]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
vm_start: apply pending changes
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'boolean',
176 description => "Allow hotplug for disk and network device",
177 default => 0,
178 },
179 reboot => {
180 optional => 1,
181 type => 'boolean',
182 description => "Allow reboot. If set to '0' the VM exit on reboot.",
183 default => 1,
184 },
185 lock => {
186 optional => 1,
187 type => 'string',
188 description => "Lock/unlock the VM.",
189 enum => [qw(migrate backup snapshot rollback)],
190 },
191 cpulimit => {
192 optional => 1,
193 type => 'integer',
194 description => "Limit of CPU usage in per cent. Note if the computer has 2 CPUs, it has total of 200% CPU time. Value '0' indicates no CPU limit.\n\nNOTE: This option is currently ignored.",
195 minimum => 0,
196 default => 0,
197 },
198 cpuunits => {
199 optional => 1,
200 type => 'integer',
201 description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.\n\nNOTE: You can disable fair-scheduler configuration by setting this to 0.",
202 minimum => 0,
203 maximum => 500000,
204 default => 1000,
205 },
206 memory => {
207 optional => 1,
208 type => 'integer',
209 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
210 minimum => 16,
211 default => 512,
212 },
213 balloon => {
214 optional => 1,
215 type => 'integer',
216 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
217 minimum => 0,
218 },
219 shares => {
220 optional => 1,
221 type => 'integer',
222 description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning",
223 minimum => 0,
224 maximum => 50000,
225 default => 1000,
226 },
227 keyboard => {
228 optional => 1,
229 type => 'string',
230 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
231 enum => PVE::Tools::kvmkeymaplist(),
232 default => 'en-us',
233 },
234 name => {
235 optional => 1,
236 type => 'string', format => 'dns-name',
237 description => "Set a name for the VM. Only used on the configuration web interface.",
238 },
239 scsihw => {
240 optional => 1,
241 type => 'string',
242 description => "scsi controller model",
243 enum => [qw(lsi lsi53c810 virtio-scsi-pci megasas pvscsi)],
244 default => 'lsi',
245 },
246 description => {
247 optional => 1,
248 type => 'string',
249 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
250 },
251 ostype => {
252 optional => 1,
253 type => 'string',
254 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
255 description => <<EODESC,
256 Used to enable special optimization/features for specific
257 operating systems:
258
259 other => unspecified OS
260 wxp => Microsoft Windows XP
261 w2k => Microsoft Windows 2000
262 w2k3 => Microsoft Windows 2003
263 w2k8 => Microsoft Windows 2008
264 wvista => Microsoft Windows Vista
265 win7 => Microsoft Windows 7
266 win8 => Microsoft Windows 8/2012
267 l24 => Linux 2.4 Kernel
268 l26 => Linux 2.6/3.X Kernel
269 solaris => solaris/opensolaris/openindiania kernel
270
271 other|l24|l26|solaris ... no special behaviour
272 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
273 EODESC
274 },
275 boot => {
276 optional => 1,
277 type => 'string',
278 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
279 pattern => '[acdn]{1,4}',
280 default => 'cdn',
281 },
282 bootdisk => {
283 optional => 1,
284 type => 'string', format => 'pve-qm-bootdisk',
285 description => "Enable booting from specified disk.",
286 pattern => '(ide|sata|scsi|virtio)\d+',
287 },
288 smp => {
289 optional => 1,
290 type => 'integer',
291 description => "The number of CPUs. Please use option -sockets instead.",
292 minimum => 1,
293 default => 1,
294 },
295 sockets => {
296 optional => 1,
297 type => 'integer',
298 description => "The number of CPU sockets.",
299 minimum => 1,
300 default => 1,
301 },
302 cores => {
303 optional => 1,
304 type => 'integer',
305 description => "The number of cores per socket.",
306 minimum => 1,
307 default => 1,
308 },
309 numa => {
310 optional => 1,
311 type => 'boolean',
312 description => "Enable/disable Numa.",
313 default => 0,
314 },
315 maxcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Maximum cpus for hotplug.",
319 minimum => 1,
320 default => 1,
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
494 my $numadesc = {
495 optional => 1,
496 type => 'string', format => 'pve-qm-numanode',
497 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>][,policy=<preferred|bind|interleave>]]",
498 description => "numa topology",
499 };
500 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
501
502 for (my $i = 0; $i < $MAX_NUMA; $i++) {
503 $confdesc->{"numa$i"} = $numadesc;
504 }
505
506 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
507 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3'];
508 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
509
510 my $netdesc = {
511 optional => 1,
512 type => 'string', format => 'pve-qm-net',
513 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>][,tag=<vlanid>][,firewall=0|1]",
514 description => <<EODESCR,
515 Specify network devices.
516
517 MODEL is one of: $nic_model_list_txt
518
519 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
520 automatically generated if not specified.
521
522 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
523
524 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'.
525
526 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
527
528 10.0.2.2 Gateway
529 10.0.2.3 DNS Server
530 10.0.2.4 SMB Server
531
532 The DHCP server assign addresses to the guest starting from 10.0.2.15.
533
534 EODESCR
535 };
536 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
537
538 for (my $i = 0; $i < $MAX_NETS; $i++) {
539 $confdesc->{"net$i"} = $netdesc;
540 }
541
542 my $drivename_hash;
543
544 my $idedesc = {
545 optional => 1,
546 type => 'string', format => 'pve-qm-drive',
547 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
548 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
549 };
550 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
551
552 my $scsidesc = {
553 optional => 1,
554 type => 'string', format => 'pve-qm-drive',
555 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
556 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
557 };
558 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
559
560 my $satadesc = {
561 optional => 1,
562 type => 'string', format => 'pve-qm-drive',
563 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
564 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
565 };
566 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
567
568 my $virtiodesc = {
569 optional => 1,
570 type => 'string', format => 'pve-qm-drive',
571 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
572 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
573 };
574 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
575
576 my $usbdesc = {
577 optional => 1,
578 type => 'string', format => 'pve-qm-usb-device',
579 typetext => 'host=HOSTUSBDEVICE|spice',
580 description => <<EODESCR,
581 Configure an USB device (n is 0 to 4). This can be used to
582 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
583
584 'bus-port(.port)*' (decimal numbers) or
585 'vendor_id:product_id' (hexadeciaml numbers)
586
587 You can use the 'lsusb -t' command to list existing usb devices.
588
589 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
590
591 The value 'spice' can be used to add a usb redirection devices for spice.
592
593 EODESCR
594 };
595 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
596
597 my $hostpcidesc = {
598 optional => 1,
599 type => 'string', format => 'pve-qm-hostpci',
600 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
601 description => <<EODESCR,
602 Map host pci devices. HOSTPCIDEVICE syntax is:
603
604 'bus:dev.func' (hexadecimal numbers)
605
606 You can us the 'lspci' command to list existing pci devices.
607
608 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
609
610 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
611
612 Experimental: user reported problems with this option.
613 EODESCR
614 };
615 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
616
617 my $serialdesc = {
618 optional => 1,
619 type => 'string',
620 pattern => '(/dev/.+|socket)',
621 description => <<EODESCR,
622 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).
623
624 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
625
626 Experimental: user reported problems with this option.
627 EODESCR
628 };
629
630 my $paralleldesc= {
631 optional => 1,
632 type => 'string',
633 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
634 description => <<EODESCR,
635 Map host parallel devices (n is 0 to 2).
636
637 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
638
639 Experimental: user reported problems with this option.
640 EODESCR
641 };
642
643 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
644 $confdesc->{"parallel$i"} = $paralleldesc;
645 }
646
647 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
648 $confdesc->{"serial$i"} = $serialdesc;
649 }
650
651 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
652 $confdesc->{"hostpci$i"} = $hostpcidesc;
653 }
654
655 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
656 $drivename_hash->{"ide$i"} = 1;
657 $confdesc->{"ide$i"} = $idedesc;
658 }
659
660 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
661 $drivename_hash->{"sata$i"} = 1;
662 $confdesc->{"sata$i"} = $satadesc;
663 }
664
665 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
666 $drivename_hash->{"scsi$i"} = 1;
667 $confdesc->{"scsi$i"} = $scsidesc ;
668 }
669
670 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
671 $drivename_hash->{"virtio$i"} = 1;
672 $confdesc->{"virtio$i"} = $virtiodesc;
673 }
674
675 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
676 $confdesc->{"usb$i"} = $usbdesc;
677 }
678
679 my $unuseddesc = {
680 optional => 1,
681 type => 'string', format => 'pve-volume-id',
682 description => "Reference to unused volumes.",
683 };
684
685 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
686 $confdesc->{"unused$i"} = $unuseddesc;
687 }
688
689 my $kvm_api_version = 0;
690
691 sub kvm_version {
692
693 return $kvm_api_version if $kvm_api_version;
694
695 my $fh = IO::File->new("</dev/kvm") ||
696 return 0;
697
698 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
699 $kvm_api_version = $v;
700 }
701
702 $fh->close();
703
704 return $kvm_api_version;
705 }
706
707 my $kvm_user_version;
708
709 sub kvm_user_version {
710
711 return $kvm_user_version if $kvm_user_version;
712
713 $kvm_user_version = 'unknown';
714
715 my $tmp = `kvm -help 2>/dev/null`;
716
717 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
718 $kvm_user_version = $2;
719 }
720
721 return $kvm_user_version;
722
723 }
724
725 my $kernel_has_vhost_net = -c '/dev/vhost-net';
726
727 sub disknames {
728 # order is important - used to autoselect boot disk
729 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
730 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
731 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
732 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
733 }
734
735 sub valid_drivename {
736 my $dev = shift;
737
738 return defined($drivename_hash->{$dev});
739 }
740
741 sub option_exists {
742 my $key = shift;
743 return defined($confdesc->{$key});
744 }
745
746 sub nic_models {
747 return $nic_model_list;
748 }
749
750 sub os_list_description {
751
752 return {
753 other => 'Other',
754 wxp => 'Windows XP',
755 w2k => 'Windows 2000',
756 w2k3 =>, 'Windows 2003',
757 w2k8 => 'Windows 2008',
758 wvista => 'Windows Vista',
759 win7 => 'Windows 7',
760 win8 => 'Windows 8/2012',
761 l24 => 'Linux 2.4',
762 l26 => 'Linux 2.6',
763 };
764 }
765
766 my $cdrom_path;
767
768 sub get_cdrom_path {
769
770 return $cdrom_path if $cdrom_path;
771
772 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
773 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
774 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
775 }
776
777 sub get_iso_path {
778 my ($storecfg, $vmid, $cdrom) = @_;
779
780 if ($cdrom eq 'cdrom') {
781 return get_cdrom_path();
782 } elsif ($cdrom eq 'none') {
783 return '';
784 } elsif ($cdrom =~ m|^/|) {
785 return $cdrom;
786 } else {
787 return PVE::Storage::path($storecfg, $cdrom);
788 }
789 }
790
791 # try to convert old style file names to volume IDs
792 sub filename_to_volume_id {
793 my ($vmid, $file, $media) = @_;
794
795 if (!($file eq 'none' || $file eq 'cdrom' ||
796 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
797
798 return undef if $file =~ m|/|;
799
800 if ($media && $media eq 'cdrom') {
801 $file = "local:iso/$file";
802 } else {
803 $file = "local:$vmid/$file";
804 }
805 }
806
807 return $file;
808 }
809
810 sub verify_media_type {
811 my ($opt, $vtype, $media) = @_;
812
813 return if !$media;
814
815 my $etype;
816 if ($media eq 'disk') {
817 $etype = 'images';
818 } elsif ($media eq 'cdrom') {
819 $etype = 'iso';
820 } else {
821 die "internal error";
822 }
823
824 return if ($vtype eq $etype);
825
826 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
827 }
828
829 sub cleanup_drive_path {
830 my ($opt, $storecfg, $drive) = @_;
831
832 # try to convert filesystem paths to volume IDs
833
834 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
835 ($drive->{file} !~ m|^/dev/.+|) &&
836 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
837 ($drive->{file} !~ m/^\d+$/)) {
838 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
839 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
840 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
841 verify_media_type($opt, $vtype, $drive->{media});
842 $drive->{file} = $volid;
843 }
844
845 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
846 }
847
848 sub create_conf_nolock {
849 my ($vmid, $settings) = @_;
850
851 my $filename = config_file($vmid);
852
853 die "configuration file '$filename' already exists\n" if -f $filename;
854
855 my $defaults = load_defaults();
856
857 $settings->{name} = "vm$vmid" if !$settings->{name};
858 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
859
860 my $data = '';
861 foreach my $opt (keys %$settings) {
862 next if !$confdesc->{$opt};
863
864 my $value = $settings->{$opt};
865 next if !$value;
866
867 $data .= "$opt: $value\n";
868 }
869
870 PVE::Tools::file_set_contents($filename, $data);
871 }
872
873 my $parse_size = sub {
874 my ($value) = @_;
875
876 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
877 my ($size, $unit) = ($1, $3);
878 if ($unit) {
879 if ($unit eq 'K') {
880 $size = $size * 1024;
881 } elsif ($unit eq 'M') {
882 $size = $size * 1024 * 1024;
883 } elsif ($unit eq 'G') {
884 $size = $size * 1024 * 1024 * 1024;
885 }
886 }
887 return int($size);
888 };
889
890 my $format_size = sub {
891 my ($size) = @_;
892
893 $size = int($size);
894
895 my $kb = int($size/1024);
896 return $size if $kb*1024 != $size;
897
898 my $mb = int($kb/1024);
899 return "${kb}K" if $mb*1024 != $kb;
900
901 my $gb = int($mb/1024);
902 return "${mb}M" if $gb*1024 != $mb;
903
904 return "${gb}G";
905 };
906
907 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
908 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
909 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
910 # [,aio=native|threads][,discard=ignore|on]
911
912 sub parse_drive {
913 my ($key, $data) = @_;
914
915 my $res = {};
916
917 # $key may be undefined - used to verify JSON parameters
918 if (!defined($key)) {
919 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
920 $res->{index} = 0;
921 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
922 $res->{interface} = $1;
923 $res->{index} = $2;
924 } else {
925 return undef;
926 }
927
928 foreach my $p (split (/,/, $data)) {
929 next if $p =~ m/^\s*$/;
930
931 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
932 my ($k, $v) = ($1, $2);
933
934 $k = 'file' if $k eq 'volume';
935
936 return undef if defined $res->{$k};
937
938 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
939 return undef if !$v || $v !~ m/^\d+/;
940 $k = "m$k";
941 $v = sprintf("%.3f", $v / (1024*1024));
942 }
943 $res->{$k} = $v;
944 } else {
945 if (!$res->{file} && $p !~ m/=/) {
946 $res->{file} = $p;
947 } else {
948 return undef;
949 }
950 }
951 }
952
953 return undef if !$res->{file};
954
955 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
956 $res->{format} = $1;
957 }
958
959 return undef if $res->{cache} &&
960 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
961 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
962 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
963 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
964 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
965 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
966 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
967 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
968 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
969 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
970 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
971 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
972 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
973
974 return undef if $res->{mbps_rd} && $res->{mbps};
975 return undef if $res->{mbps_wr} && $res->{mbps};
976
977 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
978 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
979 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
980 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
981 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
982 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
983
984 return undef if $res->{iops_rd} && $res->{iops};
985 return undef if $res->{iops_wr} && $res->{iops};
986
987
988 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
989 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
990 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
991 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
992 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
993 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
994
995
996 if ($res->{size}) {
997 return undef if !defined($res->{size} = &$parse_size($res->{size}));
998 }
999
1000 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1001 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1002 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1003 return undef if $res->{interface} eq 'virtio';
1004 }
1005
1006 # rerror does not work with scsi drives
1007 if ($res->{rerror}) {
1008 return undef if $res->{interface} eq 'scsi';
