]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
enable hotplug by default
[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::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'boolean',
176 description => "Allow hotplug for disk and network device",
177 default => 1,
178 },
179 reboot => {
180 optional => 1,
181 type => 'boolean',
182 description => "Allow reboot. If set to '0' the VM exit on reboot.",
183 default => 1,
184 },
185 lock => {
186 optional => 1,
187 type => 'string',
188 description => "Lock/unlock the VM.",
189 enum => [qw(migrate backup snapshot rollback)],
190 },
191 cpulimit => {
192 optional => 1,
193 type => 'integer',
194 description => "Limit of CPU usage in per cent. Note if the computer has 2 CPUs, it has total of 200% CPU time. Value '0' indicates no CPU limit.\n\nNOTE: This option is currently ignored.",
195 minimum => 0,
196 default => 0,
197 },
198 cpuunits => {
199 optional => 1,
200 type => 'integer',
201 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.",
202 minimum => 0,
203 maximum => 500000,
204 default => 1000,
205 },
206 memory => {
207 optional => 1,
208 type => 'integer',
209 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
210 minimum => 16,
211 default => 512,
212 },
213 balloon => {
214 optional => 1,
215 type => 'integer',
216 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
217 minimum => 0,
218 },
219 shares => {
220 optional => 1,
221 type => 'integer',
222 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",
223 minimum => 0,
224 maximum => 50000,
225 default => 1000,
226 },
227 keyboard => {
228 optional => 1,
229 type => 'string',
230 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
231 enum => PVE::Tools::kvmkeymaplist(),
232 default => 'en-us',
233 },
234 name => {
235 optional => 1,
236 type => 'string', format => 'dns-name',
237 description => "Set a name for the VM. Only used on the configuration web interface.",
238 },
239 scsihw => {
240 optional => 1,
241 type => 'string',
242 description => "scsi controller model",
243 enum => [qw(lsi lsi53c810 virtio-scsi-pci megasas pvscsi)],
244 default => 'lsi',
245 },
246 description => {
247 optional => 1,
248 type => 'string',
249 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
250 },
251 ostype => {
252 optional => 1,
253 type => 'string',
254 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
255 description => <<EODESC,
256 Used to enable special optimization/features for specific
257 operating systems:
258
259 other => unspecified OS
260 wxp => Microsoft Windows XP
261 w2k => Microsoft Windows 2000
262 w2k3 => Microsoft Windows 2003
263 w2k8 => Microsoft Windows 2008
264 wvista => Microsoft Windows Vista
265 win7 => Microsoft Windows 7
266 win8 => Microsoft Windows 8/2012
267 l24 => Linux 2.4 Kernel
268 l26 => Linux 2.6/3.X Kernel
269 solaris => solaris/opensolaris/openindiania kernel
270
271 other|l24|l26|solaris ... no special behaviour
272 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
273 EODESC
274 },
275 boot => {
276 optional => 1,
277 type => 'string',
278 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
279 pattern => '[acdn]{1,4}',
280 default => 'cdn',
281 },
282 bootdisk => {
283 optional => 1,
284 type => 'string', format => 'pve-qm-bootdisk',
285 description => "Enable booting from specified disk.",
286 pattern => '(ide|sata|scsi|virtio)\d+',
287 },
288 smp => {
289 optional => 1,
290 type => 'integer',
291 description => "The number of CPUs. Please use option -sockets instead.",
292 minimum => 1,
293 default => 1,
294 },
295 sockets => {
296 optional => 1,
297 type => 'integer',
298 description => "The number of CPU sockets.",
299 minimum => 1,
300 default => 1,
301 },
302 cores => {
303 optional => 1,
304 type => 'integer',
305 description => "The number of cores per socket.",
306 minimum => 1,
307 default => 1,
308 },
309 numa => {
310 optional => 1,
311 type => 'boolean',
312 description => "Enable/disable Numa.",
313 default => 0,
314 },
315 maxcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Maximum cpus for hotplug.",
319 minimum => 1,
320 default => 1,
321 },
322 acpi => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable ACPI.",
326 default => 1,
327 },
328 agent => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable Qemu GuestAgent.",
332 default => 0,
333 },
334 kvm => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable KVM hardware virtualization.",
338 default => 1,
339 },
340 tdf => {
341 optional => 1,
342 type => 'boolean',
343 description => "Enable/disable time drift fix.",
344 default => 0,
345 },
346 localtime => {
347 optional => 1,
348 type => 'boolean',
349 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
350 },
351 freeze => {
352 optional => 1,
353 type => 'boolean',
354 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
355 },
356 vga => {
357 optional => 1,
358 type => 'string',
359 description => "Select VGA type. If you want to use high resolution modes (>= 1280x1024x16) then you should use option 'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and 'cirrur' for other OS types. Option 'qxl' enables the SPICE display sever. You can also run without any graphic card using a serial devive as terminal.",
360 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
361 },
362 watchdog => {
363 optional => 1,
364 type => 'string', format => 'pve-qm-watchdog',
365 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
366 description => "Create a virtual hardware watchdog device. Once enabled (by a guest action), the watchdog must be periodically polled by an agent inside the guest or else the guest will be restarted (or execute the action specified)",
367 },
368 startdate => {
369 optional => 1,
370 type => 'string',
371 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
372 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'.",
373 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
374 default => 'now',
375 },
376 startup => {
377 optional => 1,
378 type => 'string', format => 'pve-qm-startup',
379 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
380 description => "Startup and shutdown behavior. Order is a non-negative number defining the general startup order. Shutdown in done with reverse ordering. Additionally you can set the 'up' or 'down' delay in seconds, which specifies a delay to wait before the next VM is started or stopped.",
381 },
382 template => {
383 optional => 1,
384 type => 'boolean',
385 description => "Enable/disable Template.",
386 default => 0,
387 },
388 args => {
389 optional => 1,
390 type => 'string',
391 description => <<EODESCR,
392 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
393
394 args: -no-reboot -no-hpet
395 EODESCR
396 },
397 tablet => {
398 optional => 1,
399 type => 'boolean',
400 default => 1,
401 description => "Enable/disable the usb tablet device. This device is usually needed to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with normal VNC clients. If you're running lots of console-only guests on one host, you may consider disabling this to save some context switches. This is turned of by default if you use spice (vga=qxl).",
402 },
403 migrate_speed => {
404 optional => 1,
405 type => 'integer',
406 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
407 minimum => 0,
408 default => 0,
409 },
410 migrate_downtime => {
411 optional => 1,
412 type => 'number',
413 description => "Set maximum tolerated downtime (in seconds) for migrations.",
414 minimum => 0,
415 default => 0.1,
416 },
417 cdrom => {
418 optional => 1,
419 type => 'string', format => 'pve-qm-drive',
420 typetext => 'volume',
421 description => "This is an alias for option -ide2",
422 },
423 cpu => {
424 optional => 1,
425 description => "Emulated CPU type.",
426 type => 'string',
427 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
428 default => 'kvm64',
429 },
430 parent => get_standard_option('pve-snapshot-name', {
431 optional => 1,
432 description => "Parent snapshot name. This is used internally, and should not be modified.",
433 }),
434 snaptime => {
435 optional => 1,
436 description => "Timestamp for snapshots.",
437 type => 'integer',
438 minimum => 0,
439 },
440 vmstate => {
441 optional => 1,
442 type => 'string', format => 'pve-volume-id',
443 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
444 },
445 machine => {
446 description => "Specific the Qemu machine type.",
447 type => 'string',
448 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
449 maxLength => 40,
450 optional => 1,
451 },
452 smbios1 => {
453 description => "Specify SMBIOS type 1 fields.",
454 type => 'string', format => 'pve-qm-smbios1',
455 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
456 maxLength => 256,
457 optional => 1,
458 },
459 };
460
461 # what about other qemu settings ?
462 #cpu => 'string',
463 #machine => 'string',
464 #fda => 'file',
465 #fdb => 'file',
466 #mtdblock => 'file',
467 #sd => 'file',
468 #pflash => 'file',
469 #snapshot => 'bool',
470 #bootp => 'file',
471 ##tftp => 'dir',
472 ##smb => 'dir',
473 #kernel => 'file',
474 #append => 'string',
475 #initrd => 'file',
476 ##soundhw => 'string',
477
478 while (my ($k, $v) = each %$confdesc) {
479 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
480 }
481
482 my $MAX_IDE_DISKS = 4;
483 my $MAX_SCSI_DISKS = 14;
484 my $MAX_VIRTIO_DISKS = 16;
485 my $MAX_SATA_DISKS = 6;
486 my $MAX_USB_DEVICES = 5;
487 my $MAX_NETS = 32;
488 my $MAX_UNUSED_DISKS = 8;
489 my $MAX_HOSTPCI_DEVICES = 4;
490 my $MAX_SERIAL_PORTS = 4;
491 my $MAX_PARALLEL_PORTS = 3;
492 my $MAX_NUMA = 8;
493
494 my $numadesc = {
495 optional => 1,
496 type => 'string', format => 'pve-qm-numanode',
497 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
498 description => "numa topology",
499 };
500 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
501
502 for (my $i = 0; $i < $MAX_NUMA; $i++) {
503 $confdesc->{"numa$i"} = $numadesc;
504 }
505
506 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
507 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
508 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
509 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
510
511 my $netdesc = {
512 optional => 1,
513 type => 'string', format => 'pve-qm-net',
514 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
515 description => <<EODESCR,
516 Specify network devices.
517
518 MODEL is one of: $nic_model_list_txt
519
520 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
521 automatically generated if not specified.
522
523 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
524
525 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'.
526
527 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
528
529 10.0.2.2 Gateway
530 10.0.2.3 DNS Server
531 10.0.2.4 SMB Server
532
533 The DHCP server assign addresses to the guest starting from 10.0.2.15.
534
535 EODESCR
536 };
537 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
538
539 for (my $i = 0; $i < $MAX_NETS; $i++) {
540 $confdesc->{"net$i"} = $netdesc;
541 }
542
543 my $drivename_hash;
544
545 my $idedesc = {
546 optional => 1,
547 type => 'string', format => 'pve-qm-drive',
548 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
549 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
550 };
551 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
552
553 my $scsidesc = {
554 optional => 1,
555 type => 'string', format => 'pve-qm-drive',
556 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
557 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
558 };
559 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
560
561 my $satadesc = {
562 optional => 1,
563 type => 'string', format => 'pve-qm-drive',
564 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
565 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
566 };
567 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
568
569 my $virtiodesc = {
570 optional => 1,
571 type => 'string', format => 'pve-qm-drive',
572 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
573 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
574 };
575 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
576
577 my $usbdesc = {
578 optional => 1,
579 type => 'string', format => 'pve-qm-usb-device',
580 typetext => 'host=HOSTUSBDEVICE|spice',
581 description => <<EODESCR,
582 Configure an USB device (n is 0 to 4). This can be used to
583 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
584
585 'bus-port(.port)*' (decimal numbers) or
586 'vendor_id:product_id' (hexadeciaml numbers)
587
588 You can use the 'lsusb -t' command to list existing usb devices.
589
590 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
591
592 The value 'spice' can be used to add a usb redirection devices for spice.
593
594 EODESCR
595 };
596 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
597
598 my $hostpcidesc = {
599 optional => 1,
600 type => 'string', format => 'pve-qm-hostpci',
601 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
602 description => <<EODESCR,
603 Map host pci devices. HOSTPCIDEVICE syntax is:
604
605 'bus:dev.func' (hexadecimal numbers)
606
607 You can us the 'lspci' command to list existing pci devices.
608
609 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
610
611 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
612
613 Experimental: user reported problems with this option.
614 EODESCR
615 };
616 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
617
618 my $serialdesc = {
619 optional => 1,
620 type => 'string',
621 pattern => '(/dev/.+|socket)',
622 description => <<EODESCR,
623 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).
624
625 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
626
627 Experimental: user reported problems with this option.
628 EODESCR
629 };
630
631 my $paralleldesc= {
632 optional => 1,
633 type => 'string',
634 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
635 description => <<EODESCR,
636 Map host parallel devices (n is 0 to 2).
637
638 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
639
640 Experimental: user reported problems with this option.
641 EODESCR
642 };
643
644 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
645 $confdesc->{"parallel$i"} = $paralleldesc;
646 }
647
648 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
649 $confdesc->{"serial$i"} = $serialdesc;
650 }
651
652 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
653 $confdesc->{"hostpci$i"} = $hostpcidesc;
654 }
655
656 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
657 $drivename_hash->{"ide$i"} = 1;
658 $confdesc->{"ide$i"} = $idedesc;
659 }
660
661 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
662 $drivename_hash->{"sata$i"} = 1;
663 $confdesc->{"sata$i"} = $satadesc;
664 }
665
666 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
667 $drivename_hash->{"scsi$i"} = 1;
668 $confdesc->{"scsi$i"} = $scsidesc ;
669 }
670
671 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
672 $drivename_hash->{"virtio$i"} = 1;
673 $confdesc->{"virtio$i"} = $virtiodesc;
674 }
675
676 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
677 $confdesc->{"usb$i"} = $usbdesc;
678 }
679
680 my $unuseddesc = {
681 optional => 1,
682 type => 'string', format => 'pve-volume-id',
683 description => "Reference to unused volumes.",
684 };
685
686 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
687 $confdesc->{"unused$i"} = $unuseddesc;
688 }
689
690 my $kvm_api_version = 0;
691
692 sub kvm_version {
693
694 return $kvm_api_version if $kvm_api_version;
695
696 my $fh = IO::File->new("</dev/kvm") ||
697 return 0;
698
699 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
700 $kvm_api_version = $v;
701 }
702
703 $fh->close();
704
705 return $kvm_api_version;
706 }
707
708 my $kvm_user_version;
709
710 sub kvm_user_version {
711
712 return $kvm_user_version if $kvm_user_version;
713
714 $kvm_user_version = 'unknown';
715
716 my $tmp = `kvm -help 2>/dev/null`;
717
718 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
719 $kvm_user_version = $2;
720 }
721
722 return $kvm_user_version;
723
724 }
725
726 my $kernel_has_vhost_net = -c '/dev/vhost-net';
727
728 sub disknames {
729 # order is important - used to autoselect boot disk
730 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
731 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
732 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
733 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
734 }
735
736 sub valid_drivename {
737 my $dev = shift;
738
739 return defined($drivename_hash->{$dev});
740 }
741
742 sub option_exists {
743 my $key = shift;
744 return defined($confdesc->{$key});
745 }
746
747 sub nic_models {
748 return $nic_model_list;
749 }
750
751 sub os_list_description {
752
753 return {
754 other => 'Other',
755 wxp => 'Windows XP',
756 w2k => 'Windows 2000',
757 w2k3 =>, 'Windows 2003',
758 w2k8 => 'Windows 2008',
759 wvista => 'Windows Vista',
760 win7 => 'Windows 7',
761 win8 => 'Windows 8/2012',
762 l24 => 'Linux 2.4',
763 l26 => 'Linux 2.6',
764 };
765 }
766
767 my $cdrom_path;
768
769 sub get_cdrom_path {
770
771 return $cdrom_path if $cdrom_path;
772
773 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
774 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
775 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
776 }
777
778 sub get_iso_path {
779 my ($storecfg, $vmid, $cdrom) = @_;
780
781 if ($cdrom eq 'cdrom') {
782 return get_cdrom_path();
783 } elsif ($cdrom eq 'none') {
784 return '';
785 } elsif ($cdrom =~ m|^/|) {
786 return $cdrom;
787 } else {
788 return PVE::Storage::path($storecfg, $cdrom);
789 }
790 }
791
792 # try to convert old style file names to volume IDs
793 sub filename_to_volume_id {
794 my ($vmid, $file, $media) = @_;
795
796 if (!($file eq 'none' || $file eq 'cdrom' ||
797 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
798
799 return undef if $file =~ m|/|;
800
801 if ($media && $media eq 'cdrom') {
802 $file = "local:iso/$file";
803 } else {
804 $file = "local:$vmid/$file";
805 }
806 }
807
808 return $file;
809 }
810
811 sub verify_media_type {
812 my ($opt, $vtype, $media) = @_;
813
814 return if !$media;
815
816 my $etype;
817 if ($media eq 'disk') {
818 $etype = 'images';
819 } elsif ($media eq 'cdrom') {
820 $etype = 'iso';
821 } else {
822 die "internal error";
823 }
824
825 return if ($vtype eq $etype);
826
827 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
828 }
829
830 sub cleanup_drive_path {
831 my ($opt, $storecfg, $drive) = @_;
832
833 # try to convert filesystem paths to volume IDs
834
835 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
836 ($drive->{file} !~ m|^/dev/.+|) &&
837 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
838 ($drive->{file} !~ m/^\d+$/)) {
839 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
840 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
841 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
842 verify_media_type($opt, $vtype, $drive->{media});
843 $drive->{file} = $volid;
844 }
845
846 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
847 }
848
849 sub create_conf_nolock {
850 my ($vmid, $settings) = @_;
851
852 my $filename = config_file($vmid);
853
854 die "configuration file '$filename' already exists\n" if -f $filename;
855
856 my $defaults = load_defaults();
857
858 $settings->{name} = "vm$vmid" if !$settings->{name};
859 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
860
861 my $data = '';
862 foreach my $opt (keys %$settings) {
863 next if !$confdesc->{$opt};
864
865 my $value = $settings->{$opt};
866 next if !$value;
867
868 $data .= "$opt: $value\n";
869 }
870
871 PVE::Tools::file_set_contents($filename, $data);
872 }
873
874 my $parse_size = sub {
875 my ($value) = @_;
876
877 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
878 my ($size, $unit) = ($1, $3);
879 if ($unit) {
880 if ($unit eq 'K') {
881 $size = $size * 1024;
882 } elsif ($unit eq 'M') {
883 $size = $size * 1024 * 1024;
884 } elsif ($unit eq 'G') {
885 $size = $size * 1024 * 1024 * 1024;
886 }
887 }
888 return int($size);
889 };
890
891 my $format_size = sub {
892 my ($size) = @_;
893
894 $size = int($size);
895
896 my $kb = int($size/1024);
897 return $size if $kb*1024 != $size;
898
899 my $mb = int($kb/1024);
900 return "${kb}K" if $mb*1024 != $kb;
901
902 my $gb = int($mb/1024);
903 return "${mb}M" if $gb*1024 != $mb;
904
905 return "${gb}G";
906 };
907
908 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
909 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
910 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
