]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
85ce28ad29e3c2d3746a80c50fd5b3ca0f92bbd2
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'boolean',
176 description => "Allow hotplug for disk and network device",
177 default => 0,
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 maxcpus => {
310 optional => 1,
311 type => 'integer',
312 description => "Maximum cpus for hotplug.",
313 minimum => 1,
314 default => 1,
315 },
316 acpi => {
317 optional => 1,
318 type => 'boolean',
319 description => "Enable/disable ACPI.",
320 default => 1,
321 },
322 agent => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable Qemu GuestAgent.",
326 default => 0,
327 },
328 kvm => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable KVM hardware virtualization.",
332 default => 1,
333 },
334 tdf => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable time drift fix.",
338 default => 0,
339 },
340 localtime => {
341 optional => 1,
342 type => 'boolean',
343 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
344 },
345 freeze => {
346 optional => 1,
347 type => 'boolean',
348 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
349 },
350 vga => {
351 optional => 1,
352 type => 'string',
353 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.",
354 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
355 },
356 watchdog => {
357 optional => 1,
358 type => 'string', format => 'pve-qm-watchdog',
359 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
360 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)",
361 },
362 startdate => {
363 optional => 1,
364 type => 'string',
365 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
366 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'.",
367 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
368 default => 'now',
369 },
370 startup => {
371 optional => 1,
372 type => 'string', format => 'pve-qm-startup',
373 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
374 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.",
375 },
376 template => {
377 optional => 1,
378 type => 'boolean',
379 description => "Enable/disable Template.",
380 default => 0,
381 },
382 args => {
383 optional => 1,
384 type => 'string',
385 description => <<EODESCR,
386 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
387
388 args: -no-reboot -no-hpet
389 EODESCR
390 },
391 tablet => {
392 optional => 1,
393 type => 'boolean',
394 default => 1,
395 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).",
396 },
397 migrate_speed => {
398 optional => 1,
399 type => 'integer',
400 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
401 minimum => 0,
402 default => 0,
403 },
404 migrate_downtime => {
405 optional => 1,
406 type => 'number',
407 description => "Set maximum tolerated downtime (in seconds) for migrations.",
408 minimum => 0,
409 default => 0.1,
410 },
411 cdrom => {
412 optional => 1,
413 type => 'string', format => 'pve-qm-drive',
414 typetext => 'volume',
415 description => "This is an alias for option -ide2",
416 },
417 cpu => {
418 optional => 1,
419 description => "Emulated CPU type.",
420 type => 'string',
421 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) ],
422 default => 'kvm64',
423 },
424 parent => get_standard_option('pve-snapshot-name', {
425 optional => 1,
426 description => "Parent snapshot name. This is used internally, and should not be modified.",
427 }),
428 snaptime => {
429 optional => 1,
430 description => "Timestamp for snapshots.",
431 type => 'integer',
432 minimum => 0,
433 },
434 vmstate => {
435 optional => 1,
436 type => 'string', format => 'pve-volume-id',
437 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
438 },
439 machine => {
440 description => "Specific the Qemu machine type.",
441 type => 'string',
442 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
443 maxLength => 40,
444 optional => 1,
445 },
446 smbios1 => {
447 description => "Specify SMBIOS type 1 fields.",
448 type => 'string', format => 'pve-qm-smbios1',
449 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
450 maxLength => 256,
451 optional => 1,
452 },
453 };
454
455 # what about other qemu settings ?
456 #cpu => 'string',
457 #machine => 'string',
458 #fda => 'file',
459 #fdb => 'file',
460 #mtdblock => 'file',
461 #sd => 'file',
462 #pflash => 'file',
463 #snapshot => 'bool',
464 #bootp => 'file',
465 ##tftp => 'dir',
466 ##smb => 'dir',
467 #kernel => 'file',
468 #append => 'string',
469 #initrd => 'file',
470 ##soundhw => 'string',
471
472 while (my ($k, $v) = each %$confdesc) {
473 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
474 }
475
476 my $MAX_IDE_DISKS = 4;
477 my $MAX_SCSI_DISKS = 14;
478 my $MAX_VIRTIO_DISKS = 16;
479 my $MAX_SATA_DISKS = 6;
480 my $MAX_USB_DEVICES = 5;
481 my $MAX_NETS = 32;
482 my $MAX_UNUSED_DISKS = 8;
483 my $MAX_HOSTPCI_DEVICES = 4;
484 my $MAX_SERIAL_PORTS = 4;
485 my $MAX_PARALLEL_PORTS = 3;
486
487 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
488 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3'];
489 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
490
491 my $netdesc = {
492 optional => 1,
493 type => 'string', format => 'pve-qm-net',
494 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>][,tag=<vlanid>][,firewall=0|1]",
495 description => <<EODESCR,
496 Specify network devices.
497
498 MODEL is one of: $nic_model_list_txt
499
500 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
501 automatically generated if not specified.
502
503 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
504
505 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'.
506
507 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
508
509 10.0.2.2 Gateway
510 10.0.2.3 DNS Server
511 10.0.2.4 SMB Server
512
513 The DHCP server assign addresses to the guest starting from 10.0.2.15.
514
515 EODESCR
516 };
517 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
518
519 for (my $i = 0; $i < $MAX_NETS; $i++) {
520 $confdesc->{"net$i"} = $netdesc;
521 }
522
523 my $drivename_hash;
524
525 my $idedesc = {
526 optional => 1,
527 type => 'string', format => 'pve-qm-drive',
528 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
529 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
530 };
531 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
532
533 my $scsidesc = {
534 optional => 1,
535 type => 'string', format => 'pve-qm-drive',
536 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
537 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
538 };
539 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
540
541 my $satadesc = {
542 optional => 1,
543 type => 'string', format => 'pve-qm-drive',
544 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
545 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
546 };
547 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
548
549 my $virtiodesc = {
550 optional => 1,
551 type => 'string', format => 'pve-qm-drive',
552 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
553 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
554 };
555 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
556
557 my $usbdesc = {
558 optional => 1,
559 type => 'string', format => 'pve-qm-usb-device',
560 typetext => 'host=HOSTUSBDEVICE|spice',
561 description => <<EODESCR,
562 Configure an USB device (n is 0 to 4). This can be used to
563 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
564
565 'bus-port(.port)*' (decimal numbers) or
566 'vendor_id:product_id' (hexadeciaml numbers)
567
568 You can use the 'lsusb -t' command to list existing usb devices.
569
570 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
571
572 The value 'spice' can be used to add a usb redirection devices for spice.
573
574 EODESCR
575 };
576 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
577
578 my $hostpcidesc = {
579 optional => 1,
580 type => 'string', format => 'pve-qm-hostpci',
581 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
582 description => <<EODESCR,
583 Map host pci devices. HOSTPCIDEVICE syntax is:
584
585 'bus:dev.func' (hexadecimal numbers)
586
587 You can us the 'lspci' command to list existing pci devices.
588
589 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
590
591 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
592
593 Experimental: user reported problems with this option.
594 EODESCR
595 };
596 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
597
598 my $serialdesc = {
599 optional => 1,
600 type => 'string',
601 pattern => '(/dev/ttyS\d+|socket)',
602 description => <<EODESCR,
603 Create a serial device inside the VM (n is 0 to 3), and pass through a host serial device, or create a unix socket on the host side (use 'qm terminal' to open a terminal connection).
604
605 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
606
607 Experimental: user reported problems with this option.
608 EODESCR
609 };
610
611 my $paralleldesc= {
612 optional => 1,
613 type => 'string',
614 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
615 description => <<EODESCR,
616 Map host parallel devices (n is 0 to 2).
617
618 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
619
620 Experimental: user reported problems with this option.
621 EODESCR
622 };
623
624 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
625 $confdesc->{"parallel$i"} = $paralleldesc;
626 }
627
628 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
629 $confdesc->{"serial$i"} = $serialdesc;
630 }
631
632 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
633 $confdesc->{"hostpci$i"} = $hostpcidesc;
634 }
635
636 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
637 $drivename_hash->{"ide$i"} = 1;
638 $confdesc->{"ide$i"} = $idedesc;
639 }
640
641 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
642 $drivename_hash->{"sata$i"} = 1;
643 $confdesc->{"sata$i"} = $satadesc;
644 }
645
646 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
647 $drivename_hash->{"scsi$i"} = 1;
648 $confdesc->{"scsi$i"} = $scsidesc ;
649 }
650
651 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
652 $drivename_hash->{"virtio$i"} = 1;
653 $confdesc->{"virtio$i"} = $virtiodesc;
654 }
655
656 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
657 $confdesc->{"usb$i"} = $usbdesc;
658 }
659
660 my $unuseddesc = {
661 optional => 1,
662 type => 'string', format => 'pve-volume-id',
663 description => "Reference to unused volumes.",
664 };
665
666 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
667 $confdesc->{"unused$i"} = $unuseddesc;
668 }
669
670 my $kvm_api_version = 0;
671
672 sub kvm_version {
673
674 return $kvm_api_version if $kvm_api_version;
675
676 my $fh = IO::File->new("</dev/kvm") ||
677 return 0;
678
679 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
680 $kvm_api_version = $v;
681 }
682
683 $fh->close();
684
685 return $kvm_api_version;
686 }
687
688 my $kvm_user_version;
689
690 sub kvm_user_version {
691
692 return $kvm_user_version if $kvm_user_version;
693
694 $kvm_user_version = 'unknown';
695
696 my $tmp = `kvm -help 2>/dev/null`;
697
698 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
699 $kvm_user_version = $2;
700 }
701
702 return $kvm_user_version;
703
704 }
705
706 my $kernel_has_vhost_net = -c '/dev/vhost-net';
707
708 sub disknames {
709 # order is important - used to autoselect boot disk
710 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
711 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
712 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
713 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
714 }
715
716 sub valid_drivename {
717 my $dev = shift;
718
719 return defined($drivename_hash->{$dev});
720 }
721
722 sub option_exists {
723 my $key = shift;
724 return defined($confdesc->{$key});
725 }
726
727 sub nic_models {
728 return $nic_model_list;
729 }
730
731 sub os_list_description {
732
733 return {
734 other => 'Other',
735 wxp => 'Windows XP',
736 w2k => 'Windows 2000',
737 w2k3 =>, 'Windows 2003',
738 w2k8 => 'Windows 2008',
739 wvista => 'Windows Vista',
740 win7 => 'Windows 7',
741 win8 => 'Windows 8/2012',
742 l24 => 'Linux 2.4',
743 l26 => 'Linux 2.6',
744 };
745 }
746
747 my $cdrom_path;
748
749 sub get_cdrom_path {
750
751 return $cdrom_path if $cdrom_path;
752
753 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
754 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
755 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
756 }
757
758 sub get_iso_path {
759 my ($storecfg, $vmid, $cdrom) = @_;
760
761 if ($cdrom eq 'cdrom') {
762 return get_cdrom_path();
763 } elsif ($cdrom eq 'none') {
764 return '';
765 } elsif ($cdrom =~ m|^/|) {
766 return $cdrom;
767 } else {
768 return PVE::Storage::path($storecfg, $cdrom);
769 }
770 }
771
772 # try to convert old style file names to volume IDs
773 sub filename_to_volume_id {
774 my ($vmid, $file, $media) = @_;
775
776 if (!($file eq 'none' || $file eq 'cdrom' ||
777 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
778
779 return undef if $file =~ m|/|;
780
781 if ($media && $media eq 'cdrom') {
782 $file = "local:iso/$file";
783 } else {
784 $file = "local:$vmid/$file";
785 }
786 }
787
788 return $file;
789 }
790
791 sub verify_media_type {
792 my ($opt, $vtype, $media) = @_;
793
794 return if !$media;
795
796 my $etype;
797 if ($media eq 'disk') {
798 $etype = 'images';
799 } elsif ($media eq 'cdrom') {
800 $etype = 'iso';
801 } else {
802 die "internal error";
803 }
804
805 return if ($vtype eq $etype);
806
807 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
808 }
809
810 sub cleanup_drive_path {
811 my ($opt, $storecfg, $drive) = @_;
812
813 # try to convert filesystem paths to volume IDs
814
815 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
816 ($drive->{file} !~ m|^/dev/.+|) &&
817 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
818 ($drive->{file} !~ m/^\d+$/)) {
819 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
820 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
821 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
822 verify_media_type($opt, $vtype, $drive->{media});
823 $drive->{file} = $volid;
824 }
825
826 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
827 }
828
829 sub create_conf_nolock {
830 my ($vmid, $settings) = @_;
831
832 my $filename = config_file($vmid);
833
834 die "configuration file '$filename' already exists\n" if -f $filename;
835
836 my $defaults = load_defaults();
837
838 $settings->{name} = "vm$vmid" if !$settings->{name};
839 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
840
841 my $data = '';
842 foreach my $opt (keys %$settings) {
843 next if !$confdesc->{$opt};
844
845 my $value = $settings->{$opt};
846 next if !$value;
847
848 $data .= "$opt: $value\n";
849 }
850
851 PVE::Tools::file_set_contents($filename, $data);
852 }
853
854 my $parse_size = sub {
855 my ($value) = @_;
856
857 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
858 my ($size, $unit) = ($1, $3);
859 if ($unit) {
860 if ($unit eq 'K') {
861 $size = $size * 1024;
862 } elsif ($unit eq 'M') {
863 $size = $size * 1024 * 1024;
864 } elsif ($unit eq 'G') {
865 $size = $size * 1024 * 1024 * 1024;
866 }
867 }
868 return int($size);
869 };
870
871 my $format_size = sub {
872 my ($size) = @_;
873
874 $size = int($size);
875
876 my $kb = int($size/1024);
877 return $size if $kb*1024 != $size;
878
879 my $mb = int($kb/1024);
880 return "${kb}K" if $mb*1024 != $kb;
881
882 my $gb = int($mb/1024);
883 return "${mb}M" if $gb*1024 != $mb;
884
885 return "${gb}G";
886 };
887
888 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
889 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
890 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
891 # [,aio=native|threads][,discard=ignore|on]
892
893 sub parse_drive {
894 my ($key, $data) = @_;
895
896 my $res = {};
897
898 # $key may be undefined - used to verify JSON parameters
899 if (!defined($key)) {
900 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
901 $res->{index} = 0;
902 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
903 $res->{interface} = $1;
904 $res->{index} = $2;
905 } else {
906 return undef;
907 }
908
909 foreach my $p (split (/,/, $data)) {
910 next if $p =~ m/^\s*$/;
911
912 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
913 my ($k, $v) = ($1, $2);
914
915 $k = 'file' if $k eq 'volume';
916
917 return undef if defined $res->{$k};
918
919 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
920 return undef if !$v || $v !~ m/^\d+/;
921 $k = "m$k";
922 $v = sprintf("%.3f", $v / (1024*1024));
923 }
924 $res->{$k} = $v;
925 } else {
926 if (!$res->{file} && $p !~ m/=/) {
927 $res->{file} = $p;
928 } else {
929 return undef;
930 }
931 }
932 }
933
934 return undef if !$res->{file};
935
936 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
937 $res->{format} = $1;
938 }
939
940 return undef if $res->{cache} &&
941 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
942 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
943 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
944 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
945 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
946 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
947 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
948 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
949 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
950 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
951 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
952 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
953 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
954
955 return undef if $res->{mbps_rd} && $res->{mbps};
