]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
qemuserver : vm_status : add extended stats (disks, nics, memory) V3
[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 $qemu_snap_storage = {rbd => 1, sheepdog => 1};
35
36 my $cpuinfo = PVE::ProcFSTools::read_cpuinfo();
37
38 # Note about locking: we use flock on the config file protect
39 # against concurent actions.
40 # Aditionaly, we have a 'lock' setting in the config file. This
41 # can be set to 'migrate', 'backup', 'snapshot' or 'rollback'. Most actions are not
42 # allowed when such lock is set. But you can ignore this kind of
43 # lock with the --skiplock flag.
44
45 cfs_register_file('/qemu-server/',
46 \&parse_vm_config,
47 \&write_vm_config);
48
49 PVE::JSONSchema::register_standard_option('skiplock', {
50 description => "Ignore locks - only root is allowed to use this option.",
51 type => 'boolean',
52 optional => 1,
53 });
54
55 PVE::JSONSchema::register_standard_option('pve-qm-stateuri', {
56 description => "Some command save/restore state from this location.",
57 type => 'string',
58 maxLength => 128,
59 optional => 1,
60 });
61
62 PVE::JSONSchema::register_standard_option('pve-snapshot-name', {
63 description => "The name of the snapshot.",
64 type => 'string', format => 'pve-configid',
65 maxLength => 40,
66 });
67
68 #no warnings 'redefine';
69
70 sub cgroups_write {
71 my ($controller, $vmid, $option, $value) = @_;
72
73 my $path = "/sys/fs/cgroup/$controller/qemu.slice/$vmid.scope/$option";
74 PVE::ProcFSTools::write_proc_entry($path, $value);
75
76 }
77
78 my $nodename = PVE::INotify::nodename();
79
80 mkdir "/etc/pve/nodes/$nodename";
81 my $confdir = "/etc/pve/nodes/$nodename/qemu-server";
82 mkdir $confdir;
83
84 my $var_run_tmpdir = "/var/run/qemu-server";
85 mkdir $var_run_tmpdir;
86
87 my $lock_dir = "/var/lock/qemu-server";
88 mkdir $lock_dir;
89
90 my $pcisysfs = "/sys/bus/pci";
91
92 my $confdesc = {
93 onboot => {
94 optional => 1,
95 type => 'boolean',
96 description => "Specifies whether a VM will be started during system bootup.",
97 default => 0,
98 },
99 autostart => {
100 optional => 1,
101 type => 'boolean',
102 description => "Automatic restart after crash (currently ignored).",
103 default => 0,
104 },
105 hotplug => {
106 optional => 1,
107 type => 'string', format => 'pve-hotplug-features',
108 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'.",
109 default => 'network,disk,usb',
110 },
111 reboot => {
112 optional => 1,
113 type => 'boolean',
114 description => "Allow reboot. If set to '0' the VM exit on reboot.",
115 default => 1,
116 },
117 lock => {
118 optional => 1,
119 type => 'string',
120 description => "Lock/unlock the VM.",
121 enum => [qw(migrate backup snapshot rollback)],
122 },
123 cpulimit => {
124 optional => 1,
125 type => 'number',
126 description => "Limit of CPU usage. Note if the computer has 2 CPUs, it has total of '2' CPU time. Value '0' indicates no CPU limit.",
127 minimum => 0,
128 maximum => 128,
129 default => 0,
130 },
131 cpuunits => {
132 optional => 1,
133 type => 'integer',
134 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.",
135 minimum => 0,
136 maximum => 500000,
137 default => 1000,
138 },
139 memory => {
140 optional => 1,
141 type => 'integer',
142 description => "Amount of RAM for the VM in MB. This is the maximum available memory when you use the balloon device.",
143 minimum => 16,
144 default => 512,
145 },
146 balloon => {
147 optional => 1,
148 type => 'integer',
149 description => "Amount of target RAM for the VM in MB. Using zero disables the ballon driver.",
150 minimum => 0,
151 },
152 shares => {
153 optional => 1,
154 type => 'integer',
155 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",
156 minimum => 0,
157 maximum => 50000,
158 default => 1000,
159 },
160 keyboard => {
161 optional => 1,
162 type => 'string',
163 description => "Keybord layout for vnc server. Default is read from the datacenter configuration file.",
164 enum => PVE::Tools::kvmkeymaplist(),
165 default => 'en-us',
166 },
167 name => {
168 optional => 1,
169 type => 'string', format => 'dns-name',
170 description => "Set a name for the VM. Only used on the configuration web interface.",
171 },
172 scsihw => {
173 optional => 1,
174 type => 'string',
175 description => "scsi controller model",
176 enum => [qw(lsi lsi53c810 virtio-scsi-pci virtio-scsi-single megasas pvscsi)],
177 default => 'lsi',
178 },
179 description => {
180 optional => 1,
181 type => 'string',
182 description => "Description for the VM. Only used on the configuration web interface. This is saved as comment inside the configuration file.",
183 },
184 ostype => {
185 optional => 1,
186 type => 'string',
187 enum => [qw(other wxp w2k w2k3 w2k8 wvista win7 win8 l24 l26 solaris)],
188 description => <<EODESC,
189 Used to enable special optimization/features for specific
190 operating systems:
191
192 other => unspecified OS
193 wxp => Microsoft Windows XP
194 w2k => Microsoft Windows 2000
195 w2k3 => Microsoft Windows 2003
196 w2k8 => Microsoft Windows 2008
197 wvista => Microsoft Windows Vista
198 win7 => Microsoft Windows 7
199 win8 => Microsoft Windows 8/2012
200 l24 => Linux 2.4 Kernel
201 l26 => Linux 2.6/3.X Kernel
202 solaris => solaris/opensolaris/openindiania kernel
203
204 other|l24|l26|solaris ... no special behaviour
205 wxp|w2k|w2k3|w2k8|wvista|win7|win8 ... use --localtime switch
206 EODESC
207 },
208 boot => {
209 optional => 1,
210 type => 'string',
211 description => "Boot on floppy (a), hard disk (c), CD-ROM (d), or network (n).",
212 pattern => '[acdn]{1,4}',
213 default => 'cdn',
214 },
215 bootdisk => {
216 optional => 1,
217 type => 'string', format => 'pve-qm-bootdisk',
218 description => "Enable booting from specified disk.",
219 pattern => '(ide|sata|scsi|virtio)\d+',
220 },
221 smp => {
222 optional => 1,
223 type => 'integer',
224 description => "The number of CPUs. Please use option -sockets instead.",
225 minimum => 1,
226 default => 1,
227 },
228 sockets => {
229 optional => 1,
230 type => 'integer',
231 description => "The number of CPU sockets.",
232 minimum => 1,
233 default => 1,
234 },
235 cores => {
236 optional => 1,
237 type => 'integer',
238 description => "The number of cores per socket.",
239 minimum => 1,
240 default => 1,
241 },
242 numa => {
243 optional => 1,
244 type => 'boolean',
245 description => "Enable/disable Numa.",
246 default => 0,
247 },
248 vcpus => {
249 optional => 1,
250 type => 'integer',
251 description => "Number of hotplugged vcpus.",
252 minimum => 1,
253 default => 0,
254 },
255 acpi => {
256 optional => 1,
257 type => 'boolean',
258 description => "Enable/disable ACPI.",
259 default => 1,
260 },
261 agent => {
262 optional => 1,
263 type => 'boolean',
264 description => "Enable/disable Qemu GuestAgent.",
265 default => 0,
266 },
267 kvm => {
268 optional => 1,
269 type => 'boolean',
270 description => "Enable/disable KVM hardware virtualization.",
271 default => 1,
272 },
273 tdf => {
274 optional => 1,
275 type => 'boolean',
276 description => "Enable/disable time drift fix.",
277 default => 0,
278 },
279 localtime => {
280 optional => 1,
281 type => 'boolean',
282 description => "Set the real time clock to local time. This is enabled by default if ostype indicates a Microsoft OS.",
283 },
284 freeze => {
285 optional => 1,
286 type => 'boolean',
287 description => "Freeze CPU at startup (use 'c' monitor command to start execution).",
288 },
289 vga => {
290 optional => 1,
291 type => 'string',
292 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.",
293 enum => [qw(std cirrus vmware qxl serial0 serial1 serial2 serial3 qxl2 qxl3 qxl4)],
294 },
295 watchdog => {
296 optional => 1,
297 type => 'string', format => 'pve-qm-watchdog',
298 typetext => '[[model=]i6300esb|ib700] [,[action=]reset|shutdown|poweroff|pause|debug|none]',
299 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)",
300 },
301 startdate => {
302 optional => 1,
303 type => 'string',
304 typetext => "(now | YYYY-MM-DD | YYYY-MM-DDTHH:MM:SS)",
305 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'.",
306 pattern => '(now|\d{4}-\d{1,2}-\d{1,2}(T\d{1,2}:\d{1,2}:\d{1,2})?)',
307 default => 'now',
308 },
309 startup => get_standard_option('pve-startup-order'),
310 template => {
311 optional => 1,
312 type => 'boolean',
313 description => "Enable/disable Template.",
314 default => 0,
315 },
316 args => {
317 optional => 1,
318 type => 'string',
319 description => <<EODESCR,
320 Note: this option is for experts only. It allows you to pass arbitrary arguments to kvm, for example:
321
322 args: -no-reboot -no-hpet
323 EODESCR
324 },
325 tablet => {
326 optional => 1,
327 type => 'boolean',
328 default => 1,
329 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).",
330 },
331 migrate_speed => {
332 optional => 1,
333 type => 'integer',
334 description => "Set maximum speed (in MB/s) for migrations. Value 0 is no limit.",
335 minimum => 0,
336 default => 0,
337 },
338 migrate_downtime => {
339 optional => 1,
340 type => 'number',
341 description => "Set maximum tolerated downtime (in seconds) for migrations.",
342 minimum => 0,
343 default => 0.1,
344 },
345 cdrom => {
346 optional => 1,
347 type => 'string', format => 'pve-qm-drive',
348 typetext => 'volume',
349 description => "This is an alias for option -ide2",
350 },
351 cpu => {
352 optional => 1,
353 description => "Emulated CPU type.",
354 type => 'string',
355 enum => [ qw(486 athlon pentium pentium2 pentium3 coreduo core2duo kvm32 kvm64 qemu32 qemu64 phenom Conroe Penryn Nehalem Westmere SandyBridge IvyBridge Haswell Broadwell Opteron_G1 Opteron_G2 Opteron_G3 Opteron_G4 Opteron_G5 host) ],
356 default => 'kvm64',
357 },
358 parent => get_standard_option('pve-snapshot-name', {
359 optional => 1,
360 description => "Parent snapshot name. This is used internally, and should not be modified.",
361 }),
362 snaptime => {
363 optional => 1,
364 description => "Timestamp for snapshots.",
365 type => 'integer',
366 minimum => 0,
367 },
368 vmstate => {
369 optional => 1,
370 type => 'string', format => 'pve-volume-id',
371 description => "Reference to a volume which stores the VM state. This is used internally for snapshots.",
372 },
373 machine => {
374 description => "Specific the Qemu machine type.",
375 type => 'string',
376 pattern => '(pc|pc(-i440fx)?-\d+\.\d+|q35|pc-q35-\d+\.\d+)',
377 maxLength => 40,
378 optional => 1,
379 },
380 smbios1 => {
381 description => "Specify SMBIOS type 1 fields.",
382 type => 'string', format => 'pve-qm-smbios1',
383 typetext => "[manufacturer=str][,product=str][,version=str][,serial=str] [,uuid=uuid][,sku=str][,family=str]",
384 maxLength => 256,
385 optional => 1,
386 },
387 };
388
389 # what about other qemu settings ?
390 #cpu => 'string',
391 #machine => 'string',
392 #fda => 'file',
393 #fdb => 'file',
394 #mtdblock => 'file',
395 #sd => 'file',
396 #pflash => 'file',
397 #snapshot => 'bool',
398 #bootp => 'file',
399 ##tftp => 'dir',
400 ##smb => 'dir',
401 #kernel => 'file',
402 #append => 'string',
403 #initrd => 'file',
404 ##soundhw => 'string',
405
406 while (my ($k, $v) = each %$confdesc) {
407 PVE::JSONSchema::register_standard_option("pve-qm-$k", $v);
408 }
409
410 my $MAX_IDE_DISKS = 4;
411 my $MAX_SCSI_DISKS = 14;
412 my $MAX_VIRTIO_DISKS = 16;
413 my $MAX_SATA_DISKS = 6;
414 my $MAX_USB_DEVICES = 5;
415 my $MAX_NETS = 32;
416 my $MAX_UNUSED_DISKS = 8;
417 my $MAX_HOSTPCI_DEVICES = 4;
418 my $MAX_SERIAL_PORTS = 4;
419 my $MAX_PARALLEL_PORTS = 3;
420 my $MAX_NUMA = 8;
421 my $MAX_MEM = 4194304;
422 my $STATICMEM = 1024;
423
424 my $numadesc = {
425 optional => 1,
426 type => 'string', format => 'pve-qm-numanode',
427 typetext => "cpus=<id[-id],memory=<mb>[[,hostnodes=<id[-id]>] [,policy=<preferred|bind|interleave>]]",
428 description => "numa topology",
429 };
430 PVE::JSONSchema::register_standard_option("pve-qm-numanode", $numadesc);
431
432 for (my $i = 0; $i < $MAX_NUMA; $i++) {
433 $confdesc->{"numa$i"} = $numadesc;
434 }
435
436 my $nic_model_list = ['rtl8139', 'ne2k_pci', 'e1000', 'pcnet', 'virtio',
437 'ne2k_isa', 'i82551', 'i82557b', 'i82559er', 'vmxnet3',
438 'e1000-82540em', 'e1000-82544gc', 'e1000-82545em'];
439 my $nic_model_list_txt = join(' ', sort @$nic_model_list);
440
441 my $netdesc = {
442 optional => 1,
443 type => 'string', format => 'pve-qm-net',
444 typetext => "MODEL=XX:XX:XX:XX:XX:XX [,bridge=<dev>][,queues=<nbqueues>][,rate=<mbps>] [,tag=<vlanid>][,firewall=0|1],link_down=0|1]",
445 description => <<EODESCR,
446 Specify network devices.
447
448 MODEL is one of: $nic_model_list_txt
449
450 XX:XX:XX:XX:XX:XX should be an unique MAC address. This is
451 automatically generated if not specified.
452
453 The bridge parameter can be used to automatically add the interface to a bridge device. The Proxmox VE standard bridge is called 'vmbr0'.
454
455 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'.
456
457 If you specify no bridge, we create a kvm 'user' (NATed) network device, which provides DHCP and DNS services. The following addresses are used:
458
459 10.0.2.2 Gateway
460 10.0.2.3 DNS Server
461 10.0.2.4 SMB Server
462
463 The DHCP server assign addresses to the guest starting from 10.0.2.15.
464
465 EODESCR
466 };
467 PVE::JSONSchema::register_standard_option("pve-qm-net", $netdesc);
468
469 for (my $i = 0; $i < $MAX_NETS; $i++) {
470 $confdesc->{"net$i"} = $netdesc;
471 }
472
473 my $drivename_hash;
474
475 my $idedesc = {
476 optional => 1,
477 type => 'string', format => 'pve-qm-drive',
478 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
479 description => "Use volume as IDE hard disk or CD-ROM (n is 0 to " .($MAX_IDE_DISKS -1) . ").",
480 };
481 PVE::JSONSchema::register_standard_option("pve-qm-ide", $idedesc);
482
483 my $scsidesc = {
484 optional => 1,
485 type => 'string', format => 'pve-qm-drive',
486 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on] [,iothread=on] [,queues=<nbqueues>]',
487 description => "Use volume as SCSI hard disk or CD-ROM (n is 0 to " . ($MAX_SCSI_DISKS - 1) . ").",
488 };
489 PVE::JSONSchema::register_standard_option("pve-qm-scsi", $scsidesc);
490
491 my $satadesc = {
492 optional => 1,
493 type => 'string', format => 'pve-qm-drive',
494 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on]',
495 description => "Use volume as SATA hard disk or CD-ROM (n is 0 to " . ($MAX_SATA_DISKS - 1). ").",
496 };
497 PVE::JSONSchema::register_standard_option("pve-qm-sata", $satadesc);
498
499 my $virtiodesc = {
500 optional => 1,
501 type => 'string', format => 'pve-qm-drive',
502 typetext => '[volume=]volume,] [,media=cdrom|disk] [,cyls=c,heads=h,secs=s[,trans=t]] [,snapshot=on|off] [,cache=none|writethrough|writeback|unsafe|directsync] [,format=f] [,backup=yes|no] [,rerror=ignore|report|stop] [,werror=enospc|ignore|report|stop] [,aio=native|threads] [,discard=ignore|on] [,iothread=on]',
503 description => "Use volume as VIRTIO hard disk (n is 0 to " . ($MAX_VIRTIO_DISKS - 1) . ").",
504 };
505 PVE::JSONSchema::register_standard_option("pve-qm-virtio", $virtiodesc);
506
507 my $usbdesc = {
508 optional => 1,
509 type => 'string', format => 'pve-qm-usb-device',
510 typetext => 'host=HOSTUSBDEVICE|spice',
511 description => <<EODESCR,
512 Configure an USB device (n is 0 to 4). This can be used to
513 pass-through usb devices to the guest. HOSTUSBDEVICE syntax is:
514
515 'bus-port(.port)*' (decimal numbers) or
516 'vendor_id:product_id' (hexadeciaml numbers)
517
518 You can use the 'lsusb -t' command to list existing usb devices.
519
520 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
521
522 The value 'spice' can be used to add a usb redirection devices for spice.
523
524 EODESCR
525 };
526 PVE::JSONSchema::register_standard_option("pve-qm-usb", $usbdesc);
527
528 my $hostpcidesc = {
529 optional => 1,
530 type => 'string', format => 'pve-qm-hostpci',
531 typetext => "[host=]HOSTPCIDEVICE [,driver=kvm|vfio] [,rombar=on|off] [,pcie=0|1] [,x-vga=on|off]",
532 description => <<EODESCR,
533 Map host pci devices. HOSTPCIDEVICE syntax is:
534
535 'bus:dev.func' (hexadecimal numbers)
536
537 You can us the 'lspci' command to list existing pci devices.
538
539 The 'rombar' option determines whether or not the device's ROM will be visible in the guest's memory map (default is 'on').
540
541 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
542
543 Experimental: user reported problems with this option.
544 EODESCR
545 };
546 PVE::JSONSchema::register_standard_option("pve-qm-hostpci", $hostpcidesc);
547
548 my $serialdesc = {
549 optional => 1,
550 type => 'string',
551 pattern => '(/dev/.+|socket)',
552 description => <<EODESCR,
553 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).
554
555 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
556
557 Experimental: user reported problems with this option.
558 EODESCR
559 };
560
561 my $paralleldesc= {
562 optional => 1,
563 type => 'string',
564 pattern => '/dev/parport\d+|/dev/usb/lp\d+',
565 description => <<EODESCR,
566 Map host parallel devices (n is 0 to 2).
567
568 Note: This option allows direct access to host hardware. So it is no longer possible to migrate such machines - use with special care.
