]> git.proxmox.com Git - pve-container.git/blob - src/PVE/LXC.pm
config: update for lxc-2.1
[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 ENOTDIR EROFS);
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::CpuSet;
22 use PVE::Network;
23 use PVE::AccessControl;
24 use PVE::ProcFSTools;
25 use PVE::Syscall;
26 use PVE::LXC::Config;
27 use Time::HiRes qw (gettimeofday);
28
29 my $nodename = PVE::INotify::nodename();
30
31 my $cpuinfo= PVE::ProcFSTools::read_cpuinfo();
32
33 sub config_list {
34 my $vmlist = PVE::Cluster::get_vmlist();
35 my $res = {};
36 return $res if !$vmlist || !$vmlist->{ids};
37 my $ids = $vmlist->{ids};
38
39 foreach my $vmid (keys %$ids) {
40 next if !$vmid; # skip CT0
41 my $d = $ids->{$vmid};
42 next if !$d->{node} || $d->{node} ne $nodename;
43 next if !$d->{type} || $d->{type} ne 'lxc';
44 $res->{$vmid}->{type} = 'lxc';
45 }
46 return $res;
47 }
48
49 sub destroy_config {
50 my ($vmid) = @_;
51
52 unlink PVE::LXC::Config->config_file($vmid, $nodename);
53 }
54
55 # container status helpers
56
57 sub list_active_containers {
58
59 my $filename = "/proc/net/unix";
60
61 # similar test is used by lcxcontainers.c: list_active_containers
62 my $res = {};
63
64 my $fh = IO::File->new ($filename, "r");
65 return $res if !$fh;
66
67 while (defined(my $line = <$fh>)) {
68 if ($line =~ m/^[a-f0-9]+:\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\d+\s+(\S+)$/) {
69 my $path = $1;
70 if ($path =~ m!^@/var/lib/lxc/(\d+)/command$!) {
71 $res->{$1} = 1;
72 }
73 }
74 }
75
76 close($fh);
77
78 return $res;
79 }
80
81 # warning: this is slow
82 sub check_running {
83 my ($vmid) = @_;
84
85 my $active_hash = list_active_containers();
86
87 return 1 if defined($active_hash->{$vmid});
88
89 return undef;
90 }
91
92 sub get_container_disk_usage {
93 my ($vmid, $pid) = @_;
94
95 return PVE::Tools::df("/proc/$pid/root/", 1);
96 }
97
98 my $last_proc_vmid_stat;
99
100 my $parse_cpuacct_stat = sub {
101 my ($vmid) = @_;
102
103 my $raw = read_cgroup_value('cpuacct', $vmid, 'cpuacct.stat', 1);
104
105 my $stat = {};
106
107 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
108
109 $stat->{utime} = $1;
110 $stat->{stime} = $2;
111
112 }
113
114 return $stat;
115 };
116
117 sub vmstatus {
118 my ($opt_vmid) = @_;
119
120 my $list = $opt_vmid ? { $opt_vmid => { type => 'lxc' }} : config_list();
121
122 my $active_hash = list_active_containers();
123
124 my $cpucount = $cpuinfo->{cpus} || 1;
125
126 my $cdtime = gettimeofday;
127
128 my $uptime = (PVE::ProcFSTools::read_proc_uptime(1))[0];
129
130 foreach my $vmid (keys %$list) {
131 my $d = $list->{$vmid};
132
133 eval { $d->{pid} = find_lxc_pid($vmid) if defined($active_hash->{$vmid}); };
134 warn $@ if $@; # ignore errors (consider them stopped)
135
136 $d->{status} = $d->{pid} ? 'running' : 'stopped';
137
138 my $cfspath = PVE::LXC::Config->cfs_config_path($vmid);
139 my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
140
141 $d->{name} = $conf->{'hostname'} || "CT$vmid";
142 $d->{name} =~ s/[\s]//g;
143
144 $d->{cpus} = $conf->{cores} || $conf->{cpulimit};
145 $d->{cpus} = $cpucount if !$d->{cpus};
146
147 $d->{lock} = $conf->{lock} || '';
148
149 if ($d->{pid}) {
150 my $res = get_container_disk_usage($vmid, $d->{pid});
151 $d->{disk} = $res->{used};
152 $d->{maxdisk} = $res->{total};
153 } else {
154 $d->{disk} = 0;
155 # use 4GB by default ??
156 if (my $rootfs = $conf->{rootfs}) {
157 my $rootinfo = PVE::LXC::Config->parse_ct_rootfs($rootfs);
158 $d->{maxdisk} = $rootinfo->{size} || (4*1024*1024*1024);
159 } else {
160 $d->{maxdisk} = 4*1024*1024*1024;
161 }
162 }
163
164 $d->{mem} = 0;
165 $d->{swap} = 0;
166 $d->{maxmem} = ($conf->{memory}||512)*1024*1024;
167 $d->{maxswap} = ($conf->{swap}//0)*1024*1024;
168
169 $d->{uptime} = 0;
170 $d->{cpu} = 0;
171
172 $d->{netout} = 0;
173 $d->{netin} = 0;
174
175 $d->{diskread} = 0;
176 $d->{diskwrite} = 0;
177
178 $d->{template} = PVE::LXC::Config->is_template($conf);
179 }
180
181 foreach my $vmid (keys %$list) {
182 my $d = $list->{$vmid};
183 my $pid = $d->{pid};
184
185 next if !$pid; # skip stopped CTs
186
187 my $ctime = (stat("/proc/$pid"))[10]; # 10 = ctime
188 $d->{uptime} = time - $ctime; # the method lxcfs uses
189
190 my $memory_stat = read_cgroup_list('memory', $vmid, 'memory.stat');
191 my $mem_usage_in_bytes = read_cgroup_value('memory', $vmid, 'memory.usage_in_bytes');
192
193 $d->{mem} = $mem_usage_in_bytes - $memory_stat->{total_cache};
194 $d->{swap} = read_cgroup_value('memory', $vmid, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
195
196 my $blkio_bytes = read_cgroup_value('blkio', $vmid, 'blkio.throttle.io_service_bytes', 1);
197 my @bytes = split(/\n/, $blkio_bytes);
198 foreach my $byte (@bytes) {
199 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
