1xfs_db(8) System Manager's Manual xfs_db(8)
2
6 xfs_db - debug an XFS filesystem
7
9 xfs_db [ -c cmd ] ... [ -i|r|x|F ] [ -f ] [ -l logdev ] [ -p progname ]
10 device
11 xfs_db -V
12
14 xfs_db is used to examine an XFS filesystem. Under rare circumstances
15 it can also be used to modify an XFS filesystem, but that task is nor‐
16 mally left to xfs_repair(8) or to scripts such as xfs_admin(8) that run
17 xfs_db.
18
20 -c cmd xfs_db commands may be run interactively (the default) or as ar‐
21 guments on the command line. Multiple -c arguments may be given.
22 The commands are run in the sequence given, then the program ex‐
23 its.
24 -f Specifies that the filesystem image to be processed is stored in
26 a regular file at device (see the mkfs.xfs(8) -d file option).
27 This might happen if an image copy of a filesystem has been made
28 into an ordinary file with xfs_copy(8).
29 -F Specifies that we want to continue even if the superblock magic
31 is not correct. For use in xfs_metadump.
32 -i Allows execution on a mounted filesystem, provided it is mounted
34 read-only. Useful for shell scripts which must only operate on
35 filesystems in a guaranteed consistent state (either unmounted
36 or mounted read-only). These semantics are slightly different to
37 that of the -r option.
38 -l logdev
40 Specifies the device where the filesystems external log resides.
41 Only for those filesystems which use an external log. See the
42 mkfs.xfs(8) -l option, and refer to xfs(5) for a detailed de‐
43 scription of the XFS log.
44 -p progname
46 Set the program name to progname for prompts and some error mes‐
47 sages, the default value is xfs_db.
48 -r Open device or filename read-only. This option is required if
50 the filesystem is mounted. It is only necessary to omit this
51 flag if a command that changes data (write, blocktrash, crc) is
52 to be used.
53 -x Specifies expert mode. This enables the (write, blocktrash, crc
55 invalidate/revalidate) commands.
56 -V Prints the version number and exits.
58
60 xfs_db commands can be broken up into two classes. Most commands are
61 for the navigation and display of data structures in the filesystem.
62 Other commands are for scanning the filesystem in some way.
63 Commands which are used to navigate the filesystem structure take argu‐
65 ments which reflect the names of filesystem structure fields. There
66 can be multiple field names separated by dots when the underlying
67 structures are nested, as in C. The field names can be indexed (as an
68 array index) if the underlying field is an array. The array indices
69 can be specified as a range, two numbers separated by a dash.
70 xfs_db maintains a current address in the filesystem. The granularity
72 of the address is a filesystem structure. This can be a filesystem
73 block, an inode or quota (smaller than a filesystem block), or a direc‐
74 tory block (could be larger than a filesystem block). There are a va‐
75 riety of commands to set the current address. Associated with the cur‐
76 rent address is the current data type, which is the structural type of
77 this data. Commands which follow the structure of the filesystem al‐
78 ways set the type as well as the address. Commands which examine
79 pieces of an individual file (inode) need the current inode to be set,
80 this is done with the inode command.
81 The current address/type information is actually maintained in a stack
83 that can be explicitly manipulated with the push, pop, and stack com‐
84 mands. This allows for easy examination of a nested filesystem struc‐
85 ture. Also, the last several locations visited are stored in a ring
86 buffer which can be manipulated with the forward, back, and ring com‐
87 mands.
88 XFS filesystems are divided into a small number of allocation groups.
90 xfs_db maintains a notion of the current allocation group which is ma‐
91 nipulated by some commands. The initial allocation group is 0.
92
94 Many commands have extensive online help. Use the help command for more
95 details on any command.
96 a See the addr command.
98 ablock filoff
100 Set current address to the offset filoff (a filesystem block
101 number) in the attribute area of the current inode.
102 addr [field-expression]
104 Set current address to the value of the field-expression. This
105 is used to "follow" a reference in one structure to the object
106 being referred to. If no argument is given, the current address
107 is printed.
108 agf [agno]
110 Set current address to the AGF block for allocation group agno.
111 If no argument is given, use the current allocation group.
112 agfl [agno]
114 Set current address to the AGFL block for allocation group agno.
115 If no argument is given, use the current allocation group.
116 agi [agno]
118 Set current address to the AGI block for allocation group agno.
119 If no argument is given, use the current allocation group.
120 agresv [agno]
122 Displays the length, free block count, per-AG reservation size,
123 and per-AG reservation usage for a given AG. If no argument is
124 given, display information for all AGs.
125 attr_remove [-r|-u|-s] [-n] name
127 Remove the specified extended attribute from the current file.
128 -r Sets the attribute in the root namespace. Only one name‐
130 space option can be specified.
131 -u Sets the attribute in the user namespace. Only one name‐
133 space option can be specified.
134 -s Sets the attribute in the secure namespace. Only one
136 namespace option can be specified.
137 -n Do not enable 'noattr2' mode on V4 filesystems.
139 attr_set [-r|-u|-s] [-n] [-R|-C] [-v namelen] name
141 Sets an extended attribute on the current file with the given
142 name.
143 -r Sets the attribute in the root namespace. Only one name‐
145 space option can be specified.
146 -u Sets the attribute in the user namespace. Only one name‐
148 space option can be specified.
149 -s Sets the attribute in the secure namespace. Only one
151 namespace option can be specified.
152 -n Do not enable 'noattr2' mode on V4 filesystems.
154 -R Replace the attribute. The command will fail if the at‐
156 tribute does not already exist.
157 -C Create the attribute. The command will fail if the at‐
159 tribute already exists.
160 -v Set the attribute value to a string of this length con‐
162 taining the letter 'v'.
163 b See the back command.
165 back Move to the previous location in the position ring.
167 blockfree
169 Free block usage information collected by the last execution of
170 the blockget command. This must be done before another blockget
171 command can be given, presumably with different arguments than
172 the previous one.
173 blockget [-npvs] [-b bno] ... [-i ino] ...
175 Get block usage and check filesystem consistency. The informa‐
176 tion is saved for use by a subsequent blockuse, ncheck, or
177 blocktrash command.
178 -b is used to specify filesystem block numbers about which
180 verbose information should be printed.
181 -i is used to specify inode numbers about which verbose in‐
183 formation should be printed.
