13 use Fcntl
qw(O_RDONLY O_NOFOLLOW O_DIRECTORY);
14 use Errno
qw(ELOOP ENOTDIR EROFS ECONNREFUSED);
17 use PVE
::Exception
qw(raise_perm_exc);
21 use PVE
::JSONSchema
qw(get_standard_option);
22 use PVE
::Tools
qw($IPV6RE $IPV4RE dir_glob_foreach lock_file lock_file_full O_PATH);
25 use PVE
::AccessControl
;
30 use Time
::HiRes qw
(gettimeofday
);
32 my $LXC_CONFIG_PATH = '/usr/share/lxc/config';
34 my $nodename = PVE
::INotify
::nodename
();
36 my $cpuinfo= PVE
::ProcFSTools
::read_cpuinfo
();
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', vmid
=> $vmid };
57 unlink PVE
::LXC
::Config-
>config_file($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 $parse_cpuacct_stat = sub {
106 my ($vmid, $unprivileged) = @_;
108 my $raw = read_cgroup_value
('cpuacct', $vmid, $unprivileged, 'cpuacct.stat', 1);
112 if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
122 our $vmstatus_return_properties = {
123 vmid
=> get_standard_option
('pve-vmid'),
125 description
=> "LXC Container status.",
127 enum
=> ['stopped', 'running'],
130 description
=> "Maximum memory in bytes.",
136 description
=> "Maximum SWAP memory in bytes.",
142 description
=> "Root disk size in bytes.",
148 description
=> "Container name.",
153 description
=> "Uptime.",
156 renderer
=> 'duration',
159 description
=> "Maximum usable CPUs.",
168 my $list = $opt_vmid ?
{ $opt_vmid => { type
=> 'lxc', vmid
=> $opt_vmid }} : config_list
();
170 my $active_hash = list_active_containers
();
172 my $cpucount = $cpuinfo->{cpus
} || 1;
174 my $cdtime = gettimeofday
;
176 my $uptime = (PVE
::ProcFSTools
::read_proc_uptime
(1))[0];
177 my $clock_ticks = POSIX
::sysconf
(&POSIX
::_SC_CLK_TCK
);
179 my $unprivileged = {};
181 foreach my $vmid (keys %$list) {
182 my $d = $list->{$vmid};
184 eval { $d->{pid
} = find_lxc_pid
($vmid) if defined($active_hash->{$vmid}); };
185 warn $@ if $@; # ignore errors (consider them stopped)
187 $d->{status
} = $d->{pid
} ?
'running' : 'stopped';
189 my $cfspath = PVE
::LXC
::Config-
>cfs_config_path($vmid);
190 my $conf = PVE
::Cluster
::cfs_read_file
($cfspath) || {};
192 $unprivileged->{$vmid} = $conf->{unprivileged
};
194 $d->{name
} = $conf->{'hostname'} || "CT$vmid";
195 $d->{name
} =~ s/[\s]//g;
197 $d->{cpus
} = $conf->{cores
} || $conf->{cpulimit
};
198 $d->{cpus
} = $cpucount if !$d->{cpus
};
200 $d->{lock} = $conf->{lock} || '';
203 my $res = get_container_disk_usage
($vmid, $d->{pid
});
204 $d->{disk
} = $res->{used
};
205 $d->{maxdisk
} = $res->{total
};
208 # use 4GB by default ??
209 if (my $rootfs = $conf->{rootfs
}) {
210 my $rootinfo = PVE
::LXC
::Config-
>parse_ct_rootfs($rootfs);
211 $d->{maxdisk
} = $rootinfo->{size
} || (4*1024*1024*1024);
213 $d->{maxdisk
} = 4*1024*1024*1024;
219 $d->{maxmem
} = ($conf->{memory
}||512)*1024*1024;
220 $d->{maxswap
} = ($conf->{swap
}//0)*1024*1024;
231 $d->{template
} = PVE
::LXC
::Config-
>is_template($conf);
234 foreach my $vmid (keys %$list) {
235 my $d = $list->{$vmid};
238 next if !$pid; # skip stopped CTs
240 my $proc_pid_stat = PVE
::ProcFSTools
::read_proc_pid_stat
($pid);
241 $d->{uptime
} = int(($uptime - $proc_pid_stat->{starttime
}) / $clock_ticks); # the method lxcfs uses
243 my $unpriv = $unprivileged->{$vmid};
245 if (-d
'/sys/fs/cgroup/memory') {
246 my $memory_stat = read_cgroup_list
('memory', $vmid, $unpriv, 'memory.stat');
247 my $mem_usage_in_bytes = read_cgroup_value
('memory', $vmid, $unpriv, 'memory.usage_in_bytes');
249 $d->{mem
} = $mem_usage_in_bytes - $memory_stat->{total_cache
};
250 $d->{swap
} = read_cgroup_value
('memory', $vmid, $unpriv, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
256 if (-d
'/sys/fs/cgroup/blkio') {
257 my $blkio_bytes = read_cgroup_value
('blkio', $vmid, $unpriv, 'blkio.throttle.io_service_bytes', 1);
258 my @bytes = split(/\n/, $blkio_bytes);
259 foreach my $byte (@bytes) {
260 if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
261 $d->{diskread
} += $2 if $key eq 'Read';
262 $d->{diskwrite
} += $2 if $key eq 'Write';
270 if (-d
'/sys/fs/cgroup/cpuacct') {
271 my $pstat = $parse_cpuacct_stat->($vmid, $unpriv);
273 my $used = $pstat->{utime} + $pstat->{stime
};
275 my $old = $last_proc_vmid_stat->{$vmid};
277 $last_proc_vmid_stat->{$vmid} = {
285 my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz
};
288 my $dutime = $used - $old->{used
};
290 $d->{cpu
} = (($dutime/$dtime)* $cpucount) / $d->{cpus
};
291 $last_proc_vmid_stat->{$vmid} = {
297 $d->{cpu
} = $old->{cpu
};
304 my $netdev = PVE
::ProcFSTools
::read_proc_net_dev
();
306 foreach my $dev (keys %$netdev) {
307 next if $dev !~ m/^veth([1-9]\d*)i/;
309 my $d = $list->{$vmid};
313 $d->{netout
} += $netdev->{$dev}->{receive
};
314 $d->{netin
} += $netdev->{$dev}->{transmit
};
321 sub read_cgroup_list
($$$$) {
322 my ($group, $vmid, $unprivileged, $name) = @_;
324 my $content = read_cgroup_value
($group, $vmid, $unprivileged, $name, 1);
326 return { split(/\s+/, $content) };
329 sub read_cgroup_value
($$$$$) {
330 my ($group, $vmid, $unprivileged, $name, $full) = @_;
332 my $nsdir = $unprivileged ?
'' : 'ns/';
333 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/${nsdir}$name";
335 return PVE
::Tools
::file_get_contents
($path) if $full;
337 return PVE
::Tools
::file_read_firstline
($path);
340 sub write_cgroup_value
{
341 my ($group, $vmid, $name, $value) = @_;
343 my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
344 PVE
::ProcFSTools
::write_proc_entry
($path, $value) if -e
$path;
348 sub find_lxc_console_pids
{
352 PVE
::Tools
::dir_glob_foreach
('/proc', '\d+', sub {
355 my $cmdline = PVE
::Tools
::file_read_firstline
("/proc/$pid/cmdline");
358 my @args = split(/\0/, $cmdline);
360 # search for lxc-console -n <vmid>
361 return if scalar(@args) != 3;
362 return if $args[1] ne '-n';
363 return if $args[2] !~ m/^\d+$/;
364 return if $args[0] !~ m
|^(/usr/bin
/)?lxc-console
$|;
368 push @{$res->{$vmid}}, $pid;
380 $pid = $1 if $line =~ m/^PID:\s+(\d+)$/;
382 PVE
::Tools
::run_command
(['lxc-info', '-n', $vmid, '-p'], outfunc
=> $parser);
384 die "unable to get PID for CT $vmid (not running?)\n" if !$pid;
389 # Note: we cannot use Net:IP, because that only allows strict
391 sub parse_ipv4_cidr
{
392 my ($cidr, $noerr) = @_;
394 if ($cidr =~ m!^($IPV4RE)(?:/(\d+))$! && ($2 > 7) && ($2 <= 32)) {
395 return { address
=> $1, netmask
=> $PVE::Network
::ipv4_reverse_mask-
>[$2] };
398 return undef if $noerr;
400 die "unable to parse ipv4 address/mask\n";
403 sub get_cgroup_subsystems
{
406 my $data = PVE
::Tools
::file_get_contents
('/proc/self/cgroup');
407 while ($data =~ /^\d+:([^:\n]*):.*$/gm) {
410 $v1->{$_} = 1 foreach split(/,/, $type);
415 return wantarray ?
