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->{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 $ctime = (stat("/proc/$pid"))[10]; # 10 = ctime
194 $d->{uptime
} = time - $ctime; # the method lxcfs uses
196 my $memory_stat = read_cgroup_list
('memory', $vmid, 'memory.stat');
197 my $mem_usage_in_bytes = read_cgroup_value
('memory', $vmid, 'memory.usage_in_bytes');
199 $d->{mem
} = $mem_usage_in_bytes - $memory_stat->{cache
};
200 $d->{swap
} = read_cgroup_value
('memory', $vmid, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
202 my $blkio_bytes = read_cgroup_value
('blkio', $vmid, 'blkio.throttle.io_service_bytes', 1);
203 my @bytes = split(/\n/, $blkio_bytes);
204 foreach my $byte (@bytes) {
205 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
206 $d->{diskread
} = $2 if $key eq 'Read';
207 $d->{diskwrite
} = $2 if $key eq 'Write';
211 my $pstat = &$parse_cpuacct_stat($vmid);
213 my $used = $pstat->{utime} + $pstat->{stime
};
215 my $old = $last_proc_vmid_stat->{$vmid};
217 $last_proc_vmid_stat->{$vmid} = {
225 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
228 my $dutime = $used - $old->{used
};
230 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
231 $last_proc_vmid_stat->{$vmid} = {
237 $d->{cpu
} = $old->{cpu
};
241 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
243 foreach my $dev (keys %$netdev) {
244 next if $dev !~ m/^veth([1-9]\d*)i/;
246 my $d = $list->{$vmid};
250 $d->{netout
} += $netdev->{$dev}->{receive
};
251 $d->{netin
} += $netdev->{$dev}->{transmit
};
258 sub read_cgroup_list
{
259 my ($group, $vmid, $name) = @_;
261 my $content = read_cgroup_value
($group, $vmid, $name, 1);
263 return { split(/\s+/, $content) };
266 sub read_cgroup_value
{
267 my ($group, $vmid, $name, $full) = @_;
269 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
271 return PVE
::Tools
::file_get_contents
($path) if $full;
273 return PVE
::Tools
::file_read_firstline
($path);
276 sub write_cgroup_value
{
277 my ($group, $vmid, $name, $value) = @_;
279 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
280 PVE
::ProcFSTools
::write_proc_entry
($path, $value) if -e
$path;
284 sub find_lxc_console_pids
{
288 PVE
::Tools
::dir_glob_foreach
('/proc', '\d+', sub {
291 my $cmdline = PVE
::Tools
::file_read_firstline
("/proc/$pid/cmdline");
294 my @args = split(/\0/, $cmdline);
296 # search for lxc-console -n <vmid>
297 return if scalar(@args) != 3;
298 return if $args[1] ne '-n';
299 return if $args[2] !~ m/^\d+$/;
300 return if $args[0] !~ m
|^(/usr/bin
/)?lxc-console
$|;
304 push @{$res->{$vmid}}, $pid;
316 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
318 PVE
::Tools
::run_command
(['lxc-info', '-n', $vmid, '-p'], outfunc
=> $parser);
320 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
325 # Note: we cannot use Net:IP, because that only allows strict
327 sub parse_ipv4_cidr
{
328 my ($cidr, $noerr) = @_;
330 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
331 return { address
=> $1, netmask
=> $PVE::Network
::ipv4_reverse_mask-
>[$2] };
334 return undef if $noerr;
336 die "unable to parse ipv4 address/mask\n";
340 sub update_lxc_config
{
341 my ($vmid, $conf) = @_;
343 my $dir = "/var/lib/lxc/$vmid";
345 if ($conf->{template
}) {
347 unlink "$dir/config";
354 die "missing 'arch' - internal error" if !$conf->{arch
};
355 $raw .= "lxc.arch = $conf->{arch}\n";
357 my $unprivileged = $conf->{unprivileged
};
358 my $custom_idmap = grep { $_->[0] eq 'lxc.id_map' } @{$conf->{lxc
}};
360 my $ostype = $conf->{ostype
} || die "missing 'ostype' - internal error";
361 if ($ostype =~ /^(?:debian | ubuntu | centos | fedora | opensuse | archlinux | alpine | gentoo | unmanaged)$/x) {
362 my $inc ="/usr/share/lxc/config/$ostype.common.conf";
363 $inc ="/usr/share/lxc/config/common.conf" if !-f
$inc;
364 $raw .= "lxc.include = $inc\n";
365 if ($unprivileged || $custom_idmap) {
366 $inc = "/usr/share/lxc/config/$ostype.userns.conf";
367 $inc = "/usr/share/lxc/config/userns.conf" if !-f
$inc;
368 $raw .= "lxc.include = $inc\n"
371 die "implement me (ostype $ostype)";
374 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
375 # cannot be exposed to the container with r/w access (cgroup perms).