184 -n is used to save pathnames for inodes visited, this is
186 used to support the xfs_ncheck(8) command. It also means
187 that pathnames will be printed for inodes that have prob‐
188 lems. This option uses a lot of memory so is not enabled
189 by default.
190 -p causes error messages to be prefixed with the filesystem
192 name being processed. This is useful if several copies of
193 xfs_db are run in parallel.
194 -s restricts output to severe errors only. This is useful if
196 the output is too long otherwise.
197 -v enables verbose output. Messages will be printed for ev‐
199 ery block and inode processed.
200 blocktrash [-z] [-o offset] [-n count] [-x min] [-y max] [-s seed]
202 [-0|1|2|3] [-t type] ...
203 Trash randomly selected filesystem metadata blocks. Trashing
204 occurs to randomly selected bits in the chosen blocks. This
205 command is available only in debugging versions of xfs_db. It
206 is useful for testing xfs_repair(8).
207 -0 | -1 | -2 | -3
209 These are used to set the operating mode for blocktrash.
210 Only one can be used: -0 changed bits are cleared; -1
211 changed bits are set; -2 changed bits are inverted; -3
212 changed bits are randomized.
213 -n supplies the count of block-trashings to perform (default
215 1).
216 -o supplies the bit offset at which to start trashing the
218 block. If the value is preceded by a '+', the trashing
219 will start at a randomly chosen offset that is larger
220 than the value supplied. The default is to randomly
221 choose an offset anywhere in the block.
222 -s supplies a seed to the random processing.
224 -t gives a type of blocks to be selected for trashing. Mul‐
226 tiple -t options may be given. If no -t options are given
227 then all metadata types can be trashed.
228 -x sets the minimum size of bit range to be trashed. The de‐
230 fault value is 1.
231 -y sets the maximum size of bit range to be trashed. The de‐
233 fault value is 1024.
234 -z trashes the block at the top of the stack. It is not
236 necessary to run blockget if this option is supplied.
237 blockuse [-n] [-c count]
239 Print usage for current filesystem block(s). For each block,
240 the type and (if any) inode are printed.
241 -c specifies a count of blocks to process. The default value
243 is 1 (the current block only).
244 -n specifies that file names should be printed. The prior
246 blockget command must have also specified the -n option.
247 bmap [-a] [-d] [block [len]]
249 Show the block map for the current inode. The map display can
250 be restricted to an area of the file with the block and len ar‐
251 guments. If block is given and len is omitted then 1 is assumed
252 for len.
253 The -a and -d options are used to select the attribute or data
255 area of the inode, if neither option is given then both areas
256 are shown.
257 btdump [-a] [-i]
259 If the cursor points to a btree node, dump the btree from that
260 block downward. If instead the cursor points to an inode, dump
261 the data fork block mapping btree if there is one. If the cur‐
262 sor points to a directory or extended attribute btree node, dump
263 that. By default, only records stored in the btree are dumped.
264 -a If the cursor points at an inode, dump the extended at‐
266 tribute block mapping btree, if present.
267 -i Dump all keys and pointers in intermediate btree nodes,
269 and all records in leaf btree nodes.
270 btheight [-b blksz] [-n recs] [-w max|-w min] btree types...
272 For a given number of btree records and a btree type, report the
273 number of records and blocks for each level of the btree, and
274 the total number of blocks. The btree type must be given after
275 the options.
276 A raw btree geometry can be provided in the format
278 "record_bytes:key_bytes:ptr_bytes:header_type", where
279 header_type is one of "short", "long", "shortcrc", or "longcrc".
280 The supported btree types are: bnobt, cntbt, inobt, finobt,
282 bmapbt, refcountbt, and rmapbt.
283 Options are as follows:
285 -b is used to override the btree block size. The default is
287 the filesystem block size.
288 -n is used to specify the number of records to store. This
290 argument is required.
291 -w max
293 shows only the best case scenario, which is when the
294 btree blocks are maximally loaded.
295 -w min
297 shows only the worst case scenario, which is when the
298 btree blocks are half full.
299 check See the blockget command.
301 convert type number [type number] ... type
303 Convert from one address form to another. The known types, with
304 alternate names, are:
305 agblock or agbno (filesystem block within an allocation
306 group)
307 agino or aginode (inode number within an allocation group)
308 agnumber or agno (allocation group number)
309 bboff or daddroff (byte offset in a daddr)
310 blkoff or fsboff or agboff (byte offset in a agblock or fs‐
311 block)
312 byte or fsbyte (byte address in filesystem)
313 daddr or bb (disk address, 512-byte blocks)
314 fsblock or fsb or fsbno (filesystem block, see the fsblock
315 command)
316 ino or inode (inode number)
317 inoidx or offset (index of inode in filesystem block)
318 inooff or inodeoff (byte offset in inode)
319 Only conversions that "make sense" are allowed. The compound
321 form (with more than three arguments) is useful for conversions
322 such as convert agno ag agbno agb fsblock.
323 crc [-i|-r|-v]
325 Invalidates, revalidates, or validates the CRC (checksum) field
326 of the current structure, if it has one. This command is avail‐
327 able only on CRC-enabled filesystems. With no argument, valida‐
328 tion is performed. Each command will display the resulting CRC
329 value and state.
330 -i Invalidate the structure's CRC value (incrementing it by
332 one), and write it to disk.
333 -r Recalculate the current structure's correct CRC value,
335 and write it to disk.
336 -v Validate and display the current value and state of the
338 structure's CRC.
339 daddr [d]
341 Set current address to the daddr (512 byte block) given by d.
342 If no value for d is given, the current address is printed, ex‐
343 pressed as a daddr. The type is set to data (uninterpreted).
344 dblock filoff
346 Set current address to the offset filoff (a filesystem block
347 number) in the data area of the current inode.
348 debug [flagbits]
350 Set debug option bits. These are used for debugging xfs_db. If
351 no value is given for flagbits, print the current debug option
352 bits. These are for the use of the implementor.
353 dquot [-g|-p|-u] id
355 Set current address to a group, project or user quota block for
356 the given ID. Defaults to user quota.
357 echo [arg] ...
359 Echo the arguments to the output.
360 f See the forward command.
362 forward
364 Move forward to the next entry in the position ring.
365 frag [-adflqRrv]
367 Get file fragmentation data. This prints information about frag‐
368 mentation of file data in the filesystem (as opposed to fragmen‐
369 tation of freespace, for which see the freesp command). Every
370 file in the filesystem is examined to see how far from ideal its
371 extent mappings are. A summary is printed giving the totals.
