]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
new scsihw_infos sub : code factorization
[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 onboot => {
156 optional => 1,
157 type => 'boolean',
158 description => "Specifies whether a VM will be started during system bootup.",
159 default => 0,
160 },
161 autostart => {
162 optional => 1,
163 type => 'boolean',
164 description => "Automatic restart after crash (currently ignored).",
165 default => 0,
166 },
167 hotplug => {
168 optional => 1,
169 type => 'string', format => 'pve-hotplug-features',
170 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
171 default => 'network,disk,usb',
172 },
173 reboot => {
174 optional => 1,
175 type => 'boolean',
176 description => "Allow reboot. If set to '0' the VM exit on reboot.",
177 default => 1,
178 },
179 lock => {
180 optional => 1,
181 type => 'string',
182 description => "Lock/unlock the VM.",
183 enum => [qw(migrate backup snapshot rollback)],
184 },
185 cpulimit => {
186 optional => 1,
187 type => 'integer',
188 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.",
189 minimum => 0,
190 default => 0,
191 },
192 cpuunits => {
193 optional => 1,
194 type => 'integer',
195 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.",
196 minimum => 0,
197 maximum => 500000,
198 default => 1000,
199 },
200 memory => {
201 optional => 1,
202 type => 'integer',
203 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
204 minimum => 16,
205 default => 512,
206 },
207 balloon => {
208 optional => 1,
209 type => 'integer',
210 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
211 minimum => 0,
212 },
213 shares => {
214 optional => 1,
215 type => 'integer',
216 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",
217 minimum => 0,
218 maximum => 50000,
219 default => 1000,
220 },
221 keyboard => {
222 optional => 1,
223 type => 'string',
224 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
225 enum => PVE::Tools::kvmkeymaplist(),
226 default => 'en-us',
227 },
228 name => {
229 optional => 1,
230 type => 'string', format => 'dns-name',
231 description => "Set a name for the VM. Only used on the configuration web interface.",
232 },
233 scsihw => {
234 optional => 1,
235 type => 'string',
236 description => "scsi controller model",
237 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
238 default => 'lsi',
239 },
240 description => {
241 optional => 1,
242 type => 'string',
243 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
244 },
245 ostype => {
246 optional => 1,
247 type => 'string',
248 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
249 description => <<EODESC,
250 Used to enable special optimization/features for specific
251 operating systems:
252
253 other => unspecified OS
254 wxp => Microsoft Windows XP
255 w2k => Microsoft Windows 2000
256 w2k3 => Microsoft Windows 2003
257 w2k8 => Microsoft Windows 2008
258 wvista => Microsoft Windows Vista
259 win7 => Microsoft Windows 7
260 win8 => Microsoft Windows 8/2012
261 l24 => Linux 2.4 Kernel
262 l26 => Linux 2.6/3.X Kernel
263 solaris => solaris/opensolaris/openindiania kernel
264
265 other|l24|l26|solaris ... no special behaviour
266 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
267 EODESC
268 },
269 boot => {
270 optional => 1,
271 type => 'string',
272 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
273 pattern => '[acdn]{1,4}',
274 default => 'cdn',
275 },
276 bootdisk => {
277 optional => 1,
278 type => 'string', format => 'pve-qm-bootdisk',
279 description => "Enable booting from specified disk.",
280 pattern => '(ide|sata|scsi|virtio)\d+',
281 },
282 smp => {
283 optional => 1,
284 type => 'integer',
285 description => "The number of CPUs. Please use option -sockets instead.",
286 minimum => 1,
287 default => 1,
288 },
289 sockets => {
290 optional => 1,
291 type => 'integer',
292 description => "The number of CPU sockets.",
293 minimum => 1,
294 default => 1,
295 },
296 cores => {
297 optional => 1,
298 type => 'integer',
299 description => "The number of cores per socket.",
300 minimum => 1,
301 default => 1,
302 },
303 numa => {
304 optional => 1,
305 type => 'boolean',
306 description => "Enable/disable Numa.",
307 default => 0,
308 },
309 vcpus => {
310 optional => 1,
311 type => 'integer',
312 description => "Number of hotplugged vcpus.",
313 minimum => 1,
314 default => 0,
315 },
316 acpi => {
317 optional => 1,
318 type => 'boolean',
319 description => "Enable/disable ACPI.",
320 default => 1,
321 },
322 agent => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable Qemu GuestAgent.",
326 default => 0,
327 },
328 kvm => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable KVM hardware virtualization.",
332 default => 1,
333 },
334 tdf => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable time drift fix.",
338 default => 0,
339 },
340 localtime => {
341 optional => 1,
342 type => 'boolean',
343 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
344 },
345 freeze => {
346 optional => 1,
347 type => 'boolean',
348 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
349 },
350 vga => {
351 optional => 1,
352 type => 'string',
353 description => "Select VGA type. If you want to use high resolution modes (>= 1280x1024x16) then you should use option 'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and 'cirrur' for other OS types. Option 'qxl' enables the SPICE display sever. You can also run without any graphic card using a serial devive as terminal.",
354 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
355 },
356 watchdog => {
357 optional => 1,
358 type => 'string', format => 'pve-qm-watchdog',
359 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
360 description => "Create a virtual hardware watchdog device. Once enabled (by a guest action), the watchdog must be periodically polled by an agent inside the guest or else the guest will be restarted (or execute the action specified)",
361 },
362 startdate => {
363 optional => 1,
364 type => 'string',
365 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
366 description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
367 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
368 default => 'now',
369 },
370 startup => {
371 optional => 1,
372 type => 'string', format => 'pve-qm-startup',
373 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
374 description => "Startup and shutdown behavior. Order is a non-negative number defining the general startup order. Shutdown in done with reverse ordering. Additionally you can set the 'up' or 'down' delay in seconds, which specifies a delay to wait before the next VM is started or stopped.",
375 },
376 template => {
377 optional => 1,
378 type => 'boolean',
379 description => "Enable/disable Template.",
380 default => 0,
381 },
382 args => {
383 optional => 1,
384 type => 'string',
385 description => <<EODESCR,
386 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
387
388 args: -no-reboot -no-hpet
389 EODESCR
390 },
391 tablet => {
392 optional => 1,
393 type => 'boolean',
394 default => 1,
395 description => "Enable/disable the usb tablet device. This device is usually needed to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with normal VNC clients. If you're running lots of console-only guests on one host, you may consider disabling this to save some context switches. This is turned of by default if you use spice (vga=qxl).",
396 },
397 migrate_speed => {
398 optional => 1,
399 type => 'integer',
400 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
401 minimum => 0,
402 default => 0,
403 },
404 migrate_downtime => {
405 optional => 1,
406 type => 'number',
407 description => "Set maximum tolerated downtime (in seconds) for migrations.",
408 minimum => 0,
409 default => 0.1,
410 },
411 cdrom => {
412 optional => 1,
413 type => 'string', format => 'pve-qm-drive',
414 typetext => 'volume',
415 description => "This is an alias for option -ide2",
416 },
417 cpu => {
418 optional => 1,
419 description => "Emulated CPU type.",
420 type => 'string',
421 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge IvyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
422 default => 'kvm64',
423 },
424 parent => get_standard_option('pve-snapshot-name', {
425 optional => 1,
426 description => "Parent snapshot name. This is used internally, and should not be modified.",
427 }),
428 snaptime => {
429 optional => 1,
430 description => "Timestamp for snapshots.",
431 type => 'integer',
432 minimum => 0,
433 },
434 vmstate => {
435 optional => 1,
436 type => 'string', format => 'pve-volume-id',
437 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
438 },
439 machine => {
440 description => "Specific the Qemu machine type.",
441 type => 'string',
442 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
443 maxLength => 40,
444 optional => 1,
445 },
446 smbios1 => {
447 description => "Specify SMBIOS type 1 fields.",
448 type => 'string', format => 'pve-qm-smbios1',
449 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
450 maxLength => 256,
451 optional => 1,
452 },
453 };
454
455 # what about other qemu settings ?
456 #cpu => 'string',
457 #machine => 'string',
458 #fda => 'file',
459 #fdb => 'file',
460 #mtdblock => 'file',
461 #sd => 'file',
462 #pflash => 'file',
463 #snapshot => 'bool',
464 #bootp => 'file',
465 ##tftp => 'dir',
466 ##smb => 'dir',
467 #kernel => 'file',
468 #append => 'string',
469 #initrd => 'file',
470 ##soundhw => 'string',
471
472 while (my ($k, $v) = each %$confdesc) {
473 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
474 }
475
476 my $MAX_IDE_DISKS = 4;
477 my $MAX_SCSI_DISKS = 14;
478 my $MAX_VIRTIO_DISKS = 16;
479 my $MAX_SATA_DISKS = 6;
480 my $MAX_USB_DEVICES = 5;
481 my $MAX_NETS = 32;
482 my $MAX_UNUSED_DISKS = 8;
483 my $MAX_HOSTPCI_DEVICES = 4;
484 my $MAX_SERIAL_PORTS = 4;
485 my $MAX_PARALLEL_PORTS = 3;
486 my $MAX_NUMA = 8;
487 my $MAX_MEM = 4194304;
488 my $STATICMEM = 1024;
489
490 my $numadesc = {
491 optional => 1,
492 type => 'string', format => 'pve-qm-numanode',
493 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
494 description => "numa topology",
495 };
496 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
497
498 for (my $i = 0; $i < $MAX_NUMA; $i++) {
499 $confdesc->{"numa$i"} = $numadesc;
500 }
501
502 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
503 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
504 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
505 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
506
507 my $netdesc = {
508 optional => 1,
509 type => 'string', format => 'pve-qm-net',
510 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
511 description => <<EODESCR,
512 Specify network devices.
513
514 MODEL is one of: $nic_model_list_txt
515
516 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
517 automatically generated if not specified.
518
519 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
520
521 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'.
522
523 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
524
525 10.0.2.2 Gateway
526 10.0.2.3 DNS Server
527 10.0.2.4 SMB Server
528
529 The DHCP server assign addresses to the guest starting from 10.0.2.15.
530
531 EODESCR
532 };
533 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
534
535 for (my $i = 0; $i < $MAX_NETS; $i++) {
536 $confdesc->{"net$i"} = $netdesc;
537 }
538
539 my $drivename_hash;
540
541 my $idedesc = {
542 optional => 1,
543 type => 'string', format => 'pve-qm-drive',
544 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
545 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
546 };
547 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
548
549 my $scsidesc = {
550 optional => 1,
551 type => 'string', format => 'pve-qm-drive',
552 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
553 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
554 };
555 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
556
557 my $satadesc = {
558 optional => 1,
559 type => 'string', format => 'pve-qm-drive',
560 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
561 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
562 };
563 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
564
565 my $virtiodesc = {
566 optional => 1,
567 type => 'string', format => 'pve-qm-drive',
568 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on] [,iothread=on]',
569 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
570 };
571 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
572
573 my $usbdesc = {
574 optional => 1,
575 type => 'string', format => 'pve-qm-usb-device',
576 typetext => 'host=HOSTUSBDEVICE|spice',
577 description => <<EODESCR,
578 Configure an USB device (n is 0 to 4). This can be used to
579 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
580
581 'bus-port(.port)*' (decimal numbers) or
582 'vendor_id:product_id' (hexadeciaml numbers)
583
584 You can use the 'lsusb -t' command to list existing usb devices.
585
586 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
587
588 The value 'spice' can be used to add a usb redirection devices for spice.
589
590 EODESCR
591 };
592 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
593
594 my $hostpcidesc = {
595 optional => 1,
596 type => 'string', format => 'pve-qm-hostpci',
597 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
598 description => <<EODESCR,
599 Map host pci devices. HOSTPCIDEVICE syntax is:
600
601 'bus:dev.func' (hexadecimal numbers)
602
603 You can us the 'lspci' command to list existing pci devices.
604
605 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
606
607 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
608
609 Experimental: user reported problems with this option.
610 EODESCR
611 };
612 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
613
614 my $serialdesc = {
615 optional => 1,
616 type => 'string',
617 pattern => '(/dev/.+|socket)',
618 description => <<EODESCR,
619 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).
620
621 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
622
623 Experimental: user reported problems with this option.
624 EODESCR
625 };
626
627 my $paralleldesc= {
628 optional => 1,
629 type => 'string',
630 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
631 description => <<EODESCR,
632 Map host parallel devices (n is 0 to 2).
633
634 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
635
636 Experimental: user reported problems with this option.
637 EODESCR
638 };
639
640 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
641 $confdesc->{"parallel$i"} = $paralleldesc;
642 }
643
644 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
645 $confdesc->{"serial$i"} = $serialdesc;
646 }
647
648 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
649 $confdesc->{"hostpci$i"} = $hostpcidesc;
650 }
651
652 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
653 $drivename_hash->{"ide$i"} = 1;
654 $confdesc->{"ide$i"} = $idedesc;
655 }
656
657 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
658 $drivename_hash->{"sata$i"} = 1;
659 $confdesc->{"sata$i"} = $satadesc;
660 }
661
662 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
663 $drivename_hash->{"scsi$i"} = 1;
664 $confdesc->{"scsi$i"} = $scsidesc ;
665 }
666
667 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
668 $drivename_hash->{"virtio$i"} = 1;
669 $confdesc->{"virtio$i"} = $virtiodesc;
670 }
671
672 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
673 $confdesc->{"usb$i"} = $usbdesc;
674 }
675
676 my $unuseddesc = {
677 optional => 1,
678 type => 'string', format => 'pve-volume-id',
679 description => "Reference to unused volumes.",
680 };
681
682 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
683 $confdesc->{"unused$i"} = $unuseddesc;
684 }
685
686 my $kvm_api_version = 0;
687
688 sub kvm_version {
689
690 return $kvm_api_version if $kvm_api_version;
691
692 my $fh = IO::File->new("</dev/kvm") ||
693 return 0;
694
695 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
696 $kvm_api_version = $v;
697 }
698
699 $fh->close();
700
701 return $kvm_api_version;
702 }
703
704 my $kvm_user_version;
705
706 sub kvm_user_version {
707
708 return $kvm_user_version if $kvm_user_version;
709
710 $kvm_user_version = 'unknown';
711
712 my $tmp = `kvm -help 2>/dev/null`;
713
714 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
715 $kvm_user_version = $2;
716 }
717
718 return $kvm_user_version;
719
720 }
721
722 my $kernel_has_vhost_net = -c '/dev/vhost-net';
723
724 sub disknames {
725 # order is important - used to autoselect boot disk
726 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
727 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
728 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
729 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
730 }
731
732 sub valid_drivename {
733 my $dev = shift;
734
735 return defined($drivename_hash->{$dev});
736 }
737
738 sub option_exists {
739 my $key = shift;
740 return defined($confdesc->{$key});
741 }
742
743 sub nic_models {
744 return $nic_model_list;
745 }
746
747 sub os_list_description {
748
749 return {
750 other => 'Other',
751 wxp => 'Windows XP',
752 w2k => 'Windows 2000',
753 w2k3 =>, 'Windows 2003',
754 w2k8 => 'Windows 2008',
755 wvista => 'Windows Vista',
756 win7 => 'Windows 7',
757 win8 => 'Windows 8/2012',
758 l24 => 'Linux 2.4',
759 l26 => 'Linux 2.6',
760 };
761 }
762
763 my $cdrom_path;
764
765 sub get_cdrom_path {
766
767 return $cdrom_path if $cdrom_path;
768
769 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
770 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
771 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
772 }
773
774 sub get_iso_path {
775 my ($storecfg, $vmid, $cdrom) = @_;
776
777 if ($cdrom eq 'cdrom') {
778 return get_cdrom_path();
779 } elsif ($cdrom eq 'none') {
780 return '';
781 } elsif ($cdrom =~ m|^/|) {
782 return $cdrom;
783 } else {
784 return PVE::Storage::path($storecfg, $cdrom);
785 }
786 }
787
788 # try to convert old style file names to volume IDs
789 sub filename_to_volume_id {
790 my ($vmid, $file, $media) = @_;
791
792 if (!($file eq 'none' || $file eq 'cdrom' ||
793 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
794
795 return undef if $file =~ m|/|;
796
797 if ($media && $media eq 'cdrom') {
798 $file = "local:iso/$file";
799 } else {
800 $file = "local:$vmid/$file";
801 }
802 }
803
804 return $file;
805 }
806
807 sub verify_media_type {
808 my ($opt, $vtype, $media) = @_;
809
810 return if !$media;
811
812 my $etype;
813 if ($media eq 'disk') {
814 $etype = 'images';
815 } elsif ($media eq 'cdrom') {
816 $etype = 'iso';
817 } else {
818 die "internal error";
819 }
820
821 return if ($vtype eq $etype);
822
823 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
824 }
825
826 sub cleanup_drive_path {
827 my ($opt, $storecfg, $drive) = @_;
828
829 # try to convert filesystem paths to volume IDs
830
831 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
832 ($drive->{file} !~ m|^/dev/.+|) &&
833 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
834 ($drive->{file} !~ m/^\d+$/)) {
835 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
836 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
837 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
838 verify_media_type($opt, $vtype, $drive->{media});
839 $drive->{file} = $volid;
840 }
841
842 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
843 }
844
845 sub create_conf_nolock {
846 my ($vmid, $settings) = @_;
847
848 my $filename = config_file($vmid);
849
850 die "configuration file '$filename' already exists\n" if -f $filename;
851
852 my $defaults = load_defaults();
853
854 $settings->{name} = "vm$vmid" if !$settings->{name};
855 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
856
857 my $data = '';
858 foreach my $opt (keys %$settings) {
859 next if !$confdesc->{$opt};
860
861 my $value = $settings->{$opt};
862 next if !$value;
863
864 $data .= "$opt: $value\n";
865 }
866
867 PVE::Tools::file_set_contents($filename, $data);
868 }
869
870 sub parse_hotplug_features {
871 my ($data) = @_;
872
873 my $res = {};
874
875 return $res if $data eq '0';
876
877 $data = $confdesc->{hotplug}->{default} if $data eq '1';
878
879 foreach my $feature (PVE::Tools::split_list($data)) {
880 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
881 $res->{$1} = 1;
882 } else {
883 warn "ignoring unknown hotplug feature '$feature'\n";
884 }
885 }
886 return $res;
887 }
888
889 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
890 sub pve_verify_hotplug_features {
891 my ($value, $noerr) = @_;
892
893 return $value if parse_hotplug_features($value);
894
895 return undef if $noerr;
896
897 die "unable to parse hotplug option\n";
898 }
899
900 my $parse_size = sub {
901 my ($value) = @_;
902
903 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
904 my ($size, $unit) = ($1, $3);
905 if ($unit) {
906 if ($unit eq 'K') {
907 $size = $size * 1024;