200 $d->{diskread} = $2 if $key eq 'Read';
201 $d->{diskwrite} = $2 if $key eq 'Write';
202 }
203 }
204
205 my $pstat = &$parse_cpuacct_stat($vmid);
206
207 my $used = $pstat->{utime} + $pstat->{stime};
208
209 my $old = $last_proc_vmid_stat->{$vmid};
210 if (!$old) {
211 $last_proc_vmid_stat->{$vmid} = {
212 time => $cdtime,
213 used => $used,
214 cpu => 0,
215 };
216 next;
217 }
218
219 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
220
221 if ($dtime > 1000) {
222 my $dutime = $used - $old->{used};
223
224 $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
225 $last_proc_vmid_stat->{$vmid} = {
226 time => $cdtime,
227 used => $used,
228 cpu => $d->{cpu},
229 };
230 } else {
231 $d->{cpu} = $old->{cpu};
232 }
233 }
234
235 my $netdev = PVE::ProcFSTools::read_proc_net_dev();
236
237 foreach my $dev (keys %$netdev) {
238 next if $dev !~ m/^veth([1-9]\d*)i/;
239 my $vmid = $1;
240 my $d = $list->{$vmid};
241
242 next if !$d;
243
244 $d->{netout} += $netdev->{$dev}->{receive};
245 $d->{netin} += $netdev->{$dev}->{transmit};
246
247 }
248
249 return $list;
250 }
251
252 sub read_cgroup_list {
253 my ($group, $vmid, $name) = @_;
254
255 my $content = read_cgroup_value($group, $vmid, $name, 1);
256
257 return { split(/\s+/, $content) };
258 }
259
260 sub read_cgroup_value {
261 my ($group, $vmid, $name, $full) = @_;
262
263 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
264
265 return PVE::Tools::file_get_contents($path) if $full;
266
267 return PVE::Tools::file_read_firstline($path);
268 }
269
270 sub write_cgroup_value {
271 my ($group, $vmid, $name, $value) = @_;
272
273 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
274 PVE::ProcFSTools::write_proc_entry($path, $value) if -e $path;
275
276 }
277
278 sub find_lxc_console_pids {
279
280 my $res = {};
281
282 PVE::Tools::dir_glob_foreach('/proc', '\d+', sub {
283 my ($pid) = @_;
284
285 my $cmdline = PVE::Tools::file_read_firstline("/proc/$pid/cmdline");
286 return if !$cmdline;
287
288 my @args = split(/\0/, $cmdline);
289
290 # search for lxc-console -n <vmid>
291 return if scalar(@args) != 3;
292 return if $args[1] ne '-n';
293 return if $args[2] !~ m/^\d+$/;
294 return if $args[0] !~ m|^(/usr/bin/)?lxc-console$|;
295
296 my $vmid = $args[2];
297
298 push @{$res->{$vmid}}, $pid;
299 });
300
301 return $res;
302 }
303
304 sub find_lxc_pid {
305 my ($vmid) = @_;
306
307 my $pid = undef;
308 my $parser = sub {
309 my $line = shift;
310 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
311 };
312 PVE::Tools::run_command(['lxc-info', '-n', $vmid, '-p'], outfunc => $parser);
313
314 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
315
316 return $pid;
317 }
318
319 # Note: we cannot use Net:IP, because that only allows strict
320 # CIDR networks
321 sub parse_ipv4_cidr {
322 my ($cidr, $noerr) = @_;
323
324 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
325 return { address => $1, netmask => $PVE::Network::ipv4_reverse_mask->[$2] };
326 }
327
328 return undef if $noerr;
329
330 die "unable to parse ipv4 address/mask\n";
331 }
332
333
334 sub update_lxc_config {
335 my ($vmid, $conf) = @_;
336
337 my $dir = "/var/lib/lxc/$vmid";
338
339 if ($conf->{template}) {
340
341 unlink "$dir/config";
342
343 return;
344 }
345
346 my $raw = '';
347
348 die "missing 'arch' - internal error" if !$conf->{arch};
349 $raw .= "lxc.arch = $conf->{arch}\n";
350
351 my $unprivileged = $conf->{unprivileged};
352 my $custom_idmap = grep { $_->[0] eq 'lxc.idmap' } @{$conf->{lxc}};
353
354 my $ostype = $conf->{ostype} || die "missing 'ostype' - internal error";
355
356 my $inc ="/usr/share/lxc/config/$ostype.common.conf";
357 $inc ="/usr/share/lxc/config/common.conf" if !-f $inc;
358 $raw .= "lxc.include = $inc\n";
359 if ($unprivileged || $custom_idmap) {
360 $inc = "/usr/share/lxc/config/$ostype.userns.conf";
361 $inc = "/usr/share/lxc/config/userns.conf" if !-f $inc;
362 $raw .= "lxc.include = $inc\n"
363 }
364
365 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
366 # cannot be exposed to the container with r/w access (cgroup perms).
367 # When this is enabled mounts will still remain in the monitor's namespace
368 # after the container unmounted them and thus will not detach from their
369 # files while the container is running!
370 $raw .= "lxc.monitor.unshare = 1\n";
371
372 # Should we read them from /etc/subuid?
373 if ($unprivileged && !$custom_idmap) {
374 $raw .= "lxc.idmap = u 0 100000 65536\n";
375 $raw .= "lxc.idmap = g 0 100000 65536\n";
376 }
377
378 if (!PVE::LXC::Config->has_dev_console($conf)) {
379 $raw .= "lxc.console.path = none\n";
380 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n";
381 }
382
383 my $ttycount = PVE::LXC::Config->get_tty_count($conf);
384 $raw .= "lxc.tty.max = $ttycount\n";
385
386 # some init scripts expect a linux terminal (turnkey).
