summaryrefslogtreecommitdiffstats
path: root/doc/tarlz.texi
blob: 00116eecf874479a1ceb339f1673e95ac0c4c627 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
\input texinfo @c -*-texinfo-*-
@c %**start of header
@setfilename tarlz.info
@documentencoding ISO-8859-15
@settitle Tarlz Manual
@finalout
@c %**end of header

@set UPDATED 30 July 2020
@set VERSION 0.17

@dircategory Data Compression
@direntry
* Tarlz: (tarlz).               Archiver with multimember lzip compression
@end direntry


@ifnothtml
@titlepage
@title Tarlz
@subtitle Archiver with multimember lzip compression
@subtitle for Tarlz version @value{VERSION}, @value{UPDATED}
@author by Antonio Diaz Diaz

@page
@vskip 0pt plus 1filll
@end titlepage

@contents
@end ifnothtml

@node Top
@top

This manual is for Tarlz (version @value{VERSION}, @value{UPDATED}).

@menu
* Introduction::              Purpose and features of tarlz
* Invoking tarlz::            Command line interface
* Portable character set::    POSIX portable filename character set
* File format::               Detailed format of the compressed archive
* Amendments to pax format::  The reasons for the differences with pax
* Program design::            Internal structure of tarlz
* Multi-threaded decoding::   Limitations of parallel tar decoding
* Minimum archive sizes::     Sizes required for full multi-threaded speed
* Examples::                  A small tutorial with examples
* Problems::                  Reporting bugs
* Concept index::             Index of concepts
@end menu

@sp 1
Copyright @copyright{} 2013-2020 Antonio Diaz Diaz.

This manual is free documentation: you have unlimited permission
to copy, distribute, and modify it.


@node Introduction
@chapter Introduction
@cindex introduction

@uref{http://www.nongnu.org/lzip/tarlz.html,,Tarlz} is a massively parallel
(multi-threaded) combined implementation of the tar archiver and the
@uref{http://www.nongnu.org/lzip/lzip.html,,lzip} compressor. Tarlz creates,
lists and extracts archives in a simplified and safer variant of the POSIX
pax format compressed with lzip, keeping the alignment between tar members
and lzip members. The resulting multimember tar.lz archive is fully backward
compatible with standard tar tools like GNU tar, which treat it like any
other tar.lz archive. Tarlz can append files to the end of such compressed
archives.

Keeping the alignment between tar members and lzip members has two
advantages. It adds an indexed lzip layer on top of the tar archive, making
it possible to decode the archive safely in parallel. It also minimizes the
amount of data lost in case of corruption. Compressing a tar archive with
plzip may even double the amount of files lost for each lzip member damaged
because it does not keep the members aligned.

Tarlz can create tar archives with five levels of compression granularity;
per file (---no-solid), per block (---bsolid, default), per directory
(---dsolid), appendable solid (---asolid), and solid (---solid). It can also
create uncompressed tar archives.

@noindent
Of course, compressing each file (or each directory) individually can't
achieve a compression ratio as high as compressing solidly the whole tar
archive, but it has the following advantages:

@itemize @bullet
@item
The resulting multimember tar.lz archive can be decompressed in
parallel, multiplying the decompression speed.

@item
New members can be appended to the archive (by removing the EOF
member), and unwanted members can be deleted from the archive. Just
like an uncompressed tar archive.

@item
It is a safe POSIX-style backup format. In case of corruption,
tarlz can extract all the undamaged members from the tar.lz
archive, skipping over the damaged members, just like the standard
(uncompressed) tar. Moreover, the option @samp{--keep-damaged} can be
used to recover as much data as possible from each damaged member,
and lziprecover can be used to recover some of the damaged members.

@item
A multimember tar.lz archive is usually smaller than the
corresponding solidly compressed tar.gz archive, except when
compressing files smaller than about 32 KiB individually.
@end itemize

Tarlz protects the extended records with a Cyclic Redundancy Check (CRC) in
a way compatible with standard tar tools. @xref{crc32}.

Tarlz does not understand other tar formats like @samp{gnu}, @samp{oldgnu},
@samp{star} or @samp{v7}. The command
@w{@samp{tarlz -tf archive.tar.lz > /dev/null}} can be used to verify that
the format of the archive is compatible with tarlz.


@node Invoking tarlz
@chapter Invoking tarlz
@cindex invoking
@cindex options
@cindex usage
@cindex version

The format for running tarlz is:

@example
tarlz [@var{options}] [@var{files}]
@end example

@noindent
All operations except @samp{--concatenate} operate on whole trees if any
@var{file} is a directory.

On archive creation or appending tarlz archives the files specified, but
removes from member names any leading and trailing slashes and any file name
prefixes containing a @samp{..} component. On extraction, leading and
trailing slashes are also removed from member names, and archive members
containing a @samp{..} component in the file name are skipped. Tarlz detects
when the archive being created or enlarged is among the files to be dumped,
appended or concatenated, and skips it.

On extraction and listing, tarlz removes leading @samp{./} strings from
member names in the archive or given in the command line, so that
@w{@samp{tarlz -xf foo ./bar baz}} extracts members @samp{bar} and
@samp{./baz} from archive @samp{foo}.

