]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
memory hotplug patch v10
[qemu-server.git] / PVE / QemuServer.pm
1 package PVE::QemuServer;
2
3 use strict;
4 use warnings;
5 use POSIX;
6 use IO::Handle;
7 use IO::Select;
8 use IO::File;
9 use IO::Dir;
10 use IO::Socket::UNIX;
11 use File::Basename;
12 use File::Path;
13 use File::stat;
14 use Getopt::Long;
15 use Digest::SHA;
16 use Fcntl ':flock';
17 use Cwd 'abs_path';
18 use IPC::Open3;
19 use JSON;
20 use Fcntl;
21 use PVE::SafeSyslog;
22 use Storable qw(dclone);
23 use PVE::Exception qw(raise raise_param_exc);
24 use PVE::Storage;
25 use PVE::Tools qw(run_command lock_file lock_file_full file_read_firstline dir_glob_foreach);
26 use PVE::JSONSchema qw(get_standard_option);
27 use PVE::Cluster qw(cfs_register_file cfs_read_file cfs_write_file cfs_lock_file);
28 use PVE::INotify;
29 use PVE::ProcFSTools;
30 use PVE::QMPClient;
31 use PVE::RPCEnvironment;
32 use Time::HiRes qw(gettimeofday);
33
34 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
35
36 # Note about locking: we use flock on the config file protect
37 # against concurent actions.
38 # Aditionaly, we have a 'lock' setting in the config file. This
39 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
40 # allowed when such lock is set. But you can ignore this kind of
41 # lock with the --skiplock flag.
42
43 cfs_register_file('/qemu-server/',
44 \&parse_vm_config,
45 \&write_vm_config);
46
47 PVE::JSONSchema::register_standard_option('skiplock', {
48 description => "Ignore locks - only root is allowed to use this option.",
49 type => 'boolean',
50 optional => 1,
51 });
52
53 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
54 description => "Some command save/restore state from this location.",
55 type => 'string',
56 maxLength => 128,
57 optional => 1,
58 });
59
60 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
61 description => "The name of the snapshot.",
62 type => 'string', format => 'pve-configid',
63 maxLength => 40,
64 });
65
66 #no warnings 'redefine';
67
68 unless(defined(&_VZSYSCALLS_H_)) {
69 eval 'sub _VZSYSCALLS_H_ () {1;}' unless defined(&_VZSYSCALLS_H_);
70 require 'sys/syscall.ph';
71 if(defined(&__x86_64__)) {
72 eval 'sub __NR_fairsched_vcpus () {499;}' unless defined(&__NR_fairsched_vcpus);
73 eval 'sub __NR_fairsched_mknod () {504;}' unless defined(&__NR_fairsched_mknod);
74 eval 'sub __NR_fairsched_rmnod () {505;}' unless defined(&__NR_fairsched_rmnod);
75 eval 'sub __NR_fairsched_chwt () {506;}' unless defined(&__NR_fairsched_chwt);
76 eval 'sub __NR_fairsched_mvpr () {507;}' unless defined(&__NR_fairsched_mvpr);
77 eval 'sub __NR_fairsched_rate () {508;}' unless defined(&__NR_fairsched_rate);
78 eval 'sub __NR_setluid () {501;}' unless defined(&__NR_setluid);
79 eval 'sub __NR_setublimit () {502;}' unless defined(&__NR_setublimit);
80 }
81 elsif(defined( &__i386__) ) {
82 eval 'sub __NR_fairsched_mknod () {500;}' unless defined(&__NR_fairsched_mknod);
83 eval 'sub __NR_fairsched_rmnod () {501;}' unless defined(&__NR_fairsched_rmnod);
84 eval 'sub __NR_fairsched_chwt () {502;}' unless defined(&__NR_fairsched_chwt);
85 eval 'sub __NR_fairsched_mvpr () {503;}' unless defined(&__NR_fairsched_mvpr);
86 eval 'sub __NR_fairsched_rate () {504;}' unless defined(&__NR_fairsched_rate);
87 eval 'sub __NR_fairsched_vcpus () {505;}' unless defined(&__NR_fairsched_vcpus);
88 eval 'sub __NR_setluid () {511;}' unless defined(&__NR_setluid);
89 eval 'sub __NR_setublimit () {512;}' unless defined(&__NR_setublimit);
90 } else {
91 die("no fairsched syscall for this arch");
92 }
93 require 'asm/ioctl.ph';
94 eval 'sub KVM_GET_API_VERSION () { &_IO(0xAE, 0x);}' unless defined(&KVM_GET_API_VERSION);
95 }
96
97 sub fairsched_mknod {
98 my ($parent, $weight, $desired) = @_;
99
100 return syscall(&__NR_fairsched_mknod, int($parent), int($weight), int($desired));
101 }
102
103 sub fairsched_rmnod {
104 my ($id) = @_;
105
106 return syscall(&__NR_fairsched_rmnod, int($id));
107 }
108
109 sub fairsched_mvpr {
110 my ($pid, $newid) = @_;
111
112 return syscall(&__NR_fairsched_mvpr, int($pid), int($newid));
113 }
114
115 sub fairsched_vcpus {
116 my ($id, $vcpus) = @_;
117
118 return syscall(&__NR_fairsched_vcpus, int($id), int($vcpus));
119 }
120
121 sub fairsched_rate {
122 my ($id, $op, $rate) = @_;
123
124 return syscall(&__NR_fairsched_rate, int($id), int($op), int($rate));
125 }
126
127 use constant FAIRSCHED_SET_RATE => 0;
128 use constant FAIRSCHED_DROP_RATE => 1;
129 use constant FAIRSCHED_GET_RATE => 2;
130
131 sub fairsched_cpulimit {
132 my ($id, $limit) = @_;
133
134 my $cpulim1024 = int($limit * 1024 / 100);
135 my $op = $cpulim1024 ? FAIRSCHED_SET_RATE : FAIRSCHED_DROP_RATE;
136
137 return fairsched_rate($id, $op, $cpulim1024);
138 }
139
140 my $nodename = PVE::INotify::nodename();
141
142 mkdir "/etc/pve/nodes/$nodename";
143 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
144 mkdir $confdir;
145
146 my $var_run_tmpdir = "/var/run/qemu-server";
147 mkdir $var_run_tmpdir;
148
149 my $lock_dir = "/var/lock/qemu-server";
150 mkdir $lock_dir;
151
152 my $pcisysfs = "/sys/bus/pci";
153
154 my $confdesc = {
155 iothread => {
156 optional => 1,
157 type => 'boolean',
158 description => "Enable iothread dataplane.",
159 default => 0,
160 },
161 onboot => {
162 optional => 1,
163 type => 'boolean',
164 description => "Specifies whether a VM will be started during system bootup.",
165 default => 0,
166 },
167 autostart => {
168 optional => 1,
169 type => 'boolean',
170 description => "Automatic restart after crash (currently ignored).",
171 default => 0,
172 },
173 hotplug => {
174 optional => 1,
175 type => 'string', format => 'pve-hotplug-features',
176 description => "Selectively enable hotplug features. This is a comma separated list of hotplug features: 'network', 'disk', 'cpu', 'memory' and 'usb'. Use '0' to disable hotplug completely. Value '1' is an alias for the default 'network,disk,usb'.",
177 default => 'network,disk,usb',
178 },
179 reboot => {
180 optional => 1,
181 type => 'boolean',
182 description => "Allow reboot. If set to '0' the VM exit on reboot.",
183 default => 1,
184 },
185 lock => {
186 optional => 1,
187 type => 'string',
188 description => "Lock/unlock the VM.",
189 enum => [qw(migrate backup snapshot rollback)],
190 },
191 cpulimit => {
192 optional => 1,
193 type => 'integer',
194 description => "Limit of CPU usage in per cent. Note if the computer has 2 CPUs, it has total of 200% CPU time. Value '0' indicates no CPU limit.\n\nNOTE: This option is currently ignored.",
195 minimum => 0,
196 default => 0,
197 },
198 cpuunits => {
199 optional => 1,
200 type => 'integer',
201 description => "CPU weight for a VM. Argument is used in the kernel fair scheduler. The larger the number is, the more CPU time this VM gets. Number is relative to weights of all the other running VMs.\n\nNOTE: You can disable fair-scheduler configuration by setting this to 0.",
202 minimum => 0,
203 maximum => 500000,
204 default => 1000,
205 },
206 memory => {
207 optional => 1,
208 type => 'integer',
209 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
210 minimum => 16,
211 default => 512,
212 },
213 balloon => {
214 optional => 1,
215 type => 'integer',
216 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
217 minimum => 0,
218 },
219 shares => {
220 optional => 1,
221 type => 'integer',
222 description => "Amount of memory shares for auto-ballooning. The larger the number is, the more memory this VM gets. Number is relative to weights of all other running VMs. Using zero disables auto-ballooning",
223 minimum => 0,
224 maximum => 50000,
225 default => 1000,
226 },
227 keyboard => {
228 optional => 1,
229 type => 'string',
230 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
231 enum => PVE::Tools::kvmkeymaplist(),
232 default => 'en-us',
233 },
234 name => {
235 optional => 1,
236 type => 'string', format => 'dns-name',
237 description => "Set a name for the VM. Only used on the configuration web interface.",
238 },
239 scsihw => {
240 optional => 1,
241 type => 'string',
242 description => "scsi controller model",
243 enum => [qw(lsi lsi53c810 virtio-scsi-pci megasas pvscsi)],
244 default => 'lsi',
245 },
246 description => {
247 optional => 1,
248 type => 'string',
249 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
250 },
251 ostype => {
252 optional => 1,
253 type => 'string',
254 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
255 description => <<EODESC,
256 Used to enable special optimization/features for specific
257 operating systems:
258
259 other => unspecified OS
260 wxp => Microsoft Windows XP
261 w2k => Microsoft Windows 2000
262 w2k3 => Microsoft Windows 2003
263 w2k8 => Microsoft Windows 2008
264 wvista => Microsoft Windows Vista
265 win7 => Microsoft Windows 7
266 win8 => Microsoft Windows 8/2012
267 l24 => Linux 2.4 Kernel
268 l26 => Linux 2.6/3.X Kernel
269 solaris => solaris/opensolaris/openindiania kernel
270
271 other|l24|l26|solaris ... no special behaviour
272 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
273 EODESC
274 },
275 boot => {
276 optional => 1,
277 type => 'string',
278 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
279 pattern => '[acdn]{1,4}',
280 default => 'cdn',
281 },
282 bootdisk => {
283 optional => 1,
284 type => 'string', format => 'pve-qm-bootdisk',
285 description => "Enable booting from specified disk.",
286 pattern => '(ide|sata|scsi|virtio)\d+',
287 },
288 smp => {
289 optional => 1,
290 type => 'integer',
291 description => "The number of CPUs. Please use option -sockets instead.",
292 minimum => 1,
293 default => 1,
294 },
295 sockets => {
296 optional => 1,
297 type => 'integer',
298 description => "The number of CPU sockets.",
299 minimum => 1,
300 default => 1,
301 },
302 cores => {
303 optional => 1,
304 type => 'integer',
305 description => "The number of cores per socket.",
306 minimum => 1,
307 default => 1,
308 },
309 numa => {
310 optional => 1,
311 type => 'boolean',
312 description => "Enable/disable Numa.",
313 default => 0,
314 },
315 vcpus => {
316 optional => 1,
317 type => 'integer',
318 description => "Number of hotplugged vcpus.",
319 minimum => 1,
320 default => 0,
321 },
322 acpi => {
323 optional => 1,
324 type => 'boolean',
325 description => "Enable/disable ACPI.",
326 default => 1,
327 },
328 agent => {
329 optional => 1,
330 type => 'boolean',
331 description => "Enable/disable Qemu GuestAgent.",
332 default => 0,
333 },
334 kvm => {
335 optional => 1,
336 type => 'boolean',
337 description => "Enable/disable KVM hardware virtualization.",
338 default => 1,
339 },
340 tdf => {
341 optional => 1,
342 type => 'boolean',
343 description => "Enable/disable time drift fix.",
344 default => 0,
345 },
346 localtime => {
347 optional => 1,
348 type => 'boolean',
349 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
350 },
351 freeze => {
352 optional => 1,
353 type => 'boolean',
354 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
355 },
356 vga => {
357 optional => 1,
358 type => 'string',
359 description => "Select VGA type. If you want to use high resolution modes (>= 1280x1024x16) then you should use option 'std' or 'vmware'. Default is 'std' for win8/win7/w2k8, and 'cirrur' for other OS types. Option 'qxl' enables the SPICE display sever. You can also run without any graphic card using a serial devive as terminal.",
360 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
361 },
362 watchdog => {
363 optional => 1,
364 type => 'string', format => 'pve-qm-watchdog',
365 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
366 description => "Create a virtual hardware watchdog device. Once enabled (by a guest action), the watchdog must be periodically polled by an agent inside the guest or else the guest will be restarted (or execute the action specified)",
367 },
368 startdate => {
369 optional => 1,
370 type => 'string',
371 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
372 description => "Set the initial date of the real time clock. Valid format for date are: 'now' or '2006-06-17T16:01:21' or '2006-06-17'.",
373 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
374 default => 'now',
375 },
376 startup => {
377 optional => 1,
378 type => 'string', format => 'pve-qm-startup',
379 typetext => '[[order=]\d+] [,up=\d+] [,down=\d+] ',
380 description => "Startup and shutdown behavior. Order is a non-negative number defining the general startup order. Shutdown in done with reverse ordering. Additionally you can set the 'up' or 'down' delay in seconds, which specifies a delay to wait before the next VM is started or stopped.",
381 },
382 template => {
383 optional => 1,
384 type => 'boolean',
385 description => "Enable/disable Template.",
386 default => 0,
387 },
388 args => {
389 optional => 1,
390 type => 'string',
391 description => <<EODESCR,
392 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
393
394 args: -no-reboot -no-hpet
395 EODESCR
396 },
397 tablet => {
398 optional => 1,
399 type => 'boolean',
400 default => 1,
401 description => "Enable/disable the usb tablet device. This device is usually needed to allow absolute mouse positioning with VNC. Else the mouse runs out of sync with normal VNC clients. If you're running lots of console-only guests on one host, you may consider disabling this to save some context switches. This is turned of by default if you use spice (vga=qxl).",
402 },
403 migrate_speed => {
404 optional => 1,
405 type => 'integer',
406 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
407 minimum => 0,
408 default => 0,
409 },
410 migrate_downtime => {
411 optional => 1,
412 type => 'number',
413 description => "Set maximum tolerated downtime (in seconds) for migrations.",
414 minimum => 0,
415 default => 0.1,
416 },
417 cdrom => {
418 optional => 1,
419 type => 'string', format => 'pve-qm-drive',
420 typetext => 'volume',
421 description => "This is an alias for option -ide2",
422 },
423 cpu => {
424 optional => 1,
425 description => "Emulated CPU type.",
426 type => 'string',
427 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
428 default => 'kvm64',
429 },
430 parent => get_standard_option('pve-snapshot-name', {
431 optional => 1,
432 description => "Parent snapshot name. This is used internally, and should not be modified.",
433 }),
434 snaptime => {
435 optional => 1,
436 description => "Timestamp for snapshots.",
437 type => 'integer',
438 minimum => 0,
439 },
440 vmstate => {
441 optional => 1,
442 type => 'string', format => 'pve-volume-id',
443 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
444 },
445 machine => {
446 description => "Specific the Qemu machine type.",
447 type => 'string',
448 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
449 maxLength => 40,
450 optional => 1,
451 },
452 smbios1 => {
453 description => "Specify SMBIOS type 1 fields.",
454 type => 'string', format => 'pve-qm-smbios1',
455 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
456 maxLength => 256,
457 optional => 1,
458 },
459 };
460
461 # what about other qemu settings ?
462 #cpu => 'string',
463 #machine => 'string',
464 #fda => 'file',
465 #fdb => 'file',
466 #mtdblock => 'file',
467 #sd => 'file',
468 #pflash => 'file',
469 #snapshot => 'bool',
470 #bootp => 'file',
471 ##tftp => 'dir',
472 ##smb => 'dir',
473 #kernel => 'file',
474 #append => 'string',
475 #initrd => 'file',
476 ##soundhw => 'string',
477
478 while (my ($k, $v) = each %$confdesc) {
479 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
480 }
481
482 my $MAX_IDE_DISKS = 4;
483 my $MAX_SCSI_DISKS = 14;
484 my $MAX_VIRTIO_DISKS = 16;
485 my $MAX_SATA_DISKS = 6;
486 my $MAX_USB_DEVICES = 5;
487 my $MAX_NETS = 32;
488 my $MAX_UNUSED_DISKS = 8;
489 my $MAX_HOSTPCI_DEVICES = 4;
490 my $MAX_SERIAL_PORTS = 4;
491 my $MAX_PARALLEL_PORTS = 3;
492 my $MAX_NUMA = 8;
493 my $MAX_MEM = 4194304;
494 my $STATICMEM = 1024;
495
496 my $numadesc = {
497 optional => 1,
498 type => 'string', format => 'pve-qm-numanode',
499 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
500 description => "numa topology",
501 };
502 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
503
504 for (my $i = 0; $i < $MAX_NUMA; $i++) {
505 $confdesc->{"numa$i"} = $numadesc;
506 }
507
508 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
509 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
510 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
511 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
512
513 my $netdesc = {
514 optional => 1,
515 type => 'string', format => 'pve-qm-net',
516 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
517 description => <<EODESCR,
518 Specify network devices.
519
520 MODEL is one of: $nic_model_list_txt
521
522 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
523 automatically generated if not specified.
524
525 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
526
527 Option 'rate' is used to limit traffic bandwidth from and to this interface. It is specified as floating point number, unit is 'Megabytes per second'.
528
529 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
530
531 10.0.2.2 Gateway
532 10.0.2.3 DNS Server
533 10.0.2.4 SMB Server
534
535 The DHCP server assign addresses to the guest starting from 10.0.2.15.
536
537 EODESCR
538 };
539 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
540
541 for (my $i = 0; $i < $MAX_NETS; $i++) {
542 $confdesc->{"net$i"} = $netdesc;
543 }
544
545 my $drivename_hash;
546
547 my $idedesc = {
548 optional => 1,
549 type => 'string', format => 'pve-qm-drive',
550 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
551 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
552 };
553 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
554
555 my $scsidesc = {
556 optional => 1,
557 type => 'string', format => 'pve-qm-drive',
558 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
559 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
560 };
561 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
562
563 my $satadesc = {
564 optional => 1,
565 type => 'string', format => 'pve-qm-drive',
566 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
567 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
568 };
569 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
570
571 my $virtiodesc = {
572 optional => 1,
573 type => 'string', format => 'pve-qm-drive',
574 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
575 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
576 };
577 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
578
579 my $usbdesc = {
580 optional => 1,
581 type => 'string', format => 'pve-qm-usb-device',
582 typetext => 'host=HOSTUSBDEVICE|spice',
583 description => <<EODESCR,
584 Configure an USB device (n is 0 to 4). This can be used to
585 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
586
587 'bus-port(.port)*' (decimal numbers) or
588 'vendor_id:product_id' (hexadeciaml numbers)
589
590 You can use the 'lsusb -t' command to list existing usb devices.
591
592 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
593
594 The value 'spice' can be used to add a usb redirection devices for spice.
595
596 EODESCR
597 };
598 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
599
600 my $hostpcidesc = {
601 optional => 1,
602 type => 'string', format => 'pve-qm-hostpci',
603 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
604 description => <<EODESCR,
605 Map host pci devices. HOSTPCIDEVICE syntax is:
606
607 'bus:dev.func' (hexadecimal numbers)
608
609 You can us the 'lspci' command to list existing pci devices.
610
611 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
612
613 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
614
615 Experimental: user reported problems with this option.
616 EODESCR
617 };
618 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
619
620 my $serialdesc = {
621 optional => 1,
622 type => 'string',
623 pattern => '(/dev/.+|socket)',
624 description => <<EODESCR,
625 Create a serial device inside the VM (n is 0 to 3), and pass through a host serial device (i.e. /dev/ttyS0), or create a unix socket on the host side (use 'qm terminal' to open a terminal connection).
