]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
correctly set and remove lock
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'string', format => 'pve-hotplug-features',
176 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
177 default => 'network,disk,usb',
178 },
179 reboot => {
180 optional => 1,
181 type => 'boolean',
182 description => "Allow reboot. If set to '0' the VM exit on reboot.",
183 default => 1,
184 },
185 lock => {
186 optional => 1,
187 type => 'string',
188 description => "Lock/unlock the VM.",
189 enum => [qw(migrate backup snapshot rollback)],
190 },
191 cpulimit => {
192 optional => 1,
193 type => 'integer',
194 description => "Limit of CPU usage in per cent. Note if the computer has 2 CPUs, it has total of 200% CPU time. Value '0' indicates no CPU limit.\n\nNOTE: This option is currently ignored.",
195 minimum => 0,
196 default => 0,
197 },
198 cpuunits => {
199 optional => 1,
200 type => 'integer',
201 description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.\n\nNOTE: You can disable fair-scheduler configuration by setting this to 0.",
202 minimum => 0,
203 maximum => 500000,
204 default => 1000,
205 },
206 memory => {
207 optional => 1,
208 type => 'integer',
209 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
210 minimum => 16,
211 default => 512,
212 },
213 balloon => {
214 optional => 1,
215 type => 'integer',
216 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
217 minimum => 0,
218 },
219 shares => {
220 optional => 1,
221 type => 'integer',
222 description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning",
223 minimum => 0,
224 maximum => 50000,
225 default => 1000,
226 },
227 keyboard => {
228 optional => 1,
229 type => 'string',
230 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
231 enum => PVE::Tools::kvmkeymaplist(),
232 default => 'en-us',
233 },
234 name => {
235 optional => 1,
236 type => 'string', format => 'dns-name',
237 description => "Set a name for the VM. Only used on the configuration web interface.",
238 },
239 scsihw => {
240 optional => 1,
241 type => 'string',
242 description => "scsi controller model",
243 enum => [qw(lsi lsi53c810 virtio-scsi-pci megasas pvscsi)],
244 default => 'lsi',
245 },
246 description => {
247 optional => 1,
248 type => 'string',
249 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
250 },
251 ostype => {
252 optional => 1,
253 type => 'string',
254 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
255 description => <<EODESC,
256 Used to enable special optimization/features for specific
257 operating systems:
258
259 other => unspecified OS
260 wxp => Microsoft Windows XP
261 w2k => Microsoft Windows 2000
262 w2k3 => Microsoft Windows 2003
263 w2k8 => Microsoft Windows 2008
264 wvista => Microsoft Windows Vista
265 win7 => Microsoft Windows 7
266 win8 => Microsoft Windows 8/2012
267 l24 => Linux 2.4 Kernel
268 l26 => Linux 2.6/3.X Kernel
269 solaris => solaris/opensolaris/openindiania kernel
270
271 other|l24|l26|solaris ... no special behaviour
272 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
273 EODESC
274 },
275 boot => {
276 optional => 1,
277 type => 'string',
278 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
279 pattern => '[acdn]{1,4}',
280 default => 'cdn',
281 },
282 bootdisk => {
283 optional => 1,
284 type => 'string', format => 'pve-qm-bootdisk',
285 description => "Enable booting from specified disk.",
286 pattern => '(ide|sata|scsi|virtio)\d+',
287 },
288 smp => {
289 optional => 1,
290 type => 'integer',
291 description => "The number of CPUs. Please use option -sockets instead.",
292 minimum => 1,
293 default => 1,
294 },
295 sockets => {
296 optional => 1,
297 type => 'integer',
298 description => "The number of CPU sockets.",
299 minimum => 1,
300 default => 1,
301 },
302 cores => {
303 optional => 1,
304 type => 'integer',
305 description => "The number of cores per socket.",
306 minimum => 1,
307 default => 1,
308 },
309 numa => {
310 optional => 1,
311 type => 'boolean',
312 description => "Enable/disable Numa.",
313 default => 0,
314 },
315 vcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Number of hotplugged vcpus.",
319 minimum => 1,
320 default => 0,
321 },
322 acpi => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable ACPI.",
326 default => 1,
327 },
328 agent => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable Qemu GuestAgent.",
332 default => 0,
333 },
334 kvm => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable KVM hardware virtualization.",
338 default => 1,
339 },
340 tdf => {
341 optional => 1,
342 type => 'boolean',
343 description => "Enable/disable time drift fix.",
344 default => 0,
345 },
346 localtime => {
347 optional => 1,
348 type => 'boolean',
349 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
350 },
351 freeze => {
352 optional => 1,
353 type => 'boolean',
354 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
355 },
356 vga => {
357 optional => 1,
358 type => 'string',
359 description => "Select VGA type. If you want to use high resolution modes (>= 1280x1024x16) then you should use option 'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and 'cirrur' for other OS types. Option 'qxl' enables the SPICE display sever. You can also run without any graphic card using a serial devive as terminal.",
360 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
361 },
362 watchdog => {
363 optional => 1,
364 type => 'string', format => 'pve-qm-watchdog',
365 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
366 description => "Create a virtual hardware watchdog device. Once enabled (by a guest action), the watchdog must be periodically polled by an agent inside the guest or else the guest will be restarted (or execute the action specified)",
367 },
368 startdate => {
369 optional => 1,
370 type => 'string',
371 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
372 description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
373 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
374 default => 'now',
375 },
376 startup => {
377 optional => 1,
378 type => 'string', format => 'pve-qm-startup',
379 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
380 description => "Startup and shutdown behavior. Order is a non-negative number defining the general startup order. Shutdown in done with reverse ordering. Additionally you can set the 'up' or 'down' delay in seconds, which specifies a delay to wait before the next VM is started or stopped.",
381 },
382 template => {
383 optional => 1,
384 type => 'boolean',
385 description => "Enable/disable Template.",
386 default => 0,
387 },
388 args => {
389 optional => 1,
390 type => 'string',
391 description => <<EODESCR,
392 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
393
394 args: -no-reboot -no-hpet
395 EODESCR
396 },
397 tablet => {
398 optional => 1,
399 type => 'boolean',
400 default => 1,
401 description => "Enable/disable the usb tablet device. This device is usually needed to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with normal VNC clients. If you're running lots of console-only guests on one host, you may consider disabling this to save some context switches. This is turned of by default if you use spice (vga=qxl).",
402 },
403 migrate_speed => {
404 optional => 1,
405 type => 'integer',
406 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
407 minimum => 0,
408 default => 0,
409 },
410 migrate_downtime => {
411 optional => 1,
412 type => 'number',
413 description => "Set maximum tolerated downtime (in seconds) for migrations.",
414 minimum => 0,
415 default => 0.1,
416 },
417 cdrom => {
418 optional => 1,
419 type => 'string', format => 'pve-qm-drive',
420 typetext => 'volume',
421 description => "This is an alias for option -ide2",
422 },
423 cpu => {
424 optional => 1,
425 description => "Emulated CPU type.",
426 type => 'string',
427 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
428 default => 'kvm64',
429 },
430 parent => get_standard_option('pve-snapshot-name', {
431 optional => 1,
432 description => "Parent snapshot name. This is used internally, and should not be modified.",
433 }),
434 snaptime => {
435 optional => 1,
436 description => "Timestamp for snapshots.",
437 type => 'integer',
438 minimum => 0,
439 },
440 vmstate => {
441 optional => 1,
442 type => 'string', format => 'pve-volume-id',
443 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
444 },
445 machine => {
446 description => "Specific the Qemu machine type.",
447 type => 'string',
448 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
449 maxLength => 40,
450 optional => 1,
451 },
452 smbios1 => {
453 description => "Specify SMBIOS type 1 fields.",
454 type => 'string', format => 'pve-qm-smbios1',
455 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
456 maxLength => 256,
457 optional => 1,
458 },
459 };
460
461 # what about other qemu settings ?
462 #cpu => 'string',
463 #machine => 'string',
464 #fda => 'file',
465 #fdb => 'file',
466 #mtdblock => 'file',
467 #sd => 'file',
468 #pflash => 'file',
469 #snapshot => 'bool',
470 #bootp => 'file',
471 ##tftp => 'dir',
472 ##smb => 'dir',
473 #kernel => 'file',
474 #append => 'string',
475 #initrd => 'file',
476 ##soundhw => 'string',
477
478 while (my ($k, $v) = each %$confdesc) {
479 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
480 }
481
482 my $MAX_IDE_DISKS = 4;
483 my $MAX_SCSI_DISKS = 14;
484 my $MAX_VIRTIO_DISKS = 16;
485 my $MAX_SATA_DISKS = 6;
486 my $MAX_USB_DEVICES = 5;
487 my $MAX_NETS = 32;
488 my $MAX_UNUSED_DISKS = 8;
489 my $MAX_HOSTPCI_DEVICES = 4;
490 my $MAX_SERIAL_PORTS = 4;
491 my $MAX_PARALLEL_PORTS = 3;
492 my $MAX_NUMA = 8;
493 my $MAX_MEM = 4194304;
494 my $STATICMEM = 1024;
495
496 my $numadesc = {
497 optional => 1,
498 type => 'string', format => 'pve-qm-numanode',
499 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
500 description => "numa topology",
501 };
502 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
503
504 for (my $i = 0; $i < $MAX_NUMA; $i++) {
505 $confdesc->{"numa$i"} = $numadesc;
506 }
507
508 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
509 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
510 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
511 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
512
513 my $netdesc = {
514 optional => 1,
515 type => 'string', format => 'pve-qm-net',
516 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
517 description => <<EODESCR,
518 Specify network devices.
519
520 MODEL is one of: $nic_model_list_txt
521
522 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
523 automatically generated if not specified.
524
525 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
526
527 Option 'rate' is used to limit traffic bandwidth from and to this interface. It is specified as floating point number, unit is 'Megabytes per second'.
528
529 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
530
531 10.0.2.2 Gateway
532 10.0.2.3 DNS Server
533 10.0.2.4 SMB Server
534
535 The DHCP server assign addresses to the guest starting from 10.0.2.15.
536
537 EODESCR
538 };
539 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
540
541 for (my $i = 0; $i < $MAX_NETS; $i++) {
542 $confdesc->{"net$i"} = $netdesc;
543 }
544
545 my $drivename_hash;
546
547 my $idedesc = {
548 optional => 1,
549 type => 'string', format => 'pve-qm-drive',
550 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
551 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
552 };
553 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
554
555 my $scsidesc = {
556 optional => 1,
557 type => 'string', format => 'pve-qm-drive',
558 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
559 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
560 };
561 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
562
563 my $satadesc = {
564 optional => 1,
565 type => 'string', format => 'pve-qm-drive',
566 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
567 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
568 };
569 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
570
571 my $virtiodesc = {
572 optional => 1,
573 type => 'string', format => 'pve-qm-drive',
574 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
575 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
576 };
577 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
578
579 my $usbdesc = {
580 optional => 1,
581 type => 'string', format => 'pve-qm-usb-device',
582 typetext => 'host=HOSTUSBDEVICE|spice',
583 description => <<EODESCR,
584 Configure an USB device (n is 0 to 4). This can be used to
585 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
586
587 'bus-port(.port)*' (decimal numbers) or
588 'vendor_id:product_id' (hexadeciaml numbers)
589
590 You can use the 'lsusb -t' command to list existing usb devices.
591
592 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
593
594 The value 'spice' can be used to add a usb redirection devices for spice.
595
596 EODESCR
597 };
598 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
599
600 my $hostpcidesc = {
601 optional => 1,
602 type => 'string', format => 'pve-qm-hostpci',
603 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
604 description => <<EODESCR,
605 Map host pci devices. HOSTPCIDEVICE syntax is:
606
607 'bus:dev.func' (hexadecimal numbers)
608
609 You can us the 'lspci' command to list existing pci devices.
610
611 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
612
613 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
614
615 Experimental: user reported problems with this option.
616 EODESCR
617 };
618 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
619
620 my $serialdesc = {
621 optional => 1,
622 type => 'string',
623 pattern => '(/dev/.+|socket)',
624 description => <<EODESCR,
625 Create a serial device inside the VM (n is 0 to 3), and pass through a host serial device (i.e. /dev/ttyS0), or create a unix socket on the host side (use 'qm terminal' to open a terminal connection).
626
627 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
628
629 Experimental: user reported problems with this option.
630 EODESCR
631 };
632
633 my $paralleldesc= {
634 optional => 1,
635 type => 'string',
636 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
637 description => <<EODESCR,
638 Map host parallel devices (n is 0 to 2).
639
640 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
641
642 Experimental: user reported problems with this option.
643 EODESCR
644 };
645
646 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
647 $confdesc->{"parallel$i"} = $paralleldesc;
648 }
649
650 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
651 $confdesc->{"serial$i"} = $serialdesc;
652 }
653
654 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
655 $confdesc->{"hostpci$i"} = $hostpcidesc;
656 }
657
658 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
659 $drivename_hash->{"ide$i"} = 1;
660 $confdesc->{"ide$i"} = $idedesc;
661 }
662
663 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
664 $drivename_hash->{"sata$i"} = 1;
665 $confdesc->{"sata$i"} = $satadesc;
666 }
667
668 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
669 $drivename_hash->{"scsi$i"} = 1;
670 $confdesc->{"scsi$i"} = $scsidesc ;
671 }
672
673 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
674 $drivename_hash->{"virtio$i"} = 1;
675 $confdesc->{"virtio$i"} = $virtiodesc;
676 }
677
678 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
679 $confdesc->{"usb$i"} = $usbdesc;
680 }
681
682 my $unuseddesc = {
683 optional => 1,
684 type => 'string', format => 'pve-volume-id',
685 description => "Reference to unused volumes.",
686 };
687
688 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
689 $confdesc->{"unused$i"} = $unuseddesc;
690 }
691
692 my $kvm_api_version = 0;
693
694 sub kvm_version {
695
696 return $kvm_api_version if $kvm_api_version;
697
698 my $fh = IO::File->new("</dev/kvm") ||
699 return 0;
700
701 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
702 $kvm_api_version = $v;
703 }
704
705 $fh->close();
706
707 return $kvm_api_version;
708 }
709
710 my $kvm_user_version;
711
712 sub kvm_user_version {
713
714 return $kvm_user_version if $kvm_user_version;
715
716 $kvm_user_version = 'unknown';
717
718 my $tmp = `kvm -help 2>/dev/null`;
719
720 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
721 $kvm_user_version = $2;
722 }
723
724 return $kvm_user_version;
725
726 }
727
728 my $kernel_has_vhost_net = -c '/dev/vhost-net';
729
730 sub disknames {
731 # order is important - used to autoselect boot disk
732 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
733 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
734 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
735 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
736 }
737
738 sub valid_drivename {
739 my $dev = shift;
740
741 return defined($drivename_hash->{$dev});
742 }
743
744 sub option_exists {
745 my $key = shift;
746 return defined($confdesc->{$key});
747 }
748
749 sub nic_models {
750 return $nic_model_list;
751 }
752
753 sub os_list_description {
754
755 return {
756 other => 'Other',
757 wxp => 'Windows XP',
758 w2k => 'Windows 2000',
759 w2k3 =>, 'Windows 2003',
760 w2k8 => 'Windows 2008',
761 wvista => 'Windows Vista',
762 win7 => 'Windows 7',
763 win8 => 'Windows 8/2012',
764 l24 => 'Linux 2.4',
765 l26 => 'Linux 2.6',
766 };
767 }
768
769 my $cdrom_path;
770
771 sub get_cdrom_path {
772
773 return $cdrom_path if $cdrom_path;
774
775 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
776 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
777 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
778 }
779
780 sub get_iso_path {
781 my ($storecfg, $vmid, $cdrom) = @_;
782
783 if ($cdrom eq 'cdrom') {
784 return get_cdrom_path();
785 } elsif ($cdrom eq 'none') {
786 return '';
787 } elsif ($cdrom =~ m|^/|) {
788 return $cdrom;
789 } else {
790 return PVE::Storage::path($storecfg, $cdrom);
791 }
792 }
793
794 # try to convert old style file names to volume IDs
795 sub filename_to_volume_id {
796 my ($vmid, $file, $media) = @_;
797
798 if (!($file eq 'none' || $file eq 'cdrom' ||
799 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
800
801 return undef if $file =~ m|/|;
802
803 if ($media && $media eq 'cdrom') {
804 $file = "local:iso/$file";
805 } else {
806 $file = "local:$vmid/$file";
807 }
808 }
809
810 return $file;
811 }
812
813 sub verify_media_type {
814 my ($opt, $vtype, $media) = @_;
815
816 return if !$media;
817
818 my $etype;
819 if ($media eq 'disk') {
820 $etype = 'images';
821 } elsif ($media eq 'cdrom') {
822 $etype = 'iso';
823 } else {
824 die "internal error";
825 }
826
827 return if ($vtype eq $etype);
828
829 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
830 }
831
832 sub cleanup_drive_path {
833 my ($opt, $storecfg, $drive) = @_;
834
835 # try to convert filesystem paths to volume IDs
836
837 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
838 ($drive->{file} !~ m|^/dev/.+|) &&
839 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
840 ($drive->{file} !~ m/^\d+$/)) {
841 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
842 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
843 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
844 verify_media_type($opt, $vtype, $drive->{media});
845 $drive->{file} = $volid;
846 }
847
848 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
849 }
850
851 sub create_conf_nolock {
852 my ($vmid, $settings) = @_;
853
854 my $filename = config_file($vmid);
855
856 die "configuration file '$filename' already exists\n" if -f $filename;
857
858 my $defaults = load_defaults();
859
860 $settings->{name} = "vm$vmid" if !$settings->{name};
861 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
862
863 my $data = '';
864 foreach my $opt (keys %$settings) {
865 next if !$confdesc->{$opt};
866
867 my $value = $settings->{$opt};
868 next if !$value;
869
870 $data .= "$opt: $value\n";
871 }
872
873 PVE::Tools::file_set_contents($filename, $data);
874 }
875
876 sub parse_hotplug_features {
877 my ($data) = @_;
878
879 my $res = {};
880
881 return $res if $data eq '0';
882
883 $data = $confdesc->{hotplug}->{default} if $data eq '1';
884
885 foreach my $feature (PVE::Tools::split_list($data)) {
886 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
887 $res->{$1} = 1;
888 } else {
889 warn "ignoring unknown hotplug feature '$feature'\n";
890 }
891 }
892 return $res;
893 }
894
895 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
896 sub pve_verify_hotplug_features {
897 my ($value, $noerr) = @_;
898
899 return $value if parse_hotplug_features($value);
900
901 return undef if $noerr;
902
903 die "unable to parse hotplug option\n";
904 }
905
906 my $parse_size = sub {
907 my ($value) = @_;
908
909 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
910 my ($size, $unit) = ($1, $3);
911 if ($unit) {
