13 use Fcntl
qw(O_RDONLY);
15 use PVE
::Exception
qw(raise_perm_exc);
19 use PVE
::Tools
qw($IPV6RE $IPV4RE dir_glob_foreach lock_file lock_file_full);
21 use PVE
::AccessControl
;
24 use Time
::HiRes qw
(gettimeofday
);
28 my $nodename = PVE
::INotify
::nodename
();
30 my $cpuinfo= PVE
::ProcFSTools
::read_cpuinfo
();
32 our $COMMON_TAR_FLAGS = [ '--sparse', '--numeric-owner', '--acls',
34 '--xattrs-include=user.*',
35 '--xattrs-include=security.capability',
36 '--warning=no-xattr-write' ];
39 my $vmlist = PVE
::Cluster
::get_vmlist
();
41 return $res if !$vmlist || !$vmlist->{ids
};
42 my $ids = $vmlist->{ids
};
44 foreach my $vmid (keys %$ids) {
45 next if !$vmid; # skip CT0
46 my $d = $ids->{$vmid};
47 next if !$d->{node
} || $d->{node
} ne $nodename;
48 next if !$d->{type
} || $d->{type
} ne 'lxc';
49 $res->{$vmid}->{type
} = 'lxc';
57 unlink PVE
::LXC
::Config-
>config_file($vmid, $nodename);
60 # container status helpers
62 sub list_active_containers
{
64 my $filename = "/proc/net/unix";
66 # similar test is used by lcxcontainers.c: list_active_containers
69 my $fh = IO
::File-
>new ($filename, "r");
72 while (defined(my $line = <$fh>)) {
73 if ($line =~ m/^[a-f0-9]+:\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\d+\s+(\S+)$/) {
75 if ($path =~ m!^@/var/lib/lxc/(\d+)/command$!) {
86 # warning: this is slow
90 my $active_hash = list_active_containers
();
92 return 1 if defined($active_hash->{$vmid});
97 sub get_container_disk_usage
{
98 my ($vmid, $pid) = @_;
100 return PVE
::Tools
::df
("/proc/$pid/root/", 1);
103 my $last_proc_vmid_stat;
105 my $parse_cpuacct_stat = sub {
108 my $raw = read_cgroup_value
('cpuacct', $vmid, 'cpuacct.stat', 1);
112 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
125 my $list = $opt_vmid ?
{ $opt_vmid => { type
=> 'lxc' }} : config_list
();
127 my $active_hash = list_active_containers
();
129 my $cpucount = $cpuinfo->{cpus
} || 1;
131 my $cdtime = gettimeofday
;
133 my $uptime = (PVE
::ProcFSTools
::read_proc_uptime
(1))[0];
135 foreach my $vmid (keys %$list) {
136 my $d = $list->{$vmid};
138 eval { $d->{pid
} = find_lxc_pid
($vmid) if defined($active_hash->{$vmid}); };
139 warn $@ if $@; # ignore errors (consider them stopped)
141 $d->{status
} = $d->{pid
} ?
'running' : 'stopped';
143 my $cfspath = PVE
::LXC
::Config-
>cfs_config_path($vmid);
144 my $conf = PVE
::Cluster
::cfs_read_file
($cfspath) || {};
146 $d->{name
} = $conf->{'hostname'} || "CT$vmid";
147 $d->{name
} =~ s/[\s]//g;
149 $d->{cpus
} = $conf->{cpulimit
} || $cpucount;
151 $d->{lock} = $conf->{lock} || '';
154 my $res = get_container_disk_usage
($vmid, $d->{pid
});
155 $d->{disk
} = $res->{used
};
156 $d->{maxdisk
} = $res->{total
};
159 # use 4GB by default ??
