]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
balloon: use qom-get for guest balloon statistics V5
[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 => 'string', format => 'pve-hotplug-features',
176 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
177 default => 'network,disk,usb',
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 vcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Number of hotplugged vcpus.",
319 minimum => 1,
320 default => 0,
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 my $MAX_MEM = 4194304;
494 my $STATICMEM = 1024;
495
496 my $numadesc = {
497 optional => 1,
498 type => 'string', format => 'pve-qm-numanode',
499 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
500 description => "numa topology",
501 };
502 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
503
504 for (my $i = 0; $i < $MAX_NUMA; $i++) {
505 $confdesc->{"numa$i"} = $numadesc;
506 }
507
508 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
509 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
510 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
511 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
512
513 my $netdesc = {
514 optional => 1,
515 type => 'string', format => 'pve-qm-net',
516 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
517 description => <<EODESCR,
518 Specify network devices.
519
520 MODEL is one of: $nic_model_list_txt
521
522 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
523 automatically generated if not specified.
524
525 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
526
527 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'.
528
529 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
530
531 10.0.2.2 Gateway
532 10.0.2.3 DNS Server
533 10.0.2.4 SMB Server
534
535 The DHCP server assign addresses to the guest starting from 10.0.2.15.
536
537 EODESCR
538 };
539 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
540
541 for (my $i = 0; $i < $MAX_NETS; $i++) {
542 $confdesc->{"net$i"} = $netdesc;
543 }
544
545 my $drivename_hash;
546
547 my $idedesc = {
548 optional => 1,
549 type => 'string', format => 'pve-qm-drive',
550 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
551 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
552 };
553 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
554
555 my $scsidesc = {
556 optional => 1,
557 type => 'string', format => 'pve-qm-drive',
558 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
559 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
560 };
561 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
562
563 my $satadesc = {
564 optional => 1,
565 type => 'string', format => 'pve-qm-drive',
566 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
567 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
568 };
569 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
570
571 my $virtiodesc = {
572 optional => 1,
573 type => 'string', format => 'pve-qm-drive',
574 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
575 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
576 };
577 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
578
579 my $usbdesc = {
580 optional => 1,
581 type => 'string', format => 'pve-qm-usb-device',
582 typetext => 'host=HOSTUSBDEVICE|spice',
583 description => <<EODESCR,
584 Configure an USB device (n is 0 to 4). This can be used to
585 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
586
587 'bus-port(.port)*' (decimal numbers) or
588 'vendor_id:product_id' (hexadeciaml numbers)
589
590 You can use the 'lsusb -t' command to list existing usb devices.
591
592 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
593
594 The value 'spice' can be used to add a usb redirection devices for spice.
595
596 EODESCR
597 };
598 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
599
600 my $hostpcidesc = {
601 optional => 1,
602 type => 'string', format => 'pve-qm-hostpci',
603 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
604 description => <<EODESCR,
605 Map host pci devices. HOSTPCIDEVICE syntax is:
606
607 'bus:dev.func' (hexadecimal numbers)
608
609 You can us the 'lspci' command to list existing pci devices.
610
611 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
612
613 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
614
615 Experimental: user reported problems with this option.
616 EODESCR
617 };
618 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
619
620 my $serialdesc = {
621 optional => 1,
622 type => 'string',
623 pattern => '(/dev/.+|socket)',
624 description => <<EODESCR,
625 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).
626
627 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
628
629 Experimental: user reported problems with this option.
630 EODESCR
631 };
632
633 my $paralleldesc= {
634 optional => 1,
635 type => 'string',
636 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
637 description => <<EODESCR,
638 Map host parallel devices (n is 0 to 2).
639
640 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
641
642 Experimental: user reported problems with this option.
643 EODESCR
644 };
645
646 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
647 $confdesc->{"parallel$i"} = $paralleldesc;
648 }
649
650 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
651 $confdesc->{"serial$i"} = $serialdesc;
652 }
653
654 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
655 $confdesc->{"hostpci$i"} = $hostpcidesc;
656 }
657
658 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
659 $drivename_hash->{"ide$i"} = 1;
660 $confdesc->{"ide$i"} = $idedesc;
661 }
662
663 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
664 $drivename_hash->{"sata$i"} = 1;
665 $confdesc->{"sata$i"} = $satadesc;
666 }
667
668 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
669 $drivename_hash->{"scsi$i"} = 1;
670 $confdesc->{"scsi$i"} = $scsidesc ;
671 }
672
673 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
674 $drivename_hash->{"virtio$i"} = 1;
675 $confdesc->{"virtio$i"} = $virtiodesc;
676 }
677
678 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
679 $confdesc->{"usb$i"} = $usbdesc;
680 }
681
682 my $unuseddesc = {
683 optional => 1,
684 type => 'string', format => 'pve-volume-id',
685 description => "Reference to unused volumes.",
686 };
687
688 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
689 $confdesc->{"unused$i"} = $unuseddesc;
690 }
691
692 my $kvm_api_version = 0;
693
694 sub kvm_version {
695
696 return $kvm_api_version if $kvm_api_version;
697
698 my $fh = IO::File->new("</dev/kvm") ||
699 return 0;
700
701 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
702 $kvm_api_version = $v;
703 }
704
705 $fh->close();
706
707 return $kvm_api_version;
708 }
709
710 my $kvm_user_version;
711
712 sub kvm_user_version {
713
714 return $kvm_user_version if $kvm_user_version;
715
716 $kvm_user_version = 'unknown';
717
718 my $tmp = `kvm -help 2>/dev/null`;
719
720 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
721 $kvm_user_version = $2;
722 }
723
724 return $kvm_user_version;
725
726 }
727
728 my $kernel_has_vhost_net = -c '/dev/vhost-net';
729
730 sub disknames {
731 # order is important - used to autoselect boot disk
732 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
733 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
734 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
735 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
736 }
737
738 sub valid_drivename {
739 my $dev = shift;
740
741 return defined($drivename_hash->{$dev});
742 }
743
744 sub option_exists {
745 my $key = shift;
746 return defined($confdesc->{$key});
747 }
748
749 sub nic_models {
750 return $nic_model_list;
751 }
752
753 sub os_list_description {
754
755 return {
756 other => 'Other',
757 wxp => 'Windows XP',
758 w2k => 'Windows 2000',
759 w2k3 =>, 'Windows 2003',
760 w2k8 => 'Windows 2008',
761 wvista => 'Windows Vista',
762 win7 => 'Windows 7',
763 win8 => 'Windows 8/2012',
764 l24 => 'Linux 2.4',
765 l26 => 'Linux 2.6',
766 };
767 }
768
769 my $cdrom_path;
770
771 sub get_cdrom_path {
772
773 return $cdrom_path if $cdrom_path;
774
775 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
776 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
777 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
778 }
779
780 sub get_iso_path {
781 my ($storecfg, $vmid, $cdrom) = @_;
782
783 if ($cdrom eq 'cdrom') {
784 return get_cdrom_path();
785 } elsif ($cdrom eq 'none') {
786 return '';
787 } elsif ($cdrom =~ m|^/|) {
788 return $cdrom;
789 } else {
790 return PVE::Storage::path($storecfg, $cdrom);
791 }
792 }
793
794 # try to convert old style file names to volume IDs
795 sub filename_to_volume_id {
796 my ($vmid, $file, $media) = @_;
797
798 if (!($file eq 'none' || $file eq 'cdrom' ||
799 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
800
801 return undef if $file =~ m|/|;
802
803 if ($media && $media eq 'cdrom') {
804 $file = "local:iso/$file";
805 } else {
806 $file = "local:$vmid/$file";
807 }
808 }
809
810 return $file;
811 }
812
813 sub verify_media_type {
814 my ($opt, $vtype, $media) = @_;
815
816 return if !$media;
817
818 my $etype;
819 if ($media eq 'disk') {
820 $etype = 'images';
821 } elsif ($media eq 'cdrom') {
822 $etype = 'iso';
823 } else {
824 die "internal error";
825 }
826
827 return if ($vtype eq $etype);
828
829 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
830 }
831
832 sub cleanup_drive_path {
833 my ($opt, $storecfg, $drive) = @_;
834
835 # try to convert filesystem paths to volume IDs
836
837 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
838 ($drive->{file} !~ m|^/dev/.+|) &&
839 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
840 ($drive->{file} !~ m/^\d+$/)) {
841 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
842 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
843 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
844 verify_media_type($opt, $vtype, $drive->{media});
845 $drive->{file} = $volid;
846 }
847
848 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
849 }
850
851 sub create_conf_nolock {
852 my ($vmid, $settings) = @_;
853
854 my $filename = config_file($vmid);
855
856 die "configuration file '$filename' already exists\n" if -f $filename;
857
858 my $defaults = load_defaults();
859
860 $settings->{name} = "vm$vmid" if !$settings->{name};
861 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
862
863 my $data = '';
864 foreach my $opt (keys %$settings) {
865 next if !$confdesc->{$opt};
866
867 my $value = $settings->{$opt};
868 next if !$value;
869
870 $data .= "$opt: $value\n";
871 }
872
873 PVE::Tools::file_set_contents($filename, $data);
874 }
875
876 sub parse_hotplug_features {
877 my ($data) = @_;
878
879 my $res = {};
880
881 return $res if $data eq '0';
882
883 $data = $confdesc->{hotplug}->{default} if $data eq '1';
884
885 foreach my $feature (PVE::Tools::split_list($data)) {
886 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
887 $res->{$1} = 1;
888 } else {
889 warn "ignoring unknown hotplug feature '$feature'\n";
890 }
891 }
892 return $res;
893 }
894
895 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
896 sub pve_verify_hotplug_features {
897 my ($value, $noerr) = @_;
898
899 return $value if parse_hotplug_features($value);
900
901 return undef if $noerr;
902
903 die "unable to parse hotplug option\n";
904 }
905
906 my $parse_size = sub {
907 my ($value) = @_;
908
909 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
910 my ($size, $unit) = ($1, $3);
911 if ($unit) {
912 if ($unit eq 'K') {
913 $size = $size * 1024;
914 } elsif ($unit eq 'M') {
915 $size = $size * 1024 * 1024;
916 } elsif ($unit eq 'G') {
917 $size = $size * 1024 * 1024 * 1024;
918 }
919 }
920 return int($size);
921 };
922
923 my $format_size = sub {
924 my ($size) = @_;
925
926 $size = int($size);
927
928 my $kb = int($size/1024);
929 return $size if $kb*1024 != $size;
930
931 my $mb = int($kb/1024);
932 return "${kb}K" if $mb*1024 != $kb;
933
934 my $gb = int($mb/1024);
935 return "${mb}M" if $gb*1024 != $mb;
936
937 return "${gb}G";
938 };
939
940 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
941 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
942 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
943 # [,aio=native|threads][,discard=ignore|on]
944
945 sub parse_drive {
946 my ($key, $data) = @_;
947
948 my $res = {};
949
950 # $key may be undefined - used to verify JSON parameters
951 if (!defined($key)) {
952 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
953 $res->{index} = 0;
954 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
955 $res->{interface} = $1;
956 $res->{index} = $2;
957 } else {
958 return undef;
959 }
960
961 foreach my $p (split (/,/, $data)) {
962 next if $p =~ m/^\s*$/;
963
964 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
965 my ($k, $v) = ($1, $2);
966
967 $k = 'file' if $k eq 'volume';
968
969 return undef if defined $res->{$k};
970
971 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
972 return undef if !$v || $v !~ m/^\d+/;
973 $k = "m$k";
974 $v = sprintf("%.3f", $v / (1024*1024));
975 }
976 $res->{$k} = $v;
977 } else {
978 if (!$res->{file} && $p !~ m/=/) {
979 $res->{file} = $p;
980 } else {
981 return undef;
982 }
983 }
984 }
985
986 return undef if !$res->{file};
987
988 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
989 $res->{format} = $1;
990 }
991
992 return undef if $res->{cache} &&
993 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
994 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
995 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
996 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
997 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
998 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
999 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
1000 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
1001 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
1002 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
1003 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
1004 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
1005 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
1006
1007 return undef if $res->{mbps_rd} && $res->{mbps};
1008 return undef if $res->{mbps_wr} && $res->{mbps};
1009
1010 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
1011 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
1012 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
1013 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
1014 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
1015 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
1016
1017 return undef if $res->{iops_rd} && $res->{iops};
1018 return undef if $res->{iops_wr} && $res->{iops};
1019
1020
1021 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
1022 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
1023 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
1024 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
1025 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
1026 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
1027
1028
1029 if ($res->{size}) {
1030 return undef if !defined($res->{size} = &$parse_size($res->{size}));
1031 }
1032
1033 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1034 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1035 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1036 return undef if $res->{interface} eq 'virtio';
1037 }
1038
1039 # rerror does not work with scsi drives
1040 if ($res->{rerror}) {
1041 return undef if $res->{interface} eq 'scsi';
1042 }
1043
1044 return $res;
1045 }
1046
1047 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1048
1049 sub print_drive {
1050 my ($vmid, $drive) = @_;
1051
1052 my $opts = '';
1053 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1054 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1055 }
1056
1057 if ($drive->{size}) {
1058 $opts .= ",size=" . &$format_size($drive->{size});
1059 }
1060
1061 return "$drive->{file}$opts";
1062 }
1063
1064 sub scsi_inquiry {
1065 my($fh, $noerr) = @_;
1066
1067 my $SG_IO = 0x2285;
1068 my $SG_GET_VERSION_NUM = 0x2282;
1069
1070 my $versionbuf = "\x00" x 8;
1071 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1072 if (!$ret) {
1073 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1074 return undef;
1075 }
1076 my $version = unpack("I", $versionbuf);
1077 if ($version < 30000) {
1078 die "scsi generic interface too old\n" if !$noerr;
1079 return undef;
1080 }
1081
1082 my $buf = "\x00" x 36;
1083 my $sensebuf = "\x00" x 8;
1084 my $cmd = pack("C x3 C x1", 0x12, 36);
1085
1086 # see /usr/include/scsi/sg.h
1087 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";
1088
1089 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1090 length($sensebuf), 0, length($buf), $buf,
1091 $cmd, $sensebuf, 6000);
1092
1093 $ret = ioctl($fh, $SG_IO, $packet);
1094 if (!$ret) {
1095 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1096 return undef;
1097 }
1098
1099 my @res = unpack($sg_io_hdr_t, $packet);
1100 if ($res[17] || $res[18]) {
1101 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1102 return undef;
1103 }
1104
1105 my $res = {};
1106 (my $byte0, my $byte1, $res->{vendor},
1107 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1108
1109 $res->{removable} = $byte1 & 128 ? 1 : 0;
1110 $res->{type} = $byte0 & 31;
1111
1112 return $res;
1113 }
1114
1115 sub path_is_scsi {
1116 my ($path) = @_;
1117
1118 my $fh = IO::File->new("+<$path") || return undef;
1119 my $res = scsi_inquiry($fh, 1);
1120 close($fh);
1121
1122 return $res;
1123 }
1124
1125 sub machine_type_is_q35 {
1126 my ($conf) = @_;
1127
1128 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1129 }
1130
1131 sub print_tabletdevice_full {
1132 my ($conf) = @_;
1133
1134 my $q35 = machine_type_is_q35($conf);
1135
1136 # we use uhci for old VMs because tablet driver was buggy in older qemu
1137 my $usbbus = $q35 ? "ehci" : "uhci";
1138
1139 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1140 }
1141
1142 sub print_drivedevice_full {
1143 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1144
1145 my $device = '';
1146 my $maxdev = 0;
1147
1148 if ($drive->{interface} eq 'virtio') {
1149 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1150 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1151 $device .= ",iothread=iothread0" if $conf->{iothread};
1152 } elsif ($drive->{interface} eq 'scsi') {
1153 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1154 my $controller = int($drive->{index} / $maxdev);
1155 my $unit = $drive->{index} % $maxdev;
1156 my $devicetype = 'hd';
1157 my $path = '';
1158 if (drive_is_cdrom($drive)) {
1159 $devicetype = 'cd';
1160 } else {
1161 if ($drive->{file} =~ m|^/|) {
1162 $path = $drive->{file};
1163 } else {
1164 $path = PVE::Storage::path($storecfg, $drive->{file});
1165 }
1166
1167 if($path =~ m/^iscsi\:\/\//){
1168 $devicetype = 'generic';
1169 } else {
1170 if (my $info = path_is_scsi($path)) {
1171 if ($info->{type} == 0) {
1172 $devicetype = 'block';
1173 } elsif ($info->{type} == 1) { # tape
1174 $devicetype = 'generic';
1175 }
1176 }
1177 }
1178 }
1179
1180 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1181 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1182 } else {
1183 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1184 }
1185
1186 } elsif ($drive->{interface} eq 'ide'){
1187 $maxdev = 2;
1188 my $controller = int($drive->{index} / $maxdev);
1189 my $unit = $drive->{index} % $maxdev;
1190 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1191
1192 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1193 } elsif ($drive->{interface} eq 'sata'){
1194 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1195 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1196 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1197 } elsif ($drive->{interface} eq 'usb') {
1198 die "implement me";
1199 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1200 } else {
1201 die "unsupported interface type";
1202 }
1203
1204 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1205
1206 return $device;
1207 }
1208
1209 sub get_initiator_name {
1210 my $initiator;
1211
1212 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1213 while (defined(my $line = <$fh>)) {
1214 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1215 $initiator = $1;
1216 last;
1217 }
1218 $fh->close();
1219
1220 return $initiator;
1221 }
1222
1223 sub print_drive_full {
1224 my ($storecfg, $vmid, $drive) = @_;
1225
1226 my $opts = '';
1227 foreach my $o (@qemu_drive_options) {
1228 next if $o eq 'bootindex';
1229 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1230 }
1231
1232 foreach my $o (qw(bps bps_rd bps_wr)) {
1233 my $v = $drive->{"m$o"};
