]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
hv_spinlocks : use 0x1fff instead 0xffff
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'string', format => 'pve-hotplug-features',
176 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
177 default => 'network,disk,usb',
178 },
179 reboot => {
180 optional => 1,
181 type => 'boolean',
182 description => "Allow reboot. If set to '0' the VM exit on reboot.",
183 default => 1,
184 },
185 lock => {
186 optional => 1,
187 type => 'string',
188 description => "Lock/unlock the VM.",
189 enum => [qw(migrate backup snapshot rollback)],
190 },
191 cpulimit => {
192 optional => 1,
193 type => 'integer',
194 description => "Limit of CPU usage in per cent. Note if the computer has 2 CPUs, it has total of 200% CPU time. Value '0' indicates no CPU limit.\n\nNOTE: This option is currently ignored.",
195 minimum => 0,
196 default => 0,
197 },
198 cpuunits => {
199 optional => 1,
200 type => 'integer',
201 description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.\n\nNOTE: You can disable fair-scheduler configuration by setting this to 0.",
202 minimum => 0,
203 maximum => 500000,
204 default => 1000,
205 },
206 memory => {
207 optional => 1,
208 type => 'integer',
209 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
210 minimum => 16,
211 default => 512,
212 },
213 balloon => {
214 optional => 1,
215 type => 'integer',
216 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
217 minimum => 0,
218 },
219 shares => {
220 optional => 1,
221 type => 'integer',
222 description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning",
223 minimum => 0,
224 maximum => 50000,
225 default => 1000,
226 },
227 keyboard => {
228 optional => 1,
229 type => 'string',
230 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
231 enum => PVE::Tools::kvmkeymaplist(),
232 default => 'en-us',
233 },
234 name => {
235 optional => 1,
236 type => 'string', format => 'dns-name',
237 description => "Set a name for the VM. Only used on the configuration web interface.",
238 },
239 scsihw => {
240 optional => 1,
241 type => 'string',
242 description => "scsi controller model",
243 enum => [qw(lsi lsi53c810 virtio-scsi-pci megasas pvscsi)],
244 default => 'lsi',
245 },
246 description => {
247 optional => 1,
248 type => 'string',
249 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
250 },
251 ostype => {
252 optional => 1,
253 type => 'string',
254 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
255 description => <<EODESC,
256 Used to enable special optimization/features for specific
257 operating systems:
258
259 other => unspecified OS
260 wxp => Microsoft Windows XP
261 w2k => Microsoft Windows 2000
262 w2k3 => Microsoft Windows 2003
263 w2k8 => Microsoft Windows 2008
264 wvista => Microsoft Windows Vista
265 win7 => Microsoft Windows 7
266 win8 => Microsoft Windows 8/2012
267 l24 => Linux 2.4 Kernel
268 l26 => Linux 2.6/3.X Kernel
269 solaris => solaris/opensolaris/openindiania kernel
270
271 other|l24|l26|solaris ... no special behaviour
272 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
273 EODESC
274 },
275 boot => {
276 optional => 1,
277 type => 'string',
278 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
279 pattern => '[acdn]{1,4}',
280 default => 'cdn',
281 },
282 bootdisk => {
283 optional => 1,
284 type => 'string', format => 'pve-qm-bootdisk',
285 description => "Enable booting from specified disk.",
286 pattern => '(ide|sata|scsi|virtio)\d+',
287 },
288 smp => {
289 optional => 1,
290 type => 'integer',
291 description => "The number of CPUs. Please use option -sockets instead.",
292 minimum => 1,
293 default => 1,
294 },
295 sockets => {
296 optional => 1,
297 type => 'integer',
298 description => "The number of CPU sockets.",
299 minimum => 1,
300 default => 1,
301 },
302 cores => {
303 optional => 1,
304 type => 'integer',
305 description => "The number of cores per socket.",
306 minimum => 1,
307 default => 1,
308 },
309 numa => {
310 optional => 1,
311 type => 'boolean',
312 description => "Enable/disable Numa.",
313 default => 0,
314 },
315 vcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Number of hotplugged vcpus.",
319 minimum => 1,
320 default => 0,
321 },
322 acpi => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable ACPI.",
326 default => 1,
327 },
328 agent => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable Qemu GuestAgent.",
332 default => 0,
333 },
334 kvm => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable KVM hardware virtualization.",
338 default => 1,
339 },
340 tdf => {
341 optional => 1,
342 type => 'boolean',
343 description => "Enable/disable time drift fix.",
344 default => 0,
345 },
346 localtime => {
347 optional => 1,
348 type => 'boolean',
349 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
350 },
351 freeze => {
352 optional => 1,
353 type => 'boolean',
354 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
355 },
356 vga => {
357 optional => 1,
358 type => 'string',
359 description => "Select VGA type. If you want to use high resolution modes (>= 1280x1024x16) then you should use option 'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and 'cirrur' for other OS types. Option 'qxl' enables the SPICE display sever. You can also run without any graphic card using a serial devive as terminal.",
360 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
361 },
362 watchdog => {
363 optional => 1,
364 type => 'string', format => 'pve-qm-watchdog',
365 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
366 description => "Create a virtual hardware watchdog device. Once enabled (by a guest action), the watchdog must be periodically polled by an agent inside the guest or else the guest will be restarted (or execute the action specified)",
367 },
368 startdate => {
369 optional => 1,
370 type => 'string',
371 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
372 description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
373 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
374 default => 'now',
375 },
376 startup => {
377 optional => 1,
378 type => 'string', format => 'pve-qm-startup',
379 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
380 description => "Startup and shutdown behavior. Order is a non-negative number defining the general startup order. Shutdown in done with reverse ordering. Additionally you can set the 'up' or 'down' delay in seconds, which specifies a delay to wait before the next VM is started or stopped.",
381 },
382 template => {
383 optional => 1,
384 type => 'boolean',
385 description => "Enable/disable Template.",
386 default => 0,
387 },
388 args => {
389 optional => 1,
390 type => 'string',
391 description => <<EODESCR,
392 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
393
394 args: -no-reboot -no-hpet
395 EODESCR
396 },
397 tablet => {
398 optional => 1,
399 type => 'boolean',
400 default => 1,
401 description => "Enable/disable the usb tablet device. This device is usually needed to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with normal VNC clients. If you're running lots of console-only guests on one host, you may consider disabling this to save some context switches. This is turned of by default if you use spice (vga=qxl).",
402 },
403 migrate_speed => {
404 optional => 1,
405 type => 'integer',
406 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
407 minimum => 0,
408 default => 0,
409 },
410 migrate_downtime => {
411 optional => 1,
412 type => 'number',
413 description => "Set maximum tolerated downtime (in seconds) for migrations.",
414 minimum => 0,
415 default => 0.1,
416 },
417 cdrom => {
418 optional => 1,
419 type => 'string', format => 'pve-qm-drive',
420 typetext => 'volume',
421 description => "This is an alias for option -ide2",
422 },
423 cpu => {
424 optional => 1,
425 description => "Emulated CPU type.",
426 type => 'string',
427 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
428 default => 'kvm64',
429 },
430 parent => get_standard_option('pve-snapshot-name', {
431 optional => 1,
432 description => "Parent snapshot name. This is used internally, and should not be modified.",
433 }),
434 snaptime => {
435 optional => 1,
436 description => "Timestamp for snapshots.",
437 type => 'integer',
438 minimum => 0,
439 },
440 vmstate => {
441 optional => 1,
442 type => 'string', format => 'pve-volume-id',
443 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
444 },
445 machine => {
446 description => "Specific the Qemu machine type.",
447 type => 'string',
448 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
449 maxLength => 40,
450 optional => 1,
451 },
452 smbios1 => {
453 description => "Specify SMBIOS type 1 fields.",
454 type => 'string', format => 'pve-qm-smbios1',
455 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
456 maxLength => 256,
457 optional => 1,
458 },
459 };
460
461 # what about other qemu settings ?
462 #cpu => 'string',
463 #machine => 'string',
464 #fda => 'file',
465 #fdb => 'file',
466 #mtdblock => 'file',
467 #sd => 'file',
468 #pflash => 'file',
469 #snapshot => 'bool',
470 #bootp => 'file',
471 ##tftp => 'dir',
472 ##smb => 'dir',
473 #kernel => 'file',
474 #append => 'string',
475 #initrd => 'file',
476 ##soundhw => 'string',
477
478 while (my ($k, $v) = each %$confdesc) {
479 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
480 }
481
482 my $MAX_IDE_DISKS = 4;
483 my $MAX_SCSI_DISKS = 14;
484 my $MAX_VIRTIO_DISKS = 16;
485 my $MAX_SATA_DISKS = 6;
486 my $MAX_USB_DEVICES = 5;
487 my $MAX_NETS = 32;
488 my $MAX_UNUSED_DISKS = 8;
489 my $MAX_HOSTPCI_DEVICES = 4;
490 my $MAX_SERIAL_PORTS = 4;
491 my $MAX_PARALLEL_PORTS = 3;
492 my $MAX_NUMA = 8;
493 my $MAX_MEM = 4194304;
494 my $STATICMEM = 1024;
495
496 my $numadesc = {
497 optional => 1,
498 type => 'string', format => 'pve-qm-numanode',
499 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
500 description => "numa topology",
501 };
502 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
503
504 for (my $i = 0; $i < $MAX_NUMA; $i++) {
505 $confdesc->{"numa$i"} = $numadesc;
506 }
507
508 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
509 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
510 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
511 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
512
513 my $netdesc = {
514 optional => 1,
515 type => 'string', format => 'pve-qm-net',
516 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
517 description => <<EODESCR,
518 Specify network devices.
519
520 MODEL is one of: $nic_model_list_txt
521
522 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
523 automatically generated if not specified.
524
525 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
526
527 Option 'rate' is used to limit traffic bandwidth from and to this interface. It is specified as floating point number, unit is 'Megabytes per second'.
528
529 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
530
531 10.0.2.2 Gateway
532 10.0.2.3 DNS Server
533 10.0.2.4 SMB Server
534
535 The DHCP server assign addresses to the guest starting from 10.0.2.15.
536
537 EODESCR
538 };
539 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
540
541 for (my $i = 0; $i < $MAX_NETS; $i++) {
542 $confdesc->{"net$i"} = $netdesc;
543 }
544
545 my $drivename_hash;
546
547 my $idedesc = {
548 optional => 1,
549 type => 'string', format => 'pve-qm-drive',
550 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
551 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
552 };
553 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
554
555 my $scsidesc = {
556 optional => 1,
557 type => 'string', format => 'pve-qm-drive',
558 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
559 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
560 };
561 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
562
563 my $satadesc = {
564 optional => 1,
565 type => 'string', format => 'pve-qm-drive',
566 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
567 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
568 };
569 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
570
571 my $virtiodesc = {
572 optional => 1,
573 type => 'string', format => 'pve-qm-drive',
574 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
575 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
576 };
577 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
578
579 my $usbdesc = {
580 optional => 1,
581 type => 'string', format => 'pve-qm-usb-device',
582 typetext => 'host=HOSTUSBDEVICE|spice',
583 description => <<EODESCR,
584 Configure an USB device (n is 0 to 4). This can be used to
585 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
586
587 'bus-port(.port)*' (decimal numbers) or
588 'vendor_id:product_id' (hexadeciaml numbers)
589
590 You can use the 'lsusb -t' command to list existing usb devices.
591
592 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
593
594 The value 'spice' can be used to add a usb redirection devices for spice.
595
596 EODESCR
597 };
598 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
599
600 my $hostpcidesc = {
601 optional => 1,
602 type => 'string', format => 'pve-qm-hostpci',
603 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
604 description => <<EODESCR,
605 Map host pci devices. HOSTPCIDEVICE syntax is:
606
607 'bus:dev.func' (hexadecimal numbers)
608
609 You can us the 'lspci' command to list existing pci devices.
610
611 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
612
613 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
614
615 Experimental: user reported problems with this option.
616 EODESCR
617 };
618 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
619
620 my $serialdesc = {
621 optional => 1,
622 type => 'string',
623 pattern => '(/dev/.+|socket)',
624 description => <<EODESCR,
625 Create a serial device inside the VM (n is 0 to 3), and pass through a host serial device (i.e. /dev/ttyS0), or create a unix socket on the host side (use 'qm terminal' to open a terminal connection).
626
627 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
628
629 Experimental: user reported problems with this option.
630 EODESCR
631 };
632
633 my $paralleldesc= {
634 optional => 1,
635 type => 'string',
636 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
637 description => <<EODESCR,
638 Map host parallel devices (n is 0 to 2).
639
640 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
641
642 Experimental: user reported problems with this option.
643 EODESCR
644 };
645
646 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
647 $confdesc->{"parallel$i"} = $paralleldesc;
648 }
649
650 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
651 $confdesc->{"serial$i"} = $serialdesc;
652 }
653
654 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
655 $confdesc->{"hostpci$i"} = $hostpcidesc;
656 }
657
658 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
659 $drivename_hash->{"ide$i"} = 1;
660 $confdesc->{"ide$i"} = $idedesc;
661 }
662
663 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
664 $drivename_hash->{"sata$i"} = 1;
665 $confdesc->{"sata$i"} = $satadesc;
666 }
667
668 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
669 $drivename_hash->{"scsi$i"} = 1;
670 $confdesc->{"scsi$i"} = $scsidesc ;
671 }
672
673 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
674 $drivename_hash->{"virtio$i"} = 1;
675 $confdesc->{"virtio$i"} = $virtiodesc;
676 }
677
678 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
679 $confdesc->{"usb$i"} = $usbdesc;
680 }
681
682 my $unuseddesc = {
683 optional => 1,
684 type => 'string', format => 'pve-volume-id',
685 description => "Reference to unused volumes.",
686 };
687
688 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
689 $confdesc->{"unused$i"} = $unuseddesc;
690 }
691
692 my $kvm_api_version = 0;
693
694 sub kvm_version {
695
696 return $kvm_api_version if $kvm_api_version;
697
698 my $fh = IO::File->new("</dev/kvm") ||
699 return 0;
700
701 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
702 $kvm_api_version = $v;
703 }
704
705 $fh->close();
706
707 return $kvm_api_version;
708 }
709
710 my $kvm_user_version;
711
712 sub kvm_user_version {
713
714 return $kvm_user_version if $kvm_user_version;
715
716 $kvm_user_version = 'unknown';
717
718 my $tmp = `kvm -help 2>/dev/null`;
719
720 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
721 $kvm_user_version = $2;
722 }
723
724 return $kvm_user_version;
725
726 }
727
728 my $kernel_has_vhost_net = -c '/dev/vhost-net';
729
730 sub disknames {
731 # order is important - used to autoselect boot disk
732 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
733 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
734 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
735 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
736 }
737
738 sub valid_drivename {
739 my $dev = shift;
740
741 return defined($drivename_hash->{$dev});
742 }
743
744 sub option_exists {
745 my $key = shift;
746 return defined($confdesc->{$key});
747 }
748
749 sub nic_models {
750 return $nic_model_list;
751 }
752
753 sub os_list_description {
754
755 return {
756 other => 'Other',
757 wxp => 'Windows XP',
758 w2k => 'Windows 2000',
759 w2k3 =>, 'Windows 2003',
760 w2k8 => 'Windows 2008',
761 wvista => 'Windows Vista',
762 win7 => 'Windows 7',
763 win8 => 'Windows 8/2012',
764 l24 => 'Linux 2.4',
765 l26 => 'Linux 2.6',
766 };
767 }
768
769 my $cdrom_path;
770
771 sub get_cdrom_path {
772
773 return $cdrom_path if $cdrom_path;
774
775 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
776 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
777 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
778 }
779
780 sub get_iso_path {
781 my ($storecfg, $vmid, $cdrom) = @_;
782
783 if ($cdrom eq 'cdrom') {
784 return get_cdrom_path();
785 } elsif ($cdrom eq 'none') {
786 return '';
787 } elsif ($cdrom =~ m|^/|) {
788 return $cdrom;
789 } else {
790 return PVE::Storage::path($storecfg, $cdrom);
791 }
792 }
793
794 # try to convert old style file names to volume IDs
795 sub filename_to_volume_id {
796 my ($vmid, $file, $media) = @_;
797
798 if (!($file eq 'none' || $file eq 'cdrom' ||
799 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
800
801 return undef if $file =~ m|/|;
802
803 if ($media && $media eq 'cdrom') {
804 $file = "local:iso/$file";
805 } else {
806 $file = "local:$vmid/$file";
807 }
808 }
809
810 return $file;
811 }
812
813 sub verify_media_type {
814 my ($opt, $vtype, $media) = @_;
815
816 return if !$media;
817
818 my $etype;
819 if ($media eq 'disk') {
820 $etype = 'images';
821 } elsif ($media eq 'cdrom') {
822 $etype = 'iso';
823 } else {
824 die "internal error";
825 }
826
827 return if ($vtype eq $etype);
828
829 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
830 }
831
832 sub cleanup_drive_path {
833 my ($opt, $storecfg, $drive) = @_;
834
835 # try to convert filesystem paths to volume IDs
836
837 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
838 ($drive->{file} !~ m|^/dev/.+|) &&
839 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
840 ($drive->{file} !~ m/^\d+$/)) {
841 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
842 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
843 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
844 verify_media_type($opt, $vtype, $drive->{media});
845 $drive->{file} = $volid;
846 }
847
848 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
849 }
850
851 sub create_conf_nolock {
852 my ($vmid, $settings) = @_;
853
854 my $filename = config_file($vmid);
855
856 die "configuration file '$filename' already exists\n" if -f $filename;
857
858 my $defaults = load_defaults();
859
860 $settings->{name} = "vm$vmid" if !$settings->{name};
861 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
862
863 my $data = '';
864 foreach my $opt (keys %$settings) {
865 next if !$confdesc->{$opt};
866
867 my $value = $settings->{$opt};
868 next if !$value;
869
870 $data .= "$opt: $value\n";
871 }
872
873 PVE::Tools::file_set_contents($filename, $data);
874 }
875
876 sub parse_hotplug_features {
877 my ($data) = @_;
878
879 my $res = {};
880
881 return $res if $data eq '0';
882
883 $data = $confdesc->{hotplug}->{default} if $data eq '1';
884
885 foreach my $feature (PVE::Tools::split_list($data)) {
886 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
887 $res->{$1} = 1;
888 } else {
889 warn "ignoring unknown hotplug feature '$feature'\n";
890 }
891 }
892 return $res;
893 }
894
895 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
896 sub pve_verify_hotplug_features {
897 my ($value, $noerr) = @_;
898
899 return $value if parse_hotplug_features($value);
900
901 return undef if $noerr;
902
903 die "unable to parse hotplug option\n";
904 }
905
906 my $parse_size = sub {
907 my ($value) = @_;
908
909 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
910 my ($size, $unit) = ($1, $3);
911 if ($unit) {
912 if ($unit eq 'K') {
