]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
drive_add : escape \ character
[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 $drive =~ s/\\/\\\\/g;
3398 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3399
3400 # If the command succeeds qemu prints: "OK"
3401 return 1 if $ret =~ m/OK/s;
3402
3403 die "adding drive failed: $ret\n";
3404 }
3405
3406 sub qemu_drivedel {
3407 my($vmid, $deviceid) = @_;
3408
3409 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3410 $ret =~ s/^\s+//;
3411
3412 return 1 if $ret eq "";
3413
3414 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3415 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3416
3417 die "deleting drive $deviceid failed : $ret\n";
3418 }
3419
3420 sub qemu_deviceaddverify {
3421 my ($vmid, $deviceid) = @_;
3422
3423 for (my $i = 0; $i <= 5; $i++) {
3424 my $devices_list = vm_devices_list($vmid);
3425 return 1 if defined($devices_list->{$deviceid});
3426 sleep 1;
3427 }
3428
3429 die "error on hotplug device '$deviceid'\n";
3430 }
3431
3432
3433 sub qemu_devicedelverify {
3434 my ($vmid, $deviceid) = @_;
3435
3436 # need to verify that the device is correctly removed as device_del
3437 # is async and empty return is not reliable
3438
3439 for (my $i = 0; $i <= 5; $i++) {
3440 my $devices_list = vm_devices_list($vmid);
3441 return 1 if !defined($devices_list->{$deviceid});
3442 sleep 1;
3443 }
3444
3445 die "error on hot-unplugging device '$deviceid'\n";
3446 }
3447
3448 sub qemu_findorcreatescsihw {
3449 my ($storecfg, $conf, $vmid, $device) = @_;
3450
3451 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3452 my $controller = int($device->{index} / $maxdev);
3453 my $scsihwid="scsihw$controller";
3454 my $devices_list = vm_devices_list($vmid);
3455
3456 if(!defined($devices_list->{$scsihwid})) {
3457 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3458 }
3459
3460 return 1;
3461 }
3462
3463 sub qemu_add_pci_bridge {
3464 my ($storecfg, $conf, $vmid, $device) = @_;
3465
3466 my $bridges = {};
3467
3468 my $bridgeid;
3469
3470 print_pci_addr($device, $bridges);
3471
3472 while (my ($k, $v) = each %$bridges) {
3473 $bridgeid = $k;
3474 }
3475 return 1 if !defined($bridgeid) || $bridgeid < 1;
3476
3477 my $bridge = "pci.$bridgeid";
3478 my $devices_list = vm_devices_list($vmid);
3479
3480 if (!defined($devices_list->{$bridge})) {
3481 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3482 }
3483
3484 return 1;
3485 }
3486
3487 sub qemu_set_link_status {
3488 my ($vmid, $device, $up) = @_;
3489
3490 vm_mon_cmd($vmid, "set_link", name => $device,
3491 up => $up ? JSON::true : JSON::false);
3492 }
3493
3494 sub qemu_netdevadd {
3495 my ($vmid, $conf, $device, $deviceid) = @_;
3496
3497 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3498 my %options = split(/[=,]/, $netdev);
3499
3500 vm_mon_cmd($vmid, "netdev_add", %options);
3501 return 1;
3502 }
3503
3504 sub qemu_netdevdel {
3505 my ($vmid, $deviceid) = @_;
3506
3507 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3508 }
3509
3510 sub qemu_cpu_hotplug {
3511 my ($vmid, $conf, $vcpus) = @_;
3512
3513 my $sockets = 1;
3514 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3515 $sockets = $conf->{sockets} if $conf->{sockets};
3516 my $cores = $conf->{cores} || 1;
3517 my $maxcpus = $sockets * $cores;
3518
3519 $vcpus = $maxcpus if !$vcpus;
3520
3521 die "you can't add more vcpus than maxcpus\n"
3522 if $vcpus > $maxcpus;
3523
3524 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3525 die "online cpu unplug is not yet possible\n"
3526 if $vcpus < $currentvcpus;
3527
3528 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3529 die "vcpus in running vm is different than configuration\n"
3530 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3531
3532 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3533 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3534 }
3535 }
3536
3537 sub qemu_memory_hotplug {
3538 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3539
3540 return $value if !check_running($vmid);
3541
3542 my $memory = $conf->{memory} || $defaults->{memory};
3543 $value = $defaults->{memory} if !$value;
3544 return $value if $value == $memory;
3545
3546 my $static_memory = $STATICMEM;
3547 my $dimm_memory = $memory - $static_memory;
3548
3549 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3550 die "memory unplug is not yet available" if $value < $memory;
3551 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3552
3553
3554 my $sockets = 1;
3555 $sockets = $conf->{sockets} if $conf->{sockets};
3556
3557 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3558 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3559
3560 return if $current_size <= $conf->{memory};
3561
3562 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3563 if (my $err = $@) {
3564 eval { qemu_objectdel($vmid, "mem-$name"); };
3565 die $err;
3566 }
3567
3568 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3569 if (my $err = $@) {
3570 eval { qemu_objectdel($vmid, "mem-$name"); };
3571 die $err;
3572 }
3573 #update conf after each succesful module hotplug
3574 $conf->{memory} = $current_size;
3575 update_config_nolock($vmid, $conf, 1);
3576 });
3577 }
3578
3579 sub qemu_block_set_io_throttle {
3580 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3581
3582 return if !check_running($vmid) ;
3583
3584 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($iops_wr));
3585
3586 }
3587
3588 # old code, only used to shutdown old VM after update
3589 sub __read_avail {
3590 my ($fh, $timeout) = @_;
3591
3592 my $sel = new IO::Select;
3593 $sel->add($fh);
3594
3595 my $res = '';
3596 my $buf;
3597
3598 my @ready;
3599 while (scalar (@ready = $sel->can_read($timeout))) {
3600 my $count;
3601 if ($count = $fh->sysread($buf, 8192)) {
3602 if ($buf =~ /^(.*)\(qemu\) $/s) {
3603 $res .= $1;
3604 last;
3605 } else {
3606 $res .= $buf;
3607 }
3608 } else {
3609 if (!defined($count)) {
3610 die "$!\n";
3611 }
3612 last;
3613 }
3614 }
3615
3616 die "monitor read timeout\n" if !scalar(@ready);
3617
3618 return $res;
3619 }
3620
3621 # old code, only used to shutdown old VM after update
3622 sub vm_monitor_command {
3623 my ($vmid, $cmdstr, $nocheck) = @_;
3624
3625 my $res;
3626
3627 eval {
3628 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3629
3630 my $sname = "${var_run_tmpdir}/$vmid.mon";
3631
3632 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3633 die "unable to connect to VM $vmid socket - $!\n";
3634
3635 my $timeout = 3;
3636
3637 # hack: migrate sometime blocks the monitor (when migrate_downtime
3638 # is set)
3639 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3640 $timeout = 60*60; # 1 hour
3641 }
3642
3643 # read banner;
3644 my $data = __read_avail($sock, $timeout);
3645
3646 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3647 die "got unexpected qemu monitor banner\n";
3648 }
3649
3650 my $sel = new IO::Select;
3651 $sel->add($sock);
3652
3653 if (!scalar(my @ready = $sel->can_write($timeout))) {
3654 die "monitor write error - timeout";
3655 }
3656
3657 my $fullcmd = "$cmdstr\r";
3658
3659 # syslog('info', "VM $vmid monitor command: $cmdstr");
3660
3661 my $b;
3662 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3663 die "monitor write error - $!";
3664 }
3665
3666 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3667
3668 $timeout = 20;
3669
3670 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3671 $timeout = 60*60; # 1 hour
3672 } elsif ($cmdstr =~ m/^(eject|change)/) {
3673 $timeout = 60; # note: cdrom mount command is slow
3674 }
3675 if ($res = __read_avail($sock, $timeout)) {
3676
3677 my @lines = split("\r?\n", $res);
3678
3679 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3680
3681 $res = join("\n", @lines);
3682 $res .= "\n";
3683 }
3684 };
3685
3686 my $err = $@;
3687
3688 if ($err) {
3689 syslog("err", "VM $vmid monitor command failed - $err");
3690 die $err;
3691 }
3692
3693 return $res;
3694 }
3695
3696 sub qemu_block_resize {
3697 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3698
3699 my $running = check_running($vmid);
3700
3701 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3702
3703 return if !$running;
3704
3705 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3706
3707 }
3708
3709 sub qemu_volume_snapshot {
3710 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3711
3712 my $running = check_running($vmid);
3713
3714 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3715
3716 return if !$running;
3717
3718 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3719
3720 }
3721
3722 sub qemu_volume_snapshot_delete {
3723 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3724
3725 my $running = check_running($vmid);
3726
3727 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3728
3729 return if !$running;
3730
3731 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3732 }
3733
3734 sub set_migration_caps {
3735 my ($vmid) = @_;
3736
3737 my $cap_ref = [];
3738
3739 my $enabled_cap = {
3740 "auto-converge" => 1,
3741 "xbzrle" => 0,
3742 "x-rdma-pin-all" => 0,
3743 "zero-blocks" => 0,
3744 };
3745
3746 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3747
3748 for my $supported_capability (@$supported_capabilities) {
3749 push @$cap_ref, {
3750 capability => $supported_capability->{capability},
3751 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3752 };
3753 }
3754
3755 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3756 }
3757
3758 my $fast_plug_option = {
3759 'lock' => 1,
3760 'name' => 1,
3761 'onboot' => 1,
3762 'shares' => 1,
3763 'startup' => 1,
3764 };
3765
3766 # hotplug changes in [PENDING]
3767 # $selection hash can be used to only apply specified options, for
3768 # example: { cores => 1 } (only apply changed 'cores')
3769 # $errors ref is used to return error messages
3770 sub vmconfig_hotplug_pending {
3771 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3772
3773 my $defaults = load_defaults();
3774
3775 # commit values which do not have any impact on running VM first
3776 # Note: those option cannot raise errors, we we do not care about
3777 # $selection and always apply them.