1234 $opts .= ",$o=" . int($v*1024*1024) if $v;
1235 }
1236
1237 # use linux-aio by default (qemu default is threads)
1238 $opts .= ",aio=native" if !$drive->{aio};
1239
1240 my $path;
1241 my $volid = $drive->{file};
1242 if (drive_is_cdrom($drive)) {
1243 $path = get_iso_path($storecfg, $vmid, $volid);
1244 } else {
1245 if ($volid =~ m|^/|) {
1246 $path = $volid;
1247 } else {
1248 $path = PVE::Storage::path($storecfg, $volid);
1249 }
1250 }
1251
1252 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1253
1254 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1255 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1256
1257 my $pathinfo = $path ? "file=$path," : '';
1258
1259 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1260 }
1261
1262 sub print_netdevice_full {
1263 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1264
1265 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1266
1267 my $device = $net->{model};
1268 if ($net->{model} eq 'virtio') {
1269 $device = 'virtio-net-pci';
1270 };
1271
1272 my $pciaddr = print_pci_addr("$netid", $bridges);
1273 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1274 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1275 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1276 my $vectors = $net->{queues} * 2 + 2;
1277 $tmpstr .= ",vectors=$vectors,mq=on";
1278 }
1279 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1280 return $tmpstr;
1281 }
1282
1283 sub print_netdev_full {
1284 my ($vmid, $conf, $net, $netid) = @_;
1285
1286 my $i = '';
1287 if ($netid =~ m/^net(\d+)$/) {
1288 $i = int($1);
1289 }
1290
1291 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1292
1293 my $ifname = "tap${vmid}i$i";
1294
1295 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1296 die "interface name '$ifname' is too long (max 15 character)\n"
1297 if length($ifname) >= 16;
1298
1299 my $vhostparam = '';
1300 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1301
1302 my $vmname = $conf->{name} || "vm$vmid";
1303
1304 my $netdev = "";
1305
1306 if ($net->{bridge}) {
1307 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1308 } else {
1309 $netdev = "type=user,id=$netid,hostname=$vmname";
1310 }
1311
1312 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1313
1314 return $netdev;
1315 }
1316
1317 sub drive_is_cdrom {
1318 my ($drive) = @_;
1319
1320 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1321
1322 }
1323
1324 sub parse_numa {
1325 my ($data) = @_;
1326
1327 my $res = {};
1328
1329 foreach my $kvp (split(/,/, $data)) {
1330
1331 if ($kvp =~ m/^memory=(\S+)$/) {
1332 $res->{memory} = $1;
1333 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1334 $res->{policy} = $1;
1335 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1336 $res->{cpus}->{start} = $1;
1337 $res->{cpus}->{end} = $3;
1338 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1339 $res->{hostnodes}->{start} = $1;
1340 $res->{hostnodes}->{end} = $3;
1341 } else {
1342 return undef;
1343 }
1344 }
1345
1346 return $res;
1347 }
1348
1349 sub parse_hostpci {
1350 my ($value) = @_;
1351
1352 return undef if !$value;
1353
1354
1355 my @list = split(/,/, $value);
1356 my $found;
1357
1358 my $res = {};
1359 foreach my $kv (@list) {
1360
1361 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1362 $found = 1;
1363 if(defined($4)){
1364 push @{$res->{pciid}}, { id => $2 , function => $4};
1365
1366 }else{
1367 my $pcidevices = lspci($2);
1368 $res->{pciid} = $pcidevices->{$2};
1369 }
1370 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1371 $res->{driver} = $1;
1372 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1373 $res->{rombar} = $1;
1374 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1375 $res->{'x-vga'} = $1;
1376 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1377 $res->{pcie} = 1 if $1 == 1;
1378 } else {
1379 warn "unknown hostpci setting '$kv'\n";
1380 }
1381 }
1382
1383 return undef if !$found;
1384
1385 return $res;
1386 }
1387
1388 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1389 sub parse_net {
1390 my ($data) = @_;
1391
1392 my $res = {};
1393
1394 foreach my $kvp (split(/,/, $data)) {
1395
1396 if ($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) {
1397 my $model = lc($1);
1398 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1399 $res->{model} = $model;
1400 $res->{macaddr} = $mac;
1401 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1402 $res->{bridge} = $1;
1403 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1404 $res->{queues} = $1;
1405 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1406 $res->{rate} = $1;
1407 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1408 $res->{tag} = $1;
1409 } elsif ($kvp =~ m/^firewall=([01])$/) {
1410 $res->{firewall} = $1;
1411 } elsif ($kvp =~ m/^link_down=([01])$/) {
1412 $res->{link_down} = $1;
1413 } else {
1414 return undef;
1415 }
1416
1417 }
1418
1419 return undef if !$res->{model};
1420
1421 return $res;
1422 }
1423
1424 sub print_net {
1425 my $net = shift;
1426
1427 my $res = "$net->{model}";
1428 $res .= "=$net->{macaddr}" if $net->{macaddr};
1429 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1430 $res .= ",rate=$net->{rate}" if $net->{rate};
1431 $res .= ",tag=$net->{tag}" if $net->{tag};
1432 $res .= ",firewall=1" if $net->{firewall};
1433 $res .= ",link_down=1" if $net->{link_down};
1434 $res .= ",queues=$net->{queues}" if $net->{queues};
1435
1436 return $res;
1437 }
1438
1439 sub add_random_macs {
1440 my ($settings) = @_;
1441
1442 foreach my $opt (keys %$settings) {
1443 next if $opt !~ m/^net(\d+)$/;
1444 my $net = parse_net($settings->{$opt});
1445 next if !$net;
1446 $settings->{$opt} = print_net($net);
1447 }
1448 }
1449
1450 sub add_unused_volume {
1451 my ($config, $volid) = @_;
1452
1453 my $key;
1454 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1455 my $test = "unused$ind";
1456 if (my $vid = $config->{$test}) {
1457 return if $vid eq $volid; # do not add duplicates
1458 } else {
1459 $key = $test;
1460 }
1461 }
1462
1463 die "To many unused volume - please delete them first.\n" if !$key;
1464
1465 $config->{$key} = $volid;
1466
1467 return $key;
1468 }
1469
1470 sub vm_is_volid_owner {
1471 my ($storecfg, $vmid, $volid) = @_;
1472
1473 if ($volid !~ m|^/|) {
1474 my ($path, $owner);
1475 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1476 if ($owner && ($owner == $vmid)) {
1477 return 1;
1478 }
1479 }
1480
1481 return undef;
1482 }
1483
1484 sub vmconfig_delete_pending_option {
1485 my ($conf, $key) = @_;
1486
1487 delete $conf->{pending}->{$key};
1488 my $pending_delete_hash = { $key => 1 };
1489 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1490 $pending_delete_hash->{$opt} = 1;
1491 }
1492 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1493 }
1494
1495 sub vmconfig_undelete_pending_option {
1496 my ($conf, $key) = @_;
1497
1498 my $pending_delete_hash = {};
1499 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1500 $pending_delete_hash->{$opt} = 1;
1501 }
1502 delete $pending_delete_hash->{$key};
1503
1504 my @keylist = keys %$pending_delete_hash;
1505 if (scalar(@keylist)) {
1506 $conf->{pending}->{delete} = join(',', @keylist);
1507 } else {
1508 delete $conf->{pending}->{delete};
1509 }
1510 }
1511
1512 sub vmconfig_register_unused_drive {
1513 my ($storecfg, $vmid, $conf, $drive) = @_;
1514
1515 if (!drive_is_cdrom($drive)) {
1516 my $volid = $drive->{file};
1517 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1518 add_unused_volume($conf, $volid, $vmid);
1519 }
1520 }
1521 }
1522
1523 sub vmconfig_cleanup_pending {
1524 my ($conf) = @_;
1525
1526 # remove pending changes when nothing changed
1527 my $changes;
1528 foreach my $opt (keys %{$conf->{pending}}) {
1529 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1530 $changes = 1;
1531 delete $conf->{pending}->{$opt};
1532 }
1533 }
1534
1535 # remove delete if option is not set
1536 my $pending_delete_hash = {};
1537 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1538 if (defined($conf->{$opt})) {
1539 $pending_delete_hash->{$opt} = 1;
1540 } else {
1541 $changes = 1;
1542 }
1543 }
1544
1545 my @keylist = keys %$pending_delete_hash;
1546 if (scalar(@keylist)) {
1547 $conf->{pending}->{delete} = join(',', @keylist);
1548 } else {
1549 delete $conf->{pending}->{delete};
1550 }
1551
1552 return $changes;
1553 }
1554
1555 my $valid_smbios1_options = {
1556 manufacturer => '\S+',
1557 product => '\S+',
1558 version => '\S+',
1559 serial => '\S+',
1560 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1561 sku => '\S+',
1562 family => '\S+',
1563 };
1564
1565 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1566 sub parse_smbios1 {
1567 my ($data) = @_;
1568
1569 my $res = {};
1570
1571 foreach my $kvp (split(/,/, $data)) {
1572 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1573 my ($k, $v) = split(/=/, $kvp);
1574 return undef if !defined($k) || !defined($v);
1575 return undef if !$valid_smbios1_options->{$k};
1576 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1577 $res->{$k} = $v;
1578 }
1579
1580 return $res;
1581 }
1582
1583 sub print_smbios1 {
1584 my ($smbios1) = @_;
1585
1586 my $data = '';
1587 foreach my $k (keys %$smbios1) {
1588 next if !defined($smbios1->{$k});
1589 next if !$valid_smbios1_options->{$k};
1590 $data .= ',' if $data;
1591 $data .= "$k=$smbios1->{$k}";
1592 }
1593 return $data;
1594 }
1595
1596 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1597 sub verify_smbios1 {
1598 my ($value, $noerr) = @_;
1599
1600 return $value if parse_smbios1($value);
1601
1602 return undef if $noerr;
1603
1604 die "unable to parse smbios (type 1) options\n";
1605 }
1606
1607 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1608 sub verify_bootdisk {
1609 my ($value, $noerr) = @_;
1610
1611 return $value if valid_drivename($value);
1612
1613 return undef if $noerr;
1614
1615 die "invalid boot disk '$value'\n";
1616 }
1617
1618 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1619 sub verify_numa {
1620 my ($value, $noerr) = @_;
1621
1622 return $value if parse_numa($value);
1623
1624 return undef if $noerr;
1625
1626 die "unable to parse numa options\n";
1627 }
1628
1629 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1630 sub verify_net {
1631 my ($value, $noerr) = @_;
1632
1633 return $value if parse_net($value);
1634
1635 return undef if $noerr;
1636
1637 die "unable to parse network options\n";
1638 }
1639
1640 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1641 sub verify_drive {
1642 my ($value, $noerr) = @_;
1643
1644 return $value if parse_drive(undef, $value);
1645
1646 return undef if $noerr;
1647
1648 die "unable to parse drive options\n";
1649 }
1650
1651 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1652 sub verify_hostpci {
1653 my ($value, $noerr) = @_;
1654
1655 return $value if parse_hostpci($value);
1656
1657 return undef if $noerr;
1658
1659 die "unable to parse pci id\n";
1660 }
1661
1662 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1663 sub verify_watchdog {
1664 my ($value, $noerr) = @_;
1665
1666 return $value if parse_watchdog($value);
1667
1668 return undef if $noerr;
1669
1670 die "unable to parse watchdog options\n";
1671 }
1672
1673 sub parse_watchdog {
1674 my ($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 ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1684 $res->{model} = $2;
1685 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1686 $res->{action} = $2;
1687 } else {
1688 return undef;
1689 }
1690 }
1691
1692 return $res;
1693 }
1694
1695 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1696 sub verify_startup {
1697 my ($value, $noerr) = @_;
1698
1699 return $value if parse_startup($value);
1700
1701 return undef if $noerr;
1702
1703 die "unable to parse startup options\n";
1704 }
1705
1706 sub parse_startup {
1707 my ($value) = @_;
1708
1709 return undef if !$value;
1710
1711 my $res = {};
1712
1713 foreach my $p (split(/,/, $value)) {
1714 next if $p =~ m/^\s*$/;
1715
1716 if ($p =~ m/^(order=)?(\d+)$/) {
1717 $res->{order} = $2;
1718 } elsif ($p =~ m/^up=(\d+)$/) {
1719 $res->{up} = $1;
1720 } elsif ($p =~ m/^down=(\d+)$/) {
1721 $res->{down} = $1;
1722 } else {
1723 return undef;
1724 }
1725 }
1726
1727 return $res;
1728 }
1729
1730 sub parse_usb_device {
1731 my ($value) = @_;
1732
1733 return undef if !$value;
1734
1735 my @dl = split(/,/, $value);
1736 my $found;
1737
1738 my $res = {};
1739 foreach my $v (@dl) {
1740 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1741 $found = 1;
1742 $res->{vendorid} = $2;
1743 $res->{productid} = $4;
1744 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1745 $found = 1;
1746 $res->{hostbus} = $1;
1747 $res->{hostport} = $2;
1748 } elsif ($v =~ m/^spice$/) {
1749 $found = 1;
1750 $res->{spice} = 1;
1751 } else {
1752 return undef;
1753 }
1754 }
1755 return undef if !$found;
1756
1757 return $res;
1758 }
1759
1760 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1761 sub verify_usb_device {
1762 my ($value, $noerr) = @_;
1763
1764 return $value if parse_usb_device($value);
1765
1766 return undef if $noerr;
1767
1768 die "unable to parse usb device\n";
1769 }
1770
1771 # add JSON properties for create and set function
1772 sub json_config_properties {
1773 my $prop = shift;
1774
1775 foreach my $opt (keys %$confdesc) {
1776 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1777 $prop->{$opt} = $confdesc->{$opt};
1778 }
1779
1780 return $prop;
1781 }
1782
1783 sub check_type {
1784 my ($key, $value) = @_;
1785
1786 die "unknown setting '$key'\n" if !$confdesc->{$key};
1787
1788 my $type = $confdesc->{$key}->{type};
1789
1790 if (!defined($value)) {
1791 die "got undefined value\n";
1792 }
1793
1794 if ($value =~ m/[\n\r]/) {
1795 die "property contains a line feed\n";
1796 }
1797
1798 if ($type eq 'boolean') {
1799 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1800 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1801 die "type check ('boolean') failed - got '$value'\n";
1802 } elsif ($type eq 'integer') {
1803 return int($1) if $value =~ m/^(\d+)$/;
1804 die "type check ('integer') failed - got '$value'\n";
1805 } elsif ($type eq 'number') {
1806 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1807 die "type check ('number') failed - got '$value'\n";
1808 } elsif ($type eq 'string') {
1809 if (my $fmt = $confdesc->{$key}->{format}) {
1810 if ($fmt eq 'pve-qm-drive') {
1811 # special case - we need to pass $key to parse_drive()
1812 my $drive = parse_drive($key, $value);
1813 return $value if $drive;
1814 die "unable to parse drive options\n";
1815 }
1816 PVE::JSONSchema::check_format($fmt, $value);
1817 return $value;
1818 }
1819 $value =~ s/^\"(.*)\"$/$1/;
1820 return $value;
1821 } else {
1822 die "internal error"
1823 }
1824 }
1825
1826 sub lock_config_full {
1827 my ($vmid, $timeout, $code, @param) = @_;
1828
1829 my $filename = config_file_lock($vmid);
1830
1831 my $res = lock_file($filename, $timeout, $code, @param);
1832
1833 die $@ if $@;
1834
1835 return $res;
1836 }
1837
1838 sub lock_config_mode {
1839 my ($vmid, $timeout, $shared, $code, @param) = @_;
1840
1841 my $filename = config_file_lock($vmid);
1842
1843 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1844
1845 die $@ if $@;
1846
1847 return $res;
1848 }
1849
1850 sub lock_config {
1851 my ($vmid, $code, @param) = @_;
1852
1853 return lock_config_full($vmid, 10, $code, @param);
1854 }
1855
1856 sub cfs_config_path {
1857 my ($vmid, $node) = @_;
1858
1859 $node = $nodename if !$node;
1860 return "nodes/$node/qemu-server/$vmid.conf";
1861 }
1862
1863 sub check_iommu_support{
1864 #fixme : need to check IOMMU support
1865 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1866
1867 my $iommu=1;
1868 return $iommu;
1869
1870 }
1871
1872 sub config_file {
1873 my ($vmid, $node) = @_;
1874
1875 my $cfspath = cfs_config_path($vmid, $node);
1876 return "/etc/pve/$cfspath";
1877 }
1878
1879 sub config_file_lock {
1880 my ($vmid) = @_;
1881
1882 return "$lock_dir/lock-$vmid.conf";
1883 }
1884
1885 sub touch_config {
1886 my ($vmid) = @_;
1887
1888 my $conf = config_file($vmid);
1889 utime undef, undef, $conf;
1890 }
1891
1892 sub destroy_vm {
1893 my ($storecfg, $vmid, $keep_empty_config) = @_;
1894
1895 my $conffile = config_file($vmid);
1896
1897 my $conf = load_config($vmid);
1898
1899 check_lock($conf);
1900
1901 # only remove disks owned by this VM
1902 foreach_drive($conf, sub {
1903 my ($ds, $drive) = @_;
1904
1905 return if drive_is_cdrom($drive);
1906
1907 my $volid = $drive->{file};
1908
1909 return if !$volid || $volid =~ m|^/|;
1910
1911 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1912 return if !$path || !$owner || ($owner != $vmid);
1913
1914 PVE::Storage::vdisk_free($storecfg, $volid);
1915 });
1916
1917 if ($keep_empty_config) {
1918 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1919 } else {
1920 unlink $conffile;
1921 }
1922
1923 # also remove unused disk
1924 eval {
1925 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1926
1927 eval {
1928 PVE::Storage::foreach_volid($dl, sub {
1929 my ($volid, $sid, $volname, $d) = @_;
1930 PVE::Storage::vdisk_free($storecfg, $volid);
1931 });
1932 };
1933 warn $@ if $@;
1934
1935 };
1936 warn $@ if $@;
1937 }
1938
1939 sub load_config {
1940 my ($vmid, $node) = @_;
1941
1942 my $cfspath = cfs_config_path($vmid, $node);
1943
1944 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1945
1946 die "no such VM ('$vmid')\n" if !defined($conf);
1947
1948 return $conf;
1949 }
1950
1951 sub parse_vm_config {
1952 my ($filename, $raw) = @_;
1953
1954 return undef if !defined($raw);
1955
1956 my $res = {
1957 digest => Digest::SHA::sha1_hex($raw),
1958 snapshots => {},
1959 pending => {},
1960 };
1961
1962 $filename =~ m|/qemu-server/(\d+)\.conf$|
1963 || die "got strange filename '$filename'";
1964
1965 my $vmid = $1;
1966
1967 my $conf = $res;
1968 my $descr = '';
1969 my $section = '';
1970
1971 my @lines = split(/\n/, $raw);
1972 foreach my $line (@lines) {
1973 next if $line =~ m/^\s*$/;
1974
1975 if ($line =~ m/^\[PENDING\]\s*$/i) {
1976 $section = 'pending';
1977 $conf->{description} = $descr if $descr;
1978 $descr = '';
1979 $conf = $res->{$section} = {};
1980 next;
1981
1982 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1983 $section = $1;
1984 $conf->{description} = $descr if $descr;
1985 $descr = '';
1986 $conf = $res->{snapshots}->{$section} = {};
1987 next;
1988 }
1989
1990 if ($line =~ m/^\#(.*)\s*$/) {
1991 $descr .= PVE::Tools::decode_text($1) . "\n";
1992 next;
1993 }
1994
1995 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1996 $descr .= PVE::Tools::decode_text($2);
1997 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1998 $conf->{snapstate} = $1;
1999 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2000 my $key = $1;
2001 my $value = $2;
2002 $conf->{$key} = $value;
2003 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2004 my $value = $1;
2005 if ($section eq 'pending') {
2006 $conf->{delete} = $value; # we parse this later
2007 } else {
2008 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2009 }
2010 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2011 my $key = $1;
2012 my $value = $2;
2013 eval { $value = check_type($key, $value); };
2014 if ($@) {
2015 warn "vm $vmid - unable to parse value of '$key' - $@";
2016 } else {
2017 my $fmt = $confdesc->{$key}->{format};
2018 if ($fmt && $fmt eq 'pve-qm-drive') {
2019 my $v = parse_drive($key, $value);
2020 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2021 $v->{file} = $volid;
2022 $value = print_drive($vmid, $v);
2023 } else {
2024 warn "vm $vmid - unable to parse value of '$key'\n";
2025 next;
2026 }
2027 }
2028
2029 if ($key eq 'cdrom') {
2030 $conf->{ide2} = $value;
2031 } else {
2032 $conf->{$key} = $value;
2033 }
2034 }
2035 }
2036 }
2037
2038 $conf->{description} = $descr if $descr;
2039
2040 delete $res->{snapstate}; # just to be sure
2041
2042 return $res;
2043 }
2044
2045 sub write_vm_config {
2046 my ($filename, $conf) = @_;
2047
2048 delete $conf->{snapstate}; # just to be sure
2049
2050 if ($conf->{cdrom}) {
2051 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2052 $conf->{ide2} = $conf->{cdrom};
2053 delete $conf->{cdrom};
2054 }
2055
2056 # we do not use 'smp' any longer
2057 if ($conf->{sockets}) {
2058 delete $conf->{smp};
2059 } elsif ($conf->{smp}) {
2060 $conf->{sockets} = $conf->{smp};
2061 delete $conf->{cores};
2062 delete $conf->{smp};
2063 }
2064
2065 my $used_volids = {};
2066
2067 my $cleanup_config = sub {
2068 my ($cref, $pending, $snapname) = @_;
2069
2070 foreach my $key (keys %$cref) {
2071 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2072 $key eq 'snapstate' || $key eq 'pending';
2073 my $value = $cref->{$key};
2074 if ($key eq 'delete') {
2075 die "propertry 'delete' is only allowed in [PENDING]\n"
2076 if !$pending;
2077 # fixme: check syntax?