913 $size = $size * 1024;
914 } elsif ($unit eq 'M') {
915 $size = $size * 1024 * 1024;
916 } elsif ($unit eq 'G') {
917 $size = $size * 1024 * 1024 * 1024;
918 }
919 }
920 return int($size);
921 };
922
923 my $format_size = sub {
924 my ($size) = @_;
925
926 $size = int($size);
927
928 my $kb = int($size/1024);
929 return $size if $kb*1024 != $size;
930
931 my $mb = int($kb/1024);
932 return "${kb}K" if $mb*1024 != $kb;
933
934 my $gb = int($mb/1024);
935 return "${mb}M" if $gb*1024 != $mb;
936
937 return "${gb}G";
938 };
939
940 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
941 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
942 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
943 # [,aio=native|threads][,discard=ignore|on]
944
945 sub parse_drive {
946 my ($key, $data) = @_;
947
948 my $res = {};
949
950 # $key may be undefined - used to verify JSON parameters
951 if (!defined($key)) {
952 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
953 $res->{index} = 0;
954 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
955 $res->{interface} = $1;
956 $res->{index} = $2;
957 } else {
958 return undef;
959 }
960
961 foreach my $p (split (/,/, $data)) {
962 next if $p =~ m/^\s*$/;
963
964 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
965 my ($k, $v) = ($1, $2);
966
967 $k = 'file' if $k eq 'volume';
968
969 return undef if defined $res->{$k};
970
971 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
972 return undef if !$v || $v !~ m/^\d+/;
973 $k = "m$k";
974 $v = sprintf("%.3f", $v / (1024*1024));
975 }
976 $res->{$k} = $v;
977 } else {
978 if (!$res->{file} && $p !~ m/=/) {
979 $res->{file} = $p;
980 } else {
981 return undef;
982 }
983 }
984 }
985
986 return undef if !$res->{file};
987
988 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
989 $res->{format} = $1;
990 }
991
992 return undef if $res->{cache} &&
993 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
994 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
995 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
996 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
997 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
998 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
999 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
1000 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
1001 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
1002 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
1003 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
1004 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
1005 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
1006
1007 return undef if $res->{mbps_rd} && $res->{mbps};
1008 return undef if $res->{mbps_wr} && $res->{mbps};
1009
1010 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
1011 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
1012 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
1013 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
1014 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
1015 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
1016
1017 return undef if $res->{iops_rd} && $res->{iops};
1018 return undef if $res->{iops_wr} && $res->{iops};
1019
1020
1021 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
1022 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
1023 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
1024 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
1025 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
1026 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
1027
1028
1029 if ($res->{size}) {
1030 return undef if !defined($res->{size} = &$parse_size($res->{size}));
1031 }
1032
1033 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1034 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1035 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1036 return undef if $res->{interface} eq 'virtio';
1037 }
1038
1039 # rerror does not work with scsi drives
1040 if ($res->{rerror}) {
1041 return undef if $res->{interface} eq 'scsi';
1042 }
1043
1044 return $res;
1045 }
1046
1047 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1048
1049 sub print_drive {
1050 my ($vmid, $drive) = @_;
1051
1052 my $opts = '';
1053 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1054 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1055 }
1056
1057 if ($drive->{size}) {
1058 $opts .= ",size=" . &$format_size($drive->{size});
1059 }
1060
1061 return "$drive->{file}$opts";
1062 }
1063
1064 sub scsi_inquiry {
1065 my($fh, $noerr) = @_;
1066
1067 my $SG_IO = 0x2285;
1068 my $SG_GET_VERSION_NUM = 0x2282;
1069
1070 my $versionbuf = "\x00" x 8;
1071 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1072 if (!$ret) {
1073 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1074 return undef;
1075 }
1076 my $version = unpack("I", $versionbuf);
1077 if ($version < 30000) {
1078 die "scsi generic interface too old\n" if !$noerr;
1079 return undef;
1080 }
1081
1082 my $buf = "\x00" x 36;
1083 my $sensebuf = "\x00" x 8;
1084 my $cmd = pack("C x3 C x1", 0x12, 36);
1085
1086 # see /usr/include/scsi/sg.h
1087 my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1088
1089 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1090 length($sensebuf), 0, length($buf), $buf,
1091 $cmd, $sensebuf, 6000);
1092
1093 $ret = ioctl($fh, $SG_IO, $packet);
1094 if (!$ret) {
1095 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1096 return undef;
1097 }
1098
1099 my @res = unpack($sg_io_hdr_t, $packet);
1100 if ($res[17] || $res[18]) {
1101 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1102 return undef;
1103 }
1104
1105 my $res = {};
1106 (my $byte0, my $byte1, $res->{vendor},
1107 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1108
1109 $res->{removable} = $byte1 & 128 ? 1 : 0;
1110 $res->{type} = $byte0 & 31;
1111
1112 return $res;
1113 }
1114
1115 sub path_is_scsi {
1116 my ($path) = @_;
1117
1118 my $fh = IO::File->new("+<$path") || return undef;
1119 my $res = scsi_inquiry($fh, 1);
1120 close($fh);
1121
1122 return $res;
1123 }
1124
1125 sub machine_type_is_q35 {
1126 my ($conf) = @_;
1127
1128 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1129 }
1130
1131 sub print_tabletdevice_full {
1132 my ($conf) = @_;
1133
1134 my $q35 = machine_type_is_q35($conf);
1135
1136 # we use uhci for old VMs because tablet driver was buggy in older qemu
1137 my $usbbus = $q35 ? "ehci" : "uhci";
1138
1139 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1140 }
1141
1142 sub print_drivedevice_full {
1143 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1144
1145 my $device = '';
1146 my $maxdev = 0;
1147
1148 if ($drive->{interface} eq 'virtio') {
1149 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1150 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1151 $device .= ",iothread=iothread0" if $conf->{iothread};
1152 } elsif ($drive->{interface} eq 'scsi') {
1153 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1154 my $controller = int($drive->{index} / $maxdev);
1155 my $unit = $drive->{index} % $maxdev;
1156 my $devicetype = 'hd';
1157 my $path = '';
1158 if (drive_is_cdrom($drive)) {
1159 $devicetype = 'cd';
1160 } else {
1161 if ($drive->{file} =~ m|^/|) {
1162 $path = $drive->{file};
1163 } else {
1164 $path = PVE::Storage::path($storecfg, $drive->{file});
1165 }
1166
1167 if($path =~ m/^iscsi\:\/\//){
1168 $devicetype = 'generic';
1169 } else {
1170 if (my $info = path_is_scsi($path)) {
1171 if ($info->{type} == 0) {
1172 $devicetype = 'block';
1173 } elsif ($info->{type} == 1) { # tape
1174 $devicetype = 'generic';
1175 }
1176 }
1177 }
1178 }
1179
1180 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1181 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1182 } else {
1183 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1184 }
1185
1186 } elsif ($drive->{interface} eq 'ide'){
1187 $maxdev = 2;
1188 my $controller = int($drive->{index} / $maxdev);
1189 my $unit = $drive->{index} % $maxdev;
1190 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1191
1192 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1193 } elsif ($drive->{interface} eq 'sata'){
1194 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1195 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1196 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1197 } elsif ($drive->{interface} eq 'usb') {
1198 die "implement me";
1199 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1200 } else {
1201 die "unsupported interface type";
1202 }
1203
1204 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1205
1206 return $device;
1207 }
1208
1209 sub get_initiator_name {
1210 my $initiator;
1211
1212 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1213 while (defined(my $line = <$fh>)) {
1214 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1215 $initiator = $1;
1216 last;
1217 }
1218 $fh->close();
1219
1220 return $initiator;
1221 }
1222
1223 sub print_drive_full {
1224 my ($storecfg, $vmid, $drive) = @_;
1225
1226 my $opts = '';
1227 foreach my $o (@qemu_drive_options) {
1228 next if $o eq 'bootindex';
1229 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1230 }
1231
1232 foreach my $o (qw(bps bps_rd bps_wr)) {
1233 my $v = $drive->{"m$o"};
1234 $opts .= ",$o=" . int($v*1024*1024) if $v;
1235 }
1236
1237 # use linux-aio by default (qemu default is threads)
1238 $opts .= ",aio=native" if !$drive->{aio};
1239
1240 my $path;
1241 my $volid = $drive->{file};
1242 if (drive_is_cdrom($drive)) {
1243 $path = get_iso_path($storecfg, $vmid, $volid);
1244 } else {
1245 if ($volid =~ m|^/|) {
1246 $path = $volid;
1247 } else {
1248 $path = PVE::Storage::path($storecfg, $volid);
1249 }
1250 }
1251
1252 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1253
1254 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1255 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1256
1257 my $pathinfo = $path ? "file=$path," : '';
1258
1259 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1260 }
1261
1262 sub print_netdevice_full {
1263 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1264
1265 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1266
1267 my $device = $net->{model};
1268 if ($net->{model} eq 'virtio') {
1269 $device = 'virtio-net-pci';
1270 };
1271
1272 my $pciaddr = print_pci_addr("$netid", $bridges);
1273 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1274 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1275 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1276 my $vectors = $net->{queues} * 2 + 2;
1277 $tmpstr .= ",vectors=$vectors,mq=on";
1278 }
1279 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1280 return $tmpstr;
1281 }
1282
1283 sub print_netdev_full {
1284 my ($vmid, $conf, $net, $netid) = @_;
1285
1286 my $i = '';
1287 if ($netid =~ m/^net(\d+)$/) {
1288 $i = int($1);
1289 }
1290
1291 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1292
1293 my $ifname = "tap${vmid}i$i";
1294
1295 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1296 die "interface name '$ifname' is too long (max 15 character)\n"
1297 if length($ifname) >= 16;
1298
1299 my $vhostparam = '';
1300 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1301
1302 my $vmname = $conf->{name} || "vm$vmid";
1303
1304 my $netdev = "";
1305
1306 if ($net->{bridge}) {
1307 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1308 } else {
1309 $netdev = "type=user,id=$netid,hostname=$vmname";
1310 }
1311
1312 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1313
1314 return $netdev;
1315 }
1316
1317 sub drive_is_cdrom {
1318 my ($drive) = @_;
1319
1320 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1321
1322 }
1323
1324 sub parse_numa {
1325 my ($data) = @_;
1326
1327 my $res = {};
1328
1329 foreach my $kvp (split(/,/, $data)) {
1330
1331 if ($kvp =~ m/^memory=(\S+)$/) {
1332 $res->{memory} = $1;
1333 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1334 $res->{policy} = $1;
1335 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1336 $res->{cpus}->{start} = $1;
1337 $res->{cpus}->{end} = $3;
1338 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1339 $res->{hostnodes}->{start} = $1;
1340 $res->{hostnodes}->{end} = $3;
1341 } else {
1342 return undef;
1343 }
1344 }
1345
1346 return $res;
1347 }
1348
1349 sub parse_hostpci {
1350 my ($value) = @_;
1351
1352 return undef if !$value;
1353
1354
1355 my @list = split(/,/, $value);
1356 my $found;
1357
1358 my $res = {};
1359 foreach my $kv (@list) {
1360
1361 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1362 $found = 1;
1363 if(defined($4)){
1364 push @{$res->{pciid}}, { id => $2 , function => $4};
1365
1366 }else{
1367 my $pcidevices = lspci($2);
1368 $res->{pciid} = $pcidevices->{$2};
1369 }
1370 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1371 $res->{driver} = $1;
1372 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1373 $res->{rombar} = $1;
1374 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1375 $res->{'x-vga'} = $1;
1376 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1377 $res->{pcie} = 1 if $1 == 1;
1378 } else {
1379 warn "unknown hostpci setting '$kv'\n";
1380 }
1381 }
1382
1383 return undef if !$found;
1384
1385 return $res;
1386 }
1387
1388 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1389 sub parse_net {
1390 my ($data) = @_;
1391
1392 my $res = {};
1393
1394 foreach my $kvp (split(/,/, $data)) {
1395
1396 if ($kvp =~ m/^(ne2k_pci|e1000|e1000-82540em|e1000-82544gc|e1000-82545em|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1397 my $model = lc($1);
1398 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1399 $res->{model} = $model;
1400 $res->{macaddr} = $mac;
1401 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1402 $res->{bridge} = $1;
1403 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1404 $res->{queues} = $1;
1405 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1406 $res->{rate} = $1;
1407 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1408 $res->{tag} = $1;
1409 } elsif ($kvp =~ m/^firewall=([01])$/) {
1410 $res->{firewall} = $1;
1411 } elsif ($kvp =~ m/^link_down=([01])$/) {
1412 $res->{link_down} = $1;
1413 } else {
1414 return undef;
1415 }
1416
1417 }
1418
1419 return undef if !$res->{model};
1420
1421 return $res;
1422 }
1423
1424 sub print_net {
1425 my $net = shift;
1426
1427 my $res = "$net->{model}";
1428 $res .= "=$net->{macaddr}" if $net->{macaddr};
1429 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1430 $res .= ",rate=$net->{rate}" if $net->{rate};
1431 $res .= ",tag=$net->{tag}" if $net->{tag};
1432 $res .= ",firewall=1" if $net->{firewall};
1433 $res .= ",link_down=1" if $net->{link_down};
1434 $res .= ",queues=$net->{queues}" if $net->{queues};
1435
1436 return $res;
1437 }
1438
1439 sub add_random_macs {
1440 my ($settings) = @_;
1441
1442 foreach my $opt (keys %$settings) {
1443 next if $opt !~ m/^net(\d+)$/;
1444 my $net = parse_net($settings->{$opt});
1445 next if !$net;
1446 $settings->{$opt} = print_net($net);
1447 }
1448 }
1449
1450 sub add_unused_volume {
1451 my ($config, $volid) = @_;
1452
1453 my $key;
1454 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1455 my $test = "unused$ind";
1456 if (my $vid = $config->{$test}) {
1457 return if $vid eq $volid; # do not add duplicates
1458 } else {
1459 $key = $test;
1460 }
1461 }
1462
1463 die "To many unused volume - please delete them first.\n" if !$key;
1464
1465 $config->{$key} = $volid;
1466
1467 return $key;
1468 }
1469
1470 sub vm_is_volid_owner {
1471 my ($storecfg, $vmid, $volid) = @_;
1472
1473 if ($volid !~ m|^/|) {
1474 my ($path, $owner);
1475 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1476 if ($owner && ($owner == $vmid)) {
1477 return 1;
1478 }
1479 }
1480
1481 return undef;
1482 }
1483
1484 sub vmconfig_delete_pending_option {
1485 my ($conf, $key) = @_;
1486
1487 delete $conf->{pending}->{$key};
1488 my $pending_delete_hash = { $key => 1 };
1489 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1490 $pending_delete_hash->{$opt} = 1;
1491 }
1492 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1493 }
1494
1495 sub vmconfig_undelete_pending_option {
1496 my ($conf, $key) = @_;
1497
1498 my $pending_delete_hash = {};
1499 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1500 $pending_delete_hash->{$opt} = 1;
1501 }
1502 delete $pending_delete_hash->{$key};
1503
1504 my @keylist = keys %$pending_delete_hash;
1505 if (scalar(@keylist)) {
1506 $conf->{pending}->{delete} = join(',', @keylist);
1507 } else {
1508 delete $conf->{pending}->{delete};
1509 }
1510 }
1511
1512 sub vmconfig_register_unused_drive {
1513 my ($storecfg, $vmid, $conf, $drive) = @_;
1514
1515 if (!drive_is_cdrom($drive)) {
1516 my $volid = $drive->{file};
1517 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1518 add_unused_volume($conf, $volid, $vmid);
1519 }
1520 }
1521 }
1522
1523 sub vmconfig_cleanup_pending {
1524 my ($conf) = @_;
1525
1526 # remove pending changes when nothing changed
1527 my $changes;
1528 foreach my $opt (keys %{$conf->{pending}}) {
1529 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1530 $changes = 1;
1531 delete $conf->{pending}->{$opt};
1532 }
1533 }
1534
1535 # remove delete if option is not set
1536 my $pending_delete_hash = {};
1537 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1538 if (defined($conf->{$opt})) {
1539 $pending_delete_hash->{$opt} = 1;
1540 } else {
1541 $changes = 1;
1542 }
1543 }
1544
1545 my @keylist = keys %$pending_delete_hash;
1546 if (scalar(@keylist)) {
1547 $conf->{pending}->{delete} = join(',', @keylist);
1548 } else {
1549 delete $conf->{pending}->{delete};
1550 }
1551
1552 return $changes;
1553 }
1554
1555 my $valid_smbios1_options = {
1556 manufacturer => '\S+',
1557 product => '\S+',
1558 version => '\S+',
1559 serial => '\S+',
1560 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1561 sku => '\S+',
1562 family => '\S+',
1563 };
1564
1565 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1566 sub parse_smbios1 {
1567 my ($data) = @_;
1568
1569 my $res = {};
1570
1571 foreach my $kvp (split(/,/, $data)) {
1572 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1573 my ($k, $v) = split(/=/, $kvp);
1574 return undef if !defined($k) || !defined($v);
1575 return undef if !$valid_smbios1_options->{$k};
1576 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1577 $res->{$k} = $v;
1578 }
1579
1580 return $res;
1581 }
1582
1583 sub print_smbios1 {
1584 my ($smbios1) = @_;
1585
1586 my $data = '';
1587 foreach my $k (keys %$smbios1) {
1588 next if !defined($smbios1->{$k});
1589 next if !$valid_smbios1_options->{$k};
1590 $data .= ',' if $data;
1591 $data .= "$k=$smbios1->{$k}";
1592 }
1593 return $data;
1594 }
1595
1596 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1597 sub verify_smbios1 {
1598 my ($value, $noerr) = @_;
1599
1600 return $value if parse_smbios1($value);
1601
1602 return undef if $noerr;
1603
1604 die "unable to parse smbios (type 1) options\n";
1605 }
1606
1607 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1608 sub verify_bootdisk {
1609 my ($value, $noerr) = @_;
1610
1611 return $value if valid_drivename($value);
1612
1613 return undef if $noerr;
1614
1615 die "invalid boot disk '$value'\n";
1616 }
1617
1618 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1619 sub verify_numa {
1620 my ($value, $noerr) = @_;
1621
1622 return $value if parse_numa($value);
1623
1624 return undef if $noerr;
1625
1626 die "unable to parse numa options\n";
1627 }
1628
1629 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1630 sub verify_net {
1631 my ($value, $noerr) = @_;
1632
1633 return $value if parse_net($value);
1634
1635 return undef if $noerr;
1636
1637 die "unable to parse network options\n";
1638 }
1639
1640 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1641 sub verify_drive {
1642 my ($value, $noerr) = @_;
1643
1644 return $value if parse_drive(undef, $value);
1645
1646 return undef if $noerr;
1647
1648 die "unable to parse drive options\n";
1649 }
1650
1651 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1652 sub verify_hostpci {
1653 my ($value, $noerr) = @_;
1654
1655 return $value if parse_hostpci($value);
1656
1657 return undef if $noerr;
1658
1659 die "unable to parse pci id\n";
1660 }
1661
1662 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1663 sub verify_watchdog {
1664 my ($value, $noerr) = @_;
1665
1666 return $value if parse_watchdog($value);
1667
1668 return undef if $noerr;
1669
1670 die "unable to parse watchdog options\n";
1671 }
1672
1673 sub parse_watchdog {
1674 my ($value) = @_;
1675
1676 return undef if !$value;
1677
1678 my $res = {};
1679
1680 foreach my $p (split(/,/, $value)) {
1681 next if $p =~ m/^\s*$/;
1682
1683 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1684 $res->{model} = $2;
1685 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1686 $res->{action} = $2;
1687 } else {
1688 return undef;
1689 }
1690 }
1691
1692 return $res;
1693 }
