use PVE::INotify;
use PVE::JSONSchema qw(get_standard_option);
use PVE::Tools qw(
- dir_glob_foreach file_get_contents file_set_contents lock_file
- lock_file_full AT_FDCWD O_PATH $IPV4RE $IPV6RE
+ run_command
+ dir_glob_foreach
+ file_get_contents
+ file_set_contents
+ AT_FDCWD
+ O_PATH
+ $IPV4RE
+ $IPV6RE
);
use PVE::CpuSet;
use PVE::Network;
use PVE::LXC::Config;
use PVE::GuestHelpers qw(safe_string_ne safe_num_ne safe_boolean_ne);
use PVE::LXC::Tools;
+use PVE::LXC::CGroup;
+use PVE::LXC::Monitor;
+use PVE::CGroup;
use Time::HiRes qw (gettimeofday);
my $have_sdn;
my $last_proc_vmid_stat;
-my $parse_cpuacct_stat = sub {
- my ($vmid, $unprivileged) = @_;
-
- my $raw = read_cgroup_value('cpuacct', $vmid, $unprivileged, 'cpuacct.stat', 1);
-
- my $stat = {};
-
- if ($raw =~ m/^user (\d+)\nsystem (\d+)\n/) {
-
- $stat->{utime} = $1;
- $stat->{stime} = $2;
-
- }
-
- return $stat;
-};
-
our $vmstatus_return_properties = {
vmid => get_standard_option('pve-vmid'),
status => {
$d->{cpus} = $conf->{cores} || $conf->{cpulimit};
$d->{cpus} = $cpucount if !$d->{cpus};
- $d->{lock} = $conf->{lock} || '';
$d->{tags} = $conf->{tags} if defined($conf->{tags});
if ($d->{pid}) {
$d->{disk} = 0;
# use 4GB by default ??
if (my $rootfs = $conf->{rootfs}) {
- my $rootinfo = PVE::LXC::Config->parse_ct_rootfs($rootfs);
+ my $rootinfo = PVE::LXC::Config->parse_volume('rootfs', $rootfs);
$d->{maxdisk} = $rootinfo->{size} || (4*1024*1024*1024);
} else {
$d->{maxdisk} = 4*1024*1024*1024;
my $unpriv = $unprivileged->{$vmid};
- if (-d '/sys/fs/cgroup/memory') {
- my $memory_stat = read_cgroup_list('memory', $vmid, $unpriv, 'memory.stat');
- my $mem_usage_in_bytes = read_cgroup_value('memory', $vmid, $unpriv, 'memory.usage_in_bytes');
+ my $cgroups = PVE::LXC::CGroup->new($vmid);
- $d->{mem} = $mem_usage_in_bytes - $memory_stat->{total_cache};
- $d->{swap} = read_cgroup_value('memory', $vmid, $unpriv, 'memory.memsw.usage_in_bytes') - $mem_usage_in_bytes;
+ if (defined(my $mem = $cgroups->get_memory_stat())) {
+ $d->{mem} = $mem->{mem};
+ $d->{swap} = $mem->{swap};
} else {
$d->{mem} = 0;
$d->{swap} = 0;
}
- if (-d '/sys/fs/cgroup/blkio') {
- 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+)/) {
- $d->{diskread} += $2 if $key eq 'Read';
- $d->{diskwrite} += $2 if $key eq 'Write';
- }
- }
+ if (defined(my $blkio = $cgroups->get_io_stats())) {
+ $d->{diskread} = $blkio->{diskread};
+ $d->{diskwrite} = $blkio->{diskwrite};
} else {
$d->{diskread} = 0;
$d->{diskwrite} = 0;
}
- if (-d '/sys/fs/cgroup/cpuacct') {
- my $pstat = $parse_cpuacct_stat->($vmid, $unpriv);
-
- my $used = $pstat->{utime} + $pstat->{stime};
+ if (defined(my $cpu = $cgroups->get_cpu_stat())) {
+ # Total time (in milliseconds) used up by the cpu.