160 if (my $rootfs = $conf->{rootfs
}) {
161 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($rootfs);
162 $d->{maxdisk
} = $rootinfo->{size
} || (4*1024*1024*1024);
164 $d->{maxdisk
} = 4*1024*1024*1024;
170 $d->{maxmem
} = ($conf->{memory
}||512)*1024*1024;
171 $d->{maxswap
} = ($conf->{swap
}//0)*1024*1024;
182 $d->{template
} = PVE
::LXC
::Config-
>is_template($conf);
185 foreach my $vmid (keys %$list) {
186 my $d = $list->{$vmid};
189 next if !$pid; # skip stopped CTs
191 my $ctime = (stat("/proc/$pid"))[10]; # 10 = ctime
192 $d->{uptime
} = time - $ctime; # the method lxcfs uses
194 $d->{mem
} = read_cgroup_value
('memory', $vmid, 'memory.usage_in_bytes');
195 $d->{swap
} = read_cgroup_value
('memory', $vmid, 'memory.memsw.usage_in_bytes') - $d->{mem
};
197 my $blkio_bytes = read_cgroup_value
('blkio', $vmid, 'blkio.throttle.io_service_bytes', 1);
198 my @bytes = split(/\n/, $blkio_bytes);
199 foreach my $byte (@bytes) {
200 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
201 $d->{diskread
} = $2 if $key eq 'Read';
202 $d->{diskwrite
} = $2 if $key eq 'Write';
206 my $pstat = &$parse_cpuacct_stat($vmid);
208 my $used = $pstat->{utime} + $pstat->{stime
};
210 my $old = $last_proc_vmid_stat->{$vmid};
212 $last_proc_vmid_stat->{$vmid} = {
220 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
223 my $dutime = $used - $old->{used
};
225 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
226 $last_proc_vmid_stat->{$vmid} = {
232 $d->{cpu
} = $old->{cpu
};
236 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
238 foreach my $dev (keys %$netdev) {
239 next if $dev !~ m/^veth([1-9]\d*)i/;
241 my $d = $list->{$vmid};
245 $d->{netout
} += $netdev->{$dev}->{receive
};
246 $d->{netin
} += $netdev->{$dev}->{transmit
};
253 sub read_cgroup_value
{
254 my ($group, $vmid, $name, $full) = @_;
256 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
258 return PVE
::Tools
::file_get_contents
($path) if $full;
260 return PVE
::Tools
::file_read_firstline
($path);
263 sub write_cgroup_value
{
264 my ($group, $vmid, $name, $value) = @_;
266 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
267 PVE
::ProcFSTools
::write_proc_entry
($path, $value) if -e
$path;
271 sub find_lxc_console_pids
{
275 PVE
::Tools
::dir_glob_foreach
('/proc', '\d+', sub {
278 my $cmdline = PVE
::Tools
::file_read_firstline
("/proc/$pid/cmdline");
281 my @args = split(/\0/, $cmdline);
283 # search for lxc-console -n <vmid>
284 return if scalar(@args) != 3;
285 return if $args[1] ne '-n';
286 return if $args[2] !~ m/^\d+$/;
287 return if $args[0] !~ m
|^(/usr/bin
/)?lxc-console
$|;
291 push @{$res->{$vmid}}, $pid;
303 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
305 PVE
::Tools
::run_command
(['lxc-info', '-n', $vmid, '-p'], outfunc
=> $parser);
307 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
312 # Note: we cannot use Net:IP, because that only allows strict
314 sub parse_ipv4_cidr
{
315 my ($cidr, $noerr) = @_;
317 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
318 return { address
=> $1, netmask
=> $PVE::Network
::ipv4_reverse_mask-
>[$2] };
321 return undef if $noerr;
323 die "unable to parse ipv4 address/mask\n";
327 sub update_lxc_config
{
328 my ($storage_cfg, $vmid, $conf) = @_;
330 my $dir = "/var/lib/lxc/$vmid";
332 if ($conf->{template
}) {
334 unlink "$dir/config";
341 die "missing 'arch' - internal error" if !$conf->{arch
};
342 $raw .= "lxc.arch = $conf->{arch}\n";
344 my $unprivileged = $conf->{unprivileged
};
345 my $custom_idmap = grep { $_->[0] eq 'lxc.id_map' } @{$conf->{lxc
}};
347 my $ostype = $conf->{ostype
} || die "missing 'ostype' - internal error";
348 if ($ostype =~ /^(?:debian | ubuntu | centos | fedora | opensuse | archlinux | alpine | gentoo | unmanaged)$/x) {
349 my $inc ="/usr/share/lxc/config/$ostype.common.conf";
350 $inc ="/usr/share/lxc/config/common.conf" if !-f
$inc;
351 $raw .= "lxc.include = $inc\n";
352 if ($unprivileged || $custom_idmap) {
353 $inc = "/usr/share/lxc/config/$ostype.userns.conf";
354 $inc = "/usr/share/lxc/config/userns.conf" if !-f
$inc;
355 $raw .= "lxc.include = $inc\n"
358 die "implement me (ostype $ostype)";
361 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
362 # cannot be exposed to the container with r/w access (cgroup perms).