1009 }
1010
1011 return $res;
1012 }
1013
1014 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1015
1016 sub print_drive {
1017 my ($vmid, $drive) = @_;
1018
1019 my $opts = '';
1020 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1021 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1022 }
1023
1024 if ($drive->{size}) {
1025 $opts .= ",size=" . &$format_size($drive->{size});
1026 }
1027
1028 return "$drive->{file}$opts";
1029 }
1030
1031 sub scsi_inquiry {
1032 my($fh, $noerr) = @_;
1033
1034 my $SG_IO = 0x2285;
1035 my $SG_GET_VERSION_NUM = 0x2282;
1036
1037 my $versionbuf = "\x00" x 8;
1038 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1039 if (!$ret) {
1040 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1041 return undef;
1042 }
1043 my $version = unpack("I", $versionbuf);
1044 if ($version < 30000) {
1045 die "scsi generic interface too old\n" if !$noerr;
1046 return undef;
1047 }
1048
1049 my $buf = "\x00" x 36;
1050 my $sensebuf = "\x00" x 8;
1051 my $cmd = pack("C x3 C x1", 0x12, 36);
1052
1053 # see /usr/include/scsi/sg.h
1054 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";
1055
1056 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1057 length($sensebuf), 0, length($buf), $buf,
1058 $cmd, $sensebuf, 6000);
1059
1060 $ret = ioctl($fh, $SG_IO, $packet);
1061 if (!$ret) {
1062 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1063 return undef;
1064 }
1065
1066 my @res = unpack($sg_io_hdr_t, $packet);
1067 if ($res[17] || $res[18]) {
1068 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1069 return undef;
1070 }
1071
1072 my $res = {};
1073 (my $byte0, my $byte1, $res->{vendor},
1074 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1075
1076 $res->{removable} = $byte1 & 128 ? 1 : 0;
1077 $res->{type} = $byte0 & 31;
1078
1079 return $res;
1080 }
1081
1082 sub path_is_scsi {
1083 my ($path) = @_;
1084
1085 my $fh = IO::File->new("+<$path") || return undef;
1086 my $res = scsi_inquiry($fh, 1);
1087 close($fh);
1088
1089 return $res;
1090 }
1091
1092 sub machine_type_is_q35 {
1093 my ($conf) = @_;
1094
1095 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1096 }
1097
1098 sub print_tabletdevice_full {
1099 my ($conf) = @_;
1100
1101 my $q35 = machine_type_is_q35($conf);
1102
1103 # we use uhci for old VMs because tablet driver was buggy in older qemu
1104 my $usbbus = $q35 ? "ehci" : "uhci";
1105
1106 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1107 }
1108
1109 sub print_drivedevice_full {
1110 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1111
1112 my $device = '';
1113 my $maxdev = 0;
1114
1115 if ($drive->{interface} eq 'virtio') {
1116 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1117 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1118 $device .= ",iothread=iothread0" if $conf->{iothread};
1119 } elsif ($drive->{interface} eq 'scsi') {
1120 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1121 my $controller = int($drive->{index} / $maxdev);
1122 my $unit = $drive->{index} % $maxdev;
1123 my $devicetype = 'hd';
1124 my $path = '';
1125 if (drive_is_cdrom($drive)) {
1126 $devicetype = 'cd';
1127 } else {
1128 if ($drive->{file} =~ m|^/|) {
1129 $path = $drive->{file};
1130 } else {
1131 $path = PVE::Storage::path($storecfg, $drive->{file});
1132 }
1133
1134 if($path =~ m/^iscsi\:\/\//){
1135 $devicetype = 'generic';
1136 } else {
1137 if (my $info = path_is_scsi($path)) {
1138 if ($info->{type} == 0) {
1139 $devicetype = 'block';
1140 } elsif ($info->{type} == 1) { # tape
1141 $devicetype = 'generic';
1142 }
1143 }
1144 }
1145 }
1146
1147 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1148 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1149 } else {
1150 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1151 }
1152
1153 } elsif ($drive->{interface} eq 'ide'){
1154 $maxdev = 2;
1155 my $controller = int($drive->{index} / $maxdev);
1156 my $unit = $drive->{index} % $maxdev;
1157 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1158
1159 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1160 } elsif ($drive->{interface} eq 'sata'){
1161 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1162 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1163 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1164 } elsif ($drive->{interface} eq 'usb') {
1165 die "implement me";
1166 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1167 } else {
1168 die "unsupported interface type";
1169 }
1170
1171 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1172
1173 return $device;
1174 }
1175
1176 sub get_initiator_name {
1177 my $initiator;
1178
1179 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1180 while (defined(my $line = <$fh>)) {
1181 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1182 $initiator = $1;
1183 last;
1184 }
1185 $fh->close();
1186
1187 return $initiator;
1188 }
1189
1190 sub print_drive_full {
1191 my ($storecfg, $vmid, $drive) = @_;
1192
1193 my $opts = '';
1194 foreach my $o (@qemu_drive_options) {
1195 next if $o eq 'bootindex';
1196 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1197 }
1198
1199 foreach my $o (qw(bps bps_rd bps_wr)) {
1200 my $v = $drive->{"m$o"};
1201 $opts .= ",$o=" . int($v*1024*1024) if $v;
1202 }
1203
1204 # use linux-aio by default (qemu default is threads)
1205 $opts .= ",aio=native" if !$drive->{aio};
1206
1207 my $path;
1208 my $volid = $drive->{file};
1209 if (drive_is_cdrom($drive)) {
1210 $path = get_iso_path($storecfg, $vmid, $volid);
1211 } else {
1212 if ($volid =~ m|^/|) {
1213 $path = $volid;
1214 } else {
1215 $path = PVE::Storage::path($storecfg, $volid);
1216 }
1217 }
1218
1219 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1220
1221 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1222 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1223
1224 my $pathinfo = $path ? "file=$path," : '';
1225
1226 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1227 }
1228
1229 sub print_netdevice_full {
1230 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1231
1232 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1233
1234 my $device = $net->{model};
1235 if ($net->{model} eq 'virtio') {
1236 $device = 'virtio-net-pci';
1237 };
1238
1239 # qemu > 0.15 always try to boot from network - we disable that by
1240 # not loading the pxe rom file
1241 my $extra = ($bootorder !~ m/n/) ? "romfile=," : '';
1242 my $pciaddr = print_pci_addr("$netid", $bridges);
1243 my $tmpstr = "$device,${extra}mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1244 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1245 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1246 my $vectors = $net->{queues} * 2 + 2;
1247 $tmpstr .= ",vectors=$vectors,mq=on";
1248 }
1249 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1250 return $tmpstr;
1251 }
1252
1253 sub print_netdev_full {
1254 my ($vmid, $conf, $net, $netid) = @_;
1255
1256 my $i = '';
1257 if ($netid =~ m/^net(\d+)$/) {
1258 $i = int($1);
1259 }
1260
1261 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1262
1263 my $ifname = "tap${vmid}i$i";
1264
1265 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1266 die "interface name '$ifname' is too long (max 15 character)\n"
1267 if length($ifname) >= 16;
1268
1269 my $vhostparam = '';
1270 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1271
1272 my $vmname = $conf->{name} || "vm$vmid";
1273
1274 my $netdev = "";
1275
1276 if ($net->{bridge}) {
1277 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1278 } else {
1279 $netdev = "type=user,id=$netid,hostname=$vmname";
1280 }
1281
1282 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1283
1284 return $netdev;
1285 }
1286
1287 sub drive_is_cdrom {
1288 my ($drive) = @_;
1289
1290 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1291
1292 }
1293
1294 sub parse_numa {
1295 my ($data) = @_;
1296
1297 my $res = {};
1298
1299 foreach my $kvp (split(/,/, $data)) {
1300
1301 if ($kvp =~ m/^memory=(\S+)$/) {
1302 $res->{memory} = $1;
1303 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1304 $res->{policy} = $1;
1305 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1306 $res->{cpus}->{start} = $1;
1307 $res->{cpus}->{end} = $3;
1308 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1309 $res->{hostnodes}->{start} = $1;
1310 $res->{hostnodes}->{end} = $3;
1311 } else {
1312 return undef;
1313 }
1314 }
1315
1316 return $res;
1317 }
1318
1319 sub parse_hostpci {
1320 my ($value) = @_;
1321
1322 return undef if !$value;
1323
1324
1325 my @list = split(/,/, $value);
1326 my $found;
1327
1328 my $res = {};
1329 foreach my $kv (@list) {
1330
1331 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1332 $found = 1;
1333 if(defined($4)){
1334 push @{$res->{pciid}}, { id => $2 , function => $4};
1335
1336 }else{
1337 my $pcidevices = lspci($2);
1338 $res->{pciid} = $pcidevices->{$2};
1339 }
1340 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1341 $res->{driver} = $1;
1342 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1343 $res->{rombar} = $1;
1344 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1345 $res->{'x-vga'} = $1;
1346 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1347 $res->{pcie} = 1 if $1 == 1;
1348 } else {
1349 warn "unknown hostpci setting '$kv'\n";
1350 }
1351 }
1352
1353 return undef if !$found;
1354
1355 return $res;
1356 }
1357
1358 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1359 sub parse_net {
1360 my ($data) = @_;
1361
1362 my $res = {};
1363
1364 foreach my $kvp (split(/,/, $data)) {
1365
1366 if ($kvp =~ m/^(ne2k_pci|e1000|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1367 my $model = lc($1);
1368 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1369 $res->{model} = $model;
1370 $res->{macaddr} = $mac;
1371 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1372 $res->{bridge} = $1;
1373 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1374 $res->{queues} = $1;
1375 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1376 $res->{rate} = $1;
1377 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1378 $res->{tag} = $1;
1379 } elsif ($kvp =~ m/^firewall=(\d+)$/) {
1380 $res->{firewall} = $1;
1381 } else {
1382 return undef;
1383 }
1384
1385 }
1386
1387 return undef if !$res->{model};
1388
1389 return $res;
1390 }
1391
1392 sub print_net {
1393 my $net = shift;
1394
1395 my $res = "$net->{model}";
1396 $res .= "=$net->{macaddr}" if $net->{macaddr};
1397 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1398 $res .= ",rate=$net->{rate}" if $net->{rate};
1399 $res .= ",tag=$net->{tag}" if $net->{tag};
1400 $res .= ",firewall=$net->{firewall}" if $net->{firewall};
1401
1402 return $res;
1403 }
1404
1405 sub add_random_macs {
1406 my ($settings) = @_;
1407
1408 foreach my $opt (keys %$settings) {
1409 next if $opt !~ m/^net(\d+)$/;
1410 my $net = parse_net($settings->{$opt});
1411 next if !$net;
1412 $settings->{$opt} = print_net($net);
1413 }
1414 }
1415
1416 sub add_unused_volume {
1417 my ($config, $volid) = @_;
1418
1419 my $key;
1420 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1421 my $test = "unused$ind";
1422 if (my $vid = $config->{$test}) {
1423 return if $vid eq $volid; # do not add duplicates
1424 } else {
1425 $key = $test;
1426 }
1427 }
1428
1429 die "To many unused volume - please delete them first.\n" if !$key;
1430
1431 $config->{$key} = $volid;
1432
1433 return $key;
1434 }
1435
1436 sub vm_is_volid_owner {
1437 my ($storecfg, $vmid, $volid) = @_;
1438
1439 if ($volid !~ m|^/|) {
1440 my ($path, $owner);
1441 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1442 if ($owner && ($owner == $vmid)) {
1443 return 1;
1444 }
1445 }
1446
1447 return undef;
1448 }
1449
1450 sub vmconfig_delete_pending_option {
1451 my ($conf, $key) = @_;
1452
1453 delete $conf->{pending}->{$key};
1454 my $pending_delete_hash = { $key => 1 };
1455 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1456 $pending_delete_hash->{$opt} = 1;
1457 }
1458 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1459 }
1460
1461 sub vmconfig_undelete_pending_option {
1462 my ($conf, $key) = @_;
1463
1464 my $pending_delete_hash = {};
1465 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1466 $pending_delete_hash->{$opt} = 1;
1467 }
1468 delete $pending_delete_hash->{$key};
1469
1470 my @keylist = keys %$pending_delete_hash;
1471 if (scalar(@keylist)) {
1472 $conf->{pending}->{delete} = join(',', @keylist);
1473 } else {
1474 delete $conf->{pending}->{delete};
1475 }
1476 }
1477
1478 sub vmconfig_register_unused_drive {
1479 my ($storecfg, $vmid, $conf, $drive) = @_;
1480
1481 if (!drive_is_cdrom($drive)) {
1482 my $volid = $drive->{file};
1483 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1484 add_unused_volume($conf, $volid, $vmid);
1485 }
1486 }
1487 }
1488
1489 my $valid_smbios1_options = {
1490 manufacturer => '\S+',
1491 product => '\S+',
1492 version => '\S+',
1493 serial => '\S+',
1494 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1495 sku => '\S+',
1496 family => '\S+',
1497 };
1498
1499 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1500 sub parse_smbios1 {
1501 my ($data) = @_;
1502
1503 my $res = {};
1504
1505 foreach my $kvp (split(/,/, $data)) {
1506 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1507 my ($k, $v) = split(/=/, $kvp);
1508 return undef if !defined($k) || !defined($v);
1509 return undef if !$valid_smbios1_options->{$k};
1510 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1511 $res->{$k} = $v;
1512 }
1513
1514 return $res;
1515 }
1516
1517 sub print_smbios1 {
1518 my ($smbios1) = @_;
1519
1520 my $data = '';
1521 foreach my $k (keys %$smbios1) {
1522 next if !defined($smbios1->{$k});
1523 next if !$valid_smbios1_options->{$k};
1524 $data .= ',' if $data;
1525 $data .= "$k=$smbios1->{$k}";
1526 }
1527 return $data;
1528 }
1529
1530 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1531 sub verify_smbios1 {
1532 my ($value, $noerr) = @_;
1533
1534 return $value if parse_smbios1($value);
1535
1536 return undef if $noerr;
1537
1538 die "unable to parse smbios (type 1) options\n";
1539 }
1540
1541 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1542 sub verify_bootdisk {
1543 my ($value, $noerr) = @_;
1544
1545 return $value if valid_drivename($value);
1546
1547 return undef if $noerr;
1548
1549 die "invalid boot disk '$value'\n";
1550 }
1551
1552 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1553 sub verify_numa {
1554 my ($value, $noerr) = @_;
1555
1556 return $value if parse_numa($value);
1557
1558 return undef if $noerr;
1559
1560 die "unable to parse numa options\n";
1561 }
1562
1563 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1564 sub verify_net {
1565 my ($value, $noerr) = @_;
1566
1567 return $value if parse_net($value);
1568
1569 return undef if $noerr;
1570
1571 die "unable to parse network options\n";
1572 }
1573
1574 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1575 sub verify_drive {
1576 my ($value, $noerr) = @_;
1577
1578 return $value if parse_drive(undef, $value);
1579
1580 return undef if $noerr;
1581
1582 die "unable to parse drive options\n";
1583 }
1584
1585 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1586 sub verify_hostpci {
1587 my ($value, $noerr) = @_;
1588
1589 return $value if parse_hostpci($value);
1590
1591 return undef if $noerr;
1592
1593 die "unable to parse pci id\n";
1594 }
1595
1596 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1597 sub verify_watchdog {
1598 my ($value, $noerr) = @_;
1599
1600 return $value if parse_watchdog($value);
1601
1602 return undef if $noerr;
1603
1604 die "unable to parse watchdog options\n";
1605 }
1606
1607 sub parse_watchdog {
1608 my ($value) = @_;
1609
1610 return undef if !$value;
1611
1612 my $res = {};
1613
1614 foreach my $p (split(/,/, $value)) {
1615 next if $p =~ m/^\s*$/;
1616
1617 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1618 $res->{model} = $2;
1619 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1620 $res->{action} = $2;
1621 } else {
1622 return undef;
1623 }
1624 }
1625
1626 return $res;
1627 }
1628
1629 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1630 sub verify_startup {
1631 my ($value, $noerr) = @_;
1632
1633 return $value if parse_startup($value);
1634
1635 return undef if $noerr;
1636
1637 die "unable to parse startup options\n";
1638 }
1639
1640 sub parse_startup {
1641 my ($value) = @_;
1642
1643 return undef if !$value;
1644
1645 my $res = {};
1646
1647 foreach my $p (split(/,/, $value)) {
1648 next if $p =~ m/^\s*$/;
1649
1650 if ($p =~ m/^(order=)?(\d+)$/) {
1651 $res->{order} = $2;
1652 } elsif ($p =~ m/^up=(\d+)$/) {
1653 $res->{up} = $1;
1654 } elsif ($p =~ m/^down=(\d+)$/) {
1655 $res->{down} = $1;
1656 } else {
1657 return undef;
1658 }
1659 }
1660
1661 return $res;
1662 }
1663
1664 sub parse_usb_device {
1665 my ($value) = @_;
1666
1667 return undef if !$value;
1668
1669 my @dl = split(/,/, $value);
1670 my $found;
1671
1672 my $res = {};
1673 foreach my $v (@dl) {
1674 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1675 $found = 1;
1676 $res->{vendorid} = $2;
1677 $res->{productid} = $4;
1678 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1679 $found = 1;
1680 $res->{hostbus} = $1;
1681 $res->{hostport} = $2;
1682 } elsif ($v =~ m/^spice$/) {
1683 $found = 1;
1684 $res->{spice} = 1;
1685 } else {
1686 return undef;
1687 }
1688 }
1689 return undef if !$found;
1690
1691 return $res;
1692 }
1693
1694 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1695 sub verify_usb_device {
1696 my ($value, $noerr) = @_;
1697
1698 return $value if parse_usb_device($value);
1699
1700 return undef if $noerr;
1701
1702 die "unable to parse usb device\n";
1703 }
1704
1705 # add JSON properties for create and set function
1706 sub json_config_properties {
1707 my $prop = shift;
1708
1709 foreach my $opt (keys %$confdesc) {
1710 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1711 $prop->{$opt} = $confdesc->{$opt};
1712 }
1713
1714 return $prop;
1715 }
1716
1717 sub check_type {
1718 my ($key, $value) = @_;
1719
1720 die "unknown setting '$key'\n" if !$confdesc->{$key};
1721
1722 my $type = $confdesc->{$key}->{type};
1723
1724 if (!defined($value)) {
1725 die "got undefined value\n";
1726 }
1727
1728 if ($value =~ m/[\n\r]/) {
1729 die "property contains a line feed\n";
1730 }
1731
1732 if ($type eq 'boolean') {
1733 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1734 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1735 die "type check ('boolean') failed - got '$value'\n";