912 if ($unit eq 'K') {
913 $size = $size * 1024;
914 } elsif ($unit eq 'M') {
915 $size = $size * 1024 * 1024;
916 } elsif ($unit eq 'G') {
917 $size = $size * 1024 * 1024 * 1024;
918 }
919 }
920 return int($size);
921 };
922
923 my $format_size = sub {
924 my ($size) = @_;
925
926 $size = int($size);
927
928 my $kb = int($size/1024);
929 return $size if $kb*1024 != $size;
930
931 my $mb = int($kb/1024);
932 return "${kb}K" if $mb*1024 != $kb;
933
934 my $gb = int($mb/1024);
935 return "${mb}M" if $gb*1024 != $mb;
936
937 return "${gb}G";
938 };
939
940 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
941 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
942 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
943 # [,aio=native|threads][,discard=ignore|on]
944
945 sub parse_drive {
946 my ($key, $data) = @_;
947
948 my $res = {};
949
950 # $key may be undefined - used to verify JSON parameters
951 if (!defined($key)) {
952 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
953 $res->{index} = 0;
954 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
955 $res->{interface} = $1;
956 $res->{index} = $2;
957 } else {
958 return undef;
959 }
960
961 foreach my $p (split (/,/, $data)) {
962 next if $p =~ m/^\s*$/;
963
964 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
965 my ($k, $v) = ($1, $2);
966
967 $k = 'file' if $k eq 'volume';
968
969 return undef if defined $res->{$k};
970
971 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
972 return undef if !$v || $v !~ m/^\d+/;
973 $k = "m$k";
974 $v = sprintf("%.3f", $v / (1024*1024));
975 }
976 $res->{$k} = $v;
977 } else {
978 if (!$res->{file} && $p !~ m/=/) {
979 $res->{file} = $p;
980 } else {
981 return undef;
982 }
983 }
984 }
985
986 return undef if !$res->{file};
987
988 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
989 $res->{format} = $1;
990 }
991
992 return undef if $res->{cache} &&
993 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
994 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
995 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
996 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
997 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
998 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
999 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
1000 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
1001 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
1002 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
1003 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
1004 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
1005 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
1006
1007 return undef if $res->{mbps_rd} && $res->{mbps};
1008 return undef if $res->{mbps_wr} && $res->{mbps};
1009
1010 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
1011 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
1012 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
1013 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
1014 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
1015 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
1016
1017 return undef if $res->{iops_rd} && $res->{iops};
1018 return undef if $res->{iops_wr} && $res->{iops};
1019
1020
1021 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
1022 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
1023 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
1024 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
1025 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
1026 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
1027
1028
1029 if ($res->{size}) {
1030 return undef if !defined($res->{size} = &$parse_size($res->{size}));
1031 }
1032
1033 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1034 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1035 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1036 return undef if $res->{interface} eq 'virtio';
1037 }
1038
1039 # rerror does not work with scsi drives
1040 if ($res->{rerror}) {
1041 return undef if $res->{interface} eq 'scsi';
1042 }
1043
1044 return $res;
1045 }
1046
1047 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1048
1049 sub print_drive {
1050 my ($vmid, $drive) = @_;
1051
1052 my $opts = '';
1053 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1054 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1055 }
1056
1057 if ($drive->{size}) {
1058 $opts .= ",size=" . &$format_size($drive->{size});
1059 }
1060
1061 return "$drive->{file}$opts";
1062 }
1063
1064 sub scsi_inquiry {
1065 my($fh, $noerr) = @_;
1066
1067 my $SG_IO = 0x2285;
1068 my $SG_GET_VERSION_NUM = 0x2282;
1069
1070 my $versionbuf = "\x00" x 8;
1071 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1072 if (!$ret) {
1073 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1074 return undef;
1075 }
1076 my $version = unpack("I", $versionbuf);
1077 if ($version < 30000) {
1078 die "scsi generic interface too old\n" if !$noerr;
1079 return undef;
1080 }
1081
1082 my $buf = "\x00" x 36;
1083 my $sensebuf = "\x00" x 8;
1084 my $cmd = pack("C x3 C x1", 0x12, 36);
1085
1086 # see /usr/include/scsi/sg.h
1087 my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1088
1089 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1090 length($sensebuf), 0, length($buf), $buf,
1091 $cmd, $sensebuf, 6000);
1092
1093 $ret = ioctl($fh, $SG_IO, $packet);
1094 if (!$ret) {
1095 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1096 return undef;
1097 }
1098
1099 my @res = unpack($sg_io_hdr_t, $packet);
1100 if ($res[17] || $res[18]) {
1101 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1102 return undef;
1103 }
1104
1105 my $res = {};
1106 (my $byte0, my $byte1, $res->{vendor},
1107 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1108
1109 $res->{removable} = $byte1 & 128 ? 1 : 0;
1110 $res->{type} = $byte0 & 31;
1111
1112 return $res;
1113 }
1114
1115 sub path_is_scsi {
1116 my ($path) = @_;
1117
1118 my $fh = IO::File->new("+<$path") || return undef;
1119 my $res = scsi_inquiry($fh, 1);
1120 close($fh);
1121
1122 return $res;
1123 }
1124
1125 sub machine_type_is_q35 {
1126 my ($conf) = @_;
1127
1128 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1129 }
1130
1131 sub print_tabletdevice_full {
1132 my ($conf) = @_;
1133
1134 my $q35 = machine_type_is_q35($conf);
1135
1136 # we use uhci for old VMs because tablet driver was buggy in older qemu
1137 my $usbbus = $q35 ? "ehci" : "uhci";
1138
1139 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1140 }
1141
1142 sub print_drivedevice_full {
1143 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1144
1145 my $device = '';
1146 my $maxdev = 0;
1147
1148 if ($drive->{interface} eq 'virtio') {
1149 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1150 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1151 $device .= ",iothread=iothread0" if $conf->{iothread};
1152 } elsif ($drive->{interface} eq 'scsi') {
1153 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1154 my $controller = int($drive->{index} / $maxdev);
1155 my $unit = $drive->{index} % $maxdev;
1156 my $devicetype = 'hd';
1157 my $path = '';
1158 if (drive_is_cdrom($drive)) {
1159 $devicetype = 'cd';
1160 } else {
1161 if ($drive->{file} =~ m|^/|) {
1162 $path = $drive->{file};
1163 } else {
1164 $path = PVE::Storage::path($storecfg, $drive->{file});
1165 }
1166
1167 if($path =~ m/^iscsi\:\/\//){
1168 $devicetype = 'generic';
1169 } else {
1170 if (my $info = path_is_scsi($path)) {
1171 if ($info->{type} == 0) {
1172 $devicetype = 'block';
1173 } elsif ($info->{type} == 1) { # tape
1174 $devicetype = 'generic';
1175 }
1176 }
1177 }
1178 }
1179
1180 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1181 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1182 } else {
1183 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1184 }
1185
1186 } elsif ($drive->{interface} eq 'ide'){
1187 $maxdev = 2;
1188 my $controller = int($drive->{index} / $maxdev);
1189 my $unit = $drive->{index} % $maxdev;
1190 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1191
1192 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1193 } elsif ($drive->{interface} eq 'sata'){
1194 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1195 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1196 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1197 } elsif ($drive->{interface} eq 'usb') {
1198 die "implement me";
1199 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1200 } else {
1201 die "unsupported interface type";
1202 }
1203
1204 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1205
1206 return $device;
1207 }
1208
1209 sub get_initiator_name {
1210 my $initiator;
1211
1212 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1213 while (defined(my $line = <$fh>)) {
1214 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1215 $initiator = $1;
1216 last;
1217 }
1218 $fh->close();
1219
1220 return $initiator;
1221 }
1222
1223 sub print_drive_full {
1224 my ($storecfg, $vmid, $drive) = @_;
1225
1226 my $opts = '';
1227 foreach my $o (@qemu_drive_options) {
1228 next if $o eq 'bootindex';
1229 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1230 }
1231
1232 foreach my $o (qw(bps bps_rd bps_wr)) {
1233 my $v = $drive->{"m$o"};
1234 $opts .= ",$o=" . int($v*1024*1024) if $v;
1235 }
1236
1237 # use linux-aio by default (qemu default is threads)
1238 $opts .= ",aio=native" if !$drive->{aio};
1239
1240 my $path;
1241 my $volid = $drive->{file};
1242 if (drive_is_cdrom($drive)) {
1243 $path = get_iso_path($storecfg, $vmid, $volid);
1244 } else {
1245 if ($volid =~ m|^/|) {
1246 $path = $volid;
1247 } else {
1248 $path = PVE::Storage::path($storecfg, $volid);
1249 }
1250 }
1251
1252 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1253
1254 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1255 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1256
1257 my $pathinfo = $path ? "file=$path," : '';
1258
1259 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1260 }
1261
1262 sub print_netdevice_full {
1263 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1264
1265 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1266
1267 my $device = $net->{model};
1268 if ($net->{model} eq 'virtio') {
1269 $device = 'virtio-net-pci';
1270 };
1271
1272 my $pciaddr = print_pci_addr("$netid", $bridges);
1273 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1274 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1275 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1276 my $vectors = $net->{queues} * 2 + 2;
1277 $tmpstr .= ",vectors=$vectors,mq=on";
1278 }
1279 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1280 return $tmpstr;
1281 }
1282
1283 sub print_netdev_full {
1284 my ($vmid, $conf, $net, $netid) = @_;
1285
1286 my $i = '';
1287 if ($netid =~ m/^net(\d+)$/) {
1288 $i = int($1);
1289 }
1290
1291 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1292
1293 my $ifname = "tap${vmid}i$i";
1294
1295 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1296 die "interface name '$ifname' is too long (max 15 character)\n"
1297 if length($ifname) >= 16;
1298
1299 my $vhostparam = '';
1300 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1301
1302 my $vmname = $conf->{name} || "vm$vmid";
1303
1304 my $netdev = "";
1305
1306 if ($net->{bridge}) {
1307 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1308 } else {
1309 $netdev = "type=user,id=$netid,hostname=$vmname";
1310 }
1311
1312 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1313
1314 return $netdev;
1315 }
1316
1317 sub drive_is_cdrom {
1318 my ($drive) = @_;
1319
1320 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1321
1322 }
1323
1324 sub parse_numa {
1325 my ($data) = @_;
1326
1327 my $res = {};
1328
1329 foreach my $kvp (split(/,/, $data)) {
1330
1331 if ($kvp =~ m/^memory=(\S+)$/) {
1332 $res->{memory} = $1;
1333 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1334 $res->{policy} = $1;
1335 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1336 $res->{cpus}->{start} = $1;
1337 $res->{cpus}->{end} = $3;
1338 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1339 $res->{hostnodes}->{start} = $1;
1340 $res->{hostnodes}->{end} = $3;
1341 } else {
1342 return undef;
1343 }
1344 }
1345
1346 return $res;
1347 }
1348
1349 sub parse_hostpci {
1350 my ($value) = @_;
1351
1352 return undef if !$value;
1353
1354
1355 my @list = split(/,/, $value);
1356 my $found;
1357
1358 my $res = {};
1359 foreach my $kv (@list) {
1360
1361 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1362 $found = 1;
1363 if(defined($4)){
1364 push @{$res->{pciid}}, { id => $2 , function => $4};
1365
1366 }else{
1367 my $pcidevices = lspci($2);
1368 $res->{pciid} = $pcidevices->{$2};
1369 }
1370 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1371 $res->{driver} = $1;
1372 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1373 $res->{rombar} = $1;
1374 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1375 $res->{'x-vga'} = $1;
1376 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1377 $res->{pcie} = 1 if $1 == 1;
1378 } else {
1379 warn "unknown hostpci setting '$kv'\n";
1380 }
1381 }
1382
1383 return undef if !$found;
1384
1385 return $res;
1386 }
1387
1388 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1389 sub parse_net {
1390 my ($data) = @_;
1391
1392 my $res = {};
1393
1394 foreach my $kvp (split(/,/, $data)) {
1395
1396 if ($kvp =~ m/^(ne2k_pci|e1000|e1000-82540em|e1000-82544gc|e1000-82545em|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1397 my $model = lc($1);
1398 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1399 $res->{model} = $model;
1400 $res->{macaddr} = $mac;
1401 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1402 $res->{bridge} = $1;
1403 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1404 $res->{queues} = $1;
1405 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1406 $res->{rate} = $1;
1407 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1408 $res->{tag} = $1;
1409 } elsif ($kvp =~ m/^firewall=([01])$/) {
1410 $res->{firewall} = $1;
1411 } elsif ($kvp =~ m/^link_down=([01])$/) {
1412 $res->{link_down} = $1;
1413 } else {
1414 return undef;
1415 }
1416
1417 }
1418
1419 return undef if !$res->{model};
1420
1421 return $res;
1422 }
1423
1424 sub print_net {
1425 my $net = shift;
1426
1427 my $res = "$net->{model}";
1428 $res .= "=$net->{macaddr}" if $net->{macaddr};
1429 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1430 $res .= ",rate=$net->{rate}" if $net->{rate};
1431 $res .= ",tag=$net->{tag}" if $net->{tag};
1432 $res .= ",firewall=1" if $net->{firewall};
1433 $res .= ",link_down=1" if $net->{link_down};
1434 $res .= ",queues=$net->{queues}" if $net->{queues};
1435
1436 return $res;
1437 }
1438
1439 sub add_random_macs {
1440 my ($settings) = @_;
1441
1442 foreach my $opt (keys %$settings) {
1443 next if $opt !~ m/^net(\d+)$/;
1444 my $net = parse_net($settings->{$opt});
1445 next if !$net;
1446 $settings->{$opt} = print_net($net);
1447 }
1448 }
1449
1450 sub add_unused_volume {
1451 my ($config, $volid) = @_;
1452
1453 my $key;
1454 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1455 my $test = "unused$ind";
1456 if (my $vid = $config->{$test}) {
1457 return if $vid eq $volid; # do not add duplicates
1458 } else {
1459 $key = $test;
1460 }
1461 }
1462
1463 die "To many unused volume - please delete them first.\n" if !$key;
1464
1465 $config->{$key} = $volid;
1466
1467 return $key;
1468 }
1469
1470 sub vm_is_volid_owner {
1471 my ($storecfg, $vmid, $volid) = @_;
1472
1473 if ($volid !~ m|^/|) {
1474 my ($path, $owner);
1475 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1476 if ($owner && ($owner == $vmid)) {
1477 return 1;
1478 }
1479 }
1480
1481 return undef;
1482 }
1483
1484 sub vmconfig_delete_pending_option {
1485 my ($conf, $key) = @_;
1486
1487 delete $conf->{pending}->{$key};
1488 my $pending_delete_hash = { $key => 1 };
1489 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1490 $pending_delete_hash->{$opt} = 1;
1491 }
1492 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1493 }
1494
1495 sub vmconfig_undelete_pending_option {
1496 my ($conf, $key) = @_;
1497
1498 my $pending_delete_hash = {};
1499 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1500 $pending_delete_hash->{$opt} = 1;
1501 }
1502 delete $pending_delete_hash->{$key};
1503
1504 my @keylist = keys %$pending_delete_hash;
1505 if (scalar(@keylist)) {
1506 $conf->{pending}->{delete} = join(',', @keylist);
1507 } else {
1508 delete $conf->{pending}->{delete};
1509 }
1510 }
1511
1512 sub vmconfig_register_unused_drive {
1513 my ($storecfg, $vmid, $conf, $drive) = @_;
1514
1515 if (!drive_is_cdrom($drive)) {
1516 my $volid = $drive->{file};
1517 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1518 add_unused_volume($conf, $volid, $vmid);
1519 }
1520 }
1521 }
1522
1523 sub vmconfig_cleanup_pending {
1524 my ($conf) = @_;
1525
1526 # remove pending changes when nothing changed
1527 my $changes;
1528 foreach my $opt (keys %{$conf->{pending}}) {
1529 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1530 $changes = 1;
1531 delete $conf->{pending}->{$opt};
1532 }
1533 }
1534
1535 # remove delete if option is not set
1536 my $pending_delete_hash = {};
1537 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1538 if (defined($conf->{$opt})) {
1539 $pending_delete_hash->{$opt} = 1;
1540 } else {
1541 $changes = 1;
1542 }
1543 }
1544
1545 my @keylist = keys %$pending_delete_hash;
1546 if (scalar(@keylist)) {
1547 $conf->{pending}->{delete} = join(',', @keylist);
1548 } else {
1549 delete $conf->{pending}->{delete};
1550 }
1551
1552 return $changes;
1553 }
1554
1555 my $valid_smbios1_options = {
1556 manufacturer => '\S+',
1557 product => '\S+',
1558 version => '\S+',
1559 serial => '\S+',
1560 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1561 sku => '\S+',
1562 family => '\S+',
1563 };
1564
1565 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1566 sub parse_smbios1 {
1567 my ($data) = @_;
1568
1569 my $res = {};
1570
1571 foreach my $kvp (split(/,/, $data)) {
1572 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1573 my ($k, $v) = split(/=/, $kvp);
1574 return undef if !defined($k) || !defined($v);
1575 return undef if !$valid_smbios1_options->{$k};
1576 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1577 $res->{$k} = $v;
1578 }
1579
1580 return $res;
1581 }
1582
1583 sub print_smbios1 {
1584 my ($smbios1) = @_;
1585
1586 my $data = '';
1587 foreach my $k (keys %$smbios1) {
1588 next if !defined($smbios1->{$k});
1589 next if !$valid_smbios1_options->{$k};
1590 $data .= ',' if $data;
1591 $data .= "$k=$smbios1->{$k}";
1592 }
1593 return $data;
1594 }
1595
1596 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1597 sub verify_smbios1 {
1598 my ($value, $noerr) = @_;
1599
1600 return $value if parse_smbios1($value);
1601
1602 return undef if $noerr;
1603
1604 die "unable to parse smbios (type 1) options\n";
1605 }
1606
1607 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1608 sub verify_bootdisk {
1609 my ($value, $noerr) = @_;
1610
1611 return $value if valid_drivename($value);
1612
1613 return undef if $noerr;
1614
1615 die "invalid boot disk '$value'\n";
1616 }
1617
1618 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1619 sub verify_numa {
1620 my ($value, $noerr) = @_;
1621
1622 return $value if parse_numa($value);
1623
1624 return undef if $noerr;
1625
1626 die "unable to parse numa options\n";
1627 }
1628
1629 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1630 sub verify_net {
1631 my ($value, $noerr) = @_;
1632
1633 return $value if parse_net($value);
1634
1635 return undef if $noerr;
1636
1637 die "unable to parse network options\n";
1638 }
1639
1640 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1641 sub verify_drive {
1642 my ($value, $noerr) = @_;
1643
1644 return $value if parse_drive(undef, $value);
1645
1646 return undef if $noerr;
1647
1648 die "unable to parse drive options\n";
1649 }
1650
1651 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1652 sub verify_hostpci {
1653 my ($value, $noerr) = @_;
1654
1655 return $value if parse_hostpci($value);
1656
1657 return undef if $noerr;
1658
1659 die "unable to parse pci id\n";
1660 }
1661
1662 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1663 sub verify_watchdog {
1664 my ($value, $noerr) = @_;
1665
1666 return $value if parse_watchdog($value);
1667
1668 return undef if $noerr;
1669
1670 die "unable to parse watchdog options\n";
1671 }
1672
1673 sub parse_watchdog {
1674 my ($value) = @_;
1675
1676 return undef if !$value;
1677
1678 my $res = {};
1679
1680 foreach my $p (split(/,/, $value)) {
1681 next if $p =~ m/^\s*$/;
1682
1683 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1684 $res->{model} = $2;
1685 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1686 $res->{action} = $2;
1687 } else {
1688 return undef;
1689 }