569
570 Experimental: user reported problems with this option.
571 EODESCR
572 };
573
574 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
575 $confdesc->{"parallel$i"} = $paralleldesc;
576 }
577
578 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
579 $confdesc->{"serial$i"} = $serialdesc;
580 }
581
582 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
583 $confdesc->{"hostpci$i"} = $hostpcidesc;
584 }
585
586 for (my $i = 0; $i < $MAX_IDE_DISKS; $i++) {
587 $drivename_hash->{"ide$i"} = 1;
588 $confdesc->{"ide$i"} = $idedesc;
589 }
590
591 for (my $i = 0; $i < $MAX_SATA_DISKS; $i++) {
592 $drivename_hash->{"sata$i"} = 1;
593 $confdesc->{"sata$i"} = $satadesc;
594 }
595
596 for (my $i = 0; $i < $MAX_SCSI_DISKS; $i++) {
597 $drivename_hash->{"scsi$i"} = 1;
598 $confdesc->{"scsi$i"} = $scsidesc ;
599 }
600
601 for (my $i = 0; $i < $MAX_VIRTIO_DISKS; $i++) {
602 $drivename_hash->{"virtio$i"} = 1;
603 $confdesc->{"virtio$i"} = $virtiodesc;
604 }
605
606 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
607 $confdesc->{"usb$i"} = $usbdesc;
608 }
609
610 my $unuseddesc = {
611 optional => 1,
612 type => 'string', format => 'pve-volume-id',
613 description => "Reference to unused volumes.",
614 };
615
616 for (my $i = 0; $i < $MAX_UNUSED_DISKS; $i++) {
617 $confdesc->{"unused$i"} = $unuseddesc;
618 }
619
620 my $kvm_api_version = 0;
621
622 sub kvm_version {
623
624 return $kvm_api_version if $kvm_api_version;
625
626 my $fh = IO::File->new("</dev/kvm") ||
627 return 0;
628
629 if (my $v = $fh->ioctl(KVM_GET_API_VERSION(), 0)) {
630 $kvm_api_version = $v;
631 }
632
633 $fh->close();
634
635 return $kvm_api_version;
636 }
637
638 my $kvm_user_version;
639
640 sub kvm_user_version {
641
642 return $kvm_user_version if $kvm_user_version;
643
644 $kvm_user_version = 'unknown';
645
646 my $tmp = `kvm -help 2>/dev/null`;
647
648 if ($tmp =~ m/^QEMU( PC)? emulator version (\d+\.\d+(\.\d+)?)[,\s]/) {
649 $kvm_user_version = $2;
650 }
651
652 return $kvm_user_version;
653
654 }
655
656 my $kernel_has_vhost_net = -c '/dev/vhost-net';
657
658 sub disknames {
659 # order is important - used to autoselect boot disk
660 return ((map { "ide$_" } (0 .. ($MAX_IDE_DISKS - 1))),
661 (map { "scsi$_" } (0 .. ($MAX_SCSI_DISKS - 1))),
662 (map { "virtio$_" } (0 .. ($MAX_VIRTIO_DISKS - 1))),
663 (map { "sata$_" } (0 .. ($MAX_SATA_DISKS - 1))));
664 }
665
666 sub valid_drivename {
667 my $dev = shift;
668
669 return defined($drivename_hash->{$dev});
670 }
671
672 sub option_exists {
673 my $key = shift;
674 return defined($confdesc->{$key});
675 }
676
677 sub nic_models {
678 return $nic_model_list;
679 }
680
681 sub os_list_description {
682
683 return {
684 other => 'Other',
685 wxp => 'Windows XP',
686 w2k => 'Windows 2000',
687 w2k3 =>, 'Windows 2003',
688 w2k8 => 'Windows 2008',
689 wvista => 'Windows Vista',
690 win7 => 'Windows 7',
691 win8 => 'Windows 8/2012',
692 l24 => 'Linux 2.4',
693 l26 => 'Linux 2.6',
694 };
695 }
696
697 my $cdrom_path;
698
699 sub get_cdrom_path {
700
701 return $cdrom_path if $cdrom_path;
702
703 return $cdrom_path = "/dev/cdrom" if -l "/dev/cdrom";
704 return $cdrom_path = "/dev/cdrom1" if -l "/dev/cdrom1";
705 return $cdrom_path = "/dev/cdrom2" if -l "/dev/cdrom2";
706 }
707
708 sub get_iso_path {
709 my ($storecfg, $vmid, $cdrom) = @_;
710
711 if ($cdrom eq 'cdrom') {
712 return get_cdrom_path();
713 } elsif ($cdrom eq 'none') {
714 return '';
715 } elsif ($cdrom =~ m|^/|) {
716 return $cdrom;
717 } else {
718 return PVE::Storage::path($storecfg, $cdrom);
719 }
720 }
721
722 # try to convert old style file names to volume IDs
723 sub filename_to_volume_id {
724 my ($vmid, $file, $media) = @_;
725
726 if (!($file eq 'none' || $file eq 'cdrom' ||
727 $file =~ m|^/dev/.+| || $file =~ m/^([^:]+):(.+)$/)) {
728
729 return undef if $file =~ m|/|;
730
731 if ($media && $media eq 'cdrom') {
732 $file = "local:iso/$file";
733 } else {
734 $file = "local:$vmid/$file";
735 }
736 }
737
738 return $file;
739 }
740
741 sub verify_media_type {
742 my ($opt, $vtype, $media) = @_;
743
744 return if !$media;
745
746 my $etype;
747 if ($media eq 'disk') {
748 $etype = 'images';
749 } elsif ($media eq 'cdrom') {
750 $etype = 'iso';
751 } else {
752 die "internal error";
753 }
754
755 return if ($vtype eq $etype);
756
757 raise_param_exc({ $opt => "unexpected media type ($vtype != $etype)" });
758 }
759
760 sub cleanup_drive_path {
761 my ($opt, $storecfg, $drive) = @_;
762
763 # try to convert filesystem paths to volume IDs
764
765 if (($drive->{file} !~ m/^(cdrom|none)$/) &&
766 ($drive->{file} !~ m|^/dev/.+|) &&
767 ($drive->{file} !~ m/^([^:]+):(.+)$/) &&
768 ($drive->{file} !~ m/^\d+$/)) {
769 my ($vtype, $volid) = PVE::Storage::path_to_volume_id($storecfg, $drive->{file});
770 raise_param_exc({ $opt => "unable to associate path '$drive->{file}' to any storage"}) if !$vtype;
771 $drive->{media} = 'cdrom' if !$drive->{media} && $vtype eq 'iso';
772 verify_media_type($opt, $vtype, $drive->{media});
773 $drive->{file} = $volid;
774 }
775
776 $drive->{media} = 'cdrom' if !$drive->{media} && $drive->{file} =~ m/^(cdrom|none)$/;
777 }
778
779 sub create_conf_nolock {
780 my ($vmid, $settings) = @_;
781
782 my $filename = config_file($vmid);
783
784 die "configuration file '$filename' already exists\n" if -f $filename;
785
786 my $defaults = load_defaults();
787
788 $settings->{name} = "vm$vmid" if !$settings->{name};
789 $settings->{memory} = $defaults->{memory} if !$settings->{memory};
790
791 my $data = '';
792 foreach my $opt (keys %$settings) {
793 next if !$confdesc->{$opt};
794
795 my $value = $settings->{$opt};
796 next if !$value;
797
798 $data .= "$opt: $value\n";
799 }
800
801 PVE::Tools::file_set_contents($filename, $data);
802 }
803
804 sub parse_hotplug_features {
805 my ($data) = @_;
806
807 my $res = {};
808
809 return $res if $data eq '0';
810
811 $data = $confdesc->{hotplug}->{default} if $data eq '1';
812
813 foreach my $feature (PVE::Tools::split_list($data)) {
814 if ($feature =~ m/^(network|disk|cpu|memory|usb)$/) {
815 $res->{$1} = 1;
816 } else {
817 warn "ignoring unknown hotplug feature '$feature'\n";
818 }
819 }
820 return $res;
821 }
822
823 PVE::JSONSchema::register_format('pve-hotplug-features', \&pve_verify_hotplug_features);
824 sub pve_verify_hotplug_features {
825 my ($value, $noerr) = @_;
826
827 return $value if parse_hotplug_features($value);
828
829 return undef if $noerr;
830
831 die "unable to parse hotplug option\n";
832 }
833
834 my $parse_size = sub {
835 my ($value) = @_;
836
837 return undef if $value !~ m/^(\d+(\.\d+)?)([KMG])?$/;
838 my ($size, $unit) = ($1, $3);
839 if ($unit) {
840 if ($unit eq 'K') {
841 $size = $size * 1024;
842 } elsif ($unit eq 'M') {
843 $size = $size * 1024 * 1024;
844 } elsif ($unit eq 'G') {
845 $size = $size * 1024 * 1024 * 1024;
846 }
847 }
848 return int($size);
849 };
850
851 my $format_size = sub {
852 my ($size) = @_;
853
854 $size = int($size);
855
856 my $kb = int($size/1024);
857 return $size if $kb*1024 != $size;
858
859 my $mb = int($kb/1024);
860 return "${kb}K" if $mb*1024 != $kb;
861
862 my $gb = int($mb/1024);
863 return "${mb}M" if $gb*1024 != $mb;
864
865 return "${gb}G";
866 };
867
868 # ideX = [volume=]volume-id[,media=d][,cyls=c,heads=h,secs=s[,trans=t]]
869 # [,snapshot=on|off][,cache=on|off][,format=f][,backup=yes|no]
870 # [,rerror=ignore|report|stop][,werror=enospc|ignore|report|stop]
871 # [,aio=native|threads][,discard=ignore|on][,iothread=on]
872
873 sub parse_drive {
874 my ($key, $data) = @_;
875
876 my $res = {};
877
878 # $key may be undefined - used to verify JSON parameters
879 if (!defined($key)) {
880 $res->{interface} = 'unknown'; # should not harm when used to verify parameters
881 $res->{index} = 0;
882 } elsif ($key =~ m/^([^\d]+)(\d+)$/) {
883 $res->{interface} = $1;
884 $res->{index} = $2;
885 } else {
886 return undef;
887 }
888
889 foreach my $p (split (/,/, $data)) {
890 next if $p =~ m/^\s*$/;
891
892 if ($p =~ m/^(file|volume|cyls|heads|secs|trans|media|snapshot|cache|format|rerror|werror|backup|aio|bps|mbps|mbps_max|bps_rd|mbps_rd|mbps_rd_max|bps_wr|mbps_wr|mbps_wr_max|iops|iops_max|iops_rd|iops_rd_max|iops_wr|iops_wr_max|size|discard|iothread|queues)=(.+)$/) {
893 my ($k, $v) = ($1, $2);
894
895 $k = 'file' if $k eq 'volume';
896
897 return undef if defined $res->{$k};
898
899 if ($k eq 'bps' || $k eq 'bps_rd' || $k eq 'bps_wr') {
900 return undef if !$v || $v !~ m/^\d+/;
901 $k = "m$k";
902 $v = sprintf("%.3f", $v / (1024*1024));
903 }
904 $res->{$k} = $v;
905 } else {
906 if (!$res->{file} && $p !~ m/=/) {
907 $res->{file} = $p;
908 } else {
909 return undef;
910 }
911 }
912 }
913
914 return undef if !$res->{file};
915
916 return undef if $res->{cache} &&
917 $res->{cache} !~ m/^(off|none|writethrough|writeback|unsafe|directsync)$/;
918 return undef if $res->{snapshot} && $res->{snapshot} !~ m/^(on|off)$/;
919 return undef if $res->{cyls} && $res->{cyls} !~ m/^\d+$/;
920 return undef if $res->{heads} && $res->{heads} !~ m/^\d+$/;
921 return undef if $res->{secs} && $res->{secs} !~ m/^\d+$/;
922 return undef if $res->{media} && $res->{media} !~ m/^(disk|cdrom)$/;
923 return undef if $res->{trans} && $res->{trans} !~ m/^(none|lba|auto)$/;
924 return undef if $res->{format} && $res->{format} !~ m/^(raw|cow|qcow|qed|qcow2|vmdk|cloop)$/;
925 return undef if $res->{rerror} && $res->{rerror} !~ m/^(ignore|report|stop)$/;
926 return undef if $res->{werror} && $res->{werror} !~ m/^(enospc|ignore|report|stop)$/;
927 return undef if $res->{backup} && $res->{backup} !~ m/^(yes|no)$/;
928 return undef if $res->{aio} && $res->{aio} !~ m/^(native|threads)$/;
929 return undef if $res->{discard} && $res->{discard} !~ m/^(ignore|on)$/;
930 return undef if $res->{iothread} && $res->{iothread} !~ m/^(on)$/;
931 return undef if $res->{queues} && ($res->{queues} !~ m/^\d+$/ || $res->{queues} < 2);
932
933 return undef if $res->{mbps_rd} && $res->{mbps};
934 return undef if $res->{mbps_wr} && $res->{mbps};
935
936 return undef if $res->{mbps} && $res->{mbps} !~ m/^\d+(\.\d+)?$/;
937 return undef if $res->{mbps_max} && $res->{mbps_max} !~ m/^\d+(\.\d+)?$/;
938 return undef if $res->{mbps_rd} && $res->{mbps_rd} !~ m/^\d+(\.\d+)?$/;
939 return undef if $res->{mbps_rd_max} && $res->{mbps_rd_max} !~ m/^\d+(\.\d+)?$/;
940 return undef if $res->{mbps_wr} && $res->{mbps_wr} !~ m/^\d+(\.\d+)?$/;
941 return undef if $res->{mbps_wr_max} && $res->{mbps_wr_max} !~ m/^\d+(\.\d+)?$/;
942
943 return undef if $res->{iops_rd} && $res->{iops};
944 return undef if $res->{iops_wr} && $res->{iops};
945
946
947 return undef if $res->{iops} && $res->{iops} !~ m/^\d+$/;
948 return undef if $res->{iops_max} && $res->{iops_max} !~ m/^\d+$/;
949 return undef if $res->{iops_rd} && $res->{iops_rd} !~ m/^\d+$/;
950 return undef if $res->{iops_rd_max} && $res->{iops_rd_max} !~ m/^\d+$/;
951 return undef if $res->{iops_wr} && $res->{iops_wr} !~ m/^\d+$/;
952 return undef if $res->{iops_wr_max} && $res->{iops_wr_max} !~ m/^\d+$/;
953
954
955 if ($res->{size}) {
956 return undef if !defined($res->{size} = &$parse_size($res->{size}));
957 }
958
959 if ($res->{media} && ($res->{media} eq 'cdrom')) {
960 return undef if $res->{snapshot} || $res->{trans} || $res->{format};
961 return undef if $res->{heads} || $res->{secs} || $res->{cyls};
962 return undef if $res->{interface} eq 'virtio';
963 }
964
965 # rerror does not work with scsi drives
966 if ($res->{rerror}) {
967 return undef if $res->{interface} eq 'scsi';
968 }
969
970 return $res;
971 }
972
973 my @qemu_drive_options = qw(heads secs cyls trans media format cache snapshot rerror werror aio discard iops iops_rd iops_wr iops_max iops_rd_max iops_wr_max);
974
975 sub print_drive {
976 my ($vmid, $drive) = @_;
977
978 my $opts = '';
979 foreach my $o (@qemu_drive_options, 'mbps', 'mbps_rd', 'mbps_wr', 'mbps_max', 'mbps_rd_max', 'mbps_wr_max', 'backup', 'iothread', 'queues') {
980 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
981 }
982
983 if ($drive->{size}) {
984 $opts .= ",size=" . &$format_size($drive->{size});
985 }
986
987 return "$drive->{file}$opts";
988 }
989
990 sub scsi_inquiry {
991 my($fh, $noerr) = @_;
992
993 my $SG_IO = 0x2285;
994 my $SG_GET_VERSION_NUM = 0x2282;
995
996 my $versionbuf = "\x00" x 8;
997 my $ret = ioctl($fh, $SG_GET_VERSION_NUM, $versionbuf);
998 if (!$ret) {
999 die "scsi ioctl SG_GET_VERSION_NUM failoed - $!\n" if !$noerr;
1000 return undef;
1001 }
1002 my $version = unpack("I", $versionbuf);
1003 if ($version < 30000) {
1004 die "scsi generic interface too old\n" if !$noerr;
1005 return undef;
1006 }
1007
1008 my $buf = "\x00" x 36;
1009 my $sensebuf = "\x00" x 8;
1010 my $cmd = pack("C x3 C x1", 0x12, 36);
1011
1012 # see /usr/include/scsi/sg.h
1013 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";
1014
1015 my $packet = pack($sg_io_hdr_t, ord('S'), -3, length($cmd),
1016 length($sensebuf), 0, length($buf), $buf,
1017 $cmd, $sensebuf, 6000);
1018
1019 $ret = ioctl($fh, $SG_IO, $packet);
1020 if (!$ret) {
1021 die "scsi ioctl SG_IO failed - $!\n" if !$noerr;
1022 return undef;
1023 }
1024
1025 my @res = unpack($sg_io_hdr_t, $packet);
1026 if ($res[17] || $res[18]) {
1027 die "scsi ioctl SG_IO status error - $!\n" if !$noerr;
1028 return undef;
1029 }
1030
1031 my $res = {};
1032 (my $byte0, my $byte1, $res->{vendor},
1033 $res->{product}, $res->{revision}) = unpack("C C x6 A8 A16 A4", $buf);
1034
1035 $res->{removable} = $byte1 & 128 ? 1 : 0;
1036 $res->{type} = $byte0 & 31;
1037
1038 return $res;
1039 }
1040
1041 sub path_is_scsi {
1042 my ($path) = @_;
1043
1044 my $fh = IO::File->new("+<$path") || return undef;
1045 my $res = scsi_inquiry($fh, 1);
1046 close($fh);
1047
1048 return $res;
1049 }
1050
1051 sub machine_type_is_q35 {
1052 my ($conf) = @_;
1053
1054 return $conf->{machine} && ($conf->{machine} =~ m/q35/) ? 1 : 0;
1055 }
1056
1057 sub print_tabletdevice_full {
1058 my ($conf) = @_;
1059
1060 my $q35 = machine_type_is_q35($conf);
1061
1062 # we use uhci for old VMs because tablet driver was buggy in older qemu
1063 my $usbbus = $q35 ? "ehci" : "uhci";
1064
1065 return "usb-tablet,id=tablet,bus=$usbbus.0,port=1";
1066 }
1067
1068 sub print_drivedevice_full {
1069 my ($storecfg, $conf, $vmid, $drive, $bridges) = @_;
1070
1071 my $device = '';
1072 my $maxdev = 0;
1073
1074 if ($drive->{interface} eq 'virtio') {
1075 my $pciaddr = print_pci_addr("$drive->{interface}$drive->{index}", $bridges);
1076 $device = "virtio-blk-pci,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}$pciaddr";
1077 $device .= ",iothread=iothread-$drive->{interface}$drive->{index}" if $drive->{iothread};
1078 } elsif ($drive->{interface} eq 'scsi') {
1079
1080 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
1081 my $unit = $drive->{index} % $maxdev;
1082 my $devicetype = 'hd';
1083 my $path = '';
1084 if (drive_is_cdrom($drive)) {
1085 $devicetype = 'cd';
1086 } else {
1087 if ($drive->{file} =~ m|^/|) {
1088 $path = $drive->{file};
1089 } else {
1090 $path = PVE::Storage::path($storecfg, $drive->{file});
1091 }
1092
1093 if($path =~ m/^iscsi\:\/\//){
1094 $devicetype = 'generic';
1095 } else {
1096 if (my $info = path_is_scsi($path)) {
1097 if ($info->{type} == 0) {
1098 $devicetype = 'block';
1099 } elsif ($info->{type} == 1) { # tape
1100 $devicetype = 'generic';
1101 }
1102 }
1103 }
1104 }
1105
1106 if (!$conf->{scsihw} || ($conf->{scsihw} =~ m/^lsi/)){
1107 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,scsi-id=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1108 } else {
1109 $device = "scsi-$devicetype,bus=$controller_prefix$controller.0,channel=0,scsi-id=0,lun=$drive->{index},drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1110 }
1111
1112 } elsif ($drive->{interface} eq 'ide'){
1113 $maxdev = 2;
1114 my $controller = int($drive->{index} / $maxdev);
1115 my $unit = $drive->{index} % $maxdev;
1116 my $devicetype = ($drive->{media} && $drive->{media} eq 'cdrom') ? "cd" : "hd";
1117
1118 $device = "ide-$devicetype,bus=ide.$controller,unit=$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1119 } elsif ($drive->{interface} eq 'sata'){
1120 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
1121 my $unit = $drive->{index} % $MAX_SATA_DISKS;
1122 $device = "ide-drive,bus=ahci$controller.$unit,drive=drive-$drive->{interface}$drive->{index},id=$drive->{interface}$drive->{index}";
1123 } elsif ($drive->{interface} eq 'usb') {
1124 die "implement me";
1125 # -device ide-drive,bus=ide.1,unit=0,drive=drive-ide0-1-0,id=ide0-1-0
1126 } else {
1127 die "unsupported interface type";
1128 }
1129
1130 $device .= ",bootindex=$drive->{bootindex}" if $drive->{bootindex};
1131
1132 return $device;
1133 }
1134
1135 sub get_initiator_name {
1136 my $initiator;
1137
1138 my $fh = IO::File->new('/etc/iscsi/initiatorname.iscsi') || return undef;
1139 while (defined(my $line = <$fh>)) {
1140 next if $line !~ m/^\s*InitiatorName\s*=\s*([\.\-:\w]+)/;
1141 $initiator = $1;
1142 last;
1143 }
1144 $fh->close();
1145
1146 return $initiator;
1147 }
1148
1149 sub print_drive_full {
1150 my ($storecfg, $vmid, $drive) = @_;
1151
1152 my $path;
1153 my $volid = $drive->{file};
1154 my $format;
1155
1156 if (drive_is_cdrom($drive)) {
1157 $path = get_iso_path($storecfg, $vmid, $volid);
1158 } else {
1159 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1160 if ($storeid) {
1161 $path = PVE::Storage::path($storecfg, $volid);
1162 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
1163 $format = qemu_img_format($scfg, $volname);
1164 } else {
1165 $path = $volid;
1166 }
1167 }
1168
1169 my $opts = '';
1170 foreach my $o (@qemu_drive_options) {
1171 next if $o eq 'bootindex';
1172 $opts .= ",$o=$drive->{$o}" if $drive->{$o};
1173 }
1174
1175 $opts .= ",format=$format" if $format && !$drive->{format};
1176
1177 foreach my $o (qw(bps bps_rd bps_wr)) {
1178 my $v = $drive->{"m$o"};
1179 $opts .= ",$o=" . int($v*1024*1024) if $v;
1180 }
1181
1182 my $cache_direct = 0;
1183
1184 if (my $cache = $drive->{cache}) {
1185 $cache_direct = $cache =~ /^(?:off|none|directsync)$/;
1186 } elsif (!drive_is_cdrom($drive)) {
1187 $opts .= ",cache=none";
1188 $cache_direct = 1;
1189 }
1190
1191 # aio native works only with O_DIRECT
1192 if (!$drive->{aio}) {
1193 if($cache_direct) {
1194 $opts .= ",aio=native";
1195 } else {
1196 $opts .= ",aio=threads";
1197 }
1198 }
1199
1200 my $detectzeroes = $drive->{discard} ? "unmap" : "on";
1201 $opts .= ",detect-zeroes=$detectzeroes" if !drive_is_cdrom($drive);
1202
1203 my $pathinfo = $path ? "file=$path," : '';
1204
1205 return "${pathinfo}if=none,id=drive-$drive->{interface}$drive->{index}$opts";
1206 }
1207
1208 sub print_netdevice_full {
1209 my ($vmid, $conf, $net, $netid, $bridges) = @_;
1210
1211 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
1212
1213 my $device = $net->{model};
1214 if ($net->{model} eq 'virtio') {
1215 $device = 'virtio-net-pci';
1216 };
1217
1218 my $pciaddr = print_pci_addr("$netid", $bridges);
1219 my $tmpstr = "$device,mac=$net->{macaddr},netdev=$netid$pciaddr,id=$netid";
1220 if ($net->{queues} && $net->{queues} > 1 && $net->{model} eq 'virtio'){
1221 #Consider we have N queues, the number of vectors needed is 2*N + 2 (plus one config interrupt and control vq)
1222 my $vectors = $net->{queues} * 2 + 2;
1223 $tmpstr .= ",vectors=$vectors,mq=on";
1224 }
1225 $tmpstr .= ",bootindex=$net->{bootindex}" if $net->{bootindex} ;