1736 } elsif ($type eq 'integer') {
1737 return int($1) if $value =~ m/^(\d+)$/;
1738 die "type check ('integer') failed - got '$value'\n";
1739 } elsif ($type eq 'number') {
1740 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1741 die "type check ('number') failed - got '$value'\n";
1742 } elsif ($type eq 'string') {
1743 if (my $fmt = $confdesc->{$key}->{format}) {
1744 if ($fmt eq 'pve-qm-drive') {
1745 # special case - we need to pass $key to parse_drive()
1746 my $drive = parse_drive($key, $value);
1747 return $value if $drive;
1748 die "unable to parse drive options\n";
1749 }
1750 PVE::JSONSchema::check_format($fmt, $value);
1751 return $value;
1752 }
1753 $value =~ s/^\"(.*)\"$/$1/;
1754 return $value;
1755 } else {
1756 die "internal error"
1757 }
1758 }
1759
1760 sub lock_config_full {
1761 my ($vmid, $timeout, $code, @param) = @_;
1762
1763 my $filename = config_file_lock($vmid);
1764
1765 my $res = lock_file($filename, $timeout, $code, @param);
1766
1767 die $@ if $@;
1768
1769 return $res;
1770 }
1771
1772 sub lock_config_mode {
1773 my ($vmid, $timeout, $shared, $code, @param) = @_;
1774
1775 my $filename = config_file_lock($vmid);
1776
1777 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1778
1779 die $@ if $@;
1780
1781 return $res;
1782 }
1783
1784 sub lock_config {
1785 my ($vmid, $code, @param) = @_;
1786
1787 return lock_config_full($vmid, 10, $code, @param);
1788 }
1789
1790 sub cfs_config_path {
1791 my ($vmid, $node) = @_;
1792
1793 $node = $nodename if !$node;
1794 return "nodes/$node/qemu-server/$vmid.conf";
1795 }
1796
1797 sub check_iommu_support{
1798 #fixme : need to check IOMMU support
1799 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1800
1801 my $iommu=1;
1802 return $iommu;
1803
1804 }
1805
1806 sub config_file {
1807 my ($vmid, $node) = @_;
1808
1809 my $cfspath = cfs_config_path($vmid, $node);
1810 return "/etc/pve/$cfspath";
1811 }
1812
1813 sub config_file_lock {
1814 my ($vmid) = @_;
1815
1816 return "$lock_dir/lock-$vmid.conf";
1817 }
1818
1819 sub touch_config {
1820 my ($vmid) = @_;
1821
1822 my $conf = config_file($vmid);
1823 utime undef, undef, $conf;
1824 }
1825
1826 sub destroy_vm {
1827 my ($storecfg, $vmid, $keep_empty_config) = @_;
1828
1829 my $conffile = config_file($vmid);
1830
1831 my $conf = load_config($vmid);
1832
1833 check_lock($conf);
1834
1835 # only remove disks owned by this VM
1836 foreach_drive($conf, sub {
1837 my ($ds, $drive) = @_;
1838
1839 return if drive_is_cdrom($drive);
1840
1841 my $volid = $drive->{file};
1842
1843 return if !$volid || $volid =~ m|^/|;
1844
1845 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1846 return if !$path || !$owner || ($owner != $vmid);
1847
1848 PVE::Storage::vdisk_free($storecfg, $volid);
1849 });
1850
1851 if ($keep_empty_config) {
1852 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1853 } else {
1854 unlink $conffile;
1855 }
1856
1857 # also remove unused disk
1858 eval {
1859 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1860
1861 eval {
1862 PVE::Storage::foreach_volid($dl, sub {
1863 my ($volid, $sid, $volname, $d) = @_;
1864 PVE::Storage::vdisk_free($storecfg, $volid);
1865 });
1866 };
1867 warn $@ if $@;
1868
1869 };
1870 warn $@ if $@;
1871 }
1872
1873 sub load_config {
1874 my ($vmid, $node) = @_;
1875
1876 my $cfspath = cfs_config_path($vmid, $node);
1877
1878 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1879
1880 die "no such VM ('$vmid')\n" if !defined($conf);
1881
1882 return $conf;
1883 }
1884
1885 sub parse_vm_config {
1886 my ($filename, $raw) = @_;
1887
1888 return undef if !defined($raw);
1889
1890 my $res = {
1891 digest => Digest::SHA::sha1_hex($raw),
1892 snapshots => {},
1893 pending => {},
1894 };
1895
1896 $filename =~ m|/qemu-server/(\d+)\.conf$|
1897 || die "got strange filename '$filename'";
1898
1899 my $vmid = $1;
1900
1901 my $conf = $res;
1902 my $descr = '';
1903 my $section = '';
1904
1905 my @lines = split(/\n/, $raw);
1906 foreach my $line (@lines) {
1907 next if $line =~ m/^\s*$/;
1908
1909 if ($line =~ m/^\[PENDING\]\s*$/i) {
1910 $section = 'pending';
1911 $conf->{description} = $descr if $descr;
1912 $descr = '';
1913 $conf = $res->{$section} = {};
1914 next;
1915
1916 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1917 $section = $1;
1918 $conf->{description} = $descr if $descr;
1919 $descr = '';
1920 $conf = $res->{snapshots}->{$section} = {};
1921 next;
1922 }
1923
1924 if ($line =~ m/^\#(.*)\s*$/) {
1925 $descr .= PVE::Tools::decode_text($1) . "\n";
1926 next;
1927 }
1928
1929 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1930 $descr .= PVE::Tools::decode_text($2);
1931 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1932 $conf->{snapstate} = $1;
1933 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1934 my $key = $1;
1935 my $value = $2;
1936 $conf->{$key} = $value;
1937 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
1938 my $value = $1;
1939 if ($section eq 'pending') {
1940 $conf->{delete} = $value; # we parse this later
1941 } else {
1942 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
1943 }
1944 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
1945 my $key = $1;
1946 my $value = $2;
1947 eval { $value = check_type($key, $value); };
1948 if ($@) {
1949 warn "vm $vmid - unable to parse value of '$key' - $@";
1950 } else {
1951 my $fmt = $confdesc->{$key}->{format};
1952 if ($fmt && $fmt eq 'pve-qm-drive') {
1953 my $v = parse_drive($key, $value);
1954 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
1955 $v->{file} = $volid;
1956 $value = print_drive($vmid, $v);
1957 } else {
1958 warn "vm $vmid - unable to parse value of '$key'\n";
1959 next;
1960 }
1961 }
1962
1963 if ($key eq 'cdrom') {
1964 $conf->{ide2} = $value;
1965 } else {
1966 $conf->{$key} = $value;
1967 }
1968 }
1969 }
1970 }
1971
1972 $conf->{description} = $descr if $descr;
1973
1974 delete $res->{snapstate}; # just to be sure
1975
1976 return $res;
1977 }
1978
1979 sub write_vm_config {
1980 my ($filename, $conf) = @_;
1981
1982 delete $conf->{snapstate}; # just to be sure
1983
1984 if ($conf->{cdrom}) {
1985 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
1986 $conf->{ide2} = $conf->{cdrom};
1987 delete $conf->{cdrom};
1988 }
1989
1990 # we do not use 'smp' any longer
1991 if ($conf->{sockets}) {
1992 delete $conf->{smp};
1993 } elsif ($conf->{smp}) {
1994 $conf->{sockets} = $conf->{smp};
1995 delete $conf->{cores};
1996 delete $conf->{smp};
1997 }
1998
1999 if ($conf->{maxcpus} && $conf->{sockets}) {
2000 delete $conf->{sockets};
2001 }
2002
2003 my $used_volids = {};
2004
2005 my $cleanup_config = sub {
2006 my ($cref, $pending, $snapname) = @_;
2007
2008 foreach my $key (keys %$cref) {
2009 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2010 $key eq 'snapstate' || $key eq 'pending';
2011 my $value = $cref->{$key};
2012 if ($key eq 'delete') {
2013 die "propertry 'delete' is only allowed in [PENDING]\n"
2014 if !$pending;
2015 # fixme: check syntax?
2016 next;
2017 }
2018 eval { $value = check_type($key, $value); };
2019 die "unable to parse value of '$key' - $@" if $@;
2020
2021 $cref->{$key} = $value;
2022
2023 if (!$snapname && valid_drivename($key)) {
2024 my $drive = parse_drive($key, $value);
2025 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2026 }
2027 }
2028 };
2029
2030 &$cleanup_config($conf);
2031
2032 &$cleanup_config($conf->{pending}, 1);
2033
2034 foreach my $snapname (keys %{$conf->{snapshots}}) {
2035 die "internal error" if $snapname eq 'pending';
2036 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2037 }
2038
2039 # remove 'unusedX' settings if we re-add a volume
2040 foreach my $key (keys %$conf) {
2041 my $value = $conf->{$key};
2042 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2043 delete $conf->{$key};
2044 }
2045 }
2046
2047 my $generate_raw_config = sub {
2048 my ($conf) = @_;
2049
2050 my $raw = '';
2051
2052 # add description as comment to top of file
2053 my $descr = $conf->{description} || '';
2054 foreach my $cl (split(/\n/, $descr)) {
2055 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2056 }
2057
2058 foreach my $key (sort keys %$conf) {
2059 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2060 $raw .= "$key: $conf->{$key}\n";
2061 }
2062 return $raw;
2063 };
2064
2065 my $raw = &$generate_raw_config($conf);
2066
2067 if (scalar(keys %{$conf->{pending}})){
2068 $raw .= "\n[PENDING]\n";
2069 $raw .= &$generate_raw_config($conf->{pending});
2070 }
2071
2072 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2073 $raw .= "\n[$snapname]\n";
2074 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2075 }
2076
2077 return $raw;
2078 }
2079
2080 sub update_config_nolock {
2081 my ($vmid, $conf, $skiplock) = @_;
2082
2083 check_lock($conf) if !$skiplock;
2084
2085 my $cfspath = cfs_config_path($vmid);
2086
2087 PVE::Cluster::cfs_write_file($cfspath, $conf);
2088 }
2089
2090 sub update_config {
2091 my ($vmid, $conf, $skiplock) = @_;
2092
2093 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2094 }
2095
2096 sub load_defaults {
2097
2098 my $res = {};
2099
2100 # we use static defaults from our JSON schema configuration
2101 foreach my $key (keys %$confdesc) {
2102 if (defined(my $default = $confdesc->{$key}->{default})) {
2103 $res->{$key} = $default;
2104 }
2105 }
2106
2107 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2108 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2109
2110 return $res;
2111 }
2112
2113 sub config_list {
2114 my $vmlist = PVE::Cluster::get_vmlist();
2115 my $res = {};
2116 return $res if !$vmlist || !$vmlist->{ids};
2117 my $ids = $vmlist->{ids};
2118
2119 foreach my $vmid (keys %$ids) {
2120 my $d = $ids->{$vmid};
2121 next if !$d->{node} || $d->{node} ne $nodename;
2122 next if !$d->{type} || $d->{type} ne 'qemu';
2123 $res->{$vmid}->{exists} = 1;
2124 }
2125 return $res;
2126 }
2127
2128 # test if VM uses local resources (to prevent migration)
2129 sub check_local_resources {
2130 my ($conf, $noerr) = @_;
2131
2132 my $loc_res = 0;
2133
2134 $loc_res = 1 if $conf->{hostusb}; # old syntax
2135 $loc_res = 1 if $conf->{hostpci}; # old syntax
2136
2137 foreach my $k (keys %$conf) {
2138 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2139 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2140 }
2141
2142 die "VM uses local resources\n" if $loc_res && !$noerr;
2143
2144 return $loc_res;
2145 }
2146
2147 # check if used storages are available on all nodes (use by migrate)
2148 sub check_storage_availability {
2149 my ($storecfg, $conf, $node) = @_;
2150
2151 foreach_drive($conf, sub {
2152 my ($ds, $drive) = @_;
2153
2154 my $volid = $drive->{file};
2155 return if !$volid;
2156
2157 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2158 return if !$sid;
2159
2160 # check if storage is available on both nodes
2161 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2162 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2163 });
2164 }
2165
2166 # list nodes where all VM images are available (used by has_feature API)
2167 sub shared_nodes {
2168 my ($conf, $storecfg) = @_;
2169
2170 my $nodelist = PVE::Cluster::get_nodelist();
2171 my $nodehash = { map { $_ => 1 } @$nodelist };
2172 my $nodename = PVE::INotify::nodename();
2173
2174 foreach_drive($conf, sub {
2175 my ($ds, $drive) = @_;
2176
2177 my $volid = $drive->{file};
2178 return if !$volid;
2179
2180 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2181 if ($storeid) {
2182 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2183 if ($scfg->{disable}) {
2184 $nodehash = {};
2185 } elsif (my $avail = $scfg->{nodes}) {
2186 foreach my $node (keys %$nodehash) {
2187 delete $nodehash->{$node} if !$avail->{$node};
2188 }
2189 } elsif (!$scfg->{shared}) {
2190 foreach my $node (keys %$nodehash) {
2191 delete $nodehash->{$node} if $node ne $nodename
2192 }
2193 }
2194 }
2195 });
2196
2197 return $nodehash
2198 }
2199
2200 sub check_lock {
2201 my ($conf) = @_;
2202
2203 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2204 }
2205
2206 sub check_cmdline {
2207 my ($pidfile, $pid) = @_;
2208
2209 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2210 if (defined($fh)) {
2211 my $line = <$fh>;
2212 $fh->close;
2213 return undef if !$line;
2214 my @param = split(/\0/, $line);
2215
2216 my $cmd = $param[0];
2217 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2218
2219 for (my $i = 0; $i < scalar (@param); $i++) {
2220 my $p = $param[$i];
2221 next if !$p;
2222 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2223 my $p = $param[$i+1];
2224 return 1 if $p && ($p eq $pidfile);
2225 return undef;
2226 }
2227 }
2228 }
2229 return undef;
2230 }
2231
2232 sub check_running {
2233 my ($vmid, $nocheck, $node) = @_;
2234
2235 my $filename = config_file($vmid, $node);
2236
2237 die "unable to find configuration file for VM $vmid - no such machine\n"
2238 if !$nocheck && ! -f $filename;
2239
2240 my $pidfile = pidfile_name($vmid);
2241
2242 if (my $fd = IO::File->new("<$pidfile")) {
2243 my $st = stat($fd);
2244 my $line = <$fd>;
2245 close($fd);
2246
2247 my $mtime = $st->mtime;
2248 if ($mtime > time()) {
2249 warn "file '$filename' modified in future\n";
2250 }
2251
2252 if ($line =~ m/^(\d+)$/) {
2253 my $pid = $1;
2254 if (check_cmdline($pidfile, $pid)) {
2255 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2256 return $pid;
2257 }
2258 }
2259 }
2260 }
2261
2262 return undef;
2263 }
2264
2265 sub vzlist {
2266
2267 my $vzlist = config_list();
2268
2269 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2270
2271 while (defined(my $de = $fd->read)) {
2272 next if $de !~ m/^(\d+)\.pid$/;
2273 my $vmid = $1;
2274 next if !defined($vzlist->{$vmid});
2275 if (my $pid = check_running($vmid)) {
2276 $vzlist->{$vmid}->{pid} = $pid;
2277 }
2278 }
2279
2280 return $vzlist;
2281 }
2282
2283 sub disksize {
2284 my ($storecfg, $conf) = @_;
2285
2286 my $bootdisk = $conf->{bootdisk};
2287 return undef if !$bootdisk;
2288 return undef if !valid_drivename($bootdisk);
2289
2290 return undef if !$conf->{$bootdisk};
2291
2292 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2293 return undef if !defined($drive);
2294
2295 return undef if drive_is_cdrom($drive);
2296
2297 my $volid = $drive->{file};
2298 return undef if !$volid;
2299
2300 return $drive->{size};
2301 }
2302
2303 my $last_proc_pid_stat;
2304
2305 # get VM status information
2306 # This must be fast and should not block ($full == false)
2307 # We only query KVM using QMP if $full == true (this can be slow)
2308 sub vmstatus {
2309 my ($opt_vmid, $full) = @_;
2310
2311 my $res = {};
2312
2313 my $storecfg = PVE::Storage::config();
2314
2315 my $list = vzlist();
2316 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2317
2318 my $cpucount = $cpuinfo->{cpus} || 1;
2319
2320 foreach my $vmid (keys %$list) {
2321 next if $opt_vmid && ($vmid ne $opt_vmid);
2322
2323 my $cfspath = cfs_config_path($vmid);
2324 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2325
2326 my $d = {};
2327 $d->{pid} = $list->{$vmid}->{pid};
2328
2329 # fixme: better status?
2330 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2331
2332 my $size = disksize($storecfg, $conf);
2333 if (defined($size)) {
2334 $d->{disk} = 0; # no info available
2335 $d->{maxdisk} = $size;
2336 } else {
2337 $d->{disk} = 0;
2338 $d->{maxdisk} = 0;
2339 }
2340
2341 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2342 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2343
2344 $d->{name} = $conf->{name} || "VM $vmid";
2345 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2346
2347 if ($conf->{balloon}) {
2348 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2349 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2350 }
2351
2352 $d->{uptime} = 0;
2353 $d->{cpu} = 0;
2354 $d->{mem} = 0;
2355
2356 $d->{netout} = 0;
2357 $d->{netin} = 0;
2358
2359 $d->{diskread} = 0;
2360 $d->{diskwrite} = 0;
2361
2362 $d->{template} = is_template($conf);
2363
2364 $res->{$vmid} = $d;
2365 }
2366
2367 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2368 foreach my $dev (keys %$netdev) {
2369 next if $dev !~ m/^tap([1-9]\d*)i/;
2370 my $vmid = $1;
2371 my $d = $res->{$vmid};
2372 next if !$d;
2373
2374 $d->{netout} += $netdev->{$dev}->{receive};
2375 $d->{netin} += $netdev->{$dev}->{transmit};
2376 }
2377
2378 my $ctime = gettimeofday;
2379
2380 foreach my $vmid (keys %$list) {
2381
2382 my $d = $res->{$vmid};
2383 my $pid = $d->{pid};
2384 next if !$pid;
2385
2386 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2387 next if !$pstat; # not running
2388
2389 my $used = $pstat->{utime} + $pstat->{stime};
2390
2391 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2392
2393 if ($pstat->{vsize}) {
2394 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2395 }
2396
2397 my $old = $last_proc_pid_stat->{$pid};
2398 if (!$old) {
2399 $last_proc_pid_stat->{$pid} = {
2400 time => $ctime,
2401 used => $used,
2402 cpu => 0,
2403 };
2404 next;
2405 }
2406
2407 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2408
2409 if ($dtime > 1000) {
2410 my $dutime = $used - $old->{used};
2411
2412 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2413 $last_proc_pid_stat->{$pid} = {
2414 time => $ctime,
2415 used => $used,
2416 cpu => $d->{cpu},
2417 };
2418 } else {
2419 $d->{cpu} = $old->{cpu};
2420 }
2421 }
2422
2423 return $res if !$full;
2424
2425 my $qmpclient = PVE::QMPClient->new();
2426
2427 my $ballooncb = sub {
2428 my ($vmid, $resp) = @_;
2429
2430 my $info = $resp->{'return'};
2431 return if !$info->{max_mem};
2432
2433 my $d = $res->{$vmid};
2434
2435 # use memory assigned to VM
2436 $d->{maxmem} = $info->{max_mem};
2437 $d->{balloon} = $info->{actual};
2438
2439 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2440 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2441 $d->{freemem} = $info->{free_mem};
2442 }
2443
2444 };
2445
2446 my $blockstatscb = sub {
2447 my ($vmid, $resp) = @_;
2448 my $data = $resp->{'return'} || [];
2449 my $totalrdbytes = 0;
2450 my $totalwrbytes = 0;
2451 for my $blockstat (@$data) {
2452 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2453 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2454 }
2455 $res->{$vmid}->{diskread} = $totalrdbytes;
2456 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2457 };
2458
2459 my $statuscb = sub {
2460 my ($vmid, $resp) = @_;
2461
2462 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2463 # this fails if ballon driver is not loaded, so this must be
2464 # the last commnand (following command are aborted if this fails).