376 # When this is enabled mounts will still remain in the monitor's namespace
377 # after the container unmounted them and thus will not detach from their
378 # files while the container is running!
379 $raw .= "lxc.monitor.unshare = 1\n";
381 # Should we read them from /etc/subuid?
382 if ($unprivileged && !$custom_idmap) {
383 $raw .= "lxc.id_map = u 0 100000 65536\n";
384 $raw .= "lxc.id_map = g 0 100000 65536\n";
387 if (!PVE
::LXC
::Config-
>has_dev_console($conf)) {
388 $raw .= "lxc.console = none\n";
389 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n";
392 my $ttycount = PVE
::LXC
::Config-
>get_tty_count($conf);
393 $raw .= "lxc.tty = $ttycount\n";
395 # some init scripts expect a linux terminal (turnkey).
396 $raw .= "lxc.environment = TERM=linux\n";
398 my $utsname = $conf->{hostname
} || "CT$vmid";
399 $raw .= "lxc.utsname = $utsname\n";
401 my $memory = $conf->{memory
} || 512;
402 my $swap = $conf->{swap
} // 0;
404 my $lxcmem = int($memory*1024*1024);
405 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
407 my $lxcswap = int(($memory + $swap)*1024*1024);
408 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
410 if (my $cpulimit = $conf->{cpulimit
}) {
411 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
412 my $value = int(100000*$cpulimit);
413 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
416 my $shares = $conf->{cpuunits
} || 1024;
417 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
419 die "missing 'rootfs' configuration\n"
420 if !defined($conf->{rootfs
});
422 my $mountpoint = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
424 $raw .= "lxc.rootfs = $dir/rootfs\n";
427 foreach my $k (sort keys %$conf) {
428 next if $k !~ m/^net(\d+)$/;
430 my $d = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$k});
432 $raw .= "lxc.network.type = veth\n";
433 $raw .= "lxc.network.veth.pair = veth${vmid}i${ind}\n";
434 $raw .= "lxc.network.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr
});
435 $raw .= "lxc.network.name = $d->{name}\n" if defined($d->{name
});
436 $raw .= "lxc.network.mtu = $d->{mtu}\n" if defined($d->{mtu
});
439 if (my $lxcconf = $conf->{lxc
}) {
440 foreach my $entry (@$lxcconf) {
441 my ($k, $v) = @$entry;
442 $netcount++ if $k eq 'lxc.network.type';
447 $raw .= "lxc.network.type = empty\n" if !$netcount;
449 File
::Path
::mkpath
("$dir/rootfs");
451 PVE
::Tools
::file_set_contents
("$dir/config", $raw);
454 # verify and cleanup nameserver list (replace \0 with ' ')
455 sub verify_nameserver_list
{
456 my ($nameserver_list) = @_;
459 foreach my $server (PVE
::Tools
::split_list
($nameserver_list)) {
460 PVE
::JSONSchema
::pve_verify_ip
($server);
464 return join(' ', @list);
467 sub verify_searchdomain_list
{
468 my ($searchdomain_list) = @_;
471 foreach my $server (PVE
::Tools
::split_list
($searchdomain_list)) {
472 # todo: should we add checks for valid dns domains?