If several compression levels or @samp{--*solid} options are given, the last
setting is used. For example @w{@samp{-9 --solid --uncompressed -1}} is
equivalent to @samp{-1 --solid}

tarlz supports the following
@uref{http://www.nongnu.org/arg-parser/manual/arg_parser_manual.html#Argument-syntax,,options}:
@ifnothtml
@xref{Argument syntax,,,arg_parser}.
@end ifnothtml

@table @code
@item --help
Print an informative help message describing the options and exit.

@item -V
@itemx --version
Print the version number of tarlz on the standard output and exit.
This version number should be included in all bug reports.

@item -A
@itemx --concatenate
Append one or more archives to the end of an archive. All the archives
involved must be regular (seekable) files, and must be either all compressed
or all uncompressed. Compressed and uncompressed archives can't be mixed.
Compressed archives must be multimember lzip files with the two end-of-file
blocks plus any zero padding contained in the last lzip member of each
archive. The intermediate end-of-file blocks are removed as each new archive
is concatenated. If the archive is uncompressed, tarlz parses and skips tar
headers until it finds the end-of-file blocks. Exit with status 0 without
modifying the archive if no @var{files} have been specified.

@anchor{--data-size}
@item -B @var{bytes}
@itemx --data-size=@var{bytes}
Set target size of input data blocks for the option @samp{--bsolid}.
@xref{--bsolid}. Valid values range from @w{8 KiB} to @w{1 GiB}. Default
value is two times the dictionary size, except for option @samp{-0} where it
defaults to @w{1 MiB}. @xref{Minimum archive sizes}.

@item -c
@itemx --create
Create a new archive from @var{files}.

@item -C @var{dir}
@itemx --directory=@var{dir}
Change to directory @var{dir}. When creating or appending, the position of
each @samp{-C} option in the command line is significant; it will change the
current working directory for the following @var{files} until a new
@samp{-C} option appears in the command line. When extracting or comparing,
all the @samp{-C} options are executed in sequence before reading the
archive. Listing ignores any @samp{-C} options specified. @var{dir} is
relative to the then current working directory, perhaps changed by a
previous @samp{-C} option.

Note that a process can only have one current working directory (CWD).
Therefore multi-threading can't be used to create an archive if a @samp{-C}
option appears after a relative file name in the command line.

@item -d
@itemx --diff
Compare and report differences between archive and file system. For each tar
member in the archive, verify that the corresponding file in the file system
exists and is of the same type (regular file, directory, etc). Report on
standard output the differences found in type, mode (permissions), owner and
group IDs, modification time, file size, file contents (of regular files),
target (of symlinks) and device number (of block/character special files).

As tarlz removes leading slashes from member names, the option @samp{-C} may
be used in combination with @samp{--diff} when absolute file names were used
on archive creation: @w{@samp{tarlz -C / -d}}. Alternatively, tarlz may be
run from the root directory to perform the comparison.

@item --ignore-ids
Make @samp{--diff} ignore differences in owner and group IDs. This option is
useful when comparing an @samp{--anonymous} archive.

@item --delete
Delete files and directories from an archive in place. It currently can
delete only from uncompressed archives and from archives with files
compressed individually (@samp{--no-solid} archives). Note that files of
about @samp{--data-size} or larger are compressed individually even if
@samp{--bsolid} is used, and can therefore be deleted. Tarlz takes care to
not delete a tar member unless it is possible to do so. For example it won't
try to delete a tar member that is not compressed individually. Even in the
case of finding a corrupt member after having deleted some member(s), tarlz
stops and copies the rest of the file as soon as corruption is found,
leaving it just as corrupt as it was, but not worse.

To delete a directory without deleting the files under it, use
@w{@samp{tarlz --delete -f foo --exclude='dir/*' dir}}. Deleting in place
may be dangerous. A corrupt archive, a power cut, or an I/O error may cause
data loss.

@item --exclude=@var{pattern}
Exclude files matching a shell pattern like @samp{*.o}. A file is considered
to match if any component of the file name matches. For example, @samp{*.o}
matches @samp{foo.o}, @samp{foo.o/bar} and @samp{foo/bar.o}. If
@var{pattern} contains a @samp{/}, it matches a corresponding @samp{/} in
the file name. For example, @samp{foo/*.o} matches @samp{foo/bar.o}.

@item -f @var{archive}
@itemx --file=@var{archive}
Use archive file @var{archive}. A hyphen @samp{-} used as an @var{archive}
argument reads from standard input or writes to standard output.

@item -h
@itemx --dereference
Follow symbolic links during archive creation, appending or comparison.
Archive or compare the files they point to instead of the links themselves.

@item --mtime=@var{date}
When creating or appending, use @var{date} as the modification time for
files added to the archive instead of their actual modification times. The
value of @var{date} may be either @samp{@@} followed by the number of
seconds since the epoch, or a date in format @w{@samp{YYYY-MM-DD HH:MM:SS}},
or the name of an existing file starting with @samp{.} or @samp{/}. In the
latter case, the modification time of that file is used.

@item -n @var{n}
@itemx --threads=@var{n}
Set the number of (de)compression threads, overriding the system's default.
Valid values range from 0 to "as many as your system can support". A value
of 0 disables threads entirely. If this option is not used, tarlz tries to
detect the number of processors in the system and use it as default value.
@w{@samp{tarlz --help}} shows the system's default value. See the note about
multi-threaded archive creation in the option @samp{-C} above.
Multi-threaded extraction of files from an archive is not yet implemented.
@xref{Multi-threaded decoding}.

