]> git.proxmox.com Git - qemu-server.git/blob - PVE/QemuServer.pm
fixed bug 662, wrong subroutine for parsing startup order
[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' || $cpu eq 'kvm32';
2844
2845 push @$cpuFlags, '-rdtscp' if $cpu =~ m/^Opteron/;
2846
2847 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
2848
2849 push @$cpuFlags , '+kvm_pv_unhalt' if !$nokvm;
2850 push @$cpuFlags , '+kvm_pv_eoi' if !$nokvm;
2851 }
2852
2853 push @$cpuFlags, 'enforce' if $cpu ne 'host' && !$nokvm;
2854
2855 $cpu .= "," . join(',', @$cpuFlags) if scalar(@$cpuFlags);
2856
2857 push @$cmd, '-cpu', $cpu;
2858
2859 my $memory = $conf->{memory} || $defaults->{memory};
2860 my $static_memory = 0;
2861 my $dimm_memory = 0;
2862
2863 if ($hotplug_features->{memory}) {
2864 die "Numa need to be enabled for memory hotplug\n" if !$conf->{numa};
2865 die "Total memory is bigger than ${MAX_MEM}MB\n" if $memory > $MAX_MEM;
2866 $static_memory = $STATICMEM;
2867 die "minimum memory must be ${static_memory}MB\n" if($memory < $static_memory);
2868 $dimm_memory = $memory - $static_memory;
2869 push @$cmd, '-m', "size=${static_memory},slots=255,maxmem=${MAX_MEM}M";
2870
2871 } else {
2872
2873 $static_memory = $memory;
2874 push @$cmd, '-m', $static_memory;
2875 }
2876
2877 if ($conf->{numa}) {
2878
2879 my $numa_totalmemory = undef;
2880 for (my $i = 0; $i < $MAX_NUMA; $i++) {
2881 next if !$conf->{"numa$i"};
2882 my $numa = parse_numa($conf->{"numa$i"});
2883 next if !$numa;
2884 # memory
2885 die "missing numa node$i memory value\n" if !$numa->{memory};
2886 my $numa_memory = $numa->{memory};
2887 $numa_totalmemory += $numa_memory;
2888 my $numa_object = "memory-backend-ram,id=ram-node$i,size=${numa_memory}M";
2889
2890 # cpus
2891 my $cpus_start = $numa->{cpus}->{start};
2892 die "missing numa node$i cpus\n" if !defined($cpus_start);
2893 my $cpus_end = $numa->{cpus}->{end} if defined($numa->{cpus}->{end});
2894 my $cpus = $cpus_start;
2895 if (defined($cpus_end)) {
2896 $cpus .= "-$cpus_end";
2897 die "numa node$i : cpu range $cpus is incorrect\n" if $cpus_end <= $cpus_start;
2898 }
2899
2900 # hostnodes
2901 my $hostnodes_start = $numa->{hostnodes}->{start};
2902 if (defined($hostnodes_start)) {
2903 my $hostnodes_end = $numa->{hostnodes}->{end} if defined($numa->{hostnodes}->{end});
2904 my $hostnodes = $hostnodes_start;
2905 if (defined($hostnodes_end)) {
2906 $hostnodes .= "-$hostnodes_end";
2907 die "host node $hostnodes range is incorrect\n" if $hostnodes_end <= $hostnodes_start;
2908 }
2909
2910 my $hostnodes_end_range = defined($hostnodes_end) ? $hostnodes_end : $hostnodes_start;
2911 for (my $i = $hostnodes_start; $i <= $hostnodes_end_range; $i++ ) {
2912 die "host numa node$i don't exist\n" if ! -d "/sys/devices/system/node/node$i/";
2913 }
2914
2915 # policy
2916 my $policy = $numa->{policy};
2917 die "you need to define a policy for hostnode $hostnodes\n" if !$policy;
2918 $numa_object .= ",host-nodes=$hostnodes,policy=$policy";
2919 }
2920
2921 push @$cmd, '-object', $numa_object;
2922 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2923 }
2924
2925 die "total memory for NUMA nodes must be equal to vm static memory\n"
2926 if $numa_totalmemory && $numa_totalmemory != $static_memory;
2927
2928 #if no custom tology, we split memory and cores across numa nodes
2929 if(!$numa_totalmemory) {
2930
2931 my $numa_memory = ($static_memory / $sockets) . "M";
2932
2933 for (my $i = 0; $i < $sockets; $i++) {
2934
2935 my $cpustart = ($cores * $i);
2936 my $cpuend = ($cpustart + $cores - 1) if $cores && $cores > 1;
2937 my $cpus = $cpustart;
2938 $cpus .= "-$cpuend" if $cpuend;
2939
2940 push @$cmd, '-object', "memory-backend-ram,size=$numa_memory,id=ram-node$i";
2941 push @$cmd, '-numa', "node,nodeid=$i,cpus=$cpus,memdev=ram-node$i";
2942 }
2943 }
2944 }
2945
2946 if ($hotplug_features->{memory}) {
2947 foreach_dimm($conf, $vmid, $memory, $sockets, sub {
2948 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
2949 push @$cmd, "-object" , "memory-backend-ram,id=mem-$name,size=${dimm_size}M";
2950 push @$cmd, "-device", "pc-dimm,id=$name,memdev=mem-$name,node=$numanode";
2951
2952 #if dimm_memory is not aligned to dimm map
2953 if($current_size > $memory) {
2954 $conf->{memory} = $current_size;
2955 update_config_nolock($vmid, $conf, 1);
2956 }
2957 });
2958 }
2959
2960 push @$cmd, '-S' if $conf->{freeze};
2961
2962 # set keyboard layout
2963 my $kb = $conf->{keyboard} || $defaults->{keyboard};
2964 push @$cmd, '-k', $kb if $kb;
2965
2966 # enable sound
2967 #my $soundhw = $conf->{soundhw} || $defaults->{soundhw};
2968 #push @$cmd, '-soundhw', 'es1370';
2969 #push @$cmd, '-soundhw', $soundhw if $soundhw;
2970
2971 if($conf->{agent}) {
2972 my $qgasocket = qmp_socket($vmid, 1);
2973 my $pciaddr = print_pci_addr("qga0", $bridges);
2974 push @$devices, '-chardev', "socket,path=$qgasocket,server,nowait,id=qga0";
2975 push @$devices, '-device', "virtio-serial,id=qga0$pciaddr";
2976 push @$devices, '-device', 'virtserialport,chardev=qga0,name=org.qemu.guest_agent.0';
2977 }
2978
2979 my $spice_port;
2980
2981 if ($qxlnum) {
2982 if ($qxlnum > 1) {
2983 if ($conf->{ostype} && $conf->{ostype} =~ m/^w/){
2984 for(my $i = 1; $i < $qxlnum; $i++){
2985 my $pciaddr = print_pci_addr("vga$i", $bridges);
2986 push @$cmd, '-device', "qxl,id=vga$i,ram_size=67108864,vram_size=33554432$pciaddr";
2987 }
2988 } else {
2989 # assume other OS works like Linux
2990 push @$cmd, '-global', 'qxl-vga.ram_size=134217728';
2991 push @$cmd, '-global', 'qxl-vga.vram_size=67108864';
2992 }
2993 }
2994
2995 my $pciaddr = print_pci_addr("spice", $bridges);
2996
2997 my $nodename = PVE::INotify::nodename();
2998 my $pfamily = PVE::Tools::get_host_address_family($nodename);
2999 $spice_port = PVE::Tools::next_spice_port($pfamily);
3000
3001 push @$devices, '-spice', "tls-port=${spice_port},addr=localhost,tls-ciphers=DES-CBC3-SHA,seamless-migration=on";
3002
3003 push @$devices, '-device', "virtio-serial,id=spice$pciaddr";
3004 push @$devices, '-chardev', "spicevmc,id=vdagent,name=vdagent";
3005 push @$devices, '-device', "virtserialport,chardev=vdagent,name=com.redhat.spice.0";
3006 }
3007
3008 # enable balloon by default, unless explicitly disabled
3009 if (!defined($conf->{balloon}) || $conf->{balloon}) {
3010 $pciaddr = print_pci_addr("balloon0", $bridges);
3011 push @$devices, '-device', "virtio-balloon-pci,id=balloon0$pciaddr";
3012 }
3013
3014 if ($conf->{watchdog}) {
3015 my $wdopts = parse_watchdog($conf->{watchdog});
3016 $pciaddr = print_pci_addr("watchdog", $bridges);
3017 my $watchdog = $wdopts->{model} || 'i6300esb';
3018 push @$devices, '-device', "$watchdog$pciaddr";
3019 push @$devices, '-watchdog-action', $wdopts->{action} if $wdopts->{action};
3020 }
3021
3022 my $vollist = [];
3023 my $scsicontroller = {};
3024 my $ahcicontroller = {};
3025 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : $defaults->{scsihw};
3026
3027 # Add iscsi initiator name if available
3028 if (my $initiator = get_initiator_name()) {
3029 push @$devices, '-iscsi', "initiator-name=$initiator";
3030 }
3031
3032 foreach_drive($conf, sub {
3033 my ($ds, $drive) = @_;
3034
3035 if (PVE::Storage::parse_volume_id($drive->{file}, 1)) {
3036 push @$vollist, $drive->{file};
3037 }
3038
3039 $use_virtio = 1 if $ds =~ m/^virtio/;
3040
3041 if (drive_is_cdrom ($drive)) {
3042 if ($bootindex_hash->{d}) {
3043 $drive->{bootindex} = $bootindex_hash->{d};
3044 $bootindex_hash->{d} += 1;
3045 }
3046 } else {
3047 if ($bootindex_hash->{c}) {
3048 $drive->{bootindex} = $bootindex_hash->{c} if $conf->{bootdisk} && ($conf->{bootdisk} eq $ds);
3049 $bootindex_hash->{c} += 1;
3050 }
3051 }
3052
3053 if($drive->{interface} eq 'virtio'){
3054 push @$cmd, '-object', "iothread,id=iothread-$ds" if $drive->{iothread};
3055 }
3056
3057 if ($drive->{interface} eq 'scsi') {
3058
3059 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $drive);
3060
3061 $pciaddr = print_pci_addr("$controller_prefix$controller", $bridges);
3062 my $scsihw_type = $scsihw =~ m/^virtio-scsi-single/ ? "virtio-scsi-pci" : $scsihw;
3063
3064 my $iothread = '';
3065 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{iothread}){
3066 $iothread .= ",iothread=iothread-$controller_prefix$controller";
3067 push @$cmd, '-object', "iothread,id=iothread-$controller_prefix$controller";
3068 }
3069
3070 my $queues = '';
3071 if($conf->{scsihw} && $conf->{scsihw} eq "virtio-scsi-single" && $drive->{queues}){
3072 $queues = ",num_queues=$drive->{queues}";
3073 }
3074
3075 push @$devices, '-device', "$scsihw_type,id=$controller_prefix$controller$pciaddr$iothread$queues" if !$scsicontroller->{$controller};
3076 $scsicontroller->{$controller}=1;
3077 }
3078
3079 if ($drive->{interface} eq 'sata') {
3080 my $controller = int($drive->{index} / $MAX_SATA_DISKS);
3081 $pciaddr = print_pci_addr("ahci$controller", $bridges);
3082 push @$devices, '-device', "ahci,id=ahci$controller,multifunction=on$pciaddr" if !$ahcicontroller->{$controller};
3083 $ahcicontroller->{$controller}=1;
3084 }
3085
3086 my $drive_cmd = print_drive_full($storecfg, $vmid, $drive);
3087 push @$devices, '-drive',$drive_cmd;
3088 push @$devices, '-device', print_drivedevice_full($storecfg, $conf, $vmid, $drive, $bridges);
3089 });
3090
3091 for (my $i = 0; $i < $MAX_NETS; $i++) {
3092 next if !$conf->{"net$i"};
3093 my $d = parse_net($conf->{"net$i"});
3094 next if !$d;
3095
3096 $use_virtio = 1 if $d->{model} eq 'virtio';
3097
3098 if ($bootindex_hash->{n}) {
3099 $d->{bootindex} = $bootindex_hash->{n};
3100 $bootindex_hash->{n} += 1;
3101 }
3102
3103 my $netdevfull = print_netdev_full($vmid,$conf,$d,"net$i");
3104 push @$devices, '-netdev', $netdevfull;
3105
3106 my $netdevicefull = print_netdevice_full($vmid,$conf,$d,"net$i",$bridges);
3107 push @$devices, '-device', $netdevicefull;
3108 }
3109
3110 if (!$q35) {
3111 # add pci bridges
3112 if (qemu_machine_feature_enabled ($machine_type, $kvmver, 2, 3)) {
3113 $bridges->{1} = 1;
3114 $bridges->{2} = 1;
3115 }
3116
3117 $bridges->{3} = 1 if $scsihw =~ m/^virtio-scsi-single/;
3118
3119 while (my ($k, $v) = each %$bridges) {
3120 $pciaddr = print_pci_addr("pci.$k");
3121 unshift @$devices, '-device', "pci-bridge,id=pci.$k,chassis_nr=$k$pciaddr" if $k > 0;
3122 }
3123 }
3124
3125 # add custom args
3126 if ($conf->{args}) {
3127 my $aa = PVE::Tools::split_args($conf->{args});
3128 push @$cmd, @$aa;
3129 }
3130
3131 push @$cmd, @$devices;
3132 push @$cmd, '-rtc', join(',', @$rtcFlags)
3133 if scalar(@$rtcFlags);
3134 push @$cmd, '-machine', join(',', @$machineFlags)
3135 if scalar(@$machineFlags);
3136 push @$cmd, '-global', join(',', @$globalFlags)
3137 if scalar(@$globalFlags);
3138
3139 return wantarray ? ($cmd, $vollist, $spice_port) : $cmd;
3140 }
3141
3142 sub vnc_socket {
3143 my ($vmid) = @_;
3144 return "${var_run_tmpdir}/$vmid.vnc";
3145 }
3146
3147 sub spice_port {
3148 my ($vmid) = @_;
3149
3150 my $res = vm_mon_cmd($vmid, 'query-spice');
3151
3152 return $res->{'tls-port'} || $res->{'port'} || die "no spice port\n";
3153 }
3154
3155 sub qmp_socket {
3156 my ($vmid, $qga) = @_;
3157 my $sockettype = $qga ? 'qga' : 'qmp';
3158 return "${var_run_tmpdir}/$vmid.$sockettype";
3159 }
3160
3161 sub pidfile_name {
3162 my ($vmid) = @_;
3163 return "${var_run_tmpdir}/$vmid.pid";
3164 }
3165
3166 sub vm_devices_list {
3167 my ($vmid) = @_;
3168
3169 my $res = vm_mon_cmd($vmid, 'query-pci');
3170 my $devices = {};
3171 foreach my $pcibus (@$res) {
3172 foreach my $device (@{$pcibus->{devices}}) {
3173 next if !$device->{'qdev_id'};
3174 if ($device->{'pci_bridge'}) {
3175 $devices->{$device->{'qdev_id'}} = 1;
3176 foreach my $bridge_device (@{$device->{'pci_bridge'}->{devices}}) {
3177 next if !$bridge_device->{'qdev_id'};
3178 $devices->{$bridge_device->{'qdev_id'}} = 1;
3179 $devices->{$device->{'qdev_id'}}++;
3180 }
3181 } else {
3182 $devices->{$device->{'qdev_id'}} = 1;
3183 }
3184 }
3185 }
3186
3187 my $resblock = vm_mon_cmd($vmid, 'query-block');
3188 foreach my $block (@$resblock) {
3189 if($block->{device} =~ m/^drive-(\S+)/){
3190 $devices->{$1} = 1;
3191 }
3192 }
3193
3194 my $resmice = vm_mon_cmd($vmid, 'query-mice');
3195 foreach my $mice (@$resmice) {
3196 if ($mice->{name} eq 'QEMU HID Tablet') {
3197 $devices->{tablet} = 1;
3198 last;
3199 }
3200 }
3201
3202 return $devices;
3203 }
3204
3205 sub vm_deviceplug {
3206 my ($storecfg, $conf, $vmid, $deviceid, $device) = @_;
3207
3208 my $q35 = machine_type_is_q35($conf);
3209
3210 my $devices_list = vm_devices_list($vmid);
3211 return 1 if defined($devices_list->{$deviceid});
3212
3213 qemu_add_pci_bridge($storecfg, $conf, $vmid, $deviceid); # add PCI bridge if we need it for the device
3214
3215 if ($deviceid eq 'tablet') {
3216
3217 qemu_deviceadd($vmid, print_tabletdevice_full($conf));
3218
3219 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3220
3221 qemu_iothread_add($vmid, $deviceid, $device);
3222
3223 qemu_driveadd($storecfg, $vmid, $device);
3224 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3225
3226 qemu_deviceadd($vmid, $devicefull);
3227 eval { qemu_deviceaddverify($vmid, $deviceid); };
3228 if (my $err = $@) {
3229 eval { qemu_drivedel($vmid, $deviceid); };
3230 warn $@ if $@;
3231 die $err;
3232 }
3233
3234 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3235
3236
3237 my $scsihw = defined($conf->{scsihw}) ? $conf->{scsihw} : "lsi";
3238 my $pciaddr = print_pci_addr($deviceid);
3239 my $scsihw_type = $scsihw eq 'virtio-scsi-single' ? "virtio-scsi-pci" : $scsihw;
3240
3241 my $devicefull = "$scsihw_type,id=$deviceid$pciaddr";
3242
3243 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{iothread}) {
3244 qemu_iothread_add($vmid, $deviceid, $device);
3245 $devicefull .= ",iothread=iothread-$deviceid";
3246 }
3247
3248 if($deviceid =~ m/^virtioscsi(\d+)$/ && $device->{queues}) {
3249 $devicefull .= ",num_queues=$device->{queues}";
3250 }
3251
3252 qemu_deviceadd($vmid, $devicefull);
3253 qemu_deviceaddverify($vmid, $deviceid);
3254
3255 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3256
3257 qemu_findorcreatescsihw($storecfg,$conf, $vmid, $device);
3258 qemu_driveadd($storecfg, $vmid, $device);
3259
3260 my $devicefull = print_drivedevice_full($storecfg, $conf, $vmid, $device);
3261 eval { qemu_deviceadd($vmid, $devicefull); };
3262 if (my $err = $@) {
3263 eval { qemu_drivedel($vmid, $deviceid); };
3264 warn $@ if $@;
3265 die $err;
3266 }
3267
3268 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3269
3270 return undef if !qemu_netdevadd($vmid, $conf, $device, $deviceid);
3271 my $netdevicefull = print_netdevice_full($vmid, $conf, $device, $deviceid);
3272 qemu_deviceadd($vmid, $netdevicefull);
3273 eval { qemu_deviceaddverify($vmid, $deviceid); };
3274 if (my $err = $@) {
3275 eval { qemu_netdevdel($vmid, $deviceid); };
3276 warn $@ if $@;
3277 die $err;
3278 }
3279
3280 } elsif (!$q35 && $deviceid =~ m/^(pci\.)(\d+)$/) {
3281
3282 my $bridgeid = $2;
3283 my $pciaddr = print_pci_addr($deviceid);
3284 my $devicefull = "pci-bridge,id=pci.$bridgeid,chassis_nr=$bridgeid$pciaddr";
3285
3286 qemu_deviceadd($vmid, $devicefull);
3287 qemu_deviceaddverify($vmid, $deviceid);
3288
3289 } else {
3290 die "can't hotplug device '$deviceid'\n";
3291 }
3292
3293 return 1;
3294 }
3295
3296 # fixme: this should raise exceptions on error!
