use PVE::ProcFSTools;
use PVE::Syscall;
use PVE::LXC::Config;
+use PVE::GuestHelpers;
use Time::HiRes qw (gettimeofday);
+my $LXC_CONFIG_PATH = '/usr/share/lxc/config';
+
my $nodename = PVE::INotify::nodename();
my $cpuinfo= PVE::ProcFSTools::read_cpuinfo();
sub destroy_config {
my ($vmid) = @_;
- unlink PVE::LXC::Config->config_file($vmid, $nodename);
+ my $config_fn = PVE::LXC::Config->config_file($vmid, $nodename);
+ unlink $config_fn or die "failed to remove config file: $!\n";
}
# container status helpers
type => 'number',
optional => 1,
},
+ lock => {
+ description => "The current config lock, if any.",
+ type => 'string',
+ optional => 1,
+ }
};
sub vmstatus {
eval { $d->{pid} = find_lxc_pid($vmid) if defined($active_hash->{$vmid}); };
warn $@ if $@; # ignore errors (consider them stopped)
- $d->{status} = $d->{pid} ? 'running' : 'stopped';
+ $d->{status} = $active_hash->{$vmid} ? 'running' : 'stopped';
my $cfspath = PVE::LXC::Config->cfs_config_path($vmid);
my $conf = PVE::Cluster::cfs_read_file($cfspath) || {};
$d->{diskwrite} = 0;
$d->{template} = PVE::LXC::Config->is_template($conf);
+ $d->{lock} = $conf->{lock} if $conf->{lock};
}
foreach my $vmid (keys %$list) {
}
if (-d '/sys/fs/cgroup/blkio') {
- my $blkio_bytes = read_cgroup_value('blkio', $vmid, $unpriv, 'blkio.throttle.io_service_bytes', 1);
+ my $blkio_bytes = read_cgroup_value('blkio', $vmid, 0, 'blkio.throttle.io_service_bytes', 1); # don't check if unpriv
my @bytes = split(/\n/, $blkio_bytes);
foreach my $byte (@bytes) {
if (my ($key, $value) = $byte =~ /(Read|Write)\s+(\d+)/) {
return wantarray ? ($v1, $v2) : $v1;
}
+# Currently we do not need to create seccomp profile 'files' as the only
+# choice our configuration actually allows is "with or without keyctl()",
+# so we distinguish between using lxc's "default" seccomp profile and our
+# added pve-userns.seccomp file.
+#
+# This returns a configuration line added to the raw lxc config.
+sub make_seccomp_config {
+ my ($conf, $unprivileged, $features) = @_;
+ # User-configured profile has precedence, note that the user's entry would
+ # be written 'after' this line anyway...
+ if (PVE::LXC::Config->has_lxc_entry($conf, 'lxc.seccomp.profile')) {
+ # Warn the user if this conflicts with a feature:
+ if ($features->{keyctl}) {
+ warn "explicitly configured lxc.seccomp.profile overrides the following settings: features:keyctl\n";
+ }
+ return '';
+ }
+
+ # Privileged containers keep using the default (which is already part of
+ # the files included via lxc.include, so we don't need to write it out,
+ # that way it stays admin-configurable via /usr/share/lxc/config/... as
+ # well)
+ return '' if !$unprivileged;
+
+ # Unprivileged containers will get keyctl() disabled by default as a
+ # workaround for systemd-networkd behavior. But we have an option to
+ # explicitly enable it:
+ return '' if $features->{keyctl};
+
+ # Finally we're in an unprivileged container without `keyctl` set
+ # explicitly. We have a file prepared for this:
+ return "lxc.seccomp.profile = $LXC_CONFIG_PATH/pve-userns.seccomp\n";
+}
+
+# Since lxc-3.0.2 we can have lxc generate a profile for the container
+# automatically. The default should be equivalent to the old
+# `lxc-container-default-cgns` profile.