956 return undef if $res->{mbps_wr} && $res->{mbps};
957
958 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
959 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
960 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
961 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
962 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
963 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
964
965 return undef if $res->{iops_rd} && $res->{iops};
966 return undef if $res->{iops_wr} && $res->{iops};
967
968
969 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
970 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
971 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
972 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
973 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
974 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
975
976
977 if ($res->{size}) {
978 return undef if !defined($res->{size} = &$parse_size($res->{size}));
979 }
980
981 if ($res->{media} && ($res->{media} eq 'cdrom')) {
982 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
983 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
984 return undef if $res->{interface} eq 'virtio';
985 }
986
987 # rerror does not work with scsi drives
988 if ($res->{rerror}) {
989 return undef if $res->{interface} eq 'scsi';
990 }
991
992 return $res;
993 }
994
995 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
996
997 sub print_drive {
998 my ($vmid, $drive) = @_;
999
1000 my $opts = '';
1001 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1002 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1003 }
1004
1005 if ($drive->{size}) {
1006 $opts .= ",size=" . &$format_size($drive->{size});
1007 }
1008
1009 return "$drive->{file}$opts";
1010 }
1011
1012 sub scsi_inquiry {
1013 my($fh, $noerr) = @_;
1014
1015 my $SG_IO = 0x2285;
1016 my $SG_GET_VERSION_NUM = 0x2282;
1017
1018 my $versionbuf = "\x00" x 8;
1019 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1020 if (!$ret) {
1021 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1022 return undef;
1023 }
1024 my $version = unpack("I", $versionbuf);
1025 if ($version < 30000) {
1026 die "scsi generic interface too old\n" if !$noerr;
1027 return undef;
1028 }
1029
1030 my $buf = "\x00" x 36;
1031 my $sensebuf = "\x00" x 8;
1032 my $cmd = pack("C x3 C x1", 0x12, 36);
1033
1034 # see /usr/include/scsi/sg.h
1035 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";
1036
1037 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1038 length($sensebuf), 0, length($buf), $buf,
1039 $cmd, $sensebuf, 6000);
1040
1041 $ret = ioctl($fh, $SG_IO, $packet);
1042 if (!$ret) {
1043 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1044 return undef;
1045 }
1046
1047 my @res = unpack($sg_io_hdr_t, $packet);
1048 if ($res[17] || $res[18]) {
1049 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1050 return undef;
1051 }
1052
1053 my $res = {};
1054 (my $byte0, my $byte1, $res->{vendor},
1055 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1056
1057 $res->{removable} = $byte1 & 128 ? 1 : 0;
1058 $res->{type} = $byte0 & 31;
1059
1060 return $res;
1061 }
1062
1063 sub path_is_scsi {
1064 my ($path) = @_;
1065
1066 my $fh = IO::File->new("+<$path") || return undef;
1067 my $res = scsi_inquiry($fh, 1);
1068 close($fh);
1069
1070 return $res;
1071 }
1072
1073 sub machine_type_is_q35 {
1074 my ($conf) = @_;
1075
1076 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1077 }
1078
1079 sub print_tabletdevice_full {
1080 my ($conf) = @_;
1081
1082 my $q35 = machine_type_is_q35($conf);
1083
1084 # we use uhci for old VMs because tablet driver was buggy in older qemu
1085 my $usbbus = $q35 ? "ehci" : "uhci";
1086
1087 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1088 }
1089
1090 sub print_drivedevice_full {
1091 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1092
1093 my $device = '';
1094 my $maxdev = 0;
1095
1096 if ($drive->{interface} eq 'virtio') {
1097 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1098 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1099 $device .= ",iothread=iothread0" if $conf->{iothread};
1100 } elsif ($drive->{interface} eq 'scsi') {
1101 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1102 my $controller = int($drive->{index} / $maxdev);
1103 my $unit = $drive->{index} % $maxdev;
1104 my $devicetype = 'hd';
1105 my $path = '';
1106 if (drive_is_cdrom($drive)) {
1107 $devicetype = 'cd';
1108 } else {
1109 if ($drive->{file} =~ m|^/|) {
1110 $path = $drive->{file};
1111 } else {
1112 $path = PVE::Storage::path($storecfg, $drive->{file});
1113 }
1114
1115 if($path =~ m/^iscsi\:\/\//){
1116 $devicetype = 'generic';
1117 } else {
1118 if (my $info = path_is_scsi($path)) {
1119 if ($info->{type} == 0) {
1120 $devicetype = 'block';
1121 } elsif ($info->{type} == 1) { # tape
1122 $devicetype = 'generic';
1123 }
1124 }
1125 }
1126 }
1127
1128 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1129 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1130 } else {
1131 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1132 }
1133
1134 } elsif ($drive->{interface} eq 'ide'){
1135 $maxdev = 2;
1136 my $controller = int($drive->{index} / $maxdev);
1137 my $unit = $drive->{index} % $maxdev;
1138 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1139
1140 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1141 } elsif ($drive->{interface} eq 'sata'){
1142 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1143 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1144 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1145 } elsif ($drive->{interface} eq 'usb') {
1146 die "implement me";
1147 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1148 } else {
1149 die "unsupported interface type";
1150 }
1151
1152 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1153
1154 return $device;
1155 }
1156
1157 sub get_initiator_name {
1158 my $initiator;
1159
1160 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1161 while (defined(my $line = <$fh>)) {
1162 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1163 $initiator = $1;
1164 last;
1165 }
1166 $fh->close();
1167
1168 return $initiator;
1169 }
1170
1171 sub print_drive_full {
1172 my ($storecfg, $vmid, $drive) = @_;
1173
1174 my $opts = '';
1175 foreach my $o (@qemu_drive_options) {
1176 next if $o eq 'bootindex';
1177 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1178 }
1179
1180 foreach my $o (qw(bps bps_rd bps_wr)) {
1181 my $v = $drive->{"m$o"};
1182 $opts .= ",$o=" . int($v*1024*1024) if $v;
1183 }
1184
1185 # use linux-aio by default (qemu default is threads)
1186 $opts .= ",aio=native" if !$drive->{aio};
1187
1188 my $path;
1189 my $volid = $drive->{file};
1190 if (drive_is_cdrom($drive)) {
1191 $path = get_iso_path($storecfg, $vmid, $volid);
1192 } else {
1193 if ($volid =~ m|^/|) {
1194 $path = $volid;
1195 } else {
1196 $path = PVE::Storage::path($storecfg, $volid);
1197 }
1198 }
1199
1200 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1201
1202 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1203 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1204
1205 my $pathinfo = $path ? "file=$path," : '';
1206
1207 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1208 }
1209
1210 sub print_netdevice_full {
1211 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1212
1213 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1214
1215 my $device = $net->{model};
1216 if ($net->{model} eq 'virtio') {
1217 $device = 'virtio-net-pci';
1218 };
1219
1220 # qemu > 0.15 always try to boot from network - we disable that by
1221 # not loading the pxe rom file
1222 my $extra = ($bootorder !~ m/n/) ? "romfile=," : '';
1223 my $pciaddr = print_pci_addr("$netid", $bridges);
1224 my $tmpstr = "$device,${extra}mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1225 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1226 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1227 my $vectors = $net->{queues} * 2 + 2;
1228 $tmpstr .= ",vectors=$vectors,mq=on";
1229 }
1230 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1231 return $tmpstr;
1232 }
1233
1234 sub print_netdev_full {
1235 my ($vmid, $conf, $net, $netid) = @_;
1236
1237 my $i = '';
1238 if ($netid =~ m/^net(\d+)$/) {
1239 $i = int($1);
1240 }
1241
1242 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1243
1244 my $ifname = "tap${vmid}i$i";
1245
1246 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1247 die "interface name '$ifname' is too long (max 15 character)\n"
1248 if length($ifname) >= 16;
1249
1250 my $vhostparam = '';
1251 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1252
1253 my $vmname = $conf->{name} || "vm$vmid";
1254
1255 my $netdev = "";
1256
1257 if ($net->{bridge}) {
1258 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1259 } else {
1260 $netdev = "type=user,id=$netid,hostname=$vmname";
1261 }
1262
1263 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1264
1265 return $netdev;
1266 }
1267
1268 sub drive_is_cdrom {
1269 my ($drive) = @_;
1270
1271 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1272
1273 }
1274
1275 sub parse_hostpci {
1276 my ($value) = @_;
1277
1278 return undef if !$value;
1279
1280
1281 my @list = split(/,/, $value);
1282 my $found;
1283
1284 my $res = {};
1285 foreach my $kv (@list) {
1286
1287 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1288 $found = 1;
1289 if(defined($4)){
1290 push @{$res->{pciid}}, { id => $2 , function => $4};
1291
1292 }else{
1293 my $pcidevices = lspci($2);
1294 $res->{pciid} = $pcidevices->{$2};
1295 }
1296 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1297 $res->{driver} = $1;
1298 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1299 $res->{rombar} = $1;
1300 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1301 $res->{'x-vga'} = $1;
1302 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1303 $res->{pcie} = 1 if $1 == 1;
1304 } else {
1305 warn "unknown hostpci setting '$kv'\n";
1306 }
1307 }
1308
1309 return undef if !$found;
1310
1311 return $res;
1312 }
1313
1314 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1315 sub parse_net {
1316 my ($data) = @_;
1317
1318 my $res = {};
1319
1320 foreach my $kvp (split(/,/, $data)) {
1321
1322 if ($kvp =~ m/^(ne2k_pci|e1000|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1323 my $model = lc($1);
1324 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1325 $res->{model} = $model;
1326 $res->{macaddr} = $mac;
1327 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1328 $res->{bridge} = $1;
1329 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1330 $res->{queues} = $1;
1331 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1332 $res->{rate} = $1;
1333 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1334 $res->{tag} = $1;
1335 } elsif ($kvp =~ m/^firewall=(\d+)$/) {
1336 $res->{firewall} = $1;
1337 } else {
1338 return undef;
1339 }
1340
1341 }
1342
1343 return undef if !$res->{model};
1344
1345 return $res;
1346 }
1347
1348 sub print_net {
1349 my $net = shift;
1350
1351 my $res = "$net->{model}";
1352 $res .= "=$net->{macaddr}" if $net->{macaddr};
1353 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1354 $res .= ",rate=$net->{rate}" if $net->{rate};
1355 $res .= ",tag=$net->{tag}" if $net->{tag};
1356 $res .= ",firewall=$net->{firewall}" if $net->{firewall};
1357
1358 return $res;
1359 }
1360
1361 sub add_random_macs {
1362 my ($settings) = @_;
1363
1364 foreach my $opt (keys %$settings) {
1365 next if $opt !~ m/^net(\d+)$/;
1366 my $net = parse_net($settings->{$opt});
1367 next if !$net;
1368 $settings->{$opt} = print_net($net);
1369 }
1370 }
1371
1372 sub add_unused_volume {
1373 my ($config, $volid) = @_;
1374
1375 my $key;
1376 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1377 my $test = "unused$ind";
1378 if (my $vid = $config->{$test}) {
1379 return if $vid eq $volid; # do not add duplicates
1380 } else {
1381 $key = $test;
1382 }
1383 }
1384
1385 die "To many unused volume - please delete them first.\n" if !$key;
1386
1387 $config->{$key} = $volid;
1388
1389 return $key;
1390 }
1391
1392 my $valid_smbios1_options = {
1393 manufacturer => '\S+',
1394 product => '\S+',
1395 version => '\S+',
1396 serial => '\S+',
1397 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1398 sku => '\S+',
1399 family => '\S+',
1400 };
1401
1402 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1403 sub parse_smbios1 {
1404 my ($data) = @_;
1405
1406 my $res = {};
1407
1408 foreach my $kvp (split(/,/, $data)) {
1409 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1410 my ($k, $v) = split(/=/, $kvp);
1411 return undef if !defined($k) || !defined($v);
1412 return undef if !$valid_smbios1_options->{$k};
1413 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1414 $res->{$k} = $v;
1415 }
1416
1417 return $res;
1418 }
1419
1420 sub print_smbios1 {
1421 my ($smbios1) = @_;
1422
1423 my $data = '';
1424 foreach my $k (keys %$smbios1) {
1425 next if !defined($smbios1->{$k});
1426 next if !$valid_smbios1_options->{$k};
1427 $data .= ',' if $data;
1428 $data .= "$k=$smbios1->{$k}";
1429 }
1430 return $data;
1431 }
1432
1433 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1434 sub verify_smbios1 {
1435 my ($value, $noerr) = @_;
1436
1437 return $value if parse_smbios1($value);
1438
1439 return undef if $noerr;
1440
1441 die "unable to parse smbios (type 1) options\n";
1442 }
1443
1444 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1445 sub verify_bootdisk {
1446 my ($value, $noerr) = @_;
1447
1448 return $value if valid_drivename($value);
1449
1450 return undef if $noerr;
1451
1452 die "invalid boot disk '$value'\n";
1453 }
1454
1455 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1456 sub verify_net {
1457 my ($value, $noerr) = @_;
1458
1459 return $value if parse_net($value);
1460
1461 return undef if $noerr;
1462
1463 die "unable to parse network options\n";
1464 }
1465
1466 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1467 sub verify_drive {
1468 my ($value, $noerr) = @_;
1469
1470 return $value if parse_drive(undef, $value);
1471
1472 return undef if $noerr;
1473
1474 die "unable to parse drive options\n";
1475 }
1476
1477 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1478 sub verify_hostpci {
1479 my ($value, $noerr) = @_;
1480
1481 return $value if parse_hostpci($value);
1482
1483 return undef if $noerr;
1484
1485 die "unable to parse pci id\n";
1486 }
1487
1488 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1489 sub verify_watchdog {
1490 my ($value, $noerr) = @_;
1491
1492 return $value if parse_watchdog($value);
1493
1494 return undef if $noerr;
1495
1496 die "unable to parse watchdog options\n";
1497 }
1498
1499 sub parse_watchdog {
1500 my ($value) = @_;
1501
1502 return undef if !$value;
1503
1504 my $res = {};
1505
1506 foreach my $p (split(/,/, $value)) {
1507 next if $p =~ m/^\s*$/;
1508
1509 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1510 $res->{model} = $2;
1511 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1512 $res->{action} = $2;
1513 } else {
1514 return undef;
1515 }
1516 }
1517
1518 return $res;
1519 }
1520
1521 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1522 sub verify_startup {
1523 my ($value, $noerr) = @_;
1524
1525 return $value if parse_startup($value);
1526
1527 return undef if $noerr;
1528
1529 die "unable to parse startup options\n";
1530 }
1531
1532 sub parse_startup {
1533 my ($value) = @_;
1534
1535 return undef if !$value;
1536
1537 my $res = {};
1538
1539 foreach my $p (split(/,/, $value)) {
1540 next if $p =~ m/^\s*$/;
1541
1542 if ($p =~ m/^(order=)?(\d+)$/) {
1543 $res->{order} = $2;
1544 } elsif ($p =~ m/^up=(\d+)$/) {
1545 $res->{up} = $1;
1546 } elsif ($p =~ m/^down=(\d+)$/) {
1547 $res->{down} = $1;
1548 } else {
1549 return undef;
1550 }
1551 }
1552
1553 return $res;
1554 }
1555
1556 sub parse_usb_device {
1557 my ($value) = @_;
1558
1559 return undef if !$value;
1560
1561 my @dl = split(/,/, $value);
1562 my $found;
1563
1564 my $res = {};
1565 foreach my $v (@dl) {
1566 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1567 $found = 1;
1568 $res->{vendorid} = $2;
1569 $res->{productid} = $4;
1570 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1571 $found = 1;
1572 $res->{hostbus} = $1;
1573 $res->{hostport} = $2;
1574 } elsif ($v =~ m/^spice$/) {
1575 $found = 1;
1576 $res->{spice} = 1;
1577 } else {
1578 return undef;
1579 }
1580 }
1581 return undef if !$found;
1582
1583 return $res;
1584 }
1585
1586 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1587 sub verify_usb_device {
1588 my ($value, $noerr) = @_;
1589
1590 return $value if parse_usb_device($value);
1591
1592 return undef if $noerr;
1593
1594 die "unable to parse usb device\n";
1595 }
1596
1597 # add JSON properties for create and set function
1598 sub json_config_properties {
1599 my $prop = shift;
1600
1601 foreach my $opt (keys %$confdesc) {
1602 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1603 $prop->{$opt} = $confdesc->{$opt};
1604 }
1605
1606 return $prop;
1607 }
1608
1609 sub check_type {
1610 my ($key, $value) = @_;
1611
1612 die "unknown setting '$key'\n" if !$confdesc->{$key};
1613
1614 my $type = $confdesc->{$key}->{type};
1615
1616 if (!defined($value)) {
1617 die "got undefined value\n";
1618 }
1619
1620 if ($value =~ m/[\n\r]/) {
1621 die "property contains a line feed\n";
1622 }
1623
1624 if ($type eq 'boolean') {
1625 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1626 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1627 die "type check ('boolean') failed - got '$value'\n";