3297 sub vm_deviceunplug {
3298 my ($vmid, $conf, $deviceid) = @_;
3299
3300 my $devices_list = vm_devices_list($vmid);
3301 return 1 if !defined($devices_list->{$deviceid});
3302
3303 die "can't unplug bootdisk" if $conf->{bootdisk} && $conf->{bootdisk} eq $deviceid;
3304
3305 if ($deviceid eq 'tablet') {
3306
3307 qemu_devicedel($vmid, $deviceid);
3308
3309 } elsif ($deviceid =~ m/^(virtio)(\d+)$/) {
3310
3311 qemu_devicedel($vmid, $deviceid);
3312 qemu_devicedelverify($vmid, $deviceid);
3313 qemu_drivedel($vmid, $deviceid);
3314 qemu_iothread_del($conf, $vmid, $deviceid);
3315
3316 } elsif ($deviceid =~ m/^(virtioscsi|scsihw)(\d+)$/) {
3317
3318 qemu_devicedel($vmid, $deviceid);
3319 qemu_devicedelverify($vmid, $deviceid);
3320 qemu_iothread_del($conf, $vmid, $deviceid);
3321
3322 } elsif ($deviceid =~ m/^(scsi)(\d+)$/) {
3323
3324 #qemu 2.3 segfault on drive_del with virtioscsi + iothread
3325 my $device = parse_drive($deviceid, $conf->{$deviceid});
3326 die "virtioscsi with iothread is not hot-unplugglable currently" if $device->{iothread};
3327
3328 qemu_devicedel($vmid, $deviceid);
3329 qemu_drivedel($vmid, $deviceid);
3330 qemu_deletescsihw($conf, $vmid, $deviceid);
3331
3332 } elsif ($deviceid =~ m/^(net)(\d+)$/) {
3333
3334 qemu_devicedel($vmid, $deviceid);
3335 qemu_devicedelverify($vmid, $deviceid);
3336 qemu_netdevdel($vmid, $deviceid);
3337
3338 } else {
3339 die "can't unplug device '$deviceid'\n";
3340 }
3341
3342 return 1;
3343 }
3344
3345 sub qemu_deviceadd {
3346 my ($vmid, $devicefull) = @_;
3347
3348 $devicefull = "driver=".$devicefull;
3349 my %options = split(/[=,]/, $devicefull);
3350
3351 vm_mon_cmd($vmid, "device_add" , %options);
3352 }
3353
3354 sub qemu_devicedel {
3355 my ($vmid, $deviceid) = @_;
3356
3357 my $ret = vm_mon_cmd($vmid, "device_del", id => $deviceid);
3358 }
3359
3360 sub qemu_iothread_add {
3361 my($vmid, $deviceid, $device) = @_;
3362
3363 if ($device->{iothread}) {
3364 my $iothreads = vm_iothreads_list($vmid);
3365 qemu_objectadd($vmid, "iothread-$deviceid", "iothread") if !$iothreads->{"iothread-$deviceid"};
3366 }
3367 }
3368
3369 sub qemu_iothread_del {
3370 my($conf, $vmid, $deviceid) = @_;
3371
3372 my $device = parse_drive($deviceid, $conf->{$deviceid});
3373 if ($device->{iothread}) {
3374 my $iothreads = vm_iothreads_list($vmid);
3375 qemu_objectdel($vmid, "iothread-$deviceid") if $iothreads->{"iothread-$deviceid"};
3376 }
3377 }
3378
3379 sub qemu_objectadd {
3380 my($vmid, $objectid, $qomtype) = @_;
3381
3382 vm_mon_cmd($vmid, "object-add", id => $objectid, "qom-type" => $qomtype);
3383
3384 return 1;
3385 }
3386
3387 sub qemu_objectdel {
3388 my($vmid, $objectid) = @_;
3389
3390 vm_mon_cmd($vmid, "object-del", id => $objectid);
3391
3392 return 1;
3393 }
3394
3395 sub qemu_driveadd {
3396 my ($storecfg, $vmid, $device) = @_;
3397
3398 my $drive = print_drive_full($storecfg, $vmid, $device);
3399 $drive =~ s/\\/\\\\/g;
3400 my $ret = vm_human_monitor_command($vmid, "drive_add auto \"$drive\"");
3401
3402 # If the command succeeds qemu prints: "OK"
3403 return 1 if $ret =~ m/OK/s;
3404
3405 die "adding drive failed: $ret\n";
3406 }
3407
3408 sub qemu_drivedel {
3409 my($vmid, $deviceid) = @_;
3410
3411 my $ret = vm_human_monitor_command($vmid, "drive_del drive-$deviceid");
3412 $ret =~ s/^\s+//;
3413
3414 return 1 if $ret eq "";
3415
3416 # NB: device not found errors mean the drive was auto-deleted and we ignore the error
3417 return 1 if $ret =~ m/Device \'.*?\' not found/s;
3418
3419 die "deleting drive $deviceid failed : $ret\n";
3420 }
3421
3422 sub qemu_deviceaddverify {
3423 my ($vmid, $deviceid) = @_;
3424
3425 for (my $i = 0; $i <= 5; $i++) {
3426 my $devices_list = vm_devices_list($vmid);
3427 return 1 if defined($devices_list->{$deviceid});
3428 sleep 1;
3429 }
3430
3431 die "error on hotplug device '$deviceid'\n";
3432 }
3433
3434
3435 sub qemu_devicedelverify {
3436 my ($vmid, $deviceid) = @_;
3437
3438 # need to verify that the device is correctly removed as device_del
3439 # is async and empty return is not reliable
3440
3441 for (my $i = 0; $i <= 5; $i++) {
3442 my $devices_list = vm_devices_list($vmid);
3443 return 1 if !defined($devices_list->{$deviceid});
3444 sleep 1;
3445 }
3446
3447 die "error on hot-unplugging device '$deviceid'\n";
3448 }
3449
3450 sub qemu_findorcreatescsihw {
3451 my ($storecfg, $conf, $vmid, $device) = @_;
3452
3453 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3454
3455 my $scsihwid="$controller_prefix$controller";
3456 my $devices_list = vm_devices_list($vmid);
3457
3458 if(!defined($devices_list->{$scsihwid})) {
3459 vm_deviceplug($storecfg, $conf, $vmid, $scsihwid, $device);
3460 }
3461
3462 return 1;
3463 }
3464
3465 sub qemu_deletescsihw {
3466 my ($conf, $vmid, $opt) = @_;
3467
3468 my $device = parse_drive($opt, $conf->{$opt});
3469
3470 if ($conf->{scsihw} && ($conf->{scsihw} eq 'virtio-scsi-single')) {
3471 vm_deviceunplug($vmid, $conf, "virtioscsi$device->{index}");
3472 return 1;
3473 }
3474
3475 my ($maxdev, $controller, $controller_prefix) = scsihw_infos($conf, $device);
3476
3477 my $devices_list = vm_devices_list($vmid);
3478 foreach my $opt (keys %{$devices_list}) {
3479 if (PVE::QemuServer::valid_drivename($opt)) {
3480 my $drive = PVE::QemuServer::parse_drive($opt, $conf->{$opt});
3481 if($drive->{interface} eq 'scsi' && $drive->{index} < (($maxdev-1)*($controller+1))) {
3482 return 1;
3483 }
3484 }
3485 }
3486
3487 my $scsihwid="scsihw$controller";
3488
3489 vm_deviceunplug($vmid, $conf, $scsihwid);
3490
3491 return 1;
3492 }
3493
3494 sub qemu_add_pci_bridge {
3495 my ($storecfg, $conf, $vmid, $device) = @_;
3496
3497 my $bridges = {};
3498
3499 my $bridgeid;
3500
3501 print_pci_addr($device, $bridges);
3502
3503 while (my ($k, $v) = each %$bridges) {
3504 $bridgeid = $k;
3505 }
3506 return 1 if !defined($bridgeid) || $bridgeid < 1;
3507
3508 my $bridge = "pci.$bridgeid";
3509 my $devices_list = vm_devices_list($vmid);
3510
3511 if (!defined($devices_list->{$bridge})) {
3512 vm_deviceplug($storecfg, $conf, $vmid, $bridge);
3513 }
3514
3515 return 1;
3516 }
3517
3518 sub qemu_set_link_status {
3519 my ($vmid, $device, $up) = @_;
3520
3521 vm_mon_cmd($vmid, "set_link", name => $device,
3522 up => $up ? JSON::true : JSON::false);
3523 }
3524
3525 sub qemu_netdevadd {
3526 my ($vmid, $conf, $device, $deviceid) = @_;
3527
3528 my $netdev = print_netdev_full($vmid, $conf, $device, $deviceid);
3529 my %options = split(/[=,]/, $netdev);
3530
3531 vm_mon_cmd($vmid, "netdev_add", %options);
3532 return 1;
3533 }
3534
3535 sub qemu_netdevdel {
3536 my ($vmid, $deviceid) = @_;
3537
3538 vm_mon_cmd($vmid, "netdev_del", id => $deviceid);
3539 }
3540
3541 sub qemu_cpu_hotplug {
3542 my ($vmid, $conf, $vcpus) = @_;
3543
3544 my $sockets = 1;
3545 $sockets = $conf->{smp} if $conf->{smp}; # old style - no longer iused
3546 $sockets = $conf->{sockets} if $conf->{sockets};
3547 my $cores = $conf->{cores} || 1;
3548 my $maxcpus = $sockets * $cores;
3549
3550 $vcpus = $maxcpus if !$vcpus;
3551
3552 die "you can't add more vcpus than maxcpus\n"
3553 if $vcpus > $maxcpus;
3554
3555 my $currentvcpus = $conf->{vcpus} || $maxcpus;
3556 die "online cpu unplug is not yet possible\n"
3557 if $vcpus < $currentvcpus;
3558
3559 my $currentrunningvcpus = vm_mon_cmd($vmid, "query-cpus");
3560 die "vcpus in running vm is different than configuration\n"
3561 if scalar(@{$currentrunningvcpus}) != $currentvcpus;
3562
3563 for (my $i = $currentvcpus; $i < $vcpus; $i++) {
3564 vm_mon_cmd($vmid, "cpu-add", id => int($i));
3565 }
3566 }
3567
3568 sub qemu_memory_hotplug {
3569 my ($vmid, $conf, $defaults, $opt, $value) = @_;
3570
3571 return $value if !check_running($vmid);
3572
3573 my $memory = $conf->{memory} || $defaults->{memory};
3574 $value = $defaults->{memory} if !$value;
3575 return $value if $value == $memory;
3576
3577 my $static_memory = $STATICMEM;
3578 my $dimm_memory = $memory - $static_memory;
3579
3580 die "memory can't be lower than $static_memory MB" if $value < $static_memory;
3581 die "memory unplug is not yet available" if $value < $memory;
3582 die "you cannot add more memory than $MAX_MEM MB!\n" if $memory > $MAX_MEM;
3583
3584
3585 my $sockets = 1;
3586 $sockets = $conf->{sockets} if $conf->{sockets};
3587
3588 foreach_dimm($conf, $vmid, $value, $sockets, sub {
3589 my ($conf, $vmid, $name, $dimm_size, $numanode, $current_size, $memory) = @_;
3590
3591 return if $current_size <= $conf->{memory};
3592
3593 eval { vm_mon_cmd($vmid, "object-add", 'qom-type' => "memory-backend-ram", id => "mem-$name", props => { size => int($dimm_size*1024*1024) } ) };
3594 if (my $err = $@) {
3595 eval { qemu_objectdel($vmid, "mem-$name"); };
3596 die $err;
3597 }
3598
3599 eval { vm_mon_cmd($vmid, "device_add", driver => "pc-dimm", id => "$name", memdev => "mem-$name", node => $numanode) };
3600 if (my $err = $@) {
3601 eval { qemu_objectdel($vmid, "mem-$name"); };
3602 die $err;
3603 }
3604 #update conf after each succesful module hotplug
3605 $conf->{memory} = $current_size;
3606 update_config_nolock($vmid, $conf, 1);
3607 });
3608 }
3609
3610 sub qemu_block_set_io_throttle {
3611 my ($vmid, $deviceid, $bps, $bps_rd, $bps_wr, $iops, $iops_rd, $iops_wr) = @_;
3612
3613 return if !check_running($vmid) ;
3614
3615 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($iops_wr));
3616
3617 }
3618
3619 # old code, only used to shutdown old VM after update
3620 sub __read_avail {
3621 my ($fh, $timeout) = @_;
3622
3623 my $sel = new IO::Select;
3624 $sel->add($fh);
3625
3626 my $res = '';
3627 my $buf;
3628
3629 my @ready;
3630 while (scalar (@ready = $sel->can_read($timeout))) {
3631 my $count;
3632 if ($count = $fh->sysread($buf, 8192)) {
3633 if ($buf =~ /^(.*)\(qemu\) $/s) {
3634 $res .= $1;
3635 last;
3636 } else {
3637 $res .= $buf;
3638 }
3639 } else {
3640 if (!defined($count)) {
3641 die "$!\n";
3642 }
3643 last;
3644 }
3645 }
3646
3647 die "monitor read timeout\n" if !scalar(@ready);
3648
3649 return $res;
3650 }
3651
3652 # old code, only used to shutdown old VM after update
3653 sub vm_monitor_command {
3654 my ($vmid, $cmdstr, $nocheck) = @_;
3655
3656 my $res;
3657
3658 eval {
3659 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
3660
3661 my $sname = "${var_run_tmpdir}/$vmid.mon";
3662
3663 my $sock = IO::Socket::UNIX->new( Peer => $sname ) ||
3664 die "unable to connect to VM $vmid socket - $!\n";
3665
3666 my $timeout = 3;
3667
3668 # hack: migrate sometime blocks the monitor (when migrate_downtime
3669 # is set)
3670 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3671 $timeout = 60*60; # 1 hour
3672 }
3673
3674 # read banner;
3675 my $data = __read_avail($sock, $timeout);
3676
3677 if ($data !~ m/^QEMU\s+(\S+)\s+monitor\s/) {
3678 die "got unexpected qemu monitor banner\n";
3679 }
3680
3681 my $sel = new IO::Select;
3682 $sel->add($sock);
3683
3684 if (!scalar(my @ready = $sel->can_write($timeout))) {
3685 die "monitor write error - timeout";
3686 }
3687
3688 my $fullcmd = "$cmdstr\r";
3689
3690 # syslog('info', "VM $vmid monitor command: $cmdstr");
3691
3692 my $b;
3693 if (!($b = $sock->syswrite($fullcmd)) || ($b != length($fullcmd))) {
3694 die "monitor write error - $!";
3695 }
3696
3697 return if ($cmdstr eq 'q') || ($cmdstr eq 'quit');
3698
3699 $timeout = 20;
3700
3701 if ($cmdstr =~ m/^(info\s+migrate|migrate\s)/) {
3702 $timeout = 60*60; # 1 hour
3703 } elsif ($cmdstr =~ m/^(eject|change)/) {
3704 $timeout = 60; # note: cdrom mount command is slow
