13 use Fcntl
qw(O_RDONLY O_NOFOLLOW O_DIRECTORY);
14 use Errno
qw(ELOOP EROFS);
16 use PVE
::Exception
qw(raise_perm_exc);
20 use PVE
::Tools
qw($IPV6RE $IPV4RE dir_glob_foreach lock_file lock_file_full O_PATH);
22 use PVE
::AccessControl
;
25 use Time
::HiRes qw
(gettimeofday
);
29 my $nodename = PVE
::INotify
::nodename
();
31 my $cpuinfo= PVE
::ProcFSTools
::read_cpuinfo
();
33 our $COMMON_TAR_FLAGS = [ '--sparse', '--numeric-owner', '--acls',
35 '--xattrs-include=user.*',
36 '--xattrs-include=security.capability',
37 '--warning=no-xattr-write' ];
40 my $vmlist = PVE
::Cluster
::get_vmlist
();
42 return $res if !$vmlist || !$vmlist->{ids
};
43 my $ids = $vmlist->{ids
};
45 foreach my $vmid (keys %$ids) {
46 next if !$vmid; # skip CT0
47 my $d = $ids->{$vmid};
48 next if !$d->{node
} || $d->{node
} ne $nodename;
49 next if !$d->{type
} || $d->{type
} ne 'lxc';
50 $res->{$vmid}->{type
} = 'lxc';
58 unlink PVE
::LXC
::Config-
>config_file($vmid, $nodename);
61 # container status helpers
63 sub list_active_containers
{
65 my $filename = "/proc/net/unix";
67 # similar test is used by lcxcontainers.c: list_active_containers
70 my $fh = IO
::File-
>new ($filename, "r");
73 while (defined(my $line = <$fh>)) {
74 if ($line =~ m/^[a-f0-9]+:\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\d+\s+(\S+)$/) {
76 if ($path =~ m!^@/var/lib/lxc/(\d+)/command$!) {
87 # warning: this is slow
91 my $active_hash = list_active_containers
();
93 return 1 if defined($active_hash->{$vmid});
98 sub get_container_disk_usage
{
99 my ($vmid, $pid) = @_;
101 return PVE
::Tools
::df
("/proc/$pid/root/", 1);
104 my $last_proc_vmid_stat;
106 my $parse_cpuacct_stat = sub {
109 my $raw = read_cgroup_value
('cpuacct', $vmid, 'cpuacct.stat', 1);
113 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
126 my $list = $opt_vmid ?
{ $opt_vmid => { type
=> 'lxc' }} : config_list
();
128 my $active_hash = list_active_containers
();
130 my $cpucount = $cpuinfo->{cpus
} || 1;
132 my $cdtime = gettimeofday
;
134 my $uptime = (PVE
::ProcFSTools
::read_proc_uptime
(1))[0];
136 foreach my $vmid (keys %$list) {
137 my $d = $list->{$vmid};
139 eval { $d->{pid
} = find_lxc_pid
($vmid) if defined($active_hash->{$vmid}); };
140 warn $@ if $@; # ignore errors (consider them stopped)
142 $d->{status
} = $d->{pid
} ?
'running' : 'stopped';
144 my $cfspath = PVE
::LXC
::Config-
>cfs_config_path($vmid);
145 my $conf = PVE
::Cluster
::cfs_read_file
($cfspath) || {};
147 $d->{name
} = $conf->{'hostname'} || "CT$vmid";
148 $d->{name
} =~ s/[\s]//g;
150 $d->{cpus
} = $conf->{cpulimit
} || $cpucount;
152 $d->{lock} = $conf->{lock} || '';
155 my $res = get_container_disk_usage
($vmid, $d->{pid
});
156 $d->{disk
} = $res->{used
};
157 $d->{maxdisk
} = $res->{total
};
160 # use 4GB by default ??