Note that the number of usable threads is limited during compression to
@w{ceil( uncompressed_size / data_size )} (@pxref{Minimum archive sizes}),
and during decompression to the number of lzip members in the tar.lz
archive, which you can find by running @w{@samp{lzip -lv archive.tar.lz}}.

@item -p
@itemx --preserve-permissions
On extraction, set file permissions as they appear in the archive. This is
the default behavior when tarlz is run by the superuser. The default for
other users is to subtract the umask of the user running tarlz from the
permissions specified in the archive.

@item -q
@itemx --quiet
Quiet operation. Suppress all messages.

@item -r
@itemx --append
Append files to the end of an archive. The archive must be a regular
(seekable) file either compressed or uncompressed. Compressed members can't
be appended to an uncompressed archive, nor vice versa. If the archive is
compressed, it must be a multimember lzip file with the two end-of-file
blocks plus any zero padding contained in the last lzip member of the
archive. It is possible to append files to an archive with a different
compression granularity. Appending works as follows; first the end-of-file
blocks are removed, then the new members are appended, and finally two new
end-of-file blocks are appended to the archive. If the archive is
uncompressed, tarlz parses and skips tar headers until it finds the
end-of-file blocks. Exit with status 0 without modifying the archive if no
@var{files} have been specified.

@item -t
@itemx --list
List the contents of an archive. If @var{files} are given, list only the
@var{files} given.

@item -v
@itemx --verbose
Verbosely list files processed.

@item -x
@itemx --extract
Extract files from an archive. If @var{files} are given, extract only the
@var{files} given. Else extract all the files in the archive. To extract a
directory without extracting the files under it, use
@w{@samp{tarlz -xf foo --exclude='dir/*' dir}}. Tarlz will not make any
special effort to extract a file over an incompatible type of file. For
example, extracting a link over a directory will usually fail. (Principle of
least surprise).

@item -0 .. -9
Set the compression level for @samp{--create} and @samp{--append}. The
default compression level is @samp{-6}. Like lzip, tarlz also minimizes the
dictionary size of the lzip members it creates, reducing the amount of
memory required for decompression.

@multitable {Level} {Dictionary size} {Match length limit}
@item Level @tab Dictionary size @tab Match length limit
@item -0 @tab 64 KiB @tab  16 bytes
@item -1 @tab  1 MiB @tab   5 bytes
@item -2 @tab  1.5 MiB @tab   6 bytes
@item -3 @tab  2 MiB @tab   8 bytes
@item -4 @tab  3 MiB @tab  12 bytes
@item -5 @tab  4 MiB @tab  20 bytes
@item -6 @tab  8 MiB @tab  36 bytes
@item -7 @tab 16 MiB @tab  68 bytes
@item -8 @tab 24 MiB @tab 132 bytes
@item -9 @tab 32 MiB @tab 273 bytes
@end multitable

@item --uncompressed
With @samp{--create}, don't compress the tar archive created. Create an
uncompressed tar archive instead. With @samp{--append}, don't compress the
new members appended to the tar archive. Compressed members can't be
appended to an uncompressed archive, nor vice versa.

@item --asolid
When creating or appending to a compressed archive, use appendable solid
compression. All the files being added to the archive are compressed into a
single lzip member, but the end-of-file blocks are compressed into a
separate lzip member. This creates a solidly compressed appendable archive.
Solid archives can't be created nor decoded in parallel.

@anchor{--bsolid}
@item --bsolid
When creating or appending to a compressed archive, use block compression.
Tar members are compressed together in a lzip member until they approximate
a target uncompressed size. The size can't be exact because each solidly
compressed data block must contain an integer number of tar members. Block
compression is the default because it improves compression ratio for
archives with many files smaller than the block size. This option allows
tarlz revert to default behavior if, for example, it is invoked through an
alias like @samp{tar='tarlz --solid'}. @xref{--data-size}, to set the target
block size.

@item --dsolid
When creating or appending to a compressed archive, compress each file
specified in the command line separately in its own lzip member, and use
solid compression for each directory specified in the command line. The
end-of-file blocks are compressed into a separate lzip member. This creates
a compressed appendable archive with a separate lzip member for each file or
top-level directory specified.

@item --no-solid
When creating or appending to a compressed archive, compress each file
separately in its own lzip member. The end-of-file blocks are compressed
into a separate lzip member. This creates a compressed appendable archive
with a lzip member for each file.

@item --solid
When creating or appending to a compressed archive, use solid compression.
The files being added to the archive, along with the end-of-file blocks, are
compressed into a single lzip member. The resulting archive is not
appendable. No more files can be later appended to the archive. Solid
archives can't be created nor decoded in parallel.

@item --anonymous
Equivalent to @samp{--owner=root --group=root}.

@item --owner=@var{owner}
When creating or appending, use @var{owner} for files added to the
archive. If @var{owner} is not a valid user name, it is decoded as a
decimal numeric user ID.

@item --group=@var{group}
When creating or appending, use @var{group} for files added to the
archive. If @var{group} is not a valid group name, it is decoded as a
decimal numeric group ID.