387 $raw .= "lxc.environment = TERM=linux\n";
388
389 my $utsname = $conf->{hostname} || "CT$vmid";
390 $raw .= "lxc.uts.name = $utsname\n";
391
392 my $memory = $conf->{memory} || 512;
393 my $swap = $conf->{swap} // 0;
394
395 my $lxcmem = int($memory*1024*1024);
396 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
397
398 my $lxcswap = int(($memory + $swap)*1024*1024);
399 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
400
401 if (my $cpulimit = $conf->{cpulimit}) {
402 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
403 my $value = int(100000*$cpulimit);
404 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
405 }
406
407 my $shares = $conf->{cpuunits} || 1024;
408 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
409
410 die "missing 'rootfs' configuration\n"
411 if !defined($conf->{rootfs});
412
413 my $mountpoint = PVE::LXC::Config->parse_ct_rootfs($conf->{rootfs});
414
415 $raw .= "lxc.rootfs.path = $dir/rootfs\n";
416
417 foreach my $k (sort keys %$conf) {
418 next if $k !~ m/^net(\d+)$/;
419 my $ind = $1;
420 my $d = PVE::LXC::Config->parse_lxc_network($conf->{$k});
421 $raw .= "lxc.net.$ind.type = veth\n";
422 $raw .= "lxc.net.$ind.veth.pair = veth${vmid}i${ind}\n";
423 $raw .= "lxc.net.$ind.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr});
424 $raw .= "lxc.net.$ind.name = $d->{name}\n" if defined($d->{name});
425 $raw .= "lxc.net.$ind.mtu = $d->{mtu}\n" if defined($d->{mtu});
426 }
427
428 my $had_cpuset = 0;
429 if (my $lxcconf = $conf->{lxc}) {
430 foreach my $entry (@$lxcconf) {
431 my ($k, $v) = @$entry;
432 $had_cpuset = 1 if $k eq 'lxc.cgroup.cpuset.cpus';
433 $raw .= "$k = $v\n";
434 }
435 }
436
437 my $cores = $conf->{cores};
438 if (!$had_cpuset && $cores) {
439 my $cpuset = eval { PVE::CpuSet->new_from_cgroup('lxc', 'effective_cpus') };
440 $cpuset = PVE::CpuSet->new_from_cgroup('', 'effective_cpus') if !$cpuset;
441 my @members = $cpuset->members();
442 while (scalar(@members) > $cores) {
443 my $randidx = int(rand(scalar(@members)));
444 $cpuset->delete($members[$randidx]);
445 splice(@members, $randidx, 1); # keep track of the changes
446 }
447 $raw .= "lxc.cgroup.cpuset.cpus = ".$cpuset->short_string()."\n";
448 }
449
450 File::Path::mkpath("$dir/rootfs");
451
452 PVE::Tools::file_set_contents("$dir/config", $raw);
453 }
454
455 # verify and cleanup nameserver list (replace \0 with ' ')
456 sub verify_nameserver_list {
457 my ($nameserver_list) = @_;
458
459 my @list = ();
460 foreach my $server (PVE::Tools::split_list($nameserver_list)) {
461 PVE::JSONSchema::pve_verify_ip($server);
462 push @list, $server;
463 }
464
465 return join(' ', @list);
466 }
467
468 sub verify_searchdomain_list {
469 my ($searchdomain_list) = @_;
470
471 my @list = ();
472 foreach my $server (PVE::Tools::split_list($searchdomain_list)) {
473 # todo: should we add checks for valid dns domains?
474 push @list, $server;
475 }
476
477 return join(' ', @list);
478 }
479
480 sub get_console_command {
481 my ($vmid, $conf) = @_;
482
483 my $cmode = PVE::LXC::Config->get_cmode($conf);
484
485 if ($cmode eq 'console') {
486 return ['lxc-console', '-n', $vmid, '-t', 0];
487 } elsif ($cmode eq 'tty') {
488 return ['lxc-console', '-n', $vmid];
489 } elsif ($cmode eq 'shell') {
490 return ['lxc-attach', '--clear-env', '-n', $vmid];
491 } else {
492 die "internal error";
493 }
494 }
495
496 sub get_primary_ips {
497 my ($conf) = @_;
498
499 # return data from net0
500
501 return undef if !defined($conf->{net0});
502 my $net = PVE::LXC::Config->parse_lxc_network($conf->{net0});
503
504 my $ipv4 = $net->{ip};
505 if ($ipv4) {
506 if ($ipv4 =~ /^(dhcp|manual)$/) {
507 $ipv4 = undef
508 } else {
509 $ipv4 =~ s!/\d+$!!;
510 }
511 }
512 my $ipv6 = $net->{ip6};
513 if ($ipv6) {
514 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
515 $ipv6 = undef;
516 } else {
517 $ipv6 =~ s!/\d+$!!;
518 }
519 }
520
521 return ($ipv4, $ipv6);
522 }
523
524 sub delete_mountpoint_volume {
525 my ($storage_cfg, $vmid, $volume) = @_;
526
527 return if PVE::LXC::Config->classify_mountpoint($volume) ne 'volume';
528
529 my ($vtype, $name, $owner) = PVE::Storage::parse_volname($storage_cfg, $volume);
530 PVE::Storage::vdisk_free($storage_cfg, $volume) if $vmid == $owner;
531 }
532
533 sub destroy_lxc_container {
534 my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
535
536 PVE::LXC::Config->foreach_mountpoint($conf, sub {
537 my ($ms, $mountpoint) = @_;
538 delete_mountpoint_volume($storage_cfg, $vmid, $mountpoint->{volume});
539 });
540
541 rmdir "/var/lib/lxc/$vmid/rootfs";
542 unlink "/var/lib/lxc/$vmid/config";
543 rmdir "/var/lib/lxc/$vmid";
544 if (defined $replacement_conf) {
545 PVE::LXC::Config->write_config($vmid, $replacement_conf);
546 } else {
547 destroy_config($vmid);
548 }
549
550 #my $cmd = ['lxc-destroy', '-n', $vmid ];
551 #PVE::Tools::run_command($cmd);
552 }
553
554 sub vm_stop_cleanup {
555 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
556
557 eval {
558 if (!$keepActive) {
559
560 my $vollist = PVE::LXC::Config->get_vm_volumes($conf);
561 PVE::Storage::deactivate_volumes($storage_cfg, $vollist);
562 }
563 };
564 warn $@ if $@; # avoid errors - just warn
565 }
566
567 my $safe_num_ne = sub {
568 my ($a, $b) = @_;
569
570 return 0 if !defined($a) && !defined($b);
571 return 1 if !defined($a);
572 return 1 if !defined($b);
573
574 return $a != $b;
575 };
576
577 my $safe_string_ne = sub {
578 my ($a, $b) = @_;
579
580 return 0 if !defined($a) && !defined($b);
581 return 1 if !defined($a);
582 return 1 if !defined($b);
583
584 return $a ne $b;
585 };
586
587 sub update_net {
588 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
589
590 if ($newnet->{type} ne 'veth') {
591 # for when there are physical interfaces
592 die "cannot update interface of type $newnet->{type}";
593 }
594
595 my $veth = "veth${vmid}i${netid}";
596 my $eth = $newnet->{name};
597
598 if (my $oldnetcfg = $conf->{$opt}) {
599 my $oldnet = PVE::LXC::Config->parse_lxc_network($oldnetcfg);
600
601 if (&$safe_string_ne($oldnet->{hwaddr}, $newnet->{hwaddr}) ||
602 &$safe_string_ne($oldnet->{name}, $newnet->{name})) {
603
604 PVE::Network::veth_delete($veth);
605 delete $conf->{$opt};
606 PVE::LXC::Config->write_config($vmid, $conf);
607
608 hotplug_net($vmid, $conf, $opt, $newnet, $netid);
609
610 } else {
611 if (&$safe_string_ne($oldnet->{bridge}, $newnet->{bridge}) ||
612 &$safe_num_ne($oldnet->{tag}, $newnet->{tag}) ||
613 &$safe_num_ne($oldnet->{firewall}, $newnet->{firewall})) {
614
615 if ($oldnet->{bridge}) {
616 PVE::Network::tap_unplug($veth);
617 foreach (qw(bridge tag firewall)) {
618 delete $oldnet->{$_};
619 }
620 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($oldnet);
621 PVE::LXC::Config->write_config($vmid, $conf);
622 }
623
624 PVE::Network::tap_plug($veth, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
625 # This includes the rate:
626 foreach (qw(bridge tag firewall rate)) {
627 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
628 }
629 } elsif (&$safe_string_ne($oldnet->{rate}, $newnet->{rate})) {
630 # Rate can be applied on its own but any change above needs to
631 # include the rate in tap_plug since OVS resets everything.