($v1, $v2) : $v1;
418 # Currently we do not need to create seccomp profile 'files' as the only
419 # choice our configuration actually allows is "with or without keyctl()",
420 # so we distinguish between using lxc's "default" seccomp profile and our
421 # added pve-userns.seccomp file.
423 # This returns a configuration line added to the raw lxc config.
424 sub make_seccomp_config
{
425 my ($conf, $unprivileged, $features) = @_;
426 # User-configured profile has precedence, note that the user's entry would
427 # be written 'after' this line anyway...
428 if (PVE
::LXC
::Config-
>has_lxc_entry($conf, 'lxc.seccomp.profile')) {
429 # Warn the user if this conflicts with a feature:
430 if ($features->{keyctl
}) {
431 warn "explicitly configured lxc.seccomp.profile overrides the following settings: features:keyctl\n";
436 # Privileged containers keep using the default (which is already part of
437 # the files included via lxc.include, so we don't need to write it out,
438 # that way it stays admin-configurable via /usr/share/lxc/config/... as
440 return '' if !$unprivileged;
442 # Unprivileged containers will get keyctl() disabled by default as a
443 # workaround for systemd-networkd behavior. But we have an option to
444 # explicitly enable it:
445 return '' if $features->{keyctl
};
447 # Finally we're in an unprivileged container without `keyctl` set
448 # explicitly. We have a file prepared for this:
449 return "lxc.seccomp.profile = $LXC_CONFIG_PATH/pve-userns.seccomp\n";
452 # Since lxc-3.0.2 we can have lxc generate a profile for the container
453 # automatically. The default should be equivalent to the old
454 # `lxc-container-default-cgns` profile.
456 # Additionally this also added `lxc.apparmor.raw` which can be used to inject
457 # additional lines into the profile. We can use that to allow mounting specific
459 sub make_apparmor_config
{
460 my ($conf, $unprivileged, $features) = @_;
462 # user-configured profile has precedence, but first we go through our own
463 # code to figure out whether we should warn the user:
465 my $raw = "lxc.apparmor.profile = generated\n";
468 if ($features->{fuse
}) {
469 # For the informational warning:
470 push @profile_uses, 'features:fuse';
473 # There's lxc.apparmor.allow_nesting now, which will add the necessary
474 # apparmor lines, create an apparmor namespace for the container, but also
475 # adds proc and sysfs mounts to /dev/.lxc/{proc,sys}. These do not have
476 # lxcfs mounted over them, because that would prevent the container from
477 # mounting new instances of them for nested containers.
478 if ($features->{nesting
}) {
479 push @profile_uses, 'features:nesting';
480 $raw .= "lxc.apparmor.allow_nesting = 1\n"
482 # In the default profile in /etc/apparmor.d we patch this in because
483 # otherwise a container can for example run `chown` on /sys, breaking
484 # access to it for non-CAP_DAC_OVERRIDE tools on the host:
485 $raw .= "lxc.apparmor.raw = deny mount -> /proc/,\n";
486 $raw .= "lxc.apparmor.raw = deny mount -> /sys/,\n";
487 # Preferably we could use the 'remount' flag but this does not sit well
488 # with apparmor_parser currently:
489 # mount options=(rw, nosuid, nodev, noexec, remount) -> /sys/,
492 if (my $mount = $features->{mount
}) {
493 push @profile_uses, 'features:mount';
494 foreach my $fs (PVE
::Tools
::split_list
($mount)) {
495 $raw .= "lxc.apparmor.raw = mount fstype=$fs,\n";
501 if (PVE
::LXC
::Config-
>has_lxc_entry($conf, 'lxc.apparmor.profile')) {
502 if (length(my $used = join(', ', @profile_uses))) {
503 warn "explicitly configured lxc.apparmor.profile overrides the following settings: $used\n";
511 sub update_lxc_config
{
512 my ($vmid, $conf) = @_;
514 my $dir = "/var/lib/lxc/$vmid";
516 if ($conf->{template
}) {
518 unlink "$dir/config";
525 die "missing 'arch' - internal error" if !$conf->{arch
};
526 $raw .= "lxc.arch = $conf->{arch}\n";
528 my $custom_idmap = PVE
::LXC
::Config-
>has_lxc_entry($conf, 'lxc.idmap');
529 my $unprivileged = $conf->{unprivileged
} || $custom_idmap;
531 my $ostype = $conf->{ostype
} || die "missing 'ostype' - internal error";
533 my $cfgpath = '/usr/share/lxc/config';
534 my $inc = "$cfgpath/$ostype.common.conf";
535 $inc ="$cfgpath/common.conf" if !-f
$inc;
536 $raw .= "lxc.include = $inc\n";
538 $inc = "$cfgpath/$ostype.userns.conf";
539 $inc = "$cfgpath/userns.conf" if !-f
$inc;
540 $raw .= "lxc.include = $inc\n";
543 my $features = PVE
::LXC
::Config-
>parse_features($conf->{features
});
545 $raw .= make_seccomp_config
($conf, $unprivileged, $features);
546 $raw .= make_apparmor_config
($conf, $unprivileged, $features);
547 if ($features->{fuse
}) {
548 $raw .= "lxc.apparmor.raw = mount fstype=fuse,\n";
549 $raw .= "lxc.mount.entry = /dev/fuse dev/fuse none bind,create=file 0 0\n";
552 # WARNING: DO NOT REMOVE this without making sure that loop device nodes
553 # cannot be exposed to the container with r/w access (cgroup perms).
554 # When this is enabled mounts will still remain in the monitor's namespace
555 # after the container unmounted them and thus will not detach from their
556 # files while the container is running!
557 $raw .= "lxc.monitor.unshare = 1\n";
559 my $cgv1 = get_cgroup_subsystems
();
561 # Should we read them from /etc/subuid?
562 if ($unprivileged && !$custom_idmap) {
563 $raw .= "lxc.idmap = u 0 100000 65536\n";
564 $raw .= "lxc.idmap = g 0 100000 65536\n";
567 if (!PVE
::LXC
::Config-
>has_dev_console($conf)) {
568 $raw .= "lxc.console.path = none\n";
569 $raw .= "lxc.cgroup.devices.deny = c 5:1 rwm\n" if $cgv1->{devices
};
572 my $ttycount = PVE
::LXC
::Config-
>get_tty_count($conf);
573 $raw .= "lxc.tty.max = $ttycount\n";
575 # some init scripts expect a linux terminal (turnkey).