@item --keep-damaged
Don't delete partially extracted files. If a decompression error happens
while extracting a file, keep the partial data extracted. Use this
option to recover as much data as possible from each damaged member.

@item --missing-crc
Exit with error status 2 if the CRC of the extended records is missing.
When this option is used, tarlz detects any corruption in the extended
records (only limited by CRC collisions). But note that a corrupt
@samp{GNU.crc32} keyword, for example @samp{GNU.crc33}, is reported as a
missing CRC instead of as a corrupt record. This misleading
@samp{Missing CRC} message is the consequence of a flaw in the POSIX pax
format; i.e., the lack of a mandatory check sequence in the extended
records. @xref{crc32}.

@item --out-slots=@var{n}
Number of @w{1 MiB} output packets buffered per worker thread during
multi-threaded creation or appending to compressed archives. Increasing the
number of packets may increase compression speed if the files being archived
are larger than @w{64 MiB} compressed, but requires more memory. Valid
values range from 1 to 1024. The default value is 64.

@ignore
@item --permissive
Allow some violations of the archive format, like consecutive extended
headers preceding a ustar header, or several records with the same
keyword appearing in the same block of extended records.
@end ignore

@end table

Exit status: 0 for a normal exit, 1 for environmental problems (file not
found, files differ, invalid flags, I/O errors, etc), 2 to indicate a
corrupt or invalid input file, 3 for an internal consistency error (eg, bug)
which caused tarlz to panic.


@node Portable character set
@chapter POSIX portable filename character set
@cindex portable character set

The set of characters from which portable file names are constructed.

@example
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
a b c d e f g h i j k l m n o p q r s t u v w x y z
0 1 2 3 4 5 6 7 8 9 . _ -
@end example

The last three characters are the period, underscore, and hyphen-minus
characters, respectively.

File names are identifiers. Therefore, archiving works better when file
names use only the portable character set without spaces added.


@node File format
@chapter File format
@cindex file format

In the diagram below, a box like this:

@verbatim
+---+
|   | <-- the vertical bars might be missing
+---+
@end verbatim

represents one byte; a box like this:

@verbatim
+==============+
|              |
+==============+
@end verbatim

represents a variable number of bytes or a fixed but large number of
bytes (for example 512).

@sp 1
A tar.lz file consists of a series of lzip members (compressed data sets).
The members simply appear one after another in the file, with no
additional information before, between, or after them.

Each lzip member contains one or more tar members in a simplified POSIX
pax interchange format. The only pax typeflag value supported by tarlz
(in addition to the typeflag values defined by the ustar format) is
@samp{x}. The pax format is an extension on top of the ustar format that
removes the size limitations of the ustar format.

Each tar member contains one file archived, and is represented by the
following sequence:

@itemize @bullet
@item
An optional extended header block with extended header records. This
header block is of the form described in pax header block, with a
typeflag value of @samp{x}. The extended header records are included as
the data for this header block.

@item
A header block in ustar format that describes the file. Any fields
defined in the preceding optional extended header records override the
associated fields in this header block for this file.

@item
Zero or more blocks that contain the contents of the file.
@end itemize

Each tar member must be contiguously stored in a lzip member for the
parallel decoding operations like @samp{--list} to work. If any tar member
is split over two or more lzip members, the archive must be decoded
sequentially. @xref{Multi-threaded decoding}.

At the end of the archive file there are two 512-byte blocks filled with
binary zeros, interpreted as an end-of-archive indicator. These EOF
blocks are either compressed in a separate lzip member or compressed
along with the tar members contained in the last lzip member.

The diagram below shows the correspondence between each tar member (formed
by one or two headers plus optional data) in the tar archive and each
@uref{http://www.nongnu.org/lzip/manual/lzip_manual.html#File-format,,lzip member}
in the resulting multimember tar.lz archive, when per file compression is
used:
@ifnothtml
@xref{File format,,,lzip}.
@end ifnothtml

@verbatim
tar
+========+======+=================+===============+========+======+========+
| header | data | extended header | extended data | header | data |   EOF  |
+========+======+=================+===============+========+======+========+

tar.lz
+===============+=================================================+========+
|     member    |                      member                     | member |
+===============+=================================================+========+
@end verbatim

@ignore
When @samp{--permissive} is used, the following violations of the
archive format are allowed:@*
If several extended headers precede an ustar header, only the last
extended header takes effect. The other extended headers are ignored.
Similarly, if several records with the same keyword appear in the same
block of extended records, only the last record for the repeated keyword
takes effect. The other records for the repeated keyword are ignored.@*
A global header inserted between an extended header and an ustar header.@*
An extended header just before the EOF blocks.
@end ignore

@sp 1
@section Pax header block

The pax header block is identical to the ustar header block described below
except that the typeflag has the value @samp{x} (extended). The field
@samp{size} is the size of the extended header data in bytes. Most other
fields in the pax header block are zeroed on archive creation to prevent
trouble if the archive is read by an ustar tool, and are ignored by tarlz on
archive extraction. @xref{flawed-compat}.

