]> git.proxmox.com Git - pve-container.git/blob - src/PVE/LXC.pm
deal with a check_mount_path race condition
[pve-container.git] / src / PVE / LXC.pm
1 package PVE::LXC;
2
3 use strict;
4 use warnings;
5
6 use POSIX qw(EINTR);
7
8 use Socket;
9
10 use File::Path;
11 use File::Spec;
12 use Cwd qw();
13 use Fcntl qw(O_RDONLY O_NOFOLLOW O_DIRECTORY);
14 use Errno qw(ELOOP);
15
16 use PVE::Exception qw(raise_perm_exc);
17 use PVE::Storage;
18 use PVE::SafeSyslog;
19 use PVE::INotify;
20 use PVE::Tools qw($IPV6RE $IPV4RE dir_glob_foreach lock_file lock_file_full O_PATH);
21 use PVE::Network;
22 use PVE::AccessControl;
23 use PVE::ProcFSTools;
24 use PVE::LXC::Config;
25 use Time::HiRes qw (gettimeofday);
26
27 use Data::Dumper;
28
29 my $nodename = PVE::INotify::nodename();
30
31 my $cpuinfo= PVE::ProcFSTools::read_cpuinfo();
32
33 our $COMMON_TAR_FLAGS = [ '--sparse', '--numeric-owner', '--acls',
34 '--xattrs',
35 '--xattrs-include=user.*',
36 '--xattrs-include=security.capability',
37 '--warning=no-xattr-write' ];
38
39 sub config_list {
40 my $vmlist = PVE::Cluster::get_vmlist();
41 my $res = {};
42 return $res if !$vmlist || !$vmlist->{ids};
43 my $ids = $vmlist->{ids};
44
45 foreach my $vmid (keys %$ids) {
46 next if !$vmid; # skip CT0
47 my $d = $ids->{$vmid};
48 next if !$d->{node} || $d->{node} ne $nodename;
49 next if !$d->{type} || $d->{type} ne 'lxc';
50 $res->{$vmid}->{type} = 'lxc';
51 }
52 return $res;
53 }
54
55 sub destroy_config {
56 my ($vmid) = @_;
57
58 unlink PVE::LXC::Config->config_file($vmid, $nodename);
59 }
60
61 # container status helpers
62
63 sub list_active_containers {
64
65 my $filename = "/proc/net/unix";
66
67 # similar test is used by lcxcontainers.c: list_active_containers
68 my $res = {};
69
70 my $fh = IO::File->new ($filename, "r");
71 return $res if !$fh;
72
73 while (defined(my $line = <$fh>)) {
74 if ($line =~ m/^[a-f0-9]+:\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\d+\s+(\S+)$/) {
75 my $path = $1;
76 if ($path =~ m!^@/var/lib/lxc/(\d+)/command$!) {
77 $res->{$1} = 1;
78 }
79 }
80 }
81
82 close($fh);
83
84 return $res;
85 }
86
87 # warning: this is slow
88 sub check_running {
89 my ($vmid) = @_;
90
91 my $active_hash = list_active_containers();
92
93 return 1 if defined($active_hash->{$vmid});
94
95 return undef;
96 }
97
98 sub get_container_disk_usage {
99 my ($vmid, $pid) = @_;
100
101 return PVE::Tools::df("/proc/$pid/root/", 1);
102 }
103
104 my $last_proc_vmid_stat;
105
106 my $parse_cpuacct_stat = sub {
107 my ($vmid) = @_;
108
109 my $raw = read_cgroup_value('cpuacct', $vmid, 'cpuacct.stat', 1);
110
111 my $stat = {};
112
113 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
114
115 $stat->{utime} = $1;
116 $stat->{stime} = $2;
117
118 }
119
120 return $stat;
121 };
122
123 sub vmstatus {
124 my ($opt_vmid) = @_;
125
126 my $list = $opt_vmid ? { $opt_vmid => { type => 'lxc' }} : config_list();
127
128 my $active_hash = list_active_containers();
129
130 my $cpucount = $cpuinfo->{cpus} || 1;
131
132 my $cdtime = gettimeofday;
133
134 my $uptime = (PVE::ProcFSTools::read_proc_uptime(1))[0];
135
136 foreach my $vmid (keys %$list) {
137 my $d = $list->{$vmid};
138
139 eval { $d->{pid} = find_lxc_pid($vmid) if defined($active_hash->{$vmid}); };
140 warn $@ if $@; # ignore errors (consider them stopped)
141
142 $d->{status} = $d->{pid} ? 'running' : 'stopped';
143
144 my $cfspath = PVE::LXC::Config->cfs_config_path($vmid);
145 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
146
147 $d->{name} = $conf->{'hostname'} || "CT$vmid";
148 $d->{name} =~ s/[\s]//g;
149
150 $d->{cpus} = $conf->{cpulimit} || $cpucount;
151
152 $d->{lock} = $conf->{lock} || '';
153
154 if ($d->{pid}) {
155 my $res = get_container_disk_usage($vmid, $d->{pid});
156 $d->{disk} = $res->{used};
157 $d->{maxdisk} = $res->{total};
158 } else {
159 $d->{disk} = 0;
160 # use 4GB by default ??