161 if (my $rootfs = $conf->{rootfs
}) {
162 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($rootfs);
163 $d->{maxdisk
} = $rootinfo->{size
} || (4*1024*1024*1024);
165 $d->{maxdisk
} = 4*1024*1024*1024;
171 $d->{maxmem
} = ($conf->{memory
}||512)*1024*1024;
172 $d->{maxswap
} = ($conf->{swap
}//0)*1024*1024;
183 $d->{template
} = PVE
::LXC
::Config-
>is_template($conf);
186 foreach my $vmid (keys %$list) {
187 my $d = $list->{$vmid};
190 next if !$pid; # skip stopped CTs
192 my $ctime = (stat("/proc/$pid"))[10]; # 10 = ctime
193 $d->{uptime
} = time - $ctime; # the method lxcfs uses
195 my $memory_stat = read_cgroup_list
('memory', $vmid, 'memory.stat');
196 my $mem_usage_in_bytes = read_cgroup_value
('memory', $vmid, 'memory.usage_in_bytes');
198 $d->{mem
} = $mem_usage_in_bytes - $memory_stat->{cache
};
199 $d->{swap
} = read_cgroup_value
('memory', $vmid, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
201 my $blkio_bytes = read_cgroup_value
('blkio', $vmid, 'blkio.throttle.io_service_bytes', 1);
202 my @bytes = split(/\n/, $blkio_bytes);
203 foreach my $byte (@bytes) {
204 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
205 $d->{diskread
} = $2 if $key eq 'Read';
206 $d->{diskwrite
} = $2 if $key eq 'Write';
210 my $pstat = &$parse_cpuacct_stat($vmid);
212 my $used = $pstat->{utime} + $pstat->{stime
};
214 my $old = $last_proc_vmid_stat->{$vmid};
216 $last_proc_vmid_stat->{$vmid} = {
224 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
227 my $dutime = $used - $old->{used
};
229 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
230 $last_proc_vmid_stat->{$vmid} = {
236 $d->{cpu
} = $old->{cpu
};
240 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
242 foreach my $dev (keys %$netdev) {
243 next if $dev !~ m/^veth([1-9]\d*)i/;
245 my $d = $list->{$vmid};
249 $d->{netout
} += $netdev->{$dev}->{receive
};
250 $d->{netin
} += $netdev->{$dev}->{transmit
};
257 sub read_cgroup_list
{
258 my ($group, $vmid, $name) = @_;
260 my $content = read_cgroup_value
($group, $vmid, $name, 1);
262 return { split(/\s+/, $content) };
265 sub read_cgroup_value
{
266 my ($group, $vmid, $name, $full) = @_;
268 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
270 return PVE
::Tools
::file_get_contents
($path) if $full;
272 return PVE
::Tools
::file_read_firstline
($path);
275 sub write_cgroup_value
{
276 my ($group, $vmid, $name, $value) = @_;
278 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
279 PVE
::ProcFSTools
::write_proc_entry
($path, $value) if -e
$path;
283 sub find_lxc_console_pids
{
287 PVE
::Tools
::dir_glob_foreach
('/proc', '\d+', sub {
290 my $cmdline = PVE
::Tools
::file_read_firstline
("/proc/$pid/cmdline");
293 my @args = split(/\0/, $cmdline);
295 # search for lxc-console -n <vmid>
296 return if scalar(@args) != 3;
297 return if $args[1] ne '-n';
298 return if $args[2] !~ m/^\d+$/;
299 return if $args[0] !~ m
|^(/usr/bin
/)?lxc-console
$|;
303 push @{$res->{$vmid}}, $pid;
315 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
317 PVE
::Tools
::run_command
(['lxc-info', '-n', $vmid, '-p'], outfunc
=> $parser);
319 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
324 # Note: we cannot use Net:IP, because that only allows strict
326 sub parse_ipv4_cidr
{
327 my ($cidr, $noerr) = @_;
329 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
330 return { address
=> $1, netmask
=> $PVE::Network
::ipv4_reverse_mask-
>[$2] };
333 return undef if $noerr;
335 die "unable to parse ipv4 address/mask\n";
339 sub update_lxc_config
{
340 my ($vmid, $conf) = @_;
342 my $dir = "/var/lib/lxc/$vmid";
344 if ($conf->{template
}) {
346 unlink "$dir/config";
353 die "missing 'arch' - internal error" if !$conf->{arch
};
354 $raw .= "lxc.arch = $conf->{arch}\n";
356 my $unprivileged = $conf->{unprivileged
};
357 my $custom_idmap = grep { $_->[0] eq 'lxc.id_map' } @{$conf->{lxc
}};
359 my $ostype = $conf->{ostype
} || die "missing 'ostype' - internal error";
360 if ($ostype =~ /^(?:debian | ubuntu | centos | fedora | opensuse | archlinux | alpine | gentoo | unmanaged)$/x) {
361 my $inc ="/usr/share/lxc/config/$ostype.common.conf";
362 $inc ="/usr/share/lxc/config/common.conf" if !-f
$inc;
363 $raw .= "lxc.include = $inc\n";
364 if ($unprivileged || $custom_idmap) {
365 $inc = "/usr/share/lxc/config/$ostype.userns.conf";
366 $inc = "/usr/share/lxc/config/userns.conf" if !-f
$inc;
367 $raw .= "lxc.include = $inc\n"
370 die "implement me (ostype $ostype)";
373 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
374 # cannot be exposed to the container with r/w access (cgroup perms).
375 # When this is enabled mounts will still remain in the monitor's namespace
376 # after the container unmounted them and thus will not detach from their
377 # files while the container is running!
378 $raw .= "lxc.monitor.unshare = 1\n";
380 # Should we read them from /etc/subuid?
381 if ($unprivileged && !$custom_idmap) {
382 $raw .= "lxc.id_map = u 0 100000 65536\n";
383 $raw .= "lxc.id_map = g 0 100000 65536\n";
386 if (!PVE
::LXC
::Config-
>has_dev_console($conf)) {
387 $raw .= "lxc.console = none\n";
388 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n";
391 my $ttycount = PVE
::LXC
::Config-
>get_tty_count($conf);
392 $raw .= "lxc.tty = $ttycount\n";
394 # some init scripts expect a linux terminal (turnkey).