372 -v sets verbosity, every inode has information printed for
374 it. The remaining options select which inodes and ex‐
375 tents are examined. If no options are given then all are
376 assumed set, otherwise just those given are enabled.
377 -a enables processing of attribute data.
379 -d enables processing of directory data.
381 -f enables processing of regular file data.
383 -l enables processing of symbolic link data.
385 -q enables processing of quota file data.
387 -R enables processing of realtime control file data.
389 -r enables processing of realtime file data.
391 freesp [-bcds] [-A alignment] [-a ag] ... [-e i] [-h h1] ... [-m m]
393 Summarize free space for the filesystem. The free blocks are ex‐
394 amined and totalled, and displayed in the form of a histogram,
395 with a count of extents in each range of free extent sizes.
396 -A reports only free extents with starting blocks aligned to
398 alignment blocks.
399 -a adds ag to the list of allocation groups to be processed.
401 If no -a options are given then all allocation groups are
402 processed.
403 -b specifies that the histogram buckets are binary-sized,
405 with the starting sizes being the powers of 2.
406 -c specifies that freesp will search the by-size (cnt) space
408 Btree instead of the default by-block (bno) space Btree.
409 -d specifies that every free extent will be displayed.
411 -e specifies that the histogram buckets are equal-sized,
413 with the size specified as i.
414 -h specifies a starting block number for a histogram bucket
416 as h1. Multiple -h's are given to specify the complete
417 set of buckets.
418 -m specifies that the histogram starting block numbers are
420 powers of m. This is the general case of -b.
421 -s specifies that a final summary of total free extents,
423 free blocks, and the average free extent size is printed.
424 fsb See the fsblock command.
426 fsblock [fsb]
428 Set current address to the fsblock value given by fsb. If no
429 value for fsb is given the current address is printed, expressed
430 as an fsb. The type is set to data (uninterpreted). XFS
431 filesystem block numbers are computed ((agno << agshift) | ag‐
432 block) where agshift depends on the size of an allocation group.
433 Use the convert command to convert to and from this form. Block
434 numbers given for file blocks (for instance from the bmap com‐
435 mand) are in this form.
436 fsmap [ start ] [ end ]
438 Prints the mapping of disk blocks used by an XFS filesystem.
439 The map lists each extent used by files, allocation group meta‐
440 data, journalling logs, and static filesystem metadata, as well
441 as any regions that are unused. All blocks, offsets, and
442 lengths are specified in units of 512-byte blocks, no matter
443 what the filesystem's block size is. The optional start and end
444 arguments can be used to constrain the output to a particular
445 range of disk blocks.
446 fuzz [-c] [-d] field action
448 Write garbage into a specific structure field on disk. Expert
449 mode must be enabled to use this command. The operation happens
450 immediately; there is no buffering.
451 The fuzz command can take the following actions against a field:
453 zeroes
455 Clears all bits in the field.
456 ones
458 Sets all bits in the field.
459 firstbit
461 Flips the first bit in the field. For a scalar value,
462 this is the highest bit.
463 middlebit
465 Flips the middle bit in the field.
466 lastbit
468 Flips the last bit in the field. For a scalar value,
469 this is the lowest bit.
470 add Adds a small value to a scalar field.
472 sub Subtracts a small value from a scalar field.
474 random
476 Randomizes the contents of the field.
477 The following switches affect the write behavior:
479 -c Skip write verifiers and CRC recalculation; allows in‐
481 valid data to be written to disk.
482 -d Skip write verifiers but perform CRC recalculation; al‐
484 lows invalid data to be written to disk to test detection
485 of invalid data.
486 hash string
488 Prints the hash value of string using the hash function of the
489 XFS directory and attribute implementation.
490 help [command]
492 Print help for one or all commands.
493 info Displays selected geometry information about the filesystem.
495 The output will have the same format that mkfs.xfs(8) prints
496 when creating a filesystem or xfs_info(8) prints when querying a
497 filesystem.
498 inode [inode#]
500 Set the current inode number. If no inode# is given, print the
501 current inode number.
502 label [label]
504 Set the filesystem label. The filesystem label can be used by
505 mount(8) instead of using a device special file. The maximum
506 length of an XFS label is 12 characters - use of a longer label
507 will result in truncation and a warning will be issued. If no
508 label is given, the current filesystem label is printed.
509 log [stop | start filename]
511 Start logging output to filename, stop logging, or print the
512 current logging status.
513 logformat [-c cycle] [-s sunit]
515 Reformats the log to the specified log cycle and log stripe
516 unit. This has the effect of clearing the log destructively.
517 If the log cycle is not specified, the log is reformatted to the
518 current cycle. If the log stripe unit is not specified, the
519 stripe unit from the filesystem superblock is used.
520 logres Print transaction reservation size information for each transac‐
522 tion type. This makes it easier to find discrepancies in the
523 reservation calculations between xfsprogs and the kernel, which
524 will help when diagnosing minimum log size calculation errors.
525 ls [-i] [paths]...
527 List the contents of a directory. If a path resolves to a di‐
528 rectory, the directory will be listed. If no paths are supplied
529 and the IO cursor points at a directory inode, the contents of
530 that directory will be listed.
531 The output format is: directory cookie, inode number, file type,
533 hash, name length, name.
534 -i Resolve each of the given paths to an inode number and
536 print that number. If no paths are given and the IO cur‐
537 sor points to an inode, print the inode number.
538 metadump [-egow] filename
540 Dumps metadata to a file. See xfs_metadump(8) for more informa‐
541 tion.
542 ncheck [-s] [-i ino] ...
544 Print name-inode pairs. A blockget -n command must be run first
545 to gather the information.
546 -i specifies an inode number to be printed. If no -i options
548 are given then all inodes are printed.
549 -s specifies that only setuid and setgid files are printed.
551 p See the print command.
553 path dir_path
555 Walk the directory tree to an inode using the supplied path.
556 Absolute and relative paths are supported.
557 pop Pop location from the stack.
559 print [field-expression] ...
561 Print field values. If no argument is given, print all fields
562 in the current structure.
563 push [command]
565 Push location to the stack. If command is supplied, set the cur‐
566 rent location to the results of command after pushing the old
567 location.
568 q See the quit command.
570 quit Exit xfs_db.