1628 } elsif ($type eq 'integer') {
1629 return int($1) if $value =~ m/^(\d+)$/;
1630 die "type check ('integer') failed - got '$value'\n";
1631 } elsif ($type eq 'number') {
1632 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1633 die "type check ('number') failed - got '$value'\n";
1634 } elsif ($type eq 'string') {
1635 if (my $fmt = $confdesc->{$key}->{format}) {
1636 if ($fmt eq 'pve-qm-drive') {
1637 # special case - we need to pass $key to parse_drive()
1638 my $drive = parse_drive($key, $value);
1639 return $value if $drive;
1640 die "unable to parse drive options\n";
1641 }
1642 PVE::JSONSchema::check_format($fmt, $value);
1643 return $value;
1644 }
1645 $value =~ s/^\"(.*)\"$/$1/;
1646 return $value;
1647 } else {
1648 die "internal error"
1649 }
1650 }
1651
1652 sub lock_config_full {
1653 my ($vmid, $timeout, $code, @param) = @_;
1654
1655 my $filename = config_file_lock($vmid);
1656
1657 my $res = lock_file($filename, $timeout, $code, @param);
1658
1659 die $@ if $@;
1660
1661 return $res;
1662 }
1663
1664 sub lock_config_mode {
1665 my ($vmid, $timeout, $shared, $code, @param) = @_;
1666
1667 my $filename = config_file_lock($vmid);
1668
1669 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1670
1671 die $@ if $@;
1672
1673 return $res;
1674 }
1675
1676 sub lock_config {
1677 my ($vmid, $code, @param) = @_;
1678
1679 return lock_config_full($vmid, 10, $code, @param);
1680 }
1681
1682 sub cfs_config_path {
1683 my ($vmid, $node) = @_;
1684
1685 $node = $nodename if !$node;
1686 return "nodes/$node/qemu-server/$vmid.conf";
1687 }
1688
1689 sub check_iommu_support{
1690 #fixme : need to check IOMMU support
1691 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1692
1693 my $iommu=1;
1694 return $iommu;
1695
1696 }
1697
1698 sub config_file {
1699 my ($vmid, $node) = @_;
1700
1701 my $cfspath = cfs_config_path($vmid, $node);
1702 return "/etc/pve/$cfspath";
1703 }
1704
1705 sub config_file_lock {
1706 my ($vmid) = @_;
1707
1708 return "$lock_dir/lock-$vmid.conf";
1709 }
1710
1711 sub touch_config {
1712 my ($vmid) = @_;
1713
1714 my $conf = config_file($vmid);
1715 utime undef, undef, $conf;
1716 }
1717
1718 sub destroy_vm {
1719 my ($storecfg, $vmid, $keep_empty_config) = @_;
1720
1721 my $conffile = config_file($vmid);
1722
1723 my $conf = load_config($vmid);
1724
1725 check_lock($conf);
1726
1727 # only remove disks owned by this VM
1728 foreach_drive($conf, sub {
1729 my ($ds, $drive) = @_;
1730
1731 return if drive_is_cdrom($drive);
1732
1733 my $volid = $drive->{file};
1734
1735 return if !$volid || $volid =~ m|^/|;
1736
1737 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1738 return if !$path || !$owner || ($owner != $vmid);
1739
1740 PVE::Storage::vdisk_free($storecfg, $volid);
1741 });
1742
1743 if ($keep_empty_config) {
1744 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1745 } else {
1746 unlink $conffile;
1747 }
1748
1749 # also remove unused disk
1750 eval {
1751 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1752
1753 eval {
1754 PVE::Storage::foreach_volid($dl, sub {
1755 my ($volid, $sid, $volname, $d) = @_;
1756 PVE::Storage::vdisk_free($storecfg, $volid);
1757 });
1758 };
1759 warn $@ if $@;
1760
1761 };
1762 warn $@ if $@;
1763 }
1764
1765 sub load_config {
1766 my ($vmid, $node) = @_;
1767
1768 my $cfspath = cfs_config_path($vmid, $node);
1769
1770 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1771
1772 die "no such VM ('$vmid')\n" if !defined($conf);
1773
1774 return $conf;
1775 }
1776
1777 sub parse_vm_config {
1778 my ($filename, $raw) = @_;
1779
1780 return undef if !defined($raw);
1781
1782 my $res = {
1783 digest => Digest::SHA::sha1_hex($raw),
1784 snapshots => {},
1785 };
1786
1787 $filename =~ m|/qemu-server/(\d+)\.conf$|
1788 || die "got strange filename '$filename'";
1789
1790 my $vmid = $1;
1791
1792 my $conf = $res;
1793 my $descr = '';
1794
1795 my @lines = split(/\n/, $raw);
1796 foreach my $line (@lines) {
1797 next if $line =~ m/^\s*$/;
1798
1799 if ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1800 my $snapname = $1;
1801 $conf->{description} = $descr if $descr;
1802 $descr = '';
1803 $conf = $res->{snapshots}->{$snapname} = {};
1804 next;
1805 }
1806
1807 if ($line =~ m/^\#(.*)\s*$/) {
1808 $descr .= PVE::Tools::decode_text($1) . "\n";
1809 next;
1810 }
1811
1812 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1813 $descr .= PVE::Tools::decode_text($2);
1814 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1815 $conf->{snapstate} = $1;
1816 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1817 my $key = $1;
1818 my $value = $2;
1819 $conf->{$key} = $value;
1820 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
1821 my $key = $1;
1822 my $value = $2;
1823 eval { $value = check_type($key, $value); };
1824 if ($@) {
1825 warn "vm $vmid - unable to parse value of '$key' - $@";
1826 } else {
1827 my $fmt = $confdesc->{$key}->{format};
1828 if ($fmt && $fmt eq 'pve-qm-drive') {
1829 my $v = parse_drive($key, $value);
1830 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
1831 $v->{file} = $volid;
1832 $value = print_drive($vmid, $v);
1833 } else {
1834 warn "vm $vmid - unable to parse value of '$key'\n";
1835 next;
1836 }
1837 }
1838
1839 if ($key eq 'cdrom') {
1840 $conf->{ide2} = $value;
1841 } else {
1842 $conf->{$key} = $value;
1843 }
1844 }
1845 }
1846 }
1847
1848 $conf->{description} = $descr if $descr;
1849
1850 delete $res->{snapstate}; # just to be sure
1851
1852 return $res;
1853 }
1854
1855 sub write_vm_config {
1856 my ($filename, $conf) = @_;
1857
1858 delete $conf->{snapstate}; # just to be sure
1859
1860 if ($conf->{cdrom}) {
1861 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
1862 $conf->{ide2} = $conf->{cdrom};
1863 delete $conf->{cdrom};
1864 }
1865
1866 # we do not use 'smp' any longer
1867 if ($conf->{sockets}) {
1868 delete $conf->{smp};
1869 } elsif ($conf->{smp}) {
1870 $conf->{sockets} = $conf->{smp};
1871 delete $conf->{cores};
1872 delete $conf->{smp};
1873 }
1874
1875 if ($conf->{maxcpus} && $conf->{sockets}) {
1876 delete $conf->{sockets};
1877 }
1878
1879 my $used_volids = {};
1880
1881 my $cleanup_config = sub {
1882 my ($cref, $snapname) = @_;
1883
1884 foreach my $key (keys %$cref) {
1885 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
1886 $key eq 'snapstate';
1887 my $value = $cref->{$key};
1888 eval { $value = check_type($key, $value); };
1889 die "unable to parse value of '$key' - $@" if $@;
1890
1891 $cref->{$key} = $value;
1892
1893 if (!$snapname && valid_drivename($key)) {
1894 my $drive = parse_drive($key, $value);
1895 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
1896 }
1897 }
1898 };
1899
1900 &$cleanup_config($conf);
1901 foreach my $snapname (keys %{$conf->{snapshots}}) {
1902 &$cleanup_config($conf->{snapshots}->{$snapname}, $snapname);
1903 }
1904
1905 # remove 'unusedX' settings if we re-add a volume
1906 foreach my $key (keys %$conf) {
1907 my $value = $conf->{$key};
1908 if ($key =~ m/^unused/ && $used_volids->{$value}) {
1909 delete $conf->{$key};
1910 }
1911 }
1912
1913 my $generate_raw_config = sub {
1914 my ($conf) = @_;
1915
1916 my $raw = '';
1917
1918 # add description as comment to top of file
1919 my $descr = $conf->{description} || '';
1920 foreach my $cl (split(/\n/, $descr)) {
1921 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
1922 }
1923
1924 foreach my $key (sort keys %$conf) {
1925 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots';
1926 $raw .= "$key: $conf->{$key}\n";
1927 }
1928 return $raw;
1929 };
1930
1931 my $raw = &$generate_raw_config($conf);
1932 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
1933 $raw .= "\n[$snapname]\n";
1934 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
1935 }
1936
1937 return $raw;
1938 }
1939
1940 sub update_config_nolock {
1941 my ($vmid, $conf, $skiplock) = @_;
1942
1943 check_lock($conf) if !$skiplock;
1944
1945 my $cfspath = cfs_config_path($vmid);
1946
1947 PVE::Cluster::cfs_write_file($cfspath, $conf);
1948 }
1949
1950 sub update_config {
1951 my ($vmid, $conf, $skiplock) = @_;
1952
1953 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
1954 }
1955
1956 sub load_defaults {
1957
1958 my $res = {};
1959
1960 # we use static defaults from our JSON schema configuration
1961 foreach my $key (keys %$confdesc) {
1962 if (defined(my $default = $confdesc->{$key}->{default})) {
1963 $res->{$key} = $default;
1964 }
1965 }
1966
1967 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
1968 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
1969
1970 return $res;
1971 }
1972
1973 sub config_list {
1974 my $vmlist = PVE::Cluster::get_vmlist();
1975 my $res = {};
1976 return $res if !$vmlist || !$vmlist->{ids};
1977 my $ids = $vmlist->{ids};
1978
1979 foreach my $vmid (keys %$ids) {
1980 my $d = $ids->{$vmid};
1981 next if !$d->{node} || $d->{node} ne $nodename;
1982 next if !$d->{type} || $d->{type} ne 'qemu';
1983 $res->{$vmid}->{exists} = 1;
1984 }
1985 return $res;
1986 }
1987
1988 # test if VM uses local resources (to prevent migration)
1989 sub check_local_resources {
1990 my ($conf, $noerr) = @_;
1991
1992 my $loc_res = 0;
1993
1994 $loc_res = 1 if $conf->{hostusb}; # old syntax
1995 $loc_res = 1 if $conf->{hostpci}; # old syntax
1996
1997 foreach my $k (keys %$conf) {
1998 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
1999 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2000 }
2001
2002 die "VM uses local resources\n" if $loc_res && !$noerr;
2003
2004 return $loc_res;
2005 }
2006
2007 # check if used storages are available on all nodes (use by migrate)
2008 sub check_storage_availability {
2009 my ($storecfg, $conf, $node) = @_;
2010
2011 foreach_drive($conf, sub {
2012 my ($ds, $drive) = @_;
2013
2014 my $volid = $drive->{file};
2015 return if !$volid;
2016
2017 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2018 return if !$sid;
2019
2020 # check if storage is available on both nodes
2021 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2022 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2023 });
2024 }
2025
2026 # list nodes where all VM images are available (used by has_feature API)
2027 sub shared_nodes {
2028 my ($conf, $storecfg) = @_;
2029
2030 my $nodelist = PVE::Cluster::get_nodelist();
2031 my $nodehash = { map { $_ => 1 } @$nodelist };
2032 my $nodename = PVE::INotify::nodename();
2033
2034 foreach_drive($conf, sub {
2035 my ($ds, $drive) = @_;
2036
2037 my $volid = $drive->{file};
2038 return if !$volid;
2039
2040 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2041 if ($storeid) {
2042 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2043 if ($scfg->{disable}) {
2044 $nodehash = {};
2045 } elsif (my $avail = $scfg->{nodes}) {
2046 foreach my $node (keys %$nodehash) {
2047 delete $nodehash->{$node} if !$avail->{$node};
2048 }
2049 } elsif (!$scfg->{shared}) {
2050 foreach my $node (keys %$nodehash) {
2051 delete $nodehash->{$node} if $node ne $nodename
2052 }
2053 }
2054 }
2055 });
2056
2057 return $nodehash
2058 }
2059
2060 sub check_lock {
2061 my ($conf) = @_;
2062
2063 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2064 }
2065
2066 sub check_cmdline {
2067 my ($pidfile, $pid) = @_;
2068
2069 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2070 if (defined($fh)) {
2071 my $line = <$fh>;
2072 $fh->close;
2073 return undef if !$line;
2074 my @param = split(/\0/, $line);
2075
2076 my $cmd = $param[0];
2077 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2078
2079 for (my $i = 0; $i < scalar (@param); $i++) {
2080 my $p = $param[$i];
2081 next if !$p;
2082 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2083 my $p = $param[$i+1];
2084 return 1 if $p && ($p eq $pidfile);
2085 return undef;
2086 }
2087 }
2088 }
2089 return undef;
2090 }
2091
2092 sub check_running {
2093 my ($vmid, $nocheck, $node) = @_;
2094
2095 my $filename = config_file($vmid, $node);
2096
2097 die "unable to find configuration file for VM $vmid - no such machine\n"
2098 if !$nocheck && ! -f $filename;
2099
2100 my $pidfile = pidfile_name($vmid);
2101
2102 if (my $fd = IO::File->new("<$pidfile")) {
2103 my $st = stat($fd);
2104 my $line = <$fd>;
2105 close($fd);
2106
2107 my $mtime = $st->mtime;
2108 if ($mtime > time()) {
2109 warn "file '$filename' modified in future\n";
2110 }
2111
2112 if ($line =~ m/^(\d+)$/) {
2113 my $pid = $1;
2114 if (check_cmdline($pidfile, $pid)) {
2115 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2116 return $pid;
2117 }
2118 }
2119 }
2120 }
2121
2122 return undef;
2123 }
2124
2125 sub vzlist {
2126
2127 my $vzlist = config_list();
2128
2129 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2130
2131 while (defined(my $de = $fd->read)) {
2132 next if $de !~ m/^(\d+)\.pid$/;
2133 my $vmid = $1;
2134 next if !defined($vzlist->{$vmid});
2135 if (my $pid = check_running($vmid)) {
2136 $vzlist->{$vmid}->{pid} = $pid;
2137 }
2138 }
2139
2140 return $vzlist;
2141 }
2142
2143 sub disksize {
2144 my ($storecfg, $conf) = @_;
2145
2146 my $bootdisk = $conf->{bootdisk};
2147 return undef if !$bootdisk;
2148 return undef if !valid_drivename($bootdisk);
2149
2150 return undef if !$conf->{$bootdisk};
2151
2152 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2153 return undef if !defined($drive);
2154
2155 return undef if drive_is_cdrom($drive);
2156
2157 my $volid = $drive->{file};
2158 return undef if !$volid;
2159
2160 return $drive->{size};
2161 }
2162
2163 my $last_proc_pid_stat;
2164
2165 # get VM status information
2166 # This must be fast and should not block ($full == false)
2167 # We only query KVM using QMP if $full == true (this can be slow)
2168 sub vmstatus {
2169 my ($opt_vmid, $full) = @_;
2170
2171 my $res = {};
2172
2173 my $storecfg = PVE::Storage::config();
2174
2175 my $list = vzlist();
2176 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2177
2178 my $cpucount = $cpuinfo->{cpus} || 1;
2179
2180 foreach my $vmid (keys %$list) {
2181 next if $opt_vmid && ($vmid ne $opt_vmid);
2182
2183 my $cfspath = cfs_config_path($vmid);
2184 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2185
2186 my $d = {};
2187 $d->{pid} = $list->{$vmid}->{pid};
2188
2189 # fixme: better status?
2190 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2191
2192 my $size = disksize($storecfg, $conf);
2193 if (defined($size)) {
2194 $d->{disk} = 0; # no info available
2195 $d->{maxdisk} = $size;
2196 } else {
2197 $d->{disk} = 0;
2198 $d->{maxdisk} = 0;
2199 }
2200
2201 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2202 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2203
2204 $d->{name} = $conf->{name} || "VM $vmid";
2205 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2206
2207 if ($conf->{balloon}) {
2208 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2209 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2210 }
2211
2212 $d->{uptime} = 0;
2213 $d->{cpu} = 0;
2214 $d->{mem} = 0;
2215
2216 $d->{netout} = 0;
2217 $d->{netin} = 0;
2218
2219 $d->{diskread} = 0;
2220 $d->{diskwrite} = 0;
2221
2222 $d->{template} = is_template($conf);
2223
2224 $res->{$vmid} = $d;
2225 }
2226
2227 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2228 foreach my $dev (keys %$netdev) {
2229 next if $dev !~ m/^tap([1-9]\d*)i/;
2230 my $vmid = $1;
2231 my $d = $res->{$vmid};
2232 next if !$d;
2233
2234 $d->{netout} += $netdev->{$dev}->{receive};
2235 $d->{netin} += $netdev->{$dev}->{transmit};
2236 }
2237
2238 my $ctime = gettimeofday;
2239
2240 foreach my $vmid (keys %$list) {
2241
2242 my $d = $res->{$vmid};
2243 my $pid = $d->{pid};
2244 next if !$pid;
2245
2246 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2247 next if !$pstat; # not running
2248
2249 my $used = $pstat->{utime} + $pstat->{stime};
2250
2251 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2252
2253 if ($pstat->{vsize}) {
2254 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2255 }
2256
2257 my $old = $last_proc_pid_stat->{$pid};
2258 if (!$old) {
2259 $last_proc_pid_stat->{$pid} = {
2260 time => $ctime,
2261 used => $used,
2262 cpu => 0,
2263 };
2264 next;
2265 }
2266
2267 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2268
2269 if ($dtime > 1000) {
2270 my $dutime = $used - $old->{used};
2271
2272 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2273 $last_proc_pid_stat->{$pid} = {
2274 time => $ctime,
2275 used => $used,
2276 cpu => $d->{cpu},
2277 };
2278 } else {
2279 $d->{cpu} = $old->{cpu};
2280 }
2281 }
2282
2283 return $res if !$full;
2284
2285 my $qmpclient = PVE::QMPClient->new();
2286
2287 my $ballooncb = sub {
2288 my ($vmid, $resp) = @_;
2289
2290 my $info = $resp->{'return'};
2291 return if !$info->{max_mem};
2292
2293 my $d = $res->{$vmid};
2294
2295 # use memory assigned to VM
2296 $d->{maxmem} = $info->{max_mem};
2297 $d->{balloon} = $info->{actual};
2298
2299 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2300 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2301 $d->{freemem} = $info->{free_mem};
2302 }
2303
2304 };
2305
2306 my $blockstatscb = sub {
2307 my ($vmid, $resp) = @_;
2308 my $data = $resp->{'return'} || [];
2309 my $totalrdbytes = 0;
2310 my $totalwrbytes = 0;
2311 for my $blockstat (@$data) {
2312 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2313 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2314 }
2315 $res->{$vmid}->{diskread} = $totalrdbytes;
2316 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2317 };
2318
2319 my $statuscb = sub {
2320 my ($vmid, $resp) = @_;
2321
2322 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2323 # this fails if ballon driver is not loaded, so this must be
2324 # the last commnand (following command are aborted if this fails).