161 if (my $rootfs = $conf->{rootfs}) {
162 my $rootinfo = PVE::LXC::Config->parse_ct_rootfs($rootfs);
163 $d->{maxdisk} = $rootinfo->{size} || (4*1024*1024*1024);
164 } else {
165 $d->{maxdisk} = 4*1024*1024*1024;
166 }
167 }
168
169 $d->{mem} = 0;
170 $d->{swap} = 0;
171 $d->{maxmem} = ($conf->{memory}||512)*1024*1024;
172 $d->{maxswap} = ($conf->{swap}//0)*1024*1024;
173
174 $d->{uptime} = 0;
175 $d->{cpu} = 0;
176
177 $d->{netout} = 0;
178 $d->{netin} = 0;
179
180 $d->{diskread} = 0;
181 $d->{diskwrite} = 0;
182
183 $d->{template} = PVE::LXC::Config->is_template($conf);
184 }
185
186 foreach my $vmid (keys %$list) {
187 my $d = $list->{$vmid};
188 my $pid = $d->{pid};
189
190 next if !$pid; # skip stopped CTs
191
192 my $ctime = (stat("/proc/$pid"))[10]; # 10 = ctime
193 $d->{uptime} = time - $ctime; # the method lxcfs uses
194
195 $d->{mem} = read_cgroup_value('memory', $vmid, 'memory.usage_in_bytes');
196 $d->{swap} = read_cgroup_value('memory', $vmid, 'memory.memsw.usage_in_bytes') - $d->{mem};
197
198 my $blkio_bytes = read_cgroup_value('blkio', $vmid, 'blkio.throttle.io_service_bytes', 1);
199 my @bytes = split(/\n/, $blkio_bytes);
200 foreach my $byte (@bytes) {
201 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
202 $d->{diskread} = $2 if $key eq 'Read';
203 $d->{diskwrite} = $2 if $key eq 'Write';
204 }
205 }
206
207 my $pstat = &$parse_cpuacct_stat($vmid);
208
209 my $used = $pstat->{utime} + $pstat->{stime};
210
211 my $old = $last_proc_vmid_stat->{$vmid};
212 if (!$old) {
213 $last_proc_vmid_stat->{$vmid} = {
214 time => $cdtime,
215 used => $used,
216 cpu => 0,
217 };
218 next;
219 }
220
221 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
222
223 if ($dtime > 1000) {
224 my $dutime = $used - $old->{used};
225
226 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
227 $last_proc_vmid_stat->{$vmid} = {
228 time => $cdtime,
229 used => $used,
230 cpu => $d->{cpu},
231 };
232 } else {
233 $d->{cpu} = $old->{cpu};
234 }
235 }
236
237 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
238
239 foreach my $dev (keys %$netdev) {
240 next if $dev !~ m/^veth([1-9]\d*)i/;
241 my $vmid = $1;
242 my $d = $list->{$vmid};
243
244 next if !$d;
245
246 $d->{netout} += $netdev->{$dev}->{receive};
247 $d->{netin} += $netdev->{$dev}->{transmit};
248
249 }
250
251 return $list;
252 }
253
254 sub read_cgroup_value {
255 my ($group, $vmid, $name, $full) = @_;
256
257 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
258
259 return PVE::Tools::file_get_contents($path) if $full;
260
261 return PVE::Tools::file_read_firstline($path);
262 }
263
264 sub write_cgroup_value {
265 my ($group, $vmid, $name, $value) = @_;
266
267 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
268 PVE::ProcFSTools::write_proc_entry($path, $value) if -e $path;
269
270 }
271
272 sub find_lxc_console_pids {
273
274 my $res = {};
275
276 PVE::Tools::dir_glob_foreach('/proc', '\d+', sub {
277 my ($pid) = @_;
278
279 my $cmdline = PVE::Tools::file_read_firstline("/proc/$pid/cmdline");
280 return if !$cmdline;
281
282 my @args = split(/\0/, $cmdline);
283
284 # search for lxc-console -n <vmid>
285 return if scalar(@args) != 3;
286 return if $args[1] ne '-n';
287 return if $args[2] !~ m/^\d+$/;
288 return if $args[0] !~ m|^(/usr/bin/)?lxc-console$|;
289
290 my $vmid = $args[2];
291
292 push @{$res->{$vmid}}, $pid;
293 });
294
295 return $res;
296 }
297
298 sub find_lxc_pid {
299 my ($vmid) = @_;
300
301 my $pid = undef;
302 my $parser = sub {
303 my $line = shift;
304 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
305 };
306 PVE::Tools::run_command(['lxc-info', '-n', $vmid, '-p'], outfunc => $parser);
307
308 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
309
310 return $pid;
311 }
312
313 # Note: we cannot use Net:IP, because that only allows strict
314 # CIDR networks
315 sub parse_ipv4_cidr {
316 my ($cidr, $noerr) = @_;
317
318 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
319 return { address => $1, netmask => $PVE::Network::ipv4_reverse_mask->[$2] };
320 }
321
322 return undef if $noerr;
323
324 die "unable to parse ipv4 address/mask\n";
325 }
326
327
328 sub update_lxc_config {
329 my ($storage_cfg, $vmid, $conf) = @_;
330
331 my $dir = "/var/lib/lxc/$vmid";
332
333 if ($conf->{template}) {
334
335 unlink "$dir/config";
336
337 return;
338 }
339
340 my $raw = '';
341
342 die "missing 'arch' - internal error" if !$conf->{arch};
343 $raw .= "lxc.arch = $conf->{arch}\n";
344
345 my $unprivileged = $conf->{unprivileged};
346 my $custom_idmap = grep { $_->[0] eq 'lxc.id_map' } @{$conf->{lxc}};
347
348 my $ostype = $conf->{ostype} || die "missing 'ostype' - internal error";
349 if ($ostype =~ /^(?:debian | ubuntu | centos | fedora | opensuse | archlinux | alpine | gentoo | unmanaged)$/x) {
350 my $inc ="/usr/share/lxc/config/$ostype.common.conf";
351 $inc ="/usr/share/lxc/config/common.conf" if !-f $inc;
352 $raw .= "lxc.include = $inc\n";
353 if ($unprivileged || $custom_idmap) {
354 $inc = "/usr/share/lxc/config/$ostype.userns.conf";
355 $inc = "/usr/share/lxc/config/userns.conf" if !-f $inc;
356 $raw .= "lxc.include = $inc\n"
357 }
358 } else {
359 die "implement me (ostype $ostype)";
360 }
361
362 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
363 # cannot be exposed to the container with r/w access (cgroup perms).
364 # When this is enabled mounts will still remain in the monitor's namespace
365 # after the container unmounted them and thus will not detach from their
366 # files while the container is running!
367 $raw .= "lxc.monitor.unshare = 1\n";
368
369 # Should we read them from /etc/subuid?
