]>
Commit | Line | Data |
---|---|---|
1 | package PVE::Diskmanage; | |
2 | ||
3 | use strict; | |
4 | use warnings; | |
5 | ||
6 | use PVE::ProcFSTools; | |
7 | use Data::Dumper; | |
8 | use Cwd qw(abs_path); | |
9 | use Fcntl ':mode'; | |
10 | use File::Basename; | |
11 | use File::stat; | |
12 | use JSON; | |
13 | ||
14 | use PVE::Tools qw(extract_param run_command file_get_contents file_read_firstline dir_glob_regex dir_glob_foreach trim); | |
15 | ||
16 | my $SMARTCTL = "/usr/sbin/smartctl"; | |
17 | my $ZPOOL = "/sbin/zpool"; | |
18 | my $SGDISK = "/sbin/sgdisk"; | |
19 | my $PVS = "/sbin/pvs"; | |
20 | my $LVS = "/sbin/lvs"; | |
21 | my $LSBLK = "/bin/lsblk"; | |
22 | ||
23 | my sub strip_dev :prototype($) { | |
24 | my ($devpath) = @_; | |
25 | $devpath =~ s|^/dev/||; | |
26 | return $devpath; | |
27 | } | |
28 | ||
29 | sub check_bin { | |
30 | my ($path) = @_; | |
31 | return -x $path; | |
32 | } | |
33 | ||
34 | sub verify_blockdev_path { | |
35 | my ($rel_path) = @_; | |
36 | ||
37 | die "missing path" if !$rel_path; | |
38 | my $path = abs_path($rel_path); | |
39 | die "failed to get absolute path to $rel_path\n" if !$path; | |
40 | ||
41 | die "got unusual device path '$path'\n" if $path !~ m|^/dev/(.*)$|; | |
42 | ||
43 | $path = "/dev/$1"; # untaint | |
44 | ||
45 | assert_blockdev($path); | |
46 | ||
47 | return $path; | |
48 | } | |
49 | ||
50 | sub assert_blockdev { | |
51 | my ($dev, $noerr) = @_; | |
52 | ||
53 | if ($dev !~ m|^/dev/| || !(-b $dev)) { | |
54 | return undef if $noerr; | |
55 | die "not a valid block device\n"; | |
56 | } | |
57 | ||
58 | return 1; | |
59 | } | |
60 | ||
61 | sub init_disk { | |
62 | my ($disk, $uuid) = @_; | |
63 | ||
64 | assert_blockdev($disk); | |
65 | ||
66 | # we should already have checked if it is in use in the api call | |
67 | # but we check again for safety | |
68 | die "disk $disk is already in use\n" if disk_is_used($disk); | |
69 | ||
70 | my $id = $uuid || 'R'; | |
71 | run_command([$SGDISK, $disk, '-U', $id]); | |
72 | return 1; | |
73 | } | |
74 | ||
75 | sub disk_is_used { | |
76 | my ($disk) = @_; | |
77 | ||
78 | my $dev = $disk; | |
79 | $dev =~ s|^/dev/||; | |
80 | ||
81 | my $disklist = get_disks($dev, 1); | |
82 | ||
83 | die "'$disk' is not a valid local disk\n" if !defined($disklist->{$dev}); | |
84 | return 1 if $disklist->{$dev}->{used}; | |
85 | ||
86 | return 0; | |
87 | } | |
88 | ||
89 | sub get_smart_data { | |
90 | my ($disk, $healthonly) = @_; | |
91 | ||
92 | assert_blockdev($disk); | |
93 | my $smartdata = {}; | |
94 | my $type; | |
95 | ||
96 | my $returncode = 0; | |
97 | ||
98 | if ($disk =~ m!^/dev/(nvme\d+n\d+)$!) { | |
99 | my $info = get_sysdir_info("/sys/block/$1"); | |
100 | $disk = "/dev/".($info->{device} | |
101 | or die "failed to get nvme controller device for $disk\n"); | |
102 | } | |
103 | ||
104 | my $cmd = [$SMARTCTL, '-H']; | |
105 | push @$cmd, '-A', '-f', 'brief' if !$healthonly; | |
106 | push @$cmd, $disk; | |
107 | ||
108 | eval { | |
109 | $returncode = run_command($cmd, noerr => 1, outfunc => sub{ | |
110 | my ($line) = @_; | |
111 | ||
112 | # ATA SMART attributes, e.g.: | |
113 | # ID# ATTRIBUTE_NAME FLAGS VALUE WORST THRESH FAIL RAW_VALUE | |
114 | # 1 Raw_Read_Error_Rate POSR-K 100 100 000 - 0 | |
115 | # | |
116 | # SAS and NVME disks, e.g.: | |
117 | # Data Units Written: 5,584,952 [2.85 TB] | |
118 | # Accumulated start-stop cycles: 34 | |
119 | ||
120 | if (defined($type) && $type eq 'ata' && $line =~ m/^([ \d]{2}\d)\s+(\S+)\s+(\S{6})\s+(\d+)\s+(\d+)\s+(\S+)\s+(\S+)\s+(.*)$/) { | |
121 | my $entry = {}; | |
122 | ||
123 | $entry->{name} = $2 if defined $2; | |
124 | $entry->{flags} = $3 if defined $3; | |
125 | # the +0 makes a number out of the strings | |
126 | $entry->{value} = $4+0 if defined $4; | |
127 | $entry->{worst} = $5+0 if defined $5; | |
128 | # some disks report the default threshold as --- instead of 000 | |
129 | if (defined($6) && $6 eq '---') { | |
130 | $entry->{threshold} = 0; | |
131 | } else { | |
