use strict;
use warnings;
+
use IO::File;
+use Net::IP;
use POSIX;
-use PVE::Tools qw(run_command);
-use PVE::Storage::Plugin;
+
+use PVE::ProcFSTools;
use PVE::RPCEnvironment;
+use PVE::Storage::Plugin;
+use PVE::Tools qw(run_command);
use base qw(PVE::Storage::Plugin);
description => "use sparse volumes",
type => 'boolean',
},
+ mountpoint => {
+ description => "mount point",
+ type => 'string', format => 'pve-storage-path',
+ },
};
}
nodes => { optional => 1 },
disable => { optional => 1 },
content => { optional => 1 },
+ bwlimit => { optional => 1 },
+ mountpoint => { optional => 1 },
};
}
# static zfs helper methods
-sub zfs_parse_size {
- my ($text) = @_;
-
- return 0 if !$text;
-
- if ($text =~ m/^(\d+(\.\d+)?)([TGMK])?$/) {
-
- my ($size, $reminder, $unit) = ($1, $2, $3);
-
- if ($unit) {
- if ($unit eq 'K') {
- $size *= 1024;
- } elsif ($unit eq 'M') {
- $size *= 1024*1024;
- } elsif ($unit eq 'G') {
- $size *= 1024*1024*1024;
- } elsif ($unit eq 'T') {
- $size *= 1024*1024*1024*1024;
- } else {
- die "got unknown zfs size unit '$unit'\n";
- }
- }
-
- if ($reminder) {
- $size = ceil($size);
- }
-
- return $size;
-
- }
-
- warn "unable to parse zfs size '$text'\n";
-
- return 0;
-}
-
sub zfs_parse_zvol_list {
my ($text) = @_;
if ($refquota eq 'none') {
$zvol->{size} = 0;
} else {
- $zvol->{size} = zfs_parse_size($refquota);
+ $zvol->{size} = $refquota + 0;
}
$zvol->{format} = 'subvol';
} else {
- $zvol->{size} = zfs_parse_size($size);
+ $zvol->{size} = $size + 0;
$zvol->{format} = 'raw';
}
if ($origin !~ /^-$/) {
# virtual zfs methods (subclass can overwrite them)
+sub on_add_hook {
+ my ($class, $storeid, $scfg, %param) = @_;
+
+ my $cfg_mountpoint = $scfg->{mountpoint};
+
+ # ignore failure, pool might currently not be imported
+ my $mountpoint;
+ eval {
+ my $res = $class->zfs_get_properties($scfg, 'mountpoint', $scfg->{pool}, 1);
+ $mountpoint = PVE::Storage::Plugin::verify_path($res, 1) if defined($res);
+ };
+
+ if (defined($cfg_mountpoint)) {
+ if (defined($mountpoint) && !($cfg_mountpoint =~ m|^\Q$mountpoint\E/?$|)) {
+ warn "warning for $storeid - mountpoint: $cfg_mountpoint " .
+ "does not match current mount point: $mountpoint\n";
+ }
+ } else {
+ $scfg->{mountpoint} = $mountpoint;
+ }
+
+ return;
+}
+
sub path {
my ($class, $scfg, $volname, $storeid, $snapname) = @_;
my ($vtype, $name, $vmid) = $class->parse_volname($volname);
my $path = '';
+ my $mountpoint = $scfg->{mountpoint} // "/$scfg->{pool}";
if ($vtype eq "images") {
if ($name =~ m/^subvol-/ || $name =~ m/^basevol-/) {
- # fixme: we currently assume standard mount point?!