370 if ($unprivileged && !$custom_idmap) {
371 $raw .= "lxc.id_map = u 0 100000 65536\n";
372 $raw .= "lxc.id_map = g 0 100000 65536\n";
373 }
374
375 if (!PVE::LXC::Config->has_dev_console($conf)) {
376 $raw .= "lxc.console = none\n";
377 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n";
378 }
379
380 my $ttycount = PVE::LXC::Config->get_tty_count($conf);
381 $raw .= "lxc.tty = $ttycount\n";
382
383 # some init scripts expect a linux terminal (turnkey).
384 $raw .= "lxc.environment = TERM=linux\n";
385
386 my $utsname = $conf->{hostname} || "CT$vmid";
387 $raw .= "lxc.utsname = $utsname\n";
388
389 my $memory = $conf->{memory} || 512;
390 my $swap = $conf->{swap} // 0;
391
392 my $lxcmem = int($memory*1024*1024);
393 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
394
395 my $lxcswap = int(($memory + $swap)*1024*1024);
396 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
397
398 if (my $cpulimit = $conf->{cpulimit}) {
399 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
400 my $value = int(100000*$cpulimit);
401 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
402 }
403
404 my $shares = $conf->{cpuunits} || 1024;
405 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
406
407 die "missing 'rootfs' configuration\n"
408 if !defined($conf->{rootfs});
409
410 my $mountpoint = PVE::LXC::Config->parse_ct_rootfs($conf->{rootfs});
411
412 $raw .= "lxc.rootfs = $dir/rootfs\n";
413
414 my $netcount = 0;
415 foreach my $k (keys %$conf) {
416 next if $k !~ m/^net(\d+)$/;
417 my $ind = $1;
418 my $d = PVE::LXC::Config->parse_lxc_network($conf->{$k});
419 $netcount++;
420 $raw .= "lxc.network.type = veth\n";
421 $raw .= "lxc.network.veth.pair = veth${vmid}i${ind}\n";
422 $raw .= "lxc.network.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr});
423 $raw .= "lxc.network.name = $d->{name}\n" if defined($d->{name});
424 $raw .= "lxc.network.mtu = $d->{mtu}\n" if defined($d->{mtu});
425 }
426
427 if (my $lxcconf = $conf->{lxc}) {
428 foreach my $entry (@$lxcconf) {
429 my ($k, $v) = @$entry;
430 $netcount++ if $k eq 'lxc.network.type';
431 $raw .= "$k = $v\n";
432 }
433 }
434
435 $raw .= "lxc.network.type = empty\n" if !$netcount;
436
437 File::Path::mkpath("$dir/rootfs");
438
439 PVE::Tools::file_set_contents("$dir/config", $raw);
440 }
441
442 # verify and cleanup nameserver list (replace \0 with ' ')
443 sub verify_nameserver_list {
444 my ($nameserver_list) = @_;
445
446 my @list = ();
447 foreach my $server (PVE::Tools::split_list($nameserver_list)) {
448 PVE::JSONSchema::pve_verify_ip($server);
449 push @list, $server;
450 }
451
452 return join(' ', @list);
453 }
454
455 sub verify_searchdomain_list {
456 my ($searchdomain_list) = @_;
457
458 my @list = ();
459 foreach my $server (PVE::Tools::split_list($searchdomain_list)) {
460 # todo: should we add checks for valid dns domains?
461 push @list, $server;
462 }
463
464 return join(' ', @list);
465 }
466
467 sub get_console_command {
468 my ($vmid, $conf) = @_;
469
470 my $cmode = PVE::LXC::Config->get_cmode($conf);
471
472 if ($cmode eq 'console') {
473 return ['lxc-console', '-n', $vmid, '-t', 0];
474 } elsif ($cmode eq 'tty') {
475 return ['lxc-console', '-n', $vmid];
476 } elsif ($cmode eq 'shell') {
477 return ['lxc-attach', '--clear-env', '-n', $vmid];
478 } else {
479 die "internal error";
480 }
481 }
482
483 sub get_primary_ips {
484 my ($conf) = @_;
485
486 # return data from net0
487
488 return undef if !defined($conf->{net0});
489 my $net = PVE::LXC::Config->parse_lxc_network($conf->{net0});
490
491 my $ipv4 = $net->{ip};
492 if ($ipv4) {
493 if ($ipv4 =~ /^(dhcp|manual)$/) {
494 $ipv4 = undef
495 } else {
496 $ipv4 =~ s!/\d+$!!;
497 }
498 }
499 my $ipv6 = $net->{ip6};
500 if ($ipv6) {
501 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
502 $ipv6 = undef;
503 } else {
504 $ipv6 =~ s!/\d+$!!;
505 }
506 }
507
508 return ($ipv4, $ipv6);
509 }
510
511 sub delete_mountpoint_volume {
512 my ($storage_cfg, $vmid, $volume) = @_;
513
514 return if PVE::LXC::Config->classify_mountpoint($volume) ne 'volume';
515
516 my ($vtype, $name, $owner) = PVE::Storage::parse_volname($storage_cfg, $volume);
517 PVE::Storage::vdisk_free($storage_cfg, $volume) if $vmid == $owner;
518 }
519
520 sub destroy_lxc_container {
521 my ($storage_cfg, $vmid, $conf) = @_;
522
523 PVE::LXC::Config->foreach_mountpoint($conf, sub {
524 my ($ms, $mountpoint) = @_;
525 delete_mountpoint_volume($storage_cfg, $vmid, $mountpoint->{volume});
526 });
527
528 rmdir "/var/lib/lxc/$vmid/rootfs";
529 unlink "/var/lib/lxc/$vmid/config";
530 rmdir "/var/lib/lxc/$vmid";
531 destroy_config($vmid);
532
533 #my $cmd = ['lxc-destroy', '-n', $vmid ];
534 #PVE::Tools::run_command($cmd);
535 }
536
537 sub vm_stop_cleanup {
538 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
539
540 eval {
541 if (!$keepActive) {
542
543 my $vollist = PVE::LXC::Config->get_vm_volumes($conf);
544 PVE::Storage::deactivate_volumes($storage_cfg, $vollist);
545 }
546 };
547 warn $@ if $@; # avoid errors - just warn
548 }
549
550 my $safe_num_ne = sub {
551 my ($a, $b) = @_;
552
553 return 0 if !defined($a) && !defined($b);
554 return 1 if !defined($a);
555 return 1 if !defined($b);
556
557 return $a != $b;
558 };
559
560 my $safe_string_ne = sub {
561 my ($a, $b) = @_;
562
563 return 0 if !defined($a) && !defined($b);
564 return 1 if !defined($a);
565 return 1 if !defined($b);
