13 use Fcntl
qw(O_RDONLY O_NOFOLLOW O_DIRECTORY);
14 use Errno
qw(ELOOP ENOTDIR EROFS ECONNREFUSED);
17 use PVE
::Exception
qw(raise_perm_exc);
21 use PVE
::Tools
qw($IPV6RE $IPV4RE dir_glob_foreach lock_file lock_file_full O_PATH);
24 use PVE
::AccessControl
;
28 use Time
::HiRes qw
(gettimeofday
);
30 my $nodename = PVE
::INotify
::nodename
();
32 my $cpuinfo= PVE
::ProcFSTools
::read_cpuinfo
();
35 my $vmlist = PVE
::Cluster
::get_vmlist
();
37 return $res if !$vmlist || !$vmlist->{ids
};
38 my $ids = $vmlist->{ids
};
40 foreach my $vmid (keys %$ids) {
41 next if !$vmid; # skip CT0
42 my $d = $ids->{$vmid};
43 next if !$d->{node
} || $d->{node
} ne $nodename;
44 next if !$d->{type
} || $d->{type
} ne 'lxc';
45 $res->{$vmid}->{type
} = 'lxc';
53 unlink PVE
::LXC
::Config-
>config_file($vmid, $nodename);
56 # container status helpers
58 sub list_active_containers
{
60 my $filename = "/proc/net/unix";
62 # similar test is used by lcxcontainers.c: list_active_containers
65 my $fh = IO
::File-
>new ($filename, "r");
68 while (defined(my $line = <$fh>)) {
69 if ($line =~ m/^[a-f0-9]+:\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\d+\s+(\S+)$/) {
71 if ($path =~ m!^@/var/lib/lxc/(\d+)/command$!) {
82 # warning: this is slow
86 my $active_hash = list_active_containers
();
88 return 1 if defined($active_hash->{$vmid});
93 sub get_container_disk_usage
{
94 my ($vmid, $pid) = @_;
96 return PVE
::Tools
::df
("/proc/$pid/root/", 1);
99 my $last_proc_vmid_stat;
101 my $parse_cpuacct_stat = sub {
102 my ($vmid, $unprivileged) = @_;
104 my $raw = read_cgroup_value
('cpuacct', $vmid, $unprivileged, 'cpuacct.stat', 1);
108 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
121 my $list = $opt_vmid ?
{ $opt_vmid => { type
=> 'lxc' }} : config_list
();
123 my $active_hash = list_active_containers
();
125 my $cpucount = $cpuinfo->{cpus
} || 1;
127 my $cdtime = gettimeofday
;
129 my $uptime = (PVE
::ProcFSTools
::read_proc_uptime
(1))[0];
130 my $clock_ticks = POSIX
::sysconf
(&POSIX
::_SC_CLK_TCK
);
132 my $unprivileged = {};
134 foreach my $vmid (keys %$list) {
135 my $d = $list->{$vmid};
137 eval { $d->{pid
} = find_lxc_pid
($vmid) if defined($active_hash->{$vmid}); };
138 warn $@ if $@; # ignore errors (consider them stopped)
140 $d->{status
} = $d->{pid
} ?
'running' : 'stopped';
142 my $cfspath = PVE
::LXC
::Config-
>cfs_config_path($vmid);
143 my $conf = PVE
::Cluster
::cfs_read_file
($cfspath) || {};
145 $unprivileged->{$vmid} = $conf->{unprivileged
};
147 $d->{name
} = $conf->{'hostname'} || "CT$vmid";
148 $d->{name
} =~ s/[\s]//g;
150 $d->{cpus
} = $conf->{cores
} || $conf->{cpulimit
};
151 $d->{cpus
} = $cpucount if !$d->{cpus
};
153 $d->{lock} = $conf->{lock} || '';
156 my $res = get_container_disk_usage
($vmid, $d->{pid
});
157 $d->{disk
} = $res->{used
};
158 $d->{maxdisk
} = $res->{total
};
161 # use 4GB by default ??
162 if (my $rootfs = $conf->{rootfs
}) {
163 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($rootfs);
164 $d->{maxdisk
} = $rootinfo->{size
} || (4*1024*1024*1024);
166 $d->{maxdisk
} = 4*1024*1024*1024;
172 $d->{maxmem
} = ($conf->{memory
}||512)*1024*1024;
173 $d->{maxswap
} = ($conf->{swap
}//0)*1024*1024;
184 $d->{template
} = PVE
::LXC
::Config-
>is_template($conf);
187 foreach my $vmid (keys %$list) {
188 my $d = $list->{$vmid};
191 next if !$pid; # skip stopped CTs
193 my $proc_pid_stat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
194 $d->{uptime
} = int(($uptime - $proc_pid_stat->{starttime
}) / $clock_ticks); # the method lxcfs uses
196 my $unpriv = $unprivileged->{$vmid};
198 if (-d
'/sys/fs/cgroup/memory') {
199 my $memory_stat = read_cgroup_list
('memory', $vmid, $unpriv, 'memory.stat');
200 my $mem_usage_in_bytes = read_cgroup_value
('memory', $vmid, $unpriv, 'memory.usage_in_bytes');
202 $d->{mem
} = $mem_usage_in_bytes - $memory_stat->{total_cache
};
203 $d->{swap
} = read_cgroup_value
('memory', $vmid, $unpriv, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
209 if (-d
'/sys/fs/cgroup/blkio') {
210 my $blkio_bytes = read_cgroup_value
('blkio', $vmid, $unpriv, 'blkio.throttle.io_service_bytes', 1);
211 my @bytes = split(/\n/, $blkio_bytes);
212 foreach my $byte (@bytes) {
213 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
214 $d->{diskread
} += $2 if $key eq 'Read';
215 $d->{diskwrite
} += $2 if $key eq 'Write';
223 if (-d
'/sys/fs/cgroup/cpuacct') {
224 my $pstat = $parse_cpuacct_stat->($vmid, $unpriv);
226 my $used = $pstat->{utime} + $pstat->{stime
};
228 my $old = $last_proc_vmid_stat->{$vmid};
230 $last_proc_vmid_stat->{$vmid} = {
238 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
241 my $dutime = $used - $old->{used
};
243 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
244 $last_proc_vmid_stat->{$vmid} = {
250 $d->{cpu
} = $old->{cpu
};
257 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
259 foreach my $dev (keys %$netdev) {
260 next if $dev !~ m/^veth([1-9]\d*)i/;
262 my $d = $list->{$vmid};
266 $d->{netout
} += $netdev->{$dev}->{receive
};
267 $d->{netin
} += $netdev->{$dev}->{transmit
};
274 sub read_cgroup_list
($$$$) {
275 my ($group, $vmid, $unprivileged, $name) = @_;
277 my $content = read_cgroup_value
($group, $vmid, $unprivileged, $name, 1);
279 return { split(/\s+/, $content) };
282 sub read_cgroup_value
($$$$$) {
283 my ($group, $vmid, $unprivileged, $name, $full) = @_;
285 my $nsdir = $unprivileged ?