2465 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2466
2467 my $status = 'unknown';
2468 if (!defined($status = $resp->{'return'}->{status})) {
2469 warn "unable to get VM status\n";
2470 return;
2471 }
2472
2473 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2474 };
2475
2476 foreach my $vmid (keys %$list) {
2477 next if $opt_vmid && ($vmid ne $opt_vmid);
2478 next if !$res->{$vmid}->{pid}; # not running
2479 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2480 }
2481
2482 $qmpclient->queue_execute(undef, 1);
2483
2484 foreach my $vmid (keys %$list) {
2485 next if $opt_vmid && ($vmid ne $opt_vmid);
2486 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2487 }
2488
2489 return $res;
2490 }
2491
2492 sub foreach_drive {
2493 my ($conf, $func) = @_;
2494
2495 foreach my $ds (keys %$conf) {
2496 next if !valid_drivename($ds);
2497
2498 my $drive = parse_drive($ds, $conf->{$ds});
2499 next if !$drive;
2500
2501 &$func($ds, $drive);
2502 }
2503 }
2504
2505 sub foreach_volid {
2506 my ($conf, $func) = @_;
2507
2508 my $volhash = {};
2509
2510 my $test_volid = sub {
2511 my ($volid, $is_cdrom) = @_;
2512
2513 return if !$volid;
2514
2515 $volhash->{$volid} = $is_cdrom || 0;
2516 };
2517
2518 foreach_drive($conf, sub {
2519 my ($ds, $drive) = @_;
2520 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2521 });
2522
2523 foreach my $snapname (keys %{$conf->{snapshots}}) {
2524 my $snap = $conf->{snapshots}->{$snapname};
2525 &$test_volid($snap->{vmstate}, 0);
2526 foreach_drive($snap, sub {
2527 my ($ds, $drive) = @_;
2528 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2529 });
2530 }
2531
2532 foreach my $volid (keys %$volhash) {
2533 &$func($volid, $volhash->{$volid});
2534 }
2535 }
2536
2537 sub vga_conf_has_spice {
2538 my ($vga) = @_;
2539
2540 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2541
2542 return $1 || 1;
2543 }
2544
2545 sub config_to_command {
2546 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2547
2548 my $cmd = [];
2549 my $globalFlags = [];
2550 my $machineFlags = [];
2551 my $rtcFlags = [];
2552 my $cpuFlags = [];
2553 my $devices = [];
2554 my $pciaddr = '';
2555 my $bridges = {};
2556 my $kvmver = kvm_user_version();
2557 my $vernum = 0; # unknown
2558 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2559 $vernum = $1*1000000+$2*1000;
2560 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2561 $vernum = $1*1000000+$2*1000+$3;
2562 }
2563
2564 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2565
2566 my $have_ovz = -f '/proc/vz/vestat';
2567
2568 my $q35 = machine_type_is_q35($conf);
2569
2570 push @$cmd, '/usr/bin/kvm';
2571
2572 push @$cmd, '-id', $vmid;
2573
2574 my $use_virtio = 0;
2575
2576 my $qmpsocket = qmp_socket($vmid);
2577 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2578 push @$cmd, '-mon', "chardev=qmp,mode=control";
2579
2580 my $socket = vnc_socket($vmid);
2581 push @$cmd, '-vnc', "unix:$socket,x509,password";
2582
2583 push @$cmd, '-pidfile' , pidfile_name($vmid);
2584
2585 push @$cmd, '-daemonize';
2586
2587 if ($conf->{smbios1}) {
2588 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2589 }
2590
2591 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2592
2593 if ($q35) {
2594 # the q35 chipset support native usb2, so we enable usb controller
2595 # by default for this machine type
2596 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2597 } else {
2598 $pciaddr = print_pci_addr("piix3", $bridges);
2599 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2600
2601 my $use_usb2 = 0;
2602 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2603 next if !$conf->{"usb$i"};
2604 $use_usb2 = 1;
2605 }
2606 # include usb device config
2607 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2608 }
2609
2610 my $vga = $conf->{vga};
2611
2612 my $qxlnum = vga_conf_has_spice($vga);
2613 $vga = 'qxl' if $qxlnum;
2614
2615 if (!$vga) {
2616 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2617 $conf->{ostype} eq 'win7' ||
2618 $conf->{ostype} eq 'w2k8')) {
2619 $vga = 'std';
2620 } else {
2621 $vga = 'cirrus';
2622 }
2623 }
2624
2625 # enable absolute mouse coordinates (needed by vnc)
2626 my $tablet;
2627 if (defined($conf->{tablet})) {
2628 $tablet = $conf->{tablet};
2629 } else {
2630 $tablet = $defaults->{tablet};
2631 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2632 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2633 }
2634
2635 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2636
2637 # host pci devices
2638 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2639 my $d = parse_hostpci($conf->{"hostpci$i"});
2640 next if !$d;
2641
2642 my $pcie = $d->{pcie};
2643 if($pcie){
2644 die "q35 machine model is not enabled" if !$q35;
2645 $pciaddr = print_pcie_addr("hostpci$i");
2646 }else{
2647 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2648 }
2649
2650 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2651 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2652 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2653 if ($xvga && $xvga ne '') {
2654 push @$cpuFlags, 'kvm=off';
2655 $vga = 'none';
2656 }
2657 $driver = "vfio-pci" if $xvga ne '';
2658 my $pcidevices = $d->{pciid};
2659 my $multifunction = 1 if @$pcidevices > 1;
2660
2661 my $j=0;
2662 foreach my $pcidevice (@$pcidevices) {
2663
2664 my $id = "hostpci$i";
2665 $id .= ".$j" if $multifunction;
2666 my $addr = $pciaddr;
2667 $addr .= ".$j" if $multifunction;
2668 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2669
2670 if($j == 0){
2671 $devicestr .= "$rombar$xvga";
2672 $devicestr .= ",multifunction=on" if $multifunction;
2673 }
2674
2675 push @$devices, '-device', $devicestr;
2676 $j++;
2677 }
2678 }
2679
2680 # usb devices
2681 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2682 my $d = parse_usb_device($conf->{"usb$i"});
2683 next if !$d;
2684 if ($d->{vendorid} && $d->{productid}) {
2685 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2686 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2687 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2688 } elsif ($d->{spice}) {
2689 # usb redir support for spice
2690 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2691 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2692 }
2693 }
2694
2695 # serial devices
2696 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2697 if (my $path = $conf->{"serial$i"}) {
2698 if ($path eq 'socket') {
2699 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2700 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2701 push @$devices, '-device', "isa-serial,chardev=serial$i";
2702 } else {
2703 die "no such serial device\n" if ! -c $path;
2704 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2705 push @$devices, '-device', "isa-serial,chardev=serial$i";
2706 }
2707 }
2708 }
2709
2710 # parallel devices
2711 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2712 if (my $path = $conf->{"parallel$i"}) {
2713 die "no such parallel device\n" if ! -c $path;
2714 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2715 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2716 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2717 }
2718 }
2719
2720 my $vmname = $conf->{name} || "vm$vmid";
2721
2722 push @$cmd, '-name', $vmname;
2723
2724 my $sockets = 1;
2725 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2726 $sockets = $conf->{sockets} if $conf->{sockets};
2727
2728 my $cores = $conf->{cores} || 1;
2729 my $maxcpus = $conf->{maxcpus} if $conf->{maxcpus};
2730
2731 my $total_cores = $sockets * $cores;
2732 my $allowed_cores = $cpuinfo->{cpus};
2733
2734 die "MAX $allowed_cores cores allowed per VM on this node\n"
2735 if ($allowed_cores < $total_cores);
2736
2737 if ($maxcpus) {
2738 push @$cmd, '-smp', "cpus=$cores,maxcpus=$maxcpus";
2739 } else {
2740 push @$cmd, '-smp', "sockets=$sockets,cores=$cores";
2741 }
2742
2743 push @$cmd, '-nodefaults';
2744
2745 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2746
2747 my $bootindex_hash = {};
2748 my $i = 1;
2749 foreach my $o (split(//, $bootorder)) {
2750 $bootindex_hash->{$o} = $i*100;
2751 $i++;
2752 }
2753
2754 push @$cmd, '-boot', "menu=on";
2755
2756 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2757
2758 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2759
2760 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2761
2762 # time drift fix
2763 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2764
2765 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2766 my $useLocaltime = $conf->{localtime};
2767
2768 if (my $ost = $conf->{ostype}) {
2769 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2770
2771 if ($ost =~ m/^w/) { # windows
2772 $useLocaltime = 1 if !defined($conf->{localtime});
2773
2774 # use time drift fix when acpi is enabled
2775 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2776 $tdf = 1 if !defined($conf->{tdf});
2777 }
2778 }
2779
2780 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2781 $ost eq 'wvista') {
2782 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2783 push @$cmd, '-no-hpet';
2784 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2785 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2786 }
2787
2788 if ($ost eq 'win7' || $ost eq 'win8') {
2789 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2790 }
2791 }
2792
2793 push @$rtcFlags, 'driftfix=slew' if $tdf;
2794
2795 if ($nokvm) {
2796 push @$machineFlags, 'accel=tcg';
2797 } else {
2798 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2799 }
2800
2801 my $machine_type = $forcemachine || $conf->{machine};
2802 if ($machine_type) {
2803 push @$machineFlags, "type=${machine_type}";
2804 }
2805
2806 if ($conf->{startdate}) {
2807 push @$rtcFlags, "base=$conf->{startdate}";
2808 } elsif ($useLocaltime) {
2809 push @$rtcFlags, 'base=localtime';
2810 }
2811
2812 my $cpu = $nokvm ? "qemu64" : "kvm64";
2813 $cpu = $conf->{cpu} if $conf->{cpu};
2814
2815 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2816
2817 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2818
2819 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2820
2821 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2822
2823 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2824
2825 # Note: enforce needs kernel 3.10, so we do not use it for now
2826 # push @$cmd, '-cpu', "$cpu,enforce";
2827 push @$cmd, '-cpu', $cpu;
2828
2829 my $memory = $conf->{memory} || $defaults->{memory};
2830 push @$cmd, '-m', $memory;
2831
2832 if ($conf->{numa}) {
2833
2834 my $numa_totalmemory = undef;
2835 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2836 next if !$conf->{"numa$i"};
2837 my $numa = parse_numa($conf->{"numa$i"});
2838 next if !$numa;
2839 # memory
2840 die "missing numa node$i memory value\n" if !$numa->{memory};
2841 my $numa_memory = $numa->{memory};
2842 $numa_totalmemory += $numa_memory;
2843 my $numa_object = "memory-backend-ram,id=ram-node$i,size=$numa_memory"."M";
2844
2845 # cpus
2846 my $cpus_start = $numa->{cpus}->{start};
2847 die "missing numa node$i cpus\n" if !defined($cpus_start);
2848 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2849 my $cpus = $cpus_start;
2850 if (defined($cpus_end)) {
2851 $cpus .= "-$cpus_end";
2852 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2853 }
2854
2855 # hostnodes
2856 my $hostnodes_start = $numa->{hostnodes}->{start};
2857 if (defined($hostnodes_start)) {
2858 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2859 my $hostnodes = $hostnodes_start;
2860 if (defined($hostnodes_end)) {
2861 $hostnodes .= "-$hostnodes_end";
2862 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2863 }
2864
2865 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2866 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2867 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2868 }
2869
2870 # policy
2871 my $policy = $numa->{policy};
2872 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2873 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2874 }
2875
2876 push @$cmd, '-object', $numa_object;
2877 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2878 }
2879
2880 die "total memory for NUMA nodes must be equal to vm memory\n"
2881 if $numa_totalmemory && $numa_totalmemory != $memory;
2882
2883 #if no custom tology, we split memory and cores across numa nodes
2884 if(!$numa_totalmemory) {
2885
2886 my $numa_memory = ($memory / $sockets) . "M";
2887
2888 for (my $i = 0; $i < $sockets; $i++) {
2889
2890 my $cpustart = ($cores * $i);
2891 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2892 my $cpus = $cpustart;
2893 $cpus .= "-$cpuend" if $cpuend;
2894
2895 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2896 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2897 }
2898 }
2899 }
2900
2901 push @$cmd, '-S' if $conf->{freeze};
2902
2903 # set keyboard layout
2904 my $kb = $conf->{keyboard} || $defaults->{keyboard};
2905 push @$cmd, '-k', $kb if $kb;
2906
2907 # enable sound
2908 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
2909 #push @$cmd, '-soundhw', 'es1370';
2910 #push @$cmd, '-soundhw', $soundhw if $soundhw;
2911
2912 if($conf->{agent}) {
2913 my $qgasocket = qmp_socket($vmid, 1);
2914 my $pciaddr = print_pci_addr("qga0", $bridges);
2915 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
2916 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
2917 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
2918 }
2919
2920 my $spice_port;
2921
2922 if ($qxlnum) {
2923 if ($qxlnum > 1) {
2924 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
2925 for(my $i = 1; $i < $qxlnum; $i++){
2926 my $pciaddr = print_pci_addr("vga$i", $bridges);
2927 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
2928 }
2929 } else {
2930 # assume other OS works like Linux
2931 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
2932 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
2933 }
2934 }
2935
2936 my $pciaddr = print_pci_addr("spice", $bridges);
2937
2938 $spice_port = PVE::Tools::next_spice_port();
2939
2940 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
2941
2942 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
2943 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
2944 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
2945 }
2946
2947 # enable balloon by default, unless explicitly disabled
2948 if (!defined($conf->{balloon}) || $conf->{balloon}) {
2949 $pciaddr = print_pci_addr("balloon0", $bridges);
2950 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
2951 }
2952
2953 if ($conf->{watchdog}) {
2954 my $wdopts = parse_watchdog($conf->{watchdog});
2955 $pciaddr = print_pci_addr("watchdog", $bridges);
2956 my $watchdog = $wdopts->{model} || 'i6300esb';
2957 push @$devices, '-device', "$watchdog$pciaddr";
2958 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
2959 }
2960
2961 my $vollist = [];
2962 my $scsicontroller = {};
2963 my $ahcicontroller = {};
2964 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
2965
2966 # Add iscsi initiator name if available
2967 if (my $initiator = get_initiator_name()) {
2968 push @$devices, '-iscsi', "initiator-name=$initiator";
2969 }
2970
2971 foreach_drive($conf, sub {
2972 my ($ds, $drive) = @_;
2973
2974 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
2975 push @$vollist, $drive->{file};
2976 }
2977
2978 $use_virtio = 1 if $ds =~ m/^virtio/;
2979
2980 if (drive_is_cdrom ($drive)) {
2981 if ($bootindex_hash->{d}) {
2982 $drive->{bootindex} = $bootindex_hash->{d};
2983 $bootindex_hash->{d} += 1;
2984 }
2985 } else {
2986 if ($bootindex_hash->{c}) {
2987 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
2988 $bootindex_hash->{c} += 1;
2989 }
2990 }
2991
2992 if ($drive->{interface} eq 'scsi') {
2993
2994 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
2995 my $controller = int($drive->{index} / $maxdev);
2996 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
2997 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
2998 $scsicontroller->{$controller}=1;
2999 }
3000
3001 if ($drive->{interface} eq 'sata') {
3002 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3003 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3004 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3005 $ahcicontroller->{$controller}=1;
3006 }
3007
3008 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3009 push @$devices, '-drive',$drive_cmd;
3010 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3011 });
3012
3013 for (my $i = 0; $i < $MAX_NETS; $i++) {
3014 next if !$conf->{"net$i"};
3015 my $d = parse_net($conf->{"net$i"});
3016 next if !$d;
3017
3018 $use_virtio = 1 if $d->{model} eq 'virtio';
3019
3020 if ($bootindex_hash->{n}) {
3021 $d->{bootindex} = $bootindex_hash->{n};
3022 $bootindex_hash->{n} += 1;
3023 }
3024
3025 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3026 push @$devices, '-netdev', $netdevfull;
3027
3028 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3029 push @$devices, '-device', $netdevicefull;
3030 }
3031
3032 if (!$q35) {
3033 # add pci bridges
3034 while (my ($k, $v) = each %$bridges) {
3035 $pciaddr = print_pci_addr("pci.$k");
3036 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3037 }
3038 }
3039
3040 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3041 # when the VM uses virtio devices.
3042 if (!$use_virtio && $have_ovz) {
3043
3044 my $cpuunits = defined($conf->{cpuunits}) ?