1226 return $tmpstr;
1227 }
1228
1229 sub print_netdev_full {
1230 my ($vmid, $conf, $net, $netid) = @_;
1231
1232 my $i = '';
1233 if ($netid =~ m/^net(\d+)$/) {
1234 $i = int($1);
1235 }
1236
1237 die "got strange net id '$i'\n" if $i >= ${MAX_NETS};
1238
1239 my $ifname = "tap${vmid}i$i";
1240
1241 # kvm uses TUNSETIFF ioctl, and that limits ifname length
1242 die "interface name '$ifname' is too long (max 15 character)\n"
1243 if length($ifname) >= 16;
1244
1245 my $vhostparam = '';
1246 $vhostparam = ',vhost=on' if $kernel_has_vhost_net && $net->{model} eq 'virtio';
1247
1248 my $vmname = $conf->{name} || "vm$vmid";
1249
1250 my $netdev = "";
1251
1252 if ($net->{bridge}) {
1253 $netdev = "type=tap,id=$netid,ifname=${ifname},script=/var/lib/qemu-server/pve-bridge,downscript=/var/lib/qemu-server/pve-bridgedown$vhostparam";
1254 } else {
1255 $netdev = "type=user,id=$netid,hostname=$vmname";
1256 }
1257
1258 $netdev .= ",queues=$net->{queues}" if ($net->{queues} && $net->{model} eq 'virtio');
1259
1260 return $netdev;
1261 }
1262
1263 sub drive_is_cdrom {
1264 my ($drive) = @_;
1265
1266 return $drive && $drive->{media} && ($drive->{media} eq 'cdrom');
1267
1268 }
1269
1270 sub parse_numa {
1271 my ($data) = @_;
1272
1273 my $res = {};
1274
1275 foreach my $kvp (split(/,/, $data)) {
1276
1277 if ($kvp =~ m/^memory=(\S+)$/) {
1278 $res->{memory} = $1;
1279 } elsif ($kvp =~ m/^policy=(preferred|bind|interleave)$/) {
1280 $res->{policy} = $1;
1281 } elsif ($kvp =~ m/^cpus=(\d+)(-(\d+))?$/) {
1282 $res->{cpus}->{start} = $1;
1283 $res->{cpus}->{end} = $3;
1284 } elsif ($kvp =~ m/^hostnodes=(\d+)(-(\d+))?$/) {
1285 $res->{hostnodes}->{start} = $1;
1286 $res->{hostnodes}->{end} = $3;
1287 } else {
1288 return undef;
1289 }
1290 }
1291
1292 return $res;
1293 }
1294
1295 sub parse_hostpci {
1296 my ($value) = @_;
1297
1298 return undef if !$value;
1299
1300
1301 my @list = split(/,/, $value);
1302 my $found;
1303
1304 my $res = {};
1305 foreach my $kv (@list) {
1306
1307 if ($kv =~ m/^(host=)?([a-f0-9]{2}:[a-f0-9]{2})(\.([a-f0-9]))?$/) {
1308 $found = 1;
1309 if(defined($4)){
1310 push @{$res->{pciid}}, { id => $2 , function => $4};
1311
1312 }else{
1313 my $pcidevices = lspci($2);
1314 $res->{pciid} = $pcidevices->{$2};
1315 }
1316 } elsif ($kv =~ m/^driver=(kvm|vfio)$/) {
1317 $res->{driver} = $1;
1318 } elsif ($kv =~ m/^rombar=(on|off)$/) {
1319 $res->{rombar} = $1;
1320 } elsif ($kv =~ m/^x-vga=(on|off)$/) {
1321 $res->{'x-vga'} = $1;
1322 } elsif ($kv =~ m/^pcie=(\d+)$/) {
1323 $res->{pcie} = 1 if $1 == 1;
1324 } else {
1325 warn "unknown hostpci setting '$kv'\n";
1326 }
1327 }
1328
1329 return undef if !$found;
1330
1331 return $res;
1332 }
1333
1334 # netX: e1000=XX:XX:XX:XX:XX:XX,bridge=vmbr0,rate=<mbps>
1335 sub parse_net {
1336 my ($data) = @_;
1337
1338 my $res = {};
1339
1340 foreach my $kvp (split(/,/, $data)) {
1341
1342 if ($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) {
1343 my $model = lc($1);
1344 my $mac = defined($3) ? uc($3) : PVE::Tools::random_ether_addr();
1345 $res->{model} = $model;
1346 $res->{macaddr} = $mac;
1347 } elsif ($kvp =~ m/^bridge=(\S+)$/) {
1348 $res->{bridge} = $1;
1349 } elsif ($kvp =~ m/^queues=(\d+)$/) {
1350 $res->{queues} = $1;
1351 } elsif ($kvp =~ m/^rate=(\d+(\.\d+)?)$/) {
1352 $res->{rate} = $1;
1353 } elsif ($kvp =~ m/^tag=(\d+)$/) {
1354 $res->{tag} = $1;
1355 } elsif ($kvp =~ m/^firewall=([01])$/) {
1356 $res->{firewall} = $1;
1357 } elsif ($kvp =~ m/^link_down=([01])$/) {
1358 $res->{link_down} = $1;
1359 } else {
1360 return undef;
1361 }
1362
1363 }
1364
1365 return undef if !$res->{model};
1366
1367 return $res;
1368 }
1369
1370 sub print_net {
1371 my $net = shift;
1372
1373 my $res = "$net->{model}";
1374 $res .= "=$net->{macaddr}" if $net->{macaddr};
1375 $res .= ",bridge=$net->{bridge}" if $net->{bridge};
1376 $res .= ",rate=$net->{rate}" if $net->{rate};
1377 $res .= ",tag=$net->{tag}" if $net->{tag};
1378 $res .= ",firewall=1" if $net->{firewall};
1379 $res .= ",link_down=1" if $net->{link_down};
1380 $res .= ",queues=$net->{queues}" if $net->{queues};
1381
1382 return $res;
1383 }
1384
1385 sub add_random_macs {
1386 my ($settings) = @_;
1387
1388 foreach my $opt (keys %$settings) {
1389 next if $opt !~ m/^net(\d+)$/;
1390 my $net = parse_net($settings->{$opt});
1391 next if !$net;
1392 $settings->{$opt} = print_net($net);
1393 }
1394 }
1395
1396 sub add_unused_volume {
1397 my ($config, $volid) = @_;
1398
1399 my $key;
1400 for (my $ind = $MAX_UNUSED_DISKS - 1; $ind >= 0; $ind--) {
1401 my $test = "unused$ind";
1402 if (my $vid = $config->{$test}) {
1403 return if $vid eq $volid; # do not add duplicates
1404 } else {
1405 $key = $test;
1406 }
1407 }
1408
1409 die "To many unused volume - please delete them first.\n" if !$key;
1410
1411 $config->{$key} = $volid;
1412
1413 return $key;
1414 }
1415
1416 sub vm_is_volid_owner {
1417 my ($storecfg, $vmid, $volid) = @_;
1418
1419 if ($volid !~ m|^/|) {
1420 my ($path, $owner);
1421 eval { ($path, $owner) = PVE::Storage::path($storecfg, $volid); };
1422 if ($owner && ($owner == $vmid)) {
1423 return 1;
1424 }
1425 }
1426
1427 return undef;
1428 }
1429
1430 sub vmconfig_delete_pending_option {
1431 my ($conf, $key) = @_;
1432
1433 delete $conf->{pending}->{$key};
1434 my $pending_delete_hash = { $key => 1 };
1435 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1436 $pending_delete_hash->{$opt} = 1;
1437 }
1438 $conf->{pending}->{delete} = join(',', keys %$pending_delete_hash);
1439 }
1440
1441 sub vmconfig_undelete_pending_option {
1442 my ($conf, $key) = @_;
1443
1444 my $pending_delete_hash = {};
1445 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1446 $pending_delete_hash->{$opt} = 1;
1447 }
1448 delete $pending_delete_hash->{$key};
1449
1450 my @keylist = keys %$pending_delete_hash;
1451 if (scalar(@keylist)) {
1452 $conf->{pending}->{delete} = join(',', @keylist);
1453 } else {
1454 delete $conf->{pending}->{delete};
1455 }
1456 }
1457
1458 sub vmconfig_register_unused_drive {
1459 my ($storecfg, $vmid, $conf, $drive) = @_;
1460
1461 if (!drive_is_cdrom($drive)) {
1462 my $volid = $drive->{file};
1463 if (vm_is_volid_owner($storecfg, $vmid, $volid)) {
1464 add_unused_volume($conf, $volid, $vmid);
1465 }
1466 }
1467 }
1468
1469 sub vmconfig_cleanup_pending {
1470 my ($conf) = @_;
1471
1472 # remove pending changes when nothing changed
1473 my $changes;
1474 foreach my $opt (keys %{$conf->{pending}}) {
1475 if (defined($conf->{$opt}) && ($conf->{pending}->{$opt} eq $conf->{$opt})) {
1476 $changes = 1;
1477 delete $conf->{pending}->{$opt};
1478 }
1479 }
1480
1481 # remove delete if option is not set
1482 my $pending_delete_hash = {};
1483 foreach my $opt (PVE::Tools::split_list($conf->{pending}->{delete})) {
1484 if (defined($conf->{$opt})) {
1485 $pending_delete_hash->{$opt} = 1;
1486 } else {
1487 $changes = 1;
1488 }
1489 }
1490
1491 my @keylist = keys %$pending_delete_hash;
1492 if (scalar(@keylist)) {
1493 $conf->{pending}->{delete} = join(',', @keylist);
1494 } else {
1495 delete $conf->{pending}->{delete};
1496 }
1497
1498 return $changes;
1499 }
1500
1501 my $valid_smbios1_options = {
1502 manufacturer => '\S+',
1503 product => '\S+',
1504 version => '\S+',
1505 serial => '\S+',
1506 uuid => '[a-fA-F0-9]{8}(?:-[a-fA-F0-9]{4}){3}-[a-fA-F0-9]{12}',
1507 sku => '\S+',
1508 family => '\S+',
1509 };
1510
1511 # smbios: [manufacturer=str][,product=str][,version=str][,serial=str][,uuid=uuid][,sku=str][,family=str]
1512 sub parse_smbios1 {
1513 my ($data) = @_;
1514
1515 my $res = {};
1516
1517 foreach my $kvp (split(/,/, $data)) {
1518 return undef if $kvp !~ m/^(\S+)=(.+)$/;
1519 my ($k, $v) = split(/=/, $kvp);
1520 return undef if !defined($k) || !defined($v);
1521 return undef if !$valid_smbios1_options->{$k};
1522 return undef if $v !~ m/^$valid_smbios1_options->{$k}$/;
1523 $res->{$k} = $v;
1524 }
1525
1526 return $res;
1527 }
1528
1529 sub print_smbios1 {
1530 my ($smbios1) = @_;
1531
1532 my $data = '';
1533 foreach my $k (keys %$smbios1) {
1534 next if !defined($smbios1->{$k});
1535 next if !$valid_smbios1_options->{$k};
1536 $data .= ',' if $data;
1537 $data .= "$k=$smbios1->{$k}";
1538 }
1539 return $data;
1540 }
1541
1542 PVE::JSONSchema::register_format('pve-qm-smbios1', \&verify_smbios1);
1543 sub verify_smbios1 {
1544 my ($value, $noerr) = @_;
1545
1546 return $value if parse_smbios1($value);
1547
1548 return undef if $noerr;
1549
1550 die "unable to parse smbios (type 1) options\n";
1551 }
1552
1553 PVE::JSONSchema::register_format('pve-qm-bootdisk', \&verify_bootdisk);
1554 sub verify_bootdisk {
1555 my ($value, $noerr) = @_;
1556
1557 return $value if valid_drivename($value);
1558
1559 return undef if $noerr;
1560
1561 die "invalid boot disk '$value'\n";
1562 }
1563
1564 PVE::JSONSchema::register_format('pve-qm-numanode', \&verify_numa);
1565 sub verify_numa {
1566 my ($value, $noerr) = @_;
1567
1568 return $value if parse_numa($value);
1569
1570 return undef if $noerr;
1571
1572 die "unable to parse numa options\n";
1573 }
1574
1575 PVE::JSONSchema::register_format('pve-qm-net', \&verify_net);
1576 sub verify_net {
1577 my ($value, $noerr) = @_;
1578
1579 return $value if parse_net($value);
1580
1581 return undef if $noerr;
1582
1583 die "unable to parse network options\n";
1584 }
1585
1586 PVE::JSONSchema::register_format('pve-qm-drive', \&verify_drive);
1587 sub verify_drive {
1588 my ($value, $noerr) = @_;
1589
1590 return $value if parse_drive(undef, $value);
1591
1592 return undef if $noerr;
1593
1594 die "unable to parse drive options\n";
1595 }
1596
1597 PVE::JSONSchema::register_format('pve-qm-hostpci', \&verify_hostpci);
1598 sub verify_hostpci {
1599 my ($value, $noerr) = @_;
1600
1601 return $value if parse_hostpci($value);
1602
1603 return undef if $noerr;
1604
1605 die "unable to parse pci id\n";
1606 }
1607
1608 PVE::JSONSchema::register_format('pve-qm-watchdog', \&verify_watchdog);
1609 sub verify_watchdog {
1610 my ($value, $noerr) = @_;
1611
1612 return $value if parse_watchdog($value);
1613
1614 return undef if $noerr;
1615
1616 die "unable to parse watchdog options\n";
1617 }
1618
1619 sub parse_watchdog {
1620 my ($value) = @_;
1621
1622 return undef if !$value;
1623
1624 my $res = {};
1625
1626 foreach my $p (split(/,/, $value)) {
1627 next if $p =~ m/^\s*$/;
1628
1629 if ($p =~ m/^(model=)?(i6300esb|ib700)$/) {
1630 $res->{model} = $2;
1631 } elsif ($p =~ m/^(action=)?(reset|shutdown|poweroff|pause|debug|none)$/) {
1632 $res->{action} = $2;
1633 } else {
1634 return undef;
1635 }
1636 }
1637
1638 return $res;
1639 }
1640
1641 sub parse_usb_device {
1642 my ($value) = @_;
1643
1644 return undef if !$value;
1645
1646 my @dl = split(/,/, $value);
1647 my $found;
1648
1649 my $res = {};
1650 foreach my $v (@dl) {
1651 if ($v =~ m/^host=(0x)?([0-9A-Fa-f]{4}):(0x)?([0-9A-Fa-f]{4})$/) {
1652 $found = 1;
1653 $res->{vendorid} = $2;
1654 $res->{productid} = $4;
1655 } elsif ($v =~ m/^host=(\d+)\-(\d+(\.\d+)*)$/) {
1656 $found = 1;
1657 $res->{hostbus} = $1;
1658 $res->{hostport} = $2;
1659 } elsif ($v =~ m/^spice$/) {
1660 $found = 1;
1661 $res->{spice} = 1;
1662 } else {
1663 return undef;
1664 }
1665 }
1666 return undef if !$found;
1667
1668 return $res;
1669 }
1670
1671 PVE::JSONSchema::register_format('pve-qm-usb-device', \&verify_usb_device);
1672 sub verify_usb_device {
1673 my ($value, $noerr) = @_;
1674
1675 return $value if parse_usb_device($value);
1676
1677 return undef if $noerr;
1678
1679 die "unable to parse usb device\n";
1680 }
1681
1682 # add JSON properties for create and set function
1683 sub json_config_properties {
1684 my $prop = shift;
1685
1686 foreach my $opt (keys %$confdesc) {
1687 next if $opt eq 'parent' || $opt eq 'snaptime' || $opt eq 'vmstate';
1688 $prop->{$opt} = $confdesc->{$opt};
1689 }
1690
1691 return $prop;
1692 }
1693
1694 sub check_type {
1695 my ($key, $value) = @_;
1696
1697 die "unknown setting '$key'\n" if !$confdesc->{$key};
1698
1699 my $type = $confdesc->{$key}->{type};
1700
1701 if (!defined($value)) {
1702 die "got undefined value\n";
1703 }
1704
1705 if ($value =~ m/[\n\r]/) {
1706 die "property contains a line feed\n";
1707 }
1708
1709 if ($type eq 'boolean') {
1710 return 1 if ($value eq '1') || ($value =~ m/^(on|yes|true)$/i);
1711 return 0 if ($value eq '0') || ($value =~ m/^(off|no|false)$/i);
1712 die "type check ('boolean') failed - got '$value'\n";
1713 } elsif ($type eq 'integer') {
1714 return int($1) if $value =~ m/^(\d+)$/;
1715 die "type check ('integer') failed - got '$value'\n";
1716 } elsif ($type eq 'number') {
1717 return $value if $value =~ m/^(\d+)(\.\d+)?$/;
1718 die "type check ('number') failed - got '$value'\n";
1719 } elsif ($type eq 'string') {
1720 if (my $fmt = $confdesc->{$key}->{format}) {
1721 if ($fmt eq 'pve-qm-drive') {
1722 # special case - we need to pass $key to parse_drive()
1723 my $drive = parse_drive($key, $value);
1724 return $value if $drive;
1725 die "unable to parse drive options\n";
1726 }
1727 PVE::JSONSchema::check_format($fmt, $value);
1728 return $value;
1729 }
1730 $value =~ s/^\"(.*)\"$/$1/;
1731 return $value;
1732 } else {
1733 die "internal error"
1734 }
1735 }
1736
1737 sub lock_config_full {
1738 my ($vmid, $timeout, $code, @param) = @_;
1739
1740 my $filename = config_file_lock($vmid);
1741
1742 my $res = lock_file($filename, $timeout, $code, @param);
1743
1744 die $@ if $@;
1745
1746 return $res;
1747 }
1748
1749 sub lock_config_mode {
1750 my ($vmid, $timeout, $shared, $code, @param) = @_;
1751
1752 my $filename = config_file_lock($vmid);
1753
1754 my $res = lock_file_full($filename, $timeout, $shared, $code, @param);
1755
1756 die $@ if $@;
1757
1758 return $res;
1759 }
1760
1761 sub lock_config {
1762 my ($vmid, $code, @param) = @_;
1763
1764 return lock_config_full($vmid, 10, $code, @param);
1765 }
1766
1767 sub cfs_config_path {
1768 my ($vmid, $node) = @_;
1769
1770 $node = $nodename if !$node;
1771 return "nodes/$node/qemu-server/$vmid.conf";
1772 }
1773
1774 sub check_iommu_support{
1775 #fixme : need to check IOMMU support
1776 #http://www.linux-kvm.org/page/How_to_assign_devices_with_VT-d_in_KVM
1777
1778 my $iommu=1;
1779 return $iommu;
1780
1781 }
1782
1783 sub config_file {
1784 my ($vmid, $node) = @_;
1785
1786 my $cfspath = cfs_config_path($vmid, $node);
1787 return "/etc/pve/$cfspath";
1788 }
1789
1790 sub config_file_lock {
1791 my ($vmid) = @_;
1792
1793 return "$lock_dir/lock-$vmid.conf";
1794 }
1795
1796 sub touch_config {
1797 my ($vmid) = @_;
1798
1799 my $conf = config_file($vmid);
1800 utime undef, undef, $conf;
1801 }
1802
1803 sub destroy_vm {
1804 my ($storecfg, $vmid, $keep_empty_config) = @_;
1805
1806 my $conffile = config_file($vmid);
1807
1808 my $conf = load_config($vmid);
1809
1810 check_lock($conf);
1811
1812 # only remove disks owned by this VM
1813 foreach_drive($conf, sub {
1814 my ($ds, $drive) = @_;
1815
1816 return if drive_is_cdrom($drive);
1817
1818 my $volid = $drive->{file};
1819
1820 return if !$volid || $volid =~ m|^/|;
1821
1822 my ($path, $owner) = PVE::Storage::path($storecfg, $volid);
1823 return if !$path || !$owner || ($owner != $vmid);
1824
1825 PVE::Storage::vdisk_free($storecfg, $volid);
1826 });
1827
1828 if ($keep_empty_config) {
1829 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
1830 } else {
1831 unlink $conffile;
1832 }
1833
1834 # also remove unused disk
1835 eval {
1836 my $dl = PVE::Storage::vdisk_list($storecfg, undef, $vmid);
1837
1838 eval {
1839 PVE::Storage::foreach_volid($dl, sub {
1840 my ($volid, $sid, $volname, $d) = @_;
1841 PVE::Storage::vdisk_free($storecfg, $volid);
1842 });
1843 };
1844 warn $@ if $@;
1845
1846 };
1847 warn $@ if $@;
1848 }
1849
1850 sub load_config {
1851 my ($vmid, $node) = @_;
1852
1853 my $cfspath = cfs_config_path($vmid, $node);
1854
1855 my $conf = PVE::Cluster::cfs_read_file($cfspath);
1856
1857 die "no such VM ('$vmid')\n" if !defined($conf);
1858
1859 return $conf;
1860 }
1861
1862 sub parse_vm_config {
1863 my ($filename, $raw) = @_;
1864
1865 return undef if !defined($raw);
1866
1867 my $res = {
1868 digest => Digest::SHA::sha1_hex($raw),
1869 snapshots => {},
1870 pending => {},
1871 };
1872
1873 $filename =~ m|/qemu-server/(\d+)\.conf$|
1874 || die "got strange filename '$filename'";
1875
1876 my $vmid = $1;
1877
1878 my $conf = $res;
1879 my $descr = '';
1880 my $section = '';
1881
1882 my @lines = split(/\n/, $raw);
1883 foreach my $line (@lines) {
1884 next if $line =~ m/^\s*$/;
1885
1886 if ($line =~ m/^\[PENDING\]\s*$/i) {
1887 $section = 'pending';
1888 $conf->{description} = $descr if $descr;
1889 $descr = '';
1890 $conf = $res->{$section} = {};
1891 next;
1892
1893 } elsif ($line =~ m/^\[([a-z][a-z0-9_\-]+)\]\s*$/i) {
1894 $section = $1;
1895 $conf->{description} = $descr if $descr;
1896 $descr = '';
1897 $conf = $res->{snapshots}->{$section} = {};
1898 next;
1899 }
1900
1901 if ($line =~ m/^\#(.*)\s*$/) {
1902 $descr .= PVE::Tools::decode_text($1) . "\n";
1903 next;
1904 }
1905
1906 if ($line =~ m/^(description):\s*(.*\S)\s*$/) {
1907 $descr .= PVE::Tools::decode_text($2);
1908 } elsif ($line =~ m/snapstate:\s*(prepare|delete)\s*$/) {
1909 $conf->{snapstate} = $1;
1910 } elsif ($line =~ m/^(args):\s*(.*\S)\s*$/) {
1911 my $key = $1;
1912 my $value = $2;
1913 $conf->{$key} = $value;
1914 } elsif ($line =~ m/^delete:\s*(.*\S)\s*$/) {
1915 my $value = $1;
1916 if ($section eq 'pending') {
1917 $conf->{delete} = $value; # we parse this later
1918 } else {
1919 warn "vm $vmid - propertry 'delete' is only allowed in [PENDING]\n";
1920 }
1921 } elsif ($line =~ m/^([a-z][a-z_]*\d*):\s*(\S+)\s*$/) {
1922 my $key = $1;
1923 my $value = $2;
1924 eval { $value = check_type($key, $value); };
1925 if ($@) {
1926 warn "vm $vmid - unable to parse value of '$key' - $@";
1927 } else {
1928 my $fmt = $confdesc->{$key}->{format};
1929 if ($fmt && $fmt eq 'pve-qm-drive') {
1930 my $v = parse_drive($key, $value);
1931 if (my $volid = filename_to_volume_id($vmid, $v->{file}, $v->{media})) {
1932 $v->{file} = $volid;
1933 $value = print_drive($vmid, $v);
1934 } else {
1935 warn "vm $vmid - unable to parse value of '$key'\n";
1936 next;
1937 }
1938 }
1939
1940 if ($key eq 'cdrom') {
1941 $conf->{ide2} = $value;
1942 } else {
1943 $conf->{$key} = $value;
1944 }
1945 }
1946 }
1947 }
1948
1949 $conf->{description} = $descr if $descr;
1950
1951 delete $res->{snapstate}; # just to be sure
1952
1953 return $res;
1954 }
1955
1956 sub write_vm_config {
1957 my ($filename, $conf) = @_;
1958
1959 delete $conf->{snapstate}; # just to be sure
1960
1961 if ($conf->{cdrom}) {
1962 die "option ide2 conflicts with cdrom\n" if $conf->{ide2};
1963 $conf->{ide2} = $conf->{cdrom};
1964 delete $conf->{cdrom};
1965 }
1966
1967 # we do not use 'smp' any longer
1968 if ($conf->{sockets}) {
1969 delete $conf->{smp};
1970 } elsif ($conf->{smp}) {
1971 $conf->{sockets} = $conf->{smp};
1972 delete $conf->{cores};
1973 delete $conf->{smp};
1974 }
1975
1976 my $used_volids = {};
1977
1978 my $cleanup_config = sub {
1979 my ($cref, $pending, $snapname) = @_;
1980
1981 foreach my $key (keys %$cref) {
1982 next if $key eq 'digest' || $key eq 'description' || $key eq 'snapshots' ||
1983 $key eq 'snapstate' || $key eq 'pending';
1984 my $value = $cref->{$key};
1985 if ($key eq 'delete') {
1986 die "propertry 'delete' is only allowed in [PENDING]\n"
1987 if !$pending;
1988 # fixme: check syntax?