908 } elsif ($unit eq 'M') {
909 $size = $size * 1024 * 1024;
910 } elsif ($unit eq 'G') {
911 $size = $size * 1024 * 1024 * 1024;
912 }
913 }
914 return int($size);
915 };
916
917 my $format_size = sub {
918 my ($size) = @_;
919
920 $size = int($size);
921
922 my $kb = int($size/1024);
923 return $size if $kb*1024 != $size;
924
925 my $mb = int($kb/1024);
926 return "${kb}K" if $mb*1024 != $kb;
927
928 my $gb = int($mb/1024);
929 return "${mb}M" if $gb*1024 != $mb;
930
931 return "${gb}G";
932 };
933
934 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
935 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
936 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
937 # [,aio=native|threads][,discard=ignore|on][,iothread=on]
938
939 sub parse_drive {
940 my ($key, $data) = @_;
941
942 my $res = {};
943
944 # $key may be undefined - used to verify JSON parameters
945 if (!defined($key)) {
946 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
947 $res->{index} = 0;
948 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
949 $res->{interface} = $1;
950 $res->{index} = $2;
951 } else {
952 return undef;
953 }
954
955 foreach my $p (split (/,/, $data)) {
956 next if $p =~ m/^\s*$/;
957
958 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard|iothread)=(.+)$/) {
959 my ($k, $v) = ($1, $2);
960
961 $k = 'file' if $k eq 'volume';
962
963 return undef if defined $res->{$k};
964
965 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
966 return undef if !$v || $v !~ m/^\d+/;
967 $k = "m$k";
968 $v = sprintf("%.3f", $v / (1024*1024));
969 }
970 $res->{$k} = $v;
971 } else {
972 if (!$res->{file} && $p !~ m/=/) {
973 $res->{file} = $p;
974 } else {
975 return undef;
976 }
977 }
978 }
979
980 return undef if !$res->{file};
981
982 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
983 $res->{format} = $1;
984 }
985
986 return undef if $res->{cache} &&
987 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
988 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
989 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
990 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
991 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
992 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
993 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
994 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
995 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
996 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
997 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
998 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
999 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
1000 return undef if $res->{iothread} && $res->{iothread} !~ m/^(on)$/;
1001
1002 return undef if $res->{mbps_rd} && $res->{mbps};
1003 return undef if $res->{mbps_wr} && $res->{mbps};
1004
1005 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
1006 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
1007 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
1008 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
1009 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
1010 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
1011
1012 return undef if $res->{iops_rd} && $res->{iops};
1013 return undef if $res->{iops_wr} && $res->{iops};
1014
1015
1016 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
1017 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
1018 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
1019 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
1020 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
1021 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
1022
1023
1024 if ($res->{size}) {
1025 return undef if !defined($res->{size} = &$parse_size($res->{size}));
1026 }
1027
1028 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1029 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1030 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1031 return undef if $res->{interface} eq 'virtio';
1032 }
1033
1034 # rerror does not work with scsi drives
1035 if ($res->{rerror}) {
1036 return undef if $res->{interface} eq 'scsi';
1037 }
1038
1039 return $res;
1040 }
1041
1042 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1043
1044 sub print_drive {
1045 my ($vmid, $drive) = @_;
1046
1047 my $opts = '';
1048 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup', 'iothread') {
1049 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1050 }
1051
1052 if ($drive->{size}) {
1053 $opts .= ",size=" . &$format_size($drive->{size});
1054 }
1055
1056 return "$drive->{file}$opts";
1057 }
1058
1059 sub scsi_inquiry {
1060 my($fh, $noerr) = @_;
1061
1062 my $SG_IO = 0x2285;
1063 my $SG_GET_VERSION_NUM = 0x2282;
1064
1065 my $versionbuf = "\x00" x 8;
1066 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1067 if (!$ret) {
1068 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1069 return undef;
1070 }
1071 my $version = unpack("I", $versionbuf);
1072 if ($version < 30000) {
1073 die "scsi generic interface too old\n" if !$noerr;
1074 return undef;
1075 }
1076
1077 my $buf = "\x00" x 36;
1078 my $sensebuf = "\x00" x 8;
1079 my $cmd = pack("C x3 C x1", 0x12, 36);
1080
1081 # see /usr/include/scsi/sg.h
1082 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";
1083
1084 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1085 length($sensebuf), 0, length($buf), $buf,
1086 $cmd, $sensebuf, 6000);
1087
1088 $ret = ioctl($fh, $SG_IO, $packet);
1089 if (!$ret) {
1090 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1091 return undef;
1092 }
1093
1094 my @res = unpack($sg_io_hdr_t, $packet);
1095 if ($res[17] || $res[18]) {
1096 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1097 return undef;
1098 }
1099
1100 my $res = {};
1101 (my $byte0, my $byte1, $res->{vendor},
1102 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1103
1104 $res->{removable} = $byte1 & 128 ? 1 : 0;
1105 $res->{type} = $byte0 & 31;
1106
1107 return $res;
1108 }
1109
1110 sub path_is_scsi {
1111 my ($path) = @_;
1112
1113 my $fh = IO::File->new("+<$path") || return undef;
1114 my $res = scsi_inquiry($fh, 1);
1115 close($fh);
1116
1117 return $res;
1118 }
1119
1120 sub machine_type_is_q35 {
1121 my ($conf) = @_;
1122
1123 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1124 }
1125
1126 sub print_tabletdevice_full {
1127 my ($conf) = @_;
1128
1129 my $q35 = machine_type_is_q35($conf);
1130
1131 # we use uhci for old VMs because tablet driver was buggy in older qemu
1132 my $usbbus = $q35 ? "ehci" : "uhci";
1133
1134 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1135 }
1136
1137 sub print_drivedevice_full {
1138 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1139
1140 my $device = '';
1141 my $maxdev = 0;
1142
1143 if ($drive->{interface} eq 'virtio') {
1144 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1145 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1146 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1147 } elsif ($drive->{interface} eq 'scsi') {
1148
1149 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1150 my $unit = $drive->{index} % $maxdev;
1151 my $devicetype = 'hd';
1152 my $path = '';
1153 if (drive_is_cdrom($drive)) {
1154 $devicetype = 'cd';
1155 } else {
1156 if ($drive->{file} =~ m|^/|) {
1157 $path = $drive->{file};
1158 } else {
1159 $path = PVE::Storage::path($storecfg, $drive->{file});
1160 }
1161
1162 if($path =~ m/^iscsi\:\/\//){
1163 $devicetype = 'generic';
1164 } else {
1165 if (my $info = path_is_scsi($path)) {
1166 if ($info->{type} == 0) {
1167 $devicetype = 'block';
1168 } elsif ($info->{type} == 1) { # tape
1169 $devicetype = 'generic';
1170 }
1171 }
1172 }
1173 }
1174
1175 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1176 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1177 } else {
1178 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1179 }
1180
1181 } elsif ($drive->{interface} eq 'ide'){
1182 $maxdev = 2;
1183 my $controller = int($drive->{index} / $maxdev);
1184 my $unit = $drive->{index} % $maxdev;
1185 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1186
1187 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1188 } elsif ($drive->{interface} eq 'sata'){
1189 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1190 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1191 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1192 } elsif ($drive->{interface} eq 'usb') {
1193 die "implement me";
1194 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1195 } else {
1196 die "unsupported interface type";
1197 }
1198
1199 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1200
1201 return $device;
1202 }
1203
1204 sub get_initiator_name {
1205 my $initiator;
1206
1207 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1208 while (defined(my $line = <$fh>)) {
1209 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1210 $initiator = $1;
1211 last;
1212 }
1213 $fh->close();
1214
1215 return $initiator;
1216 }
1217
1218 sub print_drive_full {
1219 my ($storecfg, $vmid, $drive) = @_;
1220
1221 my $opts = '';
1222 foreach my $o (@qemu_drive_options) {
1223 next if $o eq 'bootindex';
1224 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1225 }
1226
1227 foreach my $o (qw(bps bps_rd bps_wr)) {
1228 my $v = $drive->{"m$o"};
1229 $opts .= ",$o=" . int($v*1024*1024) if $v;
1230 }
1231
1232 # use linux-aio by default (qemu default is threads)
1233 $opts .= ",aio=native" if !$drive->{aio};
1234
1235 my $path;
1236 my $volid = $drive->{file};
1237 if (drive_is_cdrom($drive)) {
1238 $path = get_iso_path($storecfg, $vmid, $volid);
1239 } else {
1240 if ($volid =~ m|^/|) {
1241 $path = $volid;
1242 } else {
1243 $path = PVE::Storage::path($storecfg, $volid);
1244 }
1245 }
1246
1247 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1248
1249 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1250 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1251
1252 my $pathinfo = $path ? "file=$path," : '';
1253
1254 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1255 }
1256
1257 sub print_netdevice_full {
1258 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1259
1260 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1261
1262 my $device = $net->{model};
1263 if ($net->{model} eq 'virtio') {
1264 $device = 'virtio-net-pci';
1265 };
1266
1267 my $pciaddr = print_pci_addr("$netid", $bridges);
1268 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1269 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1270 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1271 my $vectors = $net->{queues} * 2 + 2;
1272 $tmpstr .= ",vectors=$vectors,mq=on";
1273 }
1274 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1275 return $tmpstr;
1276 }
1277
1278 sub print_netdev_full {
1279 my ($vmid, $conf, $net, $netid) = @_;
1280
1281 my $i = '';
1282 if ($netid =~ m/^net(\d+)$/) {
1283 $i = int($1);
1284 }
1285
1286 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1287
1288 my $ifname = "tap${vmid}i$i";
1289
1290 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1291 die "interface name '$ifname' is too long (max 15 character)\n"
1292 if length($ifname) >= 16;
1293
1294 my $vhostparam = '';
1295 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1296
1297 my $vmname = $conf->{name} || "vm$vmid";
1298
1299 my $netdev = "";
1300
1301 if ($net->{bridge}) {
1302 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1303 } else {
1304 $netdev = "type=user,id=$netid,hostname=$vmname";
1305 }
1306
1307 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1308
1309 return $netdev;
1310 }
1311
1312 sub drive_is_cdrom {
1313 my ($drive) = @_;
1314
1315 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1316
1317 }
1318
1319 sub parse_numa {
1320 my ($data) = @_;
1321
1322 my $res = {};
1323
1324 foreach my $kvp (split(/,/, $data)) {
1325
1326 if ($kvp =~ m/^memory=(\S+)$/) {
1327 $res->{memory} = $1;
1328 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1329 $res->{policy} = $1;
1330 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1331 $res->{cpus}->{start} = $1;
1332 $res->{cpus}->{end} = $3;
1333 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1334 $res->{hostnodes}->{start} = $1;
1335 $res->{hostnodes}->{end} = $3;
1336 } else {
1337 return undef;
1338 }
1339 }
1340
1341 return $res;
1342 }
1343
1344 sub parse_hostpci {
1345 my ($value) = @_;
1346
1347 return undef if !$value;
1348
1349
1350 my @list = split(/,/, $value);
1351 my $found;
1352
1353 my $res = {};
1354 foreach my $kv (@list) {
1355
1356 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1357 $found = 1;
1358 if(defined($4)){
1359 push @{$res->{pciid}}, { id => $2 , function => $4};
1360
1361 }else{
1362 my $pcidevices = lspci($2);
1363 $res->{pciid} = $pcidevices->{$2};
1364 }
1365 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1366 $res->{driver} = $1;
1367 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1368 $res->{rombar} = $1;
1369 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1370 $res->{'x-vga'} = $1;
1371 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1372 $res->{pcie} = 1 if $1 == 1;
1373 } else {
1374 warn "unknown hostpci setting '$kv'\n";
1375 }
1376 }
1377
1378 return undef if !$found;
1379
1380 return $res;
1381 }
1382
1383 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1384 sub parse_net {
1385 my ($data) = @_;
1386
1387 my $res = {};
1388
1389 foreach my $kvp (split(/,/, $data)) {
1390
1391 if ($kvp =~ m/^(ne2k_pci|e1000|e1000-82540em|e1000-82544gc|e1000-82545em|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1392 my $model = lc($1);
1393 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1394 $res->{model} = $model;
1395 $res->{macaddr} = $mac;
1396 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1397 $res->{bridge} = $1;
1398 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1399 $res->{queues} = $1;
1400 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1401 $res->{rate} = $1;
1402 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1403 $res->{tag} = $1;
1404 } elsif ($kvp =~ m/^firewall=([01])$/) {
1405 $res->{firewall} = $1;
1406 } elsif ($kvp =~ m/^link_down=([01])$/) {
1407 $res->{link_down} = $1;
1408 } else {
1409 return undef;
1410 }
1411
1412 }
1413
1414 return undef if !$res->{model};
1415
1416 return $res;
1417 }
1418
1419 sub print_net {
1420 my $net = shift;
1421
1422 my $res = "$net->{model}";
1423 $res .= "=$net->{macaddr}" if $net->{macaddr};
1424 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1425 $res .= ",rate=$net->{rate}" if $net->{rate};
1426 $res .= ",tag=$net->{tag}" if $net->{tag};
1427 $res .= ",firewall=1" if $net->{firewall};
1428 $res .= ",link_down=1" if $net->{link_down};
1429 $res .= ",queues=$net->{queues}" if $net->{queues};
1430
1431 return $res;
1432 }
1433
1434 sub add_random_macs {
1435 my ($settings) = @_;
1436
1437 foreach my $opt (keys %$settings) {
1438 next if $opt !~ m/^net(\d+)$/;
1439 my $net = parse_net($settings->{$opt});
1440 next if !$net;
1441 $settings->{$opt} = print_net($net);
1442 }
1443 }
1444
1445 sub add_unused_volume {
1446 my ($config, $volid) = @_;
1447
1448 my $key;
1449 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1450 my $test = "unused$ind";
1451 if (my $vid = $config->{$test}) {
1452 return if $vid eq $volid; # do not add duplicates
1453 } else {
1454 $key = $test;
1455 }
1456 }
1457
1458 die "To many unused volume - please delete them first.\n" if !$key;
1459
1460 $config->{$key} = $volid;
1461
1462 return $key;
1463 }
1464
1465 sub vm_is_volid_owner {
1466 my ($storecfg, $vmid, $volid) = @_;
1467
1468 if ($volid !~ m|^/|) {
1469 my ($path, $owner);
1470 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1471 if ($owner && ($owner == $vmid)) {
1472 return 1;
1473 }
1474 }
1475
1476 return undef;
1477 }
1478
1479 sub vmconfig_delete_pending_option {
1480 my ($conf, $key) = @_;
1481
1482 delete $conf->{pending}->{$key};
1483 my $pending_delete_hash = { $key => 1 };
1484 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1485 $pending_delete_hash->{$opt} = 1;
1486 }
1487 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1488 }
1489
1490 sub vmconfig_undelete_pending_option {
1491 my ($conf, $key) = @_;
1492
1493 my $pending_delete_hash = {};
1494 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1495 $pending_delete_hash->{$opt} = 1;
1496 }
1497 delete $pending_delete_hash->{$key};
1498
1499 my @keylist = keys %$pending_delete_hash;
1500 if (scalar(@keylist)) {
1501 $conf->{pending}->{delete} = join(',', @keylist);
1502 } else {
1503 delete $conf->{pending}->{delete};
1504 }
1505 }
1506
1507 sub vmconfig_register_unused_drive {
1508 my ($storecfg, $vmid, $conf, $drive) = @_;
1509
1510 if (!drive_is_cdrom($drive)) {
1511 my $volid = $drive->{file};
1512 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1513 add_unused_volume($conf, $volid, $vmid);
1514 }
1515 }
1516 }
1517
1518 sub vmconfig_cleanup_pending {
1519 my ($conf) = @_;
1520
1521 # remove pending changes when nothing changed
1522 my $changes;
1523 foreach my $opt (keys %{$conf->{pending}}) {
1524 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1525 $changes = 1;
1526 delete $conf->{pending}->{$opt};
1527 }
1528 }
1529
1530 # remove delete if option is not set
1531 my $pending_delete_hash = {};
1532 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1533 if (defined($conf->{$opt})) {
1534 $pending_delete_hash->{$opt} = 1;
1535 } else {
1536 $changes = 1;
1537 }
1538 }
1539
1540 my @keylist = keys %$pending_delete_hash;
1541 if (scalar(@keylist)) {
1542 $conf->{pending}->{delete} = join(',', @keylist);
1543 } else {
1544 delete $conf->{pending}->{delete};
1545 }
1546
1547 return $changes;
1548 }
1549
1550 my $valid_smbios1_options = {
1551 manufacturer => '\S+',
1552 product => '\S+',
1553 version => '\S+',
1554 serial => '\S+',
1555 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1556 sku => '\S+',
1557 family => '\S+',
1558 };
1559
1560 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1561 sub parse_smbios1 {
1562 my ($data) = @_;
1563
1564 my $res = {};
1565
1566 foreach my $kvp (split(/,/, $data)) {
1567 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1568 my ($k, $v) = split(/=/, $kvp);
1569 return undef if !defined($k) || !defined($v);
1570 return undef if !$valid_smbios1_options->{$k};
1571 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1572 $res->{$k} = $v;
1573 }
1574
1575 return $res;
1576 }
1577
1578 sub print_smbios1 {
1579 my ($smbios1) = @_;
1580
1581 my $data = '';
1582 foreach my $k (keys %$smbios1) {
1583 next if !defined($smbios1->{$k});
1584 next if !$valid_smbios1_options->{$k};
1585 $data .= ',' if $data;
1586 $data .= "$k=$smbios1->{$k}";
1587 }
1588 return $data;
1589 }
1590
1591 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1592 sub verify_smbios1 {
1593 my ($value, $noerr) = @_;
1594
1595 return $value if parse_smbios1($value);
1596
1597 return undef if $noerr;
1598
1599 die "unable to parse smbios (type 1) options\n";
1600 }
1601
1602 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1603 sub verify_bootdisk {
1604 my ($value, $noerr) = @_;
1605
1606 return $value if valid_drivename($value);
1607
1608 return undef if $noerr;
1609
1610 die "invalid boot disk '$value'\n";
1611 }
1612
1613 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1614 sub verify_numa {
1615 my ($value, $noerr) = @_;
1616
1617 return $value if parse_numa($value);
1618
1619 return undef if $noerr;
1620
1621 die "unable to parse numa options\n";
1622 }
1623
1624 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1625 sub verify_net {
1626 my ($value, $noerr) = @_;
1627
1628 return $value if parse_net($value);
1629
1630 return undef if $noerr;
1631
1632 die "unable to parse network options\n";
1633 }
1634
1635 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1636 sub verify_drive {
1637 my ($value, $noerr) = @_;
1638
1639 return $value if parse_drive(undef, $value);
1640
1641 return undef if $noerr;
1642
1643 die "unable to parse drive options\n";
1644 }
1645
1646 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1647 sub verify_hostpci {
1648 my ($value, $noerr) = @_;
1649
1650 return $value if parse_hostpci($value);
1651
1652 return undef if $noerr;
1653
1654 die "unable to parse pci id\n";
1655 }
1656
1657 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1658 sub verify_watchdog {
1659 my ($value, $noerr) = @_;
1660
1661 return $value if parse_watchdog($value);
1662
1663 return undef if $noerr;
1664
1665 die "unable to parse watchdog options\n";
1666 }
1667
1668 sub parse_watchdog {
1669 my ($value) = @_;
1670
1671 return undef if !$value;
1672
1673 my $res = {};
1674
1675 foreach my $p (split(/,/, $value)) {
1676 next if $p =~ m/^\s*$/;
1677
1678 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1679 $res->{model} = $2;
1680 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1681 $res->{action} = $2;
1682 } else {
1683 return undef;
1684 }
1685 }
1686