132 | $entry->{threshold} = $6+0 if defined $6; | |
133 | } | |
134 | $entry->{fail} = $7 if defined $7; | |
135 | $entry->{raw} = $8 if defined $8; | |
136 | $entry->{id} = $1 if defined $1; | |
137 | push @{$smartdata->{attributes}}, $entry; | |
138 | } elsif ($line =~ m/(?:Health Status|self\-assessment test result): (.*)$/ ) { | |
139 | $smartdata->{health} = $1; | |
140 | } elsif ($line =~ m/Vendor Specific SMART Attributes with Thresholds:/) { | |
141 | $type = 'ata'; | |
142 | delete $smartdata->{text}; | |
143 | } elsif ($line =~ m/=== START OF (READ )?SMART DATA SECTION ===/) { | |
144 | $type = 'text'; | |
145 | } elsif (defined($type) && $type eq 'text') { | |
146 | $smartdata->{text} = '' if !defined $smartdata->{text}; | |
147 | $smartdata->{text} .= "$line\n"; | |
148 | # extract wearout from nvme/sas text, allow for decimal values | |
149 | if ($line =~ m/Percentage Used(?: endurance indicator)?:\s*(\d+(?:\.\d+)?)\%/i) { | |
150 | $smartdata->{wearout} = 100 - $1; | |
151 | } | |
152 | } elsif ($line =~ m/SMART Disabled/) { | |
153 | $smartdata->{health} = "SMART Disabled"; | |
154 | } | |
155 | }); | |
156 | }; | |
157 | my $err = $@; | |
158 | ||
159 | # bit 0 and 1 mark an severe smartctl error | |
160 | # all others are for disk status, so ignore them | |
161 | # see smartctl(8) | |
162 | if ((defined($returncode) && ($returncode & 0b00000011)) || $err) { | |
163 | die "Error getting S.M.A.R.T. data: Exit code: $returncode\n"; | |
164 | } | |
165 | ||
166 | $smartdata->{type} = $type; | |
167 | ||
168 | return $smartdata; | |
169 | } | |
170 | ||
171 | sub get_lsblk_info() { | |
172 | my $cmd = [$LSBLK, '--json', '-o', 'path,parttype,fstype']; | |
173 | my $output = ""; | |
174 | my $res = {}; | |
175 | eval { | |
176 | run_command($cmd, outfunc => sub { | |
177 | my ($line) = @_; | |
178 | $output .= "$line\n"; | |
179 | }); | |
180 | }; | |
181 | warn "$@\n" if $@; | |
182 | return $res if $output eq ''; | |
183 | ||
184 | my $parsed = eval { decode_json($output) }; | |
185 | warn "$@\n" if $@; | |
186 | my $list = $parsed->{blockdevices} // []; | |
187 | ||
188 | $res = { map { | |
189 | $_->{path} => { | |
190 | parttype => $_->{parttype}, | |
191 | fstype => $_->{fstype} | |
192 | } | |
193 | } @{$list} }; | |
194 | ||
195 | return $res; | |
196 | } | |
197 | ||
198 | my $get_devices_by_partuuid = sub { | |
199 | my ($lsblk_info, $uuids, $res) = @_; | |
200 | ||
201 | $res = {} if !defined($res); | |
202 | ||
203 | foreach my $dev (sort keys %{$lsblk_info}) { | |
204 | my $uuid = $lsblk_info->{$dev}->{parttype}; | |
205 | next if !defined($uuid) || !defined($uuids->{$uuid}); | |
206 | $res->{$dev} = $uuids->{$uuid}; | |
207 | } | |
208 | ||
209 | return $res; | |
210 | }; | |
211 | ||
212 | sub get_zfs_devices { | |
213 | my ($lsblk_info) = @_; | |
214 | my $res = {}; | |
215 | ||
216 | return {} if !check_bin($ZPOOL); | |
217 | ||
218 | # use zpool and parttype uuid, | |
219 | # because log and cache do not have | |
220 | # zfs type uuid | |
221 | eval { | |
222 | run_command([$ZPOOL, 'list', '-HPLv'], outfunc => sub { | |
223 | my ($line) = @_; | |
224 | ||
225 | if ($line =~ m|^\t([^\t]+)\t|) { | |
226 | $res->{$1} = 1; | |
227 | } | |
228 | }); | |
229 | }; | |
230 | ||
231 | # only warn here, | |
232 | # because maybe zfs tools are not installed | |
233 | warn "$@\n" if $@; | |
234 | ||
235 | my $uuids = { | |
236 | "6a898cc3-1dd2-11b2-99a6-080020736631" => 1, # apple | |
237 | "516e7cba-6ecf-11d6-8ff8-00022d09712b" => 1, # bsd | |
238 | }; | |
239 | ||
240 | ||
241 | $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); | |
242 | ||
243 | return $res; | |
244 | } | |
245 | ||
246 | sub get_lvm_devices { | |
247 | my ($lsblk_info) = @_; | |
248 | my $res = {}; | |
249 | eval { | |
250 | run_command([$PVS, '--noheadings', '--readonly', '-o', 'pv_name'], outfunc => sub{ | |
251 | my ($line) = @_; | |
252 | $line = trim($line); | |
253 | if ($line =~ m|^/dev/|) { | |
254 | $res->{$line} = 1; | |
255 | } | |