632 PVE::Network::tap_rate_limit($veth, $newnet->{rate});
633 $oldnet->{rate} = $newnet->{rate}
634 }
635 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($oldnet);
636 PVE::LXC::Config->write_config($vmid, $conf);
637 }
638 } else {
639 hotplug_net($vmid, $conf, $opt, $newnet, $netid);
640 }
641
642 update_ipconfig($vmid, $conf, $opt, $eth, $newnet, $rootdir);
643 }
644
645 sub hotplug_net {
646 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
647
648 my $veth = "veth${vmid}i${netid}";
649 my $vethpeer = $veth . "p";
650 my $eth = $newnet->{name};
651
652 PVE::Network::veth_create($veth, $vethpeer, $newnet->{bridge}, $newnet->{hwaddr});
653 PVE::Network::tap_plug($veth, $newnet->{bridge}, $newnet->{tag}, $newnet->{firewall}, $newnet->{trunks}, $newnet->{rate});
654
655 # attach peer in container
656 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
657 PVE::Tools::run_command($cmd);
658
659 # link up peer in container
660 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
661 PVE::Tools::run_command($cmd);
662
663 my $done = { type => 'veth' };
664 foreach (qw(bridge tag firewall hwaddr name)) {
665 $done->{$_} = $newnet->{$_} if $newnet->{$_};
666 }
667 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($done);
668
669 PVE::LXC::Config->write_config($vmid, $conf);
670 }
671
672 sub update_ipconfig {
673 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
674
675 my $lxc_setup = PVE::LXC::Setup->new($conf, $rootdir);
676
677 my $optdata = PVE::LXC::Config->parse_lxc_network($conf->{$opt});
678 my $deleted = [];
679 my $added = [];
680 my $nscmd = sub {
681 my $cmdargs = shift;
682 PVE::Tools::run_command(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
683 };
684 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
685
686 my $change_ip_config = sub {
687 my ($ipversion) = @_;
688
689 my $family_opt = "-$ipversion";
690 my $suffix = $ipversion == 4 ? '' : $ipversion;
691 my $gw= "gw$suffix";
692 my $ip= "ip$suffix";
693
694 my $newip = $newnet->{$ip};
695 my $newgw = $newnet->{$gw};
696 my $oldip = $optdata->{$ip};
697
698 my $change_ip = &$safe_string_ne($oldip, $newip);
699 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
700
701 return if !$change_ip && !$change_gw;
702
703 # step 1: add new IP, if this fails we cancel
704 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
705 if ($change_ip && $is_real_ip) {
706 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
707 if (my $err = $@) {
708 warn $err;
709 return;
710 }
711 }
712
713 # step 2: replace gateway
714 # If this fails we delete the added IP and cancel.
715 # If it succeeds we save the config and delete the old IP, ignoring
716 # errors. The config is then saved.
717 # Note: 'ip route replace' can add
718 if ($change_gw) {
719 if ($newgw) {
720 eval {
721 if ($is_real_ip && !PVE::Network::is_ip_in_cidr($newgw, $newip, $ipversion)) {
722 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
723 }
724 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
725 };
726 if (my $err = $@) {
727 warn $err;
728 # the route was not replaced, the old IP is still available
729 # rollback (delete new IP) and cancel
730 if ($change_ip) {
731 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
732 warn $@ if $@; # no need to die here
733 }
734 return;
735 }
736 } else {
737 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
738 # if the route was not deleted, the guest might have deleted it manually
739 # warn and continue
740 warn $@ if $@;
741 }
742 }
743
744 # from this point on we save the configuration
745 # step 3: delete old IP ignoring errors
746 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
747 # We need to enable promote_secondaries, otherwise our newly added
748 # address will be removed along with the old one.
749 my $promote = 0;
750 eval {
751 if ($ipversion == 4) {
752 &$nscmd({ outfunc => sub { $promote = int(shift) } },
753 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
754 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
755 }
756 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
757 };
758 warn $@ if $@; # no need to die here
759
760 if ($ipversion == 4) {
761 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
762 }
763 }
764
765 foreach my $property ($ip, $gw) {
766 if ($newnet->{$property}) {
767 $optdata->{$property} = $newnet->{$property};
768 } else {
769 delete $optdata->{$property};
770 }
771 }
772 $conf->{$opt} = PVE::LXC::Config->print_lxc_network($optdata);
773 PVE::LXC::Config->write_config($vmid, $conf);
774 $lxc_setup->setup_network($conf);
775 };
776
777 &$change_ip_config(4);
778 &$change_ip_config(6);
779
780 }
781
782 my $enter_namespace = sub {
783 my ($vmid, $pid, $which, $type) = @_;
784 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
785 or die "failed to open $which namespace of container $vmid: $!\n";
786 PVE::Tools::setns(fileno($fd), $type)
787 or die "failed to enter $which namespace of container $vmid: $!\n";
788 close $fd;
789 };
790
791 my $do_syncfs = sub {
792 my ($vmid, $pid, $socket) = @_;
793
794 &$enter_namespace($vmid, $pid, 'mnt', PVE::Tools::CLONE_NEWNS);
795
796 # Tell the parent process to start reading our /proc/mounts
797 print {$socket} "go\n";
798 $socket->flush();
799
800 # Receive /proc/self/mounts
801 my $mountdata = do { local $/ = undef; <$socket> };
802 close $socket;
803
804 # Now sync all mountpoints...
