mem} = 0;
2474
2475 $d->{netout} = 0;
2476 $d->{netin} = 0;
2477
2478 $d->{diskread} = 0;
2479 $d->{diskwrite} = 0;
2480
2481 $d->{template} = PVE::QemuConfig->is_template($conf);
2482
2483 $res->{$vmid} = $d;
2484 }
2485
2486 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2487 foreach my $dev (keys %$netdev) {
2488 next if $dev !~ m/^tap([1-9]\d*)i/;
2489 my $vmid = $1;
2490 my $d = $res->{$vmid};
2491 next if !$d;
2492
2493 $d->{netout} += $netdev->{$dev}->{receive};
2494 $d->{netin} += $netdev->{$dev}->{transmit};
2495
2496 if ($full) {
2497 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
2498 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
2499 }
2500
2501 }
2502
2503 my $ctime = gettimeofday;
2504
2505 foreach my $vmid (keys %$list) {
2506
2507 my $d = $res->{$vmid};
2508 my $pid = $d->{pid};
2509 next if !$pid;
2510
2511 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2512 next if !$pstat; # not running
2513
2514 my $used = $pstat->{utime} + $pstat->{stime};
2515
2516 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2517
2518 if ($pstat->{vsize}) {
2519 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2520 }
2521
2522 my $old = $last_proc_pid_stat->{$pid};
2523 if (!$old) {
2524 $last_proc_pid_stat->{$pid} = {
2525 time => $ctime,
2526 used => $used,
2527 cpu => 0,
2528 };
2529 next;
2530 }
2531
2532 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2533
2534 if ($dtime > 1000) {
2535 my $dutime = $used - $old->{used};
2536
2537 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2538 $last_proc_pid_stat->{$pid} = {
2539 time => $ctime,
2540 used => $used,
2541 cpu => $d->{cpu},
2542 };
2543 } else {
2544 $d->{cpu} = $old->{cpu};
2545 }
2546 }
2547
2548 return $res if !$full;
2549
2550 my $qmpclient = PVE::QMPClient->new();
2551
2552 my $ballooncb = sub {
2553 my ($vmid, $resp) = @_;
2554
2555 my $info = $resp->{'return'};
2556 return if !$info->{max_mem};
2557
2558 my $d = $res->{$vmid};
2559
2560 # use memory assigned to VM
2561 $d->{maxmem} = $info->{max_mem};
2562 $d->{balloon} = $info->{actual};
2563
2564 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2565 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2566 $d->{freemem} = $info->{free_mem};
2567 }
2568
2569 $d->{ballooninfo} = $info;
2570 };
2571
2572 my $blockstatscb = sub {
2573 my ($vmid, $resp) = @_;
2574 my $data = $resp->{'return'} || [];
2575 my $totalrdbytes = 0;
2576 my $totalwrbytes = 0;
2577
2578 for my $blockstat (@$data) {
2579 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2580 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2581
2582 $blockstat->{device} =~ s/drive-//;
2583 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
2584 }
2585 $res->{$vmid}->{diskread} = $totalrdbytes;
2586 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2587 };
2588
2589 my $statuscb = sub {
2590 my ($vmid, $resp) = @_;
2591
2592 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2593 # this fails if ballon driver is not loaded, so this must be
2594 # the last commnand (following command are aborted if this fails).
2595 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2596
2597 my $status = 'unknown';
2598 if (!defined($status = $resp->{'return'}->{status})) {
2599 warn "unable to get VM status\n";
2600 return;
2601 }
2602
2603 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2604 };
2605
2606 foreach my $vmid (keys %$list) {
2607 next if $opt_vmid && ($vmid ne $opt_vmid);
2608 next if !$res->{$vmid}->{pid}; # not running
2609 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2610 }
2611
2612 $qmpclient->queue_execute(undef, 1);
2613
2614 foreach my $vmid (keys %$list) {
2615 next if $opt_vmid && ($vmid ne $opt_vmid);
2616 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2617 }
2618
2619 return $res;
2620 }
2621
2622 sub foreach_dimm {
2623 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2624
2625 my $dimm_id = 0;
2626 my $current_size = 1024;
2627 my $dimm_size = 512;
2628 return if $current_size == $memory;
2629
2630 for (my $j = 0; $j < 8; $j++) {
2631 for (my $i = 0; $i < 32; $i++) {
2632 my $name = "dimm${dimm_id}";
2633 $dimm_id++;
2634 my $numanode = $i % $sockets;
2635 $current_size += $dimm_size;
2636 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2637 return $current_size if $current_size >= $memory;
2638 }
2639 $dimm_size *= 2;
2640 }
2641 }
2642
2643 sub foreach_reverse_dimm {
2644 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2645
2646 my $dimm_id = 253;
2647 my $current_size = 4177920;
2648 my $dimm_size = 65536;
2649 return if $current_size == $memory;
2650
2651 for (my $j = 0; $j < 8; $j++) {
2652 for (my $i = 0; $i < 32; $i++) {
2653 my $name = "dimm${dimm_id}";
2654 $dimm_id--;
2655 my $numanode = $i % $sockets;
2656 $current_size -= $dimm_size;
2657 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2658 return $current_size if $current_size <= $memory;
2659 }
2660 $dimm_size /= 2;
2661 }
2662 }
2663
2664 sub foreach_drive {
2665 my ($conf, $func) = @_;
2666
2667 foreach my $ds (valid_drive_names()) {
2668 next if !defined($conf->{$ds});
2669
2670 my $drive = parse_drive($ds, $conf->{$ds});
2671 next if !$drive;
2672
2673 &$func($ds, $drive);
2674 }
2675 }
2676
2677 sub foreach_volid {
2678 my ($conf, $func) = @_;
2679
2680 my $volhash = {};
2681
2682 my $test_volid = sub {
2683 my ($volid, $is_cdrom) = @_;
2684
2685 return if !$volid;
2686
2687 $volhash->{$volid} = $is_cdrom || 0;
2688 };
2689
2690 foreach_drive($conf, sub {
2691 my ($ds, $drive) = @_;
2692 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2693 });
2694
2695 foreach my $snapname (keys %{$conf->{snapshots}}) {
2696 my $snap = $conf->{snapshots}->{$snapname};
2697 &$test_volid($snap->{vmstate}, 0);
2698 foreach_drive($snap, sub {
2699 my ($ds, $drive) = @_;
2700 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2701 });
2702 }
2703
2704 foreach my $volid (keys %$volhash) {
2705 &$func($volid, $volhash->{$volid});
2706 }
2707 }
2708
2709 sub vga_conf_has_spice {
2710 my ($vga) = @_;
2711
2712 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2713
2714 return $1 || 1;
2715 }
2716
2717 sub config_to_command {
2718 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2719
2720 my $cmd = [];
2721 my $globalFlags = [];
2722 my $machineFlags = [];
2723 my $rtcFlags = [];
2724 my $cpuFlags = [];
2725 my $devices = [];
2726 my $pciaddr = '';
2727 my $bridges = {};
2728 my $kvmver = kvm_user_version();
2729 my $vernum = 0; # unknown
2730 my $ostype = $conf->{ostype};
2731 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2732 $vernum = $1*1000000+$2*1000;
2733 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2734 $vernum = $1*1000000+$2*1000+$3;
2735 }
2736
2737 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2738
2739 my $have_ovz = -f '/proc/vz/vestat';
2740
2741 my $q35 = machine_type_is_q35($conf);
2742 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2743 my $machine_type = $forcemachine || $conf->{machine};
2744 my $use_old_bios_files = undef;
2745 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
2746
2747 my $cpuunits = defined($conf->{cpuunits}) ?
2748 $conf->{cpuunits} : $defaults->{cpuunits};
2749
2750 push @$cmd, '/usr/bin/systemd-run';
2751 push @$cmd, '--scope';
2752 push @$cmd, '--slice', "qemu";
2753 push @$cmd, '--unit', $vmid;
2754 # set KillMode=none, so that systemd don't kill those scopes
2755 # at shutdown (pve-manager service should stop the VMs instead)
2756 push @$cmd, '-p', "KillMode=none";
2757 push @$cmd, '-p', "CPUShares=$cpuunits";
2758 if ($conf->{cpulimit}) {
2759 my $cpulimit = int($conf->{cpulimit} * 100);
2760 push @$cmd, '-p', "CPUQuota=$cpulimit\%";
2761 }
2762
2763 push @$cmd, '/usr/bin/kvm';
2764
2765 push @$cmd, '-id', $vmid;
2766
2767 my $use_virtio = 0;
2768
2769 my $qmpsocket = qmp_socket($vmid);
2770 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2771 push @$cmd, '-mon', "chardev=qmp,mode=control";
2772
2773
2774 push @$cmd, '-pidfile' , pidfile_name($vmid);
2775
2776 push @$cmd, '-daemonize';
2777
2778 if ($conf->{smbios1}) {
2779 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2780 }
2781
2782 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
2783 my $ovmfvar = "OVMF_VARS-pure-efi.fd";
2784 my $ovmfvar_src = "/usr/share/kvm/$ovmfvar";
2785 my $ovmfvar_dst = "/tmp/$vmid-$ovmfvar";
2786 PVE::Tools::file_copy($ovmfvar_src, $ovmfvar_dst, 256*1024);
2787 push @$cmd, '-drive', "if=pflash,format=raw,readonly,file=/usr/share/kvm/OVMF-pure-efi.fd";
2788 push @$cmd, '-drive', "if=pflash,format=raw,file=$ovmfvar_dst";
2789 }
2790
2791 if ($q35) {
2792 # the q35 chipset support native usb2, so we enable usb controller
2793 # by default for this machine type
2794 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2795 } else {
2796 $pciaddr = print_pci_addr("piix3", $bridges);
2797 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2798
2799 my $use_usb2 = 0;
2800 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2801 next if !$conf->{"usb$i"};
2802 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format},$conf->{"usb$i"}) };
2803 next if !$d || $d->{usb3}; # do not add usb2 controller if we have only usb3 devices
2804 $use_usb2 = 1;
2805 }
2806 # include usb device config
2807 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2808 }
2809
2810 # add usb3 controller if needed
2811
2812 my $use_usb3 = 0;
2813 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2814 next if !$conf->{"usb$i"};
2815 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format},$conf->{"usb$i"}) };
2816 next if !$d || !$d->{usb3};
2817 $use_usb3 = 1;
2818 }
2819
2820 $pciaddr = print_pci_addr("xhci", $bridges);
2821 push @$devices, '-device', "nec-usb-xhci,id=xhci$pciaddr" if $use_usb3;
2822
2823 my $vga = $conf->{vga};
2824
2825 my $qxlnum = vga_conf_has_spice($vga);
2826 $vga = 'qxl' if $qxlnum;
2827
2828 if (!$vga) {
2829 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2830 $conf->{ostype} eq 'win7' ||
2831 $conf->{ostype} eq 'w2k8')) {
2832 $vga = 'std';
2833 } else {
2834 $vga = 'cirrus';
2835 }
2836 }
2837
2838 # enable absolute mouse coordinates (needed by vnc)
2839 my $tablet;
2840 if (defined($conf->{tablet})) {
2841 $tablet = $conf->{tablet};
2842 } else {
2843 $tablet = $defaults->{tablet};
2844 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2845 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2846 }
2847
2848 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2849
2850 my $kvm_off = 0;
2851 # host pci devices
2852 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2853 my $d = parse_hostpci($conf->{"hostpci$i"});
2854 next if !$d;
2855
2856 my $pcie = $d->{pcie};
2857 if($pcie){
2858 die "q35 machine model is not enabled" if !$q35;
2859 $pciaddr = print_pcie_addr("hostpci$i");
2860 }else{
2861 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2862 }
2863
2864 my $rombar = defined($d->{rombar}) && !$d->{rombar} ? ',rombar=0' : '';
2865 my $xvga = '';
2866 if ($d->{'x-vga'}) {
2867 $xvga = ',x-vga=on';
2868 $kvm_off = 1;
2869 $vga = 'none';
2870 if ($ostype eq 'win7' || $ostype eq 'win8' || $ostype eq 'w2k8') {
2871 push @$cpuFlags , 'hv_vendor_id=proxmox';
2872 }
2873 if ($conf->{bios} && $conf->{bios} eq 'ovmf') {
2874 $xvga = "";
2875 }
2876 }
2877 my $pcidevices = $d->{pciid};
2878 my $multifunction = 1 if @$pcidevices > 1;
2879
2880 my $j=0;
2881 foreach my $pcidevice (@$pcidevices) {
2882
2883 my $id = "hostpci$i";
2884 $id .= ".$j" if $multifunction;
2885 my $addr = $pciaddr;
2886 $addr .= ".$j" if $multifunction;
2887 my $devicestr = "vfio-pci,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2888
2889 if($j == 0){
2890 $devicestr .= "$rombar$xvga";
2891 $devicestr .= ",multifunction=on" if $multifunction;
2892 }
2893
2894 push @$devices, '-device', $devicestr;
2895 $j++;
2896 }
2897 }
2898
2899 # usb devices
2900 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2901 next if !$conf->{"usb$i"};
2902 my $d = eval { PVE::JSONSchema::parse_property_string($usbdesc->{format},$conf->{"usb$i"}) };
2903 next if !$d;
2904
2905 # if it is a usb3 device, attach it to the xhci controller, else omit the bus option
2906 my $usbbus = '';
2907 if (defined($d->{usb3}) && $d->{usb3}) {
2908 $usbbus = ',bus=xhci.0';
2909 }
2910
2911 if (defined($d->{host})) {
2912 $d = parse_usb_device($d->{host});
2913 if (defined($d->{vendorid}) && defined($d->{productid})) {
2914 push @$devices, '-device', "usb-host$usbbus,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2915 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2916 push @$devices, '-device', "usb-host$usbbus,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2917 } elsif (defined($d->{spice}) && $d->{spice}) {
2918 # usb redir support for spice, currently no usb3
2919 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2920 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2921 }
2922 }
2923 }
2924
2925 # serial devices
2926 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2927 if (my $path = $conf->{"serial$i"}) {
2928 if ($path eq 'socket') {
2929 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2930 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2931 push @$devices, '-device', "isa-serial,chardev=serial$i";
2932 } else {
2933 die "no such serial device\n" if ! -c $path;
2934 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2935 push @$devices, '-device', "isa-serial,chardev=serial$i";
2936 }
2937 }
2938 }
2939
2940 # parallel devices
2941 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2942 if (my $path = $conf->{"parallel$i"}) {
2943 die "no such parallel device\n" if ! -c $path;
2944 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2945 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2946 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2947 }
2948 }
2949
2950 my $vmname = $conf->{name} || "vm$vmid";
2951
2952 push @$cmd, '-name', $vmname;
2953
2954 my $sockets = 1;
2955 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2956 $sockets = $conf->{sockets} if $conf->{sockets};
2957
2958 my $cores = $conf->{cores} || 1;
2959
2960 my $maxcpus = $sockets * $cores;
2961
2962 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2963
2964 my $allowed_vcpus = $cpuinfo->{cpus};
2965
2966 die "MAX $allowed_vcpus vcpus allowed per VM on this node\n"
2967 if ($allowed_vcpus < $maxcpus);
2968
2969 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2970
2971 push @$cmd, '-nodefaults';
2972
2973 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2974
2975 my $bootindex_hash = {};
2976 my $i = 1;
2977 foreach my $o (split(//, $bootorder)) {
2978 $bootindex_hash->{$o} = $i*100;
2979 $i++;
2980 }
2981
2982 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2983
2984 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2985
2986 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2987
2988 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2989
2990 if ($vga && $vga !~ m/^serial\d+$/ && $vga ne 'none'){
2991 my $socket = vnc_socket($vmid);
2992 push @$cmd, '-vnc', "unix:$socket,x509,password";
2993 } else {
2994 push @$cmd, '-nographic';
2995 }
2996
2997 # time drift fix
2998 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2999
3000 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
3001 my $useLocaltime = $conf->{localtime};
3002
3003 if ($ostype) {
3004 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
3005
3006 if ($ostype =~ m/^w/) { # windows
3007 $useLocaltime = 1 if !defined($conf->{localtime});
3008
3009 # use time drift fix when acpi is enabled
3010 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