911 # [,aio=native|threads][,discard=ignore|on]
912
913 sub parse_drive {
914 my ($key, $data) = @_;
915
916 my $res = {};
917
918 # $key may be undefined - used to verify JSON parameters
919 if (!defined($key)) {
920 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
921 $res->{index} = 0;
922 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
923 $res->{interface} = $1;
924 $res->{index} = $2;
925 } else {
926 return undef;
927 }
928
929 foreach my $p (split (/,/, $data)) {
930 next if $p =~ m/^\s*$/;
931
932 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
933 my ($k, $v) = ($1, $2);
934
935 $k = 'file' if $k eq 'volume';
936
937 return undef if defined $res->{$k};
938
939 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
940 return undef if !$v || $v !~ m/^\d+/;
941 $k = "m$k";
942 $v = sprintf("%.3f", $v / (1024*1024));
943 }
944 $res->{$k} = $v;
945 } else {
946 if (!$res->{file} && $p !~ m/=/) {
947 $res->{file} = $p;
948 } else {
949 return undef;
950 }
951 }
952 }
953
954 return undef if !$res->{file};
955
956 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
957 $res->{format} = $1;
958 }
959
960 return undef if $res->{cache} &&
961 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
962 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
963 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
964 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
965 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
966 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
967 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
968 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
969 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
970 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
971 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
972 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
973 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
974
975 return undef if $res->{mbps_rd} && $res->{mbps};
976 return undef if $res->{mbps_wr} && $res->{mbps};
977
978 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
979 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
980 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
981 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
982 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
983 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
984
985 return undef if $res->{iops_rd} && $res->{iops};
986 return undef if $res->{iops_wr} && $res->{iops};
987
988
989 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
990 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
991 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
992 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
993 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
994 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
995
996
997 if ($res->{size}) {
998 return undef if !defined($res->{size} = &$parse_size($res->{size}));
999 }
1000
1001 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1002 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1003 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1004 return undef if $res->{interface} eq 'virtio';
1005 }
1006
1007 # rerror does not work with scsi drives
1008 if ($res->{rerror}) {
1009 return undef if $res->{interface} eq 'scsi';
1010 }
1011
1012 return $res;
1013 }
1014
1015 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1016
1017 sub print_drive {
1018 my ($vmid, $drive) = @_;
1019
1020 my $opts = '';
1021 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1022 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1023 }
1024
1025 if ($drive->{size}) {
1026 $opts .= ",size=" . &$format_size($drive->{size});
1027 }
1028
1029 return "$drive->{file}$opts";
1030 }
1031
1032 sub scsi_inquiry {
1033 my($fh, $noerr) = @_;
1034
1035 my $SG_IO = 0x2285;
1036 my $SG_GET_VERSION_NUM = 0x2282;
1037
1038 my $versionbuf = "\x00" x 8;
1039 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1040 if (!$ret) {
1041 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1042 return undef;
1043 }
1044 my $version = unpack("I", $versionbuf);
1045 if ($version < 30000) {
1046 die "scsi generic interface too old\n" if !$noerr;
1047 return undef;
1048 }
1049
1050 my $buf = "\x00" x 36;
1051 my $sensebuf = "\x00" x 8;
1052 my $cmd = pack("C x3 C x1", 0x12, 36);
1053
1054 # see /usr/include/scsi/sg.h
1055 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";
1056
1057 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1058 length($sensebuf), 0, length($buf), $buf,
1059 $cmd, $sensebuf, 6000);
1060
1061 $ret = ioctl($fh, $SG_IO, $packet);
1062 if (!$ret) {
1063 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1064 return undef;
1065 }
1066
1067 my @res = unpack($sg_io_hdr_t, $packet);
1068 if ($res[17] || $res[18]) {
1069 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1070 return undef;
1071 }
1072
1073 my $res = {};
1074 (my $byte0, my $byte1, $res->{vendor},
1075 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1076
1077 $res->{removable} = $byte1 & 128 ? 1 : 0;
1078 $res->{type} = $byte0 & 31;
1079
1080 return $res;
1081 }
1082
1083 sub path_is_scsi {
1084 my ($path) = @_;
1085
1086 my $fh = IO::File->new("+<$path") || return undef;
1087 my $res = scsi_inquiry($fh, 1);
1088 close($fh);
1089
1090 return $res;
1091 }
1092
1093 sub machine_type_is_q35 {
1094 my ($conf) = @_;
1095
1096 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1097 }
1098
1099 sub print_tabletdevice_full {
1100 my ($conf) = @_;
1101
1102 my $q35 = machine_type_is_q35($conf);
1103
1104 # we use uhci for old VMs because tablet driver was buggy in older qemu
1105 my $usbbus = $q35 ? "ehci" : "uhci";
1106
1107 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1108 }
1109
1110 sub print_drivedevice_full {
1111 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1112
1113 my $device = '';
1114 my $maxdev = 0;
1115
1116 if ($drive->{interface} eq 'virtio') {
1117 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1118 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1119 $device .= ",iothread=iothread0" if $conf->{iothread};
1120 } elsif ($drive->{interface} eq 'scsi') {
1121 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1122 my $controller = int($drive->{index} / $maxdev);
1123 my $unit = $drive->{index} % $maxdev;
1124 my $devicetype = 'hd';
1125 my $path = '';
1126 if (drive_is_cdrom($drive)) {
1127 $devicetype = 'cd';
1128 } else {
1129 if ($drive->{file} =~ m|^/|) {
1130 $path = $drive->{file};
1131 } else {
1132 $path = PVE::Storage::path($storecfg, $drive->{file});
1133 }
1134
1135 if($path =~ m/^iscsi\:\/\//){
1136 $devicetype = 'generic';
1137 } else {
1138 if (my $info = path_is_scsi($path)) {
1139 if ($info->{type} == 0) {
1140 $devicetype = 'block';
1141 } elsif ($info->{type} == 1) { # tape
1142 $devicetype = 'generic';
1143 }
1144 }
1145 }
1146 }
1147
1148 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1149 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1150 } else {
1151 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1152 }
1153
1154 } elsif ($drive->{interface} eq 'ide'){
1155 $maxdev = 2;
1156 my $controller = int($drive->{index} / $maxdev);
1157 my $unit = $drive->{index} % $maxdev;
1158 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1159
1160 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1161 } elsif ($drive->{interface} eq 'sata'){
1162 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1163 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1164 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1165 } elsif ($drive->{interface} eq 'usb') {
1166 die "implement me";
1167 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1168 } else {
1169 die "unsupported interface type";
1170 }
1171
1172 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1173
1174 return $device;
1175 }
1176
1177 sub get_initiator_name {
1178 my $initiator;
1179
1180 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1181 while (defined(my $line = <$fh>)) {
1182 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1183 $initiator = $1;
1184 last;
1185 }
1186 $fh->close();
1187
1188 return $initiator;
1189 }
1190
1191 sub print_drive_full {
1192 my ($storecfg, $vmid, $drive) = @_;
1193
1194 my $opts = '';
1195 foreach my $o (@qemu_drive_options) {
1196 next if $o eq 'bootindex';
1197 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1198 }
1199
1200 foreach my $o (qw(bps bps_rd bps_wr)) {
1201 my $v = $drive->{"m$o"};
1202 $opts .= ",$o=" . int($v*1024*1024) if $v;
1203 }
1204
1205 # use linux-aio by default (qemu default is threads)
1206 $opts .= ",aio=native" if !$drive->{aio};
1207
1208 my $path;
1209 my $volid = $drive->{file};
1210 if (drive_is_cdrom($drive)) {
1211 $path = get_iso_path($storecfg, $vmid, $volid);
1212 } else {
1213 if ($volid =~ m|^/|) {
1214 $path = $volid;
1215 } else {
1216 $path = PVE::Storage::path($storecfg, $volid);
1217 }
1218 }
1219
1220 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1221
1222 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1223 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1224
1225 my $pathinfo = $path ? "file=$path," : '';
1226
1227 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1228 }
1229
1230 sub print_netdevice_full {
1231 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1232
1233 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1234
1235 my $device = $net->{model};
1236 if ($net->{model} eq 'virtio') {
1237 $device = 'virtio-net-pci';
1238 };
1239
1240 my $pciaddr = print_pci_addr("$netid", $bridges);
1241 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1242 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1243 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1244 my $vectors = $net->{queues} * 2 + 2;
1245 $tmpstr .= ",vectors=$vectors,mq=on";
1246 }
1247 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1248 return $tmpstr;
1249 }
1250
1251 sub print_netdev_full {
1252 my ($vmid, $conf, $net, $netid) = @_;
1253
1254 my $i = '';
1255 if ($netid =~ m/^net(\d+)$/) {
1256 $i = int($1);
1257 }
1258
1259 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1260
1261 my $ifname = "tap${vmid}i$i";
1262
1263 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1264 die "interface name '$ifname' is too long (max 15 character)\n"
1265 if length($ifname) >= 16;
1266
1267 my $vhostparam = '';
1268 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1269
1270 my $vmname = $conf->{name} || "vm$vmid";
1271
1272 my $netdev = "";
1273
1274 if ($net->{bridge}) {
1275 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1276 } else {
1277 $netdev = "type=user,id=$netid,hostname=$vmname";
1278 }
1279
1280 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1281
1282 return $netdev;
1283 }
1284
1285 sub drive_is_cdrom {
1286 my ($drive) = @_;
1287
1288 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1289
1290 }
1291
1292 sub parse_numa {
1293 my ($data) = @_;
1294
1295 my $res = {};
1296
1297 foreach my $kvp (split(/,/, $data)) {
1298
1299 if ($kvp =~ m/^memory=(\S+)$/) {
1300 $res->{memory} = $1;
1301 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1302 $res->{policy} = $1;
1303 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1304 $res->{cpus}->{start} = $1;
1305 $res->{cpus}->{end} = $3;
1306 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1307 $res->{hostnodes}->{start} = $1;
1308 $res->{hostnodes}->{end} = $3;
1309 } else {
1310 return undef;
1311 }
1312 }
1313
1314 return $res;
1315 }
1316
1317 sub parse_hostpci {
1318 my ($value) = @_;
1319
1320 return undef if !$value;
1321
1322
1323 my @list = split(/,/, $value);
1324 my $found;
1325
1326 my $res = {};
1327 foreach my $kv (@list) {
1328
1329 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1330 $found = 1;
1331 if(defined($4)){
1332 push @{$res->{pciid}}, { id => $2 , function => $4};
1333
1334 }else{
1335 my $pcidevices = lspci($2);
1336 $res->{pciid} = $pcidevices->{$2};
1337 }
1338 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1339 $res->{driver} = $1;
1340 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1341 $res->{rombar} = $1;
1342 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1343 $res->{'x-vga'} = $1;
1344 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1345 $res->{pcie} = 1 if $1 == 1;
1346 } else {
1347 warn "unknown hostpci setting '$kv'\n";
1348 }
1349 }
1350
1351 return undef if !$found;
1352
1353 return $res;
1354 }
1355
1356 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1357 sub parse_net {
1358 my ($data) = @_;
1359
1360 my $res = {};
1361
1362 foreach my $kvp (split(/,/, $data)) {
1363
1364 if ($kvp =~ m/^(ne2k_pci|e1000|e1000-82540em|e1000-82544gc|e1000-82545em|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1365 my $model = lc($1);
1366 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1367 $res->{model} = $model;
1368 $res->{macaddr} = $mac;
1369 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1370 $res->{bridge} = $1;
1371 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1372 $res->{queues} = $1;
1373 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1374 $res->{rate} = $1;
1375 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1376 $res->{tag} = $1;
1377 } elsif ($kvp =~ m/^firewall=([01])$/) {
1378 $res->{firewall} = $1;
1379 } elsif ($kvp =~ m/^link_down=([01])$/) {
1380 $res->{link_down} = $1;
1381 } else {
1382 return undef;
1383 }
1384
1385 }
1386
1387 return undef if !$res->{model};
1388
1389 return $res;
1390 }
1391
1392 sub print_net {
1393 my $net = shift;
1394
1395 my $res = "$net->{model}";
1396 $res .= "=$net->{macaddr}" if $net->{macaddr};
1397 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1398 $res .= ",rate=$net->{rate}" if $net->{rate};
1399 $res .= ",tag=$net->{tag}" if $net->{tag};
1400 $res .= ",firewall=1" if $net->{firewall};
1401 $res .= ",link_down=1" if $net->{link_down};
1402
1403 return $res;
1404 }
1405
1406 sub add_random_macs {
1407 my ($settings) = @_;
1408
1409 foreach my $opt (keys %$settings) {
1410 next if $opt !~ m/^net(\d+)$/;
1411 my $net = parse_net($settings->{$opt});
1412 next if !$net;
1413 $settings->{$opt} = print_net($net);
1414 }
1415 }
1416
1417 sub add_unused_volume {
1418 my ($config, $volid) = @_;
1419
1420 my $key;
1421 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1422 my $test = "unused$ind";
1423 if (my $vid = $config->{$test}) {
1424 return if $vid eq $volid; # do not add duplicates
1425 } else {
1426 $key = $test;
1427 }
1428 }
1429
1430 die "To many unused volume - please delete them first.\n" if !$key;
1431
1432 $config->{$key} = $volid;
1433
1434 return $key;
1435 }
1436
1437 sub vm_is_volid_owner {
1438 my ($storecfg, $vmid, $volid) = @_;
1439
1440 if ($volid !~ m|^/|) {
1441 my ($path, $owner);
1442 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1443 if ($owner && ($owner == $vmid)) {
1444 return 1;
1445 }
1446 }
1447
1448 return undef;
1449 }
1450
1451 sub vmconfig_delete_pending_option {
1452 my ($conf, $key) = @_;
1453
1454 delete $conf->{pending}->{$key};
1455 my $pending_delete_hash = { $key => 1 };
1456 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1457 $pending_delete_hash->{$opt} = 1;
1458 }
1459 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1460 }
1461
1462 sub vmconfig_undelete_pending_option {
1463 my ($conf, $key) = @_;
1464
1465 my $pending_delete_hash = {};
1466 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1467 $pending_delete_hash->{$opt} = 1;
1468 }
1469 delete $pending_delete_hash->{$key};
1470
1471 my @keylist = keys %$pending_delete_hash;
1472 if (scalar(@keylist)) {
1473 $conf->{pending}->{delete} = join(',', @keylist);
1474 } else {
1475 delete $conf->{pending}->{delete};
1476 }
1477 }
1478
1479 sub vmconfig_register_unused_drive {
1480 my ($storecfg, $vmid, $conf, $drive) = @_;
1481
1482 if (!drive_is_cdrom($drive)) {
1483 my $volid = $drive->{file};
1484 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1485 add_unused_volume($conf, $volid, $vmid);
1486 }
1487 }
1488 }
1489
1490 sub vmconfig_cleanup_pending {
1491 my ($conf) = @_;
1492
1493 # remove pending changes when nothing changed
1494 my $changes;
1495 foreach my $opt (keys %{$conf->{pending}}) {
1496 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1497 $changes = 1;
1498 delete $conf->{pending}->{$opt};
1499 }
1500 }
1501
1502 # remove delete if option is not set
1503 my $pending_delete_hash = {};
1504 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1505 if (defined($conf->{$opt})) {
1506 $pending_delete_hash->{$opt} = 1;
1507 } else {
1508 $changes = 1;
1509 }
1510 }
1511
1512 my @keylist = keys %$pending_delete_hash;
1513 if (scalar(@keylist)) {
1514 $conf->{pending}->{delete} = join(',', @keylist);
1515 } else {
1516 delete $conf->{pending}->{delete};
1517 }
1518
1519 return $changes;
1520 }
1521
1522 my $valid_smbios1_options = {
1523 manufacturer => '\S+',
1524 product => '\S+',
1525 version => '\S+',
1526 serial => '\S+',
1527 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1528 sku => '\S+',
1529 family => '\S+',
1530 };
1531
1532 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1533 sub parse_smbios1 {
1534 my ($data) = @_;
1535
1536 my $res = {};
1537
1538 foreach my $kvp (split(/,/, $data)) {
1539 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1540 my ($k, $v) = split(/=/, $kvp);
1541 return undef if !defined($k) || !defined($v);
1542 return undef if !$valid_smbios1_options->{$k};
1543 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1544 $res->{$k} = $v;
1545 }
1546
1547 return $res;
1548 }
1549
1550 sub print_smbios1 {
1551 my ($smbios1) = @_;
1552
1553 my $data = '';
1554 foreach my $k (keys %$smbios1) {
1555 next if !defined($smbios1->{$k});
1556 next if !$valid_smbios1_options->{$k};
1557 $data .= ',' if $data;
1558 $data .= "$k=$smbios1->{$k}";
1559 }
1560 return $data;
1561 }
1562
1563 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1564 sub verify_smbios1 {
1565 my ($value, $noerr) = @_;
1566
1567 return $value if parse_smbios1($value);
1568
1569 return undef if $noerr;
1570
1571 die "unable to parse smbios (type 1) options\n";
1572 }
1573
1574 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1575 sub verify_bootdisk {
1576 my ($value, $noerr) = @_;
1577
1578 return $value if valid_drivename($value);
1579
1580 return undef if $noerr;
1581
1582 die "invalid boot disk '$value'\n";
1583 }
1584
1585 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1586 sub verify_numa {
1587 my ($value, $noerr) = @_;
1588
1589 return $value if parse_numa($value);
1590
1591 return undef if $noerr;
1592
1593 die "unable to parse numa options\n";
1594 }
1595
1596 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1597 sub verify_net {
1598 my ($value, $noerr) = @_;
1599
1600 return $value if parse_net($value);
1601
1602 return undef if $noerr;
1603
1604 die "unable to parse network options\n";
1605 }
1606
1607 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1608 sub verify_drive {
1609 my ($value, $noerr) = @_;
1610
1611 return $value if parse_drive(undef, $value);
1612
1613 return undef if $noerr;
1614
1615 die "unable to parse drive options\n";
1616 }
1617
1618 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1619 sub verify_hostpci {
1620 my ($value, $noerr) = @_;
1621
1622 return $value if parse_hostpci($value);
1623
1624 return undef if $noerr;
1625
1626 die "unable to parse pci id\n";
1627 }
1628
1629 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1630 sub verify_watchdog {
1631 my ($value, $noerr) = @_;
1632
1633 return $value if parse_watchdog($value);
1634
1635 return undef if $noerr;
1636
1637 die "unable to parse watchdog options\n";
1638 }
1639
1640 sub parse_watchdog {
1641 my ($value) = @_;
1642
1643 return undef if !$value;
1644
1645 my $res = {};
1646
1647 foreach my $p (split(/,/, $value)) {
1648 next if $p =~ m/^\s*$/;
1649
1650 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1651 $res->{model} = $2;
1652 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1653 $res->{action} = $2;
1654 } else {
1655 return undef;
1656 }
1657 }
1658
1659 return $res;
1660 }
1661
1662 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1663 sub verify_startup {
1664 my ($value, $noerr) = @_;
1665
1666 return $value if parse_startup($value);
1667
1668 return undef if $noerr;
1669
1670 die "unable to parse startup options\n";
1671 }
1672
1673 sub parse_startup {
1674 my ($value) = @_;
1675
1676 return undef if !$value;
1677
1678 my $res = {};
1679
1680 foreach my $p (split(/,/, $value)) {