626
627 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
628
629 Experimental: user reported problems with this option.
630 EODESCR
631 };
632
633 my $paralleldesc= {
634 optional => 1,
635 type => 'string',
636 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
637 description => <<EODESCR,
638 Map host parallel devices (n is 0 to 2).
639
640 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
641
642 Experimental: user reported problems with this option.
643 EODESCR
644 };
645
646 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
647 $confdesc->{"parallel$i"} = $paralleldesc;
648 }
649
650 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
651 $confdesc->{"serial$i"} = $serialdesc;
652 }
653
654 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
655 $confdesc->{"hostpci$i"} = $hostpcidesc;
656 }
657
658 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
659 $drivename_hash->{"ide$i"} = 1;
660 $confdesc->{"ide$i"} = $idedesc;
661 }
662
663 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
664 $drivename_hash->{"sata$i"} = 1;
665 $confdesc->{"sata$i"} = $satadesc;
666 }
667
668 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
669 $drivename_hash->{"scsi$i"} = 1;
670 $confdesc->{"scsi$i"} = $scsidesc ;
671 }
672
673 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
674 $drivename_hash->{"virtio$i"} = 1;
675 $confdesc->{"virtio$i"} = $virtiodesc;
676 }
677
678 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
679 $confdesc->{"usb$i"} = $usbdesc;
680 }
681
682 my $unuseddesc = {
683 optional => 1,
684 type => 'string', format => 'pve-volume-id',
685 description => "Reference to unused volumes.",
686 };
687
688 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
689 $confdesc->{"unused$i"} = $unuseddesc;
690 }
691
692 my $kvm_api_version = 0;
693
694 sub kvm_version {
695
696 return $kvm_api_version if $kvm_api_version;
697
698 my $fh = IO::File->new("</dev/kvm") ||
699 return 0;
700
701 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
702 $kvm_api_version = $v;
703 }
704
705 $fh->close();
706
707 return $kvm_api_version;
708 }
709
710 my $kvm_user_version;
711
712 sub kvm_user_version {
713
714 return $kvm_user_version if $kvm_user_version;
715
716 $kvm_user_version = 'unknown';
717
718 my $tmp = `kvm -help 2>/dev/null`;
719
720 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
721 $kvm_user_version = $2;
722 }
723
724 return $kvm_user_version;
725
726 }
727
728 my $kernel_has_vhost_net = -c '/dev/vhost-net';
729
730 sub disknames {
731 # order is important - used to autoselect boot disk
732 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
733 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
734 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
735 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
736 }
737
738 sub valid_drivename {
739 my $dev = shift;
740
741 return defined($drivename_hash->{$dev});
742 }
743
744 sub option_exists {
745 my $key = shift;
746 return defined($confdesc->{$key});
747 }
748
749 sub nic_models {
750 return $nic_model_list;
751 }
752
753 sub os_list_description {
754
755 return {
756 other => 'Other',
757 wxp => 'Windows XP',
758 w2k => 'Windows 2000',
759 w2k3 =>, 'Windows 2003',
760 w2k8 => 'Windows 2008',
761 wvista => 'Windows Vista',
762 win7 => 'Windows 7',
763 win8 => 'Windows 8/2012',
764 l24 => 'Linux 2.4',
765 l26 => 'Linux 2.6',
766 };
767 }
768
769 my $cdrom_path;
770
771 sub get_cdrom_path {
772
773 return $cdrom_path if $cdrom_path;
774
775 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
776 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
777 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
778 }
779
780 sub get_iso_path {
781 my ($storecfg, $vmid, $cdrom) = @_;
782
783 if ($cdrom eq 'cdrom') {
784 return get_cdrom_path();
785 } elsif ($cdrom eq 'none') {
786 return '';
787 } elsif ($cdrom =~ m|^/|) {
788 return $cdrom;
789 } else {
790 return PVE::Storage::path($storecfg, $cdrom);
791 }
792 }
793
794 # try to convert old style file names to volume IDs
795 sub filename_to_volume_id {
796 my ($vmid, $file, $media) = @_;
797
798 if (!($file eq 'none' || $file eq 'cdrom' ||
799 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
800
801 return undef if $file =~ m|/|;
802
803 if ($media && $media eq 'cdrom') {
804 $file = "local:iso/$file";
805 } else {
806 $file = "local:$vmid/$file";
807 }
808 }
809
810 return $file;
811 }
812
813 sub verify_media_type {
814 my ($opt, $vtype, $media) = @_;
815
816 return if !$media;
817
818 my $etype;
819 if ($media eq 'disk') {
820 $etype = 'images';
821 } elsif ($media eq 'cdrom') {
822 $etype = 'iso';
823 } else {
824 die "internal error";
825 }
826
827 return if ($vtype eq $etype);
828
829 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
830 }
831
832 sub cleanup_drive_path {
833 my ($opt, $storecfg, $drive) = @_;
834
835 # try to convert filesystem paths to volume IDs
836
837 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
838 ($drive->{file} !~ m|^/dev/.+|) &&
839 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
840 ($drive->{file} !~ m/^\d+$/)) {
841 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
842 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
843 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
844 verify_media_type($opt, $vtype, $drive->{media});
845 $drive->{file} = $volid;
846 }
847
848 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
849 }
850
851 sub create_conf_nolock {
852 my ($vmid, $settings) = @_;
853
854 my $filename = config_file($vmid);
855
856 die "configuration file '$filename' already exists\n" if -f $filename;
857
858 my $defaults = load_defaults();
859
860 $settings->{name} = "vm$vmid" if !$settings->{name};
861 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
862
863 my $data = '';
864 foreach my $opt (keys %$settings) {
865 next if !$confdesc->{$opt};
866
867 my $value = $settings->{$opt};
868 next if !$value;
869
870 $data .= "$opt: $value\n";
871 }
872
873 PVE::Tools::file_set_contents($filename, $data);
874 }
875
876 sub parse_hotplug_features {
877 my ($data) = @_;
878
879 my $res = {};
880
881 return $res if $data eq '0';
882
883 $data = $confdesc->{hotplug}->{default} if $data eq '1';
884
885 foreach my $feature (PVE::Tools::split_list($data)) {
886 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
887 $res->{$1} = 1;
888 } else {
889 warn "ignoring unknown hotplug feature '$feature'\n";
890 }
891 }
892 return $res;
893 }
894
895 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
896 sub pve_verify_hotplug_features {
897 my ($value, $noerr) = @_;
898
899 return $value if parse_hotplug_features($value);
900
901 return undef if $noerr;
902
903 die "unable to parse hotplug option\n";
904 }
905
906 my $parse_size = sub {
907 my ($value) = @_;
908
909 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
910 my ($size, $unit) = ($1, $3);
911 if ($unit) {
912 if ($unit eq 'K') {
913 $size = $size * 1024;
914 } elsif ($unit eq 'M') {
915 $size = $size * 1024 * 1024;
916 } elsif ($unit eq 'G') {
917 $size = $size * 1024 * 1024 * 1024;
918 }
919 }
920 return int($size);
921 };
922
923 my $format_size = sub {
924 my ($size) = @_;
925
926 $size = int($size);
927
928 my $kb = int($size/1024);
929 return $size if $kb*1024 != $size;
930
931 my $mb = int($kb/1024);
932 return "${kb}K" if $mb*1024 != $kb;
933
934 my $gb = int($mb/1024);
935 return "${mb}M" if $gb*1024 != $mb;
936
937 return "${gb}G";
938 };
939
940 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
941 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
942 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
943 # [,aio=native|threads][,discard=ignore|on]
944
945 sub parse_drive {
946 my ($key, $data) = @_;
947
948 my $res = {};
949
950 # $key may be undefined - used to verify JSON parameters
951 if (!defined($key)) {
952 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
953 $res->{index} = 0;
954 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
955 $res->{interface} = $1;
956 $res->{index} = $2;
957 } else {
958 return undef;
959 }
960
961 foreach my $p (split (/,/, $data)) {
962 next if $p =~ m/^\s*$/;
963
964 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard)=(.+)$/) {
965 my ($k, $v) = ($1, $2);
966
967 $k = 'file' if $k eq 'volume';
968
969 return undef if defined $res->{$k};
970
971 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
972 return undef if !$v || $v !~ m/^\d+/;
973 $k = "m$k";
974 $v = sprintf("%.3f", $v / (1024*1024));
975 }
976 $res->{$k} = $v;
977 } else {
978 if (!$res->{file} && $p !~ m/=/) {
979 $res->{file} = $p;
980 } else {
981 return undef;
982 }
983 }
984 }
985
986 return undef if !$res->{file};
987
988 if($res->{file} =~ m/\.(raw|cow|qcow|qcow2|vmdk|cloop)$/){
989 $res->{format} = $1;
990 }
991
992 return undef if $res->{cache} &&
993 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
994 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
995 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
996 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
997 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
998 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
999 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
1000 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qcow2|vmdk|cloop)$/;
1001 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
1002 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
1003 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
1004 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
1005 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
1006
1007 return undef if $res->{mbps_rd} && $res->{mbps};
1008 return undef if $res->{mbps_wr} && $res->{mbps};
1009
1010 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
1011 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
1012 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
1013 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
1014 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
1015 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
1016
1017 return undef if $res->{iops_rd} && $res->{iops};
1018 return undef if $res->{iops_wr} && $res->{iops};
1019
1020
1021 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
1022 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
1023 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
1024 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
1025 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
1026 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
1027
1028
1029 if ($res->{size}) {
1030 return undef if !defined($res->{size} = &$parse_size($res->{size}));
1031 }
1032
1033 if ($res->{media} && ($res->{media} eq 'cdrom')) {
1034 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
1035 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
1036 return undef if $res->{interface} eq 'virtio';
1037 }
1038
1039 # rerror does not work with scsi drives
1040 if ($res->{rerror}) {
1041 return undef if $res->{interface} eq 'scsi';
1042 }
1043
1044 return $res;
1045 }
1046
1047 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
1048
1049 sub print_drive {
1050 my ($vmid, $drive) = @_;
1051
1052 my $opts = '';
1053 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup') {
1054 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1055 }
1056
1057 if ($drive->{size}) {
1058 $opts .= ",size=" . &$format_size($drive->{size});
1059 }
1060
1061 return "$drive->{file}$opts";
1062 }
1063
1064 sub scsi_inquiry {
1065 my($fh, $noerr) = @_;
1066
1067 my $SG_IO = 0x2285;
1068 my $SG_GET_VERSION_NUM = 0x2282;
1069
1070 my $versionbuf = "\x00" x 8;
1071 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
1072 if (!$ret) {
1073 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1074 return undef;
1075 }
1076 my $version = unpack("I", $versionbuf);
1077 if ($version < 30000) {
1078 die "scsi generic interface too old\n" if !$noerr;
1079 return undef;
1080 }
1081
1082 my $buf = "\x00" x 36;
1083 my $sensebuf = "\x00" x 8;
1084 my $cmd = pack("C x3 C x1", 0x12, 36);
1085
1086 # see /usr/include/scsi/sg.h
1087 my $sg_io_hdr_t = "i i C C s I P P P I I i P C C C C S S i I I";
1088
1089 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1090 length($sensebuf), 0, length($buf), $buf,
1091 $cmd, $sensebuf, 6000);
1092
1093 $ret = ioctl($fh, $SG_IO, $packet);
1094 if (!$ret) {
1095 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1096 return undef;
1097 }
1098
1099 my @res = unpack($sg_io_hdr_t, $packet);
1100 if ($res[17] || $res[18]) {
1101 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1102 return undef;
1103 }
1104
1105 my $res = {};
1106 (my $byte0, my $byte1, $res->{vendor},
1107 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1108
1109 $res->{removable} = $byte1 & 128 ? 1 : 0;
1110 $res->{type} = $byte0 & 31;
1111
1112 return $res;
1113 }
1114
1115 sub path_is_scsi {
1116 my ($path) = @_;
1117
1118 my $fh = IO::File->new("+<$path") || return undef;
1119 my $res = scsi_inquiry($fh, 1);
1120 close($fh);
1121
1122 return $res;
1123 }
1124
1125 sub machine_type_is_q35 {
1126 my ($conf) = @_;
1127
1128 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1129 }
1130
1131 sub print_tabletdevice_full {
1132 my ($conf) = @_;
1133
1134 my $q35 = machine_type_is_q35($conf);
1135
1136 # we use uhci for old VMs because tablet driver was buggy in older qemu
1137 my $usbbus = $q35 ? "ehci" : "uhci";
1138
1139 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1140 }
1141
1142 sub print_drivedevice_full {
1143 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1144
1145 my $device = '';
1146 my $maxdev = 0;
1147
1148 if ($drive->{interface} eq 'virtio') {
1149 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1150 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1151 $device .= ",iothread=iothread0" if $conf->{iothread};
1152 } elsif ($drive->{interface} eq 'scsi') {
1153 $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
1154 my $controller = int($drive->{index} / $maxdev);
1155 my $unit = $drive->{index} % $maxdev;
1156 my $devicetype = 'hd';
1157 my $path = '';
1158 if (drive_is_cdrom($drive)) {
1159 $devicetype = 'cd';
1160 } else {
1161 if ($drive->{file} =~ m|^/|) {
1162 $path = $drive->{file};
1163 } else {
1164 $path = PVE::Storage::path($storecfg, $drive->{file});
1165 }
1166
1167 if($path =~ m/^iscsi\:\/\//){
1168 $devicetype = 'generic';
1169 } else {
1170 if (my $info = path_is_scsi($path)) {
1171 if ($info->{type} == 0) {
1172 $devicetype = 'block';
1173 } elsif ($info->{type} == 1) { # tape
1174 $devicetype = 'generic';
1175 }
1176 }
1177 }
1178 }
1179
1180 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1181 $device = "scsi-$devicetype,bus=scsihw$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1182 } else {
1183 $device = "scsi-$devicetype,bus=scsihw$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1184 }
1185
1186 } elsif ($drive->{interface} eq 'ide'){
1187 $maxdev = 2;
1188 my $controller = int($drive->{index} / $maxdev);
1189 my $unit = $drive->{index} % $maxdev;
1190 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1191
1192 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1193 } elsif ($drive->{interface} eq 'sata'){
1194 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1195 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1196 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1197 } elsif ($drive->{interface} eq 'usb') {
1198 die "implement me";
1199 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1200 } else {
1201 die "unsupported interface type";
1202 }
1203
1204 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1205
1206 return $device;
1207 }
1208
1209 sub get_initiator_name {
1210 my $initiator;
1211
1212 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1213 while (defined(my $line = <$fh>)) {
1214 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1215 $initiator = $1;
1216 last;
1217 }
1218 $fh->close();
1219
1220 return $initiator;
1221 }
1222
1223 sub print_drive_full {
1224 my ($storecfg, $vmid, $drive) = @_;
1225
1226 my $opts = '';
1227 foreach my $o (@qemu_drive_options) {
1228 next if $o eq 'bootindex';
1229 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1230 }
1231
1232 foreach my $o (qw(bps bps_rd bps_wr)) {
1233 my $v = $drive->{"m$o"};
1234 $opts .= ",$o=" . int($v*1024*1024) if $v;
1235 }
1236
1237 # use linux-aio by default (qemu default is threads)
1238 $opts .= ",aio=native" if !$drive->{aio};
1239
1240 my $path;
1241 my $volid = $drive->{file};
1242 if (drive_is_cdrom($drive)) {
1243 $path = get_iso_path($storecfg, $vmid, $volid);
1244 } else {
1245 if ($volid =~ m|^/|) {
1246 $path = $volid;
1247 } else {
1248 $path = PVE::Storage::path($storecfg, $volid);
1249 }
1250 }
1251
1252 $opts .= ",cache=none" if !$drive->{cache} && !drive_is_cdrom($drive);
1253
1254 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1255 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1256
1257 my $pathinfo = $path ? "file=$path," : '';
1258
1259 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1260 }
1261
1262 sub print_netdevice_full {
1263 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1264
1265 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1266
1267 my $device = $net->{model};
1268 if ($net->{model} eq 'virtio') {
1269 $device = 'virtio-net-pci';
1270 };
1271
1272 my $pciaddr = print_pci_addr("$netid", $bridges);
1273 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1274 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1275 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1276 my $vectors = $net->{queues} * 2 + 2;
1277 $tmpstr .= ",vectors=$vectors,mq=on";
1278 }
1279 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1280 return $tmpstr;
1281 }
1282
1283 sub print_netdev_full {
1284 my ($vmid, $conf, $net, $netid) = @_;
1285
1286 my $i = '';
1287 if ($netid =~ m/^net(\d+)$/) {
1288 $i = int($1);
1289 }
1290
1291 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1292
1293 my $ifname = "tap${vmid}i$i";
1294
1295 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1296 die "interface name '$ifname' is too long (max 15 character)\n"
1297 if length($ifname) >= 16;
1298
1299 my $vhostparam = '';
1300 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1301
1302 my $vmname = $conf->{name} || "vm$vmid";
1303
1304 my $netdev = "";
1305
1306 if ($net->{bridge}) {
1307 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1308 } else {
1309 $netdev = "type=user,id=$netid,hostname=$vmname";
1310 }
1311
1312 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1313
1314 return $netdev;
1315 }
1316
1317 sub drive_is_cdrom {
1318 my ($drive) = @_;
1319
1320 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1321
1322 }
1323
1324 sub parse_numa {
1325 my ($data) = @_;
1326
1327 my $res = {};
1328
1329 foreach my $kvp (split(/,/, $data)) {
1330
1331 if ($kvp =~ m/^memory=(\S+)$/) {
1332 $res->{memory} = $1;
1333 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1334 $res->{policy} = $1;
1335 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1336 $res->{cpus}->{start} = $1;
1337 $res->{cpus}->{end} = $3;
1338 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1339 $res->{hostnodes}->{start} = $1;
1340 $res->{hostnodes}->{end} = $3;
1341 } else {
1342 return undef;
1343 }
1344 }
1345
1346 return $res;
1347 }
1348
1349 sub parse_hostpci {
1350 my ($value) = @_;
1351
1352 return undef if !$value;
1353
1354
1355 my @list = split(/,/, $value);
1356 my $found;
1357
1358 my $res = {};
1359 foreach my $kv (@list) {
1360
1361 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1362 $found = 1;
1363 if(defined($4)){
1364 push @{$res->{pciid}}, { id => $2 , function => $4};
1365
1366 }else{
1367 my $pcidevices = lspci($2);
1368 $res->{pciid} = $pcidevices->{$2};
1369 }
1370 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1371 $res->{driver} = $1;
1372 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1373 $res->{rombar} = $1;
1374 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1375 $res->{'x-vga'} = $1;
1376 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1377 $res->{pcie} = 1 if $1 == 1;
1378 } else {
1379 warn "unknown hostpci setting '$kv'\n";
1380 }
1381 }
1382
1383 return undef if !$found;
1384
1385 return $res;
1386 }
1387
1388 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1389 sub parse_net {
1390 my ($data) = @_;
1391
1392 my $res = {};
1393
1394 foreach my $kvp (split(/,/, $data)) {
1395
1396 if ($kvp =~ m/^(ne2k_pci|e1000|e1000-82540em|e1000-82544gc|e1000-82545em|rtl8139|pcnet|virtio|ne2k_isa|i82551|i82557b|i82559er|vmxnet3)(=([0-9a-f]{2}(:[0-9a-f]{2}){5}))?$/i) {
1397 my $model = lc($1);
1398 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1399 $res->{model} = $model;
1400 $res->{macaddr} = $mac;
1401 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1402 $res->{bridge} = $1;
1403 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1404 $res->{queues} = $1;
1405 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1406 $res->{rate} = $1;
1407 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1408 $res->{tag} = $1;
1409 } elsif ($kvp =~ m/^firewall=([01])$/) {
1410 $res->{firewall} = $1;
1411 } elsif ($kvp =~ m/^link_down=([01])$/) {
1412 $res->{link_down} = $1;
1413 } else {
1414 return undef;
1415 }
1416
1417 }
1418
1419 return undef if !$res->{model};
1420
1421 return $res;
1422 }
1423
1424 sub print_net {
1425 my $net = shift;
1426
1427 my $res = "$net->{model}";
1428 $res .= "=$net->{macaddr}" if $net->{macaddr};
1429 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1430 $res .= ",rate=$net->{rate}" if $net->{rate};
1431 $res .= ",tag=$net->{tag}" if $net->{tag};
1432 $res .= ",firewall=1" if $net->{firewall};
1433 $res .= ",link_down=1" if $net->{link_down};
1434
1435 return $res;
1436 }
1437
1438 sub add_random_macs {
1439 my ($settings) = @_;
1440
1441 foreach my $opt (keys %$settings) {
1442 next if $opt !~ m/^net(\d+)$/;
1443 my $net = parse_net($settings->{$opt});
1444 next if !$net;
1445 $settings->{$opt} = print_net($net);
1446 }
1447 }
1448
1449 sub add_unused_volume {
1450 my ($config, $volid) = @_;
1451
1452 my $key;
1453 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1454 my $test = "unused$ind";
1455 if (my $vid = $config->{$test}) {
1456 return if $vid eq $volid; # do not add duplicates
1457 } else {
1458 $key = $test;
1459 }
1460 }
1461
1462 die "To many unused volume - please delete them first.\n" if !$key;
1463
1464 $config->{$key} = $volid;
1465
1466 return $key;
1467 }
1468
1469 sub vm_is_volid_owner {
1470 my ($storecfg, $vmid, $volid) = @_;
1471
1472 if ($volid !~ m|^/|) {
1473 my ($path, $owner);