3011 $tdf = 1 if !defined($conf->{tdf});
3012 }
3013 }
3014
3015 if ($ostype eq 'win7' || $ostype eq 'win8' || $ostype eq 'w2k8' ||
3016 $ostype eq 'wvista') {
3017 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
3018 push @$cmd, '-no-hpet';
3019 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3020 push @$cpuFlags , 'hv_spinlocks=0x1fff' if !$nokvm;
3021 push @$cpuFlags , 'hv_vapic' if !$nokvm;
3022 push @$cpuFlags , 'hv_time' if !$nokvm;
3023
3024 } else {
3025 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
3026 }
3027 }
3028
3029 if ($ostype eq 'win7' || $ostype eq 'win8') {
3030 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
3031 }
3032 }
3033
3034 push @$rtcFlags, 'driftfix=slew' if $tdf;
3035
3036 if ($nokvm) {
3037 push @$machineFlags, 'accel=tcg';
3038 } else {
3039 die "No accelerator found!\n" if !$cpuinfo->{hvm};
3040 }
3041
3042 if ($machine_type) {
3043 push @$machineFlags, "type=${machine_type}";
3044 }
3045
3046 if ($conf->{startdate}) {
3047 push @$rtcFlags, "base=$conf->{startdate}";
3048 } elsif ($useLocaltime) {
3049 push @$rtcFlags, 'base=localtime';
3050 }
3051
3052 my $cpu = $nokvm ? "qemu64" : "kvm64";
3053 if (my $cputype = $conf->{cpu}) {
3054 my $cpuconf = PVE::JSONSchema::parse_property_string($cpu_fmt, $cputype)
3055 or die "Cannot parse cpu description: $cputype\n";
3056 $cpu = $cpuconf->{cputype};
3057 $kvm_off = 1 if $cpuconf->{hidden};
3058 }
3059
3060 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
3061
3062 push @$cpuFlags , '-x2apic'
3063 if $conf->{ostype} && $conf->{ostype} eq 'solaris';
3064
3065 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
3066
3067 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
3068
3069 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3070
3071 push @$cpuFlags , '+kvm_pv_unhalt' if !$nokvm;
3072 push @$cpuFlags , '+kvm_pv_eoi' if !$nokvm;
3073 }
3074
3075 push @$cpuFlags, 'enforce' if $cpu ne 'host' && !$nokvm;
3076
3077 push @$cpuFlags, 'kvm=off' if $kvm_off;
3078
3079 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
3080
3081 push @$cmd, '-cpu', $cpu;
3082
3083 my $memory = $conf->{memory} || $defaults->{memory};
3084 my $static_memory = 0;
3085 my $dimm_memory = 0;
3086
3087 if ($hotplug_features->{memory}) {
3088 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
3089 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
3090 $static_memory = $STATICMEM;
3091 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
3092 $dimm_memory = $memory - $static_memory;
3093 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
3094
3095 } else {
3096
3097 $static_memory = $memory;
3098 push @$cmd, '-m', $static_memory;
3099 }
3100
3101 if ($conf->{numa}) {
3102
3103 my $numa_totalmemory = undef;
3104 for (my $i = 0; $i < $MAX_NUMA; $i++) {
3105 next if !$conf->{"numa$i"};
3106 my $numa = parse_numa($conf->{"numa$i"});
3107 next if !$numa;
3108 # memory
3109 die "missing numa node$i memory value\n" if !$numa->{memory};
3110 my $numa_memory = $numa->{memory};
3111 $numa_totalmemory += $numa_memory;
3112 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
3113
3114 # cpus
3115 my $cpulists = $numa->{cpus};
3116 die "missing numa node$i cpus\n" if !defined($cpulists);
3117 my $cpus = join(',', map {
3118 my ($start, $end) = @$_;
3119 defined($end) ? "$start-$end" : $start
3120 } @$cpulists);
3121
3122 # hostnodes
3123 my $hostnodelists = $numa->{hostnodes};
3124 if (defined($hostnodelists)) {
3125 my $hostnodes;
3126 foreach my $hostnoderange (@$hostnodelists) {
3127 my ($start, $end) = @$hostnoderange;
3128 $hostnodes .= ',' if $hostnodes;
3129 $hostnodes .= $start;
3130 $hostnodes .= "-$end" if defined($end);
3131 $end //= $start;
3132 for (my $i = $start; $i <= $end; ++$i ) {
3133 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
3134 }
3135 }
3136
3137 # policy
3138 my $policy = $numa->{policy};
3139 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
3140 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
3141 }
3142
3143 push @$cmd, '-object', $numa_object;
3144 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
3145 }
3146
3147 die "total memory for NUMA nodes must be equal to vm static memory\n"
3148 if $numa_totalmemory && $numa_totalmemory != $static_memory;
3149
3150 #if no custom tology, we split memory and cores across numa nodes
3151 if(!$numa_totalmemory) {
3152
3153 my $numa_memory = ($static_memory / $sockets) . "M";
3154
3155 for (my $i = 0; $i < $sockets; $i++) {
3156
3157 my $cpustart = ($cores * $i);
3158 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
3159 my $cpus = $cpustart;
3160 $cpus .= "-$cpuend" if $cpuend;
3161
3162 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
3163 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
3164 }
3165 }
3166 }
3167
3168 if ($hotplug_features->{memory}) {
3169 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
3170 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3171 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3172 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3173
3174 #if dimm_memory is not aligned to dimm map
3175 if($current_size > $memory) {
3176 $conf->{memory} = $current_size;
3177 PVE::QemuConfig->write_config($vmid, $conf);
3178 }
3179 });
3180 }
3181
3182 push @$cmd, '-S' if $conf->{freeze};
3183
3184 # set keyboard layout
3185 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3186 push @$cmd, '-k', $kb if $kb;
3187
3188 # enable sound
3189 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3190 #push @$cmd, '-soundhw', 'es1370';
3191 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3192
3193 if($conf->{agent}) {
3194 my $qgasocket = qmp_socket($vmid, 1);
3195 my $pciaddr = print_pci_addr("qga0", $bridges);
3196 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3197 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3198 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3199 }
3200
3201 my $spice_port;
3202
3203 if ($qxlnum) {
3204 if ($qxlnum > 1) {
3205 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3206 for(my $i = 1; $i < $qxlnum; $i++){
3207 my $pciaddr = print_pci_addr("vga$i", $bridges);
3208 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3209 }
3210 } else {
3211 # assume other OS works like Linux
3212 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3213 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3214 }
3215 }
3216
3217 my $pciaddr = print_pci_addr("spice", $bridges);
3218
3219 my $nodename = PVE::INotify::nodename();
3220 my $pfamily = PVE::Tools::get_host_address_family($nodename);
3221 $spice_port = PVE::Tools::next_spice_port($pfamily);
3222
3223 push @$devices, '-spice', "tls-port=${spice_port},addr=localhost,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3224
3225 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3226 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3227 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3228 }
3229
3230 # enable balloon by default, unless explicitly disabled
3231 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3232 $pciaddr = print_pci_addr("balloon0", $bridges);
3233 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3234 }
3235
3236 if ($conf->{watchdog}) {
3237 my $wdopts = parse_watchdog($conf->{watchdog});
3238 $pciaddr = print_pci_addr("watchdog", $bridges);
3239 my $watchdog = $wdopts->{model} || 'i6300esb';
3240 push @$devices, '-device', "$watchdog$pciaddr";
3241 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3242 }
3243
3244 my $vollist = [];
3245 my $scsicontroller = {};
3246 my $ahcicontroller = {};
3247 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3248
3249 # Add iscsi initiator name if available
3250 if (my $initiator = get_initiator_name()) {
3251 push @$devices, '-iscsi', "initiator-name=$initiator";
3252 }
3253
3254 foreach_drive($conf, sub {
3255 my ($ds, $drive) = @_;
3256
3257 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3258 push @$vollist, $drive->{file};
3259 }
3260
3261 $use_virtio = 1 if $ds =~ m/^virtio/;
3262
3263 if (drive_is_cdrom ($drive)) {
3264 if ($bootindex_hash->{d}) {
3265 $drive->{bootindex} = $bootindex_hash->{d};
3266 $bootindex_hash->{d} += 1;
3267 }
3268 } else {
3269 if ($bootindex_hash->{c}) {
3270 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3271 $bootindex_hash->{c} += 1;
3272 }
3273 }
3274
3275 if($drive->{interface} eq 'virtio'){
3276 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3277 }
3278
3279 if ($drive->{interface} eq 'scsi') {
3280
3281 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3282
3283 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3284 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3285
3286 my $iothread = '';
3287 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3288 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3289 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3290 }
3291
3292 my $queues = '';
3293 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3294 $queues = ",num_queues=$drive->{queues}";
3295 }
3296
3297 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3298 $scsicontroller->{$controller}=1;
3299 }
3300
3301 if ($drive->{interface} eq 'sata') {
3302 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3303 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3304 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3305 $ahcicontroller->{$controller}=1;
3306 }
3307
3308 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3309 push @$devices, '-drive',$drive_cmd;
3310 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3311 });
3312
3313 for (my $i = 0; $i < $MAX_NETS; $i++) {
3314 next if !$conf->{"net$i"};
3315 my $d = parse_net($conf->{"net$i"});
3316 next if !$d;
3317
3318 $use_virtio = 1 if $d->{model} eq 'virtio';
3319
3320 if ($bootindex_hash->{n}) {
3321 $d->{bootindex} = $bootindex_hash->{n};
3322 $bootindex_hash->{n} += 1;
3323 }
3324
3325 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3326 push @$devices, '-netdev', $netdevfull;
3327
3328 my $netdevicefull = print_netdevice_full($vmid, $conf, $d, "net$i", $bridges, $use_old_bios_files);
3329 push @$devices, '-device', $netdevicefull;
3330 }
3331
3332 if (!$q35) {
3333 # add pci bridges
3334 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3335 $bridges->{1} = 1;
3336 $bridges->{2} = 1;
3337 }
3338
3339 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3340
3341 while (my ($k, $v) = each %$bridges) {
3342 $pciaddr = print_pci_addr("pci.$k");
3343 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3344 }
3345 }
3346
3347 # add custom args
3348 if ($conf->{args}) {
3349 my $aa = PVE::Tools::split_args($conf->{args});
3350 push @$cmd, @$aa;
3351 }
3352
3353 push @$cmd, @$devices;
3354 push @$cmd, '-rtc', join(',', @$rtcFlags)
3355 if scalar(@$rtcFlags);
3356 push @$cmd, '-machine', join(',', @$machineFlags)
3357 if scalar(@$machineFlags);
3358 push @$cmd, '-global', join(',', @$globalFlags)
3359 if scalar(@$globalFlags);
3360
3361 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3362 }
3363
3364 sub vnc_socket {
3365 my ($vmid) = @_;
3366 return "${var_run_tmpdir}/$vmid.vnc";
3367 }
3368
3369 sub spice_port {
3370 my ($vmid) = @_;
3371
3372 my $res = vm_mon_cmd($vmid, 'query-spice');
3373
3374 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3375 }
3376
3377 sub qmp_socket {
3378 my ($vmid, $qga) = @_;
3379 my $sockettype = $qga ? 'qga' : 'qmp';
3380 return "${var_run_tmpdir}/$vmid.$sockettype";
3381 }
3382
3383 sub pidfile_name {
3384 my ($vmid) = @_;
3385 return "${var_run_tmpdir}/$vmid.pid";
3386 }
3387
3388 sub vm_devices_list {
3389 my ($vmid) = @_;
3390
3391 my $res = vm_mon_cmd($vmid, 'query-pci');
3392 my $devices = {};
3393 foreach my $pcibus (@$res) {
3394 foreach my $device (@{$pcibus->{devices}}) {
3395 next if !$device->{'qdev_id'};
3396 if ($device->{'pci_bridge'}) {
3397 $devices->{$device->{'qdev_id'}} = 1;
3398 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3399 next if !$bridge_device->{'qdev_id'};
3400 $devices->{$bridge_device->{'qdev_id'}} = 1;
3401 $devices->{$device->{'qdev_id'}}++;
3402 }
3403 } else {
3404 $devices->{$device->{'qdev_id'}} = 1;
3405 }
3406 }
3407 }
3408
3409 my $resblock = vm_mon_cmd($vmid, 'query-block');
3410 foreach my $block (@$resblock) {
3411 if($block->{device} =~ m/^drive-(\S+)/){
3412 $devices->{$1} = 1;
3413 }
3414 }
3415
3416 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3417 foreach my $mice (@$resmice) {
3418 if ($mice->{name} eq 'QEMU HID Tablet') {
3419 $devices->{tablet} = 1;
3420 last;
3421 }
3422 }
3423
3424 return $devices;
3425 }
3426
3427 sub vm_deviceplug {
3428 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3429
3430 my $q35 = machine_type_is_q35($conf);
3431
3432 my $devices_list = vm_devices_list($vmid);
3433 return 1 if defined($devices_list->{$deviceid});
3434
3435 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3436
3437 if ($deviceid eq 'tablet') {
3438
3439 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3440
3441 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3442
3443 qemu_iothread_add($vmid, $deviceid, $device);
3444
3445 qemu_driveadd($storecfg, $vmid, $device);
3446 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3447
3448 qemu_deviceadd($vmid, $devicefull);
3449 eval { qemu_deviceaddverify($vmid, $deviceid); };
3450 if (my $err = $@) {
3451 eval { qemu_drivedel($vmid, $deviceid); };
3452 warn $@ if $@;
3453 die $err;
3454 }
3455
3456 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3457
3458
3459 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3460 my $pciaddr = print_pci_addr($deviceid);
3461 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3462
3463 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3464
3465 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3466 qemu_iothread_add($vmid, $deviceid, $device);
3467 $devicefull .= ",iothread=iothread-$deviceid";
3468 }
3469
3470 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3471 $devicefull .= ",num_queues=$device->{queues}";
3472 }
3473
3474 qemu_deviceadd($vmid, $devicefull);
3475 qemu_deviceaddverify($vmid, $deviceid);
3476
3477 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3478
3479 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3480 qemu_driveadd($storecfg, $vmid, $device);
3481
3482 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3483 eval { qemu_deviceadd($vmid, $devicefull); };
3484 if (my $err = $@) {
3485 eval { qemu_drivedel($vmid, $deviceid); };
3486 warn $@ if $@;
3487 die $err;
3488 }
3489
3490 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3491
3492 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3493
3494 my $machine_type = PVE::QemuServer::qemu_machine_pxe($vmid, $conf);
3495 my $use_old_bios_files = undef;
3496 ($use_old_bios_files, $machine_type) = qemu_use_old_bios_files($machine_type);
3497
3498 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid, undef, $use_old_bios_files);
3499 qemu_deviceadd($vmid, $netdevicefull);
3500 eval { qemu_deviceaddverify($vmid, $deviceid); };
3501 if (my $err = $@) {
3502 eval { qemu_netdevdel($vmid, $deviceid); };
3503 warn $@ if $@;
3504 die $err;
3505 }
3506
3507 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3508
3509 my $bridgeid = $2;
3510 my $pciaddr = print_pci_addr($deviceid);
3511 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3512
3513 qemu_deviceadd($vmid, $devicefull);
3514 qemu_deviceaddverify($vmid, $deviceid);
3515
3516 } else {
3517 die "can't hotplug device '$deviceid'\n";
3518 }
3519
3520 return 1;
3521 }
3522
3523 # fixme: this should raise exceptions on error!