3778
3779 my $add_error = sub {
3780 my ($opt, $msg) = @_;
3781 $errors->{$opt} = "hotplug problem - $msg";
3782 };
3783
3784 my $changes = 0;
3785 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3786 if ($fast_plug_option->{$opt}) {
3787 $conf->{$opt} = $conf->{pending}->{$opt};
3788 delete $conf->{pending}->{$opt};
3789 $changes = 1;
3790 }
3791 }
3792
3793 if ($changes) {
3794 update_config_nolock($vmid, $conf, 1);
3795 $conf = load_config($vmid); # update/reload
3796 }
3797
3798 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3799
3800 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3801 foreach my $opt (@delete) {
3802 next if $selection && !$selection->{$opt};
3803 eval {
3804 if ($opt eq 'hotplug') {
3805 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3806 } elsif ($opt eq 'tablet') {
3807 die "skip\n" if !$hotplug_features->{usb};
3808 if ($defaults->{tablet}) {
3809 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3810 } else {
3811 vm_deviceunplug($vmid, $conf, $opt);
3812 }
3813 } elsif ($opt eq 'vcpus') {
3814 die "skip\n" if !$hotplug_features->{cpu};
3815 qemu_cpu_hotplug($vmid, $conf, undef);
3816 } elsif ($opt eq 'balloon') {
3817 # enable balloon device is not hotpluggable
3818 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3819 } elsif ($fast_plug_option->{$opt}) {
3820 # do nothing
3821 } elsif ($opt =~ m/^net(\d+)$/) {
3822 die "skip\n" if !$hotplug_features->{network};
3823 vm_deviceunplug($vmid, $conf, $opt);
3824 } elsif (valid_drivename($opt)) {
3825 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3826 vm_deviceunplug($vmid, $conf, $opt);
3827 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3828 } elsif ($opt =~ m/^memory$/) {
3829 die "skip\n" if !$hotplug_features->{memory};
3830 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3831 } else {
3832 die "skip\n";
3833 }
3834 };
3835 if (my $err = $@) {
3836 &$add_error($opt, $err) if $err ne "skip\n";
3837 } else {
3838 # save new config if hotplug was successful
3839 delete $conf->{$opt};
3840 vmconfig_undelete_pending_option($conf, $opt);
3841 update_config_nolock($vmid, $conf, 1);
3842 $conf = load_config($vmid); # update/reload
3843 }
3844 }
3845
3846 foreach my $opt (keys %{$conf->{pending}}) {
3847 next if $selection && !$selection->{$opt};
3848 my $value = $conf->{pending}->{$opt};
3849 eval {
3850 if ($opt eq 'hotplug') {
3851 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3852 } elsif ($opt eq 'tablet') {
3853 die "skip\n" if !$hotplug_features->{usb};
3854 if ($value == 1) {
3855 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3856 } elsif ($value == 0) {
3857 vm_deviceunplug($vmid, $conf, $opt);
3858 }
3859 } elsif ($opt eq 'vcpus') {
3860 die "skip\n" if !$hotplug_features->{cpu};
3861 qemu_cpu_hotplug($vmid, $conf, $value);
3862 } elsif ($opt eq 'balloon') {
3863 # enable/disable balloning device is not hotpluggable
3864 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3865 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3866 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3867
3868 # allow manual ballooning if shares is set to zero
3869 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3870 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3871 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3872 }
3873 } elsif ($opt =~ m/^net(\d+)$/) {
3874 # some changes can be done without hotplug
3875 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3876 $vmid, $opt, $value);
3877 } elsif (valid_drivename($opt)) {
3878 # some changes can be done without hotplug
3879 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3880 $vmid, $opt, $value, 1);
3881 } elsif ($opt =~ m/^memory$/) { #dimms
3882 die "skip\n" if !$hotplug_features->{memory};
3883 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3884 } else {
3885 die "skip\n"; # skip non-hot-pluggable options
3886 }
3887 };
3888 if (my $err = $@) {
3889 &$add_error($opt, $err) if $err ne "skip\n";
3890 } else {
3891 # save new config if hotplug was successful
3892 $conf->{$opt} = $value;
3893 delete $conf->{pending}->{$opt};
3894 update_config_nolock($vmid, $conf, 1);
3895 $conf = load_config($vmid); # update/reload
3896 }
3897 }
3898 }
3899
3900 sub vmconfig_apply_pending {
3901 my ($vmid, $conf, $storecfg) = @_;
3902
3903 # cold plug
3904
3905 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3906 foreach my $opt (@delete) { # delete
3907 die "internal error" if $opt =~ m/^unused/;
3908 $conf = load_config($vmid); # update/reload
3909 if (!defined($conf->{$opt})) {
3910 vmconfig_undelete_pending_option($conf, $opt);
3911 update_config_nolock($vmid, $conf, 1);
3912 } elsif (valid_drivename($opt)) {
3913 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3914 vmconfig_undelete_pending_option($conf, $opt);
3915 delete $conf->{$opt};
3916 update_config_nolock($vmid, $conf, 1);
3917 } else {
3918 vmconfig_undelete_pending_option($conf, $opt);
3919 delete $conf->{$opt};
3920 update_config_nolock($vmid, $conf, 1);
3921 }
3922 }
3923
3924 $conf = load_config($vmid); # update/reload
3925
3926 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3927 $conf = load_config($vmid); # update/reload
3928
3929 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3930 # skip if nothing changed
3931 } elsif (valid_drivename($opt)) {
3932 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3933 if defined($conf->{$opt});
3934 $conf->{$opt} = $conf->{pending}->{$opt};
3935 } else {
3936 $conf->{$opt} = $conf->{pending}->{$opt};
3937 }
3938
3939 delete $conf->{pending}->{$opt};
3940 update_config_nolock($vmid, $conf, 1);
3941 }
3942 }
3943
3944 my $safe_num_ne = sub {
3945 my ($a, $b) = @_;
3946
3947 return 0 if !defined($a) && !defined($b);
3948 return 1 if !defined($a);
3949 return 1 if !defined($b);
3950
3951 return $a != $b;
3952 };
3953
3954 my $safe_string_ne = sub {
3955 my ($a, $b) = @_;
3956
3957 return 0 if !defined($a) && !defined($b);
3958 return 1 if !defined($a);
3959 return 1 if !defined($b);
3960
3961 return $a ne $b;
3962 };
3963
3964 sub vmconfig_update_net {
3965 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3966
3967 my $newnet = parse_net($value);
3968
3969 if ($conf->{$opt}) {
3970 my $oldnet = parse_net($conf->{$opt});
3971
3972 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3973 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3974 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3975 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3976
3977 # for non online change, we try to hot-unplug
3978 die "skip\n" if !$hotplug;
3979 vm_deviceunplug($vmid, $conf, $opt);
3980 } else {
3981
3982 die "internal error" if $opt !~ m/net(\d+)/;
3983 my $iface = "tap${vmid}i$1";
3984
3985 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3986 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3987 }
3988
3989 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3990 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3991 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3992 PVE::Network::tap_unplug($iface);
3993 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3994 }
3995
3996 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3997 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3998 }
3999
4000 return 1;
4001 }
4002 }
4003
4004 if ($hotplug) {
4005 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4006 } else {
4007 die "skip\n";
4008 }
4009 }
4010
4011 sub vmconfig_update_disk {
4012 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4013
4014 # fixme: do we need force?