566
567 return $a ne $b;
568 };
569
570 sub update_net {
571 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
572
573 if ($newnet->{type} ne 'veth') {
574 # for when there are physical interfaces
575 die "cannot update interface of type $newnet->{type}";
576 }
577
578 my $veth = "veth${vmid}i${netid}";
579 my $eth = $newnet->{name};
580
581 if (my $oldnetcfg = $conf->{$opt}) {
582 my $oldnet = PVE::LXC::Config->parse_lxc_network($oldnetcfg);
583
584 if (&$safe_string_ne($oldnet->{hwaddr}, $newnet->{hwaddr}) ||
585 &$safe_string_ne($oldnet->{name}, $newnet->{name})) {
586
587 PVE::Network::veth_delete($veth);
588 delete $conf->{$opt};
589 PVE::LXC::Config->write_config($vmid, $conf);
590
591 hotplug_net($vmid, $conf, $opt, $newnet, $netid);
592
593 } else {
594 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
595 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
596 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
597
598 if ($oldnet->{bridge}) {
599 PVE::Network::tap_unplug($veth);
600 foreach (qw(bridge tag firewall)) {
601 delete $oldnet->{$_};
602 }
603 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($oldnet);
604 PVE::LXC::Config->write_config($vmid, $conf);
605 }
606
607 PVE::Network::tap_plug($veth, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
608 # This includes the rate:
609 foreach (qw(bridge tag firewall rate)) {
610 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
611 }
612 } elsif (&$safe_string_ne($oldnet->{rate}, $newnet->{rate})) {
613 # Rate can be applied on its own but any change above needs to
614 # include the rate in tap_plug since OVS resets everything.
615 PVE::Network::tap_rate_limit($veth, $newnet->{rate});
616 $oldnet->{rate} = $newnet->{rate}
617 }
618 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($oldnet);
619 PVE::LXC::Config->write_config($vmid, $conf);
620 }
621 } else {
622 hotplug_net($vmid, $conf, $opt, $newnet, $netid);
623 }
624
625 update_ipconfig($vmid, $conf, $opt, $eth, $newnet, $rootdir);
626 }
627
628 sub hotplug_net {
629 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
630
631 my $veth = "veth${vmid}i${netid}";
632 my $vethpeer = $veth . "p";
633 my $eth = $newnet->{name};
634
635 PVE::Network::veth_create($veth, $vethpeer, $newnet->{bridge}, $newnet->{hwaddr});
636 PVE::Network::tap_plug($veth, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
637
638 # attach peer in container
639 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
640 PVE::Tools::run_command($cmd);
641
642 # link up peer in container
643 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
644 PVE::Tools::run_command($cmd);
645
646 my $done = { type => 'veth' };
647 foreach (qw(bridge tag firewall hwaddr name)) {
648 $done->{$_} = $newnet->{$_} if $newnet->{$_};
649 }
650 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($done);
651
652 PVE::LXC::Config->write_config($vmid, $conf);
653 }
654
655 sub update_ipconfig {
656 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
657
658 my $lxc_setup = PVE::LXC::Setup->new($conf, $rootdir);
659
660 my $optdata = PVE::LXC::Config->parse_lxc_network($conf->{$opt});
661 my $deleted = [];
662 my $added = [];
663 my $nscmd = sub {
664 my $cmdargs = shift;
665 PVE::Tools::run_command(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
666 };
667 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
668
669 my $change_ip_config = sub {
670 my ($ipversion) = @_;
671
672 my $family_opt = "-$ipversion";
673 my $suffix = $ipversion == 4 ? '' : $ipversion;
674 my $gw= "gw$suffix";
675 my $ip= "ip$suffix";
676
677 my $newip = $newnet->{$ip};
678 my $newgw = $newnet->{$gw};
679 my $oldip = $optdata->{$ip};
680
681 my $change_ip = &$safe_string_ne($oldip, $newip);
682 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
683
684 return if !$change_ip && !$change_gw;
685
686 # step 1: add new IP, if this fails we cancel
687 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
688 if ($change_ip && $is_real_ip) {
689 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
690 if (my $err = $@) {
691 warn $err;
692 return;
693 }
694 }
695
696 # step 2: replace gateway
697 # If this fails we delete the added IP and cancel.
698 # If it succeeds we save the config and delete the old IP, ignoring
699 # errors. The config is then saved.
700 # Note: 'ip route replace' can add
701 if ($change_gw) {
702 if ($newgw) {
703 eval {
704 if ($is_real_ip && !PVE::Network::is_ip_in_cidr($newgw, $newip, $ipversion)) {
705 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
706 }
707 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
708 };
709 if (my $err = $@) {
710 warn $err;
711 # the route was not replaced, the old IP is still available
712 # rollback (delete new IP) and cancel
713 if ($change_ip) {
714 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
715 warn $@ if $@; # no need to die here
716 }
717 return;
718 }
719 } else {
720 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
721 # if the route was not deleted, the guest might have deleted it manually
722 # warn and continue
723 warn $@ if $@;
724 }
725 }
726
727 # from this point on we save the configuration
728 # step 3: delete old IP ignoring errors
729 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
730 # We need to enable promote_secondaries, otherwise our newly added
731 # address will be removed along with the old one.