2325 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2326
2327 my $status = 'unknown';
2328 if (!defined($status = $resp->{'return'}->{status})) {
2329 warn "unable to get VM status\n";
2330 return;
2331 }
2332
2333 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2334 };
2335
2336 foreach my $vmid (keys %$list) {
2337 next if $opt_vmid && ($vmid ne $opt_vmid);
2338 next if !$res->{$vmid}->{pid}; # not running
2339 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2340 }
2341
2342 $qmpclient->queue_execute();
2343
2344 foreach my $vmid (keys %$list) {
2345 next if $opt_vmid && ($vmid ne $opt_vmid);
2346 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2347 }
2348
2349 return $res;
2350 }
2351
2352 sub foreach_drive {
2353 my ($conf, $func) = @_;
2354
2355 foreach my $ds (keys %$conf) {
2356 next if !valid_drivename($ds);
2357
2358 my $drive = parse_drive($ds, $conf->{$ds});
2359 next if !$drive;
2360
2361 &$func($ds, $drive);
2362 }
2363 }
2364
2365 sub foreach_volid {
2366 my ($conf, $func) = @_;
2367
2368 my $volhash = {};
2369
2370 my $test_volid = sub {
2371 my ($volid, $is_cdrom) = @_;
2372
2373 return if !$volid;
2374
2375 $volhash->{$volid} = $is_cdrom || 0;
2376 };
2377
2378 foreach_drive($conf, sub {
2379 my ($ds, $drive) = @_;
2380 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2381 });
2382
2383 foreach my $snapname (keys %{$conf->{snapshots}}) {
2384 my $snap = $conf->{snapshots}->{$snapname};
2385 &$test_volid($snap->{vmstate}, 0);
2386 foreach_drive($snap, sub {
2387 my ($ds, $drive) = @_;
2388 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2389 });
2390 }
2391
2392 foreach my $volid (keys %$volhash) {
2393 &$func($volid, $volhash->{$volid});
2394 }
2395 }
2396
2397 sub vga_conf_has_spice {
2398 my ($vga) = @_;
2399
2400 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2401
2402 return $1 || 1;
2403 }
2404
2405 sub config_to_command {
2406 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2407
2408 my $cmd = [];
2409 my $globalFlags = [];
2410 my $machineFlags = [];
2411 my $rtcFlags = [];
2412 my $cpuFlags = [];
2413 my $devices = [];
2414 my $pciaddr = '';
2415 my $bridges = {};
2416 my $kvmver = kvm_user_version();
2417 my $vernum = 0; # unknown
2418 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2419 $vernum = $1*1000000+$2*1000;
2420 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2421 $vernum = $1*1000000+$2*1000+$3;
2422 }
2423
2424 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2425
2426 my $have_ovz = -f '/proc/vz/vestat';
2427
2428 my $q35 = machine_type_is_q35($conf);
2429
2430 push @$cmd, '/usr/bin/kvm';
2431
2432 push @$cmd, '-id', $vmid;
2433
2434 my $use_virtio = 0;
2435
2436 my $qmpsocket = qmp_socket($vmid);
2437 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2438 push @$cmd, '-mon', "chardev=qmp,mode=control";
2439
2440 my $socket = vnc_socket($vmid);
2441 push @$cmd, '-vnc', "unix:$socket,x509,password";
2442
2443 push @$cmd, '-pidfile' , pidfile_name($vmid);
2444
2445 push @$cmd, '-daemonize';
2446
2447 if ($conf->{smbios1}) {
2448 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2449 }
2450
2451 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2452
2453 if ($q35) {
2454 # the q35 chipset support native usb2, so we enable usb controller
2455 # by default for this machine type
2456 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2457 } else {
2458 $pciaddr = print_pci_addr("piix3", $bridges);
2459 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2460
2461 my $use_usb2 = 0;
2462 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2463 next if !$conf->{"usb$i"};
2464 $use_usb2 = 1;
2465 }
2466 # include usb device config
2467 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2468 }
2469
2470 my $vga = $conf->{vga};
2471
2472 my $qxlnum = vga_conf_has_spice($vga);
2473 $vga = 'qxl' if $qxlnum;
2474
2475 if (!$vga) {
2476 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2477 $conf->{ostype} eq 'win7' ||
2478 $conf->{ostype} eq 'w2k8')) {
2479 $vga = 'std';
2480 } else {
2481 $vga = 'cirrus';
2482 }
2483 }
2484
2485 # enable absolute mouse coordinates (needed by vnc)
2486 my $tablet;
2487 if (defined($conf->{tablet})) {
2488 $tablet = $conf->{tablet};
2489 } else {
2490 $tablet = $defaults->{tablet};
2491 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2492 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2493 }
2494
2495 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2496
2497 # host pci devices
2498 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2499 my $d = parse_hostpci($conf->{"hostpci$i"});
2500 next if !$d;
2501
2502 my $pcie = $d->{pcie};
2503 if($pcie){
2504 die "q35 machine model is not enabled" if !$q35;
2505 $pciaddr = print_pcie_addr("hostpci$i");
2506 }else{
2507 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2508 }
2509
2510 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2511 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2512 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2513 if ($xvga && $xvga ne '') {
2514 push @$cpuFlags, 'kvm=off';
2515 $vga = 'none';
2516 }
2517 $driver = "vfio-pci" if $xvga ne '';
2518 my $pcidevices = $d->{pciid};
2519 my $multifunction = 1 if @$pcidevices > 1;
2520
2521 my $j=0;
2522 foreach my $pcidevice (@$pcidevices) {
2523
2524 my $id = "hostpci$i";
2525 $id .= ".$j" if $multifunction;
2526 my $addr = $pciaddr;
2527 $addr .= ".$j" if $multifunction;
2528 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2529
2530 if($j == 0){
2531 $devicestr .= "$rombar$xvga";
2532 $devicestr .= ",multifunction=on" if $multifunction;
2533 }
2534
2535 push @$devices, '-device', $devicestr;
2536 $j++;
2537 }
2538 }
2539
2540 # usb devices
2541 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2542 my $d = parse_usb_device($conf->{"usb$i"});
2543 next if !$d;
2544 if ($d->{vendorid} && $d->{productid}) {
2545 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2546 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2547 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2548 } elsif ($d->{spice}) {
2549 # usb redir support for spice
2550 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2551 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2552 }
2553 }
2554
2555 # serial devices
2556 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2557 if (my $path = $conf->{"serial$i"}) {
2558 if ($path eq 'socket') {
2559 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2560 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2561 push @$devices, '-device', "isa-serial,chardev=serial$i";
2562 } else {
2563 die "no such serial device\n" if ! -c $path;
2564 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2565 push @$devices, '-device', "isa-serial,chardev=serial$i";
2566 }
2567 }
2568 }
2569
2570 # parallel devices
2571 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2572 if (my $path = $conf->{"parallel$i"}) {
2573 die "no such parallel device\n" if ! -c $path;
2574 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2575 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2576 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2577 }
2578 }
2579
2580 my $vmname = $conf->{name} || "vm$vmid";
2581
2582 push @$cmd, '-name', $vmname;
2583
2584 my $sockets = 1;
2585 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2586 $sockets = $conf->{sockets} if $conf->{sockets};
2587
2588 my $cores = $conf->{cores} || 1;
2589 my $maxcpus = $conf->{maxcpus} if $conf->{maxcpus};
2590
2591 if ($maxcpus) {
2592 push @$cmd, '-smp', "cpus=$cores,maxcpus=$maxcpus";
2593 } else {
2594 push @$cmd, '-smp', "sockets=$sockets,cores=$cores";
2595 }
2596
2597 push @$cmd, '-nodefaults';
2598
2599 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2600
2601 my $bootindex_hash = {};
2602 my $i = 1;
2603 foreach my $o (split(//, $bootorder)) {
2604 $bootindex_hash->{$o} = $i*100;
2605 $i++;
2606 }
2607
2608 push @$cmd, '-boot', "menu=on";
2609
2610 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2611
2612 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2613
2614 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2615
2616 # time drift fix
2617 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2618
2619 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2620 my $useLocaltime = $conf->{localtime};
2621
2622 if (my $ost = $conf->{ostype}) {
2623 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2624
2625 if ($ost =~ m/^w/) { # windows
2626 $useLocaltime = 1 if !defined($conf->{localtime});
2627
2628 # use time drift fix when acpi is enabled
2629 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2630 $tdf = 1 if !defined($conf->{tdf});
2631 }
2632 }
2633
2634 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2635 $ost eq 'wvista') {
2636 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2637 push @$cmd, '-no-hpet';
2638 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2639 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2640 }
2641
2642 if ($ost eq 'win7' || $ost eq 'win8') {
2643 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2644 }
2645 }
2646
2647 push @$rtcFlags, 'driftfix=slew' if $tdf;
2648
2649 if ($nokvm) {
2650 push @$machineFlags, 'accel=tcg';
2651 } else {
2652 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2653 }
2654
2655 my $machine_type = $forcemachine || $conf->{machine};
2656 if ($machine_type) {
2657 push @$machineFlags, "type=${machine_type}";
2658 }
2659
2660 if ($conf->{startdate}) {
2661 push @$rtcFlags, "base=$conf->{startdate}";
2662 } elsif ($useLocaltime) {
2663 push @$rtcFlags, 'base=localtime';
2664 }
2665
2666 my $cpu = $nokvm ? "qemu64" : "kvm64";
2667 $cpu = $conf->{cpu} if $conf->{cpu};
2668
2669 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2670
2671 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2672
2673 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2674
2675 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2676
2677 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2678
2679 # Note: enforce needs kernel 3.10, so we do not use it for now
2680 # push @$cmd, '-cpu', "$cpu,enforce";
2681 push @$cmd, '-cpu', $cpu;
2682
2683 push @$cmd, '-S' if $conf->{freeze};
2684
2685 # set keyboard layout
2686 my $kb = $conf->{keyboard} || $defaults->{keyboard};
2687 push @$cmd, '-k', $kb if $kb;
2688
2689 # enable sound
2690 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
2691 #push @$cmd, '-soundhw', 'es1370';
2692 #push @$cmd, '-soundhw', $soundhw if $soundhw;
2693
2694 if($conf->{agent}) {
2695 my $qgasocket = qga_socket($vmid);
2696 my $pciaddr = print_pci_addr("qga0", $bridges);
2697 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
2698 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
2699 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
2700 }
2701
2702 my $spice_port;
2703
2704 if ($qxlnum) {
2705 if ($qxlnum > 1) {
2706 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
2707 for(my $i = 1; $i < $qxlnum; $i++){
2708 my $pciaddr = print_pci_addr("vga$i", $bridges);
2709 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
2710 }
2711 } else {
2712 # assume other OS works like Linux
2713 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
2714 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
2715 }
2716 }
2717
2718 my $pciaddr = print_pci_addr("spice", $bridges);
2719
2720 $spice_port = PVE::Tools::next_spice_port();
2721
2722 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
2723
2724 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
2725 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
2726 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
2727 }
2728
2729 # enable balloon by default, unless explicitly disabled
2730 if (!defined($conf->{balloon}) || $conf->{balloon}) {
2731 $pciaddr = print_pci_addr("balloon0", $bridges);
2732 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
2733 }
2734
2735 if ($conf->{watchdog}) {
2736 my $wdopts = parse_watchdog($conf->{watchdog});
2737 $pciaddr = print_pci_addr("watchdog", $bridges);
2738 my $watchdog = $wdopts->{model} || 'i6300esb';
2739 push @$devices, '-device', "$watchdog$pciaddr";
2740 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
2741 }
2742
2743 my $vollist = [];
2744 my $scsicontroller = {};
2745 my $ahcicontroller = {};
2746 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
2747
2748 # Add iscsi initiator name if available
2749 if (my $initiator = get_initiator_name()) {
2750 push @$devices, '-iscsi', "initiator-name=$initiator";
2751 }
2752
2753 foreach_drive($conf, sub {
2754 my ($ds, $drive) = @_;
2755
2756 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
2757 push @$vollist, $drive->{file};
2758 }
2759
2760 $use_virtio = 1 if $ds =~ m/^virtio/;
2761
2762 if (drive_is_cdrom ($drive)) {
2763 if ($bootindex_hash->{d}) {
2764 $drive->{bootindex} = $bootindex_hash->{d};
2765 $bootindex_hash->{d} += 1;
2766 }
2767 } else {
2768 if ($bootindex_hash->{c}) {
2769 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
2770 $bootindex_hash->{c} += 1;
2771 }
2772 }
2773
2774 if ($drive->{interface} eq 'scsi') {
2775
2776 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
2777 my $controller = int($drive->{index} / $maxdev);
2778 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
2779 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
2780 $scsicontroller->{$controller}=1;
2781 }
2782
2783 if ($drive->{interface} eq 'sata') {
2784 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
2785 $pciaddr = print_pci_addr("ahci$controller", $bridges);
2786 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
2787 $ahcicontroller->{$controller}=1;
2788 }
2789
2790 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
2791 push @$devices, '-drive',$drive_cmd;
2792 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
2793 });
2794
2795 push @$cmd, '-m', $conf->{memory} || $defaults->{memory};
2796
2797 for (my $i = 0; $i < $MAX_NETS; $i++) {
2798 next if !$conf->{"net$i"};
2799 my $d = parse_net($conf->{"net$i"});
2800 next if !$d;
2801
2802 $use_virtio = 1 if $d->{model} eq 'virtio';
2803
2804 if ($bootindex_hash->{n}) {
2805 $d->{bootindex} = $bootindex_hash->{n};
2806 $bootindex_hash->{n} += 1;
2807 }
2808
2809 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
2810 push @$devices, '-netdev', $netdevfull;
2811
2812 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
2813 push @$devices, '-device', $netdevicefull;
2814 }
2815
2816 if (!$q35) {
2817 # add pci bridges
2818 while (my ($k, $v) = each %$bridges) {
2819 $pciaddr = print_pci_addr("pci.$k");
2820 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
2821 }
2822 }
2823
2824 # hack: virtio with fairsched is unreliable, so we do not use fairsched
2825 # when the VM uses virtio devices.
2826 if (!$use_virtio && $have_ovz) {
2827
2828 my $cpuunits = defined($conf->{cpuunits}) ?