1687 return $res;
1688 }
1689
1690 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1691 sub verify_startup {
1692 my ($value, $noerr) = @_;
1693
1694 return $value if parse_startup($value);
1695
1696 return undef if $noerr;
1697
1698 die "unable to parse startup options\n";
1699 }
1700
1701 sub parse_startup {
1702 my ($value) = @_;
1703
1704 return undef if !$value;
1705
1706 my $res = {};
1707
1708 foreach my $p (split(/,/, $value)) {
1709 next if $p =~ m/^\s*$/;
1710
1711 if ($p =~ m/^(order=)?(\d+)$/) {
1712 $res->{order} = $2;
1713 } elsif ($p =~ m/^up=(\d+)$/) {
1714 $res->{up} = $1;
1715 } elsif ($p =~ m/^down=(\d+)$/) {
1716 $res->{down} = $1;
1717 } else {
1718 return undef;
1719 }
1720 }
1721
1722 return $res;
1723 }
1724
1725 sub parse_usb_device {
1726 my ($value) = @_;
1727
1728 return undef if !$value;
1729
1730 my @dl = split(/,/, $value);
1731 my $found;
1732
1733 my $res = {};
1734 foreach my $v (@dl) {
1735 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1736 $found = 1;
1737 $res->{vendorid} = $2;
1738 $res->{productid} = $4;
1739 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1740 $found = 1;
1741 $res->{hostbus} = $1;
1742 $res->{hostport} = $2;
1743 } elsif ($v =~ m/^spice$/) {
1744 $found = 1;
1745 $res->{spice} = 1;
1746 } else {
1747 return undef;
1748 }
1749 }
1750 return undef if !$found;
1751
1752 return $res;
1753 }
1754
1755 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1756 sub verify_usb_device {
1757 my ($value, $noerr) = @_;
1758
1759 return $value if parse_usb_device($value);
1760
1761 return undef if $noerr;
1762
1763 die "unable to parse usb device\n";
1764 }
1765
1766 # add JSON properties for create and set function
1767 sub json_config_properties {
1768 my $prop = shift;
1769
1770 foreach my $opt (keys %$confdesc) {
1771 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1772 $prop->{$opt} = $confdesc->{$opt};
1773 }
1774
1775 return $prop;
1776 }
1777
1778 sub check_type {
1779 my ($key, $value) = @_;
1780
1781 die "unknown setting '$key'\n" if !$confdesc->{$key};
1782
1783 my $type = $confdesc->{$key}->{type};
1784
1785 if (!defined($value)) {
1786 die "got undefined value\n";
1787 }
1788
1789 if ($value =~ m/[\n\r]/) {
1790 die "property contains a line feed\n";
1791 }
1792
1793 if ($type eq 'boolean') {
1794 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1795 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1796 die "type check ('boolean') failed - got '$value'\n";
1797 } elsif ($type eq 'integer') {
1798 return int($1) if $value =~ m/^(\d+)$/;
1799 die "type check ('integer') failed - got '$value'\n";
1800 } elsif ($type eq 'number') {
1801 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1802 die "type check ('number') failed - got '$value'\n";
1803 } elsif ($type eq 'string') {
1804 if (my $fmt = $confdesc->{$key}->{format}) {
1805 if ($fmt eq 'pve-qm-drive') {
1806 # special case - we need to pass $key to parse_drive()
1807 my $drive = parse_drive($key, $value);
1808 return $value if $drive;
1809 die "unable to parse drive options\n";
1810 }
1811 PVE::JSONSchema::check_format($fmt, $value);
1812 return $value;
1813 }
1814 $value =~ s/^\"(.*)\"$/$1/;
1815 return $value;
1816 } else {
1817 die "internal error"
1818 }
1819 }
1820
1821 sub lock_config_full {
1822 my ($vmid, $timeout, $code, @param) = @_;
1823
1824 my $filename = config_file_lock($vmid);
1825
1826 my $res = lock_file($filename, $timeout, $code, @param);
1827
1828 die $@ if $@;
1829
1830 return $res;
1831 }
1832
1833 sub lock_config_mode {
1834 my ($vmid, $timeout, $shared, $code, @param) = @_;
1835
1836 my $filename = config_file_lock($vmid);
1837
1838 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1839
1840 die $@ if $@;
1841
1842 return $res;
1843 }
1844
1845 sub lock_config {
1846 my ($vmid, $code, @param) = @_;
1847
1848 return lock_config_full($vmid, 10, $code, @param);
1849 }
1850
1851 sub cfs_config_path {
1852 my ($vmid, $node) = @_;
1853
1854 $node = $nodename if !$node;
1855 return "nodes/$node/qemu-server/$vmid.conf";
1856 }
1857
1858 sub check_iommu_support{
1859 #fixme : need to check IOMMU support
1860 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1861
1862 my $iommu=1;
1863 return $iommu;
1864
1865 }
1866
1867 sub config_file {
1868 my ($vmid, $node) = @_;
1869
1870 my $cfspath = cfs_config_path($vmid, $node);
1871 return "/etc/pve/$cfspath";
1872 }
1873
1874 sub config_file_lock {
1875 my ($vmid) = @_;
1876
1877 return "$lock_dir/lock-$vmid.conf";
1878 }
1879
1880 sub touch_config {
1881 my ($vmid) = @_;
1882
1883 my $conf = config_file($vmid);
1884 utime undef, undef, $conf;
1885 }
1886
1887 sub destroy_vm {
1888 my ($storecfg, $vmid, $keep_empty_config) = @_;
1889
1890 my $conffile = config_file($vmid);
1891
1892 my $conf = load_config($vmid);
1893
1894 check_lock($conf);
1895
1896 # only remove disks owned by this VM
1897 foreach_drive($conf, sub {
1898 my ($ds, $drive) = @_;
1899
1900 return if drive_is_cdrom($drive);
1901
1902 my $volid = $drive->{file};
1903
1904 return if !$volid || $volid =~ m|^/|;
1905
1906 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1907 return if !$path || !$owner || ($owner != $vmid);
1908
1909 PVE::Storage::vdisk_free($storecfg, $volid);
1910 });
1911
1912 if ($keep_empty_config) {
1913 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1914 } else {
1915 unlink $conffile;
1916 }
1917
1918 # also remove unused disk
1919 eval {
1920 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1921
1922 eval {
1923 PVE::Storage::foreach_volid($dl, sub {
1924 my ($volid, $sid, $volname, $d) = @_;
1925 PVE::Storage::vdisk_free($storecfg, $volid);
1926 });
1927 };
1928 warn $@ if $@;
1929
1930 };
1931 warn $@ if $@;
1932 }
1933
1934 sub load_config {
1935 my ($vmid, $node) = @_;
1936
1937 my $cfspath = cfs_config_path($vmid, $node);
1938
1939 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1940
1941 die "no such VM ('$vmid')\n" if !defined($conf);
1942
1943 return $conf;
1944 }
1945
1946 sub parse_vm_config {
1947 my ($filename, $raw) = @_;
1948
1949 return undef if !defined($raw);
1950
1951 my $res = {
1952 digest => Digest::SHA::sha1_hex($raw),
1953 snapshots => {},
1954 pending => {},
1955 };
1956
1957 $filename =~ m|/qemu-server/(\d+)\.conf$|
1958 || die "got strange filename '$filename'";
1959
1960 my $vmid = $1;
1961
1962 my $conf = $res;
1963 my $descr = '';
1964 my $section = '';
1965
1966 my @lines = split(/\n/, $raw);
1967 foreach my $line (@lines) {
1968 next if $line =~ m/^\s*$/;
1969
1970 if ($line =~ m/^\[PENDING\]\s*$/i) {
1971 $section = 'pending';
1972 $conf->{description} = $descr if $descr;
1973 $descr = '';
1974 $conf = $res->{$section} = {};
1975 next;
1976
1977 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1978 $section = $1;
1979 $conf->{description} = $descr if $descr;
1980 $descr = '';
1981 $conf = $res->{snapshots}->{$section} = {};
1982 next;
1983 }
1984
1985 if ($line =~ m/^\#(.*)\s*$/) {
1986 $descr .= PVE::Tools::decode_text($1) . "\n";
1987 next;
1988 }
1989
1990 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1991 $descr .= PVE::Tools::decode_text($2);
1992 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1993 $conf->{snapstate} = $1;
1994 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1995 my $key = $1;
1996 my $value = $2;
1997 $conf->{$key} = $value;
1998 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
1999 my $value = $1;
2000 if ($section eq 'pending') {
2001 $conf->{delete} = $value; # we parse this later
2002 } else {
2003 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2004 }
2005 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2006 my $key = $1;
2007 my $value = $2;
2008 eval { $value = check_type($key, $value); };
2009 if ($@) {
2010 warn "vm $vmid - unable to parse value of '$key' - $@";
2011 } else {
2012 my $fmt = $confdesc->{$key}->{format};
2013 if ($fmt && $fmt eq 'pve-qm-drive') {
2014 my $v = parse_drive($key, $value);
2015 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2016 $v->{file} = $volid;
2017 $value = print_drive($vmid, $v);
2018 } else {
2019 warn "vm $vmid - unable to parse value of '$key'\n";
2020 next;
2021 }
2022 }
2023
2024 if ($key eq 'cdrom') {
2025 $conf->{ide2} = $value;
2026 } else {
2027 $conf->{$key} = $value;
2028 }
2029 }
2030 }
2031 }
2032
2033 $conf->{description} = $descr if $descr;
2034
2035 delete $res->{snapstate}; # just to be sure
2036
2037 return $res;
2038 }
2039
2040 sub write_vm_config {
2041 my ($filename, $conf) = @_;
2042
2043 delete $conf->{snapstate}; # just to be sure
2044
2045 if ($conf->{cdrom}) {
2046 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2047 $conf->{ide2} = $conf->{cdrom};
2048 delete $conf->{cdrom};
2049 }
2050
2051 # we do not use 'smp' any longer
2052 if ($conf->{sockets}) {
2053 delete $conf->{smp};
2054 } elsif ($conf->{smp}) {
2055 $conf->{sockets} = $conf->{smp};
2056 delete $conf->{cores};
2057 delete $conf->{smp};
2058 }
2059
2060 my $used_volids = {};
2061
2062 my $cleanup_config = sub {
2063 my ($cref, $pending, $snapname) = @_;
2064
2065 foreach my $key (keys %$cref) {
2066 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2067 $key eq 'snapstate' || $key eq 'pending';
2068 my $value = $cref->{$key};
2069 if ($key eq 'delete') {
2070 die "propertry 'delete' is only allowed in [PENDING]\n"
2071 if !$pending;
2072 # fixme: check syntax?
2073 next;
2074 }
2075 eval { $value = check_type($key, $value); };
2076 die "unable to parse value of '$key' - $@" if $@;
2077
2078 $cref->{$key} = $value;
2079
2080 if (!$snapname && valid_drivename($key)) {
2081 my $drive = parse_drive($key, $value);
2082 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2083 }
2084 }
2085 };
2086
2087 &$cleanup_config($conf);
2088
2089 &$cleanup_config($conf->{pending}, 1);
2090
2091 foreach my $snapname (keys %{$conf->{snapshots}}) {
2092 die "internal error" if $snapname eq 'pending';
2093 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2094 }
2095
2096 # remove 'unusedX' settings if we re-add a volume
2097 foreach my $key (keys %$conf) {
2098 my $value = $conf->{$key};
2099 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2100 delete $conf->{$key};
2101 }
2102 }
2103
2104 my $generate_raw_config = sub {
2105 my ($conf) = @_;
2106
2107 my $raw = '';
2108
2109 # add description as comment to top of file
2110 my $descr = $conf->{description} || '';
2111 foreach my $cl (split(/\n/, $descr)) {
2112 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2113 }
2114
2115 foreach my $key (sort keys %$conf) {
2116 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2117 $raw .= "$key: $conf->{$key}\n";
2118 }
2119 return $raw;
2120 };
2121
2122 my $raw = &$generate_raw_config($conf);
2123
2124 if (scalar(keys %{$conf->{pending}})){
2125 $raw .= "\n[PENDING]\n";
2126 $raw .= &$generate_raw_config($conf->{pending});
2127 }
2128
2129 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2130 $raw .= "\n[$snapname]\n";
2131 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2132 }
2133
2134 return $raw;
2135 }
2136
2137 sub update_config_nolock {
2138 my ($vmid, $conf, $skiplock) = @_;
2139
2140 check_lock($conf) if !$skiplock;
2141
2142 my $cfspath = cfs_config_path($vmid);
2143
2144 PVE::Cluster::cfs_write_file($cfspath, $conf);
2145 }
2146
2147 sub update_config {
2148 my ($vmid, $conf, $skiplock) = @_;
2149
2150 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2151 }
2152
2153 sub load_defaults {
2154
2155 my $res = {};
2156
2157 # we use static defaults from our JSON schema configuration
2158 foreach my $key (keys %$confdesc) {
2159 if (defined(my $default = $confdesc->{$key}->{default})) {
2160 $res->{$key} = $default;
2161 }
2162 }
2163
2164 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2165 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2166
2167 return $res;
2168 }
2169
2170 sub config_list {
2171 my $vmlist = PVE::Cluster::get_vmlist();
2172 my $res = {};
2173 return $res if !$vmlist || !$vmlist->{ids};
2174 my $ids = $vmlist->{ids};
2175
2176 foreach my $vmid (keys %$ids) {
2177 my $d = $ids->{$vmid};
2178 next if !$d->{node} || $d->{node} ne $nodename;
2179 next if !$d->{type} || $d->{type} ne 'qemu';
2180 $res->{$vmid}->{exists} = 1;
2181 }
2182 return $res;
2183 }
2184
2185 # test if VM uses local resources (to prevent migration)
2186 sub check_local_resources {
2187 my ($conf, $noerr) = @_;
2188
2189 my $loc_res = 0;
2190
2191 $loc_res = 1 if $conf->{hostusb}; # old syntax
2192 $loc_res = 1 if $conf->{hostpci}; # old syntax
2193
2194 foreach my $k (keys %$conf) {
2195 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2196 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2197 }
2198
2199 die "VM uses local resources\n" if $loc_res && !$noerr;
2200
2201 return $loc_res;
2202 }
2203
2204 # check if used storages are available on all nodes (use by migrate)
2205 sub check_storage_availability {
2206 my ($storecfg, $conf, $node) = @_;
2207
2208 foreach_drive($conf, sub {
2209 my ($ds, $drive) = @_;
2210
2211 my $volid = $drive->{file};
2212 return if !$volid;
2213
2214 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2215 return if !$sid;
2216
2217 # check if storage is available on both nodes
2218 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2219 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2220 });
2221 }
2222
2223 # list nodes where all VM images are available (used by has_feature API)
2224 sub shared_nodes {
2225 my ($conf, $storecfg) = @_;
2226
2227 my $nodelist = PVE::Cluster::get_nodelist();
2228 my $nodehash = { map { $_ => 1 } @$nodelist };
2229 my $nodename = PVE::INotify::nodename();
2230
2231 foreach_drive($conf, sub {
2232 my ($ds, $drive) = @_;
2233
2234 my $volid = $drive->{file};
2235 return if !$volid;
2236
2237 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2238 if ($storeid) {
2239 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2240 if ($scfg->{disable}) {
2241 $nodehash = {};
2242 } elsif (my $avail = $scfg->{nodes}) {
2243 foreach my $node (keys %$nodehash) {
2244 delete $nodehash->{$node} if !$avail->{$node};
2245 }
2246 } elsif (!$scfg->{shared}) {
2247 foreach my $node (keys %$nodehash) {
2248 delete $nodehash->{$node} if $node ne $nodename
2249 }
2250 }
2251 }
2252 });
2253
2254 return $nodehash
2255 }
2256
2257 sub check_lock {
2258 my ($conf) = @_;
2259
2260 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2261 }
2262
2263 sub check_cmdline {
2264 my ($pidfile, $pid) = @_;
2265
2266 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2267 if (defined($fh)) {
2268 my $line = <$fh>;
2269 $fh->close;
2270 return undef if !$line;
2271 my @param = split(/\0/, $line);
2272
2273 my $cmd = $param[0];
2274 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2275
2276 for (my $i = 0; $i < scalar (@param); $i++) {
2277 my $p = $param[$i];
2278 next if !$p;
2279 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2280 my $p = $param[$i+1];
2281 return 1 if $p && ($p eq $pidfile);
2282 return undef;
2283 }
2284 }
2285 }
2286 return undef;
2287 }
2288
2289 sub check_running {
2290 my ($vmid, $nocheck, $node) = @_;
2291
2292 my $filename = config_file($vmid, $node);
2293
2294 die "unable to find configuration file for VM $vmid - no such machine\n"
2295 if !$nocheck && ! -f $filename;
2296
2297 my $pidfile = pidfile_name($vmid);
2298
2299 if (my $fd = IO::File->new("<$pidfile")) {
2300 my $st = stat($fd);
2301 my $line = <$fd>;
2302 close($fd);
2303
2304 my $mtime = $st->mtime;
2305 if ($mtime > time()) {
2306 warn "file '$filename' modified in future\n";
2307 }
2308
2309 if ($line =~ m/^(\d+)$/) {
2310 my $pid = $1;
2311 if (check_cmdline($pidfile, $pid)) {
2312 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2313 return $pid;
2314 }
2315 }
2316 }
2317 }
2318
2319 return undef;
2320 }
2321
2322 sub vzlist {
2323
2324 my $vzlist = config_list();
2325
2326 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2327
2328 while (defined(my $de = $fd->read)) {
2329 next if $de !~ m/^(\d+)\.pid$/;
2330 my $vmid = $1;
2331 next if !defined($vzlist->{$vmid});
2332 if (my $pid = check_running($vmid)) {
2333 $vzlist->{$vmid}->{pid} = $pid;
2334 }
2335 }
2336
2337 return $vzlist;
2338 }
2339
2340 sub disksize {
2341 my ($storecfg, $conf) = @_;
2342
2343 my $bootdisk = $conf->{bootdisk};
2344 return undef if !$bootdisk;
2345 return undef if !valid_drivename($bootdisk);
2346
2347 return undef if !$conf->{$bootdisk};
2348
2349 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2350 return undef if !defined($drive);
2351
2352 return undef if drive_is_cdrom($drive);
2353
2354 my $volid = $drive->{file};
2355 return undef if !$volid;
2356
2357 return $drive->{size};
2358 }
2359
2360 my $last_proc_pid_stat;
2361
2362 # get VM status information
2363 # This must be fast and should not block ($full == false)
2364 # We only query KVM using QMP if $full == true (this can be slow)
2365 sub vmstatus {
2366 my ($opt_vmid, $full) = @_;
2367
2368 my $res = {};
2369
2370 my $storecfg = PVE::Storage::config();
2371
2372 my $list = vzlist();
2373 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2374
2375 my $cpucount = $cpuinfo->{cpus} || 1;
2376
2377 foreach my $vmid (keys %$list) {
2378 next if $opt_vmid && ($vmid ne $opt_vmid);
2379
2380 my $cfspath = cfs_config_path($vmid);
2381 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2382
2383 my $d = {};
2384 $d->{pid} = $list->{$vmid}->{pid};
2385
2386 # fixme: better status?
2387 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2388
2389 my $size = disksize($storecfg, $conf);
2390 if (defined($size)) {
2391 $d->{disk} = 0; # no info available
2392 $d->{maxdisk} = $size;
2393 } else {
2394 $d->{disk} = 0;
2395 $d->{maxdisk} = 0;
2396 }
2397
2398 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2399 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2400 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2401
2402 $d->{name} = $conf->{name} || "VM $vmid";
2403 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2404
2405 if ($conf->{balloon}) {
2406 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2407 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2408 }
2409
2410 $d->{uptime} = 0;
2411 $d->{cpu} = 0;
2412 $d->{mem} = 0;
2413
2414 $d->{netout} = 0;
2415 $d->{netin} = 0;
2416
2417 $d->{diskread} = 0;
2418 $d->{diskwrite} = 0;
2419
2420 $d->{template} = is_template($conf);
2421
2422 $res->{$vmid} = $d;
2423 }
2424
2425 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2426 foreach my $dev (keys %$netdev) {
2427 next if $dev !~ m/^tap([1-9]\d*)i/;
2428 my $vmid = $1;
2429 my $d = $res->{$vmid};
2430 next if !$d;
2431
2432 $d->{netout} += $netdev->{$dev}->{receive};
2433 $d->{netin} += $netdev->{$dev}->{transmit};
2434 }
2435
2436 my $ctime = gettimeofday;
2437
2438 foreach my $vmid (keys %$list) {
2439
2440 my $d = $res->{$vmid};
2441 my $pid = $d->{pid};
2442 next if !$pid;
2443
2444 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2445 next if !$pstat; # not running
2446
2447 my $used = $pstat->{utime} + $pstat->{stime};
2448
2449 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2450
2451 if ($pstat->{vsize}) {
2452 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2453 }
2454
2455 my $old = $last_proc_pid_stat->{$pid};
2456 if (!$old) {
2457 $last_proc_pid_stat->{$pid} = {
2458 time => $ctime,
2459 used => $used,
2460 cpu => 0,
2461 };
2462 next;
2463 }
2464
2465 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2466
2467 if ($dtime > 1000) {
2468 my $dutime = $used - $old->{used};
2469
2470 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2471 $last_proc_pid_stat->{$pid} = {
2472 time => $ctime,
2473 used => $used,
2474 cpu => $d->{cpu},
2475 };
2476 } else {
2477 $d->{cpu} = $old->{cpu};
2478 }
2479 }
2480
2481 return $res if !$full;
2482
2483 my $qmpclient = PVE::QMPClient->new();
2484
2485 my $ballooncb = sub {
2486 my ($vmid, $resp) = @_;
2487
2488 my $info = $resp->{'return'};
2489 return if !$info->{max_mem};
2490
2491 my $d = $res->{$vmid};
2492
2493 # use memory assigned to VM
2494 $d->{maxmem} = $info->{max_mem};
2495 $d->{balloon} = $info->{actual};
2496
2497 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2498 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2499 $d->{freemem} = $info->{free_mem};
2500 }
2501
2502 };
2503
2504 my $blockstatscb = sub {
2505 my ($vmid, $resp) = @_;
2506 my $data = $resp->{'return'} || [];
2507 my $totalrdbytes = 0;
2508 my $totalwrbytes = 0;
2509 for my $blockstat (@$data) {
2510 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2511 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2512 }
2513 $res->{$vmid}->{diskread} = $totalrdbytes;
2514 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2515 };
2516
2517 my $statuscb = sub {
2518 my ($vmid, $resp) = @_;
2519
2520 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2521 # this fails if ballon driver is not loaded, so this must be
2522 # the last commnand (following command are aborted if this fails).
