X-Git-Url: https://git.proxmox.com/?a=blobdiff_plain;f=PVE%2FDiskmanage.pm;h=18459f95291a5f70af8c051d49208cb7705fee71;hb=bc7fecb0821db8c4171e89f5d8073eb842b5213c;hp=612bc3b6be8e9b899cb338d4515690a809502540;hpb=eebcdb11194d9fadeefafaa11b16d02fc6ea3ca7;p=pve-storage.git diff --git a/PVE/Diskmanage.pm b/PVE/Diskmanage.pm index 612bc3b..18459f9 100644 --- a/PVE/Diskmanage.pm +++ b/PVE/Diskmanage.pm @@ -2,10 +2,14 @@ package PVE::Diskmanage; use strict; use warnings; + use PVE::ProcFSTools; use Data::Dumper; use Cwd qw(abs_path); use Fcntl ':mode'; +use File::Basename; +use File::stat; +use JSON; use PVE::Tools qw(extract_param run_command file_get_contents file_read_firstline dir_glob_regex dir_glob_foreach trim); @@ -13,7 +17,19 @@ my $SMARTCTL = "/usr/sbin/smartctl"; my $ZPOOL = "/sbin/zpool"; my $SGDISK = "/sbin/sgdisk"; my $PVS = "/sbin/pvs"; -my $UDEVADM = "/bin/udevadm"; +my $LVS = "/sbin/lvs"; +my $LSBLK = "/bin/lsblk"; + +my sub strip_dev :prototype($) { + my ($devpath) = @_; + $devpath =~ s|^/dev/||; + return $devpath; +} + +sub check_bin { + my ($path) = @_; + return -x $path; +} sub verify_blockdev_path { my ($rel_path) = @_; @@ -47,8 +63,8 @@ sub init_disk { assert_blockdev($disk); - # we should already have checked if it is in use in the api call - # but we check again for safety + # we should already have checked these in the api call, but we check again for safety + die "$disk is a partition\n" if is_partition($disk); die "disk $disk is already in use\n" if disk_is_used($disk); my $id = $uuid || 'R'; @@ -62,7 +78,7 @@ sub disk_is_used { my $dev = $disk; $dev =~ s|^/dev/||; - my $disklist = get_disks($dev, 1); + my $disklist = get_disks($dev, 1, 1); die "'$disk' is not a valid local disk\n" if !defined($disklist->{$dev}); return 1 if $disklist->{$dev}->{used}; @@ -104,7 +120,6 @@ sub get_smart_data { if (defined($type) && $type eq 'ata' && $line =~ m/^([ \d]{2}\d)\s+(\S+)\s+(\S{6})\s+(\d+)\s+(\d+)\s+(\S+)\s+(\S+)\s+(.*)$/) { my $entry = {}; - $entry->{name} = $2 if defined $2; $entry->{flags} = $3 if defined $3; # the +0 makes a number out of the strings @@ -130,6 +145,10 @@ sub get_smart_data { } elsif (defined($type) && $type eq 'text') { $smartdata->{text} = '' if !defined $smartdata->{text}; $smartdata->{text} .= "$line\n"; + # extract wearout from nvme/sas text, allow for decimal values + if ($line =~ m/Percentage Used(?: endurance indicator)?:\s*(\d+(?:\.\d+)?)\%/i) { + $smartdata->{wearout} = 100 - $1; + } } elsif ($line =~ m/SMART Disabled/) { $smartdata->{health} = "SMART Disabled"; } @@ -149,8 +168,52 @@ sub get_smart_data { return $smartdata; } +sub get_lsblk_info() { + my $cmd = [$LSBLK, '--json', '-o', 'path,parttype,fstype']; + my $output = ""; + my $res = {}; + eval { + run_command($cmd, outfunc => sub { + my ($line) = @_; + $output .