3705 }
3706 if ($res = __read_avail($sock, $timeout)) {
3707
3708 my @lines = split("\r?\n", $res);
3709
3710 shift @lines if $lines[0] !~ m/^unknown command/; # skip echo
3711
3712 $res = join("\n", @lines);
3713 $res .= "\n";
3714 }
3715 };
3716
3717 my $err = $@;
3718
3719 if ($err) {
3720 syslog("err", "VM $vmid monitor command failed - $err");
3721 die $err;
3722 }
3723
3724 return $res;
3725 }
3726
3727 sub qemu_block_resize {
3728 my ($vmid, $deviceid, $storecfg, $volid, $size) = @_;
3729
3730 my $running = check_running($vmid);
3731
3732 return if !PVE::Storage::volume_resize($storecfg, $volid, $size, $running);
3733
3734 return if !$running;
3735
3736 vm_mon_cmd($vmid, "block_resize", device => $deviceid, size => int($size));
3737
3738 }
3739
3740 sub qemu_volume_snapshot {
3741 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3742
3743 my $running = check_running($vmid);
3744
3745 if ($running && do_snapshots_with_qemu($storecfg, $volid)){
3746 vm_mon_cmd($vmid, "snapshot-drive", device => $deviceid, name => $snap);
3747 } else {
3748 PVE::Storage::volume_snapshot($storecfg, $volid, $snap);
3749 }
3750 }
3751
3752 sub qemu_volume_snapshot_delete {
3753 my ($vmid, $deviceid, $storecfg, $volid, $snap) = @_;
3754
3755 my $running = check_running($vmid);
3756
3757 return if !PVE::Storage::volume_snapshot_delete($storecfg, $volid, $snap, $running);
3758
3759 return if !$running;
3760
3761 vm_mon_cmd($vmid, "delete-drive-snapshot", device => $deviceid, name => $snap);
3762 }
3763
3764 sub set_migration_caps {
3765 my ($vmid) = @_;
3766
3767 my $cap_ref = [];
3768
3769 my $enabled_cap = {
3770 "auto-converge" => 1,
3771 "xbzrle" => 0,
3772 "x-rdma-pin-all" => 0,
3773 "zero-blocks" => 0,
3774 };
3775
3776 my $supported_capabilities = vm_mon_cmd_nocheck($vmid, "query-migrate-capabilities");
3777
3778 for my $supported_capability (@$supported_capabilities) {
3779 push @$cap_ref, {
3780 capability => $supported_capability->{capability},
3781 state => $enabled_cap->{$supported_capability->{capability}} ? JSON::true : JSON::false,
3782 };
3783 }
3784
3785 vm_mon_cmd_nocheck($vmid, "migrate-set-capabilities", capabilities => $cap_ref);
3786 }
3787
3788 my $fast_plug_option = {
3789 'lock' => 1,
3790 'name' => 1,
3791 'onboot' => 1,
3792 'shares' => 1,
3793 'startup' => 1,
3794 };
3795
3796 # hotplug changes in [PENDING]
3797 # $selection hash can be used to only apply specified options, for
3798 # example: { cores => 1 } (only apply changed 'cores')
3799 # $errors ref is used to return error messages
3800 sub vmconfig_hotplug_pending {
3801 my ($vmid, $conf, $storecfg, $selection, $errors) = @_;
3802
3803 my $defaults = load_defaults();
3804
3805 # commit values which do not have any impact on running VM first
3806 # Note: those option cannot raise errors, we we do not care about
3807 # $selection and always apply them.
3808
3809 my $add_error = sub {
3810 my ($opt, $msg) = @_;
3811 $errors->{$opt} = "hotplug problem - $msg";
3812 };
3813
3814 my $changes = 0;
3815 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3816 if ($fast_plug_option->{$opt}) {
3817 $conf->{$opt} = $conf->{pending}->{$opt};
3818 delete $conf->{pending}->{$opt};
3819 $changes = 1;
3820 }
3821 }
3822
3823 if ($changes) {
3824 update_config_nolock($vmid, $conf, 1);
3825 $conf = load_config($vmid); # update/reload
3826 }
3827
3828 my $hotplug_features = parse_hotplug_features(defined($conf->{hotplug}) ? $conf->{hotplug} : '1');
3829
3830 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3831 foreach my $opt (@delete) {
3832 next if $selection && !$selection->{$opt};
3833 eval {
3834 if ($opt eq 'hotplug') {
3835 die "skip\n" if ($conf->{hotplug} =~ /memory/);
3836 } elsif ($opt eq 'tablet') {
3837 die "skip\n" if !$hotplug_features->{usb};
3838 if ($defaults->{tablet}) {
3839 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3840 } else {
3841 vm_deviceunplug($vmid, $conf, $opt);
3842 }
3843 } elsif ($opt eq 'vcpus') {
3844 die "skip\n" if !$hotplug_features->{cpu};
3845 qemu_cpu_hotplug($vmid, $conf, undef);
3846 } elsif ($opt eq 'balloon') {
3847 # enable balloon device is not hotpluggable
3848 die "skip\n" if !defined($conf->{balloon}) || $conf->{balloon};
3849 } elsif ($fast_plug_option->{$opt}) {
3850 # do nothing
3851 } elsif ($opt =~ m/^net(\d+)$/) {
3852 die "skip\n" if !$hotplug_features->{network};
3853 vm_deviceunplug($vmid, $conf, $opt);
3854 } elsif (valid_drivename($opt)) {
3855 die "skip\n" if !$hotplug_features->{disk} || $opt =~ m/(ide|sata)(\d+)/;
3856 vm_deviceunplug($vmid, $conf, $opt);
3857 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3858 } elsif ($opt =~ m/^memory$/) {
3859 die "skip\n" if !$hotplug_features->{memory};
3860 qemu_memory_hotplug($vmid, $conf, $defaults, $opt);
3861 } elsif ($opt eq 'cpuunits') {
3862 cgroups_write("cpu", $vmid, "cpu.shares", $defaults->{cpuunits});
3863 } elsif ($opt eq 'cpulimit') {
3864 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", -1);
3865 } else {
3866 die "skip\n";
3867 }
3868 };
3869 if (my $err = $@) {
3870 &$add_error($opt, $err) if $err ne "skip\n";
3871 } else {
3872 # save new config if hotplug was successful
3873 delete $conf->{$opt};
3874 vmconfig_undelete_pending_option($conf, $opt);
3875 update_config_nolock($vmid, $conf, 1);
3876 $conf = load_config($vmid); # update/reload
3877 }
3878 }
3879
3880 foreach my $opt (keys %{$conf->{pending}}) {
3881 next if $selection && !$selection->{$opt};
3882 my $value = $conf->{pending}->{$opt};
3883 eval {
3884 if ($opt eq 'hotplug') {
3885 die "skip\n" if ($value =~ /memory/) || ($value !~ /memory/ && $conf->{hotplug} =~ /memory/);
3886 } elsif ($opt eq 'tablet') {
3887 die "skip\n" if !$hotplug_features->{usb};
3888 if ($value == 1) {
3889 vm_deviceplug($storecfg, $conf, $vmid, $opt);
3890 } elsif ($value == 0) {
3891 vm_deviceunplug($vmid, $conf, $opt);
3892 }
3893 } elsif ($opt eq 'vcpus') {
3894 die "skip\n" if !$hotplug_features->{cpu};
3895 qemu_cpu_hotplug($vmid, $conf, $value);
3896 } elsif ($opt eq 'balloon') {
3897 # enable/disable balloning device is not hotpluggable
3898 my $old_balloon_enabled = !!(!defined($conf->{balloon}) || $conf->{balloon});
3899 my $new_balloon_enabled = !!(!defined($conf->{pending}->{balloon}) || $conf->{pending}->{balloon});
3900 die "skip\n" if $old_balloon_enabled != $new_balloon_enabled;
3901
3902 # allow manual ballooning if shares is set to zero
3903 if ((defined($conf->{shares}) && ($conf->{shares} == 0))) {
3904 my $balloon = $conf->{pending}->{balloon} || $conf->{memory} || $defaults->{memory};
3905 vm_mon_cmd($vmid, "balloon", value => $balloon*1024*1024);
3906 }
3907 } elsif ($opt =~ m/^net(\d+)$/) {
3908 # some changes can be done without hotplug
3909 vmconfig_update_net($storecfg, $conf, $hotplug_features->{network},
3910 $vmid, $opt, $value);
3911 } elsif (valid_drivename($opt)) {
3912 # some changes can be done without hotplug
3913 vmconfig_update_disk($storecfg, $conf, $hotplug_features->{disk},
3914 $vmid, $opt, $value, 1);
3915 } elsif ($opt =~ m/^memory$/) { #dimms
3916 die "skip\n" if !$hotplug_features->{memory};
3917 $value = qemu_memory_hotplug($vmid, $conf, $defaults, $opt, $value);
3918 } elsif ($opt eq 'cpuunits') {
3919 cgroups_write("cpu", $vmid, "cpu.shares", $conf->{pending}->{$opt});
3920 } elsif ($opt eq 'cpulimit') {
3921 my $cpulimit = $conf->{pending}->{$opt} == 0 ? -1 : int($conf->{pending}->{$opt} * 100000);
3922 cgroups_write("cpu", $vmid, "cpu.cfs_quota_us", $cpulimit);
3923 } else {
3924 die "skip\n"; # skip non-hot-pluggable options
3925 }
3926 };
3927 if (my $err = $@) {
3928 &$add_error($opt, $err) if $err ne "skip\n";
3929 } else {
3930 # save new config if hotplug was successful
3931 $conf->{$opt} = $value;
3932 delete $conf->{pending}->{$opt};
3933 update_config_nolock($vmid, $conf, 1);
3934 $conf = load_config($vmid); # update/reload
3935 }
3936 }
3937 }
3938
3939 sub vmconfig_apply_pending {
3940 my ($vmid, $conf, $storecfg) = @_;
3941
3942 # cold plug
3943
3944 my @delete = PVE::Tools::split_list($conf->{pending}->{delete});
3945 foreach my $opt (@delete) { # delete
3946 die "internal error" if $opt =~ m/^unused/;
3947 $conf = load_config($vmid); # update/reload
3948 if (!defined($conf->{$opt})) {
3949 vmconfig_undelete_pending_option($conf, $opt);
3950 update_config_nolock($vmid, $conf, 1);
3951 } elsif (valid_drivename($opt)) {
3952 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}));
3953 vmconfig_undelete_pending_option($conf, $opt);
3954 delete $conf->{$opt};
3955 update_config_nolock($vmid, $conf, 1);
3956 } else {
3957 vmconfig_undelete_pending_option($conf, $opt);
3958 delete $conf->{$opt};
3959 update_config_nolock($vmid, $conf, 1);
3960 }
3961 }
3962
3963 $conf = load_config($vmid); # update/reload
3964
3965 foreach my $opt (keys %{$conf->{pending}}) { # add/change
3966 $conf = load_config($vmid); # update/reload
3967
3968 if (defined($conf->{$opt}) && ($conf->{$opt} eq $conf->{pending}->{$opt})) {
3969 # skip if nothing changed
3970 } elsif (valid_drivename($opt)) {
3971 vmconfig_register_unused_drive($storecfg, $vmid, $conf, parse_drive($opt, $conf->{$opt}))
3972 if defined($conf->{$opt});
3973 $conf->{$opt} = $conf->{pending}->{$opt};
3974 } else {
3975 $conf->{$opt} = $conf->{pending}->{$opt};
3976 }
3977
3978 delete $conf->{pending}->{$opt};
3979 update_config_nolock($vmid, $conf, 1);
3980 }
3981 }
3982
3983 my $safe_num_ne = sub {
3984 my ($a, $b) = @_;
3985
3986 return 0 if !defined($a) && !defined($b);
3987 return 1 if !defined($a);
3988 return 1 if !defined($b);
3989
3990 return $a != $b;
3991 };
3992
3993 my $safe_string_ne = sub {
3994 my ($a, $b) = @_;
3995
3996 return 0 if !defined($a) && !defined($b);
3997 return 1 if !defined($a);
3998 return 1 if !defined($b);
3999
4000 return $a ne $b;
4001 };
4002
4003 sub vmconfig_update_net {
4004 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value) = @_;
4005
4006 my $newnet = parse_net($value);
4007
4008 if ($conf->{$opt}) {
4009 my $oldnet = parse_net($conf->{$opt});
4010
4011 if (&$safe_string_ne($oldnet->{model}, $newnet->{model}) ||
4012 &$safe_string_ne($oldnet->{macaddr}, $newnet->{macaddr}) ||
4013 &$safe_num_ne($oldnet->{queues}, $newnet->{queues}) ||
4014 !($newnet->{bridge} && $oldnet->{bridge})) { # bridge/nat mode change
4015
4016 # for non online change, we try to hot-unplug
4017 die "skip\n" if !$hotplug;
4018 vm_deviceunplug($vmid, $conf, $opt);
4019 } else {
4020
4021 die "internal error" if $opt !~ m/net(\d+)/;
4022 my $iface = "tap${vmid}i$1";
4023
4024 if (&$safe_num_ne($oldnet->{rate}, $newnet->{rate})) {
4025 PVE::Network::tap_rate_limit($iface, $newnet->{rate});
4026 }
4027
4028 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
4029 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
4030 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
4031 PVE::Network::tap_unplug($iface);
4032 PVE::Network::tap_plug($iface, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall});
4033 }
4034
4035 if (&$safe_string_ne($oldnet->{link_down}, $newnet->{link_down})) {
4036 qemu_set_link_status($vmid, $opt, !$newnet->{link_down});
4037 }
4038
4039 return 1;
4040 }
4041 }
4042
4043 if ($hotplug) {
4044 vm_deviceplug($storecfg, $conf, $vmid, $opt, $newnet);
4045 } else {
4046 die "skip\n";
4047 }
4048 }
4049
4050 sub vmconfig_update_disk {
4051 my ($storecfg, $conf, $hotplug, $vmid, $opt, $value, $force) = @_;
4052
4053 # fixme: do we need force?