476 return join(' ', @list);
479 sub get_console_command
{
480 my ($vmid, $conf) = @_;
482 my $cmode = PVE
::LXC
::Config-
>get_cmode($conf);
484 if ($cmode eq 'console') {
485 return ['lxc-console', '-n', $vmid, '-t', 0];
486 } elsif ($cmode eq 'tty') {
487 return ['lxc-console', '-n', $vmid];
488 } elsif ($cmode eq 'shell') {
489 return ['lxc-attach', '--clear-env', '-n', $vmid];
491 die "internal error";
495 sub get_primary_ips
{
498 # return data from net0
500 return undef if !defined($conf->{net0
});
501 my $net = PVE
::LXC
::Config-
>parse_lxc_network($conf->{net0
});
503 my $ipv4 = $net->{ip
};
505 if ($ipv4 =~ /^(dhcp|manual)$/) {
511 my $ipv6 = $net->{ip6
};
513 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
520 return ($ipv4, $ipv6);
523 sub delete_mountpoint_volume
{
524 my ($storage_cfg, $vmid, $volume) = @_;
526 return if PVE
::LXC
::Config-
>classify_mountpoint($volume) ne 'volume';
528 my ($vtype, $name, $owner) = PVE
::Storage
::parse_volname
($storage_cfg, $volume);
529 PVE
::Storage
::vdisk_free
($storage_cfg, $volume) if $vmid == $owner;
532 sub destroy_lxc_container
{
533 my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
535 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
536 my ($ms, $mountpoint) = @_;
537 delete_mountpoint_volume
($storage_cfg, $vmid, $mountpoint->{volume
});
540 rmdir "/var/lib/lxc/$vmid/rootfs";
541 unlink "/var/lib/lxc/$vmid/config";
542 rmdir "/var/lib/lxc/$vmid";
543 if (defined $replacement_conf) {
544 PVE
::LXC
::Config-
>write_config($vmid, $replacement_conf);
546 destroy_config
($vmid);
549 #my $cmd = ['lxc-destroy', '-n', $vmid ];
550 #PVE::Tools::run_command($cmd);
553 sub vm_stop_cleanup
{
554 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
559 my $vollist = PVE
::LXC
::Config-
>get_vm_volumes($conf);
560 PVE
::Storage
::deactivate_volumes
($storage_cfg, $vollist);
563 warn $@ if $@; # avoid errors - just warn
566 my $safe_num_ne = sub {
569 return 0 if !defined($a) && !defined($b);
570 return 1 if !defined($a);
571 return 1 if !defined($b);
576 my $safe_string_ne = sub {
579 return 0 if !defined($a) && !defined($b);
580 return 1 if !defined($a);
581 return 1 if !defined($b);
587 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
589 if ($newnet->{type
} ne 'veth') {
590 # for when there are physical interfaces
591 die "cannot update interface of type $newnet->{type}";
594 my $veth = "veth${vmid}i${netid}";
595 my $eth = $newnet->{name
};
597 if (my $oldnetcfg = $conf->{$opt}) {
598 my $oldnet = PVE
::LXC
::Config-
>parse_lxc_network($oldnetcfg);
600 if (&$safe_string_ne($oldnet->{hwaddr
}, $newnet->{hwaddr
}) ||
601 &$safe_string_ne($oldnet->{name
}, $newnet->{name
})) {
603 PVE
::Network
::veth_delete
($veth);
604 delete $conf->{$opt};
605 PVE
::LXC
::Config-
>write_config($vmid, $conf);
607 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
610 if (&$safe_string_ne($oldnet->{bridge
}, $newnet->{bridge
}) ||
611 &$safe_num_ne($oldnet->{tag
}, $newnet->{tag
}) ||
612 &$safe_num_ne($oldnet->{firewall
}, $newnet->{firewall
})) {
614 if ($oldnet->{bridge
}) {
615 PVE
::Network
::tap_unplug
($veth);
616 foreach (qw(bridge tag firewall)) {
617 delete $oldnet->{$_};
619 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
620 PVE
::LXC
::Config-
>write_config($vmid, $conf);
623 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
624 # This includes the rate:
625 foreach (qw(bridge tag firewall rate)) {
626 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
628 } elsif (&$safe_string_ne($oldnet->{rate
}, $newnet->{rate
})) {
629 # Rate can be applied on its own but any change above needs to
630 # include the rate in tap_plug since OVS resets everything.
631 PVE
::Network
::tap_rate_limit
($veth, $newnet->{rate
});
632 $oldnet->{rate
} = $newnet->{rate
}
634 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
635 PVE
::LXC
::Config-
>write_config($vmid, $conf);
638 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
641 update_ipconfig
($vmid, $conf, $opt, $eth, $newnet, $rootdir);
645 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
647 my $veth = "veth${vmid}i${netid}";
648 my $vethpeer = $veth . "p";
649 my $eth = $newnet->{name
};
651 PVE
::Network
::veth_create
($veth, $vethpeer, $newnet->{bridge
}, $newnet->{hwaddr
});
652 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
654 # attach peer in container
655 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
656 PVE
::Tools
::run_command
($cmd);
658 # link up peer in container
659 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
660 PVE
::Tools
::run_command
($cmd);
662 my $done = { type
=> 'veth' };
663 foreach (qw(bridge tag firewall hwaddr name)) {
664 $done->{$_} = $newnet->{$_} if $newnet->{$_};
666 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($done);
668 PVE
::LXC
::Config-
>write_config($vmid, $conf);
671 sub update_ipconfig
{
672 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
674 my $lxc_setup = PVE
::LXC
::Setup-
>new($conf, $rootdir);
676 my $optdata = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$opt});
681 PVE
::Tools
::run_command
(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
683 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
685 my $change_ip_config = sub {
686 my ($ipversion) = @_;
688 my $family_opt = "-$ipversion";
689 my $suffix = $ipversion == 4 ?