2078 next;
2079 }
2080 eval { $value = check_type($key, $value); };
2081 die "unable to parse value of '$key' - $@" if $@;
2082
2083 $cref->{$key} = $value;
2084
2085 if (!$snapname && valid_drivename($key)) {
2086 my $drive = parse_drive($key, $value);
2087 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2088 }
2089 }
2090 };
2091
2092 &$cleanup_config($conf);
2093
2094 &$cleanup_config($conf->{pending}, 1);
2095
2096 foreach my $snapname (keys %{$conf->{snapshots}}) {
2097 die "internal error" if $snapname eq 'pending';
2098 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2099 }
2100
2101 # remove 'unusedX' settings if we re-add a volume
2102 foreach my $key (keys %$conf) {
2103 my $value = $conf->{$key};
2104 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2105 delete $conf->{$key};
2106 }
2107 }
2108
2109 my $generate_raw_config = sub {
2110 my ($conf) = @_;
2111
2112 my $raw = '';
2113
2114 # add description as comment to top of file
2115 my $descr = $conf->{description} || '';
2116 foreach my $cl (split(/\n/, $descr)) {
2117 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2118 }
2119
2120 foreach my $key (sort keys %$conf) {
2121 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2122 $raw .= "$key: $conf->{$key}\n";
2123 }
2124 return $raw;
2125 };
2126
2127 my $raw = &$generate_raw_config($conf);
2128
2129 if (scalar(keys %{$conf->{pending}})){
2130 $raw .= "\n[PENDING]\n";
2131 $raw .= &$generate_raw_config($conf->{pending});
2132 }
2133
2134 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2135 $raw .= "\n[$snapname]\n";
2136 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2137 }
2138
2139 return $raw;
2140 }
2141
2142 sub update_config_nolock {
2143 my ($vmid, $conf, $skiplock) = @_;
2144
2145 check_lock($conf) if !$skiplock;
2146
2147 my $cfspath = cfs_config_path($vmid);
2148
2149 PVE::Cluster::cfs_write_file($cfspath, $conf);
2150 }
2151
2152 sub update_config {
2153 my ($vmid, $conf, $skiplock) = @_;
2154
2155 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2156 }
2157
2158 sub load_defaults {
2159
2160 my $res = {};
2161
2162 # we use static defaults from our JSON schema configuration
2163 foreach my $key (keys %$confdesc) {
2164 if (defined(my $default = $confdesc->{$key}->{default})) {
2165 $res->{$key} = $default;
2166 }
2167 }
2168
2169 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2170 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2171
2172 return $res;
2173 }
2174
2175 sub config_list {
2176 my $vmlist = PVE::Cluster::get_vmlist();
2177 my $res = {};
2178 return $res if !$vmlist || !$vmlist->{ids};
2179 my $ids = $vmlist->{ids};
2180
2181 foreach my $vmid (keys %$ids) {
2182 my $d = $ids->{$vmid};
2183 next if !$d->{node} || $d->{node} ne $nodename;
2184 next if !$d->{type} || $d->{type} ne 'qemu';
2185 $res->{$vmid}->{exists} = 1;
2186 }
2187 return $res;
2188 }
2189
2190 # test if VM uses local resources (to prevent migration)
2191 sub check_local_resources {
2192 my ($conf, $noerr) = @_;
2193
2194 my $loc_res = 0;
2195
2196 $loc_res = 1 if $conf->{hostusb}; # old syntax
2197 $loc_res = 1 if $conf->{hostpci}; # old syntax
2198
2199 foreach my $k (keys %$conf) {
2200 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2201 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2202 }
2203
2204 die "VM uses local resources\n" if $loc_res && !$noerr;
2205
2206 return $loc_res;
2207 }
2208
2209 # check if used storages are available on all nodes (use by migrate)
2210 sub check_storage_availability {
2211 my ($storecfg, $conf, $node) = @_;
2212
2213 foreach_drive($conf, sub {
2214 my ($ds, $drive) = @_;
2215
2216 my $volid = $drive->{file};
2217 return if !$volid;
2218
2219 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2220 return if !$sid;
2221
2222 # check if storage is available on both nodes
2223 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2224 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2225 });
2226 }
2227
2228 # list nodes where all VM images are available (used by has_feature API)
2229 sub shared_nodes {
2230 my ($conf, $storecfg) = @_;
2231
2232 my $nodelist = PVE::Cluster::get_nodelist();
2233 my $nodehash = { map { $_ => 1 } @$nodelist };
2234 my $nodename = PVE::INotify::nodename();
2235
2236 foreach_drive($conf, sub {
2237 my ($ds, $drive) = @_;
2238
2239 my $volid = $drive->{file};
2240 return if !$volid;
2241
2242 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2243 if ($storeid) {
2244 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2245 if ($scfg->{disable}) {
2246 $nodehash = {};
2247 } elsif (my $avail = $scfg->{nodes}) {
2248 foreach my $node (keys %$nodehash) {
2249 delete $nodehash->{$node} if !$avail->{$node};
2250 }
2251 } elsif (!$scfg->{shared}) {
2252 foreach my $node (keys %$nodehash) {
2253 delete $nodehash->{$node} if $node ne $nodename
2254 }
2255 }
2256 }
2257 });
2258
2259 return $nodehash
2260 }
2261
2262 sub check_lock {
2263 my ($conf) = @_;
2264
2265 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2266 }
2267
2268 sub check_cmdline {
2269 my ($pidfile, $pid) = @_;
2270
2271 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2272 if (defined($fh)) {
2273 my $line = <$fh>;
2274 $fh->close;
2275 return undef if !$line;
2276 my @param = split(/\0/, $line);
2277
2278 my $cmd = $param[0];
2279 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2280
2281 for (my $i = 0; $i < scalar (@param); $i++) {
2282 my $p = $param[$i];
2283 next if !$p;
2284 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2285 my $p = $param[$i+1];
2286 return 1 if $p && ($p eq $pidfile);
2287 return undef;
2288 }
2289 }
2290 }
2291 return undef;
2292 }
2293
2294 sub check_running {
2295 my ($vmid, $nocheck, $node) = @_;
2296
2297 my $filename = config_file($vmid, $node);
2298
2299 die "unable to find configuration file for VM $vmid - no such machine\n"
2300 if !$nocheck && ! -f $filename;
2301
2302 my $pidfile = pidfile_name($vmid);
2303
2304 if (my $fd = IO::File->new("<$pidfile")) {
2305 my $st = stat($fd);
2306 my $line = <$fd>;
2307 close($fd);
2308
2309 my $mtime = $st->mtime;
2310 if ($mtime > time()) {
2311 warn "file '$filename' modified in future\n";
2312 }
2313
2314 if ($line =~ m/^(\d+)$/) {
2315 my $pid = $1;
2316 if (check_cmdline($pidfile, $pid)) {
2317 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2318 return $pid;
2319 }
2320 }
2321 }
2322 }
2323
2324 return undef;
2325 }
2326
2327 sub vzlist {
2328
2329 my $vzlist = config_list();
2330
2331 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2332
2333 while (defined(my $de = $fd->read)) {
2334 next if $de !~ m/^(\d+)\.pid$/;
2335 my $vmid = $1;
2336 next if !defined($vzlist->{$vmid});
2337 if (my $pid = check_running($vmid)) {
2338 $vzlist->{$vmid}->{pid} = $pid;
2339 }
2340 }
2341
2342 return $vzlist;
2343 }
2344
2345 sub disksize {
2346 my ($storecfg, $conf) = @_;
2347
2348 my $bootdisk = $conf->{bootdisk};
2349 return undef if !$bootdisk;
2350 return undef if !valid_drivename($bootdisk);
2351
2352 return undef if !$conf->{$bootdisk};
2353
2354 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2355 return undef if !defined($drive);
2356
2357 return undef if drive_is_cdrom($drive);
2358
2359 my $volid = $drive->{file};
2360 return undef if !$volid;
2361
2362 return $drive->{size};
2363 }
2364
2365 my $last_proc_pid_stat;
2366
2367 # get VM status information
2368 # This must be fast and should not block ($full == false)
2369 # We only query KVM using QMP if $full == true (this can be slow)
2370 sub vmstatus {
2371 my ($opt_vmid, $full) = @_;
2372
2373 my $res = {};
2374
2375 my $storecfg = PVE::Storage::config();
2376
2377 my $list = vzlist();
2378 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2379
2380 my $cpucount = $cpuinfo->{cpus} || 1;
2381
2382 foreach my $vmid (keys %$list) {
2383 next if $opt_vmid && ($vmid ne $opt_vmid);
2384
2385 my $cfspath = cfs_config_path($vmid);
2386 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2387
2388 my $d = {};
2389 $d->{pid} = $list->{$vmid}->{pid};
2390
2391 # fixme: better status?
2392 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2393
2394 my $size = disksize($storecfg, $conf);
2395 if (defined($size)) {
2396 $d->{disk} = 0; # no info available
2397 $d->{maxdisk} = $size;
2398 } else {
2399 $d->{disk} = 0;
2400 $d->{maxdisk} = 0;
2401 }
2402
2403 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2404 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2405
2406 $d->{name} = $conf->{name} || "VM $vmid";
2407 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2408
2409 if ($conf->{balloon}) {
2410 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2411 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2412 }
2413
2414 $d->{uptime} = 0;
2415 $d->{cpu} = 0;
2416 $d->{mem} = 0;
2417
2418 $d->{netout} = 0;
2419 $d->{netin} = 0;
2420
2421 $d->{diskread} = 0;
2422 $d->{diskwrite} = 0;
2423
2424 $d->{template} = is_template($conf);
2425
2426 $res->{$vmid} = $d;
2427 }
2428
2429 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2430 foreach my $dev (keys %$netdev) {
2431 next if $dev !~ m/^tap([1-9]\d*)i/;
2432 my $vmid = $1;
2433 my $d = $res->{$vmid};
2434 next if !$d;
2435
2436 $d->{netout} += $netdev->{$dev}->{receive};
2437 $d->{netin} += $netdev->{$dev}->{transmit};
2438 }
2439
2440 my $ctime = gettimeofday;
2441
2442 foreach my $vmid (keys %$list) {
2443
2444 my $d = $res->{$vmid};
2445 my $pid = $d->{pid};
2446 next if !$pid;
2447
2448 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2449 next if !$pstat; # not running
2450
2451 my $used = $pstat->{utime} + $pstat->{stime};
2452
2453 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2454
2455 if ($pstat->{vsize}) {
2456 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2457 }
2458
2459 my $old = $last_proc_pid_stat->{$pid};
2460 if (!$old) {
2461 $last_proc_pid_stat->{$pid} = {
2462 time => $ctime,
2463 used => $used,
2464 cpu => 0,
2465 };
2466 next;
2467 }
2468
2469 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2470
2471 if ($dtime > 1000) {
2472 my $dutime = $used - $old->{used};
2473
2474 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2475 $last_proc_pid_stat->{$pid} = {
2476 time => $ctime,
2477 used => $used,
2478 cpu => $d->{cpu},
2479 };
2480 } else {
2481 $d->{cpu} = $old->{cpu};
2482 }
2483 }
2484
2485 return $res if !$full;
2486
2487 my $qmpclient = PVE::QMPClient->new();
2488
2489 my $ballooncb = sub {
2490 my ($vmid, $resp) = @_;
2491
2492 my $info = $resp->{'return'};
2493 my $d = $res->{$vmid};
2494
2495 if (defined($info->{stats}->{"stat-total-memory"}) && defined($info->{stats}->{"stat-free-memory"})) {
2496 $d->{balloon} = int($info->{stats}->{"stat-total-memory"}/1024/1024);
2497 $d->{freemem} = int($info->{stats}->{"stat-free-memory"}/1024/1024);
2498 $d->{mem} = $d->{maxmem} - $d->{freemem};
2499 }
2500
2501 };
2502
2503 my $blockstatscb = sub {
2504 my ($vmid, $resp) = @_;
2505 my $data = $resp->{'return'} || [];
2506 my $totalrdbytes = 0;
2507 my $totalwrbytes = 0;
2508 for my $blockstat (@$data) {
2509 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2510 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2511 }
2512 $res->{$vmid}->{diskread} = $totalrdbytes;
2513 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2514 };
2515
2516 my $statuscb = sub {
2517 my ($vmid, $resp) = @_;
2518
2519 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2520 # this fails if ballon driver is not loaded, so this must be
2521 # the last commnand (following command are aborted if this fails).