4054
4055 my $drive = parse_drive($opt, $value);
4056
4057 if ($conf->{$opt}) {
4058
4059 if (my $old_drive = parse_drive($opt, $conf->{$opt})) {
4060
4061 my $media = $drive->{media} || 'disk';
4062 my $oldmedia = $old_drive->{media} || 'disk';
4063 die "unable to change media type\n" if $media ne $oldmedia;
4064
4065 if (!drive_is_cdrom($old_drive)) {
4066
4067 if ($drive->{file} ne $old_drive->{file}) {
4068
4069 die "skip\n" if !$hotplug;
4070
4071 # unplug and register as unused
4072 vm_deviceunplug($vmid, $conf, $opt);
4073 vmconfig_register_unused_drive($storecfg, $vmid, $conf, $old_drive)
4074
4075 } else {
4076 # update existing disk
4077
4078 # skip non hotpluggable value
4079 if (&$safe_num_ne($drive->{discard}, $old_drive->{discard}) ||
4080 &$safe_string_ne($drive->{iothread}, $old_drive->{iothread}) ||
4081 &$safe_string_ne($drive->{queues}, $old_drive->{queues}) ||
4082 &$safe_string_ne($drive->{cache}, $old_drive->{cache})) {
4083 die "skip\n";
4084 }
4085
4086 # apply throttle
4087 if (&$safe_num_ne($drive->{mbps}, $old_drive->{mbps}) ||
4088 &$safe_num_ne($drive->{mbps_rd}, $old_drive->{mbps_rd}) ||
4089 &$safe_num_ne($drive->{mbps_wr}, $old_drive->{mbps_wr}) ||
4090 &$safe_num_ne($drive->{iops}, $old_drive->{iops}) ||
4091 &$safe_num_ne($drive->{iops_rd}, $old_drive->{iops_rd}) ||
4092 &$safe_num_ne($drive->{iops_wr}, $old_drive->{iops_wr}) ||
4093 &$safe_num_ne($drive->{mbps_max}, $old_drive->{mbps_max}) ||
4094 &$safe_num_ne($drive->{mbps_rd_max}, $old_drive->{mbps_rd_max}) ||
4095 &$safe_num_ne($drive->{mbps_wr_max}, $old_drive->{mbps_wr_max}) ||
4096 &$safe_num_ne($drive->{iops_max}, $old_drive->{iops_max}) ||
4097 &$safe_num_ne($drive->{iops_rd_max}, $old_drive->{iops_rd_max}) ||
4098 &$safe_num_ne($drive->{iops_wr_max}, $old_drive->{iops_wr_max})) {
4099
4100 qemu_block_set_io_throttle($vmid,"drive-$opt",
4101 ($drive->{mbps} || 0)*1024*1024,
4102 ($drive->{mbps_rd} || 0)*1024*1024,
4103 ($drive->{mbps_wr} || 0)*1024*1024,
4104 $drive->{iops} || 0,
4105 $drive->{iops_rd} || 0,
4106 $drive->{iops_wr} || 0,
4107 ($drive->{mbps_max} || 0)*1024*1024,
4108 ($drive->{mbps_rd_max} || 0)*1024*1024,
4109 ($drive->{mbps_wr_max} || 0)*1024*1024,
4110 $drive->{iops_max} || 0,
4111 $drive->{iops_rd_max} || 0,
4112 $drive->{iops_wr_max} || 0);
4113
4114 }
4115
4116 return 1;
4117 }
4118
4119 } else { # cdrom
4120
4121 if ($drive->{file} eq 'none') {
4122 vm_mon_cmd($vmid, "eject",force => JSON::true,device => "drive-$opt");
4123 } else {
4124 my $path = get_iso_path($storecfg, $vmid, $drive->{file});
4125 vm_mon_cmd($vmid, "eject", force => JSON::true,device => "drive-$opt"); # force eject if locked
4126 vm_mon_cmd($vmid, "change", device => "drive-$opt",target => "$path") if $path;
4127 }
4128
4129 return 1;
4130 }
4131 }
4132 }
4133
4134 die "skip\n" if !$hotplug || $opt =~ m/(ide|sata)(\d+)/;
4135 # hotplug new disks
4136 vm_deviceplug($storecfg, $conf, $vmid, $opt, $drive);
4137 }
4138
4139 sub vm_start {
4140 my ($storecfg, $vmid, $statefile, $skiplock, $migratedfrom, $paused, $forcemachine, $spice_ticket) = @_;
4141
4142 lock_config($vmid, sub {
4143 my $conf = load_config($vmid, $migratedfrom);
4144
4145 die "you can't start a vm if it's a template\n" if is_template($conf);
4146
4147 check_lock($conf) if !$skiplock;
4148
4149 die "VM $vmid already running\n" if check_running($vmid, undef, $migratedfrom);
4150
4151 if (!$statefile && scalar(keys %{$conf->{pending}})) {
4152 vmconfig_apply_pending($vmid, $conf, $storecfg);
4153 $conf = load_config($vmid); # update/reload
4154 }
4155
4156 my $defaults = load_defaults();
4157
4158 # set environment variable useful inside network script
4159 $ENV{PVE_MIGRATED_FROM} = $migratedfrom if $migratedfrom;
4160
4161 my ($cmd, $vollist, $spice_port) = config_to_command($storecfg, $vmid, $conf, $defaults, $forcemachine);
4162
4163 my $migrate_port = 0;
4164 my $migrate_uri;
4165 if ($statefile) {
4166 if ($statefile eq 'tcp') {
4167 my $localip = "localhost";
4168 my $datacenterconf = PVE::Cluster::cfs_read_file('datacenter.cfg');
4169 my $nodename = PVE::INotify::nodename();
4170 if ($datacenterconf->{migration_unsecure}) {
4171 $localip = PVE::Cluster::remote_node_ip($nodename, 1);
4172 }
4173 my $pfamily = PVE::Tools::get_host_address_family($nodename);
4174 $migrate_port = PVE::Tools::next_migrate_port($pfamily);
4175 $migrate_uri = "tcp:[${localip}]:${migrate_port}";
4176 push @$cmd, '-incoming', $migrate_uri;
4177 push @$cmd, '-S';
4178 } else {
4179 push @$cmd, '-loadstate', $statefile;
4180 }
4181 } elsif ($paused) {
4182 push @$cmd, '-S';
4183 }
4184
4185 # host pci devices
4186 for (my $i = 0; $i < $MAX_HOSTPCI_DEVICES; $i++) {
4187 my $d = parse_hostpci($conf->{"hostpci$i"});
4188 next if !$d;
4189 my $pcidevices = $d->{pciid};
4190 foreach my $pcidevice (@$pcidevices) {
4191 my $pciid = $pcidevice->{id}.".".$pcidevice->{function};
4192
4193 my $info = pci_device_info("0000:$pciid");
4194 die "IOMMU not present\n" if !check_iommu_support();
4195 die "no pci device info for device '$pciid'\n" if !$info;
4196
4197 if ($d->{driver} && $d->{driver} eq "vfio") {
4198 die "can't unbind/bind pci group to vfio '$pciid'\n" if !pci_dev_group_bind_to_vfio($pciid);
4199 } else {
4200 die "can't unbind/bind to stub pci device '$pciid'\n" if !pci_dev_bind_to_stub($info);
4201 }
4202
4203 die "can't reset pci device '$pciid'\n" if $info->{has_fl_reset} and !pci_dev_reset($info);
4204 }
4205 }
4206
4207 PVE::Storage::activate_volumes($storecfg, $vollist);
4208
4209 eval { run_command($cmd, timeout => $statefile ? undef : 30,
4210 umask => 0077); };
4211 my $err = $@;
4212 die "start failed: $err" if $err;
4213
4214 print "migration listens on $migrate_uri\n" if $migrate_uri;
4215
4216 if ($statefile && $statefile ne 'tcp') {
4217 eval { vm_mon_cmd_nocheck($vmid, "cont"); };
4218 warn $@ if $@;
4219 }
4220
4221 if ($migratedfrom) {
4222
4223 eval {
4224 set_migration_caps($vmid);
4225 };
4226 warn $@ if $@;
4227
4228 if ($spice_port) {
4229 print "spice listens on port $spice_port\n";
4230 if ($spice_ticket) {
4231 vm_mon_cmd_nocheck($vmid, "set_password", protocol => 'spice', password => $spice_ticket);
4232 vm_mon_cmd_nocheck($vmid, "expire_password", protocol => 'spice', time => "+30");
4233 }
4234 }
4235
4236 } else {
4237
4238 if (!$statefile && (!defined($conf->{balloon}) || $conf->{balloon})) {
4239 vm_mon_cmd_nocheck($vmid, "balloon", value => $conf->{balloon}*1024*1024)
4240 if $conf->{balloon};
4241 }
4242
4243 foreach my $opt (keys %$conf) {
4244 next if $opt !~ m/^net\d+$/;
4245 my $nicconf = parse_net($conf->{$opt});
4246 qemu_set_link_status($vmid, $opt, 0) if $nicconf->{link_down};
4247 }
4248 }
4249
4250 vm_mon_cmd_nocheck($vmid, 'qom-set',
4251 path => "machine/peripheral/balloon0",
4252 property => "guest-stats-polling-interval",
4253 value => 2) if (!defined($conf->{balloon}) || $conf->{balloon});
4254
4255 });
4256 }
4257
4258 sub vm_mon_cmd {
4259 my ($vmid, $execute, %params) = @_;
4260
4261 my $cmd = { execute => $execute, arguments => \%params };
4262 vm_qmp_command($vmid, $cmd);
4263 }
4264
4265 sub vm_mon_cmd_nocheck {
4266 my ($vmid, $execute, %params) = @_;
4267
4268 my $cmd = { execute => $execute, arguments => \%params };
4269 vm_qmp_command($vmid, $cmd, 1);
4270 }
4271
4272 sub vm_qmp_command {
4273 my ($vmid, $cmd, $nocheck) = @_;
4274
4275 my $res;
4276
4277 my $timeout;
4278 if ($cmd->{arguments} && $cmd->{arguments}->{timeout}) {
4279 $timeout = $cmd->{arguments}->{timeout};
4280 delete $cmd->{arguments}->{timeout};
4281 }
4282
4283 eval {
4284 die "VM $vmid not running\n" if !check_running($vmid, $nocheck);
4285 my $sname = qmp_socket($vmid);
4286 if (-e $sname) { # test if VM is reasonambe new and supports qmp/qga
4287 my $qmpclient = PVE::QMPClient->new();
4288
4289 $res = $qmpclient->cmd($vmid, $cmd, $timeout);
4290 } elsif (-e "${var_run_tmpdir}/$vmid.mon") {
4291 die "can't execute complex command on old monitor - stop/start your vm to fix the problem\n"
4292 if scalar(%{$cmd->{arguments}});
4293 vm_monitor_command($vmid, $cmd->{execute}, $nocheck);
4294 } else {
4295 die "unable to open monitor socket\n";
4296 }
4297 };
4298 if (my $err = $@) {
4299 syslog("err", "VM $vmid qmp command failed - $err");
4300 die $err;
4301 }
4302
4303 return $res;
4304 }
4305
4306 sub vm_human_monitor_command {
4307 my ($vmid, $cmdline) = @_;
4308
4309 my $res;
4310
4311 my $cmd = {
4312 execute => 'human-monitor-command',
4313 arguments => { 'command-line' => $cmdline},
4314 };
4315
4316 return vm_qmp_command($vmid, $cmd);
4317 }
4318
4319 sub vm_commandline {
4320 my ($storecfg, $vmid) = @_;
4321
4322 my $conf = load_config($vmid);
4323
4324 my $defaults = load_defaults();
4325
4326 my $cmd = config_to_command($storecfg, $vmid, $conf, $defaults);
4327
4328 return join(' ', @$cmd);
4329 }
4330
4331 sub vm_reset {
4332 my ($vmid, $skiplock) = @_;
4333
4334 lock_config($vmid, sub {
4335
4336 my $conf = load_config($vmid);
4337
4338 check_lock($conf) if !$skiplock;
4339
4340 vm_mon_cmd($vmid, "system_reset");
4341 });
4342 }
4343
4344 sub get_vm_volumes {
4345 my ($conf) = @_;
4346
4347 my $vollist = [];
4348 foreach_volid($conf, sub {
4349 my ($volid, $is_cdrom) = @_;
4350
4351 return if $volid =~ m|^/|;
4352
4353 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
4354 return if !$sid;
4355
4356 push @$vollist, $volid;
4357 });
4358
4359 return $vollist;
4360 }
4361
4362 sub vm_stop_cleanup {
4363 my ($storecfg, $vmid, $conf, $keepActive, $apply_pending_changes) = @_;
4364
4365 eval {
4366
4367 if (!$keepActive) {
4368 my $vollist = get_vm_volumes($conf);
4369 PVE::Storage::deactivate_volumes($storecfg, $vollist);
4370 }
4371
4372 foreach my $ext (qw(mon qmp pid vnc qga)) {
4373 unlink "/var/run/qemu-server/${vmid}.$ext";
4374 }
4375
4376 vmconfig_apply_pending($vmid, $conf, $storecfg) if $apply_pending_changes;
4377 };
4378 warn $@ if $@; # avoid errors - just warn
4379 }
4380
4381 # Note: use $nockeck to skip tests if VM configuration file exists.