'' : $ipversion;
693 my $newip = $newnet->{$ip};
694 my $newgw = $newnet->{$gw};
695 my $oldip = $optdata->{$ip};
697 my $change_ip = &$safe_string_ne($oldip, $newip);
698 my $change_gw = &$safe_string_ne($optdata->{$gw}, $newgw);
700 return if !$change_ip && !$change_gw;
702 # step 1: add new IP, if this fails we cancel
703 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
704 if ($change_ip && $is_real_ip) {
705 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
712 # step 2: replace gateway
713 # If this fails we delete the added IP and cancel.
714 # If it succeeds we save the config and delete the old IP, ignoring
715 # errors. The config is then saved.
716 # Note: 'ip route replace' can add
720 if ($is_real_ip && !PVE
::Network
::is_ip_in_cidr
($newgw, $newip, $ipversion)) {
721 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
723 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
727 # the route was not replaced, the old IP is still available
728 # rollback (delete new IP) and cancel
730 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
731 warn $@ if $@; # no need to die here
736 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
737 # if the route was not deleted, the guest might have deleted it manually
743 # from this point on we save the configuration
744 # step 3: delete old IP ignoring errors
745 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
746 # We need to enable promote_secondaries, otherwise our newly added
747 # address will be removed along with the old one.
750 if ($ipversion == 4) {
751 &$nscmd({ outfunc
=> sub { $promote = int(shift) } },
752 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
753 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
755 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
757 warn $@ if $@; # no need to die here
759 if ($ipversion == 4) {
760 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
764 foreach my $property ($ip, $gw) {
765 if ($newnet->{$property}) {
766 $optdata->{$property} = $newnet->{$property};
768 delete $optdata->{$property};
771 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($optdata);
772 PVE
::LXC
::Config-
>write_config($vmid, $conf);
773 $lxc_setup->setup_network($conf);
776 &$change_ip_config(4);
777 &$change_ip_config(6);
781 my $enter_namespace = sub {
782 my ($vmid, $pid, $which, $type) = @_;
783 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
784 or die "failed to open $which namespace of container $vmid: $!\n";
785 PVE
::Tools
::setns
(fileno($fd), $type)
786 or die "failed to enter $which namespace of container $vmid: $!\n";
790 my $do_syncfs = sub {
791 my ($vmid, $pid, $socket) = @_;
793 &$enter_namespace($vmid, $pid, 'mnt', PVE
::Tools
::CLONE_NEWNS
);
795 # Tell the parent process to start reading our /proc/mounts
796 print {$socket} "go\n";
799 # Receive /proc/self/mounts
800 my $mountdata = do { local $/ = undef; <$socket> };
803 # Now sync all mountpoints...