1694
1695 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1696 sub verify_startup {
1697 my ($value, $noerr) = @_;
1698
1699 return $value if parse_startup($value);
1700
1701 return undef if $noerr;
1702
1703 die "unable to parse startup options\n";
1704 }
1705
1706 sub parse_startup {
1707 my ($value) = @_;
1708
1709 return undef if !$value;
1710
1711 my $res = {};
1712
1713 foreach my $p (split(/,/, $value)) {
1714 next if $p =~ m/^\s*$/;
1715
1716 if ($p =~ m/^(order=)?(\d+)$/) {
1717 $res->{order} = $2;
1718 } elsif ($p =~ m/^up=(\d+)$/) {
1719 $res->{up} = $1;
1720 } elsif ($p =~ m/^down=(\d+)$/) {
1721 $res->{down} = $1;
1722 } else {
1723 return undef;
1724 }
1725 }
1726
1727 return $res;
1728 }
1729
1730 sub parse_usb_device {
1731 my ($value) = @_;
1732
1733 return undef if !$value;
1734
1735 my @dl = split(/,/, $value);
1736 my $found;
1737
1738 my $res = {};
1739 foreach my $v (@dl) {
1740 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1741 $found = 1;
1742 $res->{vendorid} = $2;
1743 $res->{productid} = $4;
1744 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1745 $found = 1;
1746 $res->{hostbus} = $1;
1747 $res->{hostport} = $2;
1748 } elsif ($v =~ m/^spice$/) {
1749 $found = 1;
1750 $res->{spice} = 1;
1751 } else {
1752 return undef;
1753 }
1754 }
1755 return undef if !$found;
1756
1757 return $res;
1758 }
1759
1760 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1761 sub verify_usb_device {
1762 my ($value, $noerr) = @_;
1763
1764 return $value if parse_usb_device($value);
1765
1766 return undef if $noerr;
1767
1768 die "unable to parse usb device\n";
1769 }
1770
1771 # add JSON properties for create and set function
1772 sub json_config_properties {
1773 my $prop = shift;
1774
1775 foreach my $opt (keys %$confdesc) {
1776 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1777 $prop->{$opt} = $confdesc->{$opt};
1778 }
1779
1780 return $prop;
1781 }
1782
1783 sub check_type {
1784 my ($key, $value) = @_;
1785
1786 die "unknown setting '$key'\n" if !$confdesc->{$key};
1787
1788 my $type = $confdesc->{$key}->{type};
1789
1790 if (!defined($value)) {
1791 die "got undefined value\n";
1792 }
1793
1794 if ($value =~ m/[\n\r]/) {
1795 die "property contains a line feed\n";
1796 }
1797
1798 if ($type eq 'boolean') {
1799 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1800 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1801 die "type check ('boolean') failed - got '$value'\n";
1802 } elsif ($type eq 'integer') {
1803 return int($1) if $value =~ m/^(\d+)$/;
1804 die "type check ('integer') failed - got '$value'\n";
1805 } elsif ($type eq 'number') {
1806 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1807 die "type check ('number') failed - got '$value'\n";
1808 } elsif ($type eq 'string') {
1809 if (my $fmt = $confdesc->{$key}->{format}) {
1810 if ($fmt eq 'pve-qm-drive') {
1811 # special case - we need to pass $key to parse_drive()
1812 my $drive = parse_drive($key, $value);
1813 return $value if $drive;
1814 die "unable to parse drive options\n";
1815 }
1816 PVE::JSONSchema::check_format($fmt, $value);
1817 return $value;
1818 }
1819 $value =~ s/^\"(.*)\"$/$1/;
1820 return $value;
1821 } else {
1822 die "internal error"
1823 }
1824 }
1825
1826 sub lock_config_full {
1827 my ($vmid, $timeout, $code, @param) = @_;
1828
1829 my $filename = config_file_lock($vmid);
1830
1831 my $res = lock_file($filename, $timeout, $code, @param);
1832
1833 die $@ if $@;
1834
1835 return $res;
1836 }
1837
1838 sub lock_config_mode {
1839 my ($vmid, $timeout, $shared, $code, @param) = @_;
1840
1841 my $filename = config_file_lock($vmid);
1842
1843 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1844
1845 die $@ if $@;
1846
1847 return $res;
1848 }
1849
1850 sub lock_config {
1851 my ($vmid, $code, @param) = @_;
1852
1853 return lock_config_full($vmid, 10, $code, @param);
1854 }
1855
1856 sub cfs_config_path {
1857 my ($vmid, $node) = @_;
1858
1859 $node = $nodename if !$node;
1860 return "nodes/$node/qemu-server/$vmid.conf";
1861 }
1862
1863 sub check_iommu_support{
1864 #fixme : need to check IOMMU support
1865 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1866
1867 my $iommu=1;
1868 return $iommu;
1869
1870 }
1871
1872 sub config_file {
1873 my ($vmid, $node) = @_;
1874
1875 my $cfspath = cfs_config_path($vmid, $node);
1876 return "/etc/pve/$cfspath";
1877 }
1878
1879 sub config_file_lock {
1880 my ($vmid) = @_;
1881
1882 return "$lock_dir/lock-$vmid.conf";
1883 }
1884
1885 sub touch_config {
1886 my ($vmid) = @_;
1887
1888 my $conf = config_file($vmid);
1889 utime undef, undef, $conf;
1890 }
1891
1892 sub destroy_vm {
1893 my ($storecfg, $vmid, $keep_empty_config) = @_;
1894
1895 my $conffile = config_file($vmid);
1896
1897 my $conf = load_config($vmid);
1898
1899 check_lock($conf);
1900
1901 # only remove disks owned by this VM
1902 foreach_drive($conf, sub {
1903 my ($ds, $drive) = @_;
1904
1905 return if drive_is_cdrom($drive);
1906
1907 my $volid = $drive->{file};
1908
1909 return if !$volid || $volid =~ m|^/|;
1910
1911 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1912 return if !$path || !$owner || ($owner != $vmid);
1913
1914 PVE::Storage::vdisk_free($storecfg, $volid);
1915 });
1916
1917 if ($keep_empty_config) {
1918 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1919 } else {
1920 unlink $conffile;
1921 }
1922
1923 # also remove unused disk
1924 eval {
1925 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1926
1927 eval {
1928 PVE::Storage::foreach_volid($dl, sub {
1929 my ($volid, $sid, $volname, $d) = @_;
1930 PVE::Storage::vdisk_free($storecfg, $volid);
1931 });
1932 };
1933 warn $@ if $@;
1934
1935 };
1936 warn $@ if $@;
1937 }
1938
1939 sub load_config {
1940 my ($vmid, $node) = @_;
1941
1942 my $cfspath = cfs_config_path($vmid, $node);
1943
1944 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1945
1946 die "no such VM ('$vmid')\n" if !defined($conf);
1947
1948 return $conf;
1949 }
1950
1951 sub parse_vm_config {
1952 my ($filename, $raw) = @_;
1953
1954 return undef if !defined($raw);
1955
1956 my $res = {
1957 digest => Digest::SHA::sha1_hex($raw),
1958 snapshots => {},
1959 pending => {},
1960 };
1961
1962 $filename =~ m|/qemu-server/(\d+)\.conf$|
1963 || die "got strange filename '$filename'";
1964
1965 my $vmid = $1;
1966
1967 my $conf = $res;
1968 my $descr = '';
1969 my $section = '';
1970
1971 my @lines = split(/\n/, $raw);
1972 foreach my $line (@lines) {
1973 next if $line =~ m/^\s*$/;
1974
1975 if ($line =~ m/^\[PENDING\]\s*$/i) {
1976 $section = 'pending';
1977 $conf->{description} = $descr if $descr;
1978 $descr = '';
1979 $conf = $res->{$section} = {};
1980 next;
1981
1982 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1983 $section = $1;
1984 $conf->{description} = $descr if $descr;
1985 $descr = '';
1986 $conf = $res->{snapshots}->{$section} = {};
1987 next;
1988 }
1989
1990 if ($line =~ m/^\#(.*)\s*$/) {
1991 $descr .= PVE::Tools::decode_text($1) . "\n";
1992 next;
1993 }
1994
1995 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1996 $descr .= PVE::Tools::decode_text($2);
1997 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1998 $conf->{snapstate} = $1;
1999 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
2000 my $key = $1;
2001 my $value = $2;
2002 $conf->{$key} = $value;
2003 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2004 my $value = $1;
2005 if ($section eq 'pending') {
2006 $conf->{delete} = $value; # we parse this later
2007 } else {
2008 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2009 }
2010 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2011 my $key = $1;
2012 my $value = $2;
2013 eval { $value = check_type($key, $value); };
2014 if ($@) {
2015 warn "vm $vmid - unable to parse value of '$key' - $@";
2016 } else {
2017 my $fmt = $confdesc->{$key}->{format};
2018 if ($fmt && $fmt eq 'pve-qm-drive') {
2019 my $v = parse_drive($key, $value);
2020 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2021 $v->{file} = $volid;
2022 $value = print_drive($vmid, $v);
2023 } else {
2024 warn "vm $vmid - unable to parse value of '$key'\n";
2025 next;
2026 }
2027 }
2028
2029 if ($key eq 'cdrom') {
2030 $conf->{ide2} = $value;
2031 } else {
2032 $conf->{$key} = $value;
2033 }
2034 }
2035 }
2036 }
2037
2038 $conf->{description} = $descr if $descr;
2039
2040 delete $res->{snapstate}; # just to be sure
2041
2042 return $res;
2043 }
2044
2045 sub write_vm_config {
2046 my ($filename, $conf) = @_;
2047
2048 delete $conf->{snapstate}; # just to be sure
2049
2050 if ($conf->{cdrom}) {
2051 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2052 $conf->{ide2} = $conf->{cdrom};
2053 delete $conf->{cdrom};
2054 }
2055
2056 # we do not use 'smp' any longer
2057 if ($conf->{sockets}) {
2058 delete $conf->{smp};
2059 } elsif ($conf->{smp}) {
2060 $conf->{sockets} = $conf->{smp};
2061 delete $conf->{cores};
2062 delete $conf->{smp};
2063 }
2064
2065 my $used_volids = {};
2066
2067 my $cleanup_config = sub {
2068 my ($cref, $pending, $snapname) = @_;
2069
2070 foreach my $key (keys %$cref) {
2071 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2072 $key eq 'snapstate' || $key eq 'pending';
2073 my $value = $cref->{$key};
2074 if ($key eq 'delete') {
2075 die "propertry 'delete' is only allowed in [PENDING]\n"
2076 if !$pending;
2077 # fixme: check syntax?
2078 next;
2079 }
2080 eval { $value = check_type($key, $value); };
2081 die "unable to parse value of '$key' - $@" if $@;
2082
2083 $cref->{$key} = $value;
2084
2085 if (!$snapname && valid_drivename($key)) {
2086 my $drive = parse_drive($key, $value);
2087 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2088 }
2089 }
2090 };
2091
2092 &$cleanup_config($conf);
2093
2094 &$cleanup_config($conf->{pending}, 1);
2095
2096 foreach my $snapname (keys %{$conf->{snapshots}}) {
2097 die "internal error" if $snapname eq 'pending';
2098 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2099 }
2100
2101 # remove 'unusedX' settings if we re-add a volume
2102 foreach my $key (keys %$conf) {
2103 my $value = $conf->{$key};
2104 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2105 delete $conf->{$key};
2106 }
2107 }
2108
2109 my $generate_raw_config = sub {
2110 my ($conf) = @_;
2111
2112 my $raw = '';
2113
2114 # add description as comment to top of file
2115 my $descr = $conf->{description} || '';
2116 foreach my $cl (split(/\n/, $descr)) {
2117 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2118 }
2119
2120 foreach my $key (sort keys %$conf) {
2121 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2122 $raw .= "$key: $conf->{$key}\n";
2123 }
2124 return $raw;
2125 };
2126
2127 my $raw = &$generate_raw_config($conf);
2128
2129 if (scalar(keys %{$conf->{pending}})){
2130 $raw .= "\n[PENDING]\n";
2131 $raw .= &$generate_raw_config($conf->{pending});
2132 }
2133
2134 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2135 $raw .= "\n[$snapname]\n";
2136 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2137 }
2138
2139 return $raw;
2140 }
2141
2142 sub update_config_nolock {
2143 my ($vmid, $conf, $skiplock) = @_;
2144
2145 check_lock($conf) if !$skiplock;
2146
2147 my $cfspath = cfs_config_path($vmid);
2148
2149 PVE::Cluster::cfs_write_file($cfspath, $conf);
2150 }
2151
2152 sub update_config {
2153 my ($vmid, $conf, $skiplock) = @_;
2154
2155 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2156 }
2157
2158 sub load_defaults {
2159
2160 my $res = {};
2161
2162 # we use static defaults from our JSON schema configuration
2163 foreach my $key (keys %$confdesc) {
2164 if (defined(my $default = $confdesc->{$key}->{default})) {
2165 $res->{$key} = $default;
2166 }
2167 }
2168
2169 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2170 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2171
2172 return $res;
2173 }
2174
2175 sub config_list {
2176 my $vmlist = PVE::Cluster::get_vmlist();
2177 my $res = {};
2178 return $res if !$vmlist || !$vmlist->{ids};
2179 my $ids = $vmlist->{ids};
2180
2181 foreach my $vmid (keys %$ids) {
2182 my $d = $ids->{$vmid};
2183 next if !$d->{node} || $d->{node} ne $nodename;
2184 next if !$d->{type} || $d->{type} ne 'qemu';
2185 $res->{$vmid}->{exists} = 1;
2186 }
2187 return $res;
2188 }
2189
2190 # test if VM uses local resources (to prevent migration)
2191 sub check_local_resources {
2192 my ($conf, $noerr) = @_;
2193
2194 my $loc_res = 0;
2195
2196 $loc_res = 1 if $conf->{hostusb}; # old syntax
2197 $loc_res = 1 if $conf->{hostpci}; # old syntax
2198
2199 foreach my $k (keys %$conf) {
2200 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2201 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2202 }
2203
2204 die "VM uses local resources\n" if $loc_res && !$noerr;
2205
2206 return $loc_res;
2207 }
2208
2209 # check if used storages are available on all nodes (use by migrate)
2210 sub check_storage_availability {
2211 my ($storecfg, $conf, $node) = @_;
2212
2213 foreach_drive($conf, sub {
2214 my ($ds, $drive) = @_;
2215
2216 my $volid = $drive->{file};
2217 return if !$volid;
2218
2219 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2220 return if !$sid;
2221
2222 # check if storage is available on both nodes
2223 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2224 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2225 });
2226 }
2227
2228 # list nodes where all VM images are available (used by has_feature API)
2229 sub shared_nodes {
2230 my ($conf, $storecfg) = @_;
2231
2232 my $nodelist = PVE::Cluster::get_nodelist();
2233 my $nodehash = { map { $_ => 1 } @$nodelist };
2234 my $nodename = PVE::INotify::nodename();
2235
2236 foreach_drive($conf, sub {
2237 my ($ds, $drive) = @_;
2238
2239 my $volid = $drive->{file};
2240 return if !$volid;
2241
2242 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2243 if ($storeid) {
2244 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2245 if ($scfg->{disable}) {
2246 $nodehash = {};
2247 } elsif (my $avail = $scfg->{nodes}) {
2248 foreach my $node (keys %$nodehash) {
2249 delete $nodehash->{$node} if !$avail->{$node};
2250 }
2251 } elsif (!$scfg->{shared}) {
2252 foreach my $node (keys %$nodehash) {
2253 delete $nodehash->{$node} if $node ne $nodename
2254 }
2255 }
2256 }
2257 });
2258
2259 return $nodehash
2260 }
2261
2262 sub check_lock {
2263 my ($conf) = @_;
2264
2265 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2266 }
2267
2268 sub check_cmdline {
2269 my ($pidfile, $pid) = @_;
2270
2271 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2272 if (defined($fh)) {
2273 my $line = <$fh>;
2274 $fh->close;
2275 return undef if !$line;
2276 my @param = split(/\0/, $line);
2277
2278 my $cmd = $param[0];
2279 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2280
2281 for (my $i = 0; $i < scalar (@param); $i++) {
2282 my $p = $param[$i];
2283 next if !$p;
2284 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2285 my $p = $param[$i+1];
2286 return 1 if $p && ($p eq $pidfile);
2287 return undef;
2288 }
2289 }
2290 }
2291 return undef;
2292 }
2293
2294 sub check_running {
2295 my ($vmid, $nocheck, $node) = @_;
2296
2297 my $filename = config_file($vmid, $node);
2298
2299 die "unable to find configuration file for VM $vmid - no such machine\n"
2300 if !$nocheck && ! -f $filename;
2301
2302 my $pidfile = pidfile_name($vmid);
2303
2304 if (my $fd = IO::File->new("<$pidfile")) {
2305 my $st = stat($fd);
2306 my $line = <$fd>;
2307 close($fd);
2308
2309 my $mtime = $st->mtime;
2310 if ($mtime > time()) {
2311 warn "file '$filename' modified in future\n";
2312 }
2313
2314 if ($line =~ m/^(\d+)$/) {
2315 my $pid = $1;
2316 if (check_cmdline($pidfile, $pid)) {
2317 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2318 return $pid;
2319 }
2320 }
2321 }
2322 }
2323
2324 return undef;
2325 }
2326
2327 sub vzlist {
2328
2329 my $vzlist = config_list();
2330
2331 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2332
2333 while (defined(my $de = $fd->read)) {
2334 next if $de !~ m/^(\d+)\.pid$/;
2335 my $vmid = $1;
2336 next if !defined($vzlist->{$vmid});
2337 if (my $pid = check_running($vmid)) {
2338 $vzlist->{$vmid}->{pid} = $pid;
2339 }
2340 }
2341
2342 return $vzlist;
2343 }
2344
2345 sub disksize {
2346 my ($storecfg, $conf) = @_;
2347
2348 my $bootdisk = $conf->{bootdisk};
2349 return undef if !$bootdisk;
2350 return undef if !valid_drivename($bootdisk);
2351
2352 return undef if !$conf->{$bootdisk};
2353
2354 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2355 return undef if !defined($drive);
2356
2357 return undef if drive_is_cdrom($drive);
2358
2359 my $volid = $drive->{file};
2360 return undef if !$volid;
2361
2362 return $drive->{size};
2363 }
2364
2365 my $last_proc_pid_stat;
2366
2367 # get VM status information
2368 # This must be fast and should not block ($full == false)
2369 # We only query KVM using QMP if $full == true (this can be slow)
2370 sub vmstatus {
2371 my ($opt_vmid, $full) = @_;
2372
2373 my $res = {};
2374
2375 my $storecfg = PVE::Storage::config();
2376
2377 my $list = vzlist();
2378 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2379
2380 my $cpucount = $cpuinfo->{cpus} || 1;
2381
2382 foreach my $vmid (keys %$list) {
2383 next if $opt_vmid && ($vmid ne $opt_vmid);
2384
2385 my $cfspath = cfs_config_path($vmid);
2386 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2387
2388 my $d = {};
2389 $d->{pid} = $list->{$vmid}->{pid};
2390
2391 # fixme: better status?
2392 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2393
2394 my $size = disksize($storecfg, $conf);
2395 if (defined($size)) {
2396 $d->{disk} = 0; # no info available
2397 $d->{maxdisk} = $size;
2398 } else {
2399 $d->{disk} = 0;
2400 $d->{maxdisk} = 0;
2401 }
2402
2403 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2404 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2405 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2406
2407 $d->{name} = $conf->{name} || "VM $vmid";
2408 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2409
2410 if ($conf->{balloon}) {
2411 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2412 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2413 }
2414
2415 $d->{uptime} = 0;
2416 $d->{cpu} = 0;
2417 $d->{mem} = 0;
2418
2419 $d->{netout} = 0;
2420 $d->{netin} = 0;
2421
2422 $d->{diskread} = 0;
2423 $d->{diskwrite} = 0;
2424
2425 $d->{template} = is_template($conf);
2426
2427 $res->{$vmid} = $d;
2428 }
2429
2430 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2431 foreach my $dev (keys %$netdev) {
2432 next if $dev !~ m/^tap([1-9]\d*)i/;
2433 my $vmid = $1;
2434 my $d = $res->{$vmid};
2435 next if !$d;
2436
2437 $d->{netout} += $netdev->{$dev}->{receive};
2438 $d->{netin} += $netdev->{$dev}->{transmit};
2439 }
2440
2441 my $ctime = gettimeofday;
2442
2443 foreach my $vmid (keys %$list) {
2444
2445 my $d = $res->{$vmid};
2446 my $pid = $d->{pid};
2447 next if !$pid;
2448
2449 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2450 next if !$pstat; # not running
2451
2452 my $used = $pstat->{utime} + $pstat->{stime};
2453
2454 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2455
2456 if ($pstat->{vsize}) {
2457 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2458 }
2459
2460 my $old = $last_proc_pid_stat->{$pid};
2461 if (!$old) {
2462 $last_proc_pid_stat->{$pid} = {
2463 time => $ctime,
2464 used => $used,
2465 cpu => 0,
2466 };
2467 next;
2468 }
2469
2470 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2471
2472 if ($dtime > 1000) {
2473 my $dutime = $used - $old->{used};
2474
2475 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2476 $last_proc_pid_stat->{$pid} = {
2477 time => $ctime,
2478 used => $used,
2479 cpu => $d->{cpu},
2480 };
2481 } else {
2482 $d->{cpu} = $old->{cpu};
2483 }
2484 }
2485
2486 return $res if !$full;
2487
2488 my $qmpclient = PVE::QMPClient->new();
2489
2490 my $ballooncb = sub {
2491 my ($vmid, $resp) = @_;
2492
2493 my $info = $resp->{'return'};
2494 return if !$info->{max_mem};
2495
2496 my $d = $res->{$vmid};
2497
2498 # use memory assigned to VM
2499 $d->{maxmem} = $info->{max_mem};
2500 $d->{balloon} = $info->{actual};
2501
2502 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2503 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2504 $d->{freemem} = $info->{free_mem};
2505 }
2506
2507 };
2508
2509 my $blockstatscb = sub {
2510 my ($vmid, $resp) = @_;
2511 my $data = $resp->{'return'} || [];
2512 my $totalrdbytes = 0;
2513 my $totalwrbytes = 0;
2514 for my $blockstat (@$data) {
2515 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2516 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2517 }
2518 $res->{$vmid}->{diskread} = $totalrdbytes;
2519 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2520 };
2521
2522 my $statuscb = sub {
2523 my ($vmid, $resp) = @_;
2524
2525 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2526 # this fails if ballon driver is not loaded, so this must be
2527 # the last commnand (following command are aborted if this fails).