4382 # We need that when migration VMs to other nodes (files already moved)
4383 # Note: we set $keepActive in vzdump stop mode - volumes need to stay active
4384 sub vm_stop {
4385 my ($storecfg, $vmid, $skiplock, $nocheck, $timeout, $shutdown, $force, $keepActive, $migratedfrom) = @_;
4386
4387 $force = 1 if !defined($force) && !$shutdown;
4388
4389 if ($migratedfrom){
4390 my $pid = check_running($vmid, $nocheck, $migratedfrom);
4391 kill 15, $pid if $pid;
4392 my $conf = load_config($vmid, $migratedfrom);
4393 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 0);
4394 return;
4395 }
4396
4397 lock_config($vmid, sub {
4398
4399 my $pid = check_running($vmid, $nocheck);
4400 return if !$pid;
4401
4402 my $conf;
4403 if (!$nocheck) {
4404 $conf = load_config($vmid);
4405 check_lock($conf) if !$skiplock;
4406 if (!defined($timeout) && $shutdown && $conf->{startup}) {
4407 my $opts = PVE::JSONSchema::pve_parse_startup_order($conf->{startup});
4408 $timeout = $opts->{down} if $opts->{down};
4409 }
4410 }
4411
4412 $timeout = 60 if !defined($timeout);
4413
4414 eval {
4415 if ($shutdown) {
4416 if (defined($conf) && $conf->{agent}) {
4417 vm_qmp_command($vmid, { execute => "guest-shutdown" }, $nocheck);
4418 } else {
4419 vm_qmp_command($vmid, { execute => "system_powerdown" }, $nocheck);
4420 }
4421 } else {
4422 vm_qmp_command($vmid, { execute => "quit" }, $nocheck);
4423 }
4424 };
4425 my $err = $@;
4426
4427 if (!$err) {
4428 my $count = 0;
4429 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4430 $count++;
4431 sleep 1;
4432 }
4433
4434 if ($count >= $timeout) {
4435 if ($force) {
4436 warn "VM still running - terminating now with SIGTERM\n";
4437 kill 15, $pid;
4438 } else {
4439 die "VM quit/powerdown failed - got timeout\n";
4440 }
4441 } else {
4442 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4443 return;
4444 }
4445 } else {
4446 if ($force) {
4447 warn "VM quit/powerdown failed - terminating now with SIGTERM\n";
4448 kill 15, $pid;
4449 } else {
4450 die "VM quit/powerdown failed\n";
4451 }
4452 }
4453
4454 # wait again
4455 $timeout = 10;
4456
4457 my $count = 0;
4458 while (($count < $timeout) && check_running($vmid, $nocheck)) {
4459 $count++;
4460 sleep 1;
4461 }
4462
4463 if ($count >= $timeout) {
4464 warn "VM still running - terminating now with SIGKILL\n";
4465 kill 9, $pid;
4466 sleep 1;
4467 }
4468
4469 vm_stop_cleanup($storecfg, $vmid, $conf, $keepActive, 1) if $conf;
4470 });
4471 }
4472
4473 sub vm_suspend {
4474 my ($vmid, $skiplock) = @_;
4475
4476 lock_config($vmid, sub {
4477
4478 my $conf = load_config($vmid);
4479
4480 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4481
4482 vm_mon_cmd($vmid, "stop");
4483 });
4484 }
4485
4486 sub vm_resume {
4487 my ($vmid, $skiplock) = @_;
4488
4489 lock_config($vmid, sub {
4490
4491 my $conf = load_config($vmid);
4492
4493 check_lock($conf) if !($skiplock || ($conf->{lock} && $conf->{lock} eq 'backup'));
4494
4495 vm_mon_cmd($vmid, "cont");
4496 });
4497 }
4498
4499 sub vm_sendkey {
4500 my ($vmid, $skiplock, $key) = @_;
4501
4502 lock_config($vmid, sub {
4503
4504 my $conf = load_config($vmid);
4505
4506 # there is no qmp command, so we use the human monitor command
4507 vm_human_monitor_command($vmid, "sendkey $key");
4508 });
4509 }
4510
4511 sub vm_destroy {
4512 my ($storecfg, $vmid, $skiplock) = @_;
4513
4514 lock_config($vmid, sub {
4515
4516 my $conf = load_config($vmid);
4517
4518 check_lock($conf) if !$skiplock;
4519
4520 if (!check_running($vmid)) {
4521 destroy_vm($storecfg, $vmid);
4522 } else {
4523 die "VM $vmid is running - destroy failed\n";
4524 }
4525 });
4526 }
4527
4528 # pci helpers
4529
4530 sub file_write {
4531 my ($filename, $buf) = @_;
4532
4533 my $fh = IO::File->new($filename, "w");
4534 return undef if !$fh;
4535
4536 my $res = print $fh $buf;
4537
4538 $fh->close();
4539
4540 return $res;
4541 }
4542
4543 sub pci_device_info {
4544 my ($name) = @_;
4545
4546 my $res;
4547
4548 return undef if $name !~ m/^([a-f0-9]{4}):([a-f0-9]{2}):([a-f0-9]{2})\.([a-f0-9])$/;
4549 my ($domain, $bus, $slot, $func) = ($1, $2, $3, $4);
4550
4551 my $irq = file_read_firstline("$pcisysfs/devices/$name/irq");
4552 return undef if !defined($irq) || $irq !~ m/^\d+$/;
4553
4554 my $vendor = file_read_firstline("$pcisysfs/devices/$name/vendor");
4555 return undef if !defined($vendor) || $vendor !~ s/^0x//;
4556
4557 my $product = file_read_firstline("$pcisysfs/devices/$name/device");
4558 return undef if !defined($product) || $product !~ s/^0x//;
4559
4560 $res = {
4561 name => $name,
4562 vendor => $vendor,
4563 product => $product,
4564 domain => $domain,
4565 bus => $bus,
4566 slot => $slot,
4567 func => $func,
4568 irq => $irq,
4569 has_fl_reset => -f "$pcisysfs/devices/$name/reset" || 0,
4570 };
4571
4572 return $res;
4573 }
4574
4575 sub pci_dev_reset {
4576 my ($dev) = @_;
4577
4578 my $name = $dev->{name};
4579
4580 my $fn = "$pcisysfs/devices/$name/reset";
4581
4582 return file_write($fn, "1");
4583 }
4584
4585 sub pci_dev_bind_to_stub {
4586 my ($dev) = @_;
4587
4588 my $name = $dev->{name};
4589
4590 my $testdir = "$pcisysfs/drivers/pci-stub/$name";
4591 return 1 if -d $testdir;
4592
4593 my $data = "$dev->{vendor} $dev->{product}";
4594 return undef if !file_write("$pcisysfs/drivers/pci-stub/new_id", $data);
4595
4596 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4597 if (!file_write($fn, $name)) {
4598 return undef if -f $fn;
4599 }
4600
4601 $fn = "$pcisysfs/drivers/pci-stub/bind";
4602 if (! -d $testdir) {
4603 return undef if !file_write($fn, $name);
4604 }
4605
4606 return -d $testdir;
4607 }
4608
4609 sub pci_dev_bind_to_vfio {
4610 my ($dev) = @_;
4611
4612 my $name = $dev->{name};
4613
4614 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4615
4616 if (!-d $vfio_basedir) {
4617 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4618 }
4619 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4620
4621 my $testdir = "$vfio_basedir/$name";
4622 return 1 if -d $testdir;
4623
4624 my $data = "$dev->{vendor} $dev->{product}";
4625 return undef if !file_write("$vfio_basedir/new_id", $data);
4626
4627 my $fn = "$pcisysfs/devices/$name/driver/unbind";
4628 if (!file_write($fn, $name)) {
4629 return undef if -f $fn;
4630 }
4631
4632 $fn = "$vfio_basedir/bind";
4633 if (! -d $testdir) {
4634 return undef if !file_write($fn, $name);
4635 }
4636
4637 return -d $testdir;
4638 }
4639
4640 sub pci_dev_group_bind_to_vfio {
4641 my ($pciid) = @_;
4642
4643 my $vfio_basedir = "$pcisysfs/drivers/vfio-pci";
4644
4645 if (!-d $vfio_basedir) {
4646 system("/sbin/modprobe vfio-pci >/dev/null 2>/dev/null");
4647 }
4648 die "Cannot find vfio-pci module!\n" if !-d $vfio_basedir;
4649
4650 # get IOMMU group devices
4651 opendir(my $D, "$pcisysfs/devices/0000:$pciid/iommu_group/devices/") || die "Cannot open iommu_group: $!\n";
4652 my @devs = grep /^0000:/, readdir($D);
4653 closedir($D);
4654
4655 foreach my $pciid (@devs) {
4656 $pciid =~ m/^([:\.\da-f]+)$/ or die "PCI ID $pciid not valid!\n";
4657
4658 # pci bridges, switches or root ports are not supported
4659 # they have a pci_bus subdirectory so skip them
4660 next if (-e "$pcisysfs/devices/$pciid/pci_bus");
4661
4662 my $info = pci_device_info($1);
4663 pci_dev_bind_to_vfio($info) || die "Cannot bind $pciid to vfio\n";
4664 }
4665
4666 return 1;
4667 }
4668
4669 sub print_pci_addr {
4670 my ($id, $bridges) = @_;
4671
4672 my $res = '';
4673 my $devices = {
4674 piix3 => { bus => 0, addr => 1 },
4675 #addr2 : first videocard
4676 balloon0 => { bus => 0, addr => 3 },
4677 watchdog => { bus => 0, addr => 4 },
4678 scsihw0 => { bus => 0, addr => 5 },
4679 'pci.3' => { bus => 0, addr => 5 }, #can also be used for virtio-scsi-single bridge
4680 scsihw1 => { bus => 0, addr => 6 },
4681 ahci0 => { bus => 0, addr => 7 },
4682 qga0 => { bus => 0, addr => 8 },
4683 spice => { bus => 0, addr => 9 },
4684 virtio0 => { bus => 0, addr => 10 },
4685 virtio1 => { bus => 0, addr => 11 },
4686 virtio2 => { bus => 0, addr => 12 },
4687 virtio3 => { bus => 0, addr => 13 },
4688 virtio4 => { bus => 0, addr => 14 },
4689 virtio5 => { bus => 0, addr => 15 },
4690 hostpci0 => { bus => 0, addr => 16 },
4691 hostpci1 => { bus => 0, addr => 17 },
4692 net0 => { bus => 0, addr => 18 },
4693 net1 => { bus => 0, addr => 19 },
4694 net2 => { bus => 0, addr => 20 },
4695 net3 => { bus => 0, addr => 21 },
4696 net4 => { bus => 0, addr => 22 },
4697 net5 => { bus => 0, addr => 23 },
4698 vga1 => { bus => 0, addr => 24 },
4699 vga2 => { bus => 0, addr => 25 },
4700 vga3 => { bus => 0, addr => 26 },
4701 hostpci2 => { bus => 0, addr => 27 },
4702 hostpci3 => { bus => 0, addr => 28 },
4703 #addr29 : usb-host (pve-usb.cfg)
4704 'pci.1' => { bus => 0, addr => 30 },
4705 'pci.2' => { bus => 0, addr => 31 },
4706 'net6' => { bus => 1, addr => 1 },
4707 'net7' => { bus => 1, addr => 2 },
4708 'net8' => { bus => 1, addr => 3 },
4709 'net9' => { bus => 1, addr => 4 },
4710 'net10' => { bus => 1, addr => 5 },
4711 'net11' => { bus => 1, addr => 6 },
4712 'net12' => { bus => 1, addr => 7 },
4713 'net13' => { bus => 1, addr => 8 },
4714 'net14' => { bus => 1, addr => 9 },
4715 'net15' => { bus => 1, addr => 10 },
4716 'net16' => { bus => 1, addr => 11 },
4717 'net17' => { bus => 1, addr => 12 },
4718 'net18' => { bus => 1, addr => 13 },
4719 'net19' => { bus => 1, addr => 14 },
4720 'net20' => { bus => 1, addr => 15 },
4721 'net21' => { bus => 1, addr => 16 },
4722 'net22' => { bus => 1, addr => 17 },
4723 'net23' => { bus => 1, addr => 18 },
4724 'net24' => { bus => 1, addr => 19 },
4725 'net25' => { bus => 1, addr => 20 },
4726 'net26' => { bus => 1, addr => 21 },
4727 'net27' => { bus => 1, addr => 22 },
4728 'net28' => { bus => 1, addr => 23 },
4729 'net29' => { bus => 1, addr => 24 },
4730 'net30' => { bus => 1, addr => 25 },
4731 'net31' => { bus => 1, addr => 26 },
4732 'virtio6' => { bus => 2, addr => 1 },
4733 'virtio7' => { bus => 2, addr => 2 },
4734 'virtio8' => { bus => 2, addr => 3 },
4735 'virtio9' => { bus => 2, addr => 4 },
4736 'virtio10' => { bus => 2, addr => 5 },
4737 'virtio11' => { bus => 2, addr => 6 },
4738 'virtio12' => { bus => 2, addr => 7 },
4739 'virtio13' => { bus => 2, addr => 8 },
4740 'virtio14' => { bus => 2, addr => 9 },
4741 'virtio15' => { bus => 2, addr => 10 },
4742 'virtioscsi0' => { bus => 3, addr => 1 },
4743 'virtioscsi1' => { bus => 3, addr => 2 },
4744 'virtioscsi2' => { bus => 3, addr => 3 },
4745 'virtioscsi3' => { bus => 3, addr => 4 },
4746 'virtioscsi4' => { bus => 3, addr => 5 },
4747 'virtioscsi5' => { bus => 3, addr => 6 },
4748 'virtioscsi6' => { bus => 3, addr => 7 },
4749 'virtioscsi7' => { bus => 3, addr => 8 },
4750 'virtioscsi8' => { bus => 3, addr => 9 },
4751 'virtioscsi9' => { bus => 3, addr => 10 },
4752 'virtioscsi10' => { bus => 3, addr => 11 },
4753 'virtioscsi11' => { bus => 3, addr => 12 },
4754 'virtioscsi12' => { bus => 3, addr => 13 },
4755 'virtioscsi13' => { bus => 3, addr => 14 },
4756 'virtioscsi14' => { bus => 3, addr => 15 },
4757 'virtioscsi15' => { bus => 3, addr => 16 },
4758 'virtioscsi16' => { bus => 3, addr => 17 },
4759 'virtioscsi17' => { bus => 3, addr => 18 },
4760 'virtioscsi18' => { bus => 3, addr => 19 },
4761 'virtioscsi19' => { bus => 3, addr => 20 },
4762 'virtioscsi20' => { bus => 3, addr => 21 },
4763 'virtioscsi21' => { bus => 3, addr => 22 },
4764 'virtioscsi22' => { bus => 3, addr => 23 },
4765 'virtioscsi23' => { bus => 3, addr => 24 },
4766 'virtioscsi24' => { bus => 3, addr => 25 },
4767 'virtioscsi25' => { bus => 3, addr => 26 },
4768 'virtioscsi26' => { bus => 3, addr => 27 },
4769 'virtioscsi27' => { bus => 3, addr => 28 },