804 my $mounts = PVE
::ProcFSTools
::parse_mounts
($mountdata);
805 foreach my $mp (@$mounts) {
806 my ($what, $dir, $fs) = @$mp;
807 next if $fs eq 'fuse.lxcfs';
808 eval { PVE
::Tools
::sync_mountpoint
($dir); };
813 sub sync_container_namespace
{
815 my $pid = find_lxc_pid
($vmid);
817 # SOCK_DGRAM is nicer for barriers but cannot be slurped
818 socketpair my $pfd, my $cfd, AF_UNIX
, SOCK_STREAM
, PF_UNSPEC
819 or die "failed to create socketpair: $!\n";
822 die "fork failed: $!\n" if !defined($child);
827 &$do_syncfs($vmid, $pid, $cfd);
837 die "failed to enter container namespace\n" if $go ne "go\n";
839 open my $mounts, '<', "/proc/$child/mounts"
840 or die "failed to open container's /proc/mounts: $!\n";
841 my $mountdata = do { local $/ = undef; <$mounts> };
843 print {$pfd} $mountdata;
846 while (waitpid($child, 0) != $child) {}
847 die "failed to sync container namespace\n" if $? != 0;
850 sub template_create
{
851 my ($vmid, $conf) = @_;
853 my $storecfg = PVE
::Storage
::config
();
855 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
856 my $volid = $rootinfo->{volume
};
858 die "Template feature is not available for '$volid'\n"
859 if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
861 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
863 my $template_volid = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
864 $rootinfo->{volume
} = $template_volid;
865 $conf->{rootfs
} = PVE
::LXC
::Config-
>print_ct_mountpoint($rootinfo, 1);
867 PVE
::LXC
::Config-
>write_config($vmid, $conf);
870 sub check_ct_modify_config_perm
{
871 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
873 return 1 if $authuser eq 'root@pam';
876 my ($opt, $delete) = @_;
877 if ($opt eq 'cores' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
878 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
879 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
880 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
882 my $data = $opt eq 'rootfs' ? PVE
::LXC
::Config-
>parse_ct_rootfs($newconf->{$opt})
883 : PVE
::LXC
::Config-
>parse_ct_mountpoint($newconf->{$opt});
884 raise_perm_exc
("mount point type $data->{type}") if $data->{type
} ne 'volume';
885 } elsif ($opt eq 'memory' || $opt eq 'swap') {
886 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
887 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
888 $opt eq 'searchdomain' || $opt eq 'hostname') {
889 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
891 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
895 foreach my $opt (keys %$newconf) {
898 foreach my $opt (@$delete) {
906 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
908 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
909 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
911 PVE
::LXC
::Config-
>foreach_mountpoint_reverse($conf, sub {
912 my ($ms, $mountpoint) = @_;
914 my $volid = $mountpoint->{volume
};
915 my $mount = $mountpoint->{mp
};
917 return if !$volid || !$mount;
919 my $mount_path = "$rootdir/$mount";
920 $mount_path =~ s!/+!/!g;
922 return if !PVE
::ProcFSTools
::is_mounted
($mount_path);
925 PVE
::Tools
::run_command
(['umount', '-d', $mount_path]);
938 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
940 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
941 File
::Path
::make_path
($rootdir);
943 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
944 PVE
::Storage
::activate_volumes
($storage_cfg, $volid_list);
947 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
948 my ($ms, $mountpoint) = @_;
950 $mountpoint->{ro
} = 0 if $ignore_ro;
952 mountpoint_mount
($mountpoint, $rootdir, $storage_cfg);
956 warn "mounting container failed\n";
957 umount_all
($vmid, $storage_cfg, $conf, 1);
965 sub mountpoint_mount_path
{
966 my ($mountpoint, $storage_cfg, $snapname) = @_;
968 return mountpoint_mount
($mountpoint, undef, $storage_cfg, $snapname);
976 if ($line =~ m
@^(/dev/loop\d
+):@) {
980 my $cmd = ['losetup', '--associated', $path];
981 PVE
::Tools
::run_command
($cmd, outfunc
=> $parser);
985 # Run a function with a file attached to a loop device.
986 # The loop device is always detached afterwards (or set to autoclear).
987 # Returns the loop device.
988 sub run_with_loopdev
{
989 my ($func, $file) = @_;
990 my $device = query_loopdev
($file);
991 # Try to reuse an existing device
993 # We assume that whoever setup the loop device is responsible for
1001 if ($line =~ m
@^(/dev/loop\d
+)$@) {
1005 PVE
::Tools
::run_command
(['losetup', '--show', '-f', $file], outfunc
=> $parser);
1006 die "failed to setup loop device for $file\n" if !$device;
1007 eval { &$func($device); };
1009 PVE
::Tools
::run_command
(['losetup', '-d', $device]);
1014 # In scalar mode: returns a file handle to the deepest directory node.
1015 # In list context: returns a list of:
1016 # * the deepest directory node
1017 # * the 2nd deepest directory (parent of the above)
1018 # * directory name of the last directory
1019 # So that the path $2/$3 should lead to $1 afterwards.