The pax extended header data consists of one or more records, each of
them constructed as follows:@*
@samp{"%d %s=%s\n", <length>, <keyword>, <value>}

The fields <length> and <keyword> in the record must be limited to the
portable character set (@pxref{Portable character set}). The field <length>
contains the decimal length of the record in bytes, including the trailing
newline. The field <value> is stored as-is, without conversion to UTF-8 nor
any other transformation. The fields are separated by the ASCII characters
space, equal-sign, and newline.

These are the <keyword> values currently supported by tarlz:

@table @code
@item linkpath
The pathname of a link being created to another file, of any type,
previously archived. This record overrides the field @samp{linkname} in the
following ustar header block. The following ustar header block
determines the type of link created. If typeflag of the following header
block is 1, it will be a hard link. If typeflag is 2, it will be a
symbolic link and the linkpath value will be used as the contents of the
symbolic link.

@item path
The pathname of the following file. This record overrides the fields
@samp{name} and @samp{prefix} in the following ustar header block.

@item size
The size of the file in bytes, expressed as a decimal number using
digits from the ISO/IEC 646:1991 (ASCII) standard. This record overrides
the size field in the following ustar header block. The size record is
used only for files with a size value greater than 8_589_934_591
@w{(octal 77777777777)}. This is 2^33 bytes or larger.

@anchor{key_crc32}
@item GNU.crc32
CRC32-C (Castagnoli) of the extended header data excluding the 8 bytes
representing the CRC <value> itself. The <value> is represented as 8
hexadecimal digits in big endian order,
@w{@samp{22 GNU.crc32=00000000\n}}. The keyword of the CRC record is
protected by the CRC to guarante that corruption is always detected
(except in case of CRC collision). A CRC was chosen because a checksum
is too weak for a potentially large list of variable sized records. A
checksum can't detect simple errors like the swapping of two bytes.
@end table

@sp 1
@section Ustar header block

The ustar header block has a length of 512 bytes and is structured as
shown in the following table. All lengths and offsets are in decimal.

@multitable {Field Name} {Offset} {Length (in bytes)}
@item Field Name @tab Offset @tab Length (in bytes)
@item name     @tab   0 @tab 100
@item mode     @tab 100 @tab   8
@item uid      @tab 108 @tab   8
@item gid      @tab 116 @tab   8
@item size     @tab 124 @tab  12
@item mtime    @tab 136 @tab  12
@item chksum   @tab 148 @tab   8
@item typeflag @tab 156 @tab   1
@item linkname @tab 157 @tab 100
@item magic    @tab 257 @tab   6
@item version  @tab 263 @tab   2
@item uname    @tab 265 @tab  32
@item gname    @tab 297 @tab  32
@item devmajor @tab 329 @tab   8
@item devminor @tab 337 @tab   8
@item prefix   @tab 345 @tab 155
@end multitable

All characters in the header block are coded using the ISO/IEC 646:1991
(ASCII) standard, except in fields storing names for files, users, and
groups. For maximum portability between implementations, names should only
contain characters from the portable character set (@pxref{Portable
character set}), but if an implementation supports the use of characters
outside of @samp{/} and the portable character set in names for files,
users, and groups, tarlz will use the byte values in these names unmodified.

The fields @samp{name}, @samp{linkname}, and @samp{prefix} are
null-terminated character strings except when all characters in the array
contain non-null characters including the last character.

The fields @samp{prefix} and @samp{name} produce the pathname of the file. A
new pathname is formed, if prefix is not an empty string (its first
character is not null), by concatenating prefix (up to the first null
character), a slash character, and name; otherwise, name is used alone. In
either case, name is terminated at the first null character. If prefix
begins with a null character, it is ignored. In this manner, pathnames of at
most 256 characters can be supported. If a pathname does not fit in the
space provided, an extended record is used to store the pathname.

The field @samp{linkname} does not use the prefix to produce a pathname. If
the linkname does not fit in the 100 characters provided, an extended record
is used to store the linkname.

The field @samp{mode} provides 12 access permission bits. The following
table shows the symbolic name of each bit and its octal value:

@multitable {Bit Name} {Value} {Bit Name} {Value} {Bit Name} {Value}
@headitem Bit Name @tab Value @tab Bit Name @tab Value @tab Bit Name @tab Value
@item S_ISUID @tab 04000 @tab S_ISGID @tab 02000 @tab S_ISVTX @tab 01000
@item S_IRUSR @tab 00400 @tab S_IWUSR @tab 00200 @tab S_IXUSR @tab 00100
@item S_IRGRP @tab 00040 @tab S_IWGRP @tab 00020 @tab S_IXGRP @tab 00010
@item S_IROTH @tab 00004 @tab S_IWOTH @tab 00002 @tab S_IXOTH @tab 00001
@end multitable

The fields @samp{uid} and @samp{gid} are the user and group IDs of the owner
and group of the file, respectively.

The field @samp{size} contains the octal representation of the size of the
file in bytes. If the field @samp{typeflag} specifies a file of type '0'
(regular file) or '7' (high performance regular file), the number of logical
records following the header is @w{(size / 512)} rounded to the next
integer. For all other values of typeflag, tarlz either sets the size field
to 0 or ignores it, and does not store or expect any logical records
following the header. If the file size is larger than 8_589_934_591 bytes
@w{(octal 77777777777)}, an extended record is used to store the file size.

The field @samp{mtime} contains the octal representation of the modification
time of the file at the time it was archived, obtained from the function
@samp{stat}.