'' : 'ns/';
286 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/${nsdir}$name";
288 return PVE
::Tools
::file_get_contents
($path) if $full;
290 return PVE
::Tools
::file_read_firstline
($path);
293 sub write_cgroup_value
{
294 my ($group, $vmid, $name, $value) = @_;
296 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
297 PVE
::ProcFSTools
::write_proc_entry
($path, $value) if -e
$path;
301 sub find_lxc_console_pids
{
305 PVE
::Tools
::dir_glob_foreach
('/proc', '\d+', sub {
308 my $cmdline = PVE
::Tools
::file_read_firstline
("/proc/$pid/cmdline");
311 my @args = split(/\0/, $cmdline);
313 # search for lxc-console -n <vmid>
314 return if scalar(@args) != 3;
315 return if $args[1] ne '-n';
316 return if $args[2] !~ m/^\d+$/;
317 return if $args[0] !~ m
|^(/usr/bin
/)?lxc-console
$|;
321 push @{$res->{$vmid}}, $pid;
333 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
335 PVE
::Tools
::run_command
(['lxc-info', '-n', $vmid, '-p'], outfunc
=> $parser);
337 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
342 # Note: we cannot use Net:IP, because that only allows strict
344 sub parse_ipv4_cidr
{
345 my ($cidr, $noerr) = @_;
347 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
348 return { address
=> $1, netmask
=> $PVE::Network
::ipv4_reverse_mask-
>[$2] };
351 return undef if $noerr;
353 die "unable to parse ipv4 address/mask\n";
356 sub get_cgroup_subsystems
{
359 my $data = PVE
::Tools
::file_get_contents
('/proc/self/cgroup');
360 while ($data =~ /^\d+:([^:\n]*):.*$/gm) {
363 $v1->{$_} = 1 foreach split(/,/, $type);
368 return wantarray ?
($v1, $v2) : $v1;
371 sub update_lxc_config
{
372 my ($vmid, $conf) = @_;
374 my $dir = "/var/lib/lxc/$vmid";
376 if ($conf->{template
}) {
378 unlink "$dir/config";
385 die "missing 'arch' - internal error" if !$conf->{arch
};
386 $raw .= "lxc.arch = $conf->{arch}\n";
388 my $unprivileged = $conf->{unprivileged
};
389 my $custom_idmap = grep { $_->[0] eq 'lxc.idmap' } @{$conf->{lxc
}};
391 my $ostype = $conf->{ostype
} || die "missing 'ostype' - internal error";
393 my $cfgpath = '/usr/share/lxc/config';
394 my $inc = "$cfgpath/$ostype.common.conf";
395 $inc ="$cfgpath/common.conf" if !-f
$inc;
396 $raw .= "lxc.include = $inc\n";
397 if ($unprivileged || $custom_idmap) {
398 $inc = "$cfgpath/$ostype.userns.conf";
399 $inc = "$cfgpath/userns.conf" if !-f
$inc;
400 $raw .= "lxc.include = $inc\n";
401 $raw .= "lxc.seccomp.profile = $cfgpath/pve-userns.seccomp\n";
404 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
405 # cannot be exposed to the container with r/w access (cgroup perms).
406 # When this is enabled mounts will still remain in the monitor's namespace
407 # after the container unmounted them and thus will not detach from their
408 # files while the container is running!
409 $raw .= "lxc.monitor.unshare = 1\n";
411 my $cgv1 = get_cgroup_subsystems
();
413 # Should we read them from /etc/subuid?
414 if ($unprivileged && !$custom_idmap) {
415 $raw .= "lxc.idmap = u 0 100000 65536\n";
416 $raw .= "lxc.idmap = g 0 100000 65536\n";
419 if (!PVE
::LXC
::Config-
>has_dev_console($conf)) {
420 $raw .= "lxc.console.path = none\n";
421 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n" if $cgv1->{devices
};
424 my $ttycount = PVE
::LXC
::Config-
>get_tty_count($conf);
425 $raw .= "lxc.tty.max = $ttycount\n";
427 # some init scripts expect a linux terminal (turnkey).