- $path = "/$scfg->{pool}/$name";
+ $path = "$mountpoint/$name";
} else {
$path = "/dev/zvol/$scfg->{pool}/$name";
}
sub zfs_request {
my ($class, $scfg, $timeout, $method, @params) = @_;
- $timeout = PVE::RPCEnvironment::is_worker() ? 60*60 : 5
- if !$timeout;
-
my $cmd = [];
if ($method eq 'zpool_list') {
push @$cmd, 'zpool', 'list';
+ } elsif ($method eq 'zpool_import') {
+ push @$cmd, 'zpool', 'import';
+ $timeout = 15 if !$timeout || $timeout < 15;
} else {
push @$cmd, 'zfs', $method;
}
-
push @$cmd, @params;
-
+
my $msg = '';
+ my $output = sub { $msg .= "$_[0]\n" };
- my $output = sub {
- my $line = shift;
- $msg .= "$line\n";
- };
+ $timeout = PVE::RPCEnvironment->is_worker() ? 60*60 : 5 if !$timeout;
run_command($cmd, errmsg => "zfs error", outfunc => $output, timeout => $timeout);
return $msg;
}
+sub zfs_wait_for_zvol_link {
+ my ($class, $scfg, $volname, $timeout) = @_;
+
+ my $default_timeout = PVE::RPCEnvironment->is_worker() ? 60*5 : 10;
+ $timeout = $default_timeout if !defined($timeout);
+
+ my ($devname, undef, undef) = $class->path($scfg, $volname);
+
+ for (my $i = 1; $i <= $timeout; $i++) {
+ last if -b $devname;
+ die "timeout: no zvol device link for '$volname' found after $timeout sec found.\n"
+ if $i == $timeout;
+
+ sleep(1);
+ }
+}
+
sub alloc_image {
my ($class, $storeid, $scfg, $vmid, $fmt, $name, $size) = @_;
my $volname = $name;
-
+
if ($fmt eq 'raw') {
- die "illegal name '$volname' - sould be 'vm-$vmid-*'\n"
+ die "illegal name '$volname' - should be 'vm-$vmid-*'\n"
if $volname && $volname !~ m/^vm-$vmid-/;
- $volname = $class->zfs_find_free_diskname($storeid, $scfg, $vmid, $fmt)
+ $volname = $class->find_free_diskname($storeid, $scfg, $vmid, $fmt)
if !$volname;
$class->zfs_create_zvol($scfg, $volname, $size);
- my $devname = "/dev/zvol/$scfg->{pool}/$volname";
-
- run_command("udevadm trigger --subsystem-match block");
- system('udevadm', 'settle', '--timeout', '10', "--exit-if-exists=${devname}");
+ $class->zfs_wait_for_zvol_link($scfg, $volname);
} elsif ( $fmt eq 'subvol') {
- die "illegal name '$volname' - sould be 'subvol-$vmid-*'\n"
+ die "illegal name '$volname' - should be 'subvol-$vmid-*'\n"
if $volname && $volname !~ m/^subvol-$vmid-/;
- $volname = $class->zfs_find_free_diskname($storeid, $scfg, $vmid, $fmt)
+ $volname = $class->find_free_diskname($storeid, $scfg, $vmid, $fmt)
if !$volname;
- die "illegal name '$volname' - sould be 'subvol-$vmid-*'\n"
+ die "illegal name '$volname' - should be 'subvol-$vmid-*'\n"
if $volname !~ m/^subvol-$vmid-/;
- $class->zfs_create_subvol($scfg, $volname, $size);
-
+ $class->zfs_create_subvol($scfg, $volname, $size);
+
} else {
die "unsupported format '$fmt'";
}
foreach my $image (keys %$dat) {
- my $volname = $dat->{$image}->{name};
- my $parent = $dat->{$image}->{parent};
+ my $info = $dat->{$image};
- my $volid = undef;
- if ($parent && $parent =~ m/^(\S+)@(\S+)$/) {
+ my $volname = $info->{name};
+ my $parent = $info->{parent};
+ my $owner = $info->{vmid};
+
+ if ($parent && $parent =~ m/^(\S+)\@__base__$/) {
my ($basename) = ($1);
- $volid = "$storeid:$basename/$volname";
+ $info->{volid} = "$storeid:$basename/$volname";
} else {
- $volid = "$storeid:$volname";
+ $info->{volid} = "$storeid:$volname";
}
- my $owner = $dat->{$volname}->{vmid};
if ($vollist) {
- my $found = grep { $_ eq $volid } @$vollist;
+ my $found = grep { $_ eq $info->{volid} } @$vollist;
next if !$found;
} else {
next if defined ($vmid) && ($owner ne $vmid);
}
- my $info = $dat->{$volname};
- $info->{volid} = $volid;
push @$res, $info;
}
}
return $res;
}
+sub zfs_get_properties {
+ my ($class, $scfg, $properties, $dataset, $timeout) = @_;
+
+ my $result = $class->zfs_request($scfg, $timeout, 'get', '-o', 'value',
+ '-Hp', $properties, $dataset);
+ my @values = split /\n/, $result;
+ return wantarray ? @values : $values[0];
+}
+
sub zfs_get_pool_stats {
my ($class, $scfg) = @_;
my $available = 0;
my $used = 0;
- my $text = $class->zfs_request($scfg, undef, 'get', '-o', 'value', '-Hp',
- 'available,used', $scfg->{pool});