2522 # maybe is it fixed by
2523 #http://git.qemu.org/?p=qemu.git;a=commit;h=38dbd48b247ebe05bdc6ef52ccdc60cc21274877
2524
2525 $qmpclient->queue_cmd($vmid, $ballooncb, 'qom-get',
2526 path => "machine/peripheral/balloon0",
2527 property => "guest-stats");
2528
2529
2530 my $status = 'unknown';
2531 if (!defined($status = $resp->{'return'}->{status})) {
2532 warn "unable to get VM status\n";
2533 return;
2534 }
2535
2536 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2537 };
2538
2539 foreach my $vmid (keys %$list) {
2540 next if $opt_vmid && ($vmid ne $opt_vmid);
2541 next if !$res->{$vmid}->{pid}; # not running
2542 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2543 }
2544
2545 $qmpclient->queue_execute(undef, 1);
2546
2547 foreach my $vmid (keys %$list) {
2548 next if $opt_vmid && ($vmid ne $opt_vmid);
2549 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2550 }
2551
2552 return $res;
2553 }
2554
2555 sub foreach_dimm {
2556 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2557
2558 my $dimm_id = 0;
2559 my $current_size = 1024;
2560 my $dimm_size = 512;
2561 return if $current_size == $memory;
2562
2563 for (my $j = 0; $j < 8; $j++) {
2564 for (my $i = 0; $i < 32; $i++) {
2565 my $name = "dimm${dimm_id}";
2566 $dimm_id++;
2567 my $numanode = $i % $sockets;
2568 $current_size += $dimm_size;
2569 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2570 return $current_size if $current_size >= $memory;
2571 }
2572 $dimm_size *= 2;
2573 }
2574 }
2575
2576 sub foreach_drive {
2577 my ($conf, $func) = @_;
2578
2579 foreach my $ds (keys %$conf) {
2580 next if !valid_drivename($ds);
2581
2582 my $drive = parse_drive($ds, $conf->{$ds});
2583 next if !$drive;
2584
2585 &$func($ds, $drive);
2586 }
2587 }
2588
2589 sub foreach_volid {
2590 my ($conf, $func) = @_;
2591
2592 my $volhash = {};
2593
2594 my $test_volid = sub {
2595 my ($volid, $is_cdrom) = @_;
2596
2597 return if !$volid;
2598
2599 $volhash->{$volid} = $is_cdrom || 0;
2600 };
2601
2602 foreach_drive($conf, sub {
2603 my ($ds, $drive) = @_;
2604 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2605 });
2606
2607 foreach my $snapname (keys %{$conf->{snapshots}}) {
2608 my $snap = $conf->{snapshots}->{$snapname};
2609 &$test_volid($snap->{vmstate}, 0);
2610 foreach_drive($snap, sub {
2611 my ($ds, $drive) = @_;
2612 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2613 });
2614 }
2615
2616 foreach my $volid (keys %$volhash) {
2617 &$func($volid, $volhash->{$volid});
2618 }
2619 }
2620
2621 sub vga_conf_has_spice {
2622 my ($vga) = @_;
2623
2624 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2625
2626 return $1 || 1;
2627 }
2628
2629 sub config_to_command {
2630 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2631
2632 my $cmd = [];
2633 my $globalFlags = [];
2634 my $machineFlags = [];
2635 my $rtcFlags = [];
2636 my $cpuFlags = [];
2637 my $devices = [];
2638 my $pciaddr = '';
2639 my $bridges = {};
2640 my $kvmver = kvm_user_version();
2641 my $vernum = 0; # unknown
2642 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2643 $vernum = $1*1000000+$2*1000;
2644 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2645 $vernum = $1*1000000+$2*1000+$3;
2646 }
2647
2648 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2649
2650 my $have_ovz = -f '/proc/vz/vestat';
2651
2652 my $q35 = machine_type_is_q35($conf);
2653 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2654
2655 push @$cmd, '/usr/bin/kvm';
2656
2657 push @$cmd, '-id', $vmid;
2658
2659 my $use_virtio = 0;
2660
2661 my $qmpsocket = qmp_socket($vmid);
2662 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2663 push @$cmd, '-mon', "chardev=qmp,mode=control";
2664
2665 my $socket = vnc_socket($vmid);
2666 push @$cmd, '-vnc', "unix:$socket,x509,password";
2667
2668 push @$cmd, '-pidfile' , pidfile_name($vmid);
2669
2670 push @$cmd, '-daemonize';
2671
2672 if ($conf->{smbios1}) {
2673 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2674 }
2675
2676 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2677
2678 if ($q35) {
2679 # the q35 chipset support native usb2, so we enable usb controller
2680 # by default for this machine type
2681 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2682 } else {
2683 $pciaddr = print_pci_addr("piix3", $bridges);
2684 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2685
2686 my $use_usb2 = 0;
2687 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2688 next if !$conf->{"usb$i"};
2689 $use_usb2 = 1;
2690 }
2691 # include usb device config
2692 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2693 }
2694
2695 my $vga = $conf->{vga};
2696
2697 my $qxlnum = vga_conf_has_spice($vga);
2698 $vga = 'qxl' if $qxlnum;
2699
2700 if (!$vga) {
2701 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2702 $conf->{ostype} eq 'win7' ||
2703 $conf->{ostype} eq 'w2k8')) {
2704 $vga = 'std';
2705 } else {
2706 $vga = 'cirrus';
2707 }
2708 }
2709
2710 # enable absolute mouse coordinates (needed by vnc)
2711 my $tablet;
2712 if (defined($conf->{tablet})) {
2713 $tablet = $conf->{tablet};
2714 } else {
2715 $tablet = $defaults->{tablet};
2716 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2717 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2718 }
2719
2720 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2721
2722 # host pci devices
2723 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2724 my $d = parse_hostpci($conf->{"hostpci$i"});
2725 next if !$d;
2726
2727 my $pcie = $d->{pcie};
2728 if($pcie){
2729 die "q35 machine model is not enabled" if !$q35;
2730 $pciaddr = print_pcie_addr("hostpci$i");
2731 }else{
2732 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2733 }
2734
2735 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2736 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2737 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2738 if ($xvga && $xvga ne '') {
2739 push @$cpuFlags, 'kvm=off';
2740 $vga = 'none';
2741 }
2742 $driver = "vfio-pci" if $xvga ne '';
2743 my $pcidevices = $d->{pciid};
2744 my $multifunction = 1 if @$pcidevices > 1;
2745
2746 my $j=0;
2747 foreach my $pcidevice (@$pcidevices) {
2748
2749 my $id = "hostpci$i";
2750 $id .= ".$j" if $multifunction;
2751 my $addr = $pciaddr;
2752 $addr .= ".$j" if $multifunction;
2753 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2754
2755 if($j == 0){
2756 $devicestr .= "$rombar$xvga";
2757 $devicestr .= ",multifunction=on" if $multifunction;
2758 }
2759
2760 push @$devices, '-device', $devicestr;
2761 $j++;
2762 }
2763 }
2764
2765 # usb devices
2766 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2767 my $d = parse_usb_device($conf->{"usb$i"});
2768 next if !$d;
2769 if ($d->{vendorid} && $d->{productid}) {
2770 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2771 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2772 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2773 } elsif ($d->{spice}) {
2774 # usb redir support for spice
2775 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2776 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2777 }
2778 }
2779
2780 # serial devices
2781 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2782 if (my $path = $conf->{"serial$i"}) {
2783 if ($path eq 'socket') {
2784 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2785 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2786 push @$devices, '-device', "isa-serial,chardev=serial$i";
2787 } else {
2788 die "no such serial device\n" if ! -c $path;
2789 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2790 push @$devices, '-device', "isa-serial,chardev=serial$i";
2791 }
2792 }
2793 }
2794
2795 # parallel devices
2796 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2797 if (my $path = $conf->{"parallel$i"}) {
2798 die "no such parallel device\n" if ! -c $path;
2799 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2800 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2801 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2802 }
2803 }
2804
2805 my $vmname = $conf->{name} || "vm$vmid";
2806
2807 push @$cmd, '-name', $vmname;
2808
2809 my $sockets = 1;
2810 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2811 $sockets = $conf->{sockets} if $conf->{sockets};
2812
2813 my $cores = $conf->{cores} || 1;
2814
2815 my $maxcpus = $sockets * $cores;
2816
2817 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2818
2819 my $allowed_vcpus = $cpuinfo->{cpus};
2820
2821 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2822 if ($allowed_vcpus < $maxcpus);
2823
2824 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2825
2826 push @$cmd, '-nodefaults';
2827
2828 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2829
2830 my $bootindex_hash = {};
2831 my $i = 1;
2832 foreach my $o (split(//, $bootorder)) {
2833 $bootindex_hash->{$o} = $i*100;
2834 $i++;
2835 }
2836
2837 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2838
2839 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2840
2841 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2842
2843 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2844
2845 # time drift fix
2846 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2847
2848 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2849 my $useLocaltime = $conf->{localtime};
2850
2851 if (my $ost = $conf->{ostype}) {
2852 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2853
2854 if ($ost =~ m/^w/) { # windows
2855 $useLocaltime = 1 if !defined($conf->{localtime});
2856
2857 # use time drift fix when acpi is enabled
2858 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2859 $tdf = 1 if !defined($conf->{tdf});
2860 }
2861 }
2862
2863 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2864 $ost eq 'wvista') {
2865 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2866 push @$cmd, '-no-hpet';
2867 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2868 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2869 }
2870
2871 if ($ost eq 'win7' || $ost eq 'win8') {
2872 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2873 }
2874 }
2875
2876 push @$rtcFlags, 'driftfix=slew' if $tdf;
2877
2878 if ($nokvm) {
2879 push @$machineFlags, 'accel=tcg';
2880 } else {
2881 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2882 }
2883
2884 my $machine_type = $forcemachine || $conf->{machine};
2885 if ($machine_type) {
2886 push @$machineFlags, "type=${machine_type}";
2887 }
2888
2889 if ($conf->{startdate}) {
2890 push @$rtcFlags, "base=$conf->{startdate}";
2891 } elsif ($useLocaltime) {
2892 push @$rtcFlags, 'base=localtime';
2893 }
2894
2895 my $cpu = $nokvm ? "qemu64" : "kvm64";
2896 $cpu = $conf->{cpu} if $conf->{cpu};
2897
2898 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2899
2900 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2901
2902 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2903
2904 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2905
2906 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2907
2908 # Note: enforce needs kernel 3.10, so we do not use it for now
2909 # push @$cmd, '-cpu', "$cpu,enforce";
2910 push @$cmd, '-cpu', $cpu;
2911
2912 my $memory = $conf->{memory} || $defaults->{memory};
2913 my $static_memory = 0;
2914 my $dimm_memory = 0;
2915
2916 if ($hotplug_features->{memory}) {
2917 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2918 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2919 $static_memory = $STATICMEM;
2920 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2921 $dimm_memory = $memory - $static_memory;
2922 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2923
2924 } else {
2925
2926 $static_memory = $memory;
2927 push @$cmd, '-m', $static_memory;
2928 }
2929
2930 if ($conf->{numa}) {
2931
2932 my $numa_totalmemory = undef;
2933 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2934 next if !$conf->{"numa$i"};
2935 my $numa = parse_numa($conf->{"numa$i"});
2936 next if !$numa;
2937 # memory
2938 die "missing numa node$i memory value\n" if !$numa->{memory};
2939 my $numa_memory = $numa->{memory};
2940 $numa_totalmemory += $numa_memory;
2941 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2942
2943 # cpus
2944 my $cpus_start = $numa->{cpus}->{start};
2945 die "missing numa node$i cpus\n" if !defined($cpus_start);
2946 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2947 my $cpus = $cpus_start;
2948 if (defined($cpus_end)) {
2949 $cpus .= "-$cpus_end";
2950 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2951 }
2952
2953 # hostnodes
2954 my $hostnodes_start = $numa->{hostnodes}->{start};
2955 if (defined($hostnodes_start)) {
2956 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2957 my $hostnodes = $hostnodes_start;
2958 if (defined($hostnodes_end)) {
2959 $hostnodes .= "-$hostnodes_end";
2960 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2961 }
2962
2963 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2964 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2965 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2966 }
2967
2968 # policy
2969 my $policy = $numa->{policy};
2970 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2971 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2972 }
2973
2974 push @$cmd, '-object', $numa_object;
2975 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2976 }
2977
2978 die "total memory for NUMA nodes must be equal to vm static memory\n"
2979 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2980
2981 #if no custom tology, we split memory and cores across numa nodes
2982 if(!$numa_totalmemory) {
2983
2984 my $numa_memory = ($static_memory / $sockets) . "M";
2985
2986 for (my $i = 0; $i < $sockets; $i++) {
2987
2988 my $cpustart = ($cores * $i);
2989 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2990 my $cpus = $cpustart;
2991 $cpus .= "-$cpuend" if $cpuend;
2992
2993 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2994 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2995 }
2996 }
2997 }
2998
2999 if ($hotplug_features->{memory}) {
3000 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
3001 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3002 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3003 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3004
3005 #if dimm_memory is not aligned to dimm map
3006 if($current_size > $memory) {
3007 $conf->{memory} = $current_size;
3008 update_config_nolock($vmid, $conf, 1);
3009 }
3010 });
3011 }
3012
3013 push @$cmd, '-S' if $conf->{freeze};
3014
3015 # set keyboard layout
3016 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3017 push @$cmd, '-k', $kb if $kb;
3018
3019 # enable sound
3020 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3021 #push @$cmd, '-soundhw', 'es1370';
3022 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3023
3024 if($conf->{agent}) {
3025 my $qgasocket = qmp_socket($vmid, 1);
3026 my $pciaddr = print_pci_addr("qga0", $bridges);
3027 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3028 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3029 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3030 }
3031
3032 my $spice_port;
3033
3034 if ($qxlnum) {
3035 if ($qxlnum > 1) {
3036 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3037 for(my $i = 1; $i < $qxlnum; $i++){
3038 my $pciaddr = print_pci_addr("vga$i", $bridges);
3039 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3040 }
3041 } else {
3042 # assume other OS works like Linux
3043 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3044 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3045 }
3046 }
3047
3048 my $pciaddr = print_pci_addr("spice", $bridges);
3049
3050 $spice_port = PVE::Tools::next_spice_port();
3051
3052 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3053
3054 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3055 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3056 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3057 }
3058
3059 # enable balloon by default, unless explicitly disabled
3060 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3061 $pciaddr = print_pci_addr("balloon0", $bridges);
3062 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3063 }
3064
3065 if ($conf->{watchdog}) {
3066 my $wdopts = parse_watchdog($conf->{watchdog});
3067 $pciaddr = print_pci_addr("watchdog", $bridges);
3068 my $watchdog = $wdopts->{model} || 'i6300esb';
3069 push @$devices, '-device', "$watchdog$pciaddr";
3070 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3071 }
3072
3073 my $vollist = [];
3074 my $scsicontroller = {};
3075 my $ahcicontroller = {};
3076 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3077
3078 # Add iscsi initiator name if available
3079 if (my $initiator = get_initiator_name()) {
3080 push @$devices, '-iscsi', "initiator-name=$initiator";
3081 }
3082
3083 foreach_drive($conf, sub {
3084 my ($ds, $drive) = @_;
3085
3086 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3087 push @$vollist, $drive->{file};
3088 }
3089
3090 $use_virtio = 1 if $ds =~ m/^virtio/;
3091
3092 if (drive_is_cdrom ($drive)) {
3093 if ($bootindex_hash->{d}) {
3094 $drive->{bootindex} = $bootindex_hash->{d};
3095 $bootindex_hash->{d} += 1;
3096 }
3097 } else {
3098 if ($bootindex_hash->{c}) {
3099 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3100 $bootindex_hash->{c} += 1;
3101 }
3102 }
3103
3104 if ($drive->{interface} eq 'scsi') {
3105
3106 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3107 my $controller = int($drive->{index} / $maxdev);
3108 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3109 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3110 $scsicontroller->{$controller}=1;
3111 }
3112
3113 if ($drive->{interface} eq 'sata') {
3114 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3115 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3116 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3117 $ahcicontroller->{$controller}=1;
3118 }
3119
3120 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3121 push @$devices, '-drive',$drive_cmd;
3122 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3123 });
3124
3125 for (my $i = 0; $i < $MAX_NETS; $i++) {
3126 next if !$conf->{"net$i"};
3127 my $d = parse_net($conf->{"net$i"});
3128 next if !$d;
3129
3130 $use_virtio = 1 if $d->{model} eq 'virtio';
3131
3132 if ($bootindex_hash->{n}) {
3133 $d->{bootindex} = $bootindex_hash->{n};
3134 $bootindex_hash->{n} += 1;
3135 }
3136
3137 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3138 push @$devices, '-netdev', $netdevfull;
3139
3140 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3141 push @$devices, '-device', $netdevicefull;
3142 }
3143
3144 if (!$q35) {
3145 # add pci bridges
3146 while (my ($k, $v) = each %$bridges) {
3147 $pciaddr = print_pci_addr("pci.$k");
3148 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3149 }
3150 }
3151
3152 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3153 # when the VM uses virtio devices.
3154 if (!$use_virtio && $have_ovz) {
3155
3156 my $cpuunits = defined($conf->{cpuunits}) ?
3157 $conf->{cpuunits} : $defaults->{cpuunits};
3158
3159 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3160
3161 # fixme: cpulimit is currently ignored
3162 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3163 }
3164
3165 # add custom args
3166 if ($conf->{args}) {
3167 my $aa = PVE::Tools::split_args($conf->{args});
3168 push @$cmd, @$aa;
3169 }
3170
3171 push @$cmd, @$devices;
3172 push @$cmd, '-rtc', join(',', @$rtcFlags)
3173 if scalar(@$rtcFlags);
3174 push @$cmd, '-machine', join(',', @$machineFlags)
3175 if scalar(@$machineFlags);
3176 push @$cmd, '-global', join(',', @$globalFlags)
3177 if scalar(@$globalFlags);
3178
3179 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3180 }
3181
3182 sub vnc_socket {
3183 my ($vmid) = @_;
3184 return "${var_run_tmpdir}/$vmid.vnc";
3185 }
3186
3187 sub spice_port {
3188 my ($vmid) = @_;
3189
3190 my $res = vm_mon_cmd($vmid, 'query-spice');
3191
3192 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3193 }
3194
3195 sub qmp_socket {
3196 my ($vmid, $qga) = @_;
3197 my $sockettype = $qga ? 'qga' : 'qmp';
3198 return "${var_run_tmpdir}/$vmid.$sockettype";
3199 }
3200
3201 sub pidfile_name {
3202 my ($vmid) = @_;
3203 return "${var_run_tmpdir}/$vmid.pid";
3204 }
3205
3206 sub vm_devices_list {
3207 my ($vmid) = @_;
3208
3209 my $res = vm_mon_cmd($vmid, 'query-pci');
3210 my $devices = {};
3211 foreach my $pcibus (@$res) {
3212 foreach my $device (@{$pcibus->{devices}}) {
3213 next if !$device->{'qdev_id'};
3214 if ($device->{'pci_bridge'}) {
3215 $devices->{$device->{'qdev_id'}} = 1;
3216 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3217 next if !$bridge_device->{'qdev_id'};
3218 $devices->{$bridge_device->{'qdev_id'}} = 1;
3219 $devices->{$device->{'qdev_id'}}++;
3220 }
3221 } else {
3222 $devices->{$device->{'qdev_id'}} = 1;
3223 }
3224 }
3225 }
3226
3227 my $resblock = vm_mon_cmd($vmid, 'query-block');
3228 foreach my $block (@$resblock) {
3229 if($block->{device} =~ m/^drive-(\S+)/){
3230 $devices->{$1} = 1;
3231 }
3232 }
3233
3234 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3235 foreach my $mice (@$resmice) {
3236 if ($mice->{name} eq 'QEMU HID Tablet') {
3237 $devices->{tablet} = 1;
3238 last;
3239 }
3240 }
3241
3242 return $devices;
3243 }
3244
3245 sub vm_deviceplug {
3246 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3247
3248 my $q35 = machine_type_is_q35($conf);
3249
3250 my $devices_list = vm_devices_list($vmid);
3251 return 1 if defined($devices_list->{$deviceid});
3252
3253 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3254
3255 if ($deviceid eq 'tablet') {
3256
3257 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3258
3259 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3260
3261 qemu_driveadd($storecfg, $vmid, $device);
3262 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3263
3264 qemu_deviceadd($vmid, $devicefull);
3265 eval { qemu_deviceaddverify($vmid, $deviceid); };
3266 if (my $err = $@) {
3267 eval { qemu_drivedel($vmid, $deviceid); };
3268 warn $@ if $@;
3269 die $err;
3270 }
3271
3272 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3273
3274 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3275 my $pciaddr = print_pci_addr($deviceid);
3276 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3277
3278 qemu_deviceadd($vmid, $devicefull);
3279 qemu_deviceaddverify($vmid, $deviceid);
3280
3281 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3282
3283 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3284 qemu_driveadd($storecfg, $vmid, $device);
3285
3286 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3287 eval { qemu_deviceadd($vmid, $devicefull); };
3288 if (my $err = $@) {
3289 eval { qemu_drivedel($vmid, $deviceid); };
3290 warn $@ if $@;
3291 die $err;
3292 }
3293
3294 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3295
3296 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3297 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3298 qemu_deviceadd($vmid, $netdevicefull);
3299 eval { qemu_deviceaddverify($vmid, $deviceid); };
3300 if (my $err = $@) {
3301 eval { qemu_netdevdel($vmid, $deviceid); };
3302 warn $@ if $@;
3303 die $err;
3304 }
3305
3306 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3307
3308 my $bridgeid = $2;
3309 my $pciaddr = print_pci_addr($deviceid);
3310 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3311
3312 qemu_deviceadd($vmid, $devicefull);
3313 qemu_deviceaddverify($vmid, $deviceid);
3314
3315 } else {
3316 die "can't hotplug device '$deviceid'\n";
3317 }
3318
3319 return 1;
3320 }
3321
3322 # fixme: this should raise exceptions on error!