1681 next if $p =~ m/^\s*$/;
1682
1683 if ($p =~ m/^(order=)?(\d+)$/) {
1684 $res->{order} = $2;
1685 } elsif ($p =~ m/^up=(\d+)$/) {
1686 $res->{up} = $1;
1687 } elsif ($p =~ m/^down=(\d+)$/) {
1688 $res->{down} = $1;
1689 } else {
1690 return undef;
1691 }
1692 }
1693
1694 return $res;
1695 }
1696
1697 sub parse_usb_device {
1698 my ($value) = @_;
1699
1700 return undef if !$value;
1701
1702 my @dl = split(/,/, $value);
1703 my $found;
1704
1705 my $res = {};
1706 foreach my $v (@dl) {
1707 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1708 $found = 1;
1709 $res->{vendorid} = $2;
1710 $res->{productid} = $4;
1711 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1712 $found = 1;
1713 $res->{hostbus} = $1;
1714 $res->{hostport} = $2;
1715 } elsif ($v =~ m/^spice$/) {
1716 $found = 1;
1717 $res->{spice} = 1;
1718 } else {
1719 return undef;
1720 }
1721 }
1722 return undef if !$found;
1723
1724 return $res;
1725 }
1726
1727 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1728 sub verify_usb_device {
1729 my ($value, $noerr) = @_;
1730
1731 return $value if parse_usb_device($value);
1732
1733 return undef if $noerr;
1734
1735 die "unable to parse usb device\n";
1736 }
1737
1738 # add JSON properties for create and set function
1739 sub json_config_properties {
1740 my $prop = shift;
1741
1742 foreach my $opt (keys %$confdesc) {
1743 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1744 $prop->{$opt} = $confdesc->{$opt};
1745 }
1746
1747 return $prop;
1748 }
1749
1750 sub check_type {
1751 my ($key, $value) = @_;
1752
1753 die "unknown setting '$key'\n" if !$confdesc->{$key};
1754
1755 my $type = $confdesc->{$key}->{type};
1756
1757 if (!defined($value)) {
1758 die "got undefined value\n";
1759 }
1760
1761 if ($value =~ m/[\n\r]/) {
1762 die "property contains a line feed\n";
1763 }
1764
1765 if ($type eq 'boolean') {
1766 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1767 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1768 die "type check ('boolean') failed - got '$value'\n";
1769 } elsif ($type eq 'integer') {
1770 return int($1) if $value =~ m/^(\d+)$/;
1771 die "type check ('integer') failed - got '$value'\n";
1772 } elsif ($type eq 'number') {
1773 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1774 die "type check ('number') failed - got '$value'\n";
1775 } elsif ($type eq 'string') {
1776 if (my $fmt = $confdesc->{$key}->{format}) {
1777 if ($fmt eq 'pve-qm-drive') {
1778 # special case - we need to pass $key to parse_drive()
1779 my $drive = parse_drive($key, $value);
1780 return $value if $drive;
1781 die "unable to parse drive options\n";
1782 }
1783 PVE::JSONSchema::check_format($fmt, $value);
1784 return $value;
1785 }
1786 $value =~ s/^\"(.*)\"$/$1/;
1787 return $value;
1788 } else {
1789 die "internal error"
1790 }
1791 }
1792
1793 sub lock_config_full {
1794 my ($vmid, $timeout, $code, @param) = @_;
1795
1796 my $filename = config_file_lock($vmid);
1797
1798 my $res = lock_file($filename, $timeout, $code, @param);
1799
1800 die $@ if $@;
1801
1802 return $res;
1803 }
1804
1805 sub lock_config_mode {
1806 my ($vmid, $timeout, $shared, $code, @param) = @_;
1807
1808 my $filename = config_file_lock($vmid);
1809
1810 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1811
1812 die $@ if $@;
1813
1814 return $res;
1815 }
1816
1817 sub lock_config {
1818 my ($vmid, $code, @param) = @_;
1819
1820 return lock_config_full($vmid, 10, $code, @param);
1821 }
1822
1823 sub cfs_config_path {
1824 my ($vmid, $node) = @_;
1825
1826 $node = $nodename if !$node;
1827 return "nodes/$node/qemu-server/$vmid.conf";
1828 }
1829
1830 sub check_iommu_support{
1831 #fixme : need to check IOMMU support
1832 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1833
1834 my $iommu=1;
1835 return $iommu;
1836
1837 }
1838
1839 sub config_file {
1840 my ($vmid, $node) = @_;
1841
1842 my $cfspath = cfs_config_path($vmid, $node);
1843 return "/etc/pve/$cfspath";
1844 }
1845
1846 sub config_file_lock {
1847 my ($vmid) = @_;
1848
1849 return "$lock_dir/lock-$vmid.conf";
1850 }
1851
1852 sub touch_config {
1853 my ($vmid) = @_;
1854
1855 my $conf = config_file($vmid);
1856 utime undef, undef, $conf;
1857 }
1858
1859 sub destroy_vm {
1860 my ($storecfg, $vmid, $keep_empty_config) = @_;
1861
1862 my $conffile = config_file($vmid);
1863
1864 my $conf = load_config($vmid);
1865
1866 check_lock($conf);
1867
1868 # only remove disks owned by this VM
1869 foreach_drive($conf, sub {
1870 my ($ds, $drive) = @_;
1871
1872 return if drive_is_cdrom($drive);
1873
1874 my $volid = $drive->{file};
1875
1876 return if !$volid || $volid =~ m|^/|;
1877
1878 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1879 return if !$path || !$owner || ($owner != $vmid);
1880
1881 PVE::Storage::vdisk_free($storecfg, $volid);
1882 });
1883
1884 if ($keep_empty_config) {
1885 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1886 } else {
1887 unlink $conffile;
1888 }
1889
1890 # also remove unused disk
1891 eval {
1892 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1893
1894 eval {
1895 PVE::Storage::foreach_volid($dl, sub {
1896 my ($volid, $sid, $volname, $d) = @_;
1897 PVE::Storage::vdisk_free($storecfg, $volid);
1898 });
1899 };
1900 warn $@ if $@;
1901
1902 };
1903 warn $@ if $@;
1904 }
1905
1906 sub load_config {
1907 my ($vmid, $node) = @_;
1908
1909 my $cfspath = cfs_config_path($vmid, $node);
1910
1911 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1912
1913 die "no such VM ('$vmid')\n" if !defined($conf);
1914
1915 return $conf;
1916 }
1917
1918 sub parse_vm_config {
1919 my ($filename, $raw) = @_;
1920
1921 return undef if !defined($raw);
1922
1923 my $res = {
1924 digest => Digest::SHA::sha1_hex($raw),
1925 snapshots => {},
1926 pending => {},
1927 };
1928
1929 $filename =~ m|/qemu-server/(\d+)\.conf$|
1930 || die "got strange filename '$filename'";
1931
1932 my $vmid = $1;
1933
1934 my $conf = $res;
1935 my $descr = '';
1936 my $section = '';
1937
1938 my @lines = split(/\n/, $raw);
1939 foreach my $line (@lines) {
1940 next if $line =~ m/^\s*$/;
1941
1942 if ($line =~ m/^\[PENDING\]\s*$/i) {
1943 $section = 'pending';
1944 $conf->{description} = $descr if $descr;
1945 $descr = '';
1946 $conf = $res->{$section} = {};
1947 next;
1948
1949 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1950 $section = $1;
1951 $conf->{description} = $descr if $descr;
1952 $descr = '';
1953 $conf = $res->{snapshots}->{$section} = {};
1954 next;
1955 }
1956
1957 if ($line =~ m/^\#(.*)\s*$/) {
1958 $descr .= PVE::Tools::decode_text($1) . "\n";
1959 next;
1960 }
1961
1962 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1963 $descr .= PVE::Tools::decode_text($2);
1964 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1965 $conf->{snapstate} = $1;
1966 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1967 my $key = $1;
1968 my $value = $2;
1969 $conf->{$key} = $value;
1970 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
1971 my $value = $1;
1972 if ($section eq 'pending') {
1973 $conf->{delete} = $value; # we parse this later
1974 } else {
1975 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
1976 }
1977 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
1978 my $key = $1;
1979 my $value = $2;
1980 eval { $value = check_type($key, $value); };
1981 if ($@) {
1982 warn "vm $vmid - unable to parse value of '$key' - $@";
1983 } else {
1984 my $fmt = $confdesc->{$key}->{format};
1985 if ($fmt && $fmt eq 'pve-qm-drive') {
1986 my $v = parse_drive($key, $value);
1987 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
1988 $v->{file} = $volid;
1989 $value = print_drive($vmid, $v);
1990 } else {
1991 warn "vm $vmid - unable to parse value of '$key'\n";
1992 next;
1993 }
1994 }
1995
1996 if ($key eq 'cdrom') {
1997 $conf->{ide2} = $value;
1998 } else {
1999 $conf->{$key} = $value;
2000 }
2001 }
2002 }
2003 }
2004
2005 $conf->{description} = $descr if $descr;
2006
2007 delete $res->{snapstate}; # just to be sure
2008
2009 return $res;
2010 }
2011
2012 sub write_vm_config {
2013 my ($filename, $conf) = @_;
2014
2015 delete $conf->{snapstate}; # just to be sure
2016
2017 if ($conf->{cdrom}) {
2018 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2019 $conf->{ide2} = $conf->{cdrom};
2020 delete $conf->{cdrom};
2021 }
2022
2023 # we do not use 'smp' any longer
2024 if ($conf->{sockets}) {
2025 delete $conf->{smp};
2026 } elsif ($conf->{smp}) {
2027 $conf->{sockets} = $conf->{smp};
2028 delete $conf->{cores};
2029 delete $conf->{smp};
2030 }
2031
2032 if ($conf->{maxcpus} && $conf->{sockets}) {
2033 delete $conf->{sockets};
2034 }
2035
2036 my $used_volids = {};
2037
2038 my $cleanup_config = sub {
2039 my ($cref, $pending, $snapname) = @_;
2040
2041 foreach my $key (keys %$cref) {
2042 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2043 $key eq 'snapstate' || $key eq 'pending';
2044 my $value = $cref->{$key};
2045 if ($key eq 'delete') {
2046 die "propertry 'delete' is only allowed in [PENDING]\n"
2047 if !$pending;
2048 # fixme: check syntax?
2049 next;
2050 }
2051 eval { $value = check_type($key, $value); };
2052 die "unable to parse value of '$key' - $@" if $@;
2053
2054 $cref->{$key} = $value;
2055
2056 if (!$snapname && valid_drivename($key)) {
2057 my $drive = parse_drive($key, $value);
2058 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2059 }
2060 }
2061 };
2062
2063 &$cleanup_config($conf);
2064
2065 &$cleanup_config($conf->{pending}, 1);
2066
2067 foreach my $snapname (keys %{$conf->{snapshots}}) {
2068 die "internal error" if $snapname eq 'pending';
2069 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2070 }
2071
2072 # remove 'unusedX' settings if we re-add a volume
2073 foreach my $key (keys %$conf) {
2074 my $value = $conf->{$key};
2075 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2076 delete $conf->{$key};
2077 }
2078 }
2079
2080 my $generate_raw_config = sub {
2081 my ($conf) = @_;
2082
2083 my $raw = '';
2084
2085 # add description as comment to top of file
2086 my $descr = $conf->{description} || '';
2087 foreach my $cl (split(/\n/, $descr)) {
2088 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2089 }
2090
2091 foreach my $key (sort keys %$conf) {
2092 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2093 $raw .= "$key: $conf->{$key}\n";
2094 }
2095 return $raw;
2096 };
2097
2098 my $raw = &$generate_raw_config($conf);
2099
2100 if (scalar(keys %{$conf->{pending}})){
2101 $raw .= "\n[PENDING]\n";
2102 $raw .= &$generate_raw_config($conf->{pending});
2103 }
2104
2105 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2106 $raw .= "\n[$snapname]\n";
2107 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2108 }
2109
2110 return $raw;
2111 }
2112
2113 sub update_config_nolock {
2114 my ($vmid, $conf, $skiplock) = @_;
2115
2116 check_lock($conf) if !$skiplock;
2117
2118 my $cfspath = cfs_config_path($vmid);
2119
2120 PVE::Cluster::cfs_write_file($cfspath, $conf);
2121 }
2122
2123 sub update_config {
2124 my ($vmid, $conf, $skiplock) = @_;
2125
2126 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2127 }
2128
2129 sub load_defaults {
2130
2131 my $res = {};
2132
2133 # we use static defaults from our JSON schema configuration
2134 foreach my $key (keys %$confdesc) {
2135 if (defined(my $default = $confdesc->{$key}->{default})) {
2136 $res->{$key} = $default;
2137 }
2138 }
2139
2140 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2141 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2142
2143 return $res;
2144 }
2145
2146 sub config_list {
2147 my $vmlist = PVE::Cluster::get_vmlist();
2148 my $res = {};
2149 return $res if !$vmlist || !$vmlist->{ids};
2150 my $ids = $vmlist->{ids};
2151
2152 foreach my $vmid (keys %$ids) {
2153 my $d = $ids->{$vmid};
2154 next if !$d->{node} || $d->{node} ne $nodename;
2155 next if !$d->{type} || $d->{type} ne 'qemu';
2156 $res->{$vmid}->{exists} = 1;
2157 }
2158 return $res;
2159 }
2160
2161 # test if VM uses local resources (to prevent migration)
2162 sub check_local_resources {
2163 my ($conf, $noerr) = @_;
2164
2165 my $loc_res = 0;
2166
2167 $loc_res = 1 if $conf->{hostusb}; # old syntax
2168 $loc_res = 1 if $conf->{hostpci}; # old syntax
2169
2170 foreach my $k (keys %$conf) {
2171 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2172 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2173 }
2174
2175 die "VM uses local resources\n" if $loc_res && !$noerr;
2176
2177 return $loc_res;
2178 }
2179
2180 # check if used storages are available on all nodes (use by migrate)
2181 sub check_storage_availability {
2182 my ($storecfg, $conf, $node) = @_;
2183
2184 foreach_drive($conf, sub {
2185 my ($ds, $drive) = @_;
2186
2187 my $volid = $drive->{file};
2188 return if !$volid;
2189
2190 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2191 return if !$sid;
2192
2193 # check if storage is available on both nodes
2194 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2195 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2196 });
2197 }
2198
2199 # list nodes where all VM images are available (used by has_feature API)
2200 sub shared_nodes {
2201 my ($conf, $storecfg) = @_;
2202
2203 my $nodelist = PVE::Cluster::get_nodelist();
2204 my $nodehash = { map { $_ => 1 } @$nodelist };
2205 my $nodename = PVE::INotify::nodename();
2206
2207 foreach_drive($conf, sub {
2208 my ($ds, $drive) = @_;
2209
2210 my $volid = $drive->{file};
2211 return if !$volid;
2212
2213 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2214 if ($storeid) {
2215 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2216 if ($scfg->{disable}) {
2217 $nodehash = {};
2218 } elsif (my $avail = $scfg->{nodes}) {
2219 foreach my $node (keys %$nodehash) {
2220 delete $nodehash->{$node} if !$avail->{$node};
2221 }
2222 } elsif (!$scfg->{shared}) {
2223 foreach my $node (keys %$nodehash) {
2224 delete $nodehash->{$node} if $node ne $nodename
2225 }
2226 }
2227 }
2228 });
2229
2230 return $nodehash
2231 }
2232
2233 sub check_lock {
2234 my ($conf) = @_;
2235
2236 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2237 }
2238
2239 sub check_cmdline {
2240 my ($pidfile, $pid) = @_;
2241
2242 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2243 if (defined($fh)) {
2244 my $line = <$fh>;
2245 $fh->close;
2246 return undef if !$line;
2247 my @param = split(/\0/, $line);
2248
2249 my $cmd = $param[0];
2250 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2251
2252 for (my $i = 0; $i < scalar (@param); $i++) {
2253 my $p = $param[$i];
2254 next if !$p;
2255 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2256 my $p = $param[$i+1];
2257 return 1 if $p && ($p eq $pidfile);
2258 return undef;
2259 }
2260 }
2261 }
2262 return undef;
2263 }
2264
2265 sub check_running {
2266 my ($vmid, $nocheck, $node) = @_;
2267
2268 my $filename = config_file($vmid, $node);
2269
2270 die "unable to find configuration file for VM $vmid - no such machine\n"
2271 if !$nocheck && ! -f $filename;
2272
2273 my $pidfile = pidfile_name($vmid);
2274
2275 if (my $fd = IO::File->new("<$pidfile")) {
2276 my $st = stat($fd);
2277 my $line = <$fd>;
2278 close($fd);
2279
2280 my $mtime = $st->mtime;
2281 if ($mtime > time()) {
2282 warn "file '$filename' modified in future\n";
2283 }
2284
2285 if ($line =~ m/^(\d+)$/) {
2286 my $pid = $1;
2287 if (check_cmdline($pidfile, $pid)) {
2288 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2289 return $pid;
2290 }
2291 }
2292 }
2293 }
2294
2295 return undef;
2296 }
2297
2298 sub vzlist {
2299
2300 my $vzlist = config_list();
2301
2302 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2303
2304 while (defined(my $de = $fd->read)) {
2305 next if $de !~ m/^(\d+)\.pid$/;
2306 my $vmid = $1;
2307 next if !defined($vzlist->{$vmid});
2308 if (my $pid = check_running($vmid)) {
2309 $vzlist->{$vmid}->{pid} = $pid;
2310 }
2311 }
2312
2313 return $vzlist;
2314 }
2315
2316 sub disksize {
2317 my ($storecfg, $conf) = @_;
2318
2319 my $bootdisk = $conf->{bootdisk};
2320 return undef if !$bootdisk;
2321 return undef if !valid_drivename($bootdisk);
2322
2323 return undef if !$conf->{$bootdisk};
2324
2325 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2326 return undef if !defined($drive);
2327
2328 return undef if drive_is_cdrom($drive);
2329
2330 my $volid = $drive->{file};
2331 return undef if !$volid;
2332
2333 return $drive->{size};
2334 }
2335
2336 my $last_proc_pid_stat;
2337
2338 # get VM status information
2339 # This must be fast and should not block ($full == false)
2340 # We only query KVM using QMP if $full == true (this can be slow)
2341 sub vmstatus {
2342 my ($opt_vmid, $full) = @_;
2343
2344 my $res = {};
2345
2346 my $storecfg = PVE::Storage::config();
2347
2348 my $list = vzlist();
2349 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2350
2351 my $cpucount = $cpuinfo->{cpus} || 1;
2352
2353 foreach my $vmid (keys %$list) {
2354 next if $opt_vmid && ($vmid ne $opt_vmid);
2355
2356 my $cfspath = cfs_config_path($vmid);
2357 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2358
2359 my $d = {};
2360 $d->{pid} = $list->{$vmid}->{pid};
2361
2362 # fixme: better status?
2363 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2364
2365 my $size = disksize($storecfg, $conf);
2366 if (defined($size)) {
2367 $d->{disk} = 0; # no info available
2368 $d->{maxdisk} = $size;
2369 } else {
2370 $d->{disk} = 0;
2371 $d->{maxdisk} = 0;
2372 }
2373
2374 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2375 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2376
2377 $d->{name} = $conf->{name} || "VM $vmid";
2378 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2379
2380 if ($conf->{balloon}) {
2381 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2382 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2383 }
2384
2385 $d->{uptime} = 0;
2386 $d->{cpu} = 0;
2387 $d->{mem} = 0;
2388
2389 $d->{netout} = 0;
2390 $d->{netin} = 0;
2391
2392 $d->{diskread} = 0;
2393 $d->{diskwrite} = 0;
2394
2395 $d->{template} = is_template($conf);
2396
2397 $res->{$vmid} = $d;
2398 }
2399
2400 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2401 foreach my $dev (keys %$netdev) {
2402 next if $dev !~ m/^tap([1-9]\d*)i/;
2403 my $vmid = $1;
2404 my $d = $res->{$vmid};
2405 next if !$d;
2406
2407 $d->{netout} += $netdev->{$dev}->{receive};
2408 $d->{netin} += $netdev->{$dev}->{transmit};
2409 }
2410
2411 my $ctime = gettimeofday;
2412
2413 foreach my $vmid (keys %$list) {
2414
2415 my $d = $res->{$vmid};
2416 my $pid = $d->{pid};
2417 next if !$pid;
2418
2419 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2420 next if !$pstat; # not running
2421
2422 my $used = $pstat->{utime} + $pstat->{stime};
2423
2424 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2425
2426 if ($pstat->{vsize}) {
2427 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2428 }
2429
2430 my $old = $last_proc_pid_stat->{$pid};
2431 if (!$old) {
2432 $last_proc_pid_stat->{$pid} = {
2433 time => $ctime,
2434 used => $used,
2435 cpu => 0,
2436 };
2437 next;
2438 }
2439
2440 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2441
2442 if ($dtime > 1000) {
2443 my $dutime = $used - $old->{used};
2444
2445 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2446 $last_proc_pid_stat->{$pid} = {
2447 time => $ctime,
2448 used => $used,
2449 cpu => $d->{cpu},
2450 };
2451 } else {
2452 $d->{cpu} = $old->{cpu};
2453 }
2454 }
2455
2456 return $res if !$full;
2457
2458 my $qmpclient = PVE::QMPClient->new();
2459
2460 my $ballooncb = sub {
2461 my ($vmid, $resp) = @_;
2462
2463 my $info = $resp->{'return'};
2464 return if !$info->{max_mem};
2465
2466 my $d = $res->{$vmid};
2467
2468 # use memory assigned to VM
2469 $d->{maxmem} = $info->{max_mem};
2470 $d->{balloon} = $info->{actual};
2471
2472 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2473 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2474 $d->{freemem} = $info->{free_mem};
2475 }
2476
2477 };
2478
2479 my $blockstatscb = sub {
2480 my ($vmid, $resp) = @_;
2481 my $data = $resp->{'return'} || [];
2482 my $totalrdbytes = 0;
2483 my $totalwrbytes = 0;
2484 for my $blockstat (@$data) {
2485 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2486 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2487 }
2488 $res->{$vmid}->{diskread} = $totalrdbytes;
2489 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2490 };
2491
2492 my $statuscb = sub {
2493 my ($vmid, $resp) = @_;
2494
2495 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2496 # this fails if ballon driver is not loaded, so this must be
2497 # the last commnand (following command are aborted if this fails).
