lzip/tarlz
1
0
Fork 0
Code Issues Activity

Adding upstream version 0.17.

Browse source 
Signed-off-by: Daniel Baumann <daniel@debian.org>
This commit is contained in:
Daniel Baumann 2025-02-17 21:15:00 +01:00
parent bb26c2917c
commit 739f200278
Signed by: daniel
GPG key ID: FBB4F0E80A80222F
29 changed files with 2935 additions and 2272 deletions
Download patch file Download diff file Expand all files Collapse all files

24
doc/tarlz.1
View file

@ -1,5 +1,5 @@
.\" DO NOT MODIFY THIS FILE! It was generated by help2man 1.46.1.
.TH TARLZ "1" "October 2019" "tarlz 0.16" "User Commands"
.TH TARLZ "1" "July 2020" "tarlz 0.17" "User Commands"
.SH NAME
tarlz \- creates tar archives with multimember lzip compression
.SH SYNOPSIS
@ -64,9 +64,15 @@ use archive file <archive>
\fB\-h\fR, \fB\-\-dereference\fR
follow symlinks; archive the files they point to
.TP
\fB\-\-mtime=\fR<date>
use <date> as mtime for files added to archive
.TP
\fB\-n\fR, \fB\-\-threads=\fR<n>
set number of (de)compression threads [2]
.TP
\fB\-p\fR, \fB\-\-preserve\-permissions\fR
don't subtract the umask on extraction
.TP
\fB\-q\fR, \fB\-\-quiet\fR
suppress all messages
.TP
@ -107,10 +113,10 @@ create solidly compressed archive
equivalent to '\-\-owner=root \fB\-\-group\fR=\fI\,root\/\fR'
.TP
\fB\-\-owner=\fR<owner>
use <owner> name/ID for files added
use <owner> name/ID for files added to archive
.TP
\fB\-\-group=\fR<group>
use <group> name/ID for files added
use <group> name/ID for files added to archive
.TP
\fB\-\-keep\-damaged\fR
don't delete partially extracted files
@ -121,17 +127,17 @@ exit with error status if missing extended CRC
\fB\-\-out\-slots=\fR<n>
number of 1 MiB output packets buffered [64]
.PP
Exit status: 0 for a normal exit, 1 for environmental problems (file
not found, 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.
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.
.SH "REPORTING BUGS"
Report bugs to lzip\-bug@nongnu.org
.br
Tarlz home page: http://www.nongnu.org/lzip/tarlz.html
.SH COPYRIGHT
Copyright \(co 2019 Antonio Diaz Diaz.
Using lzlib 1.11
Copyright \(co 2020 Antonio Diaz Diaz.
Using lzlib 1.12\-rc1a
License GPLv2+: GNU GPL version 2 or later <http://gnu.org/licenses/gpl.html>
.br
This is free software: you are free to change and redistribute it.