4770 'virtioscsi28' => { bus => 3, addr => 29 },
4771 'virtioscsi29' => { bus => 3, addr => 30 },
4772 'virtioscsi30' => { bus => 3, addr => 31 },
4773
4774 };
4775
4776 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4777 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4778 my $bus = $devices->{$id}->{bus};
4779 $res = ",bus=pci.$bus,addr=$addr";
4780 $bridges->{$bus} = 1 if $bridges;
4781 }
4782 return $res;
4783
4784 }
4785
4786 sub print_pcie_addr {
4787 my ($id) = @_;
4788
4789 my $res = '';
4790 my $devices = {
4791 hostpci0 => { bus => "ich9-pcie-port-1", addr => 0 },
4792 hostpci1 => { bus => "ich9-pcie-port-2", addr => 0 },
4793 hostpci2 => { bus => "ich9-pcie-port-3", addr => 0 },
4794 hostpci3 => { bus => "ich9-pcie-port-4", addr => 0 },
4795 };
4796
4797 if (defined($devices->{$id}->{bus}) && defined($devices->{$id}->{addr})) {
4798 my $addr = sprintf("0x%x", $devices->{$id}->{addr});
4799 my $bus = $devices->{$id}->{bus};
4800 $res = ",bus=$bus,addr=$addr";
4801 }
4802 return $res;
4803
4804 }
4805
4806 # vzdump restore implementaion
4807
4808 sub tar_archive_read_firstfile {
4809 my $archive = shift;
4810
4811 die "ERROR: file '$archive' does not exist\n" if ! -f $archive;
4812
4813 # try to detect archive type first
4814 my $pid = open (TMP, "tar tf '$archive'|") ||
4815 die "unable to open file '$archive'\n";
4816 my $firstfile = <TMP>;
4817 kill 15, $pid;
4818 close TMP;
4819
4820 die "ERROR: archive contaions no data\n" if !$firstfile;
4821 chomp $firstfile;
4822
4823 return $firstfile;
4824 }
4825
4826 sub tar_restore_cleanup {
4827 my ($storecfg, $statfile) = @_;
4828
4829 print STDERR "starting cleanup\n";
4830
4831 if (my $fd = IO::File->new($statfile, "r")) {
4832 while (defined(my $line = <$fd>)) {
4833 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
4834 my $volid = $2;
4835 eval {
4836 if ($volid =~ m|^/|) {
4837 unlink $volid || die 'unlink failed\n';
4838 } else {
4839 PVE::Storage::vdisk_free($storecfg, $volid);
4840 }
4841 print STDERR "temporary volume '$volid' sucessfuly removed\n";
4842 };
4843 print STDERR "unable to cleanup '$volid' - $@" if $@;
4844 } else {
4845 print STDERR "unable to parse line in statfile - $line";
4846 }
4847 }
4848 $fd->close();
4849 }
4850 }
4851
4852 sub restore_archive {
4853 my ($archive, $vmid, $user, $opts) = @_;
4854
4855 my $format = $opts->{format};
4856 my $comp;
4857
4858 if ($archive =~ m/\.tgz$/ || $archive =~ m/\.tar\.gz$/) {
4859 $format = 'tar' if !$format;
4860 $comp = 'gzip';
4861 } elsif ($archive =~ m/\.tar$/) {
4862 $format = 'tar' if !$format;
4863 } elsif ($archive =~ m/.tar.lzo$/) {
4864 $format = 'tar' if !$format;
4865 $comp = 'lzop';
4866 } elsif ($archive =~ m/\.vma$/) {
4867 $format = 'vma' if !$format;
4868 } elsif ($archive =~ m/\.vma\.gz$/) {
4869 $format = 'vma' if !$format;
4870 $comp = 'gzip';
4871 } elsif ($archive =~ m/\.vma\.lzo$/) {
4872 $format = 'vma' if !$format;
4873 $comp = 'lzop';
4874 } else {
4875 $format = 'vma' if !$format; # default
4876 }
4877
4878 # try to detect archive format
4879 if ($format eq 'tar') {
4880 return restore_tar_archive($archive, $vmid, $user, $opts);
4881 } else {
4882 return restore_vma_archive($archive, $vmid, $user, $opts, $comp);
4883 }
4884 }
4885
4886 sub restore_update_config_line {
4887 my ($outfd, $cookie, $vmid, $map, $line, $unique) = @_;
4888
4889 return if $line =~ m/^\#qmdump\#/;
4890 return if $line =~ m/^\#vzdump\#/;
4891 return if $line =~ m/^lock:/;
4892 return if $line =~ m/^unused\d+:/;
4893 return if $line =~ m/^parent:/;
4894 return if $line =~ m/^template:/; # restored VM is never a template
4895
4896 if (($line =~ m/^(vlan(\d+)):\s*(\S+)\s*$/)) {
4897 # try to convert old 1.X settings
4898 my ($id, $ind, $ethcfg) = ($1, $2, $3);
4899 foreach my $devconfig (PVE::Tools::split_list($ethcfg)) {
4900 my ($model, $macaddr) = split(/\=/, $devconfig);
4901 $macaddr = PVE::Tools::random_ether_addr() if !$macaddr || $unique;
4902 my $net = {
4903 model => $model,
4904 bridge => "vmbr$ind",
4905 macaddr => $macaddr,
4906 };
4907 my $netstr = print_net($net);
4908
4909 print $outfd "net$cookie->{netcount}: $netstr\n";
4910 $cookie->{netcount}++;
4911 }
4912 } elsif (($line =~ m/^(net\d+):\s*(\S+)\s*$/) && $unique) {
4913 my ($id, $netstr) = ($1, $2);
4914 my $net = parse_net($netstr);
4915 $net->{macaddr} = PVE::Tools::random_ether_addr() if $net->{macaddr};
4916 $netstr = print_net($net);
4917 print $outfd "$id: $netstr\n";
4918 } elsif ($line =~ m/^((ide|scsi|virtio|sata)\d+):\s*(\S+)\s*$/) {
4919 my $virtdev = $1;
4920 my $value = $3;
4921 if ($line =~ m/backup=no/) {
4922 print $outfd "#$line";
4923 } elsif ($virtdev && $map->{$virtdev}) {
4924 my $di = parse_drive($virtdev, $value);
4925 delete $di->{format}; # format can change on restore
4926 $di->{file} = $map->{$virtdev};
4927 $value = print_drive($vmid, $di);
4928 print $outfd "$virtdev: $value\n";
4929 } else {
4930 print $outfd $line;
4931 }
4932 } else {
4933 print $outfd $line;
4934 }
4935 }
4936
4937 sub scan_volids {
4938 my ($cfg, $vmid) = @_;
4939
4940 my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
4941
4942 my $volid_hash = {};
4943 foreach my $storeid (keys %$info) {
4944 foreach my $item (@{$info->{$storeid}}) {
4945 next if !($item->{volid} && $item->{size});
4946 $item->{path} = PVE::Storage::path($cfg, $item->{volid});
4947 $volid_hash->{$item->{volid}} = $item;
4948 }
4949 }
4950
4951 return $volid_hash;
4952 }
4953
4954 sub get_used_paths {
4955 my ($vmid, $storecfg, $conf, $scan_snapshots, $skip_drive) = @_;
4956
4957 my $used_path = {};
4958
4959 my $scan_config = sub {
4960 my ($cref, $snapname) = @_;
4961
4962 foreach my $key (keys %$cref) {
4963 my $value = $cref->{$key};
4964 if (valid_drivename($key)) {
4965 next if $skip_drive && $key eq $skip_drive;
4966 my $drive = parse_drive($key, $value);
4967 next if !$drive || !$drive->{file} || drive_is_cdrom($drive);
4968 if ($drive->{file} =~ m!^/!) {
4969 $used_path->{$drive->{file}}++; # = 1;
4970 } else {
4971 my ($storeid, $volname) = PVE::Storage::parse_volume_id($drive->{file}, 1);
4972 next if !$storeid;
4973 my $scfg = PVE::Storage::storage_config($storecfg, $storeid, 1);
4974 next if !$scfg;
4975 my $path = PVE::Storage::path($storecfg, $drive->{file}, $snapname);
4976 $used_path->{$path}++; # = 1;
4977 }
4978 }
4979 }
4980 };
4981
4982 &$scan_config($conf);
4983
4984 undef $skip_drive;
4985
4986 if ($scan_snapshots) {
4987 foreach my $snapname (keys %{$conf->{snapshots}}) {
4988 &$scan_config($conf->{snapshots}->{$snapname}, $snapname);
4989 }
4990 }
4991
4992 return $used_path;
4993 }
4994
4995 sub update_disksize {
4996 my ($vmid, $conf, $volid_hash) = @_;
4997
4998 my $changes;
4999
5000 my $used = {};
5001
5002 # Note: it is allowed to define multiple storages with same path (alias), so
5003 # we need to check both 'volid' and real 'path' (two different volid can point
5004 # to the same path).
5005
5006 my $usedpath = {};
5007
5008 # update size info
5009 foreach my $opt (keys %$conf) {
5010 if (valid_drivename($opt)) {
5011 my $drive = parse_drive($opt, $conf->{$opt});
5012 my $volid = $drive->{file};
5013 next if !$volid;
5014
5015 $used->{$volid} = 1;
5016 if ($volid_hash->{$volid} &&
5017 (my $path = $volid_hash->{$volid}->{path})) {
5018 $usedpath->{$path} = 1;
5019 }
5020
5021 next if drive_is_cdrom($drive);
5022 next if !$volid_hash->{$volid};
5023
5024 $drive->{size} = $volid_hash->{$volid}->{size};
5025 my $new = print_drive($vmid, $drive);
5026 if ($new ne $conf->{$opt}) {
5027 $changes = 1;
5028 $conf->{$opt} = $new;
5029 }
5030 }
5031 }
5032
5033 # remove 'unusedX' entry if volume is used
5034 foreach my $opt (keys %$conf) {
5035 next if $opt !~ m/^unused\d+$/;
5036 my $volid = $conf->{$opt};
5037 my $path = $volid_hash->{$volid}->{path} if $volid_hash->{$volid};
5038 if ($used->{$volid} || ($path && $usedpath->{$path})) {
5039 $changes = 1;
5040 delete $conf->{$opt};
5041 }
5042 }
5043
5044 foreach my $volid (sort keys %$volid_hash) {
5045 next if $volid =~ m/vm-$vmid-state-/;
5046 next if $used->{$volid};
5047 my $path = $volid_hash->{$volid}->{path};
5048 next if !$path; # just to be sure
5049 next if $usedpath->{$path};
5050 $changes = 1;
5051 add_unused_volume($conf, $volid);
5052 $usedpath->{$path} = 1; # avoid to add more than once (aliases)
5053 }
5054
5055 return $changes;
5056 }
5057
5058 sub rescan {
5059 my ($vmid, $nolock) = @_;
5060
5061 my $cfg = PVE::Cluster::cfs_read_file("storage.cfg");
5062
5063 my $volid_hash = scan_volids($cfg, $vmid);
5064
5065 my $updatefn = sub {
5066 my ($vmid) = @_;
5067
5068 my $conf = load_config($vmid);
5069
5070 check_lock($conf);
5071
5072 my $vm_volids = {};
5073 foreach my $volid (keys %$volid_hash) {
5074 my $info = $volid_hash->{$volid};
5075 $vm_volids->{$volid} = $info if $info->{vmid} && $info->{vmid} == $vmid;
5076 }
5077
5078 my $changes = update_disksize($vmid, $conf, $vm_volids);
5079
5080 update_config_nolock($vmid, $conf, 1) if $changes;
5081 };
5082
5083 if (defined($vmid)) {
5084 if ($nolock) {
5085 &$updatefn($vmid);
5086 } else {
5087 lock_config($vmid, $updatefn, $vmid);
5088 }
5089 } else {
5090 my $vmlist = config_list();
5091 foreach my $vmid (keys %$vmlist) {
5092 if ($nolock) {
5093 &$updatefn($vmid);
5094 } else {
5095 lock_config($vmid, $updatefn, $vmid);
5096 }
5097 }
5098 }
5099 }
5100
5101 sub restore_vma_archive {
5102 my ($archive, $vmid, $user, $opts, $comp) = @_;
5103
5104 my $input = $archive eq '-' ? "<&STDIN" : undef;
5105 my $readfrom = $archive;
5106
5107 my $uncomp = '';
5108 if ($comp) {
5109 $readfrom = '-';
5110 my $qarchive = PVE::Tools::shellquote($archive);
5111 if ($comp eq 'gzip') {
5112 $uncomp = "zcat $qarchive|";
5113 } elsif ($comp eq 'lzop') {
5114 $uncomp = "lzop -d -c $qarchive|";
5115 } else {
5116 die "unknown compression method '$comp'\n";
5117 }
5118
5119 }
5120
5121 my $tmpdir = "/var/tmp/vzdumptmp$$";
5122 rmtree $tmpdir;
5123
5124 # disable interrupts (always do cleanups)
5125 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5126 warn "got interrupt - ignored\n";
5127 };
5128
5129 my $mapfifo = "/var/tmp/vzdumptmp$$.fifo";
5130 POSIX::mkfifo($mapfifo, 0600);
5131 my $fifofh;
5132
5133 my $openfifo = sub {
5134 open($fifofh, '>', $mapfifo) || die $!;
5135 };
5136
5137 my $cmd = "${uncomp}vma extract -v -r $mapfifo $readfrom $tmpdir";
5138
5139 my $oldtimeout;
5140 my $timeout = 5;
5141
5142 my $devinfo = {};
5143
5144 my $rpcenv = PVE::RPCEnvironment::get();
5145
5146 my $conffile = config_file($vmid);
5147 my $tmpfn = "$conffile.$$.tmp";
5148
5149 # Note: $oldconf is undef if VM does not exists
5150 my $oldconf = PVE::Cluster::cfs_read_file(cfs_config_path($vmid));
5151
5152 my $print_devmap = sub {
5153 my $virtdev_hash = {};
5154
5155 my $cfgfn = "$tmpdir/qemu-server.conf";
5156
5157 # we can read the config - that is already extracted
5158 my $fh = IO::File->new($cfgfn, "r") ||
5159 "unable to read qemu-server.conf - $!\n";
5160
5161 while (defined(my $line = <$fh>)) {
5162 if ($line =~ m/^\#qmdump\#map:(\S+):(\S+):(\S*):(\S*):$/) {
5163 my ($virtdev, $devname, $storeid, $format) = ($1, $2, $3, $4);
5164 die "archive does not contain data for drive '$virtdev'\n"
5165 if !$devinfo->{$devname};
5166 if (defined($opts->{storage})) {
5167 $storeid = $opts->{storage} || 'local';
5168 } elsif (!$storeid) {
5169 $storeid = 'local';
5170 }
5171 $format = 'raw' if !$format;
5172 $devinfo->{$devname}->{devname} = $devname;
5173 $devinfo->{$devname}->{virtdev} = $virtdev;
5174 $devinfo->{$devname}->{format} = $format;
5175 $devinfo->{$devname}->{storeid} = $storeid;
5176
5177 # check permission on storage
5178 my $pool = $opts->{pool}; # todo: do we need that?