3045 $conf->{cpuunits} : $defaults->{cpuunits};
3046
3047 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3048
3049 # fixme: cpulimit is currently ignored
3050 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3051 }
3052
3053 # add custom args
3054 if ($conf->{args}) {
3055 my $aa = PVE::Tools::split_args($conf->{args});
3056 push @$cmd, @$aa;
3057 }
3058
3059 push @$cmd, @$devices;
3060 push @$cmd, '-rtc', join(',', @$rtcFlags)
3061 if scalar(@$rtcFlags);
3062 push @$cmd, '-machine', join(',', @$machineFlags)
3063 if scalar(@$machineFlags);
3064 push @$cmd, '-global', join(',', @$globalFlags)
3065 if scalar(@$globalFlags);
3066
3067 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3068 }
3069
3070 sub vnc_socket {
3071 my ($vmid) = @_;
3072 return "${var_run_tmpdir}/$vmid.vnc";
3073 }
3074
3075 sub spice_port {
3076 my ($vmid) = @_;
3077
3078 my $res = vm_mon_cmd($vmid, 'query-spice');
3079
3080 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3081 }
3082
3083 sub qmp_socket {
3084 my ($vmid, $qga) = @_;
3085 my $sockettype = $qga ? 'qga' : 'qmp';
3086 return "${var_run_tmpdir}/$vmid.$sockettype";
3087 }
3088
3089 sub pidfile_name {
3090 my ($vmid) = @_;
3091 return "${var_run_tmpdir}/$vmid.pid";
3092 }
3093
3094 sub vm_devices_list {
3095 my ($vmid) = @_;
3096
3097 my $res = vm_mon_cmd($vmid, 'query-pci');
3098
3099 my $devices = {};
3100 foreach my $pcibus (@$res) {
3101 foreach my $device (@{$pcibus->{devices}}) {
3102 next if !$device->{'qdev_id'};
3103 $devices->{$device->{'qdev_id'}} = 1;
3104 }
3105 }
3106
3107 my $resblock = vm_mon_cmd($vmid, 'query-block');
3108 foreach my $block (@$resblock) {
3109 if($block->{device} =~ m/^drive-(\S+)/){
3110 $devices->{$1} = 1;
3111 }
3112 }
3113
3114 return $devices;
3115 }
3116
3117 sub vm_deviceplug {
3118 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3119
3120 return 1 if !check_running($vmid);
3121
3122 my $q35 = machine_type_is_q35($conf);
3123
3124 if ($deviceid eq 'tablet') {
3125 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3126 return 1;
3127 }
3128
3129 return 1 if !$conf->{hotplug};
3130
3131 my $devices_list = vm_devices_list($vmid);
3132 return 1 if defined($devices_list->{$deviceid});
3133
3134 qemu_bridgeadd($storecfg, $conf, $vmid, $deviceid); #add bridge if we need it for the device
3135
3136 if ($deviceid =~ m/^(virtio)(\d+)$/) {
3137 return undef if !qemu_driveadd($storecfg, $vmid, $device);
3138 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3139 qemu_deviceadd($vmid, $devicefull);
3140 if(!qemu_deviceaddverify($vmid, $deviceid)) {
3141 qemu_drivedel($vmid, $deviceid);
3142 return undef;
3143 }
3144 }
3145
3146 if ($deviceid =~ m/^(scsihw)(\d+)$/) {
3147 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3148 my $pciaddr = print_pci_addr($deviceid);
3149 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3150 qemu_deviceadd($vmid, $devicefull);
3151 return undef if(!qemu_deviceaddverify($vmid, $deviceid));
3152 }
3153
3154 if ($deviceid =~ m/^(scsi)(\d+)$/) {
3155 return undef if !qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3156 return undef if !qemu_driveadd($storecfg, $vmid, $device);
3157 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3158 if(!qemu_deviceadd($vmid, $devicefull)) {
3159 qemu_drivedel($vmid, $deviceid);
3160 return undef;
3161 }
3162 }
3163
3164 if ($deviceid =~ m/^(net)(\d+)$/) {
3165 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3166 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3167 qemu_deviceadd($vmid, $netdevicefull);
3168 if(!qemu_deviceaddverify($vmid, $deviceid)) {
3169 qemu_netdevdel($vmid, $deviceid);
3170 return undef;
3171 }
3172 }
3173
3174
3175 if (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3176 my $bridgeid = $2;
3177 my $pciaddr = print_pci_addr($deviceid);
3178 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3179 qemu_deviceadd($vmid, $devicefull);
3180 return undef if !qemu_deviceaddverify($vmid, $deviceid);
3181 }
3182
3183 return 1;
3184 }
3185
3186 sub vm_deviceunplug {
3187 my ($vmid, $conf, $deviceid) = @_;
3188
3189 return 1 if !check_running ($vmid);
3190
3191 if ($deviceid eq 'tablet') {
3192 qemu_devicedel($vmid, $deviceid);
3193 return 1;
3194 }
3195
3196 return 1 if !$conf->{hotplug};
3197
3198 my $devices_list = vm_devices_list($vmid);
3199 return 1 if !defined($devices_list->{$deviceid});
3200
3201 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3202
3203 if ($deviceid =~ m/^(virtio)(\d+)$/) {
3204 qemu_devicedel($vmid, $deviceid);
3205 return undef if !qemu_devicedelverify($vmid, $deviceid);
3206 return undef if !qemu_drivedel($vmid, $deviceid);
3207 }
3208
3209 if ($deviceid =~ m/^(lsi)(\d+)$/) {
3210 return undef if !qemu_devicedel($vmid, $deviceid);
3211 }
3212
3213 if ($deviceid =~ m/^(scsi)(\d+)$/) {
3214 return undef if !qemu_devicedel($vmid, $deviceid);
3215 return undef if !qemu_drivedel($vmid, $deviceid);
3216 }
3217
3218 if ($deviceid =~ m/^(net)(\d+)$/) {
3219 qemu_devicedel($vmid, $deviceid);
3220 return undef if !qemu_devicedelverify($vmid, $deviceid);
3221 return undef if !qemu_netdevdel($vmid, $deviceid);
3222 }
3223
3224 return 1;
3225 }
3226
3227 sub qemu_deviceadd {
3228 my ($vmid, $devicefull) = @_;
3229
3230 $devicefull = "driver=".$devicefull;
3231 my %options = split(/[=,]/, $devicefull);
3232
3233 vm_mon_cmd($vmid, "device_add" , %options);
3234 return 1;
3235 }
3236
3237 sub qemu_devicedel {
3238 my($vmid, $deviceid) = @_;
3239 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3240 return 1;
3241 }
3242
3243 sub qemu_driveadd {
3244 my($storecfg, $vmid, $device) = @_;
3245
3246 my $drive = print_drive_full($storecfg, $vmid, $device);
3247 my $ret = vm_human_monitor_command($vmid, "drive_add auto $drive");
3248 # If the command succeeds qemu prints: "OK"
3249 if ($ret !~ m/OK/s) {
3250 syslog("err", "adding drive failed: $ret");
3251 return undef;
3252 }
3253 return 1;
3254 }
3255
3256 sub qemu_drivedel {
3257 my($vmid, $deviceid) = @_;
3258
3259 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3260 $ret =~ s/^\s+//;
3261 if ($ret =~ m/Device \'.*?\' not found/s) {
3262 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3263 }
3264 elsif ($ret ne "") {
3265 syslog("err", "deleting drive $deviceid failed : $ret");
3266 return undef;
3267 }
3268 return 1;
3269 }
3270
3271 sub qemu_deviceaddverify {
3272 my ($vmid,$deviceid) = @_;
3273
3274 for (my $i = 0; $i <= 5; $i++) {
3275 my $devices_list = vm_devices_list($vmid);
3276 return 1 if defined($devices_list->{$deviceid});
3277 sleep 1;
3278 }
3279 syslog("err", "error on hotplug device $deviceid");
3280 return undef;
3281 }
3282
3283
3284 sub qemu_devicedelverify {
3285 my ($vmid,$deviceid) = @_;
3286
3287 #need to verify the device is correctly remove as device_del is async and empty return is not reliable
3288 for (my $i = 0; $i <= 5; $i++) {
3289 my $devices_list = vm_devices_list($vmid);
3290 return 1 if !defined($devices_list->{$deviceid});
3291 sleep 1;
3292 }
3293 syslog("err", "error on hot-unplugging device $deviceid");
3294 return undef;
3295 }
3296
3297 sub qemu_findorcreatescsihw {
3298 my ($storecfg, $conf, $vmid, $device) = @_;
3299
3300 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3301 my $controller = int($device->{index} / $maxdev);
3302 my $scsihwid="scsihw$controller";
3303 my $devices_list = vm_devices_list($vmid);
3304
3305 if(!defined($devices_list->{$scsihwid})) {
3306 return undef if !vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3307 }
3308 return 1;
3309 }
3310
3311 sub qemu_bridgeadd {
3312 my ($storecfg, $conf, $vmid, $device) = @_;
3313
3314 my $bridges = {};
3315 my $bridgeid = undef;
3316 print_pci_addr($device, $bridges);
3317
3318 while (my ($k, $v) = each %$bridges) {
3319 $bridgeid = $k;
3320 }
3321 return if !$bridgeid || $bridgeid < 1;
3322 my $bridge = "pci.$bridgeid";
3323 my $devices_list = vm_devices_list($vmid);
3324
3325 if(!defined($devices_list->{$bridge})) {
3326 return undef if !vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3327 }
3328 return 1;
3329 }
3330
3331 sub qemu_netdevadd {
3332 my ($vmid, $conf, $device, $deviceid) = @_;
3333
3334 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3335 my %options = split(/[=,]/, $netdev);
3336
3337 vm_mon_cmd($vmid, "netdev_add", %options);
3338 return 1;
3339 }
3340
3341 sub qemu_netdevdel {
3342 my ($vmid, $deviceid) = @_;
3343
3344 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3345 return 1;
3346 }
3347
3348 sub qemu_cpu_hotplug {
3349 my ($vmid, $conf, $cores) = @_;
3350
3351 die "new cores config is not defined" if !$cores;
3352 die "you can't add more cores than maxcpus"
3353 if $conf->{maxcpus} && ($cores > $conf->{maxcpus});
3354 return if !check_running($vmid);
3355
3356 my $currentcores = $conf->{cores} if $conf->{cores};
3357 die "current cores is not defined" if !$currentcores;
3358 die "maxcpus is not defined" if !$conf->{maxcpus};
3359 raise_param_exc({ 'cores' => "online cpu unplug is not yet possible" })
3360 if($cores < $currentcores);
3361
3362 my $currentrunningcores = vm_mon_cmd($vmid, "query-cpus");
3363 raise_param_exc({ 'cores' => "cores number if running vm is different than configuration" })
3364 if scalar (@{$currentrunningcores}) != $currentcores;
3365
3366 for(my $i = $currentcores; $i < $cores; $i++) {
3367 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3368 }
3369 }
3370
3371 sub qemu_block_set_io_throttle {
3372 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3373
3374 return if !check_running($vmid) ;
3375
3376 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($iops_wr));
3377
3378 }
3379
3380 # old code, only used to shutdown old VM after update
3381 sub __read_avail {
3382 my ($fh, $timeout) = @_;
3383
3384 my $sel = new IO::Select;
3385 $sel->add($fh);
3386
3387 my $res = '';
3388 my $buf;
3389
3390 my @ready;
3391 while (scalar (@ready = $sel->can_read($timeout))) {
3392 my $count;
3393 if ($count = $fh->sysread($buf, 8192)) {
3394 if ($buf =~ /^(.*)\(qemu\) $/s) {
3395 $res .= $1;
3396 last;
3397 } else {
3398 $res .= $buf;
3399 }
3400 } else {
3401 if (!defined($count)) {
3402 die "$!\n";
3403 }
3404 last;
3405 }
3406 }
3407
3408 die "monitor read timeout\n" if !scalar(@ready);
3409
3410 return $res;
3411 }
3412
3413 # old code, only used to shutdown old VM after update
3414 sub vm_monitor_command {
3415 my ($vmid, $cmdstr, $nocheck) = @_;
3416
3417 my $res;
3418
3419 eval {
3420 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3421
3422 my $sname = "${var_run_tmpdir}/$vmid.mon";
3423
3424 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3425 die "unable to connect to VM $vmid socket - $!\n";
3426
3427 my $timeout = 3;
3428
3429 # hack: migrate sometime blocks the monitor (when migrate_downtime
3430 # is set)
3431 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3432 $timeout = 60*60; # 1 hour
3433 }
3434
3435 # read banner;
3436 my $data = __read_avail($sock, $timeout);
3437
3438 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3439 die "got unexpected qemu monitor banner\n";
3440 }
3441
3442 my $sel = new IO::Select;
3443 $sel->add($sock);
3444
3445 if (!scalar(my @ready = $sel->can_write($timeout))) {
3446 die "monitor write error - timeout";
3447 }
3448
3449 my $fullcmd = "$cmdstr\r";
3450
3451 # syslog('info', "VM $vmid monitor command: $cmdstr");
3452
3453 my $b;
3454 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3455 die "monitor write error - $!";
3456 }
3457
3458 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3459
3460 $timeout = 20;
3461
3462 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3463 $timeout = 60*60; # 1 hour
3464 } elsif ($cmdstr =~ m/^(eject|change)/) {
3465 $timeout = 60; # note: cdrom mount command is slow
3466 }
3467 if ($res = __read_avail($sock, $timeout)) {
3468
3469 my @lines = split("\r?\n", $res);
3470
3471 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3472
3473 $res = join("\n", @lines);
3474 $res .= "\n";
3475 }
3476 };
3477
3478 my $err = $@;
3479
3480 if ($err) {
3481 syslog("err", "VM $vmid monitor command failed - $err");
3482 die $err;
3483 }
3484
3485 return $res;
3486 }
3487
3488 sub qemu_block_resize {
3489 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3490
3491 my $running = check_running($vmid);
3492
3493 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3494
3495 return if !$running;
3496
3497 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3498
3499 }
3500
3501 sub qemu_volume_snapshot {
3502 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3503
3504 my $running = check_running($vmid);
3505
3506 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3507
3508 return if !$running;
3509
3510 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3511
3512 }
3513
3514 sub qemu_volume_snapshot_delete {
3515 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3516
3517 my $running = check_running($vmid);
3518
3519 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3520
3521 return if !$running;
3522
3523 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3524 }
3525
3526 sub set_migration_caps {
3527 my ($vmid) = @_;
3528
3529 my $cap_ref = [];
3530
3531 my $enabled_cap = {
3532 "auto-converge" => 1,
3533 "xbzrle" => 0,
3534 "x-rdma-pin-all" => 0,
3535 "zero-blocks" => 0,
3536 };
3537
3538 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3539
3540 for my $supported_capability (@$supported_capabilities) {
3541 push @$cap_ref, {
3542 capability => $supported_capability->{capability},
3543 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3544 };
3545 }
3546
3547 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3548 }
3549
3550
3551 sub vmconfig_apply_pending {
3552 my ($vmid, $conf, $storecfg, $running) = @_;
3553
3554 die "implement me - vm is running" if $running; # fixme: if $conf->{hotplug};
3555
3556 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3557 foreach my $opt (@delete) { # delete
3558 die "internal error" if $opt =~ m/^unused/;
3559 $conf = load_config($vmid); # update/reload
3560 if (!defined($conf->{$opt})) {
3561 vmconfig_undelete_pending_option($conf, $opt);
3562 update_config_nolock($vmid, $conf, 1);
3563 } elsif (valid_drivename($opt)) {
3564 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3565 vmconfig_undelete_pending_option($conf, $opt);
3566 delete $conf->{$opt};
3567 update_config_nolock($vmid, $conf, 1);
3568 } else {
3569 vmconfig_undelete_pending_option($conf, $opt);
3570 delete $conf->{$opt};
3571 update_config_nolock($vmid, $conf, 1);
3572 }
3573 }
3574
3575 $conf = load_config($vmid); # update/reload
3576
3577 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3578 $conf = load_config($vmid); # update/reload
3579
3580 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3581 # skip if nothing changed
3582 } elsif (valid_drivename($opt)) {
3583 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3584 if defined($conf->{$opt});
3585 $conf->{$opt} = $conf->{pending}->{$opt};
3586 } else {
3587 $conf->{$opt} = $conf->{pending}->{$opt};
3588 }
3589
3590 delete $conf->{pending}->{$opt};
3591 update_config_nolock($vmid, $conf, 1);
3592 }
3593 }
3594
3595 sub vm_start {
3596 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
3597
3598 lock_config($vmid, sub {
3599 my $conf = load_config($vmid, $migratedfrom);
3600
3601 die "you can't start a vm if it's a template\n" if is_template($conf);
3602
3603 check_lock($conf) if !$skiplock;
3604
3605 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
3606
3607 if (!$statefile && scalar(keys %{$conf->{pending}})) {
3608 vmconfig_apply_pending($vmid, $conf, $storecfg, 0);
3609 $conf = load_config($vmid); # update/reload
3610 }
3611
3612 my $defaults = load_defaults();
3613
3614 # set environment variable useful inside network script
3615 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
3616
3617 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
3618
3619 my $migrate_port = 0;
3620 my $migrate_uri;
3621 if ($statefile) {
3622 if ($statefile eq 'tcp') {
3623 my $localip = "localhost";
3624 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
3625 if ($datacenterconf->{migration_unsecure}) {
3626 my $nodename = PVE::INotify::nodename();
3627 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
3628 }
3629 $migrate_port = PVE::Tools::next_migrate_port();
3630 $migrate_uri = "tcp:${localip}:${migrate_port}";
3631 push @$cmd, '-incoming', $migrate_uri;
3632 push @$cmd, '-S';
3633 } else {
3634 push @$cmd, '-loadstate', $statefile;
3635 }
3636 } elsif ($paused) {
3637 push @$cmd, '-S';
3638 }
3639
3640 # host pci devices
3641 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
3642 my $d = parse_hostpci($conf->{"hostpci$i"});
3643 next if !$d;
3644 my $pcidevices = $d->{pciid};
3645 foreach my $pcidevice (@$pcidevices) {
3646 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
3647
3648 my $info = pci_device_info("0000:$pciid");
3649 die "IOMMU not present\n" if !check_iommu_support();
3650 die "no pci device info for device '$pciid'\n" if !$info;
3651
3652 if ($d->{driver} && $d->{driver} eq "vfio") {
3653 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
3654 } else {
3655 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
3656 }
3657
3658 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
3659 }
3660 }
3661
3662 PVE::Storage::activate_volumes($storecfg, $vollist);
3663
3664 eval { run_command($cmd, timeout => $statefile ? undef : 30,
3665 umask => 0077); };
3666 my $err = $@;
3667 die "start failed: $err" if $err;
3668
3669 print "migration listens on $migrate_uri\n" if $migrate_uri;
3670
3671 if ($statefile && $statefile ne 'tcp') {
3672 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
3673 warn $@ if $@;
3674 }
3675
3676 if ($migratedfrom) {
3677
3678 eval {
3679 PVE::QemuServer::set_migration_caps($vmid);
3680 };
3681 warn $@ if $@;
3682
3683 if ($spice_port) {
3684 print "spice listens on port $spice_port\n";
3685 if ($spice_ticket) {
3686 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
3687 PVE::QemuServer::vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
3688 }
3689 }
3690
3691 } else {
3692
3693 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
3694 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
3695 if $conf->{balloon};
3696 vm_mon_cmd_nocheck($vmid, 'qom-set',
3697 path => "machine/peripheral/balloon0",
3698 property => "guest-stats-polling-interval",
3699 value => 2);
3700 }
3701 }
3702 });
3703 }
3704
3705 sub vm_mon_cmd {
3706 my ($vmid, $execute, %params) = @_;
3707
3708 my $cmd = { execute => $execute, arguments => \%params };
3709 vm_qmp_command($vmid, $cmd);
3710 }
3711
3712 sub vm_mon_cmd_nocheck {
3713 my ($vmid, $execute, %params) = @_;
3714
3715 my $cmd = { execute => $execute, arguments => \%params };
3716 vm_qmp_command($vmid, $cmd, 1);
3717 }
3718
3719 sub vm_qmp_command {
3720 my ($vmid, $cmd, $nocheck) = @_;
3721
3722 my $res;
3723
3724 my $timeout;
3725 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
3726 $timeout = $cmd->{arguments}->{timeout};
3727 delete $cmd->{arguments}->{timeout};
3728 }
3729
3730 eval {
3731 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3732 my $sname = qmp_socket($vmid);
3733 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
3734 my $qmpclient = PVE::QMPClient->new();
3735
3736 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
3737 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
3738 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
3739 if scalar(%{$cmd->{arguments}});
3740 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
3741 } else {
3742 die "unable to open monitor socket\n";
3743 }
3744 };
3745 if (my $err = $@) {
3746 syslog("err", "VM $vmid qmp command failed - $err");
3747 die $err;
3748 }
3749
3750 return $res;
3751 }
3752
3753 sub vm_human_monitor_command {
3754 my ($vmid, $cmdline) = @_;
3755
3756 my $res;
3757
3758 my $cmd = {
3759 execute => 'human-monitor-command',
3760 arguments => { 'command-line' => $cmdline},
3761 };
3762
3763 return vm_qmp_command($vmid, $cmd);
3764 }
3765
3766 sub vm_commandline {
3767 my ($storecfg, $vmid) = @_;
3768
3769 my $conf = load_config($vmid);
3770
3771 my $defaults = load_defaults();
3772
3773 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
3774
3775 return join(' ', @$cmd);
3776 }
3777
3778 sub vm_reset {
3779 my ($vmid, $skiplock) = @_;
3780
3781 lock_config($vmid, sub {
3782
3783 my $conf = load_config($vmid);
3784
3785 check_lock($conf) if !$skiplock;
3786
3787 vm_mon_cmd($vmid, "system_reset");
3788 });
3789 }
3790
3791 sub get_vm_volumes {
3792 my ($conf) = @_;
3793
3794 my $vollist = [];
3795 foreach_volid($conf, sub {
3796 my ($volid, $is_cdrom) = @_;
3797
3798 return if $volid =~ m|^/|;
3799
3800 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
3801 return if !$sid;
3802
3803 push @$vollist, $volid;
3804 });
3805
3806 return $vollist;
3807 }
3808
3809 sub vm_stop_cleanup {
3810 my ($storecfg, $vmid, $conf, $keepActive) = @_;
3811
3812 eval {
3813 fairsched_rmnod($vmid); # try to destroy group
3814
3815 if (!$keepActive) {
3816 my $vollist = get_vm_volumes($conf);
3817 PVE::Storage::deactivate_volumes($storecfg, $vollist);
3818 }
3819
3820 foreach my $ext (qw(mon qmp pid vnc qga)) {
3821 unlink "/var/run/qemu-server/${vmid}.$ext";
3822 }
3823 };
3824 warn $@ if $@; # avoid errors - just warn
3825 }
3826
3827 # Note: use $nockeck to skip tests if VM configuration file exists.