4015
4016 my $drive = parse_drive($opt, $value);
4017
4018 if ($conf->{$opt}) {
4019
4020 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4021
4022 my $media = $drive->{media} || 'disk';
4023 my $oldmedia = $old_drive->{media} || 'disk';
4024 die "unable to change media type\n" if $media ne $oldmedia;
4025
4026 if (!drive_is_cdrom($old_drive)) {
4027
4028 if ($drive->{file} ne $old_drive->{file}) {
4029
4030 die "skip\n" if !$hotplug;
4031
4032 # unplug and register as unused
4033 vm_deviceunplug($vmid, $conf, $opt);
4034 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4035
4036 } else {
4037 # update existing disk
4038
4039 # skip non hotpluggable value
4040 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4041 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4042 die "skip\n";
4043 }
4044
4045 # apply throttle
4046 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4047 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4048 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4049 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4050 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4051 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4052 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4053 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4054 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4055 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4056 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4057 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4058
4059 qemu_block_set_io_throttle($vmid,"drive-$opt",
4060 ($drive->{mbps} || 0)*1024*1024,
4061 ($drive->{mbps_rd} || 0)*1024*1024,
4062 ($drive->{mbps_wr} || 0)*1024*1024,
4063 $drive->{iops} || 0,
4064 $drive->{iops_rd} || 0,
4065 $drive->{iops_wr} || 0,
4066 ($drive->{mbps_max} || 0)*1024*1024,
4067 ($drive->{mbps_rd_max} || 0)*1024*1024,
4068 ($drive->{mbps_wr_max} || 0)*1024*1024,
4069 $drive->{iops_max} || 0,
4070 $drive->{iops_rd_max} || 0,
4071 $drive->{iops_wr_max} || 0);
4072
4073 }
4074
4075 return 1;
4076 }
4077
4078 } else { # cdrom
4079
4080 if ($drive->{file} eq 'none') {
4081 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4082 } else {
4083 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4084 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4085 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4086 }
4087
4088 return 1;
4089 }
4090 }
4091 }
4092
4093 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4094 # hotplug new disks
4095 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4096 }
4097
4098 sub vm_start {
4099 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4100
4101 lock_config($vmid, sub {
4102 my $conf = load_config($vmid, $migratedfrom);
4103
4104 die "you can't start a vm if it's a template\n" if is_template($conf);
4105
4106 check_lock($conf) if !$skiplock;
4107
4108 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4109
4110 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4111 vmconfig_apply_pending($vmid, $conf, $storecfg);
4112 $conf = load_config($vmid); # update/reload
4113 }
4114
4115 my $defaults = load_defaults();
4116
4117 # set environment variable useful inside network script
4118 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4119
4120 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4121
4122 my $migrate_port = 0;
4123 my $migrate_uri;
4124 if ($statefile) {
4125 if ($statefile eq 'tcp') {
4126 my $localip = "localhost";
4127 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4128 if ($datacenterconf->{migration_unsecure}) {
4129 my $nodename = PVE::INotify::nodename();
4130 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4131 }
4132 $migrate_port = PVE::Tools::next_migrate_port();
4133 $migrate_uri = "tcp:${localip}:${migrate_port}";
4134 push @$cmd, '-incoming', $migrate_uri;
4135 push @$cmd, '-S';
4136 } else {
4137 push @$cmd, '-loadstate', $statefile;
4138 }
4139 } elsif ($paused) {
4140 push @$cmd, '-S';
4141 }
4142
4143 # host pci devices
4144 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4145 my $d = parse_hostpci($conf->{"hostpci$i"});
4146 next if !$d;
4147 my $pcidevices = $d->{pciid};
4148 foreach my $pcidevice (@$pcidevices) {
4149 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4150
4151 my $info = pci_device_info("0000:$pciid");
4152 die "IOMMU not present\n" if !check_iommu_support();
4153 die "no pci device info for device '$pciid'\n" if !$info;
4154
4155 if ($d->{driver} && $d->{driver} eq "vfio") {
4156 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4157 } else {
4158 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4159 }
4160
4161 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4162 }
4163 }
4164
4165 PVE::Storage::activate_volumes($storecfg, $vollist);
4166
4167 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4168 umask => 0077); };
4169 my $err = $@;
4170 die "start failed: $err" if $err;
4171
4172 print "migration listens on $migrate_uri\n" if $migrate_uri;
4173
4174 if ($statefile && $statefile ne 'tcp') {
4175 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4176 warn $@ if $@;
4177 }
4178
4179 if ($migratedfrom) {
4180
4181 eval {
4182 set_migration_caps($vmid);
4183 };
4184 warn $@ if $@;
4185
4186 if ($spice_port) {
4187 print "spice listens on port $spice_port\n";
4188 if ($spice_ticket) {
4189 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4190 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4191 }
4192 }
4193
4194 } else {
4195
4196 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4197 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4198 if $conf->{balloon};
4199 vm_mon_cmd_nocheck($vmid, 'qom-set',
4200 path => "machine/peripheral/balloon0",
4201 property => "guest-stats-polling-interval",
4202 value => 2);
4203 }
4204
4205 foreach my $opt (keys %$conf) {
4206 next if $opt !~ m/^net\d+$/;
4207 my $nicconf = parse_net($conf->{$opt});
4208 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4209 }
4210 }
4211 });
4212 }
4213
4214 sub vm_mon_cmd {
4215 my ($vmid, $execute, %params) = @_;
4216
4217 my $cmd = { execute => $execute, arguments => \%params };
4218 vm_qmp_command($vmid, $cmd);
4219 }
4220
4221 sub vm_mon_cmd_nocheck {
4222 my ($vmid, $execute, %params) = @_;
4223
4224 my $cmd = { execute => $execute, arguments => \%params };
4225 vm_qmp_command($vmid, $cmd, 1);
4226 }
4227
4228 sub vm_qmp_command {
4229 my ($vmid, $cmd, $nocheck) = @_;
4230
4231 my $res;
4232
4233 my $timeout;
4234 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4235 $timeout = $cmd->{arguments}->{timeout};
4236 delete $cmd->{arguments}->{timeout};
4237 }
4238
4239 eval {
4240 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4241 my $sname = qmp_socket($vmid);
4242 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4243 my $qmpclient = PVE::QMPClient->new();
4244
4245 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4246 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4247 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4248 if scalar(%{$cmd->{arguments}});
4249 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4250 } else {
4251 die "unable to open monitor socket\n";
4252 }
4253 };
4254 if (my $err = $@) {
4255 syslog("err", "VM $vmid qmp command failed - $err");
4256 die $err;
4257 }
4258
4259 return $res;
4260 }
4261
4262 sub vm_human_monitor_command {
4263 my ($vmid, $cmdline) = @_;
4264
4265 my $res;
4266
4267 my $cmd = {
4268 execute => 'human-monitor-command',
4269 arguments => { 'command-line' => $cmdline},
4270 };
4271
4272 return vm_qmp_command($vmid, $cmd);
4273 }
4274
4275 sub vm_commandline {
4276 my ($storecfg, $vmid) = @_;
4277
4278 my $conf = load_config($vmid);
4279
4280 my $defaults = load_defaults();
4281
4282 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4283
4284 return join(' ', @$cmd);
4285 }
4286
4287 sub vm_reset {
4288 my ($vmid, $skiplock) = @_;
4289
4290 lock_config($vmid, sub {
4291
4292 my $conf = load_config($vmid);
4293
4294 check_lock($conf) if !$skiplock;
4295
4296 vm_mon_cmd($vmid, "system_reset");
4297 });
4298 }
4299
4300 sub get_vm_volumes {
4301 my ($conf) = @_;
4302
4303 my $vollist = [];
4304 foreach_volid($conf, sub {
4305 my ($volid, $is_cdrom) = @_;
4306
4307 return if $volid =~ m|^/|;
4308
4309 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4310 return if !$sid;
4311
4312 push @$vollist, $volid;
4313 });
4314
4315 return $vollist;
4316 }
4317
4318 sub vm_stop_cleanup {
4319 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4320
4321 eval {
4322 fairsched_rmnod($vmid); # try to destroy group
4323
4324 if (!$keepActive) {
4325 my $vollist = get_vm_volumes($conf);
4326 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4327 }
4328
4329 foreach my $ext (qw(mon qmp pid vnc qga)) {
4330 unlink "/var/run/qemu-server/${vmid}.$ext";
4331 }
4332
4333 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4334 };
4335 warn $@ if $@; # avoid errors - just warn
4336 }
4337
4338 # Note: use $nockeck to skip tests if VM configuration file exists.