2528 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2529
2530 my $status = 'unknown';
2531 if (!defined($status = $resp->{'return'}->{status})) {
2532 warn "unable to get VM status\n";
2533 return;
2534 }
2535
2536 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2537 };
2538
2539 foreach my $vmid (keys %$list) {
2540 next if $opt_vmid && ($vmid ne $opt_vmid);
2541 next if !$res->{$vmid}->{pid}; # not running
2542 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2543 }
2544
2545 $qmpclient->queue_execute(undef, 1);
2546
2547 foreach my $vmid (keys %$list) {
2548 next if $opt_vmid && ($vmid ne $opt_vmid);
2549 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2550 }
2551
2552 return $res;
2553 }
2554
2555 sub foreach_dimm {
2556 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2557
2558 my $dimm_id = 0;
2559 my $current_size = 1024;
2560 my $dimm_size = 512;
2561 return if $current_size == $memory;
2562
2563 for (my $j = 0; $j < 8; $j++) {
2564 for (my $i = 0; $i < 32; $i++) {
2565 my $name = "dimm${dimm_id}";
2566 $dimm_id++;
2567 my $numanode = $i % $sockets;
2568 $current_size += $dimm_size;
2569 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2570 return $current_size if $current_size >= $memory;
2571 }
2572 $dimm_size *= 2;
2573 }
2574 }
2575
2576 sub foreach_drive {
2577 my ($conf, $func) = @_;
2578
2579 foreach my $ds (keys %$conf) {
2580 next if !valid_drivename($ds);
2581
2582 my $drive = parse_drive($ds, $conf->{$ds});
2583 next if !$drive;
2584
2585 &$func($ds, $drive);
2586 }
2587 }
2588
2589 sub foreach_volid {
2590 my ($conf, $func) = @_;
2591
2592 my $volhash = {};
2593
2594 my $test_volid = sub {
2595 my ($volid, $is_cdrom) = @_;
2596
2597 return if !$volid;
2598
2599 $volhash->{$volid} = $is_cdrom || 0;
2600 };
2601
2602 foreach_drive($conf, sub {
2603 my ($ds, $drive) = @_;
2604 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2605 });
2606
2607 foreach my $snapname (keys %{$conf->{snapshots}}) {
2608 my $snap = $conf->{snapshots}->{$snapname};
2609 &$test_volid($snap->{vmstate}, 0);
2610 foreach_drive($snap, sub {
2611 my ($ds, $drive) = @_;
2612 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2613 });
2614 }
2615
2616 foreach my $volid (keys %$volhash) {
2617 &$func($volid, $volhash->{$volid});
2618 }
2619 }
2620
2621 sub vga_conf_has_spice {
2622 my ($vga) = @_;
2623
2624 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2625
2626 return $1 || 1;
2627 }
2628
2629 sub config_to_command {
2630 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2631
2632 my $cmd = [];
2633 my $globalFlags = [];
2634 my $machineFlags = [];
2635 my $rtcFlags = [];
2636 my $cpuFlags = [];
2637 my $devices = [];
2638 my $pciaddr = '';
2639 my $bridges = {};
2640 my $kvmver = kvm_user_version();
2641 my $vernum = 0; # unknown
2642 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2643 $vernum = $1*1000000+$2*1000;
2644 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2645 $vernum = $1*1000000+$2*1000+$3;
2646 }
2647
2648 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2649
2650 my $have_ovz = -f '/proc/vz/vestat';
2651
2652 my $q35 = machine_type_is_q35($conf);
2653 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2654 my $machine_type = $forcemachine || $conf->{machine};
2655
2656 push @$cmd, '/usr/bin/kvm';
2657
2658 push @$cmd, '-id', $vmid;
2659
2660 my $use_virtio = 0;
2661
2662 my $qmpsocket = qmp_socket($vmid);
2663 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2664 push @$cmd, '-mon', "chardev=qmp,mode=control";
2665
2666 my $socket = vnc_socket($vmid);
2667 push @$cmd, '-vnc', "unix:$socket,x509,password";
2668
2669 push @$cmd, '-pidfile' , pidfile_name($vmid);
2670
2671 push @$cmd, '-daemonize';
2672
2673 if ($conf->{smbios1}) {
2674 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2675 }
2676
2677 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2678
2679 if ($q35) {
2680 # the q35 chipset support native usb2, so we enable usb controller
2681 # by default for this machine type
2682 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2683 } else {
2684 $pciaddr = print_pci_addr("piix3", $bridges);
2685 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2686
2687 my $use_usb2 = 0;
2688 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2689 next if !$conf->{"usb$i"};
2690 $use_usb2 = 1;
2691 }
2692 # include usb device config
2693 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2694 }
2695
2696 my $vga = $conf->{vga};
2697
2698 my $qxlnum = vga_conf_has_spice($vga);
2699 $vga = 'qxl' if $qxlnum;
2700
2701 if (!$vga) {
2702 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2703 $conf->{ostype} eq 'win7' ||
2704 $conf->{ostype} eq 'w2k8')) {
2705 $vga = 'std';
2706 } else {
2707 $vga = 'cirrus';
2708 }
2709 }
2710
2711 # enable absolute mouse coordinates (needed by vnc)
2712 my $tablet;
2713 if (defined($conf->{tablet})) {
2714 $tablet = $conf->{tablet};
2715 } else {
2716 $tablet = $defaults->{tablet};
2717 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2718 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2719 }
2720
2721 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2722
2723 # host pci devices
2724 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2725 my $d = parse_hostpci($conf->{"hostpci$i"});
2726 next if !$d;
2727
2728 my $pcie = $d->{pcie};
2729 if($pcie){
2730 die "q35 machine model is not enabled" if !$q35;
2731 $pciaddr = print_pcie_addr("hostpci$i");
2732 }else{
2733 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2734 }
2735
2736 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2737 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2738 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2739 if ($xvga && $xvga ne '') {
2740 push @$cpuFlags, 'kvm=off';
2741 $vga = 'none';
2742 }
2743 $driver = "vfio-pci" if $xvga ne '';
2744 my $pcidevices = $d->{pciid};
2745 my $multifunction = 1 if @$pcidevices > 1;
2746
2747 my $j=0;
2748 foreach my $pcidevice (@$pcidevices) {
2749
2750 my $id = "hostpci$i";
2751 $id .= ".$j" if $multifunction;
2752 my $addr = $pciaddr;
2753 $addr .= ".$j" if $multifunction;
2754 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2755
2756 if($j == 0){
2757 $devicestr .= "$rombar$xvga";
2758 $devicestr .= ",multifunction=on" if $multifunction;
2759 }
2760
2761 push @$devices, '-device', $devicestr;
2762 $j++;
2763 }
2764 }
2765
2766 # usb devices
2767 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2768 my $d = parse_usb_device($conf->{"usb$i"});
2769 next if !$d;
2770 if ($d->{vendorid} && $d->{productid}) {
2771 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2772 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2773 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2774 } elsif ($d->{spice}) {
2775 # usb redir support for spice
2776 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2777 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2778 }
2779 }
2780
2781 # serial devices
2782 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2783 if (my $path = $conf->{"serial$i"}) {
2784 if ($path eq 'socket') {
2785 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2786 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2787 push @$devices, '-device', "isa-serial,chardev=serial$i";
2788 } else {
2789 die "no such serial device\n" if ! -c $path;
2790 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2791 push @$devices, '-device', "isa-serial,chardev=serial$i";
2792 }
2793 }
2794 }
2795
2796 # parallel devices
2797 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2798 if (my $path = $conf->{"parallel$i"}) {
2799 die "no such parallel device\n" if ! -c $path;
2800 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2801 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2802 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2803 }
2804 }
2805
2806 my $vmname = $conf->{name} || "vm$vmid";
2807
2808 push @$cmd, '-name', $vmname;
2809
2810 my $sockets = 1;
2811 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2812 $sockets = $conf->{sockets} if $conf->{sockets};
2813
2814 my $cores = $conf->{cores} || 1;
2815
2816 my $maxcpus = $sockets * $cores;
2817
2818 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2819
2820 my $allowed_vcpus = $cpuinfo->{cpus};
2821
2822 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2823 if ($allowed_vcpus < $maxcpus);
2824
2825 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2826
2827 push @$cmd, '-nodefaults';
2828
2829 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2830
2831 my $bootindex_hash = {};
2832 my $i = 1;
2833 foreach my $o (split(//, $bootorder)) {
2834 $bootindex_hash->{$o} = $i*100;
2835 $i++;
2836 }
2837
2838 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2839
2840 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2841
2842 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2843
2844 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2845
2846 # time drift fix
2847 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2848
2849 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2850 my $useLocaltime = $conf->{localtime};
2851
2852 if (my $ost = $conf->{ostype}) {
2853 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2854
2855 if ($ost =~ m/^w/) { # windows
2856 $useLocaltime = 1 if !defined($conf->{localtime});
2857
2858 # use time drift fix when acpi is enabled
2859 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2860 $tdf = 1 if !defined($conf->{tdf});
2861 }
2862 }
2863
2864 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2865 $ost eq 'wvista') {
2866 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2867 push @$cmd, '-no-hpet';
2868 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2869 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2870 push @$cpuFlags , 'hv_spinlocks=0x1fff' if !$nokvm;
2871 } else {
2872 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2873 }
2874 }
2875
2876 if ($ost eq 'win7' || $ost eq 'win8') {
2877 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2878 }
2879 }
2880
2881 push @$rtcFlags, 'driftfix=slew' if $tdf;
2882
2883 if ($nokvm) {
2884 push @$machineFlags, 'accel=tcg';
2885 } else {
2886 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2887 }
2888
2889 if ($machine_type) {
2890 push @$machineFlags, "type=${machine_type}";
2891 }
2892
2893 if ($conf->{startdate}) {
2894 push @$rtcFlags, "base=$conf->{startdate}";
2895 } elsif ($useLocaltime) {
2896 push @$rtcFlags, 'base=localtime';
2897 }
2898
2899 my $cpu = $nokvm ? "qemu64" : "kvm64";
2900 $cpu = $conf->{cpu} if $conf->{cpu};
2901
2902 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2903
2904 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2905
2906 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2907
2908 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2909
2910 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2911
2912 # Note: enforce needs kernel 3.10, so we do not use it for now
2913 # push @$cmd, '-cpu', "$cpu,enforce";
2914 push @$cmd, '-cpu', $cpu;
2915
2916 my $memory = $conf->{memory} || $defaults->{memory};
2917 my $static_memory = 0;
2918 my $dimm_memory = 0;
2919
2920 if ($hotplug_features->{memory}) {
2921 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2922 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2923 $static_memory = $STATICMEM;
2924 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2925 $dimm_memory = $memory - $static_memory;
2926 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2927
2928 } else {
2929
2930 $static_memory = $memory;
2931 push @$cmd, '-m', $static_memory;
2932 }
2933
2934 if ($conf->{numa}) {
2935
2936 my $numa_totalmemory = undef;
2937 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2938 next if !$conf->{"numa$i"};
2939 my $numa = parse_numa($conf->{"numa$i"});
2940 next if !$numa;
2941 # memory
2942 die "missing numa node$i memory value\n" if !$numa->{memory};
2943 my $numa_memory = $numa->{memory};
2944 $numa_totalmemory += $numa_memory;
2945 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2946
2947 # cpus
2948 my $cpus_start = $numa->{cpus}->{start};
2949 die "missing numa node$i cpus\n" if !defined($cpus_start);
2950 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2951 my $cpus = $cpus_start;
2952 if (defined($cpus_end)) {
2953 $cpus .= "-$cpus_end";
2954 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2955 }
2956
2957 # hostnodes
2958 my $hostnodes_start = $numa->{hostnodes}->{start};
2959 if (defined($hostnodes_start)) {
2960 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2961 my $hostnodes = $hostnodes_start;
2962 if (defined($hostnodes_end)) {
2963 $hostnodes .= "-$hostnodes_end";
2964 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2965 }
2966
2967 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2968 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2969 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2970 }
2971
2972 # policy
2973 my $policy = $numa->{policy};
2974 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2975 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2976 }
2977
2978 push @$cmd, '-object', $numa_object;
2979 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2980 }
2981
2982 die "total memory for NUMA nodes must be equal to vm static memory\n"
2983 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2984
2985 #if no custom tology, we split memory and cores across numa nodes
2986 if(!$numa_totalmemory) {
2987
2988 my $numa_memory = ($static_memory / $sockets) . "M";
2989
2990 for (my $i = 0; $i < $sockets; $i++) {
2991
2992 my $cpustart = ($cores * $i);
2993 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2994 my $cpus = $cpustart;
2995 $cpus .= "-$cpuend" if $cpuend;
2996
2997 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2998 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2999 }
3000 }
3001 }
3002
3003 if ($hotplug_features->{memory}) {
3004 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
3005 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3006 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3007 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3008
3009 #if dimm_memory is not aligned to dimm map
3010 if($current_size > $memory) {
3011 $conf->{memory} = $current_size;
3012 update_config_nolock($vmid, $conf, 1);
3013 }
3014 });
3015 }
3016
3017 push @$cmd, '-S' if $conf->{freeze};
3018
3019 # set keyboard layout
3020 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3021 push @$cmd, '-k', $kb if $kb;
3022
3023 # enable sound
3024 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3025 #push @$cmd, '-soundhw', 'es1370';
3026 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3027
3028 if($conf->{agent}) {
3029 my $qgasocket = qmp_socket($vmid, 1);
3030 my $pciaddr = print_pci_addr("qga0", $bridges);
3031 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3032 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3033 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3034 }
3035
3036 my $spice_port;
3037
3038 if ($qxlnum) {
3039 if ($qxlnum > 1) {
3040 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3041 for(my $i = 1; $i < $qxlnum; $i++){
3042 my $pciaddr = print_pci_addr("vga$i", $bridges);
3043 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3044 }
3045 } else {
3046 # assume other OS works like Linux
3047 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3048 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3049 }
3050 }
3051
3052 my $pciaddr = print_pci_addr("spice", $bridges);
3053
3054 $spice_port = PVE::Tools::next_spice_port();
3055
3056 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3057
3058 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3059 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3060 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3061 }
3062
3063 # enable balloon by default, unless explicitly disabled
3064 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3065 $pciaddr = print_pci_addr("balloon0", $bridges);
3066 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3067 }
3068
3069 if ($conf->{watchdog}) {
3070 my $wdopts = parse_watchdog($conf->{watchdog});
3071 $pciaddr = print_pci_addr("watchdog", $bridges);
3072 my $watchdog = $wdopts->{model} || 'i6300esb';
3073 push @$devices, '-device', "$watchdog$pciaddr";
3074 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3075 }
3076
3077 my $vollist = [];
3078 my $scsicontroller = {};
3079 my $ahcicontroller = {};
3080 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3081
3082 # Add iscsi initiator name if available
3083 if (my $initiator = get_initiator_name()) {
3084 push @$devices, '-iscsi', "initiator-name=$initiator";
3085 }
3086
3087 foreach_drive($conf, sub {
3088 my ($ds, $drive) = @_;
3089
3090 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3091 push @$vollist, $drive->{file};
3092 }
3093
3094 $use_virtio = 1 if $ds =~ m/^virtio/;
3095
3096 if (drive_is_cdrom ($drive)) {
3097 if ($bootindex_hash->{d}) {
3098 $drive->{bootindex} = $bootindex_hash->{d};
3099 $bootindex_hash->{d} += 1;
3100 }
3101 } else {
3102 if ($bootindex_hash->{c}) {
3103 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3104 $bootindex_hash->{c} += 1;
3105 }
3106 }
3107
3108 if ($drive->{interface} eq 'scsi') {
3109
3110 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3111 my $controller = int($drive->{index} / $maxdev);
3112 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3113 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3114 $scsicontroller->{$controller}=1;
3115 }
3116
3117 if ($drive->{interface} eq 'sata') {
3118 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3119 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3120 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3121 $ahcicontroller->{$controller}=1;
3122 }
3123
3124 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3125 push @$devices, '-drive',$drive_cmd;
3126 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3127 });
3128
3129 for (my $i = 0; $i < $MAX_NETS; $i++) {
3130 next if !$conf->{"net$i"};
3131 my $d = parse_net($conf->{"net$i"});
3132 next if !$d;
3133
3134 $use_virtio = 1 if $d->{model} eq 'virtio';
3135
3136 if ($bootindex_hash->{n}) {
3137 $d->{bootindex} = $bootindex_hash->{n};
3138 $bootindex_hash->{n} += 1;
3139 }
3140
3141 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3142 push @$devices, '-netdev', $netdevfull;
3143
3144 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3145 push @$devices, '-device', $netdevicefull;
3146 }
3147
3148 if (!$q35) {
3149 # add pci bridges
3150 while (my ($k, $v) = each %$bridges) {
3151 $pciaddr = print_pci_addr("pci.$k");
3152 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3153 }
3154 }
3155
3156 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3157 # when the VM uses virtio devices.
3158 if (!$use_virtio && $have_ovz) {
3159
3160 my $cpuunits = defined($conf->{cpuunits}) ?