1690 }
1691
1692 return $res;
1693 }
1694
1695 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1696 sub verify_startup {
1697 my ($value, $noerr) = @_;
1698
1699 return $value if parse_startup($value);
1700
1701 return undef if $noerr;
1702
1703 die "unable to parse startup options\n";
1704 }
1705
1706 sub parse_startup {
1707 my ($value) = @_;
1708
1709 return undef if !$value;
1710
1711 my $res = {};
1712
1713 foreach my $p (split(/,/, $value)) {
1714 next if $p =~ m/^\s*$/;
1715
1716 if ($p =~ m/^(order=)?(\d+)$/) {
1717 $res->{order} = $2;
1718 } elsif ($p =~ m/^up=(\d+)$/) {
1719 $res->{up} = $1;
1720 } elsif ($p =~ m/^down=(\d+)$/) {
1721 $res->{down} = $1;
1722 } else {
1723 return undef;
1724 }
1725 }
1726
1727 return $res;
1728 }
1729
1730 sub parse_usb_device {
1731 my ($value) = @_;
1732
1733 return undef if !$value;
1734
1735 my @dl = split(/,/, $value);
1736 my $found;
1737
1738 my $res = {};
1739 foreach my $v (@dl) {
1740 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1741 $found = 1;
1742 $res->{vendorid} = $2;
1743 $res->{productid} = $4;
1744 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1745 $found = 1;
1746 $res->{hostbus} = $1;
1747 $res->{hostport} = $2;
1748 } elsif ($v =~ m/^spice$/) {
1749 $found = 1;
1750 $res->{spice} = 1;
1751 } else {
1752 return undef;
1753 }
1754 }
1755 return undef if !$found;
1756
1757 return $res;
1758 }
1759
1760 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1761 sub verify_usb_device {
1762 my ($value, $noerr) = @_;
1763
1764 return $value if parse_usb_device($value);
1765
1766 return undef if $noerr;
1767
1768 die "unable to parse usb device\n";
1769 }
1770
1771 # add JSON properties for create and set function
1772 sub json_config_properties {
1773 my $prop = shift;
1774
1775 foreach my $opt (keys %$confdesc) {
1776 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1777 $prop->{$opt} = $confdesc->{$opt};
1778 }
1779
1780 return $prop;
1781 }
1782
1783 sub check_type {
1784 my ($key, $value) = @_;
1785
1786 die "unknown setting '$key'\n" if !$confdesc->{$key};
1787
1788 my $type = $confdesc->{$key}->{type};
1789
1790 if (!defined($value)) {
1791 die "got undefined value\n";
1792 }
1793
1794 if ($value =~ m/[\n\r]/) {
1795 die "property contains a line feed\n";
1796 }
1797
1798 if ($type eq 'boolean') {
1799 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1800 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1801 die "type check ('boolean') failed - got '$value'\n";
1802 } elsif ($type eq 'integer') {
1803 return int($1) if $value =~ m/^(\d+)$/;
1804 die "type check ('integer') failed - got '$value'\n";
1805 } elsif ($type eq 'number') {
1806 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1807 die "type check ('number') failed - got '$value'\n";
1808 } elsif ($type eq 'string') {
1809 if (my $fmt = $confdesc->{$key}->{format}) {
1810 if ($fmt eq 'pve-qm-drive') {
1811 # special case - we need to pass $key to parse_drive()
1812 my $drive = parse_drive($key, $value);
1813 return $value if $drive;
1814 die "unable to parse drive options\n";
1815 }
1816 PVE::JSONSchema::check_format($fmt, $value);
1817 return $value;
1818 }
1819 $value =~ s/^\"(.*)\"$/$1/;
1820 return $value;
1821 } else {
1822 die "internal error"
1823 }
1824 }
1825
1826 sub lock_config_full {
1827 my ($vmid, $timeout, $code, @param) = @_;
1828
1829 my $filename = config_file_lock($vmid);
1830
1831 my $res = lock_file($filename, $timeout, $code, @param);
1832
1833 die $@ if $@;
1834
1835 return $res;
1836 }
1837
1838 sub lock_config_mode {
1839 my ($vmid, $timeout, $shared, $code, @param) = @_;
1840
1841 my $filename = config_file_lock($vmid);
1842
1843 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1844
1845 die $@ if $@;
1846
1847 return $res;
1848 }
1849
1850 sub lock_config {
1851 my ($vmid, $code, @param) = @_;
1852
1853 return lock_config_full($vmid, 10, $code, @param);
1854 }
1855
1856 sub cfs_config_path {
1857 my ($vmid, $node) = @_;
1858
1859 $node = $nodename if !$node;
1860 return "nodes/$node/qemu-server/$vmid.conf";
1861 }
1862
1863 sub check_iommu_support{
1864 #fixme : need to check IOMMU support
1865 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1866
1867 my $iommu=1;
1868 return $iommu;
1869
1870 }
1871
1872 sub config_file {
1873 my ($vmid, $node) = @_;
1874
1875 my $cfspath = cfs_config_path($vmid, $node);
1876 return "/etc/pve/$cfspath";
1877 }
1878
1879 sub config_file_lock {
1880 my ($vmid) = @_;
1881
1882 return "$lock_dir/lock-$vmid.conf";
1883 }
1884
1885 sub touch_config {
1886 my ($vmid) = @_;
1887
1888 my $conf = config_file($vmid);
1889 utime undef, undef, $conf;
1890 }
1891
1892 sub destroy_vm {
1893 my ($storecfg, $vmid, $keep_empty_config) = @_;
1894
1895 my $conffile = config_file($vmid);
1896
1897 my $conf = load_config($vmid);
1898
1899 check_lock($conf);
1900
1901 # only remove disks owned by this VM
1902 foreach_drive($conf, sub {
1903 my ($ds, $drive) = @_;
1904
1905 return if drive_is_cdrom($drive);
1906
1907 my $volid = $drive->{file};
1908
1909 return if !$volid || $volid =~ m|^/|;
1910
1911 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1912 return if !$path || !$owner || ($owner != $vmid);
1913
1914 PVE::Storage::vdisk_free($storecfg, $volid);
1915 });
1916
1917 if ($keep_empty_config) {
1918 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1919 } else {
1920 unlink $conffile;
1921 }
1922
1923 # also remove unused disk
1924 eval {
1925 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1926
1927 eval {
1928 PVE::Storage::foreach_volid($dl, sub {
1929 my ($volid, $sid, $volname, $d) = @_;
1930 PVE::Storage::vdisk_free($storecfg, $volid);
1931 });
1932 };
1933 warn $@ if $@;
1934
1935 };
1936 warn $@ if $@;
1937 }
1938
1939 sub load_config {
1940 my ($vmid, $node) = @_;
1941
1942 my $cfspath = cfs_config_path($vmid, $node);
1943
1944 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1945
1946 die "no such VM ('$vmid')\n" if !defined($conf);
1947
1948 return $conf;
1949 }
1950
1951 sub parse_vm_config {
1952 my ($filename, $raw) = @_;
1953
1954 return undef if !defined($raw);
1955
1956 my $res = {
1957 digest => Digest::SHA::sha1_hex($raw),
1958 snapshots => {},
1959 pending => {},
1960 };
1961
1962 $filename =~ m|/qemu-server/(\d+)\.conf$|
1963 || die "got strange filename '$filename'";
1964
1965 my $vmid = $1;
1966
1967 my $conf = $res;
1968 my $descr = '';
1969 my $section = '';
1970
1971 my @lines = split(/\n/, $raw);
1972 foreach my $line (@lines) {
1973 next if $line =~ m/^\s*$/;
1974
1975 if ($line =~ m/^\[PENDING\]\s*$/i) {
1976 $section = 'pending';
1977 $conf->{description} = $descr if $descr;
1978 $descr = '';
1979 $conf = $res->{$section} = {};
1980 next;
1981
1982 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1983 $section = $1;
1984 $conf->{description} = $descr if $descr;
1985 $descr = '';
1986 $conf = $res->{snapshots}->{$section} = {};
1987 next;
1988 }
1989
1990 if ($line =~ m/^\#(.*)\s*$/) {
1991 $descr .= PVE::Tools::decode_text($1) . "\n";
1992 next;
1993 }
1994
1995 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1996 $descr .= PVE::Tools::decode_text($2);
1997 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1998 $conf->{snapstate} = $1;
1999 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2000 my $key = $1;
2001 my $value = $2;
2002 $conf->{$key} = $value;
2003 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2004 my $value = $1;
2005 if ($section eq 'pending') {
2006 $conf->{delete} = $value; # we parse this later
2007 } else {
2008 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2009 }
2010 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2011 my $key = $1;
2012 my $value = $2;
2013 eval { $value = check_type($key, $value); };
2014 if ($@) {
2015 warn "vm $vmid - unable to parse value of '$key' - $@";
2016 } else {
2017 my $fmt = $confdesc->{$key}->{format};
2018 if ($fmt && $fmt eq 'pve-qm-drive') {
2019 my $v = parse_drive($key, $value);
2020 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2021 $v->{file} = $volid;
2022 $value = print_drive($vmid, $v);
2023 } else {
2024 warn "vm $vmid - unable to parse value of '$key'\n";
2025 next;
2026 }
2027 }
2028
2029 if ($key eq 'cdrom') {
2030 $conf->{ide2} = $value;
2031 } else {
2032 $conf->{$key} = $value;
2033 }
2034 }
2035 }
2036 }
2037
2038 $conf->{description} = $descr if $descr;
2039
2040 delete $res->{snapstate}; # just to be sure
2041
2042 return $res;
2043 }
2044
2045 sub write_vm_config {
2046 my ($filename, $conf) = @_;
2047
2048 delete $conf->{snapstate}; # just to be sure
2049
2050 if ($conf->{cdrom}) {
2051 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2052 $conf->{ide2} = $conf->{cdrom};
2053 delete $conf->{cdrom};
2054 }
2055
2056 # we do not use 'smp' any longer
2057 if ($conf->{sockets}) {
2058 delete $conf->{smp};
2059 } elsif ($conf->{smp}) {
2060 $conf->{sockets} = $conf->{smp};
2061 delete $conf->{cores};
2062 delete $conf->{smp};
2063 }
2064
2065 my $used_volids = {};
2066
2067 my $cleanup_config = sub {
2068 my ($cref, $pending, $snapname) = @_;
2069
2070 foreach my $key (keys %$cref) {
2071 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2072 $key eq 'snapstate' || $key eq 'pending';
2073 my $value = $cref->{$key};
2074 if ($key eq 'delete') {
2075 die "propertry 'delete' is only allowed in [PENDING]\n"
2076 if !$pending;
2077 # fixme: check syntax?
2078 next;
2079 }
2080 eval { $value = check_type($key, $value); };
2081 die "unable to parse value of '$key' - $@" if $@;
2082
2083 $cref->{$key} = $value;
2084
2085 if (!$snapname && valid_drivename($key)) {
2086 my $drive = parse_drive($key, $value);
2087 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2088 }
2089 }
2090 };
2091
2092 &$cleanup_config($conf);
2093
2094 &$cleanup_config($conf->{pending}, 1);
2095
2096 foreach my $snapname (keys %{$conf->{snapshots}}) {
2097 die "internal error" if $snapname eq 'pending';
2098 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2099 }
2100
2101 # remove 'unusedX' settings if we re-add a volume
2102 foreach my $key (keys %$conf) {
2103 my $value = $conf->{$key};
2104 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2105 delete $conf->{$key};
2106 }
2107 }
2108
2109 my $generate_raw_config = sub {
2110 my ($conf) = @_;
2111
2112 my $raw = '';
2113
2114 # add description as comment to top of file
2115 my $descr = $conf->{description} || '';
2116 foreach my $cl (split(/\n/, $descr)) {
2117 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2118 }
2119
2120 foreach my $key (sort keys %$conf) {
2121 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2122 $raw .= "$key: $conf->{$key}\n";
2123 }
2124 return $raw;
2125 };
2126
2127 my $raw = &$generate_raw_config($conf);
2128
2129 if (scalar(keys %{$conf->{pending}})){
2130 $raw .= "\n[PENDING]\n";
2131 $raw .= &$generate_raw_config($conf->{pending});
2132 }
2133
2134 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2135 $raw .= "\n[$snapname]\n";
2136 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2137 }
2138
2139 return $raw;
2140 }
2141
2142 sub update_config_nolock {
2143 my ($vmid, $conf, $skiplock) = @_;
2144
2145 check_lock($conf) if !$skiplock;
2146
2147 my $cfspath = cfs_config_path($vmid);
2148
2149 PVE::Cluster::cfs_write_file($cfspath, $conf);
2150 }
2151
2152 sub update_config {
2153 my ($vmid, $conf, $skiplock) = @_;
2154
2155 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2156 }
2157
2158 sub load_defaults {
2159
2160 my $res = {};
2161
2162 # we use static defaults from our JSON schema configuration
2163 foreach my $key (keys %$confdesc) {
2164 if (defined(my $default = $confdesc->{$key}->{default})) {
2165 $res->{$key} = $default;
2166 }
2167 }
2168
2169 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2170 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2171
2172 return $res;
2173 }
2174
2175 sub config_list {
2176 my $vmlist = PVE::Cluster::get_vmlist();
2177 my $res = {};
2178 return $res if !$vmlist || !$vmlist->{ids};
2179 my $ids = $vmlist->{ids};
2180
2181 foreach my $vmid (keys %$ids) {
2182 my $d = $ids->{$vmid};
2183 next if !$d->{node} || $d->{node} ne $nodename;
2184 next if !$d->{type} || $d->{type} ne 'qemu';
2185 $res->{$vmid}->{exists} = 1;
2186 }
2187 return $res;
2188 }
2189
2190 # test if VM uses local resources (to prevent migration)
2191 sub check_local_resources {
2192 my ($conf, $noerr) = @_;
2193
2194 my $loc_res = 0;
2195
2196 $loc_res = 1 if $conf->{hostusb}; # old syntax
2197 $loc_res = 1 if $conf->{hostpci}; # old syntax
2198
2199 foreach my $k (keys %$conf) {
2200 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2201 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2202 }
2203
2204 die "VM uses local resources\n" if $loc_res && !$noerr;
2205
2206 return $loc_res;
2207 }
2208
2209 # check if used storages are available on all nodes (use by migrate)
2210 sub check_storage_availability {
2211 my ($storecfg, $conf, $node) = @_;
2212
2213 foreach_drive($conf, sub {
2214 my ($ds, $drive) = @_;
2215
2216 my $volid = $drive->{file};
2217 return if !$volid;
2218
2219 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2220 return if !$sid;
2221
2222 # check if storage is available on both nodes
2223 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2224 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2225 });
2226 }
2227
2228 # list nodes where all VM images are available (used by has_feature API)
2229 sub shared_nodes {
2230 my ($conf, $storecfg) = @_;
2231
2232 my $nodelist = PVE::Cluster::get_nodelist();
2233 my $nodehash = { map { $_ => 1 } @$nodelist };
2234 my $nodename = PVE::INotify::nodename();
2235
2236 foreach_drive($conf, sub {
2237 my ($ds, $drive) = @_;
2238
2239 my $volid = $drive->{file};
2240 return if !$volid;
2241
2242 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2243 if ($storeid) {
2244 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2245 if ($scfg->{disable}) {
2246 $nodehash = {};
2247 } elsif (my $avail = $scfg->{nodes}) {
2248 foreach my $node (keys %$nodehash) {
2249 delete $nodehash->{$node} if !$avail->{$node};
2250 }
2251 } elsif (!$scfg->{shared}) {
2252 foreach my $node (keys %$nodehash) {
2253 delete $nodehash->{$node} if $node ne $nodename
2254 }
2255 }
2256 }
2257 });
2258
2259 return $nodehash
2260 }
2261
2262 sub check_lock {
2263 my ($conf) = @_;
2264
2265 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2266 }
2267
2268 sub check_cmdline {
2269 my ($pidfile, $pid) = @_;
2270
2271 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2272 if (defined($fh)) {
2273 my $line = <$fh>;
2274 $fh->close;
2275 return undef if !$line;
2276 my @param = split(/\0/, $line);
2277
2278 my $cmd = $param[0];
2279 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2280
2281 for (my $i = 0; $i < scalar (@param); $i++) {
2282 my $p = $param[$i];
2283 next if !$p;
2284 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2285 my $p = $param[$i+1];
2286 return 1 if $p && ($p eq $pidfile);
2287 return undef;
2288 }
2289 }
2290 }
2291 return undef;
2292 }
2293
2294 sub check_running {
2295 my ($vmid, $nocheck, $node) = @_;
2296
2297 my $filename = config_file($vmid, $node);
2298
2299 die "unable to find configuration file for VM $vmid - no such machine\n"
2300 if !$nocheck && ! -f $filename;
2301
2302 my $pidfile = pidfile_name($vmid);
2303
2304 if (my $fd = IO::File->new("<$pidfile")) {
2305 my $st = stat($fd);
2306 my $line = <$fd>;
2307 close($fd);
2308
2309 my $mtime = $st->mtime;
2310 if ($mtime > time()) {
2311 warn "file '$filename' modified in future\n";
2312 }
2313
2314 if ($line =~ m/^(\d+)$/) {
2315 my $pid = $1;
2316 if (check_cmdline($pidfile, $pid)) {
2317 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2318 return $pid;
2319 }
2320 }
2321 }
2322 }
2323
2324 return undef;
2325 }
2326
2327 sub vzlist {
2328
2329 my $vzlist = config_list();
2330
2331 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2332
2333 while (defined(my $de = $fd->read)) {
2334 next if $de !~ m/^(\d+)\.pid$/;
2335 my $vmid = $1;
2336 next if !defined($vzlist->{$vmid});
2337 if (my $pid = check_running($vmid)) {
2338 $vzlist->{$vmid}->{pid} = $pid;
2339 }
2340 }
2341
2342 return $vzlist;
2343 }
2344
2345 sub disksize {
2346 my ($storecfg, $conf) = @_;
2347
2348 my $bootdisk = $conf->{bootdisk};
2349 return undef if !$bootdisk;
2350 return undef if !valid_drivename($bootdisk);
2351
2352 return undef if !$conf->{$bootdisk};
2353
2354 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2355 return undef if !defined($drive);
2356
2357 return undef if drive_is_cdrom($drive);
2358
2359 my $volid = $drive->{file};
2360 return undef if !$volid;
2361
2362 return $drive->{size};
2363 }
2364
2365 my $last_proc_pid_stat;
2366
2367 # get VM status information
2368 # This must be fast and should not block ($full == false)
2369 # We only query KVM using QMP if $full == true (this can be slow)
2370 sub vmstatus {
2371 my ($opt_vmid, $full) = @_;
2372
2373 my $res = {};
2374
2375 my $storecfg = PVE::Storage::config();
2376
2377 my $list = vzlist();
2378 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2379
2380 my $cpucount = $cpuinfo->{cpus} || 1;
2381
2382 foreach my $vmid (keys %$list) {
2383 next if $opt_vmid && ($vmid ne $opt_vmid);
2384
2385 my $cfspath = cfs_config_path($vmid);
2386 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2387
2388 my $d = {};
2389 $d->{pid} = $list->{$vmid}->{pid};
2390
2391 # fixme: better status?
2392 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2393
2394 my $size = disksize($storecfg, $conf);
2395 if (defined($size)) {
2396 $d->{disk} = 0; # no info available
2397 $d->{maxdisk} = $size;
2398 } else {
2399 $d->{disk} = 0;
2400 $d->{maxdisk} = 0;
2401 }
2402
2403 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2404 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2405
2406 $d->{name} = $conf->{name} || "VM $vmid";
2407 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2408
2409 if ($conf->{balloon}) {
2410 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2411 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2412 }
2413
2414 $d->{uptime} = 0;
2415 $d->{cpu} = 0;
2416 $d->{mem} = 0;
2417
2418 $d->{netout} = 0;
2419 $d->{netin} = 0;
2420
2421 $d->{diskread} = 0;
2422 $d->{diskwrite} = 0;
2423
2424 $d->{template} = is_template($conf);
2425
2426 $res->{$vmid} = $d;
2427 }
2428
2429 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2430 foreach my $dev (keys %$netdev) {
2431 next if $dev !~ m/^tap([1-9]\d*)i/;
2432 my $vmid = $1;
2433 my $d = $res->{$vmid};
2434 next if !$d;
2435
2436 $d->{netout} += $netdev->{$dev}->{receive};
2437 $d->{netin} += $netdev->{$dev}->{transmit};
2438 }
2439
2440 my $ctime = gettimeofday;
2441
2442 foreach my $vmid (keys %$list) {
2443
2444 my $d = $res->{$vmid};
2445 my $pid = $d->{pid};
2446 next if !$pid;
2447
2448 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2449 next if !$pstat; # not running
2450
2451 my $used = $pstat->{utime} + $pstat->{stime};
2452
2453 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2454
2455 if ($pstat->{vsize}) {
2456 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2457 }
2458
2459 my $old = $last_proc_pid_stat->{$pid};
2460 if (!$old) {
2461 $last_proc_pid_stat->{$pid} = {
2462 time => $ctime,
2463 used => $used,
2464 cpu => 0,
2465 };
2466 next;
2467 }
2468
2469 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2470
2471 if ($dtime > 1000) {
2472 my $dutime = $used - $old->{used};
2473
2474 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2475 $last_proc_pid_stat->{$pid} = {
2476 time => $ctime,
2477 used => $used,
2478 cpu => $d->{cpu},
2479 };
2480 } else {
2481 $d->{cpu} = $old->{cpu};
2482 }
2483 }
2484
2485 return $res if !$full;
2486
2487 my $qmpclient = PVE::QMPClient->new();
2488
2489 my $ballooncb = sub {
2490 my ($vmid, $resp) = @_;
2491
2492 my $info = $resp->{'return'};
2493 return if !$info->{max_mem};
2494
2495 my $d = $res->{$vmid};
2496
2497 # use memory assigned to VM
2498 $d->{maxmem} = $info->{max_mem};
2499 $d->{balloon} = $info->{actual};
2500
2501 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2502 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2503 $d->{freemem} = $info->{free_mem};
2504 }
2505
2506 };
2507
2508 my $blockstatscb = sub {
2509 my ($vmid, $resp) = @_;
2510 my $data = $resp->{'return'} || [];
2511 my $totalrdbytes = 0;
2512 my $totalwrbytes = 0;
2513 for my $blockstat (@$data) {
2514 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2515 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2516 }
2517 $res->{$vmid}->{diskread} = $totalrdbytes;
2518 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2519 };
2520
2521 my $statuscb = sub {
2522 my ($vmid, $resp) = @_;
2523
2524 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2525 # this fails if ballon driver is not loaded, so this must be
2526 # the last commnand (following command are aborted if this fails).