2829 $conf->{cpuunits} : $defaults->{cpuunits};
2830
2831 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
2832
2833 # fixme: cpulimit is currently ignored
2834 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
2835 }
2836
2837 # add custom args
2838 if ($conf->{args}) {
2839 my $aa = PVE::Tools::split_args($conf->{args});
2840 push @$cmd, @$aa;
2841 }
2842
2843 push @$cmd, @$devices;
2844 push @$cmd, '-rtc', join(',', @$rtcFlags)
2845 if scalar(@$rtcFlags);
2846 push @$cmd, '-machine', join(',', @$machineFlags)
2847 if scalar(@$machineFlags);
2848 push @$cmd, '-global', join(',', @$globalFlags)
2849 if scalar(@$globalFlags);
2850
2851 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
2852 }
2853
2854 sub vnc_socket {
2855 my ($vmid) = @_;
2856 return "${var_run_tmpdir}/$vmid.vnc";
2857 }
2858
2859 sub spice_port {
2860 my ($vmid) = @_;
2861
2862 my $res = vm_mon_cmd($vmid, 'query-spice');
2863
2864 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
2865 }
2866
2867 sub qmp_socket {
2868 my ($vmid) = @_;
2869 return "${var_run_tmpdir}/$vmid.qmp";
2870 }
2871
2872 sub qga_socket {
2873 my ($vmid) = @_;
2874 return "${var_run_tmpdir}/$vmid.qga";
2875 }
2876
2877 sub pidfile_name {
2878 my ($vmid) = @_;
2879 return "${var_run_tmpdir}/$vmid.pid";
2880 }
2881
2882 sub vm_devices_list {
2883 my ($vmid) = @_;
2884
2885 my $res = vm_mon_cmd($vmid, 'query-pci');
2886
2887 my $devices = {};
2888 foreach my $pcibus (@$res) {
2889 foreach my $device (@{$pcibus->{devices}}) {
2890 next if !$device->{'qdev_id'};
2891 $devices->{$device->{'qdev_id'}} = 1;
2892 }
2893 }
2894
2895 my $resblock = vm_mon_cmd($vmid, 'query-block');
2896 foreach my $block (@$resblock) {
2897 if($block->{device} =~ m/^drive-(\S+)/){
2898 $devices->{$1} = 1;
2899 }
2900 }
2901
2902 return $devices;
2903 }
2904
2905 sub vm_deviceplug {
2906 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
2907
2908 return 1 if !check_running($vmid);
2909
2910 my $q35 = machine_type_is_q35($conf);
2911
2912 if ($deviceid eq 'tablet') {
2913 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
2914 return 1;
2915 }
2916
2917 return 1 if !$conf->{hotplug};
2918
2919 my $devices_list = vm_devices_list($vmid);
2920 return 1 if defined($devices_list->{$deviceid});
2921
2922 qemu_bridgeadd($storecfg, $conf, $vmid, $deviceid); #add bridge if we need it for the device
2923
2924 if ($deviceid =~ m/^(virtio)(\d+)$/) {
2925 return undef if !qemu_driveadd($storecfg, $vmid, $device);
2926 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
2927 qemu_deviceadd($vmid, $devicefull);
2928 if(!qemu_deviceaddverify($vmid, $deviceid)) {
2929 qemu_drivedel($vmid, $deviceid);
2930 return undef;
2931 }
2932 }
2933
2934 if ($deviceid =~ m/^(scsihw)(\d+)$/) {
2935 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
2936 my $pciaddr = print_pci_addr($deviceid);
2937 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
2938 qemu_deviceadd($vmid, $devicefull);
2939 return undef if(!qemu_deviceaddverify($vmid, $deviceid));
2940 }
2941
2942 if ($deviceid =~ m/^(scsi)(\d+)$/) {
2943 return undef if !qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
2944 return undef if !qemu_driveadd($storecfg, $vmid, $device);
2945 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
2946 if(!qemu_deviceadd($vmid, $devicefull)) {
2947 qemu_drivedel($vmid, $deviceid);
2948 return undef;
2949 }
2950 }
2951
2952 if ($deviceid =~ m/^(net)(\d+)$/) {
2953 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
2954 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
2955 qemu_deviceadd($vmid, $netdevicefull);
2956 if(!qemu_deviceaddverify($vmid, $deviceid)) {
2957 qemu_netdevdel($vmid, $deviceid);
2958 return undef;
2959 }
2960 }
2961
2962
2963 if (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
2964 my $bridgeid = $2;
2965 my $pciaddr = print_pci_addr($deviceid);
2966 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
2967 qemu_deviceadd($vmid, $devicefull);
2968 return undef if !qemu_deviceaddverify($vmid, $deviceid);
2969 }
2970
2971 return 1;
2972 }
2973
2974 sub vm_deviceunplug {
2975 my ($vmid, $conf, $deviceid) = @_;
2976
2977 return 1 if !check_running ($vmid);
2978
2979 if ($deviceid eq 'tablet') {
2980 qemu_devicedel($vmid, $deviceid);
2981 return 1;
2982 }
2983
2984 return 1 if !$conf->{hotplug};
2985
2986 my $devices_list = vm_devices_list($vmid);
2987 return 1 if !defined($devices_list->{$deviceid});
2988
2989 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
2990
2991 if ($deviceid =~ m/^(virtio)(\d+)$/) {
2992 qemu_devicedel($vmid, $deviceid);
2993 return undef if !qemu_devicedelverify($vmid, $deviceid);
2994 return undef if !qemu_drivedel($vmid, $deviceid);
2995 }
2996
2997 if ($deviceid =~ m/^(lsi)(\d+)$/) {
2998 return undef if !qemu_devicedel($vmid, $deviceid);
2999 }
3000
3001 if ($deviceid =~ m/^(scsi)(\d+)$/) {
3002 return undef if !qemu_devicedel($vmid, $deviceid);
3003 return undef if !qemu_drivedel($vmid, $deviceid);
3004 }
3005
3006 if ($deviceid =~ m/^(net)(\d+)$/) {
3007 qemu_devicedel($vmid, $deviceid);
3008 return undef if !qemu_devicedelverify($vmid, $deviceid);
3009 return undef if !qemu_netdevdel($vmid, $deviceid);
3010 }
3011
3012 return 1;
3013 }
3014
3015 sub qemu_deviceadd {
3016 my ($vmid, $devicefull) = @_;
3017
3018 $devicefull = "driver=".$devicefull;
3019 my %options = split(/[=,]/, $devicefull);
3020
3021 vm_mon_cmd($vmid, "device_add" , %options);
3022 return 1;
3023 }
3024
3025 sub qemu_devicedel {
3026 my($vmid, $deviceid) = @_;
3027 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3028 return 1;
3029 }
3030
3031 sub qemu_driveadd {
3032 my($storecfg, $vmid, $device) = @_;
3033
3034 my $drive = print_drive_full($storecfg, $vmid, $device);
3035 my $ret = vm_human_monitor_command($vmid, "drive_add auto $drive");
3036 # If the command succeeds qemu prints: "OK"
3037 if ($ret !~ m/OK/s) {
3038 syslog("err", "adding drive failed: $ret");
3039 return undef;
3040 }
3041 return 1;
3042 }
3043
3044 sub qemu_drivedel {
3045 my($vmid, $deviceid) = @_;
3046
3047 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3048 $ret =~ s/^\s+//;
3049 if ($ret =~ m/Device \'.*?\' not found/s) {
3050 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3051 }
3052 elsif ($ret ne "") {
3053 syslog("err", "deleting drive $deviceid failed : $ret");
3054 return undef;
3055 }
3056 return 1;
3057 }
3058
3059 sub qemu_deviceaddverify {
3060 my ($vmid,$deviceid) = @_;
3061
3062 for (my $i = 0; $i <= 5; $i++) {
3063 my $devices_list = vm_devices_list($vmid);
3064 return 1 if defined($devices_list->{$deviceid});
3065 sleep 1;
3066 }
3067 syslog("err", "error on hotplug device $deviceid");
3068 return undef;
3069 }
3070
3071
3072 sub qemu_devicedelverify {
3073 my ($vmid,$deviceid) = @_;
3074
3075 #need to verify the device is correctly remove as device_del is async and empty return is not reliable
3076 for (my $i = 0; $i <= 5; $i++) {
3077 my $devices_list = vm_devices_list($vmid);
3078 return 1 if !defined($devices_list->{$deviceid});
3079 sleep 1;
3080 }
3081 syslog("err", "error on hot-unplugging device $deviceid");
3082 return undef;
3083 }
3084
3085 sub qemu_findorcreatescsihw {
3086 my ($storecfg, $conf, $vmid, $device) = @_;
3087
3088 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3089 my $controller = int($device->{index} / $maxdev);
3090 my $scsihwid="scsihw$controller";
3091 my $devices_list = vm_devices_list($vmid);
3092
3093 if(!defined($devices_list->{$scsihwid})) {
3094 return undef if !vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3095 }
3096 return 1;
3097 }
3098
3099 sub qemu_bridgeadd {
3100 my ($storecfg, $conf, $vmid, $device) = @_;
3101
3102 my $bridges = {};
3103 my $bridgeid = undef;
3104 print_pci_addr($device, $bridges);
3105
3106 while (my ($k, $v) = each %$bridges) {
3107 $bridgeid = $k;
3108 }
3109 return if !$bridgeid || $bridgeid < 1;
3110 my $bridge = "pci.$bridgeid";
3111 my $devices_list = vm_devices_list($vmid);
3112
3113 if(!defined($devices_list->{$bridge})) {
3114 return undef if !vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3115 }
3116 return 1;
3117 }
3118
3119 sub qemu_netdevadd {
3120 my ($vmid, $conf, $device, $deviceid) = @_;
3121
3122 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3123 my %options = split(/[=,]/, $netdev);
3124
3125 vm_mon_cmd($vmid, "netdev_add", %options);
3126 return 1;
3127 }
3128
3129 sub qemu_netdevdel {
3130 my ($vmid, $deviceid) = @_;
3131
3132 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3133 return 1;
3134 }
3135
3136 sub qemu_cpu_hotplug {
3137 my ($vmid, $conf, $cores) = @_;
3138
3139 die "new cores config is not defined" if !$cores;
3140 die "you can't add more cores than maxcpus"
3141 if $conf->{maxcpus} && ($cores > $conf->{maxcpus});
3142 return if !check_running($vmid);
3143
3144 my $currentcores = $conf->{cores} if $conf->{cores};
3145 die "current cores is not defined" if !$currentcores;
3146 die "maxcpus is not defined" if !$conf->{maxcpus};
3147 raise_param_exc({ 'cores' => "online cpu unplug is not yet possible" })
3148 if($cores < $currentcores);
3149
3150 my $currentrunningcores = vm_mon_cmd($vmid, "query-cpus");
3151 raise_param_exc({ 'cores' => "cores number if running vm is different than configuration" })
3152 if scalar (@{$currentrunningcores}) != $currentcores;
3153
3154 for(my $i = $currentcores; $i < $cores; $i++) {
3155 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3156 }
3157 }
3158
3159 sub qemu_block_set_io_throttle {
3160 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3161
3162 return if !check_running($vmid) ;
3163
3164 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($iops_wr));
3165
3166 }
3167
3168 # old code, only used to shutdown old VM after update
3169 sub __read_avail {
3170 my ($fh, $timeout) = @_;
3171
3172 my $sel = new IO::Select;
3173 $sel->add($fh);
3174
3175 my $res = '';
3176 my $buf;
3177
3178 my @ready;
3179 while (scalar (@ready = $sel->can_read($timeout))) {
3180 my $count;
3181 if ($count = $fh->sysread($buf, 8192)) {
3182 if ($buf =~ /^(.*)\(qemu\) $/s) {
3183 $res .= $1;
3184 last;
3185 } else {
3186 $res .= $buf;
3187 }
3188 } else {
3189 if (!defined($count)) {
3190 die "$!\n";
3191 }
3192 last;
3193 }
3194 }
3195
3196 die "monitor read timeout\n" if !scalar(@ready);
3197
3198 return $res;
3199 }
3200
3201 # old code, only used to shutdown old VM after update
3202 sub vm_monitor_command {
3203 my ($vmid, $cmdstr, $nocheck) = @_;
3204
3205 my $res;
3206
3207 eval {
3208 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3209
3210 my $sname = "${var_run_tmpdir}/$vmid.mon";
3211
3212 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3213 die "unable to connect to VM $vmid socket - $!\n";
3214
3215 my $timeout = 3;
3216
3217 # hack: migrate sometime blocks the monitor (when migrate_downtime
3218 # is set)
3219 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3220 $timeout = 60*60; # 1 hour
3221 }
3222
3223 # read banner;
3224 my $data = __read_avail($sock, $timeout);
3225
3226 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3227 die "got unexpected qemu monitor banner\n";
3228 }
3229
3230 my $sel = new IO::Select;
3231 $sel->add($sock);
3232
3233 if (!scalar(my @ready = $sel->can_write($timeout))) {
3234 die "monitor write error - timeout";
3235 }
3236
3237 my $fullcmd = "$cmdstr\r";
3238
3239 # syslog('info', "VM $vmid monitor command: $cmdstr");
3240
3241 my $b;
3242 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3243 die "monitor write error - $!";
3244 }
3245
3246 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3247
3248 $timeout = 20;
3249
3250 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3251 $timeout = 60*60; # 1 hour
3252 } elsif ($cmdstr =~ m/^(eject|change)/) {
3253 $timeout = 60; # note: cdrom mount command is slow
3254 }
3255 if ($res = __read_avail($sock, $timeout)) {
3256
3257 my @lines = split("\r?\n", $res);
3258
3259 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3260
3261 $res = join("\n", @lines);
3262 $res .= "\n";
3263 }
3264 };
3265
3266 my $err = $@;
3267
3268 if ($err) {
3269 syslog("err", "VM $vmid monitor command failed - $err");
3270 die $err;
3271 }
3272
3273 return $res;
3274 }
3275
3276 sub qemu_block_resize {
3277 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3278
3279 my $running = check_running($vmid);
3280
3281 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3282
3283 return if !$running;
3284
3285 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3286
3287 }
3288
3289 sub qemu_volume_snapshot {
3290 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3291
3292 my $running = check_running($vmid);
3293
3294 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3295
3296 return if !$running;
3297
3298 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3299
3300 }
3301
3302 sub qemu_volume_snapshot_delete {
3303 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3304
3305 my $running = check_running($vmid);
3306
3307 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3308
3309 return if !$running;
3310
3311 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3312 }
3313
3314 sub qga_freezefs {
3315 my ($vmid) = @_;
3316
3317 #need to impplement call to qemu-ga
3318 }
3319
3320 sub qga_unfreezefs {
3321 my ($vmid) = @_;
3322
3323 #need to impplement call to qemu-ga
3324 }
3325
3326 sub set_migration_caps {
3327 my ($vmid) = @_;
3328
3329 my $cap_ref = [];
3330
3331 my $enabled_cap = {
3332 "auto-converge" => 1,
3333 "xbzrle" => 0,
3334 "x-rdma-pin-all" => 0,
3335 "zero-blocks" => 0,
3336 };
3337
3338 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3339
3340 for my $supported_capability (@$supported_capabilities) {
3341 push @$cap_ref, {
3342 capability => $supported_capability->{capability},
3343 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3344 };
3345 }
3346
3347 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3348 }
3349
3350 sub vm_start {
3351 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
3352
3353 lock_config($vmid, sub {
3354 my $conf = load_config($vmid, $migratedfrom);
3355
3356 die "you can't start a vm if it's a template\n" if is_template($conf);
3357
3358 check_lock($conf) if !$skiplock;
3359
3360 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
3361
3362 my $defaults = load_defaults();
3363
3364 # set environment variable useful inside network script
3365 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
3366
3367 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
3368
3369 my $migrate_port = 0;
3370 my $migrate_uri;
3371 if ($statefile) {
3372 if ($statefile eq 'tcp') {
3373 my $localip = "localhost";
3374 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
3375 if ($datacenterconf->{migration_unsecure}) {
3376 my $nodename = PVE::INotify::nodename();
3377 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
3378 }
3379 $migrate_port = PVE::Tools::next_migrate_port();
3380 $migrate_uri = "tcp:${localip}:${migrate_port}";
3381 push @$cmd, '-incoming', $migrate_uri;
3382 push @$cmd, '-S';
3383 } else {
3384 push @$cmd, '-loadstate', $statefile;
3385 }
3386 } elsif ($paused) {
3387 push @$cmd, '-S';
3388 }
3389
3390 # host pci devices
3391 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3392 my $d = parse_hostpci($conf->{"hostpci$i"});
3393 next if !$d;
3394 my $pcidevices = $d->{pciid};
3395 foreach my $pcidevice (@$pcidevices) {
3396 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
3397
3398 my $info = pci_device_info("0000:$pciid");
3399 die "IOMMU not present\n" if !check_iommu_support();
3400 die "no pci device info for device '$pciid'\n" if !$info;
3401
3402 if ($d->{driver} && $d->{driver} eq "vfio") {
3403 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
3404 } else {
3405 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
3406 }
3407
3408 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
3409 }
3410 }
3411
3412 PVE::Storage::activate_volumes($storecfg, $vollist);
3413
3414 eval { run_command($cmd, timeout => $statefile ? undef : 30,
3415 umask => 0077); };
3416 my $err = $@;
3417 die "start failed: $err" if $err;
3418
3419 print "migration listens on $migrate_uri\n" if $migrate_uri;
3420
3421 if ($statefile && $statefile ne 'tcp') {
3422 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
3423 warn $@ if $@;
3424 }
3425
3426 if ($migratedfrom) {
3427
3428 eval {
3429 PVE::QemuServer::set_migration_caps($vmid);
3430 };
3431 warn $@ if $@;
3432
3433 if ($spice_port) {
3434 print "spice listens on port $spice_port\n";
3435 if ($spice_ticket) {
3436 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
3437 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
3438 }
3439 }
3440
3441 } else {
3442
3443 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
3444 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
3445 if $conf->{balloon};
3446 vm_mon_cmd_nocheck($vmid, 'qom-set',
3447 path => "machine/peripheral/balloon0",
3448 property => "guest-stats-polling-interval",
3449 value => 2);
3450 }
3451 }
3452 });
3453 }
3454
3455 sub vm_mon_cmd {
3456 my ($vmid, $execute, %params) = @_;
3457
3458 my $cmd = { execute => $execute, arguments => \%params };
3459 vm_qmp_command($vmid, $cmd);
3460 }
3461
3462 sub vm_mon_cmd_nocheck {
3463 my ($vmid, $execute, %params) = @_;
3464
3465 my $cmd = { execute => $execute, arguments => \%params };
3466 vm_qmp_command($vmid, $cmd, 1);
3467 }
3468
3469 sub vm_qmp_command {
3470 my ($vmid, $cmd, $nocheck) = @_;
3471
3472 my $res;
3473
3474 my $timeout;
3475 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
3476 $timeout = $cmd->{arguments}->{timeout};
3477 delete $cmd->{arguments}->{timeout};
3478 }
3479
3480 eval {
3481 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3482 my $sname = qmp_socket($vmid);
3483 if (-e $sname) {
3484 my $qmpclient = PVE::QMPClient->new();
3485
3486 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
3487 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
3488 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
3489 if scalar(%{$cmd->{arguments}});
3490 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
3491 } else {
3492 die "unable to open monitor socket\n";
3493 }
3494 };
3495 if (my $err = $@) {
3496 syslog("err", "VM $vmid qmp command failed - $err");
3497 die $err;
3498 }
3499
3500 return $res;
3501 }
3502
3503 sub vm_human_monitor_command {
3504 my ($vmid, $cmdline) = @_;
3505
3506 my $res;
3507
3508 my $cmd = {
3509 execute => 'human-monitor-command',
3510 arguments => { 'command-line' => $cmdline},
3511 };
3512
3513 return vm_qmp_command($vmid, $cmd);
3514 }
3515
3516 sub vm_commandline {
3517 my ($storecfg, $vmid) = @_;
3518
3519 my $conf = load_config($vmid);
3520
3521 my $defaults = load_defaults();
3522
3523 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
3524
3525 return join(' ', @$cmd);
3526 }
3527
3528 sub vm_reset {
3529 my ($vmid, $skiplock) = @_;
3530
3531 lock_config($vmid, sub {
3532
3533 my $conf = load_config($vmid);
3534
3535 check_lock($conf) if !$skiplock;
3536
3537 vm_mon_cmd($vmid, "system_reset");
3538 });
3539 }
3540
3541 sub get_vm_volumes {
3542 my ($conf) = @_;
3543
3544 my $vollist = [];
3545 foreach_volid($conf, sub {
3546 my ($volid, $is_cdrom) = @_;
3547
3548 return if $volid =~ m|^/|;
3549
3550 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
3551 return if !$sid;
3552
3553 push @$vollist, $volid;
3554 });
3555
3556 return $vollist;
3557 }
3558
3559 sub vm_stop_cleanup {
3560 my ($storecfg, $vmid, $conf, $keepActive) = @_;
3561
3562 eval {
3563 fairsched_rmnod($vmid); # try to destroy group
3564
3565 if (!$keepActive) {
3566 my $vollist = get_vm_volumes($conf);
3567 PVE::Storage::deactivate_volumes($storecfg, $vollist);
3568 }
3569
3570 foreach my $ext (qw(mon qmp pid vnc qga)) {
3571 unlink "/var/run/qemu-server/${vmid}.$ext";
3572 }
3573 };
3574 warn $@ if $@; # avoid errors - just warn
3575 }
3576
3577 # Note: use $nockeck to skip tests if VM configuration file exists.