The field @samp{chksum} contains the octal representation of the value of
the simple sum of all bytes in the header logical record. Each byte in the
header is treated as an unsigned value. When calculating the checksum, the
chksum field is treated as if it were all space characters.

The field @samp{typeflag} contains a single character specifying the type of
file archived:

@table @code
@item '0'
Regular file.

@item '1'
Hard link to another file, of any type, previously archived.

@item '2'
Symbolic link.

@item '3', '4'
Character special file and block special file respectively. In this case the
fields @samp{devmajor} and @samp{devminor} contain information defining the
device in unspecified format.

@item '5'
Directory.

@item '6'
FIFO special file.

@item '7'
Reserved to represent a file to which an implementation has associated
some high-performance attribute. Tarlz treats this type of file as a
regular file (type 0).

@end table

The field @samp{magic} contains the ASCII null-terminated string "ustar".
The field @samp{version} contains the characters "00" (0x30,0x30). The
fields @samp{uname} and @samp{gname} are null-terminated character strings
except when all characters in the array contain non-null characters
including the last character. Each numeric field contains a leading space-
or zero-filled, optionally null-terminated octal number using digits from
the ISO/IEC 646:1991 (ASCII) standard. Tarlz is able to decode numeric
fields 1 byte longer than standard ustar by not requiring a terminating null
character.


@node Amendments to pax format
@chapter The reasons for the differences with pax
@cindex Amendments to pax format

Tarlz creates safe archives that allow the reliable detection of invalid or
corrupt metadata during decoding even when the integrity checking of lzip
can't be used because the lzip members are only decompressed partially, as
it happens in parallel @samp{--diff}, @samp{--list}, and @samp{--extract}.
In order to achieve this goal, tarlz makes some changes to the variant of
the pax format that it uses. This chapter describes these changes and the
concrete reasons to implement them.

@sp 1
@anchor{crc32}
@section Add a CRC of the extended records

The POSIX pax format has a serious flaw. The metadata stored in pax extended
records are not protected by any kind of check sequence. Corruption in a
long file name may cause the extraction of the file in the wrong place
without warning. Corruption in a large file size may cause the truncation of
the file or the appending of garbage to the file, both followed by a
spurious warning about a corrupt header far from the place of the undetected
corruption.

Metadata like file name and file size must be always protected in an archive
format because of the adverse effects of undetected corruption in them,
potentially much worse that undetected corruption in the data. Even more so
in the case of pax because the amount of metadata it stores is potentially
large, making undetected corruption and archiver misbehavior more probable.

Headers and metadata must be protected separately from data because the
integrity checking of lzip may not be able to detect the corruption before
the metadata has been used, for example, to create a new file in the wrong
place.

Because of the above, tarlz protects the extended records with a CRC in a
way compatible with standard tar tools. @xref{key_crc32}.

@sp 1
@anchor{flawed-compat}
@section Remove flawed backward compatibility

In order to allow the extraction of pax archives by a tar utility conforming
to the POSIX-2:1993 standard, POSIX.1-2008 recommends selecting extended
header field values that allow such tar to create a regular file containing
the extended header records as data. This approach is broken because if the
extended header is needed because of a long file name, the fields
@samp{prefix} and @samp{name} will be unable to contain the full pathname of
the file. Therefore the files corresponding to both the extended header and
the overridden ustar header will be extracted using truncated file names,
perhaps overwriting existing files or directories. It may be a security risk
to extract a file with a truncated file name.

To avoid this problem, tarlz writes extended headers with all fields zeroed
except size, chksum, typeflag, magic and version. This prevents old tar
programs from extracting the extended records as a file in the wrong place.
Tarlz also sets to zero those fields of the ustar header overridden by
extended records.

If an extended header is required for any reason (for example a file size
larger than @w{8 GiB} or a link name longer than 100 bytes), tarlz moves the
file name also to the extended header to prevent an ustar tool from trying
to extract the file or link. This also makes easier during parallel decoding
the detection of a tar member split between two lzip members at the boundary
between the extended header and the ustar header.

@sp 1
@section As simple as possible (but not simpler)

The tarlz format is mainly ustar. Extended pax headers are used only when
needed because the length of a file name or link name, or the size of a file
exceed the limits of the ustar format. Adding @w{1 KiB} of extended headers
to each member just to record subsecond timestamps seems wasteful for a
backup format. Moreover, minimizing the overhead may help recovering the
archive with lziprecover in case of corruption.

Global pax headers are tolerated, but not supported; they are parsed and
ignored. Some operations may not behave as expected if the archive contains
global headers.

@sp 1
@section Avoid misconversions to/from UTF-8

There is no portable way to tell what charset a text string is coded into.
Therefore, tarlz stores all fields representing text strings unmodified,
without conversion to UTF-8 nor any other transformation. This prevents
accidental double UTF-8 conversions. If the need arises this behavior will
be adjusted with a command line option in the future.


@node Program design
@chapter Internal structure of tarlz
@cindex program design

The parts of tarlz related to sequential processing of the archive are more
or less similar to any other tar and won't be described here. The interesting
parts described here are those related to Multi-threaded processing.