363 # When this is enabled mounts will still remain in the monitor's namespace
364 # after the container unmounted them and thus will not detach from their
365 # files while the container is running!
366 $raw .= "lxc.monitor.unshare = 1\n";
368 # Should we read them from /etc/subuid?
369 if ($unprivileged && !$custom_idmap) {
370 $raw .= "lxc.id_map = u 0 100000 65536\n";
371 $raw .= "lxc.id_map = g 0 100000 65536\n";
374 if (!PVE
::LXC
::Config-
>has_dev_console($conf)) {
375 $raw .= "lxc.console = none\n";
376 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n";
379 my $ttycount = PVE
::LXC
::Config-
>get_tty_count($conf);
380 $raw .= "lxc.tty = $ttycount\n";
382 # some init scripts expect a linux terminal (turnkey).
383 $raw .= "lxc.environment = TERM=linux\n";
385 my $utsname = $conf->{hostname
} || "CT$vmid";
386 $raw .= "lxc.utsname = $utsname\n";
388 my $memory = $conf->{memory
} || 512;
389 my $swap = $conf->{swap
} // 0;
391 my $lxcmem = int($memory*1024*1024);
392 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
394 my $lxcswap = int(($memory + $swap)*1024*1024);
395 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
397 if (my $cpulimit = $conf->{cpulimit
}) {
398 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
399 my $value = int(100000*$cpulimit);
400 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
403 my $shares = $conf->{cpuunits
} || 1024;
404 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
406 die "missing 'rootfs' configuration\n"
407 if !defined($conf->{rootfs
});
409 my $mountpoint = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
411 $raw .= "lxc.rootfs = $dir/rootfs\n";
414 foreach my $k (keys %$conf) {
415 next if $k !~ m/^net(\d+)$/;
417 my $d = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$k});
419 $raw .= "lxc.network.type = veth\n";
420 $raw .= "lxc.network.veth.pair = veth${vmid}i${ind}\n";
421 $raw .= "lxc.network.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr
});
422 $raw .= "lxc.network.name = $d->{name}\n" if defined($d->{name
});
423 $raw .= "lxc.network.mtu = $d->{mtu}\n" if defined($d->{mtu
});
426 if (my $lxcconf = $conf->{lxc
}) {
427 foreach my $entry (@$lxcconf) {
428 my ($k, $v) = @$entry;
429 $netcount++ if $k eq 'lxc.network.type';
434 $raw .= "lxc.network.type = empty\n" if !$netcount;
436 File
::Path
::mkpath
("$dir/rootfs");
438 PVE
::Tools
::file_set_contents
("$dir/config", $raw);
441 # verify and cleanup nameserver list (replace \0 with ' ')
442 sub verify_nameserver_list
{
443 my ($nameserver_list) = @_;
446 foreach my $server (PVE
::Tools
::split_list
($nameserver_list)) {
447 PVE
::JSONSchema
::pve_verify_ip
($server);
451 return join(' ', @list);
454 sub verify_searchdomain_list
{
455 my ($searchdomain_list) = @_;
458 foreach my $server (PVE
::Tools
::split_list
($searchdomain_list)) {
459 # todo: should we add checks for valid dns domains?