3828 # We need that when migration VMs to other nodes (files already moved)
3829 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
3830 sub vm_stop {
3831 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
3832
3833 $force = 1 if !defined($force) && !$shutdown;
3834
3835 if ($migratedfrom){
3836 my $pid = check_running($vmid, $nocheck, $migratedfrom);
3837 kill 15, $pid if $pid;
3838 my $conf = load_config($vmid, $migratedfrom);
3839 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive);
3840 return;
3841 }
3842
3843 lock_config($vmid, sub {
3844
3845 my $pid = check_running($vmid, $nocheck);
3846 return if !$pid;
3847
3848 my $conf;
3849 if (!$nocheck) {
3850 $conf = load_config($vmid);
3851 check_lock($conf) if !$skiplock;
3852 if (!defined($timeout) && $shutdown && $conf->{startup}) {
3853 my $opts = parse_startup($conf->{startup});
3854 $timeout = $opts->{down} if $opts->{down};
3855 }
3856 }
3857
3858 $timeout = 60 if !defined($timeout);
3859
3860 eval {
3861 if ($shutdown) {
3862 if (!$nocheck && $conf->{agent}) {
3863 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
3864 } else {
3865 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
3866 }
3867 } else {
3868 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
3869 }
3870 };
3871 my $err = $@;
3872
3873 if (!$err) {
3874 my $count = 0;
3875 while (($count < $timeout) && check_running($vmid, $nocheck)) {
3876 $count++;
3877 sleep 1;
3878 }
3879
3880 if ($count >= $timeout) {
3881 if ($force) {
3882 warn "VM still running - terminating now with SIGTERM\n";
3883 kill 15, $pid;
3884 } else {
3885 die "VM quit/powerdown failed - got timeout\n";
3886 }
3887 } else {
3888 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
3889 return;
3890 }
3891 } else {
3892 if ($force) {
3893 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
3894 kill 15, $pid;
3895 } else {
3896 die "VM quit/powerdown failed\n";
3897 }
3898 }
3899
3900 # wait again
3901 $timeout = 10;
3902
3903 my $count = 0;
3904 while (($count < $timeout) && check_running($vmid, $nocheck)) {
3905 $count++;
3906 sleep 1;
3907 }
3908
3909 if ($count >= $timeout) {
3910 warn "VM still running - terminating now with SIGKILL\n";
3911 kill 9, $pid;
3912 sleep 1;
3913 }
3914
3915 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive) if $conf;
3916 });
3917 }
3918
3919 sub vm_suspend {
3920 my ($vmid, $skiplock) = @_;
3921
3922 lock_config($vmid, sub {
3923
3924 my $conf = load_config($vmid);
3925
3926 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
3927
3928 vm_mon_cmd($vmid, "stop");
3929 });
3930 }
3931
3932 sub vm_resume {
3933 my ($vmid, $skiplock) = @_;
3934
3935 lock_config($vmid, sub {
3936
3937 my $conf = load_config($vmid);
3938
3939 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
3940
3941 vm_mon_cmd($vmid, "cont");
3942 });
3943 }
3944
3945 sub vm_sendkey {
3946 my ($vmid, $skiplock, $key) = @_;
3947
3948 lock_config($vmid, sub {
3949
3950 my $conf = load_config($vmid);
3951
3952 # there is no qmp command, so we use the human monitor command
3953 vm_human_monitor_command($vmid, "sendkey $key");
3954 });
3955 }
3956
3957 sub vm_destroy {
3958 my ($storecfg, $vmid, $skiplock) = @_;
3959
3960 lock_config($vmid, sub {
3961
3962 my $conf = load_config($vmid);
3963
3964 check_lock($conf) if !$skiplock;
3965
3966 if (!check_running($vmid)) {
3967 fairsched_rmnod($vmid); # try to destroy group
3968 destroy_vm($storecfg, $vmid);
3969 } else {
3970 die "VM $vmid is running - destroy failed\n";
3971 }
3972 });
3973 }
3974
3975 # pci helpers
3976
3977 sub file_write {
3978 my ($filename, $buf) = @_;
3979
3980 my $fh = IO::File->new($filename, "w");
3981 return undef if !$fh;
3982
3983 my $res = print $fh $buf;
3984
3985 $fh->close();
3986
3987 return $res;
3988 }
3989
3990 sub pci_device_info {
3991 my ($name) = @_;
3992
3993 my $res;
3994
3995 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
3996 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
3997
3998 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
3999 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4000
4001 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4002 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4003
4004 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4005 return undef if !defined($product) || $product !~ s/^0x//;
4006
4007 $res = {
4008 name => $name,
4009 vendor => $vendor,
4010 product => $product,
4011 domain => $domain,
4012 bus => $bus,
4013 slot => $slot,
4014 func => $func,
4015 irq => $irq,
4016 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4017 };
4018
4019 return $res;
4020 }
4021
4022 sub pci_dev_reset {
4023 my ($dev) = @_;
4024
4025 my $name = $dev->{name};
4026
4027 my $fn = "$pcisysfs/devices/$name/reset";
4028
4029 return file_write($fn, "1");
4030 }
4031
4032 sub pci_dev_bind_to_stub {
4033 my ($dev) = @_;
4034
4035 my $name = $dev->{name};
4036
4037 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4038 return 1 if -d $testdir;
4039
4040 my $data = "$dev->{vendor} $dev->{product}";
4041 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4042
4043 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4044 if (!file_write($fn, $name)) {
4045 return undef if -f $fn;
4046 }
4047
4048 $fn = "$pcisysfs/drivers/pci-stub/bind";
4049 if (! -d $testdir) {
4050 return undef if !file_write($fn, $name);
4051 }
4052
4053 return -d $testdir;
4054 }
4055
4056 sub pci_dev_bind_to_vfio {
4057 my ($dev) = @_;
4058
4059 my $name = $dev->{name};
4060
4061 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4062
4063 if (!-d $vfio_basedir) {
4064 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4065 }
4066 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4067
4068 my $testdir = "$vfio_basedir/$name";
4069 return 1 if -d $testdir;
4070
4071 my $data = "$dev->{vendor} $dev->{product}";
4072 return undef if !file_write("$vfio_basedir/new_id", $data);
4073
4074 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4075 if (!file_write($fn, $name)) {
4076 return undef if -f $fn;
4077 }
4078
4079 $fn = "$vfio_basedir/bind";
4080 if (! -d $testdir) {
4081 return undef if !file_write($fn, $name);
4082 }
4083
4084 return -d $testdir;
4085 }
4086
4087 sub pci_dev_group_bind_to_vfio {
4088 my ($pciid) = @_;
4089
4090 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4091
4092 if (!-d $vfio_basedir) {
4093 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4094 }
4095 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4096
4097 # get IOMMU group devices
4098 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4099 my @devs = grep /^0000:/, readdir($D);
4100 closedir($D);
4101
4102 foreach my $pciid (@devs) {
4103 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4104 my $info = pci_device_info($1);
4105 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4106 }
4107
4108 return 1;
4109 }
4110
4111 sub print_pci_addr {
4112 my ($id, $bridges) = @_;
4113
4114 my $res = '';
4115 my $devices = {
4116 piix3 => { bus => 0, addr => 1 },
4117 #addr2 : first videocard
4118 balloon0 => { bus => 0, addr => 3 },
4119 watchdog => { bus => 0, addr => 4 },
4120 scsihw0 => { bus => 0, addr => 5 },
4121 scsihw1 => { bus => 0, addr => 6 },
4122 ahci0 => { bus => 0, addr => 7 },
4123 qga0 => { bus => 0, addr => 8 },
4124 spice => { bus => 0, addr => 9 },
4125 virtio0 => { bus => 0, addr => 10 },
4126 virtio1 => { bus => 0, addr => 11 },
4127 virtio2 => { bus => 0, addr => 12 },
4128 virtio3 => { bus => 0, addr => 13 },
4129 virtio4 => { bus => 0, addr => 14 },
4130 virtio5 => { bus => 0, addr => 15 },
4131 hostpci0 => { bus => 0, addr => 16 },
4132 hostpci1 => { bus => 0, addr => 17 },
4133 net0 => { bus => 0, addr => 18 },
4134 net1 => { bus => 0, addr => 19 },
4135 net2 => { bus => 0, addr => 20 },
4136 net3 => { bus => 0, addr => 21 },
4137 net4 => { bus => 0, addr => 22 },
4138 net5 => { bus => 0, addr => 23 },
4139 vga1 => { bus => 0, addr => 24 },
4140 vga2 => { bus => 0, addr => 25 },
4141 vga3 => { bus => 0, addr => 26 },
4142 hostpci2 => { bus => 0, addr => 27 },
4143 hostpci3 => { bus => 0, addr => 28 },
4144 #addr29 : usb-host (pve-usb.cfg)
4145 'pci.1' => { bus => 0, addr => 30 },
4146 'pci.2' => { bus => 0, addr => 31 },
4147 'net6' => { bus => 1, addr => 1 },
4148 'net7' => { bus => 1, addr => 2 },
4149 'net8' => { bus => 1, addr => 3 },
4150 'net9' => { bus => 1, addr => 4 },
4151 'net10' => { bus => 1, addr => 5 },
4152 'net11' => { bus => 1, addr => 6 },
4153 'net12' => { bus => 1, addr => 7 },
4154 'net13' => { bus => 1, addr => 8 },
4155 'net14' => { bus => 1, addr => 9 },
4156 'net15' => { bus => 1, addr => 10 },
4157 'net16' => { bus => 1, addr => 11 },
4158 'net17' => { bus => 1, addr => 12 },
4159 'net18' => { bus => 1, addr => 13 },
4160 'net19' => { bus => 1, addr => 14 },
4161 'net20' => { bus => 1, addr => 15 },
4162 'net21' => { bus => 1, addr => 16 },
4163 'net22' => { bus => 1, addr => 17 },
4164 'net23' => { bus => 1, addr => 18 },
4165 'net24' => { bus => 1, addr => 19 },
4166 'net25' => { bus => 1, addr => 20 },
4167 'net26' => { bus => 1, addr => 21 },
4168 'net27' => { bus => 1, addr => 22 },
4169 'net28' => { bus => 1, addr => 23 },
4170 'net29' => { bus => 1, addr => 24 },
4171 'net30' => { bus => 1, addr => 25 },
4172 'net31' => { bus => 1, addr => 26 },
4173 'virtio6' => { bus => 2, addr => 1 },
4174 'virtio7' => { bus => 2, addr => 2 },
4175 'virtio8' => { bus => 2, addr => 3 },
4176 'virtio9' => { bus => 2, addr => 4 },
4177 'virtio10' => { bus => 2, addr => 5 },
4178 'virtio11' => { bus => 2, addr => 6 },
4179 'virtio12' => { bus => 2, addr => 7 },
4180 'virtio13' => { bus => 2, addr => 8 },
4181 'virtio14' => { bus => 2, addr => 9 },
4182 'virtio15' => { bus => 2, addr => 10 },
4183 };
4184
4185 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4186 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4187 my $bus = $devices->{$id}->{bus};
4188 $res = ",bus=pci.$bus,addr=$addr";
4189 $bridges->{$bus} = 1 if $bridges;
4190 }
4191 return $res;
4192
4193 }
4194
4195 sub print_pcie_addr {
4196 my ($id) = @_;
4197
4198 my $res = '';
4199 my $devices = {
4200 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4201 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4202 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4203 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4204 };
4205
4206 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4207 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4208 my $bus = $devices->{$id}->{bus};
4209 $res = ",bus=$bus,addr=$addr";
4210 }
4211 return $res;
4212
4213 }
4214
4215 # vzdump restore implementaion
4216
4217 sub tar_archive_read_firstfile {
4218 my $archive = shift;
4219
4220 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4221
4222 # try to detect archive type first
4223 my $pid = open (TMP, "tar tf '$archive'|") ||
4224 die "unable to open file '$archive'\n";
4225 my $firstfile = <TMP>;
4226 kill 15, $pid;
4227 close TMP;
4228
4229 die "ERROR: archive contaions no data\n" if !$firstfile;
4230 chomp $firstfile;
4231
4232 return $firstfile;
4233 }
4234
4235 sub tar_restore_cleanup {
4236 my ($storecfg, $statfile) = @_;
4237
4238 print STDERR "starting cleanup\n";
4239
4240 if (my $fd = IO::File->new($statfile, "r")) {
4241 while (defined(my $line = <$fd>)) {
4242 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4243 my $volid = $2;
4244 eval {
4245 if ($volid =~ m|^/|) {
4246 unlink $volid || die 'unlink failed\n';
4247 } else {
4248 PVE::Storage::vdisk_free($storecfg, $volid);
4249 }
4250 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4251 };
4252 print STDERR "unable to cleanup '$volid' - $@" if $@;
4253 } else {
4254 print STDERR "unable to parse line in statfile - $line";
4255 }
4256 }
4257 $fd->close();
4258 }
4259 }
4260
4261 sub restore_archive {
4262 my ($archive, $vmid, $user, $opts) = @_;
4263
4264 my $format = $opts->{format};
4265 my $comp;
4266
4267 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4268 $format = 'tar' if !$format;
4269 $comp = 'gzip';
4270 } elsif ($archive =~ m/\.tar$/) {
4271 $format = 'tar' if !$format;
4272 } elsif ($archive =~ m/.tar.lzo$/) {
4273 $format = 'tar' if !$format;
4274 $comp = 'lzop';
4275 } elsif ($archive =~ m/\.vma$/) {
4276 $format = 'vma' if !$format;
4277 } elsif ($archive =~ m/\.vma\.gz$/) {
4278 $format = 'vma' if !$format;
4279 $comp = 'gzip';
4280 } elsif ($archive =~ m/\.vma\.lzo$/) {
4281 $format = 'vma' if !$format;
4282 $comp = 'lzop';
4283 } else {
4284 $format = 'vma' if !$format; # default
4285 }
4286
4287 # try to detect archive format
4288 if ($format eq 'tar') {
4289 return restore_tar_archive($archive, $vmid, $user, $opts);
4290 } else {
4291 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4292 }
4293 }
4294
4295 sub restore_update_config_line {
4296 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4297
4298 return if $line =~ m/^\#qmdump\#/;
4299 return if $line =~ m/^\#vzdump\#/;
4300 return if $line =~ m/^lock:/;
4301 return if $line =~ m/^unused\d+:/;
4302 return if $line =~ m/^parent:/;
4303 return if $line =~ m/^template:/; # restored VM is never a template
4304
4305 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4306 # try to convert old 1.X settings
4307 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4308 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4309 my ($model, $macaddr) = split(/\=/, $devconfig);
4310 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4311 my $net = {
4312 model => $model,
4313 bridge => "vmbr$ind",
4314 macaddr => $macaddr,
4315 };
4316 my $netstr = print_net($net);
4317
4318 print $outfd "net$cookie->{netcount}: $netstr\n";
4319 $cookie->{netcount}++;
4320 }
4321 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4322 my ($id, $netstr) = ($1, $2);
4323 my $net = parse_net($netstr);
4324 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4325 $netstr = print_net($net);
4326 print $outfd "$id: $netstr\n";
4327 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4328 my $virtdev = $1;
4329 my $value = $3;
4330 if ($line =~ m/backup=no/) {
4331 print $outfd "#$line";
4332 } elsif ($virtdev && $map->{$virtdev}) {
4333 my $di = parse_drive($virtdev, $value);
4334 delete $di->{format}; # format can change on restore
4335 $di->{file} = $map->{$virtdev};
4336 $value = print_drive($vmid, $di);
4337 print $outfd "$virtdev: $value\n";
4338 } else {
4339 print $outfd $line;
4340 }
4341 } else {
4342 print $outfd $line;
4343 }
4344 }
4345
4346 sub scan_volids {
4347 my ($cfg, $vmid) = @_;
4348
4349 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4350
4351 my $volid_hash = {};
4352 foreach my $storeid (keys %$info) {
4353 foreach my $item (@{$info->{$storeid}}) {
4354 next if !($item->{volid} && $item->{size});
4355 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4356 $volid_hash->{$item->{volid}} = $item;
4357 }
4358 }
4359
4360 return $volid_hash;
4361 }
4362
4363 sub get_used_paths {
4364 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4365
4366 my $used_path = {};
4367
4368 my $scan_config = sub {
4369 my ($cref, $snapname) = @_;
4370
4371 foreach my $key (keys %$cref) {
4372 my $value = $cref->{$key};
4373 if (valid_drivename($key)) {
4374 next if $skip_drive && $key eq $skip_drive;
4375 my $drive = parse_drive($key, $value);
4376 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4377 if ($drive->{file} =~ m!^/!) {
4378 $used_path->{$drive->{file}}++; # = 1;
4379 } else {
4380 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4381 next if !$storeid;
4382 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4383 next if !$scfg;
4384 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4385 $used_path->{$path}++; # = 1;
4386 }
4387 }
4388 }
4389 };
4390
4391 &$scan_config($conf);
4392
4393 undef $skip_drive;
4394
4395 if ($scan_snapshots) {
4396 foreach my $snapname (keys %{$conf->{snapshots}}) {
4397 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4398 }
4399 }
4400
4401 return $used_path;
4402 }
4403
4404 sub update_disksize {
4405 my ($vmid, $conf, $volid_hash) = @_;
4406
4407 my $changes;
4408
4409 my $used = {};
4410
4411 # Note: it is allowed to define multiple storages with same path (alias), so
4412 # we need to check both 'volid' and real 'path' (two different volid can point
4413 # to the same path).