2498 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2499
2500 my $status = 'unknown';
2501 if (!defined($status = $resp->{'return'}->{status})) {
2502 warn "unable to get VM status\n";
2503 return;
2504 }
2505
2506 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2507 };
2508
2509 foreach my $vmid (keys %$list) {
2510 next if $opt_vmid && ($vmid ne $opt_vmid);
2511 next if !$res->{$vmid}->{pid}; # not running
2512 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2513 }
2514
2515 $qmpclient->queue_execute(undef, 1);
2516
2517 foreach my $vmid (keys %$list) {
2518 next if $opt_vmid && ($vmid ne $opt_vmid);
2519 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2520 }
2521
2522 return $res;
2523 }
2524
2525 sub foreach_drive {
2526 my ($conf, $func) = @_;
2527
2528 foreach my $ds (keys %$conf) {
2529 next if !valid_drivename($ds);
2530
2531 my $drive = parse_drive($ds, $conf->{$ds});
2532 next if !$drive;
2533
2534 &$func($ds, $drive);
2535 }
2536 }
2537
2538 sub foreach_volid {
2539 my ($conf, $func) = @_;
2540
2541 my $volhash = {};
2542
2543 my $test_volid = sub {
2544 my ($volid, $is_cdrom) = @_;
2545
2546 return if !$volid;
2547
2548 $volhash->{$volid} = $is_cdrom || 0;
2549 };
2550
2551 foreach_drive($conf, sub {
2552 my ($ds, $drive) = @_;
2553 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2554 });
2555
2556 foreach my $snapname (keys %{$conf->{snapshots}}) {
2557 my $snap = $conf->{snapshots}->{$snapname};
2558 &$test_volid($snap->{vmstate}, 0);
2559 foreach_drive($snap, sub {
2560 my ($ds, $drive) = @_;
2561 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2562 });
2563 }
2564
2565 foreach my $volid (keys %$volhash) {
2566 &$func($volid, $volhash->{$volid});
2567 }
2568 }
2569
2570 sub vga_conf_has_spice {
2571 my ($vga) = @_;
2572
2573 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2574
2575 return $1 || 1;
2576 }
2577
2578 sub config_to_command {
2579 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2580
2581 my $cmd = [];
2582 my $globalFlags = [];
2583 my $machineFlags = [];
2584 my $rtcFlags = [];
2585 my $cpuFlags = [];
2586 my $devices = [];
2587 my $pciaddr = '';
2588 my $bridges = {};
2589 my $kvmver = kvm_user_version();
2590 my $vernum = 0; # unknown
2591 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2592 $vernum = $1*1000000+$2*1000;
2593 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2594 $vernum = $1*1000000+$2*1000+$3;
2595 }
2596
2597 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2598
2599 my $have_ovz = -f '/proc/vz/vestat';
2600
2601 my $q35 = machine_type_is_q35($conf);
2602
2603 push @$cmd, '/usr/bin/kvm';
2604
2605 push @$cmd, '-id', $vmid;
2606
2607 my $use_virtio = 0;
2608
2609 my $qmpsocket = qmp_socket($vmid);
2610 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2611 push @$cmd, '-mon', "chardev=qmp,mode=control";
2612
2613 my $socket = vnc_socket($vmid);
2614 push @$cmd, '-vnc', "unix:$socket,x509,password";
2615
2616 push @$cmd, '-pidfile' , pidfile_name($vmid);
2617
2618 push @$cmd, '-daemonize';
2619
2620 if ($conf->{smbios1}) {
2621 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2622 }
2623
2624 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2625
2626 if ($q35) {
2627 # the q35 chipset support native usb2, so we enable usb controller
2628 # by default for this machine type
2629 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2630 } else {
2631 $pciaddr = print_pci_addr("piix3", $bridges);
2632 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2633
2634 my $use_usb2 = 0;
2635 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2636 next if !$conf->{"usb$i"};
2637 $use_usb2 = 1;
2638 }
2639 # include usb device config
2640 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2641 }
2642
2643 my $vga = $conf->{vga};
2644
2645 my $qxlnum = vga_conf_has_spice($vga);
2646 $vga = 'qxl' if $qxlnum;
2647
2648 if (!$vga) {
2649 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2650 $conf->{ostype} eq 'win7' ||
2651 $conf->{ostype} eq 'w2k8')) {
2652 $vga = 'std';
2653 } else {
2654 $vga = 'cirrus';
2655 }
2656 }
2657
2658 # enable absolute mouse coordinates (needed by vnc)
2659 my $tablet;
2660 if (defined($conf->{tablet})) {
2661 $tablet = $conf->{tablet};
2662 } else {
2663 $tablet = $defaults->{tablet};
2664 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2665 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2666 }
2667
2668 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2669
2670 # host pci devices
2671 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2672 my $d = parse_hostpci($conf->{"hostpci$i"});
2673 next if !$d;
2674
2675 my $pcie = $d->{pcie};
2676 if($pcie){
2677 die "q35 machine model is not enabled" if !$q35;
2678 $pciaddr = print_pcie_addr("hostpci$i");
2679 }else{
2680 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2681 }
2682
2683 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2684 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2685 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2686 if ($xvga && $xvga ne '') {
2687 push @$cpuFlags, 'kvm=off';
2688 $vga = 'none';
2689 }
2690 $driver = "vfio-pci" if $xvga ne '';
2691 my $pcidevices = $d->{pciid};
2692 my $multifunction = 1 if @$pcidevices > 1;
2693
2694 my $j=0;
2695 foreach my $pcidevice (@$pcidevices) {
2696
2697 my $id = "hostpci$i";
2698 $id .= ".$j" if $multifunction;
2699 my $addr = $pciaddr;
2700 $addr .= ".$j" if $multifunction;
2701 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2702
2703 if($j == 0){
2704 $devicestr .= "$rombar$xvga";
2705 $devicestr .= ",multifunction=on" if $multifunction;
2706 }
2707
2708 push @$devices, '-device', $devicestr;
2709 $j++;
2710 }
2711 }
2712
2713 # usb devices
2714 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2715 my $d = parse_usb_device($conf->{"usb$i"});
2716 next if !$d;
2717 if ($d->{vendorid} && $d->{productid}) {
2718 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2719 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2720 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2721 } elsif ($d->{spice}) {
2722 # usb redir support for spice
2723 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2724 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2725 }
2726 }
2727
2728 # serial devices
2729 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2730 if (my $path = $conf->{"serial$i"}) {
2731 if ($path eq 'socket') {
2732 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2733 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2734 push @$devices, '-device', "isa-serial,chardev=serial$i";
2735 } else {
2736 die "no such serial device\n" if ! -c $path;
2737 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2738 push @$devices, '-device', "isa-serial,chardev=serial$i";
2739 }
2740 }
2741 }
2742
2743 # parallel devices
2744 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2745 if (my $path = $conf->{"parallel$i"}) {
2746 die "no such parallel device\n" if ! -c $path;
2747 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2748 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2749 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2750 }
2751 }
2752
2753 my $vmname = $conf->{name} || "vm$vmid";
2754
2755 push @$cmd, '-name', $vmname;
2756
2757 my $sockets = 1;
2758 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2759 $sockets = $conf->{sockets} if $conf->{sockets};
2760
2761 my $cores = $conf->{cores} || 1;
2762 my $maxcpus = $conf->{maxcpus} if $conf->{maxcpus};
2763
2764 my $total_cores = $sockets * $cores;
2765 my $allowed_cores = $cpuinfo->{cpus};
2766
2767 die "MAX $allowed_cores cores allowed per VM on this node\n"
2768 if ($allowed_cores < $total_cores);
2769
2770 if ($maxcpus) {
2771 push @$cmd, '-smp', "cpus=$cores,maxcpus=$maxcpus";
2772 } else {
2773 push @$cmd, '-smp', "sockets=$sockets,cores=$cores";
2774 }
2775
2776 push @$cmd, '-nodefaults';
2777
2778 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2779
2780 my $bootindex_hash = {};
2781 my $i = 1;
2782 foreach my $o (split(//, $bootorder)) {
2783 $bootindex_hash->{$o} = $i*100;
2784 $i++;
2785 }
2786
2787 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2788
2789 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2790
2791 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2792
2793 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2794
2795 # time drift fix
2796 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2797
2798 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2799 my $useLocaltime = $conf->{localtime};
2800
2801 if (my $ost = $conf->{ostype}) {
2802 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2803
2804 if ($ost =~ m/^w/) { # windows
2805 $useLocaltime = 1 if !defined($conf->{localtime});
2806
2807 # use time drift fix when acpi is enabled
2808 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2809 $tdf = 1 if !defined($conf->{tdf});
2810 }
2811 }
2812
2813 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2814 $ost eq 'wvista') {
2815 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2816 push @$cmd, '-no-hpet';
2817 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2818 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2819 }
2820
2821 if ($ost eq 'win7' || $ost eq 'win8') {
2822 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2823 }
2824 }
2825
2826 push @$rtcFlags, 'driftfix=slew' if $tdf;
2827
2828 if ($nokvm) {
2829 push @$machineFlags, 'accel=tcg';
2830 } else {
2831 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2832 }
2833
2834 my $machine_type = $forcemachine || $conf->{machine};
2835 if ($machine_type) {
2836 push @$machineFlags, "type=${machine_type}";
2837 }
2838
2839 if ($conf->{startdate}) {
2840 push @$rtcFlags, "base=$conf->{startdate}";
2841 } elsif ($useLocaltime) {
2842 push @$rtcFlags, 'base=localtime';
2843 }
2844
2845 my $cpu = $nokvm ? "qemu64" : "kvm64";
2846 $cpu = $conf->{cpu} if $conf->{cpu};
2847
2848 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2849
2850 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2851
2852 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2853
2854 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2855
2856 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2857
2858 # Note: enforce needs kernel 3.10, so we do not use it for now
2859 # push @$cmd, '-cpu', "$cpu,enforce";
2860 push @$cmd, '-cpu', $cpu;
2861
2862 my $memory = $conf->{memory} || $defaults->{memory};
2863 push @$cmd, '-m', $memory;
2864
2865 if ($conf->{numa}) {
2866
2867 my $numa_totalmemory = undef;
2868 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2869 next if !$conf->{"numa$i"};
2870 my $numa = parse_numa($conf->{"numa$i"});
2871 next if !$numa;
2872 # memory
2873 die "missing numa node$i memory value\n" if !$numa->{memory};
2874 my $numa_memory = $numa->{memory};
2875 $numa_totalmemory += $numa_memory;
2876 my $numa_object = "memory-backend-ram,id=ram-node$i,size=$numa_memory"."M";
2877
2878 # cpus
2879 my $cpus_start = $numa->{cpus}->{start};
2880 die "missing numa node$i cpus\n" if !defined($cpus_start);
2881 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2882 my $cpus = $cpus_start;
2883 if (defined($cpus_end)) {
2884 $cpus .= "-$cpus_end";
2885 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2886 }
2887
2888 # hostnodes
2889 my $hostnodes_start = $numa->{hostnodes}->{start};
2890 if (defined($hostnodes_start)) {
2891 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2892 my $hostnodes = $hostnodes_start;
2893 if (defined($hostnodes_end)) {
2894 $hostnodes .= "-$hostnodes_end";
2895 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2896 }
2897
2898 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2899 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2900 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2901 }
2902
2903 # policy
2904 my $policy = $numa->{policy};
2905 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2906 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2907 }
2908
2909 push @$cmd, '-object', $numa_object;
2910 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2911 }
2912
2913 die "total memory for NUMA nodes must be equal to vm memory\n"
2914 if $numa_totalmemory && $numa_totalmemory != $memory;
2915
2916 #if no custom tology, we split memory and cores across numa nodes
2917 if(!$numa_totalmemory) {
2918
2919 my $numa_memory = ($memory / $sockets) . "M";
2920
2921 for (my $i = 0; $i < $sockets; $i++) {
2922
2923 my $cpustart = ($cores * $i);
2924 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2925 my $cpus = $cpustart;
2926 $cpus .= "-$cpuend" if $cpuend;
2927
2928 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2929 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2930 }
2931 }
2932 }
2933
2934 push @$cmd, '-S' if $conf->{freeze};
2935
2936 # set keyboard layout
2937 my $kb = $conf->{keyboard} || $defaults->{keyboard};
2938 push @$cmd, '-k', $kb if $kb;
2939
2940 # enable sound
2941 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
2942 #push @$cmd, '-soundhw', 'es1370';
2943 #push @$cmd, '-soundhw', $soundhw if $soundhw;
2944
2945 if($conf->{agent}) {
2946 my $qgasocket = qmp_socket($vmid, 1);
2947 my $pciaddr = print_pci_addr("qga0", $bridges);
2948 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
2949 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
2950 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
2951 }
2952
2953 my $spice_port;
2954
2955 if ($qxlnum) {
2956 if ($qxlnum > 1) {
2957 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
2958 for(my $i = 1; $i < $qxlnum; $i++){
2959 my $pciaddr = print_pci_addr("vga$i", $bridges);
2960 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
2961 }
2962 } else {
2963 # assume other OS works like Linux
2964 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
2965 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
2966 }
2967 }
2968
2969 my $pciaddr = print_pci_addr("spice", $bridges);
2970
2971 $spice_port = PVE::Tools::next_spice_port();
2972
2973 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
2974
2975 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
2976 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
2977 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
2978 }
2979
2980 # enable balloon by default, unless explicitly disabled
2981 if (!defined($conf->{balloon}) || $conf->{balloon}) {
2982 $pciaddr = print_pci_addr("balloon0", $bridges);
2983 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
2984 }
2985
2986 if ($conf->{watchdog}) {
2987 my $wdopts = parse_watchdog($conf->{watchdog});
2988 $pciaddr = print_pci_addr("watchdog", $bridges);
2989 my $watchdog = $wdopts->{model} || 'i6300esb';
2990 push @$devices, '-device', "$watchdog$pciaddr";
2991 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
2992 }
2993
2994 my $vollist = [];
2995 my $scsicontroller = {};
2996 my $ahcicontroller = {};
2997 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
2998
2999 # Add iscsi initiator name if available
3000 if (my $initiator = get_initiator_name()) {
3001 push @$devices, '-iscsi', "initiator-name=$initiator";
3002 }
3003
3004 foreach_drive($conf, sub {
3005 my ($ds, $drive) = @_;
3006
3007 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3008 push @$vollist, $drive->{file};
3009 }
3010
3011 $use_virtio = 1 if $ds =~ m/^virtio/;
3012
3013 if (drive_is_cdrom ($drive)) {
3014 if ($bootindex_hash->{d}) {
3015 $drive->{bootindex} = $bootindex_hash->{d};
3016 $bootindex_hash->{d} += 1;
3017 }
3018 } else {
3019 if ($bootindex_hash->{c}) {
3020 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3021 $bootindex_hash->{c} += 1;
3022 }
3023 }
3024
3025 if ($drive->{interface} eq 'scsi') {
3026
3027 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3028 my $controller = int($drive->{index} / $maxdev);
3029 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3030 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3031 $scsicontroller->{$controller}=1;
3032 }
3033
3034 if ($drive->{interface} eq 'sata') {
3035 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3036 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3037 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3038 $ahcicontroller->{$controller}=1;
3039 }
3040
3041 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3042 push @$devices, '-drive',$drive_cmd;
3043 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3044 });
3045
3046 for (my $i = 0; $i < $MAX_NETS; $i++) {
3047 next if !$conf->{"net$i"};
3048 my $d = parse_net($conf->{"net$i"});
3049 next if !$d;
3050
3051 $use_virtio = 1 if $d->{model} eq 'virtio';
3052
3053 if ($bootindex_hash->{n}) {
3054 $d->{bootindex} = $bootindex_hash->{n};
3055 $bootindex_hash->{n} += 1;
3056 }
3057
3058 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3059 push @$devices, '-netdev', $netdevfull;
3060
3061 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3062 push @$devices, '-device', $netdevicefull;
3063 }
3064
3065 if (!$q35) {
3066 # add pci bridges
3067 while (my ($k, $v) = each %$bridges) {
3068 $pciaddr = print_pci_addr("pci.$k");
3069 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3070 }
3071 }
3072
3073 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3074 # when the VM uses virtio devices.
3075 if (!$use_virtio && $have_ovz) {
3076
3077 my $cpuunits = defined($conf->{cpuunits}) ?