256 | }); | |
257 | }; | |
258 | ||
259 | # if something goes wrong, we do not want | |
260 | # to give up, but indicate an error has occurred | |
261 | warn "$@\n" if $@; | |
262 | ||
263 | my $uuids = { | |
264 | "e6d6d379-f507-44c2-a23c-238f2a3df928" => 1, | |
265 | }; | |
266 | ||
267 | $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); | |
268 | ||
269 | return $res; | |
270 | } | |
271 | ||
272 | sub get_ceph_journals { | |
273 | my ($lsblk_info) = @_; | |
274 | my $res = {}; | |
275 | ||
276 | my $uuids = { | |
277 | '45b0969e-9b03-4f30-b4c6-b4b80ceff106' => 1, # journal | |
278 | '30cd0809-c2b2-499c-8879-2d6b78529876' => 2, # db | |
279 | '5ce17fce-4087-4169-b7ff-056cc58473f9' => 3, # wal | |
280 | 'cafecafe-9b03-4f30-b4c6-b4b80ceff106' => 4, # block | |
281 | }; | |
282 | ||
283 | $res = $get_devices_by_partuuid->($lsblk_info, $uuids, $res); | |
284 | ||
285 | return $res; | |
286 | } | |
287 | ||
288 | # reads the lv_tags and matches them with the devices | |
289 | sub get_ceph_volume_infos { | |
290 | my $result = {}; | |
291 | ||
292 | my $cmd = [ $LVS, '-S', 'lv_name=~^osd-', '-o', 'devices,lv_name,lv_tags', | |
293 | '--noheadings', '--readonly', '--separator', ';' ]; | |
294 | ||
295 | run_command($cmd, outfunc => sub { | |
296 | my $line = shift; | |
297 | $line =~ s/(?:^\s+)|(?:\s+$)//g; # trim whitespaces | |
298 | ||
299 | my $fields = [ split(';', $line) ]; | |
300 | ||
301 | # lvs syntax is /dev/sdX(Y) where Y is the start (which we do not need) | |
302 | my ($dev) = $fields->[0] =~ m|^(/dev/[a-z]+[^(]*)|; | |
303 | if ($fields->[1] =~ m|^osd-([^-]+)-|) { | |
304 | my $type = $1; | |
305 | # $result autovivification is wanted, to not creating empty hashes | |
306 | if (($type eq 'block' || $type eq 'data') && $fields->[2] =~ m/ceph.osd_id=([^,]+)/) { | |
307 | $result->{$dev}->{osdid} = $1; | |
308 | $result->{$dev}->{bluestore} = ($type eq 'block'); | |
309 | if ($fields->[2] =~ m/ceph\.encrypted=1/) { | |
310 | $result->{$dev}->{encrypted} = 1; | |
311 | } | |
312 | } else { | |
313 | # undef++ becomes '1' (see `perldoc perlop`: Auto-increment) | |
314 | $result->{$dev}->{$type}++; | |
315 | } | |
316 | } | |
317 | }); | |
318 | ||
319 | return $result; | |
320 | } | |
321 | ||
322 | sub get_udev_info { | |
323 | my ($dev) = @_; | |
324 | ||
325 | my $info = ""; | |
326 | my $data = {}; | |
327 | eval { | |
328 | run_command(['udevadm', 'info', '-p', $dev, '--query', 'all'], outfunc => sub { | |
329 | my ($line) = @_; | |
330 | $info .= "$line\n"; | |
331 | }); | |
332 | }; | |
333 | warn $@ if $@; | |
334 | return undef if !$info; | |
335 | ||
336 | return undef if $info !~ m/^E: DEVTYPE=(disk|partition)$/m; | |
337 | return undef if $info =~ m/^E: ID_CDROM/m; | |
338 | ||
339 | # we use this, because some disks are not simply in /dev | |
340 | # e.g. /dev/cciss/c0d0 | |
341 | if ($info =~ m/^E: DEVNAME=(\S+)$/m) { | |
342 | $data->{devpath} = $1; | |
343 | } | |
344 | return if !defined($data->{devpath}); | |
345 | ||
346 | $data->{serial} = 'unknown'; | |
347 | if ($info =~ m/^E: ID_SERIAL_SHORT=(\S+)$/m) { | |
348 | $data->{serial} = $1; | |
349 | } | |
350 | ||
351 | $data->{gpt} = 0; | |
352 | if ($info =~ m/^E: ID_PART_TABLE_TYPE=gpt$/m) { | |
353 | $data->{gpt} = 1; | |
354 | } | |
355 | ||
356 | # detect SSD | |
357 | $data->{rpm} = -1; | |
358 | if ($info =~ m/^E: ID_ATA_ROTATION_RATE_RPM=(\d+)$/m) { | |
359 | $data->{rpm} = $1; | |
360 | } | |
361 | ||
362 | if ($info =~ m/^E: ID_BUS=usb$/m) { | |
363 | $data->{usb} = 1; | |
364 | } | |
365 | ||
366 | if ($info =~ m/^E: ID_MODEL=(.+)$/m) { | |
367 | $data->{model} = $1; | |
368 | } | |
369 | ||
370 | $data->{wwn} = 'unknown'; | |
371 | if ($info =~ m/^E: ID_WWN=(.*)$/m) { | |
372 | $data->{wwn} = $1; | |
373 | } | |
374 | ||
375 | if ($info =~ m/^E: DEVLINKS=(.+)$/m) { | |
376 | my @devlinks = grep(m#^/dev/disk/by-id/(ata|scsi|nvme(?!-eui))#, split (/ /, $1)); | |
377 | $data->{by_id_link} = $devlinks[0] if defined($devlinks[0]); | |
378 | } | |
379 | ||