= "$line\n"; + }); + }; + warn "$@\n" if $@; + return $res if $output eq ''; + + my $parsed = eval { decode_json($output) }; + warn "$@\n" if $@; + my $list = $parsed->{blockdevices} // []; + + $res = { map { + $_->{path} => { + parttype => $_->{parttype}, + fstype => $_->{fstype} + } + } @{$list} }; + + return $res; +} + +my $get_devices_by_partuuid = sub { + my ($lsblk_info, $uuids, $res) = @_; + + $res = {} if !defined($res); + + foreach my $dev (sort keys %{$lsblk_info}) { + my $uuid = $lsblk_info->{$dev}->{parttype}; + next if !defined($uuid) || !defined($uuids->{$uuid}); + $res->{$dev} = $uuids->{$uuid}; + } + + return $res; +}; + sub get_zfs_devices { - my $list = {}; + my ($lsblk_info) = @_; + my $res = {}; + + return {} if !check_bin($ZPOOL); # use zpool and parttype uuid, # because log and cache do not have @@ -160,7 +223,7 @@ sub get_zfs_devices { my ($line) = @_; if ($line =~ m|^\t([^\t]+)\t|) { - $list->{$1} = 1; + $res->{$1} = 1; } }); }; @@ -169,68 +232,91 @@ sub get_zfs_devices { # because maybe zfs tools are not installed warn "$@\n" if $@; - my $applezfsuuid = "6a898cc3-1dd2-11b2-99a6-080020736631"; - my $bsdzfsuuid = "516e7cba-6ecf-11d6-8ff8-00022d09712b"; + my $uuids = { + "6a898cc3-1dd2-11b2-99a6-080020736631" => 1, # apple + "516e7cba-6ecf-11d6-8ff8-00022d09712b" => 1, # bsd + }; - dir_glob_foreach('/dev/disk/by-parttypeuuid', "($applezfsuuid|$bsdzfsuuid)\..+", sub { - my ($entry) = @_; - my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry"); - $list->{$real_dev} = 1; - }); - return $list; + $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); + + return $res; } sub get_lvm_devices { - my $list = {}; + my ($lsblk_info) = @_; + my $res = {}; eval { run_command([$PVS, '--noheadings', '--readonly', '-o', 'pv_name'], outfunc => sub{ my ($line) = @_; $line = trim($line); if ($line =~ m|^/dev/|) { - $list->{$line} = 1; + $res->{$line} = 1; } }); }; # if something goes wrong, we do not want - # to give up, but indicate an error has occured + # to give up, but indicate an error has occurred warn "$@\n" if $@; - my $lvmuuid = "e6d6d379-f507-44c2-a23c-238f2a3df928"; + my $uuids = { + "e6d6d379-f507-44c2-a23c-238f2a3df928" => 1, + }; - dir_glob_foreach('/dev/disk/by-parttypeuuid', "$lvmuuid\..+", sub { - my ($entry) = @_; - my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry"); - $list->{$real_dev} = 1; - }); + $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); - return $list; + return $res; } sub get_ceph_journals { - my $journalhash = {}; - - my $journal_uuid = '45b0969e-9b03-4f30-b4c6-b4b80ceff106'; - my $db_uuid = '30cd0809-c2b2-499c-8879-2d6b78529876'; - my $wal_uuid = '5ce17fce-4087-4169-b7ff-056cc58473f9'; - my $block_uuid = 'cafecafe-9b03-4f30-b4c6-b4b80ceff106'; - - dir_glob_foreach('/dev/disk/by-parttypeuuid', "($journal_uuid|$db_uuid|$wal_uuid|$block_uuid)\..+", sub { - my ($entry, $type) = @_; - my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry"); - if ($type eq $journal_uuid) { - $journalhash->{$real_dev} = 1; - } elsif ($type eq $db_uuid) { - $journalhash->{$real_dev} = 2; - } elsif ($type eq $wal_uuid) { - $journalhash->{$real_dev} = 3; - } elsif ($type eq $block_uuid) { - $journalhash->{$real_dev} = 4; + my ($lsblk_info) = @_; + my $res = {}; + + my $uuids = { + '45b0969e-9b03-4f30-b4c6-b4b80ceff106' => 1, # journal + '30cd0809-c2b2-499c-8879-2d6b78529876' => 2, # db + '5ce17fce-4087-4169-b7ff-056cc58473f9' => 3, # wal + 'cafecafe-9b03-4f30-b4c6-b4b80ceff106' => 4, # block + }; + + $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); + + return $res; +} + +# reads the lv_tags and matches them with the devices +sub get_ceph_volume_infos { + my $result = {}; + + my $cmd = [ $LVS, '-S', 'lv_name=~^osd-', '-o', 'devices,lv_name,lv_tags', + '--noheadings', '--readonly', '--separator', ';' ]; + + run_command($cmd, outfunc => sub { + my $line = shift; + $line =~ s/(?:^\s+)|(?:\s+$)//g; # trim whitespaces + + my $fields = [ split(';', $line) ]; + + # lvs syntax is /dev/sdX(Y) where Y is the start (which we do not need) + my ($dev) = $fields->[0] =~ m|^(/dev/[a-z]+[^(]*)|; + if ($fields->[1] =~ m|^osd-([^-]+)-|) { + my $type = $1; + # $result autovivification is wanted, to not creating empty hashes + if (($type eq 'block' || $type eq 'data') && $fields->[2] =~ m/ceph.osd_id=([^,]+)/) { + $result->{$dev}->{osdid} = $1; + $result->{$dev}->{bluestore} = ($type eq 'block'); + if ($fields->[2] =~ m/ceph\.encrypted=1/) { + $result->{$dev}->{encrypted} = 1; + } + } else { + # undef++ becomes '1' (see `perldoc perlop`: Auto-increment) + $result->{$dev}->{$type}++; + } } }); - return $journalhash; + return $result; } sub get_udev_info { @@ -239,7 +325,7 @@ sub get_udev_info { my $info = ""; my $data = {}; eval { - run_command([$UDEVADM, 'info', '-p', $dev, '--query', 'all'], outfunc => sub { + run_command(['udevadm', 'info', '-p', $dev, '--query', 'all'], outfunc => sub { my ($line) = @_; $info .= "$line\n"; }); @@ -247,7 +333,7 @@ sub get_udev_info { warn $@ if $@; return undef if !$info; - return undef if $info !~ m/^E: DEVTYPE=disk$/m; + return undef if $info !~ m/^E: DEVTYPE=(disk|partition)$/m; return undef if $info =~ m/^E: ID_CDROM/m; # we use this, because some disks are not simply in /dev @@ -286,9 +372,25 @@ sub get_udev_info { $data->{wwn} = $1; } + if ($info =~ m/^E: DEVLINKS=(.+)$/m) { + my @devlinks = grep(m#^/dev/disk/by-id/(ata|scsi|nvme(?!-eui))#, split (/ /, $1)); + $data->{by_id_link} = $devlinks[0] if defined($devlinks[0]); + } + return $data; } +sub get_sysdir_size { + my ($sysdir) = @_; + + my $size = file_read_firstline("$sysdir/size"); + return if !$size; + + # linux always considers sectors to be 512 bytes, + # independently of real block size + return $size * 512; +} + sub get_sysdir_info { my ($sysdir) = @_; @@ -296,12 +398,7 @@ sub get_sysdir_info { my $data = {}; - my $size = file_read_firstline("$sysdir/size"); - return undef if !$size; - - # linux always considers sectors to be 512 bytes, - # independently of real block size - $data->{size} = $size * 512; + $data->{size} = get_sysdir_size($sysdir) or return; # dir/queue/rotational should be 1 for hdd, 0 for ssd $data->{rotational} = file_read_firstline("$sysdir/queue/rotational") // -1; @@ -318,39 +415,42 @@ sub get_sysdir_info { } sub get_wear_leveling_info { - my ($attributes, $model) = @_; - - my $wearout; - - my $vendormap = { - 'kingston' => 231, - 'samsung' => 177, - 'intel' => 233, - 'sandisk' => 233, - 'crucial' => 202, - 'default' => 233, - }; - - # find target attr id + my ($smartdata) = @_; + my $attributes = $smartdata->{attributes}; - my $attrid; - - foreach my $vendor (keys %$vendormap) { - if ($model =~ m/$vendor/i) { - $attrid = $vendormap->{$vendor}; - # found the attribute - last; - } + if (defined($smartdata->{wearout})) { + return $smartdata->{wearout}; } - if (!