]> git.proxmox.com Git - pve-storage.git/blob - PVE/Diskmanage.pm
catch '---' in threshold output of sandisk ssds
[pve-storage.git] / PVE / Diskmanage.pm
1 package PVE::Diskmanage;
2
3 use strict;
4 use warnings;
5 use PVE::ProcFSTools;
6 use Data::Dumper;
7 use Cwd qw(abs_path);
8
9 use PVE::Tools qw(extract_param run_command file_get_contents file_read_firstline dir_glob_regex dir_glob_foreach trim);
10
11 my $SMARTCTL = "/usr/sbin/smartctl";
12 my $ZPOOL = "/sbin/zpool";
13 my $SGDISK = "/sbin/sgdisk";
14 my $PVS = "/sbin/pvs";
15 my $UDEVADM = "/bin/udevadm";
16
17 sub verify_blockdev_path {
18 my ($rel_path) = @_;
19
20 die "missing path" if !$rel_path;
21 my $path = abs_path($rel_path);
22 die "failed to get absolute path to $rel_path\n" if !$path;
23
24 die "got unusual device path '$path'\n" if $path !~ m|^/dev/(.*)$|;
25
26 $path = "/dev/$1"; # untaint
27
28 assert_blockdev($path);
29
30 return $path;
31 }
32
33 sub assert_blockdev {
34 my ($dev, $noerr) = @_;
35
36 if ($dev !~ m|^/dev/| || !(-b $dev)) {
37 return undef if $noerr;
38 die "not a valid block device\n";
39 }
40
41 return 1;
42 }
43
44 sub init_disk {
45 my ($disk, $uuid) = @_;
46
47 assert_blockdev($disk);
48
49 # we should already have checked if it is in use in the api call
50 # but we check again for safety
51 die "disk $disk is already in use\n" if disk_is_used($disk);
52
53 my $id = $uuid || 'R';
54 run_command([$SGDISK, $disk, '-U', $id]);
55 return 1;
56 }
57
58 sub disk_is_used {
59 my ($disk) = @_;
60
61 my $dev = $disk;
62 $dev =~ s|^/dev/||;
63
64 my $disklist = get_disks($dev, 1);
65
66 die "'$disk' is not a valid local disk\n" if !defined($disklist->{$dev});
67 return 1 if $disklist->{$dev}->{used};
68
69 return 0;
70 }
71
72 sub get_smart_data {
73 my ($disk, $healthonly) = @_;
74
75 assert_blockdev($disk);
76 my $smartdata = {};
77 my $type;
78
79 my $returncode = 0;
80
81 $disk =~ s/n\d+$//
82 if $disk =~ m!^/dev/nvme\d+n\d+$!;
83
84 my $cmd = [$SMARTCTL, '-H'];
85 push @$cmd, '-A', '-f', 'brief' if !$healthonly;
86 push @$cmd, $disk;
87
88 eval {
89 $returncode = run_command($cmd, noerr => 1, outfunc => sub{
90 my ($line) = @_;
91
92 # ATA SMART attributes, e.g.:
93 # ID# ATTRIBUTE_NAME FLAGS VALUE WORST THRESH FAIL RAW_VALUE
94 # 1 Raw_Read_Error_Rate POSR-K 100 100 000 - 0
95 #
96 # SAS and NVME disks, e.g.:
97 # Data Units Written: 5,584,952 [2.85 TB]
98 # Accumulated start-stop cycles: 34
99
100 if (defined($type) && $type eq 'ata' && $line =~ m/^([ \d]{2}\d)\s+(\S+)\s+(\S{6})\s+(\d+)\s+(\d+)\s+(\S+)\s+(\S+)\s+(.*)$/) {
101 my $entry = {};
102
103
104 $entry->{name} = $2 if defined $2;
105 $entry->{flags} = $3 if defined $3;
106 # the +0 makes a number out of the strings
107 $entry->{value} = $4+0 if defined $4;
108 $entry->{worst} = $5+0 if defined $5;
109 # some disks report the default threshold as --- instead of 000
110 if (defined($6) && $6 eq '---') {
111 $entry->{threshold} = 0;
112 } else {
113 $entry->{threshold} = $6+0 if defined $6;
114 }
115 $entry->{fail} = $7 if defined $7;
116 $entry->{raw} = $8 if defined $8;
117 $entry->{id} = $1 if defined $1;
118 push @{$smartdata->{attributes}}, $entry;
119 } elsif ($line =~ m/(?:Health Status|self\-assessment test result): (.*)$/ ) {