380 | return $data; | |
381 | } | |
382 | ||
383 | sub get_sysdir_size { | |
384 | my ($sysdir) = @_; | |
385 | ||
386 | my $size = file_read_firstline("$sysdir/size"); | |
387 | return if !$size; | |
388 | ||
389 | # linux always considers sectors to be 512 bytes, | |
390 | # independently of real block size | |
391 | return $size * 512; | |
392 | } | |
393 | ||
394 | sub get_sysdir_info { | |
395 | my ($sysdir) = @_; | |
396 | ||
397 | return undef if ! -d "$sysdir/device"; | |
398 | ||
399 | my $data = {}; | |
400 | ||
401 | $data->{size} = get_sysdir_size($sysdir) or return; | |
402 | ||
403 | # dir/queue/rotational should be 1 for hdd, 0 for ssd | |
404 | $data->{rotational} = file_read_firstline("$sysdir/queue/rotational") // -1; | |
405 | ||
406 | $data->{vendor} = file_read_firstline("$sysdir/device/vendor") || 'unknown'; | |
407 | $data->{model} = file_read_firstline("$sysdir/device/model") || 'unknown'; | |
408 | ||
409 | if (defined(my $device = readlink("$sysdir/device"))) { | |
410 | # strip directory and untaint: | |
411 | ($data->{device}) = $device =~ m!([^/]+)$!; | |
412 | } | |
413 | ||
414 | return $data; | |
415 | } | |
416 | ||
417 | sub get_wear_leveling_info { | |
418 | my ($smartdata) = @_; | |
419 | my $attributes = $smartdata->{attributes}; | |
420 | ||
421 | if (defined($smartdata->{wearout})) { | |
422 | return $smartdata->{wearout}; | |
423 | } | |
424 | ||
425 | my $wearout; | |
426 | ||
427 | # Common register names that represent percentage values of potential | |
428 | # failure indicators used in drivedb.h of smartmontool's. Order matters, | |
429 | # as some drives may have multiple definitions | |
430 | my @wearoutregisters = ( | |
431 | "Media_Wearout_Indicator", | |
432 | "SSD_Life_Left", | |
433 | "Wear_Leveling_Count", | |
434 | "Perc_Write\/Erase_Ct_BC", | |
435 | "Perc_Rated_Life_Remain", | |
436 | "Remaining_Lifetime_Perc", | |
437 | "Percent_Lifetime_Remain", | |
438 | "Lifetime_Left", | |
439 | "PCT_Life_Remaining", | |
440 | "Lifetime_Remaining", | |
441 | "Percent_Life_Remaining", | |
442 | "Percent_Lifetime_Used", | |
443 | "Perc_Rated_Life_Used" | |
444 | ); | |
445 | ||
446 | # Search for S.M.A.R.T. attributes for known register | |
447 | foreach my $register (@wearoutregisters) { | |
448 | last if defined $wearout; | |
449 | foreach my $attr (@$attributes) { | |
450 | next if $attr->{name} !~ m/$register/; | |
451 | $wearout = $attr->{value}; | |
452 | last; | |
453 | } | |
454 | } | |
455 | ||
456 | return $wearout; | |
457 | } | |
458 | ||
459 | sub dir_is_empty { | |
460 | my ($dir) = @_; | |
461 | ||
462 | my $dh = IO::Dir->new ($dir); | |
463 | return 1 if !$dh; | |
464 | ||
465 | while (defined(my $tmp = $dh->read)) { | |
466 | next if $tmp eq '.' || $tmp eq '..'; | |
467 | $dh->close; | |
468 | return 0; | |
469 | } | |
470 | $dh->close; | |
471 | return 1; | |
472 | } | |
473 | ||
474 | sub is_iscsi { | |
475 | my ($sysdir) = @_; | |
476 | ||
477 | if (-l $sysdir && readlink($sysdir) =~ m|host[^/]*/session[^/]*|) { | |
478 | return 1; | |
479 | } | |
480 | ||
481 | return 0; | |
482 | } | |
483 | ||
484 | my sub is_ssdlike { | |
485 | my ($type) = @_; | |
486 | return $type eq 'ssd' || $type eq 'nvme'; | |
487 | } | |
488 | ||
489 | sub mounted_blockdevs { | |
490 | my $mounted = {}; | |
491 | ||
492 | my $mounts = PVE::ProcFSTools::parse_proc_mounts(); | |
493 | ||
494 | foreach my $mount (@$mounts) { | |
495 | next if $mount->[0] !~ m|^/dev/|; | |
496 | $mounted->{abs_path($mount->[0])} = $mount->[1]; | |
497 | }; | |
498 | ||
499 | return $mounted; | |
500 | } | |
501 | ||
502 | sub get_disks { | |
503 | my ($disks, $nosmart, $include_partitions) = @_; | |
504 | my $disklist = {}; | |
505 | ||
506 | my $mounted = mounted_blockdevs(); | |
507 | ||
508 | my $lsblk_info = get_lsblk_info(); | |
509 | ||
510 | my $journalhash = get_ceph_journals($lsblk_info); | |
511 | my $ceph_volume_infos = get_ceph_volume_infos(); | |
512 | ||
513 | my $zfshash = get_zfs_devices($lsblk_info); | |
514 | ||