428 $raw .= "lxc.environment = TERM=linux\n";
430 my $utsname = $conf->{hostname
} || "CT$vmid";
431 $raw .= "lxc.uts.name = $utsname\n";
433 if ($cgv1->{memory
}) {
434 my $memory = $conf->{memory
} || 512;
435 my $swap = $conf->{swap
} // 0;
437 my $lxcmem = int($memory*1024*1024);
438 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
440 my $lxcswap = int(($memory + $swap)*1024*1024);
441 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
445 if (my $cpulimit = $conf->{cpulimit
}) {
446 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
447 my $value = int(100000*$cpulimit);
448 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
451 my $shares = $conf->{cpuunits
} || 1024;
452 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
455 die "missing 'rootfs' configuration\n"
456 if !defined($conf->{rootfs
});
458 my $mountpoint = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
460 $raw .= "lxc.rootfs.path = $dir/rootfs\n";
462 foreach my $k (sort keys %$conf) {
463 next if $k !~ m/^net(\d+)$/;
465 my $d = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$k});
466 $raw .= "lxc.net.$ind.type = veth\n";
467 $raw .= "lxc.net.$ind.veth.pair = veth${vmid}i${ind}\n";
468 $raw .= "lxc.net.$ind.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr
});
469 $raw .= "lxc.net.$ind.name = $d->{name}\n" if defined($d->{name
});
470 $raw .= "lxc.net.$ind.mtu = $d->{mtu}\n" if defined($d->{mtu
});
473 if ($cgv1->{cpuset
}) {
475 if (my $lxcconf = $conf->{lxc
}) {
476 foreach my $entry (@$lxcconf) {
477 my ($k, $v) = @$entry;
478 $had_cpuset = 1 if $k eq 'lxc.cgroup.cpuset.cpus';
483 my $cores = $conf->{cores
};
484 if (!$had_cpuset && $cores) {
485 my $cpuset = eval { PVE
::CpuSet-
>new_from_cgroup('lxc', 'effective_cpus') };
486 $cpuset = PVE
::CpuSet-
>new_from_cgroup('', 'effective_cpus') if !$cpuset;
487 my @members = $cpuset->members();
488 while (scalar(@members) > $cores) {
489 my $randidx = int(rand(scalar(@members)));
490 $cpuset->delete($members[$randidx]);
491 splice(@members, $randidx, 1); # keep track of the changes
493 $raw .= "lxc.cgroup.cpuset.cpus = ".$cpuset->short_string()."\n";
497 File
::Path
::mkpath
("$dir/rootfs");
499 PVE
::Tools
::file_set_contents
("$dir/config", $raw);
502 # verify and cleanup nameserver list (replace \0 with ' ')
503 sub verify_nameserver_list
{
504 my ($nameserver_list) = @_;
507 foreach my $server (PVE
::Tools
::split_list
($nameserver_list)) {
508 PVE
::JSONSchema
::pve_verify_ip
($server);
512 return join(' ', @list);
515 sub verify_searchdomain_list
{
516 my ($searchdomain_list) = @_;
519 foreach my $server (PVE
::Tools
::split_list
($searchdomain_list)) {
520 # todo: should we add checks for valid dns domains?
524 return join(' ', @list);
527 sub get_console_command
{
528 my ($vmid, $conf, $noescapechar) = @_;
530 my $cmode = PVE
::LXC
::Config-
>get_cmode($conf);
533 if ($cmode eq 'console') {
534 push @$cmd, 'lxc-console', '-n', $vmid, '-t', 0;
535 push @$cmd, '-e', -1 if $noescapechar;
536 } elsif ($cmode eq 'tty') {
537 push @$cmd, 'lxc-console', '-n', $vmid;
538 push @$cmd, '-e', -1 if $noescapechar;
539 } elsif ($cmode eq 'shell') {
540 push @$cmd, 'lxc-attach', '--clear-env', '-n', $vmid;
542 die "internal error";
548 sub get_primary_ips
{
551 # return data from net0
553 return undef if !defined($conf->{net0
});
554 my $net = PVE
::LXC
::Config-
>parse_lxc_network($conf->{net0
});
556 my $ipv4 = $net->{ip
};
558 if ($ipv4 =~ /^(dhcp|manual)$/) {
564 my $ipv6 = $net->{ip6
};
566 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
573 return ($ipv4, $ipv6);
576 sub delete_mountpoint_volume
{
577 my ($storage_cfg, $vmid, $volume) = @_;
579 return if PVE
::LXC
::Config-
>classify_mountpoint($volume) ne 'volume';
581 my ($vtype, $name, $owner) = PVE
::Storage
::parse_volname
($storage_cfg, $volume);
582 PVE
::Storage
::vdisk_free
($storage_cfg, $volume) if $vmid == $owner;
585 sub destroy_lxc_container
{
586 my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
588 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
589 my ($ms, $mountpoint) = @_;
590 delete_mountpoint_volume
($storage_cfg, $vmid, $mountpoint->{volume
});
593 rmdir "/var/lib/lxc/$vmid/rootfs";
594 unlink "/var/lib/lxc/$vmid/config";
595 rmdir "/var/lib/lxc/$vmid";
596 if (defined $replacement_conf) {
597 PVE
::LXC
::Config-
>write_config($vmid, $replacement_conf);
599 destroy_config
($vmid);
602 #my $cmd = ['lxc-destroy', '-n', $vmid ];
603 #PVE::Tools::run_command($cmd);
606 sub vm_stop_cleanup
{
607 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
612 my $vollist = PVE
::LXC
::Config-
>get_vm_volumes($conf);
613 PVE
::Storage
::deactivate_volumes
($storage_cfg, $vollist);
616 warn $@ if $@; # avoid errors - just warn
619 my $safe_num_ne = sub {
622 return 0 if !defined($a) && !defined($b);
623 return 1 if !defined($a);
624 return 1 if !defined($b);
629 my $safe_string_ne = sub {
632 return 0 if !defined($a) && !defined($b);
633 return 1 if !defined($a);
634 return 1 if !defined($b);
640 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
642 if ($newnet->{type
} ne 'veth') {
643 # for when there are physical interfaces
644 die "cannot update interface of type $newnet->{type}";
647 my $veth = "veth${vmid}i${netid}";
648 my $eth = $newnet->{name
};
650 if (my $oldnetcfg = $conf->{$opt}) {
651 my $oldnet = PVE
::LXC
::Config-
>parse_lxc_network($oldnetcfg);
653 if (&$safe_string_ne($oldnet->{hwaddr
}, $newnet->{hwaddr
}) ||
654 &$safe_string_ne($oldnet->{name
}, $newnet->{name
})) {
656 PVE
::Network
::veth_delete
($veth);
657 delete $conf->{$opt};
658 PVE
::LXC
::Config-
>write_config($vmid, $conf);
660 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
663 if (&$safe_string_ne($oldnet->{bridge
}, $newnet->{bridge
}) ||
664 &$safe_num_ne($oldnet->{tag
}, $newnet->{tag
}) ||
665 &$safe_num_ne($oldnet->{firewall
}, $newnet->{firewall
})) {
667 if ($oldnet->{bridge
}) {
668 PVE
::Network
::tap_unplug
($veth);
669 foreach (qw(bridge tag firewall)) {
670 delete $oldnet->{$_};
672 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
673 PVE
::LXC
::Config-
>write_config($vmid, $conf);
676 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
677 # This includes the rate:
678 foreach (qw(bridge tag firewall rate)) {
679 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
681 } elsif (&$safe_string_ne($oldnet->{rate
}, $newnet->{rate
})) {
682 # Rate can be applied on its own but any change above needs to
683 # include the rate in tap_plug since OVS resets everything.