3524 sub vm_deviceunplug {
3525 my ($vmid, $conf, $deviceid) = @_;
3526
3527 my $devices_list = vm_devices_list($vmid);
3528 return 1 if !defined($devices_list->{$deviceid});
3529
3530 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3531
3532 if ($deviceid eq 'tablet') {
3533
3534 qemu_devicedel($vmid, $deviceid);
3535
3536 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3537
3538 qemu_devicedel($vmid, $deviceid);
3539 qemu_devicedelverify($vmid, $deviceid);
3540 qemu_drivedel($vmid, $deviceid);
3541 qemu_iothread_del($conf, $vmid, $deviceid);
3542
3543 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3544
3545 qemu_devicedel($vmid, $deviceid);
3546 qemu_devicedelverify($vmid, $deviceid);
3547 qemu_iothread_del($conf, $vmid, $deviceid);
3548
3549 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3550
3551 #qemu 2.3 segfault on drive_del with virtioscsi + iothread
3552 my $device = parse_drive($deviceid, $conf->{$deviceid});
3553 die "virtioscsi with iothread is not hot-unplugglable currently" if $device->{iothread};
3554
3555 qemu_devicedel($vmid, $deviceid);
3556 qemu_drivedel($vmid, $deviceid);
3557 qemu_deletescsihw($conf, $vmid, $deviceid);
3558
3559 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3560
3561 qemu_devicedel($vmid, $deviceid);
3562 qemu_devicedelverify($vmid, $deviceid);
3563 qemu_netdevdel($vmid, $deviceid);
3564
3565 } else {
3566 die "can't unplug device '$deviceid'\n";
3567 }
3568
3569 return 1;
3570 }
3571
3572 sub qemu_deviceadd {
3573 my ($vmid, $devicefull) = @_;
3574
3575 $devicefull = "driver=".$devicefull;
3576 my %options = split(/[=,]/, $devicefull);
3577
3578 vm_mon_cmd($vmid, "device_add" , %options);
3579 }
3580
3581 sub qemu_devicedel {
3582 my ($vmid, $deviceid) = @_;
3583
3584 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3585 }
3586
3587 sub qemu_iothread_add {
3588 my($vmid, $deviceid, $device) = @_;
3589
3590 if ($device->{iothread}) {
3591 my $iothreads = vm_iothreads_list($vmid);
3592 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3593 }
3594 }
3595
3596 sub qemu_iothread_del {
3597 my($conf, $vmid, $deviceid) = @_;
3598
3599 my $device = parse_drive($deviceid, $conf->{$deviceid});
3600 if ($device->{iothread}) {
3601 my $iothreads = vm_iothreads_list($vmid);
3602 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3603 }
3604 }
3605
3606 sub qemu_objectadd {
3607 my($vmid, $objectid, $qomtype) = @_;
3608
3609 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3610
3611 return 1;
3612 }
3613
3614 sub qemu_objectdel {
3615 my($vmid, $objectid) = @_;
3616
3617 vm_mon_cmd($vmid, "object-del", id => $objectid);
3618
3619 return 1;
3620 }
3621
3622 sub qemu_driveadd {
3623 my ($storecfg, $vmid, $device) = @_;
3624
3625 my $drive = print_drive_full($storecfg, $vmid, $device);
3626 $drive =~ s/\\/\\\\/g;
3627 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3628
3629 # If the command succeeds qemu prints: "OK"
3630 return 1 if $ret =~ m/OK/s;
3631
3632 die "adding drive failed: $ret\n";
3633 }
3634
3635 sub qemu_drivedel {
3636 my($vmid, $deviceid) = @_;
3637
3638 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3639 $ret =~ s/^\s+//;
3640
3641 return 1 if $ret eq "";
3642
3643 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3644 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3645
3646 die "deleting drive $deviceid failed : $ret\n";
3647 }
3648
3649 sub qemu_deviceaddverify {
3650 my ($vmid, $deviceid) = @_;
3651
3652 for (my $i = 0; $i <= 5; $i++) {
3653 my $devices_list = vm_devices_list($vmid);
3654 return 1 if defined($devices_list->{$deviceid});
3655 sleep 1;
3656 }
3657
3658 die "error on hotplug device '$deviceid'\n";
3659 }
3660
3661
3662 sub qemu_devicedelverify {
3663 my ($vmid, $deviceid) = @_;
3664
3665 # need to verify that the device is correctly removed as device_del
3666 # is async and empty return is not reliable
3667
3668 for (my $i = 0; $i <= 5; $i++) {
3669 my $devices_list = vm_devices_list($vmid);
3670 return 1 if !defined($devices_list->{$deviceid});
3671 sleep 1;
3672 }
3673
3674 die "error on hot-unplugging device '$deviceid'\n";
3675 }
3676
3677 sub qemu_findorcreatescsihw {
3678 my ($storecfg, $conf, $vmid, $device) = @_;
3679
3680 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3681
3682 my $scsihwid="$controller_prefix$controller";
3683 my $devices_list = vm_devices_list($vmid);
3684
3685 if(!defined($devices_list->{$scsihwid})) {
3686 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
3687 }
3688
3689 return 1;
3690 }
3691
3692 sub qemu_deletescsihw {
3693 my ($conf, $vmid, $opt) = @_;
3694
3695 my $device = parse_drive($opt, $conf->{$opt});
3696
3697 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
3698 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3699 return 1;
3700 }
3701
3702 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3703
3704 my $devices_list = vm_devices_list($vmid);
3705 foreach my $opt (keys %{$devices_list}) {
3706 if (PVE::QemuServer::is_valid_drivename($opt)) {
3707 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
3708 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
3709 return 1;
3710 }
3711 }
3712 }
3713
3714 my $scsihwid="scsihw$controller";
3715
3716 vm_deviceunplug($vmid, $conf, $scsihwid);
3717
3718 return 1;
3719 }
3720
3721 sub qemu_add_pci_bridge {
3722 my ($storecfg, $conf, $vmid, $device) = @_;
3723
3724 my $bridges = {};
3725
3726 my $bridgeid;
3727
3728 print_pci_addr($device, $bridges);
3729
3730 while (my ($k, $v) = each %$bridges) {
3731 $bridgeid = $k;
3732 }
3733 return 1 if !defined($bridgeid) || $bridgeid < 1;
3734
3735 my $bridge = "pci.$bridgeid";
3736 my $devices_list = vm_devices_list($vmid);
3737
3738 if (!defined($devices_list->{$bridge})) {
3739 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3740 }
3741
3742 return 1;
3743 }
3744
3745 sub qemu_set_link_status {
3746 my ($vmid, $device, $up) = @_;
3747
3748 vm_mon_cmd($vmid, "set_link", name => $device,
3749 up => $up ? JSON::true : JSON::false);
3750 }
3751
3752 sub qemu_netdevadd {
3753 my ($vmid, $conf, $device, $deviceid) = @_;
3754
3755 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid, 1);
3756 my %options = split(/[=,]/, $netdev);
3757
3758 vm_mon_cmd($vmid, "netdev_add", %options);
3759 return 1;
3760 }
3761
3762 sub qemu_netdevdel {
3763 my ($vmid, $deviceid) = @_;
3764
3765 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3766 }
3767
3768 sub qemu_cpu_hotplug {
3769 my ($vmid, $conf, $vcpus) = @_;
3770
3771 my $sockets = 1;
3772 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3773 $sockets = $conf->{sockets} if $conf->{sockets};
3774 my $cores = $conf->{cores} || 1;
3775 my $maxcpus = $sockets * $cores;
3776
3777 $vcpus = $maxcpus if !$vcpus;
3778
3779 die "you can't add more vcpus than maxcpus\n"
3780 if $vcpus > $maxcpus;
3781
3782 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3783 die "online cpu unplug is not yet possible\n"
3784 if $vcpus < $currentvcpus;
3785
3786 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3787 die "vcpus in running vm is different than configuration\n"
3788 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3789
3790 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3791 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3792 }
3793 }
3794
3795 sub qemu_memory_hotplug {
3796 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3797
3798 return $value if !check_running($vmid);
3799
3800 my $memory = $conf->{memory} || $defaults->{memory};
3801 $value = $defaults->{memory} if !$value;
3802 return $value if $value == $memory;
3803
3804 my $static_memory = $STATICMEM;
3805 my $dimm_memory = $memory - $static_memory;
3806
3807 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3808 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3809
3810
3811 my $sockets = 1;
3812 $sockets = $conf->{sockets} if $conf->{sockets};
3813
3814 if($value > $memory) {
3815
3816 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3817 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3818
3819 return if $current_size <= $conf->{memory};
3820
3821 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3822 if (my $err = $@) {
3823 eval { qemu_objectdel($vmid, "mem-$name"); };
3824 die $err;
3825 }
3826
3827 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3828 if (my $err = $@) {
3829 eval { qemu_objectdel($vmid, "mem-$name"); };
3830 die $err;
3831 }
3832 #update conf after each succesful module hotplug
3833 $conf->{memory} = $current_size;
3834 PVE::QemuConfig->write_config($vmid, $conf);
3835 });
3836
3837 } else {
3838
3839 foreach_reverse_dimm($conf, $vmid, $value, $sockets, sub {
3840 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3841
3842 return if $current_size >= $conf->{memory};
3843 print "try to unplug memory dimm $name\n";
3844
3845 my $retry = 0;
3846 while (1) {
3847 eval { qemu_devicedel($vmid, $name) };
3848 sleep 3;
3849 my $dimm_list = qemu_dimm_list($vmid);
3850 last if !$dimm_list->{$name};
3851 raise_param_exc({ $name => "error unplug memory module" }) if $retry > 5;
3852 $retry++;
3853 }
3854
3855 #update conf after each succesful module unplug
3856 $conf->{memory} = $current_size;
3857
3858 eval { qemu_objectdel($vmid, "mem-$name"); };
3859 PVE::QemuConfig->write_config($vmid, $conf);
3860 });
3861 }
3862 }
3863
3864 sub qemu_dimm_list {
3865 my ($vmid) = @_;
3866
3867 my $dimmarray = vm_mon_cmd_nocheck($vmid, "query-memory-devices");
3868 my $dimms = {};
3869
3870 foreach my $dimm (@$dimmarray) {
3871
3872 $dimms->{$dimm->{data}->{id}}->{id} = $dimm->{data}->{id};
3873 $dimms->{$dimm->{data}->{id}}->{node} = $dimm->{data}->{node};
3874 $dimms->{$dimm->{data}->{id}}->{addr} = $dimm->{data}->{addr};
3875 $dimms->{$dimm->{data}->{id}}->{size} = $dimm->{data}->{size};
3876 $dimms->{$dimm->{data}->{id}}->{slot} = $dimm->{data}->{slot};
3877 }
3878 return $dimms;
3879 }
3880
3881 sub qemu_block_set_io_throttle {
3882 my ($vmid, $deviceid,
3883 $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr,
3884 $bps_max, $bps_rd_max, $bps_wr_max, $iops_max, $iops_rd_max, $iops_wr_max) = @_;
3885
3886 return if !check_running($vmid) ;
3887
3888 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid,
3889 bps => int($bps),
3890 bps_rd => int($bps_rd),
3891 bps_wr => int($bps_wr),
3892 iops => int($iops),
3893 iops_rd => int($iops_rd),
3894 iops_wr => int($iops_wr),
3895 bps_max => int($bps_max),
3896 bps_rd_max => int($bps_rd_max),
3897 bps_wr_max => int($bps_wr_max),
3898 iops_max => int($iops_max),
3899 iops_rd_max => int($iops_rd_max),
3900 iops_wr_max => int($iops_wr_max)
3901 );
3902
3903 }
3904
3905 # old code, only used to shutdown old VM after update
3906 sub __read_avail {
3907 my ($fh, $timeout) = @_;
3908
3909 my $sel = new IO::Select;
3910 $sel->add($fh);
3911
3912 my $res = '';
3913 my $buf;
3914
3915 my @ready;
3916 while (scalar (@ready = $sel->can_read($timeout))) {
3917 my $count;
3918 if ($count = $fh->sysread($buf, 8192)) {
3919 if ($buf =~ /^(.*)\(qemu\) $/s) {
3920 $res .= $1;
3921 last;
3922 } else {
3923 $res .= $buf;
3924 }
3925 } else {
3926 if (!defined($count)) {
3927 die "$!\n";
3928 }
3929 last;
3930 }
3931 }
3932
3933 die "monitor read timeout\n" if !scalar(@ready);
3934
3935 return $res;
3936 }
3937
3938 # old code, only used to shutdown old VM after update
3939 sub vm_monitor_command {
3940 my ($vmid, $cmdstr, $nocheck) = @_;
3941
3942 my $res;
3943
3944 eval {
3945 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3946
3947 my $sname = "${var_run_tmpdir}/$vmid.mon";
3948
3949 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3950 die "unable to connect to VM $vmid socket - $!\n";
3951
3952 my $timeout = 3;
3953
3954 # hack: migrate sometime blocks the monitor (when migrate_downtime
3955 # is set)
3956 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3957 $timeout = 60*60; # 1 hour
3958 }
3959
3960 # read banner;
3961 my $data = __read_avail($sock, $timeout);
3962
3963 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3964 die "got unexpected qemu monitor banner\n";
3965 }
3966
3967 my $sel = new IO::Select;
3968 $sel->add($sock);
3969
3970 if (!scalar(my @ready = $sel->can_write($timeout))) {
3971 die "monitor write error - timeout";
3972 }
3973
3974 my $fullcmd = "$cmdstr\r";
3975
3976 # syslog('info', "VM $vmid monitor command: $cmdstr");