1020 sub walk_tree_nofollow
($$$) {
1021 my ($start, $subdir, $mkdir) = @_;
1023 # splitdir() returns '' for empty components including the leading /
1024 my @comps = grep { length($_)>0 } File
::Spec-
>splitdir($subdir);
1026 sysopen(my $fd, $start, O_PATH
| O_DIRECTORY
)
1027 or die "failed to open start directory $start: $!\n";
1030 my $last_component = undef;
1032 foreach my $component (@comps) {
1033 $dir .= "/$component";
1034 my $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1037 # failed, check for symlinks and try to create the path
1038 die "symlink encountered at: $dir\n" if $! == ELOOP
;
1039 die "cannot open directory $dir: $!\n" if !$mkdir;
1041 # We don't check for errors on mkdirat() here and just try to
1042 # openat() again, since at least one error (EEXIST) is an
1043 # expected possibility if multiple containers start
1044 # simultaneously. If someone else injects a symlink now then
1045 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1046 PVE
::Tools
::mkdirat
(fileno($fd), $component, 0755);
1048 $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1049 die "failed to create path: $dir: $!\n" if !$next;
1052 close $second if defined($last_component);
1053 $last_component = $component;
1058 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1059 close $second if defined($last_component);
1063 # To guard against symlink attack races against other currently running
1064 # containers with shared recursive bind mount hierarchies we prepare a
1065 # directory handle for the directory we're mounting over to verify the
1066 # mountpoint afterwards.
1067 sub __bindmount_prepare
{
1068 my ($hostroot, $dir) = @_;
1069 my $srcdh = walk_tree_nofollow
($hostroot, $dir, 0);
1073 # Assuming we mount to rootfs/a/b/c, verify with the directory handle to 'b'
1074 # ($parentfd) that 'b/c' (openat($parentfd, 'c')) really leads to the directory
1075 # we intended to bind mount.
1076 sub __bindmount_verify
{
1077 my ($srcdh, $parentfd, $last_dir, $ro) = @_;
1080 # Open the mount point path coming from the parent directory since the
1081 # filehandle we would have gotten as first result of walk_tree_nofollow
1082 # earlier is still a handle to the underlying directory instead of the
1084 $destdh = PVE
::Tools
::openat
(fileno($parentfd), $last_dir, PVE
::Tools
::O_PATH
| O_NOFOLLOW
| O_DIRECTORY
);
1085 die "failed to open mount point: $!\n" if !$destdh;
1089 # no separate function because 99% of the time it's the wrong thing to use.
1090 if (syscall(269, fileno($destdh), $dot, &POSIX
::W_OK
, 0) != -1) {
1091 die "failed to mark bind mount read only\n";
1093 die "read-only check failed: $!\n" if $! != EROFS
;
1096 # For the rootfs we don't have a parentfd so we open the path directly.
1097 # Note that this means bindmounting any prefix of the host's
1098 # /var/lib/lxc/$vmid path into another container is considered a grave
1100 sysopen $destdh, $last_dir, O_PATH
| O_DIRECTORY
;
1101 die "failed to open mount point: $!\n" if !$destdh;
1104 my ($srcdev, $srcinode) = stat($srcdh);
1105 my ($dstdev, $dstinode) = stat($destdh);
1109 return ($srcdev == $dstdev && $srcinode == $dstinode);
1112 # Perform the actual bind mounting:
1113 sub __bindmount_do
{
1114 my ($dir, $dest, $ro, @extra_opts) = @_;
1115 PVE
::Tools
::run_command
(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1117 eval { PVE
::Tools
::run_command
(['mount', '-o', 'bind,remount,ro', $dest]); };
1119 warn "bindmount error\n";
1120 # don't leave writable bind-mounts behind...
1121 PVE
::Tools
::run_command
(['umount', $dest]);
1128 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1130 my $srcdh = __bindmount_prepare
('/', $dir);
1132 __bindmount_do
($dir, $dest, $ro, @extra_opts);
1134 if (!__bindmount_verify
($srcdh, $parentfd, $last_dir, $ro)) {
1135 PVE
::Tools
::run_command
(['umount', $dest]);
1136 die "detected mount path change at: $dir\n";
1140 # Cleanup $rootdir a bit (double and trailing slashes), build the mount path
1141 # from $rootdir and $mount and walk the path from $rootdir to the final
1142 # directory to check for symlinks.