463 return join(' ', @list);
466 sub get_console_command
{
467 my ($vmid, $conf) = @_;
469 my $cmode = PVE
::LXC
::Config-
>get_cmode($conf);
471 if ($cmode eq 'console') {
472 return ['lxc-console', '-n', $vmid, '-t', 0];
473 } elsif ($cmode eq 'tty') {
474 return ['lxc-console', '-n', $vmid];
475 } elsif ($cmode eq 'shell') {
476 return ['lxc-attach', '--clear-env', '-n', $vmid];
478 die "internal error";
482 sub get_primary_ips
{
485 # return data from net0
487 return undef if !defined($conf->{net0
});
488 my $net = PVE
::LXC
::Config-
>parse_lxc_network($conf->{net0
});
490 my $ipv4 = $net->{ip
};
492 if ($ipv4 =~ /^(dhcp|manual)$/) {
498 my $ipv6 = $net->{ip6
};
500 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
507 return ($ipv4, $ipv6);
510 sub delete_mountpoint_volume
{
511 my ($storage_cfg, $vmid, $volume) = @_;
513 return if PVE
::LXC
::Config-
>classify_mountpoint($volume) ne 'volume';
515 my ($vtype, $name, $owner) = PVE
::Storage
::parse_volname
($storage_cfg, $volume);
516 PVE
::Storage
::vdisk_free
($storage_cfg, $volume) if $vmid == $owner;
519 sub destroy_lxc_container
{
520 my ($storage_cfg, $vmid, $conf) = @_;
522 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
523 my ($ms, $mountpoint) = @_;
524 delete_mountpoint_volume
($storage_cfg, $vmid, $mountpoint->{volume
});
527 rmdir "/var/lib/lxc/$vmid/rootfs";
528 unlink "/var/lib/lxc/$vmid/config";
529 rmdir "/var/lib/lxc/$vmid";
530 destroy_config
($vmid);
532 #my $cmd = ['lxc-destroy', '-n', $vmid ];
533 #PVE::Tools::run_command($cmd);
536 sub vm_stop_cleanup
{
537 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
542 my $vollist = PVE
::LXC
::Config-
>get_vm_volumes($conf);
543 PVE
::Storage
::deactivate_volumes
($storage_cfg, $vollist);
546 warn $@ if $@; # avoid errors - just warn
549 my $safe_num_ne = sub {
552 return 0 if !defined($a) && !defined($b);
553 return 1 if !defined($a);
554 return 1 if !defined($b);
559 my $safe_string_ne = sub {
562 return 0 if !defined($a) && !defined($b);
563 return 1 if !defined($a);
564 return 1 if !defined($b);
570 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
572 if ($newnet->{type
} ne 'veth') {
573 # for when there are physical interfaces
574 die "cannot update interface of type $newnet->{type}";
577 my $veth = "veth${vmid}i${netid}";
578 my $eth = $newnet->{name
};
580 if (my $oldnetcfg = $conf->{$opt}) {
581 my $oldnet = PVE
::LXC
::Config-
>parse_lxc_network($oldnetcfg);
583 if (&$safe_string_ne($oldnet->{hwaddr
}, $newnet->{hwaddr
}) ||
584 &$safe_string_ne($oldnet->{name
}, $newnet->{name
})) {
586 PVE
::Network
::veth_delete
($veth);
587 delete $conf->{$opt};
588 PVE
::LXC
::Config-
>write_config($vmid, $conf);
590 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
593 if (&$safe_string_ne($oldnet->{bridge
}, $newnet->{bridge
}) ||
594 &$safe_num_ne($oldnet->{tag
}, $newnet->{tag
}) ||
595 &$safe_num_ne($oldnet->{firewall
}, $newnet->{firewall
})) {
597 if ($oldnet->{bridge
}) {
598 PVE
::Network
::tap_unplug
($veth);
599 foreach (qw(bridge tag firewall)) {
600 delete $oldnet->{$_};
602 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
603 PVE
::LXC
::Config-
>write_config($vmid, $conf);
606 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
607 # This includes the rate:
608 foreach (qw(bridge tag firewall rate)) {
609 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
611 } elsif (&$safe_string_ne($oldnet->{rate
}, $newnet->{rate
})) {
612 # Rate can be applied on its own but any change above needs to
613 # include the rate in tap_plug since OVS resets everything.
614 PVE
::Network
::tap_rate_limit
($veth, $newnet->{rate
});
615 $oldnet->{rate
} = $newnet->{rate
}
617 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
618 PVE
::LXC
::Config-
>write_config($vmid, $conf);
621 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
624 update_ipconfig
($vmid, $conf, $opt, $eth, $newnet, $rootdir);
628 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
630 my $veth = "veth${vmid}i${netid}";
631 my $vethpeer = $veth . "p";
632 my $eth = $newnet->{name
};
634 PVE
::Network
::veth_create
($veth, $vethpeer, $newnet->{bridge
}, $newnet->{hwaddr
});
635 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
637 # attach peer in container
638 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
639 PVE
::Tools
::run_command
($cmd);
641 # link up peer in container
642 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
643 PVE
::Tools
::run_command
($cmd);
645 my $done = { type
=> 'veth' };
646 foreach (qw(bridge tag firewall hwaddr name)) {
647 $done->{$_} = $newnet->{$_} if $newnet->{$_};
649 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($done);
651 PVE
::LXC
::Config-
>write_config($vmid, $conf);
654 sub update_ipconfig
{
655 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
657 my $lxc_setup = PVE
::LXC
::Setup-
>new($conf, $rootdir);
659 my $optdata = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$opt});
664 PVE
::Tools
::run_command
(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
666 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
668 my $change_ip_config = sub {
669 my ($ipversion) = @_;
671 my $family_opt = "-$ipversion";
672 my $suffix = $ipversion == 4 ?
