]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
vm_devices_list: make sure $device->{'qdev_id'} is set
[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
1435 return $res;
1436 }
1437
1438 sub add_random_macs {
1439 my ($settings) = @_;
1440
1441 foreach my $opt (keys %$settings) {
1442 next if $opt !~ m/^net(\d+)$/;
1443 my $net = parse_net($settings->{$opt});
1444 next if !$net;
1445 $settings->{$opt} = print_net($net);
1446 }
1447 }
1448
1449 sub add_unused_volume {
1450 my ($config, $volid) = @_;
1451
1452 my $key;
1453 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1454 my $test = "unused$ind";
1455 if (my $vid = $config->{$test}) {
1456 return if $vid eq $volid; # do not add duplicates
1457 } else {
1458 $key = $test;
1459 }
1460 }
1461
1462 die "To many unused volume - please delete them first.\n" if !$key;
1463
1464 $config->{$key} = $volid;
1465
1466 return $key;
1467 }
1468
1469 sub vm_is_volid_owner {
1470 my ($storecfg, $vmid, $volid) = @_;
1471
1472 if ($volid !~ m|^/|) {
1473 my ($path, $owner);
1474 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1475 if ($owner && ($owner == $vmid)) {
1476 return 1;
1477 }
1478 }
1479
1480 return undef;
1481 }
1482
1483 sub vmconfig_delete_pending_option {
1484 my ($conf, $key) = @_;
1485
1486 delete $conf->{pending}->{$key};
1487 my $pending_delete_hash = { $key => 1 };
1488 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1489 $pending_delete_hash->{$opt} = 1;
1490 }
1491 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1492 }
1493
1494 sub vmconfig_undelete_pending_option {
1495 my ($conf, $key) = @_;
1496
1497 my $pending_delete_hash = {};
1498 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1499 $pending_delete_hash->{$opt} = 1;
1500 }
1501 delete $pending_delete_hash->{$key};
1502
1503 my @keylist = keys %$pending_delete_hash;
1504 if (scalar(@keylist)) {
1505 $conf->{pending}->{delete} = join(',', @keylist);
1506 } else {
1507 delete $conf->{pending}->{delete};
1508 }
1509 }
1510
1511 sub vmconfig_register_unused_drive {
1512 my ($storecfg, $vmid, $conf, $drive) = @_;
1513
1514 if (!drive_is_cdrom($drive)) {
1515 my $volid = $drive->{file};
1516 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1517 add_unused_volume($conf, $volid, $vmid);
1518 }
1519 }
1520 }
1521
1522 sub vmconfig_cleanup_pending {
1523 my ($conf) = @_;
1524
1525 # remove pending changes when nothing changed
1526 my $changes;
1527 foreach my $opt (keys %{$conf->{pending}}) {
1528 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1529 $changes = 1;
1530 delete $conf->{pending}->{$opt};
1531 }
1532 }
1533
1534 # remove delete if option is not set
1535 my $pending_delete_hash = {};
1536 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1537 if (defined($conf->{$opt})) {
1538 $pending_delete_hash->{$opt} = 1;
1539 } else {
1540 $changes = 1;
1541 }
1542 }
1543
1544 my @keylist = keys %$pending_delete_hash;
1545 if (scalar(@keylist)) {
1546 $conf->{pending}->{delete} = join(',', @keylist);
1547 } else {
1548 delete $conf->{pending}->{delete};
1549 }
1550
1551 return $changes;
1552 }
1553
1554 my $valid_smbios1_options = {
1555 manufacturer => '\S+',
1556 product => '\S+',
1557 version => '\S+',
1558 serial => '\S+',
1559 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1560 sku => '\S+',
1561 family => '\S+',
1562 };
1563
1564 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1565 sub parse_smbios1 {
1566 my ($data) = @_;
1567
1568 my $res = {};
1569
1570 foreach my $kvp (split(/,/, $data)) {
1571 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1572 my ($k, $v) = split(/=/, $kvp);
1573 return undef if !defined($k) || !defined($v);
1574 return undef if !$valid_smbios1_options->{$k};
1575 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1576 $res->{$k} = $v;
1577 }
1578
1579 return $res;
1580 }
1581
1582 sub print_smbios1 {
1583 my ($smbios1) = @_;
1584
1585 my $data = '';
1586 foreach my $k (keys %$smbios1) {
1587 next if !defined($smbios1->{$k});
1588 next if !$valid_smbios1_options->{$k};
1589 $data .= ',' if $data;
1590 $data .= "$k=$smbios1->{$k}";
1591 }
1592 return $data;
1593 }
1594
1595 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1596 sub verify_smbios1 {
1597 my ($value, $noerr) = @_;
1598
1599 return $value if parse_smbios1($value);
1600
1601 return undef if $noerr;
1602
1603 die "unable to parse smbios (type 1) options\n";
1604 }
1605
1606 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1607 sub verify_bootdisk {
1608 my ($value, $noerr) = @_;
1609
1610 return $value if valid_drivename($value);
1611
1612 return undef if $noerr;
1613
1614 die "invalid boot disk '$value'\n";
1615 }
1616
1617 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1618 sub verify_numa {
1619 my ($value, $noerr) = @_;
1620
1621 return $value if parse_numa($value);
1622
1623 return undef if $noerr;
1624
1625 die "unable to parse numa options\n";
1626 }
1627
1628 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1629 sub verify_net {
1630 my ($value, $noerr) = @_;
1631
1632 return $value if parse_net($value);
1633
1634 return undef if $noerr;
1635
1636 die "unable to parse network options\n";
1637 }
1638
1639 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1640 sub verify_drive {
1641 my ($value, $noerr) = @_;
1642
1643 return $value if parse_drive(undef, $value);
1644
1645 return undef if $noerr;
1646
1647 die "unable to parse drive options\n";
1648 }
1649
1650 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1651 sub verify_hostpci {
1652 my ($value, $noerr) = @_;
1653
1654 return $value if parse_hostpci($value);
1655
1656 return undef if $noerr;
1657
1658 die "unable to parse pci id\n";
1659 }
1660
1661 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1662 sub verify_watchdog {
1663 my ($value, $noerr) = @_;
1664
1665 return $value if parse_watchdog($value);
1666
1667 return undef if $noerr;
1668
1669 die "unable to parse watchdog options\n";
1670 }
1671
1672 sub parse_watchdog {
1673 my ($value) = @_;
1674
1675 return undef if !$value;
1676
1677 my $res = {};
1678
1679 foreach my $p (split(/,/, $value)) {
1680 next if $p =~ m/^\s*$/;
1681
1682 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1683 $res->{model} = $2;
1684 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1685 $res->{action} = $2;
1686 } else {
1687 return undef;
1688 }
1689 }
1690
1691 return $res;
1692 }
1693
1694 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1695 sub verify_startup {
1696 my ($value, $noerr) = @_;
1697
1698 return $value if parse_startup($value);
1699
1700 return undef if $noerr;
1701
1702 die "unable to parse startup options\n";
1703 }
1704
1705 sub parse_startup {
1706 my ($value) = @_;
1707
1708 return undef if !$value;
1709
1710 my $res = {};
1711
1712 foreach my $p (split(/,/, $value)) {
1713 next if $p =~ m/^\s*$/;
1714
1715 if ($p =~ m/^(order=)?(\d+)$/) {
1716 $res->{order} = $2;
1717 } elsif ($p =~ m/^up=(\d+)$/) {
1718 $res->{up} = $1;
1719 } elsif ($p =~ m/^down=(\d+)$/) {
1720 $res->{down} = $1;
1721 } else {
1722 return undef;
1723 }
1724 }
1725
1726 return $res;
1727 }
1728
1729 sub parse_usb_device {
1730 my ($value) = @_;
1731
1732 return undef if !$value;
1733
1734 my @dl = split(/,/, $value);
1735 my $found;
1736
1737 my $res = {};
1738 foreach my $v (@dl) {
1739 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1740 $found = 1;
1741 $res->{vendorid} = $2;
1742 $res->{productid} = $4;
1743 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1744 $found = 1;
1745 $res->{hostbus} = $1;
1746 $res->{hostport} = $2;
1747 } elsif ($v =~ m/^spice$/) {
1748 $found = 1;
1749 $res->{spice} = 1;
1750 } else {
1751 return undef;
1752 }
1753 }
1754 return undef if !$found;
1755
1756 return $res;
1757 }
1758
1759 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1760 sub verify_usb_device {
1761 my ($value, $noerr) = @_;
1762
1763 return $value if parse_usb_device($value);
1764
1765 return undef if $noerr;
1766
1767 die "unable to parse usb device\n";
1768 }
1769
1770 # add JSON properties for create and set function
1771 sub json_config_properties {
1772 my $prop = shift;
1773
1774 foreach my $opt (keys %$confdesc) {
1775 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1776 $prop->{$opt} = $confdesc->{$opt};
1777 }
1778
1779 return $prop;
1780 }
1781
1782 sub check_type {
1783 my ($key, $value) = @_;
1784
1785 die "unknown setting '$key'\n" if !$confdesc->{$key};
1786
1787 my $type = $confdesc->{$key}->{type};
1788
1789 if (!defined($value)) {
1790 die "got undefined value\n";
1791 }
1792
1793 if ($value =~ m/[\n\r]/) {
1794 die "property contains a line feed\n";
1795 }
1796
1797 if ($type eq 'boolean') {
1798 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1799 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1800 die "type check ('boolean') failed - got '$value'\n";
1801 } elsif ($type eq 'integer') {
1802 return int($1) if $value =~ m/^(\d+)$/;
1803 die "type check ('integer') failed - got '$value'\n";
1804 } elsif ($type eq 'number') {
1805 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1806 die "type check ('number') failed - got '$value'\n";
1807 } elsif ($type eq 'string') {
1808 if (my $fmt = $confdesc->{$key}->{format}) {
1809 if ($fmt eq 'pve-qm-drive') {
1810 # special case - we need to pass $key to parse_drive()
1811 my $drive = parse_drive($key, $value);
1812 return $value if $drive;
1813 die "unable to parse drive options\n";
1814 }
1815 PVE::JSONSchema::check_format($fmt, $value);
1816 return $value;
1817 }
1818 $value =~ s/^\"(.*)\"$/$1/;
1819 return $value;
1820 } else {
1821 die "internal error"
1822 }
1823 }
1824
1825 sub lock_config_full {
1826 my ($vmid, $timeout, $code, @param) = @_;
1827
1828 my $filename = config_file_lock($vmid);
1829
1830 my $res = lock_file($filename, $timeout, $code, @param);
1831
1832 die $@ if $@;
1833
1834 return $res;
1835 }
1836
1837 sub lock_config_mode {
1838 my ($vmid, $timeout, $shared, $code, @param) = @_;
1839
1840 my $filename = config_file_lock($vmid);
1841
1842 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1843
1844 die $@ if $@;
1845
1846 return $res;
1847 }
1848
1849 sub lock_config {
1850 my ($vmid, $code, @param) = @_;
1851
1852 return lock_config_full($vmid, 10, $code, @param);
1853 }
1854
1855 sub cfs_config_path {
1856 my ($vmid, $node) = @_;
1857
1858 $node = $nodename if !$node;
1859 return "nodes/$node/qemu-server/$vmid.conf";
1860 }
1861
1862 sub check_iommu_support{
1863 #fixme : need to check IOMMU support
1864 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1865
1866 my $iommu=1;
1867 return $iommu;
1868
1869 }
1870
1871 sub config_file {
1872 my ($vmid, $node) = @_;
1873
1874 my $cfspath = cfs_config_path($vmid, $node);
1875 return "/etc/pve/$cfspath";
1876 }
1877
1878 sub config_file_lock {
1879 my ($vmid) = @_;
1880
1881 return "$lock_dir/lock-$vmid.conf";
1882 }
1883
1884 sub touch_config {
1885 my ($vmid) = @_;
1886
1887 my $conf = config_file($vmid);
1888 utime undef, undef, $conf;
1889 }
1890
1891 sub destroy_vm {
1892 my ($storecfg, $vmid, $keep_empty_config) = @_;
1893
1894 my $conffile = config_file($vmid);
1895
1896 my $conf = load_config($vmid);
1897
1898 check_lock($conf);
1899
1900 # only remove disks owned by this VM
1901 foreach_drive($conf, sub {
1902 my ($ds, $drive) = @_;
1903
1904 return if drive_is_cdrom($drive);
1905
1906 my $volid = $drive->{file};
1907
1908 return if !$volid || $volid =~ m|^/|;
1909
1910 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1911 return if !$path || !$owner || ($owner != $vmid);
1912
1913 PVE::Storage::vdisk_free($storecfg, $volid);
1914 });
1915
1916 if ($keep_empty_config) {
1917 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1918 } else {
1919 unlink $conffile;
1920 }
1921
1922 # also remove unused disk
1923 eval {
1924 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1925
1926 eval {
1927 PVE::Storage::foreach_volid($dl, sub {
1928 my ($volid, $sid, $volname, $d) = @_;
1929 PVE::Storage::vdisk_free($storecfg, $volid);
1930 });
1931 };
1932 warn $@ if $@;
1933
1934 };
1935 warn $@ if $@;
1936 }
1937
1938 sub load_config {
1939 my ($vmid, $node) = @_;
1940
1941 my $cfspath = cfs_config_path($vmid, $node);
1942
1943 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1944
1945 die "no such VM ('$vmid')\n" if !defined($conf);
1946
1947 return $conf;
1948 }
1949
1950 sub parse_vm_config {
1951 my ($filename, $raw) = @_;
1952
1953 return undef if !defined($raw);
1954
1955 my $res = {
1956 digest => Digest::SHA::sha1_hex($raw),
1957 snapshots => {},
1958 pending => {},
1959 };
1960
1961 $filename =~ m|/qemu-server/(\d+)\.conf$|
1962 || die "got strange filename '$filename'";
1963
1964 my $vmid = $1;
1965
1966 my $conf = $res;
1967 my $descr = '';
1968 my $section = '';
1969
1970 my @lines = split(/\n/, $raw);
1971 foreach my $line (@lines) {
1972 next if $line =~ m/^\s*$/;
1973
1974 if ($line =~ m/^\[PENDING\]\s*$/i) {
1975 $section = 'pending';
1976 $conf->{description} = $descr if $descr;
1977 $descr = '';
1978 $conf = $res->{$section} = {};
1979 next;
1980
1981 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1982 $section = $1;
1983 $conf->{description} = $descr if $descr;
1984 $descr = '';
1985 $conf = $res->{snapshots}->{$section} = {};
1986 next;
1987 }
1988
1989 if ($line =~ m/^\#(.*)\s*$/) {
1990 $descr .= PVE::Tools::decode_text($1) . "\n";
1991 next;
1992 }
1993
1994 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1995 $descr .= PVE::Tools::decode_text($2);
1996 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1997 $conf->{snapstate} = $1;
1998 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1999 my $key = $1;
2000 my $value = $2;
2001 $conf->{$key} = $value;
2002 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2003 my $value = $1;
2004 if ($section eq 'pending') {
2005 $conf->{delete} = $value; # we parse this later
2006 } else {
2007 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2008 }
2009 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2010 my $key = $1;
2011 my $value = $2;
2012 eval { $value = check_type($key, $value); };
2013 if ($@) {
2014 warn "vm $vmid - unable to parse value of '$key' - $@";
2015 } else {
2016 my $fmt = $confdesc->{$key}->{format};
2017 if ($fmt && $fmt eq 'pve-qm-drive') {
2018 my $v = parse_drive($key, $value);
2019 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2020 $v->{file} = $volid;
2021 $value = print_drive($vmid, $v);
2022 } else {
2023 warn "vm $vmid - unable to parse value of '$key'\n";
2024 next;
2025 }
2026 }
2027
2028 if ($key eq 'cdrom') {
2029 $conf->{ide2} = $value;
2030 } else {
2031 $conf->{$key} = $value;
2032 }
2033 }
2034 }
2035 }
2036
2037 $conf->{description} = $descr if $descr;
2038
2039 delete $res->{snapstate}; # just to be sure
2040
2041 return $res;
2042 }
2043
2044 sub write_vm_config {
2045 my ($filename, $conf) = @_;
2046
2047 delete $conf->{snapstate}; # just to be sure
2048
2049 if ($conf->{cdrom}) {
2050 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2051 $conf->{ide2} = $conf->{cdrom};
2052 delete $conf->{cdrom};
2053 }
2054
2055 # we do not use 'smp' any longer
2056 if ($conf->{sockets}) {
2057 delete $conf->{smp};
2058 } elsif ($conf->{smp}) {
2059 $conf->{sockets} = $conf->{smp};
2060 delete $conf->{cores};
2061 delete $conf->{smp};
2062 }
2063
2064 my $used_volids = {};
2065
2066 my $cleanup_config = sub {
2067 my ($cref, $pending, $snapname) = @_;
2068
2069 foreach my $key (keys %$cref) {
2070 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2071 $key eq 'snapstate' || $key eq 'pending';
2072 my $value = $cref->{$key};
2073 if ($key eq 'delete') {
2074 die "propertry 'delete' is only allowed in [PENDING]\n"
2075 if !$pending;
2076 # fixme: check syntax?