3323 sub vm_deviceunplug {
3324 my ($vmid, $conf, $deviceid) = @_;
3325
3326 my $devices_list = vm_devices_list($vmid);
3327 return 1 if !defined($devices_list->{$deviceid});
3328
3329 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3330
3331 if ($deviceid eq 'tablet') {
3332
3333 qemu_devicedel($vmid, $deviceid);
3334
3335 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3336
3337 qemu_devicedel($vmid, $deviceid);
3338 qemu_devicedelverify($vmid, $deviceid);
3339 qemu_drivedel($vmid, $deviceid);
3340
3341 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3342
3343 qemu_devicedel($vmid, $deviceid);
3344
3345 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3346
3347 qemu_devicedel($vmid, $deviceid);
3348 qemu_drivedel($vmid, $deviceid);
3349
3350 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3351
3352 qemu_devicedel($vmid, $deviceid);
3353 qemu_devicedelverify($vmid, $deviceid);
3354 qemu_netdevdel($vmid, $deviceid);
3355
3356 } else {
3357 die "can't unplug device '$deviceid'\n";
3358 }
3359
3360 return 1;
3361 }
3362
3363 sub qemu_deviceadd {
3364 my ($vmid, $devicefull) = @_;
3365
3366 $devicefull = "driver=".$devicefull;
3367 my %options = split(/[=,]/, $devicefull);
3368
3369 vm_mon_cmd($vmid, "device_add" , %options);
3370 }
3371
3372 sub qemu_devicedel {
3373 my ($vmid, $deviceid) = @_;
3374
3375 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3376 }
3377
3378 sub qemu_objectadd {
3379 my($vmid, $objectid, $qomtype) = @_;
3380
3381 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3382
3383 return 1;
3384 }
3385
3386 sub qemu_objectdel {
3387 my($vmid, $objectid) = @_;
3388
3389 vm_mon_cmd($vmid, "object-del", id => $objectid);
3390
3391 return 1;
3392 }
3393
3394 sub qemu_driveadd {
3395 my ($storecfg, $vmid, $device) = @_;
3396
3397 my $drive = print_drive_full($storecfg, $vmid, $device);
3398 $drive =~ s/\\/\\\\/g;
3399 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3400
3401 # If the command succeeds qemu prints: "OK"
3402 return 1 if $ret =~ m/OK/s;
3403
3404 die "adding drive failed: $ret\n";
3405 }
3406
3407 sub qemu_drivedel {
3408 my($vmid, $deviceid) = @_;
3409
3410 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3411 $ret =~ s/^\s+//;
3412
3413 return 1 if $ret eq "";
3414
3415 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3416 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3417
3418 die "deleting drive $deviceid failed : $ret\n";
3419 }
3420
3421 sub qemu_deviceaddverify {
3422 my ($vmid, $deviceid) = @_;
3423
3424 for (my $i = 0; $i <= 5; $i++) {
3425 my $devices_list = vm_devices_list($vmid);
3426 return 1 if defined($devices_list->{$deviceid});
3427 sleep 1;
3428 }
3429
3430 die "error on hotplug device '$deviceid'\n";
3431 }
3432
3433
3434 sub qemu_devicedelverify {
3435 my ($vmid, $deviceid) = @_;
3436
3437 # need to verify that the device is correctly removed as device_del
3438 # is async and empty return is not reliable
3439
3440 for (my $i = 0; $i <= 5; $i++) {
3441 my $devices_list = vm_devices_list($vmid);
3442 return 1 if !defined($devices_list->{$deviceid});
3443 sleep 1;
3444 }
3445
3446 die "error on hot-unplugging device '$deviceid'\n";
3447 }
3448
3449 sub qemu_findorcreatescsihw {
3450 my ($storecfg, $conf, $vmid, $device) = @_;
3451
3452 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3453 my $controller = int($device->{index} / $maxdev);
3454 my $scsihwid="scsihw$controller";
3455 my $devices_list = vm_devices_list($vmid);
3456
3457 if(!defined($devices_list->{$scsihwid})) {
3458 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3459 }
3460
3461 return 1;
3462 }
3463
3464 sub qemu_add_pci_bridge {
3465 my ($storecfg, $conf, $vmid, $device) = @_;
3466
3467 my $bridges = {};
3468
3469 my $bridgeid;
3470
3471 print_pci_addr($device, $bridges);
3472
3473 while (my ($k, $v) = each %$bridges) {
3474 $bridgeid = $k;
3475 }
3476 return 1 if !defined($bridgeid) || $bridgeid < 1;
3477
3478 my $bridge = "pci.$bridgeid";
3479 my $devices_list = vm_devices_list($vmid);
3480
3481 if (!defined($devices_list->{$bridge})) {
3482 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3483 }
3484
3485 return 1;
3486 }
3487
3488 sub qemu_set_link_status {
3489 my ($vmid, $device, $up) = @_;
3490
3491 vm_mon_cmd($vmid, "set_link", name => $device,
3492 up => $up ? JSON::true : JSON::false);
3493 }
3494
3495 sub qemu_netdevadd {
3496 my ($vmid, $conf, $device, $deviceid) = @_;
3497
3498 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3499 my %options = split(/[=,]/, $netdev);
3500
3501 vm_mon_cmd($vmid, "netdev_add", %options);
3502 return 1;
3503 }
3504
3505 sub qemu_netdevdel {
3506 my ($vmid, $deviceid) = @_;
3507
3508 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3509 }
3510
3511 sub qemu_cpu_hotplug {
3512 my ($vmid, $conf, $vcpus) = @_;
3513
3514 my $sockets = 1;
3515 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3516 $sockets = $conf->{sockets} if $conf->{sockets};
3517 my $cores = $conf->{cores} || 1;
3518 my $maxcpus = $sockets * $cores;
3519
3520 $vcpus = $maxcpus if !$vcpus;
3521
3522 die "you can't add more vcpus than maxcpus\n"
3523 if $vcpus > $maxcpus;
3524
3525 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3526 die "online cpu unplug is not yet possible\n"
3527 if $vcpus < $currentvcpus;
3528
3529 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3530 die "vcpus in running vm is different than configuration\n"
3531 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3532
3533 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3534 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3535 }
3536 }
3537
3538 sub qemu_memory_hotplug {
3539 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3540
3541 return $value if !check_running($vmid);
3542
3543 my $memory = $conf->{memory} || $defaults->{memory};
3544 $value = $defaults->{memory} if !$value;
3545 return $value if $value == $memory;
3546
3547 my $static_memory = $STATICMEM;
3548 my $dimm_memory = $memory - $static_memory;
3549
3550 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3551 die "memory unplug is not yet available" if $value < $memory;
3552 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3553
3554
3555 my $sockets = 1;
3556 $sockets = $conf->{sockets} if $conf->{sockets};
3557
3558 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3559 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3560
3561 return if $current_size <= $conf->{memory};
3562
3563 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3564 if (my $err = $@) {
3565 eval { qemu_objectdel($vmid, "mem-$name"); };
3566 die $err;
3567 }
3568
3569 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3570 if (my $err = $@) {
3571 eval { qemu_objectdel($vmid, "mem-$name"); };
3572 die $err;
3573 }
3574 #update conf after each succesful module hotplug
3575 $conf->{memory} = $current_size;
3576 update_config_nolock($vmid, $conf, 1);
3577 });
3578 }
3579
3580 sub qemu_block_set_io_throttle {
3581 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3582
3583 return if !check_running($vmid) ;
3584
3585 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($iops_wr));
3586
3587 }
3588
3589 # old code, only used to shutdown old VM after update
3590 sub __read_avail {
3591 my ($fh, $timeout) = @_;
3592
3593 my $sel = new IO::Select;
3594 $sel->add($fh);
3595
3596 my $res = '';
3597 my $buf;
3598
3599 my @ready;
3600 while (scalar (@ready = $sel->can_read($timeout))) {
3601 my $count;
3602 if ($count = $fh->sysread($buf, 8192)) {
3603 if ($buf =~ /^(.*)\(qemu\) $/s) {
3604 $res .= $1;
3605 last;
3606 } else {
3607 $res .= $buf;
3608 }
3609 } else {
3610 if (!defined($count)) {
3611 die "$!\n";
3612 }
3613 last;
3614 }
3615 }
3616
3617 die "monitor read timeout\n" if !scalar(@ready);
3618
3619 return $res;
3620 }
3621
3622 # old code, only used to shutdown old VM after update
3623 sub vm_monitor_command {
3624 my ($vmid, $cmdstr, $nocheck) = @_;
3625
3626 my $res;
3627
3628 eval {
3629 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3630
3631 my $sname = "${var_run_tmpdir}/$vmid.mon";
3632
3633 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3634 die "unable to connect to VM $vmid socket - $!\n";
3635
3636 my $timeout = 3;
3637
3638 # hack: migrate sometime blocks the monitor (when migrate_downtime
3639 # is set)
3640 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3641 $timeout = 60*60; # 1 hour
3642 }
3643
3644 # read banner;
3645 my $data = __read_avail($sock, $timeout);
3646
3647 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3648 die "got unexpected qemu monitor banner\n";
3649 }
3650
3651 my $sel = new IO::Select;
3652 $sel->add($sock);
3653
3654 if (!scalar(my @ready = $sel->can_write($timeout))) {
3655 die "monitor write error - timeout";
3656 }
3657
3658 my $fullcmd = "$cmdstr\r";
3659
3660 # syslog('info', "VM $vmid monitor command: $cmdstr");
3661
3662 my $b;
3663 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3664 die "monitor write error - $!";
3665 }
3666
3667 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3668
3669 $timeout = 20;
3670
3671 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3672 $timeout = 60*60; # 1 hour
3673 } elsif ($cmdstr =~ m/^(eject|change)/) {
3674 $timeout = 60; # note: cdrom mount command is slow
3675 }
3676 if ($res = __read_avail($sock, $timeout)) {
3677
3678 my @lines = split("\r?\n", $res);
3679
3680 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3681
3682 $res = join("\n", @lines);
3683 $res .= "\n";
3684 }
3685 };
3686
3687 my $err = $@;
3688
3689 if ($err) {
3690 syslog("err", "VM $vmid monitor command failed - $err");
3691 die $err;
3692 }
3693
3694 return $res;
3695 }
3696
3697 sub qemu_block_resize {
3698 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3699
3700 my $running = check_running($vmid);
3701
3702 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3703
3704 return if !$running;
3705
3706 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3707
3708 }
3709
3710 sub qemu_volume_snapshot {
3711 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3712
3713 my $running = check_running($vmid);
3714
3715 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3716
3717 return if !$running;
3718
3719 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3720
3721 }
3722
3723 sub qemu_volume_snapshot_delete {
3724 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3725
3726 my $running = check_running($vmid);
3727
3728 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3729
3730 return if !$running;
3731
3732 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3733 }
3734
3735 sub set_migration_caps {
3736 my ($vmid) = @_;
3737
3738 my $cap_ref = [];
3739
3740 my $enabled_cap = {
3741 "auto-converge" => 1,
3742 "xbzrle" => 0,
3743 "x-rdma-pin-all" => 0,
3744 "zero-blocks" => 0,
3745 };
3746
3747 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3748
3749 for my $supported_capability (@$supported_capabilities) {
3750 push @$cap_ref, {
3751 capability => $supported_capability->{capability},
3752 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3753 };
3754 }
3755
3756 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3757 }
3758
3759 my $fast_plug_option = {
3760 'lock' => 1,
3761 'name' => 1,
3762 'onboot' => 1,
3763 'shares' => 1,
3764 'startup' => 1,
3765 };
3766
3767 # hotplug changes in [PENDING]
3768 # $selection hash can be used to only apply specified options, for
3769 # example: { cores => 1 } (only apply changed 'cores')
3770 # $errors ref is used to return error messages
3771 sub vmconfig_hotplug_pending {
3772 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3773
3774 my $defaults = load_defaults();
3775
3776 # commit values which do not have any impact on running VM first
3777 # Note: those option cannot raise errors, we we do not care about
3778 # $selection and always apply them.
3779
3780 my $add_error = sub {
3781 my ($opt, $msg) = @_;
3782 $errors->{$opt} = "hotplug problem - $msg";
3783 };
3784
3785 my $changes = 0;
3786 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3787 if ($fast_plug_option->{$opt}) {
3788 $conf->{$opt} = $conf->{pending}->{$opt};
3789 delete $conf->{pending}->{$opt};
3790 $changes = 1;
3791 }
3792 }
3793
3794 if ($changes) {
3795 update_config_nolock($vmid, $conf, 1);
3796 $conf = load_config($vmid); # update/reload
3797 }
3798
3799 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3800
3801 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3802 foreach my $opt (@delete) {
3803 next if $selection && !$selection->{$opt};
3804 eval {
3805 if ($opt eq 'hotplug') {
3806 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3807 } elsif ($opt eq 'tablet') {
3808 die "skip\n" if !$hotplug_features->{usb};
3809 if ($defaults->{tablet}) {
3810 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3811 } else {
3812 vm_deviceunplug($vmid, $conf, $opt);
3813 }
3814 } elsif ($opt eq 'vcpus') {
3815 die "skip\n" if !$hotplug_features->{cpu};
3816 qemu_cpu_hotplug($vmid, $conf, undef);
3817 } elsif ($opt eq 'balloon') {
3818 # enable balloon device is not hotpluggable
3819 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3820 } elsif ($fast_plug_option->{$opt}) {
3821 # do nothing
3822 } elsif ($opt =~ m/^net(\d+)$/) {
3823 die "skip\n" if !$hotplug_features->{network};
3824 vm_deviceunplug($vmid, $conf, $opt);
3825 } elsif (valid_drivename($opt)) {
3826 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3827 vm_deviceunplug($vmid, $conf, $opt);
3828 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3829 } elsif ($opt =~ m/^memory$/) {
3830 die "skip\n" if !$hotplug_features->{memory};
3831 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3832 } else {
3833 die "skip\n";
3834 }
3835 };
3836 if (my $err = $@) {
3837 &$add_error($opt, $err) if $err ne "skip\n";
3838 } else {
3839 # save new config if hotplug was successful
3840 delete $conf->{$opt};
3841 vmconfig_undelete_pending_option($conf, $opt);
3842 update_config_nolock($vmid, $conf, 1);
3843 $conf = load_config($vmid); # update/reload
3844 }
3845 }
3846
3847 foreach my $opt (keys %{$conf->{pending}}) {
3848 next if $selection && !$selection->{$opt};
3849 my $value = $conf->{pending}->{$opt};
3850 eval {
3851 if ($opt eq 'hotplug') {
3852 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3853 } elsif ($opt eq 'tablet') {
3854 die "skip\n" if !$hotplug_features->{usb};
3855 if ($value == 1) {
3856 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3857 } elsif ($value == 0) {
3858 vm_deviceunplug($vmid, $conf, $opt);
3859 }
3860 } elsif ($opt eq 'vcpus') {
3861 die "skip\n" if !$hotplug_features->{cpu};
3862 qemu_cpu_hotplug($vmid, $conf, $value);
3863 } elsif ($opt eq 'balloon') {
3864 # enable/disable balloning device is not hotpluggable
3865 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3866 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3867 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3868
3869 # allow manual ballooning if shares is set to zero
3870 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3871 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3872 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3873 }
3874 } elsif ($opt =~ m/^net(\d+)$/) {
3875 # some changes can be done without hotplug
3876 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3877 $vmid, $opt, $value);
3878 } elsif (valid_drivename($opt)) {
3879 # some changes can be done without hotplug
3880 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3881 $vmid, $opt, $value, 1);
3882 } elsif ($opt =~ m/^memory$/) { #dimms
3883 die "skip\n" if !$hotplug_features->{memory};
3884 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3885 } else {
3886 die "skip\n"; # skip non-hot-pluggable options
3887 }
3888 };
3889 if (my $err = $@) {
3890 &$add_error($opt, $err) if $err ne "skip\n";
3891 } else {
3892 # save new config if hotplug was successful
3893 $conf->{$opt} = $value;
3894 delete $conf->{pending}->{$opt};
3895 update_config_nolock($vmid, $conf, 1);
3896 $conf = load_config($vmid); # update/reload
3897 }
3898 }
3899 }
3900
3901 sub vmconfig_apply_pending {
3902 my ($vmid, $conf, $storecfg) = @_;
3903
3904 # cold plug
3905
3906 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3907 foreach my $opt (@delete) { # delete
3908 die "internal error" if $opt =~ m/^unused/;
3909 $conf = load_config($vmid); # update/reload
3910 if (!defined($conf->{$opt})) {
3911 vmconfig_undelete_pending_option($conf, $opt);
3912 update_config_nolock($vmid, $conf, 1);
3913 } elsif (valid_drivename($opt)) {
3914 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3915 vmconfig_undelete_pending_option($conf, $opt);
3916 delete $conf->{$opt};
3917 update_config_nolock($vmid, $conf, 1);
3918 } else {
3919 vmconfig_undelete_pending_option($conf, $opt);
3920 delete $conf->{$opt};
3921 update_config_nolock($vmid, $conf, 1);
3922 }
3923 }
3924
3925 $conf = load_config($vmid); # update/reload
3926
3927 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3928 $conf = load_config($vmid); # update/reload
3929
3930 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3931 # skip if nothing changed
3932 } elsif (valid_drivename($opt)) {
3933 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3934 if defined($conf->{$opt});
3935 $conf->{$opt} = $conf->{pending}->{$opt};
3936 } else {
3937 $conf->{$opt} = $conf->{pending}->{$opt};
3938 }
3939
3940 delete $conf->{pending}->{$opt};
3941 update_config_nolock($vmid, $conf, 1);
3942 }
3943 }
3944
3945 my $safe_num_ne = sub {
3946 my ($a, $b) = @_;
3947
3948 return 0 if !defined($a) && !defined($b);
3949 return 1 if !defined($a);
3950 return 1 if !defined($b);
3951
3952 return $a != $b;
3953 };
3954
3955 my $safe_string_ne = sub {
3956 my ($a, $b) = @_;
3957
3958 return 0 if !defined($a) && !defined($b);
3959 return 1 if !defined($a);
3960 return 1 if !defined($b);
3961
3962 return $a ne $b;
3963 };
3964
3965 sub vmconfig_update_net {
3966 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3967
3968 my $newnet = parse_net($value);
3969
3970 if ($conf->{$opt}) {
3971 my $oldnet = parse_net($conf->{$opt});
3972
3973 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3974 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3975 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3976 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3977
3978 # for non online change, we try to hot-unplug
3979 die "skip\n" if !$hotplug;
3980 vm_deviceunplug($vmid, $conf, $opt);
3981 } else {
3982
3983 die "internal error" if $opt !~ m/net(\d+)/;
3984 my $iface = "tap${vmid}i$1";
3985
3986 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3987 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3988 }
3989
3990 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3991 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3992 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3993 PVE::Network::tap_unplug($iface);
3994 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3995 }
3996
3997 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3998 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3999 }
4000
4001 return 1;
4002 }
4003 }
4004
4005 if ($hotplug) {
4006 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4007 } else {
4008 die "skip\n";
4009 }
4010 }
4011
4012 sub vmconfig_update_disk {
4013 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4014
4015 # fixme: do we need force?