3977
3978 my $b;
3979 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3980 die "monitor write error - $!";
3981 }
3982
3983 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3984
3985 $timeout = 20;
3986
3987 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3988 $timeout = 60*60; # 1 hour
3989 } elsif ($cmdstr =~ m/^(eject|change)/) {
3990 $timeout = 60; # note: cdrom mount command is slow
3991 }
3992 if ($res = __read_avail($sock, $timeout)) {
3993
3994 my @lines = split("\r?\n", $res);
3995
3996 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3997
3998 $res = join("\n", @lines);
3999 $res .= "\n";
4000 }
4001 };
4002
4003 my $err = $@;
4004
4005 if ($err) {
4006 syslog("err", "VM $vmid monitor command failed - $err");
4007 die $err;
4008 }
4009
4010 return $res;
4011 }
4012
4013 sub qemu_block_resize {
4014 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
4015
4016 my $running = check_running($vmid);
4017
4018 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
4019
4020 return if !$running;
4021
4022 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
4023
4024 }
4025
4026 sub qemu_volume_snapshot {
4027 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4028
4029 my $running = check_running($vmid);
4030
4031 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
4032 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
4033 } else {
4034 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
4035 }
4036 }
4037
4038 sub qemu_volume_snapshot_delete {
4039 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
4040
4041 my $running = check_running($vmid);
4042
4043 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
4044
4045 return if !$running;
4046
4047 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
4048 }
4049
4050 sub set_migration_caps {
4051 my ($vmid) = @_;
4052
4053 my $cap_ref = [];
4054
4055 my $enabled_cap = {
4056 "auto-converge" => 1,
4057 "xbzrle" => 1,
4058 "x-rdma-pin-all" => 0,
4059 "zero-blocks" => 0,
4060 "compress" => 0
4061 };
4062
4063 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
4064
4065 for my $supported_capability (@$supported_capabilities) {
4066 push @$cap_ref, {
4067 capability => $supported_capability->{capability},
4068 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
4069 };
4070 }
4071
4072 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
4073 }
4074
4075 my $fast_plug_option = {
4076 'lock' => 1,
4077 'name' => 1,
4078 'onboot' => 1,
4079 'shares' => 1,
4080 'startup' => 1,
4081 'description' => 1,
4082 };
4083
4084 # hotplug changes in [PENDING]
4085 # $selection hash can be used to only apply specified options, for
4086 # example: { cores => 1 } (only apply changed 'cores')
4087 # $errors ref is used to return error messages
4088 sub vmconfig_hotplug_pending {
4089 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
4090
4091 my $defaults = load_defaults();
4092
4093 # commit values which do not have any impact on running VM first
4094 # Note: those option cannot raise errors, we we do not care about
4095 # $selection and always apply them.
4096
4097 my $add_error = sub {
4098 my ($opt, $msg) = @_;
4099 $errors->{$opt} = "hotplug problem - $msg";
4100 };
4101
4102 my $changes = 0;
4103 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4104 if ($fast_plug_option->{$opt}) {
4105 $conf->{$opt} = $conf->{pending}->{$opt};
4106 delete $conf->{pending}->{$opt};
4107 $changes = 1;
4108 }
4109 }
4110
4111 if ($changes) {
4112 PVE::QemuConfig->write_config($vmid, $conf);
4113 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4114 }
4115
4116 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
4117
4118 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4119 while (my ($opt, $force) = each %$pending_delete_hash) {
4120 next if $selection && !$selection->{$opt};
4121 eval {
4122 if ($opt eq 'hotplug') {
4123 die "skip\n" if ($conf->{hotplug} =~ /memory/);
4124 } elsif ($opt eq 'tablet') {
4125 die "skip\n" if !$hotplug_features->{usb};
4126 if ($defaults->{tablet}) {
4127 vm_deviceplug($storecfg, $conf, $vmid, $opt);
4128 } else {
4129 vm_deviceunplug($vmid, $conf, $opt);
4130 }
4131 } elsif ($opt eq 'vcpus') {
4132 die "skip\n" if !$hotplug_features->{cpu};
4133 qemu_cpu_hotplug($vmid, $conf, undef);
4134 } elsif ($opt eq 'balloon') {
4135 # enable balloon device is not hotpluggable
4136 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
4137 } elsif ($fast_plug_option->{$opt}) {
4138 # do nothing
4139 } elsif ($opt =~ m/^net(\d+)$/) {
4140 die "skip\n" if !$hotplug_features->{network};
4141 vm_deviceunplug($vmid, $conf, $opt);
4142 } elsif (is_valid_drivename($opt)) {
4143 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
4144 vm_deviceunplug($vmid, $conf, $opt);
4145 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4146 } elsif ($opt =~ m/^memory$/) {
4147 die "skip\n" if !$hotplug_features->{memory};
4148 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
4149 } elsif ($opt eq 'cpuunits') {
4150 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
4151 } elsif ($opt eq 'cpulimit') {
4152 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
4153 } else {
4154 die "skip\n";
4155 }
4156 };
4157 if (my $err = $@) {
4158 &$add_error($opt, $err) if $err ne "skip\n";
4159 } else {
4160 # save new config if hotplug was successful
4161 delete $conf->{$opt};
4162 vmconfig_undelete_pending_option($conf, $opt);
4163 PVE::QemuConfig->write_config($vmid, $conf);
4164 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4165 }
4166 }
4167
4168 foreach my $opt (keys %{$conf->{pending}}) {
4169 next if $selection && !$selection->{$opt};
4170 my $value = $conf->{pending}->{$opt};
4171 eval {
4172 if ($opt eq 'hotplug') {
4173 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
4174 } elsif ($opt eq 'tablet') {
4175 die "skip\n" if !$hotplug_features->{usb};
4176 if ($value == 1) {
4177 vm_deviceplug($storecfg, $conf, $vmid, $opt);
4178 } elsif ($value == 0) {
4179 vm_deviceunplug($vmid, $conf, $opt);
4180 }
4181 } elsif ($opt eq 'vcpus') {
4182 die "skip\n" if !$hotplug_features->{cpu};
4183 qemu_cpu_hotplug($vmid, $conf, $value);
4184 } elsif ($opt eq 'balloon') {
4185 # enable/disable balloning device is not hotpluggable
4186 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
4187 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
4188 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
4189
4190 # allow manual ballooning if shares is set to zero
4191 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
4192 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
4193 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
4194 }
4195 } elsif ($opt =~ m/^net(\d+)$/) {
4196 # some changes can be done without hotplug
4197 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
4198 $vmid, $opt, $value);
4199 } elsif (is_valid_drivename($opt)) {
4200 # some changes can be done without hotplug
4201 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
4202 $vmid, $opt, $value, 1);
4203 } elsif ($opt =~ m/^memory$/) { #dimms
4204 die "skip\n" if !$hotplug_features->{memory};
4205 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
4206 } elsif ($opt eq 'cpuunits') {
4207 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
4208 } elsif ($opt eq 'cpulimit') {
4209 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
4210 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
4211 } else {
4212 die "skip\n"; # skip non-hot-pluggable options
4213 }
4214 };
4215 if (my $err = $@) {
4216 &$add_error($opt, $err) if $err ne "skip\n";
4217 } else {
4218 # save new config if hotplug was successful
4219 $conf->{$opt} = $value;
4220 delete $conf->{pending}->{$opt};
4221 PVE::QemuConfig->write_config($vmid, $conf);
4222 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4223 }
4224 }
4225 }
4226
4227 sub try_deallocate_drive {
4228 my ($storecfg, $vmid, $conf, $key, $drive, $rpcenv, $authuser, $force) = @_;
4229
4230 if (($force || $key =~ /^unused/) && !drive_is_cdrom($drive, 1)) {
4231 my $volid = $drive->{file};
4232 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
4233 my $sid = PVE::Storage::parse_volume_id($volid);
4234 $rpcenv->check($authuser, "/storage/$sid", ['Datastore.AllocateSpace']);
4235
4236 # check if the disk is really unused
4237 die "unable to delete '$volid' - volume is still in use (snapshot?)\n"
4238 if is_volume_in_use($storecfg, $conf, $key, $volid);
4239 PVE::Storage::vdisk_free($storecfg, $volid);
4240 return 1;
4241 } else {
4242 # If vm is not owner of this disk remove from config
4243 return 1;
4244 }
4245 }
4246
4247 return undef;
4248 }
4249
4250 sub vmconfig_delete_or_detach_drive {
4251 my ($vmid, $storecfg, $conf, $opt, $force) = @_;
4252
4253 my $drive = parse_drive($opt, $conf->{$opt});
4254
4255 my $rpcenv = PVE::RPCEnvironment::get();
4256 my $authuser = $rpcenv->get_user();
4257
4258 if ($force) {
4259 $rpcenv->check_vm_perm($authuser, $vmid, undef, ['VM.Config.Disk']);
4260 try_deallocate_drive($storecfg, $vmid, $conf, $opt, $drive, $rpcenv, $authuser, $force);
4261 } else {
4262 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $drive);
4263 }
4264 }
4265
4266 sub vmconfig_apply_pending {
4267 my ($vmid, $conf, $storecfg) = @_;
4268
4269 # cold plug
4270
4271 my $pending_delete_hash = split_flagged_list($conf->{pending}->{delete});
4272 while (my ($opt, $force) = each %$pending_delete_hash) {
4273 die "internal error" if $opt =~ m/^unused/;
4274 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4275 if (!defined($conf->{$opt})) {
4276 vmconfig_undelete_pending_option($conf, $opt);
4277 PVE::QemuConfig->write_config($vmid, $conf);
4278 } elsif (is_valid_drivename($opt)) {
4279 vmconfig_delete_or_detach_drive($vmid, $storecfg, $conf, $opt, $force);
4280 vmconfig_undelete_pending_option($conf, $opt);
4281 delete $conf->{$opt};
4282 PVE::QemuConfig->write_config($vmid, $conf);
4283 } else {
4284 vmconfig_undelete_pending_option($conf, $opt);
4285 delete $conf->{$opt};
4286 PVE::QemuConfig->write_config($vmid, $conf);
4287 }
4288 }
4289
4290 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4291
4292 foreach my $opt (keys %{$conf->{pending}}) { # add/change
4293 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4294
4295 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
4296 # skip if nothing changed
4297 } elsif (is_valid_drivename($opt)) {
4298 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4299 if defined($conf->{$opt});
4300 $conf->{$opt} = $conf->{pending}->{$opt};
4301 } else {
4302 $conf->{$opt} = $conf->{pending}->{$opt};
4303 }
4304
4305 delete $conf->{pending}->{$opt};
4306 PVE::QemuConfig->write_config($vmid, $conf);
4307 }
4308 }
4309
4310 my $safe_num_ne = sub {
4311 my ($a, $b) = @_;
4312
4313 return 0 if !defined($a) && !defined($b);
4314 return 1 if !defined($a);
4315 return 1 if !defined($b);
4316
4317 return $a != $b;
4318 };
4319
4320 my $safe_string_ne = sub {
4321 my ($a, $b) = @_;
4322
4323 return 0 if !defined($a) && !defined($b);
4324 return 1 if !defined($a);
4325 return 1 if !defined($b);
4326
4327 return $a ne $b;
4328 };
4329
4330 sub vmconfig_update_net {
4331 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4332
4333 my $newnet = parse_net($value);
4334
4335 if ($conf->{$opt}) {
4336 my $oldnet = parse_net($conf->{$opt});
4337
4338 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4339 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4340 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4341 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4342
4343 # for non online change, we try to hot-unplug
4344 die "skip\n" if !$hotplug;
4345 vm_deviceunplug($vmid, $conf, $opt);
4346 } else {
4347
4348 die "internal error" if $opt !~ m/net(\d+)/;
4349 my $iface = "tap${vmid}i$1";
4350
4351 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4352 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4353 &$safe_string_ne($oldnet->{trunks}, $newnet->{trunks}) ||
4354 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4355 PVE::Network::tap_unplug($iface);
4356 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
4357 } elsif (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4358 # Rate can be applied on its own but any change above needs to
4359 # include the rate in tap_plug since OVS resets everything.