395 $raw .= "lxc.environment = TERM=linux\n";
397 my $utsname = $conf->{hostname
} || "CT$vmid";
398 $raw .= "lxc.utsname = $utsname\n";
400 my $memory = $conf->{memory
} || 512;
401 my $swap = $conf->{swap
} // 0;
403 my $lxcmem = int($memory*1024*1024);
404 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
406 my $lxcswap = int(($memory + $swap)*1024*1024);
407 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
409 if (my $cpulimit = $conf->{cpulimit
}) {
410 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
411 my $value = int(100000*$cpulimit);
412 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
415 my $shares = $conf->{cpuunits
} || 1024;
416 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
418 die "missing 'rootfs' configuration\n"
419 if !defined($conf->{rootfs
});
421 my $mountpoint = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
423 $raw .= "lxc.rootfs = $dir/rootfs\n";
426 foreach my $k (sort keys %$conf) {
427 next if $k !~ m/^net(\d+)$/;
429 my $d = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$k});
431 $raw .= "lxc.network.type = veth\n";
432 $raw .= "lxc.network.veth.pair = veth${vmid}i${ind}\n";
433 $raw .= "lxc.network.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr
});
434 $raw .= "lxc.network.name = $d->{name}\n" if defined($d->{name
});
435 $raw .= "lxc.network.mtu = $d->{mtu}\n" if defined($d->{mtu
});
438 if (my $lxcconf = $conf->{lxc
}) {
439 foreach my $entry (@$lxcconf) {
440 my ($k, $v) = @$entry;
441 $netcount++ if $k eq 'lxc.network.type';
446 $raw .= "lxc.network.type = empty\n" if !$netcount;
448 File
::Path
::mkpath
("$dir/rootfs");
450 PVE
::Tools
::file_set_contents
("$dir/config", $raw);
453 # verify and cleanup nameserver list (replace \0 with ' ')
454 sub verify_nameserver_list
{
455 my ($nameserver_list) = @_;
458 foreach my $server (PVE
::Tools
::split_list
($nameserver_list)) {
459 PVE
::JSONSchema
::pve_verify_ip
($server);
463 return join(' ', @list);
466 sub verify_searchdomain_list
{
467 my ($searchdomain_list) = @_;
470 foreach my $server (PVE
::Tools
::split_list
($searchdomain_list)) {
471 # todo: should we add checks for valid dns domains?
475 return join(' ', @list);
478 sub get_console_command
{
479 my ($vmid, $conf) = @_;
481 my $cmode = PVE
::LXC
::Config-
>get_cmode($conf);
483 if ($cmode eq 'console') {
484 return ['lxc-console', '-n', $vmid, '-t', 0];
485 } elsif ($cmode eq 'tty') {
486 return ['lxc-console', '-n', $vmid];
487 } elsif ($cmode eq 'shell') {
488 return ['lxc-attach', '--clear-env', '-n', $vmid];
490 die "internal error";
494 sub get_primary_ips
{
497 # return data from net0
499 return undef if !defined($conf->{net0
});
500 my $net = PVE
::LXC
::Config-
>parse_lxc_network($conf->{net0
});
502 my $ipv4 = $net->{ip
};
504 if ($ipv4 =~ /^(dhcp|manual)$/) {
510 my $ipv6 = $net->{ip6
};
512 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
519 return ($ipv4, $ipv6);
522 sub delete_mountpoint_volume
{
523 my ($storage_cfg, $vmid, $volume) = @_;
525 return if PVE
::LXC
::Config-
>classify_mountpoint($volume) ne 'volume';
527 my ($vtype, $name, $owner) = PVE
::Storage
::parse_volname
($storage_cfg, $volume);
528 PVE
::Storage
::vdisk_free
($storage_cfg, $volume) if $vmid == $owner;
531 sub destroy_lxc_container
{
532 my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
534 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
535 my ($ms, $mountpoint) = @_;
536 delete_mountpoint_volume
($storage_cfg, $vmid, $mountpoint->{volume
});
539 rmdir "/var/lib/lxc/$vmid/rootfs";
540 unlink "/var/lib/lxc/$vmid/config";
541 rmdir "/var/lib/lxc/$vmid";
542 if (defined $replacement_conf) {
543 PVE
::LXC
::Config-
>write_config($vmid, $replacement_conf);
545 destroy_config
($vmid);
548 #my $cmd = ['lxc-destroy', '-n', $vmid ];
549 #PVE::Tools::run_command($cmd);
552 sub vm_stop_cleanup
{
553 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
558 my $vollist = PVE
::LXC
::Config-
>get_vm_volumes($conf);
559 PVE
::Storage
::deactivate_volumes
($storage_cfg, $vollist);
562 warn $@ if $@; # avoid errors - just warn
565 my $safe_num_ne = sub {
568 return 0 if !defined($a) && !defined($b);
569 return 1 if !defined($a);
570 return 1 if !defined($b);
575 my $safe_string_ne = sub {
578 return 0 if !defined($a) && !defined($b);
579 return 1 if !defined($a);
580 return 1 if !defined($b);
586 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
588 if ($newnet->{type
} ne 'veth') {
589 # for when there are physical interfaces
590 die "cannot update interface of type $newnet->{type}";