5179 if ($user ne 'root@pam') {
5180 $rpcenv->check($user, "/storage/$storeid", ['Datastore.AllocateSpace']);
5181 }
5182
5183 $virtdev_hash->{$virtdev} = $devinfo->{$devname};
5184 }
5185 }
5186
5187 foreach my $devname (keys %$devinfo) {
5188 die "found no device mapping information for device '$devname'\n"
5189 if !$devinfo->{$devname}->{virtdev};
5190 }
5191
5192 my $cfg = cfs_read_file('storage.cfg');
5193
5194 # create empty/temp config
5195 if ($oldconf) {
5196 PVE::Tools::file_set_contents($conffile, "memory: 128\n");
5197 foreach_drive($oldconf, sub {
5198 my ($ds, $drive) = @_;
5199
5200 return if drive_is_cdrom($drive);
5201
5202 my $volid = $drive->{file};
5203
5204 return if !$volid || $volid =~ m|^/|;
5205
5206 my ($path, $owner) = PVE::Storage::path($cfg, $volid);
5207 return if !$path || !$owner || ($owner != $vmid);
5208
5209 # Note: only delete disk we want to restore
5210 # other volumes will become unused
5211 if ($virtdev_hash->{$ds}) {
5212 PVE::Storage::vdisk_free($cfg, $volid);
5213 }
5214 });
5215 }
5216
5217 my $map = {};
5218 foreach my $virtdev (sort keys %$virtdev_hash) {
5219 my $d = $virtdev_hash->{$virtdev};
5220 my $alloc_size = int(($d->{size} + 1024 - 1)/1024);
5221 my $scfg = PVE::Storage::storage_config($cfg, $d->{storeid});
5222
5223 # test if requested format is supported
5224 my ($defFormat, $validFormats) = PVE::Storage::storage_default_format($cfg, $d->{storeid});
5225 my $supported = grep { $_ eq $d->{format} } @$validFormats;
5226 $d->{format} = $defFormat if !$supported;
5227
5228 my $volid = PVE::Storage::vdisk_alloc($cfg, $d->{storeid}, $vmid,
5229 $d->{format}, undef, $alloc_size);
5230 print STDERR "new volume ID is '$volid'\n";
5231 $d->{volid} = $volid;
5232 my $path = PVE::Storage::path($cfg, $volid);
5233
5234 my $write_zeros = 1;
5235 # fixme: what other storages types initialize volumes with zero?
5236 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs' || $scfg->{type} eq 'glusterfs' ||
5237 $scfg->{type} eq 'sheepdog' || $scfg->{type} eq 'rbd') {
5238 $write_zeros = 0;
5239 }
5240
5241 print $fifofh "${write_zeros}:$d->{devname}=$path\n";
5242
5243 print "map '$d->{devname}' to '$path' (write zeros = ${write_zeros})\n";
5244 $map->{$virtdev} = $volid;
5245 }
5246
5247 $fh->seek(0, 0) || die "seek failed - $!\n";
5248
5249 my $outfd = new IO::File ($tmpfn, "w") ||
5250 die "unable to write config for VM $vmid\n";
5251
5252 my $cookie = { netcount => 0 };
5253 while (defined(my $line = <$fh>)) {
5254 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5255 }
5256
5257 $fh->close();
5258 $outfd->close();
5259 };
5260
5261 eval {
5262 # enable interrupts
5263 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5264 die "interrupted by signal\n";
5265 };
5266 local $SIG{ALRM} = sub { die "got timeout\n"; };
5267
5268 $oldtimeout = alarm($timeout);
5269
5270 my $parser = sub {
5271 my $line = shift;
5272
5273 print "$line\n";
5274
5275 if ($line =~ m/^DEV:\sdev_id=(\d+)\ssize:\s(\d+)\sdevname:\s(\S+)$/) {
5276 my ($dev_id, $size, $devname) = ($1, $2, $3);
5277 $devinfo->{$devname} = { size => $size, dev_id => $dev_id };
5278 } elsif ($line =~ m/^CTIME: /) {
5279 # we correctly received the vma config, so we can disable
5280 # the timeout now for disk allocation (set to 10 minutes, so
5281 # that we always timeout if something goes wrong)
5282 alarm(600);
5283 &$print_devmap();
5284 print $fifofh "done\n";
5285 my $tmp = $oldtimeout || 0;
5286 $oldtimeout = undef;
5287 alarm($tmp);
5288 close($fifofh);
5289 }
5290 };
5291
5292 print "restore vma archive: $cmd\n";
5293 run_command($cmd, input => $input, outfunc => $parser, afterfork => $openfifo);
5294 };
5295 my $err = $@;
5296
5297 alarm($oldtimeout) if $oldtimeout;
5298
5299 unlink $mapfifo;
5300
5301 if ($err) {
5302 rmtree $tmpdir;
5303 unlink $tmpfn;
5304
5305 my $cfg = cfs_read_file('storage.cfg');
5306 foreach my $devname (keys %$devinfo) {
5307 my $volid = $devinfo->{$devname}->{volid};
5308 next if !$volid;
5309 eval {
5310 if ($volid =~ m|^/|) {
5311 unlink $volid || die 'unlink failed\n';
5312 } else {
5313 PVE::Storage::vdisk_free($cfg, $volid);
5314 }
5315 print STDERR "temporary volume '$volid' sucessfuly removed\n";
5316 };
5317 print STDERR "unable to cleanup '$volid' - $@" if $@;
5318 }
5319 die $err;
5320 }
5321
5322 rmtree $tmpdir;
5323
5324 rename($tmpfn, $conffile) ||
5325 die "unable to commit configuration file '$conffile'\n";
5326
5327 PVE::Cluster::cfs_update(); # make sure we read new file
5328
5329 eval { rescan($vmid, 1); };
5330 warn $@ if $@;
5331 }
5332
5333 sub restore_tar_archive {
5334 my ($archive, $vmid, $user, $opts) = @_;
5335
5336 if ($archive ne '-') {
5337 my $firstfile = tar_archive_read_firstfile($archive);
5338 die "ERROR: file '$archive' dos not lock like a QemuServer vzdump backup\n"
5339 if $firstfile ne 'qemu-server.conf';
5340 }
5341
5342 my $storecfg = cfs_read_file('storage.cfg');
5343
5344 # destroy existing data - keep empty config
5345 my $vmcfgfn = config_file($vmid);
5346 destroy_vm($storecfg, $vmid, 1) if -f $vmcfgfn;
5347
5348 my $tocmd = "/usr/lib/qemu-server/qmextract";
5349
5350 $tocmd .= " --storage " . PVE::Tools::shellquote($opts->{storage}) if $opts->{storage};
5351 $tocmd .= " --pool " . PVE::Tools::shellquote($opts->{pool}) if $opts->{pool};
5352 $tocmd .= ' --prealloc' if $opts->{prealloc};
5353 $tocmd .= ' --info' if $opts->{info};
5354
5355 # tar option "xf" does not autodetect compression when read from STDIN,
5356 # so we pipe to zcat
5357 my $cmd = "zcat -f|tar xf " . PVE::Tools::shellquote($archive) . " " .
5358 PVE::Tools::shellquote("--to-command=$tocmd");
5359
5360 my $tmpdir = "/var/tmp/vzdumptmp$$";
5361 mkpath $tmpdir;
5362
5363 local $ENV{VZDUMP_TMPDIR} = $tmpdir;
5364 local $ENV{VZDUMP_VMID} = $vmid;
5365 local $ENV{VZDUMP_USER} = $user;
5366
5367 my $conffile = config_file($vmid);
5368 my $tmpfn = "$conffile.$$.tmp";
5369
5370 # disable interrupts (always do cleanups)
5371 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = sub {
5372 print STDERR "got interrupt - ignored\n";
5373 };
5374
5375 eval {
5376 # enable interrupts
5377 local $SIG{INT} = $SIG{TERM} = $SIG{QUIT} = $SIG{HUP} = $SIG{PIPE} = sub {
5378 die "interrupted by signal\n";
5379 };
5380
5381 if ($archive eq '-') {
5382 print "extracting archive from STDIN\n";
5383 run_command($cmd, input => "<&STDIN");
5384 } else {
5385 print "extracting archive '$archive'\n";
5386 run_command($cmd);
5387 }
5388
5389 return if $opts->{info};
5390
5391 # read new mapping
5392 my $map = {};
5393 my $statfile = "$tmpdir/qmrestore.stat";
5394 if (my $fd = IO::File->new($statfile, "r")) {
5395 while (defined (my $line = <$fd>)) {
5396 if ($line =~ m/vzdump:([^\s:]*):(\S+)$/) {
5397 $map->{$1} = $2 if $1;
5398 } else {
5399 print STDERR "unable to parse line in statfile - $line\n";
5400 }
5401 }
5402 $fd->close();
5403 }
5404
5405 my $confsrc = "$tmpdir/qemu-server.conf";
5406
5407 my $srcfd = new IO::File($confsrc, "r") ||
5408 die "unable to open file '$confsrc'\n";
5409
5410 my $outfd = new IO::File ($tmpfn, "w") ||
5411 die "unable to write config for VM $vmid\n";
5412
5413 my $cookie = { netcount => 0 };
5414 while (defined (my $line = <$srcfd>)) {
5415 restore_update_config_line($outfd, $cookie, $vmid, $map, $line, $opts->{unique});
5416 }
5417
5418 $srcfd->close();
5419 $outfd->close();
5420 };
5421 my $err = $@;
5422
5423 if ($err) {
5424
5425 unlink $tmpfn;
5426
5427 tar_restore_cleanup($storecfg, "$tmpdir/qmrestore.stat") if !$opts->{info};
5428
5429 die $err;
5430 }
5431
5432 rmtree $tmpdir;
5433
5434 rename $tmpfn, $conffile ||
5435 die "unable to commit configuration file '$conffile'\n";
5436
5437 PVE::Cluster::cfs_update(); # make sure we read new file
5438
5439 eval { rescan($vmid, 1); };
5440 warn $@ if $@;
5441 };
5442
5443
5444 # Internal snapshots
5445
5446 # NOTE: Snapshot create/delete involves several non-atomic
5447 # action, and can take a long time.
5448 # So we try to avoid locking the file and use 'lock' variable
5449 # inside the config file instead.
5450
5451 my $snapshot_copy_config = sub {
5452 my ($source, $dest) = @_;
5453
5454 foreach my $k (keys %$source) {
5455 next if $k eq 'snapshots';
5456 next if $k eq 'snapstate';
5457 next if $k eq 'snaptime';
5458 next if $k eq 'vmstate';
5459 next if $k eq 'lock';
5460 next if $k eq 'digest';
5461 next if $k eq 'description';
5462 next if $k =~ m/^unused\d+$/;
5463
5464 $dest->{$k} = $source->{$k};
5465 }
5466 };
5467
5468 my $snapshot_apply_config = sub {
5469 my ($conf, $snap) = @_;
5470
5471 # copy snapshot list
5472 my $newconf = {
5473 snapshots => $conf->{snapshots},
5474 };
5475
5476 # keep description and list of unused disks
5477 foreach my $k (keys %$conf) {
5478 next if !($k =~ m/^unused\d+$/ || $k eq 'description');
5479 $newconf->{$k} = $conf->{$k};
5480 }
5481
5482 &$snapshot_copy_config($snap, $newconf);
5483
5484 return $newconf;
5485 };
5486
5487 sub foreach_writable_storage {
5488 my ($conf, $func) = @_;
5489
5490 my $sidhash = {};
5491
5492 foreach my $ds (keys %$conf) {
5493 next if !valid_drivename($ds);
5494
5495 my $drive = parse_drive($ds, $conf->{$ds});
5496 next if !$drive;
5497 next if drive_is_cdrom($drive);
5498
5499 my $volid = $drive->{file};
5500
5501 my ($sid, $volname) = PVE::Storage::parse_volume_id($volid, 1);
5502 $sidhash->{$sid} = $sid if $sid;
5503 }
5504
5505 foreach my $sid (sort keys %$sidhash) {
5506 &$func($sid);
5507 }
5508 }
5509
5510 my $alloc_vmstate_volid = sub {
5511 my ($storecfg, $vmid, $conf, $snapname) = @_;
5512
5513 # Note: we try to be smart when selecting a $target storage
5514
5515 my $target;
5516
5517 # search shared storage first
5518 foreach_writable_storage($conf, sub {
5519 my ($sid) = @_;
5520 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5521 return if !$scfg->{shared};
5522
5523 $target = $sid if !$target || $scfg->{path}; # prefer file based storage
5524 });
5525
5526 if (!$target) {
5527 # now search local storage
5528 foreach_writable_storage($conf, sub {
5529 my ($sid) = @_;
5530 my $scfg = PVE::Storage::storage_config($storecfg, $sid);
5531 return if $scfg->{shared};
5532
5533 $target = $sid if !$target || $scfg->{path}; # prefer file based storage;
5534 });
5535 }
5536
5537 $target = 'local' if !$target;
5538
5539 my $driver_state_size = 500; # assume 32MB is enough to safe all driver state;
5540 # we abort live save after $conf->{memory}, so we need at max twice that space
5541 my $size = $conf->{memory}*2 + $driver_state_size;
5542
5543 my $name = "vm-$vmid-state-$snapname";
5544 my $scfg = PVE::Storage::storage_config($storecfg, $target);
5545 $name .= ".raw" if $scfg->{path}; # add filename extension for file base storage
5546 my $volid = PVE::Storage::vdisk_alloc($storecfg, $target, $vmid, 'raw', $name, $size*1024);
5547
5548 return $volid;
5549 };
5550
5551 my $snapshot_prepare = sub {
5552 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5553
5554 my $snap;
5555
5556 my $updatefn = sub {
5557
5558 my $conf = load_config($vmid);
5559
5560 die "you can't take a snapshot if it's a template\n"
5561 if is_template($conf);
5562
5563 check_lock($conf);
5564
5565 $conf->{lock} = 'snapshot';
5566
5567 die "snapshot name '$snapname' already used\n"
5568 if defined($conf->{snapshots}->{$snapname});
5569
5570 my $storecfg = PVE::Storage::config();
5571 die "snapshot feature is not available" if !has_feature('snapshot', $conf, $storecfg);
5572
5573 $snap = $conf->{snapshots}->{$snapname} = {};
5574
5575 if ($save_vmstate && check_running($vmid)) {
5576 $snap->{vmstate} = &$alloc_vmstate_volid($storecfg, $vmid, $conf, $snapname);
5577 }
5578
5579 &$snapshot_copy_config($conf, $snap);
5580
5581 $snap->{snapstate} = "prepare";
5582 $snap->{snaptime} = time();
5583 $snap->{description} = $comment if $comment;
5584
5585 # always overwrite machine if we save vmstate. This makes sure we
5586 # can restore it later using correct machine type
5587 $snap->{machine} = get_current_qemu_machine($vmid) if $snap->{vmstate};
5588
5589 update_config_nolock($vmid, $conf, 1);
5590 };
5591
5592 lock_config($vmid, $updatefn);
5593
5594 return $snap;
5595 };
5596
5597 my $snapshot_commit = sub {
5598 my ($vmid, $snapname) = @_;
5599
5600 my $updatefn = sub {
5601
5602 my $conf = load_config($vmid);
5603
5604 die "missing snapshot lock\n"
5605 if !($conf->{lock} && $conf->{lock} eq 'snapshot');
5606
5607 my $has_machine_config = defined($conf->{machine});
5608
5609 my $snap = $conf->{snapshots}->{$snapname};
5610
5611 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5612
5613 die "wrong snapshot state\n"
5614 if !($snap->{snapstate} && $snap->{snapstate} eq "prepare");
5615
5616 delete $snap->{snapstate};
5617 delete $conf->{lock};
5618
5619 my $newconf = &$snapshot_apply_config($conf, $snap);
5620
5621 delete $newconf->{machine} if !$has_machine_config;
5622
5623 $newconf->{parent} = $snapname;
5624
5625 update_config_nolock($vmid, $newconf, 1);
5626 };
5627
5628 lock_config($vmid, $updatefn);
5629 };
5630
5631 sub snapshot_rollback {
5632 my ($vmid, $snapname) = @_;
5633
5634 my $prepare = 1;
5635
5636 my $storecfg = PVE::Storage::config();
5637
5638 my $conf = load_config($vmid);
5639
5640 my $get_snapshot_config = sub {
5641
5642 die "you can't rollback if vm is a template\n" if is_template($conf);
5643
5644 my $res = $conf->{snapshots}->{$snapname};
5645
5646 die "snapshot '$snapname' does not exist\n" if !defined($res);
5647
5648 return $res;
5649 };
5650
5651 my $snap = &$get_snapshot_config();
5652
5653 foreach_drive($snap, sub {
5654 my ($ds, $drive) = @_;
5655
5656 return if drive_is_cdrom($drive);
5657
5658 my $volid = $drive->{file};
5659
5660 PVE::Storage::volume_rollback_is_possible($storecfg, $volid, $snapname);
5661 });
5662
5663 my $updatefn = sub {
5664
5665 $conf = load_config($vmid);
5666
5667 $snap = &$get_snapshot_config();
5668
5669 die "unable to rollback to incomplete snapshot (snapstate = $snap->{snapstate})\n"
5670 if $snap->{snapstate};
5671
5672 if ($prepare) {
5673 check_lock($conf);
5674 vm_stop($storecfg, $vmid, undef, undef, 5, undef, undef);
5675 }
5676
5677 die "unable to rollback vm $vmid: vm is running\n"
5678 if check_running($vmid);
5679
5680 if ($prepare) {
5681 $conf->{lock} = 'rollback';
5682 } else {
5683 die "got wrong lock\n" if !($conf->{lock} && $conf->{lock} eq 'rollback');
5684 delete $conf->{lock};
5685 }
5686
5687 my $forcemachine;
5688
5689 if (!$prepare) {
5690 my $has_machine_config = defined($conf->{machine});
5691
5692 # copy snapshot config to current config
5693 $conf = &$snapshot_apply_config($conf, $snap);
5694 $conf->{parent} = $snapname;
5695
5696 # Note: old code did not store 'machine', so we try to be smart
5697 # and guess the snapshot was generated with kvm 1.4 (pc-i440fx-1.4).