732 my $promote = 0;
733 eval {
734 if ($ipversion == 4) {
735 &$nscmd({ outfunc => sub { $promote = int(shift) } },
736 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
737 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
738 }
739 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
740 };
741 warn $@ if $@; # no need to die here
742
743 if ($ipversion == 4) {
744 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
745 }
746 }
747
748 foreach my $property ($ip, $gw) {
749 if ($newnet->{$property}) {
750 $optdata->{$property} = $newnet->{$property};
751 } else {
752 delete $optdata->{$property};
753 }
754 }
755 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($optdata);
756 PVE::LXC::Config->write_config($vmid, $conf);
757 $lxc_setup->setup_network($conf);
758 };
759
760 &$change_ip_config(4);
761 &$change_ip_config(6);
762
763 }
764
765 my $enter_namespace = sub {
766 my ($vmid, $pid, $which, $type) = @_;
767 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
768 or die "failed to open $which namespace of container $vmid: $!\n";
769 PVE::Tools::setns(fileno($fd), $type)
770 or die "failed to enter $which namespace of container $vmid: $!\n";
771 close $fd;
772 };
773
774 my $do_syncfs = sub {
775 my ($vmid, $pid, $socket) = @_;
776
777 &$enter_namespace($vmid, $pid, 'mnt', PVE::Tools::CLONE_NEWNS);
778
779 # Tell the parent process to start reading our /proc/mounts
780 print {$socket} "go\n";
781 $socket->flush();
782
783 # Receive /proc/self/mounts
784 my $mountdata = do { local $/ = undef; <$socket> };
785 close $socket;
786
787 # Now sync all mountpoints...
788 my $mounts = PVE::ProcFSTools::parse_mounts($mountdata);
789 foreach my $mp (@$mounts) {
790 my ($what, $dir, $fs) = @$mp;
791 next if $fs eq 'fuse.lxcfs';
792 eval { PVE::Tools::sync_mountpoint($dir); };
793 warn $@ if $@;
794 }
795 };
796
797 sub sync_container_namespace {
798 my ($vmid) = @_;
799 my $pid = find_lxc_pid($vmid);
800
801 # SOCK_DGRAM is nicer for barriers but cannot be slurped
802 socketpair my $pfd, my $cfd, AF_UNIX, SOCK_STREAM, PF_UNSPEC
803 or die "failed to create socketpair: $!\n";
804
805 my $child = fork();
806 die "fork failed: $!\n" if !defined($child);
807
808 if (!$child) {
809 eval {
810 close $pfd;
811 &$do_syncfs($vmid, $pid, $cfd);
812 };
813 if (my $err = $@) {
814 warn $err;
815 POSIX::_exit(1);
816 }
817 POSIX::_exit(0);
818 }
819 close $cfd;
820 my $go = <$pfd>;
821 die "failed to enter container namespace\n" if $go ne "go\n";
822
823 open my $mounts, '<', "/proc/$child/mounts"
824 or die "failed to open container's /proc/mounts: $!\n";
825 my $mountdata = do { local $/ = undef; <$mounts> };
826 close $mounts;
827 print {$pfd} $mountdata;
828 close $pfd;
829
830 while (waitpid($child, 0) != $child) {}
831 die "failed to sync container namespace\n" if $? != 0;
832 }
833
834 sub template_create {
835 my ($vmid, $conf) = @_;
836
837 my $storecfg = PVE::Storage::config();
838
839 my $rootinfo = PVE::LXC::Config->parse_ct_rootfs($conf->{rootfs});
840 my $volid = $rootinfo->{volume};
841
842 die "Template feature is not available for '$volid'\n"
843 if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
844
845 PVE::Storage::activate_volumes($storecfg, [$volid]);
846
847 my $template_volid = PVE::Storage::vdisk_create_base($storecfg, $volid);
848 $rootinfo->{volume} = $template_volid;
849 $conf->{rootfs} = PVE::LXC::Config->print_ct_mountpoint($rootinfo, 1);
850
851 PVE::LXC::Config->write_config($vmid, $conf);
852 }
853
854 sub check_ct_modify_config_perm {
855 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
856
857 return 1 if $authuser eq 'root@pam';
858
859 my $check = sub {
860 my ($opt, $delete) = @_;
861 if ($opt eq 'cpus' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
862 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
863 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
864 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
865 return if $delete;
866 my $data = $opt eq 'rootfs' ? PVE::LXC::Config->parse_ct_rootfs($newconf->{$opt})
867 : PVE::LXC::Config->parse_ct_mountpoint($newconf->{$opt});
868 raise_perm_exc("mountpoint type $data->{type}") if $data->{type} ne 'volume';
869 } elsif ($opt eq 'memory' || $opt eq 'swap') {
870 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
871 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
872 $opt eq 'searchdomain' || $opt eq 'hostname') {
873 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
874 } else {
875 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
876 }
877 };
878
879 foreach my $opt (keys %$newconf) {
880 &$check($opt, 0);
881 }
882 foreach my $opt (@$delete) {
883 &$check($opt, 1);
884 }
885
886 return 1;
887 }
888
889 sub umount_all {
890 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
891
892 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
893 my $volid_list = PVE::LXC::Config->get_vm_volumes($conf);
894
895 PVE::LXC::Config->foreach_mountpoint_reverse($conf, sub {
896 my ($ms, $mountpoint) = @_;
897
898 my $volid = $mountpoint->{volume};
899 my $mount = $mountpoint->{mp};
900
901 return if !$volid || !$mount;
902
903 my $mount_path = "$rootdir/$mount";
904 $mount_path =~ s!/+!/!g;
905
906 return if !PVE::ProcFSTools::is_mounted($mount_path);
907
908 eval {
909 PVE::Tools::run_command(['umount', '-d', $mount_path]);
910 };
911 if (my $err = $@) {
912 if ($noerr) {
913 warn $err;
914 } else {
915 die $err;
916 }
917 }
918 });
919 }
920
921 sub mount_all {
922 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
923
924 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
925 File::Path::make_path($rootdir);
926
927 my $volid_list = PVE::LXC::Config->get_vm_volumes($conf);
928 PVE::Storage::activate_volumes($storage_cfg, $volid_list);
929
930 eval {
931 PVE::LXC::Config->foreach_mountpoint($conf, sub {
932 my ($ms, $mountpoint) = @_;
933
934 $mountpoint->{ro} = 0 if $ignore_ro;
935
936 mountpoint_mount($mountpoint, $rootdir, $storage_cfg);
937 });
938 };
939 if (my $err = $@) {
940 warn "mounting container failed\n";
941 umount_all($vmid, $storage_cfg, $conf, 1);
942 die $err;
943 }
944
945 return $rootdir;
946 }
947
948
949 sub mountpoint_mount_path {
950 my ($mountpoint, $storage_cfg, $snapname) = @_;
951
952 return mountpoint_mount($mountpoint, undef, $storage_cfg, $snapname);
953 }
954
955 sub query_loopdev {
956 my ($path) = @_;
957 my $found;
958 my $parser = sub {
959 my $line = shift;
960 if ($line =~ m@^(/dev/loop\d+):@) {
961 $found = $1;
962 }
963 };
964 my $cmd = ['losetup', '--associated', $path];
965 PVE::Tools::run_command($cmd, outfunc => $parser);
966 return $found;
967 }
968
969 # Run a function with a file attached to a loop device.