2077 next;
2078 }
2079 eval { $value = check_type($key, $value); };
2080 die "unable to parse value of '$key' - $@" if $@;
2081
2082 $cref->{$key} = $value;
2083
2084 if (!$snapname && valid_drivename($key)) {
2085 my $drive = parse_drive($key, $value);
2086 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2087 }
2088 }
2089 };
2090
2091 &$cleanup_config($conf);
2092
2093 &$cleanup_config($conf->{pending}, 1);
2094
2095 foreach my $snapname (keys %{$conf->{snapshots}}) {
2096 die "internal error" if $snapname eq 'pending';
2097 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2098 }
2099
2100 # remove 'unusedX' settings if we re-add a volume
2101 foreach my $key (keys %$conf) {
2102 my $value = $conf->{$key};
2103 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2104 delete $conf->{$key};
2105 }
2106 }
2107
2108 my $generate_raw_config = sub {
2109 my ($conf) = @_;
2110
2111 my $raw = '';
2112
2113 # add description as comment to top of file
2114 my $descr = $conf->{description} || '';
2115 foreach my $cl (split(/\n/, $descr)) {
2116 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2117 }
2118
2119 foreach my $key (sort keys %$conf) {
2120 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2121 $raw .= "$key: $conf->{$key}\n";
2122 }
2123 return $raw;
2124 };
2125
2126 my $raw = &$generate_raw_config($conf);
2127
2128 if (scalar(keys %{$conf->{pending}})){
2129 $raw .= "\n[PENDING]\n";
2130 $raw .= &$generate_raw_config($conf->{pending});
2131 }
2132
2133 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2134 $raw .= "\n[$snapname]\n";
2135 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2136 }
2137
2138 return $raw;
2139 }
2140
2141 sub update_config_nolock {
2142 my ($vmid, $conf, $skiplock) = @_;
2143
2144 check_lock($conf) if !$skiplock;
2145
2146 my $cfspath = cfs_config_path($vmid);
2147
2148 PVE::Cluster::cfs_write_file($cfspath, $conf);
2149 }
2150
2151 sub update_config {
2152 my ($vmid, $conf, $skiplock) = @_;
2153
2154 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2155 }
2156
2157 sub load_defaults {
2158
2159 my $res = {};
2160
2161 # we use static defaults from our JSON schema configuration
2162 foreach my $key (keys %$confdesc) {
2163 if (defined(my $default = $confdesc->{$key}->{default})) {
2164 $res->{$key} = $default;
2165 }
2166 }
2167
2168 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2169 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2170
2171 return $res;
2172 }
2173
2174 sub config_list {
2175 my $vmlist = PVE::Cluster::get_vmlist();
2176 my $res = {};
2177 return $res if !$vmlist || !$vmlist->{ids};
2178 my $ids = $vmlist->{ids};
2179
2180 foreach my $vmid (keys %$ids) {
2181 my $d = $ids->{$vmid};
2182 next if !$d->{node} || $d->{node} ne $nodename;
2183 next if !$d->{type} || $d->{type} ne 'qemu';
2184 $res->{$vmid}->{exists} = 1;
2185 }
2186 return $res;
2187 }
2188
2189 # test if VM uses local resources (to prevent migration)
2190 sub check_local_resources {
2191 my ($conf, $noerr) = @_;
2192
2193 my $loc_res = 0;
2194
2195 $loc_res = 1 if $conf->{hostusb}; # old syntax
2196 $loc_res = 1 if $conf->{hostpci}; # old syntax
2197
2198 foreach my $k (keys %$conf) {
2199 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2200 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2201 }
2202
2203 die "VM uses local resources\n" if $loc_res && !$noerr;
2204
2205 return $loc_res;
2206 }
2207
2208 # check if used storages are available on all nodes (use by migrate)
2209 sub check_storage_availability {
2210 my ($storecfg, $conf, $node) = @_;
2211
2212 foreach_drive($conf, sub {
2213 my ($ds, $drive) = @_;
2214
2215 my $volid = $drive->{file};
2216 return if !$volid;
2217
2218 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2219 return if !$sid;
2220
2221 # check if storage is available on both nodes
2222 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2223 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2224 });
2225 }
2226
2227 # list nodes where all VM images are available (used by has_feature API)
2228 sub shared_nodes {
2229 my ($conf, $storecfg) = @_;
2230
2231 my $nodelist = PVE::Cluster::get_nodelist();
2232 my $nodehash = { map { $_ => 1 } @$nodelist };
2233 my $nodename = PVE::INotify::nodename();
2234
2235 foreach_drive($conf, sub {
2236 my ($ds, $drive) = @_;
2237
2238 my $volid = $drive->{file};
2239 return if !$volid;
2240
2241 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2242 if ($storeid) {
2243 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2244 if ($scfg->{disable}) {
2245 $nodehash = {};
2246 } elsif (my $avail = $scfg->{nodes}) {
2247 foreach my $node (keys %$nodehash) {
2248 delete $nodehash->{$node} if !$avail->{$node};
2249 }
2250 } elsif (!$scfg->{shared}) {
2251 foreach my $node (keys %$nodehash) {
2252 delete $nodehash->{$node} if $node ne $nodename
2253 }
2254 }
2255 }
2256 });
2257
2258 return $nodehash
2259 }
2260
2261 sub check_lock {
2262 my ($conf) = @_;
2263
2264 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2265 }
2266
2267 sub check_cmdline {
2268 my ($pidfile, $pid) = @_;
2269
2270 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2271 if (defined($fh)) {
2272 my $line = <$fh>;
2273 $fh->close;
2274 return undef if !$line;
2275 my @param = split(/\0/, $line);
2276
2277 my $cmd = $param[0];
2278 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2279
2280 for (my $i = 0; $i < scalar (@param); $i++) {
2281 my $p = $param[$i];
2282 next if !$p;
2283 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2284 my $p = $param[$i+1];
2285 return 1 if $p && ($p eq $pidfile);
2286 return undef;
2287 }
2288 }
2289 }
2290 return undef;
2291 }
2292
2293 sub check_running {
2294 my ($vmid, $nocheck, $node) = @_;
2295
2296 my $filename = config_file($vmid, $node);
2297
2298 die "unable to find configuration file for VM $vmid - no such machine\n"
2299 if !$nocheck && ! -f $filename;
2300
2301 my $pidfile = pidfile_name($vmid);
2302
2303 if (my $fd = IO::File->new("<$pidfile")) {
2304 my $st = stat($fd);
2305 my $line = <$fd>;
2306 close($fd);
2307
2308 my $mtime = $st->mtime;
2309 if ($mtime > time()) {
2310 warn "file '$filename' modified in future\n";
2311 }
2312
2313 if ($line =~ m/^(\d+)$/) {
2314 my $pid = $1;
2315 if (check_cmdline($pidfile, $pid)) {
2316 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2317 return $pid;
2318 }
2319 }
2320 }
2321 }
2322
2323 return undef;
2324 }
2325
2326 sub vzlist {
2327
2328 my $vzlist = config_list();
2329
2330 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2331
2332 while (defined(my $de = $fd->read)) {
2333 next if $de !~ m/^(\d+)\.pid$/;
2334 my $vmid = $1;
2335 next if !defined($vzlist->{$vmid});
2336 if (my $pid = check_running($vmid)) {
2337 $vzlist->{$vmid}->{pid} = $pid;
2338 }
2339 }
2340
2341 return $vzlist;
2342 }
2343
2344 sub disksize {
2345 my ($storecfg, $conf) = @_;
2346
2347 my $bootdisk = $conf->{bootdisk};
2348 return undef if !$bootdisk;
2349 return undef if !valid_drivename($bootdisk);
2350
2351 return undef if !$conf->{$bootdisk};
2352
2353 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2354 return undef if !defined($drive);
2355
2356 return undef if drive_is_cdrom($drive);
2357
2358 my $volid = $drive->{file};
2359 return undef if !$volid;
2360
2361 return $drive->{size};
2362 }
2363
2364 my $last_proc_pid_stat;
2365
2366 # get VM status information
2367 # This must be fast and should not block ($full == false)
2368 # We only query KVM using QMP if $full == true (this can be slow)
2369 sub vmstatus {
2370 my ($opt_vmid, $full) = @_;
2371
2372 my $res = {};
2373
2374 my $storecfg = PVE::Storage::config();
2375
2376 my $list = vzlist();
2377 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2378
2379 my $cpucount = $cpuinfo->{cpus} || 1;
2380
2381 foreach my $vmid (keys %$list) {
2382 next if $opt_vmid && ($vmid ne $opt_vmid);
2383
2384 my $cfspath = cfs_config_path($vmid);
2385 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2386
2387 my $d = {};
2388 $d->{pid} = $list->{$vmid}->{pid};
2389
2390 # fixme: better status?
2391 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2392
2393 my $size = disksize($storecfg, $conf);
2394 if (defined($size)) {
2395 $d->{disk} = 0; # no info available
2396 $d->{maxdisk} = $size;
2397 } else {
2398 $d->{disk} = 0;
2399 $d->{maxdisk} = 0;
2400 }
2401
2402 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2403 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2404
2405 $d->{name} = $conf->{name} || "VM $vmid";
2406 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2407
2408 if ($conf->{balloon}) {
2409 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2410 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2411 }
2412
2413 $d->{uptime} = 0;
2414 $d->{cpu} = 0;
2415 $d->{mem} = 0;
2416
2417 $d->{netout} = 0;
2418 $d->{netin} = 0;
2419
2420 $d->{diskread} = 0;
2421 $d->{diskwrite} = 0;
2422
2423 $d->{template} = is_template($conf);
2424
2425 $res->{$vmid} = $d;
2426 }
2427
2428 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2429 foreach my $dev (keys %$netdev) {
2430 next if $dev !~ m/^tap([1-9]\d*)i/;
2431 my $vmid = $1;
2432 my $d = $res->{$vmid};
2433 next if !$d;
2434
2435 $d->{netout} += $netdev->{$dev}->{receive};
2436 $d->{netin} += $netdev->{$dev}->{transmit};
2437 }
2438
2439 my $ctime = gettimeofday;
2440
2441 foreach my $vmid (keys %$list) {
2442
2443 my $d = $res->{$vmid};
2444 my $pid = $d->{pid};
2445 next if !$pid;
2446
2447 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2448 next if !$pstat; # not running
2449
2450 my $used = $pstat->{utime} + $pstat->{stime};
2451
2452 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2453
2454 if ($pstat->{vsize}) {
2455 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2456 }
2457
2458 my $old = $last_proc_pid_stat->{$pid};
2459 if (!$old) {
2460 $last_proc_pid_stat->{$pid} = {
2461 time => $ctime,
2462 used => $used,
2463 cpu => 0,
2464 };
2465 next;
2466 }
2467
2468 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2469
2470 if ($dtime > 1000) {
2471 my $dutime = $used - $old->{used};
2472
2473 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2474 $last_proc_pid_stat->{$pid} = {
2475 time => $ctime,
2476 used => $used,
2477 cpu => $d->{cpu},
2478 };
2479 } else {
2480 $d->{cpu} = $old->{cpu};
2481 }
2482 }
2483
2484 return $res if !$full;
2485
2486 my $qmpclient = PVE::QMPClient->new();
2487
2488 my $ballooncb = sub {
2489 my ($vmid, $resp) = @_;
2490
2491 my $info = $resp->{'return'};
2492 return if !$info->{max_mem};
2493
2494 my $d = $res->{$vmid};
2495
2496 # use memory assigned to VM
2497 $d->{maxmem} = $info->{max_mem};
2498 $d->{balloon} = $info->{actual};
2499
2500 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2501 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2502 $d->{freemem} = $info->{free_mem};
2503 }
2504
2505 };
2506
2507 my $blockstatscb = sub {
2508 my ($vmid, $resp) = @_;
2509 my $data = $resp->{'return'} || [];
2510 my $totalrdbytes = 0;
2511 my $totalwrbytes = 0;
2512 for my $blockstat (@$data) {
2513 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2514 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2515 }
2516 $res->{$vmid}->{diskread} = $totalrdbytes;
2517 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2518 };
2519
2520 my $statuscb = sub {
2521 my ($vmid, $resp) = @_;
2522
2523 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2524 # this fails if ballon driver is not loaded, so this must be
2525 # the last commnand (following command are aborted if this fails).