515 | my $lvmhash = get_lvm_devices($lsblk_info); | |
516 | ||
517 | my $disk_regex = ".*"; | |
518 | if (defined($disks)) { | |
519 | if (!ref($disks)) { | |
520 | $disks = [ $disks ]; | |
521 | } elsif (ref($disks) ne 'ARRAY') { | |
522 | die "disks is not a string or array reference\n"; | |
523 | } | |
524 | # we get cciss/c0d0 but need cciss!c0d0 | |
525 | $_ =~ s|cciss/|cciss!| for @$disks; | |
526 | ||
527 | if ($include_partitions) { | |
528 | # Proper blockdevice is needed for the regex, use parent for partitions. | |
529 | for my $disk ($disks->@*) { | |
530 | next if !is_partition("/dev/$disk"); | |
531 | $disk = strip_dev(get_blockdev("/dev/$disk")); | |
532 | } | |
533 | } | |
534 | ||
535 | $disk_regex = "(?:" . join('|', @$disks) . ")"; | |
536 | } | |
537 | ||
538 | dir_glob_foreach('/sys/block', $disk_regex, sub { | |
539 | my ($dev) = @_; | |
540 | # whitelisting following devices | |
541 | # hdX: ide block device | |
542 | # sdX: sd block device | |
543 | # vdX: virtual block device | |
544 | # xvdX: xen virtual block device | |
545 | # nvmeXnY: nvme devices | |
546 | # cciss!cXnY: cciss devices | |
547 | return if $dev !~ m/^(h|s|x?v)d[a-z]+$/ && | |
548 | $dev !~ m/^nvme\d+n\d+$/ && | |
549 | $dev !~ m/^cciss\!c\d+d\d+$/; | |
550 | ||
551 | my $data = get_udev_info("/sys/block/$dev"); | |
552 | return if !defined($data); | |
553 | my $devpath = $data->{devpath}; | |
554 | ||
555 | my $sysdir = "/sys/block/$dev"; | |
556 | ||
557 | # we do not want iscsi devices | |
558 | return if is_iscsi($sysdir); | |
559 | ||
560 | my $sysdata = get_sysdir_info($sysdir); | |
561 | return if !defined($sysdata); | |
562 | ||
563 | my $type = 'unknown'; | |
564 | ||
565 | if ($sysdata->{rotational} == 0) { | |
566 | $type = 'ssd'; | |
567 | $type = 'nvme' if $dev =~ m/^nvme\d+n\d+$/; | |
568 | $data->{rpm} = 0; | |
569 | } elsif ($sysdata->{rotational} == 1) { | |
570 | if ($data->{rpm} != -1) { | |
571 | $type = 'hdd'; | |
572 | } elsif ($data->{usb}) { | |
573 | $type = 'usb'; | |
574 | $data->{rpm} = 0; | |
575 | } | |
576 | } | |
577 | ||
578 | my $health = 'UNKNOWN'; | |
579 | my $wearout = 'N/A'; | |
580 | ||
581 | if (!$nosmart) { | |
582 | eval { | |
583 | my $smartdata = get_smart_data($devpath, !is_ssdlike($type)); | |
584 | $health = $smartdata->{health} if $smartdata->{health}; | |
585 | ||
586 | if (is_ssdlike($type)) { | |
587 | # if we have an ssd we try to get the wearout indicator | |
588 | my $wearval = get_wear_leveling_info($smartdata); | |
589 | $wearout = $wearval if defined($wearval); | |
590 | } | |
591 | }; | |
592 | } | |
593 | ||
594 | # we replaced cciss/ with cciss! above | |
595 | # but in the result we need cciss/ again | |
596 | # because the caller might want to check the | |
597 | # result again with the original parameter | |
598 | if ($dev =~ m|^cciss!|) { | |
599 | $dev =~ s|^cciss!|cciss/|; | |
600 | } | |
601 | ||
602 | $disklist->{$dev} = { | |
603 | vendor => $sysdata->{vendor}, | |
604 | model => $data->{model} || $sysdata->{model}, | |
605 | size => $sysdata->{size}, | |
606 | serial => $data->{serial}, | |
607 | gpt => $data->{gpt}, | |
608 | rpm => $data->{rpm}, | |
609 | type => $type, | |
610 | wwn => $data->{wwn}, | |
611 | health => $health, | |
612 | devpath => $devpath, | |
613 | wearout => $wearout, | |
614 | }; | |
615 | ||
616 | my $by_id_link = $data->{by_id_link}; | |
617 | $disklist->{$dev}->{by_id_link} = $by_id_link if defined($by_id_link); | |
618 | ||
619 | my $osdid = -1; | |
620 | my $bluestore = 0; | |
621 | my $osdencrypted = 0; | |
622 | ||
623 | my $journal_count = 0; | |
624 | my $db_count = 0; | |
625 | my $wal_count = 0; | |
626 | ||
627 | my $partpath = $devpath; | |
628 | ||
629 | # remove part after last / to | |
630 | # get the base path for the partitions | |
631 | # e.g. from /dev/cciss/c0d0 get /dev/cciss | |
632 | $partpath =~ s/\/[^\/]+$//; | |
633 | ||
634 | my $determine_usage = sub { | |
635 | my ($devpath, $sysdir, $is_partition) = @_; | |
636 | ||