1474 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1475 if ($owner && ($owner == $vmid)) {
1476 return 1;
1477 }
1478 }
1479
1480 return undef;
1481 }
1482
1483 sub vmconfig_delete_pending_option {
1484 my ($conf, $key) = @_;
1485
1486 delete $conf->{pending}->{$key};
1487 my $pending_delete_hash = { $key => 1 };
1488 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1489 $pending_delete_hash->{$opt} = 1;
1490 }
1491 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1492 }
1493
1494 sub vmconfig_undelete_pending_option {
1495 my ($conf, $key) = @_;
1496
1497 my $pending_delete_hash = {};
1498 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1499 $pending_delete_hash->{$opt} = 1;
1500 }
1501 delete $pending_delete_hash->{$key};
1502
1503 my @keylist = keys %$pending_delete_hash;
1504 if (scalar(@keylist)) {
1505 $conf->{pending}->{delete} = join(',', @keylist);
1506 } else {
1507 delete $conf->{pending}->{delete};
1508 }
1509 }
1510
1511 sub vmconfig_register_unused_drive {
1512 my ($storecfg, $vmid, $conf, $drive) = @_;
1513
1514 if (!drive_is_cdrom($drive)) {
1515 my $volid = $drive->{file};
1516 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1517 add_unused_volume($conf, $volid, $vmid);
1518 }
1519 }
1520 }
1521
1522 sub vmconfig_cleanup_pending {
1523 my ($conf) = @_;
1524
1525 # remove pending changes when nothing changed
1526 my $changes;
1527 foreach my $opt (keys %{$conf->{pending}}) {
1528 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1529 $changes = 1;
1530 delete $conf->{pending}->{$opt};
1531 }
1532 }
1533
1534 # remove delete if option is not set
1535 my $pending_delete_hash = {};
1536 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1537 if (defined($conf->{$opt})) {
1538 $pending_delete_hash->{$opt} = 1;
1539 } else {
1540 $changes = 1;
1541 }
1542 }
1543
1544 my @keylist = keys %$pending_delete_hash;
1545 if (scalar(@keylist)) {
1546 $conf->{pending}->{delete} = join(',', @keylist);
1547 } else {
1548 delete $conf->{pending}->{delete};
1549 }
1550
1551 return $changes;
1552 }
1553
1554 my $valid_smbios1_options = {
1555 manufacturer => '\S+',
1556 product => '\S+',
1557 version => '\S+',
1558 serial => '\S+',
1559 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1560 sku => '\S+',
1561 family => '\S+',
1562 };
1563
1564 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1565 sub parse_smbios1 {
1566 my ($data) = @_;
1567
1568 my $res = {};
1569
1570 foreach my $kvp (split(/,/, $data)) {
1571 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1572 my ($k, $v) = split(/=/, $kvp);
1573 return undef if !defined($k) || !defined($v);
1574 return undef if !$valid_smbios1_options->{$k};
1575 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1576 $res->{$k} = $v;
1577 }
1578
1579 return $res;
1580 }
1581
1582 sub print_smbios1 {
1583 my ($smbios1) = @_;
1584
1585 my $data = '';
1586 foreach my $k (keys %$smbios1) {
1587 next if !defined($smbios1->{$k});
1588 next if !$valid_smbios1_options->{$k};
1589 $data .= ',' if $data;
1590 $data .= "$k=$smbios1->{$k}";
1591 }
1592 return $data;
1593 }
1594
1595 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1596 sub verify_smbios1 {
1597 my ($value, $noerr) = @_;
1598
1599 return $value if parse_smbios1($value);
1600
1601 return undef if $noerr;
1602
1603 die "unable to parse smbios (type 1) options\n";
1604 }
1605
1606 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1607 sub verify_bootdisk {
1608 my ($value, $noerr) = @_;
1609
1610 return $value if valid_drivename($value);
1611
1612 return undef if $noerr;
1613
1614 die "invalid boot disk '$value'\n";
1615 }
1616
1617 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1618 sub verify_numa {
1619 my ($value, $noerr) = @_;
1620
1621 return $value if parse_numa($value);
1622
1623 return undef if $noerr;
1624
1625 die "unable to parse numa options\n";
1626 }
1627
1628 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1629 sub verify_net {
1630 my ($value, $noerr) = @_;
1631
1632 return $value if parse_net($value);
1633
1634 return undef if $noerr;
1635
1636 die "unable to parse network options\n";
1637 }
1638
1639 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1640 sub verify_drive {
1641 my ($value, $noerr) = @_;
1642
1643 return $value if parse_drive(undef, $value);
1644
1645 return undef if $noerr;
1646
1647 die "unable to parse drive options\n";
1648 }
1649
1650 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1651 sub verify_hostpci {
1652 my ($value, $noerr) = @_;
1653
1654 return $value if parse_hostpci($value);
1655
1656 return undef if $noerr;
1657
1658 die "unable to parse pci id\n";
1659 }
1660
1661 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1662 sub verify_watchdog {
1663 my ($value, $noerr) = @_;
1664
1665 return $value if parse_watchdog($value);
1666
1667 return undef if $noerr;
1668
1669 die "unable to parse watchdog options\n";
1670 }
1671
1672 sub parse_watchdog {
1673 my ($value) = @_;
1674
1675 return undef if !$value;
1676
1677 my $res = {};
1678
1679 foreach my $p (split(/,/, $value)) {
1680 next if $p =~ m/^\s*$/;
1681
1682 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1683 $res->{model} = $2;
1684 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1685 $res->{action} = $2;
1686 } else {
1687 return undef;
1688 }
1689 }
1690
1691 return $res;
1692 }
1693
1694 PVE::JSONSchema::register_format('pve-qm-startup', \&verify_startup);
1695 sub verify_startup {
1696 my ($value, $noerr) = @_;
1697
1698 return $value if parse_startup($value);
1699
1700 return undef if $noerr;
1701
1702 die "unable to parse startup options\n";
1703 }
1704
1705 sub parse_startup {
1706 my ($value) = @_;
1707
1708 return undef if !$value;
1709
1710 my $res = {};
1711
1712 foreach my $p (split(/,/, $value)) {
1713 next if $p =~ m/^\s*$/;
1714
1715 if ($p =~ m/^(order=)?(\d+)$/) {
1716 $res->{order} = $2;
1717 } elsif ($p =~ m/^up=(\d+)$/) {
1718 $res->{up} = $1;
1719 } elsif ($p =~ m/^down=(\d+)$/) {
1720 $res->{down} = $1;
1721 } else {
1722 return undef;
1723 }
1724 }
1725
1726 return $res;
1727 }
1728
1729 sub parse_usb_device {
1730 my ($value) = @_;
1731
1732 return undef if !$value;
1733
1734 my @dl = split(/,/, $value);
1735 my $found;
1736
1737 my $res = {};
1738 foreach my $v (@dl) {
1739 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1740 $found = 1;
1741 $res->{vendorid} = $2;
1742 $res->{productid} = $4;
1743 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1744 $found = 1;
1745 $res->{hostbus} = $1;
1746 $res->{hostport} = $2;
1747 } elsif ($v =~ m/^spice$/) {
1748 $found = 1;
1749 $res->{spice} = 1;
1750 } else {
1751 return undef;
1752 }
1753 }
1754 return undef if !$found;
1755
1756 return $res;
1757 }
1758
1759 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1760 sub verify_usb_device {
1761 my ($value, $noerr) = @_;
1762
1763 return $value if parse_usb_device($value);
1764
1765 return undef if $noerr;
1766
1767 die "unable to parse usb device\n";
1768 }
1769
1770 # add JSON properties for create and set function
1771 sub json_config_properties {
1772 my $prop = shift;
1773
1774 foreach my $opt (keys %$confdesc) {
1775 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1776 $prop->{$opt} = $confdesc->{$opt};
1777 }
1778
1779 return $prop;
1780 }
1781
1782 sub check_type {
1783 my ($key, $value) = @_;
1784
1785 die "unknown setting '$key'\n" if !$confdesc->{$key};
1786
1787 my $type = $confdesc->{$key}->{type};
1788
1789 if (!defined($value)) {
1790 die "got undefined value\n";
1791 }
1792
1793 if ($value =~ m/[\n\r]/) {
1794 die "property contains a line feed\n";
1795 }
1796
1797 if ($type eq 'boolean') {
1798 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1799 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1800 die "type check ('boolean') failed - got '$value'\n";
1801 } elsif ($type eq 'integer') {
1802 return int($1) if $value =~ m/^(\d+)$/;
1803 die "type check ('integer') failed - got '$value'\n";
1804 } elsif ($type eq 'number') {
1805 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1806 die "type check ('number') failed - got '$value'\n";
1807 } elsif ($type eq 'string') {
1808 if (my $fmt = $confdesc->{$key}->{format}) {
1809 if ($fmt eq 'pve-qm-drive') {
1810 # special case - we need to pass $key to parse_drive()
1811 my $drive = parse_drive($key, $value);
1812 return $value if $drive;
1813 die "unable to parse drive options\n";
1814 }
1815 PVE::JSONSchema::check_format($fmt, $value);
1816 return $value;
1817 }
1818 $value =~ s/^\"(.*)\"$/$1/;
1819 return $value;
1820 } else {
1821 die "internal error"
1822 }
1823 }
1824
1825 sub lock_config_full {
1826 my ($vmid, $timeout, $code, @param) = @_;
1827
1828 my $filename = config_file_lock($vmid);
1829
1830 my $res = lock_file($filename, $timeout, $code, @param);
1831
1832 die $@ if $@;
1833
1834 return $res;
1835 }
1836
1837 sub lock_config_mode {
1838 my ($vmid, $timeout, $shared, $code, @param) = @_;
1839
1840 my $filename = config_file_lock($vmid);
1841
1842 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1843
1844 die $@ if $@;
1845
1846 return $res;
1847 }
1848
1849 sub lock_config {
1850 my ($vmid, $code, @param) = @_;
1851
1852 return lock_config_full($vmid, 10, $code, @param);
1853 }
1854
1855 sub cfs_config_path {
1856 my ($vmid, $node) = @_;
1857
1858 $node = $nodename if !$node;
1859 return "nodes/$node/qemu-server/$vmid.conf";
1860 }
1861
1862 sub check_iommu_support{
1863 #fixme : need to check IOMMU support
1864 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1865
1866 my $iommu=1;
1867 return $iommu;
1868
1869 }
1870
1871 sub config_file {
1872 my ($vmid, $node) = @_;
1873
1874 my $cfspath = cfs_config_path($vmid, $node);
1875 return "/etc/pve/$cfspath";
1876 }
1877
1878 sub config_file_lock {
1879 my ($vmid) = @_;
1880
1881 return "$lock_dir/lock-$vmid.conf";
1882 }
1883
1884 sub touch_config {
1885 my ($vmid) = @_;
1886
1887 my $conf = config_file($vmid);
1888 utime undef, undef, $conf;
1889 }
1890
1891 sub destroy_vm {
1892 my ($storecfg, $vmid, $keep_empty_config) = @_;
1893
1894 my $conffile = config_file($vmid);
1895
1896 my $conf = load_config($vmid);
1897
1898 check_lock($conf);
1899
1900 # only remove disks owned by this VM
1901 foreach_drive($conf, sub {
1902 my ($ds, $drive) = @_;
1903
1904 return if drive_is_cdrom($drive);
1905
1906 my $volid = $drive->{file};
1907
1908 return if !$volid || $volid =~ m|^/|;
1909
1910 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1911 return if !$path || !$owner || ($owner != $vmid);
1912
1913 PVE::Storage::vdisk_free($storecfg, $volid);
1914 });
1915
1916 if ($keep_empty_config) {
1917 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1918 } else {
1919 unlink $conffile;
1920 }
1921
1922 # also remove unused disk
1923 eval {
1924 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1925
1926 eval {
1927 PVE::Storage::foreach_volid($dl, sub {
1928 my ($volid, $sid, $volname, $d) = @_;
1929 PVE::Storage::vdisk_free($storecfg, $volid);
1930 });
1931 };
1932 warn $@ if $@;
1933
1934 };
1935 warn $@ if $@;
1936 }
1937
1938 sub load_config {
1939 my ($vmid, $node) = @_;
1940
1941 my $cfspath = cfs_config_path($vmid, $node);
1942
1943 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1944
1945 die "no such VM ('$vmid')\n" if !defined($conf);
1946
1947 return $conf;
1948 }
1949
1950 sub parse_vm_config {
1951 my ($filename, $raw) = @_;
1952
1953 return undef if !defined($raw);
1954
1955 my $res = {
1956 digest => Digest::SHA::sha1_hex($raw),
1957 snapshots => {},
1958 pending => {},
1959 };
1960
1961 $filename =~ m|/qemu-server/(\d+)\.conf$|
1962 || die "got strange filename '$filename'";
1963
1964 my $vmid = $1;
1965
1966 my $conf = $res;
1967 my $descr = '';
1968 my $section = '';
1969
1970 my @lines = split(/\n/, $raw);
1971 foreach my $line (@lines) {
1972 next if $line =~ m/^\s*$/;
1973
1974 if ($line =~ m/^\[PENDING\]\s*$/i) {
1975 $section = 'pending';
1976 $conf->{description} = $descr if $descr;
1977 $descr = '';
1978 $conf = $res->{$section} = {};
1979 next;
1980
1981 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1982 $section = $1;
1983 $conf->{description} = $descr if $descr;
1984 $descr = '';
1985 $conf = $res->{snapshots}->{$section} = {};
1986 next;
1987 }
1988
1989 if ($line =~ m/^\#(.*)\s*$/) {
1990 $descr .= PVE::Tools::decode_text($1) . "\n";
1991 next;
1992 }
1993
1994 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1995 $descr .= PVE::Tools::decode_text($2);
1996 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1997 $conf->{snapstate} = $1;
1998 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1999 my $key = $1;
2000 my $value = $2;
2001 $conf->{$key} = $value;
2002 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
2003 my $value = $1;
2004 if ($section eq 'pending') {
2005 $conf->{delete} = $value; # we parse this later
2006 } else {
2007 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
2008 }
2009 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
2010 my $key = $1;
2011 my $value = $2;
2012 eval { $value = check_type($key, $value); };
2013 if ($@) {
2014 warn "vm $vmid - unable to parse value of '$key' - $@";
2015 } else {
2016 my $fmt = $confdesc->{$key}->{format};
2017 if ($fmt && $fmt eq 'pve-qm-drive') {
2018 my $v = parse_drive($key, $value);
2019 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
2020 $v->{file} = $volid;
2021 $value = print_drive($vmid, $v);
2022 } else {
2023 warn "vm $vmid - unable to parse value of '$key'\n";
2024 next;
2025 }
2026 }
2027
2028 if ($key eq 'cdrom') {
2029 $conf->{ide2} = $value;
2030 } else {
2031 $conf->{$key} = $value;
2032 }
2033 }
2034 }
2035 }
2036
2037 $conf->{description} = $descr if $descr;
2038
2039 delete $res->{snapstate}; # just to be sure
2040
2041 return $res;
2042 }
2043
2044 sub write_vm_config {
2045 my ($filename, $conf) = @_;
2046
2047 delete $conf->{snapstate}; # just to be sure
2048
2049 if ($conf->{cdrom}) {
2050 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
2051 $conf->{ide2} = $conf->{cdrom};
2052 delete $conf->{cdrom};
2053 }
2054
2055 # we do not use 'smp' any longer
2056 if ($conf->{sockets}) {
2057 delete $conf->{smp};
2058 } elsif ($conf->{smp}) {
2059 $conf->{sockets} = $conf->{smp};
2060 delete $conf->{cores};
2061 delete $conf->{smp};
2062 }
2063
2064 my $used_volids = {};
2065
2066 my $cleanup_config = sub {
2067 my ($cref, $pending, $snapname) = @_;
2068
2069 foreach my $key (keys %$cref) {
2070 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
2071 $key eq 'snapstate' || $key eq 'pending';
2072 my $value = $cref->{$key};
2073 if ($key eq 'delete') {
2074 die "propertry 'delete' is only allowed in [PENDING]\n"
2075 if !$pending;
2076 # fixme: check syntax?
2077 next;
2078 }
2079 eval { $value = check_type($key, $value); };
2080 die "unable to parse value of '$key' - $@" if $@;
2081
2082 $cref->{$key} = $value;
2083
2084 if (!$snapname && valid_drivename($key)) {
2085 my $drive = parse_drive($key, $value);
2086 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
2087 }
2088 }
2089 };
2090
2091 &$cleanup_config($conf);
2092
2093 &$cleanup_config($conf->{pending}, 1);
2094
2095 foreach my $snapname (keys %{$conf->{snapshots}}) {
2096 die "internal error" if $snapname eq 'pending';
2097 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2098 }
2099
2100 # remove 'unusedX' settings if we re-add a volume
2101 foreach my $key (keys %$conf) {
2102 my $value = $conf->{$key};
2103 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2104 delete $conf->{$key};
2105 }
2106 }
2107
2108 my $generate_raw_config = sub {
2109 my ($conf) = @_;
2110
2111 my $raw = '';
2112
2113 # add description as comment to top of file
2114 my $descr = $conf->{description} || '';
2115 foreach my $cl (split(/\n/, $descr)) {
2116 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2117 }
2118
2119 foreach my $key (sort keys %$conf) {
2120 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2121 $raw .= "$key: $conf->{$key}\n";
2122 }
2123 return $raw;
2124 };
2125
2126 my $raw = &$generate_raw_config($conf);
2127
2128 if (scalar(keys %{$conf->{pending}})){
2129 $raw .= "\n[PENDING]\n";
2130 $raw .= &$generate_raw_config($conf->{pending});
2131 }
2132
2133 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2134 $raw .= "\n[$snapname]\n";
2135 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2136 }
2137
2138 return $raw;
2139 }
2140
2141 sub update_config_nolock {
2142 my ($vmid, $conf, $skiplock) = @_;
2143
2144 check_lock($conf) if !$skiplock;
2145
2146 my $cfspath = cfs_config_path($vmid);
2147
2148 PVE::Cluster::cfs_write_file($cfspath, $conf);
2149 }
2150
2151 sub update_config {
2152 my ($vmid, $conf, $skiplock) = @_;
2153
2154 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2155 }
2156
2157 sub load_defaults {
2158
2159 my $res = {};
2160
2161 # we use static defaults from our JSON schema configuration
2162 foreach my $key (keys %$confdesc) {
2163 if (defined(my $default = $confdesc->{$key}->{default})) {
2164 $res->{$key} = $default;
2165 }
2166 }
2167
2168 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2169 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2170
2171 return $res;
2172 }
2173
2174 sub config_list {
2175 my $vmlist = PVE::Cluster::get_vmlist();
2176 my $res = {};
2177 return $res if !$vmlist || !$vmlist->{ids};
2178 my $ids = $vmlist->{ids};
2179
2180 foreach my $vmid (keys %$ids) {
2181 my $d = $ids->{$vmid};
2182 next if !$d->{node} || $d->{node} ne $nodename;
2183 next if !$d->{type} || $d->{type} ne 'qemu';
2184 $res->{$vmid}->{exists} = 1;
2185 }
2186 return $res;
2187 }
2188
2189 # test if VM uses local resources (to prevent migration)
2190 sub check_local_resources {
2191 my ($conf, $noerr) = @_;
2192
2193 my $loc_res = 0;
2194
2195 $loc_res = 1 if $conf->{hostusb}; # old syntax
2196 $loc_res = 1 if $conf->{hostpci}; # old syntax
2197
2198 foreach my $k (keys %$conf) {
2199 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2200 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2201 }
2202
2203 die "VM uses local resources\n" if $loc_res && !$noerr;
2204
2205 return $loc_res;
2206 }
2207
2208 # check if used storages are available on all nodes (use by migrate)
2209 sub check_storage_availability {
2210 my ($storecfg, $conf, $node) = @_;
2211
2212 foreach_drive($conf, sub {
2213 my ($ds, $drive) = @_;
2214
2215 my $volid = $drive->{file};
2216 return if !$volid;
2217
2218 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2219 return if !$sid;
2220
2221 # check if storage is available on both nodes
2222 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2223 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2224 });
2225 }
2226
2227 # list nodes where all VM images are available (used by has_feature API)
2228 sub shared_nodes {
2229 my ($conf, $storecfg) = @_;
2230
2231 my $nodelist = PVE::Cluster::get_nodelist();
2232 my $nodehash = { map { $_ => 1 } @$nodelist };
2233 my $nodename = PVE::INotify::nodename();
2234
2235 foreach_drive($conf, sub {
2236 my ($ds, $drive) = @_;
2237
2238 my $volid = $drive->{file};
2239 return if !$volid;
2240
2241 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2242 if ($storeid) {
2243 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2244 if ($scfg->{disable}) {
2245 $nodehash = {};
2246 } elsif (my $avail = $scfg->{nodes}) {
2247 foreach my $node (keys %$nodehash) {
2248 delete $nodehash->{$node} if !$avail->{$node};
2249 }
2250 } elsif (!$scfg->{shared}) {
2251 foreach my $node (keys %$nodehash) {
2252 delete $nodehash->{$node} if $node ne $nodename
2253 }
2254 }
2255 }
2256 });
2257
2258 return $nodehash
2259 }
2260
2261 sub check_lock {
2262 my ($conf) = @_;
2263
2264 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2265 }
2266
2267 sub check_cmdline {
2268 my ($pidfile, $pid) = @_;
2269
2270 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2271 if (defined($fh)) {
2272 my $line = <$fh>;
2273 $fh->close;
2274 return undef if !$line;
2275 my @param = split(/\0/, $line);
2276
2277 my $cmd = $param[0];
2278 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2279
2280 for (my $i = 0; $i < scalar (@param); $i++) {
2281 my $p = $param[$i];
2282 next if !$p;
2283 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2284 my $p = $param[$i+1];
2285 return 1 if $p && ($p eq $pidfile);
2286 return undef;
2287 }
2288 }
2289 }
2290 return undef;
2291 }
2292
2293 sub check_running {
2294 my ($vmid, $nocheck, $node) = @_;
2295
2296 my $filename = config_file($vmid, $node);
2297
2298 die "unable to find configuration file for VM $vmid - no such machine\n"
2299 if !$nocheck && ! -f $filename;
2300
2301 my $pidfile = pidfile_name($vmid);
2302
2303 if (my $fd = IO::File->new("<$pidfile")) {
2304 my $st = stat($fd);
2305 my $line = <$fd>;
2306 close($fd);
2307
2308 my $mtime = $st->mtime;
2309 if ($mtime > time()) {
2310 warn "file '$filename' modified in future\n";
2311 }
2312
2313 if ($line =~ m/^(\d+)$/) {
2314 my $pid = $1;
2315 if (check_cmdline($pidfile, $pid)) {
2316 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2317 return $pid;
2318 }
2319 }
2320 }
2321 }
2322
2323 return undef;
2324 }
2325
2326 sub vzlist {
2327
2328 my $vzlist = config_list();
2329
2330 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2331
2332 while (defined(my $de = $fd->read)) {
2333 next if $de !~ m/^(\d+)\.pid$/;
2334 my $vmid = $1;
2335 next if !defined($vzlist->{$vmid});
2336 if (my $pid = check_running($vmid)) {
2337 $vzlist->{$vmid}->{pid} = $pid;
2338 }
2339 }
2340
2341 return $vzlist;
2342 }
2343
2344 sub disksize {
2345 my ($storecfg, $conf) = @_;
2346
2347 my $bootdisk = $conf->{bootdisk};
2348 return undef if !$bootdisk;
2349 return undef if !valid_drivename($bootdisk);
2350
2351 return undef if !$conf->{$bootdisk};
2352
2353 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2354 return undef if !defined($drive);
2355
2356 return undef if drive_is_cdrom($drive);
2357
2358 my $volid = $drive->{file};
2359 return undef if !$volid;
2360
2361 return $drive->{size};
2362 }
2363
2364 my $last_proc_pid_stat;
2365
2366 # get VM status information
2367 # This must be fast and should not block ($full == false)
2368 # We only query KVM using QMP if $full == true (this can be slow)
2369 sub vmstatus {
2370 my ($opt_vmid, $full) = @_;
2371
2372 my $res = {};
2373
2374 my $storecfg = PVE::Storage::config();
2375
2376 my $list = vzlist();
2377 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2378
2379 my $cpucount = $cpuinfo->{cpus} || 1;
2380
2381 foreach my $vmid (keys %$list) {
2382 next if $opt_vmid && ($vmid ne $opt_vmid);
2383
2384 my $cfspath = cfs_config_path($vmid);
2385 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2386
2387 my $d = {};
2388 $d->{pid} = $list->{$vmid}->{pid};
2389
2390 # fixme: better status?