+ my $used_ms = $cpu->{utime} + $cpu->{stime};
my $old = $last_proc_vmid_stat->{$vmid};
if (!$old) {
$last_proc_vmid_stat->{$vmid} = {
time => $cdtime,
- used => $used,
+ used => $used_ms,
cpu => 0,
};
next;
}
- my $dtime = ($cdtime - $old->{time}) * $cpucount * $cpuinfo->{user_hz};
-
- if ($dtime > 1000) {
- my $dutime = $used - $old->{used};
-
- $d->{cpu} = (($dutime/$dtime)* $cpucount) / $d->{cpus};
+ my $delta_ms = ($cdtime - $old->{time}) * $cpucount * 1000.0;
+ if ($delta_ms > 1000.0) {
+ my $delta_used_ms = $used_ms - $old->{used};
+ $d->{cpu} = (($delta_used_ms / $delta_ms) * $cpucount) / $d->{cpus};
$last_proc_vmid_stat->{$vmid} = {
time => $cdtime,
- used => $used,
+ used => $used_ms,
cpu => $d->{cpu},
};
} else {
return $list;
}
-sub read_cgroup_list($$$$) {
- my ($group, $vmid, $unprivileged, $name) = @_;
-
- my $content = read_cgroup_value($group, $vmid, $unprivileged, $name, 1);
-
- return { split(/\s+/, $content) };
-}
-
-sub read_cgroup_value($$$$$) {
- my ($group, $vmid, $unprivileged, $name, $full) = @_;
-
- my $nsdir = $unprivileged ? '' : 'ns/';
- my $path = "/sys/fs/cgroup/$group/lxc/$vmid/${nsdir}$name";
-
- return PVE::Tools::file_get_contents($path) if $full;
-
- return PVE::Tools::file_read_firstline($path);
-}
-
-sub write_cgroup_value {
- my ($group, $vmid, $name, $value) = @_;
-
- my $path = "/sys/fs/cgroup/$group/lxc/$vmid/$name";
- PVE::ProcFSTools::write_proc_entry($path, $value) if -e $path;
-
-}
-
sub find_lxc_console_pids {
my $res = {};
die "unable to parse ipv4 address/mask\n";
}
+# Deprecated. Use `PVE::CGroup::get_cgroup_controllers()` instead.
sub get_cgroup_subsystems {
- my $v1 = {};
- my $v2 = 0;
- my $data = PVE::Tools::file_get_contents('/proc/self/cgroup');
- while ($data =~ /^\d+:([^:\n]*):.*$/gm) {
- my $type = $1;
- if (length($type)) {
- $v1->{$_} = 1 foreach split(/,/, $type);
- } else {
- $v2 = 1;
- }
- }
- return wantarray ? ($v1, $v2) : $v1;
+ PVE::CGroup::get_v1_controllers();
}
# With seccomp trap to userspace we now have the ability to optionally forward
return;
}
+ my ($lxc_major, $lxc_minor) = get_lxc_version();
+
my $raw = '';
+ if ($lxc_major >= 4) {
+ # Explicitly don't use relative directories, which is the default, but
+ # note that we do this mostly because they are only applied for *some*
+ # cgroups. Our pve-container@.service now starts lxc-start with `-F`,
+ # so we also don't need to worry about the new monitor cgroup to
+ # confuse systemd.
+ $raw .= "lxc.cgroup.relative = 0\n";
+
+ # To make things easier, let's keep our previous cgroup layout and
+ # simply move the monitor outside:
+ $raw .= "lxc.cgroup.dir.monitor = lxc.monitor/$vmid\n";
+ # cgroup namespace separation for stronger limits:
+ $raw .= "lxc.cgroup.dir.container = lxc/$vmid\n";
+ $raw .= "lxc.cgroup.dir.container.inner = ns\n";
+ }
+
die "missing 'arch' - internal error" if !$conf->{arch};
$raw .= "lxc.arch = $conf->{arch}\n";
$raw .= "lxc.mount.entry = /dev/fuse dev/fuse none bind,create=file 0 0\n";
}
+ if ($unprivileged && !$features->{force_rw_sys}) {
+ # unpriv. CT default to sys:rw, but that doesn't always plays well with
+ # systemd, e.g., systemd-networkd https://systemd.io/CONTAINER_INTERFACE/
+ $raw .= "lxc.mount.auto = sys:mixed\n";