576 $raw .= "lxc.environment = TERM=linux\n";
578 my $utsname = $conf->{hostname
} || "CT$vmid";
579 $raw .= "lxc.uts.name = $utsname\n";
581 if ($cgv1->{memory
}) {
582 my $memory = $conf->{memory
} || 512;
583 my $swap = $conf->{swap
} // 0;
585 my $lxcmem = int($memory*1024*1024);
586 $raw .= "lxc.cgroup.memory.limit_in_bytes = $lxcmem\n";
588 my $lxcswap = int(($memory + $swap)*1024*1024);
589 $raw .= "lxc.cgroup.memory.memsw.limit_in_bytes = $lxcswap\n";
593 if (my $cpulimit = $conf->{cpulimit
}) {
594 $raw .= "lxc.cgroup.cpu.cfs_period_us = 100000\n";
595 my $value = int(100000*$cpulimit);
596 $raw .= "lxc.cgroup.cpu.cfs_quota_us = $value\n";
599 my $shares = $conf->{cpuunits
} || 1024;
600 $raw .= "lxc.cgroup.cpu.shares = $shares\n";
603 die "missing 'rootfs' configuration\n"
604 if !defined($conf->{rootfs
});
606 my $mountpoint = PVE
::LXC
::Config-
>parse_ct_rootfs($conf->{rootfs
});
608 $raw .= "lxc.rootfs.path = $dir/rootfs\n";
610 foreach my $k (sort keys %$conf) {
611 next if $k !~ m/^net(\d+)$/;
613 my $d = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$k});
614 $raw .= "lxc.net.$ind.type = veth\n";
615 $raw .= "lxc.net.$ind.veth.pair = veth${vmid}i${ind}\n";
616 $raw .= "lxc.net.$ind.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr
});
617 $raw .= "lxc.net.$ind.name = $d->{name}\n" if defined($d->{name
});
618 $raw .= "lxc.net.$ind.mtu = $d->{mtu}\n" if defined($d->{mtu
});
621 if ($cgv1->{cpuset
}) {
623 if (my $lxcconf = $conf->{lxc
}) {
624 foreach my $entry (@$lxcconf) {
625 my ($k, $v) = @$entry;
626 $had_cpuset = 1 if $k eq 'lxc.cgroup.cpuset.cpus';
631 my $cores = $conf->{cores
};
632 if (!$had_cpuset && $cores) {
633 my $cpuset = eval { PVE
::CpuSet-
>new_from_cgroup('lxc', 'effective_cpus') };
634 $cpuset = PVE
::CpuSet-
>new_from_cgroup('', 'effective_cpus') if !$cpuset;
635 my @members = $cpuset->members();
636 while (scalar(@members) > $cores) {
637 my $randidx = int(rand(scalar(@members)));
638 $cpuset->delete($members[$randidx]);
639 splice(@members, $randidx, 1); # keep track of the changes
641 $raw .= "lxc.cgroup.cpuset.cpus = ".$cpuset->short_string()."\n";
645 File
::Path
::mkpath
("$dir/rootfs");
647 PVE
::Tools
::file_set_contents
("$dir/config", $raw);
650 # verify and cleanup nameserver list (replace \0 with ' ')
651 sub verify_nameserver_list
{
652 my ($nameserver_list) = @_;
655 foreach my $server (PVE
::Tools
::split_list
($nameserver_list)) {
656 PVE
::JSONSchema
::pve_verify_ip
($server);
660 return join(' ', @list);
663 sub verify_searchdomain_list
{
664 my ($searchdomain_list) = @_;
667 foreach my $server (PVE
::Tools
::split_list
($searchdomain_list)) {
668 # todo: should we add checks for valid dns domains?
672 return join(' ', @list);
675 sub get_console_command
{
676 my ($vmid, $conf, $escapechar) = @_;
678 # '-1' as $escapechar disables keyboard escape sequence
679 # any other passed char (a-z) will result in <Ctrl+$escapechar q>
681 my $cmode = PVE
::LXC
::Config-
>get_cmode($conf);
684 if ($cmode eq 'console') {
685 push @$cmd, 'lxc-console', '-n', $vmid, '-t', 0;
686 push @$cmd, '-e', $escapechar if $escapechar;
687 } elsif ($cmode eq 'tty') {
688 push @$cmd, 'lxc-console', '-n', $vmid;
689 push @$cmd, '-e', $escapechar if $escapechar;
690 } elsif ($cmode eq 'shell') {
691 push @$cmd, 'lxc-attach', '--clear-env', '-n', $vmid;
693 die "internal error";
699 sub get_primary_ips
{
702 # return data from net0
704 return undef if !defined($conf->{net0
});
705 my $net = PVE
::LXC
::Config-
>parse_lxc_network($conf->{net0
});
707 my $ipv4 = $net->{ip
};
709 if ($ipv4 =~ /^(dhcp|manual)$/) {
715 my $ipv6 = $net->{ip6
};
717 if ($ipv6 =~ /^(auto|dhcp|manual)$/) {
724 return ($ipv4, $ipv6);
727 sub delete_mountpoint_volume
{
728 my ($storage_cfg, $vmid, $volume) = @_;
730 return if PVE
::LXC
::Config-
>classify_mountpoint($volume) ne 'volume';
732 my ($vtype, $name, $owner) = PVE
::Storage
::parse_volname
($storage_cfg, $volume);
733 PVE
::Storage
::vdisk_free
($storage_cfg, $volume) if $vmid == $owner;
736 sub destroy_lxc_container
{
737 my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
739 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
740 my ($ms, $mountpoint) = @_;
741 delete_mountpoint_volume
($storage_cfg, $vmid, $mountpoint->{volume
});
744 rmdir "/var/lib/lxc/$vmid/rootfs";
745 unlink "/var/lib/lxc/$vmid/config";
746 rmdir "/var/lib/lxc/$vmid";
747 if (defined $replacement_conf) {
748 PVE
::LXC
::Config-
>write_config($vmid, $replacement_conf);
750 destroy_config
($vmid);
753 #my $cmd = ['lxc-destroy', '-n', $vmid ];
754 #PVE::Tools::run_command($cmd);
757 sub vm_stop_cleanup
{
758 my ($storage_cfg, $vmid, $conf, $keepActive) = @_;
763 my $vollist = PVE
::LXC
::Config-
>get_vm_volumes($conf);
764 PVE
::Storage
::deactivate_volumes
($storage_cfg, $vollist);
767 warn $@ if $@; # avoid errors - just warn
770 my $safe_num_ne = sub {
773 return 0 if !defined($a) && !defined($b);
774 return 1 if !defined($a);
775 return 1 if !defined($b);
780 my $safe_string_ne = sub {
783 return 0 if !defined($a) && !defined($b);
784 return 1 if !defined($a);
785 return 1 if !defined($b);
791 my ($vmid, $conf, $opt, $newnet, $netid, $rootdir) = @_;
793 if ($newnet->{type
} ne 'veth') {
794 # for when there are physical interfaces
795 die "cannot update interface of type $newnet->{type}";
798 my $veth = "veth${vmid}i${netid}";
799 my $eth = $newnet->{name
};
801 if (my $oldnetcfg = $conf->{$opt}) {
802 my $oldnet = PVE
::LXC
::Config-
>parse_lxc_network($oldnetcfg);
804 if (&$safe_string_ne($oldnet->{hwaddr
}, $newnet->{hwaddr
}) ||
805 &$safe_string_ne($oldnet->{name
}, $newnet->{name
})) {
807 PVE
::Network
::veth_delete
($veth);
808 delete $conf->{$opt};
809 PVE
::LXC
::Config-
>write_config($vmid, $conf);
811 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
814 if (&$safe_string_ne($oldnet->{bridge
}, $newnet->{bridge
}) ||
815 &$safe_num_ne($oldnet->{tag
}, $newnet->{tag
}) ||
816 &$safe_num_ne($oldnet->{firewall
}, $newnet->{firewall
})) {
818 if ($oldnet->{bridge
}) {
819 PVE
::Network
::tap_unplug
($veth);
820 foreach (qw(bridge tag firewall)) {
821 delete $oldnet->{$_};
823 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
824 PVE
::LXC
::Config-
>write_config($vmid, $conf);
827 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
828 # This includes the rate:
829 foreach (qw(bridge tag firewall rate)) {
830 $oldnet->{$_} = $newnet->{$_} if $newnet->{$_};
832 } elsif (&$safe_string_ne($oldnet->{rate
}, $newnet->{rate
})) {
833 # Rate can be applied on its own but any change above needs to
834 # include the rate in tap_plug since OVS resets everything.