2526 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2527
2528 my $status = 'unknown';
2529 if (!defined($status = $resp->{'return'}->{status})) {
2530 warn "unable to get VM status\n";
2531 return;
2532 }
2533
2534 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2535 };
2536
2537 foreach my $vmid (keys %$list) {
2538 next if $opt_vmid && ($vmid ne $opt_vmid);
2539 next if !$res->{$vmid}->{pid}; # not running
2540 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2541 }
2542
2543 $qmpclient->queue_execute(undef, 1);
2544
2545 foreach my $vmid (keys %$list) {
2546 next if $opt_vmid && ($vmid ne $opt_vmid);
2547 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2548 }
2549
2550 return $res;
2551 }
2552
2553 sub foreach_dimm {
2554 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2555
2556 my $dimm_id = 0;
2557 my $current_size = 1024;
2558 my $dimm_size = 512;
2559 return if $current_size == $memory;
2560
2561 for (my $j = 0; $j < 8; $j++) {
2562 for (my $i = 0; $i < 32; $i++) {
2563 my $name = "dimm${dimm_id}";
2564 $dimm_id++;
2565 my $numanode = $i % $sockets;
2566 $current_size += $dimm_size;
2567 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2568 return $current_size if $current_size >= $memory;
2569 }
2570 $dimm_size *= 2;
2571 }
2572 }
2573
2574 sub foreach_drive {
2575 my ($conf, $func) = @_;
2576
2577 foreach my $ds (keys %$conf) {
2578 next if !valid_drivename($ds);
2579
2580 my $drive = parse_drive($ds, $conf->{$ds});
2581 next if !$drive;
2582
2583 &$func($ds, $drive);
2584 }
2585 }
2586
2587 sub foreach_volid {
2588 my ($conf, $func) = @_;
2589
2590 my $volhash = {};
2591
2592 my $test_volid = sub {
2593 my ($volid, $is_cdrom) = @_;
2594
2595 return if !$volid;
2596
2597 $volhash->{$volid} = $is_cdrom || 0;
2598 };
2599
2600 foreach_drive($conf, sub {
2601 my ($ds, $drive) = @_;
2602 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2603 });
2604
2605 foreach my $snapname (keys %{$conf->{snapshots}}) {
2606 my $snap = $conf->{snapshots}->{$snapname};
2607 &$test_volid($snap->{vmstate}, 0);
2608 foreach_drive($snap, sub {
2609 my ($ds, $drive) = @_;
2610 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2611 });
2612 }
2613
2614 foreach my $volid (keys %$volhash) {
2615 &$func($volid, $volhash->{$volid});
2616 }
2617 }
2618
2619 sub vga_conf_has_spice {
2620 my ($vga) = @_;
2621
2622 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2623
2624 return $1 || 1;
2625 }
2626
2627 sub config_to_command {
2628 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2629
2630 my $cmd = [];
2631 my $globalFlags = [];
2632 my $machineFlags = [];
2633 my $rtcFlags = [];
2634 my $cpuFlags = [];
2635 my $devices = [];
2636 my $pciaddr = '';
2637 my $bridges = {};
2638 my $kvmver = kvm_user_version();
2639 my $vernum = 0; # unknown
2640 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2641 $vernum = $1*1000000+$2*1000;
2642 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2643 $vernum = $1*1000000+$2*1000+$3;
2644 }
2645
2646 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2647
2648 my $have_ovz = -f '/proc/vz/vestat';
2649
2650 my $q35 = machine_type_is_q35($conf);
2651 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2652
2653 push @$cmd, '/usr/bin/kvm';
2654
2655 push @$cmd, '-id', $vmid;
2656
2657 my $use_virtio = 0;
2658
2659 my $qmpsocket = qmp_socket($vmid);
2660 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2661 push @$cmd, '-mon', "chardev=qmp,mode=control";
2662
2663 my $socket = vnc_socket($vmid);
2664 push @$cmd, '-vnc', "unix:$socket,x509,password";
2665
2666 push @$cmd, '-pidfile' , pidfile_name($vmid);
2667
2668 push @$cmd, '-daemonize';
2669
2670 if ($conf->{smbios1}) {
2671 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2672 }
2673
2674 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2675
2676 if ($q35) {
2677 # the q35 chipset support native usb2, so we enable usb controller
2678 # by default for this machine type
2679 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2680 } else {
2681 $pciaddr = print_pci_addr("piix3", $bridges);
2682 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2683
2684 my $use_usb2 = 0;
2685 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2686 next if !$conf->{"usb$i"};
2687 $use_usb2 = 1;
2688 }
2689 # include usb device config
2690 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2691 }
2692
2693 my $vga = $conf->{vga};
2694
2695 my $qxlnum = vga_conf_has_spice($vga);
2696 $vga = 'qxl' if $qxlnum;
2697
2698 if (!$vga) {
2699 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2700 $conf->{ostype} eq 'win7' ||
2701 $conf->{ostype} eq 'w2k8')) {
2702 $vga = 'std';
2703 } else {
2704 $vga = 'cirrus';
2705 }
2706 }
2707
2708 # enable absolute mouse coordinates (needed by vnc)
2709 my $tablet;
2710 if (defined($conf->{tablet})) {
2711 $tablet = $conf->{tablet};
2712 } else {
2713 $tablet = $defaults->{tablet};
2714 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2715 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2716 }
2717
2718 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2719
2720 # host pci devices
2721 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2722 my $d = parse_hostpci($conf->{"hostpci$i"});
2723 next if !$d;
2724
2725 my $pcie = $d->{pcie};
2726 if($pcie){
2727 die "q35 machine model is not enabled" if !$q35;
2728 $pciaddr = print_pcie_addr("hostpci$i");
2729 }else{
2730 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2731 }
2732
2733 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2734 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2735 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2736 if ($xvga && $xvga ne '') {
2737 push @$cpuFlags, 'kvm=off';
2738 $vga = 'none';
2739 }
2740 $driver = "vfio-pci" if $xvga ne '';
2741 my $pcidevices = $d->{pciid};
2742 my $multifunction = 1 if @$pcidevices > 1;
2743
2744 my $j=0;
2745 foreach my $pcidevice (@$pcidevices) {
2746
2747 my $id = "hostpci$i";
2748 $id .= ".$j" if $multifunction;
2749 my $addr = $pciaddr;
2750 $addr .= ".$j" if $multifunction;
2751 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2752
2753 if($j == 0){
2754 $devicestr .= "$rombar$xvga";
2755 $devicestr .= ",multifunction=on" if $multifunction;
2756 }
2757
2758 push @$devices, '-device', $devicestr;
2759 $j++;
2760 }
2761 }
2762
2763 # usb devices
2764 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2765 my $d = parse_usb_device($conf->{"usb$i"});
2766 next if !$d;
2767 if ($d->{vendorid} && $d->{productid}) {
2768 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2769 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2770 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2771 } elsif ($d->{spice}) {
2772 # usb redir support for spice
2773 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2774 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2775 }
2776 }
2777
2778 # serial devices
2779 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2780 if (my $path = $conf->{"serial$i"}) {
2781 if ($path eq 'socket') {
2782 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2783 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2784 push @$devices, '-device', "isa-serial,chardev=serial$i";
2785 } else {
2786 die "no such serial device\n" if ! -c $path;
2787 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2788 push @$devices, '-device', "isa-serial,chardev=serial$i";
2789 }
2790 }
2791 }
2792
2793 # parallel devices
2794 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2795 if (my $path = $conf->{"parallel$i"}) {
2796 die "no such parallel device\n" if ! -c $path;
2797 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2798 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2799 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2800 }
2801 }
2802
2803 my $vmname = $conf->{name} || "vm$vmid";
2804
2805 push @$cmd, '-name', $vmname;
2806
2807 my $sockets = 1;
2808 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2809 $sockets = $conf->{sockets} if $conf->{sockets};
2810
2811 my $cores = $conf->{cores} || 1;
2812
2813 my $maxcpus = $sockets * $cores;
2814
2815 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2816
2817 my $allowed_vcpus = $cpuinfo->{cpus};
2818
2819 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2820 if ($allowed_vcpus < $maxcpus);
2821
2822 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2823
2824 push @$cmd, '-nodefaults';
2825
2826 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2827
2828 my $bootindex_hash = {};
2829 my $i = 1;
2830 foreach my $o (split(//, $bootorder)) {
2831 $bootindex_hash->{$o} = $i*100;
2832 $i++;
2833 }
2834
2835 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2836
2837 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2838
2839 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2840
2841 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2842
2843 # time drift fix
2844 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2845
2846 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2847 my $useLocaltime = $conf->{localtime};
2848
2849 if (my $ost = $conf->{ostype}) {
2850 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2851
2852 if ($ost =~ m/^w/) { # windows
2853 $useLocaltime = 1 if !defined($conf->{localtime});
2854
2855 # use time drift fix when acpi is enabled
2856 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2857 $tdf = 1 if !defined($conf->{tdf});
2858 }
2859 }
2860
2861 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2862 $ost eq 'wvista') {
2863 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2864 push @$cmd, '-no-hpet';
2865 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2866 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2867 }
2868
2869 if ($ost eq 'win7' || $ost eq 'win8') {
2870 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2871 }
2872 }
2873
2874 push @$rtcFlags, 'driftfix=slew' if $tdf;
2875
2876 if ($nokvm) {
2877 push @$machineFlags, 'accel=tcg';
2878 } else {
2879 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2880 }
2881
2882 my $machine_type = $forcemachine || $conf->{machine};
2883 if ($machine_type) {
2884 push @$machineFlags, "type=${machine_type}";
2885 }
2886
2887 if ($conf->{startdate}) {
2888 push @$rtcFlags, "base=$conf->{startdate}";
2889 } elsif ($useLocaltime) {
2890 push @$rtcFlags, 'base=localtime';
2891 }
2892
2893 my $cpu = $nokvm ? "qemu64" : "kvm64";
2894 $cpu = $conf->{cpu} if $conf->{cpu};
2895
2896 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2897
2898 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2899
2900 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2901
2902 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2903
2904 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2905
2906 # Note: enforce needs kernel 3.10, so we do not use it for now
2907 # push @$cmd, '-cpu', "$cpu,enforce";
2908 push @$cmd, '-cpu', $cpu;
2909
2910 my $memory = $conf->{memory} || $defaults->{memory};
2911 my $static_memory = 0;
2912 my $dimm_memory = 0;
2913
2914 if ($hotplug_features->{memory}) {
2915 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2916 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2917 $static_memory = $STATICMEM;
2918 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2919 $dimm_memory = $memory - $static_memory;
2920 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2921
2922 } else {
2923
2924 $static_memory = $memory;
2925 push @$cmd, '-m', $static_memory;
2926 }
2927
2928 if ($conf->{numa}) {
2929
2930 my $numa_totalmemory = undef;
2931 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2932 next if !$conf->{"numa$i"};
2933 my $numa = parse_numa($conf->{"numa$i"});
2934 next if !$numa;
2935 # memory
2936 die "missing numa node$i memory value\n" if !$numa->{memory};
2937 my $numa_memory = $numa->{memory};
2938 $numa_totalmemory += $numa_memory;
2939 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2940
2941 # cpus
2942 my $cpus_start = $numa->{cpus}->{start};
2943 die "missing numa node$i cpus\n" if !defined($cpus_start);
2944 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2945 my $cpus = $cpus_start;
2946 if (defined($cpus_end)) {
2947 $cpus .= "-$cpus_end";
2948 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2949 }
2950
2951 # hostnodes
2952 my $hostnodes_start = $numa->{hostnodes}->{start};
2953 if (defined($hostnodes_start)) {
2954 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2955 my $hostnodes = $hostnodes_start;
2956 if (defined($hostnodes_end)) {
2957 $hostnodes .= "-$hostnodes_end";
2958 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2959 }
2960
2961 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2962 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2963 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2964 }
2965
2966 # policy
2967 my $policy = $numa->{policy};
2968 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2969 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2970 }
2971
2972 push @$cmd, '-object', $numa_object;
2973 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2974 }
2975
2976 die "total memory for NUMA nodes must be equal to vm static memory\n"
2977 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2978
2979 #if no custom tology, we split memory and cores across numa nodes
2980 if(!$numa_totalmemory) {
2981
2982 my $numa_memory = ($static_memory / $sockets) . "M";
2983
2984 for (my $i = 0; $i < $sockets; $i++) {
2985
2986 my $cpustart = ($cores * $i);
2987 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2988 my $cpus = $cpustart;
2989 $cpus .= "-$cpuend" if $cpuend;
2990
2991 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2992 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2993 }
2994 }
2995 }
2996
2997 if ($hotplug_features->{memory}) {
2998 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
2999 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3000 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3001 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3002
3003 #if dimm_memory is not aligned to dimm map
3004 if($current_size > $memory) {
3005 $conf->{memory} = $current_size;
3006 update_config_nolock($vmid, $conf, 1);
3007 }
3008 });
3009 }
3010
3011 push @$cmd, '-S' if $conf->{freeze};
3012
3013 # set keyboard layout
3014 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3015 push @$cmd, '-k', $kb if $kb;
3016
3017 # enable sound
3018 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3019 #push @$cmd, '-soundhw', 'es1370';
3020 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3021
3022 if($conf->{agent}) {
3023 my $qgasocket = qmp_socket($vmid, 1);
3024 my $pciaddr = print_pci_addr("qga0", $bridges);
3025 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3026 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3027 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3028 }
3029
3030 my $spice_port;
3031
3032 if ($qxlnum) {
3033 if ($qxlnum > 1) {
3034 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3035 for(my $i = 1; $i < $qxlnum; $i++){
3036 my $pciaddr = print_pci_addr("vga$i", $bridges);
3037 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3038 }
3039 } else {
3040 # assume other OS works like Linux
3041 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3042 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3043 }
3044 }
3045
3046 my $pciaddr = print_pci_addr("spice", $bridges);
3047
3048 $spice_port = PVE::Tools::next_spice_port();
3049
3050 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3051
3052 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3053 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3054 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3055 }
3056
3057 # enable balloon by default, unless explicitly disabled
3058 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3059 $pciaddr = print_pci_addr("balloon0", $bridges);
3060 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3061 }
3062
3063 if ($conf->{watchdog}) {
3064 my $wdopts = parse_watchdog($conf->{watchdog});
3065 $pciaddr = print_pci_addr("watchdog", $bridges);
3066 my $watchdog = $wdopts->{model} || 'i6300esb';
3067 push @$devices, '-device', "$watchdog$pciaddr";
3068 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3069 }
3070
3071 my $vollist = [];
3072 my $scsicontroller = {};
3073 my $ahcicontroller = {};
3074 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3075
3076 # Add iscsi initiator name if available
3077 if (my $initiator = get_initiator_name()) {
3078 push @$devices, '-iscsi', "initiator-name=$initiator";
3079 }
3080
3081 foreach_drive($conf, sub {
3082 my ($ds, $drive) = @_;
3083
3084 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3085 push @$vollist, $drive->{file};
3086 }
3087
3088 $use_virtio = 1 if $ds =~ m/^virtio/;
3089
3090 if (drive_is_cdrom ($drive)) {
3091 if ($bootindex_hash->{d}) {
3092 $drive->{bootindex} = $bootindex_hash->{d};
3093 $bootindex_hash->{d} += 1;
3094 }
3095 } else {
3096 if ($bootindex_hash->{c}) {
3097 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3098 $bootindex_hash->{c} += 1;
3099 }
3100 }
3101
3102 if ($drive->{interface} eq 'scsi') {
3103
3104 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3105 my $controller = int($drive->{index} / $maxdev);
3106 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3107 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3108 $scsicontroller->{$controller}=1;
3109 }
3110
3111 if ($drive->{interface} eq 'sata') {
3112 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3113 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3114 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3115 $ahcicontroller->{$controller}=1;
3116 }
3117
3118 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3119 push @$devices, '-drive',$drive_cmd;
3120 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3121 });
3122
3123 for (my $i = 0; $i < $MAX_NETS; $i++) {
3124 next if !$conf->{"net$i"};
3125 my $d = parse_net($conf->{"net$i"});
3126 next if !$d;
3127
3128 $use_virtio = 1 if $d->{model} eq 'virtio';
3129
3130 if ($bootindex_hash->{n}) {
3131 $d->{bootindex} = $bootindex_hash->{n};
3132 $bootindex_hash->{n} += 1;
3133 }
3134
3135 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3136 push @$devices, '-netdev', $netdevfull;
3137
3138 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3139 push @$devices, '-device', $netdevicefull;
3140 }
3141
3142 if (!$q35) {
3143 # add pci bridges
3144 while (my ($k, $v) = each %$bridges) {
3145 $pciaddr = print_pci_addr("pci.$k");
3146 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3147 }
3148 }
3149
3150 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3151 # when the VM uses virtio devices.
3152 if (!$use_virtio && $have_ovz) {
3153
3154 my $cpuunits = defined($conf->{cpuunits}) ?