970 # The loop device is always detached afterwards (or set to autoclear).
971 # Returns the loop device.
972 sub run_with_loopdev {
973 my ($func, $file) = @_;
974 my $device = query_loopdev($file);
975 # Try to reuse an existing device
976 if ($device) {
977 # We assume that whoever setup the loop device is responsible for
978 # detaching it.
979 &$func($device);
980 return $device;
981 }
982
983 my $parser = sub {
984 my $line = shift;
985 if ($line =~ m@^(/dev/loop\d+)$@) {
986 $device = $1;
987 }
988 };
989 PVE::Tools::run_command(['losetup', '--show', '-f', $file], outfunc => $parser);
990 die "failed to setup loop device for $file\n" if !$device;
991 eval { &$func($device); };
992 my $err = $@;
993 PVE::Tools::run_command(['losetup', '-d', $device]);
994 die $err if $err;
995 return $device;
996 }
997
998 # In scalar mode: returns a file handle to the deepest directory node.
999 # In list context: returns a list of:
1000 # * the deepest directory node
1001 # * the 2nd deepest directory (parent of the above)
1002 # * directory name of the last directory
1003 # So that the path $2/$3 should lead to $1 afterwards.
1004 sub walk_tree_nofollow($$$) {
1005 my ($start, $subdir, $mkdir) = @_;
1006
1007 # splitdir() returns '' for empty components including the leading /
1008 my @comps = grep { length($_)>0 } File::Spec->splitdir($subdir);
1009
1010 sysopen(my $fd, $start, O_PATH | O_DIRECTORY)
1011 or die "failed to open start directory $start: $!\n";
1012
1013 my $dir = $start;
1014 my $last_component = undef;
1015 my $second = $fd;
1016 foreach my $component (@comps) {
1017 $dir .= "/$component";
1018 my $next = PVE::Tools::openat(fileno($fd), $component, O_NOFOLLOW | O_DIRECTORY);
1019
1020 if (!$next) {
1021 # failed, check for symlinks and try to create the path
1022 die "symlink encountered at: $dir\n" if $! == ELOOP;
1023 die "cannot open directory $dir: $!\n" if !$mkdir;
1024
1025 # We don't check for errors on mkdirat() here and just try to
1026 # openat() again, since at least one error (EEXIST) is an
1027 # expected possibility if multiple containers start
1028 # simultaneously. If someone else injects a symlink now then
1029 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1030 PVE::Tools::mkdirat(fileno($fd), $component, 0755);
1031
1032 $next = PVE::Tools::openat(fileno($fd), $component, O_NOFOLLOW | O_DIRECTORY);
1033 die "failed to create path: $dir: $!\n" if !$next;
1034 }
1035
1036 close $second if defined($last_component);
1037 $last_component = $component;
1038 $second = $fd;
1039 $fd = $next;
1040 }
1041
1042 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1043 close $second if defined($last_component);
1044 return $fd;
1045 }
1046
1047 sub bindmount {
1048 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1049
1050 my $srcdh = walk_tree_nofollow('/', $dir, 0);
1051
1052 PVE::Tools::run_command(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1053 if ($ro) {
1054 eval { PVE::Tools::run_command(['mount', '-o', 'bind,remount,ro', $dest]); };
1055 if (my $err = $@) {
1056 warn "bindmount error\n";
1057 # don't leave writable bind-mounts behind...
1058 PVE::Tools::run_command(['umount', $dest]);
1059 die $err;
1060 }
1061 }
1062
1063 my $destdh;
1064 if ($parentfd) {
1065 # Open the mount point path coming from the parent directory since the
1066 # filehandle we would have gotten as first result of walk_tree_nofollow
1067 # earlier is still a handle to the underlying directory instead of the
1068 # mounted path.
1069 $destdh = PVE::Tools::openat(fileno($parentfd), $last_dir, PVE::Tools::O_PATH)
1070 } else {
1071 # For the rootfs we don't have a parentfd so we open the path directly.
1072 # Note that this means bindmounting any prefix of the host's
1073 # /var/lib/lxc/$vmid path into another container is considered a grave
1074 # security error.