4360 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4361 }
4362
4363 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4364 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4365 }
4366
4367 return 1;
4368 }
4369 }
4370
4371 if ($hotplug) {
4372 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4373 } else {
4374 die "skip\n";
4375 }
4376 }
4377
4378 sub vmconfig_update_disk {
4379 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4380
4381 # fixme: do we need force?
4382
4383 my $drive = parse_drive($opt, $value);
4384
4385 if ($conf->{$opt}) {
4386
4387 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4388
4389 my $media = $drive->{media} || 'disk';
4390 my $oldmedia = $old_drive->{media} || 'disk';
4391 die "unable to change media type\n" if $media ne $oldmedia;
4392
4393 if (!drive_is_cdrom($old_drive)) {
4394
4395 if ($drive->{file} ne $old_drive->{file}) {
4396
4397 die "skip\n" if !$hotplug;
4398
4399 # unplug and register as unused
4400 vm_deviceunplug($vmid, $conf, $opt);
4401 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4402
4403 } else {
4404 # update existing disk
4405
4406 # skip non hotpluggable value
4407 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4408 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4409 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4410 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4411 die "skip\n";
4412 }
4413
4414 # apply throttle
4415 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4416 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4417 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4418 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4419 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4420 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4421 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4422 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4423 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4424 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4425 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4426 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4427
4428 qemu_block_set_io_throttle($vmid,"drive-$opt",
4429 ($drive->{mbps} || 0)*1024*1024,
4430 ($drive->{mbps_rd} || 0)*1024*1024,
4431 ($drive->{mbps_wr} || 0)*1024*1024,
4432 $drive->{iops} || 0,
4433 $drive->{iops_rd} || 0,
4434 $drive->{iops_wr} || 0,
4435 ($drive->{mbps_max} || 0)*1024*1024,
4436 ($drive->{mbps_rd_max} || 0)*1024*1024,
4437 ($drive->{mbps_wr_max} || 0)*1024*1024,
4438 $drive->{iops_max} || 0,
4439 $drive->{iops_rd_max} || 0,
4440 $drive->{iops_wr_max} || 0);
4441
4442 }
4443
4444 return 1;
4445 }
4446
4447 } else { # cdrom
4448
4449 if ($drive->{file} eq 'none') {
4450 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4451 } else {
4452 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4453 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4454 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4455 }
4456
4457 return 1;
4458 }
4459 }
4460 }
4461
4462 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4463 # hotplug new disks
4464 PVE::Storage::activate_volumes($storecfg, [$drive->{file}]) if $drive->{file} !~ m|^/dev/.+|;
4465 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4466 }
4467
4468 sub vm_start {
4469 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused,
4470 $forcemachine, $spice_ticket) = @_;
4471
4472 PVE::QemuConfig->lock_config($vmid, sub {
4473 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
4474
4475 die "you can't start a vm if it's a template\n" if PVE::QemuConfig->is_template($conf);
4476
4477 PVE::QemuConfig->check_lock($conf) if !$skiplock;
4478
4479 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4480
4481 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4482 vmconfig_apply_pending($vmid, $conf, $storecfg);
4483 $conf = PVE::QemuConfig->load_config($vmid); # update/reload
4484 }
4485
4486 my $defaults = load_defaults();
4487
4488 # set environment variable useful inside network script
4489 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4490
4491 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4492
4493 my $migrate_port = 0;
4494 my $migrate_uri;
4495 if ($statefile) {
4496 if ($statefile eq 'tcp') {
4497 my $localip = "localhost";
4498 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4499 my $nodename = PVE::INotify::nodename();
4500 if ($datacenterconf->{migration_unsecure}) {
4501 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4502 $localip = "[$localip]" if Net::IP::ip_is_ipv6($localip);
4503 }
4504 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4505 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
4506 $migrate_uri = "tcp:${localip}:${migrate_port}";
4507 push @$cmd, '-incoming', $migrate_uri;
4508 push @$cmd, '-S';
4509 } else {
4510 push @$cmd, '-loadstate', $statefile;
4511 }
4512 } elsif ($paused) {
4513 push @$cmd, '-S';
4514 }
4515
4516 # host pci devices
4517 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4518 my $d = parse_hostpci($conf->{"hostpci$i"});
4519 next if !$d;
4520 my $pcidevices = $d->{pciid};
4521 foreach my $pcidevice (@$pcidevices) {
4522 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4523
4524 my $info = pci_device_info("0000:$pciid");
4525 die "IOMMU not present\n" if !check_iommu_support();
4526 die "no pci device info for device '$pciid'\n" if !$info;
4527 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4528 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4529 }
4530 }
4531
4532 PVE::Storage::activate_volumes($storecfg, $vollist);
4533
4534 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4535 umask => 0077); };
4536
4537 if (my $err = $@) {
4538 # deactivate volumes if start fails
4539 eval { PVE::Storage::deactivate_volumes($storecfg, $vollist); };
4540 die "start failed: $err";
4541 }
4542
4543 print "migration listens on $migrate_uri\n" if $migrate_uri;
4544
4545 if ($statefile && $statefile ne 'tcp') {
4546 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4547 warn $@ if $@;
4548 }
4549
4550 if ($migratedfrom) {
4551
4552 eval {
4553 set_migration_caps($vmid);
4554 };
4555 warn $@ if $@;
4556
4557 if ($spice_port) {
4558 print "spice listens on port $spice_port\n";
4559 if ($spice_ticket) {
4560 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4561 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4562 }
4563 }
4564
4565 } else {
4566
4567 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4568 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4569 if $conf->{balloon};
4570 }
4571
4572 foreach my $opt (keys %$conf) {
4573 next if $opt !~ m/^net\d+$/;
4574 my $nicconf = parse_net($conf->{$opt});
4575 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4576 }
4577 }
4578
4579 vm_mon_cmd_nocheck($vmid, 'qom-set',
4580 path => "machine/peripheral/balloon0",
4581 property => "guest-stats-polling-interval",
4582 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
4583
4584 });
4585 }
4586
4587 sub vm_mon_cmd {
4588 my ($vmid, $execute, %params) = @_;
4589
4590 my $cmd = { execute => $execute, arguments => \%params };
4591 vm_qmp_command($vmid, $cmd);
4592 }
4593
4594 sub vm_mon_cmd_nocheck {
4595 my ($vmid, $execute, %params) = @_;
4596
4597 my $cmd = { execute => $execute, arguments => \%params };
4598 vm_qmp_command($vmid, $cmd, 1);
4599 }
4600
4601 sub vm_qmp_command {
4602 my ($vmid, $cmd, $nocheck) = @_;
4603
4604 my $res;
4605
4606 my $timeout;
4607 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4608 $timeout = $cmd->{arguments}->{timeout};
4609 delete $cmd->{arguments}->{timeout};
4610 }
4611
4612 eval {
4613 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4614 my $sname = qmp_socket($vmid);
4615 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4616 my $qmpclient = PVE::QMPClient->new();
4617
4618 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4619 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4620 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4621 if scalar(%{$cmd->{arguments}});
4622 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4623 } else {
4624 die "unable to open monitor socket\n";
4625 }
4626 };
4627 if (my $err = $@) {
4628 syslog("err", "VM $vmid qmp command failed - $err");
4629 die $err;
4630 }
4631
4632 return $res;
4633 }
4634
4635 sub vm_human_monitor_command {
4636 my ($vmid, $cmdline) = @_;
4637
4638 my $res;
4639
4640 my $cmd = {
4641 execute => 'human-monitor-command',
4642 arguments => { 'command-line' => $cmdline},
4643 };
4644
4645 return vm_qmp_command($vmid, $cmd);
4646 }
4647
4648 sub vm_commandline {
4649 my ($storecfg, $vmid) = @_;
4650
4651 my $conf = PVE::QemuConfig->load_config($vmid);
4652
4653 my $defaults = load_defaults();
4654
4655 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4656
4657 return join(' ', @$cmd);
4658 }
4659
4660 sub vm_reset {
4661 my ($vmid, $skiplock) = @_;
4662
4663 PVE::QemuConfig->lock_config($vmid, sub {
4664
4665 my $conf = PVE::QemuConfig->load_config($vmid);
4666
4667 PVE::QemuConfig->check_lock($conf) if !$skiplock;
4668
4669 vm_mon_cmd($vmid, "system_reset");
4670 });
4671 }
4672
4673 sub get_vm_volumes {
4674 my ($conf) = @_;
4675
4676 my $vollist = [];
4677 foreach_volid($conf, sub {
4678 my ($volid, $is_cdrom) = @_;
4679
4680 return if $volid =~ m|^/|;
4681
4682 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4683 return if !$sid;
4684
4685 push @$vollist, $volid;
4686 });
4687
4688 return $vollist;
4689 }
4690
4691 sub vm_stop_cleanup {
4692 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4693
4694 eval {
4695
4696 if (!$keepActive) {
4697 my $vollist = get_vm_volumes($conf);
4698 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4699 }
4700
4701 foreach my $ext (qw(mon qmp pid vnc qga)) {
4702 unlink "/var/run/qemu-server/${vmid}.$ext";
4703 }
4704
4705 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4706 };
4707 warn $@ if $@; # avoid errors - just warn
4708 }
4709
4710 # Note: use $nockeck to skip tests if VM configuration file exists.
4711 # We need that when migration VMs to other nodes (files already moved)
4712 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4713 sub vm_stop {
4714 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4715
4716 $force = 1 if !defined($force) && !$shutdown;
4717
4718 if ($migratedfrom){
4719 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4720 kill 15, $pid if $pid;
4721 my $conf = PVE::QemuConfig->load_config($vmid, $migratedfrom);
4722 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4723 return;
4724 }
4725
4726 PVE::QemuConfig->lock_config($vmid, sub {
4727
4728 my $pid = check_running($vmid, $nocheck);
4729 return if !$pid;
4730
4731 my $conf;
4732 if (!$nocheck) {
4733 $conf = PVE::QemuConfig->load_config($vmid);
4734 PVE::QemuConfig->check_lock($conf) if !$skiplock;
4735 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4736 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
4737 $timeout = $opts->{down} if $opts->{down};
4738 }
4739 }
4740
4741 $timeout = 60 if !defined($timeout);
4742
4743 eval {
4744 if ($shutdown) {
4745 if (defined($conf) && $conf->{agent}) {
4746 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4747 } else {
4748 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4749 }
4750 } else {
4751 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4752 }
4753 };
4754 my $err = $@;
4755
4756 if (!$err) {
4757 my $count = 0;
4758 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4759 $count++;
4760 sleep 1;
4761 }
4762
4763 if ($count >= $timeout) {
4764 if ($force) {
4765 warn "VM still running - terminating now with SIGTERM\n";
4766 kill 15, $pid;
4767 } else {
4768 die "VM quit/powerdown failed - got timeout\n";
4769 }
4770 } else {
4771 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4772 return;
4773 }
4774 } else {
4775 if ($force) {
4776 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4777 kill 15, $pid;
4778 } else {
4779 die "VM quit/powerdown failed\n";
4780 }
4781 }
4782
4783 # wait again
4784 $timeout = 10;
4785
4786 my $count = 0;
4787 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4788 $count++;
4789 sleep 1;
4790 }
4791
4792 if ($count >= $timeout) {
4793 warn "VM still running - terminating now with SIGKILL\n";
4794 kill 9, $pid;
4795 sleep 1;
4796 }
4797
4798 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4799 });
4800 }
4801
4802 sub vm_suspend {
4803 my ($vmid, $skiplock) = @_;
4804
4805 PVE::QemuConfig->lock_config($vmid, sub {
4806
4807 my $conf = PVE::QemuConfig->load_config($vmid);
4808
4809 PVE::QemuConfig->check_lock($conf)
4810 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
4811
4812 vm_mon_cmd($vmid, "stop");
4813 });
4814 }
4815
4816 sub vm_resume {
4817 my ($vmid, $skiplock, $nocheck) = @_;
4818
4819 PVE::QemuConfig->lock_config($vmid, sub {
4820
4821 if (!$nocheck) {
4822
4823 my $conf = PVE::QemuConfig->load_config($vmid);
4824
4825 PVE::QemuConfig->check_lock($conf)
4826 if !($skiplock || PVE::QemuConfig->has_lock($conf, 'backup'));
4827
4828 vm_mon_cmd($vmid, "cont");
4829
4830 } else {
4831 vm_mon_cmd_nocheck($vmid, "cont");
4832 }
4833 });
4834 }
4835
4836 sub vm_sendkey {
4837 my ($vmid, $skiplock, $key) = @_;
4838
4839 PVE::QemuConfig->lock_config($vmid, sub {
4840
4841 my $conf = PVE::QemuConfig->load_config($vmid);
4842
4843 # there is no qmp command, so we use the human monitor command
4844 vm_human_monitor_command($vmid, "sendkey $key");
4845 });
4846 }
4847
4848 sub vm_destroy {
4849 my ($storecfg, $vmid, $skiplock) = @_;
4850
4851 PVE::QemuConfig->lock_config($vmid, sub {
4852
4853 my $conf = PVE::QemuConfig->load_config($vmid);
4854
4855 if (!check_running($vmid)) {
4856 destroy_vm($storecfg, $vmid, undef, $skiplock);
4857 } else {
4858 die "VM $vmid is running - destroy failed\n";
4859 }
4860 });
4861 }
4862
4863 # pci helpers
4864
4865 sub file_write {
4866 my ($filename, $buf) = @_;
4867
4868 my $fh = IO::File->new($filename, "w");