3155 $conf->{cpuunits} : $defaults->{cpuunits};
3156
3157 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3158
3159 # fixme: cpulimit is currently ignored
3160 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3161 }
3162
3163 # add custom args
3164 if ($conf->{args}) {
3165 my $aa = PVE::Tools::split_args($conf->{args});
3166 push @$cmd, @$aa;
3167 }
3168
3169 push @$cmd, @$devices;
3170 push @$cmd, '-rtc', join(',', @$rtcFlags)
3171 if scalar(@$rtcFlags);
3172 push @$cmd, '-machine', join(',', @$machineFlags)
3173 if scalar(@$machineFlags);
3174 push @$cmd, '-global', join(',', @$globalFlags)
3175 if scalar(@$globalFlags);
3176
3177 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3178 }
3179
3180 sub vnc_socket {
3181 my ($vmid) = @_;
3182 return "${var_run_tmpdir}/$vmid.vnc";
3183 }
3184
3185 sub spice_port {
3186 my ($vmid) = @_;
3187
3188 my $res = vm_mon_cmd($vmid, 'query-spice');
3189
3190 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3191 }
3192
3193 sub qmp_socket {
3194 my ($vmid, $qga) = @_;
3195 my $sockettype = $qga ? 'qga' : 'qmp';
3196 return "${var_run_tmpdir}/$vmid.$sockettype";
3197 }
3198
3199 sub pidfile_name {
3200 my ($vmid) = @_;
3201 return "${var_run_tmpdir}/$vmid.pid";
3202 }
3203
3204 sub vm_devices_list {
3205 my ($vmid) = @_;
3206
3207 my $res = vm_mon_cmd($vmid, 'query-pci');
3208 my $devices = {};
3209 foreach my $pcibus (@$res) {
3210 foreach my $device (@{$pcibus->{devices}}) {
3211 next if !$device->{'qdev_id'};
3212 if ($device->{'pci_bridge'}) {
3213 $devices->{$device->{'qdev_id'}} = 1;
3214 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3215 next if !$bridge_device->{'qdev_id'};
3216 $devices->{$bridge_device->{'qdev_id'}} = 1;
3217 $devices->{$device->{'qdev_id'}}++;
3218 }
3219 } else {
3220 $devices->{$device->{'qdev_id'}} = 1;
3221 }
3222 }
3223 }
3224
3225 my $resblock = vm_mon_cmd($vmid, 'query-block');
3226 foreach my $block (@$resblock) {
3227 if($block->{device} =~ m/^drive-(\S+)/){
3228 $devices->{$1} = 1;
3229 }
3230 }
3231
3232 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3233 foreach my $mice (@$resmice) {
3234 if ($mice->{name} eq 'QEMU HID Tablet') {
3235 $devices->{tablet} = 1;
3236 last;
3237 }
3238 }
3239
3240 return $devices;
3241 }
3242
3243 sub vm_deviceplug {
3244 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3245
3246 my $q35 = machine_type_is_q35($conf);
3247
3248 my $devices_list = vm_devices_list($vmid);
3249 return 1 if defined($devices_list->{$deviceid});
3250
3251 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3252
3253 if ($deviceid eq 'tablet') {
3254
3255 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3256
3257 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3258
3259 qemu_driveadd($storecfg, $vmid, $device);
3260 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3261
3262 qemu_deviceadd($vmid, $devicefull);
3263 eval { qemu_deviceaddverify($vmid, $deviceid); };
3264 if (my $err = $@) {
3265 eval { qemu_drivedel($vmid, $deviceid); };
3266 warn $@ if $@;
3267 die $err;
3268 }
3269
3270 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3271
3272 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3273 my $pciaddr = print_pci_addr($deviceid);
3274 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3275
3276 qemu_deviceadd($vmid, $devicefull);
3277 qemu_deviceaddverify($vmid, $deviceid);
3278
3279 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3280
3281 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3282 qemu_driveadd($storecfg, $vmid, $device);
3283
3284 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3285 eval { qemu_deviceadd($vmid, $devicefull); };
3286 if (my $err = $@) {
3287 eval { qemu_drivedel($vmid, $deviceid); };
3288 warn $@ if $@;
3289 die $err;
3290 }
3291
3292 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3293
3294 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3295 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3296 qemu_deviceadd($vmid, $netdevicefull);
3297 eval { qemu_deviceaddverify($vmid, $deviceid); };
3298 if (my $err = $@) {
3299 eval { qemu_netdevdel($vmid, $deviceid); };
3300 warn $@ if $@;
3301 die $err;
3302 }
3303
3304 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3305
3306 my $bridgeid = $2;
3307 my $pciaddr = print_pci_addr($deviceid);
3308 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3309
3310 qemu_deviceadd($vmid, $devicefull);
3311 qemu_deviceaddverify($vmid, $deviceid);
3312
3313 } else {
3314 die "can't hotplug device '$deviceid'\n";
3315 }
3316
3317 return 1;
3318 }
3319
3320 # fixme: this should raise exceptions on error!
3321 sub vm_deviceunplug {
3322 my ($vmid, $conf, $deviceid) = @_;
3323
3324 my $devices_list = vm_devices_list($vmid);
3325 return 1 if !defined($devices_list->{$deviceid});
3326
3327 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3328
3329 if ($deviceid eq 'tablet') {
3330
3331 qemu_devicedel($vmid, $deviceid);
3332
3333 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3334
3335 qemu_devicedel($vmid, $deviceid);
3336 qemu_devicedelverify($vmid, $deviceid);
3337 qemu_drivedel($vmid, $deviceid);
3338
3339 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3340
3341 qemu_devicedel($vmid, $deviceid);
3342
3343 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3344
3345 qemu_devicedel($vmid, $deviceid);
3346 qemu_drivedel($vmid, $deviceid);
3347
3348 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3349
3350 qemu_devicedel($vmid, $deviceid);
3351 qemu_devicedelverify($vmid, $deviceid);
3352 qemu_netdevdel($vmid, $deviceid);
3353
3354 } else {
3355 die "can't unplug device '$deviceid'\n";
3356 }
3357
3358 return 1;
3359 }
3360
3361 sub qemu_deviceadd {
3362 my ($vmid, $devicefull) = @_;
3363
3364 $devicefull = "driver=".$devicefull;
3365 my %options = split(/[=,]/, $devicefull);
3366
3367 vm_mon_cmd($vmid, "device_add" , %options);
3368 }
3369
3370 sub qemu_devicedel {
3371 my ($vmid, $deviceid) = @_;
3372
3373 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3374 }
3375
3376 sub qemu_objectadd {
3377 my($vmid, $objectid, $qomtype) = @_;
3378
3379 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3380
3381 return 1;
3382 }
3383
3384 sub qemu_objectdel {
3385 my($vmid, $objectid) = @_;
3386
3387 vm_mon_cmd($vmid, "object-del", id => $objectid);
3388
3389 return 1;
3390 }
3391
3392 sub qemu_driveadd {
3393 my ($storecfg, $vmid, $device) = @_;
3394
3395 my $drive = print_drive_full($storecfg, $vmid, $device);
3396 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3397
3398 # If the command succeeds qemu prints: "OK"
3399 return 1 if $ret =~ m/OK/s;
3400
3401 die "adding drive failed: $ret\n";
3402 }
3403
3404 sub qemu_drivedel {
3405 my($vmid, $deviceid) = @_;
3406
3407 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3408 $ret =~ s/^\s+//;
3409
3410 return 1 if $ret eq "";
3411
3412 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3413 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3414
3415 die "deleting drive $deviceid failed : $ret\n";
3416 }
3417
3418 sub qemu_deviceaddverify {
3419 my ($vmid, $deviceid) = @_;
3420
3421 for (my $i = 0; $i <= 5; $i++) {
3422 my $devices_list = vm_devices_list($vmid);
3423 return 1 if defined($devices_list->{$deviceid});
3424 sleep 1;
3425 }
3426
3427 die "error on hotplug device '$deviceid'\n";
3428 }
3429
3430
3431 sub qemu_devicedelverify {
3432 my ($vmid, $deviceid) = @_;
3433
3434 # need to verify that the device is correctly removed as device_del
3435 # is async and empty return is not reliable
3436
3437 for (my $i = 0; $i <= 5; $i++) {
3438 my $devices_list = vm_devices_list($vmid);
3439 return 1 if !defined($devices_list->{$deviceid});
3440 sleep 1;
3441 }
3442
3443 die "error on hot-unplugging device '$deviceid'\n";
3444 }
3445
3446 sub qemu_findorcreatescsihw {
3447 my ($storecfg, $conf, $vmid, $device) = @_;
3448
3449 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3450 my $controller = int($device->{index} / $maxdev);
3451 my $scsihwid="scsihw$controller";
3452 my $devices_list = vm_devices_list($vmid);
3453
3454 if(!defined($devices_list->{$scsihwid})) {
3455 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3456 }
3457
3458 return 1;
3459 }
3460
3461 sub qemu_add_pci_bridge {
3462 my ($storecfg, $conf, $vmid, $device) = @_;
3463
3464 my $bridges = {};
3465
3466 my $bridgeid;
3467
3468 print_pci_addr($device, $bridges);
3469
3470 while (my ($k, $v) = each %$bridges) {
3471 $bridgeid = $k;
3472 }
3473 return 1 if !defined($bridgeid) || $bridgeid < 1;
3474
3475 my $bridge = "pci.$bridgeid";
3476 my $devices_list = vm_devices_list($vmid);
3477
3478 if (!defined($devices_list->{$bridge})) {
3479 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3480 }
3481
3482 return 1;
3483 }
3484
3485 sub qemu_set_link_status {
3486 my ($vmid, $device, $up) = @_;
3487
3488 vm_mon_cmd($vmid, "set_link", name => $device,
3489 up => $up ? JSON::true : JSON::false);
3490 }
3491
3492 sub qemu_netdevadd {
3493 my ($vmid, $conf, $device, $deviceid) = @_;
3494
3495 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3496 my %options = split(/[=,]/, $netdev);
3497
3498 vm_mon_cmd($vmid, "netdev_add", %options);
3499 return 1;
3500 }
3501
3502 sub qemu_netdevdel {
3503 my ($vmid, $deviceid) = @_;
3504
3505 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3506 }
3507
3508 sub qemu_cpu_hotplug {
3509 my ($vmid, $conf, $vcpus) = @_;
3510
3511 my $sockets = 1;
3512 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3513 $sockets = $conf->{sockets} if $conf->{sockets};
3514 my $cores = $conf->{cores} || 1;
3515 my $maxcpus = $sockets * $cores;
3516
3517 $vcpus = $maxcpus if !$vcpus;
3518
3519 die "you can't add more vcpus than maxcpus\n"
3520 if $vcpus > $maxcpus;
3521
3522 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3523 die "online cpu unplug is not yet possible\n"
3524 if $vcpus < $currentvcpus;
3525
3526 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3527 die "vcpus in running vm is different than configuration\n"
3528 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3529
3530 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3531 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3532 }
3533 }
3534
3535 sub qemu_memory_hotplug {
3536 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3537
3538 return $value if !check_running($vmid);
3539
3540 my $memory = $conf->{memory} || $defaults->{memory};
3541 $value = $defaults->{memory} if !$value;
3542 return $value if $value == $memory;
3543
3544 my $static_memory = $STATICMEM;
3545 my $dimm_memory = $memory - $static_memory;
3546
3547 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3548 die "memory unplug is not yet available" if $value < $memory;
3549 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3550
3551
3552 my $sockets = 1;
3553 $sockets = $conf->{sockets} if $conf->{sockets};
3554
3555 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3556 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3557
3558 return if $current_size <= $conf->{memory};
3559
3560 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3561 if (my $err = $@) {
3562 eval { qemu_objectdel($vmid, "mem-$name"); };
3563 die $err;
3564 }
3565
3566 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3567 if (my $err = $@) {
3568 eval { qemu_objectdel($vmid, "mem-$name"); };
3569 die $err;
3570 }
3571 #update conf after each succesful module hotplug
3572 $conf->{memory} = $current_size;
3573 update_config_nolock($vmid, $conf, 1);
3574 });
3575 }
3576
3577 sub qemu_block_set_io_throttle {
3578 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3579
3580 return if !check_running($vmid) ;
3581
3582 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($iops_wr));
3583
3584 }
3585
3586 # old code, only used to shutdown old VM after update
3587 sub __read_avail {
3588 my ($fh, $timeout) = @_;
3589
3590 my $sel = new IO::Select;
3591 $sel->add($fh);
3592
3593 my $res = '';
3594 my $buf;
3595
3596 my @ready;
3597 while (scalar (@ready = $sel->can_read($timeout))) {
3598 my $count;
3599 if ($count = $fh->sysread($buf, 8192)) {
3600 if ($buf =~ /^(.*)\(qemu\) $/s) {
3601 $res .= $1;
3602 last;
3603 } else {
3604 $res .= $buf;
3605 }
3606 } else {
3607 if (!defined($count)) {
3608 die "$!\n";
3609 }
3610 last;
3611 }
3612 }
3613
3614 die "monitor read timeout\n" if !scalar(@ready);
3615
3616 return $res;
3617 }
3618
3619 # old code, only used to shutdown old VM after update
3620 sub vm_monitor_command {
3621 my ($vmid, $cmdstr, $nocheck) = @_;
3622
3623 my $res;
3624
3625 eval {
3626 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3627
3628 my $sname = "${var_run_tmpdir}/$vmid.mon";
3629
3630 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3631 die "unable to connect to VM $vmid socket - $!\n";
3632
3633 my $timeout = 3;
3634
3635 # hack: migrate sometime blocks the monitor (when migrate_downtime
3636 # is set)
3637 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3638 $timeout = 60*60; # 1 hour
3639 }
3640
3641 # read banner;
3642 my $data = __read_avail($sock, $timeout);
3643
3644 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3645 die "got unexpected qemu monitor banner\n";
3646 }
3647
3648 my $sel = new IO::Select;
3649 $sel->add($sock);
3650
3651 if (!scalar(my @ready = $sel->can_write($timeout))) {
3652 die "monitor write error - timeout";
3653 }
3654
3655 my $fullcmd = "$cmdstr\r";
3656
3657 # syslog('info', "VM $vmid monitor command: $cmdstr");
3658
3659 my $b;
3660 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3661 die "monitor write error - $!";
3662 }
3663
3664 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3665
3666 $timeout = 20;
3667
3668 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3669 $timeout = 60*60; # 1 hour
3670 } elsif ($cmdstr =~ m/^(eject|change)/) {
3671 $timeout = 60; # note: cdrom mount command is slow
3672 }
3673 if ($res = __read_avail($sock, $timeout)) {
3674
3675 my @lines = split("\r?\n", $res);
3676
3677 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3678
3679 $res = join("\n", @lines);
3680 $res .= "\n";
3681 }
3682 };
3683
3684 my $err = $@;
3685
3686 if ($err) {
3687 syslog("err", "VM $vmid monitor command failed - $err");
3688 die $err;
3689 }
3690
3691 return $res;
3692 }
3693
3694 sub qemu_block_resize {
3695 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3696
3697 my $running = check_running($vmid);
3698
3699 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3700
3701 return if !$running;
3702
3703 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3704
3705 }
3706
3707 sub qemu_volume_snapshot {
3708 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3709
3710 my $running = check_running($vmid);
3711
3712 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3713
3714 return if !$running;
3715
3716 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3717
3718 }
3719
3720 sub qemu_volume_snapshot_delete {
3721 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3722
3723 my $running = check_running($vmid);
3724
3725 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3726
3727 return if !$running;
3728
3729 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3730 }
3731
3732 sub set_migration_caps {
3733 my ($vmid) = @_;
3734
3735 my $cap_ref = [];
3736
3737 my $enabled_cap = {
3738 "auto-converge" => 1,
3739 "xbzrle" => 0,
3740 "x-rdma-pin-all" => 0,
3741 "zero-blocks" => 0,
3742 };
3743
3744 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3745
3746 for my $supported_capability (@$supported_capabilities) {
3747 push @$cap_ref, {
3748 capability => $supported_capability->{capability},
3749 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3750 };
3751 }
3752
3753 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3754 }
3755
3756 my $fast_plug_option = {
3757 'name' => 1,
3758 'onboot' => 1,
3759 'shares' => 1,
3760 'startup' => 1,
3761 };
3762
3763 # hotplug changes in [PENDING]
3764 # $selection hash can be used to only apply specified options, for
3765 # example: { cores => 1 } (only apply changed 'cores')
3766 # $errors ref is used to return error messages
3767 sub vmconfig_hotplug_pending {
3768 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3769
3770 my $defaults = load_defaults();
3771
3772 # commit values which do not have any impact on running VM first
3773 # Note: those option cannot raise errors, we we do not care about
3774 # $selection and always apply them.