+ }
+
# WARNING: DO NOT REMOVE this without making sure that loop device nodes
# cannot be exposed to the container with r/w access (cgroup perms).
# When this is enabled mounts will still remain in the monitor's namespace
die "missing 'rootfs' configuration\n"
if !defined($conf->{rootfs});
- my $mountpoint = PVE::LXC::Config->parse_ct_rootfs($conf->{rootfs});
+ my $mountpoint = PVE::LXC::Config->parse_volume('rootfs', $conf->{rootfs});
$raw .= "lxc.rootfs.path = $dir/rootfs\n";
$raw .= "lxc.net.$ind.hwaddr = $d->{hwaddr}\n" if defined($d->{hwaddr});
$raw .= "lxc.net.$ind.name = $d->{name}\n" if defined($d->{name});
$raw .= "lxc.net.$ind.mtu = $d->{mtu}\n" if defined($d->{mtu});
+
+ # Starting with lxc 4.0, we do not patch lxc to execute our up-scripts.
+ if ($lxc_major >= 4) {
+ $raw .= "lxc.net.$ind.script.up = /usr/share/lxc/lxcnetaddbr\n";
+ }
}
- if ($cgv1->{cpuset}) {
- my $had_cpuset = 0;
- if (my $lxcconf = $conf->{lxc}) {
- foreach my $entry (@$lxcconf) {
- my ($k, $v) = @$entry;
- $had_cpuset = 1 if $k eq 'lxc.cgroup.cpuset.cpus';
- $raw .= "$k = $v\n";
- }
+ my $had_cpuset = 0;
+ if (my $lxcconf = $conf->{lxc}) {
+ foreach my $entry (@$lxcconf) {
+ my ($k, $v) = @$entry;
+ $had_cpuset = 1 if $k eq 'lxc.cgroup.cpuset.cpus' || $k eq 'lxc.cgroup2.cpuset.cpus';
+ $raw .= "$k = $v\n";
}
+ }
- my $cores = $conf->{cores};
- if (!$had_cpuset && $cores) {
- my $cpuset = eval { PVE::CpuSet->new_from_cgroup('lxc', 'effective_cpus') };
- $cpuset = PVE::CpuSet->new_from_cgroup('', 'effective_cpus') if !$cpuset;
- my @members = $cpuset->members();
- while (scalar(@members) > $cores) {
- my $randidx = int(rand(scalar(@members)));
- $cpuset->delete($members[$randidx]);
- splice(@members, $randidx, 1); # keep track of the changes
- }
- $raw .= "lxc.cgroup.cpuset.cpus = ".$cpuset->short_string()."\n";
+ my $cpuset;
+ my ($cpuset_cgroup, $cpuset_version) = eval { PVE::CGroup::cpuset_controller_path() };
+ if (defined($cpuset_cgroup)) {
+ $cpuset = eval { PVE::CpuSet->new_from_path("$cpuset_cgroup/lxc", 1) }
+ || PVE::CpuSet->new_from_path($cpuset_cgroup, 1);
+ }
+ my $cores = $conf->{cores};
+ if (!$had_cpuset && $cores && $cpuset) {
+ my @members = $cpuset->members();
+ while (scalar(@members) > $cores) {
+ my $randidx = int(rand(scalar(@members)));
+ $cpuset->delete($members[$randidx]);
+ splice(@members, $randidx, 1); # keep track of the changes
}
+ my $ver = $cpuset_version == 1 ? '' : '2';
+ $raw .= "lxc.cgroup$ver.cpuset.cpus = ".$cpuset->short_string()."\n";
}
File::Path::mkpath("$dir/rootfs");
my @list = ();
foreach my $server (PVE::Tools::split_list($nameserver_list)) {