593 my $veth = "veth${vmid}i${netid}";
594 my $eth = $newnet->{name
};
596 if (my $oldnetcfg = $conf->{$opt}) {
597 my $oldnet = PVE
::LXC
::Config-
>parse_lxc_network($oldnetcfg);
599 if (&$safe_string_ne($oldnet->{hwaddr
}, $newnet->{hwaddr
}) ||
600 &$safe_string_ne($oldnet->{name
}, $newnet->{name
})) {
602 PVE
::Network
::veth_delete
($veth);
603 delete $conf->{$opt};
604 PVE
::LXC
::Config-
>write_config($vmid, $conf);
606 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
609 if (&$safe_string_ne($oldnet->{bridge
}, $newnet->{bridge
}) ||
610 &$safe_num_ne($oldnet->{tag
}, $newnet->{tag
}) ||
611 &$safe_num_ne($oldnet->{firewall
}, $newnet->{firewall
})) {
613 if ($oldnet->{bridge
}) {
614 PVE
::Network
::tap_unplug
($veth);
615 foreach (qw(bridge tag firewall)) {
616 delete $oldnet->{$_};
618 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
619 PVE
::LXC
::Config-
>write_config($vmid, $conf);
622 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
623 # This includes the rate:
624 foreach (qw(bridge tag firewall rate)) {
625 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
627 } elsif (&$safe_string_ne($oldnet->{rate
}, $newnet->{rate
})) {
628 # Rate can be applied on its own but any change above needs to
629 # include the rate in tap_plug since OVS resets everything.
630 PVE
::Network
::tap_rate_limit
($veth, $newnet->{rate
});
631 $oldnet->{rate
} = $newnet->{rate
}
633 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
634 PVE
::LXC
::Config-
>write_config($vmid, $conf);
637 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
640 update_ipconfig
($vmid, $conf, $opt, $eth, $newnet, $rootdir);
644 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
646 my $veth = "veth${vmid}i${netid}";
647 my $vethpeer = $veth . "p";
648 my $eth = $newnet->{name
};
650 PVE
::Network
::veth_create
($veth, $vethpeer, $newnet->{bridge
}, $newnet->{hwaddr
});
651 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
653 # attach peer in container
654 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
655 PVE
::Tools
::run_command
($cmd);
657 # link up peer in container
658 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
659 PVE
::Tools
::run_command
($cmd);
661 my $done = { type
=> 'veth' };
662 foreach (qw(bridge tag firewall hwaddr name)) {
663 $done->{$_} = $newnet->{$_} if $newnet->{$_};
665 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($done);
667 PVE
::LXC
::Config-
>write_config($vmid, $conf);
670 sub update_ipconfig
{
671 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
673 my $lxc_setup = PVE
::LXC
::Setup-
>new($conf, $rootdir);
675 my $optdata = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$opt});
680 PVE
::Tools
::run_command
(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
682 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
684 my $change_ip_config = sub {
685 my ($ipversion) = @_;
687 my $family_opt = "-$ipversion";
688 my $suffix = $ipversion == 4 ?
'' : $ipversion;
692 my $newip = $newnet->{$ip};
693 my $newgw = $newnet->{$gw};
694 my $oldip = $optdata->{$ip};
696 my $change_ip = &$safe_string_ne($oldip, $newip);
697 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
699 return if !$change_ip && !$change_gw;
701 # step 1: add new IP, if this fails we cancel
702 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
703 if ($change_ip && $is_real_ip) {
704 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
711 # step 2: replace gateway
712 # If this fails we delete the added IP and cancel.
713 # If it succeeds we save the config and delete the old IP, ignoring
714 # errors. The config is then saved.
715 # Note: 'ip route replace' can add
719 if ($is_real_ip && !PVE
::Network
::is_ip_in_cidr
($newgw, $newip, $ipversion)) {
720 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
722 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
726 # the route was not replaced, the old IP is still available
727 # rollback (delete new IP) and cancel
729 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
730 warn $@ if $@; # no need to die here
735 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
736 # if the route was not deleted, the guest might have deleted it manually
742 # from this point on we save the configuration
743 # step 3: delete old IP ignoring errors
744 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
745 # We need to enable promote_secondaries, otherwise our newly added
746 # address will be removed along with the old one.