3775
3776 my $add_error = sub {
3777 my ($opt, $msg) = @_;
3778 $errors->{$opt} = "hotplug problem - $msg";
3779 };
3780
3781 my $changes = 0;
3782 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3783 if ($fast_plug_option->{$opt}) {
3784 $conf->{$opt} = $conf->{pending}->{$opt};
3785 delete $conf->{pending}->{$opt};
3786 $changes = 1;
3787 }
3788 }
3789
3790 if ($changes) {
3791 update_config_nolock($vmid, $conf, 1);
3792 $conf = load_config($vmid); # update/reload
3793 }
3794
3795 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3796
3797 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3798 foreach my $opt (@delete) {
3799 next if $selection && !$selection->{$opt};
3800 eval {
3801 if ($opt eq 'hotplug') {
3802 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3803 } elsif ($opt eq 'tablet') {
3804 die "skip\n" if !$hotplug_features->{usb};
3805 if ($defaults->{tablet}) {
3806 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3807 } else {
3808 vm_deviceunplug($vmid, $conf, $opt);
3809 }
3810 } elsif ($opt eq 'vcpus') {
3811 die "skip\n" if !$hotplug_features->{cpu};
3812 qemu_cpu_hotplug($vmid, $conf, undef);
3813 } elsif ($opt eq 'balloon') {
3814 # enable balloon device is not hotpluggable
3815 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3816 } elsif ($fast_plug_option->{$opt}) {
3817 # do nothing
3818 } elsif ($opt =~ m/^net(\d+)$/) {
3819 die "skip\n" if !$hotplug_features->{network};
3820 vm_deviceunplug($vmid, $conf, $opt);
3821 } elsif (valid_drivename($opt)) {
3822 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3823 vm_deviceunplug($vmid, $conf, $opt);
3824 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3825 } elsif ($opt =~ m/^memory$/) {
3826 die "skip\n" if !$hotplug_features->{memory};
3827 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3828 } else {
3829 die "skip\n";
3830 }
3831 };
3832 if (my $err = $@) {
3833 &$add_error($opt, $err) if $err ne "skip\n";
3834 } else {
3835 # save new config if hotplug was successful
3836 delete $conf->{$opt};
3837 vmconfig_undelete_pending_option($conf, $opt);
3838 update_config_nolock($vmid, $conf, 1);
3839 $conf = load_config($vmid); # update/reload
3840 }
3841 }
3842
3843 foreach my $opt (keys %{$conf->{pending}}) {
3844 next if $selection && !$selection->{$opt};
3845 my $value = $conf->{pending}->{$opt};
3846 eval {
3847 if ($opt eq 'hotplug') {
3848 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3849 } elsif ($opt eq 'tablet') {
3850 die "skip\n" if !$hotplug_features->{usb};
3851 if ($value == 1) {
3852 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3853 } elsif ($value == 0) {
3854 vm_deviceunplug($vmid, $conf, $opt);
3855 }
3856 } elsif ($opt eq 'vcpus') {
3857 die "skip\n" if !$hotplug_features->{cpu};
3858 qemu_cpu_hotplug($vmid, $conf, $value);
3859 } elsif ($opt eq 'balloon') {
3860 # enable/disable balloning device is not hotpluggable
3861 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3862 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3863 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3864
3865 # allow manual ballooning if shares is set to zero
3866 if (!(defined($conf->{shares}) && ($conf->{shares} == 0))) {
3867 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3868 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3869 }
3870 } elsif ($opt =~ m/^net(\d+)$/) {
3871 # some changes can be done without hotplug
3872 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3873 $vmid, $opt, $value);
3874 } elsif (valid_drivename($opt)) {
3875 # some changes can be done without hotplug
3876 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3877 $vmid, $opt, $value, 1);
3878 } elsif ($opt =~ m/^memory$/) { #dimms
3879 die "skip\n" if !$hotplug_features->{memory};
3880 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3881 } else {
3882 die "skip\n"; # skip non-hot-pluggable options
3883 }
3884 };
3885 if (my $err = $@) {
3886 &$add_error($opt, $err) if $err ne "skip\n";
3887 } else {
3888 # save new config if hotplug was successful
3889 $conf->{$opt} = $value;
3890 delete $conf->{pending}->{$opt};
3891 update_config_nolock($vmid, $conf, 1);
3892 $conf = load_config($vmid); # update/reload
3893 }
3894 }
3895 }
3896
3897 sub vmconfig_apply_pending {
3898 my ($vmid, $conf, $storecfg) = @_;
3899
3900 # cold plug
3901
3902 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3903 foreach my $opt (@delete) { # delete
3904 die "internal error" if $opt =~ m/^unused/;
3905 $conf = load_config($vmid); # update/reload
3906 if (!defined($conf->{$opt})) {
3907 vmconfig_undelete_pending_option($conf, $opt);
3908 update_config_nolock($vmid, $conf, 1);
3909 } elsif (valid_drivename($opt)) {
3910 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3911 vmconfig_undelete_pending_option($conf, $opt);
3912 delete $conf->{$opt};
3913 update_config_nolock($vmid, $conf, 1);
3914 } else {
3915 vmconfig_undelete_pending_option($conf, $opt);
3916 delete $conf->{$opt};
3917 update_config_nolock($vmid, $conf, 1);
3918 }
3919 }
3920
3921 $conf = load_config($vmid); # update/reload
3922
3923 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3924 $conf = load_config($vmid); # update/reload
3925
3926 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3927 # skip if nothing changed
3928 } elsif (valid_drivename($opt)) {
3929 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3930 if defined($conf->{$opt});
3931 $conf->{$opt} = $conf->{pending}->{$opt};
3932 } else {
3933 $conf->{$opt} = $conf->{pending}->{$opt};
3934 }
3935
3936 delete $conf->{pending}->{$opt};
3937 update_config_nolock($vmid, $conf, 1);
3938 }
3939 }
3940
3941 my $safe_num_ne = sub {
3942 my ($a, $b) = @_;
3943
3944 return 0 if !defined($a) && !defined($b);
3945 return 1 if !defined($a);
3946 return 1 if !defined($b);
3947
3948 return $a != $b;
3949 };
3950
3951 my $safe_string_ne = sub {
3952 my ($a, $b) = @_;
3953
3954 return 0 if !defined($a) && !defined($b);
3955 return 1 if !defined($a);
3956 return 1 if !defined($b);
3957
3958 return $a ne $b;
3959 };
3960
3961 sub vmconfig_update_net {
3962 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3963
3964 my $newnet = parse_net($value);
3965
3966 if ($conf->{$opt}) {
3967 my $oldnet = parse_net($conf->{$opt});
3968
3969 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3970 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3971 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3972 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3973
3974 # for non online change, we try to hot-unplug
3975 die "skip\n" if !$hotplug;
3976 vm_deviceunplug($vmid, $conf, $opt);
3977 } else {
3978
3979 die "internal error" if $opt !~ m/net(\d+)/;
3980 my $iface = "tap${vmid}i$1";
3981
3982 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3983 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3984 }
3985
3986 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3987 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3988 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3989 PVE::Network::tap_unplug($iface);
3990 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3991 }
3992
3993 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3994 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3995 }
3996
3997 return 1;
3998 }
3999 }
4000
4001 if ($hotplug) {
4002 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4003 } else {
4004 die "skip\n";
4005 }
4006 }
4007
4008 sub vmconfig_update_disk {
4009 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4010
4011 # fixme: do we need force?
4012
4013 my $drive = parse_drive($opt, $value);
4014
4015 if ($conf->{$opt}) {
4016
4017 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4018
4019 my $media = $drive->{media} || 'disk';
4020 my $oldmedia = $old_drive->{media} || 'disk';
4021 die "unable to change media type\n" if $media ne $oldmedia;
4022
4023 if (!drive_is_cdrom($old_drive)) {
4024
4025 if ($drive->{file} ne $old_drive->{file}) {
4026
4027 die "skip\n" if !$hotplug;
4028
4029 # unplug and register as unused
4030 vm_deviceunplug($vmid, $conf, $opt);
4031 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4032
4033 } else {
4034 # update existing disk
4035
4036 # skip non hotpluggable value
4037 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4038 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4039 die "skip\n";
4040 }
4041
4042 # apply throttle
4043 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4044 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4045 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4046 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4047 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4048 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4049 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4050 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4051 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4052 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4053 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4054 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4055
4056 qemu_block_set_io_throttle($vmid,"drive-$opt",
4057 ($drive->{mbps} || 0)*1024*1024,
4058 ($drive->{mbps_rd} || 0)*1024*1024,
4059 ($drive->{mbps_wr} || 0)*1024*1024,
4060 $drive->{iops} || 0,
4061 $drive->{iops_rd} || 0,
4062 $drive->{iops_wr} || 0,
4063 ($drive->{mbps_max} || 0)*1024*1024,
4064 ($drive->{mbps_rd_max} || 0)*1024*1024,
4065 ($drive->{mbps_wr_max} || 0)*1024*1024,
4066 $drive->{iops_max} || 0,
4067 $drive->{iops_rd_max} || 0,
4068 $drive->{iops_wr_max} || 0);
4069
4070 }
4071
4072 return 1;
4073 }
4074
4075 } else { # cdrom
4076
4077 if ($drive->{file} eq 'none') {
4078 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4079 } else {
4080 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4081 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4082 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4083 }
4084 }
4085 }
4086 }
4087
4088 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4089 # hotplug new disks
4090 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4091 }
4092
4093 sub vm_start {
4094 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4095
4096 lock_config($vmid, sub {
4097 my $conf = load_config($vmid, $migratedfrom);
4098
4099 die "you can't start a vm if it's a template\n" if is_template($conf);
4100
4101 check_lock($conf) if !$skiplock;
4102
4103 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4104
4105 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4106 vmconfig_apply_pending($vmid, $conf, $storecfg);
4107 $conf = load_config($vmid); # update/reload
4108 }
4109
4110 my $defaults = load_defaults();
4111
4112 # set environment variable useful inside network script
4113 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4114
4115 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4116
4117 my $migrate_port = 0;
4118 my $migrate_uri;
4119 if ($statefile) {
4120 if ($statefile eq 'tcp') {
4121 my $localip = "localhost";
4122 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4123 if ($datacenterconf->{migration_unsecure}) {
4124 my $nodename = PVE::INotify::nodename();
4125 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4126 }
4127 $migrate_port = PVE::Tools::next_migrate_port();
4128 $migrate_uri = "tcp:${localip}:${migrate_port}";
4129 push @$cmd, '-incoming', $migrate_uri;
4130 push @$cmd, '-S';
4131 } else {
4132 push @$cmd, '-loadstate', $statefile;
4133 }
4134 } elsif ($paused) {
4135 push @$cmd, '-S';
4136 }
4137
4138 # host pci devices
4139 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4140 my $d = parse_hostpci($conf->{"hostpci$i"});
4141 next if !$d;
4142 my $pcidevices = $d->{pciid};
4143 foreach my $pcidevice (@$pcidevices) {
4144 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4145
4146 my $info = pci_device_info("0000:$pciid");
4147 die "IOMMU not present\n" if !check_iommu_support();
4148 die "no pci device info for device '$pciid'\n" if !$info;
4149
4150 if ($d->{driver} && $d->{driver} eq "vfio") {
4151 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4152 } else {
4153 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4154 }
4155
4156 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4157 }
4158 }
4159
4160 PVE::Storage::activate_volumes($storecfg, $vollist);
4161
4162 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4163 umask => 0077); };
4164 my $err = $@;
4165 die "start failed: $err" if $err;
4166
4167 print "migration listens on $migrate_uri\n" if $migrate_uri;
4168
4169 if ($statefile && $statefile ne 'tcp') {
4170 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4171 warn $@ if $@;
4172 }
4173
4174 if ($migratedfrom) {
4175
4176 eval {
4177 set_migration_caps($vmid);
4178 };
4179 warn $@ if $@;
4180
4181 if ($spice_port) {
4182 print "spice listens on port $spice_port\n";
4183 if ($spice_ticket) {
4184 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4185 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4186 }
4187 }
4188
4189 } else {
4190
4191 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4192 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4193 if $conf->{balloon};
4194 vm_mon_cmd_nocheck($vmid, 'qom-set',
4195 path => "machine/peripheral/balloon0",
4196 property => "guest-stats-polling-interval",
4197 value => 2);
4198 }
4199
4200 foreach my $opt (keys %$conf) {
4201 next if $opt !~ m/^net\d+$/;
4202 my $nicconf = parse_net($conf->{$opt});
4203 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4204 }
4205 }
4206 });
4207 }
4208
4209 sub vm_mon_cmd {
4210 my ($vmid, $execute, %params) = @_;
4211
4212 my $cmd = { execute => $execute, arguments => \%params };
4213 vm_qmp_command($vmid, $cmd);
4214 }
4215
4216 sub vm_mon_cmd_nocheck {
4217 my ($vmid, $execute, %params) = @_;
4218
4219 my $cmd = { execute => $execute, arguments => \%params };
4220 vm_qmp_command($vmid, $cmd, 1);
4221 }
4222
4223 sub vm_qmp_command {
4224 my ($vmid, $cmd, $nocheck) = @_;
4225
4226 my $res;
4227
4228 my $timeout;
4229 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4230 $timeout = $cmd->{arguments}->{timeout};
4231 delete $cmd->{arguments}->{timeout};
4232 }
4233
4234 eval {
4235 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4236 my $sname = qmp_socket($vmid);
4237 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4238 my $qmpclient = PVE::QMPClient->new();
4239
4240 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4241 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4242 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4243 if scalar(%{$cmd->{arguments}});
4244 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4245 } else {
4246 die "unable to open monitor socket\n";
4247 }
4248 };
4249 if (my $err = $@) {
4250 syslog("err", "VM $vmid qmp command failed - $err");
4251 die $err;
4252 }
4253
4254 return $res;
4255 }
4256
4257 sub vm_human_monitor_command {
4258 my ($vmid, $cmdline) = @_;
4259
4260 my $res;
4261
4262 my $cmd = {
4263 execute => 'human-monitor-command',
4264 arguments => { 'command-line' => $cmdline},
4265 };
4266
4267 return vm_qmp_command($vmid, $cmd);
4268 }
4269
4270 sub vm_commandline {
4271 my ($storecfg, $vmid) = @_;
4272
4273 my $conf = load_config($vmid);
4274
4275 my $defaults = load_defaults();
4276
4277 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4278
4279 return join(' ', @$cmd);
4280 }
4281
4282 sub vm_reset {
4283 my ($vmid, $skiplock) = @_;
4284
4285 lock_config($vmid, sub {
4286
4287 my $conf = load_config($vmid);
4288
4289 check_lock($conf) if !$skiplock;
4290
4291 vm_mon_cmd($vmid, "system_reset");
4292 });
4293 }
4294
4295 sub get_vm_volumes {
4296 my ($conf) = @_;
4297
4298 my $vollist = [];
4299 foreach_volid($conf, sub {
4300 my ($volid, $is_cdrom) = @_;
4301
4302 return if $volid =~ m|^/|;
4303
4304 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4305 return if !$sid;
4306
4307 push @$vollist, $volid;
4308 });
4309
4310 return $vollist;
4311 }
4312
4313 sub vm_stop_cleanup {
4314 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4315
4316 eval {
4317 fairsched_rmnod($vmid); # try to destroy group
4318
4319 if (!$keepActive) {
4320 my $vollist = get_vm_volumes($conf);
4321 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4322 }
4323
4324 foreach my $ext (qw(mon qmp pid vnc qga)) {
4325 unlink "/var/run/qemu-server/${vmid}.$ext";
4326 }
4327
4328 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4329 };
4330 warn $@ if $@; # avoid errors - just warn
4331 }
4332
4333 # Note: use $nockeck to skip tests if VM configuration file exists.