3078 $conf->{cpuunits} : $defaults->{cpuunits};
3079
3080 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3081
3082 # fixme: cpulimit is currently ignored
3083 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3084 }
3085
3086 # add custom args
3087 if ($conf->{args}) {
3088 my $aa = PVE::Tools::split_args($conf->{args});
3089 push @$cmd, @$aa;
3090 }
3091
3092 push @$cmd, @$devices;
3093 push @$cmd, '-rtc', join(',', @$rtcFlags)
3094 if scalar(@$rtcFlags);
3095 push @$cmd, '-machine', join(',', @$machineFlags)
3096 if scalar(@$machineFlags);
3097 push @$cmd, '-global', join(',', @$globalFlags)
3098 if scalar(@$globalFlags);
3099
3100 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3101 }
3102
3103 sub vnc_socket {
3104 my ($vmid) = @_;
3105 return "${var_run_tmpdir}/$vmid.vnc";
3106 }
3107
3108 sub spice_port {
3109 my ($vmid) = @_;
3110
3111 my $res = vm_mon_cmd($vmid, 'query-spice');
3112
3113 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3114 }
3115
3116 sub qmp_socket {
3117 my ($vmid, $qga) = @_;
3118 my $sockettype = $qga ? 'qga' : 'qmp';
3119 return "${var_run_tmpdir}/$vmid.$sockettype";
3120 }
3121
3122 sub pidfile_name {
3123 my ($vmid) = @_;
3124 return "${var_run_tmpdir}/$vmid.pid";
3125 }
3126
3127 sub vm_devices_list {
3128 my ($vmid) = @_;
3129
3130 my $res = vm_mon_cmd($vmid, 'query-pci');
3131 my $devices = {};
3132 foreach my $pcibus (@$res) {
3133 foreach my $device (@{$pcibus->{devices}}) {
3134 next if !$device->{'qdev_id'};
3135 $devices->{$device->{'qdev_id'}} = 1;
3136 }
3137 }
3138
3139 my $resblock = vm_mon_cmd($vmid, 'query-block');
3140 foreach my $block (@$resblock) {
3141 if($block->{device} =~ m/^drive-(\S+)/){
3142 $devices->{$1} = 1;
3143 }
3144 }
3145
3146 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3147 foreach my $mice (@$resmice) {
3148 if ($mice->{name} eq 'QEMU HID Tablet') {
3149 $devices->{tablet} = 1;
3150 last;
3151 }
3152 }
3153
3154 return $devices;
3155 }
3156
3157 sub hotplug_enabled {
3158 my ($conf) = @_;
3159
3160 my $default = $confdesc->{'hotplug'}->{default};
3161
3162 return defined($conf->{hotplug}) ? $conf->{hotplug} : $default;
3163 }
3164
3165 sub vm_deviceplug {
3166 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3167
3168 die "internal error" if !hotplug_enabled($conf);
3169
3170 my $q35 = machine_type_is_q35($conf);
3171
3172 my $devices_list = vm_devices_list($vmid);
3173 return 1 if defined($devices_list->{$deviceid});
3174
3175 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3176
3177 if ($deviceid eq 'tablet') {
3178
3179 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3180
3181 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3182
3183 qemu_driveadd($storecfg, $vmid, $device);
3184 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3185
3186 qemu_deviceadd($vmid, $devicefull);
3187 eval { qemu_deviceaddverify($vmid, $deviceid); };
3188 if (my $err = $@) {
3189 eval { qemu_drivedel($vmid, $deviceid); };
3190 warn $@ if $@;
3191 die $err;
3192 }
3193
3194 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3195
3196 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3197 my $pciaddr = print_pci_addr($deviceid);
3198 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3199
3200 qemu_deviceadd($vmid, $devicefull);
3201 qemu_deviceaddverify($vmid, $deviceid);
3202
3203 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3204
3205 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3206 qemu_driveadd($storecfg, $vmid, $device);
3207
3208 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3209 eval { qemu_deviceadd($vmid, $devicefull); };
3210 if (my $err = $@) {
3211 eval { qemu_drivedel($vmid, $deviceid); };
3212 warn $@ if $@;
3213 die $err;
3214 }
3215
3216 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3217
3218 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3219 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3220 qemu_deviceadd($vmid, $netdevicefull);
3221 eval { qemu_deviceaddverify($vmid, $deviceid); };
3222 if (my $err = $@) {
3223 eval { qemu_netdevdel($vmid, $deviceid); };
3224 warn $@ if $@;
3225 die $err;
3226 }
3227
3228 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3229
3230 my $bridgeid = $2;
3231 my $pciaddr = print_pci_addr($deviceid);
3232 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3233
3234 qemu_deviceadd($vmid, $devicefull);
3235 qemu_deviceaddverify($vmid, $deviceid);
3236
3237 } else {
3238 die "can't hotplug device '$deviceid'\n";
3239 }
3240
3241 return 1;
3242 }
3243
3244 # fixme: this should raise exceptions on error!
3245 sub vm_deviceunplug {
3246 my ($vmid, $conf, $deviceid) = @_;
3247
3248 die "internal error" if !hotplug_enabled($conf);
3249
3250 my $devices_list = vm_devices_list($vmid);
3251 return 1 if !defined($devices_list->{$deviceid});
3252
3253 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3254
3255 if ($deviceid eq 'tablet') {
3256
3257 qemu_devicedel($vmid, $deviceid);
3258
3259 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3260
3261 qemu_devicedel($vmid, $deviceid);
3262 qemu_devicedelverify($vmid, $deviceid);
3263 qemu_drivedel($vmid, $deviceid);
3264
3265 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3266
3267 qemu_devicedel($vmid, $deviceid);
3268
3269 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3270
3271 qemu_devicedel($vmid, $deviceid);
3272 qemu_drivedel($vmid, $deviceid);
3273
3274 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3275
3276 qemu_devicedel($vmid, $deviceid);
3277 qemu_devicedelverify($vmid, $deviceid);
3278 qemu_netdevdel($vmid, $deviceid);
3279
3280 } else {
3281 die "can't unplug device '$deviceid'\n";
3282 }
3283
3284 return 1;
3285 }
3286
3287 sub qemu_deviceadd {
3288 my ($vmid, $devicefull) = @_;
3289
3290 $devicefull = "driver=".$devicefull;
3291 my %options = split(/[=,]/, $devicefull);
3292
3293 vm_mon_cmd($vmid, "device_add" , %options);
3294 }
3295
3296 sub qemu_devicedel {
3297 my ($vmid, $deviceid) = @_;
3298
3299 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3300 }
3301
3302 sub qemu_driveadd {
3303 my ($storecfg, $vmid, $device) = @_;
3304
3305 my $drive = print_drive_full($storecfg, $vmid, $device);
3306 my $ret = vm_human_monitor_command($vmid, "drive_add auto $drive");
3307
3308 # If the command succeeds qemu prints: "OK"
3309 return 1 if $ret =~ m/OK/s;
3310
3311 die "adding drive failed: $ret\n";
3312 }
3313
3314 sub qemu_drivedel {
3315 my($vmid, $deviceid) = @_;
3316
3317 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3318 $ret =~ s/^\s+//;
3319
3320 return 1 if $ret eq "";
3321
3322 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3323 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3324
3325 die "deleting drive $deviceid failed : $ret\n";
3326 }
3327
3328 sub qemu_deviceaddverify {
3329 my ($vmid, $deviceid) = @_;
3330
3331 for (my $i = 0; $i <= 5; $i++) {
3332 my $devices_list = vm_devices_list($vmid);
3333 return 1 if defined($devices_list->{$deviceid});
3334 sleep 1;
3335 }
3336
3337 die "error on hotplug device '$deviceid'\n";
3338 }
3339
3340
3341 sub qemu_devicedelverify {
3342 my ($vmid, $deviceid) = @_;
3343
3344 # need to verify that the device is correctly removed as device_del
3345 # is async and empty return is not reliable
3346
3347 for (my $i = 0; $i <= 5; $i++) {
3348 my $devices_list = vm_devices_list($vmid);
3349 return 1 if !defined($devices_list->{$deviceid});
3350 sleep 1;
3351 }
3352
3353 die "error on hot-unplugging device '$deviceid'\n";
3354 }
3355
3356 sub qemu_findorcreatescsihw {
3357 my ($storecfg, $conf, $vmid, $device) = @_;
3358
3359 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3360 my $controller = int($device->{index} / $maxdev);
3361 my $scsihwid="scsihw$controller";
3362 my $devices_list = vm_devices_list($vmid);
3363
3364 if(!defined($devices_list->{$scsihwid})) {
3365 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3366 }
3367
3368 return 1;
3369 }
3370
3371 sub qemu_add_pci_bridge {
3372 my ($storecfg, $conf, $vmid, $device) = @_;
3373
3374 my $bridges = {};
3375
3376 my $bridgeid;
3377
3378 print_pci_addr($device, $bridges);
3379
3380 while (my ($k, $v) = each %$bridges) {
3381 $bridgeid = $k;
3382 }
3383 return 1 if !defined($bridgeid) || $bridgeid < 1;
3384
3385 my $bridge = "pci.$bridgeid";
3386 my $devices_list = vm_devices_list($vmid);
3387
3388 if (!defined($devices_list->{$bridge})) {
3389 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3390 }
3391
3392 return 1;
3393 }
3394
3395 sub qemu_set_link_status {
3396 my ($vmid, $device, $up) = @_;
3397
3398 vm_mon_cmd($vmid, "set_link", name => $device,
3399 up => $up ? JSON::true : JSON::false);
3400 }
3401
3402 sub qemu_netdevadd {
3403 my ($vmid, $conf, $device, $deviceid) = @_;
3404
3405 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3406 my %options = split(/[=,]/, $netdev);
3407
3408 vm_mon_cmd($vmid, "netdev_add", %options);
3409 return 1;
3410 }
3411
3412 sub qemu_netdevdel {
3413 my ($vmid, $deviceid) = @_;
3414
3415 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3416 }
3417
3418 sub qemu_cpu_hotplug {
3419 my ($vmid, $conf, $cores) = @_;
3420
3421 my $sockets = $conf->{sockets} || 1;
3422 die "cpu hotplug only works with one socket\n"
3423 if $sockets > 1;
3424
3425 die "maxcpus is not defined\n"
3426 if !$conf->{maxcpus};
3427
3428 die "you can't add more cores than maxcpus\n"
3429 if $cores > $conf->{maxcpus};
3430
3431 my $currentcores = $conf->{cores} || 1;
3432 die "online cpu unplug is not yet possible\n"
3433 if $cores < $currentcores;
3434
3435 my $currentrunningcores = vm_mon_cmd($vmid, "query-cpus");
3436 die "cores number if running vm is different than configuration\n"
3437 if scalar(@{$currentrunningcores}) != $currentcores;
3438
3439 for (my $i = $currentcores; $i < $cores; $i++) {
3440 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3441 }
3442 }
3443
3444 sub qemu_block_set_io_throttle {
3445 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3446
3447 return if !check_running($vmid) ;
3448
3449 vm_mon_cmd($vmid, "block_set_io_throttle", device => $deviceid, bps => int($bps), bps_rd => int($bps_rd), bps_wr => int($bps_wr), iops => int($iops), iops_rd => int($iops_rd), iops_wr => int($iops_wr));
3450
3451 }
3452
3453 # old code, only used to shutdown old VM after update
3454 sub __read_avail {
3455 my ($fh, $timeout) = @_;
3456
3457 my $sel = new IO::Select;
3458 $sel->add($fh);
3459
3460 my $res = '';
3461 my $buf;
3462
3463 my @ready;
3464 while (scalar (@ready = $sel->can_read($timeout))) {
3465 my $count;
3466 if ($count = $fh->sysread($buf, 8192)) {
3467 if ($buf =~ /^(.*)\(qemu\) $/s) {
3468 $res .= $1;
3469 last;
3470 } else {
3471 $res .= $buf;
3472 }
3473 } else {
3474 if (!defined($count)) {
3475 die "$!\n";
3476 }
3477 last;
3478 }
3479 }
3480
3481 die "monitor read timeout\n" if !scalar(@ready);
3482
3483 return $res;
3484 }
3485
3486 # old code, only used to shutdown old VM after update
3487 sub vm_monitor_command {
3488 my ($vmid, $cmdstr, $nocheck) = @_;
3489
3490 my $res;
3491
3492 eval {
3493 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3494
3495 my $sname = "${var_run_tmpdir}/$vmid.mon";
3496
3497 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3498 die "unable to connect to VM $vmid socket - $!\n";
3499
3500 my $timeout = 3;
3501
3502 # hack: migrate sometime blocks the monitor (when migrate_downtime
3503 # is set)
3504 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3505 $timeout = 60*60; # 1 hour
3506 }
3507
3508 # read banner;
3509 my $data = __read_avail($sock, $timeout);
3510
3511 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3512 die "got unexpected qemu monitor banner\n";
3513 }
3514
3515 my $sel = new IO::Select;
3516 $sel->add($sock);
3517
3518 if (!scalar(my @ready = $sel->can_write($timeout))) {
3519 die "monitor write error - timeout";
3520 }
3521
3522 my $fullcmd = "$cmdstr\r";
3523
3524 # syslog('info', "VM $vmid monitor command: $cmdstr");
3525
3526 my $b;
3527 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3528 die "monitor write error - $!";
3529 }
3530
3531 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3532
3533 $timeout = 20;
3534
3535 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3536 $timeout = 60*60; # 1 hour
3537 } elsif ($cmdstr =~ m/^(eject|change)/) {
3538 $timeout = 60; # note: cdrom mount command is slow
3539 }
3540 if ($res = __read_avail($sock, $timeout)) {
3541
3542 my @lines = split("\r?\n", $res);
3543
3544 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3545
3546 $res = join("\n", @lines);
3547 $res .= "\n";
3548 }
3549 };
3550
3551 my $err = $@;
3552
3553 if ($err) {
3554 syslog("err", "VM $vmid monitor command failed - $err");
3555 die $err;
3556 }
3557
3558 return $res;
3559 }
3560
3561 sub qemu_block_resize {
3562 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3563
3564 my $running = check_running($vmid);
3565
3566 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3567
3568 return if !$running;
3569
3570 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3571
3572 }
3573
3574 sub qemu_volume_snapshot {
3575 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3576
3577 my $running = check_running($vmid);
3578
3579 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3580
3581 return if !$running;
3582
3583 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3584
3585 }
3586
3587 sub qemu_volume_snapshot_delete {
3588 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3589
3590 my $running = check_running($vmid);
3591
3592 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3593
3594 return if !$running;
3595
3596 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3597 }
3598
3599 sub set_migration_caps {
3600 my ($vmid) = @_;
3601
3602 my $cap_ref = [];
3603
3604 my $enabled_cap = {
3605 "auto-converge" => 1,
3606 "xbzrle" => 0,
3607 "x-rdma-pin-all" => 0,
3608 "zero-blocks" => 0,
3609 };
3610
3611 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3612
3613 for my $supported_capability (@$supported_capabilities) {
3614 push @$cap_ref, {
3615 capability => $supported_capability->{capability},
3616 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3617 };
3618 }
3619
3620 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3621 }
3622
3623 my $fast_plug_option = {
3624 'name' => 1,
3625 'hotplug' => 1,
3626 'onboot' => 1,
3627 'shares' => 1,
3628 'startup' => 1,
3629 };
3630
3631 # hotplug changes in [PENDING]
3632 # $selection hash can be used to only apply specified options, for
3633 # example: { cores => 1 } (only apply changed 'cores')
3634 # $errors ref is used to return error messages
3635 sub vmconfig_hotplug_pending {
3636 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3637
3638 my $defaults = load_defaults();
3639
3640 # commit values which do not have any impact on running VM first
3641 # Note: those option cannot raise errors, we we do not care about
3642 # $selection and always apply them.
3643
3644 my $add_error = sub {
3645 my ($opt, $msg) = @_;
3646 $errors->{$opt} = "hotplug problem - $msg";
3647 };
3648
3649 my $changes = 0;
3650 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3651 if ($fast_plug_option->{$opt}) {
3652 $conf->{$opt} = $conf->{pending}->{$opt};
3653 delete $conf->{pending}->{$opt};
3654 $changes = 1;
3655 }
3656 }
3657
3658 if ($changes) {
3659 update_config_nolock($vmid, $conf, 1);
3660 $conf = load_config($vmid); # update/reload
3661 }
3662
3663 my $hotplug = hotplug_enabled($conf);
3664
3665 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3666 foreach my $opt (@delete) {
3667 next if $selection && !$selection->{$opt};
3668 eval {
3669 if ($opt eq 'tablet') {
3670 die "skip\n" if !$hotplug;
3671 if ($defaults->{tablet}) {
3672 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3673 } else {
3674 vm_deviceunplug($vmid, $conf, $opt);
3675 }
3676 } elsif ($opt eq 'cores') {
3677 die "skip\n" if !$hotplug;
3678 qemu_cpu_hotplug($vmid, $conf, 1);
3679 } elsif ($opt eq 'balloon') {
3680 # enable balloon device is not hotpluggable
3681 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3682 } elsif ($fast_plug_option->{$opt}) {
3683 # do nothing
3684 } elsif ($opt =~ m/^net(\d+)$/) {
3685 die "skip\n" if !$hotplug;
3686 vm_deviceunplug($vmid, $conf, $opt);
3687 } elsif (valid_drivename($opt)) {
3688 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
3689 vm_deviceunplug($vmid, $conf, $opt);
3690 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3691 } else {
3692 die "skip\n";
3693 }
3694 };
3695 if (my $err = $@) {
3696 &$add_error($opt, $err) if $err ne "skip\n";
3697 } else {
3698 # save new config if hotplug was successful
3699 delete $conf->{$opt};
3700 vmconfig_undelete_pending_option($conf, $opt);
3701 update_config_nolock($vmid, $conf, 1);
3702 $conf = load_config($vmid); # update/reload
3703 }
3704 }
3705
3706 foreach my $opt (keys %{$conf->{pending}}) {
3707 next if $selection && !$selection->{$opt};
3708 my $value = $conf->{pending}->{$opt};
3709 eval {
3710 if ($opt eq 'tablet') {
3711 die "skip\n" if !$hotplug;
3712 if ($value == 1) {
3713 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3714 } elsif ($value == 0) {
3715 vm_deviceunplug($vmid, $conf, $opt);
3716 }
3717 } elsif ($opt eq 'cores') {
3718 die "skip\n" if !$hotplug;
3719 qemu_cpu_hotplug($vmid, $conf, $value);
3720 } elsif ($opt eq 'balloon') {
3721 # enable/disable balloning device is not hotpluggable
3722 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3723 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3724 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3725
3726 # allow manual ballooning if shares is set to zero
3727 if (!(defined($conf->{shares}) && ($conf->{shares} == 0))) {
3728 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3729 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3730 }
3731 } elsif ($opt =~ m/^net(\d+)$/) {
3732 # some changes can be done without hotplug
3733 vmconfig_update_net($storecfg, $conf, $vmid, $opt, $value);
3734 } elsif (valid_drivename($opt)) {
3735 # some changes can be done without hotplug
3736 vmconfig_update_disk($storecfg, $conf, $vmid, $opt, $value, 1);
3737 } else {
3738 die "skip\n"; # skip non-hot-pluggable options
3739 }
3740 };
3741 if (my $err = $@) {
3742 &$add_error($opt, $err) if $err ne "skip\n";
3743 } else {
3744 # save new config if hotplug was successful
3745 $conf->{$opt} = $value;
3746 delete $conf->{pending}->{$opt};
3747 update_config_nolock($vmid, $conf, 1);
3748 $conf = load_config($vmid); # update/reload
3749 }
3750 }
3751 }
3752
3753 sub vmconfig_apply_pending {
3754 my ($vmid, $conf, $storecfg) = @_;
3755
3756 # cold plug
3757
3758 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3759 foreach my $opt (@delete) { # delete
3760 die "internal error" if $opt =~ m/^unused/;
3761 $conf = load_config($vmid); # update/reload
3762 if (!defined($conf->{$opt})) {
3763 vmconfig_undelete_pending_option($conf, $opt);
3764 update_config_nolock($vmid, $conf, 1);
3765 } elsif (valid_drivename($opt)) {
3766 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3767 vmconfig_undelete_pending_option($conf, $opt);
3768 delete $conf->{$opt};
3769 update_config_nolock($vmid, $conf, 1);
3770 } else {
3771 vmconfig_undelete_pending_option($conf, $opt);
3772 delete $conf->{$opt};
3773 update_config_nolock($vmid, $conf, 1);
3774 }
3775 }
3776
3777 $conf = load_config($vmid); # update/reload
3778
3779 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3780 $conf = load_config($vmid); # update/reload
3781
3782 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3783 # skip if nothing changed
3784 } elsif (valid_drivename($opt)) {
3785 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3786 if defined($conf->{$opt});
3787 $conf->{$opt} = $conf->{pending}->{$opt};
3788 } else {
3789 $conf->{$opt} = $conf->{pending}->{$opt};
3790 }
3791
3792 delete $conf->{pending}->{$opt};
3793 update_config_nolock($vmid, $conf, 1);
3794 }
3795 }
3796
3797 my $safe_num_ne = sub {
3798 my ($a, $b) = @_;
3799
3800 return 0 if !defined($a) && !defined($b);
3801 return 1 if !defined($a);
3802 return 1 if !defined($b);
3803
3804 return $a != $b;
3805 };
3806
3807 my $safe_string_ne = sub {
3808 my ($a, $b) = @_;
3809
3810 return 0 if !defined($a) && !defined($b);
3811 return 1 if !defined($a);
3812 return 1 if !defined($b);
3813
3814 return $a ne $b;
3815 };
3816
3817 sub vmconfig_update_net {
3818 my ($storecfg, $conf, $vmid, $opt, $value) = @_;
3819
3820 my $newnet = parse_net($value);
3821
3822 my $hotplug = hotplug_enabled($conf);
3823
3824 if ($conf->{$opt}) {
3825 my $oldnet = parse_net($conf->{$opt});
3826
3827 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3828 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3829 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3830 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3831
3832 # for non online change, we try to hot-unplug
3833 die "skip\n" if !$hotplug;
3834 vm_deviceunplug($vmid, $conf, $opt);
3835 } else {
3836
3837 die "internal error" if $opt !~ m/net(\d+)/;
3838 my $iface = "tap${vmid}i$1";
3839
3840 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3841 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3842 }
3843
3844 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3845 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3846 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3847 PVE::Network::tap_unplug($iface);
3848 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3849 }
3850
3851 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3852 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3853 }
3854
3855 return 1;
3856 }
3857 }
3858
3859 if ($hotplug) {
3860 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
3861 } else {
3862 die "skip\n";
3863 }
3864 }
3865
3866 sub vmconfig_update_disk {
3867 my ($storecfg, $conf, $vmid, $opt, $value, $force) = @_;
3868
3869 # fixme: do we need force?