572 ring [index]
574 Show position ring (if no index argument is given), or move to a
575 specific entry in the position ring given by index.
576 sb [agno]
578 Set current address to SB header in allocation group agno. If
579 no agno is given, use the current allocation group number.
580 source source-file
582 Process commands from source-file. source commands can be
583 nested.
584 stack View the location stack.
586 type [type]
588 Set the current data type to type. If no argument is given,
589 show the current data type. The possible data types are: agf,
590 agfl, agi, attr, bmapbta, bmapbtd, bnobt, cntbt, data, dir,
591 dir2, dqblk, inobt, inode, log, refcntbt, rmapbt, rtbitmap, rt‐
592 summary, sb, symlink and text. See the TYPES section below for
593 more information on these data types.
594 timelimit [OPTIONS]
596 Print the minimum and maximum supported values for inode time‐
597 stamps, quota expiration timers, and quota grace periods sup‐
598 ported by this filesystem. Options include:
599 --bigtime
601 Print the time limits of an XFS filesystem with the big‐
602 time feature enabled.
603 --classic
605 Print the time limits of a classic XFS filesystem.
606 --compact
608 Print all limits as raw values on a single line.
609 --pretty
611 Print the timestamps in the current locale's date and
612 time format instead of raw seconds since the Unix epoch.
613 uuid [uuid | generate | rewrite | restore]
615 Set the filesystem universally unique identifier (UUID). The
616 filesystem UUID can be used by mount(8) instead of using a de‐
617 vice special file. The uuid can be set directly to the desired
618 UUID, or it can be automatically generated using the generate
619 option. These options will both write the UUID into every copy
620 of the superblock in the filesystem. On a CRC-enabled filesys‐
621 tem, this will set an incompatible superblock flag, and the
622 filesystem will not be mountable with older kernels. This can
623 be reverted with the restore option, which will copy the origi‐
624 nal UUID back into place and clear the incompatible flag as
625 needed. rewrite copies the current UUID from the primary su‐
626 perblock to all secondary copies of the superblock. If no argu‐
627 ment is given, the current filesystem UUID is printed.
628 version [feature | versionnum features2]
630 Enable selected features for a filesystem (certain features can
631 be enabled on an unmounted filesystem, after mkfs.xfs(8) has
632 created the filesystem). Support for unwritten extents can be
633 enabled using the extflg option. Support for version 2 log for‐
634 mat can be enabled using the log2 option. Support for extended
635 attributes can be enabled using the attr1 or attr2 option. Once
636 enabled, extended attributes cannot be disabled, but the user
637 may toggle between attr1 and attr2 at will (older kernels may
638 not support the newer version).
639 If no argument is given, the current version and feature bits
641 are printed. With one argument, this command will write the up‐
642 dated version number into every copy of the superblock in the
643 filesystem. If two arguments are given, they will be used as
644 numeric values for the versionnum and features2 bits respec‐
645 tively, and their string equivalent reported (but no modifica‐
646 tions are made).
647 write [-c|-d] [field value] ...
649 Write a value to disk. Specific fields can be set in structures
650 (struct mode), or a block can be set to data values (data mode),
651 or a block can be set to string values (string mode, for symlink
652 blocks). The operation happens immediately: there is no buffer‐
653 ing.
654 Struct mode is in effect when the current type is structural,
656 i.e. not data. For struct mode, the syntax is "write field
657 value".
658 Data mode is in effect when the current type is data. In this
660 case the contents of the block can be shifted or rotated left or
661 right, or filled with a sequence, a constant value, or a random
662 value. In this mode write with no arguments gives more informa‐
663 tion on the allowed commands.
664 -c Skip write verifiers and CRC recalculation; allows in‐
666 valid data to be written to disk.
667 -d Skip write verifiers but perform CRC recalculation. This
669 allows invalid data to be written to disk to test detec‐
670 tion of invalid data. (This is not possible for some
671 types.)
672
674 This section gives the fields in each structure type and their mean‐
675 ings. Note that some types of block cover multiple actual structures,
676 for instance directory blocks.
677 agf The AGF block is the header for block allocation information;
679 it is in the second 512-byte block of each allocation group.
680 The following fields are defined:
681 magicnum AGF block magic number, 0x58414746 ('XAGF').
682 versionnum version number, currently 1.
683 seqno sequence number starting from 0.
684 length size in filesystem blocks of the allocation
685 group. All allocation groups except the last
686 one of the filesystem have the superblock's
687 agblocks value here.
688 bnoroot block number of the root of the Btree holding
689 free space information sorted by block num‐
690 ber.
691 cntroot block number of the root of the Btree holding
692 free space information sorted by block count.
693 bnolevel number of levels in the by-block-number
694 Btree.
695 cntlevel number of levels in the by-block-count Btree.
696 flfirst index into the AGFL block of the first active
697 entry.
698 fllast index into the AGFL block of the last active
699 entry.
700 flcount count of active entries in the AGFL block.
701 freeblks count of blocks represented in the freespace
702 Btrees.
703 longest longest free space represented in the
704 freespace Btrees.
705 btreeblks number of blocks held in the AGF Btrees.
706 agfl The AGFL block contains block numbers for use of the block
708 allocator; it is in the fourth 512-byte block of each alloca‐
709 tion group. Each entry in the active list is a block number
710 within the allocation group that can be used for any purpose
711 if space runs low. The AGF block fields flfirst, fllast, and
712 flcount designate which entries are currently active. Entry
713 space is allocated in a circular manner within the AGFL
714 block. Fields defined:
715 bno array of all block numbers. Even those which
716 are not active are printed.
717 agi The AGI block is the header for inode allocation information;
719 it is in the third 512-byte block of each allocation group.
720 Fields defined:
721 magicnum AGI block magic number, 0x58414749 ('XAGI').
722 versionnum version number, currently 1.
723 seqno sequence number starting from 0.
724 length size in filesystem blocks of the allocation
725 group.
726 count count of inodes allocated.
727 root block number of the root of the Btree holding
728 inode allocation information.
729 level number of levels in the inode allocation
730 Btree.
731 freecount count of allocated inodes that are not in
732 use.
733 newino last inode number allocated.
734 dirino unused.
735 unlinked an array of inode numbers within the alloca‐
736 tion group. The entries in the AGI block are
737 the heads of lists which run through the in‐
738 ode next_unlinked field. These inodes are to
739 be unlinked the next time the filesystem is
740 mounted.