'' : $ipversion;
676 my $newip = $newnet->{$ip};
677 my $newgw = $newnet->{$gw};
678 my $oldip = $optdata->{$ip};
680 my $change_ip = &$safe_string_ne($oldip, $newip);
681 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
683 return if !$change_ip && !$change_gw;
685 # step 1: add new IP, if this fails we cancel
686 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
687 if ($change_ip && $is_real_ip) {
688 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
695 # step 2: replace gateway
696 # If this fails we delete the added IP and cancel.
697 # If it succeeds we save the config and delete the old IP, ignoring
698 # errors. The config is then saved.
699 # Note: 'ip route replace' can add
703 if ($is_real_ip && !PVE
::Network
::is_ip_in_cidr
($newgw, $newip, $ipversion)) {
704 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
706 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
710 # the route was not replaced, the old IP is still available
711 # rollback (delete new IP) and cancel
713 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
714 warn $@ if $@; # no need to die here
719 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
720 # if the route was not deleted, the guest might have deleted it manually
726 # from this point on we save the configuration
727 # step 3: delete old IP ignoring errors
728 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
729 # We need to enable promote_secondaries, otherwise our newly added
730 # address will be removed along with the old one.
733 if ($ipversion == 4) {
734 &$nscmd({ outfunc
=> sub { $promote = int(shift) } },
735 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
736 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
738 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
740 warn $@ if $@; # no need to die here
742 if ($ipversion == 4) {
743 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
747 foreach my $property ($ip, $gw) {
748 if ($newnet->{$property}) {
749 $optdata->{$property} = $newnet->{$property};
751 delete $optdata->{$property};
754 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($optdata);
755 PVE
::LXC
::Config-
>write_config($vmid, $conf);
756 $lxc_setup->setup_network($conf);
759 &$change_ip_config(4);
760 &$change_ip_config(6);
764 my $enter_namespace = sub {
765 my ($vmid, $pid, $which, $type) = @_;
766 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
767 or die "failed to open $which namespace of container $vmid: $!\n";
768 PVE
::Tools
::setns
(fileno($fd), $type)
769 or die "failed to enter $which namespace of container $vmid: $!\n";
773 my $do_syncfs = sub {
774 my ($vmid, $pid, $socket) = @_;
776 &$enter_namespace($vmid, $pid, 'mnt', PVE
::Tools
::CLONE_NEWNS
);
778 # Tell the parent process to start reading our /proc/mounts
779 print {$socket} "go\n";
782 # Receive /proc/self/mounts
783 my $mountdata = do { local $/ = undef; <$socket> };
786 # Now sync all mountpoints...