1024
doc/tarlz.info
View file

File diff suppressed because it is too large Load diff

338
doc/tarlz.texi
View file

@ -6,8 +6,8 @@
@finalout
@c %**end of header
@set UPDATED 8 October 2019
@set VERSION 0.16
@set UPDATED 30 July 2020
@set VERSION 0.17
@dircategory Data Compression
@direntry
@ -40,7 +40,8 @@ This manual is for Tarlz (version @value{VERSION}, @value{UPDATED}).
* 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
* Multi-threaded tar:: Limitations of parallel tar decoding
* 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
@ -48,10 +49,10 @@ This manual is for Tarlz (version @value{VERSION}, @value{UPDATED}).
@end menu
@sp 1
Copyright @copyright{} 2013-2019 Antonio Diaz Diaz.
Copyright @copyright{} 2013-2020 Antonio Diaz Diaz.
This manual is free documentation: you have unlimited permission
to copy, distribute and modify it.
to copy, distribute, and modify it.
@node Introduction
@ -77,7 +78,8 @@ 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).
(---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
@ -105,16 +107,16 @@ 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
individually compressing files smaller than about 32 KiB.
compressing files smaller than about 32 KiB individually.
@end itemize
Tarlz protects the extended records with a CRC in a way compatible with
standard tar tools. @xref{crc32}.
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}. @w{@samp{tarlz -tf archive.tar.lz > /dev/null}}
can be used to verify that the format of the archive is compatible with
tarlz.
@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
@ -151,7 +153,11 @@ 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 options:
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
@ -177,7 +183,7 @@ 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 @samp{--bsolid} option.
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}.
@ -210,7 +216,7 @@ 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 @samp{-C} option may
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.
@ -220,14 +226,18 @@ Make @samp{--diff} ignore differences in owner and group IDs. This option is
useful when comparing an @samp{--anonymous} archive.
@item --delete
Delete the specified files and directories from an archive in place. It
currently can delete only from uncompressed archives and from archives with
individually compressed files (@samp{--no-solid} archives). Note that files
of about @samp{--data-size} or larger are compressed individually even if
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 individually compressed. To delete a
directory without deleting the files under it, use
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.
@ -241,14 +251,22 @@ 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}. @samp{-} used as an @var{archive} argument
reads from standard input or writes to standard output.
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.
@ -256,15 +274,22 @@ 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 @samp{-C} option above.
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 tar}.
@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.
@ -298,7 +323,10 @@ Verbosely list files processed.
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}}.
@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
@ -411,9 +439,9 @@ keyword appearing in the same block of extended records.
@end table
Exit status: 0 for a normal exit, 1 for environmental problems (file not
found, 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.
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
@ -431,12 +459,16 @@ 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
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
@ -444,6 +476,7 @@ In the diagram below, a box like this:
@end verbatim
represents one byte; a box like this:
@verbatim
+==============+
| |
@ -486,7 +519,7 @@ Zero or more blocks that contain the contents of the file.
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 tar}.
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
@ -530,28 +563,29 @@ An extended header just before the EOF blocks.
@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 size field
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}.
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 <length>, <blank>, <keyword>, <equals-sign>, and <newline> in the
record must be limited to the portable character set. The <length> field
contains the decimal length of the record in bytes, including the
trailing <newline>. The <value> field is stored as-is, without
conversion to UTF-8 nor any other transformation.
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> fields currently supported by tarlz:
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 linkname field in the
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
@ -559,8 +593,8 @@ 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 name and
prefix fields in the following ustar header block.
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
@ -610,31 +644,30 @@ shown in the following table. All lengths and offsets are in decimal.
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. 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.
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 name, linkname, and prefix are null-terminated character
strings except when all characters in the array contain non-null
characters including the last character.
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 name and the prefix fields produce the pathname of the file. A new
pathname is formed, if prefix is not an empty string (its first
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.
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 linkname field does not use the prefix to produce a pathname. If the
linkname does not fit in the 100 characters provided, an extended record
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 mode field provides 12 access permission bits. The following table
shows the symbolic name of each bit and its octal value:
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
@ -644,29 +677,28 @@ shows the symbolic name of each bit and its octal value:
@item S_IROTH @tab 00004 @tab S_IWOTH @tab 00002 @tab S_IXOTH @tab 00001
@end multitable
The uid and gid fields are the user and group ID of the owner and group
of the file, respectively.
The fields @samp{uid} and @samp{gid} are the user and group IDs of the owner
and group of the file, respectively.
The size field contains the octal representation of the size of the file
in bytes. If the typeflag field specifies a file of type '0' (regular
file) or '7' (high performance regular file), the number of logical
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.
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 mtime field contains the octal representation of the modification
time of the file at the time it was archived, obtained from the stat()
function.
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 chksum field 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{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 typeflag field contains a single character specifying the type of
The field @samp{typeflag} contains a single character specifying the type of
file archived:
@table @code
@ -680,8 +712,8 @@ Hard link to another file, of any type, previously archived.
Symbolic link.
@item '3', '4'
Character special file and block special file respectively. In this case
the devmajor and devminor fields contain information defining the
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'
@ -697,14 +729,15 @@ regular file (type 0).
@end table
The magic field contains the ASCII null-terminated string "ustar". The
version field contains the characters "00" (0x30,0x30). The fields uname,
and 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.
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
@ -714,10 +747,10 @@ ustar by not requiring a terminating null character.
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{--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.
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}
@ -735,7 +768,7 @@ 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 more probable.
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
@ -753,12 +786,12 @@ 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 name and prefix
fields 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.
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
@ -778,10 +811,10 @@ between the extended header and the ustar header.
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 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.
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
@ -797,7 +830,88 @@ accidental double UTF-8 conversions. If the need arises this behavior will
be adjusted with a command line option in the future.
@node Multi-threaded tar
@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
@ -827,8 +941,8 @@ 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 @samp{--list}
option is able to do multi-threaded decoding.
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
@ -843,6 +957,10 @@ 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
@ -860,7 +978,7 @@ least as large as the number of worker threads times the block size
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 12 processors at level -9.
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