749 if ($ipversion == 4) {
750 &$nscmd({ outfunc
=> sub { $promote = int(shift) } },
751 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
752 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
754 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
756 warn $@ if $@; # no need to die here
758 if ($ipversion == 4) {
759 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
763 foreach my $property ($ip, $gw) {
764 if ($newnet->{$property}) {
765 $optdata->{$property} = $newnet->{$property};
767 delete $optdata->{$property};
770 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($optdata);
771 PVE
::LXC
::Config-
>write_config($vmid, $conf);
772 $lxc_setup->setup_network($conf);
775 &$change_ip_config(4);
776 &$change_ip_config(6);
780 my $enter_namespace = sub {
781 my ($vmid, $pid, $which, $type) = @_;
782 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
783 or die "failed to open $which namespace of container $vmid: $!\n";
784 PVE
::Tools
::setns
(fileno($fd), $type)
785 or die "failed to enter $which namespace of container $vmid: $!\n";
789 my $do_syncfs = sub {
790 my ($vmid, $pid, $socket) = @_;
792 &$enter_namespace($vmid, $pid, 'mnt', PVE
::Tools
::CLONE_NEWNS
);
794 # Tell the parent process to start reading our /proc/mounts
795 print {$socket} "go\n";
798 # Receive /proc/self/mounts
799 my $mountdata = do { local $/ = undef; <$socket> };
802 # Now sync all mountpoints...
803 my $mounts = PVE
::ProcFSTools
::parse_mounts
($mountdata);
804 foreach my $mp (@$mounts) {
805 my ($what, $dir, $fs) = @$mp;
806 next if $fs eq 'fuse.lxcfs';
807 eval { PVE
::Tools
::sync_mountpoint
($dir); };
812 sub sync_container_namespace
{
814 my $pid = find_lxc_pid
($vmid);
816 # SOCK_DGRAM is nicer for barriers but cannot be slurped
817 socketpair my $pfd, my $cfd, AF_UNIX
, SOCK_STREAM
, PF_UNSPEC
818 or die "failed to create socketpair: $!\n";
821 die "fork failed: $!\n" if !defined($child);
826 &$do_syncfs($vmid, $pid, $cfd);
836 die "failed to enter container namespace\n" if $go ne "go\n";
838 open my $mounts, '<', "/proc/$child/mounts"
839 or die "failed to open container's /proc/mounts: $!\n";
840 my $mountdata = do { local $/ = undef; <$mounts> };
842 print {$pfd} $mountdata;
845 while (waitpid($child, 0) != $child) {}
846 die "failed to sync container namespace\n" if $? != 0;
849 sub template_create
{
850 my ($vmid, $conf) = @_;
852 my $storecfg = PVE
::Storage
::config
();
854 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
855 my $volid = $rootinfo->{volume
};
857 die "Template feature is not available for '$volid'\n"
858 if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
860 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
862 my $template_volid = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
863 $rootinfo->{volume
} = $template_volid;
864 $conf->{rootfs
} = PVE
::LXC
::Config-
>print_ct_mountpoint($rootinfo, 1);
866 PVE
::LXC
::Config-
>write_config($vmid, $conf);
869 sub check_ct_modify_config_perm
{
870 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
872 return 1 if $authuser eq 'root@pam';
875 my ($opt, $delete) = @_;
876 if ($opt eq 'cpus' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
877 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
878 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
879 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
881 my $data = $opt eq 'rootfs' ? PVE
::LXC
::Config-
>parse_ct_rootfs($newconf->{$opt})
882 : PVE
::LXC
::Config-
>parse_ct_mountpoint($newconf->{$opt});
883 raise_perm_exc
("mountpoint type $data->{type}") if $data->{type
} ne 'volume';
884 } elsif ($opt eq 'memory' || $opt eq 'swap') {
885 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
886 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
887 $opt eq 'searchdomain' || $opt eq 'hostname') {
888 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
890 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
894 foreach my $opt (keys %$newconf) {
897 foreach my $opt (@$delete) {
905 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
907 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
908 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
910 PVE
::LXC
::Config-
>foreach_mountpoint_reverse($conf, sub {
911 my ($ms, $mountpoint) = @_;
913 my $volid = $mountpoint->{volume
};
914 my $mount = $mountpoint->{mp
};
916 return if !$volid || !$mount;
918 my $mount_path = "$rootdir/$mount";
919 $mount_path =~ s!/+!/!g;
921 return if !PVE
::ProcFSTools
::is_mounted
($mount_path);
924 PVE
::Tools
::run_command
(['umount', '-d', $mount_path]);
937 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
939 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
940 File
::Path
::make_path
($rootdir);
942 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
943 PVE
::Storage
::activate_volumes
($storage_cfg, $volid_list);
946 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
947 my ($ms, $mountpoint) = @_;
949 $mountpoint->{ro
} = 0 if $ignore_ro;
951 mountpoint_mount
($mountpoint, $rootdir, $storage_cfg);
955 warn "mounting container failed\n";
956 umount_all
($vmid, $storage_cfg, $conf, 1);
964 sub mountpoint_mount_path
{
965 my ($mountpoint, $storage_cfg, $snapname) = @_;
967 return mountpoint_mount
($mountpoint, undef, $storage_cfg, $snapname);
975 if ($line =~ m
@^(/dev/loop\d
+):@) {
979 my $cmd = ['losetup', '--associated', $path];
980 PVE
::Tools
::run_command
($cmd, outfunc
=> $parser);
984 # Run a function with a file attached to a loop device.