3870
3871 my $drive = parse_drive($opt, $value);
3872
3873 my $hotplug = hotplug_enabled($conf);
3874
3875 if ($conf->{$opt}) {
3876
3877 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
3878
3879 my $media = $drive->{media} || 'disk';
3880 my $oldmedia = $old_drive->{media} || 'disk';
3881 die "unable to change media type\n" if $media ne $oldmedia;
3882
3883 if (!drive_is_cdrom($old_drive)) {
3884
3885 if ($drive->{file} ne $old_drive->{file}) {
3886
3887 die "skip\n" if !$hotplug;
3888
3889 # unplug and register as unused
3890 vm_deviceunplug($vmid, $conf, $opt);
3891 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
3892
3893 } else {
3894 # update existing disk
3895
3896 # skip non hotpluggable value
3897 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
3898 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
3899 die "skip\n";
3900 }
3901
3902 # apply throttle
3903 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
3904 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
3905 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
3906 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
3907 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
3908 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
3909 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
3910 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
3911 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
3912 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
3913 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
3914 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
3915
3916 qemu_block_set_io_throttle($vmid,"drive-$opt",
3917 ($drive->{mbps} || 0)*1024*1024,
3918 ($drive->{mbps_rd} || 0)*1024*1024,
3919 ($drive->{mbps_wr} || 0)*1024*1024,
3920 $drive->{iops} || 0,
3921 $drive->{iops_rd} || 0,
3922 $drive->{iops_wr} || 0,
3923 ($drive->{mbps_max} || 0)*1024*1024,
3924 ($drive->{mbps_rd_max} || 0)*1024*1024,
3925 ($drive->{mbps_wr_max} || 0)*1024*1024,
3926 $drive->{iops_max} || 0,
3927 $drive->{iops_rd_max} || 0,
3928 $drive->{iops_wr_max} || 0);
3929
3930 }
3931
3932 return 1;
3933 }
3934 }
3935 }
3936 }
3937
3938 if (drive_is_cdrom($drive)) { # cdrom
3939
3940 if ($drive->{file} eq 'none') {
3941 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
3942 } else {
3943 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
3944 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
3945 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
3946 }
3947
3948 } else {
3949 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
3950 # hotplug new disks
3951 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
3952 }
3953 }
3954
3955 sub vm_start {
3956 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
3957
3958 lock_config($vmid, sub {
3959 my $conf = load_config($vmid, $migratedfrom);
3960
3961 die "you can't start a vm if it's a template\n" if is_template($conf);
3962
3963 check_lock($conf) if !$skiplock;
3964
3965 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
3966
3967 if (!$statefile && scalar(keys %{$conf->{pending}})) {
3968 vmconfig_apply_pending($vmid, $conf, $storecfg);
3969 $conf = load_config($vmid); # update/reload
3970 }
3971
3972 my $defaults = load_defaults();
3973
3974 # set environment variable useful inside network script
3975 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
3976
3977 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
3978
3979 my $migrate_port = 0;
3980 my $migrate_uri;
3981 if ($statefile) {
3982 if ($statefile eq 'tcp') {
3983 my $localip = "localhost";
3984 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
3985 if ($datacenterconf->{migration_unsecure}) {
3986 my $nodename = PVE::INotify::nodename();
3987 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
3988 }
3989 $migrate_port = PVE::Tools::next_migrate_port();
3990 $migrate_uri = "tcp:${localip}:${migrate_port}";
3991 push @$cmd, '-incoming', $migrate_uri;
3992 push @$cmd, '-S';
3993 } else {
3994 push @$cmd, '-loadstate', $statefile;
3995 }
3996 } elsif ($paused) {
3997 push @$cmd, '-S';
3998 }
3999
4000 # host pci devices
4001 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4002 my $d = parse_hostpci($conf->{"hostpci$i"});
4003 next if !$d;
4004 my $pcidevices = $d->{pciid};
4005 foreach my $pcidevice (@$pcidevices) {
4006 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4007
4008 my $info = pci_device_info("0000:$pciid");
4009 die "IOMMU not present\n" if !check_iommu_support();
4010 die "no pci device info for device '$pciid'\n" if !$info;
4011
4012 if ($d->{driver} && $d->{driver} eq "vfio") {
4013 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4014 } else {
4015 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4016 }
4017
4018 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4019 }
4020 }
4021
4022 PVE::Storage::activate_volumes($storecfg, $vollist);
4023
4024 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4025 umask => 0077); };
4026 my $err = $@;
4027 die "start failed: $err" if $err;
4028
4029 print "migration listens on $migrate_uri\n" if $migrate_uri;
4030
4031 if ($statefile && $statefile ne 'tcp') {
4032 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4033 warn $@ if $@;
4034 }
4035
4036 if ($migratedfrom) {
4037
4038 eval {
4039 set_migration_caps($vmid);
4040 };
4041 warn $@ if $@;
4042
4043 if ($spice_port) {
4044 print "spice listens on port $spice_port\n";
4045 if ($spice_ticket) {
4046 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4047 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4048 }
4049 }
4050
4051 } else {
4052
4053 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4054 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4055 if $conf->{balloon};
4056 vm_mon_cmd_nocheck($vmid, 'qom-set',
4057 path => "machine/peripheral/balloon0",
4058 property => "guest-stats-polling-interval",
4059 value => 2);
4060 }
4061
4062 foreach my $opt (keys %$conf) {
4063 next if $opt !~ m/^net\d+$/;
4064 my $nicconf = parse_net($conf->{$opt});
4065 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4066 }
4067 }
4068 });
4069 }
4070
4071 sub vm_mon_cmd {
4072 my ($vmid, $execute, %params) = @_;
4073
4074 my $cmd = { execute => $execute, arguments => \%params };
4075 vm_qmp_command($vmid, $cmd);
4076 }
4077
4078 sub vm_mon_cmd_nocheck {
4079 my ($vmid, $execute, %params) = @_;
4080
4081 my $cmd = { execute => $execute, arguments => \%params };
4082 vm_qmp_command($vmid, $cmd, 1);
4083 }
4084
4085 sub vm_qmp_command {
4086 my ($vmid, $cmd, $nocheck) = @_;
4087
4088 my $res;
4089
4090 my $timeout;
4091 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4092 $timeout = $cmd->{arguments}->{timeout};
4093 delete $cmd->{arguments}->{timeout};
4094 }
4095
4096 eval {
4097 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4098 my $sname = qmp_socket($vmid);
4099 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4100 my $qmpclient = PVE::QMPClient->new();
4101
4102 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4103 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4104 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4105 if scalar(%{$cmd->{arguments}});
4106 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4107 } else {
4108 die "unable to open monitor socket\n";
4109 }
4110 };
4111 if (my $err = $@) {
4112 syslog("err", "VM $vmid qmp command failed - $err");
4113 die $err;
4114 }
4115
4116 return $res;
4117 }
4118
4119 sub vm_human_monitor_command {
4120 my ($vmid, $cmdline) = @_;
4121
4122 my $res;
4123
4124 my $cmd = {
4125 execute => 'human-monitor-command',
4126 arguments => { 'command-line' => $cmdline},
4127 };
4128
4129 return vm_qmp_command($vmid, $cmd);
4130 }
4131
4132 sub vm_commandline {
4133 my ($storecfg, $vmid) = @_;
4134
4135 my $conf = load_config($vmid);
4136
4137 my $defaults = load_defaults();
4138
4139 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4140
4141 return join(' ', @$cmd);
4142 }
4143
4144 sub vm_reset {
4145 my ($vmid, $skiplock) = @_;
4146
4147 lock_config($vmid, sub {
4148
4149 my $conf = load_config($vmid);
4150
4151 check_lock($conf) if !$skiplock;
4152
4153 vm_mon_cmd($vmid, "system_reset");
4154 });
4155 }
4156
4157 sub get_vm_volumes {
4158 my ($conf) = @_;
4159
4160 my $vollist = [];
4161 foreach_volid($conf, sub {
4162 my ($volid, $is_cdrom) = @_;
4163
4164 return if $volid =~ m|^/|;
4165
4166 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4167 return if !$sid;
4168
4169 push @$vollist, $volid;
4170 });
4171
4172 return $vollist;
4173 }
4174
4175 sub vm_stop_cleanup {
4176 my ($storecfg, $vmid, $conf, $keepActive) = @_;
4177
4178 eval {
4179 fairsched_rmnod($vmid); # try to destroy group
4180
4181 if (!$keepActive) {
4182 my $vollist = get_vm_volumes($conf);
4183 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4184 }
4185
4186 foreach my $ext (qw(mon qmp pid vnc qga)) {
4187 unlink "/var/run/qemu-server/${vmid}.$ext";
4188 }
4189 };
4190 warn $@ if $@; # avoid errors - just warn
4191 }
4192
4193 # Note: use $nockeck to skip tests if VM configuration file exists.
4194 # We need that when migration VMs to other nodes (files already moved)
4195 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4196 sub vm_stop {
4197 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4198
4199 $force = 1 if !defined($force) && !$shutdown;
4200
4201 if ($migratedfrom){
4202 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4203 kill 15, $pid if $pid;
4204 my $conf = load_config($vmid, $migratedfrom);
4205 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive);
4206 return;
4207 }
4208
4209 lock_config($vmid, sub {
4210
4211 my $pid = check_running($vmid, $nocheck);
4212 return if !$pid;
4213
4214 my $conf;
4215 if (!$nocheck) {
4216 $conf = load_config($vmid);
4217 check_lock($conf) if !$skiplock;
4218 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4219 my $opts = parse_startup($conf->{startup});
4220 $timeout = $opts->{down} if $opts->{down};
4221 }
4222 }
4223
4224 $timeout = 60 if !defined($timeout);
4225
4226 eval {
4227 if ($shutdown) {
4228 if (defined($conf) && $conf->{agent}) {
4229 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4230 } else {
4231 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4232 }
4233 } else {
4234 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4235 }
4236 };
4237 my $err = $@;
4238
4239 if (!$err) {
4240 my $count = 0;
4241 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4242 $count++;
4243 sleep 1;
4244 }
4245
4246 if ($count >= $timeout) {
4247 if ($force) {
4248 warn "VM still running - terminating now with SIGTERM\n";
4249 kill 15, $pid;
4250 } else {
4251 die "VM quit/powerdown failed - got timeout\n";
4252 }
4253 } else {
4254 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
4255 return;
4256 }
4257 } else {
4258 if ($force) {
4259 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4260 kill 15, $pid;
4261 } else {
4262 die "VM quit/powerdown failed\n";
4263 }
4264 }
4265
4266 # wait again
4267 $timeout = 10;
4268
4269 my $count = 0;
4270 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4271 $count++;
4272 sleep 1;
4273 }
4274
4275 if ($count >= $timeout) {
4276 warn "VM still running - terminating now with SIGKILL\n";
4277 kill 9, $pid;
4278 sleep 1;
4279 }
4280
4281 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
4282 });
4283 }
4284
4285 sub vm_suspend {
4286 my ($vmid, $skiplock) = @_;
4287
4288 lock_config($vmid, sub {
4289
4290 my $conf = load_config($vmid);
4291
4292 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4293
4294 vm_mon_cmd($vmid, "stop");
4295 });
4296 }
4297
4298 sub vm_resume {
4299 my ($vmid, $skiplock) = @_;
4300
4301 lock_config($vmid, sub {
4302
4303 my $conf = load_config($vmid);
4304
4305 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4306
4307 vm_mon_cmd($vmid, "cont");
4308 });
4309 }
4310
4311 sub vm_sendkey {
4312 my ($vmid, $skiplock, $key) = @_;
4313
4314 lock_config($vmid, sub {
4315
4316 my $conf = load_config($vmid);
4317
4318 # there is no qmp command, so we use the human monitor command
4319 vm_human_monitor_command($vmid, "sendkey $key");
4320 });
4321 }
4322
4323 sub vm_destroy {
4324 my ($storecfg, $vmid, $skiplock) = @_;
4325
4326 lock_config($vmid, sub {
4327
4328 my $conf = load_config($vmid);
4329
4330 check_lock($conf) if !$skiplock;
4331
4332 if (!check_running($vmid)) {
4333 fairsched_rmnod($vmid); # try to destroy group
4334 destroy_vm($storecfg, $vmid);
4335 } else {
4336 die "VM $vmid is running - destroy failed\n";
4337 }
4338 });
4339 }
4340
4341 # pci helpers
4342
4343 sub file_write {
4344 my ($filename, $buf) = @_;
4345
4346 my $fh = IO::File->new($filename, "w");
4347 return undef if !$fh;
4348
4349 my $res = print $fh $buf;
4350
4351 $fh->close();
4352
4353 return $res;
4354 }
4355
4356 sub pci_device_info {
4357 my ($name) = @_;
4358
4359 my $res;
4360
4361 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4362 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4363
4364 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4365 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4366
4367 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4368 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4369
4370 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4371 return undef if !defined($product) || $product !~ s/^0x//;
4372
4373 $res = {
4374 name => $name,
4375 vendor => $vendor,
4376 product => $product,
4377 domain => $domain,
4378 bus => $bus,
4379 slot => $slot,
4380 func => $func,
4381 irq => $irq,
4382 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4383 };
4384
4385 return $res;
4386 }
4387
4388 sub pci_dev_reset {
4389 my ($dev) = @_;
4390
4391 my $name = $dev->{name};
4392
4393 my $fn = "$pcisysfs/devices/$name/reset";
4394
4395 return file_write($fn, "1");
4396 }
4397
4398 sub pci_dev_bind_to_stub {
4399 my ($dev) = @_;
4400
4401 my $name = $dev->{name};
4402
4403 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4404 return 1 if -d $testdir;
4405
4406 my $data = "$dev->{vendor} $dev->{product}";
4407 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4408
4409 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4410 if (!file_write($fn, $name)) {
4411 return undef if -f $fn;
4412 }
4413
4414 $fn = "$pcisysfs/drivers/pci-stub/bind";
4415 if (! -d $testdir) {
4416 return undef if !file_write($fn, $name);
4417 }
4418
4419 return -d $testdir;
4420 }
4421
4422 sub pci_dev_bind_to_vfio {
4423 my ($dev) = @_;
4424
4425 my $name = $dev->{name};
4426
4427 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4428
4429 if (!-d $vfio_basedir) {
4430 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4431 }
4432 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4433
4434 my $testdir = "$vfio_basedir/$name";
4435 return 1 if -d $testdir;
4436
4437 my $data = "$dev->{vendor} $dev->{product}";
4438 return undef if !file_write("$vfio_basedir/new_id", $data);
4439
4440 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4441 if (!file_write($fn, $name)) {
4442 return undef if -f $fn;
4443 }
4444
4445 $fn = "$vfio_basedir/bind";
4446 if (! -d $testdir) {
4447 return undef if !file_write($fn, $name);
4448 }
4449
4450 return -d $testdir;
4451 }
4452
4453 sub pci_dev_group_bind_to_vfio {
4454 my ($pciid) = @_;
4455
4456 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4457
4458 if (!-d $vfio_basedir) {
4459 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4460 }
4461 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4462
4463 # get IOMMU group devices
4464 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4465 my @devs = grep /^0000:/, readdir($D);
4466 closedir($D);
4467
4468 foreach my $pciid (@devs) {
4469 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4470 my $info = pci_device_info($1);
4471 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4472 }
4473
4474 return 1;
4475 }
4476
4477 sub print_pci_addr {
4478 my ($id, $bridges) = @_;
4479
4480 my $res = '';
4481 my $devices = {
4482 piix3 => { bus => 0, addr => 1 },
4483 #addr2 : first videocard
4484 balloon0 => { bus => 0, addr => 3 },
4485 watchdog => { bus => 0, addr => 4 },
4486 scsihw0 => { bus => 0, addr => 5 },
4487 scsihw1 => { bus => 0, addr => 6 },
4488 ahci0 => { bus => 0, addr => 7 },
4489 qga0 => { bus => 0, addr => 8 },
4490 spice => { bus => 0, addr => 9 },
4491 virtio0 => { bus => 0, addr => 10 },
4492 virtio1 => { bus => 0, addr => 11 },
4493 virtio2 => { bus => 0, addr => 12 },
4494 virtio3 => { bus => 0, addr => 13 },
4495 virtio4 => { bus => 0, addr => 14 },
4496 virtio5 => { bus => 0, addr => 15 },
4497 hostpci0 => { bus => 0, addr => 16 },
4498 hostpci1 => { bus => 0, addr => 17 },
4499 net0 => { bus => 0, addr => 18 },
4500 net1 => { bus => 0, addr => 19 },
4501 net2 => { bus => 0, addr => 20 },
4502 net3 => { bus => 0, addr => 21 },
4503 net4 => { bus => 0, addr => 22 },
4504 net5 => { bus => 0, addr => 23 },
4505 vga1 => { bus => 0, addr => 24 },
4506 vga2 => { bus => 0, addr => 25 },
4507 vga3 => { bus => 0, addr => 26 },
4508 hostpci2 => { bus => 0, addr => 27 },
4509 hostpci3 => { bus => 0, addr => 28 },
4510 #addr29 : usb-host (pve-usb.cfg)
4511 'pci.1' => { bus => 0, addr => 30 },
4512 'pci.2' => { bus => 0, addr => 31 },
4513 'net6' => { bus => 1, addr => 1 },
4514 'net7' => { bus => 1, addr => 2 },
4515 'net8' => { bus => 1, addr => 3 },
4516 'net9' => { bus => 1, addr => 4 },
4517 'net10' => { bus => 1, addr => 5 },
4518 'net11' => { bus => 1, addr => 6 },
4519 'net12' => { bus => 1, addr => 7 },
4520 'net13' => { bus => 1, addr => 8 },
4521 'net14' => { bus => 1, addr => 9 },
4522 'net15' => { bus => 1, addr => 10 },
4523 'net16' => { bus => 1, addr => 11 },
4524 'net17' => { bus => 1, addr => 12 },
4525 'net18' => { bus => 1, addr => 13 },
4526 'net19' => { bus => 1, addr => 14 },
4527 'net20' => { bus => 1, addr => 15 },
4528 'net21' => { bus => 1, addr => 16 },
4529 'net22' => { bus => 1, addr => 17 },
4530 'net23' => { bus => 1, addr => 18 },
4531 'net24' => { bus => 1, addr => 19 },
4532 'net25' => { bus => 1, addr => 20 },
4533 'net26' => { bus => 1, addr => 21 },
4534 'net27' => { bus => 1, addr => 22 },
4535 'net28' => { bus => 1, addr => 23 },
4536 'net29' => { bus => 1, addr => 24 },
4537 'net30' => { bus => 1, addr => 25 },
4538 'net31' => { bus => 1, addr => 26 },
4539 'virtio6' => { bus => 2, addr => 1 },
4540 'virtio7' => { bus => 2, addr => 2 },
4541 'virtio8' => { bus => 2, addr => 3 },
4542 'virtio9' => { bus => 2, addr => 4 },
4543 'virtio10' => { bus => 2, addr => 5 },
4544 'virtio11' => { bus => 2, addr => 6 },
4545 'virtio12' => { bus => 2, addr => 7 },
4546 'virtio13' => { bus => 2, addr => 8 },
4547 'virtio14' => { bus => 2, addr => 9 },
4548 'virtio15' => { bus => 2, addr => 10 },
4549 };
4550