684 PVE
::Network
::tap_rate_limit
($veth, $newnet->{rate
});
685 $oldnet->{rate
} = $newnet->{rate
}
687 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
688 PVE
::LXC
::Config-
>write_config($vmid, $conf);
691 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
694 update_ipconfig
($vmid, $conf, $opt, $eth, $newnet, $rootdir);
698 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
700 my $veth = "veth${vmid}i${netid}";
701 my $vethpeer = $veth . "p";
702 my $eth = $newnet->{name
};
704 PVE
::Network
::veth_create
($veth, $vethpeer, $newnet->{bridge
}, $newnet->{hwaddr
});
705 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
707 # attach peer in container
708 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
709 PVE
::Tools
::run_command
($cmd);
711 # link up peer in container
712 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
713 PVE
::Tools
::run_command
($cmd);
715 my $done = { type
=> 'veth' };
716 foreach (qw(bridge tag firewall hwaddr name)) {
717 $done->{$_} = $newnet->{$_} if $newnet->{$_};
719 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($done);
721 PVE
::LXC
::Config-
>write_config($vmid, $conf);
724 sub update_ipconfig
{
725 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
727 my $lxc_setup = PVE
::LXC
::Setup-
>new($conf, $rootdir);
729 my $optdata = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$opt});
734 PVE
::Tools
::run_command
(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
736 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
738 my $change_ip_config = sub {
739 my ($ipversion) = @_;
741 my $family_opt = "-$ipversion";
742 my $suffix = $ipversion == 4 ?
'' : $ipversion;
746 my $newip = $newnet->{$ip};
747 my $newgw = $newnet->{$gw};
748 my $oldip = $optdata->{$ip};
750 my $change_ip = &$safe_string_ne($oldip, $newip);
751 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
753 return if !$change_ip && !$change_gw;
755 # step 1: add new IP, if this fails we cancel
756 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
757 if ($change_ip && $is_real_ip) {
758 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
765 # step 2: replace gateway
766 # If this fails we delete the added IP and cancel.
767 # If it succeeds we save the config and delete the old IP, ignoring
768 # errors. The config is then saved.
769 # Note: 'ip route replace' can add
773 if ($is_real_ip && !PVE
::Network
::is_ip_in_cidr
($newgw, $newip, $ipversion)) {
774 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
776 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
780 # the route was not replaced, the old IP is still available
781 # rollback (delete new IP) and cancel
783 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
784 warn $@ if $@; # no need to die here
789 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
790 # if the route was not deleted, the guest might have deleted it manually
796 # from this point on we save the configuration
797 # step 3: delete old IP ignoring errors
798 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
799 # We need to enable promote_secondaries, otherwise our newly added
800 # address will be removed along with the old one.
803 if ($ipversion == 4) {
804 &$nscmd({ outfunc
=> sub { $promote = int(shift) } },
805 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
806 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
808 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
810 warn $@ if $@; # no need to die here
812 if ($ipversion == 4) {
813 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
817 foreach my $property ($ip, $gw) {
818 if ($newnet->{$property}) {
819 $optdata->{$property} = $newnet->{$property};
821 delete $optdata->{$property};
824 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($optdata);
825 PVE
::LXC
::Config-
>write_config($vmid, $conf);
826 $lxc_setup->setup_network($conf);
829 &$change_ip_config(4);
830 &$change_ip_config(6);
834 my $enter_namespace = sub {
835 my ($vmid, $pid, $which, $type) = @_;
836 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
837 or die "failed to open $which namespace of container $vmid: $!\n";
838 PVE
::Tools
::setns
(fileno($fd), $type)
839 or die "failed to enter $which namespace of container $vmid: $!\n";
843 my $do_syncfs = sub {
844 my ($vmid, $pid, $socket) = @_;
846 &$enter_namespace($vmid, $pid, 'mnt', PVE
::Tools
::CLONE_NEWNS
);
848 # Tell the parent process to start reading our /proc/mounts
849 print {$socket} "go\n";
852 # Receive /proc/self/mounts
853 my $mountdata = do { local $/ = undef; <$socket> };
856 # Now sync all mountpoints...