3161 $conf->{cpuunits} : $defaults->{cpuunits};
3162
3163 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3164
3165 # fixme: cpulimit is currently ignored
3166 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3167 }
3168
3169 # add custom args
3170 if ($conf->{args}) {
3171 my $aa = PVE::Tools::split_args($conf->{args});
3172 push @$cmd, @$aa;
3173 }
3174
3175 push @$cmd, @$devices;
3176 push @$cmd, '-rtc', join(',', @$rtcFlags)
3177 if scalar(@$rtcFlags);
3178 push @$cmd, '-machine', join(',', @$machineFlags)
3179 if scalar(@$machineFlags);
3180 push @$cmd, '-global', join(',', @$globalFlags)
3181 if scalar(@$globalFlags);
3182
3183 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3184 }
3185
3186 sub vnc_socket {
3187 my ($vmid) = @_;
3188 return "${var_run_tmpdir}/$vmid.vnc";
3189 }
3190
3191 sub spice_port {
3192 my ($vmid) = @_;
3193
3194 my $res = vm_mon_cmd($vmid, 'query-spice');
3195
3196 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3197 }
3198
3199 sub qmp_socket {
3200 my ($vmid, $qga) = @_;
3201 my $sockettype = $qga ? 'qga' : 'qmp';
3202 return "${var_run_tmpdir}/$vmid.$sockettype";
3203 }
3204
3205 sub pidfile_name {
3206 my ($vmid) = @_;
3207 return "${var_run_tmpdir}/$vmid.pid";
3208 }
3209
3210 sub vm_devices_list {
3211 my ($vmid) = @_;
3212
3213 my $res = vm_mon_cmd($vmid, 'query-pci');
3214 my $devices = {};
3215 foreach my $pcibus (@$res) {
3216 foreach my $device (@{$pcibus->{devices}}) {
3217 next if !$device->{'qdev_id'};
3218 if ($device->{'pci_bridge'}) {
3219 $devices->{$device->{'qdev_id'}} = 1;
3220 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3221 next if !$bridge_device->{'qdev_id'};
3222 $devices->{$bridge_device->{'qdev_id'}} = 1;
3223 $devices->{$device->{'qdev_id'}}++;
3224 }
3225 } else {
3226 $devices->{$device->{'qdev_id'}} = 1;
3227 }
3228 }
3229 }
3230
3231 my $resblock = vm_mon_cmd($vmid, 'query-block');
3232 foreach my $block (@$resblock) {
3233 if($block->{device} =~ m/^drive-(\S+)/){
3234 $devices->{$1} = 1;
3235 }
3236 }
3237
3238 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3239 foreach my $mice (@$resmice) {
3240 if ($mice->{name} eq 'QEMU HID Tablet') {
3241 $devices->{tablet} = 1;
3242 last;
3243 }
3244 }
3245
3246 return $devices;
3247 }
3248
3249 sub vm_deviceplug {
3250 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3251
3252 my $q35 = machine_type_is_q35($conf);
3253
3254 my $devices_list = vm_devices_list($vmid);
3255 return 1 if defined($devices_list->{$deviceid});
3256
3257 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3258
3259 if ($deviceid eq 'tablet') {
3260
3261 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3262
3263 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3264
3265 qemu_driveadd($storecfg, $vmid, $device);
3266 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3267
3268 qemu_deviceadd($vmid, $devicefull);
3269 eval { qemu_deviceaddverify($vmid, $deviceid); };
3270 if (my $err = $@) {
3271 eval { qemu_drivedel($vmid, $deviceid); };
3272 warn $@ if $@;
3273 die $err;
3274 }
3275
3276 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3277
3278 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3279 my $pciaddr = print_pci_addr($deviceid);
3280 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3281
3282 qemu_deviceadd($vmid, $devicefull);
3283 qemu_deviceaddverify($vmid, $deviceid);
3284
3285 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3286
3287 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3288 qemu_driveadd($storecfg, $vmid, $device);
3289
3290 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3291 eval { qemu_deviceadd($vmid, $devicefull); };
3292 if (my $err = $@) {
3293 eval { qemu_drivedel($vmid, $deviceid); };
3294 warn $@ if $@;
3295 die $err;
3296 }
3297
3298 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3299
3300 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3301 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3302 qemu_deviceadd($vmid, $netdevicefull);
3303 eval { qemu_deviceaddverify($vmid, $deviceid); };
3304 if (my $err = $@) {
3305 eval { qemu_netdevdel($vmid, $deviceid); };
3306 warn $@ if $@;
3307 die $err;
3308 }
3309
3310 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3311
3312 my $bridgeid = $2;
3313 my $pciaddr = print_pci_addr($deviceid);
3314 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3315
3316 qemu_deviceadd($vmid, $devicefull);
3317 qemu_deviceaddverify($vmid, $deviceid);
3318
3319 } else {
3320 die "can't hotplug device '$deviceid'\n";
3321 }
3322
3323 return 1;
3324 }
3325
3326 # fixme: this should raise exceptions on error!
3327 sub vm_deviceunplug {
3328 my ($vmid, $conf, $deviceid) = @_;
3329
3330 my $devices_list = vm_devices_list($vmid);
3331 return 1 if !defined($devices_list->{$deviceid});
3332
3333 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3334
3335 if ($deviceid eq 'tablet') {
3336
3337 qemu_devicedel($vmid, $deviceid);
3338
3339 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3340
3341 qemu_devicedel($vmid, $deviceid);
3342 qemu_devicedelverify($vmid, $deviceid);
3343 qemu_drivedel($vmid, $deviceid);
3344
3345 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3346
3347 qemu_devicedel($vmid, $deviceid);
3348
3349 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3350
3351 qemu_devicedel($vmid, $deviceid);
3352 qemu_drivedel($vmid, $deviceid);
3353
3354 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3355
3356 qemu_devicedel($vmid, $deviceid);
3357 qemu_devicedelverify($vmid, $deviceid);
3358 qemu_netdevdel($vmid, $deviceid);
3359
3360 } else {
3361 die "can't unplug device '$deviceid'\n";
3362 }
3363
3364 return 1;
3365 }
3366
3367 sub qemu_deviceadd {
3368 my ($vmid, $devicefull) = @_;
3369
3370 $devicefull = "driver=".$devicefull;
3371 my %options = split(/[=,]/, $devicefull);
3372
3373 vm_mon_cmd($vmid, "device_add" , %options);
3374 }
3375
3376 sub qemu_devicedel {
3377 my ($vmid, $deviceid) = @_;
3378
3379 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3380 }
3381
3382 sub qemu_objectadd {
3383 my($vmid, $objectid, $qomtype) = @_;
3384
3385 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3386
3387 return 1;
3388 }
3389
3390 sub qemu_objectdel {
3391 my($vmid, $objectid) = @_;
3392
3393 vm_mon_cmd($vmid, "object-del", id => $objectid);
3394
3395 return 1;
3396 }
3397
3398 sub qemu_driveadd {
3399 my ($storecfg, $vmid, $device) = @_;
3400
3401 my $drive = print_drive_full($storecfg, $vmid, $device);
3402 $drive =~ s/\\/\\\\/g;
3403 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3404
3405 # If the command succeeds qemu prints: "OK"
3406 return 1 if $ret =~ m/OK/s;
3407
3408 die "adding drive failed: $ret\n";
3409 }
3410
3411 sub qemu_drivedel {
3412 my($vmid, $deviceid) = @_;
3413
3414 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3415 $ret =~ s/^\s+//;
3416
3417 return 1 if $ret eq "";
3418
3419 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3420 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3421
3422 die "deleting drive $deviceid failed : $ret\n";
3423 }
3424
3425 sub qemu_deviceaddverify {
3426 my ($vmid, $deviceid) = @_;
3427
3428 for (my $i = 0; $i <= 5; $i++) {
3429 my $devices_list = vm_devices_list($vmid);
3430 return 1 if defined($devices_list->{$deviceid});
3431 sleep 1;
3432 }
3433
3434 die "error on hotplug device '$deviceid'\n";
3435 }
3436
3437
3438 sub qemu_devicedelverify {
3439 my ($vmid, $deviceid) = @_;
3440
3441 # need to verify that the device is correctly removed as device_del
3442 # is async and empty return is not reliable
3443
3444 for (my $i = 0; $i <= 5; $i++) {
3445 my $devices_list = vm_devices_list($vmid);
3446 return 1 if !defined($devices_list->{$deviceid});
3447 sleep 1;
3448 }
3449
3450 die "error on hot-unplugging device '$deviceid'\n";
3451 }
3452
3453 sub qemu_findorcreatescsihw {
3454 my ($storecfg, $conf, $vmid, $device) = @_;
3455
3456 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3457 my $controller = int($device->{index} / $maxdev);
3458 my $scsihwid="scsihw$controller";
3459 my $devices_list = vm_devices_list($vmid);
3460
3461 if(!defined($devices_list->{$scsihwid})) {
3462 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3463 }
3464
3465 return 1;
3466 }
3467
3468 sub qemu_add_pci_bridge {
3469 my ($storecfg, $conf, $vmid, $device) = @_;
3470
3471 my $bridges = {};
3472
3473 my $bridgeid;
3474
3475 print_pci_addr($device, $bridges);
3476
3477 while (my ($k, $v) = each %$bridges) {
3478 $bridgeid = $k;
3479 }
3480 return 1 if !defined($bridgeid) || $bridgeid < 1;
3481
3482 my $bridge = "pci.$bridgeid";
3483 my $devices_list = vm_devices_list($vmid);
3484
3485 if (!defined($devices_list->{$bridge})) {
3486 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3487 }
3488
3489 return 1;
3490 }
3491
3492 sub qemu_set_link_status {
3493 my ($vmid, $device, $up) = @_;
3494
3495 vm_mon_cmd($vmid, "set_link", name => $device,
3496 up => $up ? JSON::true : JSON::false);
3497 }
3498
3499 sub qemu_netdevadd {
3500 my ($vmid, $conf, $device, $deviceid) = @_;
3501
3502 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3503 my %options = split(/[=,]/, $netdev);
3504
3505 vm_mon_cmd($vmid, "netdev_add", %options);
3506 return 1;
3507 }
3508
3509 sub qemu_netdevdel {
3510 my ($vmid, $deviceid) = @_;
3511
3512 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3513 }
3514
3515 sub qemu_cpu_hotplug {
3516 my ($vmid, $conf, $vcpus) = @_;
3517
3518 my $sockets = 1;
3519 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3520 $sockets = $conf->{sockets} if $conf->{sockets};
3521 my $cores = $conf->{cores} || 1;
3522 my $maxcpus = $sockets * $cores;
3523
3524 $vcpus = $maxcpus if !$vcpus;
3525
3526 die "you can't add more vcpus than maxcpus\n"
3527 if $vcpus > $maxcpus;
3528
3529 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3530 die "online cpu unplug is not yet possible\n"
3531 if $vcpus < $currentvcpus;
3532
3533 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3534 die "vcpus in running vm is different than configuration\n"
3535 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3536
3537 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3538 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3539 }
3540 }
3541
3542 sub qemu_memory_hotplug {
3543 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3544
3545 return $value if !check_running($vmid);
3546
3547 my $memory = $conf->{memory} || $defaults->{memory};
3548 $value = $defaults->{memory} if !$value;
3549 return $value if $value == $memory;
3550
3551 my $static_memory = $STATICMEM;
3552 my $dimm_memory = $memory - $static_memory;
3553
3554 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3555 die "memory unplug is not yet available" if $value < $memory;
3556 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3557
3558
3559 my $sockets = 1;
3560 $sockets = $conf->{sockets} if $conf->{sockets};
3561
3562 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3563 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3564
3565 return if $current_size <= $conf->{memory};
3566
3567 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3568 if (my $err = $@) {
3569 eval { qemu_objectdel($vmid, "mem-$name"); };
3570 die $err;
3571 }
3572
3573 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3574 if (my $err = $@) {
3575 eval { qemu_objectdel($vmid, "mem-$name"); };
3576 die $err;
3577 }
3578 #update conf after each succesful module hotplug
3579 $conf->{memory} = $current_size;
3580 update_config_nolock($vmid, $conf, 1);
3581 });
3582 }
3583
3584 sub qemu_block_set_io_throttle {
3585 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3586
3587 return if !check_running($vmid) ;
3588
3589 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($iops_wr));
3590
3591 }
3592
3593 # old code, only used to shutdown old VM after update
3594 sub __read_avail {
3595 my ($fh, $timeout) = @_;
3596
3597 my $sel = new IO::Select;
3598 $sel->add($fh);
3599
3600 my $res = '';
3601 my $buf;
3602
3603 my @ready;
3604 while (scalar (@ready = $sel->can_read($timeout))) {
3605 my $count;
3606 if ($count = $fh->sysread($buf, 8192)) {
3607 if ($buf =~ /^(.*)\(qemu\) $/s) {
3608 $res .= $1;
3609 last;
3610 } else {
3611 $res .= $buf;
3612 }
3613 } else {
3614 if (!defined($count)) {
3615 die "$!\n";
3616 }
3617 last;
3618 }
3619 }
3620
3621 die "monitor read timeout\n" if !scalar(@ready);
3622
3623 return $res;
3624 }
3625
3626 # old code, only used to shutdown old VM after update
3627 sub vm_monitor_command {
3628 my ($vmid, $cmdstr, $nocheck) = @_;
3629
3630 my $res;
3631
3632 eval {
3633 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3634
3635 my $sname = "${var_run_tmpdir}/$vmid.mon";
3636
3637 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3638 die "unable to connect to VM $vmid socket - $!\n";
3639
3640 my $timeout = 3;
3641
3642 # hack: migrate sometime blocks the monitor (when migrate_downtime
3643 # is set)
3644 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3645 $timeout = 60*60; # 1 hour
3646 }
3647
3648 # read banner;
3649 my $data = __read_avail($sock, $timeout);
3650
3651 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3652 die "got unexpected qemu monitor banner\n";
3653 }
3654
3655 my $sel = new IO::Select;
3656 $sel->add($sock);
3657
3658 if (!scalar(my @ready = $sel->can_write($timeout))) {
3659 die "monitor write error - timeout";
3660 }
3661
3662 my $fullcmd = "$cmdstr\r";
3663
3664 # syslog('info', "VM $vmid monitor command: $cmdstr");
3665
3666 my $b;
3667 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3668 die "monitor write error - $!";
3669 }
3670
3671 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3672
3673 $timeout = 20;
3674
3675 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3676 $timeout = 60*60; # 1 hour
3677 } elsif ($cmdstr =~ m/^(eject|change)/) {
3678 $timeout = 60; # note: cdrom mount command is slow
3679 }
3680 if ($res = __read_avail($sock, $timeout)) {
3681
3682 my @lines = split("\r?\n", $res);
3683
3684 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3685
3686 $res = join("\n", @lines);
3687 $res .= "\n";
3688 }
3689 };
3690
3691 my $err = $@;
3692
3693 if ($err) {
3694 syslog("err", "VM $vmid monitor command failed - $err");
3695 die $err;
3696 }
3697
3698 return $res;
3699 }
3700
3701 sub qemu_block_resize {
3702 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3703
3704 my $running = check_running($vmid);
3705
3706 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3707
3708 return if !$running;
3709
3710 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3711
3712 }
3713
3714 sub qemu_volume_snapshot {
3715 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3716
3717 my $running = check_running($vmid);
3718
3719 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3720
3721 return if !$running;
3722
3723 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3724
3725 }
3726
3727 sub qemu_volume_snapshot_delete {
3728 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3729
3730 my $running = check_running($vmid);
3731
3732 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3733
3734 return if !$running;
3735
3736 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3737 }
3738
3739 sub set_migration_caps {
3740 my ($vmid) = @_;
3741
3742 my $cap_ref = [];
3743
3744 my $enabled_cap = {
3745 "auto-converge" => 1,
3746 "xbzrle" => 0,
3747 "x-rdma-pin-all" => 0,
3748 "zero-blocks" => 0,
3749 };
3750
3751 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3752
3753 for my $supported_capability (@$supported_capabilities) {
3754 push @$cap_ref, {
3755 capability => $supported_capability->{capability},
3756 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3757 };
3758 }
3759
3760 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3761 }
3762
3763 my $fast_plug_option = {
3764 'lock' => 1,
3765 'name' => 1,
3766 'onboot' => 1,
3767 'shares' => 1,
3768 'startup' => 1,
3769 };
3770
3771 # hotplug changes in [PENDING]
3772 # $selection hash can be used to only apply specified options, for
3773 # example: { cores => 1 } (only apply changed 'cores')
3774 # $errors ref is used to return error messages
3775 sub vmconfig_hotplug_pending {
3776 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3777
3778 my $defaults = load_defaults();
3779
3780 # commit values which do not have any impact on running VM first
3781 # Note: those option cannot raise errors, we we do not care about
3782 # $selection and always apply them.
3783
3784 my $add_error = sub {
3785 my ($opt, $msg) = @_;
3786 $errors->{$opt} = "hotplug problem - $msg";
3787 };
3788
3789 my $changes = 0;
3790 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3791 if ($fast_plug_option->{$opt}) {
3792 $conf->{$opt} = $conf->{pending}->{$opt};
3793 delete $conf->{pending}->{$opt};
3794 $changes = 1;
3795 }
3796 }
3797
3798 if ($changes) {
3799 update_config_nolock($vmid, $conf, 1);
3800 $conf = load_config($vmid); # update/reload
3801 }
3802
3803 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3804
3805 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3806 foreach my $opt (@delete) {
3807 next if $selection && !$selection->{$opt};
3808 eval {
3809 if ($opt eq 'hotplug') {
3810 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3811 } elsif ($opt eq 'tablet') {
3812 die "skip\n" if !$hotplug_features->{usb};
3813 if ($defaults->{tablet}) {
3814 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3815 } else {
3816 vm_deviceunplug($vmid, $conf, $opt);
3817 }
3818 } elsif ($opt eq 'vcpus') {
3819 die "skip\n" if !$hotplug_features->{cpu};
3820 qemu_cpu_hotplug($vmid, $conf, undef);
3821 } elsif ($opt eq 'balloon') {
3822 # enable balloon device is not hotpluggable
3823 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3824 } elsif ($fast_plug_option->{$opt}) {
3825 # do nothing
3826 } elsif ($opt =~ m/^net(\d+)$/) {
3827 die "skip\n" if !$hotplug_features->{network};
3828 vm_deviceunplug($vmid, $conf, $opt);
3829 } elsif (valid_drivename($opt)) {
3830 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3831 vm_deviceunplug($vmid, $conf, $opt);
3832 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3833 } elsif ($opt =~ m/^memory$/) {
3834 die "skip\n" if !$hotplug_features->{memory};
3835 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3836 } else {
3837 die "skip\n";
3838 }
3839 };
3840 if (my $err = $@) {
3841 &$add_error($opt, $err) if $err ne "skip\n";
3842 } else {
3843 # save new config if hotplug was successful
3844 delete $conf->{$opt};
3845 vmconfig_undelete_pending_option($conf, $opt);
3846 update_config_nolock($vmid, $conf, 1);
3847 $conf = load_config($vmid); # update/reload
3848 }
3849 }
3850
3851 foreach my $opt (keys %{$conf->{pending}}) {
3852 next if $selection && !$selection->{$opt};
3853 my $value = $conf->{pending}->{$opt};
3854 eval {
3855 if ($opt eq 'hotplug') {
3856 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3857 } elsif ($opt eq 'tablet') {
3858 die "skip\n" if !$hotplug_features->{usb};
3859 if ($value == 1) {
3860 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3861 } elsif ($value == 0) {
3862 vm_deviceunplug($vmid, $conf, $opt);
3863 }
3864 } elsif ($opt eq 'vcpus') {
3865 die "skip\n" if !$hotplug_features->{cpu};
3866 qemu_cpu_hotplug($vmid, $conf, $value);
3867 } elsif ($opt eq 'balloon') {
3868 # enable/disable balloning device is not hotpluggable
3869 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3870 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3871 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3872
3873 # allow manual ballooning if shares is set to zero
3874 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3875 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3876 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3877 }
3878 } elsif ($opt =~ m/^net(\d+)$/) {
3879 # some changes can be done without hotplug
3880 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3881 $vmid, $opt, $value);
3882 } elsif (valid_drivename($opt)) {
3883 # some changes can be done without hotplug
3884 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3885 $vmid, $opt, $value, 1);
3886 } elsif ($opt =~ m/^memory$/) { #dimms
3887 die "skip\n" if !$hotplug_features->{memory};
3888 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3889 } else {
3890 die "skip\n"; # skip non-hot-pluggable options
3891 }
3892 };
3893 if (my $err = $@) {
3894 &$add_error($opt, $err) if $err ne "skip\n";
3895 } else {
3896 # save new config if hotplug was successful
3897 $conf->{$opt} = $value;
3898 delete $conf->{pending}->{$opt};
3899 update_config_nolock($vmid, $conf, 1);
3900 $conf = load_config($vmid); # update/reload
3901 }
3902 }
3903 }
3904
3905 sub vmconfig_apply_pending {
3906 my ($vmid, $conf, $storecfg) = @_;
3907
3908 # cold plug
3909
3910 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3911 foreach my $opt (@delete) { # delete
3912 die "internal error" if $opt =~ m/^unused/;
3913 $conf = load_config($vmid); # update/reload
3914 if (!defined($conf->{$opt})) {
3915 vmconfig_undelete_pending_option($conf, $opt);
3916 update_config_nolock($vmid, $conf, 1);
3917 } elsif (valid_drivename($opt)) {
3918 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3919 vmconfig_undelete_pending_option($conf, $opt);
3920 delete $conf->{$opt};
3921 update_config_nolock($vmid, $conf, 1);
3922 } else {
3923 vmconfig_undelete_pending_option($conf, $opt);
3924 delete $conf->{$opt};
3925 update_config_nolock($vmid, $conf, 1);
3926 }
3927 }
3928
3929 $conf = load_config($vmid); # update/reload
3930
3931 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3932 $conf = load_config($vmid); # update/reload
3933
3934 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3935 # skip if nothing changed
3936 } elsif (valid_drivename($opt)) {
3937 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3938 if defined($conf->{$opt});
3939 $conf->{$opt} = $conf->{pending}->{$opt};
3940 } else {
3941 $conf->{$opt} = $conf->{pending}->{$opt};
3942 }
3943
3944 delete $conf->{pending}->{$opt};
3945 update_config_nolock($vmid, $conf, 1);
3946 }
3947 }
3948
3949 my $safe_num_ne = sub {
3950 my ($a, $b) = @_;
3951
3952 return 0 if !defined($a) && !defined($b);
3953 return 1 if !defined($a);
3954 return 1 if !defined($b);
3955
3956 return $a != $b;
3957 };
3958
3959 my $safe_string_ne = sub {
3960 my ($a, $b) = @_;
3961
3962 return 0 if !defined($a) && !defined($b);
3963 return 1 if !defined($a);
3964 return 1 if !defined($b);
3965
3966 return $a ne $b;
3967 };
3968
3969 sub vmconfig_update_net {
3970 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3971
3972 my $newnet = parse_net($value);
3973
3974 if ($conf->{$opt}) {
3975 my $oldnet = parse_net($conf->{$opt});
3976
3977 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3978 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3979 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3980 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3981
3982 # for non online change, we try to hot-unplug
3983 die "skip\n" if !$hotplug;
3984 vm_deviceunplug($vmid, $conf, $opt);
3985 } else {
3986
3987 die "internal error" if $opt !~ m/net(\d+)/;
3988 my $iface = "tap${vmid}i$1";
3989
3990 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3991 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3992 }
3993
3994 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3995 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3996 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3997 PVE::Network::tap_unplug($iface);
3998 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3999 }
4000
4001 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4002 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4003 }
4004
4005 return 1;
4006 }
4007 }
4008
4009 if ($hotplug) {
4010 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4011 } else {
4012 die "skip\n";
4013 }
4014 }
4015
4016 sub vmconfig_update_disk {
4017 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4018
4019 # fixme: do we need force?