4339 # We need that when migration VMs to other nodes (files already moved)
4340 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4341 sub vm_stop {
4342 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4343
4344 $force = 1 if !defined($force) && !$shutdown;
4345
4346 if ($migratedfrom){
4347 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4348 kill 15, $pid if $pid;
4349 my $conf = load_config($vmid, $migratedfrom);
4350 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4351 return;
4352 }
4353
4354 lock_config($vmid, sub {
4355
4356 my $pid = check_running($vmid, $nocheck);
4357 return if !$pid;
4358
4359 my $conf;
4360 if (!$nocheck) {
4361 $conf = load_config($vmid);
4362 check_lock($conf) if !$skiplock;
4363 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4364 my $opts = parse_startup($conf->{startup});
4365 $timeout = $opts->{down} if $opts->{down};
4366 }
4367 }
4368
4369 $timeout = 60 if !defined($timeout);
4370
4371 eval {
4372 if ($shutdown) {
4373 if (defined($conf) && $conf->{agent}) {
4374 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4375 } else {
4376 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4377 }
4378 } else {
4379 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4380 }
4381 };
4382 my $err = $@;
4383
4384 if (!$err) {
4385 my $count = 0;
4386 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4387 $count++;
4388 sleep 1;
4389 }
4390
4391 if ($count >= $timeout) {
4392 if ($force) {
4393 warn "VM still running - terminating now with SIGTERM\n";
4394 kill 15, $pid;
4395 } else {
4396 die "VM quit/powerdown failed - got timeout\n";
4397 }
4398 } else {
4399 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4400 return;
4401 }
4402 } else {
4403 if ($force) {
4404 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4405 kill 15, $pid;
4406 } else {
4407 die "VM quit/powerdown failed\n";
4408 }
4409 }
4410
4411 # wait again
4412 $timeout = 10;
4413
4414 my $count = 0;
4415 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4416 $count++;
4417 sleep 1;
4418 }
4419
4420 if ($count >= $timeout) {
4421 warn "VM still running - terminating now with SIGKILL\n";
4422 kill 9, $pid;
4423 sleep 1;
4424 }
4425
4426 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4427 });
4428 }
4429
4430 sub vm_suspend {
4431 my ($vmid, $skiplock) = @_;
4432
4433 lock_config($vmid, sub {
4434
4435 my $conf = load_config($vmid);
4436
4437 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4438
4439 vm_mon_cmd($vmid, "stop");
4440 });
4441 }
4442
4443 sub vm_resume {
4444 my ($vmid, $skiplock) = @_;
4445
4446 lock_config($vmid, sub {
4447
4448 my $conf = load_config($vmid);
4449
4450 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4451
4452 vm_mon_cmd($vmid, "cont");
4453 });
4454 }
4455
4456 sub vm_sendkey {
4457 my ($vmid, $skiplock, $key) = @_;
4458
4459 lock_config($vmid, sub {
4460
4461 my $conf = load_config($vmid);
4462
4463 # there is no qmp command, so we use the human monitor command
4464 vm_human_monitor_command($vmid, "sendkey $key");
4465 });
4466 }
4467
4468 sub vm_destroy {
4469 my ($storecfg, $vmid, $skiplock) = @_;
4470
4471 lock_config($vmid, sub {
4472
4473 my $conf = load_config($vmid);
4474
4475 check_lock($conf) if !$skiplock;
4476
4477 if (!check_running($vmid)) {
4478 fairsched_rmnod($vmid); # try to destroy group
4479 destroy_vm($storecfg, $vmid);
4480 } else {
4481 die "VM $vmid is running - destroy failed\n";
4482 }
4483 });
4484 }
4485
4486 # pci helpers
4487
4488 sub file_write {
4489 my ($filename, $buf) = @_;
4490
4491 my $fh = IO::File->new($filename, "w");
4492 return undef if !$fh;
4493
4494 my $res = print $fh $buf;
4495
4496 $fh->close();
4497
4498 return $res;
4499 }
4500
4501 sub pci_device_info {
4502 my ($name) = @_;
4503
4504 my $res;
4505
4506 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4507 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4508
4509 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4510 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4511
4512 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4513 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4514
4515 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4516 return undef if !defined($product) || $product !~ s/^0x//;
4517
4518 $res = {
4519 name => $name,
4520 vendor => $vendor,
4521 product => $product,
4522 domain => $domain,
4523 bus => $bus,
4524 slot => $slot,
4525 func => $func,
4526 irq => $irq,
4527 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4528 };
4529
4530 return $res;
4531 }
4532
4533 sub pci_dev_reset {
4534 my ($dev) = @_;
4535
4536 my $name = $dev->{name};
4537
4538 my $fn = "$pcisysfs/devices/$name/reset";
4539
4540 return file_write($fn, "1");
4541 }
4542
4543 sub pci_dev_bind_to_stub {
4544 my ($dev) = @_;
4545
4546 my $name = $dev->{name};
4547
4548 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4549 return 1 if -d $testdir;
4550
4551 my $data = "$dev->{vendor} $dev->{product}";
4552 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4553
4554 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4555 if (!file_write($fn, $name)) {
4556 return undef if -f $fn;
4557 }
4558
4559 $fn = "$pcisysfs/drivers/pci-stub/bind";
4560 if (! -d $testdir) {
4561 return undef if !file_write($fn, $name);
4562 }
4563
4564 return -d $testdir;
4565 }
4566
4567 sub pci_dev_bind_to_vfio {
4568 my ($dev) = @_;
4569
4570 my $name = $dev->{name};
4571
4572 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4573
4574 if (!-d $vfio_basedir) {
4575 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4576 }
4577 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4578
4579 my $testdir = "$vfio_basedir/$name";
4580 return 1 if -d $testdir;
4581
4582 my $data = "$dev->{vendor} $dev->{product}";
4583 return undef if !file_write("$vfio_basedir/new_id", $data);
4584
4585 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4586 if (!file_write($fn, $name)) {
4587 return undef if -f $fn;
4588 }
4589
4590 $fn = "$vfio_basedir/bind";
4591 if (! -d $testdir) {
4592 return undef if !file_write($fn, $name);
4593 }
4594
4595 return -d $testdir;
4596 }
4597
4598 sub pci_dev_group_bind_to_vfio {
4599 my ($pciid) = @_;
4600
4601 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4602
4603 if (!-d $vfio_basedir) {
4604 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4605 }
4606 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4607
4608 # get IOMMU group devices
4609 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4610 my @devs = grep /^0000:/, readdir($D);
4611 closedir($D);
4612
4613 foreach my $pciid (@devs) {
4614 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4615
4616 # pci bridges, switches or root ports are not supported
4617 # they have a pci_bus subdirectory so skip them
4618 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4619
4620 my $info = pci_device_info($1);
4621 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4622 }
4623
4624 return 1;
4625 }
4626
4627 sub print_pci_addr {
4628 my ($id, $bridges) = @_;
4629
4630 my $res = '';
4631 my $devices = {
4632 piix3 => { bus => 0, addr => 1 },
4633 #addr2 : first videocard
4634 balloon0 => { bus => 0, addr => 3 },
4635 watchdog => { bus => 0, addr => 4 },
4636 scsihw0 => { bus => 0, addr => 5 },
4637 scsihw1 => { bus => 0, addr => 6 },
4638 ahci0 => { bus => 0, addr => 7 },
4639 qga0 => { bus => 0, addr => 8 },
4640 spice => { bus => 0, addr => 9 },
4641 virtio0 => { bus => 0, addr => 10 },
4642 virtio1 => { bus => 0, addr => 11 },
4643 virtio2 => { bus => 0, addr => 12 },
4644 virtio3 => { bus => 0, addr => 13 },
4645 virtio4 => { bus => 0, addr => 14 },
4646 virtio5 => { bus => 0, addr => 15 },
4647 hostpci0 => { bus => 0, addr => 16 },
4648 hostpci1 => { bus => 0, addr => 17 },
4649 net0 => { bus => 0, addr => 18 },
4650 net1 => { bus => 0, addr => 19 },
4651 net2 => { bus => 0, addr => 20 },
4652 net3 => { bus => 0, addr => 21 },
4653 net4 => { bus => 0, addr => 22 },
4654 net5 => { bus => 0, addr => 23 },
4655 vga1 => { bus => 0, addr => 24 },
4656 vga2 => { bus => 0, addr => 25 },
4657 vga3 => { bus => 0, addr => 26 },
4658 hostpci2 => { bus => 0, addr => 27 },
4659 hostpci3 => { bus => 0, addr => 28 },
4660 #addr29 : usb-host (pve-usb.cfg)
4661 'pci.1' => { bus => 0, addr => 30 },
4662 'pci.2' => { bus => 0, addr => 31 },
4663 'net6' => { bus => 1, addr => 1 },
4664 'net7' => { bus => 1, addr => 2 },
4665 'net8' => { bus => 1, addr => 3 },
4666 'net9' => { bus => 1, addr => 4 },
4667 'net10' => { bus => 1, addr => 5 },
4668 'net11' => { bus => 1, addr => 6 },
4669 'net12' => { bus => 1, addr => 7 },
4670 'net13' => { bus => 1, addr => 8 },
4671 'net14' => { bus => 1, addr => 9 },
4672 'net15' => { bus => 1, addr => 10 },
4673 'net16' => { bus => 1, addr => 11 },
4674 'net17' => { bus => 1, addr => 12 },
4675 'net18' => { bus => 1, addr => 13 },
4676 'net19' => { bus => 1, addr => 14 },
4677 'net20' => { bus => 1, addr => 15 },
4678 'net21' => { bus => 1, addr => 16 },
4679 'net22' => { bus => 1, addr => 17 },
4680 'net23' => { bus => 1, addr => 18 },
4681 'net24' => { bus => 1, addr => 19 },
4682 'net25' => { bus => 1, addr => 20 },
4683 'net26' => { bus => 1, addr => 21 },
4684 'net27' => { bus => 1, addr => 22 },
4685 'net28' => { bus => 1, addr => 23 },
4686 'net29' => { bus => 1, addr => 24 },
4687 'net30' => { bus => 1, addr => 25 },
4688 'net31' => { bus => 1, addr => 26 },
4689 'virtio6' => { bus => 2, addr => 1 },
4690 'virtio7' => { bus => 2, addr => 2 },
4691 'virtio8' => { bus => 2, addr => 3 },
4692 'virtio9' => { bus => 2, addr => 4 },
4693 'virtio10' => { bus => 2, addr => 5 },
4694 'virtio11' => { bus => 2, addr => 6 },
4695 'virtio12' => { bus => 2, addr => 7 },
4696 'virtio13' => { bus => 2, addr => 8 },
4697 'virtio14' => { bus => 2, addr => 9 },
4698 'virtio15' => { bus => 2, addr => 10 },
4699 };
4700
4701 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4702 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4703 my $bus = $devices->{$id}->{bus};
4704 $res = ",bus=pci.$bus,addr=$addr";
4705 $bridges->{$bus} = 1 if $bridges;
4706 }
4707 return $res;
4708
4709 }
4710
4711 sub print_pcie_addr {
4712 my ($id) = @_;
4713
4714 my $res = '';
4715 my $devices = {
4716 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4717 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4718 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4719 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4720 };
4721