4869 return undef if !$fh;
4870
4871 my $res = print $fh $buf;
4872
4873 $fh->close();
4874
4875 return $res;
4876 }
4877
4878 sub pci_device_info {
4879 my ($name) = @_;
4880
4881 my $res;
4882
4883 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4884 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4885
4886 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4887 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4888
4889 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4890 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4891
4892 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4893 return undef if !defined($product) || $product !~ s/^0x//;
4894
4895 $res = {
4896 name => $name,
4897 vendor => $vendor,
4898 product => $product,
4899 domain => $domain,
4900 bus => $bus,
4901 slot => $slot,
4902 func => $func,
4903 irq => $irq,
4904 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4905 };
4906
4907 return $res;
4908 }
4909
4910 sub pci_dev_reset {
4911 my ($dev) = @_;
4912
4913 my $name = $dev->{name};
4914
4915 my $fn = "$pcisysfs/devices/$name/reset";
4916
4917 return file_write($fn, "1");
4918 }
4919
4920 sub pci_dev_bind_to_vfio {
4921 my ($dev) = @_;
4922
4923 my $name = $dev->{name};
4924
4925 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4926
4927 if (!-d $vfio_basedir) {
4928 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4929 }
4930 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4931
4932 my $testdir = "$vfio_basedir/$name";
4933 return 1 if -d $testdir;
4934
4935 my $data = "$dev->{vendor} $dev->{product}";
4936 return undef if !file_write("$vfio_basedir/new_id", $data);
4937
4938 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4939 if (!file_write($fn, $name)) {
4940 return undef if -f $fn;
4941 }
4942
4943 $fn = "$vfio_basedir/bind";
4944 if (! -d $testdir) {
4945 return undef if !file_write($fn, $name);
4946 }
4947
4948 return -d $testdir;
4949 }
4950
4951 sub pci_dev_group_bind_to_vfio {
4952 my ($pciid) = @_;
4953
4954 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4955
4956 if (!-d $vfio_basedir) {
4957 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4958 }
4959 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4960
4961 # get IOMMU group devices
4962 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4963 my @devs = grep /^0000:/, readdir($D);
4964 closedir($D);
4965
4966 foreach my $pciid (@devs) {
4967 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4968
4969 # pci bridges, switches or root ports are not supported
4970 # they have a pci_bus subdirectory so skip them
4971 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4972
4973 my $info = pci_device_info($1);
4974 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4975 }
4976
4977 return 1;
4978 }
4979
4980 sub print_pci_addr {
4981 my ($id, $bridges) = @_;
4982
4983 my $res = '';
4984 my $devices = {
4985 piix3 => { bus => 0, addr => 1 },
4986 #addr2 : first videocard
4987 balloon0 => { bus => 0, addr => 3 },
4988 watchdog => { bus => 0, addr => 4 },
4989 scsihw0 => { bus => 0, addr => 5 },
4990 'pci.3' => { bus => 0, addr => 5 }, #can also be used for virtio-scsi-single bridge
4991 scsihw1 => { bus => 0, addr => 6 },
4992 ahci0 => { bus => 0, addr => 7 },
4993 qga0 => { bus => 0, addr => 8 },
4994 spice => { bus => 0, addr => 9 },
4995 virtio0 => { bus => 0, addr => 10 },
4996 virtio1 => { bus => 0, addr => 11 },
4997 virtio2 => { bus => 0, addr => 12 },
4998 virtio3 => { bus => 0, addr => 13 },
4999 virtio4 => { bus => 0, addr => 14 },
5000 virtio5 => { bus => 0, addr => 15 },
5001 hostpci0 => { bus => 0, addr => 16 },
5002 hostpci1 => { bus => 0, addr => 17 },
5003 net0 => { bus => 0, addr => 18 },
5004 net1 => { bus => 0, addr => 19 },
5005 net2 => { bus => 0, addr => 20 },
5006 net3 => { bus => 0, addr => 21 },
5007 net4 => { bus => 0, addr => 22 },
5008 net5 => { bus => 0, addr => 23 },
5009 vga1 => { bus => 0, addr => 24 },
5010 vga2 => { bus => 0, addr => 25 },
5011 vga3 => { bus => 0, addr => 26 },
5012 hostpci2 => { bus => 0, addr => 27 },
5013 hostpci3 => { bus => 0, addr => 28 },
5014 #addr29 : usb-host (pve-usb.cfg)
5015 'pci.1' => { bus => 0, addr => 30 },
5016 'pci.2' => { bus => 0, addr => 31 },
5017 'net6' => { bus => 1, addr => 1 },
5018 'net7' => { bus => 1, addr => 2 },
5019 'net8' => { bus => 1, addr => 3 },
5020 'net9' => { bus => 1, addr => 4 },
5021 'net10' => { bus => 1, addr => 5 },
5022 'net11' => { bus => 1, addr => 6 },
5023 'net12' => { bus => 1, addr => 7 },
5024 'net13' => { bus => 1, addr => 8 },
5025 'net14' => { bus => 1, addr => 9 },
5026 'net15' => { bus => 1, addr => 10 },
5027 'net16' => { bus => 1, addr => 11 },
5028 'net17' => { bus => 1, addr => 12 },
5029 'net18' => { bus => 1, addr => 13 },
5030 'net19' => { bus => 1, addr => 14 },
5031 'net20' => { bus => 1, addr => 15 },
5032 'net21' => { bus => 1, addr => 16 },
5033 'net22' => { bus => 1, addr => 17 },
5034 'net23' => { bus => 1, addr => 18 },
5035 'net24' => { bus => 1, addr => 19 },
5036 'net25' => { bus => 1, addr => 20 },
5037 'net26' => { bus => 1, addr => 21 },
5038 'net27' => { bus => 1, addr => 22 },
5039 'net28' => { bus => 1, addr => 23 },
5040 'net29' => { bus => 1, addr => 24 },
5041 'net30' => { bus => 1, addr => 25 },
5042 'net31' => { bus => 1, addr => 26 },
5043 'xhci' => { bus => 1, addr => 27 },
5044 'virtio6' => { bus => 2, addr => 1 },
5045 'virtio7' => { bus => 2, addr => 2 },
5046 'virtio8' => { bus => 2, addr => 3 },
5047 'virtio9' => { bus => 2, addr => 4 },
5048 'virtio10' => { bus => 2, addr => 5 },
5049 'virtio11' => { bus => 2, addr => 6 },
5050 'virtio12' => { bus => 2, addr => 7 },
5051 'virtio13' => { bus => 2, addr => 8 },
5052 'virtio14' => { bus => 2, addr => 9 },
5053 'virtio15' => { bus => 2, addr => 10 },
5054 'virtioscsi0' => { bus => 3, addr => 1 },
5055 'virtioscsi1' => { bus => 3, addr => 2 },
5056 'virtioscsi2' => { bus => 3, addr => 3 },
5057 'virtioscsi3' => { bus => 3, addr => 4 },
5058 'virtioscsi4' => { bus => 3, addr => 5 },
5059 'virtioscsi5' => { bus => 3, addr => 6 },
5060 'virtioscsi6' => { bus => 3, addr => 7 },
5061 'virtioscsi7' => { bus => 3, addr => 8 },
5062 'virtioscsi8' => { bus => 3, addr => 9 },
5063 'virtioscsi9' => { bus => 3, addr => 10 },
5064 'virtioscsi10' => { bus => 3, addr => 11 },
5065 'virtioscsi11' => { bus => 3, addr => 12 },
5066 'virtioscsi12' => { bus => 3, addr => 13 },
5067 'virtioscsi13' => { bus => 3, addr => 14 },
5068 'virtioscsi14' => { bus => 3, addr => 15 },
5069 'virtioscsi15' => { bus => 3, addr => 16 },
5070 'virtioscsi16' => { bus => 3, addr => 17 },
5071 'virtioscsi17' => { bus => 3, addr => 18 },
5072 'virtioscsi18' => { bus => 3, addr => 19 },
5073 'virtioscsi19' => { bus => 3, addr => 20 },
5074 'virtioscsi20' => { bus => 3, addr => 21 },
5075 'virtioscsi21' => { bus => 3, addr => 22 },
5076 'virtioscsi22' => { bus => 3, addr => 23 },
5077 'virtioscsi23' => { bus => 3, addr => 24 },
5078 'virtioscsi24' => { bus => 3, addr => 25 },
5079 'virtioscsi25' => { bus => 3, addr => 26 },
5080 'virtioscsi26' => { bus => 3, addr => 27 },
5081 'virtioscsi27' => { bus => 3, addr => 28 },
5082 'virtioscsi28' => { bus => 3, addr => 29 },
5083 'virtioscsi29' => { bus => 3, addr => 30 },
5084 'virtioscsi30' => { bus => 3, addr => 31 },
5085
5086 };
5087
5088 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
5089 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
5090 my $bus = $devices->{$id}->{bus};
5091 $res = ",bus=pci.$bus,addr=$addr";
5092 $bridges->{$bus} = 1 if $bridges;
5093 }
5094 return $res;
5095
5096 }
5097
5098 sub print_pcie_addr {
5099 my ($id) = @_;
5100
5101 my $res = '';
5102 my $devices = {
5103 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
5104 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
5105 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
5106 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
5107 };
5108
5109 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
5110 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
5111 my $bus = $devices->{$id}->{bus};
5112 $res = ",bus=$bus,addr=$addr";
5113 }
5114 return $res;
5115
5116 }
5117
5118 # vzdump restore implementaion
5119
5120 sub tar_archive_read_firstfile {
5121 my $archive = shift;
5122
5123 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
5124
5125 # try to detect archive type first
5126 my $pid = open (TMP, "tar tf '$archive'|") ||
5127 die "unable to open file '$archive'\n";
5128 my $firstfile = <TMP>;
5129 kill 15, $pid;
5130 close TMP;
5131
5132 die "ERROR: archive contaions no data\n" if !$firstfile;
5133 chomp $firstfile;
5134
5135 return $firstfile;
5136 }
5137
5138 sub tar_restore_cleanup {
5139 my ($storecfg, $statfile) = @_;
5140
5141 print STDERR "starting cleanup\n";
5142
5143 if (my $fd = IO::File->new($statfile, "r")) {
5144 while (defined(my $line = <$fd>)) {
5145 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5146 my $volid = $2;
5147 eval {
5148 if ($volid =~ m|^/|) {
5149 unlink $volid || die 'unlink failed\n';
5150 } else {
5151 PVE::Storage::vdisk_free($storecfg, $volid);
5152 }
5153 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5154 };
5155 print STDERR "unable to cleanup '$volid' - $@" if $@;
5156 } else {
5157 print STDERR "unable to parse line in statfile - $line";
5158 }
5159 }
5160 $fd->close();
5161 }
5162 }
5163
5164 sub restore_archive {
5165 my ($archive, $vmid, $user, $opts) = @_;
5166
5167 my $format = $opts->{format};
5168 my $comp;
5169
5170 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
5171 $format = 'tar' if !$format;
5172 $comp = 'gzip';
5173 } elsif ($archive =~ m/\.tar$/) {
5174 $format = 'tar' if !$format;
5175 } elsif ($archive =~ m/.tar.lzo$/) {
5176 $format = 'tar' if !$format;
5177 $comp = 'lzop';
5178 } elsif ($archive =~ m/\.vma$/) {
5179 $format = 'vma' if !$format;
5180 } elsif ($archive =~ m/\.vma\.gz$/) {
5181 $format = 'vma' if !$format;
5182 $comp = 'gzip';
5183 } elsif ($archive =~ m/\.vma\.lzo$/) {
5184 $format = 'vma' if !$format;
5185 $comp = 'lzop';
5186 } else {
5187 $format = 'vma' if !$format; # default
5188 }
5189
5190 # try to detect archive format
5191 if ($format eq 'tar') {
5192 return restore_tar_archive($archive, $vmid, $user, $opts);
5193 } else {
5194 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
5195 }
5196 }
5197
5198 sub restore_update_config_line {
5199 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
5200
5201 return if $line =~ m/^\#qmdump\#/;
5202 return if $line =~ m/^\#vzdump\#/;
5203 return if $line =~ m/^lock:/;
5204 return if $line =~ m/^unused\d+:/;
5205 return if $line =~ m/^parent:/;
5206 return if $line =~ m/^template:/; # restored VM is never a template
5207
5208 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
5209 # try to convert old 1.X settings
5210 my ($id, $ind, $ethcfg) = ($1, $2, $3);
5211 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
5212 my ($model, $macaddr) = split(/\=/, $devconfig);
5213 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
5214 my $net = {
5215 model => $model,
5216 bridge => "vmbr$ind",
5217 macaddr => $macaddr,
5218 };
5219 my $netstr = print_net($net);
5220
5221 print $outfd "net$cookie->{netcount}: $netstr\n";
5222 $cookie->{netcount}++;
5223 }
5224 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
5225 my ($id, $netstr) = ($1, $2);
5226 my $net = parse_net($netstr);
5227 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
5228 $netstr = print_net($net);
5229 print $outfd "$id: $netstr\n";
5230 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
5231 my $virtdev = $1;
5232 my $value = $3;
5233 my $di = parse_drive($virtdev, $value);
5234 if (defined($di->{backup}) && !$di->{backup}) {
5235 print $outfd "#$line";
5236 } elsif ($map->{$virtdev}) {
5237 delete $di->{format}; # format can change on restore
5238 $di->{file} = $map->{$virtdev};
5239 $value = print_drive($vmid, $di);
5240 print $outfd "$virtdev: $value\n";
5241 } else {
5242 print $outfd $line;
5243 }
5244 } else {
5245 print $outfd $line;
5246 }
5247 }
5248
5249 sub scan_volids {
5250 my ($cfg, $vmid) = @_;
5251
5252 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
5253
5254 my $volid_hash = {};
5255 foreach my $storeid (keys %$info) {
5256 foreach my $item (@{$info->{$storeid}}) {
5257 next if !($item->{volid} && $item->{size});
5258 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
5259 $volid_hash->{$item->{volid}} = $item;
5260 }
5261 }
5262
5263 return $volid_hash;
5264 }
5265
5266 sub is_volume_in_use {
5267 my ($storecfg, $conf, $skip_drive, $volid) = @_;
5268
5269 my $path = PVE::Storage::path($storecfg, $volid);
5270
5271 my $scan_config = sub {
5272 my ($cref, $snapname) = @_;
5273
5274 foreach my $key (keys %$cref) {
5275 my $value = $cref->{$key};
5276 if (is_valid_drivename($key)) {
5277 next if $skip_drive && $key eq $skip_drive;
5278 my $drive = parse_drive($key, $value);
5279 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
5280 return 1 if $volid eq $drive->{file};
5281 if ($drive->{file} =~ m!^/!) {
5282 return 1 if $drive->{file} eq $path;
5283 } else {
5284 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
5285 next if !$storeid;
5286 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
5287 next if !$scfg;
5288 return 1 if $path eq PVE::Storage::path($storecfg, $drive->{file}, $snapname);
5289 }
5290 }
5291 }
5292
5293 return 0;
5294 };
5295
5296 return 1 if &$scan_config($conf);
5297
5298 undef $skip_drive;
5299
5300 foreach my $snapname (keys %{$conf->{snapshots}}) {
5301 return 1 if &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
5302 }
5303
5304 return 0;
5305 }
5306
5307 sub update_disksize {
5308 my ($vmid, $conf, $volid_hash) = @_;
5309
5310 my $changes;
5311
5312 my $used = {};
5313
5314 # Note: it is allowed to define multiple storages with same path (alias), so
5315 # we need to check both 'volid' and real 'path' (two different volid can point
5316 # to the same path).