4020
4021 my $drive = parse_drive($opt, $value);
4022
4023 if ($conf->{$opt}) {
4024
4025 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4026
4027 my $media = $drive->{media} || 'disk';
4028 my $oldmedia = $old_drive->{media} || 'disk';
4029 die "unable to change media type\n" if $media ne $oldmedia;
4030
4031 if (!drive_is_cdrom($old_drive)) {
4032
4033 if ($drive->{file} ne $old_drive->{file}) {
4034
4035 die "skip\n" if !$hotplug;
4036
4037 # unplug and register as unused
4038 vm_deviceunplug($vmid, $conf, $opt);
4039 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4040
4041 } else {
4042 # update existing disk
4043
4044 # skip non hotpluggable value
4045 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4046 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4047 die "skip\n";
4048 }
4049
4050 # apply throttle
4051 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4052 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4053 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4054 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4055 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4056 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4057 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4058 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4059 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4060 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4061 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4062 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4063
4064 qemu_block_set_io_throttle($vmid,"drive-$opt",
4065 ($drive->{mbps} || 0)*1024*1024,
4066 ($drive->{mbps_rd} || 0)*1024*1024,
4067 ($drive->{mbps_wr} || 0)*1024*1024,
4068 $drive->{iops} || 0,
4069 $drive->{iops_rd} || 0,
4070 $drive->{iops_wr} || 0,
4071 ($drive->{mbps_max} || 0)*1024*1024,
4072 ($drive->{mbps_rd_max} || 0)*1024*1024,
4073 ($drive->{mbps_wr_max} || 0)*1024*1024,
4074 $drive->{iops_max} || 0,
4075 $drive->{iops_rd_max} || 0,
4076 $drive->{iops_wr_max} || 0);
4077
4078 }
4079
4080 return 1;
4081 }
4082
4083 } else { # cdrom
4084
4085 if ($drive->{file} eq 'none') {
4086 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4087 } else {
4088 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4089 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4090 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4091 }
4092
4093 return 1;
4094 }
4095 }
4096 }
4097
4098 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4099 # hotplug new disks
4100 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4101 }
4102
4103 sub vm_start {
4104 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4105
4106 lock_config($vmid, sub {
4107 my $conf = load_config($vmid, $migratedfrom);
4108
4109 die "you can't start a vm if it's a template\n" if is_template($conf);
4110
4111 check_lock($conf) if !$skiplock;
4112
4113 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4114
4115 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4116 vmconfig_apply_pending($vmid, $conf, $storecfg);
4117 $conf = load_config($vmid); # update/reload
4118 }
4119
4120 my $defaults = load_defaults();
4121
4122 # set environment variable useful inside network script
4123 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4124
4125 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4126
4127 my $migrate_port = 0;
4128 my $migrate_uri;
4129 if ($statefile) {
4130 if ($statefile eq 'tcp') {
4131 my $localip = "localhost";
4132 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4133 if ($datacenterconf->{migration_unsecure}) {
4134 my $nodename = PVE::INotify::nodename();
4135 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4136 }
4137 $migrate_port = PVE::Tools::next_migrate_port();
4138 $migrate_uri = "tcp:${localip}:${migrate_port}";
4139 push @$cmd, '-incoming', $migrate_uri;
4140 push @$cmd, '-S';
4141 } else {
4142 push @$cmd, '-loadstate', $statefile;
4143 }
4144 } elsif ($paused) {
4145 push @$cmd, '-S';
4146 }
4147
4148 # host pci devices
4149 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4150 my $d = parse_hostpci($conf->{"hostpci$i"});
4151 next if !$d;
4152 my $pcidevices = $d->{pciid};
4153 foreach my $pcidevice (@$pcidevices) {
4154 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4155
4156 my $info = pci_device_info("0000:$pciid");
4157 die "IOMMU not present\n" if !check_iommu_support();
4158 die "no pci device info for device '$pciid'\n" if !$info;
4159
4160 if ($d->{driver} && $d->{driver} eq "vfio") {
4161 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4162 } else {
4163 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4164 }
4165
4166 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4167 }
4168 }
4169
4170 PVE::Storage::activate_volumes($storecfg, $vollist);
4171
4172 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4173 umask => 0077); };
4174 my $err = $@;
4175 die "start failed: $err" if $err;
4176
4177 print "migration listens on $migrate_uri\n" if $migrate_uri;
4178
4179 if ($statefile && $statefile ne 'tcp') {
4180 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4181 warn $@ if $@;
4182 }
4183
4184 if ($migratedfrom) {
4185
4186 eval {
4187 set_migration_caps($vmid);
4188 };
4189 warn $@ if $@;
4190
4191 if ($spice_port) {
4192 print "spice listens on port $spice_port\n";
4193 if ($spice_ticket) {
4194 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4195 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4196 }
4197 }
4198
4199 } else {
4200
4201 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4202 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4203 if $conf->{balloon};
4204 }
4205
4206 foreach my $opt (keys %$conf) {
4207 next if $opt !~ m/^net\d+$/;
4208 my $nicconf = parse_net($conf->{$opt});
4209 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4210 }
4211 }
4212
4213 vm_mon_cmd_nocheck($vmid, 'qom-set',
4214 path => "machine/peripheral/balloon0",
4215 property => "guest-stats-polling-interval",
4216 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
4217
4218 });
4219 }
4220
4221 sub vm_mon_cmd {
4222 my ($vmid, $execute, %params) = @_;
4223
4224 my $cmd = { execute => $execute, arguments => \%params };
4225 vm_qmp_command($vmid, $cmd);
4226 }
4227
4228 sub vm_mon_cmd_nocheck {
4229 my ($vmid, $execute, %params) = @_;
4230
4231 my $cmd = { execute => $execute, arguments => \%params };
4232 vm_qmp_command($vmid, $cmd, 1);
4233 }
4234
4235 sub vm_qmp_command {
4236 my ($vmid, $cmd, $nocheck) = @_;
4237
4238 my $res;
4239
4240 my $timeout;
4241 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4242 $timeout = $cmd->{arguments}->{timeout};
4243 delete $cmd->{arguments}->{timeout};
4244 }
4245
4246 eval {
4247 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4248 my $sname = qmp_socket($vmid);
4249 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4250 my $qmpclient = PVE::QMPClient->new();
4251
4252 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4253 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4254 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4255 if scalar(%{$cmd->{arguments}});
4256 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4257 } else {
4258 die "unable to open monitor socket\n";
4259 }
4260 };
4261 if (my $err = $@) {
4262 syslog("err", "VM $vmid qmp command failed - $err");
4263 die $err;
4264 }
4265
4266 return $res;
4267 }
4268
4269 sub vm_human_monitor_command {
4270 my ($vmid, $cmdline) = @_;
4271
4272 my $res;
4273
4274 my $cmd = {
4275 execute => 'human-monitor-command',
4276 arguments => { 'command-line' => $cmdline},
4277 };
4278
4279 return vm_qmp_command($vmid, $cmd);
4280 }
4281
4282 sub vm_commandline {
4283 my ($storecfg, $vmid) = @_;
4284
4285 my $conf = load_config($vmid);
4286
4287 my $defaults = load_defaults();
4288
4289 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4290
4291 return join(' ', @$cmd);
4292 }
4293
4294 sub vm_reset {
4295 my ($vmid, $skiplock) = @_;
4296
4297 lock_config($vmid, sub {
4298
4299 my $conf = load_config($vmid);
4300
4301 check_lock($conf) if !$skiplock;
4302
4303 vm_mon_cmd($vmid, "system_reset");
4304 });
4305 }
4306
4307 sub get_vm_volumes {
4308 my ($conf) = @_;
4309
4310 my $vollist = [];
4311 foreach_volid($conf, sub {
4312 my ($volid, $is_cdrom) = @_;
4313
4314 return if $volid =~ m|^/|;
4315
4316 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4317 return if !$sid;
4318
4319 push @$vollist, $volid;
4320 });
4321
4322 return $vollist;
4323 }
4324
4325 sub vm_stop_cleanup {
4326 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4327
4328 eval {
4329 fairsched_rmnod($vmid); # try to destroy group
4330
4331 if (!$keepActive) {
4332 my $vollist = get_vm_volumes($conf);
4333 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4334 }
4335
4336 foreach my $ext (qw(mon qmp pid vnc qga)) {
4337 unlink "/var/run/qemu-server/${vmid}.$ext";
4338 }
4339
4340 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4341 };
4342 warn $@ if $@; # avoid errors - just warn
4343 }
4344
4345 # Note: use $nockeck to skip tests if VM configuration file exists.
4346 # We need that when migration VMs to other nodes (files already moved)
4347 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4348 sub vm_stop {
4349 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4350
4351 $force = 1 if !defined($force) && !$shutdown;
4352
4353 if ($migratedfrom){
4354 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4355 kill 15, $pid if $pid;
4356 my $conf = load_config($vmid, $migratedfrom);
4357 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4358 return;
4359 }
4360
4361 lock_config($vmid, sub {
4362
4363 my $pid = check_running($vmid, $nocheck);
4364 return if !$pid;
4365
4366 my $conf;
4367 if (!$nocheck) {
4368 $conf = load_config($vmid);
4369 check_lock($conf) if !$skiplock;
4370 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4371 my $opts = parse_startup($conf->{startup});
4372 $timeout = $opts->{down} if $opts->{down};
4373 }
4374 }
4375
4376 $timeout = 60 if !defined($timeout);
4377
4378 eval {
4379 if ($shutdown) {
4380 if (defined($conf) && $conf->{agent}) {
4381 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4382 } else {
4383 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4384 }
4385 } else {
4386 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4387 }
4388 };
4389 my $err = $@;
4390
4391 if (!$err) {
4392 my $count = 0;
4393 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4394 $count++;
4395 sleep 1;
4396 }
4397
4398 if ($count >= $timeout) {
4399 if ($force) {
4400 warn "VM still running - terminating now with SIGTERM\n";
4401 kill 15, $pid;
4402 } else {
4403 die "VM quit/powerdown failed - got timeout\n";
4404 }
4405 } else {
4406 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4407 return;
4408 }
4409 } else {
4410 if ($force) {
4411 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4412 kill 15, $pid;
4413 } else {
4414 die "VM quit/powerdown failed\n";
4415 }
4416 }
4417
4418 # wait again
4419 $timeout = 10;
4420
4421 my $count = 0;
4422 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4423 $count++;
4424 sleep 1;
4425 }
4426
4427 if ($count >= $timeout) {
4428 warn "VM still running - terminating now with SIGKILL\n";
4429 kill 9, $pid;
4430 sleep 1;
4431 }
4432
4433 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4434 });
4435 }
4436
4437 sub vm_suspend {
4438 my ($vmid, $skiplock) = @_;
4439
4440 lock_config($vmid, sub {
4441
4442 my $conf = load_config($vmid);
4443
4444 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4445
4446 vm_mon_cmd($vmid, "stop");
4447 });
4448 }
4449
4450 sub vm_resume {
4451 my ($vmid, $skiplock) = @_;
4452
4453 lock_config($vmid, sub {
4454
4455 my $conf = load_config($vmid);
4456
4457 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4458
4459 vm_mon_cmd($vmid, "cont");
4460 });
4461 }
4462
4463 sub vm_sendkey {
4464 my ($vmid, $skiplock, $key) = @_;
4465
4466 lock_config($vmid, sub {
4467
4468 my $conf = load_config($vmid);
4469
4470 # there is no qmp command, so we use the human monitor command
4471 vm_human_monitor_command($vmid, "sendkey $key");
4472 });
4473 }
4474
4475 sub vm_destroy {
4476 my ($storecfg, $vmid, $skiplock) = @_;
4477
4478 lock_config($vmid, sub {
4479
4480 my $conf = load_config($vmid);
4481
4482 check_lock($conf) if !$skiplock;
4483
4484 if (!check_running($vmid)) {
4485 fairsched_rmnod($vmid); # try to destroy group
4486 destroy_vm($storecfg, $vmid);
4487 } else {
4488 die "VM $vmid is running - destroy failed\n";
4489 }
4490 });
4491 }
4492
4493 # pci helpers
4494
4495 sub file_write {
4496 my ($filename, $buf) = @_;
4497
4498 my $fh = IO::File->new($filename, "w");
4499 return undef if !$fh;
4500
4501 my $res = print $fh $buf;
4502
4503 $fh->close();
4504
4505 return $res;
4506 }
4507
4508 sub pci_device_info {
4509 my ($name) = @_;
4510
4511 my $res;
4512
4513 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4514 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4515
4516 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4517 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4518
4519 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4520 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4521
4522 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4523 return undef if !defined($product) || $product !~ s/^0x//;
4524
4525 $res = {
4526 name => $name,
4527 vendor => $vendor,
4528 product => $product,
4529 domain => $domain,
4530 bus => $bus,
4531 slot => $slot,
4532 func => $func,
4533 irq => $irq,
4534 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4535 };
4536
4537 return $res;
4538 }
4539
4540 sub pci_dev_reset {
4541 my ($dev) = @_;
4542
4543 my $name = $dev->{name};
4544
4545 my $fn = "$pcisysfs/devices/$name/reset";
4546
4547 return file_write($fn, "1");
4548 }
4549
4550 sub pci_dev_bind_to_stub {
4551 my ($dev) = @_;
4552
4553 my $name = $dev->{name};
4554
4555 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4556 return 1 if -d $testdir;
4557
4558 my $data = "$dev->{vendor} $dev->{product}";
4559 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4560
4561 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4562 if (!file_write($fn, $name)) {
4563 return undef if -f $fn;
4564 }
4565
4566 $fn = "$pcisysfs/drivers/pci-stub/bind";
4567 if (! -d $testdir) {
4568 return undef if !file_write($fn, $name);
4569 }
4570
4571 return -d $testdir;
4572 }
4573
4574 sub pci_dev_bind_to_vfio {
4575 my ($dev) = @_;
4576
4577 my $name = $dev->{name};
4578
4579 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4580
4581 if (!-d $vfio_basedir) {
4582 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4583 }
4584 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4585
4586 my $testdir = "$vfio_basedir/$name";
4587 return 1 if -d $testdir;
4588
4589 my $data = "$dev->{vendor} $dev->{product}";
4590 return undef if !file_write("$vfio_basedir/new_id", $data);
4591
4592 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4593 if (!file_write($fn, $name)) {
4594 return undef if -f $fn;
4595 }
4596
4597 $fn = "$vfio_basedir/bind";
4598 if (! -d $testdir) {
4599 return undef if !file_write($fn, $name);
4600 }
4601
4602 return -d $testdir;
4603 }
4604
4605 sub pci_dev_group_bind_to_vfio {
4606 my ($pciid) = @_;
4607
4608 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4609
4610 if (!-d $vfio_basedir) {
4611 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4612 }
4613 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4614
4615 # get IOMMU group devices
4616 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4617 my @devs = grep /^0000:/, readdir($D);
4618 closedir($D);
4619
4620 foreach my $pciid (@devs) {
4621 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4622
4623 # pci bridges, switches or root ports are not supported
4624 # they have a pci_bus subdirectory so skip them
4625 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4626
4627 my $info = pci_device_info($1);
4628 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4629 }
4630
4631 return 1;
4632 }
4633
4634 sub print_pci_addr {
4635 my ($id, $bridges) = @_;
4636
4637 my $res = '';
4638 my $devices = {
4639 piix3 => { bus => 0, addr => 1 },
4640 #addr2 : first videocard
4641 balloon0 => { bus => 0, addr => 3 },
4642 watchdog => { bus => 0, addr => 4 },
4643 scsihw0 => { bus => 0, addr => 5 },
4644 scsihw1 => { bus => 0, addr => 6 },
4645 ahci0 => { bus => 0, addr => 7 },
4646 qga0 => { bus => 0, addr => 8 },
4647 spice => { bus => 0, addr => 9 },
4648 virtio0 => { bus => 0, addr => 10 },
4649 virtio1 => { bus => 0, addr => 11 },
4650 virtio2 => { bus => 0, addr => 12 },
4651 virtio3 => { bus => 0, addr => 13 },
4652 virtio4 => { bus => 0, addr => 14 },
4653 virtio5 => { bus => 0, addr => 15 },
4654 hostpci0 => { bus => 0, addr => 16 },
4655 hostpci1 => { bus => 0, addr => 17 },
4656 net0 => { bus => 0, addr => 18 },
4657 net1 => { bus => 0, addr => 19 },
4658 net2 => { bus => 0, addr => 20 },
4659 net3 => { bus => 0, addr => 21 },
4660 net4 => { bus => 0, addr => 22 },
4661 net5 => { bus => 0, addr => 23 },
4662 vga1 => { bus => 0, addr => 24 },
4663 vga2 => { bus => 0, addr => 25 },
4664 vga3 => { bus => 0, addr => 26 },
4665 hostpci2 => { bus => 0, addr => 27 },
4666 hostpci3 => { bus => 0, addr => 28 },
4667 #addr29 : usb-host (pve-usb.cfg)
4668 'pci.1' => { bus => 0, addr => 30 },
4669 'pci.2' => { bus => 0, addr => 31 },
4670 'net6' => { bus => 1, addr => 1 },
4671 'net7' => { bus => 1, addr => 2 },
4672 'net8' => { bus => 1, addr => 3 },
4673 'net9' => { bus => 1, addr => 4 },
4674 'net10' => { bus => 1, addr => 5 },
4675 'net11' => { bus => 1, addr => 6 },
4676 'net12' => { bus => 1, addr => 7 },
4677 'net13' => { bus => 1, addr => 8 },
4678 'net14' => { bus => 1, addr => 9 },
4679 'net15' => { bus => 1, addr => 10 },
4680 'net16' => { bus => 1, addr => 11 },
4681 'net17' => { bus => 1, addr => 12 },
4682 'net18' => { bus => 1, addr => 13 },
4683 'net19' => { bus => 1, addr => 14 },
4684 'net20' => { bus => 1, addr => 15 },
4685 'net21' => { bus => 1, addr => 16 },
4686 'net22' => { bus => 1, addr => 17 },
4687 'net23' => { bus => 1, addr => 18 },
4688 'net24' => { bus => 1, addr => 19 },
4689 'net25' => { bus => 1, addr => 20 },
4690 'net26' => { bus => 1, addr => 21 },
4691 'net27' => { bus => 1, addr => 22 },
4692 'net28' => { bus => 1, addr => 23 },
4693 'net29' => { bus => 1, addr => 24 },
4694 'net30' => { bus => 1, addr => 25 },
4695 'net31' => { bus => 1, addr => 26 },
4696 'virtio6' => { bus => 2, addr => 1 },
4697 'virtio7' => { bus => 2, addr => 2 },
4698 'virtio8' => { bus => 2, addr => 3 },
4699 'virtio9' => { bus => 2, addr => 4 },
4700 'virtio10' => { bus => 2, addr => 5 },
4701 'virtio11' => { bus => 2, addr => 6 },
4702 'virtio12' => { bus => 2, addr => 7 },
4703 'virtio13' => { bus => 2, addr => 8 },
4704 'virtio14' => { bus => 2, addr => 9 },