2391 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2392
2393 my $size = disksize($storecfg, $conf);
2394 if (defined($size)) {
2395 $d->{disk} = 0; # no info available
2396 $d->{maxdisk} = $size;
2397 } else {
2398 $d->{disk} = 0;
2399 $d->{maxdisk} = 0;
2400 }
2401
2402 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2403 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2404
2405 $d->{name} = $conf->{name} || "VM $vmid";
2406 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2407
2408 if ($conf->{balloon}) {
2409 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2410 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2411 }
2412
2413 $d->{uptime} = 0;
2414 $d->{cpu} = 0;
2415 $d->{mem} = 0;
2416
2417 $d->{netout} = 0;
2418 $d->{netin} = 0;
2419
2420 $d->{diskread} = 0;
2421 $d->{diskwrite} = 0;
2422
2423 $d->{template} = is_template($conf);
2424
2425 $res->{$vmid} = $d;
2426 }
2427
2428 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2429 foreach my $dev (keys %$netdev) {
2430 next if $dev !~ m/^tap([1-9]\d*)i/;
2431 my $vmid = $1;
2432 my $d = $res->{$vmid};
2433 next if !$d;
2434
2435 $d->{netout} += $netdev->{$dev}->{receive};
2436 $d->{netin} += $netdev->{$dev}->{transmit};
2437 }
2438
2439 my $ctime = gettimeofday;
2440
2441 foreach my $vmid (keys %$list) {
2442
2443 my $d = $res->{$vmid};
2444 my $pid = $d->{pid};
2445 next if !$pid;
2446
2447 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2448 next if !$pstat; # not running
2449
2450 my $used = $pstat->{utime} + $pstat->{stime};
2451
2452 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2453
2454 if ($pstat->{vsize}) {
2455 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2456 }
2457
2458 my $old = $last_proc_pid_stat->{$pid};
2459 if (!$old) {
2460 $last_proc_pid_stat->{$pid} = {
2461 time => $ctime,
2462 used => $used,
2463 cpu => 0,
2464 };
2465 next;
2466 }
2467
2468 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2469
2470 if ($dtime > 1000) {
2471 my $dutime = $used - $old->{used};
2472
2473 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2474 $last_proc_pid_stat->{$pid} = {
2475 time => $ctime,
2476 used => $used,
2477 cpu => $d->{cpu},
2478 };
2479 } else {
2480 $d->{cpu} = $old->{cpu};
2481 }
2482 }
2483
2484 return $res if !$full;
2485
2486 my $qmpclient = PVE::QMPClient->new();
2487
2488 my $ballooncb = sub {
2489 my ($vmid, $resp) = @_;
2490
2491 my $info = $resp->{'return'};
2492 return if !$info->{max_mem};
2493
2494 my $d = $res->{$vmid};
2495
2496 # use memory assigned to VM
2497 $d->{maxmem} = $info->{max_mem};
2498 $d->{balloon} = $info->{actual};
2499
2500 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2501 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2502 $d->{freemem} = $info->{free_mem};
2503 }
2504
2505 };
2506
2507 my $blockstatscb = sub {
2508 my ($vmid, $resp) = @_;
2509 my $data = $resp->{'return'} || [];
2510 my $totalrdbytes = 0;
2511 my $totalwrbytes = 0;
2512 for my $blockstat (@$data) {
2513 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2514 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2515 }
2516 $res->{$vmid}->{diskread} = $totalrdbytes;
2517 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2518 };
2519
2520 my $statuscb = sub {
2521 my ($vmid, $resp) = @_;
2522
2523 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2524 # this fails if ballon driver is not loaded, so this must be
2525 # the last commnand (following command are aborted if this fails).
2526 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2527
2528 my $status = 'unknown';
2529 if (!defined($status = $resp->{'return'}->{status})) {
2530 warn "unable to get VM status\n";
2531 return;
2532 }
2533
2534 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2535 };
2536
2537 foreach my $vmid (keys %$list) {
2538 next if $opt_vmid && ($vmid ne $opt_vmid);
2539 next if !$res->{$vmid}->{pid}; # not running
2540 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2541 }
2542
2543 $qmpclient->queue_execute(undef, 1);
2544
2545 foreach my $vmid (keys %$list) {
2546 next if $opt_vmid && ($vmid ne $opt_vmid);
2547 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2548 }
2549
2550 return $res;
2551 }
2552
2553 sub foreach_drive {
2554 my ($conf, $func) = @_;
2555
2556 foreach my $ds (keys %$conf) {
2557 next if !valid_drivename($ds);
2558
2559 my $drive = parse_drive($ds, $conf->{$ds});
2560 next if !$drive;
2561
2562 &$func($ds, $drive);
2563 }
2564 }
2565
2566 sub foreach_volid {
2567 my ($conf, $func) = @_;
2568
2569 my $volhash = {};
2570
2571 my $test_volid = sub {
2572 my ($volid, $is_cdrom) = @_;
2573
2574 return if !$volid;
2575
2576 $volhash->{$volid} = $is_cdrom || 0;
2577 };
2578
2579 foreach_drive($conf, sub {
2580 my ($ds, $drive) = @_;
2581 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2582 });
2583
2584 foreach my $snapname (keys %{$conf->{snapshots}}) {
2585 my $snap = $conf->{snapshots}->{$snapname};
2586 &$test_volid($snap->{vmstate}, 0);
2587 foreach_drive($snap, sub {
2588 my ($ds, $drive) = @_;
2589 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2590 });
2591 }
2592
2593 foreach my $volid (keys %$volhash) {
2594 &$func($volid, $volhash->{$volid});
2595 }
2596 }
2597
2598 sub vga_conf_has_spice {
2599 my ($vga) = @_;
2600
2601 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2602
2603 return $1 || 1;
2604 }
2605
2606 sub config_to_command {
2607 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2608
2609 my $cmd = [];
2610 my $globalFlags = [];
2611 my $machineFlags = [];
2612 my $rtcFlags = [];
2613 my $cpuFlags = [];
2614 my $devices = [];
2615 my $pciaddr = '';
2616 my $bridges = {};
2617 my $kvmver = kvm_user_version();
2618 my $vernum = 0; # unknown
2619 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2620 $vernum = $1*1000000+$2*1000;
2621 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2622 $vernum = $1*1000000+$2*1000+$3;
2623 }
2624
2625 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2626
2627 my $have_ovz = -f '/proc/vz/vestat';
2628
2629 my $q35 = machine_type_is_q35($conf);
2630 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2631
2632 push @$cmd, '/usr/bin/kvm';
2633
2634 push @$cmd, '-id', $vmid;
2635
2636 my $use_virtio = 0;
2637
2638 my $qmpsocket = qmp_socket($vmid);
2639 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2640 push @$cmd, '-mon', "chardev=qmp,mode=control";
2641
2642 my $socket = vnc_socket($vmid);
2643 push @$cmd, '-vnc', "unix:$socket,x509,password";
2644
2645 push @$cmd, '-pidfile' , pidfile_name($vmid);
2646
2647 push @$cmd, '-daemonize';
2648
2649 if ($conf->{smbios1}) {
2650 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2651 }
2652
2653 push @$cmd, '-object', "iothread,id=iothread0" if $conf->{iothread};
2654
2655 if ($q35) {
2656 # the q35 chipset support native usb2, so we enable usb controller
2657 # by default for this machine type
2658 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2659 } else {
2660 $pciaddr = print_pci_addr("piix3", $bridges);
2661 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2662
2663 my $use_usb2 = 0;
2664 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2665 next if !$conf->{"usb$i"};
2666 $use_usb2 = 1;
2667 }
2668 # include usb device config
2669 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2670 }
2671
2672 my $vga = $conf->{vga};
2673
2674 my $qxlnum = vga_conf_has_spice($vga);
2675 $vga = 'qxl' if $qxlnum;
2676
2677 if (!$vga) {
2678 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2679 $conf->{ostype} eq 'win7' ||
2680 $conf->{ostype} eq 'w2k8')) {
2681 $vga = 'std';
2682 } else {
2683 $vga = 'cirrus';
2684 }
2685 }
2686
2687 # enable absolute mouse coordinates (needed by vnc)
2688 my $tablet;
2689 if (defined($conf->{tablet})) {
2690 $tablet = $conf->{tablet};
2691 } else {
2692 $tablet = $defaults->{tablet};
2693 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2694 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2695 }
2696
2697 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2698
2699 # host pci devices
2700 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2701 my $d = parse_hostpci($conf->{"hostpci$i"});
2702 next if !$d;
2703
2704 my $pcie = $d->{pcie};
2705 if($pcie){
2706 die "q35 machine model is not enabled" if !$q35;
2707 $pciaddr = print_pcie_addr("hostpci$i");
2708 }else{
2709 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2710 }
2711
2712 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2713 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2714 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2715 if ($xvga && $xvga ne '') {
2716 push @$cpuFlags, 'kvm=off';
2717 $vga = 'none';
2718 }
2719 $driver = "vfio-pci" if $xvga ne '';
2720 my $pcidevices = $d->{pciid};
2721 my $multifunction = 1 if @$pcidevices > 1;
2722
2723 my $j=0;
2724 foreach my $pcidevice (@$pcidevices) {
2725
2726 my $id = "hostpci$i";
2727 $id .= ".$j" if $multifunction;
2728 my $addr = $pciaddr;
2729 $addr .= ".$j" if $multifunction;
2730 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2731
2732 if($j == 0){
2733 $devicestr .= "$rombar$xvga";
2734 $devicestr .= ",multifunction=on" if $multifunction;
2735 }
2736
2737 push @$devices, '-device', $devicestr;
2738 $j++;
2739 }
2740 }
2741
2742 # usb devices
2743 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2744 my $d = parse_usb_device($conf->{"usb$i"});
2745 next if !$d;
2746 if ($d->{vendorid} && $d->{productid}) {
2747 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2748 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2749 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2750 } elsif ($d->{spice}) {
2751 # usb redir support for spice
2752 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2753 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2754 }
2755 }
2756
2757 # serial devices
2758 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2759 if (my $path = $conf->{"serial$i"}) {
2760 if ($path eq 'socket') {
2761 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2762 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2763 push @$devices, '-device', "isa-serial,chardev=serial$i";
2764 } else {
2765 die "no such serial device\n" if ! -c $path;
2766 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2767 push @$devices, '-device', "isa-serial,chardev=serial$i";
2768 }
2769 }
2770 }
2771
2772 # parallel devices
2773 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2774 if (my $path = $conf->{"parallel$i"}) {
2775 die "no such parallel device\n" if ! -c $path;
2776 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2777 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2778 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2779 }
2780 }
2781
2782 my $vmname = $conf->{name} || "vm$vmid";
2783
2784 push @$cmd, '-name', $vmname;
2785
2786 my $sockets = 1;
2787 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2788 $sockets = $conf->{sockets} if $conf->{sockets};
2789
2790 my $cores = $conf->{cores} || 1;
2791
2792 my $maxcpus = $sockets * $cores;
2793
2794 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2795
2796 my $allowed_vcpus = $cpuinfo->{cpus};
2797
2798 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2799 if ($allowed_vcpus < $maxcpus);
2800
2801 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2802
2803 push @$cmd, '-nodefaults';
2804
2805 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2806
2807 my $bootindex_hash = {};
2808 my $i = 1;
2809 foreach my $o (split(//, $bootorder)) {
2810 $bootindex_hash->{$o} = $i*100;
2811 $i++;
2812 }
2813
2814 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2815
2816 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2817
2818 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2819
2820 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2821
2822 # time drift fix
2823 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2824
2825 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2826 my $useLocaltime = $conf->{localtime};
2827
2828 if (my $ost = $conf->{ostype}) {
2829 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2830
2831 if ($ost =~ m/^w/) { # windows
2832 $useLocaltime = 1 if !defined($conf->{localtime});
2833
2834 # use time drift fix when acpi is enabled
2835 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2836 $tdf = 1 if !defined($conf->{tdf});
2837 }
2838 }
2839
2840 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2841 $ost eq 'wvista') {
2842 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2843 push @$cmd, '-no-hpet';
2844 #push @$cpuFlags , 'hv_vapic" if !$nokvm; #fixme, my win2008R2 hang at boot with this
2845 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2846 }
2847
2848 if ($ost eq 'win7' || $ost eq 'win8') {
2849 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2850 }
2851 }
2852
2853 push @$rtcFlags, 'driftfix=slew' if $tdf;
2854
2855 if ($nokvm) {
2856 push @$machineFlags, 'accel=tcg';
2857 } else {
2858 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2859 }
2860
2861 my $machine_type = $forcemachine || $conf->{machine};
2862 if ($machine_type) {
2863 push @$machineFlags, "type=${machine_type}";
2864 }
2865
2866 if ($conf->{startdate}) {
2867 push @$rtcFlags, "base=$conf->{startdate}";
2868 } elsif ($useLocaltime) {
2869 push @$rtcFlags, 'base=localtime';
2870 }
2871
2872 my $cpu = $nokvm ? "qemu64" : "kvm64";
2873 $cpu = $conf->{cpu} if $conf->{cpu};
2874
2875 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2876
2877 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2878
2879 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2880
2881 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2882
2883 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2884
2885 # Note: enforce needs kernel 3.10, so we do not use it for now
2886 # push @$cmd, '-cpu', "$cpu,enforce";
2887 push @$cmd, '-cpu', $cpu;
2888
2889 my $memory = $conf->{memory} || $defaults->{memory};
2890 my $static_memory = 0;
2891 my $dimm_memory = 0;
2892
2893 if ($hotplug_features->{memory}) {
2894 die "Numa need to be enabled for memory hotplug" if !$conf->{numa};
2895 die "Total memory is bigger than $MAX_MEM MB" if $memory > $MAX_MEM;
2896 die "memory should be a multiple of 128!" if ($memory % 128 != 0);
2897 $static_memory = $STATICMEM;
2898 die "minimum memory must be $static_memory"."MB" if($memory < $static_memory);
2899 $dimm_memory = $memory - $static_memory;
2900 push @$cmd, '-m', "size=".$static_memory.",slots=255,maxmem=".$MAX_MEM."M";
2901
2902 } else {
2903
2904 $static_memory = $memory;
2905 push @$cmd, '-m', $static_memory;
2906 }
2907
2908 if ($conf->{numa}) {
2909
2910 my $numa_totalmemory = undef;
2911 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2912 next if !$conf->{"numa$i"};
2913 my $numa = parse_numa($conf->{"numa$i"});
2914 next if !$numa;
2915 # memory
2916 die "missing numa node$i memory value\n" if !$numa->{memory};
2917 my $numa_memory = $numa->{memory};
2918 $numa_totalmemory += $numa_memory;
2919 my $numa_object = "memory-backend-ram,id=ram-node$i,size=$numa_memory"."M";
2920
2921 # cpus
2922 my $cpus_start = $numa->{cpus}->{start};
2923 die "missing numa node$i cpus\n" if !defined($cpus_start);
2924 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2925 my $cpus = $cpus_start;
2926 if (defined($cpus_end)) {
2927 $cpus .= "-$cpus_end";
2928 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2929 }
2930
2931 # hostnodes
2932 my $hostnodes_start = $numa->{hostnodes}->{start};
2933 if (defined($hostnodes_start)) {
2934 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2935 my $hostnodes = $hostnodes_start;
2936 if (defined($hostnodes_end)) {
2937 $hostnodes .= "-$hostnodes_end";
2938 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2939 }
2940
2941 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2942 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2943 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2944 }
2945
2946 # policy
2947 my $policy = $numa->{policy};
2948 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2949 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2950 }
2951
2952 push @$cmd, '-object', $numa_object;
2953 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2954 }
2955
2956 die "total memory for NUMA nodes must be equal to vm static memory\n"
2957 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2958
2959 #if no custom tology, we split memory and cores across numa nodes
2960 if(!$numa_totalmemory) {
2961
2962 my $numa_memory = ($static_memory / $sockets) . "M";
2963
2964 for (my $i = 0; $i < $sockets; $i++) {
2965
2966 my $cpustart = ($cores * $i);
2967 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2968 my $cpus = $cpustart;
2969 $cpus .= "-$cpuend" if $cpuend;
2970
2971 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2972 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2973 }
2974 }
2975 }
2976
2977 if ($hotplug_features->{memory}) {
2978 my $dimm_id = 0;
2979 my $dimm_size = 512;
2980 my $current_size = $static_memory;
2981 for (my $j = 0; $j < 8; $j++) {
2982 for (my $i = 0; $i < 32; $i++) {
2983 my $name = "dimm${dimm_id}";
2984 $dimm_id++;
2985 last if $current_size >= $memory;
2986 my $numanode = $i % $sockets;
2987 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=$dimm_size"."M";
2988 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
2989 $current_size += $dimm_size;
2990 #if dimm_memory is not aligned to dimm map
2991 if($current_size > $memory) {
2992 $conf->{memory} = $current_size;
2993 update_config_nolock($vmid, $conf, 1);
2994 }
2995 }
2996 $dimm_size *= 2;
2997 }
2998 }
2999
3000 push @$cmd, '-S' if $conf->{freeze};
3001
3002 # set keyboard layout
3003 my $kb = $conf->{keyboard} || $defaults->{keyboard};
3004 push @$cmd, '-k', $kb if $kb;
3005
3006 # enable sound
3007 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
3008 #push @$cmd, '-soundhw', 'es1370';
3009 #push @$cmd, '-soundhw', $soundhw if $soundhw;
3010
3011 if($conf->{agent}) {
3012 my $qgasocket = qmp_socket($vmid, 1);
3013 my $pciaddr = print_pci_addr("qga0", $bridges);
3014 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
3015 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
3016 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
3017 }
3018
3019 my $spice_port;
3020
3021 if ($qxlnum) {
3022 if ($qxlnum > 1) {
3023 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
3024 for(my $i = 1; $i < $qxlnum; $i++){
3025 my $pciaddr = print_pci_addr("vga$i", $bridges);
3026 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
3027 }
3028 } else {
3029 # assume other OS works like Linux
3030 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
3031 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
3032 }
3033 }
3034
3035 my $pciaddr = print_pci_addr("spice", $bridges);
3036
3037 $spice_port = PVE::Tools::next_spice_port();
3038
3039 push @$devices, '-spice', "tls-port=${spice_port},addr=127.0.0.1,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3040
3041 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3042 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3043 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3044 }
3045
3046 # enable balloon by default, unless explicitly disabled
3047 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3048 $pciaddr = print_pci_addr("balloon0", $bridges);
3049 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3050 }
3051
3052 if ($conf->{watchdog}) {
3053 my $wdopts = parse_watchdog($conf->{watchdog});
3054 $pciaddr = print_pci_addr("watchdog", $bridges);
3055 my $watchdog = $wdopts->{model} || 'i6300esb';
3056 push @$devices, '-device', "$watchdog$pciaddr";
3057 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3058 }
3059
3060 my $vollist = [];
3061 my $scsicontroller = {};
3062 my $ahcicontroller = {};
3063 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3064
3065 # Add iscsi initiator name if available
3066 if (my $initiator = get_initiator_name()) {
3067 push @$devices, '-iscsi', "initiator-name=$initiator";
3068 }
3069
3070 foreach_drive($conf, sub {
3071 my ($ds, $drive) = @_;
3072
3073 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3074 push @$vollist, $drive->{file};
3075 }
3076
3077 $use_virtio = 1 if $ds =~ m/^virtio/;
3078
3079 if (drive_is_cdrom ($drive)) {
3080 if ($bootindex_hash->{d}) {
3081 $drive->{bootindex} = $bootindex_hash->{d};
3082 $bootindex_hash->{d} += 1;
3083 }
3084 } else {
3085 if ($bootindex_hash->{c}) {
3086 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3087 $bootindex_hash->{c} += 1;
3088 }
3089 }
3090
3091 if ($drive->{interface} eq 'scsi') {
3092
3093 my $maxdev = ($scsihw !~ m/^lsi/) ? 256 : 7;
3094 my $controller = int($drive->{index} / $maxdev);
3095 $pciaddr = print_pci_addr("scsihw$controller", $bridges);
3096 push @$devices, '-device', "$scsihw,id=scsihw$controller$pciaddr" if !$scsicontroller->{$controller};
3097 $scsicontroller->{$controller}=1;
3098 }
3099
3100 if ($drive->{interface} eq 'sata') {
3101 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3102 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3103 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3104 $ahcicontroller->{$controller}=1;
3105 }
3106
3107 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3108 push @$devices, '-drive',$drive_cmd;
3109 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3110 });
3111
3112 for (my $i = 0; $i < $MAX_NETS; $i++) {
3113 next if !$conf->{"net$i"};
3114 my $d = parse_net($conf->{"net$i"});
3115 next if !$d;
3116
3117 $use_virtio = 1 if $d->{model} eq 'virtio';
3118
3119 if ($bootindex_hash->{n}) {
3120 $d->{bootindex} = $bootindex_hash->{n};
3121 $bootindex_hash->{n} += 1;
3122 }
3123
3124 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3125 push @$devices, '-netdev', $netdevfull;
3126
3127 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3128 push @$devices, '-device', $netdevicefull;
3129 }
3130
3131 if (!$q35) {
3132 # add pci bridges
3133 while (my ($k, $v) = each %$bridges) {
3134 $pciaddr = print_pci_addr("pci.$k");
3135 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3136 }
3137 }
3138
3139 # hack: virtio with fairsched is unreliable, so we do not use fairsched
3140 # when the VM uses virtio devices.