1989 next;
1990 }
1991 eval { $value = check_type($key, $value); };
1992 die "unable to parse value of '$key' - $@" if $@;
1993
1994 $cref->{$key} = $value;
1995
1996 if (!$snapname && valid_drivename($key)) {
1997 my $drive = parse_drive($key, $value);
1998 $used_volids->{$drive->{file}} = 1 if $drive && $drive->{file};
1999 }
2000 }
2001 };
2002
2003 &$cleanup_config($conf);
2004
2005 &$cleanup_config($conf->{pending}, 1);
2006
2007 foreach my $snapname (keys %{$conf->{snapshots}}) {
2008 die "internal error" if $snapname eq 'pending';
2009 &$cleanup_config($conf->{snapshots}->{$snapname}, undef, $snapname);
2010 }
2011
2012 # remove 'unusedX' settings if we re-add a volume
2013 foreach my $key (keys %$conf) {
2014 my $value = $conf->{$key};
2015 if ($key =~ m/^unused/ && $used_volids->{$value}) {
2016 delete $conf->{$key};
2017 }
2018 }
2019
2020 my $generate_raw_config = sub {
2021 my ($conf) = @_;
2022
2023 my $raw = '';
2024
2025 # add description as comment to top of file
2026 my $descr = $conf->{description} || '';
2027 foreach my $cl (split(/\n/, $descr)) {
2028 $raw .= '#' . PVE::Tools::encode_text($cl) . "\n";
2029 }
2030
2031 foreach my $key (sort keys %$conf) {
2032 next if $key eq 'digest' || $key eq 'description' || $key eq 'pending' || $key eq 'snapshots';
2033 $raw .= "$key: $conf->{$key}\n";
2034 }
2035 return $raw;
2036 };
2037
2038 my $raw = &$generate_raw_config($conf);
2039
2040 if (scalar(keys %{$conf->{pending}})){
2041 $raw .= "\n[PENDING]\n";
2042 $raw .= &$generate_raw_config($conf->{pending});
2043 }
2044
2045 foreach my $snapname (sort keys %{$conf->{snapshots}}) {
2046 $raw .= "\n[$snapname]\n";
2047 $raw .= &$generate_raw_config($conf->{snapshots}->{$snapname});
2048 }
2049
2050 return $raw;
2051 }
2052
2053 sub update_config_nolock {
2054 my ($vmid, $conf, $skiplock) = @_;
2055
2056 check_lock($conf) if !$skiplock;
2057
2058 my $cfspath = cfs_config_path($vmid);
2059
2060 PVE::Cluster::cfs_write_file($cfspath, $conf);
2061 }
2062
2063 sub update_config {
2064 my ($vmid, $conf, $skiplock) = @_;
2065
2066 lock_config($vmid, &update_config_nolock, $conf, $skiplock);
2067 }
2068
2069 sub load_defaults {
2070
2071 my $res = {};
2072
2073 # we use static defaults from our JSON schema configuration
2074 foreach my $key (keys %$confdesc) {
2075 if (defined(my $default = $confdesc->{$key}->{default})) {
2076 $res->{$key} = $default;
2077 }
2078 }
2079
2080 my $conf = PVE::Cluster::cfs_read_file('datacenter.cfg');
2081 $res->{keyboard} = $conf->{keyboard} if $conf->{keyboard};
2082
2083 return $res;
2084 }
2085
2086 sub config_list {
2087 my $vmlist = PVE::Cluster::get_vmlist();
2088 my $res = {};
2089 return $res if !$vmlist || !$vmlist->{ids};
2090 my $ids = $vmlist->{ids};
2091
2092 foreach my $vmid (keys %$ids) {
2093 my $d = $ids->{$vmid};
2094 next if !$d->{node} || $d->{node} ne $nodename;
2095 next if !$d->{type} || $d->{type} ne 'qemu';
2096 $res->{$vmid}->{exists} = 1;
2097 }
2098 return $res;
2099 }
2100
2101 # test if VM uses local resources (to prevent migration)
2102 sub check_local_resources {
2103 my ($conf, $noerr) = @_;
2104
2105 my $loc_res = 0;
2106
2107 $loc_res = 1 if $conf->{hostusb}; # old syntax
2108 $loc_res = 1 if $conf->{hostpci}; # old syntax
2109
2110 foreach my $k (keys %$conf) {
2111 next if $k =~ m/^usb/ && ($conf->{$k} eq 'spice');
2112 $loc_res = 1 if $k =~ m/^(usb|hostpci|serial|parallel)\d+$/;
2113 }
2114
2115 die "VM uses local resources\n" if $loc_res && !$noerr;
2116
2117 return $loc_res;
2118 }
2119
2120 # check if used storages are available on all nodes (use by migrate)
2121 sub check_storage_availability {
2122 my ($storecfg, $conf, $node) = @_;
2123
2124 foreach_drive($conf, sub {
2125 my ($ds, $drive) = @_;
2126
2127 my $volid = $drive->{file};
2128 return if !$volid;
2129
2130 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2131 return if !$sid;
2132
2133 # check if storage is available on both nodes
2134 my $scfg = PVE::Storage::storage_check_node($storecfg, $sid);
2135 PVE::Storage::storage_check_node($storecfg, $sid, $node);
2136 });
2137 }
2138
2139 # list nodes where all VM images are available (used by has_feature API)
2140 sub shared_nodes {
2141 my ($conf, $storecfg) = @_;
2142
2143 my $nodelist = PVE::Cluster::get_nodelist();
2144 my $nodehash = { map { $_ => 1 } @$nodelist };
2145 my $nodename = PVE::INotify::nodename();
2146
2147 foreach_drive($conf, sub {
2148 my ($ds, $drive) = @_;
2149
2150 my $volid = $drive->{file};
2151 return if !$volid;
2152
2153 my ($storeid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
2154 if ($storeid) {
2155 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
2156 if ($scfg->{disable}) {
2157 $nodehash = {};
2158 } elsif (my $avail = $scfg->{nodes}) {
2159 foreach my $node (keys %$nodehash) {
2160 delete $nodehash->{$node} if !$avail->{$node};
2161 }
2162 } elsif (!$scfg->{shared}) {
2163 foreach my $node (keys %$nodehash) {
2164 delete $nodehash->{$node} if $node ne $nodename
2165 }
2166 }
2167 }
2168 });
2169
2170 return $nodehash
2171 }
2172
2173 sub check_lock {
2174 my ($conf) = @_;
2175
2176 die "VM is locked ($conf->{lock})\n" if $conf->{lock};
2177 }
2178
2179 sub check_cmdline {
2180 my ($pidfile, $pid) = @_;
2181
2182 my $fh = IO::File->new("/proc/$pid/cmdline", "r");
2183 if (defined($fh)) {
2184 my $line = <$fh>;
2185 $fh->close;
2186 return undef if !$line;
2187 my @param = split(/\0/, $line);
2188
2189 my $cmd = $param[0];
2190 return if !$cmd || ($cmd !~ m|kvm$| && $cmd !~ m|qemu-system-x86_64$|);
2191
2192 for (my $i = 0; $i < scalar (@param); $i++) {
2193 my $p = $param[$i];
2194 next if !$p;
2195 if (($p eq '-pidfile') || ($p eq '--pidfile')) {
2196 my $p = $param[$i+1];
2197 return 1 if $p && ($p eq $pidfile);
2198 return undef;
2199 }
2200 }
2201 }
2202 return undef;
2203 }
2204
2205 sub check_running {
2206 my ($vmid, $nocheck, $node) = @_;
2207
2208 my $filename = config_file($vmid, $node);
2209
2210 die "unable to find configuration file for VM $vmid - no such machine\n"
2211 if !$nocheck && ! -f $filename;
2212
2213 my $pidfile = pidfile_name($vmid);
2214
2215 if (my $fd = IO::File->new("<$pidfile")) {
2216 my $st = stat($fd);
2217 my $line = <$fd>;
2218 close($fd);
2219
2220 my $mtime = $st->mtime;
2221 if ($mtime > time()) {
2222 warn "file '$filename' modified in future\n";
2223 }
2224
2225 if ($line =~ m/^(\d+)$/) {
2226 my $pid = $1;
2227 if (check_cmdline($pidfile, $pid)) {
2228 if (my $pinfo = PVE::ProcFSTools::check_process_running($pid)) {
2229 return $pid;
2230 }
2231 }
2232 }
2233 }
2234
2235 return undef;
2236 }
2237
2238 sub vzlist {
2239
2240 my $vzlist = config_list();
2241
2242 my $fd = IO::Dir->new($var_run_tmpdir) || return $vzlist;
2243
2244 while (defined(my $de = $fd->read)) {
2245 next if $de !~ m/^(\d+)\.pid$/;
2246 my $vmid = $1;
2247 next if !defined($vzlist->{$vmid});
2248 if (my $pid = check_running($vmid)) {
2249 $vzlist->{$vmid}->{pid} = $pid;
2250 }
2251 }
2252
2253 return $vzlist;
2254 }
2255
2256 sub disksize {
2257 my ($storecfg, $conf) = @_;
2258
2259 my $bootdisk = $conf->{bootdisk};
2260 return undef if !$bootdisk;
2261 return undef if !valid_drivename($bootdisk);
2262
2263 return undef if !$conf->{$bootdisk};
2264
2265 my $drive = parse_drive($bootdisk, $conf->{$bootdisk});
2266 return undef if !defined($drive);
2267
2268 return undef if drive_is_cdrom($drive);
2269
2270 my $volid = $drive->{file};
2271 return undef if !$volid;
2272
2273 return $drive->{size};
2274 }
2275
2276 my $last_proc_pid_stat;
2277
2278 # get VM status information
2279 # This must be fast and should not block ($full == false)
2280 # We only query KVM using QMP if $full == true (this can be slow)
2281 sub vmstatus {
2282 my ($opt_vmid, $full) = @_;
2283
2284 my $res = {};
2285
2286 my $storecfg = PVE::Storage::config();
2287
2288 my $list = vzlist();
2289 my ($uptime) = PVE::ProcFSTools::read_proc_uptime(1);
2290
2291 my $cpucount = $cpuinfo->{cpus} || 1;
2292
2293 foreach my $vmid (keys %$list) {
2294 next if $opt_vmid && ($vmid ne $opt_vmid);
2295
2296 my $cfspath = cfs_config_path($vmid);
2297 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
2298
2299 my $d = {};
2300 $d->{pid} = $list->{$vmid}->{pid};
2301
2302 # fixme: better status?
2303 $d->{status} = $list->{$vmid}->{pid} ? 'running' : 'stopped';
2304
2305 my $size = disksize($storecfg, $conf);
2306 if (defined($size)) {
2307 $d->{disk} = 0; # no info available
2308 $d->{maxdisk} = $size;
2309 } else {
2310 $d->{disk} = 0;
2311 $d->{maxdisk} = 0;
2312 }
2313
2314 $d->{cpus} = ($conf->{sockets} || 1) * ($conf->{cores} || 1);
2315 $d->{cpus} = $cpucount if $d->{cpus} > $cpucount;
2316 $d->{cpus} = $conf->{vcpus} if $conf->{vcpus};
2317
2318 $d->{name} = $conf->{name} || "VM $vmid";
2319 $d->{maxmem} = $conf->{memory} ? $conf->{memory}*(1024*1024) : 0;
2320
2321 if ($conf->{balloon}) {
2322 $d->{balloon_min} = $conf->{balloon}*(1024*1024);
2323 $d->{shares} = defined($conf->{shares}) ? $conf->{shares} : 1000;
2324 }
2325
2326 $d->{uptime} = 0;
2327 $d->{cpu} = 0;
2328 $d->{mem} = 0;
2329
2330 $d->{netout} = 0;
2331 $d->{netin} = 0;
2332
2333 $d->{diskread} = 0;
2334 $d->{diskwrite} = 0;
2335
2336 $d->{template} = is_template($conf);
2337
2338 $res->{$vmid} = $d;
2339 }
2340
2341 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
2342 foreach my $dev (keys %$netdev) {
2343 next if $dev !~ m/^tap([1-9]\d*)i/;
2344 my $vmid = $1;
2345 my $d = $res->{$vmid};
2346 next if !$d;
2347
2348 $d->{netout} += $netdev->{$dev}->{receive};
2349 $d->{netin} += $netdev->{$dev}->{transmit};
2350
2351 if ($full) {
2352 $d->{nics}->{$dev}->{netout} = $netdev->{$dev}->{receive};
2353 $d->{nics}->{$dev}->{netin} = $netdev->{$dev}->{transmit};
2354 }
2355
2356 }
2357
2358 my $ctime = gettimeofday;
2359
2360 foreach my $vmid (keys %$list) {
2361
2362 my $d = $res->{$vmid};
2363 my $pid = $d->{pid};
2364 next if !$pid;
2365
2366 my $pstat = PVE::ProcFSTools::read_proc_pid_stat($pid);
2367 next if !$pstat; # not running
2368
2369 my $used = $pstat->{utime} + $pstat->{stime};
2370
2371 $d->{uptime} = int(($uptime - $pstat->{starttime})/$cpuinfo->{user_hz});
2372
2373 if ($pstat->{vsize}) {
2374 $d->{mem} = int(($pstat->{rss}/$pstat->{vsize})*$d->{maxmem});
2375 }
2376
2377 my $old = $last_proc_pid_stat->{$pid};
2378 if (!$old) {
2379 $last_proc_pid_stat->{$pid} = {
2380 time => $ctime,
2381 used => $used,
2382 cpu => 0,
2383 };
2384 next;
2385 }
2386
2387 my $dtime = ($ctime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
2388
2389 if ($dtime > 1000) {
2390 my $dutime = $used - $old->{used};
2391
2392 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
2393 $last_proc_pid_stat->{$pid} = {
2394 time => $ctime,
2395 used => $used,
2396 cpu => $d->{cpu},
2397 };
2398 } else {
2399 $d->{cpu} = $old->{cpu};
2400 }
2401 }
2402
2403 return $res if !$full;
2404
2405 my $qmpclient = PVE::QMPClient->new();
2406
2407 my $ballooncb = sub {
2408 my ($vmid, $resp) = @_;
2409
2410 my $info = $resp->{'return'};
2411 return if !$info->{max_mem};
2412
2413 my $d = $res->{$vmid};
2414
2415 # use memory assigned to VM
2416 $d->{maxmem} = $info->{max_mem};
2417 $d->{balloon} = $info->{actual};
2418
2419 if (defined($info->{total_mem}) && defined($info->{free_mem})) {
2420 $d->{mem} = $info->{total_mem} - $info->{free_mem};
2421 $d->{freemem} = $info->{free_mem};
2422 }
2423
2424 $d->{ballooninfo} = $info;
2425 };
2426
2427 my $blockstatscb = sub {
2428 my ($vmid, $resp) = @_;
2429 my $data = $resp->{'return'} || [];
2430 my $totalrdbytes = 0;
2431 my $totalwrbytes = 0;
2432
2433 for my $blockstat (@$data) {
2434 $totalrdbytes = $totalrdbytes + $blockstat->{stats}->{rd_bytes};
2435 $totalwrbytes = $totalwrbytes + $blockstat->{stats}->{wr_bytes};
2436
2437 $blockstat->{device} =~ s/drive-//;
2438 $res->{$vmid}->{blockstat}->{$blockstat->{device}} = $blockstat->{stats};
2439 }
2440 $res->{$vmid}->{diskread} = $totalrdbytes;
2441 $res->{$vmid}->{diskwrite} = $totalwrbytes;
2442 };
2443
2444 my $statuscb = sub {
2445 my ($vmid, $resp) = @_;
2446
2447 $qmpclient->queue_cmd($vmid, $blockstatscb, 'query-blockstats');
2448 # this fails if ballon driver is not loaded, so this must be
2449 # the last commnand (following command are aborted if this fails).
2450 $qmpclient->queue_cmd($vmid, $ballooncb, 'query-balloon');
2451
2452 my $status = 'unknown';
2453 if (!defined($status = $resp->{'return'}->{status})) {
2454 warn "unable to get VM status\n";
2455 return;
2456 }
2457
2458 $res->{$vmid}->{qmpstatus} = $resp->{'return'}->{status};
2459 };
2460
2461 foreach my $vmid (keys %$list) {
2462 next if $opt_vmid && ($vmid ne $opt_vmid);
2463 next if !$res->{$vmid}->{pid}; # not running
2464 $qmpclient->queue_cmd($vmid, $statuscb, 'query-status');
2465 }
2466
2467 $qmpclient->queue_execute(undef, 1);
2468
2469 foreach my $vmid (keys %$list) {
2470 next if $opt_vmid && ($vmid ne $opt_vmid);
2471 $res->{$vmid}->{qmpstatus} = $res->{$vmid}->{status} if !$res->{$vmid}->{qmpstatus};
2472 }
2473
2474 return $res;
2475 }
2476
2477 sub foreach_dimm {
2478 my ($conf, $vmid, $memory, $sockets, $func) = @_;
2479
2480 my $dimm_id = 0;
2481 my $current_size = 1024;
2482 my $dimm_size = 512;
2483 return if $current_size == $memory;
2484
2485 for (my $j = 0; $j < 8; $j++) {
2486 for (my $i = 0; $i < 32; $i++) {
2487 my $name = "dimm${dimm_id}";
2488 $dimm_id++;
2489 my $numanode = $i % $sockets;
2490 $current_size += $dimm_size;
2491 &$func($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory);
2492 return $current_size if $current_size >= $memory;
2493 }
2494 $dimm_size *= 2;
2495 }
2496 }
2497
2498 sub foreach_drive {
2499 my ($conf, $func) = @_;
2500
2501 foreach my $ds (keys %$conf) {
2502 next if !valid_drivename($ds);
2503
2504 my $drive = parse_drive($ds, $conf->{$ds});
2505 next if !$drive;
2506
2507 &$func($ds, $drive);
2508 }
2509 }
2510
2511 sub foreach_volid {
2512 my ($conf, $func) = @_;
2513
2514 my $volhash = {};
2515
2516 my $test_volid = sub {
2517 my ($volid, $is_cdrom) = @_;
2518
2519 return if !$volid;
2520
2521 $volhash->{$volid} = $is_cdrom || 0;
2522 };
2523
2524 foreach_drive($conf, sub {
2525 my ($ds, $drive) = @_;
2526 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2527 });
2528
2529 foreach my $snapname (keys %{$conf->{snapshots}}) {
2530 my $snap = $conf->{snapshots}->{$snapname};
2531 &$test_volid($snap->{vmstate}, 0);
2532 foreach_drive($snap, sub {
2533 my ($ds, $drive) = @_;
2534 &$test_volid($drive->{file}, drive_is_cdrom($drive));
2535 });
2536 }
2537
2538 foreach my $volid (keys %$volhash) {
2539 &$func($volid, $volhash->{$volid});
2540 }
2541 }
2542
2543 sub vga_conf_has_spice {
2544 my ($vga) = @_;
2545
2546 return 0 if !$vga || $vga !~ m/^qxl([234])?$/;
2547
2548 return $1 || 1;
2549 }
2550
2551 sub config_to_command {
2552 my ($storecfg, $vmid, $conf, $defaults, $forcemachine) = @_;
2553
2554 my $cmd = [];
2555 my $globalFlags = [];
2556 my $machineFlags = [];
2557 my $rtcFlags = [];
2558 my $cpuFlags = [];
2559 my $devices = [];
2560 my $pciaddr = '';
2561 my $bridges = {};
2562 my $kvmver = kvm_user_version();
2563 my $vernum = 0; # unknown
2564 if ($kvmver =~ m/^(\d+)\.(\d+)$/) {
2565 $vernum = $1*1000000+$2*1000;
2566 } elsif ($kvmver =~ m/^(\d+)\.(\d+)\.(\d+)$/) {
2567 $vernum = $1*1000000+$2*1000+$3;
2568 }
2569
2570 die "detected old qemu-kvm binary ($kvmver)\n" if $vernum < 15000;
2571
2572 my $have_ovz = -f '/proc/vz/vestat';
2573
2574 my $q35 = machine_type_is_q35($conf);
2575 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
2576 my $machine_type = $forcemachine || $conf->{machine};
2577
2578 my $cpuunits = defined($conf->{cpuunits}) ?
2579 $conf->{cpuunits} : $defaults->{cpuunits};
2580
2581 push @$cmd, '/usr/bin/systemd-run';
2582 push @$cmd, '--scope';
2583 push @$cmd, '--slice', "qemu";
2584 push @$cmd, '--unit', $vmid;
2585 push @$cmd, '-p', "CPUShares=$cpuunits";
2586 if ($conf->{cpulimit}) {
2587 my $cpulimit = int($conf->{cpulimit} * 100);
2588 push @$cmd, '-p', "CPUQuota=$cpulimit\%";
2589 }
2590
2591 push @$cmd, '/usr/bin/kvm';
2592
2593 push @$cmd, '-id', $vmid;
2594
2595 my $use_virtio = 0;
2596
2597 my $qmpsocket = qmp_socket($vmid);
2598 push @$cmd, '-chardev', "socket,id=qmp,path=$qmpsocket,server,nowait";
2599 push @$cmd, '-mon', "chardev=qmp,mode=control";
2600
2601 my $socket = vnc_socket($vmid);
2602 push @$cmd, '-vnc', "unix:$socket,x509,password";
2603
2604 push @$cmd, '-pidfile' , pidfile_name($vmid);
2605
2606 push @$cmd, '-daemonize';
2607
2608 if ($conf->{smbios1}) {
2609 push @$cmd, '-smbios', "type=1,$conf->{smbios1}";
2610 }
2611
2612 if ($q35) {
2613 # the q35 chipset support native usb2, so we enable usb controller
2614 # by default for this machine type
2615 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-q35.cfg';
2616 } else {
2617 $pciaddr = print_pci_addr("piix3", $bridges);
2618 push @$devices, '-device', "piix3-usb-uhci,id=uhci$pciaddr.0x2";
2619
2620 my $use_usb2 = 0;
2621 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2622 next if !$conf->{"usb$i"};
2623 $use_usb2 = 1;
2624 }
2625 # include usb device config
2626 push @$devices, '-readconfig', '/usr/share/qemu-server/pve-usb.cfg' if $use_usb2;
2627 }
2628
2629 my $vga = $conf->{vga};
2630
2631 my $qxlnum = vga_conf_has_spice($vga);
2632 $vga = 'qxl' if $qxlnum;
2633
2634 if (!$vga) {
2635 if ($conf->{ostype} && ($conf->{ostype} eq 'win8' ||
2636 $conf->{ostype} eq 'win7' ||
2637 $conf->{ostype} eq 'w2k8')) {
2638 $vga = 'std';
2639 } else {
2640 $vga = 'cirrus';
2641 }
2642 }
2643
2644 # enable absolute mouse coordinates (needed by vnc)
2645 my $tablet;
2646 if (defined($conf->{tablet})) {
2647 $tablet = $conf->{tablet};
2648 } else {
2649 $tablet = $defaults->{tablet};
2650 $tablet = 0 if $qxlnum; # disable for spice because it is not needed
2651 $tablet = 0 if $vga =~ m/^serial\d+$/; # disable if we use serial terminal (no vga card)
2652 }
2653
2654 push @$devices, '-device', print_tabletdevice_full($conf) if $tablet;
2655
2656 # host pci devices
2657 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
2658 my $d = parse_hostpci($conf->{"hostpci$i"});
2659 next if !$d;
2660
2661 my $pcie = $d->{pcie};
2662 if($pcie){
2663 die "q35 machine model is not enabled" if !$q35;
2664 $pciaddr = print_pcie_addr("hostpci$i");
2665 }else{
2666 $pciaddr = print_pci_addr("hostpci$i", $bridges);
2667 }
2668
2669 my $rombar = $d->{rombar} && $d->{rombar} eq 'off' ? ",rombar=0" : "";
2670 my $driver = $d->{driver} && $d->{driver} eq 'vfio' ? "vfio-pci" : "pci-assign";
2671 my $xvga = $d->{'x-vga'} && $d->{'x-vga'} eq 'on' ? ",x-vga=on" : "";
2672 if ($xvga && $xvga ne '') {
2673 push @$cpuFlags, 'kvm=off';
2674 $vga = 'none';
2675 }
2676 $driver = "vfio-pci" if $xvga ne '';
2677 my $pcidevices = $d->{pciid};
2678 my $multifunction = 1 if @$pcidevices > 1;
2679
2680 my $j=0;
2681 foreach my $pcidevice (@$pcidevices) {
2682
2683 my $id = "hostpci$i";
2684 $id .= ".$j" if $multifunction;
2685 my $addr = $pciaddr;
2686 $addr .= ".$j" if $multifunction;
2687 my $devicestr = "$driver,host=$pcidevice->{id}.$pcidevice->{function},id=$id$addr";
2688
2689 if($j == 0){
2690 $devicestr .= "$rombar$xvga";
2691 $devicestr .= ",multifunction=on" if $multifunction;
2692 }
2693
2694 push @$devices, '-device', $devicestr;
2695 $j++;
2696 }
2697 }
2698
2699 # usb devices
2700 for (my $i = 0; $i < $MAX_USB_DEVICES; $i++) {
2701 my $d = parse_usb_device($conf->{"usb$i"});
2702 next if !$d;
2703 if ($d->{vendorid} && $d->{productid}) {
2704 push @$devices, '-device', "usb-host,vendorid=0x$d->{vendorid},productid=0x$d->{productid}";
2705 } elsif (defined($d->{hostbus}) && defined($d->{hostport})) {
2706 push @$devices, '-device', "usb-host,hostbus=$d->{hostbus},hostport=$d->{hostport}";
2707 } elsif ($d->{spice}) {
2708 # usb redir support for spice
2709 push @$devices, '-chardev', "spicevmc,id=usbredirchardev$i,name=usbredir";
2710 push @$devices, '-device', "usb-redir,chardev=usbredirchardev$i,id=usbredirdev$i,bus=ehci.0";
2711 }
2712 }
2713
2714 # serial devices
2715 for (my $i = 0; $i < $MAX_SERIAL_PORTS; $i++) {
2716 if (my $path = $conf->{"serial$i"}) {
2717 if ($path eq 'socket') {
2718 my $socket = "/var/run/qemu-server/${vmid}.serial$i";
2719 push @$devices, '-chardev', "socket,id=serial$i,path=$socket,server,nowait";
2720 push @$devices, '-device', "isa-serial,chardev=serial$i";
2721 } else {
2722 die "no such serial device\n" if ! -c $path;
2723 push @$devices, '-chardev', "tty,id=serial$i,path=$path";
2724 push @$devices, '-device', "isa-serial,chardev=serial$i";
2725 }
2726 }
2727 }
2728
2729 # parallel devices
2730 for (my $i = 0; $i < $MAX_PARALLEL_PORTS; $i++) {
2731 if (my $path = $conf->{"parallel$i"}) {
2732 die "no such parallel device\n" if ! -c $path;
2733 my $devtype = $path =~ m!^/dev/usb/lp! ? 'tty' : 'parport';
2734 push @$devices, '-chardev', "$devtype,id=parallel$i,path=$path";
2735 push @$devices, '-device', "isa-parallel,chardev=parallel$i";
2736 }
2737 }
2738
2739 my $vmname = $conf->{name} || "vm$vmid";
2740
2741 push @$cmd, '-name', $vmname;
2742
2743 my $sockets = 1;
2744 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
2745 $sockets = $conf->{sockets} if $conf->{sockets};
2746
2747 my $cores = $conf->{cores} || 1;
2748
2749 my $maxcpus = $sockets * $cores;
2750
2751 my $vcpus = $conf->{vcpus} ? $conf->{vcpus} : $maxcpus;
2752
2753 my $allowed_vcpus = $cpuinfo->{cpus};
2754
2755 die "MAX $maxcpus vcpus allowed per VM on this node\n"
2756 if ($allowed_vcpus < $maxcpus);
2757
2758 push @$cmd, '-smp', "$vcpus,sockets=$sockets,cores=$cores,maxcpus=$maxcpus";
2759
2760 push @$cmd, '-nodefaults';
2761
2762 my $bootorder = $conf->{boot} || $confdesc->{boot}->{default};
2763
2764 my $bootindex_hash = {};
2765 my $i = 1;
2766 foreach my $o (split(//, $bootorder)) {
2767 $bootindex_hash->{$o} = $i*100;
2768 $i++;
2769 }
2770
2771 push @$cmd, '-boot', "menu=on,strict=on,reboot-timeout=1000";
2772
2773 push @$cmd, '-no-acpi' if defined($conf->{acpi}) && $conf->{acpi} == 0;
2774
2775 push @$cmd, '-no-reboot' if defined($conf->{reboot}) && $conf->{reboot} == 0;
2776
2777 push @$cmd, '-vga', $vga if $vga && $vga !~ m/^serial\d+$/; # for kvm 77 and later
2778
2779 # time drift fix
2780 my $tdf = defined($conf->{tdf}) ? $conf->{tdf} : $defaults->{tdf};
2781
2782 my $nokvm = defined($conf->{kvm}) && $conf->{kvm} == 0 ? 1 : 0;
2783 my $useLocaltime = $conf->{localtime};
2784
2785 if (my $ost = $conf->{ostype}) {
2786 # other, wxp, w2k, w2k3, w2k8, wvista, win7, win8, l24, l26, solaris
2787
2788 if ($ost =~ m/^w/) { # windows
2789 $useLocaltime = 1 if !defined($conf->{localtime});
2790
2791 # use time drift fix when acpi is enabled
2792 if (!(defined($conf->{acpi}) && $conf->{acpi} == 0)) {
2793 $tdf = 1 if !defined($conf->{tdf});
2794 }
2795 }
2796
2797 if ($ost eq 'win7' || $ost eq 'win8' || $ost eq 'w2k8' ||
2798 $ost eq 'wvista') {
2799 push @$globalFlags, 'kvm-pit.lost_tick_policy=discard';
2800 push @$cmd, '-no-hpet';
2801 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2802 push @$cpuFlags , 'hv_spinlocks=0x1fff' if !$nokvm;
2803 push @$cpuFlags , 'hv_vapic' if !$nokvm;
2804 push @$cpuFlags , 'hv_time' if !$nokvm;
2805
2806 } else {
2807 push @$cpuFlags , 'hv_spinlocks=0xffff' if !$nokvm;
2808 }
2809 }
2810
2811 if ($ost eq 'win7' || $ost eq 'win8') {
2812 push @$cpuFlags , 'hv_relaxed' if !$nokvm;
2813 }
2814 }
2815
2816 push @$rtcFlags, 'driftfix=slew' if $tdf;
2817
2818 if ($nokvm) {
2819 push @$machineFlags, 'accel=tcg';
2820 } else {
2821 die "No accelerator found!\n" if !$cpuinfo->{hvm};
2822 }
2823
2824 if ($machine_type) {
2825 push @$machineFlags, "type=${machine_type}";
2826 }
2827
2828 if ($conf->{startdate}) {
2829 push @$rtcFlags, "base=$conf->{startdate}";
2830 } elsif ($useLocaltime) {
2831 push @$rtcFlags, 'base=localtime';
2832 }
2833
2834 my $cpu = $nokvm ? "qemu64" : "kvm64";
2835 $cpu = $conf->{cpu} if $conf->{cpu};
2836
2837 push @$cpuFlags , '+lahf_lm' if $cpu eq 'kvm64';
2838
2839 push @$cpuFlags , '+x2apic' if !$nokvm && $conf->{ostype} ne 'solaris';
2840
2841 push @$cpuFlags , '-x2apic' if $conf->{ostype} eq 'solaris';
2842
2843 push @$cpuFlags, '+sep' if $cpu eq 'kvm64' || $cpu eq 'kvm32';
2844
2845 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2846
2847 push @$cpuFlags , '+kvm_pv_unhalt' if !$nokvm;
2848 push @$cpuFlags , '+kvm_pv_eoi' if !$nokvm;
2849 }
2850
2851 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2852
2853 push @$cmd, '-cpu', "$cpu,enforce";
2854
2855 my $memory = $conf->{memory} || $defaults->{memory};
2856 my $static_memory = 0;
2857 my $dimm_memory = 0;
2858
2859 if ($hotplug_features->{memory}) {
2860 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2861 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2862 $static_memory = $STATICMEM;
2863 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2864 $dimm_memory = $memory - $static_memory;
2865 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2866
2867 } else {
2868
2869 $static_memory = $memory;
2870 push @$cmd, '-m', $static_memory;
2871 }
2872
2873 if ($conf->{numa}) {
2874
2875 my $numa_totalmemory = undef;
2876 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2877 next if !$conf->{"numa$i"};
2878 my $numa = parse_numa($conf->{"numa$i"});
2879 next if !$numa;
2880 # memory
2881 die "missing numa node$i memory value\n" if !$numa->{memory};
2882 my $numa_memory = $numa->{memory};
2883 $numa_totalmemory += $numa_memory;