4705 'virtio15' => { bus => 2, addr => 10 },
4706 };
4707
4708 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4709 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4710 my $bus = $devices->{$id}->{bus};
4711 $res = ",bus=pci.$bus,addr=$addr";
4712 $bridges->{$bus} = 1 if $bridges;
4713 }
4714 return $res;
4715
4716 }
4717
4718 sub print_pcie_addr {
4719 my ($id) = @_;
4720
4721 my $res = '';
4722 my $devices = {
4723 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4724 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4725 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4726 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4727 };
4728
4729 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4730 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4731 my $bus = $devices->{$id}->{bus};
4732 $res = ",bus=$bus,addr=$addr";
4733 }
4734 return $res;
4735
4736 }
4737
4738 # vzdump restore implementaion
4739
4740 sub tar_archive_read_firstfile {
4741 my $archive = shift;
4742
4743 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4744
4745 # try to detect archive type first
4746 my $pid = open (TMP, "tar tf '$archive'|") ||
4747 die "unable to open file '$archive'\n";
4748 my $firstfile = <TMP>;
4749 kill 15, $pid;
4750 close TMP;
4751
4752 die "ERROR: archive contaions no data\n" if !$firstfile;
4753 chomp $firstfile;
4754
4755 return $firstfile;
4756 }
4757
4758 sub tar_restore_cleanup {
4759 my ($storecfg, $statfile) = @_;
4760
4761 print STDERR "starting cleanup\n";
4762
4763 if (my $fd = IO::File->new($statfile, "r")) {
4764 while (defined(my $line = <$fd>)) {
4765 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4766 my $volid = $2;
4767 eval {
4768 if ($volid =~ m|^/|) {
4769 unlink $volid || die 'unlink failed\n';
4770 } else {
4771 PVE::Storage::vdisk_free($storecfg, $volid);
4772 }
4773 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4774 };
4775 print STDERR "unable to cleanup '$volid' - $@" if $@;
4776 } else {
4777 print STDERR "unable to parse line in statfile - $line";
4778 }
4779 }
4780 $fd->close();
4781 }
4782 }
4783
4784 sub restore_archive {
4785 my ($archive, $vmid, $user, $opts) = @_;
4786
4787 my $format = $opts->{format};
4788 my $comp;
4789
4790 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4791 $format = 'tar' if !$format;
4792 $comp = 'gzip';
4793 } elsif ($archive =~ m/\.tar$/) {
4794 $format = 'tar' if !$format;
4795 } elsif ($archive =~ m/.tar.lzo$/) {
4796 $format = 'tar' if !$format;
4797 $comp = 'lzop';
4798 } elsif ($archive =~ m/\.vma$/) {
4799 $format = 'vma' if !$format;
4800 } elsif ($archive =~ m/\.vma\.gz$/) {
4801 $format = 'vma' if !$format;
4802 $comp = 'gzip';
4803 } elsif ($archive =~ m/\.vma\.lzo$/) {
4804 $format = 'vma' if !$format;
4805 $comp = 'lzop';
4806 } else {
4807 $format = 'vma' if !$format; # default
4808 }
4809
4810 # try to detect archive format
4811 if ($format eq 'tar') {
4812 return restore_tar_archive($archive, $vmid, $user, $opts);
4813 } else {
4814 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4815 }
4816 }
4817
4818 sub restore_update_config_line {
4819 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4820
4821 return if $line =~ m/^\#qmdump\#/;
4822 return if $line =~ m/^\#vzdump\#/;
4823 return if $line =~ m/^lock:/;
4824 return if $line =~ m/^unused\d+:/;
4825 return if $line =~ m/^parent:/;
4826 return if $line =~ m/^template:/; # restored VM is never a template
4827
4828 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4829 # try to convert old 1.X settings
4830 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4831 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4832 my ($model, $macaddr) = split(/\=/, $devconfig);
4833 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4834 my $net = {
4835 model => $model,
4836 bridge => "vmbr$ind",
4837 macaddr => $macaddr,
4838 };
4839 my $netstr = print_net($net);
4840
4841 print $outfd "net$cookie->{netcount}: $netstr\n";
4842 $cookie->{netcount}++;
4843 }
4844 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4845 my ($id, $netstr) = ($1, $2);
4846 my $net = parse_net($netstr);
4847 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4848 $netstr = print_net($net);
4849 print $outfd "$id: $netstr\n";
4850 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4851 my $virtdev = $1;
4852 my $value = $3;
4853 if ($line =~ m/backup=no/) {
4854 print $outfd "#$line";
4855 } elsif ($virtdev && $map->{$virtdev}) {
4856 my $di = parse_drive($virtdev, $value);
4857 delete $di->{format}; # format can change on restore
4858 $di->{file} = $map->{$virtdev};
4859 $value = print_drive($vmid, $di);
4860 print $outfd "$virtdev: $value\n";
4861 } else {
4862 print $outfd $line;
4863 }
4864 } else {
4865 print $outfd $line;
4866 }
4867 }
4868
4869 sub scan_volids {
4870 my ($cfg, $vmid) = @_;
4871
4872 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4873
4874 my $volid_hash = {};
4875 foreach my $storeid (keys %$info) {
4876 foreach my $item (@{$info->{$storeid}}) {
4877 next if !($item->{volid} && $item->{size});
4878 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4879 $volid_hash->{$item->{volid}} = $item;
4880 }
4881 }
4882
4883 return $volid_hash;
4884 }
4885
4886 sub get_used_paths {
4887 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4888
4889 my $used_path = {};
4890
4891 my $scan_config = sub {
4892 my ($cref, $snapname) = @_;
4893
4894 foreach my $key (keys %$cref) {
4895 my $value = $cref->{$key};
4896 if (valid_drivename($key)) {
4897 next if $skip_drive && $key eq $skip_drive;
4898 my $drive = parse_drive($key, $value);
4899 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4900 if ($drive->{file} =~ m!^/!) {
4901 $used_path->{$drive->{file}}++; # = 1;
4902 } else {
4903 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4904 next if !$storeid;
4905 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4906 next if !$scfg;
4907 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4908 $used_path->{$path}++; # = 1;
4909 }
4910 }
4911 }
4912 };
4913
4914 &$scan_config($conf);
4915
4916 undef $skip_drive;
4917
4918 if ($scan_snapshots) {
4919 foreach my $snapname (keys %{$conf->{snapshots}}) {
4920 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4921 }
4922 }
4923
4924 return $used_path;
4925 }
4926
4927 sub update_disksize {
4928 my ($vmid, $conf, $volid_hash) = @_;
4929
4930 my $changes;
4931
4932 my $used = {};
4933
4934 # Note: it is allowed to define multiple storages with same path (alias), so
4935 # we need to check both 'volid' and real 'path' (two different volid can point
4936 # to the same path).
4937
4938 my $usedpath = {};
4939
4940 # update size info
4941 foreach my $opt (keys %$conf) {
4942 if (valid_drivename($opt)) {
4943 my $drive = parse_drive($opt, $conf->{$opt});
4944 my $volid = $drive->{file};
4945 next if !$volid;
4946
4947 $used->{$volid} = 1;
4948 if ($volid_hash->{$volid} &&
4949 (my $path = $volid_hash->{$volid}->{path})) {
4950 $usedpath->{$path} = 1;
4951 }
4952
4953 next if drive_is_cdrom($drive);
4954 next if !$volid_hash->{$volid};
4955
4956 $drive->{size} = $volid_hash->{$volid}->{size};
4957 my $new = print_drive($vmid, $drive);
4958 if ($new ne $conf->{$opt}) {
4959 $changes = 1;
4960 $conf->{$opt} = $new;
4961 }
4962 }
4963 }
4964
4965 # remove 'unusedX' entry if volume is used
4966 foreach my $opt (keys %$conf) {
4967 next if $opt !~ m/^unused\d+$/;
4968 my $volid = $conf->{$opt};
4969 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4970 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4971 $changes = 1;
4972 delete $conf->{$opt};
4973 }
4974 }
4975
4976 foreach my $volid (sort keys %$volid_hash) {
4977 next if $volid =~ m/vm-$vmid-state-/;
4978 next if $used->{$volid};
4979 my $path = $volid_hash->{$volid}->{path};
4980 next if !$path; # just to be sure
4981 next if $usedpath->{$path};
4982 $changes = 1;
4983 add_unused_volume($conf, $volid);
4984 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4985 }
4986
4987 return $changes;
4988 }
4989
4990 sub rescan {
4991 my ($vmid, $nolock) = @_;
4992
4993 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4994
4995 my $volid_hash = scan_volids($cfg, $vmid);
4996
4997 my $updatefn = sub {
4998 my ($vmid) = @_;
4999
5000 my $conf = load_config($vmid);
5001
5002 check_lock($conf);
5003
5004 my $vm_volids = {};
5005 foreach my $volid (keys %$volid_hash) {
5006 my $info = $volid_hash->{$volid};
5007 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5008 }
5009
5010 my $changes = update_disksize($vmid, $conf, $vm_volids);
5011
5012 update_config_nolock($vmid, $conf, 1) if $changes;
5013 };
5014
5015 if (defined($vmid)) {
5016 if ($nolock) {
5017 &$updatefn($vmid);
5018 } else {
5019 lock_config($vmid, $updatefn, $vmid);
5020 }
5021 } else {
5022 my $vmlist = config_list();
5023 foreach my $vmid (keys %$vmlist) {
5024 if ($nolock) {
5025 &$updatefn($vmid);
5026 } else {
5027 lock_config($vmid, $updatefn, $vmid);
5028 }
5029 }
5030 }
5031 }
5032
5033 sub restore_vma_archive {
5034 my ($archive, $vmid, $user, $opts, $comp) = @_;
5035
5036 my $input = $archive eq '-' ? "<&STDIN" : undef;
5037 my $readfrom = $archive;
5038
5039 my $uncomp = '';
5040 if ($comp) {
5041 $readfrom = '-';
5042 my $qarchive = PVE::Tools::shellquote($archive);
5043 if ($comp eq 'gzip') {
5044 $uncomp = "zcat $qarchive|";
5045 } elsif ($comp eq 'lzop') {
5046 $uncomp = "lzop -d -c $qarchive|";
5047 } else {
5048 die "unknown compression method '$comp'\n";
5049 }
5050
5051 }
5052
5053 my $tmpdir = "/var/tmp/vzdumptmp$$";
5054 rmtree $tmpdir;
5055
5056 # disable interrupts (always do cleanups)
5057 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5058 warn "got interrupt - ignored\n";
5059 };
5060
5061 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5062 POSIX::mkfifo($mapfifo, 0600);
5063 my $fifofh;
5064
5065 my $openfifo = sub {
5066 open($fifofh, '>', $mapfifo) || die $!;
5067 };
5068
5069 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5070
5071 my $oldtimeout;
5072 my $timeout = 5;
5073
5074 my $devinfo = {};
5075
5076 my $rpcenv = PVE::RPCEnvironment::get();
5077
5078 my $conffile = config_file($vmid);
5079 my $tmpfn = "$conffile.$$.tmp";
5080
5081 # Note: $oldconf is undef if VM does not exists
5082 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5083
5084 my $print_devmap = sub {
5085 my $virtdev_hash = {};
5086
5087 my $cfgfn = "$tmpdir/qemu-server.conf";
5088
5089 # we can read the config - that is already extracted
5090 my $fh = IO::File->new($cfgfn, "r") ||
5091 "unable to read qemu-server.conf - $!\n";
5092
5093 while (defined(my $line = <$fh>)) {
5094 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5095 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5096 die "archive does not contain data for drive '$virtdev'\n"
5097 if !$devinfo->{$devname};
5098 if (defined($opts->{storage})) {
5099 $storeid = $opts->{storage} || 'local';
5100 } elsif (!$storeid) {
5101 $storeid = 'local';
5102 }
5103 $format = 'raw' if !$format;
5104 $devinfo->{$devname}->{devname} = $devname;
5105 $devinfo->{$devname}->{virtdev} = $virtdev;
5106 $devinfo->{$devname}->{format} = $format;
5107 $devinfo->{$devname}->{storeid} = $storeid;
5108
5109 # check permission on storage
5110 my $pool = $opts->{pool}; # todo: do we need that?
5111 if ($user ne 'root@pam') {
5112 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5113 }
5114
5115 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5116 }
5117 }
5118
5119 foreach my $devname (keys %$devinfo) {
5120 die "found no device mapping information for device '$devname'\n"
5121 if !$devinfo->{$devname}->{virtdev};
5122 }
5123
5124 my $cfg = cfs_read_file('storage.cfg');
5125
5126 # create empty/temp config
5127 if ($oldconf) {
5128 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5129 foreach_drive($oldconf, sub {
5130 my ($ds, $drive) = @_;
5131
5132 return if drive_is_cdrom($drive);
5133
5134 my $volid = $drive->{file};
5135
5136 return if !$volid || $volid =~ m|^/|;
5137
5138 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5139 return if !$path || !$owner || ($owner != $vmid);
5140
5141 # Note: only delete disk we want to restore
5142 # other volumes will become unused
5143 if ($virtdev_hash->{$ds}) {
5144 PVE::Storage::vdisk_free($cfg, $volid);
5145 }
5146 });
5147 }
5148
5149 my $map = {};
5150 foreach my $virtdev (sort keys %$virtdev_hash) {
5151 my $d = $virtdev_hash->{$virtdev};
5152 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5153 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5154
5155 # test if requested format is supported
5156 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5157 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5158 $d->{format} = $defFormat if !$supported;
5159
5160 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5161 $d->{format}, undef, $alloc_size);
5162 print STDERR "new volume ID is '$volid'\n";
5163 $d->{volid} = $volid;
5164 my $path = PVE::Storage::path($cfg, $volid);
5165
5166 my $write_zeros = 1;
5167 # fixme: what other storages types initialize volumes with zero?
5168 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5169 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5170 $write_zeros = 0;
5171 }
5172
5173 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5174
5175 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5176 $map->{$virtdev} = $volid;
5177 }
5178
5179 $fh->seek(0, 0) || die "seek failed - $!\n";
5180
5181 my $outfd = new IO::File ($tmpfn, "w") ||
5182 die "unable to write config for VM $vmid\n";
5183
5184 my $cookie = { netcount => 0 };
5185 while (defined(my $line = <$fh>)) {
5186 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5187 }
5188
5189 $fh->close();
5190 $outfd->close();
5191 };
5192
5193 eval {
5194 # enable interrupts
5195 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5196 die "interrupted by signal\n";
5197 };
5198 local $SIG{ALRM} = sub { die "got timeout\n"; };
5199
5200 $oldtimeout = alarm($timeout);
5201
5202 my $parser = sub {
5203 my $line = shift;
5204
5205 print "$line\n";
5206
5207 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5208 my ($dev_id, $size, $devname) = ($1, $2, $3);
5209 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5210 } elsif ($line =~ m/^CTIME: /) {
5211 # we correctly received the vma config, so we can disable
5212 # the timeout now for disk allocation (set to 10 minutes, so
5213 # that we always timeout if something goes wrong)
5214 alarm(600);
5215 &$print_devmap();
5216 print $fifofh "done\n";
5217 my $tmp = $oldtimeout || 0;
5218 $oldtimeout = undef;
5219 alarm($tmp);
5220 close($fifofh);
5221 }
5222 };
5223
5224 print "restore vma archive: $cmd\n";
5225 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5226 };
5227 my $err = $@;
5228
5229 alarm($oldtimeout) if $oldtimeout;
5230
5231 unlink $mapfifo;
5232
5233 if ($err) {
5234 rmtree $tmpdir;
5235 unlink $tmpfn;
5236
5237 my $cfg = cfs_read_file('storage.cfg');
5238 foreach my $devname (keys %$devinfo) {
5239 my $volid = $devinfo->{$devname}->{volid};
5240 next if !$volid;
5241 eval {
5242 if ($volid =~ m|^/|) {
5243 unlink $volid || die 'unlink failed\n';
5244 } else {
5245 PVE::Storage::vdisk_free($cfg, $volid);
5246 }
5247 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5248 };
5249 print STDERR "unable to cleanup '$volid' - $@" if $@;
5250 }
5251 die $err;
5252 }
5253
5254 rmtree $tmpdir;
5255
5256 rename($tmpfn, $conffile) ||
5257 die "unable to commit configuration file '$conffile'\n";
5258
5259 PVE::Cluster::cfs_update(); # make sure we read new file
5260
5261 eval { rescan($vmid, 1); };
5262 warn $@ if $@;
5263 }
5264
5265 sub restore_tar_archive {
5266 my ($archive, $vmid, $user, $opts) = @_;
5267
5268 if ($archive ne '-') {
5269 my $firstfile = tar_archive_read_firstfile($archive);
5270 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5271 if $firstfile ne 'qemu-server.conf';
5272 }
5273
5274 my $storecfg = cfs_read_file('storage.cfg');
5275
5276 # destroy existing data - keep empty config
5277 my $vmcfgfn = config_file($vmid);
5278 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5279
5280 my $tocmd = "/usr/lib/qemu-server/qmextract";
5281
5282 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5283 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5284 $tocmd .= ' --prealloc' if $opts->{prealloc};
5285 $tocmd .= ' --info' if $opts->{info};
5286
5287 # tar option "xf" does not autodetect compression when read from STDIN,
5288 # so we pipe to zcat
5289 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5290 PVE::Tools::shellquote("--to-command=$tocmd");
5291
5292 my $tmpdir = "/var/tmp/vzdumptmp$$";
5293 mkpath $tmpdir;
5294
5295 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5296 local $ENV{VZDUMP_VMID} = $vmid;
5297 local $ENV{VZDUMP_USER} = $user;
5298
5299 my $conffile = config_file($vmid);
5300 my $tmpfn = "$conffile.$$.tmp";
5301
5302 # disable interrupts (always do cleanups)
5303 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5304 print STDERR "got interrupt - ignored\n";
5305 };
5306
5307 eval {
5308 # enable interrupts
5309 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5310 die "interrupted by signal\n";
5311 };
5312
5313 if ($archive eq '-') {
5314 print "extracting archive from STDIN\n";
5315 run_command($cmd, input => "<&STDIN");
5316 } else {
5317 print "extracting archive '$archive'\n";
5318 run_command($cmd);
5319 }
5320
5321 return if $opts->{info};
5322
5323 # read new mapping
5324 my $map = {};
5325 my $statfile = "$tmpdir/qmrestore.stat";
5326 if (my $fd = IO::File->new($statfile, "r")) {
5327 while (defined (my $line = <$fd>)) {
5328 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5329 $map->{$1} = $2 if $1;
5330 } else {
5331 print STDERR "unable to parse line in statfile - $line\n";
5332 }
5333 }
5334 $fd->close();
5335 }
5336
5337 my $confsrc = "$tmpdir/qemu-server.conf";
5338
5339 my $srcfd = new IO::File($confsrc, "r") ||
5340 die "unable to open file '$confsrc'\n";
5341
5342 my $outfd = new IO::File ($tmpfn, "w") ||
5343 die "unable to write config for VM $vmid\n";
5344
5345 my $cookie = { netcount => 0 };
5346 while (defined (my $line = <$srcfd>)) {
5347 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5348 }
5349
5350 $srcfd->close();
5351 $outfd->close();
5352 };
5353 my $err = $@;
5354
5355 if ($err) {
5356
5357 unlink $tmpfn;
5358
5359 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5360
5361 die $err;
5362 }
5363
5364 rmtree $tmpdir;
5365
5366 rename $tmpfn, $conffile ||
5367 die "unable to commit configuration file '$conffile'\n";
5368
5369 PVE::Cluster::cfs_update(); # make sure we read new file
5370
5371 eval { rescan($vmid, 1); };
5372 warn $@ if $@;
5373 };
5374
5375
5376 # Internal snapshots
5377
5378 # NOTE: Snapshot create/delete involves several non-atomic
5379 # action, and can take a long time.