5317
5318 my $usedpath = {};
5319
5320 # update size info
5321 foreach my $opt (keys %$conf) {
5322 if (is_valid_drivename($opt)) {
5323 my $drive = parse_drive($opt, $conf->{$opt});
5324 my $volid = $drive->{file};
5325 next if !$volid;
5326
5327 $used->{$volid} = 1;
5328 if ($volid_hash->{$volid} &&
5329 (my $path = $volid_hash->{$volid}->{path})) {
5330 $usedpath->{$path} = 1;
5331 }
5332
5333 next if drive_is_cdrom($drive);
5334 next if !$volid_hash->{$volid};
5335
5336 $drive->{size} = $volid_hash->{$volid}->{size};
5337 my $new = print_drive($vmid, $drive);
5338 if ($new ne $conf->{$opt}) {
5339 $changes = 1;
5340 $conf->{$opt} = $new;
5341 }
5342 }
5343 }
5344
5345 # remove 'unusedX' entry if volume is used
5346 foreach my $opt (keys %$conf) {
5347 next if $opt !~ m/^unused\d+$/;
5348 my $volid = $conf->{$opt};
5349 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5350 if ($used->{$volid} || ($path && $usedpath->{$path})) {
5351 $changes = 1;
5352 delete $conf->{$opt};
5353 }
5354 }
5355
5356 foreach my $volid (sort keys %$volid_hash) {
5357 next if $volid =~ m/vm-$vmid-state-/;
5358 next if $used->{$volid};
5359 my $path = $volid_hash->{$volid}->{path};
5360 next if !$path; # just to be sure
5361 next if $usedpath->{$path};
5362 $changes = 1;
5363 PVE::QemuConfig->add_unused_volume($conf, $volid);
5364 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
5365 }
5366
5367 return $changes;
5368 }
5369
5370 sub rescan {
5371 my ($vmid, $nolock) = @_;
5372
5373 my $cfg = PVE::Storage::config();
5374
5375 my $volid_hash = scan_volids($cfg, $vmid);
5376
5377 my $updatefn = sub {
5378 my ($vmid) = @_;
5379
5380 my $conf = PVE::QemuConfig->load_config($vmid);
5381
5382 PVE::QemuConfig->check_lock($conf);
5383
5384 my $vm_volids = {};
5385 foreach my $volid (keys %$volid_hash) {
5386 my $info = $volid_hash->{$volid};
5387 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5388 }
5389
5390 my $changes = update_disksize($vmid, $conf, $vm_volids);
5391
5392 PVE::QemuConfig->write_config($vmid, $conf) if $changes;
5393 };
5394
5395 if (defined($vmid)) {
5396 if ($nolock) {
5397 &$updatefn($vmid);
5398 } else {
5399 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5400 }
5401 } else {
5402 my $vmlist = config_list();
5403 foreach my $vmid (keys %$vmlist) {
5404 if ($nolock) {
5405 &$updatefn($vmid);
5406 } else {
5407 PVE::QemuConfig->lock_config($vmid, $updatefn, $vmid);
5408 }
5409 }
5410 }
5411 }
5412
5413 sub restore_vma_archive {
5414 my ($archive, $vmid, $user, $opts, $comp) = @_;
5415
5416 my $input = $archive eq '-' ? "<&STDIN" : undef;
5417 my $readfrom = $archive;
5418
5419 my $uncomp = '';
5420 if ($comp) {
5421 $readfrom = '-';
5422 my $qarchive = PVE::Tools::shellquote($archive);
5423 if ($comp eq 'gzip') {
5424 $uncomp = "zcat $qarchive|";
5425 } elsif ($comp eq 'lzop') {
5426 $uncomp = "lzop -d -c $qarchive|";
5427 } else {
5428 die "unknown compression method '$comp'\n";
5429 }
5430
5431 }
5432
5433 my $tmpdir = "/var/tmp/vzdumptmp$$";
5434 rmtree $tmpdir;
5435
5436 # disable interrupts (always do cleanups)
5437 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5438 warn "got interrupt - ignored\n";
5439 };
5440
5441 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5442 POSIX::mkfifo($mapfifo, 0600);
5443 my $fifofh;
5444
5445 my $openfifo = sub {
5446 open($fifofh, '>', $mapfifo) || die $!;
5447 };
5448
5449 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5450
5451 my $oldtimeout;
5452 my $timeout = 5;
5453
5454 my $devinfo = {};
5455
5456 my $rpcenv = PVE::RPCEnvironment::get();
5457
5458 my $conffile = PVE::QemuConfig->config_file($vmid);
5459 my $tmpfn = "$conffile.$$.tmp";
5460
5461 # Note: $oldconf is undef if VM does not exists
5462 my $cfs_path = PVE::QemuConfig->cfs_config_path($vmid);
5463 my $oldconf = PVE::Cluster::cfs_read_file($cfs_path);
5464
5465 my $print_devmap = sub {
5466 my $virtdev_hash = {};
5467
5468 my $cfgfn = "$tmpdir/qemu-server.conf";
5469
5470 # we can read the config - that is already extracted
5471 my $fh = IO::File->new($cfgfn, "r") ||
5472 "unable to read qemu-server.conf - $!\n";
5473
5474 my $fwcfgfn = "$tmpdir/qemu-server.fw";
5475 if (-f $fwcfgfn) {
5476 my $pve_firewall_dir = '/etc/pve/firewall';
5477 mkdir $pve_firewall_dir; # make sure the dir exists
5478 PVE::Tools::file_copy($fwcfgfn, "${pve_firewall_dir}/$vmid.fw");
5479 }
5480
5481 while (defined(my $line = <$fh>)) {
5482 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5483 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5484 die "archive does not contain data for drive '$virtdev'\n"
5485 if !$devinfo->{$devname};
5486 if (defined($opts->{storage})) {
5487 $storeid = $opts->{storage} || 'local';
5488 } elsif (!$storeid) {
5489 $storeid = 'local';
5490 }
5491 $format = 'raw' if !$format;
5492 $devinfo->{$devname}->{devname} = $devname;
5493 $devinfo->{$devname}->{virtdev} = $virtdev;
5494 $devinfo->{$devname}->{format} = $format;
5495 $devinfo->{$devname}->{storeid} = $storeid;
5496
5497 # check permission on storage
5498 my $pool = $opts->{pool}; # todo: do we need that?
5499 if ($user ne 'root@pam') {
5500 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5501 }
5502
5503 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5504 }
5505 }
5506
5507 foreach my $devname (keys %$devinfo) {
5508 die "found no device mapping information for device '$devname'\n"
5509 if !$devinfo->{$devname}->{virtdev};
5510 }
5511
5512 my $cfg = PVE::Storage::config();
5513
5514 # create empty/temp config
5515 if ($oldconf) {
5516 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5517 foreach_drive($oldconf, sub {
5518 my ($ds, $drive) = @_;
5519
5520 return if drive_is_cdrom($drive);
5521
5522 my $volid = $drive->{file};
5523
5524 return if !$volid || $volid =~ m|^/|;
5525
5526 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5527 return if !$path || !$owner || ($owner != $vmid);
5528
5529 # Note: only delete disk we want to restore
5530 # other volumes will become unused
5531 if ($virtdev_hash->{$ds}) {
5532 PVE::Storage::vdisk_free($cfg, $volid);
5533 }
5534 });
5535
5536 # delete vmstate files
5537 # since after the restore we have no snapshots anymore
5538 foreach my $snapname (keys %{$oldconf->{snapshots}}) {
5539 my $snap = $oldconf->{snapshots}->{$snapname};
5540 if ($snap->{vmstate}) {
5541 eval { PVE::Storage::vdisk_free($cfg, $snap->{vmstate}); };
5542 if (my $err = $@) {
5543 warn $err;
5544 }
5545 }
5546 }
5547 }
5548
5549 my $map = {};
5550 foreach my $virtdev (sort keys %$virtdev_hash) {
5551 my $d = $virtdev_hash->{$virtdev};
5552 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5553 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5554
5555 # test if requested format is supported
5556 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5557 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5558 $d->{format} = $defFormat if !$supported;
5559
5560 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5561 $d->{format}, undef, $alloc_size);
5562 print STDERR "new volume ID is '$volid'\n";
5563 $d->{volid} = $volid;
5564 my $path = PVE::Storage::path($cfg, $volid);
5565
5566 PVE::Storage::activate_volumes($cfg,[$volid]);
5567
5568 my $write_zeros = 1;
5569 if (PVE::Storage::volume_has_feature($cfg, 'sparseinit', $volid)) {
5570 $write_zeros = 0;
5571 }
5572
5573 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5574
5575 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5576 $map->{$virtdev} = $volid;
5577 }
5578
5579 $fh->seek(0, 0) || die "seek failed - $!\n";
5580
5581 my $outfd = new IO::File ($tmpfn, "w") ||
5582 die "unable to write config for VM $vmid\n";
5583
5584 my $cookie = { netcount => 0 };
5585 while (defined(my $line = <$fh>)) {
5586 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5587 }
5588
5589 $fh->close();
5590 $outfd->close();
5591 };
5592
5593 eval {
5594 # enable interrupts
5595 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5596 die "interrupted by signal\n";
5597 };
5598 local $SIG{ALRM} = sub { die "got timeout\n"; };
5599
5600 $oldtimeout = alarm($timeout);
5601
5602 my $parser = sub {
5603 my $line = shift;
5604
5605 print "$line\n";
5606
5607 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5608 my ($dev_id, $size, $devname) = ($1, $2, $3);
5609 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5610 } elsif ($line =~ m/^CTIME: /) {
5611 # we correctly received the vma config, so we can disable
5612 # the timeout now for disk allocation (set to 10 minutes, so
5613 # that we always timeout if something goes wrong)
5614 alarm(600);
5615 &$print_devmap();
5616 print $fifofh "done\n";
5617 my $tmp = $oldtimeout || 0;
5618 $oldtimeout = undef;
5619 alarm($tmp);
5620 close($fifofh);
5621 }
5622 };
5623
5624 print "restore vma archive: $cmd\n";
5625 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5626 };
5627 my $err = $@;
5628
5629 alarm($oldtimeout) if $oldtimeout;
5630
5631 my $vollist = [];
5632 foreach my $devname (keys %$devinfo) {
5633 my $volid = $devinfo->{$devname}->{volid};
5634 push @$vollist, $volid if $volid;
5635 }
5636
5637 my $cfg = PVE::Storage::config();
5638 PVE::Storage::deactivate_volumes($cfg, $vollist);
5639
5640 unlink $mapfifo;
5641
5642 if ($err) {
5643 rmtree $tmpdir;
5644 unlink $tmpfn;
5645
5646 foreach my $devname (keys %$devinfo) {
5647 my $volid = $devinfo->{$devname}->{volid};
5648 next if !$volid;
5649 eval {
5650 if ($volid =~ m|^/|) {
5651 unlink $volid || die 'unlink failed\n';
5652 } else {
5653 PVE::Storage::vdisk_free($cfg, $volid);
5654 }
5655 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5656 };
5657 print STDERR "unable to cleanup '$volid' - $@" if $@;
5658 }
5659 die $err;
5660 }
5661
5662 rmtree $tmpdir;
5663
5664 rename($tmpfn, $conffile) ||
5665 die "unable to commit configuration file '$conffile'\n";
5666
5667 PVE::Cluster::cfs_update(); # make sure we read new file
5668
5669 eval { rescan($vmid, 1); };
5670 warn $@ if $@;
5671 }
5672
5673 sub restore_tar_archive {
5674 my ($archive, $vmid, $user, $opts) = @_;
5675
5676 if ($archive ne '-') {
5677 my $firstfile = tar_archive_read_firstfile($archive);
5678 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5679 if $firstfile ne 'qemu-server.conf';
5680 }
5681
5682 my $storecfg = PVE::Storage::config();
5683
5684 # destroy existing data - keep empty config
5685 my $vmcfgfn = PVE::QemuConfig->config_file($vmid);
5686 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5687
5688 my $tocmd = "/usr/lib/qemu-server/qmextract";
5689
5690 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5691 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5692 $tocmd .= ' --prealloc' if $opts->{prealloc};
5693 $tocmd .= ' --info' if $opts->{info};
5694
5695 # tar option "xf" does not autodetect compression when read from STDIN,
5696 # so we pipe to zcat
5697 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5698 PVE::Tools::shellquote("--to-command=$tocmd");
5699
5700 my $tmpdir = "/var/tmp/vzdumptmp$$";
5701 mkpath $tmpdir;
5702
5703 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5704 local $ENV{VZDUMP_VMID} = $vmid;
5705 local $ENV{VZDUMP_USER} = $user;
5706
5707 my $conffile = PVE::QemuConfig->config_file($vmid);
5708 my $tmpfn = "$conffile.$$.tmp";
5709
5710 # disable interrupts (always do cleanups)
5711 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5712 print STDERR "got interrupt - ignored\n";
5713 };
5714
5715 eval {
5716 # enable interrupts
5717 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5718 die "interrupted by signal\n";
5719 };
5720
5721 if ($archive eq '-') {
5722 print "extracting archive from STDIN\n";
5723 run_command($cmd, input => "<&STDIN");
5724 } else {
5725 print "extracting archive '$archive'\n";
5726 run_command($cmd);
5727 }
5728
5729 return if $opts->{info};
5730
5731 # read new mapping
5732 my $map = {};
5733 my $statfile = "$tmpdir/qmrestore.stat";
5734 if (my $fd = IO::File->new($statfile, "r")) {
5735 while (defined (my $line = <$fd>)) {
5736 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5737 $map->{$1} = $2 if $1;
5738 } else {
5739 print STDERR "unable to parse line in statfile - $line\n";
5740 }
5741 }
5742 $fd->close();
5743 }
5744
5745 my $confsrc = "$tmpdir/qemu-server.conf";
5746
5747 my $srcfd = new IO::File($confsrc, "r") ||
5748 die "unable to open file '$confsrc'\n";
5749
5750 my $outfd = new IO::File ($tmpfn, "w") ||
5751 die "unable to write config for VM $vmid\n";
5752
5753 my $cookie = { netcount => 0 };
5754 while (defined (my $line = <$srcfd>)) {
5755 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5756 }
5757
5758 $srcfd->close();
5759 $outfd->close();
5760 };
5761 my $err = $@;
5762
5763 if ($err) {
5764
5765 unlink $tmpfn;
5766