2527 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2528
2529 my $status = 'unknown';
2530 if (!defined($status = $resp->{'return'}->{status})) {
2531 warn "unable to get VM status\n";
2532 return;
2533 }
2534
2535 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2536 };
2537
2538 foreach my $vmid (keys %$list) {
2539 next if $opt_vmid && ($vmid ne $opt_vmid);
2540 next if !$res->{$vmid}->{pid}; # not running
2541 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2542 }
2543
2544 $qmpclient->queue_execute(undef, 1);
2545
2546 foreach my $vmid (keys %$list) {
2547 next if $opt_vmid && ($vmid ne $opt_vmid);
2548 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2549 }
2550
2551 return $res;
2552 }
2553
2554 sub foreach_dimm {
2555 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2556
2557 my $dimm_id = 0;
2558 my $current_size = 1024;
2559 my $dimm_size = 512;
2560 return if $current_size == $memory;
2561
2562 for (my $j = 0; $j < 8; $j++) {
2563 for (my $i = 0; $i < 32; $i++) {
2564 my $name = "dimm${dimm_id}";
2565 $dimm_id++;
2566 my $numanode = $i % $sockets;
2567 $current_size += $dimm_size;
2568 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2569 return $current_size if $current_size >= $memory;
2570 }
2571 $dimm_size *= 2;
2572 }
2573 }
2574
2575 sub foreach_drive {
2576 my ($conf, $func) = @_;
2577
2578 foreach my $ds (keys %$conf) {
2579 next if !valid_drivename($ds);
2580
2581 my $drive = parse_drive($ds, $conf->{$ds});
2582 next if !$drive;
2583
2584 &$func($ds, $drive);
2585 }
2586 }
2587
2588 sub foreach_volid {
2589 my ($conf, $func) = @_;
2590
2591 my $volhash = {};
2592
2593 my $test_volid = sub {
2594 my ($volid, $is_cdrom) = @_;
2595
2596 return if !$volid;
2597
2598 $volhash->{$volid} = $is_cdrom || 0;
2599 };
2600
2601 foreach_drive($conf, sub {
2602 my ($ds, $drive) = @_;
2603 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2604 });
2605
2606 foreach my $snapname (keys %{$conf->{snapshots}}) {
2607 my $snap = $conf->{snapshots}->{$snapname};
2608 &$test_volid($snap->{vmstate}, 0);
2609 foreach_drive($snap, sub {
2610 my ($ds, $drive) = @_;
2611 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2612 });
2613 }
2614
2615 foreach my $volid (keys %$volhash) {
2616 &$func($volid, $volhash->{$volid});
2617 }
2618 }
2619
2620 sub vga_conf_has_spice {
2621 my ($vga) = @_;
2622
2623 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2624
2625 return $1 || 1;
2626 }
2627
2628 sub config_to_command {
2629 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2630
2631 my $cmd = [];
2632 my $globalFlags = [];
2633 my $machineFlags = [];
2634 my $rtcFlags = [];
2635 my $cpuFlags = [];
2636 my $devices = [];
2637 my $pciaddr = '';
2638 my $bridges = {};
2639 my $kvmver = kvm_user_version();
2640 my $vernum = 0; # unknown
2641 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2642 $vernum = $1*1000000+$2*1000;
2643 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2644 $vernum = $1*1000000+$2*1000+$3;
2645 }
2646
2647 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2648
2649 my $have_ovz = -f '/proc/vz/vestat';
2650
2651 my $q35 = machine_type_is_q35($conf);
2652 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2653
2654 push @$cmd, '/usr/bin/kvm';
2655
2656 push @$cmd, '-id', $vmid;
2657
2658 my $use_virtio = 0;
2659
2660 my $qmpsocket = qmp_socket($vmid);
2661 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2662 push @$cmd, '-mon', "chardev=qmp,mode=control";
2663
2664 my $socket = vnc_socket($vmid);
2665 push @$cmd, '-vnc', "unix:$socket,x509,password";
2666
2667 push @$cmd, '-pidfile' , pidfile_name($vmid);
2668
2669 push @$cmd, '-daemonize';
2670
2671 if ($conf->{smbios1}) {
2672 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2673 }
2674
2675 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2676
2677 if ($q35) {
2678 # the q35 chipset support native usb2, so we enable usb controller
2679 # by default for this machine type
2680 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2681 } else {
2682 $pciaddr = print_pci_addr("piix3", $bridges);
2683 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2684
2685 my $use_usb2 = 0;
2686 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2687 next if !$conf->{"usb$i"};
2688 $use_usb2 = 1;
2689 }
2690 # include usb device config
2691 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2692 }
2693
2694 my $vga = $conf->{vga};
2695
2696 my $qxlnum = vga_conf_has_spice($vga);
2697 $vga = 'qxl' if $qxlnum;
2698
2699 if (!$vga) {
2700 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2701 $conf->{ostype} eq 'win7' ||
2702 $conf->{ostype} eq 'w2k8')) {
2703 $vga = 'std';
2704 } else {
2705 $vga = 'cirrus';
2706 }
2707 }
2708
2709 # enable absolute mouse coordinates (needed by vnc)
2710 my $tablet;
2711 if (defined($conf->{tablet})) {
2712 $tablet = $conf->{tablet};
2713 } else {
2714 $tablet = $defaults->{tablet};
2715 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2716 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2717 }
2718
2719 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2720
2721 # host pci devices
2722 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2723 my $d = parse_hostpci($conf->{"hostpci$i"});
2724 next if !$d;
2725
2726 my $pcie = $d->{pcie};
2727 if($pcie){
2728 die "q35 machine model is not enabled" if !$q35;
2729 $pciaddr = print_pcie_addr("hostpci$i");
2730 }else{
2731 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2732 }
2733
2734 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2735 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2736 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2737 if ($xvga && $xvga ne '') {
2738 push @$cpuFlags, 'kvm=off';
2739 $vga = 'none';
2740 }
2741 $driver = "vfio-pci" if $xvga ne '';
2742 my $pcidevices = $d->{pciid};
2743 my $multifunction = 1 if @$pcidevices > 1;
2744
2745 my $j=0;
2746 foreach my $pcidevice (@$pcidevices) {
2747
2748 my $id = "hostpci$i";
2749 $id .= ".$j" if $multifunction;
2750 my $addr = $pciaddr;
2751 $addr .= ".$j" if $multifunction;
2752 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2753
2754 if($j == 0){
2755 $devicestr .= "$rombar$xvga";
2756 $devicestr .= ",multifunction=on" if $multifunction;
2757 }
2758
2759 push @$devices, '-device', $devicestr;
2760 $j++;
2761 }
2762 }
2763
2764 # usb devices
2765 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2766 my $d = parse_usb_device($conf->{"usb$i"});
2767 next if !$d;
2768 if ($d->{vendorid} && $d->{productid}) {
2769 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2770 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2771 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2772 } elsif ($d->{spice}) {
2773 # usb redir support for spice
2774 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2775 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2776 }
2777 }
2778
2779 # serial devices
2780 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2781 if (my $path = $conf->{"serial$i"}) {
2782 if ($path eq 'socket') {
2783 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2784 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2785 push @$devices, '-device', "isa-serial,chardev=serial$i";
2786 } else {
2787 die "no such serial device\n" if ! -c $path;
2788 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2789 push @$devices, '-device', "isa-serial,chardev=serial$i";
2790 }
2791 }
2792 }
2793
2794 # parallel devices
2795 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2796 if (my $path = $conf->{"parallel$i"}) {
2797 die "no such parallel device\n" if ! -c $path;
2798 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2799 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2800 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2801 }
2802 }
2803
2804 my $vmname = $conf->{name} || "vm$vmid";
2805
2806 push @$cmd, '-name', $vmname;
2807
2808 my $sockets = 1;
2809 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2810 $sockets = $conf->{sockets} if $conf->{sockets};
2811
2812 my $cores = $conf->{cores} || 1;
2813
2814 my $maxcpus = $sockets * $cores;
2815
2816 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2817
2818 my $allowed_vcpus = $cpuinfo->{cpus};
2819
2820 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2821 if ($allowed_vcpus < $maxcpus);
2822
2823 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2824
2825 push @$cmd, '-nodefaults';
2826
2827 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2828
2829 my $bootindex_hash = {};
2830 my $i = 1;
2831 foreach my $o (split(//, $bootorder)) {
2832 $bootindex_hash->{$o} = $i*100;
2833 $i++;
2834 }
2835
2836 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2837
2838 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2839
2840 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2841
2842 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2843
2844 # time drift fix
2845 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2846
2847 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2848 my $useLocaltime = $conf->{localtime};
2849
2850 if (my $ost = $conf->{ostype}) {
2851 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2852
2853 if ($ost =~ m/^w/) { # windows
2854 $useLocaltime = 1 if !defined($conf->{localtime});
2855
2856 # use time drift fix when acpi is enabled
2857 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2858 $tdf = 1 if !defined($conf->{tdf});
2859 }
2860 }
2861
2862 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2863 $ost eq 'wvista') {
2864 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2865 push @$cmd, '-no-hpet';
2866 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2867 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2868 }
2869
2870 if ($ost eq 'win7' || $ost eq 'win8') {
2871 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2872 }
2873 }
2874
2875 push @$rtcFlags, 'driftfix=slew' if $tdf;
2876
2877 if ($nokvm) {
2878 push @$machineFlags, 'accel=tcg';
2879 } else {
2880 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2881 }
2882
2883 my $machine_type = $forcemachine || $conf->{machine};
2884 if ($machine_type) {
2885 push @$machineFlags, "type=${machine_type}";
2886 }
2887
2888 if ($conf->{startdate}) {
2889 push @$rtcFlags, "base=$conf->{startdate}";
2890 } elsif ($useLocaltime) {
2891 push @$rtcFlags, 'base=localtime';
2892 }
2893
2894 my $cpu = $nokvm ? "qemu64" : "kvm64";
2895 $cpu = $conf->{cpu} if $conf->{cpu};
2896
2897 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2898
2899 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2900
2901 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2902
2903 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2904
2905 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2906
2907 # Note: enforce needs kernel 3.10, so we do not use it for now
2908 # push @$cmd, '-cpu', "$cpu,enforce";
2909 push @$cmd, '-cpu', $cpu;
2910
2911 my $memory = $conf->{memory} || $defaults->{memory};
2912 my $static_memory = 0;
2913 my $dimm_memory = 0;
2914
2915 if ($hotplug_features->{memory}) {
2916 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2917 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2918 $static_memory = $STATICMEM;
2919 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2920 $dimm_memory = $memory - $static_memory;
2921 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2922
2923 } else {
2924
2925 $static_memory = $memory;
2926 push @$cmd, '-m', $static_memory;
2927 }
2928
2929 if ($conf->{numa}) {
2930
2931 my $numa_totalmemory = undef;
2932 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2933 next if !$conf->{"numa$i"};
2934 my $numa = parse_numa($conf->{"numa$i"});
2935 next if !$numa;
2936 # memory
2937 die "missing numa node$i memory value\n" if !$numa->{memory};
2938 my $numa_memory = $numa->{memory};
2939 $numa_totalmemory += $numa_memory;
2940 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2941
2942 # cpus
2943 my $cpus_start = $numa->{cpus}->{start};
2944 die "missing numa node$i cpus\n" if !defined($cpus_start);
2945 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2946 my $cpus = $cpus_start;
2947 if (defined($cpus_end)) {
2948 $cpus .= "-$cpus_end";
2949 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2950 }
2951
2952 # hostnodes
2953 my $hostnodes_start = $numa->{hostnodes}->{start};
2954 if (defined($hostnodes_start)) {
2955 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2956 my $hostnodes = $hostnodes_start;
2957 if (defined($hostnodes_end)) {
2958 $hostnodes .= "-$hostnodes_end";
2959 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2960 }
2961
2962 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2963 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2964 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2965 }
2966
2967 # policy
2968 my $policy = $numa->{policy};
2969 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2970 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2971 }
2972
2973 push @$cmd, '-object', $numa_object;
2974 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2975 }
2976
2977 die "total memory for NUMA nodes must be equal to vm static memory\n"
2978 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2979
2980 #if no custom tology, we split memory and cores across numa nodes
2981 if(!$numa_totalmemory) {
2982
2983 my $numa_memory = ($static_memory / $sockets) . "M";
2984
2985 for (my $i = 0; $i < $sockets; $i++) {
2986
2987 my $cpustart = ($cores * $i);
2988 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2989 my $cpus = $cpustart;
2990 $cpus .= "-$cpuend" if $cpuend;
2991
2992 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2993 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2994 }
2995 }
2996 }
2997
2998 if ($hotplug_features->{memory}) {
2999 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
3000 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3001 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3002 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3003
3004 #if dimm_memory is not aligned to dimm map
3005 if($current_size > $memory) {
3006 $conf->{memory} = $current_size;
3007 update_config_nolock($vmid, $conf, 1);
3008 }
3009 });
3010 }
3011
3012 push @$cmd, '-S' if $conf->{freeze};
3013
3014 # set keyboard layout
3015 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3016 push @$cmd, '-k', $kb if $kb;
3017
3018 # enable sound
3019 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3020 #push @$cmd, '-soundhw', 'es1370';
3021 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3022
3023 if($conf->{agent}) {
3024 my $qgasocket = qmp_socket($vmid, 1);
3025 my $pciaddr = print_pci_addr("qga0", $bridges);
3026 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3027 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3028 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3029 }
3030
3031 my $spice_port;
3032
3033 if ($qxlnum) {
3034 if ($qxlnum > 1) {
3035 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3036 for(my $i = 1; $i < $qxlnum; $i++){
3037 my $pciaddr = print_pci_addr("vga$i", $bridges);
3038 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3039 }
3040 } else {
3041 # assume other OS works like Linux
3042 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3043 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3044 }
3045 }
3046
3047 my $pciaddr = print_pci_addr("spice", $bridges);
3048
3049 $spice_port = PVE::Tools::next_spice_port();
3050
3051 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3052
3053 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3054 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3055 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3056 }
3057
3058 # enable balloon by default, unless explicitly disabled
3059 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3060 $pciaddr = print_pci_addr("balloon0", $bridges);
3061 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3062 }
3063
3064 if ($conf->{watchdog}) {
3065 my $wdopts = parse_watchdog($conf->{watchdog});
3066 $pciaddr = print_pci_addr("watchdog", $bridges);
3067 my $watchdog = $wdopts->{model} || 'i6300esb';
3068 push @$devices, '-device', "$watchdog$pciaddr";
3069 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3070 }
3071
3072 my $vollist = [];
3073 my $scsicontroller = {};
3074 my $ahcicontroller = {};
3075 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3076
3077 # Add iscsi initiator name if available
3078 if (my $initiator = get_initiator_name()) {
3079 push @$devices, '-iscsi', "initiator-name=$initiator";
3080 }
3081
3082 foreach_drive($conf, sub {
3083 my ($ds, $drive) = @_;
3084
3085 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3086 push @$vollist, $drive->{file};
3087 }
3088
3089 $use_virtio = 1 if $ds =~ m/^virtio/;
3090
3091 if (drive_is_cdrom ($drive)) {
3092 if ($bootindex_hash->{d}) {
3093 $drive->{bootindex} = $bootindex_hash->{d};
3094 $bootindex_hash->{d} += 1;
3095 }
3096 } else {
3097 if ($bootindex_hash->{c}) {
3098 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3099 $bootindex_hash->{c} += 1;
3100 }
3101 }
3102
3103 if ($drive->{interface} eq 'scsi') {
3104
3105 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3106 my $controller = int($drive->{index} / $maxdev);
3107 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3108 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3109 $scsicontroller->{$controller}=1;
3110 }
3111
3112 if ($drive->{interface} eq 'sata') {
3113 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3114 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3115 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3116 $ahcicontroller->{$controller}=1;
3117 }
3118
3119 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3120 push @$devices, '-drive',$drive_cmd;
3121 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3122 });
3123
3124 for (my $i = 0; $i < $MAX_NETS; $i++) {
3125 next if !$conf->{"net$i"};
3126 my $d = parse_net($conf->{"net$i"});
3127 next if !$d;
3128
3129 $use_virtio = 1 if $d->{model} eq 'virtio';
3130
3131 if ($bootindex_hash->{n}) {
3132 $d->{bootindex} = $bootindex_hash->{n};
3133 $bootindex_hash->{n} += 1;
3134 }
3135
3136 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3137 push @$devices, '-netdev', $netdevfull;
3138
3139 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3140 push @$devices, '-device', $netdevicefull;
3141 }
3142
3143 if (!$q35) {
3144 # add pci bridges
3145 while (my ($k, $v) = each %$bridges) {
3146 $pciaddr = print_pci_addr("pci.$k");
3147 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3148 }
3149 }
3150
3151 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3152 # when the VM uses virtio devices.
3153 if (!$use_virtio && $have_ovz) {
3154
3155 my $cpuunits = defined($conf->{cpuunits}) ?