+#
+# Additionally this also added `lxc.apparmor.raw` which can be used to inject
+# additional lines into the profile. We can use that to allow mounting specific
+# file systems.
+sub make_apparmor_config {
+ my ($conf, $unprivileged, $features) = @_;
+
+ # user-configured profile has precedence, but first we go through our own
+ # code to figure out whether we should warn the user:
+
+ my $raw = "lxc.apparmor.profile = generated\n";
+ my @profile_uses;
+
+ if ($features->{fuse}) {
+ # For the informational warning:
+ push @profile_uses, 'features:fuse';
+ }
+
+ # There's lxc.apparmor.allow_nesting now, which will add the necessary
+ # apparmor lines, create an apparmor namespace for the container, but also
+ # adds proc and sysfs mounts to /dev/.lxc/{proc,sys}. These do not have
+ # lxcfs mounted over them, because that would prevent the container from
+ # mounting new instances of them for nested containers.
+ if ($features->{nesting}) {
+ push @profile_uses, 'features:nesting';
+ $raw .= "lxc.apparmor.allow_nesting = 1\n"
+ } else {
+ # In the default profile in /etc/apparmor.d we patch this in because
+ # otherwise a container can for example run `chown` on /sys, breaking
+ # access to it for non-CAP_DAC_OVERRIDE tools on the host:
+ $raw .= "lxc.apparmor.raw = deny mount -> /proc/,\n";
+ $raw .= "lxc.apparmor.raw = deny mount -> /sys/,\n";
+ # Preferably we could use the 'remount' flag but this does not sit well
+ # with apparmor_parser currently:
+ # mount options=(rw, nosuid, nodev, noexec, remount) -> /sys/,
+ }
+
+ if (my $mount = $features->{mount}) {
+ push @profile_uses, 'features:mount';
+ foreach my $fs (PVE::Tools::split_list($mount)) {
+ $raw .= "lxc.apparmor.raw = mount fstype=$fs,\n";
+ }
+ }
+
+ # More to come?
+
+ if (PVE::LXC::Config->has_lxc_entry($conf, 'lxc.apparmor.profile')) {
+ if (length(my $used = join(', ', @profile_uses))) {
+ warn "explicitly configured lxc.apparmor.profile overrides the following settings: $used\n";
+ }
+ return '';
+ }
+
+ return $raw;
+}
+
sub update_lxc_config {
my ($vmid, $conf) = @_;
die "missing 'arch' - internal error" if !$conf->{arch};
$raw .= "lxc.arch = $conf->{arch}\n";
- my $unprivileged = $conf->{unprivileged};
- my $custom_idmap = grep { $_->[0] eq 'lxc.idmap' } @{$conf->{lxc}};
+ my $custom_idmap = PVE::LXC::Config->has_lxc_entry($conf, 'lxc.idmap');
+ my $unprivileged = $conf->{unprivileged} || $custom_idmap;
my $ostype = $conf->{ostype} || die "missing 'ostype' - internal error";
my $inc = "$cfgpath/$ostype.common.conf";
$inc ="$cfgpath/common.conf" if !-f $inc;
$raw .= "lxc.include = $inc\n";
- if ($unprivileged || $custom_idmap) {
+ if ($unprivileged) {
$inc = "$cfgpath/$ostype.userns.conf";
$inc = "$cfgpath/userns.conf" if !-f $inc;
$raw .= "lxc.include = $inc\n";
- $raw .= "lxc.seccomp.profile = $cfgpath/pve-userns.seccomp\n";
+ }
+
+ my $features = PVE::LXC::Config->parse_features($conf->{features});
+
+ $raw .= make_seccomp_config($conf, $unprivileged, $features);
+ $raw .= make_apparmor_config($conf, $unprivileged, $features);