741 attr An attribute fork is organized as a Btree with the actual
743 data embedded in the leaf blocks. The root of the Btree is
744 found in block 0 of the fork. The index (sort order) of the
745 Btree is the hash value of the attribute name. All the
746 blocks contain a blkinfo structure at the beginning, see type
747 dir for a description. Nonleaf blocks are identical in format
748 to those for version 1 and version 2 directories, see type
749 dir for a description. Leaf blocks can refer to "local" or
750 "remote" attribute values. Local values are stored directly
751 in the leaf block. Leaf blocks contain the following fields:
752 hdr header containing a blkinfo structure info
753 (magic number 0xfbee), a count of active en‐
754 tries, usedbytes total bytes of names and
755 values, the firstused byte in the name area,
756 holes set if the block needs compaction, and
757 array freemap as for dir leaf blocks.
758 entries array of structures containing a hashval,
759 nameidx (index into the block of the name),
760 and flags incomplete, root, and local.
761 nvlist array of structures describing the attribute
762 names and values. Fields always present: val‐
763 uelen (length of value in bytes), namelen,
764 and name. Fields present for local values:
765 value (value string). Fields present for re‐
766 mote values: valueblk (fork block number of
767 containing the value).
768 Remote values are stored in an independent block in the at‐
770 tribute fork. Prior to v5, value blocks had no structure,
771 but in v5 they acquired a header structure with the following
772 fields:
773 magic attr3 remote block magic number, 0x5841524d
774 ('XARM').
775 offset Byte offset of this data block within the
776 overall attribute value.
777 bytes Number of bytes stored in this block.
778 crc Checksum of the attribute block contents.
779 uuid Filesystem UUID.
780 owner Inode that owns this attribute value.
781 bno Block offset of this block within the inode's
782 attribute fork.
783 lsn Log serial number of the last time this block
784 was logged.
785 data The attribute value data.
786 bmapbt Files with many extents in their data or attribute fork will
788 have the extents described by the contents of a Btree for
789 that fork, instead of being stored directly in the inode.
790 Each bmap Btree starts with a root block contained within the
791 inode. The other levels of the Btree are stored in filesys‐
792 tem blocks. The blocks are linked to sibling left and right
793 blocks at each level, as well as by pointers from parent to
794 child blocks. Each block contains the following fields:
795 magic bmap Btree block magic number, 0x424d4150
796 ('BMAP').
797 level level of this block above the leaf level.
798 numrecs number of records or keys in the block.
799 leftsib left (logically lower) sibling block, 0 if
800 none.
801 rightsib right (logically higher) sibling block, 0 if
802 none.
803 recs [leaf blocks only] array of extent records.
804 Each record contains startoff, startblock,
805 blockcount, and extentflag (1 if the extent
806 is unwritten).
807 keys [non-leaf blocks only] array of key records.
808 These are the first key value of each block
809 in the level below this one. Each record con‐
810 tains startoff.
811 ptrs [non-leaf blocks only] array of child block
812 pointers. Each pointer is a filesystem block
813 number to the next level in the Btree.
814 bnobt There is one set of filesystem blocks forming the by-block-
816 number allocation Btree for each allocation group. The root
817 block of this Btree is designated by the bnoroot field in the
818 corresponding AGF block. The blocks are linked to sibling
819 left and right blocks at each level, as well as by pointers
820 from parent to child blocks. Each block has the following
821 fields:
822 magic BNOBT block magic number, 0x41425442
823 ('ABTB').
824 level level number of this block, 0 is a leaf.
825 numrecs number of data entries in the block.
826 leftsib left (logically lower) sibling block, 0 if
827 none.
828 rightsib right (logically higher) sibling block, 0 if
829 none.
830 recs [leaf blocks only] array of freespace
831 records. Each record contains startblock and
832 blockcount.
833 keys [non-leaf blocks only] array of key records.
834 These are the first value of each block in
835 the level below this one. Each record con‐
836 tains startblock and blockcount.
837 ptrs [non-leaf blocks only] array of child block
838 pointers. Each pointer is a block number
839 within the allocation group to the next level
840 in the Btree.
841 cntbt There is one set of filesystem blocks forming the by-block-
843 count allocation Btree for each allocation group. The root
844 block of this Btree is designated by the cntroot field in the
845 corresponding AGF block. The blocks are linked to sibling
846 left and right blocks at each level, as well as by pointers
847 from parent to child blocks. Each block has the following
848 fields:
849 magic CNTBT block magic number, 0x41425443
850 ('ABTC').
851 level level number of this block, 0 is a leaf.
852 numrecs number of data entries in the block.
853 leftsib left (logically lower) sibling block, 0 if
854 none.
855 rightsib right (logically higher) sibling block, 0 if
856 none.
857 recs [leaf blocks only] array of freespace
858 records. Each record contains startblock and
859 blockcount.
860 keys [non-leaf blocks only] array of key records.
861 These are the first value of each block in
862 the level below this one. Each record con‐
863 tains blockcount and startblock.
864 ptrs [non-leaf blocks only] array of child block
865 pointers. Each pointer is a block number
866 within the allocation group to the next level
867 in the Btree.
868 data User file blocks, and other blocks whose type is unknown,
870 have this type for display purposes in xfs_db. The block
871 data is displayed in hexadecimal format.
872 dir A version 1 directory is organized as a Btree with the direc‐
874 tory data embedded in the leaf blocks. The root of the Btree
875 is found in block 0 of the file. The index (sort order) of
876 the Btree is the hash value of the entry name. All the blocks
877 contain a blkinfo structure at the beginning with the follow‐
878 ing fields:
879 forw next sibling block.
880 back previous sibling block.
881 magic magic number for this block type.
882 The non-leaf (node) blocks have the following fields:
883 hdr header containing a blkinfo structure info
884 (magic number 0xfebe), the count of active
885 entries, and the level of this block above
886 the leaves.
887 btree array of entries containing hashval and be‐
888 fore fields. The before value is a block num‐
889 ber within the directory file to the child
890 block, the hashval is the last hash value in
891 that block.
892 The leaf blocks have the following fields:
893 hdr header containing a blkinfo structure info
894 (magic number 0xfeeb), the count of active
895 entries, namebytes (total name string bytes),
896 holes flag (block needs compaction), and
897 freemap (array of base, size entries for free
898 regions).
899 entries array of structures containing hashval,
900 nameidx (byte index into the block of the
901 name string), and namelen.