-
- my @lines = split /\n/, $text;
+ my @lines = $class->zfs_get_properties($scfg, 'available,used', $scfg->{pool});
if($lines[0] =~ /^(\d+)$/) {
$available = $1;
sub zfs_create_zvol {
my ($class, $scfg, $zvol, $size) = @_;
-
+
+ # always align size to 1M as workaround until
+ # https://github.com/zfsonlinux/zfs/issues/8541 is solved
+ my $padding = (1024 - $size % 1024) % 1024;
+ $size = $size + $padding;
+
my $cmd = ['create'];
push @$cmd, '-s' if $scfg->{sparse};
my ($class, $scfg, $volname, $size) = @_;
my $dataset = "$scfg->{pool}/$volname";
-
+ my $quota = $size ? "${size}k" : "none";
+
my $cmd = ['create', '-o', 'acltype=posixacl', '-o', 'xattr=sa',
- '-o', "refquota=${size}k", $dataset];
+ '-o', "refquota=${quota}", $dataset];
$class->zfs_request($scfg, undef, @$cmd);
}
sub zfs_list_zvol {
my ($class, $scfg) = @_;
- my $text = $class->zfs_request($scfg, 10, 'list', '-o', 'name,volsize,origin,type,refquota', '-t', 'volume,filesystem', '-Hr');
+ my $text = $class->zfs_request($scfg, 10, 'list', '-o', 'name,volsize,origin,type,refquota', '-t', 'volume,filesystem', '-Hrp');
my $zvols = zfs_parse_zvol_list($text);
return undef if !$zvols;
return $list;
}
-sub zfs_find_free_diskname {
- my ($class, $storeid, $scfg, $vmid, $format) = @_;
-
- my $name = undef;
- my $volumes = $class->zfs_list_zvol($scfg);
-
- my $disk_ids = {};
- my $dat = $volumes->{$scfg->{pool}};
-
- foreach my $image (keys %$dat) {
- my $volname = $dat->{$image}->{name};
- if ($volname =~ m/(vm|base|subvol|basevol)-$vmid-disk-(\d+)/){
- $disk_ids->{$2} = 1;
- }
- }
-
- for (my $i = 1; $i < 100; $i++) {
- if (!$disk_ids->{$i}) {
- return $format eq 'subvol' ? "subvol-$vmid-disk-$i" : "vm-$vmid-disk-$i";
- }
- }
-
- die "unable to allocate an image name for VM $vmid in storage '$storeid'\n";
-}
-
sub zfs_get_latest_snapshot {
my ($class, $scfg, $volname) = @_;
$class->parse_volname($volname);
my $attr = $format eq 'subvol' ? 'refquota' : 'volsize';
- my $text = $class->zfs_request($scfg, undef, 'get', '-Hp', $attr, "$scfg->{pool}/$vname");
- if ($text =~ /\s$attr\s(\d+)\s/) {
+ my $value = $class->zfs_get_properties($scfg, $attr, "$scfg->{pool}/$vname");
+ if ($value =~ /^(\d+)$/) {
return $1;
}
sub volume_snapshot_rollback {
my ($class, $scfg, $storeid, $volname, $snap) = @_;
- my $vname = ($class->parse_volname($volname))[1];
+ my (undef, $vname, undef, undef, undef, undef, $format) = $class->parse_volname($volname);
+
+ my $msg = $class->zfs_request($scfg, undef, 'rollback', "$scfg->{pool}/$vname\@$snap");
+
+ # we have to unmount rollbacked subvols, to invalidate wrong kernel
+ # caches, they get mounted in activate volume again
+ # see zfs bug #10931 https://github.com/openzfs/zfs/issues/10931
+ if ($format eq 'subvol') {
+ eval { $class->zfs_request($scfg, undef, 'unmount', "$scfg->{pool}/$vname"); };
+ if (my $err = $@) {
+ die $err if $err !~ m/not currently mounted$/;
+ }
+ }
- $class->zfs_request($scfg, undef, 'rollback', "$scfg->{pool}/$vname\@$snap");
+ return $msg;
}
sub volume_rollback_is_possible {
- my ($class, $scfg, $storeid, $volname, $snap) = @_;
-
+ my ($class, $scfg, $storeid, $volname, $snap) = @_;
+
my $recentsnap = $class->zfs_get_latest_snapshot($scfg, $volname);
- if ($snap ne $recentsnap) {
- die "can't rollback, more recent snapshots exist\n";
- }
- return 1;
+ die "can't rollback, no snapshots exist at all\n"
+ if !defined($recentsnap);
+
+ die "can't rollback, '$snap' is not most recent snapshot\n"
+ if $snap ne $recentsnap;
+
+ return 1;
+}
+
+sub volume_snapshot_list {
+ my ($class, $scfg, $storeid, $volname) = @_;
+
+ my ($vtype, $name, $vmid) = $class->parse_volname($volname);
+
+ my $zpath = "$scfg->{pool}/$name";
+
+ my $snaps = [];
+
+ my $cmd = ['zfs', 'list', '-r', '-H', '-S', 'name', '-t', 'snap', '-o',
+ 'name', $zpath];
+
+ my $outfunc = sub {
+ my $line = shift;
+
+ if ($line =~ m/^\Q$zpath\E@(.*)$/) {
+ push @$snaps, $1;
+ }
+ };
+
+ eval { run_command( [$cmd], outfunc => $outfunc , errfunc => sub{}); };
+
+ # return an empty array if dataset does not exist.