$attrid) { - $attrid = $vendormap->{default}; - } + my $wearout; - foreach my $attr (@$attributes) { - next if $attr->{id} != $attrid; - $wearout = $attr->{value}; - last; + # Common register names that represent percentage values of potential + # failure indicators used in drivedb.h of smartmontool's. Order matters, + # as some drives may have multiple definitions + my @wearoutregisters = ( + "Media_Wearout_Indicator", + "SSD_Life_Left", + "Wear_Leveling_Count", + "Perc_Write\/Erase_Ct_BC", + "Perc_Rated_Life_Remain", + "Remaining_Lifetime_Perc", + "Percent_Lifetime_Remain", + "Lifetime_Left", + "PCT_Life_Remaining", + "Lifetime_Remaining", + "Percent_Life_Remaining", + "Percent_Lifetime_Used", + "Perc_Rated_Life_Used" + ); + + # Search for S.M.A.R.T. attributes for known register + foreach my $register (@wearoutregisters) { + last if defined $wearout; + foreach my $attr (@$attributes) { + next if $attr->{name} !~ m/$register/; + $wearout = $attr->{value}; + last; + } } return $wearout; @@ -381,10 +481,12 @@ sub is_iscsi { return 0; } -sub get_disks { - my ($disk, $nosmart) = @_; - my $disklist = {}; +my sub is_ssdlike { + my ($type) = @_; + return $type eq 'ssd' || $type eq 'nvme'; +} +sub mounted_blockdevs { my $mounted = {}; my $mounts = PVE::ProcFSTools::parse_proc_mounts(); @@ -394,25 +496,47 @@ sub get_disks { $mounted->{abs_path($mount->[0])} = $mount->[1]; }; - my $dev_is_mounted = sub { - my ($dev) = @_; - return $mounted->{$dev}; - }; + return $mounted; +} + +sub get_disks { + my ($disks, $nosmart, $include_partitions) = @_; + my $disklist = {}; + + my $mounted = mounted_blockdevs(); + + my $lsblk_info = get_lsblk_info(); - my $journalhash = get_ceph_journals(); + my $journalhash = get_ceph_journals($lsblk_info); + my $ceph_volume_infos = get_ceph_volume_infos(); - my $zfslist = get_zfs_devices(); + my $zfshash = get_zfs_devices($lsblk_info); - my $lvmlist = get_lvm_devices(); + my $lvmhash = get_lvm_devices($lsblk_info); - # we get cciss/c0d0 but need cciss!c0d0 - if (defined($disk) && $disk =~ m|^cciss/|) { - $disk =~ s|cciss/|cciss!|; + my $disk_regex = ".*"; + if (defined($disks)) { + if (!ref($disks)) { + $disks = [ $disks ]; + } elsif (ref($disks) ne 'ARRAY') { + die "disks is not a string or array reference\n"; + } + # we get cciss/c0d0 but need cciss!c0d0 + $_ =~ s|cciss/|cciss!| for @$disks; + + if ($include_partitions) { + # Proper blockdevice is needed for the regex, use parent for partitions. + for my $disk ($disks->@*) { + next if !is_partition("/dev/$disk"); + $disk = strip_dev(get_blockdev("/dev/$disk")); + } + } + + $disk_regex = "(?:" . join('|', @$disks) . ")"; } - dir_glob_foreach('/sys/block', '.