2884 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2885
2886 # cpus
2887 my $cpus_start = $numa->{cpus}->{start};
2888 die "missing numa node$i cpus\n" if !defined($cpus_start);
2889 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2890 my $cpus = $cpus_start;
2891 if (defined($cpus_end)) {
2892 $cpus .= "-$cpus_end";
2893 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2894 }
2895
2896 # hostnodes
2897 my $hostnodes_start = $numa->{hostnodes}->{start};
2898 if (defined($hostnodes_start)) {
2899 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2900 my $hostnodes = $hostnodes_start;
2901 if (defined($hostnodes_end)) {
2902 $hostnodes .= "-$hostnodes_end";
2903 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2904 }
2905
2906 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2907 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2908 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2909 }
2910
2911 # policy
2912 my $policy = $numa->{policy};
2913 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2914 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2915 }
2916
2917 push @$cmd, '-object', $numa_object;
2918 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2919 }
2920
2921 die "total memory for NUMA nodes must be equal to vm static memory\n"
2922 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2923
2924 #if no custom tology, we split memory and cores across numa nodes
2925 if(!$numa_totalmemory) {
2926
2927 my $numa_memory = ($static_memory / $sockets) . "M";
2928
2929 for (my $i = 0; $i < $sockets; $i++) {
2930
2931 my $cpustart = ($cores * $i);
2932 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2933 my $cpus = $cpustart;
2934 $cpus .= "-$cpuend" if $cpuend;
2935
2936 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2937 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2938 }
2939 }
2940 }
2941
2942 if ($hotplug_features->{memory}) {
2943 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
2944 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
2945 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
2946 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
2947
2948 #if dimm_memory is not aligned to dimm map
2949 if($current_size > $memory) {
2950 $conf->{memory} = $current_size;
2951 update_config_nolock($vmid, $conf, 1);
2952 }
2953 });
2954 }
2955
2956 push @$cmd, '-S' if $conf->{freeze};
2957
2958 # set keyboard layout
2959 my $kb = $conf->{keyboard} || $defaults->{keyboard};
2960 push @$cmd, '-k', $kb if $kb;
2961
2962 # enable sound
2963 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
2964 #push @$cmd, '-soundhw', 'es1370';
2965 #push @$cmd, '-soundhw', $soundhw if $soundhw;
2966
2967 if($conf->{agent}) {
2968 my $qgasocket = qmp_socket($vmid, 1);
2969 my $pciaddr = print_pci_addr("qga0", $bridges);
2970 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
2971 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
2972 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
2973 }
2974
2975 my $spice_port;
2976
2977 if ($qxlnum) {
2978 if ($qxlnum > 1) {
2979 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
2980 for(my $i = 1; $i < $qxlnum; $i++){
2981 my $pciaddr = print_pci_addr("vga$i", $bridges);
2982 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
2983 }
2984 } else {
2985 # assume other OS works like Linux
2986 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
2987 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
2988 }
2989 }
2990
2991 my $pciaddr = print_pci_addr("spice", $bridges);
2992
2993 my $nodename = PVE::INotify::nodename();
2994 my $pfamily = PVE::Tools::get_host_address_family($nodename);
2995 $spice_port = PVE::Tools::next_spice_port($pfamily);
2996
2997 push @$devices, '-spice', "tls-port=${spice_port},addr=localhost,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
2998
2999 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3000 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3001 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3002 }
3003
3004 # enable balloon by default, unless explicitly disabled
3005 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3006 $pciaddr = print_pci_addr("balloon0", $bridges);
3007 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3008 }
3009
3010 if ($conf->{watchdog}) {
3011 my $wdopts = parse_watchdog($conf->{watchdog});
3012 $pciaddr = print_pci_addr("watchdog", $bridges);
3013 my $watchdog = $wdopts->{model} || 'i6300esb';
3014 push @$devices, '-device', "$watchdog$pciaddr";
3015 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3016 }
3017
3018 my $vollist = [];
3019 my $scsicontroller = {};
3020 my $ahcicontroller = {};
3021 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3022
3023 # Add iscsi initiator name if available
3024 if (my $initiator = get_initiator_name()) {
3025 push @$devices, '-iscsi', "initiator-name=$initiator";
3026 }
3027
3028 foreach_drive($conf, sub {
3029 my ($ds, $drive) = @_;
3030
3031 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3032 push @$vollist, $drive->{file};
3033 }
3034
3035 $use_virtio = 1 if $ds =~ m/^virtio/;
3036
3037 if (drive_is_cdrom ($drive)) {
3038 if ($bootindex_hash->{d}) {
3039 $drive->{bootindex} = $bootindex_hash->{d};
3040 $bootindex_hash->{d} += 1;
3041 }
3042 } else {
3043 if ($bootindex_hash->{c}) {
3044 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3045 $bootindex_hash->{c} += 1;
3046 }
3047 }
3048
3049 if($drive->{interface} eq 'virtio'){
3050 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3051 }
3052
3053 if ($drive->{interface} eq 'scsi') {
3054
3055 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3056
3057 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3058 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3059
3060 my $iothread = '';
3061 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3062 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3063 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3064 }
3065
3066 my $queues = '';
3067 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3068 $queues = ",num_queues=$drive->{queues}";
3069 }
3070
3071 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3072 $scsicontroller->{$controller}=1;
3073 }
3074
3075 if ($drive->{interface} eq 'sata') {
3076 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3077 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3078 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3079 $ahcicontroller->{$controller}=1;
3080 }
3081
3082 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3083 push @$devices, '-drive',$drive_cmd;
3084 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3085 });
3086
3087 for (my $i = 0; $i < $MAX_NETS; $i++) {
3088 next if !$conf->{"net$i"};
3089 my $d = parse_net($conf->{"net$i"});
3090 next if !$d;
3091
3092 $use_virtio = 1 if $d->{model} eq 'virtio';
3093
3094 if ($bootindex_hash->{n}) {
3095 $d->{bootindex} = $bootindex_hash->{n};
3096 $bootindex_hash->{n} += 1;
3097 }
3098
3099 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3100 push @$devices, '-netdev', $netdevfull;
3101
3102 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3103 push @$devices, '-device', $netdevicefull;
3104 }
3105
3106 if (!$q35) {
3107 # add pci bridges
3108 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3109 $bridges->{1} = 1;
3110 $bridges->{2} = 1;
3111 }
3112
3113 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3114
3115 while (my ($k, $v) = each %$bridges) {
3116 $pciaddr = print_pci_addr("pci.$k");
3117 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3118 }
3119 }
3120
3121 # add custom args
3122 if ($conf->{args}) {
3123 my $aa = PVE::Tools::split_args($conf->{args});
3124 push @$cmd, @$aa;
3125 }
3126
3127 push @$cmd, @$devices;
3128 push @$cmd, '-rtc', join(',', @$rtcFlags)
3129 if scalar(@$rtcFlags);
3130 push @$cmd, '-machine', join(',', @$machineFlags)
3131 if scalar(@$machineFlags);
3132 push @$cmd, '-global', join(',', @$globalFlags)
3133 if scalar(@$globalFlags);
3134
3135 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3136 }
3137
3138 sub vnc_socket {
3139 my ($vmid) = @_;
3140 return "${var_run_tmpdir}/$vmid.vnc";
3141 }
3142
3143 sub spice_port {
3144 my ($vmid) = @_;
3145
3146 my $res = vm_mon_cmd($vmid, 'query-spice');
3147
3148 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3149 }
3150
3151 sub qmp_socket {
3152 my ($vmid, $qga) = @_;
3153 my $sockettype = $qga ? 'qga' : 'qmp';
3154 return "${var_run_tmpdir}/$vmid.$sockettype";
3155 }
3156
3157 sub pidfile_name {
3158 my ($vmid) = @_;
3159 return "${var_run_tmpdir}/$vmid.pid";
3160 }
3161
3162 sub vm_devices_list {
3163 my ($vmid) = @_;
3164
3165 my $res = vm_mon_cmd($vmid, 'query-pci');
3166 my $devices = {};
3167 foreach my $pcibus (@$res) {
3168 foreach my $device (@{$pcibus->{devices}}) {
3169 next if !$device->{'qdev_id'};
3170 if ($device->{'pci_bridge'}) {
3171 $devices->{$device->{'qdev_id'}} = 1;
3172 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3173 next if !$bridge_device->{'qdev_id'};
3174 $devices->{$bridge_device->{'qdev_id'}} = 1;
3175 $devices->{$device->{'qdev_id'}}++;
3176 }
3177 } else {
3178 $devices->{$device->{'qdev_id'}} = 1;
3179 }
3180 }
3181 }
3182
3183 my $resblock = vm_mon_cmd($vmid, 'query-block');
3184 foreach my $block (@$resblock) {
3185 if($block->{device} =~ m/^drive-(\S+)/){
3186 $devices->{$1} = 1;
3187 }
3188 }
3189
3190 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3191 foreach my $mice (@$resmice) {
3192 if ($mice->{name} eq 'QEMU HID Tablet') {
3193 $devices->{tablet} = 1;
3194 last;
3195 }
3196 }
3197
3198 return $devices;
3199 }
3200
3201 sub vm_deviceplug {
3202 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3203
3204 my $q35 = machine_type_is_q35($conf);
3205
3206 my $devices_list = vm_devices_list($vmid);
3207 return 1 if defined($devices_list->{$deviceid});
3208
3209 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3210
3211 if ($deviceid eq 'tablet') {
3212
3213 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3214
3215 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3216
3217 qemu_iothread_add($vmid, $deviceid, $device);
3218
3219 qemu_driveadd($storecfg, $vmid, $device);
3220 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3221
3222 qemu_deviceadd($vmid, $devicefull);
3223 eval { qemu_deviceaddverify($vmid, $deviceid); };
3224 if (my $err = $@) {
3225 eval { qemu_drivedel($vmid, $deviceid); };
3226 warn $@ if $@;
3227 die $err;
3228 }
3229
3230 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3231
3232
3233 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3234 my $pciaddr = print_pci_addr($deviceid);
3235 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3236
3237 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3238
3239 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3240 qemu_iothread_add($vmid, $deviceid, $device);
3241 $devicefull .= ",iothread=iothread-$deviceid";
3242 }
3243
3244 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3245 $devicefull .= ",num_queues=$device->{queues}";
3246 }
3247
3248 qemu_deviceadd($vmid, $devicefull);
3249 qemu_deviceaddverify($vmid, $deviceid);
3250
3251 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3252
3253 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3254 qemu_driveadd($storecfg, $vmid, $device);
3255
3256 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3257 eval { qemu_deviceadd($vmid, $devicefull); };
3258 if (my $err = $@) {
3259 eval { qemu_drivedel($vmid, $deviceid); };
3260 warn $@ if $@;
3261 die $err;
3262 }
3263
3264 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3265
3266 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3267 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3268 qemu_deviceadd($vmid, $netdevicefull);
3269 eval { qemu_deviceaddverify($vmid, $deviceid); };
3270 if (my $err = $@) {
3271 eval { qemu_netdevdel($vmid, $deviceid); };
3272 warn $@ if $@;
3273 die $err;
3274 }
3275
3276 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3277
3278 my $bridgeid = $2;
3279 my $pciaddr = print_pci_addr($deviceid);
3280 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3281
3282 qemu_deviceadd($vmid, $devicefull);
3283 qemu_deviceaddverify($vmid, $deviceid);
3284
3285 } else {
3286 die "can't hotplug device '$deviceid'\n";
3287 }
3288
3289 return 1;
3290 }
3291
3292 # fixme: this should raise exceptions on error!
3293 sub vm_deviceunplug {
3294 my ($vmid, $conf, $deviceid) = @_;
3295
3296 my $devices_list = vm_devices_list($vmid);
3297 return 1 if !defined($devices_list->{$deviceid});
3298
3299 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3300
3301 if ($deviceid eq 'tablet') {
3302
3303 qemu_devicedel($vmid, $deviceid);
3304
3305 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3306
3307 qemu_devicedel($vmid, $deviceid);
3308 qemu_devicedelverify($vmid, $deviceid);
3309 qemu_drivedel($vmid, $deviceid);
3310 qemu_iothread_del($conf, $vmid, $deviceid);
3311
3312 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3313
3314 qemu_devicedel($vmid, $deviceid);
3315 qemu_devicedelverify($vmid, $deviceid);
3316 qemu_iothread_del($conf, $vmid, $deviceid);
3317
3318 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3319
3320 #qemu 2.3 segfault on drive_del with virtioscsi + iothread
3321 my $device = parse_drive($deviceid, $conf->{$deviceid});
3322 die "virtioscsi with iothread is not hot-unplugglable currently" if $device->{iothread};
3323
3324 qemu_devicedel($vmid, $deviceid);
3325 qemu_drivedel($vmid, $deviceid);
3326 qemu_deletescsihw($conf, $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 => $deviceid);
3354 }
3355
3356 sub qemu_iothread_add {
3357 my($vmid, $deviceid, $device) = @_;
3358
3359 if ($device->{iothread}) {
3360 my $iothreads = vm_iothreads_list($vmid);
3361 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3362 }
3363 }
3364
3365 sub qemu_iothread_del {
3366 my($conf, $vmid, $deviceid) = @_;
3367
3368 my $device = parse_drive($deviceid, $conf->{$deviceid});
3369 if ($device->{iothread}) {
3370 my $iothreads = vm_iothreads_list($vmid);
3371 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3372 }
3373 }
3374
3375 sub qemu_objectadd {
3376 my($vmid, $objectid, $qomtype) = @_;
3377
3378 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3379
3380 return 1;
3381 }
3382
3383 sub qemu_objectdel {
3384 my($vmid, $objectid) = @_;
3385
3386 vm_mon_cmd($vmid, "object-del", id => $objectid);
3387
3388 return 1;
3389 }
3390
3391 sub qemu_driveadd {
3392 my ($storecfg, $vmid, $device) = @_;
3393
3394 my $drive = print_drive_full($storecfg, $vmid, $device);
3395 $drive =~ s/\\/\\\\/g;
3396 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3397
3398 # If the command succeeds qemu prints: "OK"
3399 return 1 if $ret =~ m/OK/s;
3400
3401 die "adding drive failed: $ret\n";
3402 }
3403
3404 sub qemu_drivedel {
3405 my($vmid, $deviceid) = @_;
3406
3407 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3408 $ret =~ s/^\s+//;
3409
3410 return 1 if $ret eq "";
3411
3412 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3413 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3414
3415 die "deleting drive $deviceid failed : $ret\n";
3416 }
3417
3418 sub qemu_deviceaddverify {
3419 my ($vmid, $deviceid) = @_;
3420
3421 for (my $i = 0; $i <= 5; $i++) {
3422 my $devices_list = vm_devices_list($vmid);
3423 return 1 if defined($devices_list->{$deviceid});
3424 sleep 1;
3425 }
3426
3427 die "error on hotplug device '$deviceid'\n";
3428 }
3429
3430
3431 sub qemu_devicedelverify {
3432 my ($vmid, $deviceid) = @_;
3433
3434 # need to verify that the device is correctly removed as device_del
3435 # is async and empty return is not reliable
3436
3437 for (my $i = 0; $i <= 5; $i++) {
3438 my $devices_list = vm_devices_list($vmid);
3439 return 1 if !defined($devices_list->{$deviceid});
3440 sleep 1;
3441 }
3442
3443 die "error on hot-unplugging device '$deviceid'\n";
3444 }
3445
3446 sub qemu_findorcreatescsihw {
3447 my ($storecfg, $conf, $vmid, $device) = @_;
3448
3449 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3450
3451 my $scsihwid="$controller_prefix$controller";
3452 my $devices_list = vm_devices_list($vmid);
3453
3454 if(!defined($devices_list->{$scsihwid})) {
3455 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
3456 }
3457
3458 return 1;
3459 }
3460
3461 sub qemu_deletescsihw {
3462 my ($conf, $vmid, $opt) = @_;
3463
3464 my $device = parse_drive($opt, $conf->{$opt});
3465
3466 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
3467 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3468 return 1;
3469 }
3470
3471 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3472
3473 my $devices_list = vm_devices_list($vmid);
3474 foreach my $opt (keys %{$devices_list}) {
3475 if (PVE::QemuServer::valid_drivename($opt)) {
3476 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
3477 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
3478 return 1;
3479 }
3480 }
3481 }
3482
3483 my $scsihwid="scsihw$controller";
3484
3485 vm_deviceunplug($vmid, $conf, $scsihwid);
3486
3487 return 1;
3488 }
3489
3490 sub qemu_add_pci_bridge {
3491 my ($storecfg, $conf, $vmid, $device) = @_;
3492
3493 my $bridges = {};
3494
3495 my $bridgeid;
3496
3497 print_pci_addr($device, $bridges);
3498
3499 while (my ($k, $v) = each %$bridges) {
3500 $bridgeid = $k;
3501 }
3502 return 1 if !defined($bridgeid) || $bridgeid < 1;
3503
3504 my $bridge = "pci.$bridgeid";
3505 my $devices_list = vm_devices_list($vmid);
3506
3507 if (!defined($devices_list->{$bridge})) {
3508 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3509 }
3510
3511 return 1;
3512 }
3513
3514 sub qemu_set_link_status {
3515 my ($vmid, $device, $up) = @_;
3516
3517 vm_mon_cmd($vmid, "set_link", name => $device,
3518 up => $up ? JSON::true : JSON::false);
3519 }
3520
3521 sub qemu_netdevadd {
3522 my ($vmid, $conf, $device, $deviceid) = @_;
3523
3524 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3525 my %options = split(/[=,]/, $netdev);
3526
3527 vm_mon_cmd($vmid, "netdev_add", %options);
3528 return 1;
3529 }
3530
3531 sub qemu_netdevdel {
3532 my ($vmid, $deviceid) = @_;
3533
3534 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3535 }
3536
3537 sub qemu_cpu_hotplug {
3538 my ($vmid, $conf, $vcpus) = @_;
3539
3540 my $sockets = 1;
3541 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3542 $sockets = $conf->{sockets} if $conf->{sockets};
3543 my $cores = $conf->{cores} || 1;
3544 my $maxcpus = $sockets * $cores;
3545
3546 $vcpus = $maxcpus if !$vcpus;
3547
3548 die "you can't add more vcpus than maxcpus\n"
3549 if $vcpus > $maxcpus;
3550
3551 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3552 die "online cpu unplug is not yet possible\n"
3553 if $vcpus < $currentvcpus;
3554
3555 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3556 die "vcpus in running vm is different than configuration\n"
3557 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3558
3559 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3560 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3561 }
3562 }
3563
3564 sub qemu_memory_hotplug {
3565 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3566
3567 return $value if !check_running($vmid);
3568
3569 my $memory = $conf->{memory} || $defaults->{memory};
3570 $value = $defaults->{memory} if !$value;
3571 return $value if $value == $memory;
3572
3573 my $static_memory = $STATICMEM;
3574 my $dimm_memory = $memory - $static_memory;
3575
3576 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3577 die "memory unplug is not yet available" if $value < $memory;
3578 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3579
3580
3581 my $sockets = 1;
3582 $sockets = $conf->{sockets} if $conf->{sockets};
3583
3584 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3585 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3586
3587 return if $current_size <= $conf->{memory};
3588
3589 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3590 if (my $err = $@) {
3591 eval { qemu_objectdel($vmid, "mem-$name"); };
3592 die $err;
3593 }
3594
3595 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3596 if (my $err = $@) {
3597 eval { qemu_objectdel($vmid, "mem-$name"); };
3598 die $err;
3599 }
3600 #update conf after each succesful module hotplug
3601 $conf->{memory} = $current_size;
3602 update_config_nolock($vmid, $conf, 1);
3603 });
3604 }
3605
3606 sub qemu_block_set_io_throttle {
3607 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3608
3609 return if !check_running($vmid) ;
3610
3611 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($iops_wr));
3612
3613 }
3614
3615 # old code, only used to shutdown old VM after update
3616 sub __read_avail {
3617 my ($fh, $timeout) = @_;
3618
3619 my $sel = new IO::Select;
3620 $sel->add($fh);
3621
3622 my $res = '';
3623 my $buf;
3624
3625 my @ready;
3626 while (scalar (@ready = $sel->can_read($timeout))) {
3627 my $count;
3628 if ($count = $fh->sysread($buf, 8192)) {
3629 if ($buf =~ /^(.*)\(qemu\) $/s) {
3630 $res .= $1;
3631 last;
3632 } else {
3633 $res .= $buf;
3634 }
3635 } else {
3636 if (!defined($count)) {
3637 die "$!\n";
3638 }
3639 last;
3640 }
3641 }
3642
3643 die "monitor read timeout\n" if !scalar(@ready);
3644
3645 return $res;
3646 }
3647
3648 # old code, only used to shutdown old VM after update
3649 sub vm_monitor_command {
3650 my ($vmid, $cmdstr, $nocheck) = @_;
3651
3652 my $res;
3653
3654 eval {
3655 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3656
3657 my $sname = "${var_run_tmpdir}/$vmid.mon";
3658
3659 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3660 die "unable to connect to VM $vmid socket - $!\n";
3661
3662 my $timeout = 3;
3663
3664 # hack: migrate sometime blocks the monitor (when migrate_downtime
3665 # is set)
3666 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3667 $timeout = 60*60; # 1 hour
3668 }
3669
3670 # read banner;
3671 my $data = __read_avail($sock, $timeout);
3672
3673 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3674 die "got unexpected qemu monitor banner\n";
3675 }
3676
3677 my $sel = new IO::Select;
3678 $sel->add($sock);
3679
3680 if (!scalar(my @ready = $sel->can_write($timeout))) {
3681 die "monitor write error - timeout";
3682 }
3683
3684 my $fullcmd = "$cmdstr\r";
3685
3686 # syslog('info', "VM $vmid monitor command: $cmdstr");
3687
3688 my $b;
3689 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3690 die "monitor write error - $!";
3691 }
3692
3693 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3694
3695 $timeout = 20;
3696
3697 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3698 $timeout = 60*60; # 1 hour
3699 } elsif ($cmdstr =~ m/^(eject|change)/) {
3700 $timeout = 60; # note: cdrom mount command is slow
3701 }
3702 if ($res = __read_avail($sock, $timeout)) {
3703
3704 my @lines = split("\r?\n", $res);
3705
3706 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3707
3708 $res = join("\n", @lines);
3709 $res .= "\n";
3710 }
3711 };
3712
3713 my $err = $@;
3714
3715 if ($err) {
3716 syslog("err", "VM $vmid monitor command failed - $err");
3717 die $err;
3718 }
3719
3720 return $res;
3721 }
3722
3723 sub qemu_block_resize {
3724 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3725
3726 my $running = check_running($vmid);
3727
3728 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3729
3730 return if !$running;
3731
3732 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3733
3734 }
3735
3736 sub qemu_volume_snapshot {
3737 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3738
3739 my $running = check_running($vmid);
3740
3741 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
3742 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3743 } else {
3744 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
3745 }
3746 }
3747
3748 sub qemu_volume_snapshot_delete {
3749 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3750
3751 my $running = check_running($vmid);
3752
3753 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3754
3755 return if !$running;
3756
3757 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3758 }
3759
3760 sub set_migration_caps {
3761 my ($vmid) = @_;
3762
3763 my $cap_ref = [];
3764
3765 my $enabled_cap = {
3766 "auto-converge" => 1,
3767 "xbzrle" => 0,
3768 "x-rdma-pin-all" => 0,
3769 "zero-blocks" => 0,
3770 };
3771
3772 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3773
3774 for my $supported_capability (@$supported_capabilities) {
3775 push @$cap_ref, {
3776 capability => $supported_capability->{capability},
3777 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3778 };
3779 }
3780
3781 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3782 }
3783
3784 my $fast_plug_option = {
3785 'lock' => 1,
3786 'name' => 1,
3787 'onboot' => 1,
3788 'shares' => 1,
3789 'startup' => 1,
3790 };
3791
3792 # hotplug changes in [PENDING]
3793 # $selection hash can be used to only apply specified options, for
3794 # example: { cores => 1 } (only apply changed 'cores')
3795 # $errors ref is used to return error messages
3796 sub vmconfig_hotplug_pending {
3797 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3798
3799 my $defaults = load_defaults();
3800
3801 # commit values which do not have any impact on running VM first
3802 # Note: those option cannot raise errors, we we do not care about
3803 # $selection and always apply them.