120 $smartdata->{health} = $1;
121 } elsif ($line =~ m/Vendor Specific SMART Attributes with Thresholds:/) {
122 $type = 'ata';
123 delete $smartdata->{text};
124 } elsif ($line =~ m/=== START OF (READ )?SMART DATA SECTION ===/) {
125 $type = 'text';
126 } elsif (defined($type) && $type eq 'text') {
127 $smartdata->{text} = '' if !defined $smartdata->{text};
128 $smartdata->{text} .= "$line\n";
129 } elsif ($line =~ m/SMART Disabled/) {
130 $smartdata->{health} = "SMART Disabled";
131 }
132 });
133 };
134 my $err = $@;
135
136 # bit 0 and 1 mark an severe smartctl error
137 # all others are for disk status, so ignore them
138 # see smartctl(8)
139 if ((defined($returncode) && ($returncode & 0b00000011)) || $err) {
140 die "Error getting S.M.A.R.T. data: Exit code: $returncode\n";
141 }
142
143 $smartdata->{type} = $type;
144
145 return $smartdata;
146 }
147
148 sub get_zfs_devices {
149 my $list = {};
150
151 # use zpool and parttype uuid,
152 # because log and cache do not have
153 # zfs type uuid
154 eval {
155 run_command([$ZPOOL, 'list', '-HPLv'], outfunc => sub {
156 my ($line) = @_;
157
158 if ($line =~ m|^\t([^\t]+)\t|) {
159 $list->{$1} = 1;
160 }
161 });
162 };
163
164 # only warn here,
165 # because maybe zfs tools are not installed
166 warn "$@\n" if $@;
167
168 my $applezfsuuid = "6a898cc3-1dd2-11b2-99a6-080020736631";
169 my $bsdzfsuuid = "516e7cba-6ecf-11d6-8ff8-00022d09712b";
170
171 dir_glob_foreach('/dev/disk/by-parttypeuuid', "($applezfsuuid|$bsdzfsuuid)\..+", sub {
172 my ($entry) = @_;
173 my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry");
174 $list->{$real_dev} = 1;
175 });
176
177 return $list;
178 }
179
180 sub get_lvm_devices {
181 my $list = {};
182 eval {
183 run_command([$PVS, '--noheadings', '--readonly', '-o', 'pv_name'], outfunc => sub{
184 my ($line) = @_;
185 $line = trim($line);
186 if ($line =~ m|^/dev/|) {
187 $list->{$line} = 1;
188 }
189 });
190 };
191
192 # if something goes wrong, we do not want
193 # to give up, but indicate an error has occured
194 warn "$@\n" if $@;
195
196 my $lvmuuid = "e6d6d379-f507-44c2-a23c-238f2a3df928";
197
198 dir_glob_foreach('/dev/disk/by-parttypeuuid', "$lvmuuid\..+", sub {
199 my ($entry) = @_;
200 my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry");
201 $list->{$real_dev} = 1;
202 });
203
204 return $list;
205 }
206
207 sub get_ceph_journals {
208 my $journalhash = {};
209
210 my $journal_uuid = '45b0969e-9b03-4f30-b4c6-b4b80ceff106';
211
212 dir_glob_foreach('/dev/disk/by-parttypeuuid', "$journal_uuid\..+", sub {
213 my ($entry) = @_;
214 my $real_dev = abs_path("/dev/disk/by-parttypeuuid/$entry");
215 $journalhash->{$real_dev} = 1;
216 });
217
218 return $journalhash;
219 }
220
221 sub get_udev_info {
222 my ($dev) = @_;
223
224 my $info = "";
225 my $data = {};
226 eval {
227 run_command([$UDEVADM, 'info', '-p', $dev, '--query', 'all'], outfunc => sub {
228 my ($line) = @_;
229 $info .= "$line\n";
230 });
231 };
232 warn $@ if $@;
233 return undef if !$info;
234
235 return undef if $info !~ m/^E: DEVTYPE=disk$/m;
236 return undef if $info =~ m/^E: ID_CDROM/m;
237
238 # we use this, because some disks are not simply in /dev
239 # e.g. /dev/cciss/c0d0
240 if ($info =~ m/^E: DEVNAME=(\S+)$/m) {
241 $data->{devpath} = $1;