2523 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2524
2525 my $status = 'unknown';
2526 if (!defined($status = $resp->{'return'}->{status})) {
2527 warn "unable to get VM status\n";
2528 return;
2529 }
2530
2531 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2532 };
2533
2534 foreach my $vmid (keys %$list) {
2535 next if $opt_vmid && ($vmid ne $opt_vmid);
2536 next if !$res->{$vmid}->{pid}; # not running
2537 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2538 }
2539
2540 $qmpclient->queue_execute(undef, 1);
2541
2542 foreach my $vmid (keys %$list) {
2543 next if $opt_vmid && ($vmid ne $opt_vmid);
2544 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2545 }
2546
2547 return $res;
2548 }
2549
2550 sub foreach_dimm {
2551 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2552
2553 my $dimm_id = 0;
2554 my $current_size = 1024;
2555 my $dimm_size = 512;
2556 return if $current_size == $memory;
2557
2558 for (my $j = 0; $j < 8; $j++) {
2559 for (my $i = 0; $i < 32; $i++) {
2560 my $name = "dimm${dimm_id}";
2561 $dimm_id++;
2562 my $numanode = $i % $sockets;
2563 $current_size += $dimm_size;
2564 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2565 return $current_size if $current_size >= $memory;
2566 }
2567 $dimm_size *= 2;
2568 }
2569 }
2570
2571 sub foreach_drive {
2572 my ($conf, $func) = @_;
2573
2574 foreach my $ds (keys %$conf) {
2575 next if !valid_drivename($ds);
2576
2577 my $drive = parse_drive($ds, $conf->{$ds});
2578 next if !$drive;
2579
2580 &$func($ds, $drive);
2581 }
2582 }
2583
2584 sub foreach_volid {
2585 my ($conf, $func) = @_;
2586
2587 my $volhash = {};
2588
2589 my $test_volid = sub {
2590 my ($volid, $is_cdrom) = @_;
2591
2592 return if !$volid;
2593
2594 $volhash->{$volid} = $is_cdrom || 0;
2595 };
2596
2597 foreach_drive($conf, sub {
2598 my ($ds, $drive) = @_;
2599 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2600 });
2601
2602 foreach my $snapname (keys %{$conf->{snapshots}}) {
2603 my $snap = $conf->{snapshots}->{$snapname};
2604 &$test_volid($snap->{vmstate}, 0);
2605 foreach_drive($snap, sub {
2606 my ($ds, $drive) = @_;
2607 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2608 });
2609 }
2610
2611 foreach my $volid (keys %$volhash) {
2612 &$func($volid, $volhash->{$volid});
2613 }
2614 }
2615
2616 sub vga_conf_has_spice {
2617 my ($vga) = @_;
2618
2619 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2620
2621 return $1 || 1;
2622 }
2623
2624 sub config_to_command {
2625 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2626
2627 my $cmd = [];
2628 my $globalFlags = [];
2629 my $machineFlags = [];
2630 my $rtcFlags = [];
2631 my $cpuFlags = [];
2632 my $devices = [];
2633 my $pciaddr = '';
2634 my $bridges = {};
2635 my $kvmver = kvm_user_version();
2636 my $vernum = 0; # unknown
2637 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2638 $vernum = $1*1000000+$2*1000;
2639 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2640 $vernum = $1*1000000+$2*1000+$3;
2641 }
2642
2643 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2644
2645 my $have_ovz = -f '/proc/vz/vestat';
2646
2647 my $q35 = machine_type_is_q35($conf);
2648 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2649 my $machine_type = $forcemachine || $conf->{machine};
2650
2651 push @$cmd, '/usr/bin/kvm';
2652
2653 push @$cmd, '-id', $vmid;
2654
2655 my $use_virtio = 0;
2656
2657 my $qmpsocket = qmp_socket($vmid);
2658 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2659 push @$cmd, '-mon', "chardev=qmp,mode=control";
2660
2661 my $socket = vnc_socket($vmid);
2662 push @$cmd, '-vnc', "unix:$socket,x509,password";
2663
2664 push @$cmd, '-pidfile' , pidfile_name($vmid);
2665
2666 push @$cmd, '-daemonize';
2667
2668 if ($conf->{smbios1}) {
2669 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2670 }
2671
2672 if ($q35) {
2673 # the q35 chipset support native usb2, so we enable usb controller
2674 # by default for this machine type
2675 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2676 } else {
2677 $pciaddr = print_pci_addr("piix3", $bridges);
2678 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2679
2680 my $use_usb2 = 0;
2681 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2682 next if !$conf->{"usb$i"};
2683 $use_usb2 = 1;
2684 }
2685 # include usb device config
2686 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2687 }
2688
2689 my $vga = $conf->{vga};
2690
2691 my $qxlnum = vga_conf_has_spice($vga);
2692 $vga = 'qxl' if $qxlnum;
2693
2694 if (!$vga) {
2695 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2696 $conf->{ostype} eq 'win7' ||
2697 $conf->{ostype} eq 'w2k8')) {
2698 $vga = 'std';
2699 } else {
2700 $vga = 'cirrus';
2701 }
2702 }
2703
2704 # enable absolute mouse coordinates (needed by vnc)
2705 my $tablet;
2706 if (defined($conf->{tablet})) {
2707 $tablet = $conf->{tablet};
2708 } else {
2709 $tablet = $defaults->{tablet};
2710 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2711 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2712 }
2713
2714 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2715
2716 # host pci devices
2717 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2718 my $d = parse_hostpci($conf->{"hostpci$i"});
2719 next if !$d;
2720
2721 my $pcie = $d->{pcie};
2722 if($pcie){
2723 die "q35 machine model is not enabled" if !$q35;
2724 $pciaddr = print_pcie_addr("hostpci$i");
2725 }else{
2726 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2727 }
2728
2729 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2730 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2731 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2732 if ($xvga && $xvga ne '') {
2733 push @$cpuFlags, 'kvm=off';
2734 $vga = 'none';
2735 }
2736 $driver = "vfio-pci" if $xvga ne '';
2737 my $pcidevices = $d->{pciid};
2738 my $multifunction = 1 if @$pcidevices > 1;
2739
2740 my $j=0;
2741 foreach my $pcidevice (@$pcidevices) {
2742
2743 my $id = "hostpci$i";
2744 $id .= ".$j" if $multifunction;
2745 my $addr = $pciaddr;
2746 $addr .= ".$j" if $multifunction;
2747 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2748
2749 if($j == 0){
2750 $devicestr .= "$rombar$xvga";
2751 $devicestr .= ",multifunction=on" if $multifunction;
2752 }
2753
2754 push @$devices, '-device', $devicestr;
2755 $j++;
2756 }
2757 }
2758
2759 # usb devices
2760 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2761 my $d = parse_usb_device($conf->{"usb$i"});
2762 next if !$d;
2763 if ($d->{vendorid} && $d->{productid}) {
2764 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2765 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2766 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2767 } elsif ($d->{spice}) {
2768 # usb redir support for spice
2769 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2770 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2771 }
2772 }
2773
2774 # serial devices
2775 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2776 if (my $path = $conf->{"serial$i"}) {
2777 if ($path eq 'socket') {
2778 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2779 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2780 push @$devices, '-device', "isa-serial,chardev=serial$i";
2781 } else {
2782 die "no such serial device\n" if ! -c $path;
2783 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2784 push @$devices, '-device', "isa-serial,chardev=serial$i";
2785 }
2786 }
2787 }
2788
2789 # parallel devices
2790 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2791 if (my $path = $conf->{"parallel$i"}) {
2792 die "no such parallel device\n" if ! -c $path;
2793 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2794 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2795 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2796 }
2797 }
2798
2799 my $vmname = $conf->{name} || "vm$vmid";
2800
2801 push @$cmd, '-name', $vmname;
2802
2803 my $sockets = 1;
2804 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2805 $sockets = $conf->{sockets} if $conf->{sockets};
2806
2807 my $cores = $conf->{cores} || 1;
2808
2809 my $maxcpus = $sockets * $cores;
2810
2811 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2812
2813 my $allowed_vcpus = $cpuinfo->{cpus};
2814
2815 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2816 if ($allowed_vcpus < $maxcpus);
2817
2818 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2819
2820 push @$cmd, '-nodefaults';
2821
2822 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2823
2824 my $bootindex_hash = {};
2825 my $i = 1;
2826 foreach my $o (split(//, $bootorder)) {
2827 $bootindex_hash->{$o} = $i*100;
2828 $i++;
2829 }
2830
2831 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2832
2833 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2834
2835 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2836
2837 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2838
2839 # time drift fix
2840 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2841
2842 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2843 my $useLocaltime = $conf->{localtime};
2844
2845 if (my $ost = $conf->{ostype}) {
2846 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2847
2848 if ($ost =~ m/^w/) { # windows
2849 $useLocaltime = 1 if !defined($conf->{localtime});
2850
2851 # use time drift fix when acpi is enabled
2852 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2853 $tdf = 1 if !defined($conf->{tdf});
2854 }
2855 }
2856
2857 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2858 $ost eq 'wvista') {
2859 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2860 push @$cmd, '-no-hpet';
2861 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2862 push @$cpuFlags , 'hv_spinlocks=0x1fff' if !$nokvm;
2863 push @$cpuFlags , 'hv_vapic' if !$nokvm;
2864 push @$cpuFlags , 'hv_time' if !$nokvm;
2865
2866 } else {
2867 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2868 }
2869 }
2870
2871 if ($ost eq 'win7' || $ost eq 'win8') {
2872 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2873 }
2874 }
2875
2876 push @$rtcFlags, 'driftfix=slew' if $tdf;
2877
2878 if ($nokvm) {
2879 push @$machineFlags, 'accel=tcg';
2880 } else {
2881 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2882 }
2883
2884 if ($machine_type) {
2885 push @$machineFlags, "type=${machine_type}";
2886 }
2887
2888 if ($conf->{startdate}) {
2889 push @$rtcFlags, "base=$conf->{startdate}";
2890 } elsif ($useLocaltime) {
2891 push @$rtcFlags, 'base=localtime';
2892 }
2893
2894 my $cpu = $nokvm ? "qemu64" : "kvm64";
2895 $cpu = $conf->{cpu} if $conf->{cpu};
2896
2897 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2898
2899 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2900
2901 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2902
2903 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2904
2905 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2906
2907 push @$cpuFlags , '+kvm_pv_unhalt' if !$nokvm;
2908 push @$cpuFlags , '+kvm_pv_eoi' if !$nokvm;
2909 }
2910
2911 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2912
2913 push @$cmd, '-cpu', "$cpu,enforce";
2914
2915 my $memory = $conf->{memory} || $defaults->{memory};
2916 my $static_memory = 0;
2917 my $dimm_memory = 0;
2918
2919 if ($hotplug_features->{memory}) {
2920 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2921 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2922 $static_memory = $STATICMEM;
2923 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2924 $dimm_memory = $memory - $static_memory;
2925 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2926
2927 } else {
2928
2929 $static_memory = $memory;
2930 push @$cmd, '-m', $static_memory;
2931 }
2932
2933 if ($conf->{numa}) {
2934
2935 my $numa_totalmemory = undef;
2936 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2937 next if !$conf->{"numa$i"};
2938 my $numa = parse_numa($conf->{"numa$i"});
2939 next if !$numa;
2940 # memory
2941 die "missing numa node$i memory value\n" if !$numa->{memory};
2942 my $numa_memory = $numa->{memory};
2943 $numa_totalmemory += $numa_memory;
2944 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2945
2946 # cpus
2947 my $cpus_start = $numa->{cpus}->{start};
2948 die "missing numa node$i cpus\n" if !defined($cpus_start);
2949 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2950 my $cpus = $cpus_start;
2951 if (defined($cpus_end)) {
2952 $cpus .= "-$cpus_end";
2953 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2954 }
2955
2956 # hostnodes
2957 my $hostnodes_start = $numa->{hostnodes}->{start};
2958 if (defined($hostnodes_start)) {
2959 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2960 my $hostnodes = $hostnodes_start;
2961 if (defined($hostnodes_end)) {
2962 $hostnodes .= "-$hostnodes_end";
2963 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2964 }
2965
2966 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2967 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2968 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2969 }
2970
2971 # policy
2972 my $policy = $numa->{policy};
2973 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2974 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2975 }
2976
2977 push @$cmd, '-object', $numa_object;
2978 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2979 }
2980
2981 die "total memory for NUMA nodes must be equal to vm static memory\n"
2982 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2983
2984 #if no custom tology, we split memory and cores across numa nodes
2985 if(!$numa_totalmemory) {
2986
2987 my $numa_memory = ($static_memory / $sockets) . "M";
2988
2989 for (my $i = 0; $i < $sockets; $i++) {
2990
2991 my $cpustart = ($cores * $i);
2992 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2993 my $cpus = $cpustart;
2994 $cpus .= "-$cpuend" if $cpuend;
2995
2996 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2997 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2998 }
2999 }
3000 }
3001
3002 if ($hotplug_features->{memory}) {
3003 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
3004 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3005 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
3006 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
3007
3008 #if dimm_memory is not aligned to dimm map
3009 if($current_size > $memory) {
3010 $conf->{memory} = $current_size;
3011 update_config_nolock($vmid, $conf, 1);
3012 }
3013 });
3014 }
3015
3016 push @$cmd, '-S' if $conf->{freeze};
3017
3018 # set keyboard layout
3019 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3020 push @$cmd, '-k', $kb if $kb;
3021
3022 # enable sound
3023 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3024 #push @$cmd, '-soundhw', 'es1370';
3025 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3026
3027 if($conf->{agent}) {
3028 my $qgasocket = qmp_socket($vmid, 1);
3029 my $pciaddr = print_pci_addr("qga0", $bridges);
3030 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3031 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3032 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3033 }
3034
3035 my $spice_port;
3036
3037 if ($qxlnum) {
3038 if ($qxlnum > 1) {
3039 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3040 for(my $i = 1; $i < $qxlnum; $i++){
3041 my $pciaddr = print_pci_addr("vga$i", $bridges);
3042 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3043 }
3044 } else {
3045 # assume other OS works like Linux
3046 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3047 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3048 }
3049 }
3050
3051 my $pciaddr = print_pci_addr("spice", $bridges);
3052
3053 $spice_port = PVE::Tools::next_spice_port();
3054
3055 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3056
3057 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3058 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3059 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3060 }
3061
3062 # enable balloon by default, unless explicitly disabled
3063 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3064 $pciaddr = print_pci_addr("balloon0", $bridges);
3065 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3066 }
3067
3068 if ($conf->{watchdog}) {
3069 my $wdopts = parse_watchdog($conf->{watchdog});
3070 $pciaddr = print_pci_addr("watchdog", $bridges);
3071 my $watchdog = $wdopts->{model} || 'i6300esb';
3072 push @$devices, '-device', "$watchdog$pciaddr";
3073 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3074 }
3075
3076 my $vollist = [];
3077 my $scsicontroller = {};
3078 my $ahcicontroller = {};
3079 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3080
3081 # Add iscsi initiator name if available
3082 if (my $initiator = get_initiator_name()) {
3083 push @$devices, '-iscsi', "initiator-name=$initiator";
3084 }
3085
3086 foreach_drive($conf, sub {
3087 my ($ds, $drive) = @_;
3088
3089 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3090 push @$vollist, $drive->{file};
3091 }
3092
3093 $use_virtio = 1 if $ds =~ m/^virtio/;
3094
3095 if (drive_is_cdrom ($drive)) {
3096 if ($bootindex_hash->{d}) {
3097 $drive->{bootindex} = $bootindex_hash->{d};
3098 $bootindex_hash->{d} += 1;
3099 }
3100 } else {
3101 if ($bootindex_hash->{c}) {
3102 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3103 $bootindex_hash->{c} += 1;
3104 }
3105 }
3106
3107 if($drive->{interface} eq 'virtio'){
3108 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3109 }
3110
3111 if ($drive->{interface} eq 'scsi') {
3112
3113 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3114
3115 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3116 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3117 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr" if !$scsicontroller->{$controller};
3118 $scsicontroller->{$controller}=1;
3119 }
3120
3121 if ($drive->{interface} eq 'sata') {
3122 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3123 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3124 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3125 $ahcicontroller->{$controller}=1;
3126 }
3127
3128 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3129 push @$devices, '-drive',$drive_cmd;
3130 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3131 });
3132
3133 for (my $i = 0; $i < $MAX_NETS; $i++) {
3134 next if !$conf->{"net$i"};
3135 my $d = parse_net($conf->{"net$i"});
3136 next if !$d;
3137
3138 $use_virtio = 1 if $d->{model} eq 'virtio';
3139
3140 if ($bootindex_hash->{n}) {
3141 $d->{bootindex} = $bootindex_hash->{n};
3142 $bootindex_hash->{n} += 1;
3143 }
3144
3145 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3146 push @$devices, '-netdev', $netdevfull;
3147
3148 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3149 push @$devices, '-device', $netdevicefull;
3150 }
3151
3152 if (!$q35) {
3153 # add pci bridges
3154 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3155 $bridges->{1} = 1;
3156 $bridges->{2} = 1;
3157 }
3158
3159 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3160
3161 while (my ($k, $v) = each %$bridges) {
3162 $pciaddr = print_pci_addr("pci.$k");
3163 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3164 }
3165 }
3166
3167 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3168 # when the VM uses virtio devices.
3169 if (!$use_virtio && $have_ovz) {
3170
3171 my $cpuunits = defined($conf->{cpuunits}) ?
3172 $conf->{cpuunits} : $defaults->{cpuunits};
3173
3174 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3175
3176 # fixme: cpulimit is currently ignored
3177 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3178 }
3179
3180 # add custom args
3181 if ($conf->{args}) {
3182 my $aa = PVE::Tools::split_args($conf->{args});
3183 push @$cmd, @$aa;
3184 }
3185
3186 push @$cmd, @$devices;
3187 push @$cmd, '-rtc', join(',', @$rtcFlags)
3188 if scalar(@$rtcFlags);
3189 push @$cmd, '-machine', join(',', @$machineFlags)
3190 if scalar(@$machineFlags);
3191 push @$cmd, '-global', join(',', @$globalFlags)
3192 if scalar(@$globalFlags);
3193
3194 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3195 }
3196
3197 sub vnc_socket {
3198 my ($vmid) = @_;
3199 return "${var_run_tmpdir}/$vmid.vnc";
3200 }
3201
3202 sub spice_port {
3203 my ($vmid) = @_;
3204
3205 my $res = vm_mon_cmd($vmid, 'query-spice');
3206
3207 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3208 }
3209
3210 sub qmp_socket {
3211 my ($vmid, $qga) = @_;
3212 my $sockettype = $qga ? 'qga' : 'qmp';
3213 return "${var_run_tmpdir}/$vmid.$sockettype";
3214 }
3215
3216 sub pidfile_name {
3217 my ($vmid) = @_;
3218 return "${var_run_tmpdir}/$vmid.pid";
3219 }
3220
3221 sub vm_devices_list {
3222 my ($vmid) = @_;
3223
3224 my $res = vm_mon_cmd($vmid, 'query-pci');
3225 my $devices = {};
3226 foreach my $pcibus (@$res) {
3227 foreach my $device (@{$pcibus->{devices}}) {
3228 next if !$device->{'qdev_id'};
3229 if ($device->{'pci_bridge'}) {
3230 $devices->{$device->{'qdev_id'}} = 1;
3231 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3232 next if !$bridge_device->{'qdev_id'};
3233 $devices->{$bridge_device->{'qdev_id'}} = 1;
3234 $devices->{$device->{'qdev_id'}}++;
3235 }
3236 } else {
3237 $devices->{$device->{'qdev_id'}} = 1;
3238 }
3239 }
3240 }
3241
3242 my $resblock = vm_mon_cmd($vmid, 'query-block');
3243 foreach my $block (@$resblock) {
3244 if($block->{device} =~ m/^drive-(\S+)/){
3245 $devices->{$1} = 1;
3246 }
3247 }
3248
3249 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3250 foreach my $mice (@$resmice) {
3251 if ($mice->{name} eq 'QEMU HID Tablet') {
3252 $devices->{tablet} = 1;
3253 last;
3254 }
3255 }
3256
3257 return $devices;
3258 }
3259
3260 sub vm_deviceplug {
3261 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3262
3263 my $q35 = machine_type_is_q35($conf);
3264
3265 my $devices_list = vm_devices_list($vmid);
3266 return 1 if defined($devices_list->{$deviceid});
3267
3268 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3269
3270 if ($deviceid eq 'tablet') {
3271
3272 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3273
3274 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3275
3276 qemu_iothread_add($vmid, $deviceid, $device);
3277
3278 qemu_driveadd($storecfg, $vmid, $device);
3279 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3280
3281 qemu_deviceadd($vmid, $devicefull);
3282 eval { qemu_deviceaddverify($vmid, $deviceid); };
3283 if (my $err = $@) {
3284 eval { qemu_drivedel($vmid, $deviceid); };
3285 warn $@ if $@;
3286 die $err;
3287 }
3288
3289 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3290
3291 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3292 my $pciaddr = print_pci_addr($deviceid);
3293 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3294
3295 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3296
3297 qemu_deviceadd($vmid, $devicefull);
3298 qemu_deviceaddverify($vmid, $deviceid);
3299
3300 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3301
3302 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3303 qemu_driveadd($storecfg, $vmid, $device);
3304
3305 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3306 eval { qemu_deviceadd($vmid, $devicefull); };
3307 if (my $err = $@) {
3308 eval { qemu_drivedel($vmid, $deviceid); };
3309 warn $@ if $@;
3310 die $err;
3311 }
3312
3313 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3314
3315 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3316 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3317 qemu_deviceadd($vmid, $netdevicefull);
3318 eval { qemu_deviceaddverify($vmid, $deviceid); };
3319 if (my $err = $@) {
3320 eval { qemu_netdevdel($vmid, $deviceid); };
3321 warn $@ if $@;
3322 die $err;
3323 }
3324
3325 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3326
3327 my $bridgeid = $2;
3328 my $pciaddr = print_pci_addr($deviceid);
3329 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3330
3331 qemu_deviceadd($vmid, $devicefull);
3332 qemu_deviceaddverify($vmid, $deviceid);
3333
3334 } else {
3335 die "can't hotplug device '$deviceid'\n";
3336 }
3337
3338 return 1;
3339 }
3340
3341 # fixme: this should raise exceptions on error!