835 PVE
::Network
::tap_rate_limit
($veth, $newnet->{rate
});
836 $oldnet->{rate
} = $newnet->{rate
}
838 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($oldnet);
839 PVE
::LXC
::Config-
>write_config($vmid, $conf);
842 hotplug_net
($vmid, $conf, $opt, $newnet, $netid);
845 update_ipconfig
($vmid, $conf, $opt, $eth, $newnet, $rootdir);
849 my ($vmid, $conf, $opt, $newnet, $netid) = @_;
851 my $veth = "veth${vmid}i${netid}";
852 my $vethpeer = $veth . "p";
853 my $eth = $newnet->{name
};
855 PVE
::Network
::veth_create
($veth, $vethpeer, $newnet->{bridge
}, $newnet->{hwaddr
});
856 PVE
::Network
::tap_plug
($veth, $newnet->{bridge
}, $newnet->{tag
}, $newnet->{firewall
}, $newnet->{trunks
}, $newnet->{rate
});
858 # attach peer in container
859 my $cmd = ['lxc-device', '-n', $vmid, 'add', $vethpeer, "$eth" ];
860 PVE
::Tools
::run_command
($cmd);
862 # link up peer in container
863 $cmd = ['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', '/sbin/ip', 'link', 'set', $eth ,'up' ];
864 PVE
::Tools
::run_command
($cmd);
866 my $done = { type
=> 'veth' };
867 foreach (qw(bridge tag firewall hwaddr name)) {
868 $done->{$_} = $newnet->{$_} if $newnet->{$_};
870 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($done);
872 PVE
::LXC
::Config-
>write_config($vmid, $conf);
875 sub update_ipconfig
{
876 my ($vmid, $conf, $opt, $eth, $newnet, $rootdir) = @_;
878 my $lxc_setup = PVE
::LXC
::Setup-
>new($conf, $rootdir);
880 my $optdata = PVE
::LXC
::Config-
>parse_lxc_network($conf->{$opt});
885 PVE
::Tools
::run_command
(['lxc-attach', '-n', $vmid, '-s', 'NETWORK', '--', @_], %$cmdargs);
887 my $ipcmd = sub { &$nscmd({}, '/sbin/ip', @_) };
889 my $change_ip_config = sub {
890 my ($ipversion) = @_;
892 my $family_opt = "-$ipversion";
893 my $suffix = $ipversion == 4 ?
'' : $ipversion;
897 my $newip = $newnet->{$ip};
898 my $newgw = $newnet->{$gw};
899 my $oldip = $optdata->{$ip};
900 my $oldgw = $optdata->{$gw};
902 my $change_ip = &$safe_string_ne($oldip, $newip);
903 my $change_gw = &$safe_string_ne($oldgw, $newgw);
905 return if !$change_ip && !$change_gw;
907 # step 1: add new IP, if this fails we cancel
908 my $is_real_ip = ($newip && $newip !~ /^(?:auto|dhcp|manual)$/);
909 if ($change_ip && $is_real_ip) {
910 eval { &$ipcmd($family_opt, 'addr', 'add', $newip, 'dev', $eth); };
917 # step 2: replace gateway
918 # If this fails we delete the added IP and cancel.
919 # If it succeeds we save the config and delete the old IP, ignoring
920 # errors. The config is then saved.
921 # Note: 'ip route replace' can add
925 if ($is_real_ip && !PVE
::Network
::is_ip_in_cidr
($newgw, $newip, $ipversion)) {
926 &$ipcmd($family_opt, 'route', 'add', $newgw, 'dev', $eth);
928 &$ipcmd($family_opt, 'route', 'replace', 'default', 'via', $newgw);
932 # the route was not replaced, the old IP is still available
933 # rollback (delete new IP) and cancel
935 eval { &$ipcmd($family_opt, 'addr', 'del', $newip, 'dev', $eth); };
936 warn $@ if $@; # no need to die here
941 eval { &$ipcmd($family_opt, 'route', 'del', 'default'); };
942 # if the route was not deleted, the guest might have deleted it manually
946 if ($oldgw && $oldip && !PVE
::Network
::is_ip_in_cidr
($oldgw, $oldip)) {
947 eval { &$ipcmd($family_opt, 'route', 'del', $oldgw, 'dev', $eth); };
948 # warn if the route was deleted manually
953 # from this point on we save the configuration
954 # step 3: delete old IP ignoring errors
955 if ($change_ip && $oldip && $oldip !~ /^(?:auto|dhcp)$/) {
956 # We need to enable promote_secondaries, otherwise our newly added
957 # address will be removed along with the old one.
960 if ($ipversion == 4) {
961 &$nscmd({ outfunc
=> sub { $promote = int(shift) } },
962 'cat', "/proc/sys/net/ipv4/conf/$eth/promote_secondaries");
963 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=1");
965 &$ipcmd($family_opt, 'addr', 'del', $oldip, 'dev', $eth);
967 warn $@ if $@; # no need to die here
969 if ($ipversion == 4) {
970 &$nscmd({}, 'sysctl', "net.ipv4.conf.$eth.promote_secondaries=$promote");
974 foreach my $property ($ip, $gw) {
975 if ($newnet->{$property}) {
976 $optdata->{$property} = $newnet->{$property};
978 delete $optdata->{$property};
981 $conf->{$opt} = PVE
::LXC
::Config-
>print_lxc_network($optdata);
982 PVE
::LXC
::Config-
>write_config($vmid, $conf);
983 $lxc_setup->setup_network($conf);
986 &$change_ip_config(4);
987 &$change_ip_config(6);
991 my $enter_namespace = sub {
992 my ($vmid, $pid, $which, $type) = @_;
993 sysopen my $fd, "/proc/$pid/ns/$which", O_RDONLY
994 or die "failed to open $which namespace of container $vmid: $!\n";
995 PVE
::Tools
::setns
(fileno($fd), $type)
996 or die "failed to enter $which namespace of container $vmid: $!\n";
1000 my $do_syncfs = sub {
1001 my ($vmid, $pid, $socket) = @_;
1003 &$enter_namespace($vmid, $pid, 'mnt', PVE
::Tools
::CLONE_NEWNS
);
1005 # Tell the parent process to start reading our /proc/mounts
1006 print {$socket} "go\n";
1009 # Receive /proc/self/mounts
1010 my $mountdata = do { local $/ = undef; <$socket> };
1013 # Now sync all mountpoints...