4551 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4552 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4553 my $bus = $devices->{$id}->{bus};
4554 $res = ",bus=pci.$bus,addr=$addr";
4555 $bridges->{$bus} = 1 if $bridges;
4556 }
4557 return $res;
4558
4559 }
4560
4561 sub print_pcie_addr {
4562 my ($id) = @_;
4563
4564 my $res = '';
4565 my $devices = {
4566 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4567 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4568 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4569 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4570 };
4571
4572 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4573 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4574 my $bus = $devices->{$id}->{bus};
4575 $res = ",bus=$bus,addr=$addr";
4576 }
4577 return $res;
4578
4579 }
4580
4581 # vzdump restore implementaion
4582
4583 sub tar_archive_read_firstfile {
4584 my $archive = shift;
4585
4586 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4587
4588 # try to detect archive type first
4589 my $pid = open (TMP, "tar tf '$archive'|") ||
4590 die "unable to open file '$archive'\n";
4591 my $firstfile = <TMP>;
4592 kill 15, $pid;
4593 close TMP;
4594
4595 die "ERROR: archive contaions no data\n" if !$firstfile;
4596 chomp $firstfile;
4597
4598 return $firstfile;
4599 }
4600
4601 sub tar_restore_cleanup {
4602 my ($storecfg, $statfile) = @_;
4603
4604 print STDERR "starting cleanup\n";
4605
4606 if (my $fd = IO::File->new($statfile, "r")) {
4607 while (defined(my $line = <$fd>)) {
4608 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4609 my $volid = $2;
4610 eval {
4611 if ($volid =~ m|^/|) {
4612 unlink $volid || die 'unlink failed\n';
4613 } else {
4614 PVE::Storage::vdisk_free($storecfg, $volid);
4615 }
4616 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4617 };
4618 print STDERR "unable to cleanup '$volid' - $@" if $@;
4619 } else {
4620 print STDERR "unable to parse line in statfile - $line";
4621 }
4622 }
4623 $fd->close();
4624 }
4625 }
4626
4627 sub restore_archive {
4628 my ($archive, $vmid, $user, $opts) = @_;
4629
4630 my $format = $opts->{format};
4631 my $comp;
4632
4633 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4634 $format = 'tar' if !$format;
4635 $comp = 'gzip';
4636 } elsif ($archive =~ m/\.tar$/) {
4637 $format = 'tar' if !$format;
4638 } elsif ($archive =~ m/.tar.lzo$/) {
4639 $format = 'tar' if !$format;
4640 $comp = 'lzop';
4641 } elsif ($archive =~ m/\.vma$/) {
4642 $format = 'vma' if !$format;
4643 } elsif ($archive =~ m/\.vma\.gz$/) {
4644 $format = 'vma' if !$format;
4645 $comp = 'gzip';
4646 } elsif ($archive =~ m/\.vma\.lzo$/) {
4647 $format = 'vma' if !$format;
4648 $comp = 'lzop';
4649 } else {
4650 $format = 'vma' if !$format; # default
4651 }
4652
4653 # try to detect archive format
4654 if ($format eq 'tar') {
4655 return restore_tar_archive($archive, $vmid, $user, $opts);
4656 } else {
4657 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4658 }
4659 }
4660
4661 sub restore_update_config_line {
4662 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4663
4664 return if $line =~ m/^\#qmdump\#/;
4665 return if $line =~ m/^\#vzdump\#/;
4666 return if $line =~ m/^lock:/;
4667 return if $line =~ m/^unused\d+:/;
4668 return if $line =~ m/^parent:/;
4669 return if $line =~ m/^template:/; # restored VM is never a template
4670
4671 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4672 # try to convert old 1.X settings
4673 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4674 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4675 my ($model, $macaddr) = split(/\=/, $devconfig);
4676 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4677 my $net = {
4678 model => $model,
4679 bridge => "vmbr$ind",
4680 macaddr => $macaddr,
4681 };
4682 my $netstr = print_net($net);
4683
4684 print $outfd "net$cookie->{netcount}: $netstr\n";
4685 $cookie->{netcount}++;
4686 }
4687 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4688 my ($id, $netstr) = ($1, $2);
4689 my $net = parse_net($netstr);
4690 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4691 $netstr = print_net($net);
4692 print $outfd "$id: $netstr\n";
4693 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4694 my $virtdev = $1;
4695 my $value = $3;
4696 if ($line =~ m/backup=no/) {
4697 print $outfd "#$line";
4698 } elsif ($virtdev && $map->{$virtdev}) {
4699 my $di = parse_drive($virtdev, $value);
4700 delete $di->{format}; # format can change on restore
4701 $di->{file} = $map->{$virtdev};
4702 $value = print_drive($vmid, $di);
4703 print $outfd "$virtdev: $value\n";
4704 } else {
4705 print $outfd $line;
4706 }
4707 } else {
4708 print $outfd $line;
4709 }
4710 }
4711
4712 sub scan_volids {
4713 my ($cfg, $vmid) = @_;
4714
4715 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4716
4717 my $volid_hash = {};
4718 foreach my $storeid (keys %$info) {
4719 foreach my $item (@{$info->{$storeid}}) {
4720 next if !($item->{volid} && $item->{size});
4721 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4722 $volid_hash->{$item->{volid}} = $item;
4723 }
4724 }
4725
4726 return $volid_hash;
4727 }
4728
4729 sub get_used_paths {
4730 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4731
4732 my $used_path = {};
4733
4734 my $scan_config = sub {
4735 my ($cref, $snapname) = @_;
4736
4737 foreach my $key (keys %$cref) {
4738 my $value = $cref->{$key};
4739 if (valid_drivename($key)) {
4740 next if $skip_drive && $key eq $skip_drive;
4741 my $drive = parse_drive($key, $value);
4742 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4743 if ($drive->{file} =~ m!^/!) {
4744 $used_path->{$drive->{file}}++; # = 1;
4745 } else {
4746 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4747 next if !$storeid;
4748 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4749 next if !$scfg;
4750 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4751 $used_path->{$path}++; # = 1;
4752 }
4753 }
4754 }
4755 };
4756
4757 &$scan_config($conf);
4758
4759 undef $skip_drive;
4760
4761 if ($scan_snapshots) {
4762 foreach my $snapname (keys %{$conf->{snapshots}}) {
4763 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4764 }
4765 }
4766
4767 return $used_path;
4768 }
4769
4770 sub update_disksize {
4771 my ($vmid, $conf, $volid_hash) = @_;
4772
4773 my $changes;
4774
4775 my $used = {};
4776
4777 # Note: it is allowed to define multiple storages with same path (alias), so
4778 # we need to check both 'volid' and real 'path' (two different volid can point
4779 # to the same path).
4780
4781 my $usedpath = {};
4782
4783 # update size info
4784 foreach my $opt (keys %$conf) {
4785 if (valid_drivename($opt)) {
4786 my $drive = parse_drive($opt, $conf->{$opt});
4787 my $volid = $drive->{file};
4788 next if !$volid;
4789
4790 $used->{$volid} = 1;
4791 if ($volid_hash->{$volid} &&
4792 (my $path = $volid_hash->{$volid}->{path})) {
4793 $usedpath->{$path} = 1;
4794 }
4795
4796 next if drive_is_cdrom($drive);
4797 next if !$volid_hash->{$volid};
4798
4799 $drive->{size} = $volid_hash->{$volid}->{size};
4800 my $new = print_drive($vmid, $drive);
4801 if ($new ne $conf->{$opt}) {
4802 $changes = 1;
4803 $conf->{$opt} = $new;
4804 }
4805 }
4806 }
4807
4808 # remove 'unusedX' entry if volume is used
4809 foreach my $opt (keys %$conf) {
4810 next if $opt !~ m/^unused\d+$/;
4811 my $volid = $conf->{$opt};
4812 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4813 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4814 $changes = 1;
4815 delete $conf->{$opt};
4816 }
4817 }
4818
4819 foreach my $volid (sort keys %$volid_hash) {
4820 next if $volid =~ m/vm-$vmid-state-/;
4821 next if $used->{$volid};
4822 my $path = $volid_hash->{$volid}->{path};
4823 next if !$path; # just to be sure
4824 next if $usedpath->{$path};
4825 $changes = 1;
4826 add_unused_volume($conf, $volid);
4827 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4828 }
4829
4830 return $changes;
4831 }
4832
4833 sub rescan {
4834 my ($vmid, $nolock) = @_;
4835
4836 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4837
4838 my $volid_hash = scan_volids($cfg, $vmid);
4839
4840 my $updatefn = sub {
4841 my ($vmid) = @_;
4842
4843 my $conf = load_config($vmid);
4844
4845 check_lock($conf);
4846
4847 my $vm_volids = {};
4848 foreach my $volid (keys %$volid_hash) {
4849 my $info = $volid_hash->{$volid};
4850 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
4851 }
4852
4853 my $changes = update_disksize($vmid, $conf, $vm_volids);
4854
4855 update_config_nolock($vmid, $conf, 1) if $changes;
4856 };
4857
4858 if (defined($vmid)) {
4859 if ($nolock) {
4860 &$updatefn($vmid);
4861 } else {
4862 lock_config($vmid, $updatefn, $vmid);
4863 }
4864 } else {
4865 my $vmlist = config_list();
4866 foreach my $vmid (keys %$vmlist) {
4867 if ($nolock) {
4868 &$updatefn($vmid);
4869 } else {
4870 lock_config($vmid, $updatefn, $vmid);
4871 }
4872 }
4873 }
4874 }
4875
4876 sub restore_vma_archive {
4877 my ($archive, $vmid, $user, $opts, $comp) = @_;
4878
4879 my $input = $archive eq '-' ? "<&STDIN" : undef;
4880 my $readfrom = $archive;
4881
4882 my $uncomp = '';
4883 if ($comp) {
4884 $readfrom = '-';
4885 my $qarchive = PVE::Tools::shellquote($archive);
4886 if ($comp eq 'gzip') {
4887 $uncomp = "zcat $qarchive|";
4888 } elsif ($comp eq 'lzop') {
4889 $uncomp = "lzop -d -c $qarchive|";
4890 } else {
4891 die "unknown compression method '$comp'\n";
4892 }
4893
4894 }
4895
4896 my $tmpdir = "/var/tmp/vzdumptmp$$";
4897 rmtree $tmpdir;
4898
4899 # disable interrupts (always do cleanups)
4900 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
4901 warn "got interrupt - ignored\n";
4902 };
4903
4904 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
4905 POSIX::mkfifo($mapfifo, 0600);
4906 my $fifofh;
4907
4908 my $openfifo = sub {
4909 open($fifofh, '>', $mapfifo) || die $!;
4910 };
4911
4912 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
4913
4914 my $oldtimeout;
4915 my $timeout = 5;
4916
4917 my $devinfo = {};
4918
4919 my $rpcenv = PVE::RPCEnvironment::get();
4920
4921 my $conffile = config_file($vmid);
4922 my $tmpfn = "$conffile.$$.tmp";
4923
4924 # Note: $oldconf is undef if VM does not exists
4925 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
4926
4927 my $print_devmap = sub {
4928 my $virtdev_hash = {};
4929
4930 my $cfgfn = "$tmpdir/qemu-server.conf";
4931
4932 # we can read the config - that is already extracted
4933 my $fh = IO::File->new($cfgfn, "r") ||
4934 "unable to read qemu-server.conf - $!\n";
4935
4936 while (defined(my $line = <$fh>)) {
4937 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
4938 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
4939 die "archive does not contain data for drive '$virtdev'\n"
4940 if !$devinfo->{$devname};
4941 if (defined($opts->{storage})) {
4942 $storeid = $opts->{storage} || 'local';
4943 } elsif (!$storeid) {
4944 $storeid = 'local';
4945 }
4946 $format = 'raw' if !$format;
4947 $devinfo->{$devname}->{devname} = $devname;
4948 $devinfo->{$devname}->{virtdev} = $virtdev;
4949 $devinfo->{$devname}->{format} = $format;
4950 $devinfo->{$devname}->{storeid} = $storeid;
4951
4952 # check permission on storage
4953 my $pool = $opts->{pool}; # todo: do we need that?
4954 if ($user ne 'root@pam') {
4955 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
4956 }
4957
4958 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
4959 }
4960 }
4961
4962 foreach my $devname (keys %$devinfo) {
4963 die "found no device mapping information for device '$devname'\n"
4964 if !$devinfo->{$devname}->{virtdev};
4965 }
4966
4967 my $cfg = cfs_read_file('storage.cfg');
4968
4969 # create empty/temp config
4970 if ($oldconf) {
4971 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
4972 foreach_drive($oldconf, sub {
4973 my ($ds, $drive) = @_;
4974
4975 return if drive_is_cdrom($drive);
4976
4977 my $volid = $drive->{file};
4978
4979 return if !$volid || $volid =~ m|^/|;
4980
4981 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
4982 return if !$path || !$owner || ($owner != $vmid);
4983
4984 # Note: only delete disk we want to restore
4985 # other volumes will become unused
4986 if ($virtdev_hash->{$ds}) {
4987 PVE::Storage::vdisk_free($cfg, $volid);
4988 }
4989 });
4990 }
4991
4992 my $map = {};
4993 foreach my $virtdev (sort keys %$virtdev_hash) {
4994 my $d = $virtdev_hash->{$virtdev};
4995 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
4996 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
4997
4998 # test if requested format is supported
4999 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5000 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5001 $d->{format} = $defFormat if !$supported;
5002
5003 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5004 $d->{format}, undef, $alloc_size);
5005 print STDERR "new volume ID is '$volid'\n";
5006 $d->{volid} = $volid;
5007 my $path = PVE::Storage::path($cfg, $volid);
5008
5009 my $write_zeros = 1;
5010 # fixme: what other storages types initialize volumes with zero?
5011 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5012 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5013 $write_zeros = 0;
5014 }
5015
5016 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5017
5018 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5019 $map->{$virtdev} = $volid;
5020 }
5021
5022 $fh->seek(0, 0) || die "seek failed - $!\n";
5023
5024 my $outfd = new IO::File ($tmpfn, "w") ||
5025 die "unable to write config for VM $vmid\n";
5026
5027 my $cookie = { netcount => 0 };
5028 while (defined(my $line = <$fh>)) {
5029 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5030 }
5031
5032 $fh->close();
5033 $outfd->close();
5034 };
5035
5036 eval {
5037 # enable interrupts
5038 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5039 die "interrupted by signal\n";
5040 };
5041 local $SIG{ALRM} = sub { die "got timeout\n"; };
5042
5043 $oldtimeout = alarm($timeout);
5044
5045 my $parser = sub {
5046 my $line = shift;
5047
5048 print "$line\n";
5049
5050 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5051 my ($dev_id, $size, $devname) = ($1, $2, $3);
5052 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5053 } elsif ($line =~ m/^CTIME: /) {
5054 # we correctly received the vma config, so we can disable
5055 # the timeout now for disk allocation (set to 10 minutes, so
5056 # that we always timeout if something goes wrong)
5057 alarm(600);
5058 &$print_devmap();
5059 print $fifofh "done\n";
5060 my $tmp = $oldtimeout || 0;
5061 $oldtimeout = undef;
5062 alarm($tmp);
5063 close($fifofh);
5064 }
5065 };
5066
5067 print "restore vma archive: $cmd\n";
5068 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5069 };
5070 my $err = $@;
5071
5072 alarm($oldtimeout) if $oldtimeout;
5073
5074 unlink $mapfifo;
5075
5076 if ($err) {
5077 rmtree $tmpdir;
5078 unlink $tmpfn;
5079
5080 my $cfg = cfs_read_file('storage.cfg');
5081 foreach my $devname (keys %$devinfo) {
5082 my $volid = $devinfo->{$devname}->{volid};
5083 next if !$volid;
5084 eval {
5085 if ($volid =~ m|^/|) {
5086 unlink $volid || die 'unlink failed\n';
5087 } else {
5088 PVE::Storage::vdisk_free($cfg, $volid);
5089 }
5090 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5091 };
5092 print STDERR "unable to cleanup '$volid' - $@" if $@;
5093 }
5094 die $err;
5095 }
5096
5097 rmtree $tmpdir;
5098
5099 rename($tmpfn, $conffile) ||
5100 die "unable to commit configuration file '$conffile'\n";
5101
5102 PVE::Cluster::cfs_update(); # make sure we read new file
5103
5104 eval { rescan($vmid, 1); };
5105 warn $@ if $@;
5106 }
5107
5108 sub restore_tar_archive {
5109 my ($archive, $vmid, $user, $opts) = @_;
5110
5111 if ($archive ne '-') {
5112 my $firstfile = tar_archive_read_firstfile($archive);
5113 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5114 if $firstfile ne 'qemu-server.conf';
5115 }
5116
5117 my $storecfg = cfs_read_file('storage.cfg');
5118
5119 # destroy existing data - keep empty config
5120 my $vmcfgfn = config_file($vmid);
5121 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5122
5123 my $tocmd = "/usr/lib/qemu-server/qmextract";
5124
5125 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5126 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5127 $tocmd .= ' --prealloc' if $opts->{prealloc};
5128 $tocmd .= ' --info' if $opts->{info};
5129
5130 # tar option "xf" does not autodetect compression when read from STDIN,
5131 # so we pipe to zcat
5132 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5133 PVE::Tools::shellquote("--to-command=$tocmd");
5134
5135 my $tmpdir = "/var/tmp/vzdumptmp$$";
5136 mkpath $tmpdir;
5137
5138 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5139 local $ENV{VZDUMP_VMID} = $vmid;
5140 local $ENV{VZDUMP_USER} = $user;
5141
5142 my $conffile = config_file($vmid);
5143 my $tmpfn = "$conffile.$$.tmp";
5144
5145 # disable interrupts (always do cleanups)
5146 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5147 print STDERR "got interrupt - ignored\n";
5148 };
5149
5150 eval {
5151 # enable interrupts
5152 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5153 die "interrupted by signal\n";
5154 };
5155
5156 if ($archive eq '-') {
5157 print "extracting archive from STDIN\n";
5158 run_command($cmd, input => "<&STDIN");
5159 } else {
5160 print "extracting archive '$archive'\n";
5161 run_command($cmd);
5162 }
5163
5164 return if $opts->{info};
5165
5166 # read new mapping
5167 my $map = {};
5168 my $statfile = "$tmpdir/qmrestore.stat";
5169 if (my $fd = IO::File->new($statfile, "r")) {
5170 while (defined (my $line = <$fd>)) {
5171 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5172 $map->{$1} = $2 if $1;
5173 } else {
5174 print STDERR "unable to parse line in statfile - $line\n";
5175 }
5176 }
5177 $fd->close();
5178 }
5179
5180 my $confsrc = "$tmpdir/qemu-server.conf";
5181
5182 my $srcfd = new IO::File($confsrc, "r") ||
5183 die "unable to open file '$confsrc'\n";
5184
5185 my $outfd = new IO::File ($tmpfn, "w") ||
5186 die "unable to write config for VM $vmid\n";
5187
5188 my $cookie = { netcount => 0 };
5189 while (defined (my $line = <$srcfd>)) {
5190 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5191 }
5192
5193 $srcfd->close();
5194 $outfd->close();
5195 };
5196 my $err = $@;
5197
5198 if ($err) {
5199
5200 unlink $tmpfn;
5201
5202 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5203
5204 die $err;
5205 }
5206
5207 rmtree $tmpdir;
5208
5209 rename $tmpfn, $conffile ||
5210 die "unable to commit configuration file '$conffile'\n";
5211
5212 PVE::Cluster::cfs_update(); # make sure we read new file
5213
5214 eval { rescan($vmid, 1); };
5215 warn $@ if $@;
5216 };
5217
5218
5219 # Internal snapshots
5220
5221 # NOTE: Snapshot create/delete involves several non-atomic
5222 # action, and can take a long time.