+ if ($features->{fuse}) {
+ $raw .= "lxc.apparmor.raw = mount fstype=fuse,\n";
+ $raw .= "lxc.mount.entry = /dev/fuse dev/fuse none bind,create=file 0 0\n";
}
# WARNING: DO NOT REMOVE this without making sure that loop device nodes
}
sub get_console_command {
- my ($vmid, $conf, $noescapechar) = @_;
+ my ($vmid, $conf, $escapechar) = @_;
+
+ # '-1' as $escapechar disables keyboard escape sequence
+ # any other passed char (a-z) will result in <Ctrl+$escapechar q>
my $cmode = PVE::LXC::Config->get_cmode($conf);
my $cmd = [];
if ($cmode eq 'console') {
push @$cmd, 'lxc-console', '-n', $vmid, '-t', 0;
- push @$cmd, '-e', -1 if $noescapechar;
+ push @$cmd, '-e', $escapechar if $escapechar;
} elsif ($cmode eq 'tty') {
push @$cmd, 'lxc-console', '-n', $vmid;
- push @$cmd, '-e', -1 if $noescapechar;
+ push @$cmd, '-e', $escapechar if $escapechar;
} elsif ($cmode eq 'shell') {
push @$cmd, 'lxc-attach', '--clear-env', '-n', $vmid;
} else {
} elsif ($opt =~ m/^net\d+$/ || $opt eq 'nameserver' ||
$opt eq 'searchdomain' || $opt eq 'hostname') {
$rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Network']);
+ } elsif ($opt eq 'features') {
+ # For now this is restricted to root@pam
+ raise_perm_exc("changing feature flags is only allowed for root\@pam");
+ } elsif ($opt eq 'hookscript') {
+ # For now this is restricted to root@pam
+ raise_perm_exc("changing the hookscript is only allowed for root\@pam");
} else {
$rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Options']);
}
# The loop device is always detached afterwards (or set to autoclear).
# Returns the loop device.
sub run_with_loopdev {
- my ($func, $file) = @_;
+ my ($func, $file, $readonly) = @_;
my $device = query_loopdev($file);
# Try to reuse an existing device
if ($device) {
$device = $1;
}
};
- PVE::Tools::run_command(['losetup', '--show', '-f', $file], outfunc => $parser);
+ my $losetup_cmd = [
+ 'losetup',
+ '--show',
+ '-f',
+ $file,
+ ];
+ push @$losetup_cmd, '-r' if $readonly;
+ PVE::Tools::run_command($losetup_cmd, outfunc => $parser);
die "failed to setup loop device for $file\n" if !$device;
eval { &$func($device); };
my $err = $@;
my $scfg = PVE::Storage::storage_config($storage_cfg, $storage);
- # early sanity checks:
- # we otherwise call realpath on the rbd url
- die "containers on rbd storage without krbd are not supported\n"
- if $scfg->{type} eq 'rbd' && !$scfg->{krbd};
+ my $path = PVE::Storage::map_volume($storage_cfg, $volid, $snapname);
- my $path = PVE::Storage::path($storage_cfg, $volid, $snapname);
+ $path = PVE::Storage::path($storage_cfg, $volid, $snapname) if !defined($path);
my ($vtype, undef, undef, undef, undef, $isBase, $format) =
PVE::Storage::parse_volname($storage_cfg, $volid);
};
my $use_loopdev = 0;
if ($scfg->{path}) {
- $mounted_dev = run_with_loopdev($domount, $path);
+ $mounted_dev = run_with_loopdev($domount, $path, $readonly);
$use_loopdev = 1;
} elsif ($scfg->{type} eq 'drbd' || $scfg->{type} eq 'lvm' ||
$scfg->{type} eq 'rbd' || $scfg->{type} eq 'lvmthin') {
PVE::Storage::activate_volumes($storage_cfg, [$volid]);