3578 # We need that when migration VMs to other nodes (files already moved)
3579 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
3580 sub vm_stop {
3581 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
3582
3583 $force = 1 if !defined($force) && !$shutdown;
3584
3585 if ($migratedfrom){
3586 my $pid = check_running($vmid, $nocheck, $migratedfrom);
3587 kill 15, $pid if $pid;
3588 my $conf = load_config($vmid, $migratedfrom);
3589 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive);
3590 return;
3591 }
3592
3593 lock_config($vmid, sub {
3594
3595 my $pid = check_running($vmid, $nocheck);
3596 return if !$pid;
3597
3598 my $conf;
3599 if (!$nocheck) {
3600 $conf = load_config($vmid);
3601 check_lock($conf) if !$skiplock;
3602 if (!defined($timeout) && $shutdown && $conf->{startup}) {
3603 my $opts = parse_startup($conf->{startup});
3604 $timeout = $opts->{down} if $opts->{down};
3605 }
3606 }
3607
3608 $timeout = 60 if !defined($timeout);
3609
3610 eval {
3611 if ($shutdown) {
3612 $nocheck ? vm_mon_cmd_nocheck($vmid, "system_powerdown") : vm_mon_cmd($vmid, "system_powerdown");
3613
3614 } else {
3615 $nocheck ? vm_mon_cmd_nocheck($vmid, "quit") : vm_mon_cmd($vmid, "quit");
3616 }
3617 };
3618 my $err = $@;
3619
3620 if (!$err) {
3621 my $count = 0;
3622 while (($count < $timeout) && check_running($vmid, $nocheck)) {
3623 $count++;
3624 sleep 1;
3625 }
3626
3627 if ($count >= $timeout) {
3628 if ($force) {
3629 warn "VM still running - terminating now with SIGTERM\n";
3630 kill 15, $pid;
3631 } else {
3632 die "VM quit/powerdown failed - got timeout\n";
3633 }
3634 } else {
3635 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
3636 return;
3637 }
3638 } else {
3639 if ($force) {
3640 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
3641 kill 15, $pid;
3642 } else {
3643 die "VM quit/powerdown failed\n";
3644 }
3645 }
3646
3647 # wait again
3648 $timeout = 10;
3649
3650 my $count = 0;
3651 while (($count < $timeout) && check_running($vmid, $nocheck)) {
3652 $count++;
3653 sleep 1;
3654 }
3655
3656 if ($count >= $timeout) {
3657 warn "VM still running - terminating now with SIGKILL\n";
3658 kill 9, $pid;
3659 sleep 1;
3660 }
3661
3662 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
3663 });
3664 }
3665
3666 sub vm_suspend {
3667 my ($vmid, $skiplock) = @_;
3668
3669 lock_config($vmid, sub {
3670
3671 my $conf = load_config($vmid);
3672
3673 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
3674
3675 vm_mon_cmd($vmid, "stop");
3676 });
3677 }
3678
3679 sub vm_resume {
3680 my ($vmid, $skiplock) = @_;
3681
3682 lock_config($vmid, sub {
3683
3684 my $conf = load_config($vmid);
3685
3686 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
3687
3688 vm_mon_cmd($vmid, "cont");
3689 });
3690 }
3691
3692 sub vm_sendkey {
3693 my ($vmid, $skiplock, $key) = @_;
3694
3695 lock_config($vmid, sub {
3696
3697 my $conf = load_config($vmid);
3698
3699 # there is no qmp command, so we use the human monitor command
3700 vm_human_monitor_command($vmid, "sendkey $key");
3701 });
3702 }
3703
3704 sub vm_destroy {
3705 my ($storecfg, $vmid, $skiplock) = @_;
3706
3707 lock_config($vmid, sub {
3708
3709 my $conf = load_config($vmid);
3710
3711 check_lock($conf) if !$skiplock;
3712
3713 if (!check_running($vmid)) {
3714 fairsched_rmnod($vmid); # try to destroy group
3715 destroy_vm($storecfg, $vmid);
3716 } else {
3717 die "VM $vmid is running - destroy failed\n";
3718 }
3719 });
3720 }
3721
3722 # pci helpers
3723
3724 sub file_write {
3725 my ($filename, $buf) = @_;
3726
3727 my $fh = IO::File->new($filename, "w");
3728 return undef if !$fh;
3729
3730 my $res = print $fh $buf;
3731
3732 $fh->close();
3733
3734 return $res;
3735 }
3736
3737 sub pci_device_info {
3738 my ($name) = @_;
3739
3740 my $res;
3741
3742 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
3743 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
3744
3745 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
3746 return undef if !defined($irq) || $irq !~ m/^\d+$/;
3747
3748 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
3749 return undef if !defined($vendor) || $vendor !~ s/^0x//;
3750
3751 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
3752 return undef if !defined($product) || $product !~ s/^0x//;
3753
3754 $res = {
3755 name => $name,
3756 vendor => $vendor,
3757 product => $product,
3758 domain => $domain,
3759 bus => $bus,
3760 slot => $slot,
3761 func => $func,
3762 irq => $irq,
3763 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
3764 };
3765
3766 return $res;
3767 }
3768
3769 sub pci_dev_reset {
3770 my ($dev) = @_;
3771
3772 my $name = $dev->{name};
3773
3774 my $fn = "$pcisysfs/devices/$name/reset";
3775
3776 return file_write($fn, "1");
3777 }
3778
3779 sub pci_dev_bind_to_stub {
3780 my ($dev) = @_;
3781
3782 my $name = $dev->{name};
3783
3784 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
3785 return 1 if -d $testdir;
3786
3787 my $data = "$dev->{vendor} $dev->{product}";
3788 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
3789
3790 my $fn = "$pcisysfs/devices/$name/driver/unbind";
3791 if (!file_write($fn, $name)) {
3792 return undef if -f $fn;
3793 }
3794
3795 $fn = "$pcisysfs/drivers/pci-stub/bind";
3796 if (! -d $testdir) {
3797 return undef if !file_write($fn, $name);
3798 }
3799
3800 return -d $testdir;
3801 }
3802
3803 sub pci_dev_bind_to_vfio {
3804 my ($dev) = @_;
3805
3806 my $name = $dev->{name};
3807
3808 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
3809
3810 if (!-d $vfio_basedir) {
3811 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
3812 }
3813 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
3814
3815 my $testdir = "$vfio_basedir/$name";
3816 return 1 if -d $testdir;
3817
3818 my $data = "$dev->{vendor} $dev->{product}";
3819 return undef if !file_write("$vfio_basedir/new_id", $data);
3820
3821 my $fn = "$pcisysfs/devices/$name/driver/unbind";
3822 if (!file_write($fn, $name)) {
3823 return undef if -f $fn;
3824 }
3825
3826 $fn = "$vfio_basedir/bind";
3827 if (! -d $testdir) {
3828 return undef if !file_write($fn, $name);
3829 }
3830
3831 return -d $testdir;
3832 }
3833
3834 sub pci_dev_group_bind_to_vfio {
3835 my ($pciid) = @_;
3836
3837 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
3838
3839 if (!-d $vfio_basedir) {
3840 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
3841 }
3842 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
3843
3844 # get IOMMU group devices
3845 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
3846 my @devs = grep /^0000:/, readdir($D);
3847 closedir($D);
3848
3849 foreach my $pciid (@devs) {
3850 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
3851 my $info = pci_device_info($1);
3852 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
3853 }
3854
3855 return 1;
3856 }
3857
3858 sub print_pci_addr {
3859 my ($id, $bridges) = @_;
3860
3861 my $res = '';
3862 my $devices = {
3863 piix3 => { bus => 0, addr => 1 },
3864 #addr2 : first videocard
3865 balloon0 => { bus => 0, addr => 3 },
3866 watchdog => { bus => 0, addr => 4 },
3867 scsihw0 => { bus => 0, addr => 5 },
3868 scsihw1 => { bus => 0, addr => 6 },
3869 ahci0 => { bus => 0, addr => 7 },
3870 qga0 => { bus => 0, addr => 8 },
3871 spice => { bus => 0, addr => 9 },
3872 virtio0 => { bus => 0, addr => 10 },
3873 virtio1 => { bus => 0, addr => 11 },
3874 virtio2 => { bus => 0, addr => 12 },
3875 virtio3 => { bus => 0, addr => 13 },
3876 virtio4 => { bus => 0, addr => 14 },
3877 virtio5 => { bus => 0, addr => 15 },
3878 hostpci0 => { bus => 0, addr => 16 },
3879 hostpci1 => { bus => 0, addr => 17 },
3880 net0 => { bus => 0, addr => 18 },
3881 net1 => { bus => 0, addr => 19 },
3882 net2 => { bus => 0, addr => 20 },
3883 net3 => { bus => 0, addr => 21 },
3884 net4 => { bus => 0, addr => 22 },
3885 net5 => { bus => 0, addr => 23 },
3886 vga1 => { bus => 0, addr => 24 },
3887 vga2 => { bus => 0, addr => 25 },
3888 vga3 => { bus => 0, addr => 26 },
3889 hostpci2 => { bus => 0, addr => 27 },
3890 hostpci3 => { bus => 0, addr => 28 },
3891 #addr29 : usb-host (pve-usb.cfg)
3892 'pci.1' => { bus => 0, addr => 30 },
3893 'pci.2' => { bus => 0, addr => 31 },
3894 'net6' => { bus => 1, addr => 1 },
3895 'net7' => { bus => 1, addr => 2 },
3896 'net8' => { bus => 1, addr => 3 },
3897 'net9' => { bus => 1, addr => 4 },
3898 'net10' => { bus => 1, addr => 5 },
3899 'net11' => { bus => 1, addr => 6 },
3900 'net12' => { bus => 1, addr => 7 },
3901 'net13' => { bus => 1, addr => 8 },
3902 'net14' => { bus => 1, addr => 9 },
3903 'net15' => { bus => 1, addr => 10 },
3904 'net16' => { bus => 1, addr => 11 },
3905 'net17' => { bus => 1, addr => 12 },
3906 'net18' => { bus => 1, addr => 13 },
3907 'net19' => { bus => 1, addr => 14 },
3908 'net20' => { bus => 1, addr => 15 },
3909 'net21' => { bus => 1, addr => 16 },
3910 'net22' => { bus => 1, addr => 17 },
3911 'net23' => { bus => 1, addr => 18 },
3912 'net24' => { bus => 1, addr => 19 },
3913 'net25' => { bus => 1, addr => 20 },
3914 'net26' => { bus => 1, addr => 21 },
3915 'net27' => { bus => 1, addr => 22 },
3916 'net28' => { bus => 1, addr => 23 },
3917 'net29' => { bus => 1, addr => 24 },
3918 'net30' => { bus => 1, addr => 25 },
3919 'net31' => { bus => 1, addr => 26 },
3920 'virtio6' => { bus => 2, addr => 1 },
3921 'virtio7' => { bus => 2, addr => 2 },
3922 'virtio8' => { bus => 2, addr => 3 },
3923 'virtio9' => { bus => 2, addr => 4 },
3924 'virtio10' => { bus => 2, addr => 5 },
3925 'virtio11' => { bus => 2, addr => 6 },
3926 'virtio12' => { bus => 2, addr => 7 },
3927 'virtio13' => { bus => 2, addr => 8 },
3928 'virtio14' => { bus => 2, addr => 9 },
3929 'virtio15' => { bus => 2, addr => 10 },
3930 };
3931
3932 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
3933 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
3934 my $bus = $devices->{$id}->{bus};
3935 $res = ",bus=pci.$bus,addr=$addr";
3936 $bridges->{$bus} = 1 if $bridges;
3937 }
3938 return $res;
3939
3940 }
3941
3942 sub print_pcie_addr {
3943 my ($id) = @_;
3944
3945 my $res = '';
3946 my $devices = {
3947 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
3948 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
3949 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
3950 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
3951 };
3952
3953 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
3954 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
3955 my $bus = $devices->{$id}->{bus};
3956 $res = ",bus=$bus,addr=$addr";
3957 }
3958 return $res;
3959
3960 }
3961
3962 # vzdump restore implementaion
3963
3964 sub tar_archive_read_firstfile {
3965 my $archive = shift;
3966
3967 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
3968
3969 # try to detect archive type first
3970 my $pid = open (TMP, "tar tf '$archive'|") ||
3971 die "unable to open file '$archive'\n";
3972 my $firstfile = <TMP>;
3973 kill 15, $pid;
3974 close TMP;
3975
3976 die "ERROR: archive contaions no data\n" if !$firstfile;
3977 chomp $firstfile;
3978
3979 return $firstfile;
3980 }
3981
3982 sub tar_restore_cleanup {
3983 my ($storecfg, $statfile) = @_;
3984
3985 print STDERR "starting cleanup\n";
3986
3987 if (my $fd = IO::File->new($statfile, "r")) {
3988 while (defined(my $line = <$fd>)) {
3989 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
3990 my $volid = $2;
3991 eval {
3992 if ($volid =~ m|^/|) {
3993 unlink $volid || die 'unlink failed\n';
3994 } else {
3995 PVE::Storage::vdisk_free($storecfg, $volid);
3996 }
3997 print STDERR "temporary volume '$volid' sucessfuly removed\n";
3998 };
3999 print STDERR "unable to cleanup '$volid' - $@" if $@;
4000 } else {
4001 print STDERR "unable to parse line in statfile - $line";
4002 }
4003 }
4004 $fd->close();
4005 }
4006 }
4007
4008 sub restore_archive {
4009 my ($archive, $vmid, $user, $opts) = @_;
4010
4011 my $format = $opts->{format};
4012 my $comp;
4013
4014 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4015 $format = 'tar' if !$format;
4016 $comp = 'gzip';
4017 } elsif ($archive =~ m/\.tar$/) {
4018 $format = 'tar' if !$format;
4019 } elsif ($archive =~ m/.tar.lzo$/) {
4020 $format = 'tar' if !$format;
4021 $comp = 'lzop';
4022 } elsif ($archive =~ m/\.vma$/) {
4023 $format = 'vma' if !$format;
4024 } elsif ($archive =~ m/\.vma\.gz$/) {
4025 $format = 'vma' if !$format;
4026 $comp = 'gzip';
4027 } elsif ($archive =~ m/\.vma\.lzo$/) {
4028 $format = 'vma' if !$format;
4029 $comp = 'lzop';
4030 } else {
4031 $format = 'vma' if !$format; # default
4032 }
4033
4034 # try to detect archive format
4035 if ($format eq 'tar') {
4036 return restore_tar_archive($archive, $vmid, $user, $opts);
4037 } else {
4038 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4039 }
4040 }
4041
4042 sub restore_update_config_line {
4043 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4044
4045 return if $line =~ m/^\#qmdump\#/;
4046 return if $line =~ m/^\#vzdump\#/;
4047 return if $line =~ m/^lock:/;
4048 return if $line =~ m/^unused\d+:/;
4049 return if $line =~ m/^parent:/;
4050 return if $line =~ m/^template:/; # restored VM is never a template
4051
4052 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4053 # try to convert old 1.X settings
4054 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4055 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4056 my ($model, $macaddr) = split(/\=/, $devconfig);
4057 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4058 my $net = {
4059 model => $model,
4060 bridge => "vmbr$ind",
4061 macaddr => $macaddr,
4062 };
4063 my $netstr = print_net($net);
4064
4065 print $outfd "net$cookie->{netcount}: $netstr\n";
4066 $cookie->{netcount}++;
4067 }
4068 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4069 my ($id, $netstr) = ($1, $2);
4070 my $net = parse_net($netstr);
4071 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4072 $netstr = print_net($net);
4073 print $outfd "$id: $netstr\n";
4074 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4075 my $virtdev = $1;
4076 my $value = $3;
4077 if ($line =~ m/backup=no/) {
4078 print $outfd "#$line";
4079 } elsif ($virtdev && $map->{$virtdev}) {
4080 my $di = parse_drive($virtdev, $value);
4081 delete $di->{format}; # format can change on restore
4082 $di->{file} = $map->{$virtdev};
4083 $value = print_drive($vmid, $di);
4084 print $outfd "$virtdev: $value\n";
4085 } else {
4086 print $outfd $line;
4087 }
4088 } else {
4089 print $outfd $line;
4090 }
4091 }
4092
4093 sub scan_volids {
4094 my ($cfg, $vmid) = @_;
4095
4096 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4097
4098 my $volid_hash = {};
4099 foreach my $storeid (keys %$info) {
4100 foreach my $item (@{$info->{$storeid}}) {
4101 next if !($item->{volid} && $item->{size});
4102 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4103 $volid_hash->{$item->{volid}} = $item;
4104 }
4105 }
4106
4107 return $volid_hash;
4108 }
4109
4110 sub get_used_paths {
4111 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4112
4113 my $used_path = {};
4114
4115 my $scan_config = sub {
4116 my ($cref, $snapname) = @_;
4117
4118 foreach my $key (keys %$cref) {
4119 my $value = $cref->{$key};
4120 if (valid_drivename($key)) {
4121 next if $skip_drive && $key eq $skip_drive;
4122 my $drive = parse_drive($key, $value);
4123 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4124 if ($drive->{file} =~ m!^/!) {
4125 $used_path->{$drive->{file}}++; # = 1;
4126 } else {
4127 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4128 next if !$storeid;
4129 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4130 next if !$scfg;
4131 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4132 $used_path->{$path}++; # = 1;
4133 }
4134 }
4135 }
4136 };
4137
4138 &$scan_config($conf);
4139
4140 undef $skip_drive;
4141
4142 if ($scan_snapshots) {
4143 foreach my $snapname (keys %{$conf->{snapshots}}) {
4144 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4145 }
4146 }
4147
4148 return $used_path;
4149 }
4150
4151 sub update_disksize {
4152 my ($vmid, $conf, $volid_hash) = @_;
4153
4154 my $changes;
4155
4156 my $used = {};
4157
4158 # Note: it is allowed to define multiple storages with same path (alias), so
4159 # we need to check both 'volid' and real 'path' (two different volid can point
4160 # to the same path).