4016
4017 my $drive = parse_drive($opt, $value);
4018
4019 if ($conf->{$opt}) {
4020
4021 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4022
4023 my $media = $drive->{media} || 'disk';
4024 my $oldmedia = $old_drive->{media} || 'disk';
4025 die "unable to change media type\n" if $media ne $oldmedia;
4026
4027 if (!drive_is_cdrom($old_drive)) {
4028
4029 if ($drive->{file} ne $old_drive->{file}) {
4030
4031 die "skip\n" if !$hotplug;
4032
4033 # unplug and register as unused
4034 vm_deviceunplug($vmid, $conf, $opt);
4035 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4036
4037 } else {
4038 # update existing disk
4039
4040 # skip non hotpluggable value
4041 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4042 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4043 die "skip\n";
4044 }
4045
4046 # apply throttle
4047 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4048 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4049 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4050 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4051 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4052 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4053 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4054 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4055 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4056 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4057 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4058 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4059
4060 qemu_block_set_io_throttle($vmid,"drive-$opt",
4061 ($drive->{mbps} || 0)*1024*1024,
4062 ($drive->{mbps_rd} || 0)*1024*1024,
4063 ($drive->{mbps_wr} || 0)*1024*1024,
4064 $drive->{iops} || 0,
4065 $drive->{iops_rd} || 0,
4066 $drive->{iops_wr} || 0,
4067 ($drive->{mbps_max} || 0)*1024*1024,
4068 ($drive->{mbps_rd_max} || 0)*1024*1024,
4069 ($drive->{mbps_wr_max} || 0)*1024*1024,
4070 $drive->{iops_max} || 0,
4071 $drive->{iops_rd_max} || 0,
4072 $drive->{iops_wr_max} || 0);
4073
4074 }
4075
4076 return 1;
4077 }
4078
4079 } else { # cdrom
4080
4081 if ($drive->{file} eq 'none') {
4082 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4083 } else {
4084 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4085 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4086 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4087 }
4088
4089 return 1;
4090 }
4091 }
4092 }
4093
4094 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4095 # hotplug new disks
4096 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4097 }
4098
4099 sub vm_start {
4100 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4101
4102 lock_config($vmid, sub {
4103 my $conf = load_config($vmid, $migratedfrom);
4104
4105 die "you can't start a vm if it's a template\n" if is_template($conf);
4106
4107 check_lock($conf) if !$skiplock;
4108
4109 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4110
4111 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4112 vmconfig_apply_pending($vmid, $conf, $storecfg);
4113 $conf = load_config($vmid); # update/reload
4114 }
4115
4116 my $defaults = load_defaults();
4117
4118 # set environment variable useful inside network script
4119 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4120
4121 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4122
4123 my $migrate_port = 0;
4124 my $migrate_uri;
4125 if ($statefile) {
4126 if ($statefile eq 'tcp') {
4127 my $localip = "localhost";
4128 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4129 if ($datacenterconf->{migration_unsecure}) {
4130 my $nodename = PVE::INotify::nodename();
4131 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4132 }
4133 $migrate_port = PVE::Tools::next_migrate_port();
4134 $migrate_uri = "tcp:${localip}:${migrate_port}";
4135 push @$cmd, '-incoming', $migrate_uri;
4136 push @$cmd, '-S';
4137 } else {
4138 push @$cmd, '-loadstate', $statefile;
4139 }
4140 } elsif ($paused) {
4141 push @$cmd, '-S';
4142 }
4143
4144 # host pci devices
4145 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4146 my $d = parse_hostpci($conf->{"hostpci$i"});
4147 next if !$d;
4148 my $pcidevices = $d->{pciid};
4149 foreach my $pcidevice (@$pcidevices) {
4150 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4151
4152 my $info = pci_device_info("0000:$pciid");
4153 die "IOMMU not present\n" if !check_iommu_support();
4154 die "no pci device info for device '$pciid'\n" if !$info;
4155
4156 if ($d->{driver} && $d->{driver} eq "vfio") {
4157 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4158 } else {
4159 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4160 }
4161
4162 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4163 }
4164 }
4165
4166 PVE::Storage::activate_volumes($storecfg, $vollist);
4167
4168 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4169 umask => 0077); };
4170 my $err = $@;
4171 die "start failed: $err" if $err;
4172
4173 print "migration listens on $migrate_uri\n" if $migrate_uri;
4174
4175 if ($statefile && $statefile ne 'tcp') {
4176 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4177 warn $@ if $@;
4178 }
4179
4180 if ($migratedfrom) {
4181
4182 eval {
4183 set_migration_caps($vmid);
4184 };
4185 warn $@ if $@;
4186
4187 if ($spice_port) {
4188 print "spice listens on port $spice_port\n";
4189 if ($spice_ticket) {
4190 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4191 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4192 }
4193 }
4194
4195 } else {
4196
4197 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4198 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4199 if $conf->{balloon};
4200 vm_mon_cmd_nocheck($vmid, 'qom-set',
4201 path => "machine/peripheral/balloon0",
4202 property => "guest-stats-polling-interval",
4203 value => 2);
4204 }
4205
4206 foreach my $opt (keys %$conf) {
4207 next if $opt !~ m/^net\d+$/;
4208 my $nicconf = parse_net($conf->{$opt});
4209 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4210 }
4211 }
4212 });
4213 }
4214
4215 sub vm_mon_cmd {
4216 my ($vmid, $execute, %params) = @_;
4217
4218 my $cmd = { execute => $execute, arguments => \%params };
4219 vm_qmp_command($vmid, $cmd);
4220 }
4221
4222 sub vm_mon_cmd_nocheck {
4223 my ($vmid, $execute, %params) = @_;
4224
4225 my $cmd = { execute => $execute, arguments => \%params };
4226 vm_qmp_command($vmid, $cmd, 1);
4227 }
4228
4229 sub vm_qmp_command {
4230 my ($vmid, $cmd, $nocheck) = @_;
4231
4232 my $res;
4233
4234 my $timeout;
4235 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4236 $timeout = $cmd->{arguments}->{timeout};
4237 delete $cmd->{arguments}->{timeout};
4238 }
4239
4240 eval {
4241 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4242 my $sname = qmp_socket($vmid);
4243 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4244 my $qmpclient = PVE::QMPClient->new();
4245
4246 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4247 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4248 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4249 if scalar(%{$cmd->{arguments}});
4250 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4251 } else {
4252 die "unable to open monitor socket\n";
4253 }
4254 };
4255 if (my $err = $@) {
4256 syslog("err", "VM $vmid qmp command failed - $err");
4257 die $err;
4258 }
4259
4260 return $res;
4261 }
4262
4263 sub vm_human_monitor_command {
4264 my ($vmid, $cmdline) = @_;
4265
4266 my $res;
4267
4268 my $cmd = {
4269 execute => 'human-monitor-command',
4270 arguments => { 'command-line' => $cmdline},
4271 };
4272
4273 return vm_qmp_command($vmid, $cmd);
4274 }
4275
4276 sub vm_commandline {
4277 my ($storecfg, $vmid) = @_;
4278
4279 my $conf = load_config($vmid);
4280
4281 my $defaults = load_defaults();
4282
4283 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4284
4285 return join(' ', @$cmd);
4286 }
4287
4288 sub vm_reset {
4289 my ($vmid, $skiplock) = @_;
4290
4291 lock_config($vmid, sub {
4292
4293 my $conf = load_config($vmid);
4294
4295 check_lock($conf) if !$skiplock;
4296
4297 vm_mon_cmd($vmid, "system_reset");
4298 });
4299 }
4300
4301 sub get_vm_volumes {
4302 my ($conf) = @_;
4303
4304 my $vollist = [];
4305 foreach_volid($conf, sub {
4306 my ($volid, $is_cdrom) = @_;
4307
4308 return if $volid =~ m|^/|;
4309
4310 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4311 return if !$sid;
4312
4313 push @$vollist, $volid;
4314 });
4315
4316 return $vollist;
4317 }
4318
4319 sub vm_stop_cleanup {
4320 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4321
4322 eval {
4323 fairsched_rmnod($vmid); # try to destroy group
4324
4325 if (!$keepActive) {
4326 my $vollist = get_vm_volumes($conf);
4327 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4328 }
4329
4330 foreach my $ext (qw(mon qmp pid vnc qga)) {
4331 unlink "/var/run/qemu-server/${vmid}.$ext";
4332 }
4333
4334 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4335 };
4336 warn $@ if $@; # avoid errors - just warn
4337 }
4338
4339 # Note: use $nockeck to skip tests if VM configuration file exists.
4340 # We need that when migration VMs to other nodes (files already moved)
4341 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4342 sub vm_stop {
4343 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4344
4345 $force = 1 if !defined($force) && !$shutdown;
4346
4347 if ($migratedfrom){
4348 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4349 kill 15, $pid if $pid;
4350 my $conf = load_config($vmid, $migratedfrom);
4351 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4352 return;
4353 }
4354
4355 lock_config($vmid, sub {
4356
4357 my $pid = check_running($vmid, $nocheck);
4358 return if !$pid;
4359
4360 my $conf;
4361 if (!$nocheck) {
4362 $conf = load_config($vmid);
4363 check_lock($conf) if !$skiplock;
4364 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4365 my $opts = parse_startup($conf->{startup});
4366 $timeout = $opts->{down} if $opts->{down};
4367 }
4368 }
4369
4370 $timeout = 60 if !defined($timeout);
4371
4372 eval {
4373 if ($shutdown) {
4374 if (defined($conf) && $conf->{agent}) {
4375 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4376 } else {
4377 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4378 }
4379 } else {
4380 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4381 }
4382 };
4383 my $err = $@;
4384
4385 if (!$err) {
4386 my $count = 0;
4387 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4388 $count++;
4389 sleep 1;
4390 }
4391
4392 if ($count >= $timeout) {
4393 if ($force) {
4394 warn "VM still running - terminating now with SIGTERM\n";
4395 kill 15, $pid;
4396 } else {
4397 die "VM quit/powerdown failed - got timeout\n";
4398 }
4399 } else {
4400 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4401 return;
4402 }
4403 } else {
4404 if ($force) {
4405 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4406 kill 15, $pid;
4407 } else {
4408 die "VM quit/powerdown failed\n";
4409 }
4410 }
4411
4412 # wait again
4413 $timeout = 10;
4414
4415 my $count = 0;
4416 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4417 $count++;
4418 sleep 1;
4419 }
4420
4421 if ($count >= $timeout) {
4422 warn "VM still running - terminating now with SIGKILL\n";
4423 kill 9, $pid;
4424 sleep 1;
4425 }
4426
4427 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4428 });
4429 }
4430
4431 sub vm_suspend {
4432 my ($vmid, $skiplock) = @_;
4433
4434 lock_config($vmid, sub {
4435
4436 my $conf = load_config($vmid);
4437
4438 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4439
4440 vm_mon_cmd($vmid, "stop");
4441 });
4442 }
4443
4444 sub vm_resume {
4445 my ($vmid, $skiplock) = @_;
4446
4447 lock_config($vmid, sub {
4448
4449 my $conf = load_config($vmid);
4450
4451 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4452
4453 vm_mon_cmd($vmid, "cont");
4454 });
4455 }
4456
4457 sub vm_sendkey {
4458 my ($vmid, $skiplock, $key) = @_;
4459
4460 lock_config($vmid, sub {
4461
4462 my $conf = load_config($vmid);
4463
4464 # there is no qmp command, so we use the human monitor command
4465 vm_human_monitor_command($vmid, "sendkey $key");
4466 });
4467 }
4468
4469 sub vm_destroy {
4470 my ($storecfg, $vmid, $skiplock) = @_;
4471
4472 lock_config($vmid, sub {
4473
4474 my $conf = load_config($vmid);
4475
4476 check_lock($conf) if !$skiplock;
4477
4478 if (!check_running($vmid)) {
4479 fairsched_rmnod($vmid); # try to destroy group
4480 destroy_vm($storecfg, $vmid);
4481 } else {
4482 die "VM $vmid is running - destroy failed\n";
4483 }
4484 });
4485 }
4486
4487 # pci helpers
4488
4489 sub file_write {
4490 my ($filename, $buf) = @_;
4491
4492 my $fh = IO::File->new($filename, "w");
4493 return undef if !$fh;
4494
4495 my $res = print $fh $buf;
4496
4497 $fh->close();
4498
4499 return $res;
4500 }
4501
4502 sub pci_device_info {
4503 my ($name) = @_;
4504
4505 my $res;
4506
4507 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4508 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4509
4510 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4511 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4512
4513 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4514 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4515
4516 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4517 return undef if !defined($product) || $product !~ s/^0x//;
4518
4519 $res = {
4520 name => $name,
4521 vendor => $vendor,
4522 product => $product,
4523 domain => $domain,
4524 bus => $bus,
4525 slot => $slot,
4526 func => $func,
4527 irq => $irq,
4528 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4529 };
4530
4531 return $res;
4532 }
4533
4534 sub pci_dev_reset {
4535 my ($dev) = @_;
4536
4537 my $name = $dev->{name};
4538
4539 my $fn = "$pcisysfs/devices/$name/reset";
4540
4541 return file_write($fn, "1");
4542 }
4543
4544 sub pci_dev_bind_to_stub {
4545 my ($dev) = @_;
4546
4547 my $name = $dev->{name};
4548
4549 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4550 return 1 if -d $testdir;
4551
4552 my $data = "$dev->{vendor} $dev->{product}";
4553 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4554
4555 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4556 if (!file_write($fn, $name)) {
4557 return undef if -f $fn;
4558 }
4559
4560 $fn = "$pcisysfs/drivers/pci-stub/bind";
4561 if (! -d $testdir) {
4562 return undef if !file_write($fn, $name);
4563 }
4564
4565 return -d $testdir;
4566 }
4567
4568 sub pci_dev_bind_to_vfio {
4569 my ($dev) = @_;
4570
4571 my $name = $dev->{name};
4572
4573 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4574
4575 if (!-d $vfio_basedir) {
4576 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4577 }
4578 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4579
4580 my $testdir = "$vfio_basedir/$name";
4581 return 1 if -d $testdir;
4582
4583 my $data = "$dev->{vendor} $dev->{product}";
4584 return undef if !file_write("$vfio_basedir/new_id", $data);
4585
4586 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4587 if (!file_write($fn, $name)) {
4588 return undef if -f $fn;
4589 }
4590
4591 $fn = "$vfio_basedir/bind";
4592 if (! -d $testdir) {
4593 return undef if !file_write($fn, $name);
4594 }
4595
4596 return -d $testdir;
4597 }
4598
4599 sub pci_dev_group_bind_to_vfio {
4600 my ($pciid) = @_;
4601
4602 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4603
4604 if (!-d $vfio_basedir) {
4605 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4606 }
4607 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4608
4609 # get IOMMU group devices
4610 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4611 my @devs = grep /^0000:/, readdir($D);
4612 closedir($D);
4613
4614 foreach my $pciid (@devs) {
4615 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4616
4617 # pci bridges, switches or root ports are not supported
4618 # they have a pci_bus subdirectory so skip them
4619 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4620
4621 my $info = pci_device_info($1);
4622 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4623 }
4624
4625 return 1;
4626 }
4627
4628 sub print_pci_addr {
4629 my ($id, $bridges) = @_;
4630
4631 my $res = '';
4632 my $devices = {
4633 piix3 => { bus => 0, addr => 1 },
4634 #addr2 : first videocard
4635 balloon0 => { bus => 0, addr => 3 },
4636 watchdog => { bus => 0, addr => 4 },
4637 scsihw0 => { bus => 0, addr => 5 },
4638 scsihw1 => { bus => 0, addr => 6 },
4639 ahci0 => { bus => 0, addr => 7 },
4640 qga0 => { bus => 0, addr => 8 },
4641 spice => { bus => 0, addr => 9 },
4642 virtio0 => { bus => 0, addr => 10 },
4643 virtio1 => { bus => 0, addr => 11 },
4644 virtio2 => { bus => 0, addr => 12 },
4645 virtio3 => { bus => 0, addr => 13 },
4646 virtio4 => { bus => 0, addr => 14 },
4647 virtio5 => { bus => 0, addr => 15 },
4648 hostpci0 => { bus => 0, addr => 16 },
4649 hostpci1 => { bus => 0, addr => 17 },
4650 net0 => { bus => 0, addr => 18 },
4651 net1 => { bus => 0, addr => 19 },
4652 net2 => { bus => 0, addr => 20 },
4653 net3 => { bus => 0, addr => 21 },
4654 net4 => { bus => 0, addr => 22 },
4655 net5 => { bus => 0, addr => 23 },
4656 vga1 => { bus => 0, addr => 24 },
4657 vga2 => { bus => 0, addr => 25 },
4658 vga3 => { bus => 0, addr => 26 },
4659 hostpci2 => { bus => 0, addr => 27 },
4660 hostpci3 => { bus => 0, addr => 28 },
4661 #addr29 : usb-host (pve-usb.cfg)
4662 'pci.1' => { bus => 0, addr => 30 },
4663 'pci.2' => { bus => 0, addr => 31 },
4664 'net6' => { bus => 1, addr => 1 },
4665 'net7' => { bus => 1, addr => 2 },
4666 'net8' => { bus => 1, addr => 3 },
4667 'net9' => { bus => 1, addr => 4 },
4668 'net10' => { bus => 1, addr => 5 },
4669 'net11' => { bus => 1, addr => 6 },
4670 'net12' => { bus => 1, addr => 7 },
4671 'net13' => { bus => 1, addr => 8 },
4672 'net14' => { bus => 1, addr => 9 },
4673 'net15' => { bus => 1, addr => 10 },
4674 'net16' => { bus => 1, addr => 11 },
4675 'net17' => { bus => 1, addr => 12 },
4676 'net18' => { bus => 1, addr => 13 },
4677 'net19' => { bus => 1, addr => 14 },
4678 'net20' => { bus => 1, addr => 15 },
4679 'net21' => { bus => 1, addr => 16 },
4680 'net22' => { bus => 1, addr => 17 },
4681 'net23' => { bus => 1, addr => 18 },
4682 'net24' => { bus => 1, addr => 19 },
4683 'net25' => { bus => 1, addr => 20 },
4684 'net26' => { bus => 1, addr => 21 },
4685 'net27' => { bus => 1, addr => 22 },
4686 'net28' => { bus => 1, addr => 23 },
4687 'net29' => { bus => 1, addr => 24 },
4688 'net30' => { bus => 1, addr => 25 },
4689 'net31' => { bus => 1, addr => 26 },
4690 'virtio6' => { bus => 2, addr => 1 },
4691 'virtio7' => { bus => 2, addr => 2 },
4692 'virtio8' => { bus => 2, addr => 3 },
4693 'virtio9' => { bus => 2, addr => 4 },
4694 'virtio10' => { bus => 2, addr => 5 },
4695 'virtio11' => { bus => 2, addr => 6 },
4696 'virtio12' => { bus => 2, addr => 7 },
4697 'virtio13' => { bus => 2, addr => 8 },
4698 'virtio14' => { bus => 2, addr => 9 },
4699 'virtio15' => { bus => 2, addr => 10 },
4700 };
4701
4702 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4703 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4704 my $bus = $devices->{$id}->{bus};
4705 $res = ",bus=pci.$bus,addr=$addr";
4706 $bridges->{$bus} = 1 if $bridges;
4707 }
4708 return $res;
4709
4710 }
4711
4712 sub print_pcie_addr {
4713 my ($id) = @_;
4714
4715 my $res = '';
4716 my $devices = {
4717 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4718 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4719 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4720 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4721 };
4722
4723 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4724 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4725 my $bus = $devices->{$id}->{bus};
4726 $res = ",bus=$bus,addr=$addr";
4727 }
4728 return $res;
4729
4730 }
4731
4732 # vzdump restore implementaion
4733
4734 sub tar_archive_read_firstfile {
4735 my $archive = shift;
4736
4737 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4738
4739 # try to detect archive type first
4740 my $pid = open (TMP, "tar tf '$archive'|") ||
4741 die "unable to open file '$archive'\n";
4742 my $firstfile = <TMP>;
4743 kill 15, $pid;
4744 close TMP;
4745
4746 die "ERROR: archive contaions no data\n" if !$firstfile;
4747 chomp $firstfile;
4748
4749 return $firstfile;
4750 }
4751
4752 sub tar_restore_cleanup {
4753 my ($storecfg, $statfile) = @_;
4754
4755 print STDERR "starting cleanup\n";
4756
4757 if (my $fd = IO::File->new($statfile, "r")) {
4758 while (defined(my $line = <$fd>)) {
4759 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4760 my $volid = $2;
4761 eval {
4762 if ($volid =~ m|^/|) {
4763 unlink $volid || die 'unlink failed\n';
4764 } else {
4765 PVE::Storage::vdisk_free($storecfg, $volid);
4766 }
4767 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4768 };
4769 print STDERR "unable to cleanup '$volid' - $@" if $@;
4770 } else {
4771 print STDERR "unable to parse line in statfile - $line";
4772 }
4773 }
4774 $fd->close();
4775 }
4776 }
4777
4778 sub restore_archive {
4779 my ($archive, $vmid, $user, $opts) = @_;
4780
4781 my $format = $opts->{format};
4782 my $comp;
4783
4784 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4785 $format = 'tar' if !$format;
4786 $comp = 'gzip';
4787 } elsif ($archive =~ m/\.tar$/) {
4788 $format = 'tar' if !$format;
4789 } elsif ($archive =~ m/.tar.lzo$/) {
4790 $format = 'tar' if !$format;
4791 $comp = 'lzop';
4792 } elsif ($archive =~ m/\.vma$/) {
4793 $format = 'vma' if !$format;
4794 } elsif ($archive =~ m/\.vma\.gz$/) {
4795 $format = 'vma' if !$format;
4796 $comp = 'gzip';
4797 } elsif ($archive =~ m/\.vma\.lzo$/) {
4798 $format = 'vma' if !$format;
4799 $comp = 'lzop';
4800 } else {
4801 $format = 'vma' if !$format; # default
4802 }
4803
4804 # try to detect archive format
4805 if ($format eq 'tar') {
4806 return restore_tar_archive($archive, $vmid, $user, $opts);
4807 } else {
4808 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4809 }
4810 }
4811
4812 sub restore_update_config_line {
4813 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4814
4815 return if $line =~ m/^\#qmdump\#/;
4816 return if $line =~ m/^\#vzdump\#/;
4817 return if $line =~ m/^lock:/;
4818 return if $line =~ m/^unused\d+:/;
4819 return if $line =~ m/^parent:/;
4820 return if $line =~ m/^template:/; # restored VM is never a template
4821
4822 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4823 # try to convert old 1.X settings
4824 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4825 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4826 my ($model, $macaddr) = split(/\=/, $devconfig);
4827 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4828 my $net = {
4829 model => $model,
4830 bridge => "vmbr$ind",
4831 macaddr => $macaddr,
4832 };
4833 my $netstr = print_net($net);
4834
4835 print $outfd "net$cookie->{netcount}: $netstr\n";
4836 $cookie->{netcount}++;
4837 }
4838 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4839 my ($id, $netstr) = ($1, $2);
4840 my $net = parse_net($netstr);
4841 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4842 $netstr = print_net($net);
4843 print $outfd "$id: $netstr\n";
4844 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4845 my $virtdev = $1;
4846 my $value = $3;
4847 if ($line =~ m/backup=no/) {
4848 print $outfd "#$line";
4849 } elsif ($virtdev && $map->{$virtdev}) {
4850 my $di = parse_drive($virtdev, $value);
4851 delete $di->{format}; # format can change on restore
4852 $di->{file} = $map->{$virtdev};
4853 $value = print_drive($vmid, $di);
4854 print $outfd "$virtdev: $value\n";
4855 } else {
4856 print $outfd $line;
4857 }
4858 } else {
4859 print $outfd $line;
4860 }
4861 }
4862
4863 sub scan_volids {
4864 my ($cfg, $vmid) = @_;
4865
4866 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4867
4868 my $volid_hash = {};
4869 foreach my $storeid (keys %$info) {
4870 foreach my $item (@{$info->{$storeid}}) {
4871 next if !($item->{volid} && $item->{size});
4872 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4873 $volid_hash->{$item->{volid}} = $item;
4874 }
4875 }
4876
4877 return $volid_hash;
4878 }
4879
4880 sub get_used_paths {
4881 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4882
4883 my $used_path = {};
4884
4885 my $scan_config = sub {
4886 my ($cref, $snapname) = @_;
4887
4888 foreach my $key (keys %$cref) {
4889 my $value = $cref->{$key};
4890 if (valid_drivename($key)) {
4891 next if $skip_drive && $key eq $skip_drive;
4892 my $drive = parse_drive($key, $value);
4893 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4894 if ($drive->{file} =~ m!^/!) {
4895 $used_path->{$drive->{file}}++; # = 1;
4896 } else {
4897 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4898 next if !$storeid;
4899 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4900 next if !$scfg;
4901 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4902 $used_path->{$path}++; # = 1;
4903 }
4904 }
4905 }
4906 };
4907
4908 &$scan_config($conf);
4909
4910 undef $skip_drive;
4911
4912 if ($scan_snapshots) {
4913 foreach my $snapname (keys %{$conf->{snapshots}}) {
4914 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4915 }
4916 }
4917
4918 return $used_path;
4919 }
4920
4921 sub update_disksize {
4922 my ($vmid, $conf, $volid_hash) = @_;
4923
4924 my $changes;
4925
4926 my $used = {};
4927
4928 # Note: it is allowed to define multiple storages with same path (alias), so
4929 # we need to check both 'volid' and real 'path' (two different volid can point
4930 # to the same path).