4334 # We need that when migration VMs to other nodes (files already moved)
4335 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4336 sub vm_stop {
4337 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4338
4339 $force = 1 if !defined($force) && !$shutdown;
4340
4341 if ($migratedfrom){
4342 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4343 kill 15, $pid if $pid;
4344 my $conf = load_config($vmid, $migratedfrom);
4345 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4346 return;
4347 }
4348
4349 lock_config($vmid, sub {
4350
4351 my $pid = check_running($vmid, $nocheck);
4352 return if !$pid;
4353
4354 my $conf;
4355 if (!$nocheck) {
4356 $conf = load_config($vmid);
4357 check_lock($conf) if !$skiplock;
4358 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4359 my $opts = parse_startup($conf->{startup});
4360 $timeout = $opts->{down} if $opts->{down};
4361 }
4362 }
4363
4364 $timeout = 60 if !defined($timeout);
4365
4366 eval {
4367 if ($shutdown) {
4368 if (defined($conf) && $conf->{agent}) {
4369 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4370 } else {
4371 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4372 }
4373 } else {
4374 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4375 }
4376 };
4377 my $err = $@;
4378
4379 if (!$err) {
4380 my $count = 0;
4381 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4382 $count++;
4383 sleep 1;
4384 }
4385
4386 if ($count >= $timeout) {
4387 if ($force) {
4388 warn "VM still running - terminating now with SIGTERM\n";
4389 kill 15, $pid;
4390 } else {
4391 die "VM quit/powerdown failed - got timeout\n";
4392 }
4393 } else {
4394 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4395 return;
4396 }
4397 } else {
4398 if ($force) {
4399 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4400 kill 15, $pid;
4401 } else {
4402 die "VM quit/powerdown failed\n";
4403 }
4404 }
4405
4406 # wait again
4407 $timeout = 10;
4408
4409 my $count = 0;
4410 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4411 $count++;
4412 sleep 1;
4413 }
4414
4415 if ($count >= $timeout) {
4416 warn "VM still running - terminating now with SIGKILL\n";
4417 kill 9, $pid;
4418 sleep 1;
4419 }
4420
4421 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4422 });
4423 }
4424
4425 sub vm_suspend {
4426 my ($vmid, $skiplock) = @_;
4427
4428 lock_config($vmid, sub {
4429
4430 my $conf = load_config($vmid);
4431
4432 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4433
4434 vm_mon_cmd($vmid, "stop");
4435 });
4436 }
4437
4438 sub vm_resume {
4439 my ($vmid, $skiplock) = @_;
4440
4441 lock_config($vmid, sub {
4442
4443 my $conf = load_config($vmid);
4444
4445 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4446
4447 vm_mon_cmd($vmid, "cont");
4448 });
4449 }
4450
4451 sub vm_sendkey {
4452 my ($vmid, $skiplock, $key) = @_;
4453
4454 lock_config($vmid, sub {
4455
4456 my $conf = load_config($vmid);
4457
4458 # there is no qmp command, so we use the human monitor command
4459 vm_human_monitor_command($vmid, "sendkey $key");
4460 });
4461 }
4462
4463 sub vm_destroy {
4464 my ($storecfg, $vmid, $skiplock) = @_;
4465
4466 lock_config($vmid, sub {
4467
4468 my $conf = load_config($vmid);
4469
4470 check_lock($conf) if !$skiplock;
4471
4472 if (!check_running($vmid)) {
4473 fairsched_rmnod($vmid); # try to destroy group
4474 destroy_vm($storecfg, $vmid);
4475 } else {
4476 die "VM $vmid is running - destroy failed\n";
4477 }
4478 });
4479 }
4480
4481 # pci helpers
4482
4483 sub file_write {
4484 my ($filename, $buf) = @_;
4485
4486 my $fh = IO::File->new($filename, "w");
4487 return undef if !$fh;
4488
4489 my $res = print $fh $buf;
4490
4491 $fh->close();
4492
4493 return $res;
4494 }
4495
4496 sub pci_device_info {
4497 my ($name) = @_;
4498
4499 my $res;
4500
4501 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4502 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4503
4504 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4505 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4506
4507 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4508 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4509
4510 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4511 return undef if !defined($product) || $product !~ s/^0x//;
4512
4513 $res = {
4514 name => $name,
4515 vendor => $vendor,
4516 product => $product,
4517 domain => $domain,
4518 bus => $bus,
4519 slot => $slot,
4520 func => $func,
4521 irq => $irq,
4522 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4523 };
4524
4525 return $res;
4526 }
4527
4528 sub pci_dev_reset {
4529 my ($dev) = @_;
4530
4531 my $name = $dev->{name};
4532
4533 my $fn = "$pcisysfs/devices/$name/reset";
4534
4535 return file_write($fn, "1");
4536 }
4537
4538 sub pci_dev_bind_to_stub {
4539 my ($dev) = @_;
4540
4541 my $name = $dev->{name};
4542
4543 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4544 return 1 if -d $testdir;
4545
4546 my $data = "$dev->{vendor} $dev->{product}";
4547 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4548
4549 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4550 if (!file_write($fn, $name)) {
4551 return undef if -f $fn;
4552 }
4553
4554 $fn = "$pcisysfs/drivers/pci-stub/bind";
4555 if (! -d $testdir) {
4556 return undef if !file_write($fn, $name);
4557 }
4558
4559 return -d $testdir;
4560 }
4561
4562 sub pci_dev_bind_to_vfio {
4563 my ($dev) = @_;
4564
4565 my $name = $dev->{name};
4566
4567 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4568
4569 if (!-d $vfio_basedir) {
4570 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4571 }
4572 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4573
4574 my $testdir = "$vfio_basedir/$name";
4575 return 1 if -d $testdir;
4576
4577 my $data = "$dev->{vendor} $dev->{product}";
4578 return undef if !file_write("$vfio_basedir/new_id", $data);
4579
4580 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4581 if (!file_write($fn, $name)) {
4582 return undef if -f $fn;
4583 }
4584
4585 $fn = "$vfio_basedir/bind";
4586 if (! -d $testdir) {
4587 return undef if !file_write($fn, $name);
4588 }
4589
4590 return -d $testdir;
4591 }
4592
4593 sub pci_dev_group_bind_to_vfio {
4594 my ($pciid) = @_;
4595
4596 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4597
4598 if (!-d $vfio_basedir) {
4599 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4600 }
4601 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4602
4603 # get IOMMU group devices
4604 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4605 my @devs = grep /^0000:/, readdir($D);
4606 closedir($D);
4607
4608 foreach my $pciid (@devs) {
4609 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4610
4611 # pci bridges, switches or root ports are not supported
4612 # they have a pci_bus subdirectory so skip them
4613 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4614
4615 my $info = pci_device_info($1);
4616 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4617 }
4618
4619 return 1;
4620 }
4621
4622 sub print_pci_addr {
4623 my ($id, $bridges) = @_;
4624
4625 my $res = '';
4626 my $devices = {
4627 piix3 => { bus => 0, addr => 1 },
4628 #addr2 : first videocard
4629 balloon0 => { bus => 0, addr => 3 },
4630 watchdog => { bus => 0, addr => 4 },
4631 scsihw0 => { bus => 0, addr => 5 },
4632 scsihw1 => { bus => 0, addr => 6 },
4633 ahci0 => { bus => 0, addr => 7 },
4634 qga0 => { bus => 0, addr => 8 },
4635 spice => { bus => 0, addr => 9 },
4636 virtio0 => { bus => 0, addr => 10 },
4637 virtio1 => { bus => 0, addr => 11 },
4638 virtio2 => { bus => 0, addr => 12 },
4639 virtio3 => { bus => 0, addr => 13 },
4640 virtio4 => { bus => 0, addr => 14 },
4641 virtio5 => { bus => 0, addr => 15 },
4642 hostpci0 => { bus => 0, addr => 16 },
4643 hostpci1 => { bus => 0, addr => 17 },
4644 net0 => { bus => 0, addr => 18 },
4645 net1 => { bus => 0, addr => 19 },
4646 net2 => { bus => 0, addr => 20 },
4647 net3 => { bus => 0, addr => 21 },
4648 net4 => { bus => 0, addr => 22 },
4649 net5 => { bus => 0, addr => 23 },
4650 vga1 => { bus => 0, addr => 24 },
4651 vga2 => { bus => 0, addr => 25 },
4652 vga3 => { bus => 0, addr => 26 },
4653 hostpci2 => { bus => 0, addr => 27 },
4654 hostpci3 => { bus => 0, addr => 28 },
4655 #addr29 : usb-host (pve-usb.cfg)
4656 'pci.1' => { bus => 0, addr => 30 },
4657 'pci.2' => { bus => 0, addr => 31 },
4658 'net6' => { bus => 1, addr => 1 },
4659 'net7' => { bus => 1, addr => 2 },
4660 'net8' => { bus => 1, addr => 3 },
4661 'net9' => { bus => 1, addr => 4 },
4662 'net10' => { bus => 1, addr => 5 },
4663 'net11' => { bus => 1, addr => 6 },
4664 'net12' => { bus => 1, addr => 7 },
4665 'net13' => { bus => 1, addr => 8 },
4666 'net14' => { bus => 1, addr => 9 },
4667 'net15' => { bus => 1, addr => 10 },
4668 'net16' => { bus => 1, addr => 11 },
4669 'net17' => { bus => 1, addr => 12 },
4670 'net18' => { bus => 1, addr => 13 },
4671 'net19' => { bus => 1, addr => 14 },
4672 'net20' => { bus => 1, addr => 15 },
4673 'net21' => { bus => 1, addr => 16 },
4674 'net22' => { bus => 1, addr => 17 },
4675 'net23' => { bus => 1, addr => 18 },
4676 'net24' => { bus => 1, addr => 19 },
4677 'net25' => { bus => 1, addr => 20 },
4678 'net26' => { bus => 1, addr => 21 },
4679 'net27' => { bus => 1, addr => 22 },
4680 'net28' => { bus => 1, addr => 23 },
4681 'net29' => { bus => 1, addr => 24 },
4682 'net30' => { bus => 1, addr => 25 },
4683 'net31' => { bus => 1, addr => 26 },
4684 'virtio6' => { bus => 2, addr => 1 },
4685 'virtio7' => { bus => 2, addr => 2 },
4686 'virtio8' => { bus => 2, addr => 3 },
4687 'virtio9' => { bus => 2, addr => 4 },
4688 'virtio10' => { bus => 2, addr => 5 },
4689 'virtio11' => { bus => 2, addr => 6 },
4690 'virtio12' => { bus => 2, addr => 7 },
4691 'virtio13' => { bus => 2, addr => 8 },
4692 'virtio14' => { bus => 2, addr => 9 },
4693 'virtio15' => { bus => 2, addr => 10 },
4694 };
4695
4696 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4697 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4698 my $bus = $devices->{$id}->{bus};
4699 $res = ",bus=pci.$bus,addr=$addr";
4700 $bridges->{$bus} = 1 if $bridges;
4701 }
4702 return $res;
4703
4704 }
4705
4706 sub print_pcie_addr {
4707 my ($id) = @_;
4708
4709 my $res = '';
4710 my $devices = {
4711 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4712 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4713 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4714 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4715 };
4716
4717 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4718 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4719 my $bus = $devices->{$id}->{bus};
4720 $res = ",bus=$bus,addr=$addr";
4721 }
4722 return $res;
4723
4724 }
4725
4726 # vzdump restore implementaion
4727
4728 sub tar_archive_read_firstfile {
4729 my $archive = shift;
4730
4731 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4732
4733 # try to detect archive type first
4734 my $pid = open (TMP, "tar tf '$archive'|") ||
4735 die "unable to open file '$archive'\n";
4736 my $firstfile = <TMP>;
4737 kill 15, $pid;
4738 close TMP;
4739
4740 die "ERROR: archive contaions no data\n" if !$firstfile;
4741 chomp $firstfile;
4742
4743 return $firstfile;
4744 }
4745
4746 sub tar_restore_cleanup {
4747 my ($storecfg, $statfile) = @_;
4748
4749 print STDERR "starting cleanup\n";
4750
4751 if (my $fd = IO::File->new($statfile, "r")) {
4752 while (defined(my $line = <$fd>)) {
4753 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4754 my $volid = $2;
4755 eval {
4756 if ($volid =~ m|^/|) {
4757 unlink $volid || die 'unlink failed\n';
4758 } else {
4759 PVE::Storage::vdisk_free($storecfg, $volid);
4760 }
4761 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4762 };
4763 print STDERR "unable to cleanup '$volid' - $@" if $@;
4764 } else {
4765 print STDERR "unable to parse line in statfile - $line";
4766 }
4767 }
4768 $fd->close();
4769 }
4770 }
4771
4772 sub restore_archive {
4773 my ($archive, $vmid, $user, $opts) = @_;
4774
4775 my $format = $opts->{format};
4776 my $comp;
4777
4778 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4779 $format = 'tar' if !$format;
4780 $comp = 'gzip';
4781 } elsif ($archive =~ m/\.tar$/) {
4782 $format = 'tar' if !$format;
4783 } elsif ($archive =~ m/.tar.lzo$/) {
4784 $format = 'tar' if !$format;
4785 $comp = 'lzop';
4786 } elsif ($archive =~ m/\.vma$/) {
4787 $format = 'vma' if !$format;
4788 } elsif ($archive =~ m/\.vma\.gz$/) {
4789 $format = 'vma' if !$format;
4790 $comp = 'gzip';
4791 } elsif ($archive =~ m/\.vma\.lzo$/) {
4792 $format = 'vma' if !$format;
4793 $comp = 'lzop';
4794 } else {
4795 $format = 'vma' if !$format; # default
4796 }
4797
4798 # try to detect archive format
4799 if ($format eq 'tar') {
4800 return restore_tar_archive($archive, $vmid, $user, $opts);
4801 } else {
4802 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4803 }
4804 }
4805
4806 sub restore_update_config_line {
4807 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4808
4809 return if $line =~ m/^\#qmdump\#/;
4810 return if $line =~ m/^\#vzdump\#/;
4811 return if $line =~ m/^lock:/;
4812 return if $line =~ m/^unused\d+:/;
4813 return if $line =~ m/^parent:/;
4814 return if $line =~ m/^template:/; # restored VM is never a template
4815
4816 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4817 # try to convert old 1.X settings
4818 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4819 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4820 my ($model, $macaddr) = split(/\=/, $devconfig);
4821 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4822 my $net = {
4823 model => $model,
4824 bridge => "vmbr$ind",
4825 macaddr => $macaddr,
4826 };
4827 my $netstr = print_net($net);
4828
4829 print $outfd "net$cookie->{netcount}: $netstr\n";
4830 $cookie->{netcount}++;
4831 }
4832 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4833 my ($id, $netstr) = ($1, $2);
4834 my $net = parse_net($netstr);
4835 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4836 $netstr = print_net($net);
4837 print $outfd "$id: $netstr\n";
4838 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4839 my $virtdev = $1;
4840 my $value = $3;
4841 if ($line =~ m/backup=no/) {
4842 print $outfd "#$line";
4843 } elsif ($virtdev && $map->{$virtdev}) {
4844 my $di = parse_drive($virtdev, $value);
4845 delete $di->{format}; # format can change on restore
4846 $di->{file} = $map->{$virtdev};
4847 $value = print_drive($vmid, $di);
4848 print $outfd "$virtdev: $value\n";
4849 } else {
4850 print $outfd $line;
4851 }
4852 } else {
4853 print $outfd $line;
4854 }
4855 }
4856
4857 sub scan_volids {
4858 my ($cfg, $vmid) = @_;
4859
4860 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4861
4862 my $volid_hash = {};
4863 foreach my $storeid (keys %$info) {
4864 foreach my $item (@{$info->{$storeid}}) {
4865 next if !($item->{volid} && $item->{size});
4866 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4867 $volid_hash->{$item->{volid}} = $item;
4868 }
4869 }
4870
4871 return $volid_hash;
4872 }
4873
4874 sub get_used_paths {
4875 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4876
4877 my $used_path = {};
4878
4879 my $scan_config = sub {
4880 my ($cref, $snapname) = @_;
4881
4882 foreach my $key (keys %$cref) {
4883 my $value = $cref->{$key};
4884 if (valid_drivename($key)) {
4885 next if $skip_drive && $key eq $skip_drive;
4886 my $drive = parse_drive($key, $value);
4887 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4888 if ($drive->{file} =~ m!^/!) {
4889 $used_path->{$drive->{file}}++; # = 1;
4890 } else {
4891 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4892 next if !$storeid;
4893 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4894 next if !$scfg;
4895 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4896 $used_path->{$path}++; # = 1;
4897 }
4898 }
4899 }
4900 };
4901
4902 &$scan_config($conf);
4903
4904 undef $skip_drive;
4905
4906 if ($scan_snapshots) {
4907 foreach my $snapname (keys %{$conf->{snapshots}}) {
4908 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4909 }
4910 }
4911
4912 return $used_path;
4913 }
4914
4915 sub update_disksize {
4916 my ($vmid, $conf, $volid_hash) = @_;
4917
4918 my $changes;
4919
4920 my $used = {};
4921
4922 # Note: it is allowed to define multiple storages with same path (alias), so
4923 # we need to check both 'volid' and real 'path' (two different volid can point
4924 # to the same path).