*', sub { + dir_glob_foreach('/sys/block', $disk_regex, sub { my ($dev) = @_; - return if defined($disk) && $disk ne $dev; # whitelisting following devices # hdX: ide block device # sdX: sd block device @@ -440,6 +564,7 @@ sub get_disks { if ($sysdata->{rotational} == 0) { $type = 'ssd'; + $type = 'nvme' if $dev =~ m/^nvme\d+n\d+$/; $data->{rpm} = 0; } elsif ($sysdata->{rotational} == 1) { if ($data->{rpm} != -1) { @@ -455,25 +580,17 @@ sub get_disks { if (!$nosmart) { eval { - my $smartdata = get_smart_data($devpath, ($type ne 'ssd')); + my $smartdata = get_smart_data($devpath, !is_ssdlike($type)); $health = $smartdata->{health} if $smartdata->{health}; - if ($type eq 'ssd') { + if (is_ssdlike($type)) { # if we have an ssd we try to get the wearout indicator - my $wearval = get_wear_leveling_info($smartdata->{attributes}, $data->{model} || $sysdir->{model}); - $wearout = $wearval if $wearval; + my $wearval = get_wear_leveling_info($smartdata); + $wearout = $wearval if defined($wearval); } }; } - my $used; - - $used = 'LVM' if $lvmlist->{$devpath}; - - $used = 'mounted' if &$dev_is_mounted($devpath); - - $used = 'ZFS' if $zfslist->{$devpath}; - # we replaced cciss/ with cciss! above # but in the result we need cciss/ again # because the caller might want to check the @@ -496,18 +613,17 @@ sub get_disks { wearout => $wearout, }; + my $by_id_link = $data->{by_id_link}; + $disklist->{$dev}->{by_id_link} = $by_id_link if defined($by_id_link); + my $osdid = -1; my $bluestore = 0; + my $osdencrypted = 0; my $journal_count = 0; my $db_count = 0; my $wal_count = 0; - my $found_partitions; - my $found_lvm; - my $found_mountpoints; - my $found_zfs; - my $found_dm; my $partpath = $devpath; # remove part after last / to @@ -515,55 +631,133 @@ sub get_disks { # e.g. from /dev/cciss/c0d0 get /dev/cciss $partpath =~ s/\/[^\/]+$//; - dir_glob_foreach("$sysdir", "$dev.+", sub { - my ($part) = @_; + my $determine_usage = sub { + my ($devpath, $sysdir, $is_partition) = @_; - $found_partitions = 1; + return 'LVM' if $lvmhash->{$devpath}; + return 'ZFS' if $zfshash->{$devpath}; - if (my $mp = &$dev_is_mounted("$partpath/$part")) { - $found_mountpoints = 1; - if ($mp =~ m|^/var/lib/ceph/osd/ceph-(\d+)$|) { - $osdid = $1; - } + my $info = $lsblk_info->{$devpath} // {}; + + my $parttype = $info->{parttype}; + if (defined($parttype)) { + return 'BIOS boot' + if $parttype eq '21686148-6449-6e6f-744e-656564454649'; + return 'EFI' + if $parttype eq 'c12a7328-f81f-11d2-ba4b-00a0c93ec93b'; + return 'ZFS reserved' + if $parttype eq '6a945a3b-1dd2-11b2-99a6-080020736631'; } - if ($lvmlist->{"$partpath/$part"}) { - $found_lvm = 1; + my $fstype = $info->{fstype}; + if (defined($fstype)) { + return "${fstype} (mounted)" if $mounted->{$devpath}; + return "${fstype}"; } + return 'mounted' if $mounted->{$devpath}; + + return if !