787 my $mounts = PVE
::ProcFSTools
::parse_mounts
($mountdata);
788 foreach my $mp (@$mounts) {
789 my ($what, $dir, $fs) = @$mp;
790 next if $fs eq 'fuse.lxcfs';
791 eval { PVE
::Tools
::sync_mountpoint
($dir); };
796 sub sync_container_namespace
{
798 my $pid = find_lxc_pid
($vmid);
800 # SOCK_DGRAM is nicer for barriers but cannot be slurped
801 socketpair my $pfd, my $cfd, AF_UNIX
, SOCK_STREAM
, PF_UNSPEC
802 or die "failed to create socketpair: $!\n";
805 die "fork failed: $!\n" if !defined($child);
810 &$do_syncfs($vmid, $pid, $cfd);
820 die "failed to enter container namespace\n" if $go ne "go\n";
822 open my $mounts, '<', "/proc/$child/mounts"
823 or die "failed to open container's /proc/mounts: $!\n";
824 my $mountdata = do { local $/ = undef; <$mounts> };
826 print {$pfd} $mountdata;
829 while (waitpid($child, 0) != $child) {}
830 die "failed to sync container namespace\n" if $? != 0;
833 sub template_create
{
834 my ($vmid, $conf) = @_;
836 my $storecfg = PVE
::Storage
::config
();
838 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
839 my $volid = $rootinfo->{volume
};
841 die "Template feature is not available for '$volid'\n"
842 if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
844 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
846 my $template_volid = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
847 $rootinfo->{volume
} = $template_volid;
848 $conf->{rootfs
} = PVE
::LXC
::Config-
>print_ct_mountpoint($rootinfo, 1);
850 PVE
::LXC
::Config-
>write_config($vmid, $conf);
853 sub check_ct_modify_config_perm
{
854 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
856 return 1 if $authuser eq 'root@pam';
859 my ($opt, $delete) = @_;
860 if ($opt eq 'cpus' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
861 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
862 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
863 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
865 my $data = $opt eq 'rootfs' ? PVE
::LXC
::Config-
>parse_ct_rootfs($newconf->{$opt})
866 : PVE
::LXC
::Config-
>parse_ct_mountpoint($newconf->{$opt});
867 raise_perm_exc
("mountpoint type $data->{type}") if $data->{type
} ne 'volume';
868 } elsif ($opt eq 'memory' || $opt eq 'swap') {
869 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
870 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
871 $opt eq 'searchdomain' || $opt eq 'hostname') {
872 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
874 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
878 foreach my $opt (keys %$newconf) {
881 foreach my $opt (@$delete) {
889 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
891 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
892 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
894 PVE
::LXC
::Config-
>foreach_mountpoint_reverse($conf, sub {
895 my ($ms, $mountpoint) = @_;
897 my $volid = $mountpoint->{volume
};
898 my $mount = $mountpoint->{mp
};
900 return if !$volid || !$mount;
902 my $mount_path = "$rootdir/$mount";
903 $mount_path =~ s!/+!/!g;
905 return if !PVE
::ProcFSTools
::is_mounted
($mount_path);
908 PVE
::Tools
::run_command
(['umount', '-d', $mount_path]);
921 my ($vmid, $storage_cfg, $conf) = @_;
923 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
924 File
::Path
::make_path
($rootdir);
926 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
927 PVE
::Storage
::activate_volumes
($storage_cfg, $volid_list);
930 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
931 my ($ms, $mountpoint) = @_;
933 mountpoint_mount
($mountpoint, $rootdir, $storage_cfg);
937 warn "mounting container failed\n";
938 umount_all
($vmid, $storage_cfg, $conf, 1);
946 sub mountpoint_mount_path
{
947 my ($mountpoint, $storage_cfg, $snapname) = @_;
949 return mountpoint_mount
($mountpoint, undef, $storage_cfg, $snapname);
952 my $check_mount_path = sub {
954 $path = File
::Spec-
>canonpath($path);
955 my $real = Cwd
::realpath
($path);
956 if ($real ne $path) {
957 die "mount path modified by symlink: $path != $real";
966 if ($line =~ m
@^(/dev/loop\d
+):@) {
970 my $cmd = ['losetup', '--associated', $path];
971 PVE
::Tools
::run_command
($cmd, outfunc
=> $parser);
975 # Run a function with a file attached to a loop device.
976 # The loop device is always detached afterwards (or set to autoclear).
977 # Returns the loop device.
978 sub run_with_loopdev
{
979 my ($func, $file) = @_;
980 my $device = query_loopdev
($file);
981 # Try to reuse an existing device
983 # We assume that whoever setup the loop device is responsible for
991 if ($line =~ m
@^(/dev/loop\d
+)$@) {
995 PVE
::Tools
::run_command
(['losetup', '--show', '-f', $file], outfunc
=> $parser);
996 die "failed to setup loop device for $file\n" if !$device;
997 eval { &$func($device); };
999 PVE
::Tools
::run_command
(['losetup', '-d', $device]);
1005 my ($dir, $dest, $ro, @extra_opts) = @_;
1006 PVE
::Tools
::run_command
(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1008 eval { PVE
::Tools
::run_command
(['mount', '-o', 'bind,remount,ro', $dest]); };
1010 warn "bindmount error\n";
1011 # don't leave writable bind-mounts behind...