4722 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4723 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4724 my $bus = $devices->{$id}->{bus};
4725 $res = ",bus=$bus,addr=$addr";
4726 }
4727 return $res;
4728
4729 }
4730
4731 # vzdump restore implementaion
4732
4733 sub tar_archive_read_firstfile {
4734 my $archive = shift;
4735
4736 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4737
4738 # try to detect archive type first
4739 my $pid = open (TMP, "tar tf '$archive'|") ||
4740 die "unable to open file '$archive'\n";
4741 my $firstfile = <TMP>;
4742 kill 15, $pid;
4743 close TMP;
4744
4745 die "ERROR: archive contaions no data\n" if !$firstfile;
4746 chomp $firstfile;
4747
4748 return $firstfile;
4749 }
4750
4751 sub tar_restore_cleanup {
4752 my ($storecfg, $statfile) = @_;
4753
4754 print STDERR "starting cleanup\n";
4755
4756 if (my $fd = IO::File->new($statfile, "r")) {
4757 while (defined(my $line = <$fd>)) {
4758 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4759 my $volid = $2;
4760 eval {
4761 if ($volid =~ m|^/|) {
4762 unlink $volid || die 'unlink failed\n';
4763 } else {
4764 PVE::Storage::vdisk_free($storecfg, $volid);
4765 }
4766 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4767 };
4768 print STDERR "unable to cleanup '$volid' - $@" if $@;
4769 } else {
4770 print STDERR "unable to parse line in statfile - $line";
4771 }
4772 }
4773 $fd->close();
4774 }
4775 }
4776
4777 sub restore_archive {
4778 my ($archive, $vmid, $user, $opts) = @_;
4779
4780 my $format = $opts->{format};
4781 my $comp;
4782
4783 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4784 $format = 'tar' if !$format;
4785 $comp = 'gzip';
4786 } elsif ($archive =~ m/\.tar$/) {
4787 $format = 'tar' if !$format;
4788 } elsif ($archive =~ m/.tar.lzo$/) {
4789 $format = 'tar' if !$format;
4790 $comp = 'lzop';
4791 } elsif ($archive =~ m/\.vma$/) {
4792 $format = 'vma' if !$format;
4793 } elsif ($archive =~ m/\.vma\.gz$/) {
4794 $format = 'vma' if !$format;
4795 $comp = 'gzip';
4796 } elsif ($archive =~ m/\.vma\.lzo$/) {
4797 $format = 'vma' if !$format;
4798 $comp = 'lzop';
4799 } else {
4800 $format = 'vma' if !$format; # default
4801 }
4802
4803 # try to detect archive format
4804 if ($format eq 'tar') {
4805 return restore_tar_archive($archive, $vmid, $user, $opts);
4806 } else {
4807 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4808 }
4809 }
4810
4811 sub restore_update_config_line {
4812 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4813
4814 return if $line =~ m/^\#qmdump\#/;
4815 return if $line =~ m/^\#vzdump\#/;
4816 return if $line =~ m/^lock:/;
4817 return if $line =~ m/^unused\d+:/;
4818 return if $line =~ m/^parent:/;
4819 return if $line =~ m/^template:/; # restored VM is never a template
4820
4821 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4822 # try to convert old 1.X settings
4823 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4824 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4825 my ($model, $macaddr) = split(/\=/, $devconfig);
4826 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4827 my $net = {
4828 model => $model,
4829 bridge => "vmbr$ind",
4830 macaddr => $macaddr,
4831 };
4832 my $netstr = print_net($net);
4833
4834 print $outfd "net$cookie->{netcount}: $netstr\n";
4835 $cookie->{netcount}++;
4836 }
4837 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4838 my ($id, $netstr) = ($1, $2);
4839 my $net = parse_net($netstr);
4840 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4841 $netstr = print_net($net);
4842 print $outfd "$id: $netstr\n";
4843 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4844 my $virtdev = $1;
4845 my $value = $3;
4846 if ($line =~ m/backup=no/) {
4847 print $outfd "#$line";
4848 } elsif ($virtdev && $map->{$virtdev}) {
4849 my $di = parse_drive($virtdev, $value);
4850 delete $di->{format}; # format can change on restore
4851 $di->{file} = $map->{$virtdev};
4852 $value = print_drive($vmid, $di);
4853 print $outfd "$virtdev: $value\n";
4854 } else {
4855 print $outfd $line;
4856 }
4857 } else {
4858 print $outfd $line;
4859 }
4860 }
4861
4862 sub scan_volids {
4863 my ($cfg, $vmid) = @_;
4864
4865 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4866
4867 my $volid_hash = {};
4868 foreach my $storeid (keys %$info) {
4869 foreach my $item (@{$info->{$storeid}}) {
4870 next if !($item->{volid} && $item->{size});
4871 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4872 $volid_hash->{$item->{volid}} = $item;
4873 }
4874 }
4875
4876 return $volid_hash;
4877 }
4878
4879 sub get_used_paths {
4880 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4881
4882 my $used_path = {};
4883
4884 my $scan_config = sub {
4885 my ($cref, $snapname) = @_;
4886
4887 foreach my $key (keys %$cref) {
4888 my $value = $cref->{$key};
4889 if (valid_drivename($key)) {
4890 next if $skip_drive && $key eq $skip_drive;
4891 my $drive = parse_drive($key, $value);
4892 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4893 if ($drive->{file} =~ m!^/!) {
4894 $used_path->{$drive->{file}}++; # = 1;
4895 } else {
4896 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4897 next if !$storeid;
4898 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4899 next if !$scfg;
4900 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4901 $used_path->{$path}++; # = 1;
4902 }
4903 }
4904 }
4905 };
4906
4907 &$scan_config($conf);
4908
4909 undef $skip_drive;
4910
4911 if ($scan_snapshots) {
4912 foreach my $snapname (keys %{$conf->{snapshots}}) {
4913 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4914 }
4915 }
4916
4917 return $used_path;
4918 }
4919
4920 sub update_disksize {
4921 my ($vmid, $conf, $volid_hash) = @_;
4922
4923 my $changes;
4924
4925 my $used = {};
4926
4927 # Note: it is allowed to define multiple storages with same path (alias), so
4928 # we need to check both 'volid' and real 'path' (two different volid can point
4929 # to the same path).
4930
4931 my $usedpath = {};
4932
4933 # update size info
4934 foreach my $opt (keys %$conf) {
4935 if (valid_drivename($opt)) {
4936 my $drive = parse_drive($opt, $conf->{$opt});
4937 my $volid = $drive->{file};
4938 next if !$volid;
4939
4940 $used->{$volid} = 1;
4941 if ($volid_hash->{$volid} &&
4942 (my $path = $volid_hash->{$volid}->{path})) {
4943 $usedpath->{$path} = 1;
4944 }
4945
4946 next if drive_is_cdrom($drive);
4947 next if !$volid_hash->{$volid};
4948
4949 $drive->{size} = $volid_hash->{$volid}->{size};
4950 my $new = print_drive($vmid, $drive);
4951 if ($new ne $conf->{$opt}) {
4952 $changes = 1;
4953 $conf->{$opt} = $new;
4954 }
4955 }
4956 }
4957
4958 # remove 'unusedX' entry if volume is used
4959 foreach my $opt (keys %$conf) {
4960 next if $opt !~ m/^unused\d+$/;
4961 my $volid = $conf->{$opt};
4962 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4963 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4964 $changes = 1;
4965 delete $conf->{$opt};
4966 }
4967 }
4968
4969 foreach my $volid (sort keys %$volid_hash) {
4970 next if $volid =~ m/vm-$vmid-state-/;
4971 next if $used->{$volid};
4972 my $path = $volid_hash->{$volid}->{path};
4973 next if !$path; # just to be sure
4974 next if $usedpath->{$path};
4975 $changes = 1;
4976 add_unused_volume($conf, $volid);
4977 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4978 }
4979
4980 return $changes;
4981 }
4982
4983 sub rescan {
4984 my ($vmid, $nolock) = @_;
4985
4986 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4987
4988 my $volid_hash = scan_volids($cfg, $vmid);
4989
4990 my $updatefn = sub {
4991 my ($vmid) = @_;
4992
4993 my $conf = load_config($vmid);
4994
4995 check_lock($conf);
4996
4997 my $vm_volids = {};
4998 foreach my $volid (keys %$volid_hash) {
4999 my $info = $volid_hash->{$volid};
5000 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5001 }
5002
5003 my $changes = update_disksize($vmid, $conf, $vm_volids);
5004
5005 update_config_nolock($vmid, $conf, 1) if $changes;
5006 };
5007
5008 if (defined($vmid)) {
5009 if ($nolock) {
5010 &$updatefn($vmid);
5011 } else {
5012 lock_config($vmid, $updatefn, $vmid);
5013 }
5014 } else {
5015 my $vmlist = config_list();
5016 foreach my $vmid (keys %$vmlist) {
5017 if ($nolock) {
5018 &$updatefn($vmid);
5019 } else {
5020 lock_config($vmid, $updatefn, $vmid);
5021 }
5022 }
5023 }
5024 }
5025
5026 sub restore_vma_archive {
5027 my ($archive, $vmid, $user, $opts, $comp) = @_;
5028
5029 my $input = $archive eq '-' ? "<&STDIN" : undef;
5030 my $readfrom = $archive;
5031
5032 my $uncomp = '';
5033 if ($comp) {
5034 $readfrom = '-';
5035 my $qarchive = PVE::Tools::shellquote($archive);
5036 if ($comp eq 'gzip') {
5037 $uncomp = "zcat $qarchive|";
5038 } elsif ($comp eq 'lzop') {
5039 $uncomp = "lzop -d -c $qarchive|";
5040 } else {
5041 die "unknown compression method '$comp'\n";
5042 }
5043
5044 }
5045
5046 my $tmpdir = "/var/tmp/vzdumptmp$$";
5047 rmtree $tmpdir;
5048
5049 # disable interrupts (always do cleanups)
5050 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5051 warn "got interrupt - ignored\n";
5052 };
5053
5054 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5055 POSIX::mkfifo($mapfifo, 0600);
5056 my $fifofh;
5057
5058 my $openfifo = sub {
5059 open($fifofh, '>', $mapfifo) || die $!;
5060 };
5061
5062 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5063
5064 my $oldtimeout;
5065 my $timeout = 5;
5066
5067 my $devinfo = {};
5068
5069 my $rpcenv = PVE::RPCEnvironment::get();
5070
5071 my $conffile = config_file($vmid);
5072 my $tmpfn = "$conffile.$$.tmp";
5073
5074 # Note: $oldconf is undef if VM does not exists
5075 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5076
5077 my $print_devmap = sub {
5078 my $virtdev_hash = {};
5079
5080 my $cfgfn = "$tmpdir/qemu-server.conf";
5081
5082 # we can read the config - that is already extracted
5083 my $fh = IO::File->new($cfgfn, "r") ||
5084 "unable to read qemu-server.conf - $!\n";
5085
5086 while (defined(my $line = <$fh>)) {
5087 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5088 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5089 die "archive does not contain data for drive '$virtdev'\n"
5090 if !$devinfo->{$devname};
5091 if (defined($opts->{storage})) {
5092 $storeid = $opts->{storage} || 'local';
5093 } elsif (!$storeid) {
5094 $storeid = 'local';
5095 }
5096 $format = 'raw' if !$format;
5097 $devinfo->{$devname}->{devname} = $devname;
5098 $devinfo->{$devname}->{virtdev} = $virtdev;
5099 $devinfo->{$devname}->{format} = $format;
5100 $devinfo->{$devname}->{storeid} = $storeid;
5101
5102 # check permission on storage
5103 my $pool = $opts->{pool}; # todo: do we need that?