1075 sysopen $destdh, $last_dir, O_PATH | O_DIRECTORY;
1076 }
1077 die "failed to open directory $dir: $!\n" if !$destdh;
1078
1079 my ($srcdev, $srcinode) = stat($srcdh);
1080 my ($dstdev, $dstinode) = stat($destdh);
1081 close $srcdh;
1082 close $destdh;
1083
1084 if ($srcdev != $dstdev || $srcinode != $dstinode) {
1085 PVE::Tools::run_command(['umount', $dest]);
1086 die "detected mount path change at: $dir\n";
1087 }
1088 }
1089
1090 # use $rootdir = undef to just return the corresponding mount path
1091 sub mountpoint_mount {
1092 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1093
1094 my $volid = $mountpoint->{volume};
1095 my $mount = $mountpoint->{mp};
1096 my $type = $mountpoint->{type};
1097 my $quota = !$snapname && !$mountpoint->{ro} && $mountpoint->{quota};
1098 my $mounted_dev;
1099
1100 return if !$volid || !$mount;
1101
1102 $mount =~ s!/+!/!g;
1103
1104 my $mount_path;
1105 my ($mpfd, $parentfd, $last_dir);
1106
1107 if (defined($rootdir)) {
1108 $rootdir =~ s!/+!/!g;
1109 $rootdir =~ s!/+$!!;
1110 $mount_path = "$rootdir/$mount";
1111 ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow($rootdir, $mount, 1);
1112 }
1113
1114 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1115
1116 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1117
1118 my $optstring = '';
1119 my $acl = $mountpoint->{acl};
1120 if (defined($acl)) {
1121 $optstring .= ($acl ? 'acl' : 'noacl');
1122 }
1123 my $readonly = $mountpoint->{ro};
1124
1125 my @extra_opts = ('-o', $optstring) if $optstring;
1126
1127 if ($storage) {
1128
1129 my $scfg = PVE::Storage::storage_config($storage_cfg, $storage);
1130 my $path = PVE::Storage::path($storage_cfg, $volid, $snapname);
1131
1132 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1133 PVE::Storage::parse_volname($storage_cfg, $volid);
1134
1135 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1136
1137 if ($format eq 'subvol') {
1138 if ($mount_path) {
1139 if ($snapname) {
1140 if ($scfg->{type} eq 'zfspool') {
1141 my $path_arg = $path;
1142 $path_arg =~ s!^/+!!;
1143 PVE::Tools::run_command(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1144 } else {
1145 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1146 }
1147 } else {
1148 if (defined($acl) && $scfg->{type} eq 'zfspool') {
1149 my $acltype = ($acl ? 'acltype=posixacl' : 'acltype=noacl');
1150 my (undef, $name) = PVE::Storage::parse_volname($storage_cfg, $volid);
1151 $name .= "\@$snapname" if defined($snapname);
1152 PVE::Tools::run_command(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1153 }
1154 bindmount($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1155 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1156 }
1157 }
1158 return wantarray ? ($path, 0, undef) : $path;
1159 } elsif ($format eq 'raw' || $format eq 'iso') {
1160 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1161 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1162 # Our autodev hook expects the /dev/dm-* device currently
1163 # and will create the /dev/mapper symlink accordingly
1164 ($path) = (Cwd::realpath($path) =~ /^(.*)$/s); # realpath() taints
1165 my $domount = sub {
1166 my ($path) = @_;
1167 if ($mount_path) {
1168 if ($format eq 'iso') {
1169 PVE::Tools::run_command(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1170 } elsif ($isBase || defined($snapname)) {
1171 PVE::Tools::run_command(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1172 } else {
1173 if ($quota) {
1174 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1175 }
1176 push @extra_opts, '-o', 'ro' if $readonly;
1177 PVE::Tools::run_command(['mount', @extra_opts, $path, $mount_path]);
1178 }
1179 }
1180 };
1181 my $use_loopdev = 0;
1182 if ($scfg->{path}) {
1183 $mounted_dev = run_with_loopdev($domount, $path);
1184 $use_loopdev = 1;
1185 } elsif ($scfg->{type} eq 'drbd' || $scfg->{type} eq 'lvm' ||
1186 $scfg->{type} eq 'rbd' || $scfg->{type} eq 'lvmthin') {
1187 $mounted_dev = $path;
1188 &$domount($path);
1189 } else {
1190 die "unsupported storage type '$scfg->{type}'\n";
1191 }
1192 return wantarray ? ($path, $use_loopdev, $mounted_dev) : $path;
1193 } else {
1194 die "unsupported image format '$format'\n";
1195 }
1196 } elsif ($type eq 'device') {
1197 push @extra_opts, '-o', 'ro' if $readonly;
1198 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1199 # See the NOTE above about devicemapper canonicalization
1200 my ($devpath) = (Cwd::realpath($volid) =~ /^(.*)$/s); # realpath() taints
1201 PVE::Tools::run_command(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1202 return wantarray ? ($volid, 0, $devpath) : $volid;
1203 } elsif ($type eq 'bind') {
1204 die "directory '$volid' does not exist\n" if ! -d $volid;
1205 bindmount($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1206 warn "cannot enable quota control for bind mounts\n" if $quota;
1207 return wantarray ? ($volid, 0, undef) : $volid;
1208 }
1209
1210 die "unsupported storage";
1211 }
1212
1213 sub mkfs {
1214 my ($dev, $rootuid, $rootgid) = @_;
1215
1216 PVE::Tools::run_command(['mkfs.ext4', '-O', 'mmp',
1217 '-E', "root_owner=$rootuid:$rootgid",
1218 $dev]);
1219 }
1220
1221 sub format_disk {
1222 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1223
1224 if ($volid =~ m!^/dev/.+!) {
1225 mkfs($volid);
1226 return;
1227 }
1228
1229 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1230
1231 die "cannot format volume '$volid' with no storage\n" if !$storage;
1232
1233 PVE::Storage::activate_volumes($storage_cfg, [$volid]);
1234
1235 my $path = PVE::Storage::path($storage_cfg, $volid);
1236
1237 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1238 PVE::Storage::parse_volname($storage_cfg, $volid);
1239
1240 die "cannot format volume '$volid' (format == $format)\n"
1241 if $format ne 'raw';
1242
1243 mkfs($path, $rootuid, $rootgid);
1244 }
1245
1246 sub destroy_disks {
1247 my ($storecfg, $vollist) = @_;
1248
1249 foreach my $volid (@$vollist) {
1250 eval { PVE::Storage::vdisk_free($storecfg, $volid); };
1251 warn $@ if $@;
1252 }
1253 }
1254
1255 sub create_disks {
1256 my ($storecfg, $vmid, $settings, $conf) = @_;
1257
1258 my $vollist = [];
1259
1260 eval {
1261 my (undef, $rootuid, $rootgid) = PVE::LXC::parse_id_maps($conf);
1262 my $chown_vollist = [];
1263
1264 PVE::LXC::Config->foreach_mountpoint($settings, sub {
1265 my ($ms, $mountpoint) = @_;
1266
1267 my $volid = $mountpoint->{volume};
1268 my $mp = $mountpoint->{mp};
1269
1270 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1271
1272 if ($storage && ($volid =~ m/^([^:\s]+):(\d+(\.\d+)?)$/)) {
1273 my ($storeid, $size_gb) = ($1, $2);
1274
1275 my $size_kb = int(${size_gb}*1024) * 1024;
1276
1277 my $scfg = PVE::Storage::storage_config($storecfg, $storage);
1278 # fixme: use better naming ct-$vmid-disk-X.raw?