1012 PVE
::Tools
::run_command
(['umount', $dest]);
1018 # use $rootdir = undef to just return the corresponding mount path
1019 sub mountpoint_mount
{
1020 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1022 my $volid = $mountpoint->{volume
};
1023 my $mount = $mountpoint->{mp
};
1024 my $type = $mountpoint->{type
};
1025 my $quota = !$snapname && !$mountpoint->{ro
} && $mountpoint->{quota
};
1028 return if !$volid || !$mount;
1032 if (defined($rootdir)) {
1033 $rootdir =~ s!/+$!!;
1034 $mount_path = "$rootdir/$mount";
1035 $mount_path =~ s!/+!/!g;
1036 &$check_mount_path($mount_path);
1037 File
::Path
::mkpath
($mount_path);
1040 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1042 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1045 my $acl = $mountpoint->{acl
};
1046 if (defined($acl)) {
1047 $optstring .= ($acl ?
'acl' : 'noacl');
1049 my $readonly = $mountpoint->{ro
};
1051 my @extra_opts = ('-o', $optstring) if $optstring;
1055 my $scfg = PVE
::Storage
::storage_config
($storage_cfg, $storage);
1056 my $path = PVE
::Storage
::path
($storage_cfg, $volid, $snapname);
1058 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1059 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1061 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1063 if ($format eq 'subvol') {
1066 if ($scfg->{type
} eq 'zfspool') {
1067 my $path_arg = $path;
1068 $path_arg =~ s!^/+!!;
1069 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1071 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1074 if (defined($acl) && $scfg->{type
} eq 'zfspool') {
1075 my $acltype = ($acl ?
'acltype=posixacl' : 'acltype=noacl');
1076 my (undef, $name) = PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1077 $name .= "\@$snapname" if defined($snapname);
1078 PVE
::Tools
::run_command
(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1080 bindmount
($path, $mount_path, $readonly, @extra_opts);
1081 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1084 return wantarray ?
($path, 0, undef) : $path;
1085 } elsif ($format eq 'raw' || $format eq 'iso') {
1086 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1087 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1088 # Our autodev hook expects the /dev/dm-* device currently
1089 # and will create the /dev/mapper symlink accordingly
1090 ($path) = (Cwd
::realpath
($path) =~ /^(.*)$/s); # realpath() taints
1094 if ($format eq 'iso') {
1095 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1096 } elsif ($isBase || defined($snapname)) {
1097 PVE
::Tools
::run_command
(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1100 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1102 push @extra_opts, '-o', 'ro' if $readonly;
1103 PVE
::Tools
::run_command
(['mount', @extra_opts, $path, $mount_path]);
1107 my $use_loopdev = 0;
1108 if ($scfg->{path
}) {
1109 $mounted_dev = run_with_loopdev
($domount, $path);
1111 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' ||
1112 $scfg->{type
} eq 'rbd' || $scfg->{type
} eq 'lvmthin') {
1113 $mounted_dev = $path;
1116 die "unsupported storage type '$scfg->{type}'\n";
1118 return wantarray ?
($path, $use_loopdev, $mounted_dev) : $path;
1120 die "unsupported image format '$format'\n";
1122 } elsif ($type eq 'device') {
1123 push @extra_opts, '-o', 'ro' if $readonly;
1124 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1125 # See the NOTE above about devicemapper canonicalization
1126 my ($devpath) = (Cwd
::realpath
($volid) =~ /^(.*)$/s); # realpath() taints
1127 PVE
::Tools
::run_command
(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1128 return wantarray ?
($volid, 0, $devpath) : $volid;
1129 } elsif ($type eq 'bind') {
1130 die "directory '$volid' does not exist\n" if ! -d
$volid;
1131 &$check_mount_path($volid);
1132 bindmount
($volid, $mount_path, $readonly, @extra_opts) if $mount_path;
1133 warn "cannot enable quota control for bind mounts\n" if $quota;
1134 return wantarray ?