$is_partition; - if ($zfslist->{"$partpath/$part"}) { - $found_zfs = 1; + # for devices, this check is done explicitly later + return 'Device Mapper' if !dir_is_empty("$sysdir/holders"); + + return; # unused partition + }; + + my $collect_ceph_info = sub { + my ($devpath) = @_; + + my $ceph_volume = $ceph_volume_infos->{$devpath} or return; + $journal_count += $ceph_volume->{journal} // 0; + $db_count += $ceph_volume->{db} // 0; + $wal_count += $ceph_volume->{wal} // 0; + if (defined($ceph_volume->{osdid})) { + $osdid = $ceph_volume->{osdid}; + $bluestore = 1 if $ceph_volume->{bluestore}; + $osdencrypted = 1 if $ceph_volume->{encrypted}; } - if ($journalhash->{"$partpath/$part"}) { - $journal_count++ if $journalhash->{"$partpath/$part"} == 1; - $db_count++ if $journalhash->{"$partpath/$part"} == 2; - $wal_count++ if $journalhash->{"$partpath/$part"} == 3; - $bluestore = 1 if $journalhash->{"$partpath/$part"} == 4; + my $result = { %{$ceph_volume} }; + $result->{journals} = delete $result->{journal} + if $result->{journal}; + return $result; + }; + + my $partitions = {}; + + dir_glob_foreach("$sysdir", "$dev.+", sub { + my ($part) = @_; + + $partitions->{$part} = $collect_ceph_info->("$partpath/$part"); + my $lvm_based_osd = defined($partitions->{$part}); + + $partitions->{$part}->{devpath} = "$partpath/$part"; + $partitions->{$part}->{parent} = "$devpath"; + $partitions->{$part}->{gpt} = $data->{gpt}; + $partitions->{$part}->{type} = 'partition'; + $partitions->{$part}->{size} = + get_sysdir_size("$sysdir/$part") // 0; + $partitions->{$part}->{used} = + $determine_usage->("$partpath/$part", "$sysdir/$part", 1); + $partitions->{$part}->{osdid} //= -1; + + # Avoid counting twice (e.g. partition on which the LVM for the + # DB OSD resides is present in the $journalhash) + return if $lvm_based_osd; + + # Legacy handling for non-LVM based OSDs + + if (my $mp = $mounted->{"$partpath/$part"}) { + if ($mp =~ m|^/var/lib/ceph/osd/ceph-(\d+)$|) { + $osdid = $1; + $partitions->{$part}->{osdid} = $osdid; + } } - if (!dir_is_empty("$sysdir/$part/holders") && !$found_lvm) { - $found_dm = 1; + if (my $journal_part = $journalhash->{"$partpath/$part"}) { + $journal_count++ if $journal_part == 1; + $db_count++ if $journal_part == 2; + $wal_count++ if $journal_part == 3; + $bluestore = 1 if $journal_part == 4; + + $partitions->{$part}->{journals} = 1 if $journal_part == 1; + $partitions->{$part}->{db} = 1 if $journal_part == 2; + $partitions->{$part}->{wal} = 1 if $journal_part == 3; + $partitions->{$part}->{bluestore} = 1 if $journal_part == 4; } }); - $used = 'mounted' if $found_mountpoints && !$used; - $used = 'LVM' if $found_lvm && !$used; - $used = 'ZFS' if $found_zfs && !$used; - $used = 'Device Mapper' if $found_dm && !$used; - $used = 'partitions' if $found_partitions && !$used; - + my $used = $determine_usage->($devpath, $sysdir, 0); + if (!$include_partitions) { + foreach my $part (sort keys %{$partitions}) { + $used //= $partitions->{$part}->{used}; + } + } else { + # fstype might be set even if there are partitions, but showing that is confusing + $used = 'partitions' if scalar(keys %{$partitions}); + } + $used //= 'partitions' if scalar(keys %{$partitions}); # multipath, software raid, etc. # this check comes in last, to show more specific info # if we have it - $used = 'Device Mapper' if !