4931
4932 my $usedpath = {};
4933
4934 # update size info
4935 foreach my $opt (keys %$conf) {
4936 if (valid_drivename($opt)) {
4937 my $drive = parse_drive($opt, $conf->{$opt});
4938 my $volid = $drive->{file};
4939 next if !$volid;
4940
4941 $used->{$volid} = 1;
4942 if ($volid_hash->{$volid} &&
4943 (my $path = $volid_hash->{$volid}->{path})) {
4944 $usedpath->{$path} = 1;
4945 }
4946
4947 next if drive_is_cdrom($drive);
4948 next if !$volid_hash->{$volid};
4949
4950 $drive->{size} = $volid_hash->{$volid}->{size};
4951 my $new = print_drive($vmid, $drive);
4952 if ($new ne $conf->{$opt}) {
4953 $changes = 1;
4954 $conf->{$opt} = $new;
4955 }
4956 }
4957 }
4958
4959 # remove 'unusedX' entry if volume is used
4960 foreach my $opt (keys %$conf) {
4961 next if $opt !~ m/^unused\d+$/;
4962 my $volid = $conf->{$opt};
4963 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4964 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4965 $changes = 1;
4966 delete $conf->{$opt};
4967 }
4968 }
4969
4970 foreach my $volid (sort keys %$volid_hash) {
4971 next if $volid =~ m/vm-$vmid-state-/;
4972 next if $used->{$volid};
4973 my $path = $volid_hash->{$volid}->{path};
4974 next if !$path; # just to be sure
4975 next if $usedpath->{$path};
4976 $changes = 1;
4977 add_unused_volume($conf, $volid);
4978 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4979 }
4980
4981 return $changes;
4982 }
4983
4984 sub rescan {
4985 my ($vmid, $nolock) = @_;
4986
4987 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4988
4989 my $volid_hash = scan_volids($cfg, $vmid);
4990
4991 my $updatefn = sub {
4992 my ($vmid) = @_;
4993
4994 my $conf = load_config($vmid);
4995
4996 check_lock($conf);
4997
4998 my $vm_volids = {};
4999 foreach my $volid (keys %$volid_hash) {
5000 my $info = $volid_hash->{$volid};
5001 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5002 }
5003
5004 my $changes = update_disksize($vmid, $conf, $vm_volids);
5005
5006 update_config_nolock($vmid, $conf, 1) if $changes;
5007 };
5008
5009 if (defined($vmid)) {
5010 if ($nolock) {
5011 &$updatefn($vmid);
5012 } else {
5013 lock_config($vmid, $updatefn, $vmid);
5014 }
5015 } else {
5016 my $vmlist = config_list();
5017 foreach my $vmid (keys %$vmlist) {
5018 if ($nolock) {
5019 &$updatefn($vmid);
5020 } else {
5021 lock_config($vmid, $updatefn, $vmid);
5022 }
5023 }
5024 }
5025 }
5026
5027 sub restore_vma_archive {
5028 my ($archive, $vmid, $user, $opts, $comp) = @_;
5029
5030 my $input = $archive eq '-' ? "<&STDIN" : undef;
5031 my $readfrom = $archive;
5032
5033 my $uncomp = '';
5034 if ($comp) {
5035 $readfrom = '-';
5036 my $qarchive = PVE::Tools::shellquote($archive);
5037 if ($comp eq 'gzip') {
5038 $uncomp = "zcat $qarchive|";
5039 } elsif ($comp eq 'lzop') {
5040 $uncomp = "lzop -d -c $qarchive|";
5041 } else {
5042 die "unknown compression method '$comp'\n";
5043 }
5044
5045 }
5046
5047 my $tmpdir = "/var/tmp/vzdumptmp$$";
5048 rmtree $tmpdir;
5049
5050 # disable interrupts (always do cleanups)
5051 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5052 warn "got interrupt - ignored\n";
5053 };
5054
5055 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5056 POSIX::mkfifo($mapfifo, 0600);
5057 my $fifofh;
5058
5059 my $openfifo = sub {
5060 open($fifofh, '>', $mapfifo) || die $!;
5061 };
5062
5063 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5064
5065 my $oldtimeout;
5066 my $timeout = 5;
5067
5068 my $devinfo = {};
5069
5070 my $rpcenv = PVE::RPCEnvironment::get();
5071
5072 my $conffile = config_file($vmid);
5073 my $tmpfn = "$conffile.$$.tmp";
5074
5075 # Note: $oldconf is undef if VM does not exists
5076 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5077
5078 my $print_devmap = sub {
5079 my $virtdev_hash = {};
5080
5081 my $cfgfn = "$tmpdir/qemu-server.conf";
5082
5083 # we can read the config - that is already extracted
5084 my $fh = IO::File->new($cfgfn, "r") ||
5085 "unable to read qemu-server.conf - $!\n";
5086
5087 while (defined(my $line = <$fh>)) {
5088 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5089 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5090 die "archive does not contain data for drive '$virtdev'\n"
5091 if !$devinfo->{$devname};
5092 if (defined($opts->{storage})) {
5093 $storeid = $opts->{storage} || 'local';
5094 } elsif (!$storeid) {
5095 $storeid = 'local';
5096 }
5097 $format = 'raw' if !$format;
5098 $devinfo->{$devname}->{devname} = $devname;
5099 $devinfo->{$devname}->{virtdev} = $virtdev;
5100 $devinfo->{$devname}->{format} = $format;
5101 $devinfo->{$devname}->{storeid} = $storeid;
5102
5103 # check permission on storage
5104 my $pool = $opts->{pool}; # todo: do we need that?
5105 if ($user ne 'root@pam') {
5106 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5107 }
5108
5109 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5110 }
5111 }
5112
5113 foreach my $devname (keys %$devinfo) {
5114 die "found no device mapping information for device '$devname'\n"
5115 if !$devinfo->{$devname}->{virtdev};
5116 }
5117
5118 my $cfg = cfs_read_file('storage.cfg');
5119
5120 # create empty/temp config
5121 if ($oldconf) {
5122 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5123 foreach_drive($oldconf, sub {
5124 my ($ds, $drive) = @_;
5125
5126 return if drive_is_cdrom($drive);
5127
5128 my $volid = $drive->{file};
5129
5130 return if !$volid || $volid =~ m|^/|;
5131
5132 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5133 return if !$path || !$owner || ($owner != $vmid);
5134
5135 # Note: only delete disk we want to restore
5136 # other volumes will become unused
5137 if ($virtdev_hash->{$ds}) {
5138 PVE::Storage::vdisk_free($cfg, $volid);
5139 }
5140 });
5141 }
5142
5143 my $map = {};
5144 foreach my $virtdev (sort keys %$virtdev_hash) {
5145 my $d = $virtdev_hash->{$virtdev};
5146 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5147 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5148
5149 # test if requested format is supported
5150 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5151 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5152 $d->{format} = $defFormat if !$supported;
5153
5154 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5155 $d->{format}, undef, $alloc_size);
5156 print STDERR "new volume ID is '$volid'\n";
5157 $d->{volid} = $volid;
5158 my $path = PVE::Storage::path($cfg, $volid);
5159
5160 my $write_zeros = 1;
5161 # fixme: what other storages types initialize volumes with zero?
5162 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5163 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5164 $write_zeros = 0;
5165 }
5166
5167 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5168
5169 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5170 $map->{$virtdev} = $volid;
5171 }
5172
5173 $fh->seek(0, 0) || die "seek failed - $!\n";
5174
5175 my $outfd = new IO::File ($tmpfn, "w") ||
5176 die "unable to write config for VM $vmid\n";
5177
5178 my $cookie = { netcount => 0 };
5179 while (defined(my $line = <$fh>)) {
5180 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5181 }
5182
5183 $fh->close();
5184 $outfd->close();
5185 };
5186
5187 eval {
5188 # enable interrupts
5189 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5190 die "interrupted by signal\n";
5191 };
5192 local $SIG{ALRM} = sub { die "got timeout\n"; };
5193
5194 $oldtimeout = alarm($timeout);
5195
5196 my $parser = sub {
5197 my $line = shift;
5198
5199 print "$line\n";
5200
5201 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5202 my ($dev_id, $size, $devname) = ($1, $2, $3);
5203 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5204 } elsif ($line =~ m/^CTIME: /) {
5205 # we correctly received the vma config, so we can disable
5206 # the timeout now for disk allocation (set to 10 minutes, so
5207 # that we always timeout if something goes wrong)
5208 alarm(600);
5209 &$print_devmap();
5210 print $fifofh "done\n";
5211 my $tmp = $oldtimeout || 0;
5212 $oldtimeout = undef;
5213 alarm($tmp);
5214 close($fifofh);
5215 }
5216 };
5217
5218 print "restore vma archive: $cmd\n";
5219 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5220 };
5221 my $err = $@;
5222
5223 alarm($oldtimeout) if $oldtimeout;
5224
5225 unlink $mapfifo;
5226
5227 if ($err) {
5228 rmtree $tmpdir;
5229 unlink $tmpfn;
5230
5231 my $cfg = cfs_read_file('storage.cfg');
5232 foreach my $devname (keys %$devinfo) {
5233 my $volid = $devinfo->{$devname}->{volid};
5234 next if !$volid;
5235 eval {
5236 if ($volid =~ m|^/|) {
5237 unlink $volid || die 'unlink failed\n';
5238 } else {
5239 PVE::Storage::vdisk_free($cfg, $volid);
5240 }
5241 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5242 };
5243 print STDERR "unable to cleanup '$volid' - $@" if $@;
5244 }
5245 die $err;
5246 }
5247
5248 rmtree $tmpdir;
5249
5250 rename($tmpfn, $conffile) ||
5251 die "unable to commit configuration file '$conffile'\n";
5252
5253 PVE::Cluster::cfs_update(); # make sure we read new file
5254
5255 eval { rescan($vmid, 1); };
5256 warn $@ if $@;
5257 }
5258
5259 sub restore_tar_archive {
5260 my ($archive, $vmid, $user, $opts) = @_;
5261
5262 if ($archive ne '-') {
5263 my $firstfile = tar_archive_read_firstfile($archive);
5264 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5265 if $firstfile ne 'qemu-server.conf';
5266 }
5267
5268 my $storecfg = cfs_read_file('storage.cfg');
5269
5270 # destroy existing data - keep empty config
5271 my $vmcfgfn = config_file($vmid);
5272 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5273
5274 my $tocmd = "/usr/lib/qemu-server/qmextract";
5275
5276 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5277 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5278 $tocmd .= ' --prealloc' if $opts->{prealloc};
5279 $tocmd .= ' --info' if $opts->{info};
5280
5281 # tar option "xf" does not autodetect compression when read from STDIN,
5282 # so we pipe to zcat
5283 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5284 PVE::Tools::shellquote("--to-command=$tocmd");
5285
5286 my $tmpdir = "/var/tmp/vzdumptmp$$";
5287 mkpath $tmpdir;
5288
5289 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5290 local $ENV{VZDUMP_VMID} = $vmid;
5291 local $ENV{VZDUMP_USER} = $user;
5292
5293 my $conffile = config_file($vmid);
5294 my $tmpfn = "$conffile.$$.tmp";
5295
5296 # disable interrupts (always do cleanups)
5297 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5298 print STDERR "got interrupt - ignored\n";
5299 };
5300
5301 eval {
5302 # enable interrupts
5303 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5304 die "interrupted by signal\n";
5305 };
5306
5307 if ($archive eq '-') {
5308 print "extracting archive from STDIN\n";
5309 run_command($cmd, input => "<&STDIN");
5310 } else {
5311 print "extracting archive '$archive'\n";
5312 run_command($cmd);
5313 }
5314
5315 return if $opts->{info};
5316
5317 # read new mapping
5318 my $map = {};
5319 my $statfile = "$tmpdir/qmrestore.stat";
5320 if (my $fd = IO::File->new($statfile, "r")) {
5321 while (defined (my $line = <$fd>)) {
5322 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5323 $map->{$1} = $2 if $1;
5324 } else {
5325 print STDERR "unable to parse line in statfile - $line\n";
5326 }
5327 }
5328 $fd->close();
5329 }
5330
5331 my $confsrc = "$tmpdir/qemu-server.conf";
5332
5333 my $srcfd = new IO::File($confsrc, "r") ||
5334 die "unable to open file '$confsrc'\n";
5335
5336 my $outfd = new IO::File ($tmpfn, "w") ||
5337 die "unable to write config for VM $vmid\n";
5338
5339 my $cookie = { netcount => 0 };
5340 while (defined (my $line = <$srcfd>)) {
5341 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5342 }
5343
5344 $srcfd->close();
5345 $outfd->close();
5346 };
5347 my $err = $@;
5348
5349 if ($err) {
5350
5351 unlink $tmpfn;
5352
5353 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5354
5355 die $err;
5356 }
5357
5358 rmtree $tmpdir;
5359
5360 rename $tmpfn, $conffile ||
5361 die "unable to commit configuration file '$conffile'\n";
5362
5363 PVE::Cluster::cfs_update(); # make sure we read new file
5364
5365 eval { rescan($vmid, 1); };
5366 warn $@ if $@;
5367 };
5368
5369
5370 # Internal snapshots
5371
5372 # NOTE: Snapshot create/delete involves several non-atomic
5373 # action, and can take a long time.
5374 # So we try to avoid locking the file and use 'lock' variable
5375 # inside the config file instead.