4414
4415 my $usedpath = {};
4416
4417 # update size info
4418 foreach my $opt (keys %$conf) {
4419 if (valid_drivename($opt)) {
4420 my $drive = parse_drive($opt, $conf->{$opt});
4421 my $volid = $drive->{file};
4422 next if !$volid;
4423
4424 $used->{$volid} = 1;
4425 if ($volid_hash->{$volid} &&
4426 (my $path = $volid_hash->{$volid}->{path})) {
4427 $usedpath->{$path} = 1;
4428 }
4429
4430 next if drive_is_cdrom($drive);
4431 next if !$volid_hash->{$volid};
4432
4433 $drive->{size} = $volid_hash->{$volid}->{size};
4434 my $new = print_drive($vmid, $drive);
4435 if ($new ne $conf->{$opt}) {
4436 $changes = 1;
4437 $conf->{$opt} = $new;
4438 }
4439 }
4440 }
4441
4442 # remove 'unusedX' entry if volume is used
4443 foreach my $opt (keys %$conf) {
4444 next if $opt !~ m/^unused\d+$/;
4445 my $volid = $conf->{$opt};
4446 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4447 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4448 $changes = 1;
4449 delete $conf->{$opt};
4450 }
4451 }
4452
4453 foreach my $volid (sort keys %$volid_hash) {
4454 next if $volid =~ m/vm-$vmid-state-/;
4455 next if $used->{$volid};
4456 my $path = $volid_hash->{$volid}->{path};
4457 next if !$path; # just to be sure
4458 next if $usedpath->{$path};
4459 $changes = 1;
4460 add_unused_volume($conf, $volid);
4461 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4462 }
4463
4464 return $changes;
4465 }
4466
4467 sub rescan {
4468 my ($vmid, $nolock) = @_;
4469
4470 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4471
4472 my $volid_hash = scan_volids($cfg, $vmid);
4473
4474 my $updatefn = sub {
4475 my ($vmid) = @_;
4476
4477 my $conf = load_config($vmid);
4478
4479 check_lock($conf);
4480
4481 my $vm_volids = {};
4482 foreach my $volid (keys %$volid_hash) {
4483 my $info = $volid_hash->{$volid};
4484 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
4485 }
4486
4487 my $changes = update_disksize($vmid, $conf, $vm_volids);
4488
4489 update_config_nolock($vmid, $conf, 1) if $changes;
4490 };
4491
4492 if (defined($vmid)) {
4493 if ($nolock) {
4494 &$updatefn($vmid);
4495 } else {
4496 lock_config($vmid, $updatefn, $vmid);
4497 }
4498 } else {
4499 my $vmlist = config_list();
4500 foreach my $vmid (keys %$vmlist) {
4501 if ($nolock) {
4502 &$updatefn($vmid);
4503 } else {
4504 lock_config($vmid, $updatefn, $vmid);
4505 }
4506 }
4507 }
4508 }
4509
4510 sub restore_vma_archive {
4511 my ($archive, $vmid, $user, $opts, $comp) = @_;
4512
4513 my $input = $archive eq '-' ? "<&STDIN" : undef;
4514 my $readfrom = $archive;
4515
4516 my $uncomp = '';
4517 if ($comp) {
4518 $readfrom = '-';
4519 my $qarchive = PVE::Tools::shellquote($archive);
4520 if ($comp eq 'gzip') {
4521 $uncomp = "zcat $qarchive|";
4522 } elsif ($comp eq 'lzop') {
4523 $uncomp = "lzop -d -c $qarchive|";
4524 } else {
4525 die "unknown compression method '$comp'\n";
4526 }
4527
4528 }
4529
4530 my $tmpdir = "/var/tmp/vzdumptmp$$";
4531 rmtree $tmpdir;
4532
4533 # disable interrupts (always do cleanups)
4534 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
4535 warn "got interrupt - ignored\n";
4536 };
4537
4538 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
4539 POSIX::mkfifo($mapfifo, 0600);
4540 my $fifofh;
4541
4542 my $openfifo = sub {
4543 open($fifofh, '>', $mapfifo) || die $!;
4544 };
4545
4546 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
4547
4548 my $oldtimeout;
4549 my $timeout = 5;
4550
4551 my $devinfo = {};
4552
4553 my $rpcenv = PVE::RPCEnvironment::get();
4554
4555 my $conffile = config_file($vmid);
4556 my $tmpfn = "$conffile.$$.tmp";
4557
4558 # Note: $oldconf is undef if VM does not exists
4559 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
4560
4561 my $print_devmap = sub {
4562 my $virtdev_hash = {};
4563
4564 my $cfgfn = "$tmpdir/qemu-server.conf";
4565
4566 # we can read the config - that is already extracted
4567 my $fh = IO::File->new($cfgfn, "r") ||
4568 "unable to read qemu-server.conf - $!\n";
4569
4570 while (defined(my $line = <$fh>)) {
4571 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
4572 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
4573 die "archive does not contain data for drive '$virtdev'\n"
4574 if !$devinfo->{$devname};
4575 if (defined($opts->{storage})) {
4576 $storeid = $opts->{storage} || 'local';
4577 } elsif (!$storeid) {
4578 $storeid = 'local';
4579 }
4580 $format = 'raw' if !$format;
4581 $devinfo->{$devname}->{devname} = $devname;
4582 $devinfo->{$devname}->{virtdev} = $virtdev;
4583 $devinfo->{$devname}->{format} = $format;
4584 $devinfo->{$devname}->{storeid} = $storeid;
4585
4586 # check permission on storage
4587 my $pool = $opts->{pool}; # todo: do we need that?
4588 if ($user ne 'root@pam') {
4589 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
4590 }
4591
4592 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
4593 }
4594 }
4595
4596 foreach my $devname (keys %$devinfo) {
4597 die "found no device mapping information for device '$devname'\n"
4598 if !$devinfo->{$devname}->{virtdev};
4599 }
4600
4601 my $cfg = cfs_read_file('storage.cfg');
4602
4603 # create empty/temp config
4604 if ($oldconf) {
4605 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
4606 foreach_drive($oldconf, sub {
4607 my ($ds, $drive) = @_;
4608
4609 return if drive_is_cdrom($drive);
4610
4611 my $volid = $drive->{file};
4612
4613 return if !$volid || $volid =~ m|^/|;
4614
4615 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
4616 return if !$path || !$owner || ($owner != $vmid);
4617
4618 # Note: only delete disk we want to restore
4619 # other volumes will become unused
4620 if ($virtdev_hash->{$ds}) {
4621 PVE::Storage::vdisk_free($cfg, $volid);
4622 }
4623 });
4624 }
4625
4626 my $map = {};
4627 foreach my $virtdev (sort keys %$virtdev_hash) {
4628 my $d = $virtdev_hash->{$virtdev};
4629 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
4630 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
4631
4632 # test if requested format is supported
4633 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
4634 my $supported = grep { $_ eq $d->{format} } @$validFormats;
4635 $d->{format} = $defFormat if !$supported;
4636
4637 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
4638 $d->{format}, undef, $alloc_size);
4639 print STDERR "new volume ID is '$volid'\n";
4640 $d->{volid} = $volid;
4641 my $path = PVE::Storage::path($cfg, $volid);
4642
4643 my $write_zeros = 1;
4644 # fixme: what other storages types initialize volumes with zero?
4645 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
4646 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
4647 $write_zeros = 0;
4648 }
4649
4650 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
4651
4652 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
4653 $map->{$virtdev} = $volid;
4654 }
4655
4656 $fh->seek(0, 0) || die "seek failed - $!\n";
4657
4658 my $outfd = new IO::File ($tmpfn, "w") ||
4659 die "unable to write config for VM $vmid\n";
4660
4661 my $cookie = { netcount => 0 };
4662 while (defined(my $line = <$fh>)) {
4663 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
4664 }
4665
4666 $fh->close();
4667 $outfd->close();
4668 };
4669
4670 eval {
4671 # enable interrupts
4672 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
4673 die "interrupted by signal\n";
4674 };
4675 local $SIG{ALRM} = sub { die "got timeout\n"; };
4676
4677 $oldtimeout = alarm($timeout);
4678
4679 my $parser = sub {
4680 my $line = shift;
4681
4682 print "$line\n";
4683
4684 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
4685 my ($dev_id, $size, $devname) = ($1, $2, $3);
4686 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
4687 } elsif ($line =~ m/^CTIME: /) {
4688 # we correctly received the vma config, so we can disable
4689 # the timeout now for disk allocation (set to 10 minutes, so
4690 # that we always timeout if something goes wrong)
4691 alarm(600);
4692 &$print_devmap();
4693 print $fifofh "done\n";
4694 my $tmp = $oldtimeout || 0;
4695 $oldtimeout = undef;
4696 alarm($tmp);
4697 close($fifofh);
4698 }
4699 };
4700
4701 print "restore vma archive: $cmd\n";
4702 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
4703 };
4704 my $err = $@;
4705
4706 alarm($oldtimeout) if $oldtimeout;
4707
4708 unlink $mapfifo;
4709
4710 if ($err) {
4711 rmtree $tmpdir;
4712 unlink $tmpfn;
4713
4714 my $cfg = cfs_read_file('storage.cfg');
4715 foreach my $devname (keys %$devinfo) {
4716 my $volid = $devinfo->{$devname}->{volid};
4717 next if !$volid;
4718 eval {
4719 if ($volid =~ m|^/|) {
4720 unlink $volid || die 'unlink failed\n';
4721 } else {
4722 PVE::Storage::vdisk_free($cfg, $volid);
4723 }
4724 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4725 };
4726 print STDERR "unable to cleanup '$volid' - $@" if $@;
4727 }
4728 die $err;
4729 }
4730
4731 rmtree $tmpdir;
4732
4733 rename($tmpfn, $conffile) ||
4734 die "unable to commit configuration file '$conffile'\n";
4735
4736 PVE::Cluster::cfs_update(); # make sure we read new file
4737
4738 eval { rescan($vmid, 1); };
4739 warn $@ if $@;
4740 }
4741
4742 sub restore_tar_archive {
4743 my ($archive, $vmid, $user, $opts) = @_;
4744
4745 if ($archive ne '-') {
4746 my $firstfile = tar_archive_read_firstfile($archive);
4747 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
4748 if $firstfile ne 'qemu-server.conf';
4749 }
4750
4751 my $storecfg = cfs_read_file('storage.cfg');
4752
4753 # destroy existing data - keep empty config
4754 my $vmcfgfn = PVE::QemuServer::config_file($vmid);
4755 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
4756
4757 my $tocmd = "/usr/lib/qemu-server/qmextract";
4758
4759 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
4760 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
4761 $tocmd .= ' --prealloc' if $opts->{prealloc};
4762 $tocmd .= ' --info' if $opts->{info};
4763
4764 # tar option "xf" does not autodetect compression when read from STDIN,
4765 # so we pipe to zcat
4766 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
4767 PVE::Tools::shellquote("--to-command=$tocmd");
4768
4769 my $tmpdir = "/var/tmp/vzdumptmp$$";
4770 mkpath $tmpdir;
4771
4772 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
4773 local $ENV{VZDUMP_VMID} = $vmid;
4774 local $ENV{VZDUMP_USER} = $user;
4775
4776 my $conffile = config_file($vmid);
4777 my $tmpfn = "$conffile.$$.tmp";
4778
4779 # disable interrupts (always do cleanups)
4780 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
4781 print STDERR "got interrupt - ignored\n";
4782 };
4783
4784 eval {
4785 # enable interrupts
4786 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
4787 die "interrupted by signal\n";
4788 };
4789
4790 if ($archive eq '-') {
4791 print "extracting archive from STDIN\n";
4792 run_command($cmd, input => "<&STDIN");
4793 } else {
4794 print "extracting archive '$archive'\n";
4795 run_command($cmd);
4796 }
4797
4798 return if $opts->{info};
4799
4800 # read new mapping
4801 my $map = {};
4802 my $statfile = "$tmpdir/qmrestore.stat";
4803 if (my $fd = IO::File->new($statfile, "r")) {
4804 while (defined (my $line = <$fd>)) {
4805 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4806 $map->{$1} = $2 if $1;
4807 } else {
4808 print STDERR "unable to parse line in statfile - $line\n";
4809 }
4810 }
4811 $fd->close();
4812 }
4813
4814 my $confsrc = "$tmpdir/qemu-server.conf";
4815
4816 my $srcfd = new IO::File($confsrc, "r") ||
4817 die "unable to open file '$confsrc'\n";
4818
4819 my $outfd = new IO::File ($tmpfn, "w") ||
4820 die "unable to write config for VM $vmid\n";
4821
4822 my $cookie = { netcount => 0 };
4823 while (defined (my $line = <$srcfd>)) {
4824 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
4825 }
4826
4827 $srcfd->close();
4828 $outfd->close();
4829 };
4830 my $err = $@;
4831
4832 if ($err) {
4833
4834 unlink $tmpfn;
4835
4836 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
4837
4838 die $err;
4839 }
4840
4841 rmtree $tmpdir;
4842
4843 rename $tmpfn, $conffile ||
4844 die "unable to commit configuration file '$conffile'\n";
4845
4846 PVE::Cluster::cfs_update(); # make sure we read new file
4847
4848 eval { rescan($vmid, 1); };
4849 warn $@ if $@;
4850 };
4851
4852
4853 # Internal snapshots
4854
4855 # NOTE: Snapshot create/delete involves several non-atomic
4856 # action, and can take a long time.
4857 # So we try to avoid locking the file and use 'lock' variable
4858 # inside the config file instead.