$used && !dir_is_empty("$sysdir/holders"); + $used //= 'Device Mapper' if !dir_is_empty("$sysdir/holders"); $disklist->{$dev}->{used} = $used if $used; + + $collect_ceph_info->($devpath); + $disklist->{$dev}->{osdid} = $osdid; $disklist->{$dev}->{journals} = $journal_count if $journal_count; $disklist->{$dev}->{bluestore} = $bluestore if $osdid != -1; + $disklist->{$dev}->{osdencrypted} = $osdencrypted if $osdid != -1; $disklist->{$dev}->{db} = $db_count if $db_count; $disklist->{$dev}->{wal} = $wal_count if $wal_count; + + if ($include_partitions) { + foreach my $part (keys %{$partitions}) { + $disklist->{$part} = $partitions->{$part}; + } + } }); return $disklist; @@ -573,18 +767,20 @@ sub get_disks { sub get_partnum { my ($part_path) = @_; - my ($mode, $rdev) = (stat($part_path))[2,6]; + my $st = stat($part_path); - next if !$mode || !S_ISBLK($mode) || !$rdev; - my $major = PVE::Tools::dev_t_major($rdev); - my $minor = PVE::Tools::dev_t_minor($rdev); + die "error detecting block device '$part_path'\n" + if !$st || !$st->mode || !S_ISBLK($st->mode) || !$st->rdev; + + my $major = PVE::Tools::dev_t_major($st->rdev); + my $minor = PVE::Tools::dev_t_minor($st->rdev); my $partnum_path = "/sys/dev/block/$major:$minor/"; my $partnum; $partnum = file_read_firstline("${partnum_path}partition"); - die "Partition does not exists\n" if !defined($partnum); + die "Partition does not exist\n" if !defined($partnum); #untaint and ensure it is a int if ($partnum =~ m/(\d+)/) { @@ -600,19 +796,28 @@ sub get_partnum { sub get_blockdev { my ($part_path) = @_; - my $dev = $1 if $part_path =~ m|^/dev/(.*)$|; - my $link = readlink "/sys/class/block/$dev"; - my $block_dev = $1 if $link =~ m|([^/]*)/$dev$|; + my ($dev, $block_dev); + if ($part_path =~ m|^/dev/(.*)$|) { + $dev = $1; + my $link = readlink "/sys/class/block/$dev"; + $block_dev = $1 if $link =~ m|([^/]*)/$dev$|; + } die "Can't parse parent device\n" if !defined($block_dev); die "No valid block device\n" if index($dev, $block_dev) == -1; $block_dev = "/dev/$block_dev"; - die "Block device does not exsists\n" if !(-b $block_dev); + die "Block device does not exists\n" if !(-b $block_dev); return $block_dev; } +sub is_partition { + my ($dev_path) = @_; + + return defined(eval { get_partnum($dev_path) }); +} + sub locked_disk_action { my ($sub) = @_; my $res = PVE::Tools::lock_file('/run/lock/pve-diskmanage.lck', undef, $sub); @@ -628,4 +833,137 @@ sub assert_disk_unused { return undef; } +sub append_partition { + my ($dev, $size) = @_; + + my $devname = $dev; + $devname =~ s|^/dev/||; + + my $newpartid = 1; + dir_glob_foreach("/sys/block/$devname", qr/\Q$devname\E.*?(\d+)/, sub { + my ($part, $partid) = @_; + + if ($partid >= $newpartid) { + $newpartid = $partid + 1; + } + }); + + $size = PVE::Tools::convert_size($size, 'b' => 'mb'); + + run_command([ $SGDISK, '-n', "$newpartid:0:+${size}M", $dev ], + errmsg => "error creating partition '$newpartid' on '$dev'"); + + my $partition; + + # loop again to detect the real partition device which does not always follow + # a strict $devname$partition scheme like /dev/nvme0n1 -> /dev/nvme0n1p1 + dir_glob_foreach("/sys/block/$devname", qr/\Q$devname\E.