3342 sub vm_deviceunplug {
3343 my ($vmid, $conf, $deviceid) = @_;
3344
3345 my $devices_list = vm_devices_list($vmid);
3346 return 1 if !defined($devices_list->{$deviceid});
3347
3348 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3349
3350 if ($deviceid eq 'tablet') {
3351
3352 qemu_devicedel($vmid, $deviceid);
3353
3354 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3355
3356 qemu_devicedel($vmid, $deviceid);
3357 qemu_devicedelverify($vmid, $deviceid);
3358 qemu_drivedel($vmid, $deviceid);
3359 qemu_iothread_del($conf, $vmid, $deviceid);
3360
3361 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3362
3363 qemu_devicedel($vmid, $deviceid);
3364 qemu_devicedelverify($vmid, $deviceid);
3365
3366 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3367
3368 qemu_devicedel($vmid, $deviceid);
3369 qemu_drivedel($vmid, $deviceid);
3370 qemu_deletescsihw($conf, $vmid, $deviceid);
3371
3372 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3373
3374 qemu_devicedel($vmid, $deviceid);
3375 qemu_devicedelverify($vmid, $deviceid);
3376 qemu_netdevdel($vmid, $deviceid);
3377
3378 } else {
3379 die "can't unplug device '$deviceid'\n";
3380 }
3381
3382 return 1;
3383 }
3384
3385 sub qemu_deviceadd {
3386 my ($vmid, $devicefull) = @_;
3387
3388 $devicefull = "driver=".$devicefull;
3389 my %options = split(/[=,]/, $devicefull);
3390
3391 vm_mon_cmd($vmid, "device_add" , %options);
3392 }
3393
3394 sub qemu_devicedel {
3395 my ($vmid, $deviceid) = @_;
3396
3397 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3398 }
3399
3400 sub qemu_iothread_add {
3401 my($vmid, $deviceid, $device) = @_;
3402
3403 if ($device->{iothread}) {
3404 my $iothreads = vm_iothreads_list($vmid);
3405 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3406 }
3407 }
3408
3409 sub qemu_iothread_del {
3410 my($conf, $vmid, $deviceid) = @_;
3411
3412 my $device = parse_drive($deviceid, $conf->{$deviceid});
3413 if ($device->{iothread}) {
3414 my $iothreads = vm_iothreads_list($vmid);
3415 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3416 }
3417 }
3418
3419 sub qemu_objectadd {
3420 my($vmid, $objectid, $qomtype) = @_;
3421
3422 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3423
3424 return 1;
3425 }
3426
3427 sub qemu_objectdel {
3428 my($vmid, $objectid) = @_;
3429
3430 vm_mon_cmd($vmid, "object-del", id => $objectid);
3431
3432 return 1;
3433 }
3434
3435 sub qemu_driveadd {
3436 my ($storecfg, $vmid, $device) = @_;
3437
3438 my $drive = print_drive_full($storecfg, $vmid, $device);
3439 $drive =~ s/\\/\\\\/g;
3440 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3441
3442 # If the command succeeds qemu prints: "OK"
3443 return 1 if $ret =~ m/OK/s;
3444
3445 die "adding drive failed: $ret\n";
3446 }
3447
3448 sub qemu_drivedel {
3449 my($vmid, $deviceid) = @_;
3450
3451 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3452 $ret =~ s/^\s+//;
3453
3454 return 1 if $ret eq "";
3455
3456 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3457 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3458
3459 die "deleting drive $deviceid failed : $ret\n";
3460 }
3461
3462 sub qemu_deviceaddverify {
3463 my ($vmid, $deviceid) = @_;
3464
3465 for (my $i = 0; $i <= 5; $i++) {
3466 my $devices_list = vm_devices_list($vmid);
3467 return 1 if defined($devices_list->{$deviceid});
3468 sleep 1;
3469 }
3470
3471 die "error on hotplug device '$deviceid'\n";
3472 }
3473
3474
3475 sub qemu_devicedelverify {
3476 my ($vmid, $deviceid) = @_;
3477
3478 # need to verify that the device is correctly removed as device_del
3479 # is async and empty return is not reliable
3480
3481 for (my $i = 0; $i <= 5; $i++) {
3482 my $devices_list = vm_devices_list($vmid);
3483 return 1 if !defined($devices_list->{$deviceid});
3484 sleep 1;
3485 }
3486
3487 die "error on hot-unplugging device '$deviceid'\n";
3488 }
3489
3490 sub qemu_findorcreatescsihw {
3491 my ($storecfg, $conf, $vmid, $device) = @_;
3492
3493 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3494
3495 my $scsihwid="$controller_prefix$controller";
3496 my $devices_list = vm_devices_list($vmid);
3497
3498 if(!defined($devices_list->{$scsihwid})) {
3499 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3500 }
3501
3502 return 1;
3503 }
3504
3505 sub qemu_deletescsihw {
3506 my ($conf, $vmid, $opt) = @_;
3507
3508 my $device = parse_drive($opt, $conf->{$opt});
3509
3510 if ($conf->{scsihw} && ($conf->{scsihw} =~ m/^virtio-scsi-single/)) {
3511 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3512 return 1;
3513 }
3514
3515 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3516
3517 my $devices_list = vm_devices_list($vmid);
3518 foreach my $opt (keys %{$devices_list}) {
3519 if (PVE::QemuServer::valid_drivename($opt)) {
3520 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
3521 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
3522 return 1;
3523 }
3524 }
3525 }
3526
3527 my $scsihwid="scsihw$controller";
3528
3529 vm_deviceunplug($vmid, $conf, $scsihwid);
3530
3531 return 1;
3532 }
3533
3534 sub qemu_add_pci_bridge {
3535 my ($storecfg, $conf, $vmid, $device) = @_;
3536
3537 my $bridges = {};
3538
3539 my $bridgeid;
3540
3541 print_pci_addr($device, $bridges);
3542
3543 while (my ($k, $v) = each %$bridges) {
3544 $bridgeid = $k;
3545 }
3546 return 1 if !defined($bridgeid) || $bridgeid < 1;
3547
3548 my $bridge = "pci.$bridgeid";
3549 my $devices_list = vm_devices_list($vmid);
3550
3551 if (!defined($devices_list->{$bridge})) {
3552 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3553 }
3554
3555 return 1;
3556 }
3557
3558 sub qemu_set_link_status {
3559 my ($vmid, $device, $up) = @_;
3560
3561 vm_mon_cmd($vmid, "set_link", name => $device,
3562 up => $up ? JSON::true : JSON::false);
3563 }
3564
3565 sub qemu_netdevadd {
3566 my ($vmid, $conf, $device, $deviceid) = @_;
3567
3568 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3569 my %options = split(/[=,]/, $netdev);
3570
3571 vm_mon_cmd($vmid, "netdev_add", %options);
3572 return 1;
3573 }
3574
3575 sub qemu_netdevdel {
3576 my ($vmid, $deviceid) = @_;
3577
3578 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3579 }
3580
3581 sub qemu_cpu_hotplug {
3582 my ($vmid, $conf, $vcpus) = @_;
3583
3584 my $sockets = 1;
3585 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3586 $sockets = $conf->{sockets} if $conf->{sockets};
3587 my $cores = $conf->{cores} || 1;
3588 my $maxcpus = $sockets * $cores;
3589
3590 $vcpus = $maxcpus if !$vcpus;
3591
3592 die "you can't add more vcpus than maxcpus\n"
3593 if $vcpus > $maxcpus;
3594
3595 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3596 die "online cpu unplug is not yet possible\n"
3597 if $vcpus < $currentvcpus;
3598
3599 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3600 die "vcpus in running vm is different than configuration\n"
3601 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3602
3603 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3604 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3605 }
3606 }
3607
3608 sub qemu_memory_hotplug {
3609 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3610
3611 return $value if !check_running($vmid);
3612
3613 my $memory = $conf->{memory} || $defaults->{memory};
3614 $value = $defaults->{memory} if !$value;
3615 return $value if $value == $memory;
3616
3617 my $static_memory = $STATICMEM;
3618 my $dimm_memory = $memory - $static_memory;
3619
3620 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3621 die "memory unplug is not yet available" if $value < $memory;
3622 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3623
3624
3625 my $sockets = 1;
3626 $sockets = $conf->{sockets} if $conf->{sockets};
3627
3628 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3629 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3630
3631 return if $current_size <= $conf->{memory};
3632
3633 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3634 if (my $err = $@) {
3635 eval { qemu_objectdel($vmid, "mem-$name"); };
3636 die $err;
3637 }
3638
3639 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3640 if (my $err = $@) {
3641 eval { qemu_objectdel($vmid, "mem-$name"); };
3642 die $err;
3643 }
3644 #update conf after each succesful module hotplug
3645 $conf->{memory} = $current_size;
3646 update_config_nolock($vmid, $conf, 1);
3647 });
3648 }
3649
3650 sub qemu_block_set_io_throttle {
3651 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3652
3653 return if !check_running($vmid) ;
3654
3655 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($iops_wr));
3656
3657 }
3658
3659 # old code, only used to shutdown old VM after update
3660 sub __read_avail {
3661 my ($fh, $timeout) = @_;
3662
3663 my $sel = new IO::Select;
3664 $sel->add($fh);
3665
3666 my $res = '';
3667 my $buf;
3668
3669 my @ready;
3670 while (scalar (@ready = $sel->can_read($timeout))) {
3671 my $count;
3672 if ($count = $fh->sysread($buf, 8192)) {
3673 if ($buf =~ /^(.*)\(qemu\) $/s) {
3674 $res .= $1;
3675 last;
3676 } else {
3677 $res .= $buf;
3678 }
3679 } else {
3680 if (!defined($count)) {
3681 die "$!\n";
3682 }
3683 last;
3684 }
3685 }
3686
3687 die "monitor read timeout\n" if !scalar(@ready);
3688
3689 return $res;
3690 }
3691
3692 # old code, only used to shutdown old VM after update
3693 sub vm_monitor_command {
3694 my ($vmid, $cmdstr, $nocheck) = @_;
3695
3696 my $res;
3697
3698 eval {
3699 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3700
3701 my $sname = "${var_run_tmpdir}/$vmid.mon";
3702
3703 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3704 die "unable to connect to VM $vmid socket - $!\n";
3705
3706 my $timeout = 3;
3707
3708 # hack: migrate sometime blocks the monitor (when migrate_downtime
3709 # is set)
3710 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3711 $timeout = 60*60; # 1 hour
3712 }
3713
3714 # read banner;
3715 my $data = __read_avail($sock, $timeout);
3716
3717 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3718 die "got unexpected qemu monitor banner\n";
3719 }
3720
3721 my $sel = new IO::Select;
3722 $sel->add($sock);
3723
3724 if (!scalar(my @ready = $sel->can_write($timeout))) {
3725 die "monitor write error - timeout";
3726 }
3727
3728 my $fullcmd = "$cmdstr\r";
3729
3730 # syslog('info', "VM $vmid monitor command: $cmdstr");
3731
3732 my $b;
3733 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3734 die "monitor write error - $!";
3735 }
3736
3737 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3738
3739 $timeout = 20;
3740
3741 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3742 $timeout = 60*60; # 1 hour
3743 } elsif ($cmdstr =~ m/^(eject|change)/) {
3744 $timeout = 60; # note: cdrom mount command is slow
3745 }
3746 if ($res = __read_avail($sock, $timeout)) {
3747
3748 my @lines = split("\r?\n", $res);
3749
3750 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3751
3752 $res = join("\n", @lines);
3753 $res .= "\n";
3754 }
3755 };
3756
3757 my $err = $@;
3758
3759 if ($err) {
3760 syslog("err", "VM $vmid monitor command failed - $err");
3761 die $err;
3762 }
3763
3764 return $res;
3765 }
3766
3767 sub qemu_block_resize {
3768 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3769
3770 my $running = check_running($vmid);
3771
3772 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3773
3774 return if !$running;
3775
3776 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3777
3778 }
3779
3780 sub qemu_volume_snapshot {
3781 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3782
3783 my $running = check_running($vmid);
3784
3785 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3786
3787 return if !$running;
3788
3789 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3790
3791 }
3792
3793 sub qemu_volume_snapshot_delete {
3794 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3795
3796 my $running = check_running($vmid);
3797
3798 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3799
3800 return if !$running;
3801
3802 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3803 }
3804
3805 sub set_migration_caps {
3806 my ($vmid) = @_;
3807
3808 my $cap_ref = [];
3809
3810 my $enabled_cap = {
3811 "auto-converge" => 1,
3812 "xbzrle" => 0,
3813 "x-rdma-pin-all" => 0,
3814 "zero-blocks" => 0,
3815 };
3816
3817 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3818
3819 for my $supported_capability (@$supported_capabilities) {
3820 push @$cap_ref, {
3821 capability => $supported_capability->{capability},
3822 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3823 };
3824 }
3825
3826 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3827 }
3828
3829 my $fast_plug_option = {
3830 'lock' => 1,
3831 'name' => 1,
3832 'onboot' => 1,
3833 'shares' => 1,
3834 'startup' => 1,
3835 };
3836
3837 # hotplug changes in [PENDING]
3838 # $selection hash can be used to only apply specified options, for
3839 # example: { cores => 1 } (only apply changed 'cores')
3840 # $errors ref is used to return error messages
3841 sub vmconfig_hotplug_pending {
3842 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3843
3844 my $defaults = load_defaults();
3845
3846 # commit values which do not have any impact on running VM first
3847 # Note: those option cannot raise errors, we we do not care about
3848 # $selection and always apply them.
3849
3850 my $add_error = sub {
3851 my ($opt, $msg) = @_;
3852 $errors->{$opt} = "hotplug problem - $msg";
3853 };
3854
3855 my $changes = 0;
3856 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3857 if ($fast_plug_option->{$opt}) {
3858 $conf->{$opt} = $conf->{pending}->{$opt};
3859 delete $conf->{pending}->{$opt};
3860 $changes = 1;
3861 }
3862 }
3863
3864 if ($changes) {
3865 update_config_nolock($vmid, $conf, 1);
3866 $conf = load_config($vmid); # update/reload
3867 }
3868
3869 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3870
3871 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3872 foreach my $opt (@delete) {
3873 next if $selection && !$selection->{$opt};
3874 eval {
3875 if ($opt eq 'hotplug') {
3876 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3877 } elsif ($opt eq 'tablet') {
3878 die "skip\n" if !$hotplug_features->{usb};
3879 if ($defaults->{tablet}) {
3880 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3881 } else {
3882 vm_deviceunplug($vmid, $conf, $opt);
3883 }
3884 } elsif ($opt eq 'vcpus') {
3885 die "skip\n" if !$hotplug_features->{cpu};
3886 qemu_cpu_hotplug($vmid, $conf, undef);
3887 } elsif ($opt eq 'balloon') {
3888 # enable balloon device is not hotpluggable
3889 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3890 } elsif ($fast_plug_option->{$opt}) {
3891 # do nothing
3892 } elsif ($opt =~ m/^net(\d+)$/) {
3893 die "skip\n" if !$hotplug_features->{network};
3894 vm_deviceunplug($vmid, $conf, $opt);
3895 } elsif (valid_drivename($opt)) {
3896 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3897 vm_deviceunplug($vmid, $conf, $opt);
3898 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3899 } elsif ($opt =~ m/^memory$/) {
3900 die "skip\n" if !$hotplug_features->{memory};
3901 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3902 } else {
3903 die "skip\n";
3904 }
3905 };
3906 if (my $err = $@) {
3907 &$add_error($opt, $err) if $err ne "skip\n";
3908 } else {
3909 # save new config if hotplug was successful
3910 delete $conf->{$opt};
3911 vmconfig_undelete_pending_option($conf, $opt);
3912 update_config_nolock($vmid, $conf, 1);
3913 $conf = load_config($vmid); # update/reload
3914 }
3915 }
3916
3917 foreach my $opt (keys %{$conf->{pending}}) {
3918 next if $selection && !$selection->{$opt};
3919 my $value = $conf->{pending}->{$opt};
3920 eval {
3921 if ($opt eq 'hotplug') {
3922 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3923 } elsif ($opt eq 'tablet') {
3924 die "skip\n" if !$hotplug_features->{usb};
3925 if ($value == 1) {
3926 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3927 } elsif ($value == 0) {
3928 vm_deviceunplug($vmid, $conf, $opt);
3929 }
3930 } elsif ($opt eq 'vcpus') {
3931 die "skip\n" if !$hotplug_features->{cpu};
3932 qemu_cpu_hotplug($vmid, $conf, $value);
3933 } elsif ($opt eq 'balloon') {
3934 # enable/disable balloning device is not hotpluggable
3935 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3936 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3937 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3938
3939 # allow manual ballooning if shares is set to zero
3940 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3941 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3942 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3943 }
3944 } elsif ($opt =~ m/^net(\d+)$/) {
3945 # some changes can be done without hotplug
3946 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3947 $vmid, $opt, $value);
3948 } elsif (valid_drivename($opt)) {
3949 # some changes can be done without hotplug
3950 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3951 $vmid, $opt, $value, 1);
3952 } elsif ($opt =~ m/^memory$/) { #dimms
3953 die "skip\n" if !$hotplug_features->{memory};
3954 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3955 } else {
3956 die "skip\n"; # skip non-hot-pluggable options
3957 }
3958 };
3959 if (my $err = $@) {
3960 &$add_error($opt, $err) if $err ne "skip\n";
3961 } else {
3962 # save new config if hotplug was successful
3963 $conf->{$opt} = $value;
3964 delete $conf->{pending}->{$opt};
3965 update_config_nolock($vmid, $conf, 1);
3966 $conf = load_config($vmid); # update/reload
3967 }
3968 }
3969 }
3970
3971 sub vmconfig_apply_pending {
3972 my ($vmid, $conf, $storecfg) = @_;
3973
3974 # cold plug
3975
3976 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3977 foreach my $opt (@delete) { # delete
3978 die "internal error" if $opt =~ m/^unused/;
3979 $conf = load_config($vmid); # update/reload
3980 if (!defined($conf->{$opt})) {
3981 vmconfig_undelete_pending_option($conf, $opt);
3982 update_config_nolock($vmid, $conf, 1);
3983 } elsif (valid_drivename($opt)) {
3984 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3985 vmconfig_undelete_pending_option($conf, $opt);
3986 delete $conf->{$opt};
3987 update_config_nolock($vmid, $conf, 1);
3988 } else {
3989 vmconfig_undelete_pending_option($conf, $opt);
3990 delete $conf->{$opt};
3991 update_config_nolock($vmid, $conf, 1);
3992 }
3993 }
3994
3995 $conf = load_config($vmid); # update/reload
3996
3997 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3998 $conf = load_config($vmid); # update/reload
3999
4000 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
4001 # skip if nothing changed
4002 } elsif (valid_drivename($opt)) {
4003 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
4004 if defined($conf->{$opt});
4005 $conf->{$opt} = $conf->{pending}->{$opt};
4006 } else {
4007 $conf->{$opt} = $conf->{pending}->{$opt};
4008 }
4009
4010 delete $conf->{pending}->{$opt};
4011 update_config_nolock($vmid, $conf, 1);
4012 }
4013 }
4014
4015 my $safe_num_ne = sub {
4016 my ($a, $b) = @_;
4017
4018 return 0 if !defined($a) && !defined($b);
4019 return 1 if !defined($a);
4020 return 1 if !defined($b);
4021
4022 return $a != $b;
4023 };
4024
4025 my $safe_string_ne = sub {
4026 my ($a, $b) = @_;
4027
4028 return 0 if !defined($a) && !defined($b);
4029 return 1 if !defined($a);
4030 return 1 if !defined($b);
4031
4032 return $a ne $b;
4033 };
4034
4035 sub vmconfig_update_net {
4036 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4037
4038 my $newnet = parse_net($value);
4039
4040 if ($conf->{$opt}) {
4041 my $oldnet = parse_net($conf->{$opt});
4042
4043 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4044 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4045 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4046 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4047
4048 # for non online change, we try to hot-unplug
4049 die "skip\n" if !$hotplug;
4050 vm_deviceunplug($vmid, $conf, $opt);
4051 } else {
4052
4053 die "internal error" if $opt !~ m/net(\d+)/;
4054 my $iface = "tap${vmid}i$1";
4055
4056 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4057 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4058 }
4059
4060 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4061 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4062 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4063 PVE::Network::tap_unplug($iface);
4064 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
4065 }
4066
4067 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4068 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4069 }
4070
4071 return 1;
4072 }
4073 }
4074
4075 if ($hotplug) {
4076 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4077 } else {
4078 die "skip\n";
4079 }
4080 }
4081
4082 sub vmconfig_update_disk {
4083 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4084
4085 # fixme: do we need force?