5104 if ($user ne 'root@pam') {
5105 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5106 }
5107
5108 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5109 }
5110 }
5111
5112 foreach my $devname (keys %$devinfo) {
5113 die "found no device mapping information for device '$devname'\n"
5114 if !$devinfo->{$devname}->{virtdev};
5115 }
5116
5117 my $cfg = cfs_read_file('storage.cfg');
5118
5119 # create empty/temp config
5120 if ($oldconf) {
5121 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5122 foreach_drive($oldconf, sub {
5123 my ($ds, $drive) = @_;
5124
5125 return if drive_is_cdrom($drive);
5126
5127 my $volid = $drive->{file};
5128
5129 return if !$volid || $volid =~ m|^/|;
5130
5131 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5132 return if !$path || !$owner || ($owner != $vmid);
5133
5134 # Note: only delete disk we want to restore
5135 # other volumes will become unused
5136 if ($virtdev_hash->{$ds}) {
5137 PVE::Storage::vdisk_free($cfg, $volid);
5138 }
5139 });
5140 }
5141
5142 my $map = {};
5143 foreach my $virtdev (sort keys %$virtdev_hash) {
5144 my $d = $virtdev_hash->{$virtdev};
5145 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5146 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5147
5148 # test if requested format is supported
5149 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5150 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5151 $d->{format} = $defFormat if !$supported;
5152
5153 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5154 $d->{format}, undef, $alloc_size);
5155 print STDERR "new volume ID is '$volid'\n";
5156 $d->{volid} = $volid;
5157 my $path = PVE::Storage::path($cfg, $volid);
5158
5159 my $write_zeros = 1;
5160 # fixme: what other storages types initialize volumes with zero?
5161 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5162 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5163 $write_zeros = 0;
5164 }
5165
5166 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5167
5168 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5169 $map->{$virtdev} = $volid;
5170 }
5171
5172 $fh->seek(0, 0) || die "seek failed - $!\n";
5173
5174 my $outfd = new IO::File ($tmpfn, "w") ||
5175 die "unable to write config for VM $vmid\n";
5176
5177 my $cookie = { netcount => 0 };
5178 while (defined(my $line = <$fh>)) {
5179 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5180 }
5181
5182 $fh->close();
5183 $outfd->close();
5184 };
5185
5186 eval {
5187 # enable interrupts
5188 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5189 die "interrupted by signal\n";
5190 };
5191 local $SIG{ALRM} = sub { die "got timeout\n"; };
5192
5193 $oldtimeout = alarm($timeout);
5194
5195 my $parser = sub {
5196 my $line = shift;
5197
5198 print "$line\n";
5199
5200 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5201 my ($dev_id, $size, $devname) = ($1, $2, $3);
5202 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5203 } elsif ($line =~ m/^CTIME: /) {
5204 # we correctly received the vma config, so we can disable
5205 # the timeout now for disk allocation (set to 10 minutes, so
5206 # that we always timeout if something goes wrong)
5207 alarm(600);
5208 &$print_devmap();
5209 print $fifofh "done\n";
5210 my $tmp = $oldtimeout || 0;
5211 $oldtimeout = undef;
5212 alarm($tmp);
5213 close($fifofh);
5214 }
5215 };
5216
5217 print "restore vma archive: $cmd\n";
5218 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5219 };
5220 my $err = $@;
5221
5222 alarm($oldtimeout) if $oldtimeout;
5223
5224 unlink $mapfifo;
5225
5226 if ($err) {
5227 rmtree $tmpdir;
5228 unlink $tmpfn;
5229
5230 my $cfg = cfs_read_file('storage.cfg');
5231 foreach my $devname (keys %$devinfo) {
5232 my $volid = $devinfo->{$devname}->{volid};
5233 next if !$volid;
5234 eval {
5235 if ($volid =~ m|^/|) {
5236 unlink $volid || die 'unlink failed\n';
5237 } else {
5238 PVE::Storage::vdisk_free($cfg, $volid);
5239 }
5240 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5241 };
5242 print STDERR "unable to cleanup '$volid' - $@" if $@;
5243 }
5244 die $err;
5245 }
5246
5247 rmtree $tmpdir;
5248
5249 rename($tmpfn, $conffile) ||
5250 die "unable to commit configuration file '$conffile'\n";
5251
5252 PVE::Cluster::cfs_update(); # make sure we read new file
5253
5254 eval { rescan($vmid, 1); };
5255 warn $@ if $@;
5256 }
5257
5258 sub restore_tar_archive {
5259 my ($archive, $vmid, $user, $opts) = @_;
5260
5261 if ($archive ne '-') {
5262 my $firstfile = tar_archive_read_firstfile($archive);
5263 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5264 if $firstfile ne 'qemu-server.conf';
5265 }
5266
5267 my $storecfg = cfs_read_file('storage.cfg');
5268
5269 # destroy existing data - keep empty config
5270 my $vmcfgfn = config_file($vmid);
5271 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5272
5273 my $tocmd = "/usr/lib/qemu-server/qmextract";
5274
5275 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5276 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5277 $tocmd .= ' --prealloc' if $opts->{prealloc};
5278 $tocmd .= ' --info' if $opts->{info};
5279
5280 # tar option "xf" does not autodetect compression when read from STDIN,
5281 # so we pipe to zcat
5282 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5283 PVE::Tools::shellquote("--to-command=$tocmd");
5284
5285 my $tmpdir = "/var/tmp/vzdumptmp$$";
5286 mkpath $tmpdir;
5287
5288 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5289 local $ENV{VZDUMP_VMID} = $vmid;
5290 local $ENV{VZDUMP_USER} = $user;
5291
5292 my $conffile = config_file($vmid);
5293 my $tmpfn = "$conffile.$$.tmp";
5294
5295 # disable interrupts (always do cleanups)
5296 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5297 print STDERR "got interrupt - ignored\n";
5298 };
5299
5300 eval {
5301 # enable interrupts
5302 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5303 die "interrupted by signal\n";
5304 };
5305
5306 if ($archive eq '-') {
5307 print "extracting archive from STDIN\n";
5308 run_command($cmd, input => "<&STDIN");
5309 } else {
5310 print "extracting archive '$archive'\n";
5311 run_command($cmd);
5312 }
5313
5314 return if $opts->{info};
5315
5316 # read new mapping
5317 my $map = {};
5318 my $statfile = "$tmpdir/qmrestore.stat";
5319 if (my $fd = IO::File->new($statfile, "r")) {
5320 while (defined (my $line = <$fd>)) {
5321 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5322 $map->{$1} = $2 if $1;
5323 } else {
5324 print STDERR "unable to parse line in statfile - $line\n";
5325 }
5326 }
5327 $fd->close();
5328 }
5329
5330 my $confsrc = "$tmpdir/qemu-server.conf";
5331
5332 my $srcfd = new IO::File($confsrc, "r") ||
5333 die "unable to open file '$confsrc'\n";
5334
5335 my $outfd = new IO::File ($tmpfn, "w") ||
5336 die "unable to write config for VM $vmid\n";
5337
5338 my $cookie = { netcount => 0 };
5339 while (defined (my $line = <$srcfd>)) {
5340 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5341 }
5342
5343 $srcfd->close();
5344 $outfd->close();
5345 };
5346 my $err = $@;
5347
5348 if ($err) {
5349
5350 unlink $tmpfn;
5351
5352 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5353
5354 die $err;
5355 }
5356
5357 rmtree $tmpdir;
5358
5359 rename $tmpfn, $conffile ||
5360 die "unable to commit configuration file '$conffile'\n";
5361
5362 PVE::Cluster::cfs_update(); # make sure we read new file
5363
5364 eval { rescan($vmid, 1); };
5365 warn $@ if $@;
5366 };
5367
5368
5369 # Internal snapshots
5370
5371 # NOTE: Snapshot create/delete involves several non-atomic
5372 # action, and can take a long time.
5373 # So we try to avoid locking the file and use 'lock' variable
5374 # inside the config file instead.