+ return $snaps;
+}
+
+my sub dataset_mounted_heuristic {
+ my ($dataset) = @_;
+
+ my $mounts = PVE::ProcFSTools::parse_proc_mounts();
+ for my $mp (@$mounts) {
+ my ($what, $dir, $fs) = $mp->@*;
+ next if $fs ne 'zfs';
+ # check for root-dataset or any child-dataset (root-dataset could have 'canmount=off')
+ # If any child is mounted heuristically assume that `zfs mount -a` was successful
+ next if $what !~ m!^$dataset(?:/|$)!;
+ return 1;
+ }
+ return 0;
}
sub activate_storage {
my ($class, $storeid, $scfg, $cache) = @_;
- my @param = ('-o', 'name', '-H');
+ # Note: $scfg->{pool} can include dataset <pool>/<dataset>
+ my $dataset = $scfg->{pool};
+ my $pool = ($dataset =~ s!/.*$!!r);
- my $text = $class->zfs_request($scfg, undef, 'zpool_list', @param);
+ return 1 if dataset_mounted_heuristic($dataset); # early return
- # Note: $scfg->{pool} can include dataset <pool>/<dataset>
- my $pool = $scfg->{pool};
- $pool =~ s!/.*$!!;
+ my $pool_imported = sub {
+ my @param = ('-o', 'name', '-H', $pool);
+ my $res = eval { $class->zfs_request($scfg, undef, 'zpool_list', @param) };
+ warn "$@\n" if $@;
- if ($text !~ $pool) {
- run_command("zpool import -d /dev/disk/by-id/ -a");
+ return defined($res) && $res =~ m/$pool/;
+ };
+
+ if (!$pool_imported->()) {
+ # import can only be done if not yet imported!
+ my @param = ('-d', '/dev/disk/by-id/', '-o', 'cachefile=none', $pool);
+ eval { $class->zfs_request($scfg, undef, 'zpool_import', @param) };
+ if (my $err = $@) {
+ # just could've raced with another import, so recheck if it is imported
+ die "could not activate storage '$storeid', $err\n" if !$pool_imported->();
+ }
}
+ eval { $class->zfs_request($scfg, undef, 'mount', '-a') };
+ die "could not activate storage '$storeid', $@\n" if $@;
return 1;
}
sub activate_volume {
my ($class, $storeid, $scfg, $volname, $snapname, $cache) = @_;
+
+ return 1 if defined($snapname);
+
+ my (undef, $dataset, undef, undef, undef, undef, $format) = $class->parse_volname($volname);
+
+ if ($format eq 'raw') {
+ $class->zfs_wait_for_zvol_link($scfg, $volname);
+ } elsif ($format eq 'subvol') {
+ my $mounted = $class->zfs_get_properties($scfg, 'mounted', "$scfg->{pool}/$dataset");
+ if ($mounted !~ m/^yes$/) {
+ $class->zfs_request($scfg, undef, 'mount', "$scfg->{pool}/$dataset");
+ }
+ }
+
return 1;
}
die "clone_image only works on base images\n" if !$isBase;
- my $name = $class->zfs_find_free_diskname($storeid, $scfg, $vmid, $format);
+ my $name = $class->find_free_diskname($storeid, $scfg, $vmid, $format);
if ($format eq 'subvol') {
- my $size = $class->zfs_request($scfg, undef, 'list', '-H', '-o', 'refquota', "$scfg->{pool}/$basename");
+ my $size = $class->zfs_request($scfg, undef, 'list', '-Hp', '-o', 'refquota', "$scfg->{pool}/$basename");
chomp($size);
$class->zfs_request($scfg, undef, 'clone', "$scfg->{pool}/$basename\@$snap", "$scfg->{pool}/$name", '-o', "refquota=$size");
} else {
my $attr = $format eq 'subvol' ? 'refquota' : 'volsize';
+ # align size to 1M so we always have a valid multiple of the volume block size
+ if ($format eq 'raw') {
+ my $padding = (1024 - $new_size % 1024) % 1024;
+ $new_size = $new_size + $padding;
+ }
+
$class->zfs_request($scfg, undef, 'set', "$attr=${new_size}k", "$scfg->{pool}/$vname");
return $new_size;
}
+sub storage_can_replicate {