4086
4087 my $drive = parse_drive($opt, $value);
4088
4089 if ($conf->{$opt}) {
4090
4091 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4092
4093 my $media = $drive->{media} || 'disk';
4094 my $oldmedia = $old_drive->{media} || 'disk';
4095 die "unable to change media type\n" if $media ne $oldmedia;
4096
4097 if (!drive_is_cdrom($old_drive)) {
4098
4099 if ($drive->{file} ne $old_drive->{file}) {
4100
4101 die "skip\n" if !$hotplug;
4102
4103 # unplug and register as unused
4104 vm_deviceunplug($vmid, $conf, $opt);
4105 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4106
4107 } else {
4108 # update existing disk
4109
4110 # skip non hotpluggable value
4111 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4112 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4113 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4114 die "skip\n";
4115 }
4116
4117 # apply throttle
4118 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4119 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4120 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4121 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4122 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4123 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4124 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4125 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4126 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4127 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4128 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4129 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4130
4131 qemu_block_set_io_throttle($vmid,"drive-$opt",
4132 ($drive->{mbps} || 0)*1024*1024,
4133 ($drive->{mbps_rd} || 0)*1024*1024,
4134 ($drive->{mbps_wr} || 0)*1024*1024,
4135 $drive->{iops} || 0,
4136 $drive->{iops_rd} || 0,
4137 $drive->{iops_wr} || 0,
4138 ($drive->{mbps_max} || 0)*1024*1024,
4139 ($drive->{mbps_rd_max} || 0)*1024*1024,
4140 ($drive->{mbps_wr_max} || 0)*1024*1024,
4141 $drive->{iops_max} || 0,
4142 $drive->{iops_rd_max} || 0,
4143 $drive->{iops_wr_max} || 0);
4144
4145 }
4146
4147 return 1;
4148 }
4149
4150 } else { # cdrom
4151
4152 if ($drive->{file} eq 'none') {
4153 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4154 } else {
4155 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4156 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4157 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4158 }
4159
4160 return 1;
4161 }
4162 }
4163 }
4164
4165 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4166 # hotplug new disks
4167 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4168 }
4169
4170 sub vm_start {
4171 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4172
4173 lock_config($vmid, sub {
4174 my $conf = load_config($vmid, $migratedfrom);
4175
4176 die "you can't start a vm if it's a template\n" if is_template($conf);
4177
4178 check_lock($conf) if !$skiplock;
4179
4180 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4181
4182 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4183 vmconfig_apply_pending($vmid, $conf, $storecfg);
4184 $conf = load_config($vmid); # update/reload
4185 }
4186
4187 my $defaults = load_defaults();
4188
4189 # set environment variable useful inside network script
4190 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4191
4192 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4193
4194 my $migrate_port = 0;
4195 my $migrate_uri;
4196 if ($statefile) {
4197 if ($statefile eq 'tcp') {
4198 my $localip = "localhost";
4199 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4200 if ($datacenterconf->{migration_unsecure}) {
4201 my $nodename = PVE::INotify::nodename();
4202 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4203 }
4204 $migrate_port = PVE::Tools::next_migrate_port();
4205 $migrate_uri = "tcp:${localip}:${migrate_port}";
4206 push @$cmd, '-incoming', $migrate_uri;
4207 push @$cmd, '-S';
4208 } else {
4209 push @$cmd, '-loadstate', $statefile;
4210 }
4211 } elsif ($paused) {
4212 push @$cmd, '-S';
4213 }
4214
4215 # host pci devices
4216 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4217 my $d = parse_hostpci($conf->{"hostpci$i"});
4218 next if !$d;
4219 my $pcidevices = $d->{pciid};
4220 foreach my $pcidevice (@$pcidevices) {
4221 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4222
4223 my $info = pci_device_info("0000:$pciid");
4224 die "IOMMU not present\n" if !check_iommu_support();
4225 die "no pci device info for device '$pciid'\n" if !$info;
4226
4227 if ($d->{driver} && $d->{driver} eq "vfio") {
4228 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4229 } else {
4230 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4231 }
4232
4233 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4234 }
4235 }
4236
4237 PVE::Storage::activate_volumes($storecfg, $vollist);
4238
4239 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4240 umask => 0077); };
4241 my $err = $@;
4242 die "start failed: $err" if $err;
4243
4244 print "migration listens on $migrate_uri\n" if $migrate_uri;
4245
4246 if ($statefile && $statefile ne 'tcp') {
4247 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4248 warn $@ if $@;
4249 }
4250
4251 if ($migratedfrom) {
4252
4253 eval {
4254 set_migration_caps($vmid);
4255 };
4256 warn $@ if $@;
4257
4258 if ($spice_port) {
4259 print "spice listens on port $spice_port\n";
4260 if ($spice_ticket) {
4261 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4262 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4263 }
4264 }
4265
4266 } else {
4267
4268 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4269 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4270 if $conf->{balloon};
4271 }
4272
4273 foreach my $opt (keys %$conf) {
4274 next if $opt !~ m/^net\d+$/;
4275 my $nicconf = parse_net($conf->{$opt});
4276 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4277 }
4278 }
4279
4280 vm_mon_cmd_nocheck($vmid, 'qom-set',
4281 path => "machine/peripheral/balloon0",
4282 property => "guest-stats-polling-interval",
4283 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
4284
4285 });
4286 }
4287
4288 sub vm_mon_cmd {
4289 my ($vmid, $execute, %params) = @_;
4290
4291 my $cmd = { execute => $execute, arguments => \%params };
4292 vm_qmp_command($vmid, $cmd);
4293 }
4294
4295 sub vm_mon_cmd_nocheck {
4296 my ($vmid, $execute, %params) = @_;
4297
4298 my $cmd = { execute => $execute, arguments => \%params };
4299 vm_qmp_command($vmid, $cmd, 1);
4300 }
4301
4302 sub vm_qmp_command {
4303 my ($vmid, $cmd, $nocheck) = @_;
4304
4305 my $res;
4306
4307 my $timeout;
4308 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4309 $timeout = $cmd->{arguments}->{timeout};
4310 delete $cmd->{arguments}->{timeout};
4311 }
4312
4313 eval {
4314 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4315 my $sname = qmp_socket($vmid);
4316 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4317 my $qmpclient = PVE::QMPClient->new();
4318
4319 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4320 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4321 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4322 if scalar(%{$cmd->{arguments}});
4323 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4324 } else {
4325 die "unable to open monitor socket\n";
4326 }
4327 };
4328 if (my $err = $@) {
4329 syslog("err", "VM $vmid qmp command failed - $err");
4330 die $err;
4331 }
4332
4333 return $res;
4334 }
4335
4336 sub vm_human_monitor_command {
4337 my ($vmid, $cmdline) = @_;
4338
4339 my $res;
4340
4341 my $cmd = {
4342 execute => 'human-monitor-command',
4343 arguments => { 'command-line' => $cmdline},
4344 };
4345
4346 return vm_qmp_command($vmid, $cmd);
4347 }
4348
4349 sub vm_commandline {
4350 my ($storecfg, $vmid) = @_;
4351
4352 my $conf = load_config($vmid);
4353
4354 my $defaults = load_defaults();
4355
4356 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4357
4358 return join(' ', @$cmd);
4359 }
4360
4361 sub vm_reset {
4362 my ($vmid, $skiplock) = @_;
4363
4364 lock_config($vmid, sub {
4365
4366 my $conf = load_config($vmid);
4367
4368 check_lock($conf) if !$skiplock;
4369
4370 vm_mon_cmd($vmid, "system_reset");
4371 });
4372 }
4373
4374 sub get_vm_volumes {
4375 my ($conf) = @_;
4376
4377 my $vollist = [];
4378 foreach_volid($conf, sub {
4379 my ($volid, $is_cdrom) = @_;
4380
4381 return if $volid =~ m|^/|;
4382
4383 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4384 return if !$sid;
4385
4386 push @$vollist, $volid;
4387 });
4388
4389 return $vollist;
4390 }
4391
4392 sub vm_stop_cleanup {
4393 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4394
4395 eval {
4396 fairsched_rmnod($vmid); # try to destroy group
4397
4398 if (!$keepActive) {
4399 my $vollist = get_vm_volumes($conf);
4400 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4401 }
4402
4403 foreach my $ext (qw(mon qmp pid vnc qga)) {
4404 unlink "/var/run/qemu-server/${vmid}.$ext";
4405 }
4406
4407 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4408 };
4409 warn $@ if $@; # avoid errors - just warn
4410 }
4411
4412 # Note: use $nockeck to skip tests if VM configuration file exists.
4413 # We need that when migration VMs to other nodes (files already moved)
4414 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4415 sub vm_stop {
4416 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4417
4418 $force = 1 if !defined($force) && !$shutdown;
4419
4420 if ($migratedfrom){
4421 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4422 kill 15, $pid if $pid;
4423 my $conf = load_config($vmid, $migratedfrom);
4424 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4425 return;
4426 }
4427
4428 lock_config($vmid, sub {
4429
4430 my $pid = check_running($vmid, $nocheck);
4431 return if !$pid;
4432
4433 my $conf;
4434 if (!$nocheck) {
4435 $conf = load_config($vmid);
4436 check_lock($conf) if !$skiplock;
4437 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4438 my $opts = parse_startup($conf->{startup});
4439 $timeout = $opts->{down} if $opts->{down};
4440 }
4441 }
4442
4443 $timeout = 60 if !defined($timeout);
4444
4445 eval {
4446 if ($shutdown) {
4447 if (defined($conf) && $conf->{agent}) {
4448 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4449 } else {
4450 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4451 }
4452 } else {
4453 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4454 }
4455 };
4456 my $err = $@;
4457
4458 if (!$err) {
4459 my $count = 0;
4460 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4461 $count++;
4462 sleep 1;
4463 }
4464
4465 if ($count >= $timeout) {
4466 if ($force) {
4467 warn "VM still running - terminating now with SIGTERM\n";
4468 kill 15, $pid;
4469 } else {
4470 die "VM quit/powerdown failed - got timeout\n";
4471 }
4472 } else {
4473 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4474 return;
4475 }
4476 } else {
4477 if ($force) {
4478 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4479 kill 15, $pid;
4480 } else {
4481 die "VM quit/powerdown failed\n";
4482 }
4483 }
4484
4485 # wait again
4486 $timeout = 10;
4487
4488 my $count = 0;
4489 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4490 $count++;
4491 sleep 1;
4492 }
4493
4494 if ($count >= $timeout) {
4495 warn "VM still running - terminating now with SIGKILL\n";
4496 kill 9, $pid;
4497 sleep 1;
4498 }
4499
4500 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4501 });
4502 }
4503
4504 sub vm_suspend {
4505 my ($vmid, $skiplock) = @_;
4506
4507 lock_config($vmid, sub {
4508
4509 my $conf = load_config($vmid);
4510
4511 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4512
4513 vm_mon_cmd($vmid, "stop");
4514 });
4515 }
4516
4517 sub vm_resume {
4518 my ($vmid, $skiplock) = @_;
4519
4520 lock_config($vmid, sub {
4521
4522 my $conf = load_config($vmid);
4523
4524 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4525
4526 vm_mon_cmd($vmid, "cont");
4527 });
4528 }
4529
4530 sub vm_sendkey {
4531 my ($vmid, $skiplock, $key) = @_;
4532
4533 lock_config($vmid, sub {
4534
4535 my $conf = load_config($vmid);
4536
4537 # there is no qmp command, so we use the human monitor command
4538 vm_human_monitor_command($vmid, "sendkey $key");
4539 });
4540 }
4541
4542 sub vm_destroy {
4543 my ($storecfg, $vmid, $skiplock) = @_;
4544
4545 lock_config($vmid, sub {
4546
4547 my $conf = load_config($vmid);
4548
4549 check_lock($conf) if !$skiplock;
4550
4551 if (!check_running($vmid)) {
4552 fairsched_rmnod($vmid); # try to destroy group
4553 destroy_vm($storecfg, $vmid);
4554 } else {
4555 die "VM $vmid is running - destroy failed\n";
4556 }
4557 });
4558 }
4559
4560 # pci helpers
4561
4562 sub file_write {
4563 my ($filename, $buf) = @_;
4564
4565 my $fh = IO::File->new($filename, "w");
4566 return undef if !$fh;
4567
4568 my $res = print $fh $buf;
4569
4570 $fh->close();
4571
4572 return $res;
4573 }
4574
4575 sub pci_device_info {
4576 my ($name) = @_;
4577
4578 my $res;
4579
4580 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4581 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4582
4583 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4584 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4585
4586 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4587 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4588
4589 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4590 return undef if !defined($product) || $product !~ s/^0x//;
4591
4592 $res = {
4593 name => $name,
4594 vendor => $vendor,
4595 product => $product,
4596 domain => $domain,
4597 bus => $bus,
4598 slot => $slot,
4599 func => $func,
4600 irq => $irq,
4601 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4602 };
4603
4604 return $res;
4605 }
4606
4607 sub pci_dev_reset {
4608 my ($dev) = @_;
4609
4610 my $name = $dev->{name};
4611
4612 my $fn = "$pcisysfs/devices/$name/reset";
4613
4614 return file_write($fn, "1");
4615 }
4616
4617 sub pci_dev_bind_to_stub {
4618 my ($dev) = @_;
4619
4620 my $name = $dev->{name};
4621
4622 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4623 return 1 if -d $testdir;
4624
4625 my $data = "$dev->{vendor} $dev->{product}";
4626 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4627
4628 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4629 if (!file_write($fn, $name)) {
4630 return undef if -f $fn;
4631 }
4632
4633 $fn = "$pcisysfs/drivers/pci-stub/bind";
4634 if (! -d $testdir) {
4635 return undef if !file_write($fn, $name);
4636 }
4637
4638 return -d $testdir;
4639 }
4640
4641 sub pci_dev_bind_to_vfio {
4642 my ($dev) = @_;
4643
4644 my $name = $dev->{name};
4645
4646 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4647
4648 if (!-d $vfio_basedir) {
4649 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4650 }
4651 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4652
4653 my $testdir = "$vfio_basedir/$name";
4654 return 1 if -d $testdir;
4655
4656 my $data = "$dev->{vendor} $dev->{product}";
4657 return undef if !file_write("$vfio_basedir/new_id", $data);
4658
4659 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4660 if (!file_write($fn, $name)) {
4661 return undef if -f $fn;
4662 }
4663
4664 $fn = "$vfio_basedir/bind";
4665 if (! -d $testdir) {
4666 return undef if !file_write($fn, $name);
4667 }
4668
4669 return -d $testdir;
4670 }
4671
4672 sub pci_dev_group_bind_to_vfio {
4673 my ($pciid) = @_;
4674
4675 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4676
4677 if (!-d $vfio_basedir) {
4678 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4679 }
4680 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4681
4682 # get IOMMU group devices
4683 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4684 my @devs = grep /^0000:/, readdir($D);
4685 closedir($D);
4686
4687 foreach my $pciid (@devs) {
4688 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4689
4690 # pci bridges, switches or root ports are not supported
4691 # they have a pci_bus subdirectory so skip them
4692 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4693
4694 my $info = pci_device_info($1);
4695 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4696 }
4697
4698 return 1;
4699 }
4700
4701 sub print_pci_addr {
4702 my ($id, $bridges) = @_;
4703
4704 my $res = '';
4705 my $devices = {
4706 piix3 => { bus => 0, addr => 1 },
4707 #addr2 : first videocard
4708 balloon0 => { bus => 0, addr => 3 },
4709 watchdog => { bus => 0, addr => 4 },
4710 scsihw0 => { bus => 0, addr => 5 },
4711 'pci.3' => { bus => 0, addr => 5 }, #can also be used for virtio-scsi-single bridge
4712 scsihw1 => { bus => 0, addr => 6 },
4713 ahci0 => { bus => 0, addr => 7 },
4714 qga0 => { bus => 0, addr => 8 },
4715 spice => { bus => 0, addr => 9 },
4716 virtio0 => { bus => 0, addr => 10 },
4717 virtio1 => { bus => 0, addr => 11 },
4718 virtio2 => { bus => 0, addr => 12 },
4719 virtio3 => { bus => 0, addr => 13 },
4720 virtio4 => { bus => 0, addr => 14 },
4721 virtio5 => { bus => 0, addr => 15 },
4722 hostpci0 => { bus => 0, addr => 16 },
4723 hostpci1 => { bus => 0, addr => 17 },
4724 net0 => { bus => 0, addr => 18 },
4725 net1 => { bus => 0, addr => 19 },
4726 net2 => { bus => 0, addr => 20 },
4727 net3 => { bus => 0, addr => 21 },
4728 net4 => { bus => 0, addr => 22 },
4729 net5 => { bus => 0, addr => 23 },
4730 vga1 => { bus => 0, addr => 24 },
4731 vga2 => { bus => 0, addr => 25 },
4732 vga3 => { bus => 0, addr => 26 },
4733 hostpci2 => { bus => 0, addr => 27 },
4734 hostpci3 => { bus => 0, addr => 28 },
4735 #addr29 : usb-host (pve-usb.cfg)
4736 'pci.1' => { bus => 0, addr => 30 },
4737 'pci.2' => { bus => 0, addr => 31 },
4738 'net6' => { bus => 1, addr => 1 },
4739 'net7' => { bus => 1, addr => 2 },
4740 'net8' => { bus => 1, addr => 3 },
4741 'net9' => { bus => 1, addr => 4 },
4742 'net10' => { bus => 1, addr => 5 },
4743 'net11' => { bus => 1, addr => 6 },
4744 'net12' => { bus => 1, addr => 7 },
4745 'net13' => { bus => 1, addr => 8 },
4746 'net14' => { bus => 1, addr => 9 },
4747 'net15' => { bus => 1, addr => 10 },
4748 'net16' => { bus => 1, addr => 11 },
4749 'net17' => { bus => 1, addr => 12 },
4750 'net18' => { bus => 1, addr => 13 },
4751 'net19' => { bus => 1, addr => 14 },
4752 'net20' => { bus => 1, addr => 15 },
4753 'net21' => { bus => 1, addr => 16 },
4754 'net22' => { bus => 1, addr => 17 },
4755 'net23' => { bus => 1, addr => 18 },
4756 'net24' => { bus => 1, addr => 19 },
4757 'net25' => { bus => 1, addr => 20 },
4758 'net26' => { bus => 1, addr => 21 },
4759 'net27' => { bus => 1, addr => 22 },
4760 'net28' => { bus => 1, addr => 23 },
4761 'net29' => { bus => 1, addr => 24 },
4762 'net30' => { bus => 1, addr => 25 },
4763 'net31' => { bus => 1, addr => 26 },
4764 'virtio6' => { bus => 2, addr => 1 },
4765 'virtio7' => { bus => 2, addr => 2 },
4766 'virtio8' => { bus => 2, addr => 3 },
4767 'virtio9' => { bus => 2, addr => 4 },
4768 'virtio10' => { bus => 2, addr => 5 },
4769 'virtio11' => { bus => 2, addr => 6 },
4770 'virtio12' => { bus => 2, addr => 7 },
4771 'virtio13' => { bus => 2, addr => 8 },
4772 'virtio14' => { bus => 2, addr => 9 },
4773 'virtio15' => { bus => 2, addr => 10 },
4774 'virtioscsi0' => { bus => 3, addr => 1 },
4775 'virtioscsi1' => { bus => 3, addr => 2 },
4776 'virtioscsi2' => { bus => 3, addr => 3 },
4777 'virtioscsi3' => { bus => 3, addr => 4 },
4778 'virtioscsi4' => { bus => 3, addr => 5 },
4779 'virtioscsi5' => { bus => 3, addr => 6 },
4780 'virtioscsi6' => { bus => 3, addr => 7 },
4781 'virtioscsi7' => { bus => 3, addr => 8 },
4782 'virtioscsi8' => { bus => 3, addr => 9 },
4783 'virtioscsi9' => { bus => 3, addr => 10 },
4784 'virtioscsi10' => { bus => 3, addr => 11 },
4785 'virtioscsi11' => { bus => 3, addr => 12 },
4786 'virtioscsi12' => { bus => 3, addr => 13 },
4787 'virtioscsi13' => { bus => 3, addr => 14 },