4161
4162 my $usedpath = {};
4163
4164 # update size info
4165 foreach my $opt (keys %$conf) {
4166 if (valid_drivename($opt)) {
4167 my $drive = parse_drive($opt, $conf->{$opt});
4168 my $volid = $drive->{file};
4169 next if !$volid;
4170
4171 $used->{$volid} = 1;
4172 if ($volid_hash->{$volid} &&
4173 (my $path = $volid_hash->{$volid}->{path})) {
4174 $usedpath->{$path} = 1;
4175 }
4176
4177 next if drive_is_cdrom($drive);
4178 next if !$volid_hash->{$volid};
4179
4180 $drive->{size} = $volid_hash->{$volid}->{size};
4181 my $new = print_drive($vmid, $drive);
4182 if ($new ne $conf->{$opt}) {
4183 $changes = 1;
4184 $conf->{$opt} = $new;
4185 }
4186 }
4187 }
4188
4189 # remove 'unusedX' entry if volume is used
4190 foreach my $opt (keys %$conf) {
4191 next if $opt !~ m/^unused\d+$/;
4192 my $volid = $conf->{$opt};
4193 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4194 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4195 $changes = 1;
4196 delete $conf->{$opt};
4197 }
4198 }
4199
4200 foreach my $volid (sort keys %$volid_hash) {
4201 next if $volid =~ m/vm-$vmid-state-/;
4202 next if $used->{$volid};
4203 my $path = $volid_hash->{$volid}->{path};
4204 next if !$path; # just to be sure
4205 next if $usedpath->{$path};
4206 $changes = 1;
4207 add_unused_volume($conf, $volid);
4208 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4209 }
4210
4211 return $changes;
4212 }
4213
4214 sub rescan {
4215 my ($vmid, $nolock) = @_;
4216
4217 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4218
4219 my $volid_hash = scan_volids($cfg, $vmid);
4220
4221 my $updatefn = sub {
4222 my ($vmid) = @_;
4223
4224 my $conf = load_config($vmid);
4225
4226 check_lock($conf);
4227
4228 my $vm_volids = {};
4229 foreach my $volid (keys %$volid_hash) {
4230 my $info = $volid_hash->{$volid};
4231 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
4232 }
4233
4234 my $changes = update_disksize($vmid, $conf, $vm_volids);
4235
4236 update_config_nolock($vmid, $conf, 1) if $changes;
4237 };
4238
4239 if (defined($vmid)) {
4240 if ($nolock) {
4241 &$updatefn($vmid);
4242 } else {
4243 lock_config($vmid, $updatefn, $vmid);
4244 }
4245 } else {
4246 my $vmlist = config_list();
4247 foreach my $vmid (keys %$vmlist) {
4248 if ($nolock) {
4249 &$updatefn($vmid);
4250 } else {
4251 lock_config($vmid, $updatefn, $vmid);
4252 }
4253 }
4254 }
4255 }
4256
4257 sub restore_vma_archive {
4258 my ($archive, $vmid, $user, $opts, $comp) = @_;
4259
4260 my $input = $archive eq '-' ? "<&STDIN" : undef;
4261 my $readfrom = $archive;
4262
4263 my $uncomp = '';
4264 if ($comp) {
4265 $readfrom = '-';
4266 my $qarchive = PVE::Tools::shellquote($archive);
4267 if ($comp eq 'gzip') {
4268 $uncomp = "zcat $qarchive|";
4269 } elsif ($comp eq 'lzop') {
4270 $uncomp = "lzop -d -c $qarchive|";
4271 } else {
4272 die "unknown compression method '$comp'\n";
4273 }
4274
4275 }
4276
4277 my $tmpdir = "/var/tmp/vzdumptmp$$";
4278 rmtree $tmpdir;
4279
4280 # disable interrupts (always do cleanups)
4281 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
4282 warn "got interrupt - ignored\n";
4283 };
4284
4285 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
4286 POSIX::mkfifo($mapfifo, 0600);
4287 my $fifofh;
4288
4289 my $openfifo = sub {
4290 open($fifofh, '>', $mapfifo) || die $!;
4291 };
4292
4293 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
4294
4295 my $oldtimeout;
4296 my $timeout = 5;
4297
4298 my $devinfo = {};
4299
4300 my $rpcenv = PVE::RPCEnvironment::get();
4301
4302 my $conffile = config_file($vmid);
4303 my $tmpfn = "$conffile.$$.tmp";
4304
4305 # Note: $oldconf is undef if VM does not exists
4306 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
4307
4308 my $print_devmap = sub {
4309 my $virtdev_hash = {};
4310
4311 my $cfgfn = "$tmpdir/qemu-server.conf";
4312
4313 # we can read the config - that is already extracted
4314 my $fh = IO::File->new($cfgfn, "r") ||
4315 "unable to read qemu-server.conf - $!\n";
4316
4317 while (defined(my $line = <$fh>)) {
4318 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
4319 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
4320 die "archive does not contain data for drive '$virtdev'\n"
4321 if !$devinfo->{$devname};
4322 if (defined($opts->{storage})) {
4323 $storeid = $opts->{storage} || 'local';
4324 } elsif (!$storeid) {
4325 $storeid = 'local';
4326 }
4327 $format = 'raw' if !$format;
4328 $devinfo->{$devname}->{devname} = $devname;
4329 $devinfo->{$devname}->{virtdev} = $virtdev;
4330 $devinfo->{$devname}->{format} = $format;
4331 $devinfo->{$devname}->{storeid} = $storeid;
4332
4333 # check permission on storage
4334 my $pool = $opts->{pool}; # todo: do we need that?
4335 if ($user ne 'root@pam') {
4336 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
4337 }
4338
4339 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
4340 }
4341 }
4342
4343 foreach my $devname (keys %$devinfo) {
4344 die "found no device mapping information for device '$devname'\n"
4345 if !$devinfo->{$devname}->{virtdev};
4346 }
4347
4348 my $cfg = cfs_read_file('storage.cfg');
4349
4350 # create empty/temp config
4351 if ($oldconf) {
4352 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
4353 foreach_drive($oldconf, sub {
4354 my ($ds, $drive) = @_;
4355
4356 return if drive_is_cdrom($drive);
4357
4358 my $volid = $drive->{file};
4359
4360 return if !$volid || $volid =~ m|^/|;
4361
4362 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
4363 return if !$path || !$owner || ($owner != $vmid);
4364
4365 # Note: only delete disk we want to restore
4366 # other volumes will become unused
4367 if ($virtdev_hash->{$ds}) {
4368 PVE::Storage::vdisk_free($cfg, $volid);
4369 }
4370 });
4371 }
4372
4373 my $map = {};
4374 foreach my $virtdev (sort keys %$virtdev_hash) {
4375 my $d = $virtdev_hash->{$virtdev};
4376 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
4377 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
4378
4379 # test if requested format is supported
4380 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
4381 my $supported = grep { $_ eq $d->{format} } @$validFormats;
4382 $d->{format} = $defFormat if !$supported;
4383
4384 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
4385 $d->{format}, undef, $alloc_size);
4386 print STDERR "new volume ID is '$volid'\n";
4387 $d->{volid} = $volid;
4388 my $path = PVE::Storage::path($cfg, $volid);
4389
4390 my $write_zeros = 1;
4391 # fixme: what other storages types initialize volumes with zero?
4392 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
4393 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
4394 $write_zeros = 0;
4395 }
4396
4397 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
4398
4399 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
4400 $map->{$virtdev} = $volid;
4401 }
4402
4403 $fh->seek(0, 0) || die "seek failed - $!\n";
4404
4405 my $outfd = new IO::File ($tmpfn, "w") ||
4406 die "unable to write config for VM $vmid\n";
4407
4408 my $cookie = { netcount => 0 };
4409 while (defined(my $line = <$fh>)) {
4410 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
4411 }
4412
4413 $fh->close();
4414 $outfd->close();
4415 };
4416
4417 eval {
4418 # enable interrupts
4419 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
4420 die "interrupted by signal\n";
4421 };
4422 local $SIG{ALRM} = sub { die "got timeout\n"; };
4423
4424 $oldtimeout = alarm($timeout);
4425
4426 my $parser = sub {
4427 my $line = shift;
4428
4429 print "$line\n";
4430
4431 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
4432 my ($dev_id, $size, $devname) = ($1, $2, $3);
4433 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
4434 } elsif ($line =~ m/^CTIME: /) {
4435 # we correctly received the vma config, so we can disable
4436 # the timeout now for disk allocation (set to 10 minutes, so
4437 # that we always timeout if something goes wrong)
4438 alarm(600);
4439 &$print_devmap();
4440 print $fifofh "done\n";
4441 my $tmp = $oldtimeout || 0;
4442 $oldtimeout = undef;
4443 alarm($tmp);
4444 close($fifofh);
4445 }
4446 };
4447
4448 print "restore vma archive: $cmd\n";
4449 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
4450 };
4451 my $err = $@;
4452
4453 alarm($oldtimeout) if $oldtimeout;
4454
4455 unlink $mapfifo;
4456
4457 if ($err) {
4458 rmtree $tmpdir;
4459 unlink $tmpfn;
4460
4461 my $cfg = cfs_read_file('storage.cfg');
4462 foreach my $devname (keys %$devinfo) {
4463 my $volid = $devinfo->{$devname}->{volid};
4464 next if !$volid;
4465 eval {
4466 if ($volid =~ m|^/|) {
4467 unlink $volid || die 'unlink failed\n';
4468 } else {
4469 PVE::Storage::vdisk_free($cfg, $volid);
4470 }
4471 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4472 };
4473 print STDERR "unable to cleanup '$volid' - $@" if $@;
4474 }
4475 die $err;
4476 }
4477
4478 rmtree $tmpdir;
4479
4480 rename($tmpfn, $conffile) ||
4481 die "unable to commit configuration file '$conffile'\n";
4482
4483 PVE::Cluster::cfs_update(); # make sure we read new file
4484
4485 eval { rescan($vmid, 1); };
4486 warn $@ if $@;
4487 }
4488
4489 sub restore_tar_archive {
4490 my ($archive, $vmid, $user, $opts) = @_;
4491
4492 if ($archive ne '-') {
4493 my $firstfile = tar_archive_read_firstfile($archive);
4494 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
4495 if $firstfile ne 'qemu-server.conf';
4496 }
4497
4498 my $storecfg = cfs_read_file('storage.cfg');
4499
4500 # destroy existing data - keep empty config
4501 my $vmcfgfn = PVE::QemuServer::config_file($vmid);
4502 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
4503
4504 my $tocmd = "/usr/lib/qemu-server/qmextract";
4505
4506 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
4507 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
4508 $tocmd .= ' --prealloc' if $opts->{prealloc};
4509 $tocmd .= ' --info' if $opts->{info};
4510
4511 # tar option "xf" does not autodetect compression when read from STDIN,
4512 # so we pipe to zcat
4513 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
4514 PVE::Tools::shellquote("--to-command=$tocmd");
4515
4516 my $tmpdir = "/var/tmp/vzdumptmp$$";
4517 mkpath $tmpdir;
4518
4519 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
4520 local $ENV{VZDUMP_VMID} = $vmid;
4521 local $ENV{VZDUMP_USER} = $user;
4522
4523 my $conffile = config_file($vmid);
4524 my $tmpfn = "$conffile.$$.tmp";
4525
4526 # disable interrupts (always do cleanups)
4527 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
4528 print STDERR "got interrupt - ignored\n";
4529 };
4530
4531 eval {
4532 # enable interrupts
4533 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
4534 die "interrupted by signal\n";
4535 };
4536
4537 if ($archive eq '-') {
4538 print "extracting archive from STDIN\n";
4539 run_command($cmd, input => "<&STDIN");
4540 } else {
4541 print "extracting archive '$archive'\n";
4542 run_command($cmd);
4543 }
4544
4545 return if $opts->{info};
4546
4547 # read new mapping
4548 my $map = {};
4549 my $statfile = "$tmpdir/qmrestore.stat";
4550 if (my $fd = IO::File->new($statfile, "r")) {
4551 while (defined (my $line = <$fd>)) {
4552 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4553 $map->{$1} = $2 if $1;
4554 } else {
4555 print STDERR "unable to parse line in statfile - $line\n";
4556 }
4557 }
4558 $fd->close();
4559 }
4560
4561 my $confsrc = "$tmpdir/qemu-server.conf";
4562
4563 my $srcfd = new IO::File($confsrc, "r") ||
4564 die "unable to open file '$confsrc'\n";
4565
4566 my $outfd = new IO::File ($tmpfn, "w") ||
4567 die "unable to write config for VM $vmid\n";
4568
4569 my $cookie = { netcount => 0 };
4570 while (defined (my $line = <$srcfd>)) {
4571 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
4572 }
4573
4574 $srcfd->close();
4575 $outfd->close();
4576 };
4577 my $err = $@;
4578
4579 if ($err) {
4580
4581 unlink $tmpfn;
4582
4583 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
4584
4585 die $err;
4586 }
4587
4588 rmtree $tmpdir;
4589
4590 rename $tmpfn, $conffile ||
4591 die "unable to commit configuration file '$conffile'\n";
4592
4593 PVE::Cluster::cfs_update(); # make sure we read new file
4594
4595 eval { rescan($vmid, 1); };
4596 warn $@ if $@;
4597 };
4598
4599
4600 # Internal snapshots
4601
4602 # NOTE: Snapshot create/delete involves several non-atomic
4603 # action, and can take a long time.
4604 # So we try to avoid locking the file and use 'lock' variable
4605 # inside the config file instead.