5375
5376 my $snapshot_copy_config = sub {
5377 my ($source, $dest) = @_;
5378
5379 foreach my $k (keys %$source) {
5380 next if $k eq 'snapshots';
5381 next if $k eq 'snapstate';
5382 next if $k eq 'snaptime';
5383 next if $k eq 'vmstate';
5384 next if $k eq 'lock';
5385 next if $k eq 'digest';
5386 next if $k eq 'description';
5387 next if $k =~ m/^unused\d+$/;
5388
5389 $dest->{$k} = $source->{$k};
5390 }
5391 };
5392
5393 my $snapshot_apply_config = sub {
5394 my ($conf, $snap) = @_;
5395
5396 # copy snapshot list
5397 my $newconf = {
5398 snapshots => $conf->{snapshots},
5399 };
5400
5401 # keep description and list of unused disks
5402 foreach my $k (keys %$conf) {
5403 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5404 $newconf->{$k} = $conf->{$k};
5405 }
5406
5407 &$snapshot_copy_config($snap, $newconf);
5408
5409 return $newconf;
5410 };
5411
5412 sub foreach_writable_storage {
5413 my ($conf, $func) = @_;
5414
5415 my $sidhash = {};
5416
5417 foreach my $ds (keys %$conf) {
5418 next if !valid_drivename($ds);
5419
5420 my $drive = parse_drive($ds, $conf->{$ds});
5421 next if !$drive;
5422 next if drive_is_cdrom($drive);
5423
5424 my $volid = $drive->{file};
5425
5426 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5427 $sidhash->{$sid} = $sid if $sid;
5428 }
5429
5430 foreach my $sid (sort keys %$sidhash) {
5431 &$func($sid);
5432 }
5433 }
5434
5435 my $alloc_vmstate_volid = sub {
5436 my ($storecfg, $vmid, $conf, $snapname) = @_;
5437
5438 # Note: we try to be smart when selecting a $target storage
5439
5440 my $target;
5441
5442 # search shared storage first
5443 foreach_writable_storage($conf, sub {
5444 my ($sid) = @_;
5445 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5446 return if !$scfg->{shared};
5447
5448 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5449 });
5450
5451 if (!$target) {
5452 # now search local storage
5453 foreach_writable_storage($conf, sub {
5454 my ($sid) = @_;
5455 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5456 return if $scfg->{shared};
5457
5458 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5459 });
5460 }
5461
5462 $target = 'local' if !$target;
5463
5464 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5465 # we abort live save after $conf->{memory}, so we need at max twice that space
5466 my $size = $conf->{memory}*2 + $driver_state_size;
5467
5468 my $name = "vm-$vmid-state-$snapname";
5469 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5470 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5471 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5472
5473 return $volid;
5474 };
5475
5476 my $snapshot_prepare = sub {
5477 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5478
5479 my $snap;
5480
5481 my $updatefn = sub {
5482
5483 my $conf = load_config($vmid);
5484
5485 die "you can't take a snapshot if it's a template\n"
5486 if is_template($conf);
5487
5488 check_lock($conf);
5489
5490 $conf->{lock} = 'snapshot';
5491
5492 die "snapshot name '$snapname' already used\n"
5493 if defined($conf->{snapshots}->{$snapname});
5494
5495 my $storecfg = PVE::Storage::config();
5496 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5497
5498 $snap = $conf->{snapshots}->{$snapname} = {};
5499
5500 if ($save_vmstate && check_running($vmid)) {
5501 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5502 }
5503
5504 &$snapshot_copy_config($conf, $snap);
5505
5506 $snap->{snapstate} = "prepare";
5507 $snap->{snaptime} = time();
5508 $snap->{description} = $comment if $comment;
5509
5510 # always overwrite machine if we save vmstate. This makes sure we
5511 # can restore it later using correct machine type
5512 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5513
5514 update_config_nolock($vmid, $conf, 1);
5515 };
5516
5517 lock_config($vmid, $updatefn);
5518
5519 return $snap;
5520 };
5521
5522 my $snapshot_commit = sub {
5523 my ($vmid, $snapname) = @_;
5524
5525 my $updatefn = sub {
5526
5527 my $conf = load_config($vmid);
5528
5529 die "missing snapshot lock\n"
5530 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5531
5532 my $has_machine_config = defined($conf->{machine});
5533
5534 my $snap = $conf->{snapshots}->{$snapname};
5535
5536 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5537
5538 die "wrong snapshot state\n"
5539 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5540
5541 delete $snap->{snapstate};
5542 delete $conf->{lock};
5543
5544 my $newconf = &$snapshot_apply_config($conf, $snap);
5545
5546 delete $newconf->{machine} if !$has_machine_config;
5547
5548 $newconf->{parent} = $snapname;
5549
5550 update_config_nolock($vmid, $newconf, 1);
5551 };
5552
5553 lock_config($vmid, $updatefn);
5554 };
5555
5556 sub snapshot_rollback {
5557 my ($vmid, $snapname) = @_;
5558
5559 my $prepare = 1;
5560
5561 my $storecfg = PVE::Storage::config();
5562
5563 my $conf = load_config($vmid);
5564
5565 my $get_snapshot_config = sub {
5566
5567 die "you can't rollback if vm is a template\n" if is_template($conf);
5568
5569 my $res = $conf->{snapshots}->{$snapname};
5570
5571 die "snapshot '$snapname' does not exist\n" if !defined($res);
5572
5573 return $res;
5574 };
5575
5576 my $snap = &$get_snapshot_config();
5577
5578 foreach_drive($snap, sub {
5579 my ($ds, $drive) = @_;
5580
5581 return if drive_is_cdrom($drive);
5582
5583 my $volid = $drive->{file};
5584
5585 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5586 });
5587
5588 my $updatefn = sub {
5589
5590 $conf = load_config($vmid);
5591
5592 $snap = &$get_snapshot_config();
5593
5594 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5595 if $snap->{snapstate};
5596
5597 if ($prepare) {
5598 check_lock($conf);
5599 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5600 }
5601
5602 die "unable to rollback vm $vmid: vm is running\n"
5603 if check_running($vmid);
5604
5605 if ($prepare) {
5606 $conf->{lock} = 'rollback';
5607 } else {
5608 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5609 delete $conf->{lock};
5610 }
5611
5612 my $forcemachine;
5613
5614 if (!$prepare) {
5615 my $has_machine_config = defined($conf->{machine});
5616
5617 # copy snapshot config to current config
5618 $conf = &$snapshot_apply_config($conf, $snap);
5619 $conf->{parent} = $snapname;
5620
5621 # Note: old code did not store 'machine', so we try to be smart
5622 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5623 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5624 # we remove the 'machine' configuration if not explicitly specified
5625 # in the original config.
5626 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5627 }
5628
5629 update_config_nolock($vmid, $conf, 1);
5630
5631 if (!$prepare && $snap->{vmstate}) {
5632 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5633 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5634 }
5635 };
5636
5637 lock_config($vmid, $updatefn);
5638
5639 foreach_drive($snap, sub {
5640 my ($ds, $drive) = @_;
5641
5642 return if drive_is_cdrom($drive);
5643
5644 my $volid = $drive->{file};
5645 my $device = "drive-$ds";
5646
5647 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5648 });
5649
5650 $prepare = 0;
5651 lock_config($vmid, $updatefn);
5652 }
5653
5654 my $savevm_wait = sub {
5655 my ($vmid) = @_;
5656
5657 for(;;) {
5658 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5659 if (!$stat->{status}) {
5660 die "savevm not active\n";
5661 } elsif ($stat->{status} eq 'active') {
5662 sleep(1);
5663 next;
5664 } elsif ($stat->{status} eq 'completed') {
5665 last;
5666 } else {
5667 die "query-savevm returned status '$stat->{status}'\n";
5668 }
5669 }
5670 };
5671
5672 sub snapshot_create {
5673 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5674
5675 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5676
5677 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5678
5679 my $config = load_config($vmid);
5680
5681 my $running = check_running($vmid);
5682
5683 my $freezefs = $running && $config->{agent};
5684 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5685
5686 my $drivehash = {};
5687
5688 if ($freezefs) {
5689 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5690 warn "guest-fsfreeze-freeze problems - $@" if $@;
5691 }
5692
5693 eval {
5694 # create internal snapshots of all drives
5695
5696 my $storecfg = PVE::Storage::config();
5697
5698 if ($running) {
5699 if ($snap->{vmstate}) {
5700 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5701 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5702 &$savevm_wait($vmid);
5703 } else {
5704 vm_mon_cmd($vmid, "savevm-start");
5705 }
5706 };
5707
5708 foreach_drive($snap, sub {
5709 my ($ds, $drive) = @_;
5710
5711 return if drive_is_cdrom($drive);
5712
5713 my $volid = $drive->{file};
5714 my $device = "drive-$ds";
5715
5716 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5717 $drivehash->{$ds} = 1;
5718 });
5719 };
5720 my $err = $@;
5721
5722 if ($running) {
5723 eval { vm_mon_cmd($vmid, "savevm-end") };
5724 warn $@ if $@;
5725
5726 if ($freezefs) {
5727 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5728 warn "guest-fsfreeze-thaw problems - $@" if $@;
5729 }
5730
5731 # savevm-end is async, we need to wait
5732 for (;;) {
5733 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5734 if (!$stat->{bytes}) {
5735 last;
5736 } else {
5737 print "savevm not yet finished\n";
5738 sleep(1);
5739 next;
5740 }
5741 }
5742 }
5743
5744 if ($err) {
5745 warn "snapshot create failed: starting cleanup\n";
5746 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5747 warn $@ if $@;
5748 die $err;
5749 }
5750
5751 &$snapshot_commit($vmid, $snapname);
5752 }
5753
5754 # Note: $drivehash is only set when called from snapshot_create.