3804
3805 my $add_error = sub {
3806 my ($opt, $msg) = @_;
3807 $errors->{$opt} = "hotplug problem - $msg";
3808 };
3809
3810 my $changes = 0;
3811 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3812 if ($fast_plug_option->{$opt}) {
3813 $conf->{$opt} = $conf->{pending}->{$opt};
3814 delete $conf->{pending}->{$opt};
3815 $changes = 1;
3816 }
3817 }
3818
3819 if ($changes) {
3820 update_config_nolock($vmid, $conf, 1);
3821 $conf = load_config($vmid); # update/reload
3822 }
3823
3824 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3825
3826 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3827 foreach my $opt (@delete) {
3828 next if $selection && !$selection->{$opt};
3829 eval {
3830 if ($opt eq 'hotplug') {
3831 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3832 } elsif ($opt eq 'tablet') {
3833 die "skip\n" if !$hotplug_features->{usb};
3834 if ($defaults->{tablet}) {
3835 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3836 } else {
3837 vm_deviceunplug($vmid, $conf, $opt);
3838 }
3839 } elsif ($opt eq 'vcpus') {
3840 die "skip\n" if !$hotplug_features->{cpu};
3841 qemu_cpu_hotplug($vmid, $conf, undef);
3842 } elsif ($opt eq 'balloon') {
3843 # enable balloon device is not hotpluggable
3844 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3845 } elsif ($fast_plug_option->{$opt}) {
3846 # do nothing
3847 } elsif ($opt =~ m/^net(\d+)$/) {
3848 die "skip\n" if !$hotplug_features->{network};
3849 vm_deviceunplug($vmid, $conf, $opt);
3850 } elsif (valid_drivename($opt)) {
3851 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3852 vm_deviceunplug($vmid, $conf, $opt);
3853 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3854 } elsif ($opt =~ m/^memory$/) {
3855 die "skip\n" if !$hotplug_features->{memory};
3856 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3857 } elsif ($opt eq 'cpuunits') {
3858 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
3859 } elsif ($opt eq 'cpulimit') {
3860 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
3861 } else {
3862 die "skip\n";
3863 }
3864 };
3865 if (my $err = $@) {
3866 &$add_error($opt, $err) if $err ne "skip\n";
3867 } else {
3868 # save new config if hotplug was successful
3869 delete $conf->{$opt};
3870 vmconfig_undelete_pending_option($conf, $opt);
3871 update_config_nolock($vmid, $conf, 1);
3872 $conf = load_config($vmid); # update/reload
3873 }
3874 }
3875
3876 foreach my $opt (keys %{$conf->{pending}}) {
3877 next if $selection && !$selection->{$opt};
3878 my $value = $conf->{pending}->{$opt};
3879 eval {
3880 if ($opt eq 'hotplug') {
3881 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3882 } elsif ($opt eq 'tablet') {
3883 die "skip\n" if !$hotplug_features->{usb};
3884 if ($value == 1) {
3885 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3886 } elsif ($value == 0) {
3887 vm_deviceunplug($vmid, $conf, $opt);
3888 }
3889 } elsif ($opt eq 'vcpus') {
3890 die "skip\n" if !$hotplug_features->{cpu};
3891 qemu_cpu_hotplug($vmid, $conf, $value);
3892 } elsif ($opt eq 'balloon') {
3893 # enable/disable balloning device is not hotpluggable
3894 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3895 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3896 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3897
3898 # allow manual ballooning if shares is set to zero
3899 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3900 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3901 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3902 }
3903 } elsif ($opt =~ m/^net(\d+)$/) {
3904 # some changes can be done without hotplug
3905 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3906 $vmid, $opt, $value);
3907 } elsif (valid_drivename($opt)) {
3908 # some changes can be done without hotplug
3909 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3910 $vmid, $opt, $value, 1);
3911 } elsif ($opt =~ m/^memory$/) { #dimms
3912 die "skip\n" if !$hotplug_features->{memory};
3913 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3914 } elsif ($opt eq 'cpuunits') {
3915 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
3916 } elsif ($opt eq 'cpulimit') {
3917 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
3918 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
3919 } else {
3920 die "skip\n"; # skip non-hot-pluggable options
3921 }
3922 };
3923 if (my $err = $@) {
3924 &$add_error($opt, $err) if $err ne "skip\n";
3925 } else {
3926 # save new config if hotplug was successful
3927 $conf->{$opt} = $value;
3928 delete $conf->{pending}->{$opt};
3929 update_config_nolock($vmid, $conf, 1);
3930 $conf = load_config($vmid); # update/reload
3931 }
3932 }
3933 }
3934
3935 sub vmconfig_apply_pending {
3936 my ($vmid, $conf, $storecfg) = @_;
3937
3938 # cold plug
3939
3940 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3941 foreach my $opt (@delete) { # delete
3942 die "internal error" if $opt =~ m/^unused/;
3943 $conf = load_config($vmid); # update/reload
3944 if (!defined($conf->{$opt})) {
3945 vmconfig_undelete_pending_option($conf, $opt);
3946 update_config_nolock($vmid, $conf, 1);
3947 } elsif (valid_drivename($opt)) {
3948 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3949 vmconfig_undelete_pending_option($conf, $opt);
3950 delete $conf->{$opt};
3951 update_config_nolock($vmid, $conf, 1);
3952 } else {
3953 vmconfig_undelete_pending_option($conf, $opt);
3954 delete $conf->{$opt};
3955 update_config_nolock($vmid, $conf, 1);
3956 }
3957 }
3958
3959 $conf = load_config($vmid); # update/reload
3960
3961 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3962 $conf = load_config($vmid); # update/reload
3963
3964 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3965 # skip if nothing changed
3966 } elsif (valid_drivename($opt)) {
3967 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3968 if defined($conf->{$opt});
3969 $conf->{$opt} = $conf->{pending}->{$opt};
3970 } else {
3971 $conf->{$opt} = $conf->{pending}->{$opt};
3972 }
3973
3974 delete $conf->{pending}->{$opt};
3975 update_config_nolock($vmid, $conf, 1);
3976 }
3977 }
3978
3979 my $safe_num_ne = sub {
3980 my ($a, $b) = @_;
3981
3982 return 0 if !defined($a) && !defined($b);
3983 return 1 if !defined($a);
3984 return 1 if !defined($b);
3985
3986 return $a != $b;
3987 };
3988
3989 my $safe_string_ne = sub {
3990 my ($a, $b) = @_;
3991
3992 return 0 if !defined($a) && !defined($b);
3993 return 1 if !defined($a);
3994 return 1 if !defined($b);
3995
3996 return $a ne $b;
3997 };
3998
3999 sub vmconfig_update_net {
4000 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4001
4002 my $newnet = parse_net($value);
4003
4004 if ($conf->{$opt}) {
4005 my $oldnet = parse_net($conf->{$opt});
4006
4007 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4008 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4009 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4010 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4011
4012 # for non online change, we try to hot-unplug
4013 die "skip\n" if !$hotplug;
4014 vm_deviceunplug($vmid, $conf, $opt);
4015 } else {
4016
4017 die "internal error" if $opt !~ m/net(\d+)/;
4018 my $iface = "tap${vmid}i$1";
4019
4020 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4021 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4022 }
4023
4024 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4025 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4026 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4027 PVE::Network::tap_unplug($iface);
4028 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
4029 }
4030
4031 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4032 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4033 }
4034
4035 return 1;
4036 }
4037 }
4038
4039 if ($hotplug) {
4040 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4041 } else {
4042 die "skip\n";
4043 }
4044 }
4045
4046 sub vmconfig_update_disk {
4047 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4048
4049 # fixme: do we need force?
4050
4051 my $drive = parse_drive($opt, $value);
4052
4053 if ($conf->{$opt}) {
4054
4055 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4056
4057 my $media = $drive->{media} || 'disk';
4058 my $oldmedia = $old_drive->{media} || 'disk';
4059 die "unable to change media type\n" if $media ne $oldmedia;
4060
4061 if (!drive_is_cdrom($old_drive)) {
4062
4063 if ($drive->{file} ne $old_drive->{file}) {
4064
4065 die "skip\n" if !$hotplug;
4066
4067 # unplug and register as unused
4068 vm_deviceunplug($vmid, $conf, $opt);
4069 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4070
4071 } else {
4072 # update existing disk
4073
4074 # skip non hotpluggable value
4075 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4076 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4077 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4078 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4079 die "skip\n";
4080 }
4081
4082 # apply throttle
4083 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4084 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4085 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4086 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4087 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4088 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4089 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4090 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4091 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4092 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4093 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4094 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4095
4096 qemu_block_set_io_throttle($vmid,"drive-$opt",
4097 ($drive->{mbps} || 0)*1024*1024,
4098 ($drive->{mbps_rd} || 0)*1024*1024,
4099 ($drive->{mbps_wr} || 0)*1024*1024,
4100 $drive->{iops} || 0,
4101 $drive->{iops_rd} || 0,
4102 $drive->{iops_wr} || 0,
4103 ($drive->{mbps_max} || 0)*1024*1024,
4104 ($drive->{mbps_rd_max} || 0)*1024*1024,
4105 ($drive->{mbps_wr_max} || 0)*1024*1024,
4106 $drive->{iops_max} || 0,
4107 $drive->{iops_rd_max} || 0,
4108 $drive->{iops_wr_max} || 0);
4109
4110 }
4111
4112 return 1;
4113 }
4114
4115 } else { # cdrom
4116
4117 if ($drive->{file} eq 'none') {
4118 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4119 } else {
4120 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4121 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4122 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4123 }
4124
4125 return 1;
4126 }
4127 }
4128 }
4129
4130 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4131 # hotplug new disks
4132 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4133 }
4134
4135 sub vm_start {
4136 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4137
4138 lock_config($vmid, sub {
4139 my $conf = load_config($vmid, $migratedfrom);
4140
4141 die "you can't start a vm if it's a template\n" if is_template($conf);
4142
4143 check_lock($conf) if !$skiplock;
4144
4145 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4146
4147 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4148 vmconfig_apply_pending($vmid, $conf, $storecfg);
4149 $conf = load_config($vmid); # update/reload
4150 }
4151
4152 my $defaults = load_defaults();
4153
4154 # set environment variable useful inside network script
4155 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4156
4157 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4158
4159 my $migrate_port = 0;
4160 my $migrate_uri;
4161 if ($statefile) {
4162 if ($statefile eq 'tcp') {
4163 my $localip = "localhost";
4164 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4165 my $nodename = PVE::INotify::nodename();
4166 if ($datacenterconf->{migration_unsecure}) {
4167 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4168 }
4169 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4170 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
4171 $migrate_uri = "tcp:[${localip}]:${migrate_port}";
4172 push @$cmd, '-incoming', $migrate_uri;
4173 push @$cmd, '-S';
4174 } else {
4175 push @$cmd, '-loadstate', $statefile;
4176 }
4177 } elsif ($paused) {
4178 push @$cmd, '-S';
4179 }
4180
4181 # host pci devices
4182 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4183 my $d = parse_hostpci($conf->{"hostpci$i"});
4184 next if !$d;
4185 my $pcidevices = $d->{pciid};
4186 foreach my $pcidevice (@$pcidevices) {
4187 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4188
4189 my $info = pci_device_info("0000:$pciid");
4190 die "IOMMU not present\n" if !check_iommu_support();
4191 die "no pci device info for device '$pciid'\n" if !$info;
4192
4193 if ($d->{driver} && $d->{driver} eq "vfio") {
4194 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4195 } else {
4196 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4197 }
4198
4199 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4200 }
4201 }
4202
4203 PVE::Storage::activate_volumes($storecfg, $vollist);
4204
4205 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4206 umask => 0077); };
4207 my $err = $@;
4208 die "start failed: $err" if $err;
4209
4210 print "migration listens on $migrate_uri\n" if $migrate_uri;
4211
4212 if ($statefile && $statefile ne 'tcp') {
4213 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4214 warn $@ if $@;
4215 }
4216
4217 if ($migratedfrom) {
4218
4219 eval {
4220 set_migration_caps($vmid);
4221 };
4222 warn $@ if $@;
4223
4224 if ($spice_port) {
4225 print "spice listens on port $spice_port\n";
4226 if ($spice_ticket) {
4227 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4228 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4229 }
4230 }
4231
4232 } else {
4233
4234 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4235 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4236 if $conf->{balloon};
4237 }
4238
4239 foreach my $opt (keys %$conf) {
4240 next if $opt !~ m/^net\d+$/;
4241 my $nicconf = parse_net($conf->{$opt});
4242 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4243 }
4244 }
4245
4246 vm_mon_cmd_nocheck($vmid, 'qom-set',
4247 path => "machine/peripheral/balloon0",
4248 property => "guest-stats-polling-interval",
4249 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
4250
4251 });
4252 }
4253
4254 sub vm_mon_cmd {
4255 my ($vmid, $execute, %params) = @_;
4256
4257 my $cmd = { execute => $execute, arguments => \%params };
4258 vm_qmp_command($vmid, $cmd);
4259 }
4260
4261 sub vm_mon_cmd_nocheck {
4262 my ($vmid, $execute, %params) = @_;
4263
4264 my $cmd = { execute => $execute, arguments => \%params };
4265 vm_qmp_command($vmid, $cmd, 1);
4266 }
4267
4268 sub vm_qmp_command {
4269 my ($vmid, $cmd, $nocheck) = @_;
4270
4271 my $res;
4272
4273 my $timeout;
4274 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4275 $timeout = $cmd->{arguments}->{timeout};
4276 delete $cmd->{arguments}->{timeout};
4277 }
4278
4279 eval {
4280 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4281 my $sname = qmp_socket($vmid);
4282 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4283 my $qmpclient = PVE::QMPClient->new();
4284
4285 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4286 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4287 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4288 if scalar(%{$cmd->{arguments}});
4289 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4290 } else {
4291 die "unable to open monitor socket\n";
4292 }
4293 };
4294 if (my $err = $@) {
4295 syslog("err", "VM $vmid qmp command failed - $err");
4296 die $err;
4297 }
4298
4299 return $res;
4300 }
4301
4302 sub vm_human_monitor_command {
4303 my ($vmid, $cmdline) = @_;
4304
4305 my $res;
4306
4307 my $cmd = {
4308 execute => 'human-monitor-command',
4309 arguments => { 'command-line' => $cmdline},
4310 };
4311
4312 return vm_qmp_command($vmid, $cmd);
4313 }
4314
4315 sub vm_commandline {
4316 my ($storecfg, $vmid) = @_;
4317
4318 my $conf = load_config($vmid);
4319
4320 my $defaults = load_defaults();
4321
4322 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4323
4324 return join(' ', @$cmd);
4325 }
4326
4327 sub vm_reset {
4328 my ($vmid, $skiplock) = @_;
4329
4330 lock_config($vmid, sub {
4331
4332 my $conf = load_config($vmid);
4333
4334 check_lock($conf) if !$skiplock;
4335
4336 vm_mon_cmd($vmid, "system_reset");
4337 });
4338 }
4339
4340 sub get_vm_volumes {
4341 my ($conf) = @_;
4342
4343 my $vollist = [];
4344 foreach_volid($conf, sub {
4345 my ($volid, $is_cdrom) = @_;
4346
4347 return if $volid =~ m|^/|;
4348
4349 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4350 return if !$sid;
4351
4352 push @$vollist, $volid;
4353 });
4354
4355 return $vollist;
4356 }
4357
4358 sub vm_stop_cleanup {
4359 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4360
4361 eval {
4362
4363 if (!$keepActive) {
4364 my $vollist = get_vm_volumes($conf);
4365 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4366 }
4367
4368 foreach my $ext (qw(mon qmp pid vnc qga)) {
4369 unlink "/var/run/qemu-server/${vmid}.$ext";
4370 }
4371
4372 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4373 };
4374 warn $@ if $@; # avoid errors - just warn
4375 }
4376
4377 # Note: use $nockeck to skip tests if VM configuration file exists.