857 my $mounts = PVE
::ProcFSTools
::parse_mounts
($mountdata);
858 foreach my $mp (@$mounts) {
859 my ($what, $dir, $fs) = @$mp;
860 next if $fs eq 'fuse.lxcfs';
861 eval { PVE
::Tools
::sync_mountpoint
($dir); };
866 sub sync_container_namespace
{
868 my $pid = find_lxc_pid
($vmid);
870 # SOCK_DGRAM is nicer for barriers but cannot be slurped
871 socketpair my $pfd, my $cfd, AF_UNIX
, SOCK_STREAM
, PF_UNSPEC
872 or die "failed to create socketpair: $!\n";
875 die "fork failed: $!\n" if !defined($child);
880 &$do_syncfs($vmid, $pid, $cfd);
890 die "failed to enter container namespace\n" if $go ne "go\n";
892 open my $mounts, '<', "/proc/$child/mounts"
893 or die "failed to open container's /proc/mounts: $!\n";
894 my $mountdata = do { local $/ = undef; <$mounts> };
896 print {$pfd} $mountdata;
899 while (waitpid($child, 0) != $child) {}
900 die "failed to sync container namespace\n" if $? != 0;
903 sub template_create
{
904 my ($vmid, $conf) = @_;
906 my $storecfg = PVE
::Storage
::config
();
908 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
909 my ($ms, $mountpoint) = @_;
911 my $volid = $mountpoint->{volume
};
913 die "Template feature is not available for '$volid'\n"
914 if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
917 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
918 my ($ms, $mountpoint) = @_;
920 my $volid = $mountpoint->{volume
};
922 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
924 my $template_volid = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
925 $mountpoint->{volume
} = $template_volid;
926 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq "rootfs");
929 PVE
::LXC
::Config-
>write_config($vmid, $conf);
932 sub check_ct_modify_config_perm
{
933 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
935 return 1 if $authuser eq 'root@pam';
938 my ($opt, $delete) = @_;
939 if ($opt eq 'cores' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
940 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
941 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
942 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
944 my $data = $opt eq 'rootfs' ? PVE
::LXC
::Config-
>parse_ct_rootfs($newconf->{$opt})
945 : PVE
::LXC
::Config-
>parse_ct_mountpoint($newconf->{$opt});
946 raise_perm_exc
("mount point type $data->{type} is only allowed for root\@pam")
947 if $data->{type
} ne 'volume';
948 } elsif ($opt eq 'memory' || $opt eq 'swap') {
949 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
950 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
951 $opt eq 'searchdomain' || $opt eq 'hostname') {
952 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
954 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
958 foreach my $opt (keys %$newconf) {
961 foreach my $opt (@$delete) {
969 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
971 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
972 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
974 PVE
::LXC
::Config-
>foreach_mountpoint_reverse($conf, sub {
975 my ($ms, $mountpoint) = @_;
977 my $volid = $mountpoint->{volume
};
978 my $mount = $mountpoint->{mp
};
980 return if !$volid || !$mount;
982 my $mount_path = "$rootdir/$mount";
983 $mount_path =~ s!/+!/!g;
985 return if !PVE
::ProcFSTools
::is_mounted
($mount_path);
988 PVE
::Tools
::run_command
(['umount', '-d', $mount_path]);
1001 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
1003 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
1004 File
::Path
::make_path
($rootdir);
1006 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
1007 PVE
::Storage
::activate_volumes
($storage_cfg, $volid_list);
1010 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
1011 my ($ms, $mountpoint) = @_;
1013 $mountpoint->{ro
} = 0 if $ignore_ro;
1015 mountpoint_mount
($mountpoint, $rootdir, $storage_cfg);
1019 warn "mounting container failed\n";
1020 umount_all
($vmid, $storage_cfg, $conf, 1);
1028 sub mountpoint_mount_path
{
1029 my ($mountpoint, $storage_cfg, $snapname) = @_;
1031 return mountpoint_mount
($mountpoint, undef, $storage_cfg, $snapname);
1039 if ($line =~ m
@^(/dev/loop\d
+):@) {
1043 my $cmd = ['losetup', '--associated', $path];
1044 PVE
::Tools
::run_command
($cmd, outfunc
=> $parser);
1048 # Run a function with a file attached to a loop device.
1049 # The loop device is always detached afterwards (or set to autoclear).
1050 # Returns the loop device.
1051 sub run_with_loopdev
{
1052 my ($func, $file) = @_;
1053 my $device = query_loopdev
($file);
1054 # Try to reuse an existing device
1056 # We assume that whoever setup the loop device is responsible for
1064 if ($line =~ m
@^(/dev/loop\d
+)$@) {
1068 PVE
::Tools
::run_command
(['losetup', '--show', '-f', $file], outfunc
=> $parser);
1069 die "failed to setup loop device for $file\n" if !$device;
1070 eval { &$func($device); };
1072 PVE
::Tools
::run_command
(['losetup', '-d', $device]);
1077 # In scalar mode: returns a file handle to the deepest directory node.
1078 # In list context: returns a list of:
1079 # * the deepest directory node
1080 # * the 2nd deepest directory (parent of the above)
1081 # * directory name of the last directory
1082 # So that the path $2/$3 should lead to $1 afterwards.
1083 sub walk_tree_nofollow
($$$) {
1084 my ($start, $subdir, $mkdir) = @_;
1086 # splitdir() returns '' for empty components including the leading /
1087 my @comps = grep { length($_)>0 } File
::Spec-
>splitdir($subdir);
1089 sysopen(my $fd, $start, O_PATH
| O_DIRECTORY
)
1090 or die "failed to open start directory $start: $!\n";
1093 my $last_component = undef;
1095 foreach my $component (@comps) {
1096 $dir .= "/$component";
1097 my $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1100 # failed, check for symlinks and try to create the path
1101 die "symlink encountered at: $dir\n" if $! == ELOOP
|| $! == ENOTDIR
;
1102 die "cannot open directory $dir: $!\n" if !$mkdir;
1104 # We don't check for errors on mkdirat() here and just try to
1105 # openat() again, since at least one error (EEXIST) is an
1106 # expected possibility if multiple containers start
1107 # simultaneously. If someone else injects a symlink now then
1108 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1109 PVE
::Tools
::mkdirat
(fileno($fd), $component, 0755);
1111 $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1112 die "failed to create path: $dir: $!\n" if !$next;
1115 close $second if defined($last_component);
1116 $last_component = $component;
1121 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1122 close $second if defined($last_component);
1126 # To guard against symlink attack races against other currently running
1127 # containers with shared recursive bind mount hierarchies we prepare a
1128 # directory handle for the directory we're mounting over to verify the
1129 # mountpoint afterwards.
1130 sub __bindmount_prepare
{
1131 my ($hostroot, $dir) = @_;
1132 my $srcdh = walk_tree_nofollow
($hostroot, $dir, 0);
1136 # Assuming we mount to rootfs/a/b/c, verify with the directory handle to 'b'
1137 # ($parentfd) that 'b/c' (openat($parentfd, 'c')) really leads to the directory
1138 # we intended to bind mount.