3141 if (!$use_virtio && $have_ovz) {
3142
3143 my $cpuunits = defined($conf->{cpuunits}) ?
3144 $conf->{cpuunits} : $defaults->{cpuunits};
3145
3146 push @$cmd, '-cpuunits', $cpuunits if $cpuunits;
3147
3148 # fixme: cpulimit is currently ignored
3149 #push @$cmd, '-cpulimit', $conf->{cpulimit} if $conf->{cpulimit};
3150 }
3151
3152 # add custom args
3153 if ($conf->{args}) {
3154 my $aa = PVE::Tools::split_args($conf->{args});
3155 push @$cmd, @$aa;
3156 }
3157
3158 push @$cmd, @$devices;
3159 push @$cmd, '-rtc', join(',', @$rtcFlags)
3160 if scalar(@$rtcFlags);
3161 push @$cmd, '-machine', join(',', @$machineFlags)
3162 if scalar(@$machineFlags);
3163 push @$cmd, '-global', join(',', @$globalFlags)
3164 if scalar(@$globalFlags);
3165
3166 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3167 }
3168
3169 sub vnc_socket {
3170 my ($vmid) = @_;
3171 return "${var_run_tmpdir}/$vmid.vnc";
3172 }
3173
3174 sub spice_port {
3175 my ($vmid) = @_;
3176
3177 my $res = vm_mon_cmd($vmid, 'query-spice');
3178
3179 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3180 }
3181
3182 sub qmp_socket {
3183 my ($vmid, $qga) = @_;
3184 my $sockettype = $qga ? 'qga' : 'qmp';
3185 return "${var_run_tmpdir}/$vmid.$sockettype";
3186 }
3187
3188 sub pidfile_name {
3189 my ($vmid) = @_;
3190 return "${var_run_tmpdir}/$vmid.pid";
3191 }
3192
3193 sub vm_devices_list {
3194 my ($vmid) = @_;
3195
3196 my $res = vm_mon_cmd($vmid, 'query-pci');
3197 my $devices = {};
3198 foreach my $pcibus (@$res) {
3199 foreach my $device (@{$pcibus->{devices}}) {
3200 next if !$device->{'qdev_id'};
3201 $devices->{$device->{'qdev_id'}} = 1;
3202 }
3203 }
3204
3205 my $resblock = vm_mon_cmd($vmid, 'query-block');
3206 foreach my $block (@$resblock) {
3207 if($block->{device} =~ m/^drive-(\S+)/){
3208 $devices->{$1} = 1;
3209 }
3210 }
3211
3212 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3213 foreach my $mice (@$resmice) {
3214 if ($mice->{name} eq 'QEMU HID Tablet') {
3215 $devices->{tablet} = 1;
3216 last;
3217 }
3218 }
3219
3220 return $devices;
3221 }
3222
3223 sub vm_deviceplug {
3224 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3225
3226 my $q35 = machine_type_is_q35($conf);
3227
3228 my $devices_list = vm_devices_list($vmid);
3229 return 1 if defined($devices_list->{$deviceid});
3230
3231 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3232
3233 if ($deviceid eq 'tablet') {
3234
3235 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3236
3237 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3238
3239 qemu_driveadd($storecfg, $vmid, $device);
3240 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3241
3242 qemu_deviceadd($vmid, $devicefull);
3243 eval { qemu_deviceaddverify($vmid, $deviceid); };
3244 if (my $err = $@) {
3245 eval { qemu_drivedel($vmid, $deviceid); };
3246 warn $@ if $@;
3247 die $err;
3248 }
3249
3250 } elsif ($deviceid =~ m/^(scsihw)(\d+)$/) {
3251
3252 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3253 my $pciaddr = print_pci_addr($deviceid);
3254 my $devicefull = "$scsihw,id=$deviceid$pciaddr";
3255
3256 qemu_deviceadd($vmid, $devicefull);
3257 qemu_deviceaddverify($vmid, $deviceid);
3258
3259 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3260
3261 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3262 qemu_driveadd($storecfg, $vmid, $device);
3263
3264 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3265 eval { qemu_deviceadd($vmid, $devicefull); };
3266 if (my $err = $@) {
3267 eval { qemu_drivedel($vmid, $deviceid); };
3268 warn $@ if $@;
3269 die $err;
3270 }
3271
3272 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3273
3274 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3275 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3276 qemu_deviceadd($vmid, $netdevicefull);
3277 eval { qemu_deviceaddverify($vmid, $deviceid); };
3278 if (my $err = $@) {
3279 eval { qemu_netdevdel($vmid, $deviceid); };
3280 warn $@ if $@;
3281 die $err;
3282 }
3283
3284 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3285
3286 my $bridgeid = $2;
3287 my $pciaddr = print_pci_addr($deviceid);
3288 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3289
3290 qemu_deviceadd($vmid, $devicefull);
3291 qemu_deviceaddverify($vmid, $deviceid);
3292
3293 } else {
3294 die "can't hotplug device '$deviceid'\n";
3295 }
3296
3297 return 1;
3298 }
3299
3300 # fixme: this should raise exceptions on error!
3301 sub vm_deviceunplug {
3302 my ($vmid, $conf, $deviceid) = @_;
3303
3304 my $devices_list = vm_devices_list($vmid);
3305 return 1 if !defined($devices_list->{$deviceid});
3306
3307 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3308
3309 if ($deviceid eq 'tablet') {
3310
3311 qemu_devicedel($vmid, $deviceid);
3312
3313 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3314
3315 qemu_devicedel($vmid, $deviceid);
3316 qemu_devicedelverify($vmid, $deviceid);
3317 qemu_drivedel($vmid, $deviceid);
3318
3319 } elsif ($deviceid =~ m/^(lsi)(\d+)$/) {
3320
3321 qemu_devicedel($vmid, $deviceid);
3322
3323 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3324
3325 qemu_devicedel($vmid, $deviceid);
3326 qemu_drivedel($vmid, $deviceid);
3327
3328 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3329
3330 qemu_devicedel($vmid, $deviceid);
3331 qemu_devicedelverify($vmid, $deviceid);
3332 qemu_netdevdel($vmid, $deviceid);
3333
3334 } else {
3335 die "can't unplug device '$deviceid'\n";
3336 }
3337
3338 return 1;
3339 }
3340
3341 sub qemu_deviceadd {
3342 my ($vmid, $devicefull) = @_;
3343
3344 $devicefull = "driver=".$devicefull;
3345 my %options = split(/[=,]/, $devicefull);
3346
3347 vm_mon_cmd($vmid, "device_add" , %options);
3348 }
3349
3350 sub qemu_devicedel {
3351 my ($vmid, $deviceid) = @_;
3352
3353 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3354 }
3355
3356 sub qemu_objectadd {
3357 my($vmid, $objectid, $qomtype) = @_;
3358
3359 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3360
3361 return 1;
3362 }
3363
3364 sub qemu_objectdel {
3365 my($vmid, $objectid) = @_;
3366
3367 vm_mon_cmd($vmid, "object-del", id => $objectid);
3368
3369 return 1;
3370 }
3371
3372 sub qemu_driveadd {
3373 my ($storecfg, $vmid, $device) = @_;
3374
3375 my $drive = print_drive_full($storecfg, $vmid, $device);
3376 my $ret = vm_human_monitor_command($vmid, "drive_add auto $drive");
3377
3378 # If the command succeeds qemu prints: "OK"
3379 return 1 if $ret =~ m/OK/s;
3380
3381 die "adding drive failed: $ret\n";
3382 }
3383
3384 sub qemu_drivedel {
3385 my($vmid, $deviceid) = @_;
3386
3387 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3388 $ret =~ s/^\s+//;
3389
3390 return 1 if $ret eq "";
3391
3392 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3393 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3394
3395 die "deleting drive $deviceid failed : $ret\n";
3396 }
3397
3398 sub qemu_deviceaddverify {
3399 my ($vmid, $deviceid) = @_;
3400
3401 for (my $i = 0; $i <= 5; $i++) {
3402 my $devices_list = vm_devices_list($vmid);
3403 return 1 if defined($devices_list->{$deviceid});
3404 sleep 1;
3405 }
3406
3407 die "error on hotplug device '$deviceid'\n";
3408 }
3409
3410
3411 sub qemu_devicedelverify {
3412 my ($vmid, $deviceid) = @_;
3413
3414 # need to verify that the device is correctly removed as device_del
3415 # is async and empty return is not reliable
3416
3417 for (my $i = 0; $i <= 5; $i++) {
3418 my $devices_list = vm_devices_list($vmid);
3419 return 1 if !defined($devices_list->{$deviceid});
3420 sleep 1;
3421 }
3422
3423 die "error on hot-unplugging device '$deviceid'\n";
3424 }
3425
3426 sub qemu_findorcreatescsihw {
3427 my ($storecfg, $conf, $vmid, $device) = @_;
3428
3429 my $maxdev = ($conf->{scsihw} && ($conf->{scsihw} !~ m/^lsi/)) ? 256 : 7;
3430 my $controller = int($device->{index} / $maxdev);
3431 my $scsihwid="scsihw$controller";
3432 my $devices_list = vm_devices_list($vmid);
3433
3434 if(!defined($devices_list->{$scsihwid})) {
3435 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid);
3436 }
3437
3438 return 1;
3439 }
3440
3441 sub qemu_add_pci_bridge {
3442 my ($storecfg, $conf, $vmid, $device) = @_;
3443
3444 my $bridges = {};
3445
3446 my $bridgeid;
3447
3448 print_pci_addr($device, $bridges);
3449
3450 while (my ($k, $v) = each %$bridges) {
3451 $bridgeid = $k;
3452 }
3453 return 1 if !defined($bridgeid) || $bridgeid < 1;
3454
3455 my $bridge = "pci.$bridgeid";
3456 my $devices_list = vm_devices_list($vmid);
3457
3458 if (!defined($devices_list->{$bridge})) {
3459 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3460 }
3461
3462 return 1;
3463 }
3464
3465 sub qemu_set_link_status {
3466 my ($vmid, $device, $up) = @_;
3467
3468 vm_mon_cmd($vmid, "set_link", name => $device,
3469 up => $up ? JSON::true : JSON::false);
3470 }
3471
3472 sub qemu_netdevadd {
3473 my ($vmid, $conf, $device, $deviceid) = @_;
3474
3475 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3476 my %options = split(/[=,]/, $netdev);
3477
3478 vm_mon_cmd($vmid, "netdev_add", %options);
3479 return 1;
3480 }
3481
3482 sub qemu_netdevdel {
3483 my ($vmid, $deviceid) = @_;
3484
3485 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3486 }
3487
3488 sub qemu_cpu_hotplug {
3489 my ($vmid, $conf, $vcpus) = @_;
3490
3491 my $sockets = 1;
3492 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3493 $sockets = $conf->{sockets} if $conf->{sockets};
3494 my $cores = $conf->{cores} || 1;
3495 my $maxcpus = $sockets * $cores;
3496
3497 $vcpus = $maxcpus if !$vcpus;
3498
3499 die "you can't add more vcpus than maxcpus\n"
3500 if $vcpus > $maxcpus;
3501
3502 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3503 die "online cpu unplug is not yet possible\n"
3504 if $vcpus < $currentvcpus;
3505
3506 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3507 die "vcpus in running vm is different than configuration\n"
3508 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3509
3510 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3511 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3512 }
3513 }
3514
3515 sub qemu_memory_hotplug {
3516 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3517
3518 return $value if !check_running($vmid);
3519
3520 my $memory = $conf->{memory} || $defaults->{memory};
3521 $value = $defaults->{memory} if !$value;
3522 return $value if $value == $memory;
3523
3524 my $static_memory = $STATICMEM;
3525 my $dimm_memory = $memory - $static_memory;
3526
3527 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3528 die "memory unplug is not yet available" if $value < $memory;
3529 die "memory should be a multiple of 128!\n" if ($value % 128 != 0);
3530 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3531
3532
3533 my $sockets = 1;
3534 $sockets = $conf->{sockets} if $conf->{sockets};
3535
3536 my $dimm_id = 0;
3537 my $current_size = $static_memory;
3538 my $dimm_size = 512;
3539 for (my $j = 0; $j < 8; $j++) {
3540 for (my $i = 0; $i < 32; $i++) {
3541 my $name = "dimm${dimm_id}";
3542 $dimm_id++;
3543 $current_size += $dimm_size;
3544 next if $current_size <= $memory;
3545 my $numanode = $i % $sockets;
3546
3547 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3548 if (my $err = $@) {
3549 eval { qemu_objectdel($vmid, "mem-$name"); };
3550 die $err;
3551 }
3552
3553 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3554 if (my $err = $@) {
3555 eval { qemu_objectdel($vmid, "mem-$name"); };
3556 die $err;
3557 }
3558 #update conf after each succesful module hotplug
3559 $conf->{$opt} = $current_size;
3560 update_config_nolock($vmid, $conf, 1);
3561
3562 return $current_size if $current_size >= $value;
3563 }
3564 $dimm_size *= 2;
3565 }
3566
3567 }
3568
3569 sub qemu_block_set_io_throttle {
3570 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3571
3572 return if !check_running($vmid) ;
3573
3574 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($iops_wr));
3575
3576 }
3577
3578 # old code, only used to shutdown old VM after update
3579 sub __read_avail {
3580 my ($fh, $timeout) = @_;
3581
3582 my $sel = new IO::Select;
3583 $sel->add($fh);
3584
3585 my $res = '';
3586 my $buf;
3587
3588 my @ready;
3589 while (scalar (@ready = $sel->can_read($timeout))) {
3590 my $count;
3591 if ($count = $fh->sysread($buf, 8192)) {
3592 if ($buf =~ /^(.*)\(qemu\) $/s) {
3593 $res .= $1;
3594 last;
3595 } else {
3596 $res .= $buf;
3597 }
3598 } else {
3599 if (!defined($count)) {
3600 die "$!\n";
3601 }
3602 last;
3603 }
3604 }
3605
3606 die "monitor read timeout\n" if !scalar(@ready);
3607
3608 return $res;
3609 }
3610
3611 # old code, only used to shutdown old VM after update
3612 sub vm_monitor_command {
3613 my ($vmid, $cmdstr, $nocheck) = @_;
3614
3615 my $res;
3616
3617 eval {
3618 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3619
3620 my $sname = "${var_run_tmpdir}/$vmid.mon";
3621
3622 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3623 die "unable to connect to VM $vmid socket - $!\n";
3624
3625 my $timeout = 3;
3626
3627 # hack: migrate sometime blocks the monitor (when migrate_downtime
3628 # is set)
3629 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3630 $timeout = 60*60; # 1 hour
3631 }
3632
3633 # read banner;
3634 my $data = __read_avail($sock, $timeout);
3635
3636 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3637 die "got unexpected qemu monitor banner\n";
3638 }
3639
3640 my $sel = new IO::Select;
3641 $sel->add($sock);
3642
3643 if (!scalar(my @ready = $sel->can_write($timeout))) {
3644 die "monitor write error - timeout";
3645 }
3646
3647 my $fullcmd = "$cmdstr\r";
3648
3649 # syslog('info', "VM $vmid monitor command: $cmdstr");
3650
3651 my $b;
3652 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3653 die "monitor write error - $!";
3654 }
3655
3656 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3657
3658 $timeout = 20;
3659
3660 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3661 $timeout = 60*60; # 1 hour
3662 } elsif ($cmdstr =~ m/^(eject|change)/) {
3663 $timeout = 60; # note: cdrom mount command is slow
3664 }
3665 if ($res = __read_avail($sock, $timeout)) {
3666
3667 my @lines = split("\r?\n", $res);
3668
3669 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3670
3671 $res = join("\n", @lines);
3672 $res .= "\n";
3673 }
3674 };
3675
3676 my $err = $@;
3677
3678 if ($err) {
3679 syslog("err", "VM $vmid monitor command failed - $err");
3680 die $err;
3681 }
3682
3683 return $res;
3684 }
3685
3686 sub qemu_block_resize {
3687 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3688
3689 my $running = check_running($vmid);
3690
3691 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3692
3693 return if !$running;
3694
3695 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3696
3697 }
3698
3699 sub qemu_volume_snapshot {
3700 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3701
3702 my $running = check_running($vmid);
3703
3704 return if !PVE::Storage::volume_snapshot($storecfg, $volid, $snap, $running);
3705
3706 return if !$running;
3707
3708 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3709
3710 }
3711
3712 sub qemu_volume_snapshot_delete {
3713 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3714
3715 my $running = check_running($vmid);
3716
3717 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3718
3719 return if !$running;
3720
3721 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3722 }
3723
3724 sub set_migration_caps {
3725 my ($vmid) = @_;
3726
3727 my $cap_ref = [];
3728
3729 my $enabled_cap = {
3730 "auto-converge" => 1,
3731 "xbzrle" => 0,
3732 "x-rdma-pin-all" => 0,
3733 "zero-blocks" => 0,
3734 };
3735
3736 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3737
3738 for my $supported_capability (@$supported_capabilities) {
3739 push @$cap_ref, {
3740 capability => $supported_capability->{capability},
3741 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3742 };
3743 }
3744
3745 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3746 }
3747
3748 my $fast_plug_option = {
3749 'name' => 1,
3750 'hotplug' => 1,
3751 'onboot' => 1,
3752 'shares' => 1,
3753 'startup' => 1,
3754 };
3755
3756 # hotplug changes in [PENDING]
3757 # $selection hash can be used to only apply specified options, for
3758 # example: { cores => 1 } (only apply changed 'cores')
3759 # $errors ref is used to return error messages
3760 sub vmconfig_hotplug_pending {
3761 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3762
3763 my $defaults = load_defaults();
3764
3765 # commit values which do not have any impact on running VM first
3766 # Note: those option cannot raise errors, we we do not care about
3767 # $selection and always apply them.