805 my $mounts = PVE::ProcFSTools::parse_mounts($mountdata);
806 foreach my $mp (@$mounts) {
807 my ($what, $dir, $fs) = @$mp;
808 next if $fs eq 'fuse.lxcfs';
809 eval { PVE::Tools::sync_mountpoint($dir); };
810 warn $@ if $@;
811 }
812 };
813
814 sub sync_container_namespace {
815 my ($vmid) = @_;
816 my $pid = find_lxc_pid($vmid);
817
818 # SOCK_DGRAM is nicer for barriers but cannot be slurped
819 socketpair my $pfd, my $cfd, AF_UNIX, SOCK_STREAM, PF_UNSPEC
820 or die "failed to create socketpair: $!\n";
821
822 my $child = fork();
823 die "fork failed: $!\n" if !defined($child);
824
825 if (!$child) {
826 eval {
827 close $pfd;
828 &$do_syncfs($vmid, $pid, $cfd);
829 };
830 if (my $err = $@) {
831 warn $err;
832 POSIX::_exit(1);
833 }
834 POSIX::_exit(0);
835 }
836 close $cfd;
837 my $go = <$pfd>;
838 die "failed to enter container namespace\n" if $go ne "go\n";
839
840 open my $mounts, '<', "/proc/$child/mounts"
841 or die "failed to open container's /proc/mounts: $!\n";
842 my $mountdata = do { local $/ = undef; <$mounts> };
843 close $mounts;
844 print {$pfd} $mountdata;
845 close $pfd;
846
847 while (waitpid($child, 0) != $child) {}
848 die "failed to sync container namespace\n" if $? != 0;
849 }
850
851 sub template_create {
852 my ($vmid, $conf) = @_;
853
854 my $storecfg = PVE::Storage::config();
855
856 my $rootinfo = PVE::LXC::Config->parse_ct_rootfs($conf->{rootfs});
857 my $volid = $rootinfo->{volume};
858
859 die "Template feature is not available for '$volid'\n"
860 if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
861
862 PVE::Storage::activate_volumes($storecfg, [$volid]);
863
864 my $template_volid = PVE::Storage::vdisk_create_base($storecfg, $volid);
865 $rootinfo->{volume} = $template_volid;
866 $conf->{rootfs} = PVE::LXC::Config->print_ct_mountpoint($rootinfo, 1);
867
868 PVE::LXC::Config->write_config($vmid, $conf);
869 }
870
871 sub check_ct_modify_config_perm {
872 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
873
874 return 1 if $authuser eq 'root@pam';
875
876 my $check = sub {
877 my ($opt, $delete) = @_;
878 if ($opt eq 'cores' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
879 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
880 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
881 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
882 return if $delete;
883 my $data = $opt eq 'rootfs' ? PVE::LXC::Config->parse_ct_rootfs($newconf->{$opt})
884 : PVE::LXC::Config->parse_ct_mountpoint($newconf->{$opt});
885 raise_perm_exc("mount point type $data->{type} is only allowed for root\@pam")
886 if $data->{type} ne 'volume';
887 } elsif ($opt eq 'memory' || $opt eq 'swap') {
888 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
889 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
890 $opt eq 'searchdomain' || $opt eq 'hostname') {
891 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
892 } else {
893 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
894 }
895 };
896
897 foreach my $opt (keys %$newconf) {
898 &$check($opt, 0);
899 }
900 foreach my $opt (@$delete) {
901 &$check($opt, 1);
902 }
903
904 return 1;
905 }
906
907 sub umount_all {
908 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
909
910 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
911 my $volid_list = PVE::LXC::Config->get_vm_volumes($conf);
912
913 PVE::LXC::Config->foreach_mountpoint_reverse($conf, sub {
914 my ($ms, $mountpoint) = @_;
915
916 my $volid = $mountpoint->{volume};
917 my $mount = $mountpoint->{mp};
918
919 return if !$volid || !$mount;
920
921 my $mount_path = "$rootdir/$mount";
922 $mount_path =~ s!/+!/!g;
923
924 return if !PVE::ProcFSTools::is_mounted($mount_path);
925
926 eval {
927 PVE::Tools::run_command(['umount', '-d', $mount_path]);
928 };
929 if (my $err = $@) {
930 if ($noerr) {
931 warn $err;
932 } else {
933 die $err;
934 }
935 }
936 });
937 }
938
939 sub mount_all {
940 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
941
942 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
943 File::Path::make_path($rootdir);
944
945 my $volid_list = PVE::LXC::Config->get_vm_volumes($conf);
946 PVE::Storage::activate_volumes($storage_cfg, $volid_list);
947
948 eval {
949 PVE::LXC::Config->foreach_mountpoint($conf, sub {
950 my ($ms, $mountpoint) = @_;
951
952 $mountpoint->{ro} = 0 if $ignore_ro;
953
954 mountpoint_mount($mountpoint, $rootdir, $storage_cfg);
955 });
956 };
957 if (my $err = $@) {
958 warn "mounting container failed\n";
959 umount_all($vmid, $storage_cfg, $conf, 1);
960 die $err;
961 }
962
963 return $rootdir;
964 }
965
966
967 sub mountpoint_mount_path {
968 my ($mountpoint, $storage_cfg, $snapname) = @_;
969
970 return mountpoint_mount($mountpoint, undef, $storage_cfg, $snapname);
971 }
972
973 sub query_loopdev {
974 my ($path) = @_;
975 my $found;
976 my $parser = sub {
977 my $line = shift;
978 if ($line =~ m@^(/dev/loop\d+):@) {
979 $found = $1;
980 }
981 };
982 my $cmd = ['losetup', '--associated', $path];
983 PVE::Tools::run_command($cmd, outfunc => $parser);
984 return $found;
985 }
986
987 # Run a function with a file attached to a loop device.
988 # The loop device is always detached afterwards (or set to autoclear).
989 # Returns the loop device.