1279
1280 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs') {
1281 if ($size_kb > 0) {
1282 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw',
1283 undef, $size_kb);
1284 format_disk($storecfg, $volid, $rootuid, $rootgid);
1285 } else {
1286 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'subvol',
1287 undef, 0);
1288 push @$chown_vollist, $volid;
1289 }
1290 } elsif ($scfg->{type} eq 'zfspool') {
1291
1292 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'subvol',
1293 undef, $size_kb);
1294 push @$chown_vollist, $volid;
1295 } elsif ($scfg->{type} eq 'drbd' || $scfg->{type} eq 'lvm' || $scfg->{type} eq 'lvmthin') {
1296
1297 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1298 format_disk($storecfg, $volid, $rootuid, $rootgid);
1299
1300 } elsif ($scfg->{type} eq 'rbd') {
1301
1302 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd};
1303 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1304 format_disk($storecfg, $volid, $rootuid, $rootgid);
1305 } else {
1306 die "unable to create containers on storage type '$scfg->{type}'\n";
1307 }
1308 push @$vollist, $volid;
1309 $mountpoint->{volume} = $volid;
1310 $mountpoint->{size} = $size_kb * 1024;
1311 $conf->{$ms} = PVE::LXC::Config->print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1312 } else {
1313 # use specified/existing volid/dir/device
1314 $conf->{$ms} = PVE::LXC::Config->print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1315 }
1316 });
1317
1318 PVE::Storage::activate_volumes($storecfg, $chown_vollist, undef);
1319 foreach my $volid (@$chown_vollist) {
1320 my $path = PVE::Storage::path($storecfg, $volid, undef);
1321 chown($rootuid, $rootgid, $path);
1322 }
1323 PVE::Storage::deactivate_volumes($storecfg, $chown_vollist, undef);
1324 };
1325 # free allocated images on error
1326 if (my $err = $@) {
1327 destroy_disks($storecfg, $vollist);
1328 die $err;
1329 }
1330 return $vollist;
1331 }
1332
1333 # bash completion helper
1334
1335 sub complete_os_templates {
1336 my ($cmdname, $pname, $cvalue) = @_;
1337
1338 my $cfg = PVE::Storage::config();
1339
1340 my $storeid;
1341
1342 if ($cvalue =~ m/^([^:]+):/) {
1343 $storeid = $1;
1344 }
1345
1346 my $vtype = $cmdname eq 'restore' ? 'backup' : 'vztmpl';
1347 my $data = PVE::Storage::template_list($cfg, $storeid, $vtype);
1348
1349 my $res = [];
1350 foreach my $id (keys %$data) {
1351 foreach my $item (@{$data->{$id}}) {
1352 push @$res, $item->{volid} if defined($item->{volid});
1353 }
1354 }
1355
1356 return $res;
1357 }
1358
1359 my $complete_ctid_full = sub {
1360 my ($running) = @_;
1361
1362 my $idlist = vmstatus();
1363
1364 my $active_hash = list_active_containers();
1365
1366 my $res = [];
1367
1368 foreach my $id (keys %$idlist) {
1369 my $d = $idlist->{$id};
1370 if (defined($running)) {
1371 next if $d->{template};
1372 next if $running && !$active_hash->{$id};
1373 next if !$running && $active_hash->{$id};
1374 }
1375 push @$res, $id;
1376
1377 }
1378 return $res;
1379 };
1380
1381 sub complete_ctid {
1382 return &$complete_ctid_full();
1383 }
1384
1385 sub complete_ctid_stopped {
1386 return &$complete_ctid_full(0);
1387 }
1388
1389 sub complete_ctid_running {
1390 return &$complete_ctid_full(1);
1391 }
1392
1393 sub parse_id_maps {
1394 my ($conf) = @_;
1395
1396 my $id_map = [];
1397 my $rootuid = 0;
1398 my $rootgid = 0;
1399
1400 my $lxc = $conf->{lxc};
1401 foreach my $entry (@$lxc) {
1402 my ($key, $value) = @$entry;
1403 next if $key ne 'lxc.id_map';
1404 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1405 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1406 push @$id_map, [$type, $ct, $host, $length];
1407 if ($ct == 0) {
1408 $rootuid = $host if $type eq 'u';
1409 $rootgid = $host if $type eq 'g';
1410 }
1411 } else {
1412 die "failed to parse id_map: $value\n";
1413 }
1414 }
1415
1416 if (!@$id_map && $conf->{unprivileged}) {
1417 # Should we read them from /etc/subuid?
1418 $id_map = [ ['u', '0', '100000', '65536'],
1419 ['g', '0', '100000', '65536'] ];
1420 $rootuid = $rootgid = 100000;
1421 }
1422
1423 return ($id_map, $rootuid, $rootgid);
1424 }
1425
1426 sub userns_command {
1427 my ($id_map) = @_;
1428 if (@$id_map) {
1429 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];
1430 }
1431 return [];
1432 }
1433
1434
1435 1;