5755 sub snapshot_delete {
5756 my ($vmid, $snapname, $force, $drivehash) = @_;
5757
5758 my $prepare = 1;
5759
5760 my $snap;
5761 my $unused = [];
5762
5763 my $unlink_parent = sub {
5764 my ($confref, $new_parent) = @_;
5765
5766 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5767 if ($new_parent) {
5768 $confref->{parent} = $new_parent;
5769 } else {
5770 delete $confref->{parent};
5771 }
5772 }
5773 };
5774
5775 my $updatefn = sub {
5776 my ($remove_drive) = @_;
5777
5778 my $conf = load_config($vmid);
5779
5780 if (!$drivehash) {
5781 check_lock($conf);
5782 die "you can't delete a snapshot if vm is a template\n"
5783 if is_template($conf);
5784 }
5785
5786 $snap = $conf->{snapshots}->{$snapname};
5787
5788 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5789
5790 # remove parent refs
5791 if (!$prepare) {
5792 &$unlink_parent($conf, $snap->{parent});
5793 foreach my $sn (keys %{$conf->{snapshots}}) {
5794 next if $sn eq $snapname;
5795 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5796 }
5797 }
5798
5799 if ($remove_drive) {
5800 if ($remove_drive eq 'vmstate') {
5801 delete $snap->{$remove_drive};
5802 } else {
5803 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5804 my $volid = $drive->{file};
5805 delete $snap->{$remove_drive};
5806 add_unused_volume($conf, $volid);
5807 }
5808 }
5809
5810 if ($prepare) {
5811 $snap->{snapstate} = 'delete';
5812 } else {
5813 delete $conf->{snapshots}->{$snapname};
5814 delete $conf->{lock} if $drivehash;
5815 foreach my $volid (@$unused) {
5816 add_unused_volume($conf, $volid);
5817 }
5818 }
5819
5820 update_config_nolock($vmid, $conf, 1);
5821 };
5822
5823 lock_config($vmid, $updatefn);
5824
5825 # now remove vmstate file
5826
5827 my $storecfg = PVE::Storage::config();
5828
5829 if ($snap->{vmstate}) {
5830 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5831 if (my $err = $@) {
5832 die $err if !$force;
5833 warn $err;
5834 }
5835 # save changes (remove vmstate from snapshot)
5836 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5837 };
5838
5839 # now remove all internal snapshots
5840 foreach_drive($snap, sub {
5841 my ($ds, $drive) = @_;
5842
5843 return if drive_is_cdrom($drive);
5844
5845 my $volid = $drive->{file};
5846 my $device = "drive-$ds";
5847
5848 if (!$drivehash || $drivehash->{$ds}) {
5849 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5850 if (my $err = $@) {
5851 die $err if !$force;
5852 warn $err;
5853 }
5854 }
5855
5856 # save changes (remove drive fron snapshot)
5857 lock_config($vmid, $updatefn, $ds) if !$force;
5858 push @$unused, $volid;
5859 });
5860
5861 # now cleanup config
5862 $prepare = 0;
5863 lock_config($vmid, $updatefn);
5864 }
5865
5866 sub has_feature {
5867 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5868
5869 my $err;
5870 foreach_drive($conf, sub {
5871 my ($ds, $drive) = @_;
5872
5873 return if drive_is_cdrom($drive);
5874 my $volid = $drive->{file};
5875 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5876 });
5877
5878 return $err ? 0 : 1;
5879 }
5880
5881 sub template_create {
5882 my ($vmid, $conf, $disk) = @_;
5883
5884 my $storecfg = PVE::Storage::config();
5885
5886 foreach_drive($conf, sub {
5887 my ($ds, $drive) = @_;
5888
5889 return if drive_is_cdrom($drive);
5890 return if $disk && $ds ne $disk;
5891
5892 my $volid = $drive->{file};
5893 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5894
5895 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5896 $drive->{file} = $voliddst;
5897 $conf->{$ds} = print_drive($vmid, $drive);
5898 update_config_nolock($vmid, $conf, 1);
5899 });
5900 }
5901
5902 sub is_template {
5903 my ($conf) = @_;
5904
5905 return 1 if defined $conf->{template} && $conf->{template} == 1;
5906 }
5907
5908 sub qemu_img_convert {
5909 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5910
5911 my $storecfg = PVE::Storage::config();
5912 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5913 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5914
5915 if ($src_storeid && $dst_storeid) {
5916 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5917 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5918
5919 my $src_format = qemu_img_format($src_scfg, $src_volname);
5920 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5921
5922 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5923 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5924
5925 my $cmd = [];
5926 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5927 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5928 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5929
5930 my $parser = sub {
5931 my $line = shift;
5932 if($line =~ m/\((\S+)\/100\%\)/){
5933 my $percent = $1;
5934 my $transferred = int($size * $percent / 100);
5935 my $remaining = $size - $transferred;
5936
5937 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5938 }
5939
5940 };
5941
5942 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5943 my $err = $@;
5944 die "copy failed: $err" if $err;
5945 }
5946 }
5947
5948 sub qemu_img_format {
5949 my ($scfg, $volname) = @_;
5950
5951 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5952 return $1;
5953 } elsif ($scfg->{type} eq 'iscsi') {
5954 return "host_device";
5955 } else {
5956 return "raw";
5957 }
5958 }
5959
5960 sub qemu_drive_mirror {
5961 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5962
5963 my $count = 0;
5964 my $old_len = 0;
5965 my $frozen = undef;
5966 my $maxwait = 120;
5967
5968 my $storecfg = PVE::Storage::config();
5969 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5970
5971 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5972
5973 my $format;
5974 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5975 $format = $1;
5976 }
5977
5978 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5979
5980 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5981 $opts->{format} = $format if $format;
5982
5983 #fixme : sometime drive-mirror timeout, but works fine after.
5984 # (I have see the problem with big volume > 200GB), so we need to eval
5985 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5986 # ignore errors here
5987
5988 eval {
5989 while (1) {
5990 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5991 my $stat = @$stats[0];
5992 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5993 die "error job is not mirroring" if $stat->{type} ne "mirror";
5994
5995 my $busy = $stat->{busy};
5996
5997 if (my $total = $stat->{len}) {
5998 my $transferred = $stat->{offset} || 0;
5999 my $remaining = $total - $transferred;
6000 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6001
6002 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
6003 }
6004
6005 if ($stat->{len} == $stat->{offset}) {
6006 if ($busy eq 'false') {
6007
6008 last if $vmiddst != $vmid;
6009
6010 # try to switch the disk if source and destination are on the same guest
6011 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6012 last if !$@;
6013 die $@ if $@ !~ m/cannot be completed/;
6014 }
6015
6016 if ($count > $maxwait) {
6017 # if too much writes to disk occurs at the end of migration
6018 #the disk needs to be freezed to be able to complete the migration
6019 vm_suspend($vmid,1);
6020 $frozen = 1;
6021 }
6022 $count ++
6023 }
6024 $old_len = $stat->{offset};
6025 sleep 1;
6026 }
6027
6028 vm_resume($vmid, 1) if $frozen;
6029
6030 };
6031 my $err = $@;
6032
6033 my $cancel_job = sub {
6034 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6035 while (1) {
6036 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6037 my $stat = @$stats[0];
6038 last if !$stat;
6039 sleep 1;
6040 }
6041 };
6042
6043 if ($err) {
6044 eval { &$cancel_job(); };
6045 die "mirroring error: $err";
6046 }
6047
6048 if ($vmiddst != $vmid) {
6049 # if we clone a disk for a new target vm, we don't switch the disk
6050 &$cancel_job(); # so we call block-job-cancel
6051 }
6052 }
6053
6054 sub clone_disk {
6055 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6056 $newvmid, $storage, $format, $full, $newvollist) = @_;
6057
6058 my $newvolid;
6059
6060 if (!$full) {
6061 print "create linked clone of drive $drivename ($drive->{file})\n";
6062 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6063 push @$newvollist, $newvolid;
6064 } else {
6065 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6066 $storeid = $storage if $storage;
6067
6068 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6069 if (!$format) {
6070 $format = $drive->{format} || $defFormat;
6071 }
6072
6073 # test if requested format is supported - else use default
6074 my $supported = grep { $_ eq $format } @$validFormats;
6075 $format = $defFormat if !$supported;
6076
6077 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6078
6079 print "create full clone of drive $drivename ($drive->{file})\n";
6080 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6081 push @$newvollist, $newvolid;
6082
6083 if (!$running || $snapname) {
6084 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6085 } else {
6086 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6087 }
6088 }
6089
6090 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6091
6092 my $disk = $drive;
6093 $disk->{format} = undef;
6094 $disk->{file} = $newvolid;
6095 $disk->{size} = $size;
6096
6097 return $disk;
6098 }
6099
6100 # this only works if VM is running
6101 sub get_current_qemu_machine {
6102 my ($vmid) = @_;
6103
6104 my $cmd = { execute => 'query-machines', arguments => {} };
6105 my $res = vm_qmp_command($vmid, $cmd);
6106
6107 my ($current, $default);
6108 foreach my $e (@$res) {
6109 $default = $e->{name} if $e->{'is-default'};
6110 $current = $e->{name} if $e->{'is-current'};
6111 }
6112
6113 # fallback to the default machine if current is not supported by qemu
6114 return $current || $default || 'pc';
6115 }
6116
6117 sub lspci {
6118
6119 my $devices = {};
6120
6121 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6122 my (undef, $id, $function) = @_;
6123 my $res = { id => $id, function => $function};
6124 push @{$devices->{$id}}, $res;
6125 });
6126
6127 return $devices;
6128 }
6129
6130 1;