4925
4926 my $usedpath = {};
4927
4928 # update size info
4929 foreach my $opt (keys %$conf) {
4930 if (valid_drivename($opt)) {
4931 my $drive = parse_drive($opt, $conf->{$opt});
4932 my $volid = $drive->{file};
4933 next if !$volid;
4934
4935 $used->{$volid} = 1;
4936 if ($volid_hash->{$volid} &&
4937 (my $path = $volid_hash->{$volid}->{path})) {
4938 $usedpath->{$path} = 1;
4939 }
4940
4941 next if drive_is_cdrom($drive);
4942 next if !$volid_hash->{$volid};
4943
4944 $drive->{size} = $volid_hash->{$volid}->{size};
4945 my $new = print_drive($vmid, $drive);
4946 if ($new ne $conf->{$opt}) {
4947 $changes = 1;
4948 $conf->{$opt} = $new;
4949 }
4950 }
4951 }
4952
4953 # remove 'unusedX' entry if volume is used
4954 foreach my $opt (keys %$conf) {
4955 next if $opt !~ m/^unused\d+$/;
4956 my $volid = $conf->{$opt};
4957 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4958 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4959 $changes = 1;
4960 delete $conf->{$opt};
4961 }
4962 }
4963
4964 foreach my $volid (sort keys %$volid_hash) {
4965 next if $volid =~ m/vm-$vmid-state-/;
4966 next if $used->{$volid};
4967 my $path = $volid_hash->{$volid}->{path};
4968 next if !$path; # just to be sure
4969 next if $usedpath->{$path};
4970 $changes = 1;
4971 add_unused_volume($conf, $volid);
4972 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4973 }
4974
4975 return $changes;
4976 }
4977
4978 sub rescan {
4979 my ($vmid, $nolock) = @_;
4980
4981 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4982
4983 my $volid_hash = scan_volids($cfg, $vmid);
4984
4985 my $updatefn = sub {
4986 my ($vmid) = @_;
4987
4988 my $conf = load_config($vmid);
4989
4990 check_lock($conf);
4991
4992 my $vm_volids = {};
4993 foreach my $volid (keys %$volid_hash) {
4994 my $info = $volid_hash->{$volid};
4995 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
4996 }
4997
4998 my $changes = update_disksize($vmid, $conf, $vm_volids);
4999
5000 update_config_nolock($vmid, $conf, 1) if $changes;
5001 };
5002
5003 if (defined($vmid)) {
5004 if ($nolock) {
5005 &$updatefn($vmid);
5006 } else {
5007 lock_config($vmid, $updatefn, $vmid);
5008 }
5009 } else {
5010 my $vmlist = config_list();
5011 foreach my $vmid (keys %$vmlist) {
5012 if ($nolock) {
5013 &$updatefn($vmid);
5014 } else {
5015 lock_config($vmid, $updatefn, $vmid);
5016 }
5017 }
5018 }
5019 }
5020
5021 sub restore_vma_archive {
5022 my ($archive, $vmid, $user, $opts, $comp) = @_;
5023
5024 my $input = $archive eq '-' ? "<&STDIN" : undef;
5025 my $readfrom = $archive;
5026
5027 my $uncomp = '';
5028 if ($comp) {
5029 $readfrom = '-';
5030 my $qarchive = PVE::Tools::shellquote($archive);
5031 if ($comp eq 'gzip') {
5032 $uncomp = "zcat $qarchive|";
5033 } elsif ($comp eq 'lzop') {
5034 $uncomp = "lzop -d -c $qarchive|";
5035 } else {
5036 die "unknown compression method '$comp'\n";
5037 }
5038
5039 }
5040
5041 my $tmpdir = "/var/tmp/vzdumptmp$$";
5042 rmtree $tmpdir;
5043
5044 # disable interrupts (always do cleanups)
5045 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5046 warn "got interrupt - ignored\n";
5047 };
5048
5049 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5050 POSIX::mkfifo($mapfifo, 0600);
5051 my $fifofh;
5052
5053 my $openfifo = sub {
5054 open($fifofh, '>', $mapfifo) || die $!;
5055 };
5056
5057 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5058
5059 my $oldtimeout;
5060 my $timeout = 5;
5061
5062 my $devinfo = {};
5063
5064 my $rpcenv = PVE::RPCEnvironment::get();
5065
5066 my $conffile = config_file($vmid);
5067 my $tmpfn = "$conffile.$$.tmp";
5068
5069 # Note: $oldconf is undef if VM does not exists
5070 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5071
5072 my $print_devmap = sub {
5073 my $virtdev_hash = {};
5074
5075 my $cfgfn = "$tmpdir/qemu-server.conf";
5076
5077 # we can read the config - that is already extracted
5078 my $fh = IO::File->new($cfgfn, "r") ||
5079 "unable to read qemu-server.conf - $!\n";
5080
5081 while (defined(my $line = <$fh>)) {
5082 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5083 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5084 die "archive does not contain data for drive '$virtdev'\n"
5085 if !$devinfo->{$devname};
5086 if (defined($opts->{storage})) {
5087 $storeid = $opts->{storage} || 'local';
5088 } elsif (!$storeid) {
5089 $storeid = 'local';
5090 }
5091 $format = 'raw' if !$format;
5092 $devinfo->{$devname}->{devname} = $devname;
5093 $devinfo->{$devname}->{virtdev} = $virtdev;
5094 $devinfo->{$devname}->{format} = $format;
5095 $devinfo->{$devname}->{storeid} = $storeid;
5096
5097 # check permission on storage
5098 my $pool = $opts->{pool}; # todo: do we need that?
5099 if ($user ne 'root@pam') {
5100 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5101 }
5102
5103 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5104 }
5105 }
5106
5107 foreach my $devname (keys %$devinfo) {
5108 die "found no device mapping information for device '$devname'\n"
5109 if !$devinfo->{$devname}->{virtdev};
5110 }
5111
5112 my $cfg = cfs_read_file('storage.cfg');
5113
5114 # create empty/temp config
5115 if ($oldconf) {
5116 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5117 foreach_drive($oldconf, sub {
5118 my ($ds, $drive) = @_;
5119
5120 return if drive_is_cdrom($drive);
5121
5122 my $volid = $drive->{file};
5123
5124 return if !$volid || $volid =~ m|^/|;
5125
5126 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5127 return if !$path || !$owner || ($owner != $vmid);
5128
5129 # Note: only delete disk we want to restore
5130 # other volumes will become unused
5131 if ($virtdev_hash->{$ds}) {
5132 PVE::Storage::vdisk_free($cfg, $volid);
5133 }
5134 });
5135 }
5136
5137 my $map = {};
5138 foreach my $virtdev (sort keys %$virtdev_hash) {
5139 my $d = $virtdev_hash->{$virtdev};
5140 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5141 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5142
5143 # test if requested format is supported
5144 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5145 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5146 $d->{format} = $defFormat if !$supported;
5147
5148 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5149 $d->{format}, undef, $alloc_size);
5150 print STDERR "new volume ID is '$volid'\n";
5151 $d->{volid} = $volid;
5152 my $path = PVE::Storage::path($cfg, $volid);
5153
5154 my $write_zeros = 1;
5155 # fixme: what other storages types initialize volumes with zero?
5156 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5157 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5158 $write_zeros = 0;
5159 }
5160
5161 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5162
5163 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5164 $map->{$virtdev} = $volid;
5165 }
5166
5167 $fh->seek(0, 0) || die "seek failed - $!\n";
5168
5169 my $outfd = new IO::File ($tmpfn, "w") ||
5170 die "unable to write config for VM $vmid\n";
5171
5172 my $cookie = { netcount => 0 };
5173 while (defined(my $line = <$fh>)) {
5174 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5175 }
5176
5177 $fh->close();
5178 $outfd->close();
5179 };
5180
5181 eval {
5182 # enable interrupts
5183 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5184 die "interrupted by signal\n";
5185 };
5186 local $SIG{ALRM} = sub { die "got timeout\n"; };
5187
5188 $oldtimeout = alarm($timeout);
5189
5190 my $parser = sub {
5191 my $line = shift;
5192
5193 print "$line\n";
5194
5195 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5196 my ($dev_id, $size, $devname) = ($1, $2, $3);
5197 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5198 } elsif ($line =~ m/^CTIME: /) {
5199 # we correctly received the vma config, so we can disable
5200 # the timeout now for disk allocation (set to 10 minutes, so
5201 # that we always timeout if something goes wrong)
5202 alarm(600);
5203 &$print_devmap();
5204 print $fifofh "done\n";
5205 my $tmp = $oldtimeout || 0;
5206 $oldtimeout = undef;
5207 alarm($tmp);
5208 close($fifofh);
5209 }
5210 };
5211
5212 print "restore vma archive: $cmd\n";
5213 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5214 };
5215 my $err = $@;
5216
5217 alarm($oldtimeout) if $oldtimeout;
5218
5219 unlink $mapfifo;
5220
5221 if ($err) {
5222 rmtree $tmpdir;
5223 unlink $tmpfn;
5224
5225 my $cfg = cfs_read_file('storage.cfg');
5226 foreach my $devname (keys %$devinfo) {
5227 my $volid = $devinfo->{$devname}->{volid};
5228 next if !$volid;
5229 eval {
5230 if ($volid =~ m|^/|) {
5231 unlink $volid || die 'unlink failed\n';
5232 } else {
5233 PVE::Storage::vdisk_free($cfg, $volid);
5234 }
5235 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5236 };
5237 print STDERR "unable to cleanup '$volid' - $@" if $@;
5238 }
5239 die $err;
5240 }
5241
5242 rmtree $tmpdir;
5243
5244 rename($tmpfn, $conffile) ||
5245 die "unable to commit configuration file '$conffile'\n";
5246
5247 PVE::Cluster::cfs_update(); # make sure we read new file
5248
5249 eval { rescan($vmid, 1); };
5250 warn $@ if $@;
5251 }
5252
5253 sub restore_tar_archive {
5254 my ($archive, $vmid, $user, $opts) = @_;
5255
5256 if ($archive ne '-') {
5257 my $firstfile = tar_archive_read_firstfile($archive);
5258 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5259 if $firstfile ne 'qemu-server.conf';
5260 }
5261
5262 my $storecfg = cfs_read_file('storage.cfg');
5263
5264 # destroy existing data - keep empty config
5265 my $vmcfgfn = config_file($vmid);
5266 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5267
5268 my $tocmd = "/usr/lib/qemu-server/qmextract";
5269
5270 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5271 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5272 $tocmd .= ' --prealloc' if $opts->{prealloc};
5273 $tocmd .= ' --info' if $opts->{info};
5274
5275 # tar option "xf" does not autodetect compression when read from STDIN,
5276 # so we pipe to zcat
5277 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5278 PVE::Tools::shellquote("--to-command=$tocmd");
5279
5280 my $tmpdir = "/var/tmp/vzdumptmp$$";
5281 mkpath $tmpdir;
5282
5283 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5284 local $ENV{VZDUMP_VMID} = $vmid;
5285 local $ENV{VZDUMP_USER} = $user;
5286
5287 my $conffile = config_file($vmid);
5288 my $tmpfn = "$conffile.$$.tmp";
5289
5290 # disable interrupts (always do cleanups)
5291 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5292 print STDERR "got interrupt - ignored\n";
5293 };
5294
5295 eval {
5296 # enable interrupts
5297 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5298 die "interrupted by signal\n";
5299 };
5300
5301 if ($archive eq '-') {
5302 print "extracting archive from STDIN\n";
5303 run_command($cmd, input => "<&STDIN");
5304 } else {
5305 print "extracting archive '$archive'\n";
5306 run_command($cmd);
5307 }
5308
5309 return if $opts->{info};
5310
5311 # read new mapping
5312 my $map = {};
5313 my $statfile = "$tmpdir/qmrestore.stat";
5314 if (my $fd = IO::File->new($statfile, "r")) {
5315 while (defined (my $line = <$fd>)) {
5316 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5317 $map->{$1} = $2 if $1;
5318 } else {
5319 print STDERR "unable to parse line in statfile - $line\n";
5320 }
5321 }
5322 $fd->close();
5323 }
5324
5325 my $confsrc = "$tmpdir/qemu-server.conf";
5326
5327 my $srcfd = new IO::File($confsrc, "r") ||
5328 die "unable to open file '$confsrc'\n";
5329
5330 my $outfd = new IO::File ($tmpfn, "w") ||
5331 die "unable to write config for VM $vmid\n";
5332
5333 my $cookie = { netcount => 0 };
5334 while (defined (my $line = <$srcfd>)) {
5335 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5336 }
5337
5338 $srcfd->close();
5339 $outfd->close();
5340 };
5341 my $err = $@;
5342
5343 if ($err) {
5344
5345 unlink $tmpfn;
5346
5347 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5348
5349 die $err;
5350 }
5351
5352 rmtree $tmpdir;
5353
5354 rename $tmpfn, $conffile ||
5355 die "unable to commit configuration file '$conffile'\n";
5356
5357 PVE::Cluster::cfs_update(); # make sure we read new file
5358
5359 eval { rescan($vmid, 1); };
5360 warn $@ if $@;
5361 };
5362
5363
5364 # Internal snapshots
5365
5366 # NOTE: Snapshot create/delete involves several non-atomic
5367 # action, and can take a long time.
5368 # So we try to avoid locking the file and use 'lock' variable
5369 # inside the config file instead.