4859
4860 my $snapshot_copy_config = sub {
4861 my ($source, $dest) = @_;
4862
4863 foreach my $k (keys %$source) {
4864 next if $k eq 'snapshots';
4865 next if $k eq 'snapstate';
4866 next if $k eq 'snaptime';
4867 next if $k eq 'vmstate';
4868 next if $k eq 'lock';
4869 next if $k eq 'digest';
4870 next if $k eq 'description';
4871 next if $k =~ m/^unused\d+$/;
4872
4873 $dest->{$k} = $source->{$k};
4874 }
4875 };
4876
4877 my $snapshot_apply_config = sub {
4878 my ($conf, $snap) = @_;
4879
4880 # copy snapshot list
4881 my $newconf = {
4882 snapshots => $conf->{snapshots},
4883 };
4884
4885 # keep description and list of unused disks
4886 foreach my $k (keys %$conf) {
4887 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
4888 $newconf->{$k} = $conf->{$k};
4889 }
4890
4891 &$snapshot_copy_config($snap, $newconf);
4892
4893 return $newconf;
4894 };
4895
4896 sub foreach_writable_storage {
4897 my ($conf, $func) = @_;
4898
4899 my $sidhash = {};
4900
4901 foreach my $ds (keys %$conf) {
4902 next if !valid_drivename($ds);
4903
4904 my $drive = parse_drive($ds, $conf->{$ds});
4905 next if !$drive;
4906 next if drive_is_cdrom($drive);
4907
4908 my $volid = $drive->{file};
4909
4910 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4911 $sidhash->{$sid} = $sid if $sid;
4912 }
4913
4914 foreach my $sid (sort keys %$sidhash) {
4915 &$func($sid);
4916 }
4917 }
4918
4919 my $alloc_vmstate_volid = sub {
4920 my ($storecfg, $vmid, $conf, $snapname) = @_;
4921
4922 # Note: we try to be smart when selecting a $target storage
4923
4924 my $target;
4925
4926 # search shared storage first
4927 foreach_writable_storage($conf, sub {
4928 my ($sid) = @_;
4929 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
4930 return if !$scfg->{shared};
4931
4932 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
4933 });
4934
4935 if (!$target) {
4936 # now search local storage
4937 foreach_writable_storage($conf, sub {
4938 my ($sid) = @_;
4939 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
4940 return if $scfg->{shared};
4941
4942 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
4943 });
4944 }
4945
4946 $target = 'local' if !$target;
4947
4948 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
4949 # we abort live save after $conf->{memory}, so we need at max twice that space
4950 my $size = $conf->{memory}*2 + $driver_state_size;
4951
4952 my $name = "vm-$vmid-state-$snapname";
4953 my $scfg = PVE::Storage::storage_config($storecfg, $target);
4954 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
4955 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
4956
4957 return $volid;
4958 };
4959
4960 my $snapshot_prepare = sub {
4961 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
4962
4963 my $snap;
4964
4965 my $updatefn = sub {
4966
4967 my $conf = load_config($vmid);
4968
4969 die "you can't take a snapshot if it's a template\n"
4970 if is_template($conf);
4971
4972 check_lock($conf);
4973
4974 $conf->{lock} = 'snapshot';
4975
4976 die "snapshot name '$snapname' already used\n"
4977 if defined($conf->{snapshots}->{$snapname});
4978
4979 my $storecfg = PVE::Storage::config();
4980 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
4981
4982 $snap = $conf->{snapshots}->{$snapname} = {};
4983
4984 if ($save_vmstate && check_running($vmid)) {
4985 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
4986 }
4987
4988 &$snapshot_copy_config($conf, $snap);
4989
4990 $snap->{snapstate} = "prepare";
4991 $snap->{snaptime} = time();
4992 $snap->{description} = $comment if $comment;
4993
4994 # always overwrite machine if we save vmstate. This makes sure we
4995 # can restore it later using correct machine type
4996 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
4997
4998 update_config_nolock($vmid, $conf, 1);
4999 };
5000
5001 lock_config($vmid, $updatefn);
5002
5003 return $snap;
5004 };
5005
5006 my $snapshot_commit = sub {
5007 my ($vmid, $snapname) = @_;
5008
5009 my $updatefn = sub {
5010
5011 my $conf = load_config($vmid);
5012
5013 die "missing snapshot lock\n"
5014 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5015
5016 my $has_machine_config = defined($conf->{machine});
5017
5018 my $snap = $conf->{snapshots}->{$snapname};
5019
5020 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5021
5022 die "wrong snapshot state\n"
5023 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5024
5025 delete $snap->{snapstate};
5026 delete $conf->{lock};
5027
5028 my $newconf = &$snapshot_apply_config($conf, $snap);
5029
5030 delete $newconf->{machine} if !$has_machine_config;
5031
5032 $newconf->{parent} = $snapname;
5033
5034 update_config_nolock($vmid, $newconf, 1);
5035 };
5036
5037 lock_config($vmid, $updatefn);
5038 };
5039
5040 sub snapshot_rollback {
5041 my ($vmid, $snapname) = @_;
5042
5043 my $snap;
5044
5045 my $prepare = 1;
5046
5047 my $storecfg = PVE::Storage::config();
5048
5049 my $updatefn = sub {
5050
5051 my $conf = load_config($vmid);
5052
5053 die "you can't rollback if vm is a template\n" if is_template($conf);
5054
5055 $snap = $conf->{snapshots}->{$snapname};
5056
5057 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5058
5059 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5060 if $snap->{snapstate};
5061
5062 if ($prepare) {
5063 check_lock($conf);
5064 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5065 }
5066
5067 die "unable to rollback vm $vmid: vm is running\n"
5068 if check_running($vmid);
5069
5070 if ($prepare) {
5071 $conf->{lock} = 'rollback';
5072 } else {
5073 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5074 delete $conf->{lock};
5075 }
5076
5077 my $forcemachine;
5078
5079 if (!$prepare) {
5080 my $has_machine_config = defined($conf->{machine});
5081
5082 # copy snapshot config to current config
5083 $conf = &$snapshot_apply_config($conf, $snap);
5084 $conf->{parent} = $snapname;
5085
5086 # Note: old code did not store 'machine', so we try to be smart
5087 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5088 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5089 # we remove the 'machine' configuration if not explicitly specified
5090 # in the original config.
5091 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5092 }
5093
5094 update_config_nolock($vmid, $conf, 1);
5095
5096 if (!$prepare && $snap->{vmstate}) {
5097 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5098 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5099 }
5100 };
5101
5102 lock_config($vmid, $updatefn);
5103
5104 foreach_drive($snap, sub {
5105 my ($ds, $drive) = @_;
5106
5107 return if drive_is_cdrom($drive);
5108
5109 my $volid = $drive->{file};
5110 my $device = "drive-$ds";
5111
5112 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5113 });
5114
5115 $prepare = 0;
5116 lock_config($vmid, $updatefn);
5117 }
5118
5119 my $savevm_wait = sub {
5120 my ($vmid) = @_;
5121
5122 for(;;) {
5123 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5124 if (!$stat->{status}) {
5125 die "savevm not active\n";
5126 } elsif ($stat->{status} eq 'active') {
5127 sleep(1);
5128 next;
5129 } elsif ($stat->{status} eq 'completed') {
5130 last;
5131 } else {
5132 die "query-savevm returned status '$stat->{status}'\n";
5133 }
5134 }
5135 };
5136
5137 sub snapshot_create {
5138 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5139
5140 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5141
5142 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5143
5144 my $config = load_config($vmid);
5145
5146 my $running = check_running($vmid);
5147
5148 my $freezefs = $running && $config->{agent};
5149 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5150
5151 my $drivehash = {};
5152
5153 if ($freezefs) {
5154 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5155 warn "guest-fsfreeze-freeze problems - $@" if $@;
5156 }
5157
5158 eval {
5159 # create internal snapshots of all drives
5160
5161 my $storecfg = PVE::Storage::config();
5162
5163 if ($running) {
5164 if ($snap->{vmstate}) {
5165 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5166 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5167 &$savevm_wait($vmid);
5168 } else {
5169 vm_mon_cmd($vmid, "savevm-start");
5170 }
5171 };
5172
5173 foreach_drive($snap, sub {
5174 my ($ds, $drive) = @_;
5175
5176 return if drive_is_cdrom($drive);
5177
5178 my $volid = $drive->{file};
5179 my $device = "drive-$ds";
5180
5181 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5182 $drivehash->{$ds} = 1;
5183 });
5184 };
5185 my $err = $@;
5186
5187 if ($running) {
5188 eval { vm_mon_cmd($vmid, "savevm-end") };
5189 warn $@ if $@;
5190
5191 if ($freezefs) {
5192 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5193 warn "guest-fsfreeze-thaw problems - $@" if $@;
5194 }
5195
5196 # savevm-end is async, we need to wait
5197 for (;;) {
5198 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5199 if (!$stat->{bytes}) {
5200 last;
5201 } else {
5202 print "savevm not yet finished\n";
5203 sleep(1);
5204 next;
5205 }
5206 }
5207 }
5208
5209 if ($err) {
5210 warn "snapshot create failed: starting cleanup\n";
5211 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5212 warn $@ if $@;
5213 die $err;
5214 }
5215
5216 &$snapshot_commit($vmid, $snapname);
5217 }
5218
5219 # Note: $drivehash is only set when called from snapshot_create.
5220 sub snapshot_delete {
5221 my ($vmid, $snapname, $force, $drivehash) = @_;
5222
5223 my $prepare = 1;
5224
5225 my $snap;
5226 my $unused = [];
5227
5228 my $unlink_parent = sub {
5229 my ($confref, $new_parent) = @_;
5230
5231 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5232 if ($new_parent) {
5233 $confref->{parent} = $new_parent;
5234 } else {
5235 delete $confref->{parent};
5236 }
5237 }
5238 };
5239
5240 my $updatefn = sub {
5241 my ($remove_drive) = @_;
5242
5243 my $conf = load_config($vmid);
5244
5245 if (!$drivehash) {
5246 check_lock($conf);
5247 die "you can't delete a snapshot if vm is a template\n"
5248 if is_template($conf);
5249 }
5250
5251 $snap = $conf->{snapshots}->{$snapname};
5252
5253 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5254
5255 # remove parent refs
5256 if (!$prepare) {
5257 &$unlink_parent($conf, $snap->{parent});
5258 foreach my $sn (keys %{$conf->{snapshots}}) {
5259 next if $sn eq $snapname;
5260 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5261 }
5262 }
5263
5264 if ($remove_drive) {
5265 if ($remove_drive eq 'vmstate') {
5266 delete $snap->{$remove_drive};
5267 } else {
5268 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5269 my $volid = $drive->{file};
5270 delete $snap->{$remove_drive};
5271 add_unused_volume($conf, $volid);
5272 }
5273 }
5274
5275 if ($prepare) {
5276 $snap->{snapstate} = 'delete';
5277 } else {
5278 delete $conf->{snapshots}->{$snapname};
5279 delete $conf->{lock} if $drivehash;
5280 foreach my $volid (@$unused) {
5281 add_unused_volume($conf, $volid);
5282 }
5283 }
5284
5285 update_config_nolock($vmid, $conf, 1);
5286 };
5287
5288 lock_config($vmid, $updatefn);
5289
5290 # now remove vmstate file
5291
5292 my $storecfg = PVE::Storage::config();
5293
5294 if ($snap->{vmstate}) {
5295 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5296 if (my $err = $@) {
5297 die $err if !$force;
5298 warn $err;
5299 }
5300 # save changes (remove vmstate from snapshot)
5301 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5302 };
5303
5304 # now remove all internal snapshots
5305 foreach_drive($snap, sub {
5306 my ($ds, $drive) = @_;
5307
5308 return if drive_is_cdrom($drive);
5309
5310 my $volid = $drive->{file};
5311 my $device = "drive-$ds";
5312
5313 if (!$drivehash || $drivehash->{$ds}) {
5314 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5315 if (my $err = $@) {
5316 die $err if !$force;
5317 warn $err;
5318 }
5319 }
5320
5321 # save changes (remove drive fron snapshot)
5322 lock_config($vmid, $updatefn, $ds) if !$force;
5323 push @$unused, $volid;
5324 });
5325
5326 # now cleanup config
5327 $prepare = 0;
5328 lock_config($vmid, $updatefn);
5329 }
5330
5331 sub has_feature {
5332 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5333
5334 my $err;
5335 foreach_drive($conf, sub {
5336 my ($ds, $drive) = @_;
5337
5338 return if drive_is_cdrom($drive);
5339 my $volid = $drive->{file};
5340 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5341 });
5342
5343 return $err ? 0 : 1;
5344 }
5345
5346 sub template_create {
5347 my ($vmid, $conf, $disk) = @_;
5348
5349 my $storecfg = PVE::Storage::config();
5350
5351 foreach_drive($conf, sub {
5352 my ($ds, $drive) = @_;
5353
5354 return if drive_is_cdrom($drive);
5355 return if $disk && $ds ne $disk;
5356
5357 my $volid = $drive->{file};
5358 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5359
5360 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5361 $drive->{file} = $voliddst;
5362 $conf->{$ds} = print_drive($vmid, $drive);
5363 update_config_nolock($vmid, $conf, 1);
5364 });
5365 }
5366
5367 sub is_template {
5368 my ($conf) = @_;
5369
5370 return 1 if defined $conf->{template} && $conf->{template} == 1;
5371 }
5372
5373 sub qemu_img_convert {
5374 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5375
5376 my $storecfg = PVE::Storage::config();
5377 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5378 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5379
5380 if ($src_storeid && $dst_storeid) {
5381 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5382 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5383
5384 my $src_format = qemu_img_format($src_scfg, $src_volname);
5385 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5386
5387 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5388 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5389
5390 my $cmd = [];
5391 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5392 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5393 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5394
5395 my $parser = sub {
5396 my $line = shift;
5397 if($line =~ m/\((\S+)\/100\%\)/){
5398 my $percent = $1;
5399 my $transferred = int($size * $percent / 100);
5400 my $remaining = $size - $transferred;
5401
5402 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5403 }
5404
5405 };
5406
5407 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5408 my $err = $@;
5409 die "copy failed: $err" if $err;
5410 }
5411 }
5412
5413 sub qemu_img_format {
5414 my ($scfg, $volname) = @_;
5415
5416 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5417 return $1;
5418 } elsif ($scfg->{type} eq 'iscsi') {
5419 return "host_device";
5420 } else {
5421 return "raw";
5422 }
5423 }
5424
5425 sub qemu_drive_mirror {
5426 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5427
5428 my $count = 0;
5429 my $old_len = 0;
5430 my $frozen = undef;
5431 my $maxwait = 120;
5432
5433 my $storecfg = PVE::Storage::config();
5434 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5435
5436 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5437
5438 my $format;
5439 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5440 $format = $1;
5441 }
5442
5443 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5444
5445 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5446 $opts->{format} = $format if $format;
5447
5448 #fixme : sometime drive-mirror timeout, but works fine after.
5449 # (I have see the problem with big volume > 200GB), so we need to eval
5450 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5451 # ignore errors here
5452
5453 eval {
5454 while (1) {
5455 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5456 my $stat = @$stats[0];
5457 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5458 die "error job is not mirroring" if $stat->{type} ne "mirror";
5459
5460 my $busy = $stat->{busy};
5461
5462 if (my $total = $stat->{len}) {
5463 my $transferred = $stat->{offset} || 0;
5464 my $remaining = $total - $transferred;
5465 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5466
5467 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
5468 }
5469
5470 if ($stat->{len} == $stat->{offset}) {
5471 if ($busy eq 'false') {
5472
5473 last if $vmiddst != $vmid;
5474
5475 # try to switch the disk if source and destination are on the same guest
5476 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
5477 last if !$@;
5478 die $@ if $@ !~ m/cannot be completed/;
5479 }
5480
5481 if ($count > $maxwait) {
5482 # if too much writes to disk occurs at the end of migration
5483 #the disk needs to be freezed to be able to complete the migration
5484 vm_suspend($vmid,1);
5485 $frozen = 1;
5486 }
5487 $count ++
5488 }
5489 $old_len = $stat->{offset};
5490 sleep 1;
5491 }
5492
5493 vm_resume($vmid, 1) if $frozen;
5494
5495 };
5496 my $err = $@;
5497
5498 my $cancel_job = sub {
5499 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
5500 while (1) {
5501 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5502 my $stat = @$stats[0];
5503 last if !$stat;
5504 sleep 1;
5505 }
5506 };
5507
5508 if ($err) {
5509 eval { &$cancel_job(); };
5510 die "mirroring error: $err";
5511 }
5512
5513 if ($vmiddst != $vmid) {
5514 # if we clone a disk for a new target vm, we don't switch the disk
5515 &$cancel_job(); # so we call block-job-cancel
5516 }
5517 }
5518
5519 sub clone_disk {
5520 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
5521 $newvmid, $storage, $format, $full, $newvollist) = @_;
5522
5523 my $newvolid;
5524
5525 if (!$full) {
5526 print "create linked clone of drive $drivename ($drive->{file})\n";
5527 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
5528 push @$newvollist, $newvolid;
5529 } else {
5530 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
5531 $storeid = $storage if $storage;
5532
5533 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
5534 if (!$format) {
5535 $format = $drive->{format} || $defFormat;
5536 }
5537
5538 # test if requested format is supported - else use default
5539 my $supported = grep { $_ eq $format } @$validFormats;
5540 $format = $defFormat if !$supported;
5541
5542 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
5543
5544 print "create full clone of drive $drivename ($drive->{file})\n";
5545 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
5546 push @$newvollist, $newvolid;
5547
5548 if (!$running || $snapname) {
5549 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
5550 } else {
5551 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
5552 }
5553 }
5554
5555 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
5556
5557 my $disk = $drive;
5558 $disk->{format} = undef;
5559 $disk->{file} = $newvolid;
5560 $disk->{size} = $size;
5561
5562 return $disk;
5563 }
5564
5565 # this only works if VM is running
5566 sub get_current_qemu_machine {
5567 my ($vmid) = @_;
5568
5569 my $cmd = { execute => 'query-machines', arguments => {} };
5570 my $res = PVE::QemuServer::vm_qmp_command($vmid, $cmd);
5571
5572 my ($current, $default);
5573 foreach my $e (@$res) {
5574 $default = $e->{name} if $e->{'is-default'};
5575 $current = $e->{name} if $e->{'is-current'};
5576 }
5577
5578 # fallback to the default machine if current is not supported by qemu
5579 return $current || $default || 'pc';
5580 }
5581
5582 sub lspci {
5583
5584 my $devices = {};
5585
5586 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
5587 my (undef, $id, $function) = @_;
5588 my $res = { id => $id, function => $function};
5589 push @{$devices->{$id}}, $res;
5590 });
5591
5592 return $devices;
5593 }
5594
5595 1;