5698 $forcemachine = $conf->{machine} || 'pc-i440fx-1.4';
5699 # we remove the 'machine' configuration if not explicitly specified
5700 # in the original config.
5701 delete $conf->{machine} if $snap->{vmstate} && !$has_machine_config;
5702 }
5703
5704 update_config_nolock($vmid, $conf, 1);
5705
5706 if (!$prepare && $snap->{vmstate}) {
5707 my $statefile = PVE::Storage::path($storecfg, $snap->{vmstate});
5708 vm_start($storecfg, $vmid, $statefile, undef, undef, undef, $forcemachine);
5709 }
5710 };
5711
5712 lock_config($vmid, $updatefn);
5713
5714 foreach_drive($snap, sub {
5715 my ($ds, $drive) = @_;
5716
5717 return if drive_is_cdrom($drive);
5718
5719 my $volid = $drive->{file};
5720 my $device = "drive-$ds";
5721
5722 PVE::Storage::volume_snapshot_rollback($storecfg, $volid, $snapname);
5723 });
5724
5725 $prepare = 0;
5726 lock_config($vmid, $updatefn);
5727 }
5728
5729 my $savevm_wait = sub {
5730 my ($vmid) = @_;
5731
5732 for(;;) {
5733 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5734 if (!$stat->{status}) {
5735 die "savevm not active\n";
5736 } elsif ($stat->{status} eq 'active') {
5737 sleep(1);
5738 next;
5739 } elsif ($stat->{status} eq 'completed') {
5740 last;
5741 } else {
5742 die "query-savevm returned status '$stat->{status}'\n";
5743 }
5744 }
5745 };
5746
5747 sub do_snapshots_with_qemu {
5748 my ($storecfg, $volid) = @_;
5749
5750 my $storage_name = PVE::Storage::parse_volume_id($volid);
5751
5752 if ($qemu_snap_storage->{$storecfg->{ids}->{$storage_name}->{type}} ){
5753 return 1;
5754 }
5755
5756 if ($volid =~ m/\.(qcow2|qed)$/){
5757 return 1;
5758 }
5759
5760 return undef;
5761 }
5762
5763 sub snapshot_create {
5764 my ($vmid, $snapname, $save_vmstate, $comment) = @_;
5765
5766 my $snap = &$snapshot_prepare($vmid, $snapname, $save_vmstate, $comment);
5767
5768 $save_vmstate = 0 if !$snap->{vmstate}; # vm is not running
5769
5770 my $config = load_config($vmid);
5771
5772 my $running = check_running($vmid);
5773
5774 my $freezefs = $running && $config->{agent};
5775 $freezefs = 0 if $snap->{vmstate}; # not needed if we save RAM
5776
5777 my $drivehash = {};
5778
5779 if ($freezefs) {
5780 eval { vm_mon_cmd($vmid, "guest-fsfreeze-freeze"); };
5781 warn "guest-fsfreeze-freeze problems - $@" if $@;
5782 }
5783
5784 eval {
5785 # create internal snapshots of all drives
5786
5787 my $storecfg = PVE::Storage::config();
5788
5789 if ($running) {
5790 if ($snap->{vmstate}) {
5791 my $path = PVE::Storage::path($storecfg, $snap->{vmstate});
5792 vm_mon_cmd($vmid, "savevm-start", statefile => $path);
5793 &$savevm_wait($vmid);
5794 } else {
5795 vm_mon_cmd($vmid, "savevm-start");
5796 }
5797 };
5798
5799 foreach_drive($snap, sub {
5800 my ($ds, $drive) = @_;
5801
5802 return if drive_is_cdrom($drive);
5803
5804 my $volid = $drive->{file};
5805 my $device = "drive-$ds";
5806
5807 qemu_volume_snapshot($vmid, $device, $storecfg, $volid, $snapname);
5808 $drivehash->{$ds} = 1;
5809 });
5810 };
5811 my $err = $@;
5812
5813 if ($running) {
5814 eval { vm_mon_cmd($vmid, "savevm-end") };
5815 warn $@ if $@;
5816
5817 if ($freezefs) {
5818 eval { vm_mon_cmd($vmid, "guest-fsfreeze-thaw"); };
5819 warn "guest-fsfreeze-thaw problems - $@" if $@;
5820 }
5821
5822 # savevm-end is async, we need to wait
5823 for (;;) {
5824 my $stat = vm_mon_cmd_nocheck($vmid, "query-savevm");
5825 if (!$stat->{bytes}) {
5826 last;
5827 } else {
5828 print "savevm not yet finished\n";
5829 sleep(1);
5830 next;
5831 }
5832 }
5833 }
5834
5835 if ($err) {
5836 warn "snapshot create failed: starting cleanup\n";
5837 eval { snapshot_delete($vmid, $snapname, 0, $drivehash); };
5838 warn $@ if $@;
5839 die $err;
5840 }
5841
5842 &$snapshot_commit($vmid, $snapname);
5843 }
5844
5845 # Note: $drivehash is only set when called from snapshot_create.
5846 sub snapshot_delete {
5847 my ($vmid, $snapname, $force, $drivehash) = @_;
5848
5849 my $prepare = 1;
5850
5851 my $snap;
5852 my $unused = [];
5853
5854 my $unlink_parent = sub {
5855 my ($confref, $new_parent) = @_;
5856
5857 if ($confref->{parent} && $confref->{parent} eq $snapname) {
5858 if ($new_parent) {
5859 $confref->{parent} = $new_parent;
5860 } else {
5861 delete $confref->{parent};
5862 }
5863 }
5864 };
5865
5866 my $updatefn = sub {
5867 my ($remove_drive) = @_;
5868
5869 my $conf = load_config($vmid);
5870
5871 if (!$drivehash) {
5872 check_lock($conf);
5873 die "you can't delete a snapshot if vm is a template\n"
5874 if is_template($conf);
5875 }
5876
5877 $snap = $conf->{snapshots}->{$snapname};
5878
5879 die "snapshot '$snapname' does not exist\n" if !defined($snap);
5880
5881 # remove parent refs
5882 if (!$prepare) {
5883 &$unlink_parent($conf, $snap->{parent});
5884 foreach my $sn (keys %{$conf->{snapshots}}) {
5885 next if $sn eq $snapname;
5886 &$unlink_parent($conf->{snapshots}->{$sn}, $snap->{parent});
5887 }
5888 }
5889
5890 if ($remove_drive) {
5891 if ($remove_drive eq 'vmstate') {
5892 delete $snap->{$remove_drive};
5893 } else {
5894 my $drive = parse_drive($remove_drive, $snap->{$remove_drive});
5895 my $volid = $drive->{file};
5896 delete $snap->{$remove_drive};
5897 add_unused_volume($conf, $volid);
5898 }
5899 }
5900
5901 if ($prepare) {
5902 $snap->{snapstate} = 'delete';
5903 } else {
5904 delete $conf->{snapshots}->{$snapname};
5905 delete $conf->{lock} if $drivehash;
5906 foreach my $volid (@$unused) {
5907 add_unused_volume($conf, $volid);
5908 }
5909 }
5910
5911 update_config_nolock($vmid, $conf, 1);
5912 };
5913
5914 lock_config($vmid, $updatefn);
5915
5916 # now remove vmstate file
5917
5918 my $storecfg = PVE::Storage::config();
5919
5920 if ($snap->{vmstate}) {
5921 eval { PVE::Storage::vdisk_free($storecfg, $snap->{vmstate}); };
5922 if (my $err = $@) {
5923 die $err if !$force;
5924 warn $err;
5925 }
5926 # save changes (remove vmstate from snapshot)
5927 lock_config($vmid, $updatefn, 'vmstate') if !$force;
5928 };
5929
5930 # now remove all internal snapshots
5931 foreach_drive($snap, sub {
5932 my ($ds, $drive) = @_;
5933
5934 return if drive_is_cdrom($drive);
5935
5936 my $volid = $drive->{file};
5937 my $device = "drive-$ds";
5938
5939 if (!$drivehash || $drivehash->{$ds}) {
5940 eval { qemu_volume_snapshot_delete($vmid, $device, $storecfg, $volid, $snapname); };
5941 if (my $err = $@) {
5942 die $err if !$force;
5943 warn $err;
5944 }
5945 }
5946
5947 # save changes (remove drive fron snapshot)
5948 lock_config($vmid, $updatefn, $ds) if !$force;
5949 push @$unused, $volid;
5950 });
5951
5952 # now cleanup config
5953 $prepare = 0;
5954 lock_config($vmid, $updatefn);
5955 }
5956
5957 sub has_feature {
5958 my ($feature, $conf, $storecfg, $snapname, $running) = @_;
5959
5960 my $err;
5961 foreach_drive($conf, sub {
5962 my ($ds, $drive) = @_;
5963
5964 return if drive_is_cdrom($drive);
5965 my $volid = $drive->{file};
5966 $err = 1 if !PVE::Storage::volume_has_feature($storecfg, $feature, $volid, $snapname, $running);
5967 });
5968
5969 return $err ? 0 : 1;
5970 }
5971
5972 sub template_create {
5973 my ($vmid, $conf, $disk) = @_;
5974
5975 my $storecfg = PVE::Storage::config();
5976
5977 foreach_drive($conf, sub {
5978 my ($ds, $drive) = @_;
5979
5980 return if drive_is_cdrom($drive);
5981 return if $disk && $ds ne $disk;
5982
5983 my $volid = $drive->{file};
5984 return if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
5985
5986 my $voliddst = PVE::Storage::vdisk_create_base($storecfg, $volid);
5987 $drive->{file} = $voliddst;
5988 $conf->{$ds} = print_drive($vmid, $drive);
5989 update_config_nolock($vmid, $conf, 1);
5990 });
5991 }
5992
5993 sub is_template {
5994 my ($conf) = @_;
5995
5996 return 1 if defined $conf->{template} && $conf->{template} == 1;
5997 }
5998
5999 sub qemu_img_convert {
6000 my ($src_volid, $dst_volid, $size, $snapname) = @_;
6001
6002 my $storecfg = PVE::Storage::config();
6003 my ($src_storeid, $src_volname) = PVE::Storage::parse_volume_id($src_volid, 1);
6004 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid, 1);
6005
6006 if ($src_storeid && $dst_storeid) {
6007 my $src_scfg = PVE::Storage::storage_config($storecfg, $src_storeid);
6008 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6009
6010 my $src_format = qemu_img_format($src_scfg, $src_volname);
6011 my $dst_format = qemu_img_format($dst_scfg, $dst_volname);
6012
6013 my $src_path = PVE::Storage::path($storecfg, $src_volid, $snapname);
6014 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6015
6016 my $cmd = [];
6017 push @$cmd, '/usr/bin/qemu-img', 'convert', '-t', 'writeback', '-p', '-n';
6018 push @$cmd, '-s', $snapname if($snapname && $src_format eq "qcow2");
6019 push @$cmd, '-f', $src_format, '-O', $dst_format, $src_path, $dst_path;
6020
6021 my $parser = sub {
6022 my $line = shift;
6023 if($line =~ m/\((\S+)\/100\%\)/){
6024 my $percent = $1;
6025 my $transferred = int($size * $percent / 100);
6026 my $remaining = $size - $transferred;
6027
6028 print "transferred: $transferred bytes remaining: $remaining bytes total: $size bytes progression: $percent %\n";
6029 }
6030
6031 };
6032
6033 eval { run_command($cmd, timeout => undef, outfunc => $parser); };
6034 my $err = $@;
6035 die "copy failed: $err" if $err;
6036 }
6037 }
6038
6039 sub qemu_img_format {
6040 my ($scfg, $volname) = @_;
6041
6042 if ($scfg->{path} && $volname =~ m/\.(raw|cow|qcow|qcow2|qed|vmdk|cloop)$/) {
6043 return $1;
6044 } else {
6045 return "raw";
6046 }
6047 }
6048
6049 sub qemu_drive_mirror {
6050 my ($vmid, $drive, $dst_volid, $vmiddst) = @_;
6051
6052 my $storecfg = PVE::Storage::config();
6053 my ($dst_storeid, $dst_volname) = PVE::Storage::parse_volume_id($dst_volid);
6054
6055 my $dst_scfg = PVE::Storage::storage_config($storecfg, $dst_storeid);
6056
6057 my $format = qemu_img_format($dst_scfg, $dst_volname);
6058
6059 my $dst_path = PVE::Storage::path($storecfg, $dst_volid);
6060
6061 my $opts = { timeout => 10, device => "drive-$drive", mode => "existing", sync => "full", target => $dst_path };
6062 $opts->{format} = $format if $format;
6063
6064 print "drive mirror is starting (scanning bitmap) : this step can take some minutes/hours, depend of disk size and storage speed\n";
6065
6066 eval {
6067 vm_mon_cmd($vmid, "drive-mirror", %$opts);
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;