+ my ($class, $scfg, $storeid, $format) = @_;
+
+ return 1 if $format eq 'raw' || $format eq 'subvol';
+
+ return 0;
+}
+
sub volume_has_feature {
my ($class, $scfg, $feature, $storeid, $volname, $snapname, $running) = @_;
template => { current => 1},
copy => { base => 1, current => 1},
sparseinit => { base => 1, current => 1},
+ replicate => { base => 1, current => 1},
};
my ($vtype, $name, $vmid, $basename, $basevmid, $isBase) =
return undef;
}
+sub volume_export {
+ my ($class, $scfg, $storeid, $fh, $volname, $format, $snapshot, $base_snapshot, $with_snapshots) = @_;
+
+ die "unsupported export stream format for $class: $format\n"
+ if $format ne 'zfs';
+
+ die "$class storage can only export snapshots\n"
+ if !defined($snapshot);
+
+ my $dataset = ($class->parse_volname($volname))[1];
+
+ my $fd = fileno($fh);
+ die "internal error: invalid file handle for volume_export\n"
+ if !defined($fd);
+ $fd = ">&$fd";
+
+ # For zfs we always create a replication stream (-R) which means the remote
+ # side will always delete non-existing source snapshots. This should work
+ # for all our use cases.
+ my $cmd = ['zfs', 'send', '-Rpv'];
+ if (defined($base_snapshot)) {
+ my $arg = $with_snapshots ? '-I' : '-i';
+ push @$cmd, $arg, $base_snapshot;
+ }
+ push @$cmd, '--', "$scfg->{pool}/$dataset\@$snapshot";
+
+ run_command($cmd, output => $fd);
+
+ return;
+}
+
+sub volume_export_formats {
+ my ($class, $scfg, $storeid, $volname, $snapshot, $base_snapshot, $with_snapshots) = @_;
+
+ my @formats = ('zfs');
+ # TODOs:
+ # push @formats, 'fies' if $volname !~ /^(?:basevol|subvol)-/;
+ # push @formats, 'raw' if !$base_snapshot && !$with_snapshots;
+ return @formats;
+}
+
+sub volume_import {
+ my ($class, $scfg, $storeid, $fh, $volname, $format, $snapshot, $base_snapshot, $with_snapshots, $allow_rename) = @_;
+
+ die "unsupported import stream format for $class: $format\n"
+ if $format ne 'zfs';
+
+ my $fd = fileno($fh);
+ die "internal error: invalid file handle for volume_import\n"
+ if !defined($fd);
+
+ my (undef, $dataset, $vmid) = $class->parse_volname($volname);
+ my $zfspath = "$scfg->{pool}/$dataset";
+ my $suffix = defined($base_snapshot) ? "\@$base_snapshot" : '';
+ my $exists = 0 == run_command(['zfs', 'get', '-H', 'name', $zfspath.$suffix],
+ noerr => 1, quiet => 1);
+ if (defined($base_snapshot)) {
+ die "base snapshot '$zfspath\@$base_snapshot' doesn't exist\n" if !$exists;
+ } elsif ($exists) {
+ die "volume '$zfspath' already exists\n" if !$allow_rename;
+ warn "volume '$zfspath' already exists - importing with a different name\n";
+ $dataset = $class->find_free_diskname($storeid, $scfg, $vmid, $format);
+ $zfspath = "$scfg->{pool}/$dataset";
+ }
+
+ eval { run_command(['zfs', 'recv', '-F', '--', $zfspath], input => "<&$fd") };
+ if (my $err = $@) {
+ if (defined($base_snapshot)) {
+ eval { run_command(['zfs', 'rollback', '-r', '--', "$zfspath\@$base_snapshot"]) };
+ } else {
+ eval { run_command(['zfs', 'destroy', '-r', '--', $zfspath]) };
+ }
+ die $err;
+ }
+
+ return "$storeid:$dataset";
+}
+
+sub volume_import_formats {
+ my ($class, $scfg, $storeid, $volname, $snapshot, $base_snapshot, $with_snapshots) = @_;
+
+ return $class->volume_export_formats($scfg, $storeid, $volname, $snapshot, $base_snapshot, $with_snapshots);
+}
+
1;