902 namelist array of structures containing inumber and
903 name.
904 dir2 A version 2 directory has four kinds of blocks. Data blocks
906 start at offset 0 in the file. There are two kinds of data
907 blocks: single-block directories have the leaf information
908 embedded at the end of the block, data blocks in multi-block
909 directories do not. Node and leaf blocks start at offset
910 32GiB (with either a single leaf block or the root node
911 block). Freespace blocks start at offset 64GiB. The node
912 and leaf blocks form a Btree, with references to the data in
913 the data blocks. The freespace blocks form an index of long‐
914 est free spaces within the data blocks.
915 A single-block directory block contains the following fields:
917 bhdr header containing magic number 0x58443242
918 ('XD2B') and an array bestfree of the longest
919 3 free spaces in the block (offset, length).
920 bu array of union structures. Each element is
921 either an entry or a freespace. For entries,
922 there are the following fields: inumber,
923 namelen, name, and tag. For freespace, there
924 are the following fields: freetag [22m(0xffff),
925 length, and tag. The tag value is the byte
926 offset in the block of the start of the entry
927 it is contained in.
928 bleaf array of leaf entries containing hashval and
929 address. The address is a 64-bit word offset
930 into the file.
931 btail tail structure containing the total count of
932 leaf entries and stale count of unused leaf
933 entries.
934 A data block contains the following fields:
935 dhdr header containing magic number 0x58443244
936 ('XD2D') and an array bestfree of the longest
937 3 free spaces in the block (offset, length).
938 du array of union structures as for bu.
939 Leaf blocks have two possible forms. If the Btree consists of
940 a single leaf then the freespace information is in the leaf
941 block, otherwise it is in separate blocks and the root of the
942 Btree is a node block. A leaf block contains the following
943 fields:
944 lhdr header containing a blkinfo structure info
945 (magic number 0xd2f1 for the single leaf
946 case, 0xd2ff for the true Btree case), the
947 total count of leaf entries, and stale count
948 of unused leaf entries.
949 lents leaf entries, as for bleaf.
950 lbests [single leaf only] array of values which rep‐
951 resent the longest freespace in each data
952 block in the directory.
953 ltail [single leaf only] tail structure containing
954 bestcount count of lbests.
955 A node block is identical to that for types attr and dir.
956 A freespace block contains the following fields:
958 fhdr header containing magic number 0x58443246
959 ('XD2F'), firstdb first data block number
960 covered by this freespace block, nvalid num‐
961 ber of valid entries, and nused number of en‐
962 tries representing real data blocks.
963 fbests array of values as for lbests.
964 dqblk The quota information is stored in files referred to by the
966 superblock uquotino and pquotino fields. Each filesystem
967 block in a quota file contains a constant number of quota en‐
968 tries. The quota entry size is currently 136 bytes, so with a
969 4KiB filesystem block size there are 30 quota entries per
970 block. The dquot command is used to locate these entries in
971 the filesystem. The file entries are indexed by the user or
972 project identifier to determine the block and offset. Each
973 quota entry has the following fields:
974 magic magic number, 0x4451 ('DQ').
975 version version number, currently 1.
976 flags flags, values include 0x01 for user quota,
977 0x02 for project quota.
978 id user or project identifier.
979 blk_hardlimit absolute limit on blocks in use.
980 blk_softlimit preferred limit on blocks in use.
981 ino_hardlimit absolute limit on inodes in use.
982 ino_softlimit preferred limit on inodes in use.
983 bcount blocks actually in use.
984 icount inodes actually in use.
985 itimer time when service will be refused if soft
986 limit is violated for inodes.
987 btimer time when service will be refused if soft
988 limit is violated for blocks.
989 iwarns number of warnings issued about inode
990 limit violations.
991 bwarns number of warnings issued about block
992 limit violations.
993 rtb_hardlimit absolute limit on realtime blocks in use.
994 rtb_softlimit preferred limit on realtime blocks in use.
995 rtbcount realtime blocks actually in use.
996 rtbtimer time when service will be refused if soft
997 limit is violated for realtime blocks.
998 rtbwarns number of warnings issued about realtime
999 block limit violations.
1000 inobt There is one set of filesystem blocks forming the inode allo‐
1002 cation Btree for each allocation group. The root block of
1003 this Btree is designated by the root field in the correspond‐
1004 ing AGI block. The blocks are linked to sibling left and
1005 right blocks at each level, as well as by pointers from par‐
1006 ent to child blocks. Each block has the following fields:
1007 magic INOBT block magic number, 0x49414254
1008 ('IABT').
1009 level level number of this block, 0 is a leaf.
1010 numrecs number of data entries in the block.
1011 leftsib left (logically lower) sibling block, 0 if
1012 none.
1013 rightsib right (logically higher) sibling block, 0 if
1014 none.
1015 recs [leaf blocks only] array of inode records.
1016 Each record contains startino allocation-
1017 group relative inode number, freecount count
1018 of free inodes in this chunk, and free bit‐
1019 map, LSB corresponds to inode 0.
1020 keys [non-leaf blocks only] array of key records.
1021 These are the first value of each block in
1022 the level below this one. Each record con‐
1023 tains startino.
1024 ptrs [non-leaf blocks only] array of child block
1025 pointers. Each pointer is a block number
1026 within the allocation group to the next level
1027 in the Btree.
1028 inode Inodes are allocated in "chunks" of 64 inodes each. Usually a
1030 chunk is multiple filesystem blocks, although there are cases
1031 with large filesystem blocks where a chunk is less than one
1032 block. The inode Btree (see inobt above) refers to the inode
1033 numbers per allocation group. The inode numbers directly re‐
1034 flect the location of the inode block on disk. Use the inode
1035 command to point xfs_db to a specific inode. Each inode con‐
1036 tains four regions: core, next_unlinked, u, and a. core con‐
1037 tains the fixed information. next_unlinked is separated from
1038 the core due to journaling considerations, see type agi field
1039 unlinked. u is a union structure that is different in size
1040 and format depending on the type and representation of the
1041 file data ("data fork"). a is an optional union structure to
1042 describe attribute data, that is different in size, format,
1043 and location depending on the presence and representation of
1044 attribute data, and the size of the u data ("attribute
1045 fork"). xfs_db automatically selects the proper union mem‐
1046 bers based on information in the inode.
1047 The following are fields in the inode core:
1049 magic inode magic number, 0x494e ('IN').