5376
5377 my $snapshot_copy_config = sub {
5378 my ($source, $dest) = @_;
5379
5380 foreach my $k (keys %$source) {
5381 next if $k eq 'snapshots';
5382 next if $k eq 'snapstate';
5383 next if $k eq 'snaptime';
5384 next if $k eq 'vmstate';
5385 next if $k eq 'lock';
5386 next if $k eq 'digest';
5387 next if $k eq 'description';
5388 next if $k =~ m/^unused\d+$/;
5389
5390 $dest->{$k} = $source->{$k};
5391 }
5392 };
5393
5394 my $snapshot_apply_config = sub {
5395 my ($conf, $snap) = @_;
5396
5397 # copy snapshot list
5398 my $newconf = {
5399 snapshots => $conf->{snapshots},
5400 };
5401
5402 # keep description and list of unused disks
5403 foreach my $k (keys %$conf) {
5404 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5405 $newconf->{$k} = $conf->{$k};
5406 }
5407
5408 &$snapshot_copy_config($snap, $newconf);
5409
5410 return $newconf;
5411 };
5412
5413 sub foreach_writable_storage {
5414 my ($conf, $func) = @_;
5415
5416 my $sidhash = {};
5417
5418 foreach my $ds (keys %$conf) {
5419 next if !valid_drivename($ds);
5420
5421 my $drive = parse_drive($ds, $conf->{$ds});
5422 next if !$drive;
5423 next if drive_is_cdrom($drive);
5424
5425 my $volid = $drive->{file};
5426
5427 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5428 $sidhash->{$sid} = $sid if $sid;
5429 }
5430
5431 foreach my $sid (sort keys %$sidhash) {
5432 &$func($sid);
5433 }
5434 }
5435
5436 my $alloc_vmstate_volid = sub {
5437 my ($storecfg, $vmid, $conf, $snapname) = @_;
5438
5439 # Note: we try to be smart when selecting a $target storage
5440
5441 my $target;
5442
5443 # search shared storage first
5444 foreach_writable_storage($conf, sub {
5445 my ($sid) = @_;
5446 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5447 return if !$scfg->{shared};
5448
5449 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5450 });
5451
5452 if (!$target) {
5453 # now search local storage
5454 foreach_writable_storage($conf, sub {
5455 my ($sid) = @_;
5456 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5457 return if $scfg->{shared};
5458
5459 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5460 });
5461 }
5462
5463 $target = 'local' if !$target;
5464
5465 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5466 # we abort live save after $conf->{memory}, so we need at max twice that space
5467 my $size = $conf->{memory}*2 + $driver_state_size;
5468
5469 my $name = "vm-$vmid-state-$snapname";
5470 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5471 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5472 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5473
5474 return $volid;
5475 };
5476
5477 my $snapshot_prepare = sub {
5478 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5479
5480 my $snap;
5481
5482 my $updatefn = sub {
5483
5484 my $conf = load_config($vmid);
5485
5486 die "you can't take a snapshot if it's a template\n"
5487 if is_template($conf);
5488
5489 check_lock($conf);
5490
5491 $conf->{lock} = 'snapshot';
5492
5493 die "snapshot name '$snapname' already used\n"
5494 if defined($conf->{snapshots}->{$snapname});
5495
5496 my $storecfg = PVE::Storage::config();
5497 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5498
5499 $snap = $conf->{snapshots}->{$snapname} = {};
5500
5501 if ($save_vmstate && check_running($vmid)) {
5502 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5503 }
5504
5505 &$snapshot_copy_config($conf, $snap);
5506
5507 $snap->{snapstate} = "prepare";
5508 $snap->{snaptime} = time();
5509 $snap->{description} = $comment if $comment;
5510
5511 # always overwrite machine if we save vmstate. This makes sure we
5512 # can restore it later using correct machine type
5513 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5514
5515 update_config_nolock($vmid, $conf, 1);
5516 };
5517
5518 lock_config($vmid, $updatefn);
5519
5520 return $snap;
5521 };
5522
5523 my $snapshot_commit = sub {
5524 my ($vmid, $snapname) = @_;
5525
5526 my $updatefn = sub {
5527
5528 my $conf = load_config($vmid);
5529
5530 die "missing snapshot lock\n"
5531 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5532
5533 my $has_machine_config = defined($conf->{machine});
5534
5535 my $snap = $conf->{snapshots}->{$snapname};
5536
5537 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5538
5539 die "wrong snapshot state\n"
5540 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5541
5542 delete $snap->{snapstate};
5543 delete $conf->{lock};
5544
5545 my $newconf = &$snapshot_apply_config($conf, $snap);
5546
5547 delete $newconf->{machine} if !$has_machine_config;
5548
5549 $newconf->{parent} = $snapname;
5550
5551 update_config_nolock($vmid, $newconf, 1);
5552 };
5553
5554 lock_config($vmid, $updatefn);
5555 };
5556
5557 sub snapshot_rollback {
5558 my ($vmid, $snapname) = @_;
5559
5560 my $prepare = 1;
5561
5562 my $storecfg = PVE::Storage::config();
5563
5564 my $conf = load_config($vmid);
5565
5566 my $get_snapshot_config = sub {
5567
5568 die "you can't rollback if vm is a template\n" if is_template($conf);
5569
5570 my $res = $conf->{snapshots}->{$snapname};
5571
5572 die "snapshot '$snapname' does not exist\n" if !defined($res);
5573
5574 return $res;
5575 };
5576
5577 my $snap = &$get_snapshot_config();
5578
5579 foreach_drive($snap, sub {
5580 my ($ds, $drive) = @_;
5581
5582 return if drive_is_cdrom($drive);
5583
5584 my $volid = $drive->{file};
5585
5586 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5587 });
5588
5589 my $updatefn = sub {
5590
5591 $conf = load_config($vmid);
5592
5593 $snap = &$get_snapshot_config();
5594
5595 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5596 if $snap->{snapstate};
5597
5598 if ($prepare) {
5599 check_lock($conf);
5600 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5601 }
5602
5603 die "unable to rollback vm $vmid: vm is running\n"
5604 if check_running($vmid);
5605
5606 if ($prepare) {
5607 $conf->{lock} = 'rollback';
5608 } else {
5609 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5610 delete $conf->{lock};
5611 }
5612
5613 my $forcemachine;
5614
5615 if (!$prepare) {
5616 my $has_machine_config = defined($conf->{machine});
5617
5618 # copy snapshot config to current config
5619 $conf = &$snapshot_apply_config($conf, $snap);
5620 $conf->{parent} = $snapname;
5621
5622 # Note: old code did not store 'machine', so we try to be smart
5623 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5624 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5625 # we remove the 'machine' configuration if not explicitly specified
5626 # in the original config.
5627 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5628 }
5629
5630 update_config_nolock($vmid, $conf, 1);
5631
5632 if (!$prepare && $snap->{vmstate}) {
5633 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5634 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5635 }
5636 };
5637
5638 lock_config($vmid, $updatefn);
5639
5640 foreach_drive($snap, sub {
5641 my ($ds, $drive) = @_;
5642
5643 return if drive_is_cdrom($drive);
5644
5645 my $volid = $drive->{file};
5646 my $device = "drive-$ds";
5647
5648 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5649 });
5650
5651 $prepare = 0;
5652 lock_config($vmid, $updatefn);
5653 }
5654
5655 my $savevm_wait = sub {
5656 my ($vmid) = @_;
5657
5658 for(;;) {
5659 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5660 if (!$stat->{status}) {
5661 die "savevm not active\n";
5662 } elsif ($stat->{status} eq 'active') {
5663 sleep(1);
5664 next;
5665 } elsif ($stat->{status} eq 'completed') {
5666 last;
5667 } else {
5668 die "query-savevm returned status '$stat->{status}'\n";
5669 }
5670 }
5671 };
5672
5673 sub snapshot_create {
5674 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5675
5676 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5677
5678 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5679
5680 my $config = load_config($vmid);
5681
5682 my $running = check_running($vmid);
5683
5684 my $freezefs = $running && $config->{agent};
5685 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5686
5687 my $drivehash = {};
5688
5689 if ($freezefs) {
5690 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5691 warn "guest-fsfreeze-freeze problems - $@" if $@;
5692 }
5693
5694 eval {
5695 # create internal snapshots of all drives
5696
5697 my $storecfg = PVE::Storage::config();
5698
5699 if ($running) {
5700 if ($snap->{vmstate}) {
5701 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5702 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5703 &$savevm_wait($vmid);
5704 } else {
5705 vm_mon_cmd($vmid, "savevm-start");
5706 }
5707 };
5708
5709 foreach_drive($snap, sub {
5710 my ($ds, $drive) = @_;
5711
5712 return if drive_is_cdrom($drive);
5713
5714 my $volid = $drive->{file};
5715 my $device = "drive-$ds";
5716
5717 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5718 $drivehash->{$ds} = 1;
5719 });
5720 };
5721 my $err = $@;
5722
5723 if ($running) {
5724 eval { vm_mon_cmd($vmid, "savevm-end") };
5725 warn $@ if $@;
5726
5727 if ($freezefs) {
5728 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5729 warn "guest-fsfreeze-thaw problems - $@" if $@;
5730 }
5731
5732 # savevm-end is async, we need to wait
5733 for (;;) {
5734 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5735 if (!$stat->{bytes}) {
5736 last;
5737 } else {
5738 print "savevm not yet finished\n";
5739 sleep(1);
5740 next;
5741 }
5742 }
5743 }
5744
5745 if ($err) {
5746 warn "snapshot create failed: starting cleanup\n";
5747 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5748 warn $@ if $@;
5749 die $err;
5750 }
5751
5752 &$snapshot_commit($vmid, $snapname);
5753 }
5754
5755 # Note: $drivehash is only set when called from snapshot_create.
5756 sub snapshot_delete {
5757 my ($vmid, $snapname, $force, $drivehash) = @_;
5758
5759 my $prepare = 1;
5760
5761 my $snap;
5762 my $unused = [];
5763
5764 my $unlink_parent = sub {
5765 my ($confref, $new_parent) = @_;
5766
5767 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5768 if ($new_parent) {
5769 $confref->{parent} = $new_parent;
5770 } else {
5771 delete $confref->{parent};
5772 }
5773 }
5774 };
5775
5776 my $updatefn = sub {
5777 my ($remove_drive) = @_;
5778
5779 my $conf = load_config($vmid);
5780
5781 if (!$drivehash) {
5782 check_lock($conf);
5783 die "you can't delete a snapshot if vm is a template\n"
5784 if is_template($conf);
5785 }
5786
5787 $snap = $conf->{snapshots}->{$snapname};
5788
5789 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5790
5791 # remove parent refs
5792 if (!$prepare) {
5793 &$unlink_parent($conf, $snap->{parent});
5794 foreach my $sn (keys %{$conf->{snapshots}}) {
5795 next if $sn eq $snapname;
5796 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5797 }
5798 }
5799
5800 if ($remove_drive) {
5801 if ($remove_drive eq 'vmstate') {
5802 delete $snap->{$remove_drive};
5803 } else {
5804 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5805 my $volid = $drive->{file};
5806 delete $snap->{$remove_drive};
5807 add_unused_volume($conf, $volid);
5808 }
5809 }
5810
5811 if ($prepare) {
5812 $snap->{snapstate} = 'delete';
5813 } else {
5814 delete $conf->{snapshots}->{$snapname};
5815 delete $conf->{lock} if $drivehash;
5816 foreach my $volid (@$unused) {
5817 add_unused_volume($conf, $volid);
5818 }
5819 }
5820
5821 update_config_nolock($vmid, $conf, 1);
5822 };
5823
5824 lock_config($vmid, $updatefn);
5825
5826 # now remove vmstate file
5827
5828 my $storecfg = PVE::Storage::config();
5829
5830 if ($snap->{vmstate}) {
5831 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5832 if (my $err = $@) {
5833 die $err if !$force;
5834 warn $err;
5835 }
5836 # save changes (remove vmstate from snapshot)
5837 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5838 };
5839
5840 # now remove all internal snapshots
5841 foreach_drive($snap, sub {
5842 my ($ds, $drive) = @_;
5843
5844 return if drive_is_cdrom($drive);
5845
5846 my $volid = $drive->{file};
5847 my $device = "drive-$ds";
5848
5849 if (!$drivehash || $drivehash->{$ds}) {
5850 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5851 if (my $err = $@) {
5852 die $err if !$force;
5853 warn $err;
5854 }
5855 }
5856
5857 # save changes (remove drive fron snapshot)
5858 lock_config($vmid, $updatefn, $ds) if !$force;
5859 push @$unused, $volid;
5860 });
5861
5862 # now cleanup config
5863 $prepare = 0;
5864 lock_config($vmid, $updatefn);
5865 }
5866
5867 sub has_feature {
5868 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5869
5870 my $err;
5871 foreach_drive($conf, sub {
5872 my ($ds, $drive) = @_;
5873
5874 return if drive_is_cdrom($drive);
5875 my $volid = $drive->{file};
5876 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5877 });
5878
5879 return $err ? 0 : 1;
5880 }
5881
5882 sub template_create {
5883 my ($vmid, $conf, $disk) = @_;
5884
5885 my $storecfg = PVE::Storage::config();
5886
5887 foreach_drive($conf, sub {
5888 my ($ds, $drive) = @_;
5889
5890 return if drive_is_cdrom($drive);
5891 return if $disk && $ds ne $disk;
5892
5893 my $volid = $drive->{file};
5894 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5895
5896 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5897 $drive->{file} = $voliddst;
5898 $conf->{$ds} = print_drive($vmid, $drive);
5899 update_config_nolock($vmid, $conf, 1);
5900 });
5901 }
5902
5903 sub is_template {
5904 my ($conf) = @_;
5905
5906 return 1 if defined $conf->{template} && $conf->{template} == 1;
5907 }
5908
5909 sub qemu_img_convert {
5910 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5911
5912 my $storecfg = PVE::Storage::config();
5913 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5914 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5915
5916 if ($src_storeid && $dst_storeid) {
5917 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5918 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5919
5920 my $src_format = qemu_img_format($src_scfg, $src_volname);
5921 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5922
5923 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5924 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5925
5926 my $cmd = [];
5927 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5928 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5929 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5930
5931 my $parser = sub {
5932 my $line = shift;
5933 if($line =~ m/\((\S+)\/100\%\)/){
5934 my $percent = $1;
5935 my $transferred = int($size * $percent / 100);
5936 my $remaining = $size - $transferred;
5937
5938 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5939 }
5940
5941 };
5942
5943 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5944 my $err = $@;
5945 die "copy failed: $err" if $err;
5946 }
5947 }
5948
5949 sub qemu_img_format {
5950 my ($scfg, $volname) = @_;
5951
5952 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5953 return $1;
5954 } elsif ($scfg->{type} eq 'iscsi') {
5955 return "host_device";
5956 } else {
5957 return "raw";
5958 }
5959 }
5960
5961 sub qemu_drive_mirror {
5962 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5963
5964 my $count = 0;
5965 my $old_len = 0;
5966 my $frozen = undef;
5967 my $maxwait = 120;
5968
5969 my $storecfg = PVE::Storage::config();
5970 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5971
5972 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5973
5974 my $format;
5975 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5976 $format = $1;
5977 }
5978
5979 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5980
5981 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5982 $opts->{format} = $format if $format;
5983
5984 #fixme : sometime drive-mirror timeout, but works fine after.
5985 # (I have see the problem with big volume > 200GB), so we need to eval
5986 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5987 # ignore errors here
5988
5989 eval {
5990 while (1) {
5991 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5992 my $stat = @$stats[0];
5993 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5994 die "error job is not mirroring" if $stat->{type} ne "mirror";
5995
5996 my $busy = $stat->{busy};
5997
5998 if (my $total = $stat->{len}) {
5999 my $transferred = $stat->{offset} || 0;
6000 my $remaining = $total - $transferred;
6001 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6002
6003 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
6004 }
6005
6006 if ($stat->{len} == $stat->{offset}) {
6007 if ($busy eq 'false') {
6008
6009 last if $vmiddst != $vmid;
6010
6011 # try to switch the disk if source and destination are on the same guest
6012 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6013 last if !$@;
6014 die $@ if $@ !~ m/cannot be completed/;
6015 }
6016
6017 if ($count > $maxwait) {
6018 # if too much writes to disk occurs at the end of migration
6019 #the disk needs to be freezed to be able to complete the migration
6020 vm_suspend($vmid,1);
6021 $frozen = 1;
6022 }
6023 $count ++
6024 }
6025 $old_len = $stat->{offset};
6026 sleep 1;
6027 }
6028
6029 vm_resume($vmid, 1) if $frozen;
6030
6031 };
6032 my $err = $@;
6033
6034 my $cancel_job = sub {
6035 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6036 while (1) {
6037 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6038 my $stat = @$stats[0];
6039 last if !$stat;
6040 sleep 1;
6041 }
6042 };
6043
6044 if ($err) {
6045 eval { &$cancel_job(); };
6046 die "mirroring error: $err";
6047 }
6048
6049 if ($vmiddst != $vmid) {
6050 # if we clone a disk for a new target vm, we don't switch the disk
6051 &$cancel_job(); # so we call block-job-cancel
6052 }
6053 }
6054
6055 sub clone_disk {
6056 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6057 $newvmid, $storage, $format, $full, $newvollist) = @_;
6058
6059 my $newvolid;
6060
6061 if (!$full) {
6062 print "create linked clone of drive $drivename ($drive->{file})\n";
6063 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6064 push @$newvollist, $newvolid;
6065 } else {
6066 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6067 $storeid = $storage if $storage;
6068
6069 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6070 if (!$format) {
6071 $format = $drive->{format} || $defFormat;
6072 }
6073
6074 # test if requested format is supported - else use default
6075 my $supported = grep { $_ eq $format } @$validFormats;
6076 $format = $defFormat if !$supported;
6077
6078 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6079
6080 print "create full clone of drive $drivename ($drive->{file})\n";
6081 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6082 push @$newvollist, $newvolid;
6083
6084 if (!$running || $snapname) {
6085 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6086 } else {
6087 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6088 }
6089 }
6090
6091 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6092
6093 my $disk = $drive;
6094 $disk->{format} = undef;
6095 $disk->{file} = $newvolid;
6096 $disk->{size} = $size;
6097
6098 return $disk;
6099 }
6100
6101 # this only works if VM is running
6102 sub get_current_qemu_machine {
6103 my ($vmid) = @_;
6104
6105 my $cmd = { execute => 'query-machines', arguments => {} };
6106 my $res = vm_qmp_command($vmid, $cmd);
6107
6108 my ($current, $default);
6109 foreach my $e (@$res) {
6110 $default = $e->{name} if $e->{'is-default'};
6111 $current = $e->{name} if $e->{'is-current'};
6112 }
6113
6114 # fallback to the default machine if current is not supported by qemu
6115 return $current || $default || 'pc';
6116 }
6117
6118 sub lspci {
6119
6120 my $devices = {};
6121
6122 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6123 my (undef, $id, $function) = @_;
6124 my $res = { id => $id, function => $function};
6125 push @{$devices->{$id}}, $res;
6126 });
6127
6128 return $devices;
6129 }
6130
6131 1;