4378 # We need that when migration VMs to other nodes (files already moved)
4379 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4380 sub vm_stop {
4381 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4382
4383 $force = 1 if !defined($force) && !$shutdown;
4384
4385 if ($migratedfrom){
4386 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4387 kill 15, $pid if $pid;
4388 my $conf = load_config($vmid, $migratedfrom);
4389 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4390 return;
4391 }
4392
4393 lock_config($vmid, sub {
4394
4395 my $pid = check_running($vmid, $nocheck);
4396 return if !$pid;
4397
4398 my $conf;
4399 if (!$nocheck) {
4400 $conf = load_config($vmid);
4401 check_lock($conf) if !$skiplock;
4402 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4403 my $opts = parse_startup($conf->{startup});
4404 $timeout = $opts->{down} if $opts->{down};
4405 }
4406 }
4407
4408 $timeout = 60 if !defined($timeout);
4409
4410 eval {
4411 if ($shutdown) {
4412 if (defined($conf) && $conf->{agent}) {
4413 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4414 } else {
4415 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4416 }
4417 } else {
4418 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4419 }
4420 };
4421 my $err = $@;
4422
4423 if (!$err) {
4424 my $count = 0;
4425 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4426 $count++;
4427 sleep 1;
4428 }
4429
4430 if ($count >= $timeout) {
4431 if ($force) {
4432 warn "VM still running - terminating now with SIGTERM\n";
4433 kill 15, $pid;
4434 } else {
4435 die "VM quit/powerdown failed - got timeout\n";
4436 }
4437 } else {
4438 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4439 return;
4440 }
4441 } else {
4442 if ($force) {
4443 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4444 kill 15, $pid;
4445 } else {
4446 die "VM quit/powerdown failed\n";
4447 }
4448 }
4449
4450 # wait again
4451 $timeout = 10;
4452
4453 my $count = 0;
4454 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4455 $count++;
4456 sleep 1;
4457 }
4458
4459 if ($count >= $timeout) {
4460 warn "VM still running - terminating now with SIGKILL\n";
4461 kill 9, $pid;
4462 sleep 1;
4463 }
4464
4465 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4466 });
4467 }
4468
4469 sub vm_suspend {
4470 my ($vmid, $skiplock) = @_;
4471
4472 lock_config($vmid, sub {
4473
4474 my $conf = load_config($vmid);
4475
4476 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4477
4478 vm_mon_cmd($vmid, "stop");
4479 });
4480 }
4481
4482 sub vm_resume {
4483 my ($vmid, $skiplock) = @_;
4484
4485 lock_config($vmid, sub {
4486
4487 my $conf = load_config($vmid);
4488
4489 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4490
4491 vm_mon_cmd($vmid, "cont");
4492 });
4493 }
4494
4495 sub vm_sendkey {
4496 my ($vmid, $skiplock, $key) = @_;
4497
4498 lock_config($vmid, sub {
4499
4500 my $conf = load_config($vmid);
4501
4502 # there is no qmp command, so we use the human monitor command
4503 vm_human_monitor_command($vmid, "sendkey $key");
4504 });
4505 }
4506
4507 sub vm_destroy {
4508 my ($storecfg, $vmid, $skiplock) = @_;
4509
4510 lock_config($vmid, sub {
4511
4512 my $conf = load_config($vmid);
4513
4514 check_lock($conf) if !$skiplock;
4515
4516 if (!check_running($vmid)) {
4517 destroy_vm($storecfg, $vmid);
4518 } else {
4519 die "VM $vmid is running - destroy failed\n";
4520 }
4521 });
4522 }
4523
4524 # pci helpers
4525
4526 sub file_write {
4527 my ($filename, $buf) = @_;
4528
4529 my $fh = IO::File->new($filename, "w");
4530 return undef if !$fh;
4531
4532 my $res = print $fh $buf;
4533
4534 $fh->close();
4535
4536 return $res;
4537 }
4538
4539 sub pci_device_info {
4540 my ($name) = @_;
4541
4542 my $res;
4543
4544 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4545 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4546
4547 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4548 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4549
4550 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4551 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4552
4553 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4554 return undef if !defined($product) || $product !~ s/^0x//;
4555
4556 $res = {
4557 name => $name,
4558 vendor => $vendor,
4559 product => $product,
4560 domain => $domain,
4561 bus => $bus,
4562 slot => $slot,
4563 func => $func,
4564 irq => $irq,
4565 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4566 };
4567
4568 return $res;
4569 }
4570
4571 sub pci_dev_reset {
4572 my ($dev) = @_;
4573
4574 my $name = $dev->{name};
4575
4576 my $fn = "$pcisysfs/devices/$name/reset";
4577
4578 return file_write($fn, "1");
4579 }
4580
4581 sub pci_dev_bind_to_stub {
4582 my ($dev) = @_;
4583
4584 my $name = $dev->{name};
4585
4586 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4587 return 1 if -d $testdir;
4588
4589 my $data = "$dev->{vendor} $dev->{product}";
4590 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4591
4592 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4593 if (!file_write($fn, $name)) {
4594 return undef if -f $fn;
4595 }
4596
4597 $fn = "$pcisysfs/drivers/pci-stub/bind";
4598 if (! -d $testdir) {
4599 return undef if !file_write($fn, $name);
4600 }
4601
4602 return -d $testdir;
4603 }
4604
4605 sub pci_dev_bind_to_vfio {
4606 my ($dev) = @_;
4607
4608 my $name = $dev->{name};
4609
4610 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4611
4612 if (!-d $vfio_basedir) {
4613 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4614 }
4615 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4616
4617 my $testdir = "$vfio_basedir/$name";
4618 return 1 if -d $testdir;
4619
4620 my $data = "$dev->{vendor} $dev->{product}";
4621 return undef if !file_write("$vfio_basedir/new_id", $data);
4622
4623 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4624 if (!file_write($fn, $name)) {
4625 return undef if -f $fn;
4626 }
4627
4628 $fn = "$vfio_basedir/bind";
4629 if (! -d $testdir) {
4630 return undef if !file_write($fn, $name);
4631 }
4632
4633 return -d $testdir;
4634 }
4635
4636 sub pci_dev_group_bind_to_vfio {
4637 my ($pciid) = @_;
4638
4639 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4640
4641 if (!-d $vfio_basedir) {
4642 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4643 }
4644 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4645
4646 # get IOMMU group devices
4647 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4648 my @devs = grep /^0000:/, readdir($D);
4649 closedir($D);
4650
4651 foreach my $pciid (@devs) {
4652 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4653
4654 # pci bridges, switches or root ports are not supported
4655 # they have a pci_bus subdirectory so skip them
4656 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4657
4658 my $info = pci_device_info($1);
4659 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4660 }
4661
4662 return 1;
4663 }
4664
4665 sub print_pci_addr {
4666 my ($id, $bridges) = @_;
4667
4668 my $res = '';
4669 my $devices = {
4670 piix3 => { bus => 0, addr => 1 },
4671 #addr2 : first videocard
4672 balloon0 => { bus => 0, addr => 3 },
4673 watchdog => { bus => 0, addr => 4 },
4674 scsihw0 => { bus => 0, addr => 5 },
4675 'pci.3' => { bus => 0, addr => 5 }, #can also be used for virtio-scsi-single bridge
4676 scsihw1 => { bus => 0, addr => 6 },
4677 ahci0 => { bus => 0, addr => 7 },
4678 qga0 => { bus => 0, addr => 8 },
4679 spice => { bus => 0, addr => 9 },
4680 virtio0 => { bus => 0, addr => 10 },
4681 virtio1 => { bus => 0, addr => 11 },
4682 virtio2 => { bus => 0, addr => 12 },
4683 virtio3 => { bus => 0, addr => 13 },
4684 virtio4 => { bus => 0, addr => 14 },
4685 virtio5 => { bus => 0, addr => 15 },
4686 hostpci0 => { bus => 0, addr => 16 },
4687 hostpci1 => { bus => 0, addr => 17 },
4688 net0 => { bus => 0, addr => 18 },
4689 net1 => { bus => 0, addr => 19 },
4690 net2 => { bus => 0, addr => 20 },
4691 net3 => { bus => 0, addr => 21 },
4692 net4 => { bus => 0, addr => 22 },
4693 net5 => { bus => 0, addr => 23 },
4694 vga1 => { bus => 0, addr => 24 },
4695 vga2 => { bus => 0, addr => 25 },
4696 vga3 => { bus => 0, addr => 26 },
4697 hostpci2 => { bus => 0, addr => 27 },
4698 hostpci3 => { bus => 0, addr => 28 },
4699 #addr29 : usb-host (pve-usb.cfg)
4700 'pci.1' => { bus => 0, addr => 30 },
4701 'pci.2' => { bus => 0, addr => 31 },
4702 'net6' => { bus => 1, addr => 1 },
4703 'net7' => { bus => 1, addr => 2 },
4704 'net8' => { bus => 1, addr => 3 },
4705 'net9' => { bus => 1, addr => 4 },
4706 'net10' => { bus => 1, addr => 5 },
4707 'net11' => { bus => 1, addr => 6 },
4708 'net12' => { bus => 1, addr => 7 },
4709 'net13' => { bus => 1, addr => 8 },
4710 'net14' => { bus => 1, addr => 9 },
4711 'net15' => { bus => 1, addr => 10 },
4712 'net16' => { bus => 1, addr => 11 },
4713 'net17' => { bus => 1, addr => 12 },
4714 'net18' => { bus => 1, addr => 13 },
4715 'net19' => { bus => 1, addr => 14 },
4716 'net20' => { bus => 1, addr => 15 },
4717 'net21' => { bus => 1, addr => 16 },
4718 'net22' => { bus => 1, addr => 17 },
4719 'net23' => { bus => 1, addr => 18 },
4720 'net24' => { bus => 1, addr => 19 },
4721 'net25' => { bus => 1, addr => 20 },
4722 'net26' => { bus => 1, addr => 21 },
4723 'net27' => { bus => 1, addr => 22 },
4724 'net28' => { bus => 1, addr => 23 },
4725 'net29' => { bus => 1, addr => 24 },
4726 'net30' => { bus => 1, addr => 25 },
4727 'net31' => { bus => 1, addr => 26 },
4728 'virtio6' => { bus => 2, addr => 1 },
4729 'virtio7' => { bus => 2, addr => 2 },
4730 'virtio8' => { bus => 2, addr => 3 },
4731 'virtio9' => { bus => 2, addr => 4 },
4732 'virtio10' => { bus => 2, addr => 5 },
4733 'virtio11' => { bus => 2, addr => 6 },
4734 'virtio12' => { bus => 2, addr => 7 },
4735 'virtio13' => { bus => 2, addr => 8 },
4736 'virtio14' => { bus => 2, addr => 9 },
4737 'virtio15' => { bus => 2, addr => 10 },
4738 'virtioscsi0' => { bus => 3, addr => 1 },
4739 'virtioscsi1' => { bus => 3, addr => 2 },
4740 'virtioscsi2' => { bus => 3, addr => 3 },
4741 'virtioscsi3' => { bus => 3, addr => 4 },
4742 'virtioscsi4' => { bus => 3, addr => 5 },
4743 'virtioscsi5' => { bus => 3, addr => 6 },
4744 'virtioscsi6' => { bus => 3, addr => 7 },
4745 'virtioscsi7' => { bus => 3, addr => 8 },
4746 'virtioscsi8' => { bus => 3, addr => 9 },
4747 'virtioscsi9' => { bus => 3, addr => 10 },
4748 'virtioscsi10' => { bus => 3, addr => 11 },
4749 'virtioscsi11' => { bus => 3, addr => 12 },
4750 'virtioscsi12' => { bus => 3, addr => 13 },
4751 'virtioscsi13' => { bus => 3, addr => 14 },
4752 'virtioscsi14' => { bus => 3, addr => 15 },
4753 'virtioscsi15' => { bus => 3, addr => 16 },
4754 'virtioscsi16' => { bus => 3, addr => 17 },
4755 'virtioscsi17' => { bus => 3, addr => 18 },
4756 'virtioscsi18' => { bus => 3, addr => 19 },
4757 'virtioscsi19' => { bus => 3, addr => 20 },
4758 'virtioscsi20' => { bus => 3, addr => 21 },
4759 'virtioscsi21' => { bus => 3, addr => 22 },
4760 'virtioscsi22' => { bus => 3, addr => 23 },
4761 'virtioscsi23' => { bus => 3, addr => 24 },
4762 'virtioscsi24' => { bus => 3, addr => 25 },
4763 'virtioscsi25' => { bus => 3, addr => 26 },
4764 'virtioscsi26' => { bus => 3, addr => 27 },
4765 'virtioscsi27' => { bus => 3, addr => 28 },
4766 'virtioscsi28' => { bus => 3, addr => 29 },
4767 'virtioscsi29' => { bus => 3, addr => 30 },
4768 'virtioscsi30' => { bus => 3, addr => 31 },
4769
4770 };
4771
4772 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4773 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4774 my $bus = $devices->{$id}->{bus};
4775 $res = ",bus=pci.$bus,addr=$addr";
4776 $bridges->{$bus} = 1 if $bridges;
4777 }
4778 return $res;
4779
4780 }
4781
4782 sub print_pcie_addr {
4783 my ($id) = @_;
4784
4785 my $res = '';
4786 my $devices = {
4787 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4788 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4789 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4790 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4791 };
4792
4793 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4794 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4795 my $bus = $devices->{$id}->{bus};
4796 $res = ",bus=$bus,addr=$addr";
4797 }
4798 return $res;
4799
4800 }
4801
4802 # vzdump restore implementaion
4803
4804 sub tar_archive_read_firstfile {
4805 my $archive = shift;
4806
4807 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4808
4809 # try to detect archive type first
4810 my $pid = open (TMP, "tar tf '$archive'|") ||
4811 die "unable to open file '$archive'\n";
4812 my $firstfile = <TMP>;
4813 kill 15, $pid;
4814 close TMP;
4815
4816 die "ERROR: archive contaions no data\n" if !$firstfile;
4817 chomp $firstfile;
4818
4819 return $firstfile;
4820 }
4821
4822 sub tar_restore_cleanup {
4823 my ($storecfg, $statfile) = @_;
4824
4825 print STDERR "starting cleanup\n";
4826
4827 if (my $fd = IO::File->new($statfile, "r")) {
4828 while (defined(my $line = <$fd>)) {
4829 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4830 my $volid = $2;
4831 eval {
4832 if ($volid =~ m|^/|) {
4833 unlink $volid || die 'unlink failed\n';
4834 } else {
4835 PVE::Storage::vdisk_free($storecfg, $volid);
4836 }
4837 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4838 };
4839 print STDERR "unable to cleanup '$volid' - $@" if $@;
4840 } else {
4841 print STDERR "unable to parse line in statfile - $line";
4842 }
4843 }
4844 $fd->close();
4845 }
4846 }
4847
4848 sub restore_archive {
4849 my ($archive, $vmid, $user, $opts) = @_;
4850
4851 my $format = $opts->{format};
4852 my $comp;
4853
4854 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4855 $format = 'tar' if !$format;
4856 $comp = 'gzip';
4857 } elsif ($archive =~ m/\.tar$/) {
4858 $format = 'tar' if !$format;
4859 } elsif ($archive =~ m/.tar.lzo$/) {
4860 $format = 'tar' if !$format;
4861 $comp = 'lzop';
4862 } elsif ($archive =~ m/\.vma$/) {
4863 $format = 'vma' if !$format;
4864 } elsif ($archive =~ m/\.vma\.gz$/) {
4865 $format = 'vma' if !$format;
4866 $comp = 'gzip';
4867 } elsif ($archive =~ m/\.vma\.lzo$/) {
4868 $format = 'vma' if !$format;
4869 $comp = 'lzop';
4870 } else {
4871 $format = 'vma' if !$format; # default
4872 }
4873
4874 # try to detect archive format
4875 if ($format eq 'tar') {
4876 return restore_tar_archive($archive, $vmid, $user, $opts);
4877 } else {
4878 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4879 }
4880 }
4881
4882 sub restore_update_config_line {
4883 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4884
4885 return if $line =~ m/^\#qmdump\#/;
4886 return if $line =~ m/^\#vzdump\#/;
4887 return if $line =~ m/^lock:/;
4888 return if $line =~ m/^unused\d+:/;
4889 return if $line =~ m/^parent:/;
4890 return if $line =~ m/^template:/; # restored VM is never a template
4891
4892 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4893 # try to convert old 1.X settings
4894 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4895 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4896 my ($model, $macaddr) = split(/\=/, $devconfig);
4897 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4898 my $net = {
4899 model => $model,
4900 bridge => "vmbr$ind",
4901 macaddr => $macaddr,
4902 };
4903 my $netstr = print_net($net);
4904
4905 print $outfd "net$cookie->{netcount}: $netstr\n";
4906 $cookie->{netcount}++;
4907 }
4908 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4909 my ($id, $netstr) = ($1, $2);
4910 my $net = parse_net($netstr);
4911 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4912 $netstr = print_net($net);
4913 print $outfd "$id: $netstr\n";
4914 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4915 my $virtdev = $1;
4916 my $value = $3;
4917 if ($line =~ m/backup=no/) {
4918 print $outfd "#$line";
4919 } elsif ($virtdev && $map->{$virtdev}) {
4920 my $di = parse_drive($virtdev, $value);
4921 delete $di->{format}; # format can change on restore
4922 $di->{file} = $map->{$virtdev};
4923 $value = print_drive($vmid, $di);
4924 print $outfd "$virtdev: $value\n";
4925 } else {
4926 print $outfd $line;
4927 }
4928 } else {
4929 print $outfd $line;
4930 }
4931 }
4932
4933 sub scan_volids {
4934 my ($cfg, $vmid) = @_;
4935
4936 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4937
4938 my $volid_hash = {};
4939 foreach my $storeid (keys %$info) {
4940 foreach my $item (@{$info->{$storeid}}) {
4941 next if !($item->{volid} && $item->{size});
4942 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4943 $volid_hash->{$item->{volid}} = $item;
4944 }
4945 }
4946
4947 return $volid_hash;
4948 }
4949
4950 sub get_used_paths {
4951 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4952
4953 my $used_path = {};
4954
4955 my $scan_config = sub {
4956 my ($cref, $snapname) = @_;
4957
4958 foreach my $key (keys %$cref) {
4959 my $value = $cref->{$key};
4960 if (valid_drivename($key)) {
4961 next if $skip_drive && $key eq $skip_drive;
4962 my $drive = parse_drive($key, $value);
4963 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4964 if ($drive->{file} =~ m!^/!) {
4965 $used_path->{$drive->{file}}++; # = 1;
4966 } else {
4967 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4968 next if !$storeid;
4969 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4970 next if !$scfg;
4971 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4972 $used_path->{$path}++; # = 1;
4973 }
4974 }
4975 }
4976 };
4977
4978 &$scan_config($conf);
4979
4980 undef $skip_drive;
4981
4982 if ($scan_snapshots) {
4983 foreach my $snapname (keys %{$conf->{snapshots}}) {
4984 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4985 }
4986 }
4987
4988 return $used_path;
4989 }
4990
4991 sub update_disksize {
4992 my ($vmid, $conf, $volid_hash) = @_;
4993
4994 my $changes;
4995
4996 my $used = {};
4997
4998 # Note: it is allowed to define multiple storages with same path (alias), so
4999 # we need to check both 'volid' and real 'path' (two different volid can point
5000 # to the same path).
5001
5002 my $usedpath = {};
5003
5004 # update size info
5005 foreach my $opt (keys %$conf) {
5006 if (valid_drivename($opt)) {
5007 my $drive = parse_drive($opt, $conf->{$opt});
5008 my $volid = $drive->{file};
5009 next if !$volid;
5010
5011 $used->{$volid} = 1;
5012 if ($volid_hash->{$volid} &&
5013 (my $path = $volid_hash->{$volid}->{path})) {
5014 $usedpath->{$path} = 1;
5015 }
5016
5017 next if drive_is_cdrom($drive);
5018 next if !$volid_hash->{$volid};
5019
5020 $drive->{size} = $volid_hash->{$volid}->{size};
5021 my $new = print_drive($vmid, $drive);
5022 if ($new ne $conf->{$opt}) {
5023 $changes = 1;
5024 $conf->{$opt} = $new;
5025 }
5026 }
5027 }
5028
5029 # remove 'unusedX' entry if volume is used
5030 foreach my $opt (keys %$conf) {
5031 next if $opt !~ m/^unused\d+$/;
5032 my $volid = $conf->{$opt};
5033 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5034 if ($used->{$volid} || ($path && $usedpath->{$path})) {
5035 $changes = 1;
5036 delete $conf->{$opt};
5037 }
5038 }
5039
5040 foreach my $volid (sort keys %$volid_hash) {
5041 next if $volid =~ m/vm-$vmid-state-/;
5042 next if $used->{$volid};
5043 my $path = $volid_hash->{$volid}->{path};
5044 next if !$path; # just to be sure
5045 next if $usedpath->{$path};
5046 $changes = 1;
5047 add_unused_volume($conf, $volid);
5048 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
5049 }
5050
5051 return $changes;
5052 }
5053
5054 sub rescan {
5055 my ($vmid, $nolock) = @_;
5056
5057 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
5058
5059 my $volid_hash = scan_volids($cfg, $vmid);
5060
5061 my $updatefn = sub {
5062 my ($vmid) = @_;
5063
5064 my $conf = load_config($vmid);
5065
5066 check_lock($conf);
5067
5068 my $vm_volids = {};
5069 foreach my $volid (keys %$volid_hash) {
5070 my $info = $volid_hash->{$volid};
5071 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5072 }
5073
5074 my $changes = update_disksize($vmid, $conf, $vm_volids);
5075
5076 update_config_nolock($vmid, $conf, 1) if $changes;
5077 };
5078
5079 if (defined($vmid)) {
5080 if ($nolock) {
5081 &$updatefn($vmid);
5082 } else {
5083 lock_config($vmid, $updatefn, $vmid);
5084 }
5085 } else {
5086 my $vmlist = config_list();
5087 foreach my $vmid (keys %$vmlist) {
5088 if ($nolock) {
5089 &$updatefn($vmid);
5090 } else {
5091 lock_config($vmid, $updatefn, $vmid);
5092 }
5093 }
5094 }
5095 }
5096
5097 sub restore_vma_archive {
5098 my ($archive, $vmid, $user, $opts, $comp) = @_;
5099
5100 my $input = $archive eq '-' ? "<&STDIN" : undef;
5101 my $readfrom = $archive;
5102
5103 my $uncomp = '';
5104 if ($comp) {
5105 $readfrom = '-';
5106 my $qarchive = PVE::Tools::shellquote($archive);
5107 if ($comp eq 'gzip') {
5108 $uncomp = "zcat $qarchive|";
5109 } elsif ($comp eq 'lzop') {
5110 $uncomp = "lzop -d -c $qarchive|";
5111 } else {
5112 die "unknown compression method '$comp'\n";
5113 }
5114
5115 }
5116
5117 my $tmpdir = "/var/tmp/vzdumptmp$$";
5118 rmtree $tmpdir;
5119
5120 # disable interrupts (always do cleanups)
5121 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5122 warn "got interrupt - ignored\n";
5123 };
5124
5125 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5126 POSIX::mkfifo($mapfifo, 0600);
5127 my $fifofh;
5128
5129 my $openfifo = sub {
5130 open($fifofh, '>', $mapfifo) || die $!;
5131 };
5132
5133 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5134
5135 my $oldtimeout;
5136 my $timeout = 5;
5137
5138 my $devinfo = {};
5139
5140 my $rpcenv = PVE::RPCEnvironment::get();
5141
5142 my $conffile = config_file($vmid);
5143 my $tmpfn = "$conffile.$$.tmp";
5144
5145 # Note: $oldconf is undef if VM does not exists
5146 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5147
5148 my $print_devmap = sub {
5149 my $virtdev_hash = {};
5150
5151 my $cfgfn = "$tmpdir/qemu-server.conf";
5152
5153 # we can read the config - that is already extracted
5154 my $fh = IO::File->new($cfgfn, "r") ||
5155 "unable to read qemu-server.conf - $!\n";
5156
5157 while (defined(my $line = <$fh>)) {
5158 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5159 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5160 die "archive does not contain data for drive '$virtdev'\n"
5161 if !$devinfo->{$devname};
5162 if (defined($opts->{storage})) {
5163 $storeid = $opts->{storage} || 'local';
5164 } elsif (!$storeid) {
5165 $storeid = 'local';
5166 }
5167 $format = 'raw' if !$format;
5168 $devinfo->{$devname}->{devname} = $devname;
5169 $devinfo->{$devname}->{virtdev} = $virtdev;
5170 $devinfo->{$devname}->{format} = $format;
5171 $devinfo->{$devname}->{storeid} = $storeid;
5172
5173 # check permission on storage
5174 my $pool = $opts->{pool}; # todo: do we need that?
5175 if ($user ne 'root@pam') {
5176 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5177 }
5178
5179 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5180 }
5181 }
5182
5183 foreach my $devname (keys %$devinfo) {
5184 die "found no device mapping information for device '$devname'\n"
5185 if !$devinfo->{$devname}->{virtdev};
5186 }
5187
5188 my $cfg = cfs_read_file('storage.cfg');
5189
5190 # create empty/temp config
5191 if ($oldconf) {
5192 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5193 foreach_drive($oldconf, sub {
5194 my ($ds, $drive) = @_;
5195
5196 return if drive_is_cdrom($drive);
5197
5198 my $volid = $drive->{file};
5199
5200 return if !$volid || $volid =~ m|^/|;
5201
5202 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5203 return if !$path || !$owner || ($owner != $vmid);
5204
5205 # Note: only delete disk we want to restore
5206 # other volumes will become unused
5207 if ($virtdev_hash->{$ds}) {
5208 PVE::Storage::vdisk_free($cfg, $volid);
5209 }
5210 });
5211 }
5212
5213 my $map = {};
5214 foreach my $virtdev (sort keys %$virtdev_hash) {
5215 my $d = $virtdev_hash->{$virtdev};
5216 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5217 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5218
5219 # test if requested format is supported
5220 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5221 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5222 $d->{format} = $defFormat if !$supported;
5223
5224 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5225 $d->{format}, undef, $alloc_size);
5226 print STDERR "new volume ID is '$volid'\n";
5227 $d->{volid} = $volid;
5228 my $path = PVE::Storage::path($cfg, $volid);
5229
5230 my $write_zeros = 1;
5231 # fixme: what other storages types initialize volumes with zero?
5232 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5233 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5234 $write_zeros = 0;
5235 }
5236
5237 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5238
5239 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5240 $map->{$virtdev} = $volid;
5241 }
5242
5243 $fh->seek(0, 0) || die "seek failed - $!\n";
5244
5245 my $outfd = new IO::File ($tmpfn, "w") ||
5246 die "unable to write config for VM $vmid\n";
5247
5248 my $cookie = { netcount => 0 };
5249 while (defined(my $line = <$fh>)) {
5250 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5251 }
5252
5253 $fh->close();
5254 $outfd->close();
5255 };
5256
5257 eval {
5258 # enable interrupts
5259 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5260 die "interrupted by signal\n";
5261 };
5262 local $SIG{ALRM} = sub { die "got timeout\n"; };
5263
5264 $oldtimeout = alarm($timeout);
5265
5266 my $parser = sub {
5267 my $line = shift;
5268
5269 print "$line\n";
5270
5271 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5272 my ($dev_id, $size, $devname) = ($1, $2, $3);
5273 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5274 } elsif ($line =~ m/^CTIME: /) {
5275 # we correctly received the vma config, so we can disable
5276 # the timeout now for disk allocation (set to 10 minutes, so
5277 # that we always timeout if something goes wrong)
5278 alarm(600);
5279 &$print_devmap();
5280 print $fifofh "done\n";
5281 my $tmp = $oldtimeout || 0;
5282 $oldtimeout = undef;
5283 alarm($tmp);
5284 close($fifofh);
5285 }
5286 };
5287
5288 print "restore vma archive: $cmd\n";
5289 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5290 };
5291 my $err = $@;
5292
5293 alarm($oldtimeout) if $oldtimeout;
5294
5295 unlink $mapfifo;
5296
5297 if ($err) {
5298 rmtree $tmpdir;
5299 unlink $tmpfn;
5300
5301 my $cfg = cfs_read_file('storage.cfg');
5302 foreach my $devname (keys %$devinfo) {
5303 my $volid = $devinfo->{$devname}->{volid};
5304 next if !$volid;
5305 eval {
5306 if ($volid =~ m|^/|) {
5307 unlink $volid || die 'unlink failed\n';
5308 } else {
5309 PVE::Storage::vdisk_free($cfg, $volid);
5310 }
5311 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5312 };
5313 print STDERR "unable to cleanup '$volid' - $@" if $@;
5314 }
5315 die $err;
5316 }
5317
5318 rmtree $tmpdir;
5319
5320 rename($tmpfn, $conffile) ||
5321 die "unable to commit configuration file '$conffile'\n";
5322
5323 PVE::Cluster::cfs_update(); # make sure we read new file
5324
5325 eval { rescan($vmid, 1); };
5326 warn $@ if $@;
5327 }
5328
5329 sub restore_tar_archive {
5330 my ($archive, $vmid, $user, $opts) = @_;
5331
5332 if ($archive ne '-') {
5333 my $firstfile = tar_archive_read_firstfile($archive);
5334 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5335 if $firstfile ne 'qemu-server.conf';
5336 }
5337
5338 my $storecfg = cfs_read_file('storage.cfg');
5339
5340 # destroy existing data - keep empty config
5341 my $vmcfgfn = config_file($vmid);
5342 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5343
5344 my $tocmd = "/usr/lib/qemu-server/qmextract";
5345
5346 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5347 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5348 $tocmd .= ' --prealloc' if $opts->{prealloc};
5349 $tocmd .= ' --info' if $opts->{info};
5350
5351 # tar option "xf" does not autodetect compression when read from STDIN,
5352 # so we pipe to zcat
5353 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5354 PVE::Tools::shellquote("--to-command=$tocmd");
5355
5356 my $tmpdir = "/var/tmp/vzdumptmp$$";
5357 mkpath $tmpdir;
5358
5359 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5360 local $ENV{VZDUMP_VMID} = $vmid;
5361 local $ENV{VZDUMP_USER} = $user;
5362
5363 my $conffile = config_file($vmid);
5364 my $tmpfn = "$conffile.$$.tmp";
5365
5366 # disable interrupts (always do cleanups)
5367 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5368 print STDERR "got interrupt - ignored\n";
5369 };
5370
5371 eval {
5372 # enable interrupts
5373 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5374 die "interrupted by signal\n";
5375 };
5376
5377 if ($archive eq '-') {
5378 print "extracting archive from STDIN\n";
5379 run_command($cmd, input => "<&STDIN");
5380 } else {
5381 print "extracting archive '$archive'\n";
5382 run_command($cmd);
5383 }
5384
5385 return if $opts->{info};
5386
5387 # read new mapping
5388 my $map = {};
5389 my $statfile = "$tmpdir/qmrestore.stat";
5390 if (my $fd = IO::File->new($statfile, "r")) {
5391 while (defined (my $line = <$fd>)) {
5392 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5393 $map->{$1} = $2 if $1;
5394 } else {
5395 print STDERR "unable to parse line in statfile - $line\n";
5396 }
5397 }
5398 $fd->close();
5399 }
5400
5401 my $confsrc = "$tmpdir/qemu-server.conf";
5402
5403 my $srcfd = new IO::File($confsrc, "r") ||
5404 die "unable to open file '$confsrc'\n";
5405
5406 my $outfd = new IO::File ($tmpfn, "w") ||
5407 die "unable to write config for VM $vmid\n";
5408
5409 my $cookie = { netcount => 0 };
5410 while (defined (my $line = <$srcfd>)) {
5411 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5412 }
5413
5414 $srcfd->close();
5415 $outfd->close();
5416 };
5417 my $err = $@;
5418
5419 if ($err) {
5420
5421 unlink $tmpfn;
5422
5423 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5424
5425 die $err;
5426 }
5427
5428 rmtree $tmpdir;
5429
5430 rename $tmpfn, $conffile ||
5431 die "unable to commit configuration file '$conffile'\n";
5432
5433 PVE::Cluster::cfs_update(); # make sure we read new file
5434
5435 eval { rescan($vmid, 1); };
5436 warn $@ if $@;
5437 };
5438
5439
5440 # Internal snapshots
5441
5442 # NOTE: Snapshot create/delete involves several non-atomic
5443 # action, and can take a long time.