5767 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5768
5769 die $err;
5770 }
5771
5772 rmtree $tmpdir;
5773
5774 rename $tmpfn, $conffile ||
5775 die "unable to commit configuration file '$conffile'\n";
5776
5777 PVE::Cluster::cfs_update(); # make sure we read new file
5778
5779 eval { rescan($vmid, 1); };
5780 warn $@ if $@;
5781 };
5782
5783 sub foreach_writable_storage {
5784 my ($conf, $func) = @_;
5785
5786 my $sidhash = {};
5787
5788 foreach my $ds (keys %$conf) {
5789 next if !is_valid_drivename($ds);
5790
5791 my $drive = parse_drive($ds, $conf->{$ds});
5792 next if !$drive;
5793 next if drive_is_cdrom($drive);
5794
5795 my $volid = $drive->{file};
5796
5797 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5798 $sidhash->{$sid} = $sid if $sid;
5799 }
5800
5801 foreach my $sid (sort keys %$sidhash) {
5802 &$func($sid);
5803 }
5804 }
5805
5806 sub do_snapshots_with_qemu {
5807 my ($storecfg, $volid) = @_;
5808
5809 my $storage_name = PVE::Storage::parse_volume_id($volid);
5810
5811 if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}}
5812 && !$storecfg->{ids}->{$storage_name}->{krbd}){
5813 return 1;
5814 }
5815
5816 if ($volid =~ m/\.(qcow2|qed)$/){
5817 return 1;
5818 }
5819
5820 return undef;
5821 }
5822
5823 sub qga_check_running {
5824 my ($vmid) = @_;
5825
5826 eval { vm_mon_cmd($vmid, "guest-ping", timeout => 3); };
5827 if ($@) {
5828 warn "Qemu Guest Agent are not running - $@";
5829 return 0;
5830 }
5831 return 1;
5832 }
5833
5834 sub template_create {
5835 my ($vmid, $conf, $disk) = @_;
5836
5837 my $storecfg = PVE::Storage::config();
5838
5839 foreach_drive($conf, sub {
5840 my ($ds, $drive) = @_;
5841
5842 return if drive_is_cdrom($drive);
5843 return if $disk && $ds ne $disk;
5844
5845 my $volid = $drive->{file};
5846 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5847
5848 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5849 $drive->{file} = $voliddst;
5850 $conf->{$ds} = print_drive($vmid, $drive);
5851 PVE::QemuConfig->write_config($vmid, $conf);
5852 });
5853 }
5854
5855 sub qemu_img_convert {
5856 my ($src_volid, $dst_volid, $size, $snapname, $is_zero_initialized) = @_;
5857
5858 my $storecfg = PVE::Storage::config();
5859 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5860 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5861
5862 if ($src_storeid && $dst_storeid) {
5863
5864 PVE::Storage::activate_volumes($storecfg, [$src_volid], $snapname);
5865
5866 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5867 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5868
5869 my $src_format = qemu_img_format($src_scfg, $src_volname);
5870 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5871
5872 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5873 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5874
5875 my $cmd = [];
5876 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5877 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5878 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path;
5879 if ($is_zero_initialized) {
5880 push @$cmd, "zeroinit:$dst_path";
5881 } else {
5882 push @$cmd, $dst_path;
5883 }
5884
5885 my $parser = sub {
5886 my $line = shift;
5887 if($line =~ m/\((\S+)\/100\%\)/){
5888 my $percent = $1;
5889 my $transferred = int($size * $percent / 100);
5890 my $remaining = $size - $transferred;
5891
5892 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5893 }
5894
5895 };
5896
5897 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5898 my $err = $@;
5899 die "copy failed: $err" if $err;
5900 }
5901 }
5902
5903 sub qemu_img_format {
5904 my ($scfg, $volname) = @_;
5905
5906 if ($scfg->{path} && $volname =~ m/\.(raw|cow|qcow|qcow2|qed|vmdk|cloop)$/) {
5907 return $1;
5908 } else {
5909 return "raw";
5910 }
5911 }
5912
5913 sub qemu_drive_mirror {
5914 my ($vmid, $drive, $dst_volid, $vmiddst, $is_zero_initialized) = @_;
5915
5916 my $storecfg = PVE::Storage::config();
5917 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5918
5919 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5920
5921 my $format = qemu_img_format($dst_scfg, $dst_volname);
5922
5923 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5924
5925 my $qemu_target = $is_zero_initialized ? "zeroinit:$dst_path" : $dst_path;
5926
5927 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $qemu_target };
5928 $opts->{format} = $format if $format;
5929
5930 print "drive mirror is starting (scanning bitmap) : this step can take some minutes/hours, depend of disk size and storage speed\n";
5931
5932 my $finish_job = sub {
5933 while (1) {
5934 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5935 my $stat = @$stats[0];
5936 last if !$stat;
5937 sleep 1;
5938 }
5939 };
5940
5941 eval {
5942 vm_mon_cmd($vmid, "drive-mirror", %$opts);
5943 while (1) {
5944 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5945 my $stat = @$stats[0];
5946 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5947 die "error job is not mirroring" if $stat->{type} ne "mirror";
5948
5949 my $busy = $stat->{busy};
5950 my $ready = $stat->{ready};
5951
5952 if (my $total = $stat->{len}) {
5953 my $transferred = $stat->{offset} || 0;
5954 my $remaining = $total - $transferred;
5955 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5956
5957 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
5958 }
5959
5960
5961 if ($stat->{ready} eq 'true') {
5962
5963 last if $vmiddst != $vmid;
5964
5965 # try to switch the disk if source and destination are on the same guest
5966 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
5967 if (!$@) {
5968 &$finish_job();
5969 last;
5970 }
5971 die $@ if $@ !~ m/cannot be completed/;
5972 }
5973 sleep 1;
5974 }
5975
5976
5977 };
5978 my $err = $@;
5979
5980 my $cancel_job = sub {
5981 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
5982 &$finish_job();
5983 };
5984
5985 if ($err) {
5986 eval { &$cancel_job(); };
5987 die "mirroring error: $err";
5988 }
5989
5990 if ($vmiddst != $vmid) {
5991 # if we clone a disk for a new target vm, we don't switch the disk
5992 &$cancel_job(); # so we call block-job-cancel
5993 }
5994 }
5995
5996 sub clone_disk {
5997 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
5998 $newvmid, $storage, $format, $full, $newvollist) = @_;
5999
6000 my $newvolid;
6001
6002 if (!$full) {
6003 print "create linked clone of drive $drivename ($drive->{file})\n";
6004 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6005 push @$newvollist, $newvolid;
6006 } else {
6007 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6008 $storeid = $storage if $storage;
6009
6010 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6011 if (!$format) {
6012 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6013 $format = qemu_img_format($scfg, $volname);
6014 }
6015
6016 # test if requested format is supported - else use default
6017 my $supported = grep { $_ eq $format } @$validFormats;
6018 $format = $defFormat if !$supported;
6019
6020 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6021
6022 print "create full clone of drive $drivename ($drive->{file})\n";
6023 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6024 push @$newvollist, $newvolid;
6025
6026 PVE::Storage::activate_volumes($storecfg, $newvollist);
6027
6028 my $sparseinit = PVE::Storage::volume_has_feature($storecfg, 'sparseinit', $newvolid);
6029 if (!$running || $snapname) {
6030 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname, $sparseinit);
6031 } else {
6032 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid, $sparseinit);
6033 }
6034 }
6035
6036 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6037
6038 my $disk = $drive;
6039 $disk->{format} = undef;
6040 $disk->{file} = $newvolid;
6041 $disk->{size} = $size;
6042
6043 return $disk;
6044 }
6045
6046 # this only works if VM is running
6047 sub get_current_qemu_machine {
6048 my ($vmid) = @_;
6049
6050 my $cmd = { execute => 'query-machines', arguments => {} };
6051 my $res = vm_qmp_command($vmid, $cmd);
6052
6053 my ($current, $default);
6054 foreach my $e (@$res) {
6055 $default = $e->{name} if $e->{'is-default'};
6056 $current = $e->{name} if $e->{'is-current'};
6057 }
6058
6059 # fallback to the default machine if current is not supported by qemu
6060 return $current || $default || 'pc';
6061 }
6062
6063 sub qemu_machine_feature_enabled {
6064 my ($machine, $kvmver, $version_major, $version_minor) = @_;
6065
6066 my $current_major;
6067 my $current_minor;
6068
6069 if ($machine && $machine =~ m/^(pc(-i440fx|-q35)?-(\d+)\.(\d+))/) {
6070
6071 $current_major = $3;
6072 $current_minor = $4;
6073
6074 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6075
6076 $current_major = $1;
6077 $current_minor = $2;
6078 }
6079
6080 return 1 if $current_major >= $version_major && $current_minor >= $version_minor;
6081
6082
6083 }
6084
6085 sub qemu_machine_pxe {
6086 my ($vmid, $conf, $machine) = @_;
6087
6088 $machine = PVE::QemuServer::get_current_qemu_machine($vmid) if !$machine;
6089
6090 foreach my $opt (keys %$conf) {
6091 next if $opt !~ m/^net(\d+)$/;
6092 my $net = PVE::QemuServer::parse_net($conf->{$opt});
6093 next if !$net;
6094 my $romfile = PVE::QemuServer::vm_mon_cmd_nocheck($vmid, 'qom-get', path => $opt, property => 'romfile');
6095 return $machine.".pxe" if $romfile =~ m/pxe/;
6096 last;
6097 }
6098
6099 return $machine;
6100 }
6101
6102 sub qemu_use_old_bios_files {
6103 my ($machine_type) = @_;
6104
6105 return if !$machine_type;
6106
6107 my $use_old_bios_files = undef;
6108
6109 if ($machine_type =~ m/^(\S+)\.pxe$/) {
6110 $machine_type = $1;
6111 $use_old_bios_files = 1;
6112 } else {
6113 my $kvmver = kvm_user_version();
6114 # Note: kvm version < 2.4 use non-efi pxe files, and have problems when we
6115 # load new efi bios files on migration. So this hack is required to allow
6116 # live migration from qemu-2.2 to qemu-2.4, which is sometimes used when
6117 # updrading from proxmox-ve-3.X to proxmox-ve 4.0
6118 $use_old_bios_files = !qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 4);
6119 }
6120
6121 return ($use_old_bios_files, $machine_type);
6122 }
6123
6124 sub lspci {
6125
6126 my $devices = {};
6127
6128 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6129 my (undef, $id, $function) = @_;
6130 my $res = { id => $id, function => $function};
6131 push @{$devices->{$id}}, $res;
6132 });
6133
6134 return $devices;
6135 }
6136
6137 sub vm_iothreads_list {
6138 my ($vmid) = @_;
6139
6140 my $res = vm_mon_cmd($vmid, 'query-iothreads');
6141
6142 my $iothreads = {};
6143 foreach my $iothread (@$res) {
6144 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6145 }
6146
6147 return $iothreads;
6148 }
6149
6150 sub scsihw_infos {
6151 my ($conf, $drive) = @_;
6152
6153 my $maxdev = 0;
6154
6155 if ($conf->{scsihw} && ($conf->{scsihw} =~ m/^lsi/)) {
6156 $maxdev = 7;
6157 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
6158 $maxdev = 1;
6159 } else {
6160 $maxdev = 256;
6161 }
6162
6163 my $controller = int($drive->{index} / $maxdev);
6164 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
6165
6166 return ($maxdev, $controller, $controller_prefix);
6167 }
6168
6169 # bash completion helper
6170
6171 sub complete_backup_archives {
6172 my ($cmdname, $pname, $cvalue) = @_;
6173
6174 my $cfg = PVE::Storage::config();
6175
6176 my $storeid;
6177
6178 if ($cvalue =~ m/^([^:]+):/) {
6179 $storeid = $1;
6180 }
6181
6182 my $data = PVE::Storage::template_list($cfg, $storeid, 'backup');
6183
6184 my $res = [];
6185 foreach my $id (keys %$data) {
6186 foreach my $item (@{$data->{$id}}) {
6187 next if $item->{format} !~ m/^vma\.(gz|lzo)$/;
6188 push @$res, $item->{volid} if defined($item->{volid});
6189 }
6190 }
6191
6192 return $res;
6193 }
6194
6195 my $complete_vmid_full = sub {
6196 my ($running) = @_;
6197
6198 my $idlist = vmstatus();
6199
6200 my $res = [];
6201
6202 foreach my $id (keys %$idlist) {
6203 my $d = $idlist->{$id};
6204 if (defined($running)) {
6205 next if $d->{template};
6206 next if $running && $d->{status} ne 'running';
6207 next if !$running && $d->{status} eq 'running';
6208 }
6209 push @$res, $id;
6210
6211 }
6212 return $res;
6213 };
6214
6215 sub complete_vmid {
6216 return &$complete_vmid_full();
6217 }
6218
6219 sub complete_vmid_stopped {
6220 return &$complete_vmid_full(0);
6221 }
6222
6223 sub complete_vmid_running {
6224 return &$complete_vmid_full(1);
6225 }
6226
6227 sub complete_storage {
6228
6229 my $cfg = PVE::Storage::config();
6230 my $ids = $cfg->{ids};
6231
6232 my $res = [];
6233 foreach my $sid (keys %$ids) {
6234 next if !PVE::Storage::storage_check_enabled($cfg, $sid, undef, 1);
6235 next if !$ids->{$sid}->{content}->{images};
6236 push @$res, $sid;
6237 }
6238
6239 return $res;
6240 }
6241
6242 1;