- PVE::JSONSchema::pve_verify_ip($server);
+ PVE::LXC::Config::verify_ip_with_ll_iface($server);
push @list, $server;
}
}
sub destroy_lxc_container {
- my ($storage_cfg, $vmid, $conf, $replacement_conf) = @_;
+ my ($storage_cfg, $vmid, $conf, $replacement_conf, $purge_unreferenced) = @_;
- PVE::LXC::Config->foreach_mountpoint($conf, sub {
+ PVE::LXC::Config->foreach_volume_full($conf, {include_unused => 1}, sub {
my ($ms, $mountpoint) = @_;
delete_mountpoint_volume($storage_cfg, $vmid, $mountpoint->{volume});
});
+ if ($purge_unreferenced) { # also remove unreferenced disk
+ my $vmdisks = PVE::Storage::vdisk_list($storage_cfg, undef, $vmid, undef, 'rootdir');
+ PVE::Storage::foreach_volid($vmdisks, sub {
+ my ($volid, $sid, $volname, $d) = @_;
+ eval { PVE::Storage::vdisk_free($storage_cfg, $volid) };
+ warn $@ if $@;
+ });
+ }
+
rmdir "/var/lib/lxc/$vmid/rootfs";
unlink "/var/lib/lxc/$vmid/config";
rmdir "/var/lib/lxc/$vmid";
my $mountdata = do { local $/ = undef; <$socket> };
close $socket;
+ my %nosyncfs = (
+ cgroup => 1,
+ cgroup2 => 1,
+ devtmpfs => 1,
+ devpts => 1,
+ 'fuse.lxcfs' => 1,
+ fusectl => 1,
+ mqueue => 1,
+ proc => 1,
+ sysfs => 1,
+ tmpfs => 1,
+ );
+
# Now sync all mountpoints...
my $mounts = PVE::ProcFSTools::parse_mounts($mountdata);
foreach my $mp (@$mounts) {
my ($what, $dir, $fs) = @$mp;
- next if $fs eq 'fuse.lxcfs';
+ next if $nosyncfs{$fs};
eval { PVE::Tools::sync_mountpoint($dir); };
warn $@ if $@;
}
my $storecfg = PVE::Storage::config();
- PVE::LXC::Config->foreach_mountpoint($conf, sub {
+ PVE::LXC::Config->foreach_volume($conf, sub {
my ($ms, $mountpoint) = @_;
my $volid = $mountpoint->{volume};
if !PVE::Storage::volume_has_feature($storecfg, 'template', $volid);
});
- PVE::LXC::Config->foreach_mountpoint($conf, sub {
+ PVE::LXC::Config->foreach_volume($conf, sub {
my ($ms, $mountpoint) = @_;
my $volid = $mountpoint->{volume};
} elsif ($opt eq 'rootfs' || $opt =~ /^mp\d+$/) {
$rpcenv->check_vm_perm($authuser, $vmid, $pool, ['VM.Config.Disk']);
return if $delete;
- my $data = $opt eq 'rootfs' ? PVE::LXC::Config->parse_ct_rootfs($newconf->{$opt})
- : PVE::LXC::Config->parse_ct_mountpoint($newconf->{$opt});
+ my $data = PVE::LXC::Config->parse_volume($opt, $newconf->{$opt});
raise_perm_exc("mount point type $data->{type} is only allowed for root\@pam")
if $data->{type} ne 'volume';
} elsif ($opt eq 'memory' || $opt eq 'swap') {
my $res = 1;
- PVE::LXC::Config->foreach_mountpoint_reverse($conf, sub {
+ PVE::LXC::Config->foreach_volume_full($conf, {'reverse' => 1}, sub {
my ($ms, $mountpoint) = @_;