3156 $conf->{cpuunits} : $defaults->{cpuunits};
3157
3158 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3159
3160 # fixme: cpulimit is currently ignored
3161 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3162 }
3163
3164 # add custom args
3165 if ($conf->{args}) {
3166 my $aa = PVE::Tools::split_args($conf->{args});
3167 push @$cmd, @$aa;
3168 }
3169
3170 push @$cmd, @$devices;
3171 push @$cmd, '-rtc', join(',', @$rtcFlags)
3172 if scalar(@$rtcFlags);
3173 push @$cmd, '-machine', join(',', @$machineFlags)
3174 if scalar(@$machineFlags);
3175 push @$cmd, '-global', join(',', @$globalFlags)
3176 if scalar(@$globalFlags);
3177
3178 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3179 }
3180
3181 sub vnc_socket {
3182 my ($vmid) = @_;
3183 return "${var_run_tmpdir}/$vmid.vnc";
3184 }
3185
3186 sub spice_port {
3187 my ($vmid) = @_;
3188
3189 my $res = vm_mon_cmd($vmid, 'query-spice');
3190
3191 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3192 }
3193
3194 sub qmp_socket {
3195 my ($vmid, $qga) = @_;
3196 my $sockettype = $qga ? 'qga' : 'qmp';
3197 return "${var_run_tmpdir}/$vmid.$sockettype";
3198 }
3199
3200 sub pidfile_name {
3201 my ($vmid) = @_;
3202 return "${var_run_tmpdir}/$vmid.pid";
3203 }
3204
3205 sub vm_devices_list {
3206 my ($vmid) = @_;
3207
3208 my $res = vm_mon_cmd($vmid, 'query-pci');
3209 my $devices = {};
3210 foreach my $pcibus (@$res) {
3211 foreach my $device (@{$pcibus->{devices}}) {
3212 next if !$device->{'qdev_id'};
3213 if ($device->{'pci_bridge'}) {
3214 $devices->{$device->{'qdev_id'}} = 1;
3215 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3216 next if !$bridge_device->{'qdev_id'};
3217 $devices->{$bridge_device->{'qdev_id'}} = 1;
3218 $devices->{$device->{'qdev_id'}}++;
3219 }
3220 } else {
3221 $devices->{$device->{'qdev_id'}} = 1;
3222 }
3223 }
3224 }
3225
3226 my $resblock = vm_mon_cmd($vmid, 'query-block');
3227 foreach my $block (@$resblock) {
3228 if($block->{device} =~ m/^drive-(\S+)/){
3229 $devices->{$1} = 1;
3230 }
3231 }
3232
3233 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3234 foreach my $mice (@$resmice) {
3235 if ($mice->{name} eq 'QEMU HID Tablet') {
3236 $devices->{tablet} = 1;
3237 last;
3238 }
3239 }
3240
3241 return $devices;
3242 }
3243
3244 sub vm_deviceplug {
3245 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3246
3247 my $q35 = machine_type_is_q35($conf);
3248
3249 my $devices_list = vm_devices_list($vmid);
3250 return 1 if defined($devices_list->{$deviceid});
3251
3252 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3253
3254 if ($deviceid eq 'tablet') {
3255
3256 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3257
3258 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3259
3260 qemu_driveadd($storecfg, $vmid, $device);
3261 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3262
3263 qemu_deviceadd($vmid, $devicefull);
3264 eval { qemu_deviceaddverify($vmid, $deviceid); };
3265 if (my $err = $@) {
3266 eval { qemu_drivedel($vmid, $deviceid); };
3267 warn $@ if $@;
3268 die $err;
3269 }
3270
3271 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3272
3273 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3274 my $pciaddr = print_pci_addr($deviceid);
3275 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3276
3277 qemu_deviceadd($vmid, $devicefull);
3278 qemu_deviceaddverify($vmid, $deviceid);
3279
3280 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3281
3282 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3283 qemu_driveadd($storecfg, $vmid, $device);
3284
3285 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3286 eval { qemu_deviceadd($vmid, $devicefull); };
3287 if (my $err = $@) {
3288 eval { qemu_drivedel($vmid, $deviceid); };
3289 warn $@ if $@;
3290 die $err;
3291 }
3292
3293 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3294
3295 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3296 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3297 qemu_deviceadd($vmid, $netdevicefull);
3298 eval { qemu_deviceaddverify($vmid, $deviceid); };
3299 if (my $err = $@) {
3300 eval { qemu_netdevdel($vmid, $deviceid); };
3301 warn $@ if $@;
3302 die $err;
3303 }
3304
3305 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3306
3307 my $bridgeid = $2;
3308 my $pciaddr = print_pci_addr($deviceid);
3309 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3310
3311 qemu_deviceadd($vmid, $devicefull);
3312 qemu_deviceaddverify($vmid, $deviceid);
3313
3314 } else {
3315 die "can't hotplug device '$deviceid'\n";
3316 }
3317
3318 return 1;
3319 }
3320
3321 # fixme: this should raise exceptions on error!
3322 sub vm_deviceunplug {
3323 my ($vmid, $conf, $deviceid) = @_;
3324
3325 my $devices_list = vm_devices_list($vmid);
3326 return 1 if !defined($devices_list->{$deviceid});
3327
3328 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3329
3330 if ($deviceid eq 'tablet') {
3331
3332 qemu_devicedel($vmid, $deviceid);
3333
3334 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3335
3336 qemu_devicedel($vmid, $deviceid);
3337 qemu_devicedelverify($vmid, $deviceid);
3338 qemu_drivedel($vmid, $deviceid);
3339
3340 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3341
3342 qemu_devicedel($vmid, $deviceid);
3343
3344 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3345
3346 qemu_devicedel($vmid, $deviceid);
3347 qemu_drivedel($vmid, $deviceid);
3348
3349 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3350
3351 qemu_devicedel($vmid, $deviceid);
3352 qemu_devicedelverify($vmid, $deviceid);
3353 qemu_netdevdel($vmid, $deviceid);
3354
3355 } else {
3356 die "can't unplug device '$deviceid'\n";
3357 }
3358
3359 return 1;
3360 }
3361
3362 sub qemu_deviceadd {
3363 my ($vmid, $devicefull) = @_;
3364
3365 $devicefull = "driver=".$devicefull;
3366 my %options = split(/[=,]/, $devicefull);
3367
3368 vm_mon_cmd($vmid, "device_add" , %options);
3369 }
3370
3371 sub qemu_devicedel {
3372 my ($vmid, $deviceid) = @_;
3373
3374 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3375 }
3376
3377 sub qemu_objectadd {
3378 my($vmid, $objectid, $qomtype) = @_;
3379
3380 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3381
3382 return 1;
3383 }
3384
3385 sub qemu_objectdel {
3386 my($vmid, $objectid) = @_;
3387
3388 vm_mon_cmd($vmid, "object-del", id => $objectid);
3389
3390 return 1;
3391 }
3392
3393 sub qemu_driveadd {
3394 my ($storecfg, $vmid, $device) = @_;
3395
3396 my $drive = print_drive_full($storecfg, $vmid, $device);
3397 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3398
3399 # If the command succeeds qemu prints: "OK"
3400 return 1 if $ret =~ m/OK/s;
3401
3402 die "adding drive failed: $ret\n";
3403 }
3404
3405 sub qemu_drivedel {
3406 my($vmid, $deviceid) = @_;
3407
3408 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3409 $ret =~ s/^\s+//;
3410
3411 return 1 if $ret eq "";
3412
3413 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3414 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3415
3416 die "deleting drive $deviceid failed : $ret\n";
3417 }
3418
3419 sub qemu_deviceaddverify {
3420 my ($vmid, $deviceid) = @_;
3421
3422 for (my $i = 0; $i <= 5; $i++) {
3423 my $devices_list = vm_devices_list($vmid);
3424 return 1 if defined($devices_list->{$deviceid});
3425 sleep 1;
3426 }
3427
3428 die "error on hotplug device '$deviceid'\n";
3429 }
3430
3431
3432 sub qemu_devicedelverify {
3433 my ($vmid, $deviceid) = @_;
3434
3435 # need to verify that the device is correctly removed as device_del
3436 # is async and empty return is not reliable
3437
3438 for (my $i = 0; $i <= 5; $i++) {
3439 my $devices_list = vm_devices_list($vmid);
3440 return 1 if !defined($devices_list->{$deviceid});
3441 sleep 1;
3442 }
3443
3444 die "error on hot-unplugging device '$deviceid'\n";
3445 }
3446
3447 sub qemu_findorcreatescsihw {
3448 my ($storecfg, $conf, $vmid, $device) = @_;
3449
3450 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3451 my $controller = int($device->{index} / $maxdev);
3452 my $scsihwid="scsihw$controller";
3453 my $devices_list = vm_devices_list($vmid);
3454
3455 if(!defined($devices_list->{$scsihwid})) {
3456 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3457 }
3458
3459 return 1;
3460 }
3461
3462 sub qemu_add_pci_bridge {
3463 my ($storecfg, $conf, $vmid, $device) = @_;
3464
3465 my $bridges = {};
3466
3467 my $bridgeid;
3468
3469 print_pci_addr($device, $bridges);
3470
3471 while (my ($k, $v) = each %$bridges) {
3472 $bridgeid = $k;
3473 }
3474 return 1 if !defined($bridgeid) || $bridgeid < 1;
3475
3476 my $bridge = "pci.$bridgeid";
3477 my $devices_list = vm_devices_list($vmid);
3478
3479 if (!defined($devices_list->{$bridge})) {
3480 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3481 }
3482
3483 return 1;
3484 }
3485
3486 sub qemu_set_link_status {
3487 my ($vmid, $device, $up) = @_;
3488
3489 vm_mon_cmd($vmid, "set_link", name => $device,
3490 up => $up ? JSON::true : JSON::false);
3491 }
3492
3493 sub qemu_netdevadd {
3494 my ($vmid, $conf, $device, $deviceid) = @_;
3495
3496 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3497 my %options = split(/[=,]/, $netdev);
3498
3499 vm_mon_cmd($vmid, "netdev_add", %options);
3500 return 1;
3501 }
3502
3503 sub qemu_netdevdel {
3504 my ($vmid, $deviceid) = @_;
3505
3506 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3507 }
3508
3509 sub qemu_cpu_hotplug {
3510 my ($vmid, $conf, $vcpus) = @_;
3511
3512 my $sockets = 1;
3513 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3514 $sockets = $conf->{sockets} if $conf->{sockets};
3515 my $cores = $conf->{cores} || 1;
3516 my $maxcpus = $sockets * $cores;
3517
3518 $vcpus = $maxcpus if !$vcpus;
3519
3520 die "you can't add more vcpus than maxcpus\n"
3521 if $vcpus > $maxcpus;
3522
3523 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3524 die "online cpu unplug is not yet possible\n"
3525 if $vcpus < $currentvcpus;
3526
3527 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3528 die "vcpus in running vm is different than configuration\n"
3529 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3530
3531 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3532 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3533 }
3534 }
3535
3536 sub qemu_memory_hotplug {
3537 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3538
3539 return $value if !check_running($vmid);
3540
3541 my $memory = $conf->{memory} || $defaults->{memory};
3542 $value = $defaults->{memory} if !$value;
3543 return $value if $value == $memory;
3544
3545 my $static_memory = $STATICMEM;
3546 my $dimm_memory = $memory - $static_memory;
3547
3548 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3549 die "memory unplug is not yet available" if $value < $memory;
3550 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3551
3552
3553 my $sockets = 1;
3554 $sockets = $conf->{sockets} if $conf->{sockets};
3555
3556 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3557 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3558
3559 return if $current_size <= $conf->{memory};
3560
3561 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3562 if (my $err = $@) {
3563 eval { qemu_objectdel($vmid, "mem-$name"); };
3564 die $err;
3565 }
3566
3567 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3568 if (my $err = $@) {
3569 eval { qemu_objectdel($vmid, "mem-$name"); };
3570 die $err;
3571 }
3572 #update conf after each succesful module hotplug
3573 $conf->{memory} = $current_size;
3574 update_config_nolock($vmid, $conf, 1);
3575 });
3576 }
3577
3578 sub qemu_block_set_io_throttle {
3579 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3580
3581 return if !check_running($vmid) ;
3582
3583 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($iops_wr));
3584
3585 }
3586
3587 # old code, only used to shutdown old VM after update
3588 sub __read_avail {
3589 my ($fh, $timeout) = @_;
3590
3591 my $sel = new IO::Select;
3592 $sel->add($fh);
3593
3594 my $res = '';
3595 my $buf;
3596
3597 my @ready;
3598 while (scalar (@ready = $sel->can_read($timeout))) {
3599 my $count;
3600 if ($count = $fh->sysread($buf, 8192)) {
3601 if ($buf =~ /^(.*)\(qemu\) $/s) {
3602 $res .= $1;
3603 last;
3604 } else {
3605 $res .= $buf;
3606 }
3607 } else {
3608 if (!defined($count)) {
3609 die "$!\n";
3610 }
3611 last;
3612 }
3613 }
3614
3615 die "monitor read timeout\n" if !scalar(@ready);
3616
3617 return $res;
3618 }
3619
3620 # old code, only used to shutdown old VM after update
3621 sub vm_monitor_command {
3622 my ($vmid, $cmdstr, $nocheck) = @_;
3623
3624 my $res;
3625
3626 eval {
3627 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3628
3629 my $sname = "${var_run_tmpdir}/$vmid.mon";
3630
3631 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3632 die "unable to connect to VM $vmid socket - $!\n";
3633
3634 my $timeout = 3;
3635
3636 # hack: migrate sometime blocks the monitor (when migrate_downtime
3637 # is set)
3638 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3639 $timeout = 60*60; # 1 hour
3640 }
3641
3642 # read banner;
3643 my $data = __read_avail($sock, $timeout);
3644
3645 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3646 die "got unexpected qemu monitor banner\n";
3647 }
3648
3649 my $sel = new IO::Select;
3650 $sel->add($sock);
3651
3652 if (!scalar(my @ready = $sel->can_write($timeout))) {
3653 die "monitor write error - timeout";
3654 }
3655
3656 my $fullcmd = "$cmdstr\r";
3657
3658 # syslog('info', "VM $vmid monitor command: $cmdstr");
3659
3660 my $b;
3661 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3662 die "monitor write error - $!";
3663 }
3664
3665 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3666
3667 $timeout = 20;
3668
3669 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3670 $timeout = 60*60; # 1 hour
3671 } elsif ($cmdstr =~ m/^(eject|change)/) {
3672 $timeout = 60; # note: cdrom mount command is slow
3673 }
3674 if ($res = __read_avail($sock, $timeout)) {
3675
3676 my @lines = split("\r?\n", $res);
3677
3678 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3679
3680 $res = join("\n", @lines);
3681 $res .= "\n";
3682 }
3683 };
3684
3685 my $err = $@;
3686
3687 if ($err) {
3688 syslog("err", "VM $vmid monitor command failed - $err");
3689 die $err;
3690 }
3691
3692 return $res;
3693 }
3694
3695 sub qemu_block_resize {
3696 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3697
3698 my $running = check_running($vmid);
3699
3700 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3701
3702 return if !$running;
3703
3704 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3705
3706 }
3707
3708 sub qemu_volume_snapshot {
3709 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3710
3711 my $running = check_running($vmid);
3712
3713 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3714
3715 return if !$running;
3716
3717 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3718
3719 }
3720
3721 sub qemu_volume_snapshot_delete {
3722 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3723
3724 my $running = check_running($vmid);
3725
3726 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3727
3728 return if !$running;
3729
3730 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3731 }
3732
3733 sub set_migration_caps {
3734 my ($vmid) = @_;
3735
3736 my $cap_ref = [];
3737
3738 my $enabled_cap = {
3739 "auto-converge" => 1,
3740 "xbzrle" => 0,
3741 "x-rdma-pin-all" => 0,
3742 "zero-blocks" => 0,
3743 };
3744
3745 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3746
3747 for my $supported_capability (@$supported_capabilities) {
3748 push @$cap_ref, {
3749 capability => $supported_capability->{capability},
3750 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3751 };
3752 }
3753
3754 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3755 }
3756
3757 my $fast_plug_option = {
3758 'lock' => 1,
3759 'name' => 1,
3760 'onboot' => 1,
3761 'shares' => 1,
3762 'startup' => 1,
3763 };
3764
3765 # hotplug changes in [PENDING]
3766 # $selection hash can be used to only apply specified options, for
3767 # example: { cores => 1 } (only apply changed 'cores')
3768 # $errors ref is used to return error messages
3769 sub vmconfig_hotplug_pending {
3770 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3771
3772 my $defaults = load_defaults();
3773
3774 # commit values which do not have any impact on running VM first
3775 # Note: those option cannot raise errors, we we do not care about
3776 # $selection and always apply them.
3777
3778 my $add_error = sub {
3779 my ($opt, $msg) = @_;
3780 $errors->{$opt} = "hotplug problem - $msg";
3781 };
3782
3783 my $changes = 0;
3784 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3785 if ($fast_plug_option->{$opt}) {
3786 $conf->{$opt} = $conf->{pending}->{$opt};
3787 delete $conf->{pending}->{$opt};
3788 $changes = 1;
3789 }
3790 }
3791
3792 if ($changes) {
3793 update_config_nolock($vmid, $conf, 1);
3794 $conf = load_config($vmid); # update/reload
3795 }
3796
3797 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3798
3799 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3800 foreach my $opt (@delete) {
3801 next if $selection && !$selection->{$opt};
3802 eval {
3803 if ($opt eq 'hotplug') {
3804 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3805 } elsif ($opt eq 'tablet') {
3806 die "skip\n" if !$hotplug_features->{usb};
3807 if ($defaults->{tablet}) {
3808 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3809 } else {
3810 vm_deviceunplug($vmid, $conf, $opt);
3811 }
3812 } elsif ($opt eq 'vcpus') {
3813 die "skip\n" if !$hotplug_features->{cpu};
3814 qemu_cpu_hotplug($vmid, $conf, undef);
3815 } elsif ($opt eq 'balloon') {
3816 # enable balloon device is not hotpluggable
3817 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3818 } elsif ($fast_plug_option->{$opt}) {
3819 # do nothing
3820 } elsif ($opt =~ m/^net(\d+)$/) {
3821 die "skip\n" if !$hotplug_features->{network};
3822 vm_deviceunplug($vmid, $conf, $opt);
3823 } elsif (valid_drivename($opt)) {
3824 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3825 vm_deviceunplug($vmid, $conf, $opt);
3826 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3827 } elsif ($opt =~ m/^memory$/) {
3828 die "skip\n" if !$hotplug_features->{memory};
3829 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3830 } else {
3831 die "skip\n";
3832 }
3833 };
3834 if (my $err = $@) {
3835 &$add_error($opt, $err) if $err ne "skip\n";
3836 } else {
3837 # save new config if hotplug was successful
3838 delete $conf->{$opt};
3839 vmconfig_undelete_pending_option($conf, $opt);
3840 update_config_nolock($vmid, $conf, 1);
3841 $conf = load_config($vmid); # update/reload
3842 }
3843 }
3844
3845 foreach my $opt (keys %{$conf->{pending}}) {
3846 next if $selection && !$selection->{$opt};
3847 my $value = $conf->{pending}->{$opt};
3848 eval {
3849 if ($opt eq 'hotplug') {
3850 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3851 } elsif ($opt eq 'tablet') {
3852 die "skip\n" if !$hotplug_features->{usb};
3853 if ($value == 1) {
3854 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3855 } elsif ($value == 0) {
3856 vm_deviceunplug($vmid, $conf, $opt);
3857 }
3858 } elsif ($opt eq 'vcpus') {
3859 die "skip\n" if !$hotplug_features->{cpu};
3860 qemu_cpu_hotplug($vmid, $conf, $value);
3861 } elsif ($opt eq 'balloon') {
3862 # enable/disable balloning device is not hotpluggable
3863 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3864 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3865 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3866
3867 # allow manual ballooning if shares is set to zero
3868 if (!(defined($conf->{shares}) && ($conf->{shares} == 0))) {
3869 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3870 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3871 }
3872 } elsif ($opt =~ m/^net(\d+)$/) {
3873 # some changes can be done without hotplug
3874 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3875 $vmid, $opt, $value);
3876 } elsif (valid_drivename($opt)) {
3877 # some changes can be done without hotplug
3878 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3879 $vmid, $opt, $value, 1);
3880 } elsif ($opt =~ m/^memory$/) { #dimms
3881 die "skip\n" if !$hotplug_features->{memory};
3882 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3883 } else {
3884 die "skip\n"; # skip non-hot-pluggable options
3885 }
3886 };
3887 if (my $err = $@) {
3888 &$add_error($opt, $err) if $err ne "skip\n";
3889 } else {
3890 # save new config if hotplug was successful
3891 $conf->{$opt} = $value;
3892 delete $conf->{pending}->{$opt};
3893 update_config_nolock($vmid, $conf, 1);
3894 $conf = load_config($vmid); # update/reload
3895 }
3896 }
3897 }
3898
3899 sub vmconfig_apply_pending {
3900 my ($vmid, $conf, $storecfg) = @_;
3901
3902 # cold plug
3903
3904 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3905 foreach my $opt (@delete) { # delete
3906 die "internal error" if $opt =~ m/^unused/;
3907 $conf = load_config($vmid); # update/reload
3908 if (!defined($conf->{$opt})) {
3909 vmconfig_undelete_pending_option($conf, $opt);
3910 update_config_nolock($vmid, $conf, 1);
3911 } elsif (valid_drivename($opt)) {
3912 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3913 vmconfig_undelete_pending_option($conf, $opt);
3914 delete $conf->{$opt};
3915 update_config_nolock($vmid, $conf, 1);
3916 } else {
3917 vmconfig_undelete_pending_option($conf, $opt);
3918 delete $conf->{$opt};
3919 update_config_nolock($vmid, $conf, 1);
3920 }
3921 }
3922
3923 $conf = load_config($vmid); # update/reload
3924
3925 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3926 $conf = load_config($vmid); # update/reload
3927
3928 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3929 # skip if nothing changed
3930 } elsif (valid_drivename($opt)) {
3931 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3932 if defined($conf->{$opt});
3933 $conf->{$opt} = $conf->{pending}->{$opt};
3934 } else {
3935 $conf->{$opt} = $conf->{pending}->{$opt};
3936 }
3937
3938 delete $conf->{pending}->{$opt};
3939 update_config_nolock($vmid, $conf, 1);
3940 }
3941 }
3942
3943 my $safe_num_ne = sub {
3944 my ($a, $b) = @_;
3945
3946 return 0 if !defined($a) && !defined($b);
3947 return 1 if !defined($a);
3948 return 1 if !defined($b);
3949
3950 return $a != $b;
3951 };
3952
3953 my $safe_string_ne = sub {
3954 my ($a, $b) = @_;
3955
3956 return 0 if !defined($a) && !defined($b);
3957 return 1 if !defined($a);
3958 return 1 if !defined($b);
3959
3960 return $a ne $b;
3961 };
3962
3963 sub vmconfig_update_net {
3964 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3965
3966 my $newnet = parse_net($value);
3967
3968 if ($conf->{$opt}) {
3969 my $oldnet = parse_net($conf->{$opt});
3970
3971 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3972 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3973 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3974 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3975
3976 # for non online change, we try to hot-unplug
3977 die "skip\n" if !$hotplug;
3978 vm_deviceunplug($vmid, $conf, $opt);
3979 } else {
3980
3981 die "internal error" if $opt !~ m/net(\d+)/;
3982 my $iface = "tap${vmid}i$1";
3983
3984 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3985 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3986 }
3987
3988 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3989 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3990 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3991 PVE::Network::tap_unplug($iface);
3992 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3993 }
3994
3995 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3996 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3997 }
3998
3999 return 1;
4000 }
4001 }
4002
4003 if ($hotplug) {
4004 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4005 } else {
4006 die "skip\n";
4007 }
4008 }
4009
4010 sub vmconfig_update_disk {
4011 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4012
4013 # fixme: do we need force?