- my $path = PVE::Storage::path($storage_cfg, $volid);
+ my $path = PVE::Storage::map_volume($storage_cfg, $volid);
+
+ $path = PVE::Storage::path($storage_cfg, $volid) if !defined($path);
my ($vtype, undef, undef, undef, undef, $isBase, $format) =
PVE::Storage::parse_volname($storage_cfg, $volid);
} elsif ($scfg->{type} eq 'rbd') {
- die "krbd option must be enabled on storage type '$scfg->{type}'\n" if !$scfg->{krbd};
$volid = PVE::Storage::vdisk_alloc($storecfg, $storage, $vmid, 'raw', undef, $size_kb);
$do_format = 1;
} else {
return $vollist;
}
+sub update_disksize {
+ my ($vmid, $conf, $all_volumes) = @_;
+
+ my $changes;
+ my $prefix = "CT $vmid:";
+
+ my $update_mp = sub {
+ my ($key, $mp, @param) = @_;
+ my $size = $all_volumes->{$mp->{volume}}->{size} // 0;
+
+ if (!defined($mp->{size}) || $size != $mp->{size}) {
+ $changes = 1;
+ print "$prefix updated volume size of '$mp->{volume}' in config.\n";
+ $mp->{size} = $size;
+ my $nomp = 1 if ($key eq 'rootfs');
+ $conf->{$key} = PVE::LXC::Config->print_ct_mountpoint($mp, $nomp);
+ }
+ };
+
+ PVE::LXC::Config->foreach_mountpoint($conf, $update_mp);
+
+ return $changes;
+}
+
+sub update_unused {
+ my ($vmid, $conf, $all_volumes) = @_;
+
+ my $changes;
+ my $prefix = "CT $vmid:";
+
+ # Note: it is allowed to define multiple storage entries with the same path
+ # (alias), so we need to check both 'volid' and real 'path' (two different
+ # volid can point to the same path).
+
+ # used and unused disks
+ my $refpath = {};
+ my $orphans = {};
+
+ foreach my $opt (keys %$conf) {
+ next if ($opt !~ m/^unused\d+$/);
+ my $vol = $all_volumes->{$conf->{$opt}};
+ $refpath->{$vol->{path}} = $vol->{volid};
+ }
+
+ foreach my $key (keys %$all_volumes) {
+ my $vol = $all_volumes->{$key};
+ my $in_use = PVE::LXC::Config->is_volume_in_use($conf, $vol->{volid});
+ my $path = $vol->{path};
+
+ if ($in_use) {
+ $refpath->{$path} = $key;
+ delete $orphans->{$path};
+ } else {
+ if ((!$orphans->{$path}) && (!$refpath->{$path})) {
+ $orphans->{$path} = $key;
+ }
+ }
+ }
+
+ for my $key (keys %$orphans) {
+ my $disk = $orphans->{$key};
+ my $unused = PVE::LXC::Config->add_unused_volume($conf, $disk);
+
+ if ($unused) {
+ $changes = 1;
+ print "$prefix add unreferenced volume '$disk' as '$unused' to config.\n";
+ }
+ }
+
+ return $changes;
+}
+
+sub scan_volids {
+ my ($cfg, $vmid) = @_;
+
+ my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
+
+ my $all_volumes = {};
+ foreach my $storeid (keys %$info) {
+ foreach my $item (@{$info->{$storeid}}) {
+ my $volid = $item->{volid};
+ next if !($volid && $item->{size});
+ $item->{path} = PVE::Storage::path($cfg, $volid);
+ $all_volumes->{$volid} = $item;
+ }
+ }
+
+ return $all_volumes;
+}
+
+sub rescan {
+ my ($vmid, $nolock, $dryrun) = @_;
+
+ my $cfg = PVE::Storage::config();
+
+ # FIXME: Remove once our RBD plugin can handle CT and VM on a single storage
+ # see: https://pve.proxmox.com/pipermail/pve-devel/2018-July/032900.html
+ foreach my $stor (keys %{$cfg->{ids}}) {
+ delete($cfg->{ids}->{$stor}) if !$cfg->{ids}->{$stor}->{content}->{rootdir};