5223 # So we try to avoid locking the file and use 'lock' variable
5224 # inside the config file instead.
5225
5226 my $snapshot_copy_config = sub {
5227 my ($source, $dest) = @_;
5228
5229 foreach my $k (keys %$source) {
5230 next if $k eq 'snapshots';
5231 next if $k eq 'snapstate';
5232 next if $k eq 'snaptime';
5233 next if $k eq 'vmstate';
5234 next if $k eq 'lock';
5235 next if $k eq 'digest';
5236 next if $k eq 'description';
5237 next if $k =~ m/^unused\d+$/;
5238
5239 $dest->{$k} = $source->{$k};
5240 }
5241 };
5242
5243 my $snapshot_apply_config = sub {
5244 my ($conf, $snap) = @_;
5245
5246 # copy snapshot list
5247 my $newconf = {
5248 snapshots => $conf->{snapshots},
5249 };
5250
5251 # keep description and list of unused disks
5252 foreach my $k (keys %$conf) {
5253 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5254 $newconf->{$k} = $conf->{$k};
5255 }
5256
5257 &$snapshot_copy_config($snap, $newconf);
5258
5259 return $newconf;
5260 };
5261
5262 sub foreach_writable_storage {
5263 my ($conf, $func) = @_;
5264
5265 my $sidhash = {};
5266
5267 foreach my $ds (keys %$conf) {
5268 next if !valid_drivename($ds);
5269
5270 my $drive = parse_drive($ds, $conf->{$ds});
5271 next if !$drive;
5272 next if drive_is_cdrom($drive);
5273
5274 my $volid = $drive->{file};
5275
5276 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5277 $sidhash->{$sid} = $sid if $sid;
5278 }
5279
5280 foreach my $sid (sort keys %$sidhash) {
5281 &$func($sid);
5282 }
5283 }
5284
5285 my $alloc_vmstate_volid = sub {
5286 my ($storecfg, $vmid, $conf, $snapname) = @_;
5287
5288 # Note: we try to be smart when selecting a $target storage
5289
5290 my $target;
5291
5292 # search shared storage first
5293 foreach_writable_storage($conf, sub {
5294 my ($sid) = @_;
5295 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5296 return if !$scfg->{shared};
5297
5298 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5299 });
5300
5301 if (!$target) {
5302 # now search local storage
5303 foreach_writable_storage($conf, sub {
5304 my ($sid) = @_;
5305 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5306 return if $scfg->{shared};
5307
5308 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5309 });
5310 }
5311
5312 $target = 'local' if !$target;
5313
5314 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5315 # we abort live save after $conf->{memory}, so we need at max twice that space
5316 my $size = $conf->{memory}*2 + $driver_state_size;
5317
5318 my $name = "vm-$vmid-state-$snapname";
5319 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5320 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5321 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5322
5323 return $volid;
5324 };
5325
5326 my $snapshot_prepare = sub {
5327 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5328
5329 my $snap;
5330
5331 my $updatefn = sub {
5332
5333 my $conf = load_config($vmid);
5334
5335 die "you can't take a snapshot if it's a template\n"
5336 if is_template($conf);
5337
5338 check_lock($conf);
5339
5340 $conf->{lock} = 'snapshot';
5341
5342 die "snapshot name '$snapname' already used\n"
5343 if defined($conf->{snapshots}->{$snapname});
5344
5345 my $storecfg = PVE::Storage::config();
5346 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5347
5348 $snap = $conf->{snapshots}->{$snapname} = {};
5349
5350 if ($save_vmstate && check_running($vmid)) {
5351 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5352 }
5353
5354 &$snapshot_copy_config($conf, $snap);
5355
5356 $snap->{snapstate} = "prepare";
5357 $snap->{snaptime} = time();
5358 $snap->{description} = $comment if $comment;
5359
5360 # always overwrite machine if we save vmstate. This makes sure we
5361 # can restore it later using correct machine type
5362 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5363
5364 update_config_nolock($vmid, $conf, 1);
5365 };
5366
5367 lock_config($vmid, $updatefn);
5368
5369 return $snap;
5370 };
5371
5372 my $snapshot_commit = sub {
5373 my ($vmid, $snapname) = @_;
5374
5375 my $updatefn = sub {
5376
5377 my $conf = load_config($vmid);
5378
5379 die "missing snapshot lock\n"
5380 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5381
5382 my $has_machine_config = defined($conf->{machine});
5383
5384 my $snap = $conf->{snapshots}->{$snapname};
5385
5386 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5387
5388 die "wrong snapshot state\n"
5389 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5390
5391 delete $snap->{snapstate};
5392 delete $conf->{lock};
5393
5394 my $newconf = &$snapshot_apply_config($conf, $snap);
5395
5396 delete $newconf->{machine} if !$has_machine_config;
5397
5398 $newconf->{parent} = $snapname;
5399
5400 update_config_nolock($vmid, $newconf, 1);
5401 };
5402
5403 lock_config($vmid, $updatefn);
5404 };
5405
5406 sub snapshot_rollback {
5407 my ($vmid, $snapname) = @_;
5408
5409 my $snap;
5410
5411 my $prepare = 1;
5412
5413 my $storecfg = PVE::Storage::config();
5414
5415 my $updatefn = sub {
5416
5417 my $conf = load_config($vmid);
5418
5419 die "you can't rollback if vm is a template\n" if is_template($conf);
5420
5421 $snap = $conf->{snapshots}->{$snapname};
5422
5423 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5424
5425 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5426 if $snap->{snapstate};
5427
5428 if ($prepare) {
5429 check_lock($conf);
5430 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5431 }
5432
5433 die "unable to rollback vm $vmid: vm is running\n"
5434 if check_running($vmid);
5435
5436 if ($prepare) {
5437 $conf->{lock} = 'rollback';
5438 } else {
5439 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5440 delete $conf->{lock};
5441 }
5442
5443 my $forcemachine;
5444
5445 if (!$prepare) {
5446 my $has_machine_config = defined($conf->{machine});
5447
5448 # copy snapshot config to current config
5449 $conf = &$snapshot_apply_config($conf, $snap);
5450 $conf->{parent} = $snapname;
5451
5452 # Note: old code did not store 'machine', so we try to be smart
5453 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5454 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5455 # we remove the 'machine' configuration if not explicitly specified
5456 # in the original config.
5457 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5458 }
5459
5460 update_config_nolock($vmid, $conf, 1);
5461
5462 if (!$prepare && $snap->{vmstate}) {
5463 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5464 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5465 }
5466 };
5467
5468 lock_config($vmid, $updatefn);
5469
5470 foreach_drive($snap, sub {
5471 my ($ds, $drive) = @_;
5472
5473 return if drive_is_cdrom($drive);
5474
5475 my $volid = $drive->{file};
5476 my $device = "drive-$ds";
5477
5478 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5479 });
5480
5481 $prepare = 0;
5482 lock_config($vmid, $updatefn);
5483 }
5484
5485 my $savevm_wait = sub {
5486 my ($vmid) = @_;
5487
5488 for(;;) {
5489 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5490 if (!$stat->{status}) {
5491 die "savevm not active\n";
5492 } elsif ($stat->{status} eq 'active') {
5493 sleep(1);
5494 next;
5495 } elsif ($stat->{status} eq 'completed') {
5496 last;
5497 } else {
5498 die "query-savevm returned status '$stat->{status}'\n";
5499 }
5500 }
5501 };
5502
5503 sub snapshot_create {
5504 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5505
5506 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5507
5508 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5509
5510 my $config = load_config($vmid);
5511
5512 my $running = check_running($vmid);
5513
5514 my $freezefs = $running && $config->{agent};
5515 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5516
5517 my $drivehash = {};
5518
5519 if ($freezefs) {
5520 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5521 warn "guest-fsfreeze-freeze problems - $@" if $@;
5522 }
5523
5524 eval {
5525 # create internal snapshots of all drives
5526
5527 my $storecfg = PVE::Storage::config();
5528
5529 if ($running) {
5530 if ($snap->{vmstate}) {
5531 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5532 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5533 &$savevm_wait($vmid);
5534 } else {
5535 vm_mon_cmd($vmid, "savevm-start");
5536 }
5537 };
5538
5539 foreach_drive($snap, sub {
5540 my ($ds, $drive) = @_;
5541
5542 return if drive_is_cdrom($drive);
5543
5544 my $volid = $drive->{file};
5545 my $device = "drive-$ds";
5546
5547 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5548 $drivehash->{$ds} = 1;
5549 });
5550 };
5551 my $err = $@;
5552
5553 if ($running) {
5554 eval { vm_mon_cmd($vmid, "savevm-end") };
5555 warn $@ if $@;
5556
5557 if ($freezefs) {
5558 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5559 warn "guest-fsfreeze-thaw problems - $@" if $@;
5560 }
5561
5562 # savevm-end is async, we need to wait
5563 for (;;) {
5564 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5565 if (!$stat->{bytes}) {
5566 last;
5567 } else {
5568 print "savevm not yet finished\n";
5569 sleep(1);
5570 next;
5571 }
5572 }
5573 }
5574
5575 if ($err) {
5576 warn "snapshot create failed: starting cleanup\n";
5577 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5578 warn $@ if $@;
5579 die $err;
5580 }
5581
5582 &$snapshot_commit($vmid, $snapname);
5583 }
5584
5585 # Note: $drivehash is only set when called from snapshot_create.
5586 sub snapshot_delete {
5587 my ($vmid, $snapname, $force, $drivehash) = @_;
5588
5589 my $prepare = 1;
5590
5591 my $snap;
5592 my $unused = [];
5593
5594 my $unlink_parent = sub {
5595 my ($confref, $new_parent) = @_;
5596
5597 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5598 if ($new_parent) {
5599 $confref->{parent} = $new_parent;
5600 } else {
5601 delete $confref->{parent};
5602 }
5603 }
5604 };
5605
5606 my $updatefn = sub {
5607 my ($remove_drive) = @_;
5608
5609 my $conf = load_config($vmid);
5610
5611 if (!$drivehash) {
5612 check_lock($conf);
5613 die "you can't delete a snapshot if vm is a template\n"
5614 if is_template($conf);
5615 }
5616
5617 $snap = $conf->{snapshots}->{$snapname};
5618
5619 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5620
5621 # remove parent refs
5622 if (!$prepare) {
5623 &$unlink_parent($conf, $snap->{parent});
5624 foreach my $sn (keys %{$conf->{snapshots}}) {
5625 next if $sn eq $snapname;
5626 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5627 }
5628 }
5629
5630 if ($remove_drive) {
5631 if ($remove_drive eq 'vmstate') {
5632 delete $snap->{$remove_drive};
5633 } else {
5634 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5635 my $volid = $drive->{file};
5636 delete $snap->{$remove_drive};
5637 add_unused_volume($conf, $volid);
5638 }
5639 }
5640
5641 if ($prepare) {
5642 $snap->{snapstate} = 'delete';
5643 } else {
5644 delete $conf->{snapshots}->{$snapname};
5645 delete $conf->{lock} if $drivehash;
5646 foreach my $volid (@$unused) {
5647 add_unused_volume($conf, $volid);
5648 }
5649 }
5650
5651 update_config_nolock($vmid, $conf, 1);
5652 };
5653
5654 lock_config($vmid, $updatefn);
5655
5656 # now remove vmstate file
5657
5658 my $storecfg = PVE::Storage::config();
5659
5660 if ($snap->{vmstate}) {
5661 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5662 if (my $err = $@) {
5663 die $err if !$force;
5664 warn $err;
5665 }
5666 # save changes (remove vmstate from snapshot)
5667 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5668 };
5669
5670 # now remove all internal snapshots
5671 foreach_drive($snap, sub {
5672 my ($ds, $drive) = @_;
5673
5674 return if drive_is_cdrom($drive);
5675
5676 my $volid = $drive->{file};
5677 my $device = "drive-$ds";
5678
5679 if (!$drivehash || $drivehash->{$ds}) {
5680 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5681 if (my $err = $@) {
5682 die $err if !$force;
5683 warn $err;
5684 }
5685 }
5686
5687 # save changes (remove drive fron snapshot)
5688 lock_config($vmid, $updatefn, $ds) if !$force;
5689 push @$unused, $volid;
5690 });
5691
5692 # now cleanup config
5693 $prepare = 0;
5694 lock_config($vmid, $updatefn);
5695 }
5696
5697 sub has_feature {
5698 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5699
5700 my $err;
5701 foreach_drive($conf, sub {
5702 my ($ds, $drive) = @_;
5703
5704 return if drive_is_cdrom($drive);
5705 my $volid = $drive->{file};
5706 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5707 });
5708
5709 return $err ? 0 : 1;
5710 }
5711
5712 sub template_create {
5713 my ($vmid, $conf, $disk) = @_;
5714
5715 my $storecfg = PVE::Storage::config();
5716
5717 foreach_drive($conf, sub {
5718 my ($ds, $drive) = @_;
5719
5720 return if drive_is_cdrom($drive);
5721 return if $disk && $ds ne $disk;
5722
5723 my $volid = $drive->{file};
5724 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5725
5726 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5727 $drive->{file} = $voliddst;
5728 $conf->{$ds} = print_drive($vmid, $drive);
5729 update_config_nolock($vmid, $conf, 1);
5730 });
5731 }
5732
5733 sub is_template {
5734 my ($conf) = @_;
5735
5736 return 1 if defined $conf->{template} && $conf->{template} == 1;
5737 }
5738
5739 sub qemu_img_convert {
5740 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5741
5742 my $storecfg = PVE::Storage::config();
5743 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5744 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5745
5746 if ($src_storeid && $dst_storeid) {
5747 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5748 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5749
5750 my $src_format = qemu_img_format($src_scfg, $src_volname);
5751 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5752
5753 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5754 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5755
5756 my $cmd = [];
5757 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5758 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5759 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5760
5761 my $parser = sub {
5762 my $line = shift;
5763 if($line =~ m/\((\S+)\/100\%\)/){
5764 my $percent = $1;
5765 my $transferred = int($size * $percent / 100);
5766 my $remaining = $size - $transferred;
5767
5768 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5769 }
5770
5771 };
5772
5773 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5774 my $err = $@;
5775 die "copy failed: $err" if $err;
5776 }
5777 }
5778
5779 sub qemu_img_format {
5780 my ($scfg, $volname) = @_;
5781
5782 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5783 return $1;
5784 } elsif ($scfg->{type} eq 'iscsi') {
5785 return "host_device";
5786 } else {
5787 return "raw";
5788 }
5789 }
5790
5791 sub qemu_drive_mirror {
5792 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5793
5794 my $count = 0;
5795 my $old_len = 0;
5796 my $frozen = undef;
5797 my $maxwait = 120;
5798
5799 my $storecfg = PVE::Storage::config();
5800 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5801
5802 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5803
5804 my $format;
5805 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5806 $format = $1;
5807 }
5808
5809 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5810
5811 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5812 $opts->{format} = $format if $format;
5813
5814 #fixme : sometime drive-mirror timeout, but works fine after.
5815 # (I have see the problem with big volume > 200GB), so we need to eval
5816 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5817 # ignore errors here
5818
5819 eval {
5820 while (1) {
5821 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5822 my $stat = @$stats[0];
5823 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5824 die "error job is not mirroring" if $stat->{type} ne "mirror";
5825
5826 my $busy = $stat->{busy};
5827
5828 if (my $total = $stat->{len}) {
5829 my $transferred = $stat->{offset} || 0;
5830 my $remaining = $total - $transferred;
5831 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5832
5833 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
5834 }
5835
5836 if ($stat->{len} == $stat->{offset}) {
5837 if ($busy eq 'false') {
5838
5839 last if $vmiddst != $vmid;
5840
5841 # try to switch the disk if source and destination are on the same guest
5842 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
5843 last if !$@;
5844 die $@ if $@ !~ m/cannot be completed/;
5845 }
5846
5847 if ($count > $maxwait) {
5848 # if too much writes to disk occurs at the end of migration
5849 #the disk needs to be freezed to be able to complete the migration
5850 vm_suspend($vmid,1);
5851 $frozen = 1;
5852 }
5853 $count ++
5854 }
5855 $old_len = $stat->{offset};
5856 sleep 1;
5857 }
5858
5859 vm_resume($vmid, 1) if $frozen;
5860
5861 };
5862 my $err = $@;
5863
5864 my $cancel_job = sub {
5865 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
5866 while (1) {
5867 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5868 my $stat = @$stats[0];
5869 last if !$stat;
5870 sleep 1;
5871 }
5872 };
5873
5874 if ($err) {
5875 eval { &$cancel_job(); };
5876 die "mirroring error: $err";
5877 }
5878
5879 if ($vmiddst != $vmid) {
5880 # if we clone a disk for a new target vm, we don't switch the disk
5881 &$cancel_job(); # so we call block-job-cancel
5882 }
5883 }
5884
5885 sub clone_disk {
5886 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
5887 $newvmid, $storage, $format, $full, $newvollist) = @_;
5888
5889 my $newvolid;
5890
5891 if (!$full) {
5892 print "create linked clone of drive $drivename ($drive->{file})\n";
5893 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
5894 push @$newvollist, $newvolid;
5895 } else {
5896 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5897 $storeid = $storage if $storage;
5898
5899 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5900 if (!$format) {
5901 $format = $drive->{format} || $defFormat;
5902 }
5903
5904 # test if requested format is supported - else use default
5905 my $supported = grep { $_ eq $format } @$validFormats;
5906 $format = $defFormat if !$supported;
5907
5908 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
5909
5910 print "create full clone of drive $drivename ($drive->{file})\n";
5911 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
5912 push @$newvollist, $newvolid;
5913
5914 if (!$running || $snapname) {
5915 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
5916 } else {
5917 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
5918 }
5919 }
5920
5921 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
5922
5923 my $disk = $drive;
5924 $disk->{format} = undef;
5925 $disk->{file} = $newvolid;
5926 $disk->{size} = $size;
5927
5928 return $disk;
5929 }
5930
5931 # this only works if VM is running
5932 sub get_current_qemu_machine {
5933 my ($vmid) = @_;
5934
5935 my $cmd = { execute => 'query-machines', arguments => {} };
5936 my $res = vm_qmp_command($vmid, $cmd);
5937
5938 my ($current, $default);
5939 foreach my $e (@$res) {
5940 $default = $e->{name} if $e->{'is-default'};
5941 $current = $e->{name} if $e->{'is-current'};
5942 }
5943
5944 # fallback to the default machine if current is not supported by qemu
5945 return $current || $default || 'pc';
5946 }
5947
5948 sub lspci {
5949
5950 my $devices = {};
5951
5952 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
5953 my (undef, $id, $function) = @_;
5954 my $res = { id => $id, function => $function};
5955 push @{$devices->{$id}}, $res;
5956 });
5957
5958 return $devices;
5959 }
5960
5961 1;