my $volid = $mountpoint->{volume};
my (undef, $rootuid, $rootgid) = parse_id_maps($conf);
eval {
- PVE::LXC::Config->foreach_mountpoint($conf, sub {
+ PVE::LXC::Config->foreach_volume($conf, sub {
my ($ms, $mountpoint) = @_;
$mountpoint->{ro} = 0 if $ignore_ro;
sub mkfs {
my ($dev, $rootuid, $rootgid) = @_;
- PVE::Tools::run_command(['mkfs.ext4', '-O', 'mmp',
- '-E', "root_owner=$rootuid:$rootgid",
- $dev]);
+ run_command(
+ [
+ 'mkfs.ext4',
+ '-O',
+ 'mmp',
+ '-E',
+ "root_owner=$rootuid:$rootgid",
+ $dev,
+ ],
+ outfunc => sub {
+ my $line = shift;
+ # a hack to print only the relevant stuff, i.e., the one which could help on repair
+ if ($line =~ /^(Creating filesystem|Filesystem UUID|Superblock backups|\s+\d+, \d)/) {
+ print "$line\n";
+ }
+ },
+ errfunc => sub {
+ my $line = shift;
+ print STDERR "$line\n" if $line && $line !~ /^mke2fs \d\.\d/;
+ }
+ );
}
sub format_disk {
my (undef, $rootuid, $rootgid) = PVE::LXC::parse_id_maps($conf);
my $chown_vollist = [];
- PVE::LXC::Config->foreach_mountpoint($settings, sub {
+ PVE::LXC::Config->foreach_volume($settings, sub {
my ($ms, $mountpoint) = @_;
my $volid = $mountpoint->{volume};
}
};
- PVE::LXC::Config->foreach_mountpoint($conf, $update_mp);
+ PVE::LXC::Config->foreach_volume($conf, $update_mp);
return $changes;
}
sub scan_volids {
my ($cfg, $vmid) = @_;
- my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid);
+ my $info = PVE::Storage::vdisk_list($cfg, undef, $vmid, undef, 'rootdir');
my $all_volumes = {};
foreach my $storeid (keys %$info) {
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);
return [];
}
+my sub print_ct_stderr_log {
+ my ($vmid) = @_;
+ my $log = eval { file_get_contents("/run/pve/ct-$vmid.stderr") };
+ return if !$log;
+
+ while ($log =~ /^\h*(lxc-start:?\s+$vmid:?\s*\S+\s*)?(.*?)\h*$/gm) {
+ my $line = $2;
+ print STDERR "$line\n";
+ }
+}
+
+my sub monitor_state_change($$) {
+ my ($monitor_socket, $vmid) = @_;
+ die "no monitor socket\n" if !defined($monitor_socket);
+
+ while (1) {
+ my ($type, $name, $value) = PVE::LXC::Monitor::read_lxc_message($monitor_socket);
+
+ die "monitor socket: got EOF\n" if !defined($type);
+
+ next if $name ne "$vmid" || $type ne 'STATE';
+
+ if ($value eq PVE::LXC::Monitor::STATE_STARTING) {
+ alarm(0); # don't timeout after seeing the starting state
+ } elsif ($value eq PVE::LXC::Monitor::STATE_ABORTING ||
+ $value eq PVE::LXC::Monitor::STATE_STOPPING ||
+ $value eq PVE::LXC::Monitor::STATE_STOPPED) {
+ return 0;
+ } elsif ($value eq PVE::LXC::Monitor::STATE_RUNNING) {
+ return 1;
+ } else {
+ warn "unexpected message from monitor socket - " .