3768
3769 my $add_error = sub {
3770 my ($opt, $msg) = @_;
3771 $errors->{$opt} = "hotplug problem - $msg";
3772 };
3773
3774 my $changes = 0;
3775 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3776 if ($fast_plug_option->{$opt}) {
3777 $conf->{$opt} = $conf->{pending}->{$opt};
3778 delete $conf->{pending}->{$opt};
3779 $changes = 1;
3780 }
3781 }
3782
3783 if ($changes) {
3784 update_config_nolock($vmid, $conf, 1);
3785 $conf = load_config($vmid); # update/reload
3786 }
3787
3788 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3789
3790 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3791 foreach my $opt (@delete) {
3792 next if $selection && !$selection->{$opt};
3793 eval {
3794 if ($opt eq 'tablet') {
3795 die "skip\n" if !$hotplug_features->{usb};
3796 if ($defaults->{tablet}) {
3797 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3798 } else {
3799 vm_deviceunplug($vmid, $conf, $opt);
3800 }
3801 } elsif ($opt eq 'vcpus') {
3802 die "skip\n" if !$hotplug_features->{cpu};
3803 qemu_cpu_hotplug($vmid, $conf, undef);
3804 } elsif ($opt eq 'balloon') {
3805 # enable balloon device is not hotpluggable
3806 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3807 } elsif ($fast_plug_option->{$opt}) {
3808 # do nothing
3809 } elsif ($opt =~ m/^net(\d+)$/) {
3810 die "skip\n" if !$hotplug_features->{network};
3811 vm_deviceunplug($vmid, $conf, $opt);
3812 } elsif (valid_drivename($opt)) {
3813 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3814 vm_deviceunplug($vmid, $conf, $opt);
3815 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3816 } elsif ($opt =~ m/^memory$/) {
3817 die "skip\n" if !$hotplug_features->{memory};
3818 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3819 } else {
3820 die "skip\n";
3821 }
3822 };
3823 if (my $err = $@) {
3824 &$add_error($opt, $err) if $err ne "skip\n";
3825 } else {
3826 # save new config if hotplug was successful
3827 delete $conf->{$opt};
3828 vmconfig_undelete_pending_option($conf, $opt);
3829 update_config_nolock($vmid, $conf, 1);
3830 $conf = load_config($vmid); # update/reload
3831 }
3832 }
3833
3834 foreach my $opt (keys %{$conf->{pending}}) {
3835 next if $selection && !$selection->{$opt};
3836 my $value = $conf->{pending}->{$opt};
3837 eval {
3838 if ($opt eq 'tablet') {
3839 die "skip\n" if !$hotplug_features->{usb};
3840 if ($value == 1) {
3841 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3842 } elsif ($value == 0) {
3843 vm_deviceunplug($vmid, $conf, $opt);
3844 }
3845 } elsif ($opt eq 'vcpus') {
3846 die "skip\n" if !$hotplug_features->{cpu};
3847 qemu_cpu_hotplug($vmid, $conf, $value);
3848 } elsif ($opt eq 'balloon') {
3849 # enable/disable balloning device is not hotpluggable
3850 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3851 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3852 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3853
3854 # allow manual ballooning if shares is set to zero
3855 if (!(defined($conf->{shares}) && ($conf->{shares} == 0))) {
3856 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3857 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3858 }
3859 } elsif ($opt =~ m/^net(\d+)$/) {
3860 # some changes can be done without hotplug
3861 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3862 $vmid, $opt, $value);
3863 } elsif (valid_drivename($opt)) {
3864 # some changes can be done without hotplug
3865 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3866 $vmid, $opt, $value, 1);
3867 } elsif ($opt =~ m/^memory$/) { #dimms
3868 die "skip\n" if !$hotplug_features->{memory};
3869 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3870 } else {
3871 die "skip\n"; # skip non-hot-pluggable options
3872 }
3873 };
3874 if (my $err = $@) {
3875 &$add_error($opt, $err) if $err ne "skip\n";
3876 } else {
3877 # save new config if hotplug was successful
3878 $conf->{$opt} = $value;
3879 delete $conf->{pending}->{$opt};
3880 update_config_nolock($vmid, $conf, 1);
3881 $conf = load_config($vmid); # update/reload
3882 }
3883 }
3884 }
3885
3886 sub vmconfig_apply_pending {
3887 my ($vmid, $conf, $storecfg) = @_;
3888
3889 # cold plug
3890
3891 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3892 foreach my $opt (@delete) { # delete
3893 die "internal error" if $opt =~ m/^unused/;
3894 $conf = load_config($vmid); # update/reload
3895 if (!defined($conf->{$opt})) {
3896 vmconfig_undelete_pending_option($conf, $opt);
3897 update_config_nolock($vmid, $conf, 1);
3898 } elsif (valid_drivename($opt)) {
3899 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3900 vmconfig_undelete_pending_option($conf, $opt);
3901 delete $conf->{$opt};
3902 update_config_nolock($vmid, $conf, 1);
3903 } else {
3904 vmconfig_undelete_pending_option($conf, $opt);
3905 delete $conf->{$opt};
3906 update_config_nolock($vmid, $conf, 1);
3907 }
3908 }
3909
3910 $conf = load_config($vmid); # update/reload
3911
3912 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3913 $conf = load_config($vmid); # update/reload
3914
3915 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3916 # skip if nothing changed
3917 } elsif (valid_drivename($opt)) {
3918 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3919 if defined($conf->{$opt});
3920 $conf->{$opt} = $conf->{pending}->{$opt};
3921 } else {
3922 $conf->{$opt} = $conf->{pending}->{$opt};
3923 }
3924
3925 delete $conf->{pending}->{$opt};
3926 update_config_nolock($vmid, $conf, 1);
3927 }
3928 }
3929
3930 my $safe_num_ne = sub {
3931 my ($a, $b) = @_;
3932
3933 return 0 if !defined($a) && !defined($b);
3934 return 1 if !defined($a);
3935 return 1 if !defined($b);
3936
3937 return $a != $b;
3938 };
3939
3940 my $safe_string_ne = sub {
3941 my ($a, $b) = @_;
3942
3943 return 0 if !defined($a) && !defined($b);
3944 return 1 if !defined($a);
3945 return 1 if !defined($b);
3946
3947 return $a ne $b;
3948 };
3949
3950 sub vmconfig_update_net {
3951 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
3952
3953 my $newnet = parse_net($value);
3954
3955 if ($conf->{$opt}) {
3956 my $oldnet = parse_net($conf->{$opt});
3957
3958 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
3959 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
3960 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
3961 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
3962
3963 # for non online change, we try to hot-unplug
3964 die "skip\n" if !$hotplug;
3965 vm_deviceunplug($vmid, $conf, $opt);
3966 } else {
3967
3968 die "internal error" if $opt !~ m/net(\d+)/;
3969 my $iface = "tap${vmid}i$1";
3970
3971 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
3972 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
3973 }
3974
3975 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
3976 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
3977 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
3978 PVE::Network::tap_unplug($iface);
3979 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
3980 }
3981
3982 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
3983 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
3984 }
3985
3986 return 1;
3987 }
3988 }
3989
3990 if ($hotplug) {
3991 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
3992 } else {
3993 die "skip\n";
3994 }
3995 }
3996
3997 sub vmconfig_update_disk {
3998 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
3999
4000 # fixme: do we need force?
4001
4002 my $drive = parse_drive($opt, $value);
4003
4004 if ($conf->{$opt}) {
4005
4006 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4007
4008 my $media = $drive->{media} || 'disk';
4009 my $oldmedia = $old_drive->{media} || 'disk';
4010 die "unable to change media type\n" if $media ne $oldmedia;
4011
4012 if (!drive_is_cdrom($old_drive)) {
4013
4014 if ($drive->{file} ne $old_drive->{file}) {
4015
4016 die "skip\n" if !$hotplug;
4017
4018 # unplug and register as unused
4019 vm_deviceunplug($vmid, $conf, $opt);
4020 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4021
4022 } else {
4023 # update existing disk
4024
4025 # skip non hotpluggable value
4026 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4027 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4028 die "skip\n";
4029 }
4030
4031 # apply throttle
4032 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4033 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4034 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4035 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4036 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4037 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4038 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4039 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4040 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4041 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4042 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4043 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4044
4045 qemu_block_set_io_throttle($vmid,"drive-$opt",
4046 ($drive->{mbps} || 0)*1024*1024,
4047 ($drive->{mbps_rd} || 0)*1024*1024,
4048 ($drive->{mbps_wr} || 0)*1024*1024,
4049 $drive->{iops} || 0,
4050 $drive->{iops_rd} || 0,
4051 $drive->{iops_wr} || 0,
4052 ($drive->{mbps_max} || 0)*1024*1024,
4053 ($drive->{mbps_rd_max} || 0)*1024*1024,
4054 ($drive->{mbps_wr_max} || 0)*1024*1024,
4055 $drive->{iops_max} || 0,
4056 $drive->{iops_rd_max} || 0,
4057 $drive->{iops_wr_max} || 0);
4058
4059 }
4060
4061 return 1;
4062 }
4063 }
4064 }
4065 }
4066
4067 if (drive_is_cdrom($drive)) { # cdrom
4068
4069 if ($drive->{file} eq 'none') {
4070 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4071 } else {
4072 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4073 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4074 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4075 }
4076
4077 } else {
4078 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4079 # hotplug new disks
4080 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4081 }
4082 }
4083
4084 sub vm_start {
4085 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4086
4087 lock_config($vmid, sub {
4088 my $conf = load_config($vmid, $migratedfrom);
4089
4090 die "you can't start a vm if it's a template\n" if is_template($conf);
4091
4092 check_lock($conf) if !$skiplock;
4093
4094 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4095
4096 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4097 vmconfig_apply_pending($vmid, $conf, $storecfg);
4098 $conf = load_config($vmid); # update/reload
4099 }
4100
4101 my $defaults = load_defaults();
4102
4103 # set environment variable useful inside network script
4104 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4105
4106 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4107
4108 my $migrate_port = 0;
4109 my $migrate_uri;
4110 if ($statefile) {
4111 if ($statefile eq 'tcp') {
4112 my $localip = "localhost";
4113 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4114 if ($datacenterconf->{migration_unsecure}) {
4115 my $nodename = PVE::INotify::nodename();
4116 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4117 }
4118 $migrate_port = PVE::Tools::next_migrate_port();
4119 $migrate_uri = "tcp:${localip}:${migrate_port}";
4120 push @$cmd, '-incoming', $migrate_uri;
4121 push @$cmd, '-S';
4122 } else {
4123 push @$cmd, '-loadstate', $statefile;
4124 }
4125 } elsif ($paused) {
4126 push @$cmd, '-S';
4127 }
4128
4129 # host pci devices
4130 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4131 my $d = parse_hostpci($conf->{"hostpci$i"});
4132 next if !$d;
4133 my $pcidevices = $d->{pciid};
4134 foreach my $pcidevice (@$pcidevices) {
4135 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4136
4137 my $info = pci_device_info("0000:$pciid");
4138 die "IOMMU not present\n" if !check_iommu_support();
4139 die "no pci device info for device '$pciid'\n" if !$info;
4140
4141 if ($d->{driver} && $d->{driver} eq "vfio") {
4142 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4143 } else {
4144 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4145 }
4146
4147 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4148 }
4149 }
4150
4151 PVE::Storage::activate_volumes($storecfg, $vollist);
4152
4153 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4154 umask => 0077); };
4155 my $err = $@;
4156 die "start failed: $err" if $err;
4157
4158 print "migration listens on $migrate_uri\n" if $migrate_uri;
4159
4160 if ($statefile && $statefile ne 'tcp') {
4161 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4162 warn $@ if $@;
4163 }
4164
4165 if ($migratedfrom) {
4166
4167 eval {
4168 set_migration_caps($vmid);
4169 };
4170 warn $@ if $@;
4171
4172 if ($spice_port) {
4173 print "spice listens on port $spice_port\n";
4174 if ($spice_ticket) {
4175 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4176 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4177 }
4178 }
4179
4180 } else {
4181
4182 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4183 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4184 if $conf->{balloon};
4185 vm_mon_cmd_nocheck($vmid, 'qom-set',
4186 path => "machine/peripheral/balloon0",
4187 property => "guest-stats-polling-interval",
4188 value => 2);
4189 }
4190
4191 foreach my $opt (keys %$conf) {
4192 next if $opt !~ m/^net\d+$/;
4193 my $nicconf = parse_net($conf->{$opt});
4194 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4195 }
4196 }
4197 });
4198 }
4199
4200 sub vm_mon_cmd {
4201 my ($vmid, $execute, %params) = @_;
4202
4203 my $cmd = { execute => $execute, arguments => \%params };
4204 vm_qmp_command($vmid, $cmd);
4205 }
4206
4207 sub vm_mon_cmd_nocheck {
4208 my ($vmid, $execute, %params) = @_;
4209
4210 my $cmd = { execute => $execute, arguments => \%params };
4211 vm_qmp_command($vmid, $cmd, 1);
4212 }
4213
4214 sub vm_qmp_command {
4215 my ($vmid, $cmd, $nocheck) = @_;
4216
4217 my $res;
4218
4219 my $timeout;
4220 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4221 $timeout = $cmd->{arguments}->{timeout};
4222 delete $cmd->{arguments}->{timeout};
4223 }
4224
4225 eval {
4226 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4227 my $sname = qmp_socket($vmid);
4228 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4229 my $qmpclient = PVE::QMPClient->new();
4230
4231 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4232 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4233 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4234 if scalar(%{$cmd->{arguments}});
4235 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4236 } else {
4237 die "unable to open monitor socket\n";
4238 }
4239 };
4240 if (my $err = $@) {
4241 syslog("err", "VM $vmid qmp command failed - $err");
4242 die $err;
4243 }
4244
4245 return $res;
4246 }
4247
4248 sub vm_human_monitor_command {
4249 my ($vmid, $cmdline) = @_;
4250
4251 my $res;
4252
4253 my $cmd = {
4254 execute => 'human-monitor-command',
4255 arguments => { 'command-line' => $cmdline},
4256 };
4257
4258 return vm_qmp_command($vmid, $cmd);
4259 }
4260
4261 sub vm_commandline {
4262 my ($storecfg, $vmid) = @_;
4263
4264 my $conf = load_config($vmid);
4265
4266 my $defaults = load_defaults();
4267
4268 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4269
4270 return join(' ', @$cmd);
4271 }
4272
4273 sub vm_reset {
4274 my ($vmid, $skiplock) = @_;
4275
4276 lock_config($vmid, sub {
4277
4278 my $conf = load_config($vmid);
4279
4280 check_lock($conf) if !$skiplock;
4281
4282 vm_mon_cmd($vmid, "system_reset");
4283 });
4284 }
4285
4286 sub get_vm_volumes {
4287 my ($conf) = @_;
4288
4289 my $vollist = [];
4290 foreach_volid($conf, sub {
4291 my ($volid, $is_cdrom) = @_;
4292
4293 return if $volid =~ m|^/|;
4294
4295 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4296 return if !$sid;
4297
4298 push @$vollist, $volid;
4299 });
4300
4301 return $vollist;
4302 }
4303
4304 sub vm_stop_cleanup {
4305 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4306
4307 eval {
4308 fairsched_rmnod($vmid); # try to destroy group
4309
4310 if (!$keepActive) {
4311 my $vollist = get_vm_volumes($conf);
4312 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4313 }
4314
4315 foreach my $ext (qw(mon qmp pid vnc qga)) {
4316 unlink "/var/run/qemu-server/${vmid}.$ext";
4317 }
4318
4319 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4320 };
4321 warn $@ if $@; # avoid errors - just warn
4322 }
4323
4324 # Note: use $nockeck to skip tests if VM configuration file exists.
4325 # We need that when migration VMs to other nodes (files already moved)
4326 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4327 sub vm_stop {
4328 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4329
4330 $force = 1 if !defined($force) && !$shutdown;
4331
4332 if ($migratedfrom){
4333 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4334 kill 15, $pid if $pid;
4335 my $conf = load_config($vmid, $migratedfrom);
4336 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4337 return;
4338 }
4339
4340 lock_config($vmid, sub {
4341
4342 my $pid = check_running($vmid, $nocheck);
4343 return if !$pid;
4344
4345 my $conf;
4346 if (!$nocheck) {
4347 $conf = load_config($vmid);
4348 check_lock($conf) if !$skiplock;
4349 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4350 my $opts = parse_startup($conf->{startup});
4351 $timeout = $opts->{down} if $opts->{down};
4352 }
4353 }
4354
4355 $timeout = 60 if !defined($timeout);
4356
4357 eval {
4358 if ($shutdown) {
4359 if (defined($conf) && $conf->{agent}) {
4360 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4361 } else {
4362 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4363 }
4364 } else {
4365 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4366 }
4367 };
4368 my $err = $@;
4369
4370 if (!$err) {
4371 my $count = 0;
4372 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4373 $count++;
4374 sleep 1;
4375 }
4376
4377 if ($count >= $timeout) {
4378 if ($force) {
4379 warn "VM still running - terminating now with SIGTERM\n";
4380 kill 15, $pid;
4381 } else {
4382 die "VM quit/powerdown failed - got timeout\n";
4383 }
4384 } else {
4385 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4386 return;
4387 }
4388 } else {
4389 if ($force) {
4390 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4391 kill 15, $pid;
4392 } else {
4393 die "VM quit/powerdown failed\n";
4394 }
4395 }
4396
4397 # wait again
4398 $timeout = 10;
4399
4400 my $count = 0;
4401 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4402 $count++;
4403 sleep 1;
4404 }
4405
4406 if ($count >= $timeout) {
4407 warn "VM still running - terminating now with SIGKILL\n";
4408 kill 9, $pid;
4409 sleep 1;
4410 }
4411
4412 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4413 });
4414 }
4415
4416 sub vm_suspend {
4417 my ($vmid, $skiplock) = @_;
4418
4419 lock_config($vmid, sub {
4420
4421 my $conf = load_config($vmid);
4422
4423 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4424
4425 vm_mon_cmd($vmid, "stop");
4426 });
4427 }
4428
4429 sub vm_resume {
4430 my ($vmid, $skiplock) = @_;
4431
4432 lock_config($vmid, sub {
4433
4434 my $conf = load_config($vmid);
4435
4436 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4437
4438 vm_mon_cmd($vmid, "cont");
4439 });
4440 }
4441
4442 sub vm_sendkey {
4443 my ($vmid, $skiplock, $key) = @_;
4444
4445 lock_config($vmid, sub {
4446
4447 my $conf = load_config($vmid);
4448
4449 # there is no qmp command, so we use the human monitor command
4450 vm_human_monitor_command($vmid, "sendkey $key");
4451 });
4452 }
4453
4454 sub vm_destroy {
4455 my ($storecfg, $vmid, $skiplock) = @_;
4456
4457 lock_config($vmid, sub {
4458
4459 my $conf = load_config($vmid);
4460
4461 check_lock($conf) if !$skiplock;
4462
4463 if (!check_running($vmid)) {
4464 fairsched_rmnod($vmid); # try to destroy group
4465 destroy_vm($storecfg, $vmid);
4466 } else {
4467 die "VM $vmid is running - destroy failed\n";
4468 }
4469 });
4470 }
4471
4472 # pci helpers
4473
4474 sub file_write {
4475 my ($filename, $buf) = @_;
4476
4477 my $fh = IO::File->new($filename, "w");
4478 return undef if !$fh;
4479
4480 my $res = print $fh $buf;
4481
4482 $fh->close();
4483
4484 return $res;
4485 }
4486
4487 sub pci_device_info {
4488 my ($name) = @_;
4489
4490 my $res;
4491
4492 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4493 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4494
4495 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4496 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4497
4498 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4499 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4500
4501 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4502 return undef if !defined($product) || $product !~ s/^0x//;
4503
4504 $res = {
4505 name => $name,
4506 vendor => $vendor,
4507 product => $product,
4508 domain => $domain,
4509 bus => $bus,
4510 slot => $slot,
4511 func => $func,
4512 irq => $irq,
4513 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4514 };
4515
4516 return $res;
4517 }
4518
4519 sub pci_dev_reset {
4520 my ($dev) = @_;
4521
4522 my $name = $dev->{name};
4523
4524 my $fn = "$pcisysfs/devices/$name/reset";
4525
4526 return file_write($fn, "1");
4527 }
4528
4529 sub pci_dev_bind_to_stub {
4530 my ($dev) = @_;
4531
4532 my $name = $dev->{name};
4533
4534 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4535 return 1 if -d $testdir;
4536
4537 my $data = "$dev->{vendor} $dev->{product}";
4538 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4539
4540 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4541 if (!file_write($fn, $name)) {
4542 return undef if -f $fn;
4543 }
4544
4545 $fn = "$pcisysfs/drivers/pci-stub/bind";
4546 if (! -d $testdir) {
4547 return undef if !file_write($fn, $name);
4548 }
4549
4550 return -d $testdir;
4551 }
4552
4553 sub pci_dev_bind_to_vfio {
4554 my ($dev) = @_;
4555
4556 my $name = $dev->{name};
4557
4558 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4559
4560 if (!-d $vfio_basedir) {
4561 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4562 }
4563 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4564
4565 my $testdir = "$vfio_basedir/$name";
4566 return 1 if -d $testdir;
4567
4568 my $data = "$dev->{vendor} $dev->{product}";
4569 return undef if !file_write("$vfio_basedir/new_id", $data);
4570
4571 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4572 if (!file_write($fn, $name)) {
4573 return undef if -f $fn;
4574 }
4575
4576 $fn = "$vfio_basedir/bind";