637 | return 'LVM' if $lvmhash->{$devpath}; | |
638 | return 'ZFS' if $zfshash->{$devpath}; | |
639 | ||
640 | my $info = $lsblk_info->{$devpath} // {}; | |
641 | ||
642 | my $parttype = $info->{parttype}; | |
643 | if (defined($parttype)) { | |
644 | return 'BIOS boot' | |
645 | if $parttype eq '21686148-6449-6e6f-744e-656564454649'; | |
646 | return 'EFI' | |
647 | if $parttype eq 'c12a7328-f81f-11d2-ba4b-00a0c93ec93b'; | |
648 | return 'ZFS reserved' | |
649 | if $parttype eq '6a945a3b-1dd2-11b2-99a6-080020736631'; | |
650 | } | |
651 | ||
652 | my $fstype = $info->{fstype}; | |
653 | if (defined($fstype)) { | |
654 | return "${fstype} (mounted)" if $mounted->{$devpath}; | |
655 | return "${fstype}"; | |
656 | } | |
657 | return 'mounted' if $mounted->{$devpath}; | |
658 | ||
659 | return if !$is_partition; | |
660 | ||
661 | # for devices, this check is done explicitly later | |
662 | return 'Device Mapper' if !dir_is_empty("$sysdir/holders"); | |
663 | ||
664 | return 'partition'; | |
665 | }; | |
666 | ||
667 | my $collect_ceph_info = sub { | |
668 | my ($devpath) = @_; | |
669 | ||
670 | my $ceph_volume = $ceph_volume_infos->{$devpath} or return; | |
671 | $journal_count += $ceph_volume->{journal} // 0; | |
672 | $db_count += $ceph_volume->{db} // 0; | |
673 | $wal_count += $ceph_volume->{wal} // 0; | |
674 | if (defined($ceph_volume->{osdid})) { | |
675 | $osdid = $ceph_volume->{osdid}; | |
676 | $bluestore = 1 if $ceph_volume->{bluestore}; | |
677 | $osdencrypted = 1 if $ceph_volume->{encrypted}; | |
678 | } | |
679 | ||
680 | my $result = { %{$ceph_volume} }; | |
681 | $result->{journals} = delete $result->{journal} | |
682 | if $result->{journal}; | |
683 | return $result; | |
684 | }; | |
685 | ||
686 | my $partitions = {}; | |
687 | ||
688 | dir_glob_foreach("$sysdir", "$dev.+", sub { | |
689 | my ($part) = @_; | |
690 | ||
691 | $partitions->{$part} = $collect_ceph_info->("$partpath/$part"); | |
692 | my $lvm_based_osd = defined($partitions->{$part}); | |
693 | ||
694 | $partitions->{$part}->{devpath} = "$partpath/$part"; | |
695 | $partitions->{$part}->{parent} = "$devpath"; | |
696 | $partitions->{$part}->{gpt} = $data->{gpt}; | |
697 | $partitions->{$part}->{type} = 'partition'; | |
698 | $partitions->{$part}->{size} = | |
699 | get_sysdir_size("$sysdir/$part") // 0; | |
700 | $partitions->{$part}->{used} = | |
701 | $determine_usage->("$partpath/$part", "$sysdir/$part", 1); | |
702 | $partitions->{$part}->{osdid} //= -1; | |
703 | ||
704 | # Avoid counting twice (e.g. partition on which the LVM for the | |
705 | # DB OSD resides is present in the $journalhash) | |
706 | return if $lvm_based_osd; | |
707 | ||
708 | # Legacy handling for non-LVM based OSDs | |
709 | ||
710 | if (my $mp = $mounted->{"$partpath/$part"}) { | |
711 | if ($mp =~ m|^/var/lib/ceph/osd/ceph-(\d+)$|) { | |
712 | $osdid = $1; | |
713 | $partitions->{$part}->{osdid} = $osdid; | |
714 | } | |
715 | } | |
716 | ||
717 | if (my $journal_part = $journalhash->{"$partpath/$part"}) { | |
718 | $journal_count++ if $journal_part == 1; | |
719 | $db_count++ if $journal_part == 2; | |
720 | $wal_count++ if $journal_part == 3; | |
721 | $bluestore = 1 if $journal_part == 4; | |
722 | ||
723 | $partitions->{$part}->{journals} = 1 if $journal_part == 1; | |
724 | $partitions->{$part}->{db} = 1 if $journal_part == 2; | |
725 | $partitions->{$part}->{wal} = 1 if $journal_part == 3; | |
726 | $partitions->{$part}->{bluestore} = 1 if $journal_part == 4; | |
727 | } | |
728 | }); | |
729 | ||
730 | my $used = $determine_usage->($devpath, $sysdir, 0); | |
731 | if (!$include_partitions) { | |
732 | foreach my $part (sort keys %{$partitions}) { | |
733 | next if $partitions->{$part}->{used} eq 'partition'; | |
734 | $used //= $partitions->{$part}->{used}; | |
735 | } | |
736 | } else { | |
737 | # fstype might be set even if there are partitions, but showing that is confusing | |
738 | $used = 'partitions' if scalar(keys %{$partitions}); | |
739 | } | |