The structure of the part of tarlz performing Multi-threaded archive
creation is somewhat similar to that of plzip with the added complication of
the solidity levels. A grouper thread and several worker threads are
created, acting the main thread as muxer (multiplexer) thread. A "packet
courier" takes care of data transfers among threads and limits the maximum
number of data blocks (packets) being processed simultaneously.

The grouper traverses the directory tree, groups together the metadata of
the files to be archived in each lzip member, and distributes them to the
workers. The workers compress the metadata received from the grouper along
with the file data read from the file system. The muxer collects processed
packets from the workers, and writes them to the archive.

@verbatim
,--------,
|    data|---> to each worker below
|        |                    ,------------,
| file   |                ,-->| worker   0 |--,
| system |                |   `------------'  |
|        |    ,---------, |   ,------------,  |   ,-------,   ,---------,
|metadata|--->| grouper |-+-->| worker   1 |--+-->| muxer |-->| archive |
`--------'    `---------' |   `------------'  |   `-------'   `---------'
                          |        ...        |
                          |   ,------------,  |
                          `-->| worker N-1 |--'
                              `------------'
@end verbatim

Decoding an archive is somewhat similar to how plzip decompresses a regular
file to standard output, with the differences that it is not the data but
only messages what is written to stdout/stderr, and that each worker may
access files in the file system either to read them (diff) or write them
(extract). As in plzip, each worker reads members directly from the archive.

@verbatim
,--------,
| file   |<---> data to/from each worker below
| system |
`--------'
                ,------------,
            ,-->| worker   0 |--,
            |   `------------'  |
,---------, |   ,------------,  |   ,-------,   ,--------,
| archive |-+-->| worker   1 |--+-->| muxer |-->| stdout |
`---------' |   `------------'  |   `-------'   | stderr |
            |        ...        |               `--------'
            |   ,------------,  |
            `-->| worker N-1 |--'
                `------------'
@end verbatim

As misaligned tar.lz archives can't be decoded in parallel, and the
misalignment can't be detected until after decoding has started, a
"mastership request" mechanism has been designed that allows the decoding to
continue instead of signalling an error.

During parallel decoding, if a worker finds a misalignment, it requests
mastership to decode the rest of the archive. When mastership is requested,
an error_member_id is set, and all subsequently received packets with
member_id > error_member_id are rejected. All workers requesting mastership
are blocked at the request_mastership call until mastership is granted.
Mastership is granted to the delivering worker when its queue is empty to
make sure that all preceding packets have been processed. When mastership is
granted, all packets are deleted and all subsequently received packets not
coming from the master are rejected.

If a worker can't continue decoding for any cause (for example lack of
memory or finding a split tar member at the beginning of a lzip member), it
requests mastership to print an error and terminate the program. Only if
some other worker requests mastership in a previous lzip member can this
error be avoided.


@node Multi-threaded decoding
@chapter Limitations of parallel tar decoding
@cindex parallel tar decoding

Safely decoding an arbitrary tar archive in parallel is impossible. For
example, if a tar archive containing another tar archive is decoded starting
from some position other than the beginning, there is no way to know if the
first header found there belongs to the outer tar archive or to the inner
tar archive. Tar is a format inherently serial; it was designed for tapes.

In the case of compressed tar archives, the start of each compressed block
determines one point through which the tar archive can be decoded in
parallel. Therefore, in tar.lz archives the decoding operations can't be
parallelized if the tar members are not aligned with the lzip members. Tar
archives compressed with plzip can't be decoded in parallel because tar and
plzip do not have a way to align both sets of members. Certainly one can
decompress one such archive with a multi-threaded tool like plzip, but the
increase in speed is not as large as it could be because plzip must
serialize the decompressed data and pass them to tar, which decodes them
sequentially, one tar member at a time.

On the other hand, if the tar.lz archive is created with a tool like tarlz,
which can guarantee the alignment between tar members and lzip members
because it controls both archiving and compression, then the lzip format
becomes an indexed layer on top of the tar archive which makes possible
decoding it safely in parallel.

Tarlz is able to automatically decode aligned and unaligned multimember
tar.lz archives, keeping backwards compatibility. If tarlz finds a member
misalignment during multi-threaded decoding, it switches to single-threaded
mode and continues decoding the archive. Currently only the options
@samp{--diff} and @samp{--list} are able to do multi-threaded decoding.

If the files in the archive are large, multi-threaded @samp{--list} on a
regular (seekable) tar.lz archive can be hundreds of times faster than
sequential @samp{--list} because, in addition to using several processors,
it only needs to decompress part of each lzip member. See the following
example listing the Silesia corpus on a dual core machine:

@example
tarlz -9 --no-solid -cf silesia.tar.lz silesia
time lzip -cd silesia.tar.lz | tar -tf -            (5.032s)
time plzip -cd silesia.tar.lz | tar -tf -           (3.256s)
time tarlz -tf silesia.tar.lz                       (0.020s)
@end example

On the other hand, multi-threaded @samp{--list} won't detect corruption in
the tar member data because it only decodes the part of each lzip member
corresponding to the tar member header.


@node Minimum archive sizes
@chapter Minimum archive sizes required for multi-threaded block compression
@cindex minimum archive sizes

When creating or appending to a compressed archive using multi-threaded
block compression, tarlz puts tar members together in blocks and compresses
as many blocks simultaneously as worker threads are chosen, creating a
multimember compressed archive.