4788 'virtioscsi14' => { bus => 3, addr => 15 },
4789 'virtioscsi15' => { bus => 3, addr => 16 },
4790 'virtioscsi16' => { bus => 3, addr => 17 },
4791 'virtioscsi17' => { bus => 3, addr => 18 },
4792 'virtioscsi18' => { bus => 3, addr => 19 },
4793 'virtioscsi19' => { bus => 3, addr => 20 },
4794 'virtioscsi20' => { bus => 3, addr => 21 },
4795 'virtioscsi21' => { bus => 3, addr => 22 },
4796 'virtioscsi22' => { bus => 3, addr => 23 },
4797 'virtioscsi23' => { bus => 3, addr => 24 },
4798 'virtioscsi24' => { bus => 3, addr => 25 },
4799 'virtioscsi25' => { bus => 3, addr => 26 },
4800 'virtioscsi26' => { bus => 3, addr => 27 },
4801 'virtioscsi27' => { bus => 3, addr => 28 },
4802 'virtioscsi28' => { bus => 3, addr => 29 },
4803 'virtioscsi29' => { bus => 3, addr => 30 },
4804 'virtioscsi30' => { bus => 3, addr => 31 },
4805
4806 };
4807
4808 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4809 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4810 my $bus = $devices->{$id}->{bus};
4811 $res = ",bus=pci.$bus,addr=$addr";
4812 $bridges->{$bus} = 1 if $bridges;
4813 }
4814 return $res;
4815
4816 }
4817
4818 sub print_pcie_addr {
4819 my ($id) = @_;
4820
4821 my $res = '';
4822 my $devices = {
4823 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4824 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4825 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4826 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4827 };
4828
4829 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4830 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4831 my $bus = $devices->{$id}->{bus};
4832 $res = ",bus=$bus,addr=$addr";
4833 }
4834 return $res;
4835
4836 }
4837
4838 # vzdump restore implementaion
4839
4840 sub tar_archive_read_firstfile {
4841 my $archive = shift;
4842
4843 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4844
4845 # try to detect archive type first
4846 my $pid = open (TMP, "tar tf '$archive'|") ||
4847 die "unable to open file '$archive'\n";
4848 my $firstfile = <TMP>;
4849 kill 15, $pid;
4850 close TMP;
4851
4852 die "ERROR: archive contaions no data\n" if !$firstfile;
4853 chomp $firstfile;
4854
4855 return $firstfile;
4856 }
4857
4858 sub tar_restore_cleanup {
4859 my ($storecfg, $statfile) = @_;
4860
4861 print STDERR "starting cleanup\n";
4862
4863 if (my $fd = IO::File->new($statfile, "r")) {
4864 while (defined(my $line = <$fd>)) {
4865 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4866 my $volid = $2;
4867 eval {
4868 if ($volid =~ m|^/|) {
4869 unlink $volid || die 'unlink failed\n';
4870 } else {
4871 PVE::Storage::vdisk_free($storecfg, $volid);
4872 }
4873 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4874 };
4875 print STDERR "unable to cleanup '$volid' - $@" if $@;
4876 } else {
4877 print STDERR "unable to parse line in statfile - $line";
4878 }
4879 }
4880 $fd->close();
4881 }
4882 }
4883
4884 sub restore_archive {
4885 my ($archive, $vmid, $user, $opts) = @_;
4886
4887 my $format = $opts->{format};
4888 my $comp;
4889
4890 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4891 $format = 'tar' if !$format;
4892 $comp = 'gzip';
4893 } elsif ($archive =~ m/\.tar$/) {
4894 $format = 'tar' if !$format;
4895 } elsif ($archive =~ m/.tar.lzo$/) {
4896 $format = 'tar' if !$format;
4897 $comp = 'lzop';
4898 } elsif ($archive =~ m/\.vma$/) {
4899 $format = 'vma' if !$format;
4900 } elsif ($archive =~ m/\.vma\.gz$/) {
4901 $format = 'vma' if !$format;
4902 $comp = 'gzip';
4903 } elsif ($archive =~ m/\.vma\.lzo$/) {
4904 $format = 'vma' if !$format;
4905 $comp = 'lzop';
4906 } else {
4907 $format = 'vma' if !$format; # default
4908 }
4909
4910 # try to detect archive format
4911 if ($format eq 'tar') {
4912 return restore_tar_archive($archive, $vmid, $user, $opts);
4913 } else {
4914 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4915 }
4916 }
4917
4918 sub restore_update_config_line {
4919 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4920
4921 return if $line =~ m/^\#qmdump\#/;
4922 return if $line =~ m/^\#vzdump\#/;
4923 return if $line =~ m/^lock:/;
4924 return if $line =~ m/^unused\d+:/;
4925 return if $line =~ m/^parent:/;
4926 return if $line =~ m/^template:/; # restored VM is never a template
4927
4928 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4929 # try to convert old 1.X settings
4930 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4931 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4932 my ($model, $macaddr) = split(/\=/, $devconfig);
4933 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4934 my $net = {
4935 model => $model,
4936 bridge => "vmbr$ind",
4937 macaddr => $macaddr,
4938 };
4939 my $netstr = print_net($net);
4940
4941 print $outfd "net$cookie->{netcount}: $netstr\n";
4942 $cookie->{netcount}++;
4943 }
4944 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4945 my ($id, $netstr) = ($1, $2);
4946 my $net = parse_net($netstr);
4947 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4948 $netstr = print_net($net);
4949 print $outfd "$id: $netstr\n";
4950 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4951 my $virtdev = $1;
4952 my $value = $3;
4953 if ($line =~ m/backup=no/) {
4954 print $outfd "#$line";
4955 } elsif ($virtdev && $map->{$virtdev}) {
4956 my $di = parse_drive($virtdev, $value);
4957 delete $di->{format}; # format can change on restore
4958 $di->{file} = $map->{$virtdev};
4959 $value = print_drive($vmid, $di);
4960 print $outfd "$virtdev: $value\n";
4961 } else {
4962 print $outfd $line;
4963 }
4964 } else {
4965 print $outfd $line;
4966 }
4967 }
4968
4969 sub scan_volids {
4970 my ($cfg, $vmid) = @_;
4971
4972 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4973
4974 my $volid_hash = {};
4975 foreach my $storeid (keys %$info) {
4976 foreach my $item (@{$info->{$storeid}}) {
4977 next if !($item->{volid} && $item->{size});
4978 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4979 $volid_hash->{$item->{volid}} = $item;
4980 }
4981 }
4982
4983 return $volid_hash;
4984 }
4985
4986 sub get_used_paths {
4987 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4988
4989 my $used_path = {};
4990
4991 my $scan_config = sub {
4992 my ($cref, $snapname) = @_;
4993
4994 foreach my $key (keys %$cref) {
4995 my $value = $cref->{$key};
4996 if (valid_drivename($key)) {
4997 next if $skip_drive && $key eq $skip_drive;
4998 my $drive = parse_drive($key, $value);
4999 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
5000 if ($drive->{file} =~ m!^/!) {
5001 $used_path->{$drive->{file}}++; # = 1;
5002 } else {
5003 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
5004 next if !$storeid;
5005 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
5006 next if !$scfg;
5007 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
5008 $used_path->{$path}++; # = 1;
5009 }
5010 }
5011 }
5012 };
5013
5014 &$scan_config($conf);
5015
5016 undef $skip_drive;
5017
5018 if ($scan_snapshots) {
5019 foreach my $snapname (keys %{$conf->{snapshots}}) {
5020 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
5021 }
5022 }
5023
5024 return $used_path;
5025 }
5026
5027 sub update_disksize {
5028 my ($vmid, $conf, $volid_hash) = @_;
5029
5030 my $changes;
5031
5032 my $used = {};
5033
5034 # Note: it is allowed to define multiple storages with same path (alias), so
5035 # we need to check both 'volid' and real 'path' (two different volid can point
5036 # to the same path).
5037
5038 my $usedpath = {};
5039
5040 # update size info
5041 foreach my $opt (keys %$conf) {
5042 if (valid_drivename($opt)) {
5043 my $drive = parse_drive($opt, $conf->{$opt});
5044 my $volid = $drive->{file};
5045 next if !$volid;
5046
5047 $used->{$volid} = 1;
5048 if ($volid_hash->{$volid} &&
5049 (my $path = $volid_hash->{$volid}->{path})) {
5050 $usedpath->{$path} = 1;
5051 }
5052
5053 next if drive_is_cdrom($drive);
5054 next if !$volid_hash->{$volid};
5055
5056 $drive->{size} = $volid_hash->{$volid}->{size};
5057 my $new = print_drive($vmid, $drive);
5058 if ($new ne $conf->{$opt}) {
5059 $changes = 1;
5060 $conf->{$opt} = $new;
5061 }
5062 }
5063 }
5064
5065 # remove 'unusedX' entry if volume is used
5066 foreach my $opt (keys %$conf) {
5067 next if $opt !~ m/^unused\d+$/;
5068 my $volid = $conf->{$opt};
5069 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5070 if ($used->{$volid} || ($path && $usedpath->{$path})) {
5071 $changes = 1;
5072 delete $conf->{$opt};
5073 }
5074 }
5075
5076 foreach my $volid (sort keys %$volid_hash) {
5077 next if $volid =~ m/vm-$vmid-state-/;
5078 next if $used->{$volid};
5079 my $path = $volid_hash->{$volid}->{path};
5080 next if !$path; # just to be sure
5081 next if $usedpath->{$path};
5082 $changes = 1;
5083 add_unused_volume($conf, $volid);
5084 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
5085 }
5086
5087 return $changes;
5088 }
5089
5090 sub rescan {
5091 my ($vmid, $nolock) = @_;
5092
5093 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
5094
5095 my $volid_hash = scan_volids($cfg, $vmid);
5096
5097 my $updatefn = sub {
5098 my ($vmid) = @_;
5099
5100 my $conf = load_config($vmid);
5101
5102 check_lock($conf);
5103
5104 my $vm_volids = {};
5105 foreach my $volid (keys %$volid_hash) {
5106 my $info = $volid_hash->{$volid};
5107 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5108 }
5109
5110 my $changes = update_disksize($vmid, $conf, $vm_volids);
5111
5112 update_config_nolock($vmid, $conf, 1) if $changes;
5113 };
5114
5115 if (defined($vmid)) {
5116 if ($nolock) {
5117 &$updatefn($vmid);
5118 } else {
5119 lock_config($vmid, $updatefn, $vmid);
5120 }
5121 } else {
5122 my $vmlist = config_list();
5123 foreach my $vmid (keys %$vmlist) {
5124 if ($nolock) {
5125 &$updatefn($vmid);
5126 } else {
5127 lock_config($vmid, $updatefn, $vmid);
5128 }
5129 }
5130 }
5131 }
5132
5133 sub restore_vma_archive {
5134 my ($archive, $vmid, $user, $opts, $comp) = @_;
5135
5136 my $input = $archive eq '-' ? "<&STDIN" : undef;
5137 my $readfrom = $archive;
5138
5139 my $uncomp = '';
5140 if ($comp) {
5141 $readfrom = '-';
5142 my $qarchive = PVE::Tools::shellquote($archive);
5143 if ($comp eq 'gzip') {
5144 $uncomp = "zcat $qarchive|";
5145 } elsif ($comp eq 'lzop') {
5146 $uncomp = "lzop -d -c $qarchive|";
5147 } else {
5148 die "unknown compression method '$comp'\n";
5149 }
5150
5151 }
5152
5153 my $tmpdir = "/var/tmp/vzdumptmp$$";
5154 rmtree $tmpdir;
5155
5156 # disable interrupts (always do cleanups)
5157 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5158 warn "got interrupt - ignored\n";
5159 };
5160
5161 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5162 POSIX::mkfifo($mapfifo, 0600);
5163 my $fifofh;
5164
5165 my $openfifo = sub {
5166 open($fifofh, '>', $mapfifo) || die $!;
5167 };
5168
5169 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5170
5171 my $oldtimeout;
5172 my $timeout = 5;
5173
5174 my $devinfo = {};
5175
5176 my $rpcenv = PVE::RPCEnvironment::get();
5177
5178 my $conffile = config_file($vmid);
5179 my $tmpfn = "$conffile.$$.tmp";
5180
5181 # Note: $oldconf is undef if VM does not exists
5182 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5183
5184 my $print_devmap = sub {
5185 my $virtdev_hash = {};
5186
5187 my $cfgfn = "$tmpdir/qemu-server.conf";
5188
5189 # we can read the config - that is already extracted
5190 my $fh = IO::File->new($cfgfn, "r") ||
5191 "unable to read qemu-server.conf - $!\n";
5192
5193 while (defined(my $line = <$fh>)) {
5194 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5195 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5196 die "archive does not contain data for drive '$virtdev'\n"
5197 if !$devinfo->{$devname};
5198 if (defined($opts->{storage})) {
5199 $storeid = $opts->{storage} || 'local';
5200 } elsif (!$storeid) {
5201 $storeid = 'local';
5202 }
5203 $format = 'raw' if !$format;
5204 $devinfo->{$devname}->{devname} = $devname;
5205 $devinfo->{$devname}->{virtdev} = $virtdev;
5206 $devinfo->{$devname}->{format} = $format;
5207 $devinfo->{$devname}->{storeid} = $storeid;
5208
5209 # check permission on storage
5210 my $pool = $opts->{pool}; # todo: do we need that?
5211 if ($user ne 'root@pam') {
5212 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5213 }
5214
5215 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5216 }
5217 }
5218
5219 foreach my $devname (keys %$devinfo) {
5220 die "found no device mapping information for device '$devname'\n"
5221 if !$devinfo->{$devname}->{virtdev};
5222 }
5223
5224 my $cfg = cfs_read_file('storage.cfg');
5225
5226 # create empty/temp config
5227 if ($oldconf) {
5228 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5229 foreach_drive($oldconf, sub {
5230 my ($ds, $drive) = @_;
5231
5232 return if drive_is_cdrom($drive);
5233
5234 my $volid = $drive->{file};
5235
5236 return if !$volid || $volid =~ m|^/|;
5237
5238 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5239 return if !$path || !$owner || ($owner != $vmid);
5240
5241 # Note: only delete disk we want to restore
5242 # other volumes will become unused
5243 if ($virtdev_hash->{$ds}) {
5244 PVE::Storage::vdisk_free($cfg, $volid);
5245 }
5246 });
5247 }
5248
5249 my $map = {};
5250 foreach my $virtdev (sort keys %$virtdev_hash) {
5251 my $d = $virtdev_hash->{$virtdev};
5252 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5253 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5254
5255 # test if requested format is supported
5256 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5257 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5258 $d->{format} = $defFormat if !$supported;
5259
5260 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5261 $d->{format}, undef, $alloc_size);
5262 print STDERR "new volume ID is '$volid'\n";
5263 $d->{volid} = $volid;
5264 my $path = PVE::Storage::path($cfg, $volid);
5265
5266 my $write_zeros = 1;
5267 # fixme: what other storages types initialize volumes with zero?
5268 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5269 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5270 $write_zeros = 0;
5271 }
5272
5273 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5274
5275 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5276 $map->{$virtdev} = $volid;
5277 }
5278
5279 $fh->seek(0, 0) || die "seek failed - $!\n";
5280
5281 my $outfd = new IO::File ($tmpfn, "w") ||
5282 die "unable to write config for VM $vmid\n";
5283
5284 my $cookie = { netcount => 0 };
5285 while (defined(my $line = <$fh>)) {
5286 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5287 }
5288
5289 $fh->close();
5290 $outfd->close();
5291 };
5292
5293 eval {
5294 # enable interrupts
5295 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5296 die "interrupted by signal\n";
5297 };
5298 local $SIG{ALRM} = sub { die "got timeout\n"; };
5299
5300 $oldtimeout = alarm($timeout);
5301
5302 my $parser = sub {
5303 my $line = shift;
5304
5305 print "$line\n";
5306
5307 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5308 my ($dev_id, $size, $devname) = ($1, $2, $3);
5309 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5310 } elsif ($line =~ m/^CTIME: /) {
5311 # we correctly received the vma config, so we can disable
5312 # the timeout now for disk allocation (set to 10 minutes, so
5313 # that we always timeout if something goes wrong)
5314 alarm(600);
5315 &$print_devmap();
5316 print $fifofh "done\n";
5317 my $tmp = $oldtimeout || 0;
5318 $oldtimeout = undef;
5319 alarm($tmp);
5320 close($fifofh);
5321 }
5322 };
5323
5324 print "restore vma archive: $cmd\n";
5325 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5326 };
5327 my $err = $@;
5328
5329 alarm($oldtimeout) if $oldtimeout;
5330
5331 unlink $mapfifo;
5332
5333 if ($err) {
5334 rmtree $tmpdir;
5335 unlink $tmpfn;
5336
5337 my $cfg = cfs_read_file('storage.cfg');
5338 foreach my $devname (keys %$devinfo) {
5339 my $volid = $devinfo->{$devname}->{volid};
5340 next if !$volid;
5341 eval {
5342 if ($volid =~ m|^/|) {
5343 unlink $volid || die 'unlink failed\n';
5344 } else {
5345 PVE::Storage::vdisk_free($cfg, $volid);
5346 }
5347 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5348 };
5349 print STDERR "unable to cleanup '$volid' - $@" if $@;
5350 }
5351 die $err;
5352 }
5353
5354 rmtree $tmpdir;
5355
5356 rename($tmpfn, $conffile) ||
5357 die "unable to commit configuration file '$conffile'\n";
5358
5359 PVE::Cluster::cfs_update(); # make sure we read new file
5360
5361 eval { rescan($vmid, 1); };
5362 warn $@ if $@;
5363 }
5364
5365 sub restore_tar_archive {
5366 my ($archive, $vmid, $user, $opts) = @_;
5367
5368 if ($archive ne '-') {
5369 my $firstfile = tar_archive_read_firstfile($archive);
5370 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5371 if $firstfile ne 'qemu-server.conf';
5372 }
5373
5374 my $storecfg = cfs_read_file('storage.cfg');
5375
5376 # destroy existing data - keep empty config
5377 my $vmcfgfn = config_file($vmid);
5378 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5379
5380 my $tocmd = "/usr/lib/qemu-server/qmextract";
5381
5382 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5383 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5384 $tocmd .= ' --prealloc' if $opts->{prealloc};
5385 $tocmd .= ' --info' if $opts->{info};
5386
5387 # tar option "xf" does not autodetect compression when read from STDIN,
5388 # so we pipe to zcat
5389 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5390 PVE::Tools::shellquote("--to-command=$tocmd");
5391
5392 my $tmpdir = "/var/tmp/vzdumptmp$$";
5393 mkpath $tmpdir;
5394
5395 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5396 local $ENV{VZDUMP_VMID} = $vmid;
5397 local $ENV{VZDUMP_USER} = $user;
5398
5399 my $conffile = config_file($vmid);
5400 my $tmpfn = "$conffile.$$.tmp";
5401
5402 # disable interrupts (always do cleanups)
5403 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5404 print STDERR "got interrupt - ignored\n";
5405 };
5406
5407 eval {
5408 # enable interrupts
5409 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5410 die "interrupted by signal\n";
5411 };
5412
5413 if ($archive eq '-') {
5414 print "extracting archive from STDIN\n";
5415 run_command($cmd, input => "<&STDIN");
5416 } else {
5417 print "extracting archive '$archive'\n";
5418 run_command($cmd);
5419 }
5420
5421 return if $opts->{info};
5422
5423 # read new mapping
5424 my $map = {};
5425 my $statfile = "$tmpdir/qmrestore.stat";
5426 if (my $fd = IO::File->new($statfile, "r")) {
5427 while (defined (my $line = <$fd>)) {
5428 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5429 $map->{$1} = $2 if $1;
5430 } else {
5431 print STDERR "unable to parse line in statfile - $line\n";
5432 }
5433 }
5434 $fd->close();
5435 }
5436
5437 my $confsrc = "$tmpdir/qemu-server.conf";
5438
5439 my $srcfd = new IO::File($confsrc, "r") ||
5440 die "unable to open file '$confsrc'\n";
5441
5442 my $outfd = new IO::File ($tmpfn, "w") ||
5443 die "unable to write config for VM $vmid\n";
5444
5445 my $cookie = { netcount => 0 };
5446 while (defined (my $line = <$srcfd>)) {
5447 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5448 }
5449
5450 $srcfd->close();
5451 $outfd->close();
5452 };
5453 my $err = $@;
5454
5455 if ($err) {
5456
5457 unlink $tmpfn;
5458
5459 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5460
5461 die $err;
5462 }
5463
5464 rmtree $tmpdir;
5465
5466 rename $tmpfn, $conffile ||
5467 die "unable to commit configuration file '$conffile'\n";
5468
5469 PVE::Cluster::cfs_update(); # make sure we read new file
5470
5471 eval { rescan($vmid, 1); };
5472 warn $@ if $@;
5473 };
5474
5475
5476 # Internal snapshots
5477
5478 # NOTE: Snapshot create/delete involves several non-atomic
5479 # action, and can take a long time.
5480 # So we try to avoid locking the file and use 'lock' variable
5481 # inside the config file instead.