4014
4015 my $drive = parse_drive($opt, $value);
4016
4017 if ($conf->{$opt}) {
4018
4019 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4020
4021 my $media = $drive->{media} || 'disk';
4022 my $oldmedia = $old_drive->{media} || 'disk';
4023 die "unable to change media type\n" if $media ne $oldmedia;
4024
4025 if (!drive_is_cdrom($old_drive)) {
4026
4027 if ($drive->{file} ne $old_drive->{file}) {
4028
4029 die "skip\n" if !$hotplug;
4030
4031 # unplug and register as unused
4032 vm_deviceunplug($vmid, $conf, $opt);
4033 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4034
4035 } else {
4036 # update existing disk
4037
4038 # skip non hotpluggable value
4039 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4040 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4041 die "skip\n";
4042 }
4043
4044 # apply throttle
4045 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4046 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4047 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4048 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4049 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4050 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4051 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4052 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4053 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4054 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4055 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4056 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4057
4058 qemu_block_set_io_throttle($vmid,"drive-$opt",
4059 ($drive->{mbps} || 0)*1024*1024,
4060 ($drive->{mbps_rd} || 0)*1024*1024,
4061 ($drive->{mbps_wr} || 0)*1024*1024,
4062 $drive->{iops} || 0,
4063 $drive->{iops_rd} || 0,
4064 $drive->{iops_wr} || 0,
4065 ($drive->{mbps_max} || 0)*1024*1024,
4066 ($drive->{mbps_rd_max} || 0)*1024*1024,
4067 ($drive->{mbps_wr_max} || 0)*1024*1024,
4068 $drive->{iops_max} || 0,
4069 $drive->{iops_rd_max} || 0,
4070 $drive->{iops_wr_max} || 0);
4071
4072 }
4073
4074 return 1;
4075 }
4076
4077 } else { # cdrom
4078
4079 if ($drive->{file} eq 'none') {
4080 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4081 } else {
4082 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4083 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4084 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4085 }
4086
4087 return 1;
4088 }
4089 }
4090 }
4091
4092 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4093 # hotplug new disks
4094 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4095 }
4096
4097 sub vm_start {
4098 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4099
4100 lock_config($vmid, sub {
4101 my $conf = load_config($vmid, $migratedfrom);
4102
4103 die "you can't start a vm if it's a template\n" if is_template($conf);
4104
4105 check_lock($conf) if !$skiplock;
4106
4107 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4108
4109 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4110 vmconfig_apply_pending($vmid, $conf, $storecfg);
4111 $conf = load_config($vmid); # update/reload
4112 }
4113
4114 my $defaults = load_defaults();
4115
4116 # set environment variable useful inside network script
4117 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4118
4119 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4120
4121 my $migrate_port = 0;
4122 my $migrate_uri;
4123 if ($statefile) {
4124 if ($statefile eq 'tcp') {
4125 my $localip = "localhost";
4126 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4127 if ($datacenterconf->{migration_unsecure}) {
4128 my $nodename = PVE::INotify::nodename();
4129 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4130 }
4131 $migrate_port = PVE::Tools::next_migrate_port();
4132 $migrate_uri = "tcp:${localip}:${migrate_port}";
4133 push @$cmd, '-incoming', $migrate_uri;
4134 push @$cmd, '-S';
4135 } else {
4136 push @$cmd, '-loadstate', $statefile;
4137 }
4138 } elsif ($paused) {
4139 push @$cmd, '-S';
4140 }
4141
4142 # host pci devices
4143 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4144 my $d = parse_hostpci($conf->{"hostpci$i"});
4145 next if !$d;
4146 my $pcidevices = $d->{pciid};
4147 foreach my $pcidevice (@$pcidevices) {
4148 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4149
4150 my $info = pci_device_info("0000:$pciid");
4151 die "IOMMU not present\n" if !check_iommu_support();
4152 die "no pci device info for device '$pciid'\n" if !$info;
4153
4154 if ($d->{driver} && $d->{driver} eq "vfio") {
4155 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4156 } else {
4157 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4158 }
4159
4160 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4161 }
4162 }
4163
4164 PVE::Storage::activate_volumes($storecfg, $vollist);
4165
4166 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4167 umask => 0077); };
4168 my $err = $@;
4169 die "start failed: $err" if $err;
4170
4171 print "migration listens on $migrate_uri\n" if $migrate_uri;
4172
4173 if ($statefile && $statefile ne 'tcp') {
4174 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4175 warn $@ if $@;
4176 }
4177
4178 if ($migratedfrom) {
4179
4180 eval {
4181 set_migration_caps($vmid);
4182 };
4183 warn $@ if $@;
4184
4185 if ($spice_port) {
4186 print "spice listens on port $spice_port\n";
4187 if ($spice_ticket) {
4188 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4189 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4190 }
4191 }
4192
4193 } else {
4194
4195 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4196 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4197 if $conf->{balloon};
4198 vm_mon_cmd_nocheck($vmid, 'qom-set',
4199 path => "machine/peripheral/balloon0",
4200 property => "guest-stats-polling-interval",
4201 value => 2);
4202 }
4203
4204 foreach my $opt (keys %$conf) {
4205 next if $opt !~ m/^net\d+$/;
4206 my $nicconf = parse_net($conf->{$opt});
4207 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4208 }
4209 }
4210 });
4211 }
4212
4213 sub vm_mon_cmd {
4214 my ($vmid, $execute, %params) = @_;
4215
4216 my $cmd = { execute => $execute, arguments => \%params };
4217 vm_qmp_command($vmid, $cmd);
4218 }
4219
4220 sub vm_mon_cmd_nocheck {
4221 my ($vmid, $execute, %params) = @_;
4222
4223 my $cmd = { execute => $execute, arguments => \%params };
4224 vm_qmp_command($vmid, $cmd, 1);
4225 }
4226
4227 sub vm_qmp_command {
4228 my ($vmid, $cmd, $nocheck) = @_;
4229
4230 my $res;
4231
4232 my $timeout;
4233 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4234 $timeout = $cmd->{arguments}->{timeout};
4235 delete $cmd->{arguments}->{timeout};
4236 }
4237
4238 eval {
4239 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4240 my $sname = qmp_socket($vmid);
4241 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4242 my $qmpclient = PVE::QMPClient->new();
4243
4244 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4245 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4246 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4247 if scalar(%{$cmd->{arguments}});
4248 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4249 } else {
4250 die "unable to open monitor socket\n";
4251 }
4252 };
4253 if (my $err = $@) {
4254 syslog("err", "VM $vmid qmp command failed - $err");
4255 die $err;
4256 }
4257
4258 return $res;
4259 }
4260
4261 sub vm_human_monitor_command {
4262 my ($vmid, $cmdline) = @_;
4263
4264 my $res;
4265
4266 my $cmd = {
4267 execute => 'human-monitor-command',
4268 arguments => { 'command-line' => $cmdline},
4269 };
4270
4271 return vm_qmp_command($vmid, $cmd);
4272 }
4273
4274 sub vm_commandline {
4275 my ($storecfg, $vmid) = @_;
4276
4277 my $conf = load_config($vmid);
4278
4279 my $defaults = load_defaults();
4280
4281 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4282
4283 return join(' ', @$cmd);
4284 }
4285
4286 sub vm_reset {
4287 my ($vmid, $skiplock) = @_;
4288
4289 lock_config($vmid, sub {
4290
4291 my $conf = load_config($vmid);
4292
4293 check_lock($conf) if !$skiplock;
4294
4295 vm_mon_cmd($vmid, "system_reset");
4296 });
4297 }
4298
4299 sub get_vm_volumes {
4300 my ($conf) = @_;
4301
4302 my $vollist = [];
4303 foreach_volid($conf, sub {
4304 my ($volid, $is_cdrom) = @_;
4305
4306 return if $volid =~ m|^/|;
4307
4308 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4309 return if !$sid;
4310
4311 push @$vollist, $volid;
4312 });
4313
4314 return $vollist;
4315 }
4316
4317 sub vm_stop_cleanup {
4318 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4319
4320 eval {
4321 fairsched_rmnod($vmid); # try to destroy group
4322
4323 if (!$keepActive) {
4324 my $vollist = get_vm_volumes($conf);
4325 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4326 }
4327
4328 foreach my $ext (qw(mon qmp pid vnc qga)) {
4329 unlink "/var/run/qemu-server/${vmid}.$ext";
4330 }
4331
4332 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4333 };
4334 warn $@ if $@; # avoid errors - just warn
4335 }
4336
4337 # Note: use $nockeck to skip tests if VM configuration file exists.
4338 # We need that when migration VMs to other nodes (files already moved)
4339 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4340 sub vm_stop {
4341 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4342
4343 $force = 1 if !defined($force) && !$shutdown;
4344
4345 if ($migratedfrom){
4346 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4347 kill 15, $pid if $pid;
4348 my $conf = load_config($vmid, $migratedfrom);
4349 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4350 return;
4351 }
4352
4353 lock_config($vmid, sub {
4354
4355 my $pid = check_running($vmid, $nocheck);
4356 return if !$pid;
4357
4358 my $conf;
4359 if (!$nocheck) {
4360 $conf = load_config($vmid);
4361 check_lock($conf) if !$skiplock;
4362 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4363 my $opts = parse_startup($conf->{startup});
4364 $timeout = $opts->{down} if $opts->{down};
4365 }
4366 }
4367
4368 $timeout = 60 if !defined($timeout);
4369
4370 eval {
4371 if ($shutdown) {
4372 if (defined($conf) && $conf->{agent}) {
4373 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4374 } else {
4375 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4376 }
4377 } else {
4378 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4379 }
4380 };
4381 my $err = $@;
4382
4383 if (!$err) {
4384 my $count = 0;
4385 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4386 $count++;
4387 sleep 1;
4388 }
4389
4390 if ($count >= $timeout) {
4391 if ($force) {
4392 warn "VM still running - terminating now with SIGTERM\n";
4393 kill 15, $pid;
4394 } else {
4395 die "VM quit/powerdown failed - got timeout\n";
4396 }
4397 } else {
4398 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4399 return;
4400 }
4401 } else {
4402 if ($force) {
4403 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4404 kill 15, $pid;
4405 } else {
4406 die "VM quit/powerdown failed\n";
4407 }
4408 }
4409
4410 # wait again
4411 $timeout = 10;
4412
4413 my $count = 0;
4414 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4415 $count++;
4416 sleep 1;
4417 }
4418
4419 if ($count >= $timeout) {
4420 warn "VM still running - terminating now with SIGKILL\n";
4421 kill 9, $pid;
4422 sleep 1;
4423 }
4424
4425 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4426 });
4427 }
4428
4429 sub vm_suspend {
4430 my ($vmid, $skiplock) = @_;
4431
4432 lock_config($vmid, sub {
4433
4434 my $conf = load_config($vmid);
4435
4436 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4437
4438 vm_mon_cmd($vmid, "stop");
4439 });
4440 }
4441
4442 sub vm_resume {
4443 my ($vmid, $skiplock) = @_;
4444
4445 lock_config($vmid, sub {
4446
4447 my $conf = load_config($vmid);
4448
4449 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4450
4451 vm_mon_cmd($vmid, "cont");
4452 });
4453 }
4454
4455 sub vm_sendkey {
4456 my ($vmid, $skiplock, $key) = @_;
4457
4458 lock_config($vmid, sub {
4459
4460 my $conf = load_config($vmid);
4461
4462 # there is no qmp command, so we use the human monitor command
4463 vm_human_monitor_command($vmid, "sendkey $key");
4464 });
4465 }
4466
4467 sub vm_destroy {
4468 my ($storecfg, $vmid, $skiplock) = @_;
4469
4470 lock_config($vmid, sub {
4471
4472 my $conf = load_config($vmid);
4473
4474 check_lock($conf) if !$skiplock;
4475
4476 if (!check_running($vmid)) {
4477 fairsched_rmnod($vmid); # try to destroy group
4478 destroy_vm($storecfg, $vmid);
4479 } else {
4480 die "VM $vmid is running - destroy failed\n";
4481 }
4482 });
4483 }
4484
4485 # pci helpers
4486
4487 sub file_write {
4488 my ($filename, $buf) = @_;
4489
4490 my $fh = IO::File->new($filename, "w");
4491 return undef if !$fh;
4492
4493 my $res = print $fh $buf;
4494
4495 $fh->close();
4496
4497 return $res;
4498 }
4499
4500 sub pci_device_info {
4501 my ($name) = @_;
4502
4503 my $res;
4504
4505 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4506 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4507
4508 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4509 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4510
4511 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4512 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4513
4514 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4515 return undef if !defined($product) || $product !~ s/^0x//;
4516
4517 $res = {
4518 name => $name,
4519 vendor => $vendor,
4520 product => $product,
4521 domain => $domain,
4522 bus => $bus,
4523 slot => $slot,
4524 func => $func,
4525 irq => $irq,
4526 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4527 };
4528
4529 return $res;
4530 }
4531
4532 sub pci_dev_reset {
4533 my ($dev) = @_;
4534
4535 my $name = $dev->{name};
4536
4537 my $fn = "$pcisysfs/devices/$name/reset";
4538
4539 return file_write($fn, "1");
4540 }
4541
4542 sub pci_dev_bind_to_stub {
4543 my ($dev) = @_;
4544
4545 my $name = $dev->{name};
4546
4547 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4548 return 1 if -d $testdir;
4549
4550 my $data = "$dev->{vendor} $dev->{product}";
4551 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4552
4553 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4554 if (!file_write($fn, $name)) {
4555 return undef if -f $fn;
4556 }
4557
4558 $fn = "$pcisysfs/drivers/pci-stub/bind";
4559 if (! -d $testdir) {
4560 return undef if !file_write($fn, $name);
4561 }
4562
4563 return -d $testdir;
4564 }
4565
4566 sub pci_dev_bind_to_vfio {
4567 my ($dev) = @_;
4568
4569 my $name = $dev->{name};
4570
4571 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4572
4573 if (!-d $vfio_basedir) {
4574 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4575 }
4576 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4577
4578 my $testdir = "$vfio_basedir/$name";
4579 return 1 if -d $testdir;
4580
4581 my $data = "$dev->{vendor} $dev->{product}";
4582 return undef if !file_write("$vfio_basedir/new_id", $data);
4583
4584 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4585 if (!file_write($fn, $name)) {
4586 return undef if -f $fn;
4587 }
4588
4589 $fn = "$vfio_basedir/bind";
4590 if (! -d $testdir) {
4591 return undef if !file_write($fn, $name);
4592 }
4593
4594 return -d $testdir;
4595 }
4596
4597 sub pci_dev_group_bind_to_vfio {
4598 my ($pciid) = @_;
4599
4600 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4601
4602 if (!-d $vfio_basedir) {
4603 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4604 }
4605 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4606
4607 # get IOMMU group devices
4608 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4609 my @devs = grep /^0000:/, readdir($D);
4610 closedir($D);
4611
4612 foreach my $pciid (@devs) {
4613 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4614
4615 # pci bridges, switches or root ports are not supported
4616 # they have a pci_bus subdirectory so skip them
4617 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4618
4619 my $info = pci_device_info($1);
4620 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4621 }
4622
4623 return 1;
4624 }
4625
4626 sub print_pci_addr {
4627 my ($id, $bridges) = @_;
4628
4629 my $res = '';
4630 my $devices = {
4631 piix3 => { bus => 0, addr => 1 },
4632 #addr2 : first videocard
4633 balloon0 => { bus => 0, addr => 3 },
4634 watchdog => { bus => 0, addr => 4 },
4635 scsihw0 => { bus => 0, addr => 5 },
4636 scsihw1 => { bus => 0, addr => 6 },
4637 ahci0 => { bus => 0, addr => 7 },
4638 qga0 => { bus => 0, addr => 8 },
4639 spice => { bus => 0, addr => 9 },
4640 virtio0 => { bus => 0, addr => 10 },
4641 virtio1 => { bus => 0, addr => 11 },
4642 virtio2 => { bus => 0, addr => 12 },
4643 virtio3 => { bus => 0, addr => 13 },
4644 virtio4 => { bus => 0, addr => 14 },
4645 virtio5 => { bus => 0, addr => 15 },
4646 hostpci0 => { bus => 0, addr => 16 },
4647 hostpci1 => { bus => 0, addr => 17 },
4648 net0 => { bus => 0, addr => 18 },
4649 net1 => { bus => 0, addr => 19 },
4650 net2 => { bus => 0, addr => 20 },
4651 net3 => { bus => 0, addr => 21 },
4652 net4 => { bus => 0, addr => 22 },
4653 net5 => { bus => 0, addr => 23 },
4654 vga1 => { bus => 0, addr => 24 },
4655 vga2 => { bus => 0, addr => 25 },
4656 vga3 => { bus => 0, addr => 26 },
4657 hostpci2 => { bus => 0, addr => 27 },
4658 hostpci3 => { bus => 0, addr => 28 },
4659 #addr29 : usb-host (pve-usb.cfg)
4660 'pci.1' => { bus => 0, addr => 30 },
4661 'pci.2' => { bus => 0, addr => 31 },
4662 'net6' => { bus => 1, addr => 1 },
4663 'net7' => { bus => 1, addr => 2 },
4664 'net8' => { bus => 1, addr => 3 },
4665 'net9' => { bus => 1, addr => 4 },
4666 'net10' => { bus => 1, addr => 5 },
4667 'net11' => { bus => 1, addr => 6 },
4668 'net12' => { bus => 1, addr => 7 },
4669 'net13' => { bus => 1, addr => 8 },
4670 'net14' => { bus => 1, addr => 9 },
4671 'net15' => { bus => 1, addr => 10 },
4672 'net16' => { bus => 1, addr => 11 },
4673 'net17' => { bus => 1, addr => 12 },
4674 'net18' => { bus => 1, addr => 13 },
4675 'net19' => { bus => 1, addr => 14 },
4676 'net20' => { bus => 1, addr => 15 },
4677 'net21' => { bus => 1, addr => 16 },
4678 'net22' => { bus => 1, addr => 17 },
4679 'net23' => { bus => 1, addr => 18 },
4680 'net24' => { bus => 1, addr => 19 },
4681 'net25' => { bus => 1, addr => 20 },
4682 'net26' => { bus => 1, addr => 21 },
4683 'net27' => { bus => 1, addr => 22 },
4684 'net28' => { bus => 1, addr => 23 },
4685 'net29' => { bus => 1, addr => 24 },
4686 'net30' => { bus => 1, addr => 25 },
4687 'net31' => { bus => 1, addr => 26 },
4688 'virtio6' => { bus => 2, addr => 1 },
4689 'virtio7' => { bus => 2, addr => 2 },
4690 'virtio8' => { bus => 2, addr => 3 },
4691 'virtio9' => { bus => 2, addr => 4 },
4692 'virtio10' => { bus => 2, addr => 5 },
4693 'virtio11' => { bus => 2, addr => 6 },
4694 'virtio12' => { bus => 2, addr => 7 },
4695 'virtio13' => { bus => 2, addr => 8 },
4696 'virtio14' => { bus => 2, addr => 9 },