242 }
243 return if !defined($data->{devpath});
244
245 $data->{serial} = 'unknown';
246 if ($info =~ m/^E: ID_SERIAL_SHORT=(\S+)$/m) {
247 $data->{serial} = $1;
248 }
249
250 $data->{gpt} = 0;
251 if ($info =~ m/^E: ID_PART_TABLE_TYPE=gpt$/m) {
252 $data->{gpt} = 1;
253 }
254
255 # detect SSD
256 $data->{rpm} = -1;
257 if ($info =~ m/^E: ID_ATA_ROTATION_RATE_RPM=(\d+)$/m) {
258 $data->{rpm} = $1;
259 }
260
261 if ($info =~ m/^E: ID_BUS=usb$/m) {
262 $data->{usb} = 1;
263 }
264
265 if ($info =~ m/^E: ID_MODEL=(.+)$/m) {
266 $data->{model} = $1;
267 }
268
269 $data->{wwn} = 'unknown';
270 if ($info =~ m/^E: ID_WWN=(.*)$/m) {
271 $data->{wwn} = $1;
272 }
273
274 return $data;
275 }
276
277 sub get_sysdir_info {
278 my ($sysdir) = @_;
279
280 return undef if ! -d "$sysdir/device";
281
282 my $data = {};
283
284 my $size = file_read_firstline("$sysdir/size");
285 return undef if !$size;
286
287 # linux always considers sectors to be 512 bytes,
288 # independently of real block size
289 $data->{size} = $size * 512;
290
291 # dir/queue/rotational should be 1 for hdd, 0 for ssd
292 $data->{rotational} = file_read_firstline("$sysdir/queue/rotational") // -1;
293
294 $data->{vendor} = file_read_firstline("$sysdir/device/vendor") || 'unknown';
295 $data->{model} = file_read_firstline("$sysdir/device/model") || 'unknown';
296
297 return $data;
298 }
299
300 sub get_wear_leveling_info {
301 my ($attributes, $model) = @_;
302
303 my $wearout;
304
305 my $vendormap = {
306 'kingston' => 231,
307 'samsung' => 177,
308 'intel' => 233,
309 'sandisk' => 233,
310 'default' => 233,
311 };
312
313 # find target attr id
314
315 my $attrid;
316
317 foreach my $vendor (keys %$vendormap) {
318 if ($model =~ m/$vendor/i) {
319 $attrid = $vendormap->{$vendor};
320 # found the attribute
321 last;
322 }
323 }
324
325 if (!$attrid) {
326 $attrid = $vendormap->{default};
327 }
328
329 foreach my $attr (@$attributes) {
330 next if $attr->{id} != $attrid;
331 $wearout = $attr->{value};
332 last;
333 }
334
335 return $wearout;
336 }
337
338 sub dir_is_empty {
339 my ($dir) = @_;
340
341 my $dh = IO::Dir->new ($dir);
342 return 1 if !$dh;
343
344 while (defined(my $tmp = $dh->read)) {
345 next if $tmp eq '.' || $tmp eq '..';
346 $dh->close;
347 return 0;
348 }
349 $dh->close;
350 return 1;
351 }
352
353 sub get_disks {
354 my ($disk, $nosmart) = @_;
355 my $disklist = {};
356
357 my $mounted = {};
358
359 my $mounts = PVE::ProcFSTools::parse_proc_mounts();
360
361 foreach my $mount (@$mounts) {
362 next if $mount->[0] !~ m|^/dev/|;
363 $mounted->{abs_path($mount->[0])} = $mount->[1];
364 };
365
366 my $dev_is_mounted = sub {
367 my ($dev) = @_;
368 return $mounted->{$dev};
369 };
370
371 my $journalhash = get_ceph_journals();
372
373 my $zfslist = get_zfs_devices();
374
375 my $lvmlist = get_lvm_devices();
376
377 dir_glob_foreach('/sys/block', '.*', sub {
378 my ($dev) = @_;
379 return if defined($disk) && $disk ne $dev;
380 # whitelisting following devices
381 # hdX: ide block device
382 # sdX: sd block device
383 # vdX: virtual block device
384 # xvdX: xen virtual block device
385 # nvmeXnY: nvme devices
386 # cciss!cXnY: cciss devices
387 return if $dev !~ m/^(h|s|x?v)d[a-z]+$/ &&
388 $dev !~ m/^nvme\d+n\d+$/ &&