5370
5371 my $snapshot_copy_config = sub {
5372 my ($source, $dest) = @_;
5373
5374 foreach my $k (keys %$source) {
5375 next if $k eq 'snapshots';
5376 next if $k eq 'snapstate';
5377 next if $k eq 'snaptime';
5378 next if $k eq 'vmstate';
5379 next if $k eq 'lock';
5380 next if $k eq 'digest';
5381 next if $k eq 'description';
5382 next if $k =~ m/^unused\d+$/;
5383
5384 $dest->{$k} = $source->{$k};
5385 }
5386 };
5387
5388 my $snapshot_apply_config = sub {
5389 my ($conf, $snap) = @_;
5390
5391 # copy snapshot list
5392 my $newconf = {
5393 snapshots => $conf->{snapshots},
5394 };
5395
5396 # keep description and list of unused disks
5397 foreach my $k (keys %$conf) {
5398 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5399 $newconf->{$k} = $conf->{$k};
5400 }
5401
5402 &$snapshot_copy_config($snap, $newconf);
5403
5404 return $newconf;
5405 };
5406
5407 sub foreach_writable_storage {
5408 my ($conf, $func) = @_;
5409
5410 my $sidhash = {};
5411
5412 foreach my $ds (keys %$conf) {
5413 next if !valid_drivename($ds);
5414
5415 my $drive = parse_drive($ds, $conf->{$ds});
5416 next if !$drive;
5417 next if drive_is_cdrom($drive);
5418
5419 my $volid = $drive->{file};
5420
5421 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5422 $sidhash->{$sid} = $sid if $sid;
5423 }
5424
5425 foreach my $sid (sort keys %$sidhash) {
5426 &$func($sid);
5427 }
5428 }
5429
5430 my $alloc_vmstate_volid = sub {
5431 my ($storecfg, $vmid, $conf, $snapname) = @_;
5432
5433 # Note: we try to be smart when selecting a $target storage
5434
5435 my $target;
5436
5437 # search shared storage first
5438 foreach_writable_storage($conf, sub {
5439 my ($sid) = @_;
5440 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5441 return if !$scfg->{shared};
5442
5443 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5444 });
5445
5446 if (!$target) {
5447 # now search local storage
5448 foreach_writable_storage($conf, sub {
5449 my ($sid) = @_;
5450 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5451 return if $scfg->{shared};
5452
5453 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5454 });
5455 }
5456
5457 $target = 'local' if !$target;
5458
5459 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5460 # we abort live save after $conf->{memory}, so we need at max twice that space
5461 my $size = $conf->{memory}*2 + $driver_state_size;
5462
5463 my $name = "vm-$vmid-state-$snapname";
5464 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5465 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5466 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5467
5468 return $volid;
5469 };
5470
5471 my $snapshot_prepare = sub {
5472 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5473
5474 my $snap;
5475
5476 my $updatefn = sub {
5477
5478 my $conf = load_config($vmid);
5479
5480 die "you can't take a snapshot if it's a template\n"
5481 if is_template($conf);
5482
5483 check_lock($conf);
5484
5485 $conf->{lock} = 'snapshot';
5486
5487 die "snapshot name '$snapname' already used\n"
5488 if defined($conf->{snapshots}->{$snapname});
5489
5490 my $storecfg = PVE::Storage::config();
5491 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5492
5493 $snap = $conf->{snapshots}->{$snapname} = {};
5494
5495 if ($save_vmstate && check_running($vmid)) {
5496 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5497 }
5498
5499 &$snapshot_copy_config($conf, $snap);
5500
5501 $snap->{snapstate} = "prepare";
5502 $snap->{snaptime} = time();
5503 $snap->{description} = $comment if $comment;
5504
5505 # always overwrite machine if we save vmstate. This makes sure we
5506 # can restore it later using correct machine type
5507 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5508
5509 update_config_nolock($vmid, $conf, 1);
5510 };
5511
5512 lock_config($vmid, $updatefn);
5513
5514 return $snap;
5515 };
5516
5517 my $snapshot_commit = sub {
5518 my ($vmid, $snapname) = @_;
5519
5520 my $updatefn = sub {
5521
5522 my $conf = load_config($vmid);
5523
5524 die "missing snapshot lock\n"
5525 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5526
5527 my $has_machine_config = defined($conf->{machine});
5528
5529 my $snap = $conf->{snapshots}->{$snapname};
5530
5531 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5532
5533 die "wrong snapshot state\n"
5534 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5535
5536 delete $snap->{snapstate};
5537 delete $conf->{lock};
5538
5539 my $newconf = &$snapshot_apply_config($conf, $snap);
5540
5541 delete $newconf->{machine} if !$has_machine_config;
5542
5543 $newconf->{parent} = $snapname;
5544
5545 update_config_nolock($vmid, $newconf, 1);
5546 };
5547
5548 lock_config($vmid, $updatefn);
5549 };
5550
5551 sub snapshot_rollback {
5552 my ($vmid, $snapname) = @_;
5553
5554 my $snap;
5555
5556 my $prepare = 1;
5557
5558 my $storecfg = PVE::Storage::config();
5559
5560 my $updatefn = sub {
5561
5562 my $conf = load_config($vmid);
5563
5564 die "you can't rollback if vm is a template\n" if is_template($conf);
5565
5566 $snap = $conf->{snapshots}->{$snapname};
5567
5568 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5569
5570 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5571 if $snap->{snapstate};
5572
5573 if ($prepare) {
5574 check_lock($conf);
5575 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5576 }
5577
5578 die "unable to rollback vm $vmid: vm is running\n"
5579 if check_running($vmid);
5580
5581 if ($prepare) {
5582 $conf->{lock} = 'rollback';
5583 } else {
5584 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5585 delete $conf->{lock};
5586 }
5587
5588 my $forcemachine;
5589
5590 if (!$prepare) {
5591 my $has_machine_config = defined($conf->{machine});
5592
5593 # copy snapshot config to current config
5594 $conf = &$snapshot_apply_config($conf, $snap);
5595 $conf->{parent} = $snapname;
5596
5597 # Note: old code did not store 'machine', so we try to be smart
5598 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5599 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5600 # we remove the 'machine' configuration if not explicitly specified
5601 # in the original config.
5602 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5603 }
5604
5605 update_config_nolock($vmid, $conf, 1);
5606
5607 if (!$prepare && $snap->{vmstate}) {
5608 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5609 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5610 }
5611 };
5612
5613 lock_config($vmid, $updatefn);
5614
5615 foreach_drive($snap, sub {
5616 my ($ds, $drive) = @_;
5617
5618 return if drive_is_cdrom($drive);
5619
5620 my $volid = $drive->{file};
5621 my $device = "drive-$ds";
5622
5623 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5624 });
5625
5626 $prepare = 0;
5627 lock_config($vmid, $updatefn);
5628 }
5629
5630 my $savevm_wait = sub {
5631 my ($vmid) = @_;
5632
5633 for(;;) {
5634 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5635 if (!$stat->{status}) {
5636 die "savevm not active\n";
5637 } elsif ($stat->{status} eq 'active') {
5638 sleep(1);
5639 next;
5640 } elsif ($stat->{status} eq 'completed') {
5641 last;
5642 } else {
5643 die "query-savevm returned status '$stat->{status}'\n";
5644 }
5645 }
5646 };
5647
5648 sub snapshot_create {
5649 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5650
5651 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5652
5653 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5654
5655 my $config = load_config($vmid);
5656
5657 my $running = check_running($vmid);
5658
5659 my $freezefs = $running && $config->{agent};
5660 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5661
5662 my $drivehash = {};
5663
5664 if ($freezefs) {
5665 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5666 warn "guest-fsfreeze-freeze problems - $@" if $@;
5667 }
5668
5669 eval {
5670 # create internal snapshots of all drives
5671
5672 my $storecfg = PVE::Storage::config();
5673
5674 if ($running) {
5675 if ($snap->{vmstate}) {
5676 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5677 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5678 &$savevm_wait($vmid);
5679 } else {
5680 vm_mon_cmd($vmid, "savevm-start");
5681 }
5682 };
5683
5684 foreach_drive($snap, sub {
5685 my ($ds, $drive) = @_;
5686
5687 return if drive_is_cdrom($drive);
5688
5689 my $volid = $drive->{file};
5690 my $device = "drive-$ds";
5691
5692 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5693 $drivehash->{$ds} = 1;
5694 });
5695 };
5696 my $err = $@;
5697
5698 if ($running) {
5699 eval { vm_mon_cmd($vmid, "savevm-end") };
5700 warn $@ if $@;
5701
5702 if ($freezefs) {
5703 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5704 warn "guest-fsfreeze-thaw problems - $@" if $@;
5705 }
5706
5707 # savevm-end is async, we need to wait
5708 for (;;) {
5709 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5710 if (!$stat->{bytes}) {
5711 last;
5712 } else {
5713 print "savevm not yet finished\n";
5714 sleep(1);
5715 next;
5716 }
5717 }
5718 }
5719
5720 if ($err) {
5721 warn "snapshot create failed: starting cleanup\n";
5722 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5723 warn $@ if $@;
5724 die $err;
5725 }
5726
5727 &$snapshot_commit($vmid, $snapname);
5728 }
5729
5730 # Note: $drivehash is only set when called from snapshot_create.
5731 sub snapshot_delete {
5732 my ($vmid, $snapname, $force, $drivehash) = @_;
5733
5734 my $prepare = 1;
5735
5736 my $snap;
5737 my $unused = [];
5738
5739 my $unlink_parent = sub {
5740 my ($confref, $new_parent) = @_;
5741
5742 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5743 if ($new_parent) {
5744 $confref->{parent} = $new_parent;
5745 } else {
5746 delete $confref->{parent};
5747 }
5748 }
5749 };
5750
5751 my $updatefn = sub {
5752 my ($remove_drive) = @_;
5753
5754 my $conf = load_config($vmid);
5755
5756 if (!$drivehash) {
5757 check_lock($conf);
5758 die "you can't delete a snapshot if vm is a template\n"
5759 if is_template($conf);
5760 }
5761
5762 $snap = $conf->{snapshots}->{$snapname};
5763
5764 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5765
5766 # remove parent refs
5767 if (!$prepare) {
5768 &$unlink_parent($conf, $snap->{parent});
5769 foreach my $sn (keys %{$conf->{snapshots}}) {
5770 next if $sn eq $snapname;
5771 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5772 }
5773 }
5774
5775 if ($remove_drive) {
5776 if ($remove_drive eq 'vmstate') {
5777 delete $snap->{$remove_drive};
5778 } else {
5779 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5780 my $volid = $drive->{file};
5781 delete $snap->{$remove_drive};
5782 add_unused_volume($conf, $volid);
5783 }
5784 }
5785
5786 if ($prepare) {
5787 $snap->{snapstate} = 'delete';
5788 } else {
5789 delete $conf->{snapshots}->{$snapname};
5790 delete $conf->{lock} if $drivehash;
5791 foreach my $volid (@$unused) {
5792 add_unused_volume($conf, $volid);
5793 }
5794 }
5795
5796 update_config_nolock($vmid, $conf, 1);
5797 };
5798
5799 lock_config($vmid, $updatefn);
5800
5801 # now remove vmstate file
5802
5803 my $storecfg = PVE::Storage::config();
5804
5805 if ($snap->{vmstate}) {
5806 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5807 if (my $err = $@) {
5808 die $err if !$force;
5809 warn $err;
5810 }
5811 # save changes (remove vmstate from snapshot)
5812 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5813 };
5814
5815 # now remove all internal snapshots
5816 foreach_drive($snap, sub {
5817 my ($ds, $drive) = @_;
5818
5819 return if drive_is_cdrom($drive);
5820
5821 my $volid = $drive->{file};
5822 my $device = "drive-$ds";
5823
5824 if (!$drivehash || $drivehash->{$ds}) {
5825 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5826 if (my $err = $@) {
5827 die $err if !$force;
5828 warn $err;
5829 }
5830 }
5831
5832 # save changes (remove drive fron snapshot)
5833 lock_config($vmid, $updatefn, $ds) if !$force;
5834 push @$unused, $volid;
5835 });
5836
5837 # now cleanup config
5838 $prepare = 0;
5839 lock_config($vmid, $updatefn);
5840 }
5841
5842 sub has_feature {
5843 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5844
5845 my $err;
5846 foreach_drive($conf, sub {
5847 my ($ds, $drive) = @_;
5848
5849 return if drive_is_cdrom($drive);
5850 my $volid = $drive->{file};
5851 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5852 });
5853
5854 return $err ? 0 : 1;
5855 }
5856
5857 sub template_create {
5858 my ($vmid, $conf, $disk) = @_;
5859
5860 my $storecfg = PVE::Storage::config();
5861
5862 foreach_drive($conf, sub {
5863 my ($ds, $drive) = @_;
5864
5865 return if drive_is_cdrom($drive);
5866 return if $disk && $ds ne $disk;
5867
5868 my $volid = $drive->{file};
5869 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5870
5871 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5872 $drive->{file} = $voliddst;
5873 $conf->{$ds} = print_drive($vmid, $drive);
5874 update_config_nolock($vmid, $conf, 1);
5875 });
5876 }
5877
5878 sub is_template {
5879 my ($conf) = @_;
5880
5881 return 1 if defined $conf->{template} && $conf->{template} == 1;
5882 }
5883
5884 sub qemu_img_convert {
5885 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5886
5887 my $storecfg = PVE::Storage::config();
5888 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5889 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5890
5891 if ($src_storeid && $dst_storeid) {
5892 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5893 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5894
5895 my $src_format = qemu_img_format($src_scfg, $src_volname);
5896 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5897
5898 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5899 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5900
5901 my $cmd = [];
5902 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5903 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5904 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5905
5906 my $parser = sub {
5907 my $line = shift;
5908 if($line =~ m/\((\S+)\/100\%\)/){
5909 my $percent = $1;
5910 my $transferred = int($size * $percent / 100);
5911 my $remaining = $size - $transferred;
5912
5913 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5914 }
5915
5916 };
5917
5918 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5919 my $err = $@;
5920 die "copy failed: $err" if $err;
5921 }
5922 }
5923
5924 sub qemu_img_format {
5925 my ($scfg, $volname) = @_;
5926
5927 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5928 return $1;
5929 } elsif ($scfg->{type} eq 'iscsi') {
5930 return "host_device";
5931 } else {
5932 return "raw";
5933 }
5934 }
5935
5936 sub qemu_drive_mirror {
5937 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5938
5939 my $count = 0;
5940 my $old_len = 0;
5941 my $frozen = undef;
5942 my $maxwait = 120;
5943
5944 my $storecfg = PVE::Storage::config();
5945 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5946
5947 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5948
5949 my $format;
5950 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5951 $format = $1;
5952 }
5953
5954 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5955
5956 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5957 $opts->{format} = $format if $format;
5958
5959 #fixme : sometime drive-mirror timeout, but works fine after.
5960 # (I have see the problem with big volume > 200GB), so we need to eval
5961 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5962 # ignore errors here
5963
5964 eval {
5965 while (1) {
5966 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5967 my $stat = @$stats[0];
5968 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5969 die "error job is not mirroring" if $stat->{type} ne "mirror";
5970
5971 my $busy = $stat->{busy};
5972
5973 if (my $total = $stat->{len}) {
5974 my $transferred = $stat->{offset} || 0;
5975 my $remaining = $total - $transferred;
5976 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5977
5978 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
5979 }
5980
5981 if ($stat->{len} == $stat->{offset}) {
5982 if ($busy eq 'false') {
5983
5984 last if $vmiddst != $vmid;
5985
5986 # try to switch the disk if source and destination are on the same guest
5987 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
5988 last if !$@;
5989 die $@ if $@ !~ m/cannot be completed/;
5990 }
5991
5992 if ($count > $maxwait) {
5993 # if too much writes to disk occurs at the end of migration
5994 #the disk needs to be freezed to be able to complete the migration
5995 vm_suspend($vmid,1);
5996 $frozen = 1;
5997 }
5998 $count ++
5999 }
6000 $old_len = $stat->{offset};
6001 sleep 1;
6002 }
6003
6004 vm_resume($vmid, 1) if $frozen;
6005
6006 };
6007 my $err = $@;
6008
6009 my $cancel_job = sub {
6010 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6011 while (1) {
6012 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6013 my $stat = @$stats[0];
6014 last if !$stat;
6015 sleep 1;
6016 }
6017 };
6018
6019 if ($err) {
6020 eval { &$cancel_job(); };
6021 die "mirroring error: $err";
6022 }
6023
6024 if ($vmiddst != $vmid) {
6025 # if we clone a disk for a new target vm, we don't switch the disk
6026 &$cancel_job(); # so we call block-job-cancel
6027 }
6028 }
6029
6030 sub clone_disk {
6031 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6032 $newvmid, $storage, $format, $full, $newvollist) = @_;
6033
6034 my $newvolid;
6035
6036 if (!$full) {
6037 print "create linked clone of drive $drivename ($drive->{file})\n";
6038 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6039 push @$newvollist, $newvolid;
6040 } else {
6041 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6042 $storeid = $storage if $storage;
6043
6044 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6045 if (!$format) {
6046 $format = $drive->{format} || $defFormat;
6047 }
6048
6049 # test if requested format is supported - else use default
6050 my $supported = grep { $_ eq $format } @$validFormats;
6051 $format = $defFormat if !$supported;
6052
6053 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6054
6055 print "create full clone of drive $drivename ($drive->{file})\n";
6056 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6057 push @$newvollist, $newvolid;
6058
6059 if (!$running || $snapname) {
6060 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6061 } else {
6062 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6063 }
6064 }
6065
6066 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6067
6068 my $disk = $drive;
6069 $disk->{format} = undef;
6070 $disk->{file} = $newvolid;
6071 $disk->{size} = $size;
6072
6073 return $disk;
6074 }
6075
6076 # this only works if VM is running
6077 sub get_current_qemu_machine {
6078 my ($vmid) = @_;
6079
6080 my $cmd = { execute => 'query-machines', arguments => {} };
6081 my $res = vm_qmp_command($vmid, $cmd);
6082
6083 my ($current, $default);
6084 foreach my $e (@$res) {
6085 $default = $e->{name} if $e->{'is-default'};
6086 $current = $e->{name} if $e->{'is-current'};
6087 }
6088
6089 # fallback to the default machine if current is not supported by qemu
6090 return $current || $default || 'pc';
6091 }
6092
6093 sub lspci {
6094
6095 my $devices = {};
6096
6097 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6098 my (undef, $id, $function) = @_;
6099 my $res = { id => $id, function => $function};
6100 push @{$devices->{$id}}, $res;
6101 });
6102
6103 return $devices;
6104 }
6105
6106 1;