740 | $used //= 'partitions' if scalar(keys %{$partitions}); | |
741 | # multipath, software raid, etc. | |
742 | # this check comes in last, to show more specific info | |
743 | # if we have it | |
744 | $used //= 'Device Mapper' if !dir_is_empty("$sysdir/holders"); | |
745 | ||
746 | $disklist->{$dev}->{used} = $used if $used; | |
747 | ||
748 | $collect_ceph_info->($devpath); | |
749 | ||
750 | $disklist->{$dev}->{osdid} = $osdid; | |
751 | $disklist->{$dev}->{journals} = $journal_count if $journal_count; | |
752 | $disklist->{$dev}->{bluestore} = $bluestore if $osdid != -1; | |
753 | $disklist->{$dev}->{osdencrypted} = $osdencrypted if $osdid != -1; | |
754 | $disklist->{$dev}->{db} = $db_count if $db_count; | |
755 | $disklist->{$dev}->{wal} = $wal_count if $wal_count; | |
756 | ||
757 | if ($include_partitions) { | |
758 | foreach my $part (keys %{$partitions}) { | |
759 | $disklist->{$part} = $partitions->{$part}; | |
760 | } | |
761 | } | |
762 | }); | |
763 | ||
764 | return $disklist; | |
765 | ||
766 | } | |
767 | ||
768 | sub get_partnum { | |
769 | my ($part_path) = @_; | |
770 | ||
771 | my $st = stat($part_path); | |
772 | ||
773 | die "error detecting block device '$part_path'\n" | |
774 | if !$st || !$st->mode || !S_ISBLK($st->mode) || !$st->rdev; | |
775 | ||
776 | my $major = PVE::Tools::dev_t_major($st->rdev); | |
777 | my $minor = PVE::Tools::dev_t_minor($st->rdev); | |
778 | my $partnum_path = "/sys/dev/block/$major:$minor/"; | |
779 | ||
780 | my $partnum; | |
781 | ||
782 | $partnum = file_read_firstline("${partnum_path}partition"); | |
783 | ||
784 | die "Partition does not exist\n" if !defined($partnum); | |
785 | ||
786 | #untaint and ensure it is a int | |
787 | if ($partnum =~ m/(\d+)/) { | |
788 | $partnum = $1; | |
789 | die "Partition number $partnum is invalid\n" if $partnum > 128; | |
790 | } else { | |
791 | die "Failed to get partition number\n"; | |
792 | } | |
793 | ||
794 | return $partnum; | |
795 | } | |
796 | ||
797 | sub get_blockdev { | |
798 | my ($part_path) = @_; | |
799 | ||
800 | my ($dev, $block_dev); | |
801 | if ($part_path =~ m|^/dev/(.*)$|) { | |
802 | $dev = $1; | |
803 | my $link = readlink "/sys/class/block/$dev"; | |
804 | $block_dev = $1 if $link =~ m|([^/]*)/$dev$|; | |
805 | } | |
806 | ||
807 | die "Can't parse parent device\n" if !defined($block_dev); | |
808 | die "No valid block device\n" if index($dev, $block_dev) == -1; | |
809 | ||
810 | $block_dev = "/dev/$block_dev"; | |
811 | die "Block device does not exists\n" if !(-b $block_dev); | |
812 | ||
813 | return $block_dev; | |
814 | } | |
815 | ||
816 | sub is_partition { | |
817 | my ($dev_path) = @_; | |
818 | ||
819 | return defined(eval { get_partnum($dev_path) }); | |
820 | } | |
821 | ||
822 | sub locked_disk_action { | |
823 | my ($sub) = @_; | |
824 | my $res = PVE::Tools::lock_file('/run/lock/pve-diskmanage.lck', undef, $sub); | |
825 | die $@ if $@; | |
826 | return $res; | |
827 | } | |
828 | ||
829 | sub assert_disk_unused { | |
830 | my ($dev) = @_; | |
831 | ||
832 | die "device '$dev' is already in use\n" if disk_is_used($dev); | |
833 | ||
834 | return undef; | |
835 | } | |
836 | ||
837 | sub append_partition { | |
838 | my ($dev, $size) = @_; | |
839 | ||
840 | my $devname = $dev; | |
841 | $devname =~ s|^/dev/||; | |
842 | ||
843 | my $newpartid = 1; | |
844 | dir_glob_foreach("/sys/block/$devname", qr/\Q$devname\E.*?(\d+)/, sub { | |
845 | my ($part, $partid) = @_; | |
846 | ||
847 | if ($partid >= $newpartid) { | |
848 | $newpartid = $partid + 1; | |
849 | } | |
850 | }); | |
851 | ||
852 | $size = PVE::Tools::convert_size($size, 'b' => 'mb'); | |
853 | ||
854 | run_command([ $SGDISK, '-n', "$newpartid:0:+${size}M", $dev ], | |
855 | errmsg => "error creating partition '$newpartid' on '$dev'"); | |
856 | ||
857 | my $partition; | |
858 | ||
859 | # loop again to detect the real partition device which does not always follow | |
860 | # a strict $devname$partition scheme like /dev/nvme0n1 -> /dev/nvme0n1p1 | |