1014 my $mounts = PVE
::ProcFSTools
::parse_mounts
($mountdata);
1015 foreach my $mp (@$mounts) {
1016 my ($what, $dir, $fs) = @$mp;
1017 next if $fs eq 'fuse.lxcfs';
1018 eval { PVE
::Tools
::sync_mountpoint
($dir); };
1023 sub sync_container_namespace
{
1025 my $pid = find_lxc_pid
($vmid);
1027 # SOCK_DGRAM is nicer for barriers but cannot be slurped
1028 socketpair my $pfd, my $cfd, AF_UNIX
, SOCK_STREAM
, PF_UNSPEC
1029 or die "failed to create socketpair: $!\n";
1032 die "fork failed: $!\n" if !defined($child);
1037 &$do_syncfs($vmid, $pid, $cfd);
1047 die "failed to enter container namespace\n" if $go ne "go\n";
1049 open my $mounts, '<', "/proc/$child/mounts"
1050 or die "failed to open container's /proc/mounts: $!\n";
1051 my $mountdata = do { local $/ = undef; <$mounts> };
1053 print {$pfd} $mountdata;
1056 while (waitpid($child, 0) != $child) {}
1057 die "failed to sync container namespace\n" if $? != 0;
1060 sub template_create
{
1061 my ($vmid, $conf) = @_;
1063 my $storecfg = PVE
::Storage
::config
();
1065 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
1066 my ($ms, $mountpoint) = @_;
1068 my $volid = $mountpoint->{volume
};
1070 die "Template feature is not available for '$volid'\n"
1071 if !PVE
::Storage
::volume_has_feature
($storecfg, 'template', $volid);
1074 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
1075 my ($ms, $mountpoint) = @_;
1077 my $volid = $mountpoint->{volume
};
1079 PVE
::Storage
::activate_volumes
($storecfg, [$volid]);
1081 my $template_volid = PVE
::Storage
::vdisk_create_base
($storecfg, $volid);
1082 $mountpoint->{volume
} = $template_volid;
1083 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq "rootfs");
1086 PVE
::LXC
::Config-
>write_config($vmid, $conf);
1089 sub check_ct_modify_config_perm
{
1090 my ($rpcenv, $authuser, $vmid, $pool, $newconf, $delete) = @_;
1092 return 1 if $authuser eq 'root@pam';
1095 my ($opt, $delete) = @_;
1096 if ($opt eq 'cores' || $opt eq 'cpuunits' || $opt eq 'cpulimit') {
1097 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.CPU']);
1098 } elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
1099 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
1101 my $data = $opt eq 'rootfs' ? PVE
::LXC
::Config-
>parse_ct_rootfs($newconf->{$opt})
1102 : PVE
::LXC
::Config-
>parse_ct_mountpoint($newconf->{$opt});
1103 raise_perm_exc
("mount point type $data->{type} is only allowed for root\@pam")
1104 if $data->{type
} ne 'volume';
1105 } elsif ($opt eq 'memory' || $opt eq 'swap') {
1106 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Memory']);
1107 } elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
1108 $opt eq 'searchdomain' || $opt eq 'hostname') {
1109 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
1110 } elsif ($opt eq 'features') {
1111 # For now this is restricted to root@pam
1112 raise_perm_exc
("changing feature flags is only allowed for root\@pam");
1114 $rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
1118 foreach my $opt (keys %$newconf) {
1121 foreach my $opt (@$delete) {
1129 my ($vmid, $storage_cfg, $conf, $noerr) = @_;
1131 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
1132 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
1134 PVE
::LXC
::Config-
>foreach_mountpoint_reverse($conf, sub {
1135 my ($ms, $mountpoint) = @_;
1137 my $volid = $mountpoint->{volume
};
1138 my $mount = $mountpoint->{mp
};
1140 return if !$volid || !$mount;
1142 my $mount_path = "$rootdir/$mount";
1143 $mount_path =~ s!/+!/!g;
1145 return if !PVE
::ProcFSTools
::is_mounted
($mount_path);
1148 PVE
::Tools
::run_command
(['umount', '-d', $mount_path]);
1161 my ($vmid, $storage_cfg, $conf, $ignore_ro) = @_;
1163 my $rootdir = "/var/lib/lxc/$vmid/rootfs";
1164 File
::Path
::make_path
($rootdir);
1166 my $volid_list = PVE
::LXC
::Config-
>get_vm_volumes($conf);
1167 PVE
::Storage
::activate_volumes
($storage_cfg, $volid_list);
1170 PVE
::LXC
::Config-
>foreach_mountpoint($conf, sub {
1171 my ($ms, $mountpoint) = @_;
1173 $mountpoint->{ro
} = 0 if $ignore_ro;
1175 mountpoint_mount
($mountpoint, $rootdir, $storage_cfg);
1179 warn "mounting container failed\n";
1180 umount_all
($vmid, $storage_cfg, $conf, 1);
1188 sub mountpoint_mount_path
{
1189 my ($mountpoint, $storage_cfg, $snapname) = @_;
1191 return mountpoint_mount
($mountpoint, undef, $storage_cfg, $snapname);
1199 if ($line =~ m
@^(/dev/loop\d
+):@) {
1203 my $cmd = ['losetup', '--associated', $path];
1204 PVE
::Tools
::run_command
($cmd, outfunc
=> $parser);
1208 # Run a function with a file attached to a loop device.
1209 # The loop device is always detached afterwards (or set to autoclear).
1210 # Returns the loop device.
1211 sub run_with_loopdev
{
1212 my ($func, $file, $readonly) = @_;
1213 my $device = query_loopdev
($file);
1214 # Try to reuse an existing device
1216 # We assume that whoever setup the loop device is responsible for
1224 if ($line =~ m
@^(/dev/loop\d
+)$@) {
1234 push @$losetup_cmd, '-r' if $readonly;
1235 PVE
::Tools
::run_command
($losetup_cmd, outfunc
=> $parser);
1236 die "failed to setup loop device for $file\n" if !$device;
1237 eval { &$func($device); };
1239 PVE
::Tools
::run_command
(['losetup', '-d', $device]);
1244 # In scalar mode: returns a file handle to the deepest directory node.
1245 # In list context: returns a list of:
1246 # * the deepest directory node
1247 # * the 2nd deepest directory (parent of the above)
1248 # * directory name of the last directory
1249 # So that the path $2/$3 should lead to $1 afterwards.
1250 sub walk_tree_nofollow
($$$) {
1251 my ($start, $subdir, $mkdir) = @_;
1253 # splitdir() returns '' for empty components including the leading /
1254 my @comps = grep { length($_)>0 } File
::Spec-
>splitdir($subdir);
1256 sysopen(my $fd, $start, O_PATH
| O_DIRECTORY
)
1257 or die "failed to open start directory $start: $!\n";
1260 my $last_component = undef;
1262 foreach my $component (@comps) {
1263 $dir .= "/$component";
1264 my $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1267 # failed, check for symlinks and try to create the path
1268 die "symlink encountered at: $dir\n" if $! == ELOOP
|| $! == ENOTDIR
;
1269 die "cannot open directory $dir: $!\n" if !$mkdir;
1271 # We don't check for errors on mkdirat() here and just try to
1272 # openat() again, since at least one error (EEXIST) is an
1273 # expected possibility if multiple containers start
1274 # simultaneously. If someone else injects a symlink now then
1275 # the subsequent openat() will fail due to O_NOFOLLOW anyway.
1276 PVE
::Tools
::mkdirat
(fileno($fd), $component, 0755);
1278 $next = PVE
::Tools
::openat
(fileno($fd), $component, O_NOFOLLOW
| O_DIRECTORY
);
1279 die "failed to create path: $dir: $!\n" if !$next;
1282 close $second if defined($last_component);
1283 $last_component = $component;
1288 return ($fd, defined($last_component) && $second, $last_component) if wantarray;
1289 close $second if defined($last_component);
1293 # To guard against symlink attack races against other currently running
1294 # containers with shared recursive bind mount hierarchies we prepare a
1295 # directory handle for the directory we're mounting over to verify the
1296 # mountpoint afterwards.
1297 sub __bindmount_prepare
{
1298 my ($hostroot, $dir) = @_;
1299 my $srcdh = walk_tree_nofollow
($hostroot, $dir, 0);
1303 # Assuming we mount to rootfs/a/b/c, verify with the directory handle to 'b'
1304 # ($parentfd) that 'b/c' (openat($parentfd, 'c')) really leads to the directory
1305 # we intended to bind mount.
1306 sub __bindmount_verify
{
1307 my ($srcdh, $parentfd, $last_dir, $ro) = @_;
1310 # Open the mount point path coming from the parent directory since the
1311 # filehandle we would have gotten as first result of walk_tree_nofollow
1312 # earlier is still a handle to the underlying directory instead of the
1314 $destdh = PVE
::Tools
::openat
(fileno($parentfd), $last_dir, PVE
::Tools
::O_PATH
| O_NOFOLLOW
| O_DIRECTORY
);
1315 die "failed to open mount point: $!\n" if !$destdh;
1318 # no separate function because 99% of the time it's the wrong thing to use.