5380 # So we try to avoid locking the file and use 'lock' variable
5381 # inside the config file instead.
5382
5383 my $snapshot_copy_config = sub {
5384 my ($source, $dest) = @_;
5385
5386 foreach my $k (keys %$source) {
5387 next if $k eq 'snapshots';
5388 next if $k eq 'snapstate';
5389 next if $k eq 'snaptime';
5390 next if $k eq 'vmstate';
5391 next if $k eq 'lock';
5392 next if $k eq 'digest';
5393 next if $k eq 'description';
5394 next if $k =~ m/^unused\d+$/;
5395
5396 $dest->{$k} = $source->{$k};
5397 }
5398 };
5399
5400 my $snapshot_apply_config = sub {
5401 my ($conf, $snap) = @_;
5402
5403 # copy snapshot list
5404 my $newconf = {
5405 snapshots => $conf->{snapshots},
5406 };
5407
5408 # keep description and list of unused disks
5409 foreach my $k (keys %$conf) {
5410 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5411 $newconf->{$k} = $conf->{$k};
5412 }
5413
5414 &$snapshot_copy_config($snap, $newconf);
5415
5416 return $newconf;
5417 };
5418
5419 sub foreach_writable_storage {
5420 my ($conf, $func) = @_;
5421
5422 my $sidhash = {};
5423
5424 foreach my $ds (keys %$conf) {
5425 next if !valid_drivename($ds);
5426
5427 my $drive = parse_drive($ds, $conf->{$ds});
5428 next if !$drive;
5429 next if drive_is_cdrom($drive);
5430
5431 my $volid = $drive->{file};
5432
5433 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5434 $sidhash->{$sid} = $sid if $sid;
5435 }
5436
5437 foreach my $sid (sort keys %$sidhash) {
5438 &$func($sid);
5439 }
5440 }
5441
5442 my $alloc_vmstate_volid = sub {
5443 my ($storecfg, $vmid, $conf, $snapname) = @_;
5444
5445 # Note: we try to be smart when selecting a $target storage
5446
5447 my $target;
5448
5449 # search shared storage first
5450 foreach_writable_storage($conf, sub {
5451 my ($sid) = @_;
5452 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5453 return if !$scfg->{shared};
5454
5455 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5456 });
5457
5458 if (!$target) {
5459 # now search local storage
5460 foreach_writable_storage($conf, sub {
5461 my ($sid) = @_;
5462 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5463 return if $scfg->{shared};
5464
5465 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5466 });
5467 }
5468
5469 $target = 'local' if !$target;
5470
5471 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5472 # we abort live save after $conf->{memory}, so we need at max twice that space
5473 my $size = $conf->{memory}*2 + $driver_state_size;
5474
5475 my $name = "vm-$vmid-state-$snapname";
5476 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5477 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5478 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5479
5480 return $volid;
5481 };
5482
5483 my $snapshot_prepare = sub {
5484 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5485
5486 my $snap;
5487
5488 my $updatefn = sub {
5489
5490 my $conf = load_config($vmid);
5491
5492 die "you can't take a snapshot if it's a template\n"
5493 if is_template($conf);
5494
5495 check_lock($conf);
5496
5497 $conf->{lock} = 'snapshot';
5498
5499 die "snapshot name '$snapname' already used\n"
5500 if defined($conf->{snapshots}->{$snapname});
5501
5502 my $storecfg = PVE::Storage::config();
5503 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5504
5505 $snap = $conf->{snapshots}->{$snapname} = {};
5506
5507 if ($save_vmstate && check_running($vmid)) {
5508 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5509 }
5510
5511 &$snapshot_copy_config($conf, $snap);
5512
5513 $snap->{snapstate} = "prepare";
5514 $snap->{snaptime} = time();
5515 $snap->{description} = $comment if $comment;
5516
5517 # always overwrite machine if we save vmstate. This makes sure we
5518 # can restore it later using correct machine type
5519 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5520
5521 update_config_nolock($vmid, $conf, 1);
5522 };
5523
5524 lock_config($vmid, $updatefn);
5525
5526 return $snap;
5527 };
5528
5529 my $snapshot_commit = sub {
5530 my ($vmid, $snapname) = @_;
5531
5532 my $updatefn = sub {
5533
5534 my $conf = load_config($vmid);
5535
5536 die "missing snapshot lock\n"
5537 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5538
5539 my $has_machine_config = defined($conf->{machine});
5540
5541 my $snap = $conf->{snapshots}->{$snapname};
5542
5543 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5544
5545 die "wrong snapshot state\n"
5546 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5547
5548 delete $snap->{snapstate};
5549 delete $conf->{lock};
5550
5551 my $newconf = &$snapshot_apply_config($conf, $snap);
5552
5553 delete $newconf->{machine} if !$has_machine_config;
5554
5555 $newconf->{parent} = $snapname;
5556
5557 update_config_nolock($vmid, $newconf, 1);
5558 };
5559
5560 lock_config($vmid, $updatefn);
5561 };
5562
5563 sub snapshot_rollback {
5564 my ($vmid, $snapname) = @_;
5565
5566 my $prepare = 1;
5567
5568 my $storecfg = PVE::Storage::config();
5569
5570 my $conf = load_config($vmid);
5571
5572 my $get_snapshot_config = sub {
5573
5574 die "you can't rollback if vm is a template\n" if is_template($conf);
5575
5576 my $res = $conf->{snapshots}->{$snapname};
5577
5578 die "snapshot '$snapname' does not exist\n" if !defined($res);
5579
5580 return $res;
5581 };
5582
5583 my $snap = &$get_snapshot_config();
5584
5585 foreach_drive($snap, sub {
5586 my ($ds, $drive) = @_;
5587
5588 return if drive_is_cdrom($drive);
5589
5590 my $volid = $drive->{file};
5591
5592 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5593 });
5594
5595 my $updatefn = sub {
5596
5597 $conf = load_config($vmid);
5598
5599 $snap = &$get_snapshot_config();
5600
5601 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5602 if $snap->{snapstate};
5603
5604 if ($prepare) {
5605 check_lock($conf);
5606 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5607 }
5608
5609 die "unable to rollback vm $vmid: vm is running\n"
5610 if check_running($vmid);
5611
5612 if ($prepare) {
5613 $conf->{lock} = 'rollback';
5614 } else {
5615 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5616 delete $conf->{lock};
5617 }
5618
5619 my $forcemachine;
5620
5621 if (!$prepare) {
5622 my $has_machine_config = defined($conf->{machine});
5623
5624 # copy snapshot config to current config
5625 $conf = &$snapshot_apply_config($conf, $snap);
5626 $conf->{parent} = $snapname;
5627
5628 # Note: old code did not store 'machine', so we try to be smart
5629 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5630 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5631 # we remove the 'machine' configuration if not explicitly specified
5632 # in the original config.
5633 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5634 }
5635
5636 update_config_nolock($vmid, $conf, 1);
5637
5638 if (!$prepare && $snap->{vmstate}) {
5639 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5640 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5641 }
5642 };
5643
5644 lock_config($vmid, $updatefn);
5645
5646 foreach_drive($snap, sub {
5647 my ($ds, $drive) = @_;
5648
5649 return if drive_is_cdrom($drive);
5650
5651 my $volid = $drive->{file};
5652 my $device = "drive-$ds";
5653
5654 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5655 });
5656
5657 $prepare = 0;
5658 lock_config($vmid, $updatefn);
5659 }
5660
5661 my $savevm_wait = sub {
5662 my ($vmid) = @_;
5663
5664 for(;;) {
5665 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5666 if (!$stat->{status}) {
5667 die "savevm not active\n";
5668 } elsif ($stat->{status} eq 'active') {
5669 sleep(1);
5670 next;
5671 } elsif ($stat->{status} eq 'completed') {
5672 last;
5673 } else {
5674 die "query-savevm returned status '$stat->{status}'\n";
5675 }
5676 }
5677 };
5678
5679 sub snapshot_create {
5680 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5681
5682 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5683
5684 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5685
5686 my $config = load_config($vmid);
5687
5688 my $running = check_running($vmid);
5689
5690 my $freezefs = $running && $config->{agent};
5691 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5692
5693 my $drivehash = {};
5694
5695 if ($freezefs) {
5696 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5697 warn "guest-fsfreeze-freeze problems - $@" if $@;
5698 }
5699
5700 eval {
5701 # create internal snapshots of all drives
5702
5703 my $storecfg = PVE::Storage::config();
5704
5705 if ($running) {
5706 if ($snap->{vmstate}) {
5707 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5708 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5709 &$savevm_wait($vmid);
5710 } else {
5711 vm_mon_cmd($vmid, "savevm-start");
5712 }
5713 };
5714
5715 foreach_drive($snap, sub {
5716 my ($ds, $drive) = @_;
5717
5718 return if drive_is_cdrom($drive);
5719
5720 my $volid = $drive->{file};
5721 my $device = "drive-$ds";
5722
5723 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5724 $drivehash->{$ds} = 1;
5725 });
5726 };
5727 my $err = $@;
5728
5729 if ($running) {
5730 eval { vm_mon_cmd($vmid, "savevm-end") };
5731 warn $@ if $@;
5732
5733 if ($freezefs) {
5734 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5735 warn "guest-fsfreeze-thaw problems - $@" if $@;
5736 }
5737
5738 # savevm-end is async, we need to wait
5739 for (;;) {
5740 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5741 if (!$stat->{bytes}) {
5742 last;
5743 } else {
5744 print "savevm not yet finished\n";
5745 sleep(1);
5746 next;
5747 }
5748 }
5749 }
5750
5751 if ($err) {
5752 warn "snapshot create failed: starting cleanup\n";
5753 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5754 warn $@ if $@;
5755 die $err;
5756 }
5757
5758 &$snapshot_commit($vmid, $snapname);
5759 }
5760
5761 # Note: $drivehash is only set when called from snapshot_create.
5762 sub snapshot_delete {
5763 my ($vmid, $snapname, $force, $drivehash) = @_;
5764
5765 my $prepare = 1;
5766
5767 my $snap;
5768 my $unused = [];
5769
5770 my $unlink_parent = sub {
5771 my ($confref, $new_parent) = @_;
5772
5773 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5774 if ($new_parent) {
5775 $confref->{parent} = $new_parent;
5776 } else {
5777 delete $confref->{parent};
5778 }
5779 }
5780 };
5781
5782 my $updatefn = sub {
5783 my ($remove_drive) = @_;
5784
5785 my $conf = load_config($vmid);
5786
5787 if (!$drivehash) {
5788 check_lock($conf);
5789 die "you can't delete a snapshot if vm is a template\n"
5790 if is_template($conf);
5791 }
5792
5793 $snap = $conf->{snapshots}->{$snapname};
5794
5795 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5796
5797 # remove parent refs
5798 if (!$prepare) {
5799 &$unlink_parent($conf, $snap->{parent});
5800 foreach my $sn (keys %{$conf->{snapshots}}) {
5801 next if $sn eq $snapname;
5802 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5803 }
5804 }
5805
5806 if ($remove_drive) {
5807 if ($remove_drive eq 'vmstate') {
5808 delete $snap->{$remove_drive};
5809 } else {
5810 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5811 my $volid = $drive->{file};
5812 delete $snap->{$remove_drive};
5813 add_unused_volume($conf, $volid);
5814 }
5815 }
5816
5817 if ($prepare) {
5818 $snap->{snapstate} = 'delete';
5819 } else {
5820 delete $conf->{snapshots}->{$snapname};
5821 delete $conf->{lock} if $drivehash;
5822 foreach my $volid (@$unused) {
5823 add_unused_volume($conf, $volid);
5824 }
5825 }
5826
5827 update_config_nolock($vmid, $conf, 1);
5828 };
5829
5830 lock_config($vmid, $updatefn);
5831
5832 # now remove vmstate file
5833
5834 my $storecfg = PVE::Storage::config();
5835
5836 if ($snap->{vmstate}) {
5837 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5838 if (my $err = $@) {
5839 die $err if !$force;
5840 warn $err;
5841 }
5842 # save changes (remove vmstate from snapshot)
5843 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5844 };
5845
5846 # now remove all internal snapshots
5847 foreach_drive($snap, sub {
5848 my ($ds, $drive) = @_;
5849
5850 return if drive_is_cdrom($drive);
5851
5852 my $volid = $drive->{file};
5853 my $device = "drive-$ds";
5854
5855 if (!$drivehash || $drivehash->{$ds}) {
5856 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5857 if (my $err = $@) {
5858 die $err if !$force;
5859 warn $err;
5860 }
5861 }
5862
5863 # save changes (remove drive fron snapshot)
5864 lock_config($vmid, $updatefn, $ds) if !$force;
5865 push @$unused, $volid;
5866 });
5867
5868 # now cleanup config
5869 $prepare = 0;
5870 lock_config($vmid, $updatefn);
5871 }
5872
5873 sub has_feature {
5874 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5875
5876 my $err;
5877 foreach_drive($conf, sub {
5878 my ($ds, $drive) = @_;
5879
5880 return if drive_is_cdrom($drive);
5881 my $volid = $drive->{file};
5882 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5883 });
5884
5885 return $err ? 0 : 1;
5886 }
5887
5888 sub template_create {
5889 my ($vmid, $conf, $disk) = @_;
5890
5891 my $storecfg = PVE::Storage::config();
5892
5893 foreach_drive($conf, sub {
5894 my ($ds, $drive) = @_;
5895
5896 return if drive_is_cdrom($drive);
5897 return if $disk && $ds ne $disk;
5898
5899 my $volid = $drive->{file};
5900 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5901
5902 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5903 $drive->{file} = $voliddst;
5904 $conf->{$ds} = print_drive($vmid, $drive);
5905 update_config_nolock($vmid, $conf, 1);
5906 });
5907 }
5908
5909 sub is_template {
5910 my ($conf) = @_;
5911
5912 return 1 if defined $conf->{template} && $conf->{template} == 1;
5913 }
5914
5915 sub qemu_img_convert {
5916 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5917
5918 my $storecfg = PVE::Storage::config();
5919 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5920 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5921
5922 if ($src_storeid && $dst_storeid) {
5923 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5924 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5925
5926 my $src_format = qemu_img_format($src_scfg, $src_volname);
5927 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5928
5929 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5930 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5931
5932 my $cmd = [];
5933 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5934 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5935 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5936
5937 my $parser = sub {
5938 my $line = shift;
5939 if($line =~ m/\((\S+)\/100\%\)/){
5940 my $percent = $1;
5941 my $transferred = int($size * $percent / 100);
5942 my $remaining = $size - $transferred;
5943
5944 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5945 }
5946
5947 };
5948
5949 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5950 my $err = $@;
5951 die "copy failed: $err" if $err;
5952 }
5953 }
5954
5955 sub qemu_img_format {
5956 my ($scfg, $volname) = @_;
5957
5958 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5959 return $1;
5960 } elsif ($scfg->{type} eq 'iscsi') {
5961 return "host_device";
5962 } else {
5963 return "raw";
5964 }
5965 }
5966
5967 sub qemu_drive_mirror {
5968 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5969
5970 my $count = 0;
5971 my $old_len = 0;
5972 my $frozen = undef;
5973 my $maxwait = 120;
5974
5975 my $storecfg = PVE::Storage::config();
5976 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5977
5978 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5979
5980 my $format;
5981 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5982 $format = $1;
5983 }
5984
5985 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5986
5987 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5988 $opts->{format} = $format if $format;
5989
5990 #fixme : sometime drive-mirror timeout, but works fine after.
5991 # (I have see the problem with big volume > 200GB), so we need to eval
5992 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5993 # ignore errors here
5994
5995 eval {
5996 while (1) {
5997 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5998 my $stat = @$stats[0];
5999 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
6000 die "error job is not mirroring" if $stat->{type} ne "mirror";
6001
6002 my $busy = $stat->{busy};
6003
6004 if (my $total = $stat->{len}) {
6005 my $transferred = $stat->{offset} || 0;
6006 my $remaining = $total - $transferred;
6007 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6008
6009 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
6010 }
6011
6012 if ($stat->{len} == $stat->{offset}) {
6013 if ($busy eq 'false') {
6014
6015 last if $vmiddst != $vmid;
6016
6017 # try to switch the disk if source and destination are on the same guest
6018 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6019 last if !$@;
6020 die $@ if $@ !~ m/cannot be completed/;
6021 }
6022
6023 if ($count > $maxwait) {
6024 # if too much writes to disk occurs at the end of migration
6025 #the disk needs to be freezed to be able to complete the migration
6026 vm_suspend($vmid,1);
6027 $frozen = 1;
6028 }
6029 $count ++
6030 }
6031 $old_len = $stat->{offset};
6032 sleep 1;
6033 }
6034
6035 vm_resume($vmid, 1) if $frozen;
6036
6037 };
6038 my $err = $@;
6039
6040 my $cancel_job = sub {
6041 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6042 while (1) {
6043 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6044 my $stat = @$stats[0];
6045 last if !$stat;
6046 sleep 1;
6047 }
6048 };
6049
6050 if ($err) {
6051 eval { &$cancel_job(); };
6052 die "mirroring error: $err";
6053 }
6054
6055 if ($vmiddst != $vmid) {
6056 # if we clone a disk for a new target vm, we don't switch the disk
6057 &$cancel_job(); # so we call block-job-cancel
6058 }
6059 }
6060
6061 sub clone_disk {
6062 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6063 $newvmid, $storage, $format, $full, $newvollist) = @_;
6064
6065 my $newvolid;
6066
6067 if (!$full) {
6068 print "create linked clone of drive $drivename ($drive->{file})\n";
6069 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6070 push @$newvollist, $newvolid;
6071 } else {
6072 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6073 $storeid = $storage if $storage;
6074
6075 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6076 if (!$format) {
6077 $format = $drive->{format} || $defFormat;
6078 }
6079
6080 # test if requested format is supported - else use default
6081 my $supported = grep { $_ eq $format } @$validFormats;
6082 $format = $defFormat if !$supported;
6083
6084 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6085
6086 print "create full clone of drive $drivename ($drive->{file})\n";
6087 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6088 push @$newvollist, $newvolid;
6089
6090 if (!$running || $snapname) {
6091 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6092 } else {
6093 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6094 }
6095 }
6096
6097 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6098
6099 my $disk = $drive;
6100 $disk->{format} = undef;
6101 $disk->{file} = $newvolid;
6102 $disk->{size} = $size;
6103
6104 return $disk;
6105 }
6106
6107 # this only works if VM is running
6108 sub get_current_qemu_machine {
6109 my ($vmid) = @_;
6110
6111 my $cmd = { execute => 'query-machines', arguments => {} };
6112 my $res = vm_qmp_command($vmid, $cmd);
6113
6114 my ($current, $default);
6115 foreach my $e (@$res) {
6116 $default = $e->{name} if $e->{'is-default'};
6117 $current = $e->{name} if $e->{'is-current'};
6118 }
6119
6120 # fallback to the default machine if current is not supported by qemu
6121 return $current || $default || 'pc';
6122 }
6123
6124 sub qemu_machine_feature_enabled {
6125 my ($machine, $kvmver, $version_major, $version_minor) = @_;
6126
6127 my $current_major;
6128 my $current_minor;
6129
6130 if ($machine && $machine =~ m/^(pc(-i440fx|-q35)?-(\d+)\.(\d+))/) {
6131
6132 $current_major = $3;
6133 $current_minor = $4;
6134
6135 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6136
6137 $current_major = $1;
6138 $current_minor = $2;
6139 }
6140
6141 return 1 if $current_major >= $version_major && $current_minor >= $version_minor;
6142
6143
6144 }
6145
6146 sub lspci {
6147
6148 my $devices = {};
6149
6150 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6151 my (undef, $id, $function) = @_;
6152 my $res = { id => $id, function => $function};
6153 push @{$devices->{$id}}, $res;
6154 });
6155
6156 return $devices;
6157 }
6158
6159 1;