4697 'virtio15' => { bus => 2, addr => 10 },
4698 };
4699
4700 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4701 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4702 my $bus = $devices->{$id}->{bus};
4703 $res = ",bus=pci.$bus,addr=$addr";
4704 $bridges->{$bus} = 1 if $bridges;
4705 }
4706 return $res;
4707
4708 }
4709
4710 sub print_pcie_addr {
4711 my ($id) = @_;
4712
4713 my $res = '';
4714 my $devices = {
4715 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4716 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4717 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4718 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4719 };
4720
4721 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4722 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4723 my $bus = $devices->{$id}->{bus};
4724 $res = ",bus=$bus,addr=$addr";
4725 }
4726 return $res;
4727
4728 }
4729
4730 # vzdump restore implementaion
4731
4732 sub tar_archive_read_firstfile {
4733 my $archive = shift;
4734
4735 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4736
4737 # try to detect archive type first
4738 my $pid = open (TMP, "tar tf '$archive'|") ||
4739 die "unable to open file '$archive'\n";
4740 my $firstfile = <TMP>;
4741 kill 15, $pid;
4742 close TMP;
4743
4744 die "ERROR: archive contaions no data\n" if !$firstfile;
4745 chomp $firstfile;
4746
4747 return $firstfile;
4748 }
4749
4750 sub tar_restore_cleanup {
4751 my ($storecfg, $statfile) = @_;
4752
4753 print STDERR "starting cleanup\n";
4754
4755 if (my $fd = IO::File->new($statfile, "r")) {
4756 while (defined(my $line = <$fd>)) {
4757 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4758 my $volid = $2;
4759 eval {
4760 if ($volid =~ m|^/|) {
4761 unlink $volid || die 'unlink failed\n';
4762 } else {
4763 PVE::Storage::vdisk_free($storecfg, $volid);
4764 }
4765 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4766 };
4767 print STDERR "unable to cleanup '$volid' - $@" if $@;
4768 } else {
4769 print STDERR "unable to parse line in statfile - $line";
4770 }
4771 }
4772 $fd->close();
4773 }
4774 }
4775
4776 sub restore_archive {
4777 my ($archive, $vmid, $user, $opts) = @_;
4778
4779 my $format = $opts->{format};
4780 my $comp;
4781
4782 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4783 $format = 'tar' if !$format;
4784 $comp = 'gzip';
4785 } elsif ($archive =~ m/\.tar$/) {
4786 $format = 'tar' if !$format;
4787 } elsif ($archive =~ m/.tar.lzo$/) {
4788 $format = 'tar' if !$format;
4789 $comp = 'lzop';
4790 } elsif ($archive =~ m/\.vma$/) {
4791 $format = 'vma' if !$format;
4792 } elsif ($archive =~ m/\.vma\.gz$/) {
4793 $format = 'vma' if !$format;
4794 $comp = 'gzip';
4795 } elsif ($archive =~ m/\.vma\.lzo$/) {
4796 $format = 'vma' if !$format;
4797 $comp = 'lzop';
4798 } else {
4799 $format = 'vma' if !$format; # default
4800 }
4801
4802 # try to detect archive format
4803 if ($format eq 'tar') {
4804 return restore_tar_archive($archive, $vmid, $user, $opts);
4805 } else {
4806 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4807 }
4808 }
4809
4810 sub restore_update_config_line {
4811 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4812
4813 return if $line =~ m/^\#qmdump\#/;
4814 return if $line =~ m/^\#vzdump\#/;
4815 return if $line =~ m/^lock:/;
4816 return if $line =~ m/^unused\d+:/;
4817 return if $line =~ m/^parent:/;
4818 return if $line =~ m/^template:/; # restored VM is never a template
4819
4820 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4821 # try to convert old 1.X settings
4822 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4823 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4824 my ($model, $macaddr) = split(/\=/, $devconfig);
4825 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4826 my $net = {
4827 model => $model,
4828 bridge => "vmbr$ind",
4829 macaddr => $macaddr,
4830 };
4831 my $netstr = print_net($net);
4832
4833 print $outfd "net$cookie->{netcount}: $netstr\n";
4834 $cookie->{netcount}++;
4835 }
4836 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4837 my ($id, $netstr) = ($1, $2);
4838 my $net = parse_net($netstr);
4839 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4840 $netstr = print_net($net);
4841 print $outfd "$id: $netstr\n";
4842 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4843 my $virtdev = $1;
4844 my $value = $3;
4845 if ($line =~ m/backup=no/) {
4846 print $outfd "#$line";
4847 } elsif ($virtdev && $map->{$virtdev}) {
4848 my $di = parse_drive($virtdev, $value);
4849 delete $di->{format}; # format can change on restore
4850 $di->{file} = $map->{$virtdev};
4851 $value = print_drive($vmid, $di);
4852 print $outfd "$virtdev: $value\n";
4853 } else {
4854 print $outfd $line;
4855 }
4856 } else {
4857 print $outfd $line;
4858 }
4859 }
4860
4861 sub scan_volids {
4862 my ($cfg, $vmid) = @_;
4863
4864 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4865
4866 my $volid_hash = {};
4867 foreach my $storeid (keys %$info) {
4868 foreach my $item (@{$info->{$storeid}}) {
4869 next if !($item->{volid} && $item->{size});
4870 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4871 $volid_hash->{$item->{volid}} = $item;
4872 }
4873 }
4874
4875 return $volid_hash;
4876 }
4877
4878 sub get_used_paths {
4879 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4880
4881 my $used_path = {};
4882
4883 my $scan_config = sub {
4884 my ($cref, $snapname) = @_;
4885
4886 foreach my $key (keys %$cref) {
4887 my $value = $cref->{$key};
4888 if (valid_drivename($key)) {
4889 next if $skip_drive && $key eq $skip_drive;
4890 my $drive = parse_drive($key, $value);
4891 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4892 if ($drive->{file} =~ m!^/!) {
4893 $used_path->{$drive->{file}}++; # = 1;
4894 } else {
4895 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4896 next if !$storeid;
4897 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4898 next if !$scfg;
4899 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4900 $used_path->{$path}++; # = 1;
4901 }
4902 }
4903 }
4904 };
4905
4906 &$scan_config($conf);
4907
4908 undef $skip_drive;
4909
4910 if ($scan_snapshots) {
4911 foreach my $snapname (keys %{$conf->{snapshots}}) {
4912 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4913 }
4914 }
4915
4916 return $used_path;
4917 }
4918
4919 sub update_disksize {
4920 my ($vmid, $conf, $volid_hash) = @_;
4921
4922 my $changes;
4923
4924 my $used = {};
4925
4926 # Note: it is allowed to define multiple storages with same path (alias), so
4927 # we need to check both 'volid' and real 'path' (two different volid can point
4928 # to the same path).
4929
4930 my $usedpath = {};
4931
4932 # update size info
4933 foreach my $opt (keys %$conf) {
4934 if (valid_drivename($opt)) {
4935 my $drive = parse_drive($opt, $conf->{$opt});
4936 my $volid = $drive->{file};
4937 next if !$volid;
4938
4939 $used->{$volid} = 1;
4940 if ($volid_hash->{$volid} &&
4941 (my $path = $volid_hash->{$volid}->{path})) {
4942 $usedpath->{$path} = 1;
4943 }
4944
4945 next if drive_is_cdrom($drive);
4946 next if !$volid_hash->{$volid};
4947
4948 $drive->{size} = $volid_hash->{$volid}->{size};
4949 my $new = print_drive($vmid, $drive);
4950 if ($new ne $conf->{$opt}) {
4951 $changes = 1;
4952 $conf->{$opt} = $new;
4953 }
4954 }
4955 }
4956
4957 # remove 'unusedX' entry if volume is used
4958 foreach my $opt (keys %$conf) {
4959 next if $opt !~ m/^unused\d+$/;
4960 my $volid = $conf->{$opt};
4961 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4962 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4963 $changes = 1;
4964 delete $conf->{$opt};
4965 }
4966 }
4967
4968 foreach my $volid (sort keys %$volid_hash) {
4969 next if $volid =~ m/vm-$vmid-state-/;
4970 next if $used->{$volid};
4971 my $path = $volid_hash->{$volid}->{path};
4972 next if !$path; # just to be sure
4973 next if $usedpath->{$path};
4974 $changes = 1;
4975 add_unused_volume($conf, $volid);
4976 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4977 }
4978
4979 return $changes;
4980 }
4981
4982 sub rescan {
4983 my ($vmid, $nolock) = @_;
4984
4985 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4986
4987 my $volid_hash = scan_volids($cfg, $vmid);
4988
4989 my $updatefn = sub {
4990 my ($vmid) = @_;
4991
4992 my $conf = load_config($vmid);
4993
4994 check_lock($conf);
4995
4996 my $vm_volids = {};
4997 foreach my $volid (keys %$volid_hash) {
4998 my $info = $volid_hash->{$volid};
4999 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5000 }
5001
5002 my $changes = update_disksize($vmid, $conf, $vm_volids);
5003
5004 update_config_nolock($vmid, $conf, 1) if $changes;
5005 };
5006
5007 if (defined($vmid)) {
5008 if ($nolock) {
5009 &$updatefn($vmid);
5010 } else {
5011 lock_config($vmid, $updatefn, $vmid);
5012 }
5013 } else {
5014 my $vmlist = config_list();
5015 foreach my $vmid (keys %$vmlist) {
5016 if ($nolock) {
5017 &$updatefn($vmid);
5018 } else {
5019 lock_config($vmid, $updatefn, $vmid);
5020 }
5021 }
5022 }
5023 }
5024
5025 sub restore_vma_archive {
5026 my ($archive, $vmid, $user, $opts, $comp) = @_;
5027
5028 my $input = $archive eq '-' ? "<&STDIN" : undef;
5029 my $readfrom = $archive;
5030
5031 my $uncomp = '';
5032 if ($comp) {
5033 $readfrom = '-';
5034 my $qarchive = PVE::Tools::shellquote($archive);
5035 if ($comp eq 'gzip') {
5036 $uncomp = "zcat $qarchive|";
5037 } elsif ($comp eq 'lzop') {
5038 $uncomp = "lzop -d -c $qarchive|";
5039 } else {
5040 die "unknown compression method '$comp'\n";
5041 }
5042
5043 }
5044
5045 my $tmpdir = "/var/tmp/vzdumptmp$$";
5046 rmtree $tmpdir;
5047
5048 # disable interrupts (always do cleanups)
5049 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5050 warn "got interrupt - ignored\n";
5051 };
5052
5053 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5054 POSIX::mkfifo($mapfifo, 0600);
5055 my $fifofh;
5056
5057 my $openfifo = sub {
5058 open($fifofh, '>', $mapfifo) || die $!;
5059 };
5060
5061 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5062
5063 my $oldtimeout;
5064 my $timeout = 5;
5065
5066 my $devinfo = {};
5067
5068 my $rpcenv = PVE::RPCEnvironment::get();
5069
5070 my $conffile = config_file($vmid);
5071 my $tmpfn = "$conffile.$$.tmp";
5072
5073 # Note: $oldconf is undef if VM does not exists
5074 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5075
5076 my $print_devmap = sub {
5077 my $virtdev_hash = {};
5078
5079 my $cfgfn = "$tmpdir/qemu-server.conf";
5080
5081 # we can read the config - that is already extracted
5082 my $fh = IO::File->new($cfgfn, "r") ||
5083 "unable to read qemu-server.conf - $!\n";
5084
5085 while (defined(my $line = <$fh>)) {
5086 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5087 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5088 die "archive does not contain data for drive '$virtdev'\n"
5089 if !$devinfo->{$devname};
5090 if (defined($opts->{storage})) {
5091 $storeid = $opts->{storage} || 'local';
5092 } elsif (!$storeid) {
5093 $storeid = 'local';
5094 }
5095 $format = 'raw' if !$format;
5096 $devinfo->{$devname}->{devname} = $devname;
5097 $devinfo->{$devname}->{virtdev} = $virtdev;
5098 $devinfo->{$devname}->{format} = $format;
5099 $devinfo->{$devname}->{storeid} = $storeid;
5100
5101 # check permission on storage
5102 my $pool = $opts->{pool}; # todo: do we need that?
5103 if ($user ne 'root@pam') {
5104 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5105 }
5106
5107 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5108 }
5109 }
5110
5111 foreach my $devname (keys %$devinfo) {
5112 die "found no device mapping information for device '$devname'\n"
5113 if !$devinfo->{$devname}->{virtdev};
5114 }
5115
5116 my $cfg = cfs_read_file('storage.cfg');
5117
5118 # create empty/temp config
5119 if ($oldconf) {
5120 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5121 foreach_drive($oldconf, sub {
5122 my ($ds, $drive) = @_;
5123
5124 return if drive_is_cdrom($drive);
5125
5126 my $volid = $drive->{file};
5127
5128 return if !$volid || $volid =~ m|^/|;
5129
5130 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5131 return if !$path || !$owner || ($owner != $vmid);
5132
5133 # Note: only delete disk we want to restore
5134 # other volumes will become unused
5135 if ($virtdev_hash->{$ds}) {
5136 PVE::Storage::vdisk_free($cfg, $volid);
5137 }
5138 });
5139 }
5140
5141 my $map = {};
5142 foreach my $virtdev (sort keys %$virtdev_hash) {
5143 my $d = $virtdev_hash->{$virtdev};
5144 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5145 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5146
5147 # test if requested format is supported
5148 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5149 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5150 $d->{format} = $defFormat if !$supported;
5151
5152 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5153 $d->{format}, undef, $alloc_size);
5154 print STDERR "new volume ID is '$volid'\n";
5155 $d->{volid} = $volid;
5156 my $path = PVE::Storage::path($cfg, $volid);
5157
5158 my $write_zeros = 1;
5159 # fixme: what other storages types initialize volumes with zero?
5160 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5161 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5162 $write_zeros = 0;
5163 }
5164
5165 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5166
5167 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5168 $map->{$virtdev} = $volid;
5169 }
5170
5171 $fh->seek(0, 0) || die "seek failed - $!\n";
5172
5173 my $outfd = new IO::File ($tmpfn, "w") ||
5174 die "unable to write config for VM $vmid\n";
5175
5176 my $cookie = { netcount => 0 };
5177 while (defined(my $line = <$fh>)) {
5178 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5179 }
5180
5181 $fh->close();
5182 $outfd->close();
5183 };
5184
5185 eval {
5186 # enable interrupts
5187 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5188 die "interrupted by signal\n";
5189 };
5190 local $SIG{ALRM} = sub { die "got timeout\n"; };
5191
5192 $oldtimeout = alarm($timeout);
5193
5194 my $parser = sub {
5195 my $line = shift;
5196
5197 print "$line\n";
5198
5199 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5200 my ($dev_id, $size, $devname) = ($1, $2, $3);
5201 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5202 } elsif ($line =~ m/^CTIME: /) {
5203 # we correctly received the vma config, so we can disable
5204 # the timeout now for disk allocation (set to 10 minutes, so
5205 # that we always timeout if something goes wrong)
5206 alarm(600);
5207 &$print_devmap();
5208 print $fifofh "done\n";
5209 my $tmp = $oldtimeout || 0;
5210 $oldtimeout = undef;
5211 alarm($tmp);
5212 close($fifofh);
5213 }
5214 };
5215
5216 print "restore vma archive: $cmd\n";
5217 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5218 };
5219 my $err = $@;
5220
5221 alarm($oldtimeout) if $oldtimeout;
5222
5223 unlink $mapfifo;
5224
5225 if ($err) {
5226 rmtree $tmpdir;
5227 unlink $tmpfn;
5228
5229 my $cfg = cfs_read_file('storage.cfg');
5230 foreach my $devname (keys %$devinfo) {
5231 my $volid = $devinfo->{$devname}->{volid};
5232 next if !$volid;
5233 eval {
5234 if ($volid =~ m|^/|) {
5235 unlink $volid || die 'unlink failed\n';
5236 } else {
5237 PVE::Storage::vdisk_free($cfg, $volid);
5238 }
5239 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5240 };
5241 print STDERR "unable to cleanup '$volid' - $@" if $@;
5242 }
5243 die $err;
5244 }
5245
5246 rmtree $tmpdir;
5247
5248 rename($tmpfn, $conffile) ||
5249 die "unable to commit configuration file '$conffile'\n";
5250
5251 PVE::Cluster::cfs_update(); # make sure we read new file
5252
5253 eval { rescan($vmid, 1); };
5254 warn $@ if $@;
5255 }
5256
5257 sub restore_tar_archive {
5258 my ($archive, $vmid, $user, $opts) = @_;
5259
5260 if ($archive ne '-') {
5261 my $firstfile = tar_archive_read_firstfile($archive);
5262 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5263 if $firstfile ne 'qemu-server.conf';
5264 }
5265
5266 my $storecfg = cfs_read_file('storage.cfg');
5267
5268 # destroy existing data - keep empty config
5269 my $vmcfgfn = config_file($vmid);
5270 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5271
5272 my $tocmd = "/usr/lib/qemu-server/qmextract";
5273
5274 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5275 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5276 $tocmd .= ' --prealloc' if $opts->{prealloc};
5277 $tocmd .= ' --info' if $opts->{info};
5278
5279 # tar option "xf" does not autodetect compression when read from STDIN,
5280 # so we pipe to zcat
5281 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5282 PVE::Tools::shellquote("--to-command=$tocmd");
5283
5284 my $tmpdir = "/var/tmp/vzdumptmp$$";
5285 mkpath $tmpdir;
5286
5287 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5288 local $ENV{VZDUMP_VMID} = $vmid;
5289 local $ENV{VZDUMP_USER} = $user;
5290
5291 my $conffile = config_file($vmid);
5292 my $tmpfn = "$conffile.$$.tmp";
5293
5294 # disable interrupts (always do cleanups)
5295 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5296 print STDERR "got interrupt - ignored\n";
5297 };
5298
5299 eval {
5300 # enable interrupts
5301 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5302 die "interrupted by signal\n";
5303 };
5304
5305 if ($archive eq '-') {
5306 print "extracting archive from STDIN\n";
5307 run_command($cmd, input => "<&STDIN");
5308 } else {
5309 print "extracting archive '$archive'\n";
5310 run_command($cmd);
5311 }
5312
5313 return if $opts->{info};
5314
5315 # read new mapping
5316 my $map = {};
5317 my $statfile = "$tmpdir/qmrestore.stat";
5318 if (my $fd = IO::File->new($statfile, "r")) {
5319 while (defined (my $line = <$fd>)) {
5320 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5321 $map->{$1} = $2 if $1;
5322 } else {
5323 print STDERR "unable to parse line in statfile - $line\n";
5324 }
5325 }
5326 $fd->close();
5327 }
5328
5329 my $confsrc = "$tmpdir/qemu-server.conf";
5330
5331 my $srcfd = new IO::File($confsrc, "r") ||
5332 die "unable to open file '$confsrc'\n";
5333
5334 my $outfd = new IO::File ($tmpfn, "w") ||
5335 die "unable to write config for VM $vmid\n";
5336
5337 my $cookie = { netcount => 0 };
5338 while (defined (my $line = <$srcfd>)) {
5339 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5340 }
5341
5342 $srcfd->close();
5343 $outfd->close();
5344 };
5345 my $err = $@;
5346
5347 if ($err) {
5348
5349 unlink $tmpfn;
5350
5351 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5352
5353 die $err;
5354 }
5355
5356 rmtree $tmpdir;
5357
5358 rename $tmpfn, $conffile ||
5359 die "unable to commit configuration file '$conffile'\n";
5360
5361 PVE::Cluster::cfs_update(); # make sure we read new file
5362
5363 eval { rescan($vmid, 1); };
5364 warn $@ if $@;
5365 };
5366
5367
5368 # Internal snapshots
5369
5370 # NOTE: Snapshot create/delete involves several non-atomic
5371 # action, and can take a long time.