5444 # So we try to avoid locking the file and use 'lock' variable
5445 # inside the config file instead.
5446
5447 my $snapshot_copy_config = sub {
5448 my ($source, $dest) = @_;
5449
5450 foreach my $k (keys %$source) {
5451 next if $k eq 'snapshots';
5452 next if $k eq 'snapstate';
5453 next if $k eq 'snaptime';
5454 next if $k eq 'vmstate';
5455 next if $k eq 'lock';
5456 next if $k eq 'digest';
5457 next if $k eq 'description';
5458 next if $k =~ m/^unused\d+$/;
5459
5460 $dest->{$k} = $source->{$k};
5461 }
5462 };
5463
5464 my $snapshot_apply_config = sub {
5465 my ($conf, $snap) = @_;
5466
5467 # copy snapshot list
5468 my $newconf = {
5469 snapshots => $conf->{snapshots},
5470 };
5471
5472 # keep description and list of unused disks
5473 foreach my $k (keys %$conf) {
5474 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5475 $newconf->{$k} = $conf->{$k};
5476 }
5477
5478 &$snapshot_copy_config($snap, $newconf);
5479
5480 return $newconf;
5481 };
5482
5483 sub foreach_writable_storage {
5484 my ($conf, $func) = @_;
5485
5486 my $sidhash = {};
5487
5488 foreach my $ds (keys %$conf) {
5489 next if !valid_drivename($ds);
5490
5491 my $drive = parse_drive($ds, $conf->{$ds});
5492 next if !$drive;
5493 next if drive_is_cdrom($drive);
5494
5495 my $volid = $drive->{file};
5496
5497 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5498 $sidhash->{$sid} = $sid if $sid;
5499 }
5500
5501 foreach my $sid (sort keys %$sidhash) {
5502 &$func($sid);
5503 }
5504 }
5505
5506 my $alloc_vmstate_volid = sub {
5507 my ($storecfg, $vmid, $conf, $snapname) = @_;
5508
5509 # Note: we try to be smart when selecting a $target storage
5510
5511 my $target;
5512
5513 # search shared storage first
5514 foreach_writable_storage($conf, sub {
5515 my ($sid) = @_;
5516 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5517 return if !$scfg->{shared};
5518
5519 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5520 });
5521
5522 if (!$target) {
5523 # now search local storage
5524 foreach_writable_storage($conf, sub {
5525 my ($sid) = @_;
5526 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5527 return if $scfg->{shared};
5528
5529 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5530 });
5531 }
5532
5533 $target = 'local' if !$target;
5534
5535 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5536 # we abort live save after $conf->{memory}, so we need at max twice that space
5537 my $size = $conf->{memory}*2 + $driver_state_size;
5538
5539 my $name = "vm-$vmid-state-$snapname";
5540 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5541 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5542 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5543
5544 return $volid;
5545 };
5546
5547 my $snapshot_prepare = sub {
5548 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5549
5550 my $snap;
5551
5552 my $updatefn = sub {
5553
5554 my $conf = load_config($vmid);
5555
5556 die "you can't take a snapshot if it's a template\n"
5557 if is_template($conf);
5558
5559 check_lock($conf);
5560
5561 $conf->{lock} = 'snapshot';
5562
5563 die "snapshot name '$snapname' already used\n"
5564 if defined($conf->{snapshots}->{$snapname});
5565
5566 my $storecfg = PVE::Storage::config();
5567 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5568
5569 $snap = $conf->{snapshots}->{$snapname} = {};
5570
5571 if ($save_vmstate && check_running($vmid)) {
5572 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5573 }
5574
5575 &$snapshot_copy_config($conf, $snap);
5576
5577 $snap->{snapstate} = "prepare";
5578 $snap->{snaptime} = time();
5579 $snap->{description} = $comment if $comment;
5580
5581 # always overwrite machine if we save vmstate. This makes sure we
5582 # can restore it later using correct machine type
5583 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5584
5585 update_config_nolock($vmid, $conf, 1);
5586 };
5587
5588 lock_config($vmid, $updatefn);
5589
5590 return $snap;
5591 };
5592
5593 my $snapshot_commit = sub {
5594 my ($vmid, $snapname) = @_;
5595
5596 my $updatefn = sub {
5597
5598 my $conf = load_config($vmid);
5599
5600 die "missing snapshot lock\n"
5601 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5602
5603 my $has_machine_config = defined($conf->{machine});
5604
5605 my $snap = $conf->{snapshots}->{$snapname};
5606
5607 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5608
5609 die "wrong snapshot state\n"
5610 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5611
5612 delete $snap->{snapstate};
5613 delete $conf->{lock};
5614
5615 my $newconf = &$snapshot_apply_config($conf, $snap);
5616
5617 delete $newconf->{machine} if !$has_machine_config;
5618
5619 $newconf->{parent} = $snapname;
5620
5621 update_config_nolock($vmid, $newconf, 1);
5622 };
5623
5624 lock_config($vmid, $updatefn);
5625 };
5626
5627 sub snapshot_rollback {
5628 my ($vmid, $snapname) = @_;
5629
5630 my $prepare = 1;
5631
5632 my $storecfg = PVE::Storage::config();
5633
5634 my $conf = load_config($vmid);
5635
5636 my $get_snapshot_config = sub {
5637
5638 die "you can't rollback if vm is a template\n" if is_template($conf);
5639
5640 my $res = $conf->{snapshots}->{$snapname};
5641
5642 die "snapshot '$snapname' does not exist\n" if !defined($res);
5643
5644 return $res;
5645 };
5646
5647 my $snap = &$get_snapshot_config();
5648
5649 foreach_drive($snap, sub {
5650 my ($ds, $drive) = @_;
5651
5652 return if drive_is_cdrom($drive);
5653
5654 my $volid = $drive->{file};
5655
5656 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5657 });
5658
5659 my $updatefn = sub {
5660
5661 $conf = load_config($vmid);
5662
5663 $snap = &$get_snapshot_config();
5664
5665 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5666 if $snap->{snapstate};
5667
5668 if ($prepare) {
5669 check_lock($conf);
5670 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5671 }
5672
5673 die "unable to rollback vm $vmid: vm is running\n"
5674 if check_running($vmid);
5675
5676 if ($prepare) {
5677 $conf->{lock} = 'rollback';
5678 } else {
5679 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5680 delete $conf->{lock};
5681 }
5682
5683 my $forcemachine;
5684
5685 if (!$prepare) {
5686 my $has_machine_config = defined($conf->{machine});
5687
5688 # copy snapshot config to current config
5689 $conf = &$snapshot_apply_config($conf, $snap);
5690 $conf->{parent} = $snapname;
5691
5692 # Note: old code did not store 'machine', so we try to be smart
5693 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5694 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5695 # we remove the 'machine' configuration if not explicitly specified
5696 # in the original config.
5697 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5698 }
5699
5700 update_config_nolock($vmid, $conf, 1);
5701
5702 if (!$prepare && $snap->{vmstate}) {
5703 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5704 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5705 }
5706 };
5707
5708 lock_config($vmid, $updatefn);
5709
5710 foreach_drive($snap, sub {
5711 my ($ds, $drive) = @_;
5712
5713 return if drive_is_cdrom($drive);
5714
5715 my $volid = $drive->{file};
5716 my $device = "drive-$ds";
5717
5718 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5719 });
5720
5721 $prepare = 0;
5722 lock_config($vmid, $updatefn);
5723 }
5724
5725 my $savevm_wait = sub {
5726 my ($vmid) = @_;
5727
5728 for(;;) {
5729 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5730 if (!$stat->{status}) {
5731 die "savevm not active\n";
5732 } elsif ($stat->{status} eq 'active') {
5733 sleep(1);
5734 next;
5735 } elsif ($stat->{status} eq 'completed') {
5736 last;
5737 } else {
5738 die "query-savevm returned status '$stat->{status}'\n";
5739 }
5740 }
5741 };
5742
5743 sub do_snapshots_with_qemu {
5744 my ($storecfg, $volid) = @_;
5745
5746 my $storage_name = PVE::Storage::parse_volume_id($volid);
5747
5748 if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}} ){
5749 return 1;
5750 }
5751
5752 if ($volid =~ m/\.(qcow2|qed)$/){
5753 return 1;
5754 }
5755
5756 return undef;
5757 }
5758
5759 sub snapshot_create {
5760 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5761
5762 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5763
5764 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5765
5766 my $config = load_config($vmid);
5767
5768 my $running = check_running($vmid);
5769
5770 my $freezefs = $running && $config->{agent};
5771 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5772
5773 my $drivehash = {};
5774
5775 if ($freezefs) {
5776 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5777 warn "guest-fsfreeze-freeze problems - $@" if $@;
5778 }
5779
5780 eval {
5781 # create internal snapshots of all drives
5782
5783 my $storecfg = PVE::Storage::config();
5784
5785 if ($running) {
5786 if ($snap->{vmstate}) {
5787 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5788 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5789 &$savevm_wait($vmid);
5790 } else {
5791 vm_mon_cmd($vmid, "savevm-start");
5792 }
5793 };
5794
5795 foreach_drive($snap, sub {
5796 my ($ds, $drive) = @_;
5797
5798 return if drive_is_cdrom($drive);
5799
5800 my $volid = $drive->{file};
5801 my $device = "drive-$ds";
5802
5803 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5804 $drivehash->{$ds} = 1;
5805 });
5806 };
5807 my $err = $@;
5808
5809 if ($running) {
5810 eval { vm_mon_cmd($vmid, "savevm-end") };
5811 warn $@ if $@;
5812
5813 if ($freezefs) {
5814 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5815 warn "guest-fsfreeze-thaw problems - $@" if $@;
5816 }
5817
5818 # savevm-end is async, we need to wait
5819 for (;;) {
5820 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5821 if (!$stat->{bytes}) {
5822 last;
5823 } else {
5824 print "savevm not yet finished\n";
5825 sleep(1);
5826 next;
5827 }
5828 }
5829 }
5830
5831 if ($err) {
5832 warn "snapshot create failed: starting cleanup\n";
5833 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5834 warn $@ if $@;
5835 die $err;
5836 }
5837
5838 &$snapshot_commit($vmid, $snapname);
5839 }
5840
5841 # Note: $drivehash is only set when called from snapshot_create.
5842 sub snapshot_delete {
5843 my ($vmid, $snapname, $force, $drivehash) = @_;
5844
5845 my $prepare = 1;
5846
5847 my $snap;
5848 my $unused = [];
5849
5850 my $unlink_parent = sub {
5851 my ($confref, $new_parent) = @_;
5852
5853 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5854 if ($new_parent) {
5855 $confref->{parent} = $new_parent;
5856 } else {
5857 delete $confref->{parent};
5858 }
5859 }
5860 };
5861
5862 my $updatefn = sub {
5863 my ($remove_drive) = @_;
5864
5865 my $conf = load_config($vmid);
5866
5867 if (!$drivehash) {
5868 check_lock($conf);
5869 die "you can't delete a snapshot if vm is a template\n"
5870 if is_template($conf);
5871 }
5872
5873 $snap = $conf->{snapshots}->{$snapname};
5874
5875 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5876
5877 # remove parent refs
5878 if (!$prepare) {
5879 &$unlink_parent($conf, $snap->{parent});
5880 foreach my $sn (keys %{$conf->{snapshots}}) {
5881 next if $sn eq $snapname;
5882 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5883 }
5884 }
5885
5886 if ($remove_drive) {
5887 if ($remove_drive eq 'vmstate') {
5888 delete $snap->{$remove_drive};
5889 } else {
5890 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5891 my $volid = $drive->{file};
5892 delete $snap->{$remove_drive};
5893 add_unused_volume($conf, $volid);
5894 }
5895 }
5896
5897 if ($prepare) {
5898 $snap->{snapstate} = 'delete';
5899 } else {
5900 delete $conf->{snapshots}->{$snapname};
5901 delete $conf->{lock} if $drivehash;
5902 foreach my $volid (@$unused) {
5903 add_unused_volume($conf, $volid);
5904 }
5905 }
5906
5907 update_config_nolock($vmid, $conf, 1);
5908 };
5909
5910 lock_config($vmid, $updatefn);
5911
5912 # now remove vmstate file
5913
5914 my $storecfg = PVE::Storage::config();
5915
5916 if ($snap->{vmstate}) {
5917 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5918 if (my $err = $@) {
5919 die $err if !$force;
5920 warn $err;
5921 }
5922 # save changes (remove vmstate from snapshot)
5923 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5924 };
5925
5926 # now remove all internal snapshots
5927 foreach_drive($snap, sub {
5928 my ($ds, $drive) = @_;
5929
5930 return if drive_is_cdrom($drive);
5931
5932 my $volid = $drive->{file};
5933 my $device = "drive-$ds";
5934
5935 if (!$drivehash || $drivehash->{$ds}) {
5936 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5937 if (my $err = $@) {
5938 die $err if !$force;
5939 warn $err;
5940 }
5941 }
5942
5943 # save changes (remove drive fron snapshot)
5944 lock_config($vmid, $updatefn, $ds) if !$force;
5945 push @$unused, $volid;
5946 });
5947
5948 # now cleanup config
5949 $prepare = 0;
5950 lock_config($vmid, $updatefn);
5951 }
5952
5953 sub has_feature {
5954 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5955
5956 my $err;
5957 foreach_drive($conf, sub {
5958 my ($ds, $drive) = @_;
5959
5960 return if drive_is_cdrom($drive);
5961 my $volid = $drive->{file};
5962 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5963 });
5964
5965 return $err ? 0 : 1;
5966 }
5967
5968 sub template_create {
5969 my ($vmid, $conf, $disk) = @_;
5970
5971 my $storecfg = PVE::Storage::config();
5972
5973 foreach_drive($conf, sub {
5974 my ($ds, $drive) = @_;
5975
5976 return if drive_is_cdrom($drive);
5977 return if $disk && $ds ne $disk;
5978
5979 my $volid = $drive->{file};
5980 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5981
5982 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5983 $drive->{file} = $voliddst;
5984 $conf->{$ds} = print_drive($vmid, $drive);
5985 update_config_nolock($vmid, $conf, 1);
5986 });
5987 }
5988
5989 sub is_template {
5990 my ($conf) = @_;
5991
5992 return 1 if defined $conf->{template} && $conf->{template} == 1;
5993 }
5994
5995 sub qemu_img_convert {
5996 my ($src_volid, $dst_volid, $size, $snapname) = @_;
5997
5998 my $storecfg = PVE::Storage::config();
5999 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6000 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6001
6002 if ($src_storeid && $dst_storeid) {
6003 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6004 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6005
6006 my $src_format = qemu_img_format($src_scfg, $src_volname);
6007 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6008
6009 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6010 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6011
6012 my $cmd = [];
6013 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
6014 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
6015 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
6016
6017 my $parser = sub {
6018 my $line = shift;
6019 if($line =~ m/\((\S+)\/100\%\)/){
6020 my $percent = $1;
6021 my $transferred = int($size * $percent / 100);
6022 my $remaining = $size - $transferred;
6023
6024 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6025 }
6026
6027 };
6028
6029 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6030 my $err = $@;
6031 die "copy failed: $err" if $err;
6032 }
6033 }
6034
6035 sub qemu_img_format {
6036 my ($scfg, $volname) = @_;
6037
6038 if ($scfg->{path} && $volname =~ m/\.(raw|cow|qcow|qcow2|qed|vmdk|cloop)$/) {
6039 return $1;
6040 } else {
6041 return "raw";
6042 }
6043 }
6044
6045 sub qemu_drive_mirror {
6046 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
6047
6048 my $storecfg = PVE::Storage::config();
6049 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6050
6051 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6052
6053 my $format = qemu_img_format($dst_scfg, $dst_volname);
6054
6055 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6056
6057 #drive-mirror is doing lseek on source image before starting, and this can take a lot of time for big nfs volume
6058 #during this time, qmp socket is hanging
6059 #http://lists.nongnu.org/archive/html/qemu-devel/2015-05/msg01838.html
6060 #so we need to setup a big timeout
6061 my $opts = { timeout => 14400, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
6062 $opts->{format} = $format if $format;
6063
6064 print "drive mirror is starting : this step can take some minutes/hours, depend of disk size and storage speed\n";
6065
6066 vm_mon_cmd($vmid, "drive-mirror", %$opts);
6067 eval {
6068 while (1) {
6069 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6070 my $stat = @$stats[0];
6071 die "mirroring job seem to have die. Maybe do you have bad sectors?" if !$stat;
6072 die "error job is not mirroring" if $stat->{type} ne "mirror";
6073
6074 my $busy = $stat->{busy};
6075 my $ready = $stat->{ready};
6076
6077 if (my $total = $stat->{len}) {
6078 my $transferred = $stat->{offset} || 0;
6079 my $remaining = $total - $transferred;
6080 my $percent = sprintf "%.2f", ($transferred * 100 / $total);
6081
6082 print "transferred: $transferred bytes remaining: $remaining bytes total: $total bytes progression: $percent % busy: $busy ready: $ready \n";
6083 }
6084
6085
6086 if ($stat->{ready} eq 'true') {
6087
6088 last if $vmiddst != $vmid;
6089
6090 # try to switch the disk if source and destination are on the same guest
6091 eval { vm_mon_cmd($vmid, "block-job-complete", device => "drive-$drive") };
6092 last if !$@;
6093 die $@ if $@ !~ m/cannot be completed/;
6094 }
6095 sleep 1;
6096 }
6097
6098
6099 };
6100 my $err = $@;
6101
6102 my $cancel_job = sub {
6103 vm_mon_cmd($vmid, "block-job-cancel", device => "drive-$drive");
6104 while (1) {
6105 my $stats = vm_mon_cmd($vmid, "query-block-jobs");
6106 my $stat = @$stats[0];
6107 last if !$stat;
6108 sleep 1;
6109 }
6110 };
6111
6112 if ($err) {
6113 eval { &$cancel_job(); };
6114 die "mirroring error: $err";
6115 }
6116
6117 if ($vmiddst != $vmid) {
6118 # if we clone a disk for a new target vm, we don't switch the disk
6119 &$cancel_job(); # so we call block-job-cancel
6120 }
6121 }
6122
6123 sub clone_disk {
6124 my ($storecfg, $vmid, $running, $drivename, $drive, $snapname,
6125 $newvmid, $storage, $format, $full, $newvollist) = @_;
6126
6127 my $newvolid;
6128
6129 if (!$full) {
6130 print "create linked clone of drive $drivename ($drive->{file})\n";
6131 $newvolid = PVE::Storage::vdisk_clone($storecfg, $drive->{file}, $newvmid, $snapname);
6132 push @$newvollist, $newvolid;
6133 } else {
6134 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file});
6135 $storeid = $storage if $storage;
6136
6137 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($storecfg, $storeid);
6138 if (!$format) {
6139 my $scfg = PVE::Storage::storage_config($storecfg, $storeid);
6140 $format = qemu_img_format($scfg, $volname);
6141 }
6142
6143 # test if requested format is supported - else use default
6144 my $supported = grep { $_ eq $format } @$validFormats;
6145 $format = $defFormat if !$supported;
6146
6147 my ($size) = PVE::Storage::volume_size_info($storecfg, $drive->{file}, 3);
6148
6149 print "create full clone of drive $drivename ($drive->{file})\n";
6150 $newvolid = PVE::Storage::vdisk_alloc($storecfg, $storeid, $newvmid, $format, undef, ($size/1024));
6151 push @$newvollist, $newvolid;
6152
6153 if (!$running || $snapname) {
6154 qemu_img_convert($drive->{file}, $newvolid, $size, $snapname);
6155 } else {
6156 qemu_drive_mirror($vmid, $drivename, $newvolid, $newvmid);
6157 }
6158 }
6159
6160 my ($size) = PVE::Storage::volume_size_info($storecfg, $newvolid, 3);
6161
6162 my $disk = $drive;
6163 $disk->{format} = undef;
6164 $disk->{file} = $newvolid;
6165 $disk->{size} = $size;
6166
6167 return $disk;
6168 }
6169
6170 # this only works if VM is running
6171 sub get_current_qemu_machine {
6172 my ($vmid) = @_;
6173
6174 my $cmd = { execute => 'query-machines', arguments => {} };
6175 my $res = vm_qmp_command($vmid, $cmd);
6176
6177 my ($current, $default);
6178 foreach my $e (@$res) {
6179 $default = $e->{name} if $e->{'is-default'};
6180 $current = $e->{name} if $e->{'is-current'};
6181 }
6182
6183 # fallback to the default machine if current is not supported by qemu
6184 return $current || $default || 'pc';
6185 }
6186
6187 sub qemu_machine_feature_enabled {
6188 my ($machine, $kvmver, $version_major, $version_minor) = @_;
6189
6190 my $current_major;
6191 my $current_minor;
6192
6193 if ($machine && $machine =~ m/^(pc(-i440fx|-q35)?-(\d+)\.(\d+))/) {
6194
6195 $current_major = $3;
6196 $current_minor = $4;
6197
6198 } elsif ($kvmver =~ m/^(\d+)\.(\d+)/) {
6199
6200 $current_major = $1;
6201 $current_minor = $2;
6202 }
6203
6204 return 1 if $current_major >= $version_major && $current_minor >= $version_minor;
6205
6206
6207 }
6208
6209 sub lspci {
6210
6211 my $devices = {};
6212
6213 dir_glob_foreach("$pcisysfs/devices", '[a-f0-9]{4}:([a-f0-9]{2}:[a-f0-9]{2})\.([0-9])', sub {
6214 my (undef, $id, $function) = @_;
6215 my $res = { id => $id, function => $function};
6216 push @{$devices->{$id}}, $res;
6217 });
6218
6219 return $devices;
6220 }
6221
6222 sub vm_iothreads_list {
6223 my ($vmid) = @_;
6224
6225 my $res = vm_mon_cmd($vmid, 'query-iothreads');
6226
6227 my $iothreads = {};
6228 foreach my $iothread (@$res) {
6229 $iothreads->{ $iothread->{id} } = $iothread->{"thread-id"};
6230 }
6231
6232 return $iothreads;
6233 }
6234
6235 sub scsihw_infos {
6236 my ($conf, $drive) = @_;
6237
6238 my $maxdev = 0;
6239
6240 if ($conf->{scsihw} && ($conf->{scsihw} =~ m/^lsi/)) {
6241 $maxdev = 7;
6242 } elsif ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
6243 $maxdev = 1;
6244 } else {
6245 $maxdev = 256;
6246 }
6247
6248 my $controller = int($drive->{index} / $maxdev);
6249 my $controller_prefix = ($conf->{scsihw} && $conf->{scsihw} eq 'virtio-scsi-single') ? "virtioscsi" : "scsihw";
6250
6251 return ($maxdev, $controller, $controller_prefix);
6252 }
6253
6254 1;