389 $dev !~ m/^cciss\!c\d+d\d+$/;
390
391 my $data = get_udev_info("/sys/block/$dev");
392 return if !defined($data);
393 my $devpath = $data->{devpath};
394
395 my $sysdir = "/sys/block/$dev";
396
397 # we do not want iscsi devices
398 return if -l $sysdir && readlink($sysdir) =~ m|host[^/]*/session[^/]*|;
399
400 my $sysdata = get_sysdir_info($sysdir);
401 return if !defined($sysdata);
402
403 my $type = 'unknown';
404
405 if ($sysdata->{rotational} == 0) {
406 $type = 'ssd';
407 $data->{rpm} = 0;
408 } elsif ($sysdata->{rotational} == 1) {
409 if ($data->{rpm} != -1) {
410 $type = 'hdd';
411 } elsif ($data->{usb}) {
412 $type = 'usb';
413 $data->{rpm} = 0;
414 }
415 }
416
417 my $health = 'UNKNOWN';
418 my $wearout = 'N/A';
419
420 if (!$nosmart) {
421 eval {
422 my $smartdata = get_smart_data($devpath, ($type ne 'ssd'));
423 $health = $smartdata->{health} if $smartdata->{health};
424
425 if ($type eq 'ssd') {
426 # if we have an ssd we try to get the wearout indicator
427 my $wearval = get_wear_leveling_info($smartdata->{attributes}, $data->{model} || $sysdir->{model});
428 $wearout = $wearval if $wearval;
429 }
430 };
431 }
432
433 my $used;
434
435 $used = 'LVM' if $lvmlist->{$devpath};
436
437 $used = 'mounted' if &$dev_is_mounted($devpath);
438
439 $used = 'ZFS' if $zfslist->{$devpath};
440
441 $disklist->{$dev} = {
442 vendor => $sysdata->{vendor},
443 model => $data->{model} || $sysdata->{model},
444 size => $sysdata->{size},
445 serial => $data->{serial},
446 gpt => $data->{gpt},
447 rpm => $data->{rpm},
448 type => $type,
449 wwn => $data->{wwn},
450 health => $health,
451 devpath => $devpath,
452 wearout => $wearout,
453 };
454
455 my $osdid = -1;
456
457 my $journal_count = 0;
458
459 my $found_partitions;
460 my $found_lvm;
461 my $found_mountpoints;
462 my $found_zfs;
463 my $found_dm;
464 my $partpath = $devpath;
465
466 # remove part after last / to
467 # get the base path for the partitions
468 # e.g. from /dev/cciss/c0d0 get /dev/cciss
469 $partpath =~ s/\/[^\/]+$//;
470
471 dir_glob_foreach("$sysdir", "$dev.+", sub {
472 my ($part) = @_;
473
474 $found_partitions = 1;
475
476 if (my $mp = &$dev_is_mounted("$partpath/$part")) {
477 $found_mountpoints = 1;
478 if ($mp =~ m|^/var/lib/ceph/osd/ceph-(\d+)$|) {
479 $osdid = $1;
480 }
481 }
482
483 if ($lvmlist->{"$partpath/$part"}) {
484 $found_lvm = 1;
485 }
486
487 if ($zfslist->{"$partpath/$part"}) {
488 $found_zfs = 1;
489 }
490
491 $journal_count++ if $journalhash->{"$partpath/$part"};
492
493 if (!dir_is_empty("$sysdir/$part/holders") && !$found_lvm) {
494 $found_dm = 1;
495 }
496 });
497
498 $used = 'mounted' if $found_mountpoints && !$used;
499 $used = 'LVM' if $found_lvm && !$used;
500 $used = 'ZFS' if $found_zfs && !$used;
501 $used = 'Device Mapper' if $found_dm && !$used;
502 $used = 'partitions' if $found_partitions && !$used;
503
504 # multipath, software raid, etc.
505 # this check comes in last, to show more specific info
506 # if we have it
507 $used = 'Device Mapper' if !$used && !dir_is_empty("$sysdir/holders");
508
509 $disklist->{$dev}->{used} = $used if $used;
510 $disklist->{$dev}->{osdid} = $osdid;
511 $disklist->{$dev}->{journals} = $journal_count;
512 });
513
514 return $disklist;
515
516 }
517
518 1;