985 # The loop device is always detached afterwards (or set to autoclear).
986 # Returns the loop device.
987 sub run_with_loopdev
{
988 my ($func, $file) = @_;
989 my $device = query_loopdev
($file);
990 # Try to reuse an existing device
992 # We assume that whoever setup the loop device is responsible for
1000 if ($line =~ m
@^(/dev/loop\d
+)$@) {
1004 PVE
::Tools
::run_command
(['losetup', '--show', '-f', $file], outfunc
=> $parser);
1005 die "failed to setup loop device for $file\n" if !$device;
1006 eval { &$func($device); };
1008 PVE
::Tools
::run_command
(['losetup', '-d', $device]);
1013 # In scalar mode: returns a file handle to the deepest directory node.
1014 # In list context: returns a list of:
1015 # * the deepest directory node
1016 # * the 2nd deepest directory (parent of the above)
1017 # * directory name of the last directory
1018 # So that the path $2/$3 should lead to $1 afterwards.
1019 sub walk_tree_nofollow
($$$) {
1020 my ($start, $subdir, $mkdir) = @_;
1022 # splitdir() returns '' for empty components including the leading /
1023 my @comps = grep { length($_)>0 } File
::Spec-
>splitdir($subdir);
1025 sysopen(my $fd, $start, O_PATH
| O_DIRECTORY
)
1026 or die "failed to open start directory $start: $!\n";
1029 my $last_component = undef;
1031 foreach my $component (@comps) {
1032 $dir .= "/$component";
1033 my $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1036 # failed, check for symlinks and try to create the path
1037 die "symlink encountered at: $dir\n" if $! == ELOOP
;
1038 die "cannot open directory $dir: $!\n" if !$mkdir;
1040 # We don't check for errors on mkdirat() here and just try to
1041 # openat() again, since at least one error (EEXIST) is an
1042 # expected possibility if multiple containers start
1043 # simultaneously. If someone else injects a symlink now then
1044 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1045 PVE
::Tools
::mkdirat
(fileno($fd), $component, 0755);
1047 $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1048 die "failed to create path: $dir: $!\n" if !$next;
1051 close $second if defined($last_component);
1052 $last_component = $component;
1057 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1058 close $second if defined($last_component);
1062 # To guard against symlink attack races against other currently running
1063 # containers with shared recursive bind mount hierarchies we prepare a
1064 # directory handle for the directory we're mounting over to verify the
1065 # mountpoint afterwards.
1066 sub __bindmount_prepare
{
1067 my ($hostroot, $dir) = @_;
1068 my $srcdh = walk_tree_nofollow
($hostroot, $dir, 0);
1072 # Assuming we mount to rootfs/a/b/c, verify with the directory handle to 'b'
1073 # ($parentfd) that 'b/c' (openat($parentfd, 'c')) really leads to the directory
1074 # we intended to bind mount.
1075 sub __bindmount_verify
{
1076 my ($srcdh, $parentfd, $last_dir, $ro) = @_;
1079 # Open the mount point path coming from the parent directory since the
1080 # filehandle we would have gotten as first result of walk_tree_nofollow
1081 # earlier is still a handle to the underlying directory instead of the
1083 $destdh = PVE
::Tools
::openat
(fileno($parentfd), $last_dir, PVE
::Tools
::O_PATH
| O_NOFOLLOW
| O_DIRECTORY
);
1084 die "failed to open mount point: $!\n" if !$destdh;
1088 # no separate function because 99% of the time it's the wrong thing to use.
1089 if (syscall(269, fileno($destdh), $dot, &POSIX
::W_OK
, 0) != -1) {
1090 die "failed to mark bind mount read only\n";
1092 die "read-only check failed: $!\n" if $! != EROFS
;
1095 # For the rootfs we don't have a parentfd so we open the path directly.
1096 # Note that this means bindmounting any prefix of the host's
1097 # /var/lib/lxc/$vmid path into another container is considered a grave
1099 sysopen $destdh, $last_dir, O_PATH
| O_DIRECTORY
;
1100 die "failed to open mount point: $!\n" if !$destdh;
1103 my ($srcdev, $srcinode) = stat($srcdh);
1104 my ($dstdev, $dstinode) = stat($destdh);
1108 return ($srcdev == $dstdev && $srcinode == $dstinode);
1111 # Perform the actual bind mounting:
1112 sub __bindmount_do
{
1113 my ($dir, $dest, $ro, @extra_opts) = @_;
1114 PVE
::Tools
::run_command
(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1116 eval { PVE
::Tools
::run_command
(['mount', '-o', 'bind,remount,ro', $dest]); };
1118 warn "bindmount error\n";
1119 # don't leave writable bind-mounts behind...