4577 if (! -d $testdir) {
4578 return undef if !file_write($fn, $name);
4579 }
4580
4581 return -d $testdir;
4582 }
4583
4584 sub pci_dev_group_bind_to_vfio {
4585 my ($pciid) = @_;
4586
4587 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4588
4589 if (!-d $vfio_basedir) {
4590 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4591 }
4592 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4593
4594 # get IOMMU group devices
4595 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4596 my @devs = grep /^0000:/, readdir($D);
4597 closedir($D);
4598
4599 foreach my $pciid (@devs) {
4600 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4601 my $info = pci_device_info($1);
4602 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4603 }
4604
4605 return 1;
4606 }
4607
4608 sub print_pci_addr {
4609 my ($id, $bridges) = @_;
4610
4611 my $res = '';
4612 my $devices = {
4613 piix3 => { bus => 0, addr => 1 },
4614 #addr2 : first videocard
4615 balloon0 => { bus => 0, addr => 3 },
4616 watchdog => { bus => 0, addr => 4 },
4617 scsihw0 => { bus => 0, addr => 5 },
4618 scsihw1 => { bus => 0, addr => 6 },
4619 ahci0 => { bus => 0, addr => 7 },
4620 qga0 => { bus => 0, addr => 8 },
4621 spice => { bus => 0, addr => 9 },
4622 virtio0 => { bus => 0, addr => 10 },
4623 virtio1 => { bus => 0, addr => 11 },
4624 virtio2 => { bus => 0, addr => 12 },
4625 virtio3 => { bus => 0, addr => 13 },
4626 virtio4 => { bus => 0, addr => 14 },
4627 virtio5 => { bus => 0, addr => 15 },
4628 hostpci0 => { bus => 0, addr => 16 },
4629 hostpci1 => { bus => 0, addr => 17 },
4630 net0 => { bus => 0, addr => 18 },
4631 net1 => { bus => 0, addr => 19 },
4632 net2 => { bus => 0, addr => 20 },
4633 net3 => { bus => 0, addr => 21 },
4634 net4 => { bus => 0, addr => 22 },
4635 net5 => { bus => 0, addr => 23 },
4636 vga1 => { bus => 0, addr => 24 },
4637 vga2 => { bus => 0, addr => 25 },
4638 vga3 => { bus => 0, addr => 26 },
4639 hostpci2 => { bus => 0, addr => 27 },
4640 hostpci3 => { bus => 0, addr => 28 },
4641 #addr29 : usb-host (pve-usb.cfg)
4642 'pci.1' => { bus => 0, addr => 30 },
4643 'pci.2' => { bus => 0, addr => 31 },
4644 'net6' => { bus => 1, addr => 1 },
4645 'net7' => { bus => 1, addr => 2 },
4646 'net8' => { bus => 1, addr => 3 },
4647 'net9' => { bus => 1, addr => 4 },
4648 'net10' => { bus => 1, addr => 5 },
4649 'net11' => { bus => 1, addr => 6 },
4650 'net12' => { bus => 1, addr => 7 },
4651 'net13' => { bus => 1, addr => 8 },
4652 'net14' => { bus => 1, addr => 9 },
4653 'net15' => { bus => 1, addr => 10 },
4654 'net16' => { bus => 1, addr => 11 },
4655 'net17' => { bus => 1, addr => 12 },
4656 'net18' => { bus => 1, addr => 13 },
4657 'net19' => { bus => 1, addr => 14 },
4658 'net20' => { bus => 1, addr => 15 },
4659 'net21' => { bus => 1, addr => 16 },
4660 'net22' => { bus => 1, addr => 17 },
4661 'net23' => { bus => 1, addr => 18 },
4662 'net24' => { bus => 1, addr => 19 },
4663 'net25' => { bus => 1, addr => 20 },
4664 'net26' => { bus => 1, addr => 21 },
4665 'net27' => { bus => 1, addr => 22 },
4666 'net28' => { bus => 1, addr => 23 },
4667 'net29' => { bus => 1, addr => 24 },
4668 'net30' => { bus => 1, addr => 25 },
4669 'net31' => { bus => 1, addr => 26 },
4670 'virtio6' => { bus => 2, addr => 1 },
4671 'virtio7' => { bus => 2, addr => 2 },
4672 'virtio8' => { bus => 2, addr => 3 },
4673 'virtio9' => { bus => 2, addr => 4 },
4674 'virtio10' => { bus => 2, addr => 5 },
4675 'virtio11' => { bus => 2, addr => 6 },
4676 'virtio12' => { bus => 2, addr => 7 },
4677 'virtio13' => { bus => 2, addr => 8 },
4678 'virtio14' => { bus => 2, addr => 9 },
4679 'virtio15' => { bus => 2, addr => 10 },
4680 };
4681
4682 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4683 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4684 my $bus = $devices->{$id}->{bus};
4685 $res = ",bus=pci.$bus,addr=$addr";
4686 $bridges->{$bus} = 1 if $bridges;
4687 }
4688 return $res;
4689
4690 }
4691
4692 sub print_pcie_addr {
4693 my ($id) = @_;
4694
4695 my $res = '';
4696 my $devices = {
4697 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4698 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4699 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4700 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4701 };
4702
4703 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4704 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4705 my $bus = $devices->{$id}->{bus};
4706 $res = ",bus=$bus,addr=$addr";
4707 }
4708 return $res;
4709
4710 }
4711
4712 # vzdump restore implementaion
4713
4714 sub tar_archive_read_firstfile {
4715 my $archive = shift;
4716
4717 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4718
4719 # try to detect archive type first
4720 my $pid = open (TMP, "tar tf '$archive'|") ||
4721 die "unable to open file '$archive'\n";
4722 my $firstfile = <TMP>;
4723 kill 15, $pid;
4724 close TMP;
4725
4726 die "ERROR: archive contaions no data\n" if !$firstfile;
4727 chomp $firstfile;
4728
4729 return $firstfile;
4730 }
4731
4732 sub tar_restore_cleanup {
4733 my ($storecfg, $statfile) = @_;
4734
4735 print STDERR "starting cleanup\n";
4736
4737 if (my $fd = IO::File->new($statfile, "r")) {
4738 while (defined(my $line = <$fd>)) {
4739 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4740 my $volid = $2;
4741 eval {
4742 if ($volid =~ m|^/|) {
4743 unlink $volid || die 'unlink failed\n';
4744 } else {
4745 PVE::Storage::vdisk_free($storecfg, $volid);
4746 }
4747 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4748 };
4749 print STDERR "unable to cleanup '$volid' - $@" if $@;
4750 } else {
4751 print STDERR "unable to parse line in statfile - $line";
4752 }
4753 }
4754 $fd->close();
4755 }
4756 }
4757
4758 sub restore_archive {
4759 my ($archive, $vmid, $user, $opts) = @_;
4760
4761 my $format = $opts->{format};
4762 my $comp;
4763
4764 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4765 $format = 'tar' if !$format;
4766 $comp = 'gzip';
4767 } elsif ($archive =~ m/\.tar$/) {
4768 $format = 'tar' if !$format;
4769 } elsif ($archive =~ m/.tar.lzo$/) {
4770 $format = 'tar' if !$format;
4771 $comp = 'lzop';
4772 } elsif ($archive =~ m/\.vma$/) {
4773 $format = 'vma' if !$format;
4774 } elsif ($archive =~ m/\.vma\.gz$/) {
4775 $format = 'vma' if !$format;
4776 $comp = 'gzip';
4777 } elsif ($archive =~ m/\.vma\.lzo$/) {
4778 $format = 'vma' if !$format;
4779 $comp = 'lzop';
4780 } else {
4781 $format = 'vma' if !$format; # default
4782 }
4783
4784 # try to detect archive format
4785 if ($format eq 'tar') {
4786 return restore_tar_archive($archive, $vmid, $user, $opts);
4787 } else {
4788 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4789 }
4790 }
4791
4792 sub restore_update_config_line {
4793 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4794
4795 return if $line =~ m/^\#qmdump\#/;
4796 return if $line =~ m/^\#vzdump\#/;
4797 return if $line =~ m/^lock:/;
4798 return if $line =~ m/^unused\d+:/;
4799 return if $line =~ m/^parent:/;
4800 return if $line =~ m/^template:/; # restored VM is never a template
4801
4802 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4803 # try to convert old 1.X settings
4804 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4805 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4806 my ($model, $macaddr) = split(/\=/, $devconfig);
4807 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4808 my $net = {
4809 model => $model,
4810 bridge => "vmbr$ind",
4811 macaddr => $macaddr,
4812 };
4813 my $netstr = print_net($net);
4814
4815 print $outfd "net$cookie->{netcount}: $netstr\n";
4816 $cookie->{netcount}++;
4817 }
4818 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4819 my ($id, $netstr) = ($1, $2);
4820 my $net = parse_net($netstr);
4821 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4822 $netstr = print_net($net);
4823 print $outfd "$id: $netstr\n";
4824 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4825 my $virtdev = $1;
4826 my $value = $3;
4827 if ($line =~ m/backup=no/) {
4828 print $outfd "#$line";
4829 } elsif ($virtdev && $map->{$virtdev}) {
4830 my $di = parse_drive($virtdev, $value);
4831 delete $di->{format}; # format can change on restore
4832 $di->{file} = $map->{$virtdev};
4833 $value = print_drive($vmid, $di);
4834 print $outfd "$virtdev: $value\n";
4835 } else {
4836 print $outfd $line;
4837 }
4838 } else {
4839 print $outfd $line;
4840 }
4841 }
4842
4843 sub scan_volids {
4844 my ($cfg, $vmid) = @_;
4845
4846 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4847
4848 my $volid_hash = {};
4849 foreach my $storeid (keys %$info) {
4850 foreach my $item (@{$info->{$storeid}}) {
4851 next if !($item->{volid} && $item->{size});
4852 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4853 $volid_hash->{$item->{volid}} = $item;
4854 }
4855 }
4856
4857 return $volid_hash;
4858 }
4859
4860 sub get_used_paths {
4861 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4862
4863 my $used_path = {};
4864
4865 my $scan_config = sub {
4866 my ($cref, $snapname) = @_;
4867
4868 foreach my $key (keys %$cref) {
4869 my $value = $cref->{$key};
4870 if (valid_drivename($key)) {
4871 next if $skip_drive && $key eq $skip_drive;
4872 my $drive = parse_drive($key, $value);
4873 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4874 if ($drive->{file} =~ m!^/!) {
4875 $used_path->{$drive->{file}}++; # = 1;
4876 } else {
4877 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4878 next if !$storeid;
4879 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4880 next if !$scfg;
4881 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4882 $used_path->{$path}++; # = 1;
4883 }
4884 }
4885 }
4886 };
4887
4888 &$scan_config($conf);
4889
4890 undef $skip_drive;
4891
4892 if ($scan_snapshots) {
4893 foreach my $snapname (keys %{$conf->{snapshots}}) {
4894 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4895 }
4896 }
4897
4898 return $used_path;
4899 }
4900
4901 sub update_disksize {
4902 my ($vmid, $conf, $volid_hash) = @_;
4903
4904 my $changes;
4905
4906 my $used = {};
4907
4908 # Note: it is allowed to define multiple storages with same path (alias), so
4909 # we need to check both 'volid' and real 'path' (two different volid can point
4910 # to the same path).
4911
4912 my $usedpath = {};
4913
4914 # update size info
4915 foreach my $opt (keys %$conf) {
4916 if (valid_drivename($opt)) {
4917 my $drive = parse_drive($opt, $conf->{$opt});
4918 my $volid = $drive->{file};
4919 next if !$volid;
4920
4921 $used->{$volid} = 1;
4922 if ($volid_hash->{$volid} &&
4923 (my $path = $volid_hash->{$volid}->{path})) {
4924 $usedpath->{$path} = 1;
4925 }
4926
4927 next if drive_is_cdrom($drive);
4928 next if !$volid_hash->{$volid};
4929
4930 $drive->{size} = $volid_hash->{$volid}->{size};
4931 my $new = print_drive($vmid, $drive);
4932 if ($new ne $conf->{$opt}) {
4933 $changes = 1;
4934 $conf->{$opt} = $new;
4935 }
4936 }
4937 }
4938
4939 # remove 'unusedX' entry if volume is used
4940 foreach my $opt (keys %$conf) {
4941 next if $opt !~ m/^unused\d+$/;
4942 my $volid = $conf->{$opt};
4943 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
4944 if ($used->{$volid} || ($path && $usedpath->{$path})) {
4945 $changes = 1;
4946 delete $conf->{$opt};
4947 }
4948 }
4949
4950 foreach my $volid (sort keys %$volid_hash) {
4951 next if $volid =~ m/vm-$vmid-state-/;
4952 next if $used->{$volid};
4953 my $path = $volid_hash->{$volid}->{path};
4954 next if !$path; # just to be sure
4955 next if $usedpath->{$path};
4956 $changes = 1;
4957 add_unused_volume($conf, $volid);
4958 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
4959 }
4960
4961 return $changes;
4962 }
4963
4964 sub rescan {
4965 my ($vmid, $nolock) = @_;
4966
4967 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
4968
4969 my $volid_hash = scan_volids($cfg, $vmid);
4970
4971 my $updatefn = sub {
4972 my ($vmid) = @_;
4973
4974 my $conf = load_config($vmid);
4975
4976 check_lock($conf);
4977
4978 my $vm_volids = {};
4979 foreach my $volid (keys %$volid_hash) {
4980 my $info = $volid_hash->{$volid};
4981 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
4982 }
4983
4984 my $changes = update_disksize($vmid, $conf, $vm_volids);
4985
4986 update_config_nolock($vmid, $conf, 1) if $changes;
4987 };
4988
4989 if (defined($vmid)) {
4990 if ($nolock) {
4991 &$updatefn($vmid);
4992 } else {
4993 lock_config($vmid, $updatefn, $vmid);
4994 }
4995 } else {
4996 my $vmlist = config_list();
4997 foreach my $vmid (keys %$vmlist) {
4998 if ($nolock) {
4999 &$updatefn($vmid);
5000 } else {
5001 lock_config($vmid, $updatefn, $vmid);
5002 }
5003 }
5004 }
5005 }
5006
5007 sub restore_vma_archive {
5008 my ($archive, $vmid, $user, $opts, $comp) = @_;
5009
5010 my $input = $archive eq '-' ? "<&STDIN" : undef;
5011 my $readfrom = $archive;
5012
5013 my $uncomp = '';
5014 if ($comp) {
5015 $readfrom = '-';
5016 my $qarchive = PVE::Tools::shellquote($archive);
5017 if ($comp eq 'gzip') {
5018 $uncomp = "zcat $qarchive|";
5019 } elsif ($comp eq 'lzop') {
5020 $uncomp = "lzop -d -c $qarchive|";
5021 } else {
5022 die "unknown compression method '$comp'\n";
5023 }
5024
5025 }
5026
5027 my $tmpdir = "/var/tmp/vzdumptmp$$";
5028 rmtree $tmpdir;
5029
5030 # disable interrupts (always do cleanups)
5031 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5032 warn "got interrupt - ignored\n";
5033 };
5034
5035 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5036 POSIX::mkfifo($mapfifo, 0600);
5037 my $fifofh;
5038
5039 my $openfifo = sub {
5040 open($fifofh, '>', $mapfifo) || die $!;
5041 };
5042
5043 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5044
5045 my $oldtimeout;
5046 my $timeout = 5;
5047
5048 my $devinfo = {};
5049
5050 my $rpcenv = PVE::RPCEnvironment::get();
5051
5052 my $conffile = config_file($vmid);
5053 my $tmpfn = "$conffile.$$.tmp";
5054
5055 # Note: $oldconf is undef if VM does not exists
5056 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5057
5058 my $print_devmap = sub {
5059 my $virtdev_hash = {};
5060
5061 my $cfgfn = "$tmpdir/qemu-server.conf";
5062
5063 # we can read the config - that is already extracted
5064 my $fh = IO::File->new($cfgfn, "r") ||
5065 "unable to read qemu-server.conf - $!\n";
5066
5067 while (defined(my $line = <$fh>)) {
5068 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5069 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5070 die "archive does not contain data for drive '$virtdev'\n"
5071 if !$devinfo->{$devname};
5072 if (defined($opts->{storage})) {
5073 $storeid = $opts->{storage} || 'local';
5074 } elsif (!$storeid) {
5075 $storeid = 'local';
5076 }
5077 $format = 'raw' if !$format;
5078 $devinfo->{$devname}->{devname} = $devname;
5079 $devinfo->{$devname}->{virtdev} = $virtdev;
5080 $devinfo->{$devname}->{format} = $format;
5081 $devinfo->{$devname}->{storeid} = $storeid;
5082
5083 # check permission on storage
5084 my $pool = $opts->{pool}; # todo: do we need that?
5085 if ($user ne 'root@pam') {
5086 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5087 }
5088
5089 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5090 }
5091 }
5092
5093 foreach my $devname (keys %$devinfo) {
5094 die "found no device mapping information for device '$devname'\n"
5095 if !$devinfo->{$devname}->{virtdev};
5096 }
5097
5098 my $cfg = cfs_read_file('storage.cfg');
5099
5100 # create empty/temp config
5101 if ($oldconf) {
5102 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5103 foreach_drive($oldconf, sub {
5104 my ($ds, $drive) = @_;
5105
5106 return if drive_is_cdrom($drive);
5107
5108 my $volid = $drive->{file};
5109
5110 return if !$volid || $volid =~ m|^/|;
5111
5112 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5113 return if !$path || !$owner || ($owner != $vmid);
5114
5115 # Note: only delete disk we want to restore
5116 # other volumes will become unused
5117 if ($virtdev_hash->{$ds}) {
5118 PVE::Storage::vdisk_free($cfg, $volid);
5119 }
5120 });
5121 }
5122
5123 my $map = {};
5124 foreach my $virtdev (sort keys %$virtdev_hash) {
5125 my $d = $virtdev_hash->{$virtdev};
5126 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5127 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5128
5129 # test if requested format is supported
5130 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5131 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5132 $d->{format} = $defFormat if !$supported;
5133
5134 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5135 $d->{format}, undef, $alloc_size);
5136 print STDERR "new volume ID is '$volid'\n";
5137 $d->{volid} = $volid;
5138 my $path = PVE::Storage::path($cfg, $volid);
5139
5140 my $write_zeros = 1;
5141 # fixme: what other storages types initialize volumes with zero?
5142 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5143 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5144 $write_zeros = 0;
5145 }
5146
5147 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5148
5149 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5150 $map->{$virtdev} = $volid;
5151 }
5152
5153 $fh->seek(0, 0) || die "seek failed - $!\n";
5154
5155 my $outfd = new IO::File ($tmpfn, "w") ||
5156 die "unable to write config for VM $vmid\n";
5157
5158 my $cookie = { netcount => 0 };
5159 while (defined(my $line = <$fh>)) {
5160 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5161 }
5162
5163 $fh->close();
5164 $outfd->close();
5165 };
5166
5167 eval {
5168 # enable interrupts
5169 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5170 die "interrupted by signal\n";
5171 };
5172 local $SIG{ALRM} = sub { die "got timeout\n"; };
5173
5174 $oldtimeout = alarm($timeout);
5175
5176 my $parser = sub {
5177 my $line = shift;
5178
5179 print "$line\n";
5180
5181 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5182 my ($dev_id, $size, $devname) = ($1, $2, $3);
5183 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5184 } elsif ($line =~ m/^CTIME: /) {
5185 # we correctly received the vma config, so we can disable
5186 # the timeout now for disk allocation (set to 10 minutes, so
5187 # that we always timeout if something goes wrong)
5188 alarm(600);
5189 &$print_devmap();
5190 print $fifofh "done\n";
5191 my $tmp = $oldtimeout || 0;
5192 $oldtimeout = undef;
5193 alarm($tmp);
5194 close($fifofh);
5195 }
5196 };
5197
5198 print "restore vma archive: $cmd\n";
5199 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5200 };
5201 my $err = $@;
5202
5203 alarm($oldtimeout) if $oldtimeout;
5204
5205 unlink $mapfifo;
5206
5207 if ($err) {
5208 rmtree $tmpdir;
5209 unlink $tmpfn;
5210
5211 my $cfg = cfs_read_file('storage.cfg');
5212 foreach my $devname (keys %$devinfo) {
5213 my $volid = $devinfo->{$devname}->{volid};
5214 next if !$volid;
5215 eval {
5216 if ($volid =~ m|^/|) {
5217 unlink $volid || die 'unlink failed\n';
5218 } else {
5219 PVE::Storage::vdisk_free($cfg, $volid);
5220 }
5221 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5222 };
5223 print STDERR "unable to cleanup '$volid' - $@" if $@;
5224 }
5225 die $err;
5226 }
5227
5228 rmtree $tmpdir;
5229
5230 rename($tmpfn, $conffile) ||
5231 die "unable to commit configuration file '$conffile'\n";
5232
5233 PVE::Cluster::cfs_update(); # make sure we read new file
5234
5235 eval { rescan($vmid, 1); };
5236 warn $@ if $@;
5237 }
5238
5239 sub restore_tar_archive {
5240 my ($archive, $vmid, $user, $opts) = @_;
5241
5242 if ($archive ne '-') {
5243 my $firstfile = tar_archive_read_firstfile($archive);
5244 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5245 if $firstfile ne 'qemu-server.conf';
5246 }
5247
5248 my $storecfg = cfs_read_file('storage.cfg');
5249
5250 # destroy existing data - keep empty config
5251 my $vmcfgfn = config_file($vmid);
5252 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5253
5254 my $tocmd = "/usr/lib/qemu-server/qmextract";
5255
5256 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5257 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5258 $tocmd .= ' --prealloc' if $opts->{prealloc};
5259 $tocmd .= ' --info' if $opts->{info};
5260
5261 # tar option "xf" does not autodetect compression when read from STDIN,
5262 # so we pipe to zcat
5263 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5264 PVE::Tools::shellquote("--to-command=$tocmd");
5265
5266 my $tmpdir = "/var/tmp/vzdumptmp$$";
5267 mkpath $tmpdir;
5268
5269 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5270 local $ENV{VZDUMP_VMID} = $vmid;
5271 local $ENV{VZDUMP_USER} = $user;
5272
5273 my $conffile = config_file($vmid);
5274 my $tmpfn = "$conffile.$$.tmp";
5275
5276 # disable interrupts (always do cleanups)
5277 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5278 print STDERR "got interrupt - ignored\n";
5279 };
5280
5281 eval {
5282 # enable interrupts
5283 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5284 die "interrupted by signal\n";
5285 };
5286
5287 if ($archive eq '-') {
5288 print "extracting archive from STDIN\n";
5289 run_command($cmd, input => "<&STDIN");
5290 } else {
5291 print "extracting archive '$archive'\n";
5292 run_command($cmd);
5293 }
5294
5295 return if $opts->{info};
5296
5297 # read new mapping
5298 my $map = {};
5299 my $statfile = "$tmpdir/qmrestore.stat";
5300 if (my $fd = IO::File->new($statfile, "r")) {
5301 while (defined (my $line = <$fd>)) {
5302 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5303 $map->{$1} = $2 if $1;
5304 } else {
5305 print STDERR "unable to parse line in statfile - $line\n";
5306 }
5307 }
5308 $fd->close();
5309 }
5310
5311 my $confsrc = "$tmpdir/qemu-server.conf";
5312
5313 my $srcfd = new IO::File($confsrc, "r") ||
5314 die "unable to open file '$confsrc'\n";
5315
5316 my $outfd = new IO::File ($tmpfn, "w") ||
5317 die "unable to write config for VM $vmid\n";
5318
5319 my $cookie = { netcount => 0 };
5320 while (defined (my $line = <$srcfd>)) {
5321 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5322 }
5323
5324 $srcfd->close();
5325 $outfd->close();
5326 };
5327 my $err = $@;
5328
5329 if ($err) {
5330
5331 unlink $tmpfn;
5332
5333 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5334
5335 die $err;
5336 }
5337
5338 rmtree $tmpdir;
5339
5340 rename $tmpfn, $conffile ||
5341 die "unable to commit configuration file '$conffile'\n";
5342
5343 PVE::Cluster::cfs_update(); # make sure we read new file
5344
5345 eval { rescan($vmid, 1); };
5346 warn $@ if $@;
5347 };
5348
5349
5350 # Internal snapshots
5351
5352 # NOTE: Snapshot create/delete involves several non-atomic
5353 # action, and can take a long time.