990 sub run_with_loopdev {
991 my ($func, $file) = @_;
992 my $device = query_loopdev($file);
993 # Try to reuse an existing device
994 if ($device) {
995 # We assume that whoever setup the loop device is responsible for
996 # detaching it.
997 &$func($device);
998 return $device;
999 }
1000
1001 my $parser = sub {
1002 my $line = shift;
1003 if ($line =~ m@^(/dev/loop\d+)$@) {
1004 $device = $1;
1005 }
1006 };
1007 PVE::Tools::run_command(['losetup', '--show', '-f', $file], outfunc => $parser);
1008 die "failed to setup loop device for $file\n" if !$device;
1009 eval { &$func($device); };
1010 my $err = $@;
1011 PVE::Tools::run_command(['losetup', '-d', $device]);
1012 die $err if $err;
1013 return $device;
1014 }
1015
1016 # In scalar mode: returns a file handle to the deepest directory node.
1017 # In list context: returns a list of:
1018 # * the deepest directory node
1019 # * the 2nd deepest directory (parent of the above)
1020 # * directory name of the last directory
1021 # So that the path $2/$3 should lead to $1 afterwards.
1022 sub walk_tree_nofollow($$$) {
1023 my ($start, $subdir, $mkdir) = @_;
1024
1025 # splitdir() returns '' for empty components including the leading /
1026 my @comps = grep { length($_)>0 } File::Spec->splitdir($subdir);
1027
1028 sysopen(my $fd, $start, O_PATH | O_DIRECTORY)
1029 or die "failed to open start directory $start: $!\n";
1030
1031 my $dir = $start;
1032 my $last_component = undef;
1033 my $second = $fd;
1034 foreach my $component (@comps) {
1035 $dir .= "/$component";
1036 my $next = PVE::Tools::openat(fileno($fd), $component, O_NOFOLLOW | O_DIRECTORY);
1037
1038 if (!$next) {
1039 # failed, check for symlinks and try to create the path
1040 die "symlink encountered at: $dir\n" if $! == ELOOP || $! == ENOTDIR;
1041 die "cannot open directory $dir: $!\n" if !$mkdir;
1042
1043 # We don't check for errors on mkdirat() here and just try to
1044 # openat() again, since at least one error (EEXIST) is an
1045 # expected possibility if multiple containers start
1046 # simultaneously. If someone else injects a symlink now then
1047 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1048 PVE::Tools::mkdirat(fileno($fd), $component, 0755);
1049
1050 $next = PVE::Tools::openat(fileno($fd), $component, O_NOFOLLOW | O_DIRECTORY);
1051 die "failed to create path: $dir: $!\n" if !$next;
1052 }
1053
1054 close $second if defined($last_component);
1055 $last_component = $component;
1056 $second = $fd;
1057 $fd = $next;
1058 }
1059
1060 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1061 close $second if defined($last_component);
1062 return $fd;
1063 }
1064
1065 # To guard against symlink attack races against other currently running
1066 # containers with shared recursive bind mount hierarchies we prepare a
1067 # directory handle for the directory we're mounting over to verify the
1068 # mountpoint afterwards.
1069 sub __bindmount_prepare {
1070 my ($hostroot, $dir) = @_;
1071 my $srcdh = walk_tree_nofollow($hostroot, $dir, 0);
1072 return $srcdh;
1073 }
1074
1075 # Assuming we mount to rootfs/a/b/c, verify with the directory handle to 'b'
1076 # ($parentfd) that 'b/c' (openat($parentfd, 'c')) really leads to the directory
1077 # we intended to bind mount.
1078 sub __bindmount_verify {
1079 my ($srcdh, $parentfd, $last_dir, $ro) = @_;
1080 my $destdh;
1081 if ($parentfd) {
1082 # Open the mount point path coming from the parent directory since the
1083 # filehandle we would have gotten as first result of walk_tree_nofollow
1084 # earlier is still a handle to the underlying directory instead of the
1085 # mounted path.
1086 $destdh = PVE::Tools::openat(fileno($parentfd), $last_dir, PVE::Tools::O_PATH | O_NOFOLLOW | O_DIRECTORY);
1087 die "failed to open mount point: $!\n" if !$destdh;
1088 if ($ro) {
1089 my $dot = '.';
1090 # no separate function because 99% of the time it's the wrong thing to use.
1091 if (syscall(PVE::Syscall::faccessat, fileno($destdh), $dot, &POSIX::W_OK, 0) != -1) {
1092 die "failed to mark bind mount read only\n";
1093 }
1094 die "read-only check failed: $!\n" if $! != EROFS;
1095 }
1096 } else {
1097 # For the rootfs we don't have a parentfd so we open the path directly.
1098 # Note that this means bindmounting any prefix of the host's
1099 # /var/lib/lxc/$vmid path into another container is considered a grave
1100 # security error.
1101 sysopen $destdh, $last_dir, O_PATH | O_DIRECTORY;
1102 die "failed to open mount point: $!\n" if !$destdh;
1103 }
1104
1105 my ($srcdev, $srcinode) = stat($srcdh);
1106 my ($dstdev, $dstinode) = stat($destdh);
1107 close $srcdh;
1108 close $destdh;
1109
1110 return ($srcdev == $dstdev && $srcinode == $dstinode);
1111 }
1112
1113 # Perform the actual bind mounting:
1114 sub __bindmount_do {
1115 my ($dir, $dest, $ro, @extra_opts) = @_;
1116 PVE::Tools::run_command(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1117 if ($ro) {
1118 eval { PVE::Tools::run_command(['mount', '-o', 'bind,remount,ro', $dest]); };
1119 if (my $err = $@) {
1120 warn "bindmount error\n";
1121 # don't leave writable bind-mounts behind...
1122 PVE::Tools::run_command(['umount', $dest]);
1123 die $err;
1124 }
1125 }
1126 }
1127
1128 sub bindmount {
1129 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1130
1131 my $srcdh = __bindmount_prepare('/', $dir);
1132
1133 __bindmount_do($dir, $dest, $ro, @extra_opts);
1134
1135 if (!__bindmount_verify($srcdh, $parentfd, $last_dir, $ro)) {
1136 PVE::Tools::run_command(['umount', $dest]);
1137 die "detected mount path change at: $dir\n";
1138 }
1139 }
1140
1141 # Cleanup $rootdir a bit (double and trailing slashes), build the mount path
1142 # from $rootdir and $mount and walk the path from $rootdir to the final
1143 # directory to check for symlinks.