1120 PVE
::Tools
::run_command
(['umount', $dest]);
1127 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1129 my $srcdh = __bindmount_prepare
('/', $dir);
1131 __bindmount_do
($dir, $dest, $ro, @extra_opts);
1133 if (!__bindmount_verify
($srcdh, $parentfd, $last_dir, $ro)) {
1134 PVE
::Tools
::run_command
(['umount', $dest]);
1135 die "detected mount path change at: $dir\n";
1139 # Cleanup $rootdir a bit (double and trailing slashes), build the mount path
1140 # from $rootdir and $mount and walk the path from $rootdir to the final
1141 # directory to check for symlinks.
1142 sub __mount_prepare_rootdir
{
1143 my ($rootdir, $mount) = @_;
1144 $rootdir =~ s!/+!/!g;
1145 $rootdir =~ s!/+$!!;
1146 my $mount_path = "$rootdir/$mount";
1147 my ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow
($rootdir, $mount, 1);
1148 return ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir);
1151 # use $rootdir = undef to just return the corresponding mount path
1152 sub mountpoint_mount
{
1153 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1155 my $volid = $mountpoint->{volume
};
1156 my $mount = $mountpoint->{mp
};
1157 my $type = $mountpoint->{type
};
1158 my $quota = !$snapname && !$mountpoint->{ro
} && $mountpoint->{quota
};
1161 return if !$volid || !$mount;
1166 my ($mpfd, $parentfd, $last_dir);
1168 if (defined($rootdir)) {
1169 ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir) =
1170 __mount_prepare_rootdir
($rootdir, $mount);
1173 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1175 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1178 my $acl = $mountpoint->{acl
};
1179 if (defined($acl)) {
1180 $optstring .= ($acl ?
'acl' : 'noacl');
1182 my $readonly = $mountpoint->{ro
};
1184 my @extra_opts = ('-o', $optstring) if $optstring;
1188 my $scfg = PVE
::Storage
::storage_config
($storage_cfg, $storage);
1190 # early sanity checks:
1191 # we otherwise call realpath on the rbd url
1192 die "containers on rbd storage without krbd are not supported\n"
1193 if $scfg->{type
} eq 'rbd' && !$scfg->{krbd
};
1195 my $path = PVE
::Storage
::path
($storage_cfg, $volid, $snapname);
1197 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1198 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1200 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1202 if ($format eq 'subvol') {
1205 if ($scfg->{type
} eq 'zfspool') {
1206 my $path_arg = $path;
1207 $path_arg =~ s!^/+!!;
1208 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1210 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1213 if (defined($acl) && $scfg->{type
} eq 'zfspool') {
1214 my $acltype = ($acl ?
'acltype=posixacl' : 'acltype=noacl');
1215 my (undef, $name) = PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1216 $name .= "\@$snapname" if defined($snapname);
1217 PVE
::Tools
::run_command
(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1219 bindmount
($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1220 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1223 return wantarray ?
($path, 0, undef) : $path;
1224 } elsif ($format eq 'raw' || $format eq 'iso') {
1225 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1226 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1227 # Our autodev hook expects the /dev/dm-* device currently
1228 # and will create the /dev/mapper symlink accordingly
1229 $path = Cwd
::realpath
($path);
1230 die "failed to get device path\n" if !$path;
1231 ($path) = ($path =~ /^(.*)$/s); #untaint
1235 if ($format eq 'iso') {
1236 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1237 } elsif ($isBase || defined($snapname)) {
1238 PVE
::Tools
::run_command
(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1241 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1243 push @extra_opts, '-o', 'ro' if $readonly;
1244 PVE
::Tools
::run_command
(['mount', @extra_opts, $path, $mount_path]);
1248 my $use_loopdev = 0;
1249 if ($scfg->{path
}) {
1250 $mounted_dev = run_with_loopdev
($domount, $path);
1252 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' ||
1253 $scfg->{type
} eq 'rbd' || $scfg->{type
} eq 'lvmthin') {
1254 $mounted_dev = $path;
1257 die "unsupported storage type '$scfg->{type}'\n";
1259 return wantarray ?
($path, $use_loopdev, $mounted_dev) : $path;
1261 die "unsupported image format '$format'\n";
1263 } elsif ($type eq 'device') {
1264 push @extra_opts, '-o', 'ro' if $readonly;
1265 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1266 # See the NOTE above about devicemapper canonicalization
1267 my ($devpath) = (Cwd
::realpath
($volid) =~ /^(.*)$/s); # realpath() taints
1268 PVE
::Tools
::run_command
(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1269 return wantarray ?
($volid, 0, $devpath) : $volid;
1270 } elsif ($type eq 'bind') {
1271 die "directory '$volid' does not exist\n" if ! -d
$volid;
1272 bindmount
($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1273 warn "cannot enable quota control for bind mounts\n" if $quota;
1274 return wantarray ?