5354 # So we try to avoid locking the file and use 'lock' variable
5355 # inside the config file instead.
5356
5357 my $snapshot_copy_config = sub {
5358 my ($source, $dest) = @_;
5359
5360 foreach my $k (keys %$source) {
5361 next if $k eq 'snapshots';
5362 next if $k eq 'snapstate';
5363 next if $k eq 'snaptime';
5364 next if $k eq 'vmstate';
5365 next if $k eq 'lock';
5366 next if $k eq 'digest';
5367 next if $k eq 'description';
5368 next if $k =~ m/^unused\d+$/;
5369
5370 $dest->{$k} = $source->{$k};
5371 }
5372 };
5373
5374 my $snapshot_apply_config = sub {
5375 my ($conf, $snap) = @_;
5376
5377 # copy snapshot list
5378 my $newconf = {
5379 snapshots => $conf->{snapshots},
5380 };
5381
5382 # keep description and list of unused disks
5383 foreach my $k (keys %$conf) {
5384 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5385 $newconf->{$k} = $conf->{$k};
5386 }
5387
5388 &$snapshot_copy_config($snap, $newconf);
5389
5390 return $newconf;
5391 };
5392
5393 sub foreach_writable_storage {
5394 my ($conf, $func) = @_;
5395
5396 my $sidhash = {};
5397
5398 foreach my $ds (keys %$conf) {
5399 next if !valid_drivename($ds);
5400
5401 my $drive = parse_drive($ds, $conf->{$ds});
5402 next if !$drive;
5403 next if drive_is_cdrom($drive);
5404
5405 my $volid = $drive->{file};
5406
5407 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5408 $sidhash->{$sid} = $sid if $sid;
5409 }
5410
5411 foreach my $sid (sort keys %$sidhash) {
5412 &$func($sid);
5413 }
5414 }
5415
5416 my $alloc_vmstate_volid = sub {
5417 my ($storecfg, $vmid, $conf, $snapname) = @_;
5418
5419 # Note: we try to be smart when selecting a $target storage
5420
5421 my $target;
5422
5423 # search shared storage first
5424 foreach_writable_storage($conf, sub {
5425 my ($sid) = @_;
5426 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5427 return if !$scfg->{shared};
5428
5429 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5430 });
5431
5432 if (!$target) {
5433 # now search local storage
5434 foreach_writable_storage($conf, sub {
5435 my ($sid) = @_;
5436 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5437 return if $scfg->{shared};
5438
5439 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5440 });
5441 }
5442
5443 $target = 'local' if !$target;
5444
5445 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5446 # we abort live save after $conf->{memory}, so we need at max twice that space
5447 my $size = $conf->{memory}*2 + $driver_state_size;
5448
5449 my $name = "vm-$vmid-state-$snapname";
5450 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5451 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5452 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5453
5454 return $volid;
5455 };
5456
5457 my $snapshot_prepare = sub {
5458 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5459
5460 my $snap;
5461
5462 my $updatefn = sub {
5463
5464 my $conf = load_config($vmid);
5465
5466 die "you can't take a snapshot if it's a template\n"
5467 if is_template($conf);
5468
5469 check_lock($conf);
5470
5471 $conf->{lock} = 'snapshot';
5472
5473 die "snapshot name '$snapname' already used\n"
5474 if defined($conf->{snapshots}->{$snapname});
5475
5476 my $storecfg = PVE::Storage::config();
5477 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5478
5479 $snap = $conf->{snapshots}->{$snapname} = {};
5480
5481 if ($save_vmstate && check_running($vmid)) {
5482 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5483 }
5484
5485 &$snapshot_copy_config($conf, $snap);
5486
5487 $snap->{snapstate} = "prepare";
5488 $snap->{snaptime} = time();
5489 $snap->{description} = $comment if $comment;
5490
5491 # always overwrite machine if we save vmstate. This makes sure we
5492 # can restore it later using correct machine type
5493 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5494
5495 update_config_nolock($vmid, $conf, 1);
5496 };
5497
5498 lock_config($vmid, $updatefn);
5499
5500 return $snap;
5501 };
5502
5503 my $snapshot_commit = sub {
5504 my ($vmid, $snapname) = @_;
5505
5506 my $updatefn = sub {
5507
5508 my $conf = load_config($vmid);
5509
5510 die "missing snapshot lock\n"
5511 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5512
5513 my $has_machine_config = defined($conf->{machine});
5514
5515 my $snap = $conf->{snapshots}->{$snapname};
5516
5517 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5518
5519 die "wrong snapshot state\n"
5520 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5521
5522 delete $snap->{snapstate};
5523 delete $conf->{lock};
5524
5525 my $newconf = &$snapshot_apply_config($conf, $snap);
5526
5527 delete $newconf->{machine} if !$has_machine_config;
5528
5529 $newconf->{parent} = $snapname;
5530
5531 update_config_nolock($vmid, $newconf, 1);
5532 };
5533
5534 lock_config($vmid, $updatefn);
5535 };
5536
5537 sub snapshot_rollback {
5538 my ($vmid, $snapname) = @_;
5539
5540 my $snap;
5541
5542 my $prepare = 1;
5543
5544 my $storecfg = PVE::Storage::config();
5545
5546 my $updatefn = sub {
5547
5548 my $conf = load_config($vmid);
5549
5550 die "you can't rollback if vm is a template\n" if is_template($conf);
5551
5552 $snap = $conf->{snapshots}->{$snapname};
5553
5554 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5555
5556 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5557 if $snap->{snapstate};
5558
5559 if ($prepare) {
5560 check_lock($conf);
5561 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5562 }
5563
5564 die "unable to rollback vm $vmid: vm is running\n"
5565 if check_running($vmid);
5566
5567 if ($prepare) {
5568 $conf->{lock} = 'rollback';
5569 } else {
5570 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5571 delete $conf->{lock};
5572 }
5573
5574 my $forcemachine;
5575
5576 if (!$prepare) {
5577 my $has_machine_config = defined($conf->{machine});
5578
5579 # copy snapshot config to current config
5580 $conf = &$snapshot_apply_config($conf, $snap);
5581 $conf->{parent} = $snapname;
5582
5583 # Note: old code did not store 'machine', so we try to be smart
5584 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5585 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5586 # we remove the 'machine' configuration if not explicitly specified
5587 # in the original config.
5588 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5589 }
5590
5591 update_config_nolock($vmid, $conf, 1);
5592
5593 if (!$prepare && $snap->{vmstate}) {
5594 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5595 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5596 }
5597 };
5598
5599 lock_config($vmid, $updatefn);
5600
5601 foreach_drive($snap, sub {
5602 my ($ds, $drive) = @_;
5603
5604 return if drive_is_cdrom($drive);
5605
5606 my $volid = $drive->{file};
5607 my $device = "drive-$ds";
5608
5609 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5610 });
5611
5612 $prepare = 0;
5613 lock_config($vmid, $updatefn);
5614 }
5615
5616 my $savevm_wait = sub {
5617 my ($vmid) = @_;
5618
5619 for(;;) {
5620 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5621 if (!$stat->{status}) {
5622 die "savevm not active\n";
5623 } elsif ($stat->{status} eq 'active') {
5624 sleep(1);
5625 next;
5626 } elsif ($stat->{status} eq 'completed') {
5627 last;
5628 } else {
5629 die "query-savevm returned status '$stat->{status}'\n";
5630 }
5631 }
5632 };
5633
5634 sub snapshot_create {
5635 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5636
5637 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5638
5639 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5640
5641 my $config = load_config($vmid);
5642
5643 my $running = check_running($vmid);
5644
5645 my $freezefs = $running && $config->{agent};
5646 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5647
5648 my $drivehash = {};
5649
5650 if ($freezefs) {
5651 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5652 warn "guest-fsfreeze-freeze problems - $@" if $@;
5653 }
5654
5655 eval {
5656 # create internal snapshots of all drives
5657
5658 my $storecfg = PVE::Storage::config();
5659
5660 if ($running) {
5661 if ($snap->{vmstate}) {
5662 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5663 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5664 &$savevm_wait($vmid);
5665 } else {
5666 vm_mon_cmd($vmid, "savevm-start");
5667 }
5668 };
5669
5670 foreach_drive($snap, sub {
5671 my ($ds, $drive) = @_;
5672
5673 return if drive_is_cdrom($drive);
5674
5675 my $volid = $drive->{file};
5676 my $device = "drive-$ds";
5677
5678 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5679 $drivehash->{$ds} = 1;
5680 });
5681 };
5682 my $err = $@;
5683
5684 if ($running) {
5685 eval { vm_mon_cmd($vmid, "savevm-end") };
5686 warn $@ if $@;
5687
5688 if ($freezefs) {
5689 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5690 warn "guest-fsfreeze-thaw problems - $@" if $@;
5691 }
5692
5693 # savevm-end is async, we need to wait
5694 for (;;) {
5695 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5696 if (!$stat->{bytes}) {
5697 last;
5698 } else {
5699 print "savevm not yet finished\n";
5700 sleep(1);
5701 next;
5702 }
5703 }
5704 }
5705
5706 if ($err) {
5707 warn "snapshot create failed: starting cleanup\n";
5708 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5709 warn $@ if $@;
5710 die $err;
5711 }
5712
5713 &$snapshot_commit($vmid, $snapname);
5714 }
5715
5716 # Note: $drivehash is only set when called from snapshot_create.
5717 sub snapshot_delete {
5718 my ($vmid, $snapname, $force, $drivehash) = @_;
5719
5720 my $prepare = 1;
5721
5722 my $snap;
5723 my $unused = [];
5724
5725 my $unlink_parent = sub {
5726 my ($confref, $new_parent) = @_;
5727
5728 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5729 if ($new_parent) {
5730 $confref->{parent} = $new_parent;
5731 } else {
5732 delete $confref->{parent};
5733 }
5734 }
5735 };
5736
5737 my $updatefn = sub {
5738 my ($remove_drive) = @_;
5739
5740 my $conf = load_config($vmid);
5741
5742 if (!$drivehash) {
5743 check_lock($conf);
5744 die "you can't delete a snapshot if vm is a template\n"
5745 if is_template($conf);
5746 }
5747
5748 $snap = $conf->{snapshots}->{$snapname};
5749
5750 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5751
5752 # remove parent refs
5753 if (!$prepare) {
5754 &$unlink_parent($conf, $snap->{parent});
5755 foreach my $sn (keys %{$conf->{snapshots}}) {
5756 next if $sn eq $snapname;
5757 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5758 }
5759 }
5760
5761 if ($remove_drive) {
5762 if ($remove_drive eq 'vmstate') {
5763 delete $snap->{$remove_drive};
5764 } else {
5765 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5766 my $volid = $drive->{file};
5767 delete $snap->{$remove_drive};
5768 add_unused_volume($conf, $volid);
5769 }
5770 }
5771
5772 if ($prepare) {
5773 $snap->{snapstate} = 'delete';
5774 } else {
5775 delete $conf->{snapshots}->{$snapname};
5776 delete $conf->{lock} if $drivehash;
5777 foreach my $volid (@$unused) {
5778 add_unused_volume($conf, $volid);
5779 }
5780 }
5781
5782 update_config_nolock($vmid, $conf, 1);
5783 };
5784
5785 lock_config($vmid, $updatefn);
5786
5787 # now remove vmstate file
5788
5789 my $storecfg = PVE::Storage::config();
5790
5791 if ($snap->{vmstate}) {
5792 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5793 if (my $err = $@) {
5794 die $err if !$force;
5795 warn $err;
5796 }
5797 # save changes (remove vmstate from snapshot)
5798 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5799 };
5800
5801 # now remove all internal snapshots
5802 foreach_drive($snap, sub {
5803 my ($ds, $drive) = @_;
5804
5805 return if drive_is_cdrom($drive);
5806
5807 my $volid = $drive->{file};
5808 my $device = "drive-$ds";
5809
5810 if (!$drivehash || $drivehash->{$ds}) {
5811 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5812 if (my $err = $@) {
5813 die $err if !$force;
5814 warn $err;
5815 }
5816 }
5817
5818 # save changes (remove drive fron snapshot)
5819 lock_config($vmid, $updatefn, $ds) if !$force;
5820 push @$unused, $volid;
5821 });
5822
5823 # now cleanup config
5824 $prepare = 0;
5825 lock_config($vmid, $updatefn);
5826 }
5827
5828 sub has_feature {
5829 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5830
5831 my $err;
5832 foreach_drive($conf, sub {
5833 my ($ds, $drive) = @_;
5834
5835 return if drive_is_cdrom($drive);
5836 my $volid = $drive->{file};
5837 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5838 });
5839
5840 return $err ? 0 : 1;
5841 }
5842
5843 sub template_create {
5844 my ($vmid, $conf, $disk) = @_;
5845
5846 my $storecfg = PVE::Storage::config();
5847
5848 foreach_drive($conf, sub {
5849 my ($ds, $drive) = @_;
5850
5851 return if drive_is_cdrom($drive);
5852 return if $disk && $ds ne $disk;
5853
5854 my $volid = $drive->{file};
5855 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5856
5857 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5858 $drive->{file} = $voliddst;
5859 $conf->{$ds} = print_drive($vmid, $drive);
5860 update_config_nolock($vmid, $conf, 1);
5861 });
5862 }
5863
5864 sub is_template {
5865 my ($conf) = @_;
5866
5867 return 1 if defined $conf->{template} && $conf->{template} == 1;
5868 }
5869
5870 sub qemu_img_convert {
5871 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5872
5873 my $storecfg = PVE::Storage::config();
5874 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
5875 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
5876
5877 if ($src_storeid && $dst_storeid) {
5878 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
5879 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5880
5881 my $src_format = qemu_img_format($src_scfg, $src_volname);
5882 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
5883
5884 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
5885 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5886
5887 my $cmd = [];
5888 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
5889 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
5890 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
5891
5892 my $parser = sub {
5893 my $line = shift;
5894 if($line =~ m/\((\S+)\/100\%\)/){
5895 my $percent = $1;
5896 my $transferred = int($size * $percent / 100);
5897 my $remaining = $size - $transferred;
5898
5899 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
5900 }
5901
5902 };
5903
5904 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
5905 my $err = $@;
5906 die "copy failed: $err" if $err;
5907 }
5908 }
5909
5910 sub qemu_img_format {
5911 my ($scfg, $volname) = @_;
5912
5913 if ($scfg->{path} && $volname =~ m/\.(raw|qcow2|qed|vmdk)$/) {
5914 return $1;
5915 } elsif ($scfg->{type} eq 'iscsi') {
5916 return "host_device";
5917 } else {
5918 return "raw";
5919 }
5920 }
5921
5922 sub qemu_drive_mirror {
5923 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
5924
5925 my $count = 0;
5926 my $old_len = 0;
5927 my $frozen = undef;
5928 my $maxwait = 120;
5929
5930 my $storecfg = PVE::Storage::config();
5931 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
5932
5933 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
5934
5935 my $format;
5936 if ($dst_volname =~ m/\.(raw|qcow2)$/){
5937 $format = $1;
5938 }
5939
5940 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
5941
5942 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
5943 $opts->{format} = $format if $format;
5944
5945 #fixme : sometime drive-mirror timeout, but works fine after.
5946 # (I have see the problem with big volume > 200GB), so we need to eval
5947 eval { vm_mon_cmd($vmid, "drive-mirror", %$opts); };
5948 # ignore errors here
5949
5950 eval {
5951 while (1) {
5952 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5953 my $stat = @$stats[0];
5954 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
5955 die "error job is not mirroring" if $stat->{type} ne "mirror";
5956
5957 my $busy = $stat->{busy};
5958
5959 if (my $total = $stat->{len}) {
5960 my $transferred = $stat->{offset} || 0;
5961 my $remaining = $total - $transferred;
5962 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
5963
5964 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy\n";
5965 }
5966
5967 if ($stat->{len} == $stat->{offset}) {
5968 if ($busy eq 'false') {
5969
5970 last if $vmiddst != $vmid;
5971
5972 # try to switch the disk if source and destination are on the same guest
5973 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
5974 last if !$@;
5975 die $@ if $@ !~ m/cannot be completed/;
5976 }
5977
5978 if ($count > $maxwait) {
5979 # if too much writes to disk occurs at the end of migration
5980 #the disk needs to be freezed to be able to complete the migration
5981 vm_suspend($vmid,1);
5982 $frozen = 1;
5983 }
5984 $count ++
5985 }
5986 $old_len = $stat->{offset};
5987 sleep 1;
5988 }
5989
5990 vm_resume($vmid, 1) if $frozen;
5991
5992 };
5993 my $err = $@;
5994
5995 my $cancel_job = sub {
5996 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
5997 while (1) {
5998 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
5999 my $stat = @$stats[0];
6000 last if !$stat;
6001 sleep 1;
6002 }
6003 };
6004
6005 if ($err) {
6006 eval { &$cancel_job(); };
6007 die "mirroring error: $err";
6008 }
6009
6010 if ($vmiddst != $vmid) {
6011 # if we clone a disk for a new target vm, we don't switch the disk
6012 &$cancel_job(); # so we call block-job-cancel
6013 }
6014 }
6015
6016 sub clone_disk {
6017 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6018 $newvmid, $storage, $format, $full, $newvollist) = @_;
6019
6020 my $newvolid;
6021
6022 if (!$full) {
6023 print "create linked clone of drive $drivename ($drive->{file})\n";
6024 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6025 push @$newvollist, $newvolid;
6026 } else {
6027 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6028 $storeid = $storage if $storage;
6029
6030 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6031 if (!$format) {
6032 $format = $drive->{format} || $defFormat;
6033 }
6034
6035 # test if requested format is supported - else use default
6036 my $supported = grep { $_ eq $format } @$validFormats;
6037 $format = $defFormat if !$supported;
6038
6039 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6040
6041 print "create full clone of drive $drivename ($drive->{file})\n";
6042 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6043 push @$newvollist, $newvolid;
6044
6045 if (!$running || $snapname) {
6046 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6047 } else {
6048 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6049 }
6050 }
6051
6052 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6053
6054 my $disk = $drive;
6055 $disk->{format} = undef;
6056 $disk->{file} = $newvolid;
6057 $disk->{size} = $size;
6058
6059 return $disk;
6060 }
6061
6062 # this only works if VM is running
6063 sub get_current_qemu_machine {
6064 my ($vmid) = @_;
6065
6066 my $cmd = { execute => 'query-machines', arguments => {} };
6067 my $res = vm_qmp_command($vmid, $cmd);
6068
6069 my ($current, $default);
6070 foreach my $e (@$res) {
6071 $default = $e->{name} if $e->{'is-default'};
6072 $current = $e->{name} if $e->{'is-current'};
6073 }
6074
6075 # fallback to the default machine if current is not supported by qemu
6076 return $current || $default || 'pc';
6077 }
6078
6079 sub lspci {
6080
6081 my $devices = {};
6082
6083 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6084 my (undef, $id, $function) = @_;
6085 my $res = { id => $id, function => $function};
6086 push @{$devices->{$id}}, $res;
6087 });
6088
6089 return $devices;
6090 }
6091
6092 1;