1319 if (syscall(PVE
::Syscall
::faccessat
, fileno($destdh), $dot, &POSIX
::W_OK
, 0) != -1) {
1320 die "failed to mark bind mount read only\n";
1322 die "read-only check failed: $!\n" if $! != EROFS
;
1325 # For the rootfs we don't have a parentfd so we open the path directly.
1326 # Note that this means bindmounting any prefix of the host's
1327 # /var/lib/lxc/$vmid path into another container is considered a grave
1329 sysopen $destdh, $last_dir, O_PATH
| O_DIRECTORY
;
1330 die "failed to open mount point: $!\n" if !$destdh;
1333 my ($srcdev, $srcinode) = stat($srcdh);
1334 my ($dstdev, $dstinode) = stat($destdh);
1338 return ($srcdev == $dstdev && $srcinode == $dstinode);
1341 # Perform the actual bind mounting:
1342 sub __bindmount_do
{
1343 my ($dir, $dest, $ro, @extra_opts) = @_;
1344 PVE
::Tools
::run_command
(['mount', '-o', 'bind', @extra_opts, $dir, $dest]);
1346 eval { PVE
::Tools
::run_command
(['mount', '-o', 'bind,remount,ro', $dest]); };
1348 warn "bindmount error\n";
1349 # don't leave writable bind-mounts behind...
1350 PVE
::Tools
::run_command
(['umount', $dest]);
1357 my ($dir, $parentfd, $last_dir, $dest, $ro, @extra_opts) = @_;
1359 my $srcdh = __bindmount_prepare
('/', $dir);
1361 __bindmount_do
($dir, $dest, $ro, @extra_opts);
1363 if (!__bindmount_verify
($srcdh, $parentfd, $last_dir, $ro)) {
1364 PVE
::Tools
::run_command
(['umount', $dest]);
1365 die "detected mount path change at: $dir\n";
1369 # Cleanup $rootdir a bit (double and trailing slashes), build the mount path
1370 # from $rootdir and $mount and walk the path from $rootdir to the final
1371 # directory to check for symlinks.
1372 sub __mount_prepare_rootdir
{
1373 my ($rootdir, $mount) = @_;
1374 $rootdir =~ s!/+!/!g;
1375 $rootdir =~ s!/+$!!;
1376 my $mount_path = "$rootdir/$mount";
1377 my ($mpfd, $parentfd, $last_dir) = walk_tree_nofollow
($rootdir, $mount, 1);
1378 return ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir);
1381 # use $rootdir = undef to just return the corresponding mount path
1382 sub mountpoint_mount
{
1383 my ($mountpoint, $rootdir, $storage_cfg, $snapname) = @_;
1385 my $volid = $mountpoint->{volume
};
1386 my $mount = $mountpoint->{mp
};
1387 my $type = $mountpoint->{type
};
1388 my $quota = !$snapname && !$mountpoint->{ro
} && $mountpoint->{quota
};
1391 return if !$volid || !$mount;
1396 my ($mpfd, $parentfd, $last_dir);
1398 if (defined($rootdir)) {
1399 ($rootdir, $mount_path, $mpfd, $parentfd, $last_dir) =
1400 __mount_prepare_rootdir
($rootdir, $mount);
1403 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1405 die "unknown snapshot path for '$volid'" if !$storage && defined($snapname);
1408 my $acl = $mountpoint->{acl
};
1409 if (defined($acl)) {
1410 $optstring .= ($acl ?
'acl' : 'noacl');
1412 my $readonly = $mountpoint->{ro
};
1415 @extra_opts = ('-o', $optstring) if $optstring;
1419 my $scfg = PVE
::Storage
::storage_config
($storage_cfg, $storage);
1421 my $path = PVE
::Storage
::map_volume
($storage_cfg, $volid, $snapname);
1423 $path = PVE
::Storage
::path
($storage_cfg, $volid, $snapname) if !defined($path);
1425 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1426 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1428 $format = 'iso' if $vtype eq 'iso'; # allow to handle iso files
1430 if ($format eq 'subvol') {
1433 if ($scfg->{type
} eq 'zfspool') {
1434 my $path_arg = $path;
1435 $path_arg =~ s!^/+!!;
1436 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, '-t', 'zfs', $path_arg, $mount_path]);
1438 die "cannot mount subvol snapshots for storage type '$scfg->{type}'\n";
1441 if (defined($acl) && $scfg->{type
} eq 'zfspool') {
1442 my $acltype = ($acl ?
'acltype=posixacl' : 'acltype=noacl');
1443 my (undef, $name) = PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1444 $name .= "\@$snapname" if defined($snapname);
1445 PVE
::Tools
::run_command
(['zfs', 'set', $acltype, "$scfg->{pool}/$name"]);
1447 bindmount
($path, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts);
1448 warn "cannot enable quota control for bind mounted subvolumes\n" if $quota;
1451 return wantarray ?
($path, 0, undef) : $path;
1452 } elsif ($format eq 'raw' || $format eq 'iso') {
1453 # NOTE: 'mount' performs canonicalization without the '-c' switch, which for
1454 # device-mapper devices is special-cased to use the /dev/mapper symlinks.
1455 # Our autodev hook expects the /dev/dm-* device currently
1456 # and will create the /dev/mapper symlink accordingly
1457 $path = Cwd
::realpath
($path);
1458 die "failed to get device path\n" if !$path;
1459 ($path) = ($path =~ /^(.*)$/s); #untaint
1463 if ($format eq 'iso') {
1464 PVE
::Tools
::run_command
(['mount', '-o', 'ro', @extra_opts, $path, $mount_path]);
1465 } elsif ($isBase || defined($snapname)) {
1466 PVE
::Tools
::run_command
(['mount', '-o', 'ro,noload', @extra_opts, $path, $mount_path]);
1469 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0';
1471 push @extra_opts, '-o', 'ro' if $readonly;
1472 PVE
::Tools
::run_command
(['mount', @extra_opts, $path, $mount_path]);
1476 my $use_loopdev = 0;
1477 if ($scfg->{path
}) {
1478 $mounted_dev = run_with_loopdev
($domount, $path, $readonly);
1480 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' ||
1481 $scfg->{type
} eq 'rbd' || $scfg->{type
} eq 'lvmthin') {
1482 $mounted_dev = $path;
1485 die "unsupported storage type '$scfg->{type}'\n";
1487 return wantarray ?
($path, $use_loopdev, $mounted_dev) : $path;
1489 die "unsupported image format '$format'\n";
1491 } elsif ($type eq 'device') {
1492 push @extra_opts, '-o', 'ro' if $readonly;
1493 push @extra_opts, '-o', 'usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0' if $quota;
1494 # See the NOTE above about devicemapper canonicalization
1495 my ($devpath) = (Cwd
::realpath
($volid) =~ /^(.*)$/s); # realpath() taints
1496 PVE
::Tools
::run_command
(['mount', @extra_opts, $volid, $mount_path]) if $mount_path;
1497 return wantarray ?
($volid, 0, $devpath) : $volid;
1498 } elsif ($type eq 'bind') {
1499 die "directory '$volid' does not exist\n" if ! -d
$volid;
1500 bindmount
($volid, $parentfd, $last_dir//$rootdir, $mount_path, $readonly, @extra_opts) if $mount_path;
1501 warn "cannot enable quota control for bind mounts\n" if $quota;
1502 return wantarray ?