1050 mode mode and type of file, as described in
1051 chmod(2), mknod(2), and stat(2).
1052 version inode version, 1 or 2.
1053 format format of u union data (0: xfs_dev_t, 1: lo‐
1054 cal file - in-inode directory or symlink, 2:
1055 extent list, 3: Btree root, 4: unique id [un‐
1056 used]).
1057 nlinkv1 number of links to the file in a version 1
1058 inode.
1059 nlinkv2 number of links to the file in a version 2
1060 inode.
1061 projid_lo owner's project id (low word; version 2 inode
1062 only). projid_hi owner's project id (high
1063 word; version 2 inode only).
1064 uid owner's user id.
1065 gid owner's group id.
1066 atime time last accessed (seconds and nanoseconds).
1067 mtime time last modified.
1068 ctime time created or inode last modified.
1069 size number of bytes in the file.
1070 nblocks total number of blocks in the file including
1071 indirect and attribute.
1072 extsize basic/minimum extent size for the file.
1073 nextents number of extents in the data fork.
1074 naextents number of extents in the attribute fork.
1075 forkoff attribute fork offset in the inode, in 64-bit
1076 words from the start of u.
1077 aformat format of a data (1: local attribute data, 2:
1078 extent list, 3: Btree root).
1079 dmevmask DMAPI event mask.
1080 dmstate DMAPI state information.
1081 newrtbm file is the realtime bitmap and is "new" for‐
1082 mat.
1083 prealloc file has preallocated data space after EOF.
1084 realtime file data is in the realtime subvolume.
1085 gen inode generation number.
1086 The following fields are in the u data fork union:
1087 bmbt bmap Btree root. This looks like a bmapbtd
1088 block with redundant information removed.
1089 bmx array of extent descriptors.
1090 dev dev_t for the block or character device.
1091 sfdir shortform (in-inode) version 1 directory.
1092 This consists of a hdr containing the parent
1093 inode number and a count of active entries in
1094 the directory, followed by an array list of
1095 hdr.count entries. Each such entry contains
1096 inumber, namelen, and name string.
1097 sfdir2 shortform (in-inode) version 2 directory.
1098 This consists of a hdr containing a count of
1099 active entries in the directory, an i8count
1100 of entries with inumbers that don't fit in a
1101 32-bit value, and the parent inode number,
1102 followed by an array list of hdr.count en‐
1103 tries. Each such entry contains namelen, a
1104 saved offset used when the directory is con‐
1105 verted to a larger form, a name string, and
1106 the inumber.
1107 symlink symbolic link string value.
1108 The following fields are in the a attribute fork union if it
1109 exists:
1110 bmbt bmap Btree root, as above.
1111 bmx array of extent descriptors.
1112 sfattr shortform (in-inode) attribute values. This
1113 consists of a hdr containing a totsize (total
1114 size in bytes) and a count of active entries,
1115 followed by an array list of hdr.count en‐
1116 tries. Each such entry contains namelen, val‐
1117 uelen, root flag, name, and value.
1118 log Log blocks contain the journal entries for XFS. It's not
1120 useful to examine these with xfs_db, use xfs_logprint(8) in‐
1121 stead.
1122 refcntbt There is one set of filesystem blocks forming the reference
1124 count Btree for each allocation group. The root block of this
1125 Btree is designated by the refcntroot field in the corre‐
1126 sponding AGF block. The blocks are linked to sibling left
1127 and right blocks at each level, as well as by pointers from
1128 parent to child blocks. Each block has the following fields:
1129 magic REFC block magic number, 0x52334643 ('R3FC').
1130 level level number of this block, 0 is a leaf.
1131 numrecs number of data entries in the block.
1132 leftsib left (logically lower) sibling block, 0 if
1133 none.
1134 rightsib right (logically higher) sibling block, 0 if
1135 none.
1136 recs [leaf blocks only] array of reference count
1137 records. Each record contains startblock,
1138 blockcount, and refcount.
1139 keys [non-leaf blocks only] array of key records.
1140 These are the first value of each block in
1141 the level below this one. Each record con‐
1142 tains startblock.
1143 ptrs [non-leaf blocks only] array of child block
1144 pointers. Each pointer is a block number
1145 within the allocation group to the next level
1146 in the Btree.
1147 rmapbt There is one set of filesystem blocks forming the reverse
1149 mapping Btree for each allocation group. The root block of
1150 this Btree is designated by the rmaproot field in the corre‐
1151 sponding AGF block. The blocks are linked to sibling left
1152 and right blocks at each level, as well as by pointers from
1153 parent to child blocks. Each block has the following fields:
1154 magic RMAP block magic number, 0x524d4233 ('RMB3').
1155 level level number of this block, 0 is a leaf.
1156 numrecs number of data entries in the block.
1157 leftsib left (logically lower) sibling block, 0 if
1158 none.
1159 rightsib right (logically higher) sibling block, 0 if
1160 none.
1161 recs [leaf blocks only] array of reference count
1162 records. Each record contains startblock,
1163 blockcount, owner, offset, attr_fork,
1164 bmbt_block, and unwritten.
1165 keys [non-leaf blocks only] array of double-key
1166 records. The first ("low") key contains the
1167 first value of each block in the level below
1168 this one. The second ("high") key contains
1169 the largest key that can be used to identify
1170 any record in the subtree. Each record con‐
1171 tains startblock, owner, offset, attr_fork,
1172 and bmbt_block.
1173 ptrs [non-leaf blocks only] array of child block
1174 pointers. Each pointer is a block number
1175 within the allocation group to the next level
1176 in the Btree.
1177 rtbitmap If the filesystem has a realtime subvolume, then the rbmino
1179 field in the superblock refers to a file that contains the
1180 realtime bitmap. Each bit in the bitmap file controls the
1181 allocation of a single realtime extent (set == free). The
1182 bitmap is processed in 32-bit words, the LSB of a word is
1183 used for the first extent controlled by that bitmap word. The
1184 atime field of the realtime bitmap inode contains a counter
1185 that is used to control where the next new realtime file will
1186 start.
1187 rtsummary If the filesystem has a realtime subvolume, then the rsumino
1189 field in the superblock refers to a file that contains the
1190 realtime summary data. The summary file contains a two-dimen‐
1191 sional array of 16-bit values. Each value counts the number
1192 of free extent runs (consecutive free realtime extents) of a
1193 given range of sizes that starts in a given bitmap block.