($volid, 0, undef) : $volid;
1277 die "unsupported storage";
1281 my ($dev, $rootuid, $rootgid) = @_;
1283 PVE
::Tools
::run_command
(['mkfs.ext4', '-O', 'mmp',
1284 '-E', "root_owner=$rootuid:$rootgid",
1289 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1291 if ($volid =~ m!^/dev/.+!) {
1296 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1298 die "cannot format volume '$volid' with no storage\n" if !$storage;
1300 PVE
::Storage
::activate_volumes
($storage_cfg, [$volid]);
1302 my $path = PVE
::Storage
::path
($storage_cfg, $volid);
1304 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1305 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1307 die "cannot format volume '$volid' (format == $format)\n"
1308 if $format ne 'raw';
1310 mkfs
($path, $rootuid, $rootgid);
1314 my ($storecfg, $vollist) = @_;
1316 foreach my $volid (@$vollist) {
1317 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1323 my ($storecfg, $vmid, $settings, $conf) = @_;
1328 my (undef, $rootuid, $rootgid) = PVE
::LXC
::parse_id_maps
($conf);
1329 my $chown_vollist = [];
1331 PVE
::LXC
::Config-
>foreach_mountpoint($settings, sub {
1332 my ($ms, $mountpoint) = @_;
1334 my $volid = $mountpoint->{volume
};
1335 my $mp = $mountpoint->{mp
};
1337 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1339 if ($storage && ($volid =~ m/^([^:\s]+):(\d+(\.\d+)?)$/)) {
1340 my ($storeid, $size_gb) = ($1, $2);
1342 my $size_kb = int(${size_gb
}*1024) * 1024;
1344 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storage);
1345 # fixme: use better naming ct-$vmid-disk-X.raw?
1347 if ($scfg->{type
} eq 'dir' || $scfg->{type
} eq 'nfs') {
1349 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw',
1351 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1353 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1355 push @$chown_vollist, $volid;
1357 } elsif ($scfg->{type
} eq 'zfspool') {
1359 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1361 push @$chown_vollist, $volid;
1362 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' || $scfg->{type
} eq 'lvmthin') {
1364 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1365 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1367 } elsif ($scfg->{type
} eq 'rbd') {
1369 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd
};
1370 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1371 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1373 die "unable to create containers on storage type '$scfg->{type}'\n";
1375 push @$vollist, $volid;
1376 $mountpoint->{volume
} = $volid;
1377 $mountpoint->{size
} = $size_kb * 1024;
1378 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1380 # use specified/existing volid/dir/device
1381 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1385 PVE
::Storage
::activate_volumes
($storecfg, $chown_vollist, undef);
1386 foreach my $volid (@$chown_vollist) {
1387 my $path = PVE
::Storage
::path
($storecfg, $volid, undef);
1388 chown($rootuid, $rootgid, $path);
1390 PVE
::Storage
::deactivate_volumes
($storecfg, $chown_vollist, undef);
1392 # free allocated images on error
1394 destroy_disks
($storecfg, $vollist);
1400 # bash completion helper
1402 sub complete_os_templates
{
1403 my ($cmdname, $pname, $cvalue) = @_;
1405 my $cfg = PVE
::Storage
::config
();
1409 if ($cvalue =~ m/^([^:]+):/) {
1413 my $vtype = $cmdname eq 'restore' ?
'backup' : 'vztmpl';
1414 my $data = PVE
::Storage
::template_list
($cfg, $storeid, $vtype);
1417 foreach my $id (keys %$data) {
1418 foreach my $item (@{$data->{$id}}) {
1419 push @$res, $item->{volid
} if defined($item->{volid
});
1426 my $complete_ctid_full = sub {
1429 my $idlist = vmstatus
();
1431 my $active_hash = list_active_containers
();
1435 foreach my $id (keys %$idlist) {
1436 my $d = $idlist->{$id};
1437 if (defined($running)) {
1438 next if $d->{template
};
1439 next if $running && !$active_hash->{$id};
1440 next if !$running && $active_hash->{$id};
1449 return &$complete_ctid_full();
1452 sub complete_ctid_stopped
{
1453 return &$complete_ctid_full(0);
1456 sub complete_ctid_running
{
1457 return &$complete_ctid_full(1);
1467 my $lxc = $conf->{lxc
};
1468 foreach my $entry (@$lxc) {
1469 my ($key, $value) = @$entry;
1470 next if $key ne 'lxc.id_map';
1471 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1472 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1473 push @$id_map, [$type, $ct, $host, $length];
1475 $rootuid = $host if $type eq 'u';
1476 $rootgid = $host if $type eq 'g';
1479 die "failed to parse id_map: $value\n";
1483 if (!@$id_map && $conf->{unprivileged
}) {
1484 # Should we read them from /etc/subuid?
1485 $id_map = [ ['u', '0', '100000', '65536'],
1486 ['g', '0', '100000', '65536'] ];
1487 $rootuid = $rootgid = 100000;
1490 return ($id_map, $rootuid, $rootgid);
1493 sub userns_command
{
1496 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];