861 | dir_glob_foreach("/sys/block/$devname", qr/\Q$devname\E.*$newpartid/, sub { | |
862 | my ($part) = @_; | |
863 | ||
864 | $partition = "/dev/$part"; | |
865 | }); | |
866 | ||
867 | return $partition; | |
868 | } | |
869 | ||
870 | # Check if a disk or any of its partitions has a holder. | |
871 | # Can also be called with a partition. | |
872 | # Expected to be called with a result of verify_blockdev_path(). | |
873 | sub has_holder { | |
874 | my ($devpath) = @_; | |
875 | ||
876 | my $dev = strip_dev($devpath); | |
877 | ||
878 | return $devpath if !dir_is_empty("/sys/class/block/${dev}/holders"); | |
879 | ||
880 | my $found; | |
881 | dir_glob_foreach("/sys/block/${dev}", "${dev}.+", sub { | |
882 | my ($part) = @_; | |
883 | $found = "/dev/${part}" if !dir_is_empty("/sys/class/block/${part}/holders"); | |
884 | }); | |
885 | ||
886 | return $found; | |
887 | } | |
888 | ||
889 | # Basic check if a disk or any of its partitions is mounted. | |
890 | # Can also be called with a partition. | |
891 | # Expected to be called with a result of verify_blockdev_path(). | |
892 | sub is_mounted { | |
893 | my ($devpath) = @_; | |
894 | ||
895 | my $mounted = mounted_blockdevs(); | |
896 | ||
897 | return $devpath if $mounted->{$devpath}; | |
898 | ||
899 | my $dev = strip_dev($devpath); | |
900 | ||
901 | my $found; | |
902 | dir_glob_foreach("/sys/block/${dev}", "${dev}.+", sub { | |
903 | my ($part) = @_; | |
904 | my $partpath = "/dev/${part}"; | |
905 | ||
906 | $found = $partpath if $mounted->{$partpath}; | |
907 | }); | |
908 | ||
909 | return $found; | |
910 | } | |
911 | ||
912 | # Currently only supports GPT-partitioned disks. | |
913 | sub change_parttype { | |
914 | my ($partpath, $parttype) = @_; | |
915 | ||
916 | my $err = "unable to change partition type for $partpath"; | |
917 | ||
918 | my $partnum = get_partnum($partpath); | |
919 | my $blockdev = get_blockdev($partpath); | |
920 | my $dev = strip_dev($blockdev); | |
921 | ||
922 | my $info = get_disks($dev, 1); | |
923 | die "$err - unable to get disk info for '$blockdev'\n" if !defined($info->{$dev}); | |
924 | die "$err - disk '$blockdev' is not GPT partitioned\n" if !$info->{$dev}->{gpt}; | |
925 | ||
926 | run_command(['sgdisk', "-t${partnum}:${parttype}", $blockdev], errmsg => $err); | |
927 | } | |
928 | ||
929 | # Wipes all labels and the first 200 MiB of a disk/partition (or the whole if it is smaller). | |
930 | # If called with a partition, also sets the partition type to 0x83 'Linux filesystem'. | |
931 | # Expected to be called with a result of verify_blockdev_path(). | |
932 | sub wipe_blockdev { | |
933 | my ($devpath) = @_; | |
934 | ||
935 | my $devname = basename($devpath); | |
936 | my $dev_size = PVE::Tools::file_get_contents("/sys/class/block/$devname/size"); | |
937 | ||
938 | ($dev_size) = $dev_size =~ m|(\d+)|; # untaint $dev_size | |
939 | die "Couldn't get the size of the device $devname\n" if !defined($dev_size); | |
940 | ||
941 | my $size = ($dev_size * 512 / 1024 / 1024); | |
942 | my $count = ($size < 200) ? $size : 200; | |
943 | ||
944 | my $to_wipe = []; | |
945 | dir_glob_foreach("/sys/class/block/${devname}", "${devname}.+", sub { | |
946 | my ($part) = @_; | |
947 | push $to_wipe->@*, "/dev/${part}" if -b "/dev/${part}"; | |
948 | }); | |
949 | ||
950 | if (scalar($to_wipe->@*) > 0) { | |
951 | print "found child partitions to wipe: ". join(', ', $to_wipe->@*) ."\n"; | |
952 | } | |
953 | push $to_wipe->@*, $devpath; # put actual device last | |
954 | ||
955 | print "wiping block device ${devpath}\n"; | |
956 | ||
957 | run_command(['wipefs', '--all', $to_wipe->@*], errmsg => "error wiping '${devpath}'"); | |
958 | ||
959 | run_command( | |
960 | ['dd', 'if=/dev/zero', "of=${devpath}", 'bs=1M', 'conv=fdatasync', "count=${count}"], | |
961 | errmsg => "error wiping '${devpath}'", | |
962 | ); | |
963 | ||
964 | if (is_partition($devpath)) { | |
965 | eval { change_parttype($devpath, '8300'); }; | |
966 | warn $@ if $@; | |
967 | } | |
968 | } | |
969 | ||
970 | 1; |