For this to work as expected (and roughly multiply the compression speed by
the number of available processors), the uncompressed archive must be at
least as large as the number of worker threads times the block size
(@pxref{--data-size}). Else some processors will not get any data to
compress, and compression will be proportionally slower. The maximum speed
increase achievable on a given archive is limited by the ratio
@w{(uncompressed_size / data_size)}. For example, a tarball the size of gcc
or linux will scale up to 10 or 14 processors at level -9.

The following table shows the minimum uncompressed archive size needed for
full use of N processors at a given compression level, using the default
data size for each level:

@multitable {Processors} {512 MiB} {512 MiB} {512 MiB} {512 MiB} {512 MiB} {512 MiB}
@headitem Processors @tab 2 @tab 4 @tab 8 @tab 16 @tab 64 @tab 256
@item Level
@item -0 @tab   2 MiB @tab   4 MiB @tab   8 MiB @tab  16 MiB @tab  64 MiB @tab 256 MiB
@item -1 @tab   4 MiB @tab   8 MiB @tab  16 MiB @tab  32 MiB @tab 128 MiB @tab 512 MiB
@item -2 @tab   6 MiB @tab  12 MiB @tab  24 MiB @tab  48 MiB @tab 192 MiB @tab 768 MiB
@item -3 @tab   8 MiB @tab  16 MiB @tab  32 MiB @tab  64 MiB @tab 256 MiB @tab   1 GiB
@item -4 @tab  12 MiB @tab  24 MiB @tab  48 MiB @tab  96 MiB @tab 384 MiB @tab 1.5 GiB
@item -5 @tab  16 MiB @tab  32 MiB @tab  64 MiB @tab 128 MiB @tab 512 MiB @tab   2 GiB
@item -6 @tab  32 MiB @tab  64 MiB @tab 128 MiB @tab 256 MiB @tab   1 GiB @tab   4 GiB
@item -7 @tab  64 MiB @tab 128 MiB @tab 256 MiB @tab 512 MiB @tab   2 GiB @tab   8 GiB
@item -8 @tab  96 MiB @tab 192 MiB @tab 384 MiB @tab 768 MiB @tab   3 GiB @tab  12 GiB
@item -9 @tab 128 MiB @tab 256 MiB @tab 512 MiB @tab   1 GiB @tab   4 GiB @tab  16 GiB
@end multitable


@node Examples
@chapter A small tutorial with examples
@cindex examples

@noindent
Example 1: Create a multimember compressed archive @samp{archive.tar.lz}
containing files @samp{a}, @samp{b} and @samp{c}.

@example
tarlz -cf archive.tar.lz a b c
@end example

@sp 1
@noindent
Example 2: Append files @samp{d} and @samp{e} to the multimember
compressed archive @samp{archive.tar.lz}.

@example
tarlz -rf archive.tar.lz d e
@end example

@sp 1
@noindent
Example 3: Create a solidly compressed appendable archive
@samp{archive.tar.lz} containing files @samp{a}, @samp{b} and @samp{c}.
Then append files @samp{d} and @samp{e} to the archive.

@example
tarlz --asolid -cf archive.tar.lz a b c
tarlz --asolid -rf archive.tar.lz d e
@end example

@sp 1
@noindent
Example 4: Create a compressed appendable archive containing directories
@samp{dir1}, @samp{dir2} and @samp{dir3} with a separate lzip member per
directory. Then append files @samp{a}, @samp{b}, @samp{c}, @samp{d} and
@samp{e} to the archive, all of them contained in a single lzip member.
The resulting archive @samp{archive.tar.lz} contains 5 lzip members
(including the EOF member).

@example
tarlz --dsolid -cf archive.tar.lz dir1 dir2 dir3
tarlz --asolid -rf archive.tar.lz a b c d e
@end example

@sp 1
@noindent
Example 5: Create a solidly compressed archive @samp{archive.tar.lz}
containing files @samp{a}, @samp{b} and @samp{c}. Note that no more
files can be later appended to the archive.

@example
tarlz --solid -cf archive.tar.lz a b c
@end example

@sp 1
@noindent
Example 6: Extract all files from archive @samp{archive.tar.lz}.

@example
tarlz -xf archive.tar.lz
@end example

@sp 1
@noindent
Example 7: Extract files @samp{a} and @samp{c}, and the whole tree under
directory @samp{dir1} from archive @samp{archive.tar.lz}.

@example
tarlz -xf archive.tar.lz a c dir1
@end example

@sp 1
@noindent
Example 8: Copy the contents of directory @samp{sourcedir} to the
directory @samp{destdir}.

@example
tarlz -C sourcedir -c . | tarlz -C destdir -x
@end example


@node Problems
@chapter Reporting bugs
@cindex bugs
@cindex getting help

There are probably bugs in tarlz. There are certainly errors and
omissions in this manual. If you report them, they will get fixed. If
you don't, no one will ever know about them and they will remain unfixed
for all eternity, if not longer.

If you find a bug in tarlz, please send electronic mail to
@email{lzip-bug@@nongnu.org}. Include the version number, which you can
find by running @w{@samp{tarlz --version}}.


@node Concept index
@unnumbered Concept index

@printindex cp

@bye