5372 # So we try to avoid locking the file and use 'lock' variable
5373 # inside the config file instead.
5374
5375 my $snapshot_copy_config = sub {
5376 my ($source, $dest) = @_;
5377
5378 foreach my $k (keys %$source) {
5379 next if $k eq 'snapshots';
5380 next if $k eq 'snapstate';
5381 next if $k eq 'snaptime';
5382 next if $k eq 'vmstate';
5383 next if $k eq 'lock';
5384 next if $k eq 'digest';
5385 next if $k eq 'description';
5386 next if $k =~ m/^unused\d+$/;
5387
5388 $dest->{$k} = $source->{$k};
5389 }
5390 };
5391
5392 my $snapshot_apply_config = sub {
5393 my ($conf, $snap) = @_;
5394
5395 # copy snapshot list
5396 my $newconf = {
5397 snapshots => $conf->{snapshots},
5398 };
5399
5400 # keep description and list of unused disks
5401 foreach my $k (keys %$conf) {
5402 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5403 $newconf->{$k} = $conf->{$k};
5404 }
5405
5406 &$snapshot_copy_config($snap, $newconf);
5407
5408 return $newconf;
5409 };
5410
5411 sub foreach_writable_storage {
5412 my ($conf, $func) = @_;
5413
5414 my $sidhash = {};
5415
5416 foreach my $ds (keys %$conf) {
5417 next if !valid_drivename($ds);
5418
5419 my $drive = parse_drive($ds, $conf->{$ds});
5420 next if !$drive;
5421 next if drive_is_cdrom($drive);
5422
5423 my $volid = $drive->{file};
5424
5425 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5426 $sidhash->{$sid} = $sid if $sid;
5427 }
5428
5429 foreach my $sid (sort keys %$sidhash) {
5430 &$func($sid);
5431 }
5432 }
5433
5434 my $alloc_vmstate_volid = sub {
5435 my ($storecfg, $vmid, $conf, $snapname) = @_;
5436
5437 # Note: we try to be smart when selecting a $target storage
5438
5439 my $target;
5440
5441 # search shared storage first
5442 foreach_writable_storage($conf, sub {
5443 my ($sid) = @_;
5444 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5445 return if !$scfg->{shared};
5446
5447 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5448 });
5449
5450 if (!$target) {
5451 # now search local storage
5452 foreach_writable_storage($conf, sub {
5453 my ($sid) = @_;
5454 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5455 return if $scfg->{shared};
5456
5457 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5458 });
5459 }
5460
5461 $target = 'local' if !$target;
5462
5463 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5464 # we abort live save after $conf->{memory}, so we need at max twice that space
5465 my $size = $conf->{memory}*2 + $driver_state_size;
5466
5467 my $name = "vm-$vmid-state-$snapname";
5468 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5469 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5470 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5471
5472 return $volid;
5473 };
5474
5475 my $snapshot_prepare = sub {
5476 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5477
5478 my $snap;
5479
5480 my $updatefn = sub {
5481
5482 my $conf = load_config($vmid);
5483
5484 die "you can't take a snapshot if it's a template\n"
5485 if is_template($conf);
5486
5487 check_lock($conf);
5488
5489 $conf->{lock} = 'snapshot';
5490
5491 die "snapshot name '$snapname' already used\n"
5492 if defined($conf->{snapshots}->{$snapname});
5493
5494 my $storecfg = PVE::Storage::config();
5495 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5496
5497 $snap = $conf->{snapshots}->{$snapname} = {};
5498
5499 if ($save_vmstate && check_running($vmid)) {
5500 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5501 }
5502
5503 &$snapshot_copy_config($conf, $snap);
5504
5505 $snap->{snapstate} = "prepare";
5506 $snap->{snaptime} = time();
5507 $snap->{description} = $comment if $comment;
5508
5509 # always overwrite machine if we save vmstate. This makes sure we
5510 # can restore it later using correct machine type
5511 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5512
5513 update_config_nolock($vmid, $conf, 1);
5514 };
5515
5516 lock_config($vmid, $updatefn);
5517
5518 return $snap;
5519 };
5520
5521 my $snapshot_commit = sub {
5522 my ($vmid, $snapname) = @_;
5523
5524 my $updatefn = sub {
5525
5526 my $conf = load_config($vmid);
5527
5528 die "missing snapshot lock\n"
5529 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5530
5531 my $has_machine_config = defined($conf->{machine});
5532
5533 my $snap = $conf->{snapshots}->{$snapname};
5534
5535 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5536
5537 die "wrong snapshot state\n"
5538 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5539
5540 delete $snap->{snapstate};
5541 delete $conf->{lock};
5542
5543 my $newconf = &$snapshot_apply_config($conf, $snap);
5544
5545 delete $newconf->{machine} if !$has_machine_config;
5546
5547 $newconf->{parent} = $snapname;
5548
5549 update_config_nolock($vmid, $newconf, 1);
5550 };
5551
5552 lock_config($vmid, $updatefn);
5553 };
5554
5555 sub snapshot_rollback {
5556 my ($vmid, $snapname) = @_;
5557
5558 my $prepare = 1;
5559
5560 my $storecfg = PVE::Storage::config();
5561
5562 my $conf = load_config($vmid);
5563
5564 my $get_snapshot_config = sub {
5565
5566 die "you can't rollback if vm is a template\n" if is_template($conf);
5567
5568 my $res = $conf->{snapshots}->{$snapname};
5569
5570 die "snapshot '$snapname' does not exist\n" if !defined($res);
5571
5572 return $res;
5573 };
5574
5575 my $snap = &$get_snapshot_config();
5576
5577 foreach_drive($snap, sub {
5578 my ($ds, $drive) = @_;
5579
5580 return if drive_is_cdrom($drive);
5581
5582 my $volid = $drive->{file};
5583
5584 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5585 });
5586
5587 my $updatefn = sub {
5588
5589 $conf = load_config($vmid);
5590
5591 $snap = &$get_snapshot_config();
5592
5593 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5594 if $snap->{snapstate};
5595
5596 if ($prepare) {
5597 check_lock($conf);
5598 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5599 }
5600
5601 die "unable to rollback vm $vmid: vm is running\n"
5602 if check_running($vmid);
5603
5604 if ($prepare) {
5605 $conf->{lock} = 'rollback';
5606 } else {
5607 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5608 delete $conf->{lock};
5609 }
5610
5611 my $forcemachine;
5612
5613 if (!$prepare) {
5614 my $has_machine_config = defined($conf->{machine});
5615
5616 # copy snapshot config to current config
5617 $conf = &$snapshot_apply_config($conf, $snap);
5618 $conf->{parent} = $snapname;
5619
5620 # Note: old code did not store 'machine', so we try to be smart
5621 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5622 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5623 # we remove the 'machine' configuration if not explicitly specified
5624 # in the original config.
5625 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5626 }
5627
5628 update_config_nolock($vmid, $conf, 1);
5629
5630 if (!$prepare && $snap->{vmstate}) {
5631 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5632 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5633 }
5634 };
5635
5636 lock_config($vmid, $updatefn);
5637
5638 foreach_drive($snap, sub {
5639 my ($ds, $drive) = @_;
5640
5641 return if drive_is_cdrom($drive);
5642
5643 my $volid = $drive->{file};
5644 my $device = "drive-$ds";
5645
5646 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5647 });
5648
5649 $prepare = 0;
5650 lock_config($vmid, $updatefn);
5651 }
5652
5653 my $savevm_wait = sub {
5654 my ($vmid) = @_;
5655
5656 for(;;) {
5657 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5658 if (!$stat->{status}) {
5659 die "savevm not active\n";
5660 } elsif ($stat->{status} eq 'active') {
5661 sleep(1);
5662 next;
5663 } elsif ($stat->{status} eq 'completed') {
5664 last;
5665 } else {
5666 die "query-savevm returned status '$stat->{status}'\n";
5667 }
5668 }
5669 };
5670
5671 sub snapshot_create {
5672 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5673
5674 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5675
5676 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5677
5678 my $config = load_config($vmid);
5679
5680 my $running = check_running($vmid);
5681
5682 my $freezefs = $running && $config->{agent};
5683 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5684
5685 my $drivehash = {};
5686
5687 if ($freezefs) {
5688 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5689 warn "guest-fsfreeze-freeze problems - $@" if $@;
5690 }
5691
5692 eval {
5693 # create internal snapshots of all drives
5694
5695 my $storecfg = PVE::Storage::config();
5696
5697 if ($running) {
5698 if ($snap->{vmstate}) {
5699 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5700 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5701 &$savevm_wait($vmid);
5702 } else {
5703 vm_mon_cmd($vmid, "savevm-start");
5704 }
5705 };
5706
5707 foreach_drive($snap, sub {
5708 my ($ds, $drive) = @_;
5709
5710 return if drive_is_cdrom($drive);
5711
5712 my $volid = $drive->{file};
5713 my $device = "drive-$ds";
5714
5715 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5716 $drivehash->{$ds} = 1;
5717 });
5718 };
5719 my $err = $@;
5720
5721 if ($running) {
5722 eval { vm_mon_cmd($vmid, "savevm-end") };
5723 warn $@ if $@;
5724
5725 if ($freezefs) {
5726 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5727 warn "guest-fsfreeze-thaw problems - $@" if $@;
5728 }
5729
5730 # savevm-end is async, we need to wait
5731 for (;;) {
5732 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5733 if (!$stat->{bytes}) {
5734 last;
5735 } else {
5736 print "savevm not yet finished\n";
5737 sleep(1);
5738 next;
5739 }
5740 }
5741 }
5742
5743 if ($err) {
5744 warn "snapshot create failed: starting cleanup\n";
5745 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5746 warn $@ if $@;
5747 die $err;
5748 }
5749
5750 &$snapshot_commit($vmid, $snapname);
5751 }
5752
5753 # Note: $drivehash is only set when called from snapshot_create.
5754 sub snapshot_delete {
5755 my ($vmid, $snapname, $force, $drivehash) = @_;
5756
5757 my $prepare = 1;
5758
5759 my $snap;
5760 my $unused = [];
5761
5762 my $unlink_parent = sub {
5763 my ($confref, $new_parent) = @_;
5764
5765 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5766 if ($new_parent) {
5767 $confref->{parent} = $new_parent;
5768 } else {
5769 delete $confref->{parent};
5770 }
5771 }
5772 };
5773
5774 my $updatefn = sub {
5775 my ($remove_drive) = @_;
5776
5777 my $conf = load_config($vmid);
5778
5779 if (!$drivehash) {
5780 check_lock($conf);
5781 die "you can't delete a snapshot if vm is a template\n"
5782 if is_template($conf);
5783 }
5784
5785 $snap = $conf->{snapshots}->{$snapname};
5786
5787 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5788
5789 # remove parent refs
5790 if (!$prepare) {
5791 &$unlink_parent($conf, $snap->{parent});
5792 foreach my $sn (keys %{$conf->{snapshots}}) {
5793 next if $sn eq $snapname;
5794 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5795 }
5796 }
5797
5798 if ($remove_drive) {
5799 if ($remove_drive eq 'vmstate') {
5800 delete $snap->{$remove_drive};
5801 } else {
5802 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5803 my $volid = $drive->{file};
5804 delete $snap->{$remove_drive};
5805 add_unused_volume($conf, $volid);
5806 }
5807 }
5808
5809 if ($prepare) {
5810 $snap->{snapstate} = 'delete';
5811 } else {
5812 delete $conf->{snapshots}->{$snapname};
5813 delete $conf->{lock} if $drivehash;
5814 foreach my $volid (@$unused) {
5815 add_unused_volume($conf, $volid);
5816 }
5817 }
5818
5819 update_config_nolock($vmid, $conf, 1);
5820 };
5821
5822 lock_config($vmid, $updatefn);
5823
5824 # now remove vmstate file
5825
5826 my $storecfg = PVE::Storage::config();
5827
5828 if ($snap->{vmstate}) {
5829 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5830 if (my $err = $@) {
5831 die $err if !$force;
5832 warn $err;
5833 }
5834 # save changes (remove vmstate from snapshot)
5835 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5836 };
5837
5838 # now remove all internal snapshots
5839 foreach_drive($snap, sub {
5840 my ($ds, $drive) = @_;
5841
5842 return if drive_is_cdrom($drive);
5843
5844 my $volid = $drive->{file};
5845 my $device = "drive-$ds";
5846
5847 if (!$drivehash || $drivehash->{$ds}) {
5848 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5849 if (my $err = $@) {
5850 die $err if !$force;
5851 warn $err;
5852 }
5853 }
5854
5855 # save changes (remove drive fron snapshot)
5856 lock_config($vmid, $updatefn, $ds) if !$force;
5857 push @$unused, $volid;
5858 });
5859
5860 # now cleanup config
5861 $prepare = 0;
5862 lock_config($vmid, $updatefn);
5863 }
5864
5865 sub has_feature {
5866 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5867
5868 my $err;
5869 foreach_drive($conf, sub {
5870 my ($ds, $drive) = @_;
5871
5872 return if drive_is_cdrom($drive);
5873 my $volid = $drive->{file};
5874 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5875 });
5876
5877 return $err ? 0 : 1;
5878 }
5879
5880 sub template_create {
5881 my ($vmid, $conf, $disk) = @_;
5882
5883 my $storecfg = PVE::Storage::config();
5884
5885 foreach_drive($conf, sub {
5886 my ($ds, $drive) = @_;
5887
5888 return if drive_is_cdrom($drive);
5889 return if $disk && $ds ne $disk;
5890
5891 my $volid = $drive->{file};
5892 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5893
5894 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5895 $drive->{file} = $voliddst;
5896 $conf->{$ds} = print_drive($vmid, $drive);
5897 update_config_nolock($vmid, $conf, 1);
5898 });
5899 }
5900
5901 sub is_template {
5902 my ($conf) = @_;
5903
5904 return 1 if defined $conf->{template} && $conf->{template} == 1;
5905 }
5906
5907 sub qemu_img_convert {
5908 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5909
5910 my $storecfg = PVE::Storage::config();
5911 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5912 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5913
5914 if ($src_storeid && $dst_storeid) {
5915 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5916 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5917
5918 my $src_format = qemu_img_format($src_scfg, $src_volname);
5919 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5920
5921 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5922 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5923
5924 my $cmd = [];
5925 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5926 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5927 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5928
5929 my $parser = sub {
5930 my $line = shift;
5931 if($line =~ m/\((\S+)\/100\%\)/){
5932 my $percent = $1;
5933 my $transferred = int($size * $percent / 100);
5934 my $remaining = $size - $transferred;
5935
5936 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5937 }
5938
5939 };
5940
5941 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5942 my $err = $@;
5943 die "copy failed: $err" if $err;
5944 }
5945 }
5946
5947 sub qemu_img_format {
5948 my ($scfg, $volname) = @_;
5949
5950 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5951 return $1;
5952 } elsif ($scfg->{type} eq 'iscsi') {
5953 return "host_device";
5954 } else {
5955 return "raw";
5956 }
5957 }
5958
5959 sub qemu_drive_mirror {
5960 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5961
5962 my $count = 0;
5963 my $old_len = 0;
5964 my $frozen = undef;
5965 my $maxwait = 120;
5966
5967 my $storecfg = PVE::Storage::config();
5968 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5969
5970 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5971
5972 my $format;
5973 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5974 $format = $1;
5975 }
5976
5977 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5978
5979 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5980 $opts->{format} = $format if $format;
5981
5982 #fixme : sometime drive-mirror timeout, but works fine after.
5983 # (I have see the problem with big volume > 200GB), so we need to eval
5984 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5985 # ignore errors here
5986
5987 eval {
5988 while (1) {
5989 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5990 my $stat = @$stats[0];
5991 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5992 die "error job is not mirroring" if $stat->{type} ne "mirror";
5993
5994 my $busy = $stat->{busy};
5995
5996 if (my $total = $stat->{len}) {
5997 my $transferred = $stat->{offset} || 0;
5998 my $remaining = $total - $transferred;
5999 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6000
6001 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
6002 }
6003
6004 if ($stat->{len} == $stat->{offset}) {
6005 if ($busy eq 'false') {
6006
6007 last if $vmiddst != $vmid;
6008
6009 # try to switch the disk if source and destination are on the same guest
6010 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6011 last if !$@;
6012 die $@ if $@ !~ m/cannot be completed/;
6013 }
6014
6015 if ($count > $maxwait) {
6016 # if too much writes to disk occurs at the end of migration
6017 #the disk needs to be freezed to be able to complete the migration
6018 vm_suspend($vmid,1);
6019 $frozen = 1;
6020 }
6021 $count ++
6022 }
6023 $old_len = $stat->{offset};
6024 sleep 1;
6025 }
6026
6027 vm_resume($vmid, 1) if $frozen;
6028
6029 };
6030 my $err = $@;
6031
6032 my $cancel_job = sub {
6033 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6034 while (1) {
6035 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6036 my $stat = @$stats[0];
6037 last if !$stat;
6038 sleep 1;
6039 }
6040 };
6041
6042 if ($err) {
6043 eval { &$cancel_job(); };
6044 die "mirroring error: $err";
6045 }
6046
6047 if ($vmiddst != $vmid) {
6048 # if we clone a disk for a new target vm, we don't switch the disk
6049 &$cancel_job(); # so we call block-job-cancel
6050 }
6051 }
6052
6053 sub clone_disk {
6054 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6055 $newvmid, $storage, $format, $full, $newvollist) = @_;
6056
6057 my $newvolid;
6058
6059 if (!$full) {
6060 print "create linked clone of drive $drivename ($drive->{file})\n";
6061 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6062 push @$newvollist, $newvolid;
6063 } else {
6064 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6065 $storeid = $storage if $storage;
6066
6067 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6068 if (!$format) {
6069 $format = $drive->{format} || $defFormat;
6070 }
6071
6072 # test if requested format is supported - else use default
6073 my $supported = grep { $_ eq $format } @$validFormats;
6074 $format = $defFormat if !$supported;
6075
6076 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6077
6078 print "create full clone of drive $drivename ($drive->{file})\n";
6079 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6080 push @$newvollist, $newvolid;
6081
6082 if (!$running || $snapname) {
6083 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6084 } else {
6085 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6086 }
6087 }
6088
6089 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6090
6091 my $disk = $drive;
6092 $disk->{format} = undef;
6093 $disk->{file} = $newvolid;
6094 $disk->{size} = $size;
6095
6096 return $disk;
6097 }
6098
6099 # this only works if VM is running
6100 sub get_current_qemu_machine {
6101 my ($vmid) = @_;
6102
6103 my $cmd = { execute => 'query-machines', arguments => {} };
6104 my $res = vm_qmp_command($vmid, $cmd);
6105
6106 my ($current, $default);
6107 foreach my $e (@$res) {
6108 $default = $e->{name} if $e->{'is-default'};
6109 $current = $e->{name} if $e->{'is-current'};
6110 }
6111
6112 # fallback to the default machine if current is not supported by qemu
6113 return $current || $default || 'pc';
6114 }
6115
6116 sub lspci {
6117
6118 my $devices = {};
6119
6120 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6121 my (undef, $id, $function) = @_;
6122 my $res = { id => $id, function => $function};
6123 push @{$devices->{$id}}, $res;
6124 });
6125
6126 return $devices;
6127 }
6128
6129 1;