1194 The size ranges are binary buckets (low size in the bucket is
1195 a power of 2). There are as many size ranges as are neces‐
1196 sary given the size of the realtime subvolume. The first di‐
1197 mension is the size range, the second dimension is the start‐
1198 ing bitmap block number (adjacent entries are for the same
1199 size, adjacent bitmap blocks).
1200 sb There is one sb (superblock) structure per allocation group.
1202 It is the first disk block in the allocation group. Only the
1203 first one (block 0 of the filesystem) is actually used; the
1204 other blocks are redundant information for xfs_repair(8) to
1205 use if the first superblock is damaged. Fields defined:
1206 magicnum superblock magic number, 0x58465342 ('XFSB').
1207 blocksize filesystem block size in bytes.
1208 dblocks number of filesystem blocks present in the
1209 data subvolume.
1210 rblocks number of filesystem blocks present in the
1211 realtime subvolume.
1212 rextents number of realtime extents that rblocks con‐
1213 tain.
1214 uuid unique identifier of the filesystem.
1215 logstart starting filesystem block number of the log
1216 (journal). If this value is 0 the log is
1217 "external".
1218 rootino root inode number.
1219 rbmino realtime bitmap inode number.
1220 rsumino realtime summary data inode number.
1221 rextsize realtime extent size in filesystem blocks.
1222 agblocks size of an allocation group in filesystem
1223 blocks.
1224 agcount number of allocation groups.
1225 rbmblocks number of realtime bitmap blocks.
1226 logblocks number of log blocks (filesystem blocks).
1227 versionnum filesystem version information. This value
1228 is currently 1, 2, 3, or 4 in the low 4 bits.
1229 If the low bits are 4 then the other bits
1230 have additional meanings. 1 is the original
1231 value. 2 means that attributes were used. 3
1232 means that version 2 inodes (large link
1233 counts) were used. 4 is the bitmask version
1234 of the version number. In this case, the
1235 other bits are used as flags (0x0010: at‐
1236 tributes were used, 0x0020: version 2 inodes
1237 were used, 0x0040: quotas were used, 0x0080:
1238 inode cluster alignment is in force, 0x0100:
1239 data stripe alignment is in force, 0x0200:
1240 the shared_vn field is used, 0x1000: unwrit‐
1241 ten extent tracking is on, 0x2000: version 2
1242 directories are in use).
1243 sectsize sector size in bytes, currently always 512.
1244 This is the size of the superblock and the
1245 other header blocks.
1246 inodesize inode size in bytes.
1247 inopblock number of inodes per filesystem block.
1248 fname obsolete, filesystem name.
1249 fpack obsolete, filesystem pack name.
1250 blocklog log2 of blocksize.
1251 sectlog log2 of sectsize.
1252 inodelog log2 of inodesize.
1253 inopblog log2 of inopblock.
1254 agblklog log2 of agblocks (rounded up).
1255 rextslog log2 of rextents.
1256 inprogress mkfs.xfs(8) or xfs_copy(8) aborted before
1257 completing this filesystem.
1258 imax_pct maximum percentage of filesystem space used
1259 for inode blocks.
1260 icount number of allocated inodes.
1261 ifree number of allocated inodes that are not in
1262 use.
1263 fdblocks number of free data blocks.
1264 frextents number of free realtime extents.
1265 uquotino user quota inode number.
1266 pquotino project quota inode number; this is currently
1267 unused.
1268 qflags quota status flags (0x01: user quota account‐
1269 ing is on, 0x02: user quota limits are en‐
1270 forced, 0x04: quotacheck has been run on user
1271 quotas, 0x08: project quota accounting is on,
1272 0x10: project quota limits are enforced,
1273 0x20: quotacheck has been run on project quo‐
1274 tas).
1275 flags random flags. 0x01: only read-only mounts are
1276 allowed.
1277 shared_vn shared version number (shared readonly
1278 filesystems).
1279 inoalignmt inode chunk alignment in filesystem blocks.
1280 unit stripe or RAID unit.
1281 width stripe or RAID width.
1282 dirblklog log2 of directory block size (filesystem
1283 blocks).
1284 symlink Symbolic link blocks are used only when the symbolic link
1286 value does not fit inside the inode. The block content is
1287 just the string value. Bytes past the logical end of the
1288 symbolic link value have arbitrary values.
1289 text User file blocks, and other blocks whose type is unknown,
1291 have this type for display purposes in xfs_db. The block
1292 data is displayed in two columns: Hexadecimal format and
1293 printable ASCII chars.
1294
1296 Many messages can come from the check (blockget) command. If the
1297 filesystem is completely corrupt, a core dump might be produced instead
1298 of the message
1299 device is not a valid filesystem
1300 If the filesystem is very large (has many files) then check might run
1302 out of memory. In this case the message
1303 out of memory
1304 is printed.
1305 The following is a description of the most likely problems and the as‐
1307 sociated messages. Most of the diagnostics produced are only meaning‐
1308 ful with an understanding of the structure of the filesystem.
1309 agf_freeblks n, counted m in ag a
1311 The freeblocks count in the allocation group header for alloca‐
1312 tion group a doesn't match the number of blocks counted free.
1313 agf_longest n, counted m in ag a
1315 The longest free extent in the allocation group header for allo‐
1316 cation group a doesn't match the longest free extent found in
1317 the allocation group.
1318 agi_count n, counted m in ag a
1320 The allocated inode count in the allocation group header for al‐
1321 location group a doesn't match the number of inodes counted in
1322 the allocation group.
1323 agi_freecount n, counted m in ag a
1325 The free inode count in the allocation group header for alloca‐
1326 tion group a doesn't match the number of inodes counted free in
1327 the allocation group.
1328 block a/b expected inum 0 got i
1330 The block number is specified as a pair (allocation group num‐
1331 ber, block in the allocation group). The block is used multiple
1332 times (shared), between multiple inodes. This message usually
1333 follows a message of the next type.
1334 block a/b expected type unknown got y
1336 The block is used multiple times (shared).
1337 block a/b type unknown not expected
1339
1341 mkfs.xfs(8), xfs_admin(8), xfs_copy(8), xfs_logprint(8), xfs_metad‐
1342 ump(8), xfs_ncheck(8), xfs_repair(8), mount(8), chmod(2), mknod(2),
1343 stat(2), xfs(5).
1344 xfs_db(8)