1144 sub __mount_prepare_rootdir {
1145 my ($rootdir, $mount) = @_;
1146 $rootdir =~ s!/+!/!g;
1147 $rootdir =~ s!/+$!!;
1148 my $mount_path = "$rootdir/$mount";
1149 my ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow($rootdir, $mount, 1);
1150 return ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir);
1151 }
1152
1153 # use $rootdir = undef to just return the corresponding mount path
1154 sub mountpoint_mount {
1155 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1156
1157 my $volid = $mountpoint->{volume};
1158 my $mount = $mountpoint->{mp};
1159 my $type = $mountpoint->{type};
1160 my $quota = !$snapname && !$mountpoint->{ro} && $mountpoint->{quota};
1161 my $mounted_dev;
1162
1163 return if !$volid || !$mount;
1164
1165 $mount =~ s!/+!/!g;
1166
1167 my $mount_path;
1168 my ($mpfd, $parentfd, $last_dir);
1169
1170 if (defined($rootdir)) {
1171 ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir) =
1172 __mount_prepare_rootdir($rootdir, $mount);
1173 }
1174
1175 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1176
1177 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1178
1179 my $optstring = '';
1180 my $acl = $mountpoint->{acl};
1181 if (defined($acl)) {
1182 $optstring .= ($acl ? 'acl' : 'noacl');
1183 }
1184 my $readonly = $mountpoint->{ro};
1185
1186 my @extra_opts = ('-o', $optstring) if $optstring;
1187
1188 if ($storage) {
1189
1190 my $scfg = PVE::Storage::storage_config($storage_cfg, $storage);
1191
1192 # early sanity checks:
1193 # we otherwise call realpath on the rbd url
1194 die "containers on rbd storage without krbd are not supported\n"
1195 if $scfg->{type} eq 'rbd' && !$scfg->{krbd};
1196
1197 my $path = PVE::Storage::path($storage_cfg, $volid, $snapname);
1198
1199 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1200 PVE::Storage::parse_volname($storage_cfg, $volid);
1201
1202 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1203
1204 if ($format eq 'subvol') {
1205 if ($mount_path) {
1206 if ($snapname) {
1207 if ($scfg->{type} eq 'zfspool') {
1208 my $path_arg = $path;
1209 $path_arg =~ s!^/+!!;
1210 PVE::Tools::run_command(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1211 } else {
1212 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1213 }
1214 } else {
1215 if (defined($acl) && $scfg->{type} eq 'zfspool') {
1216 my $acltype = ($acl ? 'acltype=posixacl' : 'acltype=noacl');
1217 my (undef, $name) = PVE::Storage::parse_volname($storage_cfg, $volid);
1218 $name .= "\@$snapname" if defined($snapname);
1219 PVE::Tools::run_command(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1220 }
1221 bindmount($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1222 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1223 }
1224 }
1225 return wantarray ? ($path, 0, undef) : $path;
1226 } elsif ($format eq 'raw' || $format eq 'iso') {
1227 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1228 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1229 # Our autodev hook expects the /dev/dm-* device currently
1230 # and will create the /dev/mapper symlink accordingly
1231 $path = Cwd::realpath($path);
1232 die "failed to get device path\n" if !$path;
1233 ($path) = ($path =~ /^(.*)$/s); #untaint
1234 my $domount = sub {
1235 my ($path) = @_;
1236 if ($mount_path) {
1237 if ($format eq 'iso') {
1238 PVE::Tools::run_command(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1239 } elsif ($isBase || defined($snapname)) {
1240 PVE::Tools::run_command(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1241 } else {
1242 if ($quota) {
1243 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1244 }
1245 push @extra_opts, '-o', 'ro' if $readonly;
1246 PVE::Tools::run_command(['mount', @extra_opts, $path, $mount_path]);
1247 }
1248 }
1249 };
1250 my $use_loopdev = 0;
1251 if ($scfg->{path}) {
1252 $mounted_dev = run_with_loopdev($domount, $path);
1253 $use_loopdev = 1;
1254 } elsif ($scfg->{type} eq 'drbd' || $scfg->{type} eq 'lvm' ||
1255 $scfg->{type} eq 'rbd' || $scfg->{type} eq 'lvmthin') {
1256 $mounted_dev = $path;
1257 &$domount($path);
1258 } else {
1259 die "unsupported storage type '$scfg->{type}'\n";
1260 }
1261 return wantarray ? ($path, $use_loopdev, $mounted_dev) : $path;
1262 } else {
1263 die "unsupported image format '$format'\n";
1264 }
1265 } elsif ($type eq 'device') {
1266 push @extra_opts, '-o', 'ro' if $readonly;
1267 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1268 # See the NOTE above about devicemapper canonicalization
1269 my ($devpath) = (Cwd::realpath($volid) =~ /^(.*)$/s); # realpath() taints
1270 PVE::Tools::run_command(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1271 return wantarray ? ($volid, 0, $devpath) : $volid;
1272 } elsif ($type eq 'bind') {
1273 die "directory '$volid' does not exist\n" if ! -d $volid;
1274 bindmount($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1275 warn "cannot enable quota control for bind mounts\n" if $quota;
1276 return wantarray ? ($volid, 0, undef) : $volid;
1277 }
1278
1279 die "unsupported storage";
1280 }
1281
1282 sub mkfs {
1283 my ($dev, $rootuid, $rootgid) = @_;
1284
1285 PVE::Tools::run_command(['mkfs.ext4', '-O', 'mmp',
1286 '-E', "root_owner=$rootuid:$rootgid",
1287 $dev]);
1288 }
1289
1290 sub format_disk {
1291 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1292
1293 if ($volid =~ m!^/dev/.+!) {
1294 mkfs($volid);
1295 return;
1296 }
1297
1298 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1299
1300 die "cannot format volume '$volid' with no storage\n" if !$storage;
1301
1302 PVE::Storage::activate_volumes($storage_cfg, [$volid]);
1303
1304 my $path = PVE::Storage::path($storage_cfg, $volid);
1305
1306 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1307 PVE::Storage::parse_volname($storage_cfg, $volid);
1308
1309 die "cannot format volume '$volid' (format == $format)\n"
1310 if $format ne 'raw';
1311
1312 mkfs($path, $rootuid, $rootgid);
1313 }
1314
1315 sub destroy_disks {
1316 my ($storecfg, $vollist) = @_;
1317
1318 foreach my $volid (@$vollist) {
1319 eval { PVE::Storage::vdisk_free($storecfg, $volid); };
1320 warn $@ if $@;
1321 }
1322 }
1323
1324 our $NEW_DISK_RE = qr/^([^:\s]+):(\d+(\.\d+)?)$/;
1325 sub create_disks {
1326 my ($storecfg, $vmid, $settings, $conf) = @_;
1327
1328 my $vollist = [];
1329
1330 eval {
1331 my (undef, $rootuid, $rootgid) = PVE::LXC::parse_id_maps($conf);
1332 my $chown_vollist = [];
1333
1334 PVE::LXC::Config->foreach_mountpoint($settings, sub {
1335 my ($ms, $mountpoint) = @_;
1336
1337 my $volid = $mountpoint->{volume};
1338 my $mp = $mountpoint->{mp};
1339
1340 my ($storage, $volname) = PVE::Storage::parse_volume_id($volid, 1);
1341
1342 if ($storage && ($volid =~ $NEW_DISK_RE)) {
1343 my ($storeid, $size_gb) = ($1, $2);
1344
1345 my $size_kb = int(${size_gb}*1024) * 1024;
1346
1347 my $scfg = PVE::Storage::storage_config($storecfg, $storage);
1348 # fixme: use better naming ct-$vmid-disk-X.raw?