($volid, 0, undef) : $volid;
1505 die "unsupported storage";
1509 my ($dev, $rootuid, $rootgid) = @_;
1511 PVE
::Tools
::run_command
(['mkfs.ext4', '-O', 'mmp',
1512 '-E', "root_owner=$rootuid:$rootgid",
1517 my ($storage_cfg, $volid, $rootuid, $rootgid) = @_;
1519 if ($volid =~ m!^/dev/.+!) {
1524 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1526 die "cannot format volume '$volid' with no storage\n" if !$storage;
1528 PVE
::Storage
::activate_volumes
($storage_cfg, [$volid]);
1530 my $path = PVE
::Storage
::map_volume
($storage_cfg, $volid);
1532 $path = PVE
::Storage
::path
($storage_cfg, $volid) if !defined($path);
1534 my ($vtype, undef, undef, undef, undef, $isBase, $format) =
1535 PVE
::Storage
::parse_volname
($storage_cfg, $volid);
1537 die "cannot format volume '$volid' (format == $format)\n"
1538 if $format ne 'raw';
1540 mkfs
($path, $rootuid, $rootgid);
1544 my ($storecfg, $vollist) = @_;
1546 foreach my $volid (@$vollist) {
1547 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1553 my ($storecfg, $vmid, $storage, $size_kb, $rootuid, $rootgid) = @_;
1555 my $needs_chown = 0;
1558 my $scfg = PVE
::Storage
::storage_config
($storecfg, $storage);
1559 # fixme: use better naming ct-$vmid-disk-X.raw?
1563 if ($scfg->{type
} eq 'dir' || $scfg->{type
} eq 'nfs' || $scfg->{type
} eq 'cifs' ) {
1565 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw',
1569 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1573 } elsif ($scfg->{type
} eq 'zfspool') {
1575 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'subvol',
1578 } elsif ($scfg->{type
} eq 'drbd' || $scfg->{type
} eq 'lvm' || $scfg->{type
} eq 'lvmthin') {
1580 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1583 } elsif ($scfg->{type
} eq 'rbd') {
1585 $volid = PVE
::Storage
::vdisk_alloc
($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
1588 die "unable to create containers on storage type '$scfg->{type}'\n";
1590 format_disk
($storecfg, $volid, $rootuid, $rootgid) if $do_format;
1593 # in case formatting got interrupted:
1594 if (defined($volid)) {
1595 eval { PVE
::Storage
::vdisk_free
($storecfg, $volid); };
1601 return ($volid, $needs_chown);
1604 our $NEW_DISK_RE = qr/^([^:\s]+):(\d+(\.\d+)?)$/;
1606 my ($storecfg, $vmid, $settings, $conf) = @_;
1611 my (undef, $rootuid, $rootgid) = PVE
::LXC
::parse_id_maps
($conf);
1612 my $chown_vollist = [];
1614 PVE
::LXC
::Config-
>foreach_mountpoint($settings, sub {
1615 my ($ms, $mountpoint) = @_;
1617 my $volid = $mountpoint->{volume
};
1618 my $mp = $mountpoint->{mp
};
1620 my ($storage, $volname) = PVE
::Storage
::parse_volume_id
($volid, 1);
1622 if ($storage && ($volid =~ $NEW_DISK_RE)) {
1623 my ($storeid, $size_gb) = ($1, $2);
1625 my $size_kb = int(${size_gb
}*1024) * 1024;
1627 my $needs_chown = 0;
1628 ($volid, $needs_chown) = alloc_disk
($storecfg, $vmid, $storage, $size_kb, $rootuid, $rootgid);
1629 push @$chown_vollist, $volid if $needs_chown;
1630 push @$vollist, $volid;
1631 $mountpoint->{volume
} = $volid;
1632 $mountpoint->{size
} = $size_kb * 1024;
1633 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1635 # use specified/existing volid/dir/device
1636 $conf->{$ms} = PVE
::LXC
::Config-
>print_ct_mountpoint($mountpoint, $ms eq 'rootfs');
1640 PVE
::Storage
::activate_volumes
($storecfg, $chown_vollist, undef);
1641 foreach my $volid (@$chown_vollist) {
1642 my $path = PVE
::Storage
::path
($storecfg, $volid, undef);
1643 chown($rootuid, $rootgid, $path);
1645 PVE
::Storage
::deactivate_volumes
($storecfg, $chown_vollist, undef);
1647 # free allocated images on error
1649 destroy_disks
($storecfg, $vollist);
1655 sub update_disksize
{
1656 my ($vmid, $conf, $all_volumes) = @_;
1659 my $prefix = "CT $vmid:";
1661 my $update_mp = sub {
1662 my ($key, $mp, @param) = @_;
1663 my $size = $all_volumes->{$mp->{volume
}}->{size
} // 0;
1665 if (!defined($mp->{size
}) || $size != $mp->{size
}) {
1667 print "$prefix updated volume size of '$mp->{volume}' in config.\n";
1668 $mp->{size
} = $size;
1669 my $nomp = 1 if ($key eq 'rootfs');
1670 $conf->{$key} = PVE
::LXC
::Config-
>print_ct_mountpoint($mp, $nomp);
1674 PVE
::LXC
::Config-
>foreach_mountpoint($conf, $update_mp);
1680 my ($vmid, $conf, $all_volumes) = @_;
1683 my $prefix = "CT $vmid:";
1685 # Note: it is allowed to define multiple storage entries with the same path
1686 # (alias), so we need to check both 'volid' and real 'path' (two different
1687 # volid can point to the same path).
1689 # used and unused disks
1693 foreach my $opt (keys %$conf) {
1694 next if ($opt !~ m/^unused\d+$/);
1695 my $vol = $all_volumes->{$conf->{$opt}};
1696 $refpath->{$vol->{path
}} = $vol->{volid
};
1699 foreach my $key (keys %$all_volumes) {
1700 my $vol = $all_volumes->{$key};
1701 my $in_use = PVE
::LXC
::Config-
>is_volume_in_use($conf, $vol->{volid
});
1702 my $path = $vol->{path
};
1705 $refpath->{$path} = $key;
1706 delete $orphans->{$path};
1708 if ((!$orphans->{$path}) && (!$refpath->{$path})) {
1709 $orphans->{$path} = $key;
1714 for my $key (keys %$orphans) {
1715 my $disk = $orphans->{$key};
1716 my $unused = PVE
::LXC
::Config-
>add_unused_volume($conf, $disk);
1720 print "$prefix add unreferenced volume '$disk' as '$unused' to config.\n";
1728 my ($cfg, $vmid) = @_;
1730 my $info = PVE
::Storage
::vdisk_list
($cfg, undef, $vmid);
1732 my $all_volumes = {};
1733 foreach my $storeid (keys %$info) {
1734 foreach my $item (@{$info->{$storeid}}) {
1735 my $volid = $item->{volid
};
1736 next if !($volid && $item->{size
});
1737 $item->{path
} = PVE
::Storage
::path
($cfg, $volid);
1738 $all_volumes->{$volid} = $item;
1742 return $all_volumes;
1746 my ($vmid, $nolock, $dryrun) = @_;
1748 my $cfg = PVE
::Storage
::config
();
1750 # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
1751 # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
1752 foreach my $stor (keys %{$cfg->{ids
}}) {
1753 delete($cfg->{ids
}->{$stor}) if !$cfg->{ids
}->{$stor}->{content
}->{rootdir
};
1756 print "rescan volumes...\n";
1757 my $all_volumes = scan_volids
($cfg, $vmid);
1759 my $updatefn = sub {
1763 my $conf = PVE
::LXC
::Config-
>load_config($vmid);
1765 PVE
::LXC
::Config-
>check_lock($conf);
1768 foreach my $volid (keys %$all_volumes) {
1769 my $info = $all_volumes->{$volid};
1770 $vm_volids->{$volid} = $info if $info->{vmid
} == $vmid;
1773 my $upu = update_unused
($vmid, $conf, $vm_volids);
1774 my $upd = update_disksize
($vmid, $conf, $vm_volids);
1775 $changes = $upu || $upd;
1777 PVE
::LXC
::Config-
>write_config($vmid, $conf) if $changes && !$dryrun;
1780 if (defined($vmid)) {
1784 PVE
::LXC
::Config-
>lock_config($vmid, $updatefn, $vmid);
1787 my $vmlist = config_list
();
1788 foreach my $vmid (keys %$vmlist) {
1792 PVE
::LXC
::Config-
>lock_config($vmid, $updatefn, $vmid);
1799 # bash completion helper
1801 sub complete_os_templates
{
1802 my ($cmdname, $pname, $cvalue) = @_;
1804 my $cfg = PVE
::Storage
::config
();
1808 if ($cvalue =~ m/^([^:]+):/) {
1812 my $vtype = $cmdname eq 'restore' ?