+ "type: '$type' - value: '$value'\n";
+ }
+ }
+}
+my sub monitor_start($$) {
+ my ($monitor_socket, $vmid) = @_;
+
+ my $success = eval {
+ PVE::Tools::run_with_timeout(10, \&monitor_state_change, $monitor_socket, $vmid)
+ };
+ if (my $err = $@) {
+ warn "problem with monitor socket, but continuing anyway: $err\n";
+ } elsif (!$success) {
+ print_ct_stderr_log($vmid);
+ die "startup for container '$vmid' failed\n";
+ }
+}
+
sub vm_start {
- my ($vmid, $conf, $skiplock) = @_;
+ my ($vmid, $conf, $skiplock, $debug) = @_;
# apply pending changes while starting
if (scalar(keys %{$conf->{pending}})) {
my $storecfg = PVE::Storage::config();
PVE::LXC::Config->vmconfig_apply_pending($vmid, $conf, $storecfg);
+ PVE::LXC::Config->write_config($vmid, $conf);
$conf = PVE::LXC::Config->load_config($vmid); # update/reload
}
close($fh);
}
- my $cmd = ['systemctl', 'start', "pve-container\@$vmid"];
+ my $storage_cfg = PVE::Storage::config();
+ my $vollist = PVE::LXC::Config->get_vm_volumes($conf);
+
+ PVE::Storage::activate_volumes($storage_cfg, $vollist);
+
+ my $monitor_socket = eval { PVE::LXC::Monitor::get_monitor_socket() };
+ warn $@ if $@;
+
+ unlink "/run/pve/ct-$vmid.stderr"; # systemd does not truncate log files
+
+ my $is_debug = $debug || (!defined($debug) && $conf->{debug});
+ my $base_unit = $is_debug ? 'pve-container-debug' : 'pve-container';
+
+ my $cmd = ['systemctl', 'start', "$base_unit\@$vmid"];
PVE::GuestHelpers::exec_hookscript($conf, $vmid, 'pre-start', 1);
- eval { PVE::Tools::run_command($cmd); };
+ eval {
+ run_command($cmd);
+
+ monitor_start($monitor_socket, $vmid) if defined($monitor_socket);
+
+ # if debug is requested, print the log it also when the start succeeded
+ print_ct_stderr_log($vmid) if $is_debug;
+ };
if (my $err = $@) {
unlink $skiplock_flag_fn;
die $err;
}
}
- eval { PVE::Tools::run_command($cmd, timeout => $shutdown_timeout) };
+ eval { run_command($cmd, timeout => $shutdown_timeout) };
if (my $err = $@) {
warn $@ if $@;
}
# Unshare the mount namespace
die "failed to unshare mount namespace: $!\n"
if !PVE::Tools::unshare(PVE::Tools::CLONE_NEWNS);
- PVE::Tools::run_command(['mount', '--make-rslave', '/']);
+ run_command(['mount', '--make-rslave', '/']);
return $code->();
});
}
$bwlimit //= 0;
- PVE::Tools::run_command(['/usr/bin/rsync', '--stats', '-X', '-A', '--numeric-ids',
- '-aH', '--whole-file', '--sparse', '--one-file-system',
- "--bwlimit=$bwlimit", "$src/", $dest]);
+ run_command([
+ 'rsync',
+ '--stats',
+ '-X',
+ '-A',
+ '--numeric-ids',
+ '-aH',
+ '--whole-file',
+ '--sparse',
+ '--one-file-system',
+ "--bwlimit=$bwlimit",
+ "$src/",
+ $dest
+ ]);
};
my $err = $@;
while ((system {"fuser"} "fuser", "-s", $dest) == 0) {sleep 1};
foreach my $mount (reverse @mounted) {
- eval { PVE::Tools::run_command(['/bin/umount', $mount], errfunc => sub{})};
+ eval { run_command(['/bin/umount', $mount], errfunc => sub{})};
warn "Can't umount $mount\n" if $@;
}
return $new_volid;
}
+sub get_lxc_version() {
+ my $version;
+ run_command([qw(lxc-start --version)], outfunc => sub {
+ my ($line) = @_;
+ # We only parse out major & minor version numbers.
+ if ($line =~ /^(\d+)\.(\d+)(?:\D.*)?$/) {
+ $version = [$1, $2];
+ }
+ });
+
+ die "failed to get lxc version\n" if !defined($version);
+
+ # return as a list:
+ return $version->@*;
+}
+
+sub freeze($) {
+ my ($vmid) = @_;
+ if (PVE::CGroup::cgroup_mode() == 2) {
+ PVE::LXC::Command::freeze($vmid, 30);
+ } else {
+ PVE::LXC::CGroup->new($vmid)->freeze_thaw(1);
+ }
+}
+
+sub thaw($) {
+ my ($vmid) = @_;
+ if (PVE::CGroup::cgroup_mode() == 2) {
+ PVE::LXC::Command::unfreeze($vmid, 30);
+ } else {
+ PVE::LXC::CGroup->new($vmid)->freeze_thaw(0);
+ }
+}
+
1;
projects / pve-container.git / blobdiff