1349
1350 if ($scfg->{type} eq 'dir' || $scfg->{type} eq 'nfs') {
1351 if ($size_kb > 0) {
1352 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw',
1353 undef, $size_kb);
1354 format_disk($storecfg, $volid, $rootuid, $rootgid);
1355 } else {
1356 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'subvol',
1357 undef, 0);
1358 push @$chown_vollist, $volid;
1359 }
1360 } elsif ($scfg->{type} eq 'zfspool') {
1361
1362 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'subvol',
1363 undef, $size_kb);
1364 push @$chown_vollist, $volid;
1365 } elsif ($scfg->{type} eq 'drbd' || $scfg->{type} eq 'lvm' || $scfg->{type} eq 'lvmthin') {
1366
1367 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1368 format_disk($storecfg, $volid, $rootuid, $rootgid);
1369
1370 } elsif ($scfg->{type} eq 'rbd') {
1371
1372 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd};
1373 $volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1374 format_disk($storecfg, $volid, $rootuid, $rootgid);
1375 } else {
1376 die "unable to create containers on storage type '$scfg->{type}'\n";
1377 }
1378 push @$vollist, $volid;
1379 $mountpoint->{volume} = $volid;
1380 $mountpoint->{size} = $size_kb * 1024;
1381 $conf->{$ms} = PVE::LXC::Config->print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1382 } else {
1383 # use specified/existing volid/dir/device
1384 $conf->{$ms} = PVE::LXC::Config->print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1385 }
1386 });
1387
1388 PVE::Storage::activate_volumes($storecfg, $chown_vollist, undef);
1389 foreach my $volid (@$chown_vollist) {
1390 my $path = PVE::Storage::path($storecfg, $volid, undef);
1391 chown($rootuid, $rootgid, $path);
1392 }
1393 PVE::Storage::deactivate_volumes($storecfg, $chown_vollist, undef);
1394 };
1395 # free allocated images on error
1396 if (my $err = $@) {
1397 destroy_disks($storecfg, $vollist);
1398 die $err;
1399 }
1400 return $vollist;
1401 }
1402
1403 # bash completion helper
1404
1405 sub complete_os_templates {
1406 my ($cmdname, $pname, $cvalue) = @_;
1407
1408 my $cfg = PVE::Storage::config();
1409
1410 my $storeid;
1411
1412 if ($cvalue =~ m/^([^:]+):/) {
1413 $storeid = $1;
1414 }
1415
1416 my $vtype = $cmdname eq 'restore' ? 'backup' : 'vztmpl';
1417 my $data = PVE::Storage::template_list($cfg, $storeid, $vtype);
1418
1419 my $res = [];
1420 foreach my $id (keys %$data) {
1421 foreach my $item (@{$data->{$id}}) {
1422 push @$res, $item->{volid} if defined($item->{volid});
1423 }
1424 }
1425
1426 return $res;
1427 }
1428
1429 my $complete_ctid_full = sub {
1430 my ($running) = @_;
1431
1432 my $idlist = vmstatus();
1433
1434 my $active_hash = list_active_containers();
1435
1436 my $res = [];
1437
1438 foreach my $id (keys %$idlist) {
1439 my $d = $idlist->{$id};
1440 if (defined($running)) {
1441 next if $d->{template};
1442 next if $running && !$active_hash->{$id};
1443 next if !$running && $active_hash->{$id};
1444 }
1445 push @$res, $id;
1446
1447 }
1448 return $res;
1449 };
1450
1451 sub complete_ctid {
1452 return &$complete_ctid_full();
1453 }
1454
1455 sub complete_ctid_stopped {
1456 return &$complete_ctid_full(0);
1457 }
1458
1459 sub complete_ctid_running {
1460 return &$complete_ctid_full(1);
1461 }
1462
1463 sub parse_id_maps {
1464 my ($conf) = @_;
1465
1466 my $id_map = [];
1467 my $rootuid = 0;
1468 my $rootgid = 0;
1469
1470 my $lxc = $conf->{lxc};
1471 foreach my $entry (@$lxc) {
1472 my ($key, $value) = @$entry;
1473 # FIXME: remove the 'id_map' variant when lxc-3.0 arrives
1474 next if $key ne 'lxc.idmap' && $key ne 'lxc.id_map';
1475 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1476 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1477 push @$id_map, [$type, $ct, $host, $length];
1478 if ($ct == 0) {
1479 $rootuid = $host if $type eq 'u';
1480 $rootgid = $host if $type eq 'g';
1481 }
1482 } else {
1483 die "failed to parse idmap: $value\n";
1484 }
1485 }
1486
1487 if (!@$id_map && $conf->{unprivileged}) {
1488 # Should we read them from /etc/subuid?
1489 $id_map = [ ['u', '0', '100000', '65536'],
1490 ['g', '0', '100000', '65536'] ];
1491 $rootuid = $rootgid = 100000;
1492 }
1493
1494 return ($id_map, $rootuid, $rootgid);
1495 }
1496
1497 sub userns_command {
1498 my ($id_map) = @_;
1499 if (@$id_map) {
1500 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];
1501 }
1502 return [];
1503 }
1504
1505
1506 1;