+ }
+
+ print "rescan volumes...\n";
+ my $all_volumes = scan_volids($cfg, $vmid);
+
+ my $updatefn = sub {
+ my ($vmid) = @_;
+
+ my $changes;
+ my $conf = PVE::LXC::Config->load_config($vmid);
+
+ PVE::LXC::Config->check_lock($conf);
+
+ my $vm_volids = {};
+ foreach my $volid (keys %$all_volumes) {
+ my $info = $all_volumes->{$volid};
+ $vm_volids->{$volid} = $info if $info->{vmid} == $vmid;
+ }
+
+ my $upu = update_unused($vmid, $conf, $vm_volids);
+ my $upd = update_disksize($vmid, $conf, $vm_volids);
+ $changes = $upu || $upd;
+
+ PVE::LXC::Config->write_config($vmid, $conf) if $changes && !$dryrun;
+ };
+
+ if (defined($vmid)) {
+ if ($nolock) {
+ &$updatefn($vmid);
+ } else {
+ PVE::LXC::Config->lock_config($vmid, $updatefn, $vmid);
+ }
+ } else {
+ my $vmlist = config_list();
+ foreach my $vmid (keys %$vmlist) {
+ if ($nolock) {
+ &$updatefn($vmid);
+ } else {
+ PVE::LXC::Config->lock_config($vmid, $updatefn, $vmid);
+ }
+ }
+ }
+}
+
+
# bash completion helper
sub complete_os_templates {
my $cmd = ['systemctl', 'start', "pve-container\@$vmid"];
+ PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
eval { PVE::Tools::run_command($cmd); };
if (my $err = $@) {
unlink $skiplock_flag_fn;
die $err;
}
+ PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'post-start');
return;
}
die "failed to open container ${vmid}'s command socket: $!\n";
}
+ my $conf = PVE::LXC::Config->load_config($vmid);
+ PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-stop');
+
# Stop the container:
my $cmd = ['lxc-stop', '-n', $vmid];
}
my $copy_volume = sub {
- my ($src_volid, $src, $dst_volid, $dest, $storage_cfg, $snapname) = @_;
+ my ($src_volid, $src, $dst_volid, $dest, $storage_cfg, $snapname, $bwlimit) = @_;
- my $src_mp = { volume => $src_volid, mp => '/' };
+ my $src_mp = { volume => $src_volid, mp => '/', ro => 1 };
$src_mp->{type} = PVE::LXC::Config->classify_mountpoint($src_volid);
- my $dst_mp = { volume => $dst_volid, mp => '/' };
+ my $dst_mp = { volume => $dst_volid, mp => '/', ro => 0 };
$dst_mp->{type} = PVE::LXC::Config->classify_mountpoint($dst_volid);
my @mounted;
mountpoint_mount($dst_mp, $dest, $storage_cfg);
push @mounted, $dest;
+ $bwlimit //= 0;
+
PVE::Tools::run_command(['/usr/bin/rsync', '--stats', '-X', '-A', '--numeric-ids',
'-aH', '--whole-file', '--sparse', '--one-file-system',
- "$src/", $dest]);
+ "--bwlimit=$bwlimit", "$src/", $dest]);
};
my $err = $@;
foreach my $mount (reverse @mounted) {
# Should not be called after unsharing the mount namespace!
sub copy_volume {
- my ($mp, $vmid, $storage, $storage_cfg, $conf, $snapname) = @_;
+ my ($mp, $vmid, $storage, $storage_cfg, $conf, $snapname, $bwlimit) = @_;
die "cannot copy volumes of type $mp->{type}\n" if $mp->{type} ne 'volume';
File::Path::make_path("/var/lib/lxc/$vmid");
}
run_unshared(sub {
- $copy_volume->($mp->{volume}, $src, $new_volid, $dest, $storage_cfg, $snapname);
+ $copy_volume->($mp->{volume}, $src, $new_volid, $dest, $storage_cfg, $snapname, $bwlimit);
});
};
if (my $err = $@) {