1139 sub __bindmount_verify
{
1140 my ($srcdh, $parentfd, $last_dir, $ro) = @_;
1143 # Open the mount point path coming from the parent directory since the
1144 # filehandle we would have gotten as first result of walk_tree_nofollow
1145 # earlier is still a handle to the underlying directory instead of the
1147 $destdh = PVE
::Tools
::openat
(fileno($parentfd), $last_dir, PVE
::Tools
::O_PATH
| O_NOFOLLOW
| O_DIRECTORY
);
1148 die "failed to open mount point: $!\n" if !$destdh;
1151 # no separate function because 99% of the time it's the wrong thing to use.
1152 if (syscall(PVE
::Syscall
::faccessat
, fileno($destdh), $dot, &POSIX
::W_OK
, 0) != -1) {
1153 die "failed to mark bind mount read only\n";
1155 die "read-only check failed: $!\n" if $! != EROFS
;
1158 # For the rootfs we don't have a parentfd so we open the path directly.
1159 # Note that this means bindmounting any prefix of the host's
1160 # /var/lib/lxc/$vmid path into another container is considered a grave
1162 sysopen $destdh, $last_dir, O_PATH
| O_DIRECTORY
;
1163 die "failed to open mount point: $!\n" if !$destdh;
1166 my ($srcdev, $srcinode) = stat($srcdh);
1167 my ($dstdev, $dstinode) = stat($destdh);
1171 return ($srcdev == $dstdev && $srcinode == $dstinode);
1174 # Perform the actual bind mounting:
1175 sub __bindmount_do
{
1176 my ($dir, $dest, $ro, @extra_opts) = @_;
1177 PVE
::Tools
::run_command
(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1179 eval { PVE
::Tools
::run_command
(['mount', '-o', 'bind,remount,ro', $dest]); };
1181 warn "bindmount error\n";
1182 # don't leave writable bind-mounts behind...
1183 PVE
::Tools
::run_command
(['umount', $dest]);
1190 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1192 my $srcdh = __bindmount_prepare
('/', $dir);
1194 __bindmount_do
($dir, $dest, $ro, @extra_opts);
1196 if (!__bindmount_verify
($srcdh, $parentfd, $last_dir, $ro)) {
1197 PVE
::Tools
::run_command
(['umount', $dest]);
1198 die "detected mount path change at: $dir\n";
1202 # Cleanup $rootdir a bit (double and trailing slashes), build the mount path
1203 # from $rootdir and $mount and walk the path from $rootdir to the final
1204 # directory to check for symlinks.
1205 sub __mount_prepare_rootdir
{
1206 my ($rootdir, $mount) = @_;
1207 $rootdir =~ s!/+!/!g;
1208 $rootdir =~ s!/+$!!;
1209 my $mount_path = "$rootdir/$mount";
1210 my ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow
($rootdir, $mount, 1);
1211 return ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir);
1214 # use $rootdir = undef to just return the corresponding mount path
1215 sub mountpoint_mount
{
1216 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1218 my $volid = $mountpoint->{volume
};
1219 my $mount = $mountpoint->{mp
};
1220 my $type = $mountpoint->{type
};
1221 my $quota = !$snapname && !$mountpoint->{ro
} && $mountpoint->{quota
};
1224 return if !$volid || !$mount;
1229 my ($mpfd, $parentfd, $last_dir);
1231 if (defined($rootdir)) {
1232 ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir) =
1233 __mount_prepare_rootdir
($rootdir, $mount);
1236 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1238 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1241 my $acl = $mountpoint->{acl
};
1242 if (defined($acl)) {
1243 $optstring .= ($acl ?
'acl' : 'noacl');
1245 my $readonly = $mountpoint->{ro
};
1248 @extra_opts = ('-o', $optstring) if $optstring;
1252 my $scfg = PVE
::Storage
::storage_config
($storage_cfg, $storage);
1254 # early sanity checks:
1255 # we otherwise call realpath on the rbd url
1256 die "containers on rbd storage without krbd are not supported\n"
1257 if $scfg->{type
} eq 'rbd' && !$scfg->{krbd
};
1259 my $path = PVE
::Storage
::path
($storage_cfg, $volid, $snapname);
1261 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1262 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1264 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1266 if ($format eq 'subvol') {
1269 if ($scfg->{type
} eq 'zfspool') {
1270 my $path_arg = $path;
1271 $path_arg =~ s!^/+!!;
1272 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1274 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1277 if (defined($acl) && $scfg->{type
} eq 'zfspool') {
1278 my $acltype = ($acl ?
'acltype=posixacl' : 'acltype=noacl');
1279 my (undef, $name) = PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1280 $name .= "\@$snapname" if defined($snapname);
1281 PVE
::Tools
::run_command
(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1283 bindmount
($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1284 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1287 return wantarray ?
($path, 0, undef) : $path;
1288 } elsif ($format eq 'raw' || $format eq 'iso') {
1289 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1290 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1291 # Our autodev hook expects the /dev/dm-* device currently
1292 # and will create the /dev/mapper symlink accordingly
1293 $path = Cwd
::realpath
($path);
1294 die "failed to get device path\n" if !$path;
1295 ($path) = ($path =~ /^(.*)$/s); #untaint
1299 if ($format eq 'iso') {
1300 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1301 } elsif ($isBase || defined($snapname)) {
1302 PVE
::Tools
::run_command
(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1305 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1307 push @extra_opts, '-o', 'ro' if $readonly;
1308 PVE
::Tools
::run_command
(['mount', @extra_opts, $path, $mount_path]);
1312 my $use_loopdev = 0;
1313 if ($scfg->{path
}) {
1314 $mounted_dev = run_with_loopdev
($domount, $path);
1316 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' ||
1317 $scfg->{type
} eq 'rbd' || $scfg->{type
} eq 'lvmthin') {
1318 $mounted_dev = $path;
1321 die "unsupported storage type '$scfg->{type}'\n";
1323 return wantarray ?