1143 sub __mount_prepare_rootdir
{
1144 my ($rootdir, $mount) = @_;
1145 $rootdir =~ s!/+!/!g;
1146 $rootdir =~ s!/+$!!;
1147 my $mount_path = "$rootdir/$mount";
1148 my ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow
($rootdir, $mount, 1);
1149 return ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir);
1152 # use $rootdir = undef to just return the corresponding mount path
1153 sub mountpoint_mount
{
1154 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1156 my $volid = $mountpoint->{volume
};
1157 my $mount = $mountpoint->{mp
};
1158 my $type = $mountpoint->{type
};
1159 my $quota = !$snapname && !$mountpoint->{ro
} && $mountpoint->{quota
};
1162 return if !$volid || !$mount;
1167 my ($mpfd, $parentfd, $last_dir);
1169 if (defined($rootdir)) {
1170 ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir) =
1171 __mount_prepare_rootdir
($rootdir, $mount);
1174 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1176 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1179 my $acl = $mountpoint->{acl
};
1180 if (defined($acl)) {
1181 $optstring .= ($acl ?
'acl' : 'noacl');
1183 my $readonly = $mountpoint->{ro
};
1185 my @extra_opts = ('-o', $optstring) if $optstring;
1189 my $scfg = PVE
::Storage
::storage_config
($storage_cfg, $storage);
1191 # early sanity checks:
1192 # we otherwise call realpath on the rbd url
1193 die "containers on rbd storage without krbd are not supported\n"
1194 if $scfg->{type
} eq 'rbd' && !$scfg->{krbd
};
1196 my $path = PVE
::Storage
::path
($storage_cfg, $volid, $snapname);
1198 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1199 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1201 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1203 if ($format eq 'subvol') {
1206 if ($scfg->{type
} eq 'zfspool') {
1207 my $path_arg = $path;
1208 $path_arg =~ s!^/+!!;
1209 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1211 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1214 if (defined($acl) && $scfg->{type
} eq 'zfspool') {
1215 my $acltype = ($acl ?
'acltype=posixacl' : 'acltype=noacl');
1216 my (undef, $name) = PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1217 $name .= "\@$snapname" if defined($snapname);
1218 PVE
::Tools
::run_command
(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1220 bindmount
($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1221 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1224 return wantarray ?
($path, 0, undef) : $path;
1225 } elsif ($format eq 'raw' || $format eq 'iso') {
1226 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1227 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1228 # Our autodev hook expects the /dev/dm-* device currently
1229 # and will create the /dev/mapper symlink accordingly
1230 $path = Cwd
::realpath
($path);
1231 die "failed to get device path\n" if !$path;
1232 ($path) = ($path =~ /^(.*)$/s); #untaint
1236 if ($format eq 'iso') {
1237 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1238 } elsif ($isBase || defined($snapname)) {
1239 PVE
::Tools
::run_command
(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1242 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1244 push @extra_opts, '-o', 'ro' if $readonly;
1245 PVE
::Tools
::run_command
(['mount', @extra_opts, $path, $mount_path]);
1249 my $use_loopdev = 0;
1250 if ($scfg->{path
}) {
1251 $mounted_dev = run_with_loopdev
($domount, $path);
1253 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' ||
1254 $scfg->{type
} eq 'rbd' || $scfg->{type
} eq 'lvmthin') {
1255 $mounted_dev = $path;
1258 die "unsupported storage type '$scfg->{type}'\n";
1260 return wantarray ?
($path, $use_loopdev, $mounted_dev) : $path;
1262 die "unsupported image format '$format'\n";
1264 } elsif ($type eq 'device') {
1265 push @extra_opts, '-o', 'ro' if $readonly;
1266 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1267 # See the NOTE above about devicemapper canonicalization
1268 my ($devpath) = (Cwd
::realpath
($volid) =~ /^(.*)$/s); # realpath() taints
1269 PVE
::Tools
::run_command
(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1270 return wantarray ?
($volid, 0, $devpath) : $volid;
1271 } elsif ($type eq 'bind') {
1272 die "directory '$volid' does not exist\n" if ! -d
$volid;
1273 bindmount
($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1274 warn "cannot enable quota control for bind mounts\n" if $quota;
1275 return wantarray ?