($volid, 0, undef) : $volid;
1137 die "unsupported storage";
1141 my ($dev, $rootuid, $rootgid) = @_;
1143 PVE
::Tools
::run_command
(['mkfs.ext4', '-O', 'mmp',
1144 '-E', "root_owner=$rootuid:$rootgid",
1149 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1151 if ($volid =~ m!^/dev/.+!) {
1156 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1158 die "cannot format volume '$volid' with no storage\n" if !$storage;
1160 PVE
::Storage
::activate_volumes
($storage_cfg, [$volid]);
1162 my $path = PVE
::Storage
::path
($storage_cfg, $volid);
1164 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1165 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1167 die "cannot format volume '$volid' (format == $format)\n"
1168 if $format ne 'raw';
1170 mkfs
($path, $rootuid, $rootgid);
1174 my ($storecfg, $vollist) = @_;
1176 foreach my $volid (@$vollist) {
1177 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1183 my ($storecfg, $vmid, $settings, $conf) = @_;
1188 my (undef, $rootuid, $rootgid) = PVE
::LXC
::parse_id_maps
($conf);
1189 my $chown_vollist = [];
1191 PVE
::LXC
::Config-
>foreach_mountpoint($settings, sub {
1192 my ($ms, $mountpoint) = @_;
1194 my $volid = $mountpoint->{volume
};
1195 my $mp = $mountpoint->{mp
};
1197 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1199 if ($storage && ($volid =~ m/^([^:\s]+):(\d+(\.\d+)?)$/)) {
1200 my ($storeid, $size_gb) = ($1, $2);
1202 my $size_kb = int(${size_gb
}*1024) * 1024;
1204 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storage);
1205 # fixme: use better naming ct-$vmid-disk-X.raw?
1207 if ($scfg->{type
} eq 'dir' || $scfg->{type
} eq 'nfs') {
1209 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw',
1211 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1213 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1215 push @$chown_vollist, $volid;
1217 } elsif ($scfg->{type
} eq 'zfspool') {
1219 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1221 push @$chown_vollist, $volid;
1222 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' || $scfg->{type
} eq 'lvmthin') {
1224 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1225 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1227 } elsif ($scfg->{type
} eq 'rbd') {
1229 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd
};
1230 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1231 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1233 die "unable to create containers on storage type '$scfg->{type}'\n";
1235 push @$vollist, $volid;
1236 $mountpoint->{volume
} = $volid;
1237 $mountpoint->{size
} = $size_kb * 1024;
1238 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1240 # use specified/existing volid/dir/device
1241 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1245 PVE
::Storage
::activate_volumes
($storecfg, $chown_vollist, undef);
1246 foreach my $volid (@$chown_vollist) {
1247 my $path = PVE
::Storage
::path
($storecfg, $volid, undef);
1248 chown($rootuid, $rootgid, $path);
1250 PVE
::Storage
::deactivate_volumes
($storecfg, $chown_vollist, undef);
1252 # free allocated images on error
1254 destroy_disks
($storecfg, $vollist);
1260 # bash completion helper
1262 sub complete_os_templates
{
1263 my ($cmdname, $pname, $cvalue) = @_;
1265 my $cfg = PVE
::Storage
::config
();
1269 if ($cvalue =~ m/^([^:]+):/) {
1273 my $vtype = $cmdname eq 'restore' ?
'backup' : 'vztmpl';
1274 my $data = PVE
::Storage
::template_list
($cfg, $storeid, $vtype);
1277 foreach my $id (keys %$data) {
1278 foreach my $item (@{$data->{$id}}) {
1279 push @$res, $item->{volid
} if defined($item->{volid
});
1286 my $complete_ctid_full = sub {
1289 my $idlist = vmstatus
();
1291 my $active_hash = list_active_containers
();
1295 foreach my $id (keys %$idlist) {
1296 my $d = $idlist->{$id};
1297 if (defined($running)) {
1298 next if $d->{template
};
1299 next if $running && !$active_hash->{$id};
1300 next if !$running && $active_hash->{$id};
1309 return &$complete_ctid_full();
1312 sub complete_ctid_stopped
{
1313 return &$complete_ctid_full(0);
1316 sub complete_ctid_running
{
1317 return &$complete_ctid_full(1);
1327 my $lxc = $conf->{lxc
};
1328 foreach my $entry (@$lxc) {
1329 my ($key, $value) = @$entry;
1330 next if $key ne 'lxc.id_map';
1331 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1332 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1333 push @$id_map, [$type, $ct, $host, $length];
1335 $rootuid = $host if $type eq 'u';
1336 $rootgid = $host if $type eq 'g';
1339 die "failed to parse id_map: $value\n";
1343 if (!@$id_map && $conf->{unprivileged
}) {
1344 # Should we read them from /etc/subuid?
1345 $id_map = [ ['u', '0', '100000', '65536'],
1346 ['g', '0', '100000', '65536'] ];
1347 $rootuid = $rootgid = 100000;
1350 return ($id_map, $rootuid, $rootgid);
1353 sub userns_command
{
1356 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];