($path, $use_loopdev, $mounted_dev) : $path;
1325 die "unsupported image format '$format'\n";
1327 } elsif ($type eq 'device') {
1328 push @extra_opts, '-o', 'ro' if $readonly;
1329 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1330 # See the NOTE above about devicemapper canonicalization
1331 my ($devpath) = (Cwd
::realpath
($volid) =~ /^(.*)$/s); # realpath() taints
1332 PVE
::Tools
::run_command
(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1333 return wantarray ?
($volid, 0, $devpath) : $volid;
1334 } elsif ($type eq 'bind') {
1335 die "directory '$volid' does not exist\n" if ! -d
$volid;
1336 bindmount
($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1337 warn "cannot enable quota control for bind mounts\n" if $quota;
1338 return wantarray ?
($volid, 0, undef) : $volid;
1341 die "unsupported storage";
1345 my ($dev, $rootuid, $rootgid) = @_;
1347 PVE
::Tools
::run_command
(['mkfs.ext4', '-O', 'mmp',
1348 '-E', "root_owner=$rootuid:$rootgid",
1353 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1355 if ($volid =~ m!^/dev/.+!) {
1360 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1362 die "cannot format volume '$volid' with no storage\n" if !$storage;
1364 PVE
::Storage
::activate_volumes
($storage_cfg, [$volid]);
1366 my $path = PVE
::Storage
::path
($storage_cfg, $volid);
1368 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1369 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1371 die "cannot format volume '$volid' (format == $format)\n"
1372 if $format ne 'raw';
1374 mkfs
($path, $rootuid, $rootgid);
1378 my ($storecfg, $vollist) = @_;
1380 foreach my $volid (@$vollist) {
1381 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1387 my ($storecfg, $vmid, $storage, $size_kb, $rootuid, $rootgid) = @_;
1389 my $needs_chown = 0;
1392 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storage);
1393 # fixme: use better naming ct-$vmid-disk-X.raw?
1397 if ($scfg->{type
} eq 'dir' || $scfg->{type
} eq 'nfs' || $scfg->{type
} eq 'cifs' ) {
1399 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw',
1403 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1407 } elsif ($scfg->{type
} eq 'zfspool') {
1409 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1412 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' || $scfg->{type
} eq 'lvmthin') {
1414 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1417 } elsif ($scfg->{type
} eq 'rbd') {
1419 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd
};
1420 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1423 die "unable to create containers on storage type '$scfg->{type}'\n";
1425 format_disk
($storecfg, $volid, $rootuid, $rootgid) if $do_format;
1428 # in case formatting got interrupted:
1429 if (defined($volid)) {
1430 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1436 return ($volid, $needs_chown);
1439 our $NEW_DISK_RE = qr/^([^:\s]+):(\d+(\.\d+)?)$/;
1441 my ($storecfg, $vmid, $settings, $conf) = @_;
1446 my (undef, $rootuid, $rootgid) = PVE
::LXC
::parse_id_maps
($conf);
1447 my $chown_vollist = [];
1449 PVE
::LXC
::Config-
>foreach_mountpoint($settings, sub {
1450 my ($ms, $mountpoint) = @_;
1452 my $volid = $mountpoint->{volume
};
1453 my $mp = $mountpoint->{mp
};
1455 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1457 if ($storage && ($volid =~ $NEW_DISK_RE)) {
1458 my ($storeid, $size_gb) = ($1, $2);
1460 my $size_kb = int(${size_gb
}*1024) * 1024;
1462 my $needs_chown = 0;
1463 ($volid, $needs_chown) = alloc_disk
($storecfg, $vmid, $storage, $size_kb, $rootuid, $rootgid);
1464 push @$chown_vollist, $volid if $needs_chown;
1465 push @$vollist, $volid;
1466 $mountpoint->{volume
} = $volid;
1467 $mountpoint->{size
} = $size_kb * 1024;
1468 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1470 # use specified/existing volid/dir/device
1471 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1475 PVE
::Storage
::activate_volumes
($storecfg, $chown_vollist, undef);
1476 foreach my $volid (@$chown_vollist) {
1477 my $path = PVE
::Storage
::path
($storecfg, $volid, undef);
1478 chown($rootuid, $rootgid, $path);
1480 PVE
::Storage
::deactivate_volumes
($storecfg, $chown_vollist, undef);
1482 # free allocated images on error
1484 destroy_disks
($storecfg, $vollist);
1490 # bash completion helper
1492 sub complete_os_templates
{
1493 my ($cmdname, $pname, $cvalue) = @_;
1495 my $cfg = PVE
::Storage
::config
();
1499 if ($cvalue =~ m/^([^:]+):/) {
1503 my $vtype = $cmdname eq 'restore' ?
'backup' : 'vztmpl';
1504 my $data = PVE
::Storage
::template_list
($cfg, $storeid, $vtype);
1507 foreach my $id (keys %$data) {
1508 foreach my $item (@{$data->{$id}}) {
1509 push @$res, $item->{volid
} if defined($item->{volid
});
1516 my $complete_ctid_full = sub {
1519 my $idlist = vmstatus
();
1521 my $active_hash = list_active_containers
();
1525 foreach my $id (keys %$idlist) {
1526 my $d = $idlist->{$id};
1527 if (defined($running)) {
1528 next if $d->{template
};
1529 next if $running && !$active_hash->{$id};
1530 next if !$running && $active_hash->{$id};
1539 return &$complete_ctid_full();
1542 sub complete_ctid_stopped
{
1543 return &$complete_ctid_full(0);
1546 sub complete_ctid_running
{
1547 return &$complete_ctid_full(1);
1557 my $lxc = $conf->{lxc
};
1558 foreach my $entry (@$lxc) {
1559 my ($key, $value) = @$entry;
1560 # FIXME: remove the 'id_map' variant when lxc-3.0 arrives
1561 next if $key ne 'lxc.idmap' && $key ne 'lxc.id_map';
1562 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1563 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1564 push @$id_map, [$type, $ct, $host, $length];
1566 $rootuid = $host if $type eq 'u';
1567 $rootgid = $host if $type eq 'g';
1570 die "failed to parse idmap: $value\n";
1574 if (!@$id_map && $conf->{unprivileged
}) {
1575 # Should we read them from /etc/subuid?