($volid, 0, undef) : $volid;
1278 die "unsupported storage";
1282 my ($dev, $rootuid, $rootgid) = @_;
1284 PVE
::Tools
::run_command
(['mkfs.ext4', '-O', 'mmp',
1285 '-E', "root_owner=$rootuid:$rootgid",
1290 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1292 if ($volid =~ m!^/dev/.+!) {
1297 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1299 die "cannot format volume '$volid' with no storage\n" if !$storage;
1301 PVE
::Storage
::activate_volumes
($storage_cfg, [$volid]);
1303 my $path = PVE
::Storage
::path
($storage_cfg, $volid);
1305 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1306 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1308 die "cannot format volume '$volid' (format == $format)\n"
1309 if $format ne 'raw';
1311 mkfs
($path, $rootuid, $rootgid);
1315 my ($storecfg, $vollist) = @_;
1317 foreach my $volid (@$vollist) {
1318 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1324 my ($storecfg, $vmid, $settings, $conf) = @_;
1329 my (undef, $rootuid, $rootgid) = PVE
::LXC
::parse_id_maps
($conf);
1330 my $chown_vollist = [];
1332 PVE
::LXC
::Config-
>foreach_mountpoint($settings, sub {
1333 my ($ms, $mountpoint) = @_;
1335 my $volid = $mountpoint->{volume
};
1336 my $mp = $mountpoint->{mp
};
1338 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1340 if ($storage && ($volid =~ m/^([^:\s]+):(\d+(\.\d+)?)$/)) {
1341 my ($storeid, $size_gb) = ($1, $2);
1343 my $size_kb = int(${size_gb
}*1024) * 1024;
1345 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storage);
1346 # fixme: use better naming ct-$vmid-disk-X.raw?
1348 if ($scfg->{type
} eq 'dir' || $scfg->{type
} eq 'nfs') {
1350 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw',
1352 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1354 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1356 push @$chown_vollist, $volid;
1358 } elsif ($scfg->{type
} eq 'zfspool') {
1360 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1362 push @$chown_vollist, $volid;
1363 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' || $scfg->{type
} eq 'lvmthin') {
1365 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1366 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1368 } elsif ($scfg->{type
} eq 'rbd') {
1370 die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd
};
1371 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1372 format_disk
($storecfg, $volid, $rootuid, $rootgid);
1374 die "unable to create containers on storage type '$scfg->{type}'\n";
1376 push @$vollist, $volid;
1377 $mountpoint->{volume
} = $volid;
1378 $mountpoint->{size
} = $size_kb * 1024;
1379 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1381 # use specified/existing volid/dir/device
1382 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1386 PVE
::Storage
::activate_volumes
($storecfg, $chown_vollist, undef);
1387 foreach my $volid (@$chown_vollist) {
1388 my $path = PVE
::Storage
::path
($storecfg, $volid, undef);
1389 chown($rootuid, $rootgid, $path);
1391 PVE
::Storage
::deactivate_volumes
($storecfg, $chown_vollist, undef);
1393 # free allocated images on error
1395 destroy_disks
($storecfg, $vollist);
1401 # bash completion helper
1403 sub complete_os_templates
{
1404 my ($cmdname, $pname, $cvalue) = @_;
1406 my $cfg = PVE
::Storage
::config
();
1410 if ($cvalue =~ m/^([^:]+):/) {
1414 my $vtype = $cmdname eq 'restore' ?
'backup' : 'vztmpl';
1415 my $data = PVE
::Storage
::template_list
($cfg, $storeid, $vtype);
1418 foreach my $id (keys %$data) {
1419 foreach my $item (@{$data->{$id}}) {
1420 push @$res, $item->{volid
} if defined($item->{volid
});
1427 my $complete_ctid_full = sub {
1430 my $idlist = vmstatus
();
1432 my $active_hash = list_active_containers
();
1436 foreach my $id (keys %$idlist) {
1437 my $d = $idlist->{$id};
1438 if (defined($running)) {
1439 next if $d->{template
};
1440 next if $running && !$active_hash->{$id};
1441 next if !$running && $active_hash->{$id};
1450 return &$complete_ctid_full();
1453 sub complete_ctid_stopped
{
1454 return &$complete_ctid_full(0);
1457 sub complete_ctid_running
{
1458 return &$complete_ctid_full(1);
1468 my $lxc = $conf->{lxc
};
1469 foreach my $entry (@$lxc) {
1470 my ($key, $value) = @$entry;
1471 next if $key ne 'lxc.id_map';
1472 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1473 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1474 push @$id_map, [$type, $ct, $host, $length];
1476 $rootuid = $host if $type eq 'u';
1477 $rootgid = $host if $type eq 'g';
1480 die "failed to parse id_map: $value\n";
1484 if (!@$id_map && $conf->{unprivileged
}) {
1485 # Should we read them from /etc/subuid?
1486 $id_map = [ ['u', '0', '100000', '65536'],
1487 ['g', '0', '100000', '65536'] ];
1488 $rootuid = $rootgid = 100000;
1491 return ($id_map, $rootuid, $rootgid);
1494 sub userns_command
{
1497 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];