5482
5483 my $snapshot_copy_config = sub {
5484 my ($source, $dest) = @_;
5485
5486 foreach my $k (keys %$source) {
5487 next if $k eq 'snapshots';
5488 next if $k eq 'snapstate';
5489 next if $k eq 'snaptime';
5490 next if $k eq 'vmstate';
5491 next if $k eq 'lock';
5492 next if $k eq 'digest';
5493 next if $k eq 'description';
5494 next if $k =~ m/^unused\d+$/;
5495
5496 $dest->{$k} = $source->{$k};
5497 }
5498 };
5499
5500 my $snapshot_apply_config = sub {
5501 my ($conf, $snap) = @_;
5502
5503 # copy snapshot list
5504 my $newconf = {
5505 snapshots => $conf->{snapshots},
5506 };
5507
5508 # keep description and list of unused disks
5509 foreach my $k (keys %$conf) {
5510 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5511 $newconf->{$k} = $conf->{$k};
5512 }
5513
5514 &$snapshot_copy_config($snap, $newconf);
5515
5516 return $newconf;
5517 };
5518
5519 sub foreach_writable_storage {
5520 my ($conf, $func) = @_;
5521
5522 my $sidhash = {};
5523
5524 foreach my $ds (keys %$conf) {
5525 next if !valid_drivename($ds);
5526
5527 my $drive = parse_drive($ds, $conf->{$ds});
5528 next if !$drive;
5529 next if drive_is_cdrom($drive);
5530
5531 my $volid = $drive->{file};
5532
5533 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5534 $sidhash->{$sid} = $sid if $sid;
5535 }
5536
5537 foreach my $sid (sort keys %$sidhash) {
5538 &$func($sid);
5539 }
5540 }
5541
5542 my $alloc_vmstate_volid = sub {
5543 my ($storecfg, $vmid, $conf, $snapname) = @_;
5544
5545 # Note: we try to be smart when selecting a $target storage
5546
5547 my $target;
5548
5549 # search shared storage first
5550 foreach_writable_storage($conf, sub {
5551 my ($sid) = @_;
5552 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5553 return if !$scfg->{shared};
5554
5555 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5556 });
5557
5558 if (!$target) {
5559 # now search local storage
5560 foreach_writable_storage($conf, sub {
5561 my ($sid) = @_;
5562 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5563 return if $scfg->{shared};
5564
5565 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5566 });
5567 }
5568
5569 $target = 'local' if !$target;
5570
5571 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5572 # we abort live save after $conf->{memory}, so we need at max twice that space
5573 my $size = $conf->{memory}*2 + $driver_state_size;
5574
5575 my $name = "vm-$vmid-state-$snapname";
5576 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5577 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5578 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5579
5580 return $volid;
5581 };
5582
5583 my $snapshot_prepare = sub {
5584 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5585
5586 my $snap;
5587
5588 my $updatefn = sub {
5589
5590 my $conf = load_config($vmid);
5591
5592 die "you can't take a snapshot if it's a template\n"
5593 if is_template($conf);
5594
5595 check_lock($conf);
5596
5597 $conf->{lock} = 'snapshot';
5598
5599 die "snapshot name '$snapname' already used\n"
5600 if defined($conf->{snapshots}->{$snapname});
5601
5602 my $storecfg = PVE::Storage::config();
5603 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5604
5605 $snap = $conf->{snapshots}->{$snapname} = {};
5606
5607 if ($save_vmstate && check_running($vmid)) {
5608 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5609 }
5610
5611 &$snapshot_copy_config($conf, $snap);
5612
5613 $snap->{snapstate} = "prepare";
5614 $snap->{snaptime} = time();
5615 $snap->{description} = $comment if $comment;
5616
5617 # always overwrite machine if we save vmstate. This makes sure we
5618 # can restore it later using correct machine type
5619 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5620
5621 update_config_nolock($vmid, $conf, 1);
5622 };
5623
5624 lock_config($vmid, $updatefn);
5625
5626 return $snap;
5627 };
5628
5629 my $snapshot_commit = sub {
5630 my ($vmid, $snapname) = @_;
5631
5632 my $updatefn = sub {
5633
5634 my $conf = load_config($vmid);
5635
5636 die "missing snapshot lock\n"
5637 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5638
5639 my $has_machine_config = defined($conf->{machine});
5640
5641 my $snap = $conf->{snapshots}->{$snapname};
5642
5643 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5644
5645 die "wrong snapshot state\n"
5646 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5647
5648 delete $snap->{snapstate};
5649 delete $conf->{lock};
5650
5651 my $newconf = &$snapshot_apply_config($conf, $snap);
5652
5653 delete $newconf->{machine} if !$has_machine_config;
5654
5655 $newconf->{parent} = $snapname;
5656
5657 update_config_nolock($vmid, $newconf, 1);
5658 };
5659
5660 lock_config($vmid, $updatefn);
5661 };
5662
5663 sub snapshot_rollback {
5664 my ($vmid, $snapname) = @_;
5665
5666 my $prepare = 1;
5667
5668 my $storecfg = PVE::Storage::config();
5669
5670 my $conf = load_config($vmid);
5671
5672 my $get_snapshot_config = sub {
5673
5674 die "you can't rollback if vm is a template\n" if is_template($conf);
5675
5676 my $res = $conf->{snapshots}->{$snapname};
5677
5678 die "snapshot '$snapname' does not exist\n" if !defined($res);
5679
5680 return $res;
5681 };
5682
5683 my $snap = &$get_snapshot_config();
5684
5685 foreach_drive($snap, sub {
5686 my ($ds, $drive) = @_;
5687
5688 return if drive_is_cdrom($drive);
5689
5690 my $volid = $drive->{file};
5691
5692 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5693 });
5694
5695 my $updatefn = sub {
5696
5697 $conf = load_config($vmid);
5698
5699 $snap = &$get_snapshot_config();
5700
5701 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5702 if $snap->{snapstate};
5703
5704 if ($prepare) {
5705 check_lock($conf);
5706 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5707 }
5708
5709 die "unable to rollback vm $vmid: vm is running\n"
5710 if check_running($vmid);
5711
5712 if ($prepare) {
5713 $conf->{lock} = 'rollback';
5714 } else {
5715 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5716 delete $conf->{lock};
5717 }
5718
5719 my $forcemachine;
5720
5721 if (!$prepare) {
5722 my $has_machine_config = defined($conf->{machine});
5723
5724 # copy snapshot config to current config
5725 $conf = &$snapshot_apply_config($conf, $snap);
5726 $conf->{parent} = $snapname;
5727
5728 # Note: old code did not store 'machine', so we try to be smart
5729 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5730 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5731 # we remove the 'machine' configuration if not explicitly specified
5732 # in the original config.
5733 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5734 }
5735
5736 update_config_nolock($vmid, $conf, 1);
5737
5738 if (!$prepare && $snap->{vmstate}) {
5739 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5740 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5741 }
5742 };
5743
5744 lock_config($vmid, $updatefn);
5745
5746 foreach_drive($snap, sub {
5747 my ($ds, $drive) = @_;
5748
5749 return if drive_is_cdrom($drive);
5750
5751 my $volid = $drive->{file};
5752 my $device = "drive-$ds";
5753
5754 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5755 });
5756
5757 $prepare = 0;
5758 lock_config($vmid, $updatefn);
5759 }
5760
5761 my $savevm_wait = sub {
5762 my ($vmid) = @_;
5763
5764 for(;;) {
5765 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5766 if (!$stat->{status}) {
5767 die "savevm not active\n";
5768 } elsif ($stat->{status} eq 'active') {
5769 sleep(1);
5770 next;
5771 } elsif ($stat->{status} eq 'completed') {
5772 last;
5773 } else {
5774 die "query-savevm returned status '$stat->{status}'\n";
5775 }
5776 }
5777 };
5778
5779 sub snapshot_create {
5780 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5781
5782 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5783
5784 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5785
5786 my $config = load_config($vmid);
5787
5788 my $running = check_running($vmid);
5789
5790 my $freezefs = $running && $config->{agent};
5791 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5792
5793 my $drivehash = {};
5794
5795 if ($freezefs) {
5796 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5797 warn "guest-fsfreeze-freeze problems - $@" if $@;
5798 }
5799
5800 eval {
5801 # create internal snapshots of all drives
5802
5803 my $storecfg = PVE::Storage::config();
5804
5805 if ($running) {
5806 if ($snap->{vmstate}) {
5807 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5808 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5809 &$savevm_wait($vmid);
5810 } else {
5811 vm_mon_cmd($vmid, "savevm-start");
5812 }
5813 };
5814
5815 foreach_drive($snap, sub {
5816 my ($ds, $drive) = @_;
5817
5818 return if drive_is_cdrom($drive);
5819
5820 my $volid = $drive->{file};
5821 my $device = "drive-$ds";
5822
5823 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5824 $drivehash->{$ds} = 1;
5825 });
5826 };
5827 my $err = $@;
5828
5829 if ($running) {
5830 eval { vm_mon_cmd($vmid, "savevm-end") };
5831 warn $@ if $@;
5832
5833 if ($freezefs) {
5834 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5835 warn "guest-fsfreeze-thaw problems - $@" if $@;
5836 }
5837
5838 # savevm-end is async, we need to wait
5839 for (;;) {
5840 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5841 if (!$stat->{bytes}) {
5842 last;
5843 } else {
5844 print "savevm not yet finished\n";
5845 sleep(1);
5846 next;
5847 }
5848 }
5849 }
5850
5851 if ($err) {
5852 warn "snapshot create failed: starting cleanup\n";
5853 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5854 warn $@ if $@;
5855 die $err;
5856 }
5857
5858 &$snapshot_commit($vmid, $snapname);
5859 }
5860
5861 # Note: $drivehash is only set when called from snapshot_create.
5862 sub snapshot_delete {
5863 my ($vmid, $snapname, $force, $drivehash) = @_;
5864
5865 my $prepare = 1;
5866
5867 my $snap;
5868 my $unused = [];
5869
5870 my $unlink_parent = sub {
5871 my ($confref, $new_parent) = @_;
5872
5873 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5874 if ($new_parent) {
5875 $confref->{parent} = $new_parent;
5876 } else {
5877 delete $confref->{parent};
5878 }
5879 }
5880 };
5881
5882 my $updatefn = sub {
5883 my ($remove_drive) = @_;
5884
5885 my $conf = load_config($vmid);
5886
5887 if (!$drivehash) {
5888 check_lock($conf);
5889 die "you can't delete a snapshot if vm is a template\n"
5890 if is_template($conf);
5891 }
5892
5893 $snap = $conf->{snapshots}->{$snapname};
5894
5895 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5896
5897 # remove parent refs
5898 if (!$prepare) {
5899 &$unlink_parent($conf, $snap->{parent});
5900 foreach my $sn (keys %{$conf->{snapshots}}) {
5901 next if $sn eq $snapname;
5902 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5903 }
5904 }
5905
5906 if ($remove_drive) {
5907 if ($remove_drive eq 'vmstate') {
5908 delete $snap->{$remove_drive};
5909 } else {
5910 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5911 my $volid = $drive->{file};
5912 delete $snap->{$remove_drive};
5913 add_unused_volume($conf, $volid);
5914 }
5915 }
5916
5917 if ($prepare) {
5918 $snap->{snapstate} = 'delete';
5919 } else {
5920 delete $conf->{snapshots}->{$snapname};
5921 delete $conf->{lock} if $drivehash;
5922 foreach my $volid (@$unused) {
5923 add_unused_volume($conf, $volid);
5924 }
5925 }
5926
5927 update_config_nolock($vmid, $conf, 1);
5928 };
5929
5930 lock_config($vmid, $updatefn);
5931
5932 # now remove vmstate file
5933
5934 my $storecfg = PVE::Storage::config();
5935
5936 if ($snap->{vmstate}) {
5937 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5938 if (my $err = $@) {
5939 die $err if !$force;
5940 warn $err;
5941 }
5942 # save changes (remove vmstate from snapshot)
5943 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5944 };
5945
5946 # now remove all internal snapshots
5947 foreach_drive($snap, sub {
5948 my ($ds, $drive) = @_;
5949
5950 return if drive_is_cdrom($drive);
5951
5952 my $volid = $drive->{file};
5953 my $device = "drive-$ds";
5954
5955 if (!$drivehash || $drivehash->{$ds}) {
5956 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5957 if (my $err = $@) {
5958 die $err if !$force;
5959 warn $err;
5960 }
5961 }
5962
5963 # save changes (remove drive fron snapshot)
5964 lock_config($vmid, $updatefn, $ds) if !$force;
5965 push @$unused, $volid;
5966 });
5967
5968 # now cleanup config
5969 $prepare = 0;
5970 lock_config($vmid, $updatefn);
5971 }
5972
5973 sub has_feature {
5974 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5975
5976 my $err;
5977 foreach_drive($conf, sub {
5978 my ($ds, $drive) = @_;
5979
5980 return if drive_is_cdrom($drive);
5981 my $volid = $drive->{file};
5982 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5983 });
5984
5985 return $err ? 0 : 1;
5986 }
5987
5988 sub template_create {
5989 my ($vmid, $conf, $disk) = @_;
5990
5991 my $storecfg = PVE::Storage::config();
5992
5993 foreach_drive($conf, sub {
5994 my ($ds, $drive) = @_;
5995
5996 return if drive_is_cdrom($drive);
5997 return if $disk && $ds ne $disk;
5998
5999 my $volid = $drive->{file};
6000 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
6001
6002 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
6003 $drive->{file} = $voliddst;
6004 $conf->{$ds} = print_drive($vmid, $drive);
6005 update_config_nolock($vmid, $conf, 1);
6006 });
6007 }
6008
6009 sub is_template {
6010 my ($conf) = @_;
6011
6012 return 1 if defined $conf->{template} && $conf->{template} == 1;
6013 }
6014
6015 sub qemu_img_convert {
6016 my ($src_volid, $dst_volid, $size, $snapname) = @_;
6017
6018 my $storecfg = PVE::Storage::config();
6019 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6020 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6021
6022 if ($src_storeid && $dst_storeid) {
6023 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6024 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6025
6026 my $src_format = qemu_img_format($src_scfg, $src_volname);
6027 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6028
6029 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6030 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6031
6032 my $cmd = [];
6033 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
6034 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
6035 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
6036
6037 my $parser = sub {
6038 my $line = shift;
6039 if($line =~ m/\((\S+)\/100\%\)/){
6040 my $percent = $1;
6041 my $transferred = int($size * $percent / 100);
6042 my $remaining = $size - $transferred;
6043
6044 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6045 }
6046
6047 };
6048
6049 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6050 my $err = $@;
6051 die "copy failed: $err" if $err;
6052 }
6053 }
6054
6055 sub qemu_img_format {
6056 my ($scfg, $volname) = @_;
6057
6058 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
6059 return $1;
6060 } elsif ($scfg->{type} eq 'iscsi') {
6061 return "host_device";
6062 } else {
6063 return "raw";
6064 }
6065 }
6066
6067 sub qemu_drive_mirror {
6068 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
6069
6070 my $count = 0;
6071 my $old_len = 0;
6072 my $frozen = undef;
6073 my $maxwait = 120;
6074
6075 my $storecfg = PVE::Storage::config();
6076 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6077
6078 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6079
6080 my $format;
6081 if ($dst_volname =~ m/\.(raw|qcow2)$/){
6082 $format = $1;
6083 }
6084
6085 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6086
6087 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
6088 $opts->{format} = $format if $format;
6089
6090 #fixme : sometime drive-mirror timeout, but works fine after.
6091 # (I have see the problem with big volume > 200GB), so we need to eval
6092 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
6093 # ignore errors here
6094
6095 eval {
6096 while (1) {
6097 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6098 my $stat = @$stats[0];
6099 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
6100 die "error job is not mirroring" if $stat->{type} ne "mirror";
6101
6102 my $busy = $stat->{busy};
6103
6104 if (my $total = $stat->{len}) {
6105 my $transferred = $stat->{offset} || 0;
6106 my $remaining = $total - $transferred;
6107 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6108
6109 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
6110 }
6111
6112 if ($stat->{len} == $stat->{offset}) {
6113 if ($busy eq 'false') {
6114
6115 last if $vmiddst != $vmid;
6116
6117 # try to switch the disk if source and destination are on the same guest
6118 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6119 last if !$@;
6120 die $@ if $@ !~ m/cannot be completed/;
6121 }
6122
6123 if ($count > $maxwait) {
6124 # if too much writes to disk occurs at the end of migration
6125 #the disk needs to be freezed to be able to complete the migration
6126 vm_suspend($vmid,1);
6127 $frozen = 1;
6128 }
6129 $count ++
6130 }
6131 $old_len = $stat->{offset};
6132 sleep 1;
6133 }
6134
6135 vm_resume($vmid, 1) if $frozen;
6136
6137 };
6138 my $err = $@;
6139
6140 my $cancel_job = sub {
6141 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6142 while (1) {
6143 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6144 my $stat = @$stats[0];
6145 last if !$stat;
6146 sleep 1;
6147 }
6148 };
6149
6150 if ($err) {
6151 eval { &$cancel_job(); };
6152 die "mirroring error: $err";
6153 }
6154
6155 if ($vmiddst != $vmid) {
6156 # if we clone a disk for a new target vm, we don't switch the disk
6157 &$cancel_job(); # so we call block-job-cancel
6158 }
6159 }
6160
6161 sub clone_disk {
6162 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6163 $newvmid, $storage, $format, $full, $newvollist) = @_;
6164
6165 my $newvolid;
6166
6167 if (!$full) {
6168 print "create linked clone of drive $drivename ($drive->{file})\n";
6169 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6170 push @$newvollist, $newvolid;
6171 } else {
6172 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6173 $storeid = $storage if $storage;
6174
6175 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6176 if (!$format) {
6177 $format = $drive->{format} || $defFormat;
6178 }
6179
6180 # test if requested format is supported - else use default
6181 my $supported = grep { $_ eq $format } @$validFormats;
6182 $format = $defFormat if !$supported;
6183
6184 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6185
6186 print "create full clone of drive $drivename ($drive->{file})\n";
6187 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6188 push @$newvollist, $newvolid;
6189
6190 if (!$running || $snapname) {
6191 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6192 } else {
6193 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6194 }
6195 }
6196
6197 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6198
6199 my $disk = $drive;
6200 $disk->{format} = undef;
6201 $disk->{file} = $newvolid;
6202 $disk->{size} = $size;
6203
6204 return $disk;
6205 }
6206
6207 # this only works if VM is running
6208 sub get_current_qemu_machine {
6209 my ($vmid) = @_;
6210
6211 my $cmd = { execute => 'query-machines', arguments => {} };
6212 my $res = vm_qmp_command($vmid, $cmd);
6213
6214 my ($current, $default);
6215 foreach my $e (@$res) {
6216 $default = $e->{name} if $e->{'is-default'};
6217 $current = $e->{name} if $e->{'is-current'};
6218 }
6219
6220 # fallback to the default machine if current is not supported by qemu
6221 return $current || $default || 'pc';
6222 }
6223
6224 sub qemu_machine_feature_enabled {
6225 my ($machine, $kvmver, $version_major, $version_minor) = @_;
6226
6227 my $current_major;
6228 my $current_minor;
6229
6230 if ($machine && $machine =~ m/^(pc(-i440fx|-q35)?-(\d+)\.(\d+))/) {
6231
6232 $current_major = $3;
6233 $current_minor = $4;
6234
6235 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6236
6237 $current_major = $1;
6238 $current_minor = $2;
6239 }
6240
6241 return 1 if $current_major >= $version_major && $current_minor >= $version_minor;
6242
6243
6244 }
6245
6246 sub lspci {
6247
6248 my $devices = {};
6249
6250 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6251 my (undef, $id, $function) = @_;
6252 my $res = { id => $id, function => $function};
6253 push @{$devices->{$id}}, $res;
6254 });
6255
6256 return $devices;
6257 }
6258
6259 sub vm_iothreads_list {
6260 my ($vmid) = @_;
6261
6262 my $res = vm_mon_cmd($vmid, 'query-iothreads');
6263
6264 my $iothreads = {};
6265 foreach my $iothread (@$res) {
6266 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6267 }
6268
6269 return $iothreads;
6270 }
6271
6272 sub scsihw_infos {
6273 my ($conf, $drive) = @_;
6274
6275 my $maxdev = 0;
6276
6277 if ($conf->{scsihw} && ($conf->{scsihw} =~ m/^lsi/)) {
6278 $maxdev = 7;
6279 } elsif ($conf->{scsihw} && ($conf->{scsihw} =~ m/^virtio-scsi-single/)) {
6280 $maxdev = 1;
6281 } else {
6282 $maxdev = 256;
6283 }
6284
6285 my $controller = int($drive->{index} / $maxdev);
6286 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} =~ m/^virtio-scsi-single/) ? "virtioscsi" : "scsihw";
6287
6288 return ($maxdev, $controller, $controller_prefix);
6289 }
6290 1;