1576 $id_map = [ ['u', '0', '100000', '65536'],
1577 ['g', '0', '100000', '65536'] ];
1578 $rootuid = $rootgid = 100000;
1581 return ($id_map, $rootuid, $rootgid);
1584 sub userns_command
{
1587 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];
1593 my ($vmid, $conf, $skiplock) = @_;
1595 update_lxc_config
($vmid, $conf);
1597 my $skiplock_flag_fn = "/run/lxc/skiplock-$vmid";
1600 open(my $fh, '>', $skiplock_flag_fn) || die "failed to open $skiplock_flag_fn for writing: $!\n";
1604 my $cmd = ['systemctl', 'start', "pve-container\@$vmid"];
1606 eval { PVE
::Tools
::run_command
($cmd); };
1608 unlink $skiplock_flag_fn;
1615 # Helper to stop a container completely and make sure it has stopped completely.
1616 # This is necessary because we want the post-stop hook to have completed its
1617 # unmount-all step, but post-stop happens after lxc puts the container into the
1620 my ($vmid, $kill, $shutdown_timeout, $exit_timeout) = @_;
1622 # Open the container's command socket.
1623 my $path = "\0/var/lib/lxc/$vmid/command";
1624 my $sock = IO
::Socket
::UNIX-
>new(
1625 Type
=> SOCK_STREAM
(),
1629 return if $! == ECONNREFUSED
; # The container is not running
1630 die "failed to open container ${vmid}'s command socket: $!\n";
1633 # Stop the container:
1635 my $cmd = ['lxc-stop', '-n', $vmid];
1638 push @$cmd, '--kill'; # doesn't allow timeouts
1639 } elsif (defined($shutdown_timeout)) {
1640 push @$cmd, '--timeout', $shutdown_timeout;
1641 # Give run_command 5 extra seconds
1642 $shutdown_timeout += 5;
1645 eval { PVE
::Tools
::run_command
($cmd, timeout
=> $shutdown_timeout) };
1651 my $wait = sub { $result = <$sock>; };
1652 if (defined($exit_timeout)) {
1653 PVE
::Tools
::run_with_timeout
($exit_timeout, $wait);
1658 return if !defined $result; # monitor is gone and the ct has stopped.
1659 die "container did not stop\n";
1665 return PVE
::Tools
::run_fork
(sub {
1666 # Unshare the mount namespace
1667 die "failed to unshare mount namespace: $!\n"
1668 if !PVE
::Tools
::unshare
(PVE
::Tools
::CLONE_NEWNS
);
1669 PVE
::Tools
::run_command
(['mount', '--make-rslave', '/']);
1674 my $copy_volume = sub {
1675 my ($src_volid, $src, $dst_volid, $dest, $storage_cfg, $snapname) = @_;
1677 my $src_mp = { volume
=> $src_volid, mp
=> '/' };
1678 $src_mp->{type
} = PVE
::LXC
::Config-
>classify_mountpoint($src_volid);
1680 my $dst_mp = { volume
=> $dst_volid, mp
=> '/' };
1681 $dst_mp->{type
} = PVE
::LXC
::Config-
>classify_mountpoint($dst_volid);
1687 mountpoint_mount
($src_mp, $src, $storage_cfg, $snapname);
1688 push @mounted, $src;
1690 mountpoint_mount
($dst_mp, $dest, $storage_cfg);
1691 push @mounted, $dest;
1693 PVE
::Tools
::run_command
(['/usr/bin/rsync', '--stats', '-X', '-A', '--numeric-ids',
1694 '-aH', '--whole-file', '--sparse', '--one-file-system',
1698 foreach my $mount (reverse @mounted) {
1699 eval { PVE
::Tools
::run_command
(['/bin/umount', '--lazy', $mount], errfunc
=> sub{})};
1700 warn "Can't umount $mount\n" if $@;
1703 # If this fails they're used as mount points in a concurrent operation
1704 # (which should not happen but there's also no real need to get rid of them).
1711 # Should not be called after unsharing the mount namespace!
1713 my ($mp, $vmid, $storage, $storage_cfg, $conf, $snapname) = @_;
1715 die "cannot copy volumes of type $mp->{type}\n" if $mp->{type
} ne 'volume';
1716 File
::Path
::make_path
("/var/lib/lxc/$vmid");
1717 my $dest = "/var/lib/lxc/$vmid/.copy-volume-1";
1718 my $src = "/var/lib/lxc/$vmid/.copy-volume-2";
1720 # get id's for unprivileged container
1721 my (undef, $rootuid, $rootgid) = parse_id_maps
($conf);
1723 # Allocate the disk before unsharing in order to make sure zfs subvolumes
1724 # are visible in this namespace, otherwise the host only sees the empty
1725 # (not-mounted) directory.
1728 # Make sure $mp contains a correct size.
1729 $mp->{size
} = PVE
::Storage
::volume_size_info
($storage_cfg, $mp->{volume
});
1731 ($new_volid, $needs_chown) = alloc_disk
($storage_cfg, $vmid, $storage, $mp->{size
}/1024, $rootuid, $rootgid);
1733 PVE
::Storage
::activate_volumes
($storage_cfg, [$new_volid], undef);
1734 my $path = PVE
::Storage
::path
($storage_cfg, $new_volid, undef);
1735 chown($rootuid, $rootgid, $path);
1739 $copy_volume->($mp->{volume
}, $src, $new_volid, $dest, $storage_cfg, $snapname);
1743 PVE
::Storage
::vdisk_free
($storage_cfg, $new_volid)
1744 if defined($new_volid);