*$newpartid/, sub { + my ($part) = @_; + + $partition = "/dev/$part"; + }); + + return $partition; +} + +# Check if a disk or any of its partitions has a holder. +# Can also be called with a partition. +# Expected to be called with a result of verify_blockdev_path(). +sub has_holder { + my ($devpath) = @_; + + my $dev = strip_dev($devpath); + + return $devpath if !dir_is_empty("/sys/class/block/${dev}/holders"); + + my $found; + dir_glob_foreach("/sys/block/${dev}", "${dev}.+", sub { + my ($part) = @_; + $found = "/dev/${part}" if !dir_is_empty("/sys/class/block/${part}/holders"); + }); + + return $found; +} + +# Basic check if a disk or any of its partitions is mounted. +# Can also be called with a partition. +# Expected to be called with a result of verify_blockdev_path(). +sub is_mounted { + my ($devpath) = @_; + + my $mounted = mounted_blockdevs(); + + return $devpath if $mounted->{$devpath}; + + my $dev = strip_dev($devpath); + + my $found; + dir_glob_foreach("/sys/block/${dev}", "${dev}.+", sub { + my ($part) = @_; + my $partpath = "/dev/${part}"; + + $found = $partpath if $mounted->{$partpath}; + }); + + return $found; +} + +# Currently only supports GPT-partitioned disks. +sub change_parttype { + my ($partpath, $parttype) = @_; + + my $err = "unable to change partition type for $partpath"; + + my $partnum = get_partnum($partpath); + my $blockdev = get_blockdev($partpath); + my $dev = strip_dev($blockdev); + + my $info = get_disks($dev, 1); + die "$err - unable to get disk info for '$blockdev'\n" if !defined($info->{$dev}); + die "$err - disk '$blockdev' is not GPT partitioned\n" if !$info->{$dev}->{gpt}; + + run_command(['sgdisk', "-t${partnum}:${parttype}", $blockdev], errmsg => $err); +} + +# Wipes all labels and the first 200 MiB of a disk/partition (or the whole if it is smaller). +# If called with a partition, also sets the partition type to 0x83 'Linux filesystem'. +# Expected to be called with a result of verify_blockdev_path(). +sub wipe_blockdev { + my ($devpath) = @_; + + my $devname = basename($devpath); + my $dev_size = PVE::Tools::file_get_contents("/sys/class/block/$devname/size"); + + ($dev_size) = $dev_size =~ m|(\d+)|; # untaint $dev_size + die "Couldn't get the size of the device $devname\n" if !defined($dev_size); + + my $size = ($dev_size * 512 / 1024 / 1024); + my $count = ($size < 200) ? $size : 200; + + my $to_wipe = []; + dir_glob_foreach("/sys/class/block/${devname}", "${devname}.+", sub { + my ($part) = @_; + push $to_wipe->@*, "/dev/${part}" if -b "/dev/${part}"; + }); + + if (scalar($to_wipe->@*) > 0) { + print "found child partitions to wipe: ". join(', ', $to_wipe->@*) ."\n"; + } + push $to_wipe->@*, $devpath; # put actual device last + + print "wiping block device ${devpath}\n"; + + run_command(['wipefs', '--all', $to_wipe->@*], errmsg => "error wiping '${devpath}'"); + + run_command( + ['dd', 'if=/dev/zero', "of=${devpath}", 'bs=1M', 'conv=fdatasync', "count=${count}"], + errmsg => "error wiping '${devpath}'", + ); + + if (is_partition($devpath)) { + eval { change_parttype($devpath, '8300'); }; + warn $@ if $@; + } +} + 1;