'backup' : 'vztmpl';
1813 my $data = PVE
::Storage
::template_list
($cfg, $storeid, $vtype);
1816 foreach my $id (keys %$data) {
1817 foreach my $item (@{$data->{$id}}) {
1818 push @$res, $item->{volid
} if defined($item->{volid
});
1825 my $complete_ctid_full = sub {
1828 my $idlist = vmstatus
();
1830 my $active_hash = list_active_containers
();
1834 foreach my $id (keys %$idlist) {
1835 my $d = $idlist->{$id};
1836 if (defined($running)) {
1837 next if $d->{template
};
1838 next if $running && !$active_hash->{$id};
1839 next if !$running && $active_hash->{$id};
1848 return &$complete_ctid_full();
1851 sub complete_ctid_stopped
{
1852 return &$complete_ctid_full(0);
1855 sub complete_ctid_running
{
1856 return &$complete_ctid_full(1);
1866 my $lxc = $conf->{lxc
};
1867 foreach my $entry (@$lxc) {
1868 my ($key, $value) = @$entry;
1869 # FIXME: remove the 'id_map' variant when lxc-3.0 arrives
1870 next if $key ne 'lxc.idmap' && $key ne 'lxc.id_map';
1871 if ($value =~ /^([ug])\s+(\d+)\s+(\d+)\s+(\d+)\s*$/) {
1872 my ($type, $ct, $host, $length) = ($1, $2, $3, $4);
1873 push @$id_map, [$type, $ct, $host, $length];
1875 $rootuid = $host if $type eq 'u';
1876 $rootgid = $host if $type eq 'g';
1879 die "failed to parse idmap: $value\n";
1883 if (!@$id_map && $conf->{unprivileged
}) {
1884 # Should we read them from /etc/subuid?
1885 $id_map = [ ['u', '0', '100000', '65536'],
1886 ['g', '0', '100000', '65536'] ];
1887 $rootuid = $rootgid = 100000;
1890 return ($id_map, $rootuid, $rootgid);
1893 sub userns_command
{
1896 return ['lxc-usernsexec', (map { ('-m', join(':', @$_)) } @$id_map), '--'];
1902 my ($vmid, $conf, $skiplock) = @_;
1904 update_lxc_config
($vmid, $conf);
1906 my $skiplock_flag_fn = "/run/lxc/skiplock-$vmid";
1909 open(my $fh, '>', $skiplock_flag_fn) || die "failed to open $skiplock_flag_fn for writing: $!\n";
1913 my $cmd = ['systemctl', 'start', "pve-container\@$vmid"];
1915 eval { PVE
::Tools
::run_command
($cmd); };
1917 unlink $skiplock_flag_fn;
1924 # Helper to stop a container completely and make sure it has stopped completely.
1925 # This is necessary because we want the post-stop hook to have completed its
1926 # unmount-all step, but post-stop happens after lxc puts the container into the
1929 my ($vmid, $kill, $shutdown_timeout, $exit_timeout) = @_;
1931 # Open the container's command socket.
1932 my $path = "\0/var/lib/lxc/$vmid/command";
1933 my $sock = IO
::Socket
::UNIX-
>new(
1934 Type
=> SOCK_STREAM
(),
1938 return if $! == ECONNREFUSED
; # The container is not running
1939 die "failed to open container ${vmid}'s command socket: $!\n";
1942 # Stop the container:
1944 my $cmd = ['lxc-stop', '-n', $vmid];
1947 push @$cmd, '--kill'; # doesn't allow timeouts
1948 } elsif (defined($shutdown_timeout)) {
1949 push @$cmd, '--timeout', $shutdown_timeout;
1950 # Give run_command 5 extra seconds
1951 $shutdown_timeout += 5;
1954 eval { PVE
::Tools
::run_command
($cmd, timeout
=> $shutdown_timeout) };
1960 my $wait = sub { $result = <$sock>; };
1961 if (defined($exit_timeout)) {
1962 PVE
::Tools
::run_with_timeout
($exit_timeout, $wait);
1967 return if !defined $result; # monitor is gone and the ct has stopped.
1968 die "container did not stop\n";
1974 return PVE
::Tools
::run_fork
(sub {
1975 # Unshare the mount namespace
1976 die "failed to unshare mount namespace: $!\n"
1977 if !PVE
::Tools
::unshare
(PVE
::Tools
::CLONE_NEWNS
);
1978 PVE
::Tools
::run_command
(['mount', '--make-rslave', '/']);
1983 my $copy_volume = sub {
1984 my ($src_volid, $src, $dst_volid, $dest, $storage_cfg, $snapname) = @_;
1986 my $src_mp = { volume
=> $src_volid, mp
=> '/', ro
=> 1 };
1987 $src_mp->{type
} = PVE
::LXC
::Config-
>classify_mountpoint($src_volid);
1989 my $dst_mp = { volume
=> $dst_volid, mp
=> '/', ro
=> 0 };
1990 $dst_mp->{type
} = PVE
::LXC
::Config-
>classify_mountpoint($dst_volid);
1996 mountpoint_mount
($src_mp, $src, $storage_cfg, $snapname);
1997 push @mounted, $src;
1999 mountpoint_mount
($dst_mp, $dest, $storage_cfg);
2000 push @mounted, $dest;
2002 PVE
::Tools
::run_command
(['/usr/bin/rsync', '--stats', '-X', '-A', '--numeric-ids',
2003 '-aH', '--whole-file', '--sparse', '--one-file-system',
2007 foreach my $mount (reverse @mounted) {
2008 eval { PVE
::Tools
::run_command
(['/bin/umount', '--lazy', $mount], errfunc
=> sub{})};
2009 warn "Can't umount $mount\n" if $@;
2012 # If this fails they're used as mount points in a concurrent operation
2013 # (which should not happen but there's also no real need to get rid of them).
2020 # Should not be called after unsharing the mount namespace!
2022 my ($mp, $vmid, $storage, $storage_cfg, $conf, $snapname) = @_;
2024 die "cannot copy volumes of type $mp->{type}\n" if $mp->{type
} ne 'volume';
2025 File
::Path
::make_path
("/var/lib/lxc/$vmid");
2026 my $dest = "/var/lib/lxc/$vmid/.copy-volume-1";
2027 my $src = "/var/lib/lxc/$vmid/.copy-volume-2";
2029 # get id's for unprivileged container
2030 my (undef, $rootuid, $rootgid) = parse_id_maps
($conf);
2032 # Allocate the disk before unsharing in order to make sure zfs subvolumes
2033 # are visible in this namespace, otherwise the host only sees the empty
2034 # (not-mounted) directory.
2037 # Make sure $mp contains a correct size.
2038 $mp->{size
} = PVE
::Storage
::volume_size_info
($storage_cfg, $mp->{volume
});
2040 ($new_volid, $needs_chown) = alloc_disk
($storage_cfg, $vmid, $storage, $mp->{size
}/1024, $rootuid, $rootgid);
2042 PVE
::Storage
::activate_volumes
($storage_cfg, [$new_volid], undef);
2043 my $path = PVE
::Storage
::path
($storage_cfg, $new_volid, undef);
2044 chown($rootuid, $rootgid, $path);
2048 $copy_volume->($mp->{volume
}, $src, $new_volid, $dest, $storage_cfg, $snapname);
2052 PVE
::Storage
::vdisk_free
($storage_cfg, $new_volid)
2053 if defined($new_volid);