4606
4607 my $snapshot_copy_config = sub {
4608 my ($source, $dest) = @_;
4609
4610 foreach my $k (keys %$source) {
4611 next if $k eq 'snapshots';
4612 next if $k eq 'snapstate';
4613 next if $k eq 'snaptime';
4614 next if $k eq 'vmstate';
4615 next if $k eq 'lock';
4616 next if $k eq 'digest';
4617 next if $k eq 'description';
4618 next if $k =~ m/^unused\d+$/;
4619
4620 $dest->{$k} = $source->{$k};
4621 }
4622 };
4623
4624 my $snapshot_apply_config = sub {
4625 my ($conf, $snap) = @_;
4626
4627 # copy snapshot list
4628 my $newconf = {
4629 snapshots => $conf->{snapshots},
4630 };
4631
4632 # keep description and list of unused disks
4633 foreach my $k (keys %$conf) {
4634 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
4635 $newconf->{$k} = $conf->{$k};
4636 }
4637
4638 &$snapshot_copy_config($snap, $newconf);
4639
4640 return $newconf;
4641 };
4642
4643 sub foreach_writable_storage {
4644 my ($conf, $func) = @_;
4645
4646 my $sidhash = {};
4647
4648 foreach my $ds (keys %$conf) {
4649 next if !valid_drivename($ds);
4650
4651 my $drive = parse_drive($ds, $conf->{$ds});
4652 next if !$drive;
4653 next if drive_is_cdrom($drive);
4654
4655 my $volid = $drive->{file};
4656
4657 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4658 $sidhash->{$sid} = $sid if $sid;
4659 }
4660
4661 foreach my $sid (sort keys %$sidhash) {
4662 &$func($sid);
4663 }
4664 }
4665
4666 my $alloc_vmstate_volid = sub {
4667 my ($storecfg, $vmid, $conf, $snapname) = @_;
4668
4669 # Note: we try to be smart when selecting a $target storage
4670
4671 my $target;
4672
4673 # search shared storage first
4674 foreach_writable_storage($conf, sub {
4675 my ($sid) = @_;
4676 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
4677 return if !$scfg->{shared};
4678
4679 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
4680 });
4681
4682 if (!$target) {
4683 # now search local storage
4684 foreach_writable_storage($conf, sub {
4685 my ($sid) = @_;
4686 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
4687 return if $scfg->{shared};
4688
4689 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
4690 });
4691 }
4692
4693 $target = 'local' if !$target;
4694
4695 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
4696 # we abort live save after $conf->{memory}, so we need at max twice that space
4697 my $size = $conf->{memory}*2 + $driver_state_size;
4698
4699 my $name = "vm-$vmid-state-$snapname";
4700 my $scfg = PVE::Storage::storage_config($storecfg, $target);
4701 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
4702 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
4703
4704 return $volid;
4705 };
4706
4707 my $snapshot_prepare = sub {
4708 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
4709
4710 my $snap;
4711
4712 my $updatefn = sub {
4713
4714 my $conf = load_config($vmid);
4715
4716 die "you can't take a snapshot if it's a template\n"
4717 if is_template($conf);
4718
4719 check_lock($conf);
4720
4721 $conf->{lock} = 'snapshot';
4722
4723 die "snapshot name '$snapname' already used\n"
4724 if defined($conf->{snapshots}->{$snapname});
4725
4726 my $storecfg = PVE::Storage::config();
4727 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
4728
4729 $snap = $conf->{snapshots}->{$snapname} = {};
4730
4731 if ($save_vmstate && check_running($vmid)) {
4732 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
4733 }
4734
4735 &$snapshot_copy_config($conf, $snap);
4736
4737 $snap->{snapstate} = "prepare";
4738 $snap->{snaptime} = time();
4739 $snap->{description} = $comment if $comment;
4740
4741 # always overwrite machine if we save vmstate. This makes sure we
4742 # can restore it later using correct machine type
4743 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
4744
4745 update_config_nolock($vmid, $conf, 1);
4746 };
4747
4748 lock_config($vmid, $updatefn);
4749
4750 return $snap;
4751 };
4752
4753 my $snapshot_commit = sub {
4754 my ($vmid, $snapname) = @_;
4755
4756 my $updatefn = sub {
4757
4758 my $conf = load_config($vmid);
4759
4760 die "missing snapshot lock\n"
4761 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
4762
4763 my $has_machine_config = defined($conf->{machine});
4764
4765 my $snap = $conf->{snapshots}->{$snapname};
4766
4767 die "snapshot '$snapname' does not exist\n" if !defined($snap);
4768
4769 die "wrong snapshot state\n"
4770 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
4771
4772 delete $snap->{snapstate};
4773 delete $conf->{lock};
4774
4775 my $newconf = &$snapshot_apply_config($conf, $snap);
4776
4777 delete $newconf->{machine} if !$has_machine_config;
4778
4779 $newconf->{parent} = $snapname;
4780
4781 update_config_nolock($vmid, $newconf, 1);
4782 };
4783
4784 lock_config($vmid, $updatefn);
4785 };
4786
4787 sub snapshot_rollback {
4788 my ($vmid, $snapname) = @_;
4789
4790 my $snap;
4791
4792 my $prepare = 1;
4793
4794 my $storecfg = PVE::Storage::config();
4795
4796 my $updatefn = sub {
4797
4798 my $conf = load_config($vmid);
4799
4800 die "you can't rollback if vm is a template\n" if is_template($conf);
4801
4802 $snap = $conf->{snapshots}->{$snapname};
4803
4804 die "snapshot '$snapname' does not exist\n" if !defined($snap);
4805
4806 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
4807 if $snap->{snapstate};
4808
4809 if ($prepare) {
4810 check_lock($conf);
4811 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
4812 }
4813
4814 die "unable to rollback vm $vmid: vm is running\n"
4815 if check_running($vmid);
4816
4817 if ($prepare) {
4818 $conf->{lock} = 'rollback';
4819 } else {
4820 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
4821 delete $conf->{lock};
4822 }
4823
4824 my $forcemachine;
4825
4826 if (!$prepare) {
4827 my $has_machine_config = defined($conf->{machine});
4828
4829 # copy snapshot config to current config
4830 $conf = &$snapshot_apply_config($conf, $snap);
4831 $conf->{parent} = $snapname;
4832
4833 # Note: old code did not store 'machine', so we try to be smart
4834 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
4835 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
4836 # we remove the 'machine' configuration if not explicitly specified
4837 # in the original config.
4838 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
4839 }
4840
4841 update_config_nolock($vmid, $conf, 1);
4842
4843 if (!$prepare && $snap->{vmstate}) {
4844 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
4845 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
4846 }
4847 };
4848
4849 lock_config($vmid, $updatefn);
4850
4851 foreach_drive($snap, sub {
4852 my ($ds, $drive) = @_;
4853
4854 return if drive_is_cdrom($drive);
4855
4856 my $volid = $drive->{file};
4857 my $device = "drive-$ds";
4858
4859 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
4860 });
4861
4862 $prepare = 0;
4863 lock_config($vmid, $updatefn);
4864 }
4865
4866 my $savevm_wait = sub {
4867 my ($vmid) = @_;
4868
4869 for(;;) {
4870 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
4871 if (!$stat->{status}) {
4872 die "savevm not active\n";
4873 } elsif ($stat->{status} eq 'active') {
4874 sleep(1);
4875 next;
4876 } elsif ($stat->{status} eq 'completed') {
4877 last;
4878 } else {
4879 die "query-savevm returned status '$stat->{status}'\n";
4880 }
4881 }
4882 };
4883
4884 sub snapshot_create {
4885 my ($vmid, $snapname, $save_vmstate, $freezefs, $comment) = @_;
4886
4887 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
4888
4889 $freezefs = $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
4890
4891 my $drivehash = {};
4892
4893 my $running = check_running($vmid);
4894
4895 eval {
4896 # create internal snapshots of all drives
4897
4898 my $storecfg = PVE::Storage::config();
4899
4900 if ($running) {
4901 if ($snap->{vmstate}) {
4902 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
4903 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
4904 &$savevm_wait($vmid);
4905 } else {
4906 vm_mon_cmd($vmid, "savevm-start");
4907 }
4908 };
4909
4910 qga_freezefs($vmid) if $running && $freezefs;
4911
4912 foreach_drive($snap, sub {
4913 my ($ds, $drive) = @_;
4914
4915 return if drive_is_cdrom($drive);
4916
4917 my $volid = $drive->{file};
4918 my $device = "drive-$ds";
4919
4920 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
4921 $drivehash->{$ds} = 1;
4922 });
4923 };
4924 my $err = $@;
4925
4926 eval { qga_unfreezefs($vmid) if $running && $freezefs; };
4927 warn $@ if $@;
4928
4929 eval { vm_mon_cmd($vmid, "savevm-end") if $running; };
4930 warn $@ if $@;
4931
4932 if ($err) {
4933 warn "snapshot create failed: starting cleanup\n";
4934 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
4935 warn $@ if $@;
4936 die $err;
4937 }
4938
4939 &$snapshot_commit($vmid, $snapname);
4940 }
4941
4942 # Note: $drivehash is only set when called from snapshot_create.
4943 sub snapshot_delete {
4944 my ($vmid, $snapname, $force, $drivehash) = @_;
4945
4946 my $prepare = 1;
4947
4948 my $snap;
4949 my $unused = [];
4950
4951 my $unlink_parent = sub {
4952 my ($confref, $new_parent) = @_;
4953
4954 if ($confref->{parent} && $confref->{parent} eq $snapname) {
4955 if ($new_parent) {
4956 $confref->{parent} = $new_parent;
4957 } else {
4958 delete $confref->{parent};
4959 }
4960 }
4961 };
4962
4963 my $updatefn = sub {
4964 my ($remove_drive) = @_;
4965
4966 my $conf = load_config($vmid);
4967
4968 if (!$drivehash) {
4969 check_lock($conf);
4970 die "you can't delete a snapshot if vm is a template\n"
4971 if is_template($conf);
4972 }
4973
4974 $snap = $conf->{snapshots}->{$snapname};
4975
4976 die "snapshot '$snapname' does not exist\n" if !defined($snap);
4977
4978 # remove parent refs
4979 if (!$prepare) {
4980 &$unlink_parent($conf, $snap->{parent});
4981 foreach my $sn (keys %{$conf->{snapshots}}) {
4982 next if $sn eq $snapname;
4983 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
4984 }
4985 }
4986
4987 if ($remove_drive) {
4988 if ($remove_drive eq 'vmstate') {
4989 delete $snap->{$remove_drive};
4990 } else {
4991 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
4992 my $volid = $drive->{file};
4993 delete $snap->{$remove_drive};
4994 add_unused_volume($conf, $volid);
4995 }
4996 }
4997
4998 if ($prepare) {
4999 $snap->{snapstate} = 'delete';
5000 } else {
5001 delete $conf->{snapshots}->{$snapname};
5002 delete $conf->{lock} if $drivehash;
5003 foreach my $volid (@$unused) {
5004 add_unused_volume($conf, $volid);
5005 }
5006 }
5007
5008 update_config_nolock($vmid, $conf, 1);
5009 };
5010
5011 lock_config($vmid, $updatefn);
5012
5013 # now remove vmstate file
5014
5015 my $storecfg = PVE::Storage::config();
5016
5017 if ($snap->{vmstate}) {
5018 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5019 if (my $err = $@) {
5020 die $err if !$force;
5021 warn $err;
5022 }
5023 # save changes (remove vmstate from snapshot)
5024 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5025 };
5026
5027 # now remove all internal snapshots
5028 foreach_drive($snap, sub {
5029 my ($ds, $drive) = @_;
5030
5031 return if drive_is_cdrom($drive);
5032
5033 my $volid = $drive->{file};
5034 my $device = "drive-$ds";
5035
5036 if (!$drivehash || $drivehash->{$ds}) {
5037 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5038 if (my $err = $@) {
5039 die $err if !$force;
5040 warn $err;
5041 }
5042 }
5043
5044 # save changes (remove drive fron snapshot)
5045 lock_config($vmid, $updatefn, $ds) if !$force;
5046 push @$unused, $volid;
5047 });
5048
5049 # now cleanup config
5050 $prepare = 0;
5051 lock_config($vmid, $updatefn);
5052 }
5053
5054 sub has_feature {
5055 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5056
5057 my $err;
5058 foreach_drive($conf, sub {
5059 my ($ds, $drive) = @_;
5060
5061 return if drive_is_cdrom($drive);
5062 my $volid = $drive->{file};
5063 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5064 });
5065
5066 return $err ? 0 : 1;
5067 }
5068
5069 sub template_create {
5070 my ($vmid, $conf, $disk) = @_;
5071
5072 my $storecfg = PVE::Storage::config();
5073
5074 foreach_drive($conf, sub {
5075 my ($ds, $drive) = @_;
5076
5077 return if drive_is_cdrom($drive);
5078 return if $disk && $ds ne $disk;
5079
5080 my $volid = $drive->{file};
5081 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5082
5083 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5084 $drive->{file} = $voliddst;
5085 $conf->{$ds} = print_drive($vmid, $drive);
5086 update_config_nolock($vmid, $conf, 1);
5087 });
5088 }
5089
5090 sub is_template {
5091 my ($conf) = @_;
5092
5093 return 1 if defined $conf->{template} && $conf->{template} == 1;
5094 }
5095
5096 sub qemu_img_convert {
5097 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5098
5099 my $storecfg = PVE::Storage::config();
5100 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5101 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5102
5103 if ($src_storeid && $dst_storeid) {
5104 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5105 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5106
5107 my $src_format = qemu_img_format($src_scfg, $src_volname);
5108 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5109
5110 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5111 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5112
5113 my $cmd = [];
5114 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5115 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5116 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5117
5118 my $parser = sub {
5119 my $line = shift;
5120 if($line =~ m/\((\S+)\/100\%\)/){
5121 my $percent = $1;
5122 my $transferred = int($size * $percent / 100);
5123 my $remaining = $size - $transferred;
5124
5125 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5126 }
5127
5128 };
5129
5130 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5131 my $err = $@;
5132 die "copy failed: $err" if $err;
5133 }
5134 }
5135
5136 sub qemu_img_format {
5137 my ($scfg, $volname) = @_;
5138
5139 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5140 return $1;
5141 } elsif ($scfg->{type} eq 'iscsi') {
5142 return "host_device";
5143 } else {
5144 return "raw";
5145 }
5146 }
5147
5148 sub qemu_drive_mirror {
5149 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5150
5151 my $count = 0;
5152 my $old_len = 0;
5153 my $frozen = undef;
5154 my $maxwait = 120;
5155
5156 my $storecfg = PVE::Storage::config();
5157 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5158
5159 if ($dst_storeid) {
5160 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5161
5162 my $format;
5163 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5164 $format = $1;
5165 }
5166
5167 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5168
5169 if ($format) {
5170 #fixme : sometime drive-mirror timeout, but works fine after.
5171 # (I have see the problem with big volume > 200GB), so we need to eval
5172 eval { vm_mon_cmd($vmid, "drive-mirror", timeout => 10, device => "drive-$drive", mode => "existing",
5173 sync => "full", target => $dst_path, format => $format); };
5174 } else {
5175 eval { vm_mon_cmd($vmid, "drive-mirror", timeout => 10, device => "drive-$drive", mode => "existing",
5176 sync => "full", target => $dst_path); };
5177 }
5178
5179 eval {
5180 while (1) {
5181 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5182 my $stat = @$stats[0];
5183 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5184 die "error job is not mirroring" if $stat->{type} ne "mirror";
5185
5186 my $transferred = $stat->{offset};
5187 my $total = $stat->{len};
5188 my $remaining = $total - $transferred;
5189 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5190 my $busy = $stat->{busy};
5191
5192 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
5193
5194 if ($stat->{len} == $stat->{offset}) {
5195 last if $busy eq 'false';
5196 if ($count > $maxwait) {
5197 # if too much writes to disk occurs at the end of migration
5198 #the disk needs to be freezed to be able to complete the migration
5199 vm_suspend($vmid,1);
5200 $frozen = 1;
5201 }
5202 $count ++
5203 }
5204 $old_len = $stat->{offset};
5205 sleep 1;
5206 }
5207
5208 if ($vmiddst == $vmid) {
5209 # switch the disk if source and destination are on the same guest
5210 vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive");
5211 }
5212
5213 vm_resume($vmid, 1) if $frozen;
5214
5215 };
5216 if (my $err = $@) {
5217 eval { vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive"); };
5218 while (1) {
5219 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5220 my $stat = @$stats[0];
5221 last if !$stat;
5222 sleep 1;
5223 }
5224 die "mirroring error: $err";
5225 }
5226
5227 if ($vmiddst != $vmid) {
5228 # if we clone a disk for a new target vm, we don't switch the disk
5229 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
5230 while (1) {
5231 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5232 my $stat = @$stats[0];
5233 last if !$stat;
5234 sleep 1;
5235 }
5236 }
5237 }
5238 }
5239
5240 sub clone_disk {
5241 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
5242 $newvmid, $storage, $format, $full, $newvollist) = @_;
5243
5244 my $newvolid;
5245
5246 if (!$full) {
5247 print "create linked clone of drive $drivename ($drive->{file})\n";
5248 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
5249 push @$newvollist, $newvolid;
5250 } else {
5251 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5252 $storeid = $storage if $storage;
5253
5254 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5255 if (!$format) {
5256 $format = $drive->{format} || $defFormat;
5257 }
5258
5259 # test if requested format is supported - else use default
5260 my $supported = grep { $_ eq $format } @$validFormats;
5261 $format = $defFormat if !$supported;
5262
5263 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
5264
5265 print "create full clone of drive $drivename ($drive->{file})\n";
5266 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
5267 push @$newvollist, $newvolid;
5268
5269 if (!$running || $snapname) {
5270 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
5271 } else {
5272 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
5273 }
5274 }
5275
5276 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
5277
5278 my $disk = $drive;
5279 $disk->{format} = undef;
5280 $disk->{file} = $newvolid;
5281 $disk->{size} = $size;
5282
5283 return $disk;
5284 }
5285
5286 # this only works if VM is running
5287 sub get_current_qemu_machine {
5288 my ($vmid) = @_;
5289
5290 my $cmd = { execute => 'query-machines', arguments => {} };
5291 my $res = PVE::QemuServer::vm_qmp_command($vmid, $cmd);
5292
5293 my ($current, $default);
5294 foreach my $e (@$res) {
5295 $default = $e->{name} if $e->{'is-default'};
5296 $current = $e->{name} if $e->{'is-current'};
5297 }
5298
5299 # fallback to the default machine if current is not supported by qemu
5300 return $current || $default || 'pc';
5301 }
5302
5303 sub lspci {
5304
5305 my $devices = {};
5306
5307 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
5308 my (undef, $id, $function) = @_;
5309 my $res = { id => $id, function => $function};
5310 push @{$devices->{$id}}, $res;
5311 });
5312
5313 return $devices;
5314 }
5315
5316 1;