From aa976a517088ba75fa0168b47803b3988f89c9c3 Mon Sep 17 00:00:00 2001 From: Paul Eggert Date: Tue, 29 Nov 2005 19:14:21 +0000 Subject: [PATCH] * NEWS: New option --atime-preserve=system, which uses O_NOATIME. * configure.ac: Prefer AC_CHECK_HEADERS_ONCE to AC_CHECK_HEADERS. Check for stropts.h and sys/filio.h too, for _FIOSATIME. * doc/tar.texi: Change "modification time" to "data modification time", "change time" to "status change time", and "filesystem" to "file system", so that we use terminology consistent with POSIX. Use American spacing rather than French for sentence ends. "non-dependable" -> "undependable". (Option Summary, Attributes): Explain better the pitfalls of the --atime-preserve option, and suggest read-only mounts,loopback mounts, and noatime mounts for older systems. * doc/value.texi (op-atime-preserve-system): Renamed from op-atime-preserver-system to fix a misspelling. * src/common.h (enum atime_preserve): Use lower case for enum values. * src/compare.c: Don't include utimens.h; no longer needed. (diff_file): Use set_file_atime rather than utimens; avoid closing diff_handle until after this, so that we can set the file time stamp via the file descriptor rather than via its name. * src/create.c: Don't include utimens.h; no longer needed. (dump_regular_finish): Remove. All callers now do its work inline. (dump_dir): New arg FD. All callers changed. Use fdsavedir rather than savedir. (unknown_file_error): Arg is a const pointer now. (dump_file0): 2nd arg is a const pointer now. Treat directories more like files, with respect to --atime-preserve. For example, also warn if a directory changes while we are dumping it. Prefer file descriptors to file names when retrieving/setting file attributes; this saves path-resolution time and allows us to avoid changing mtime/ctime on Solaris when restoring atime as root. Use O_DIRECTORY when opening directories, to avoid some race conditions. Do not reset atime if mtime has changed. Report an error if we cannot reset atime. First cut at adding support for --atime-preserve=system. * doc/tar.texi (Option Summary): First cut at documenting it. All other uses of --atime-preserve changed to --atime-preserve=replace. * doc/value.texi (op-atime-preserve-replace, op-atime-preserver-system): New. (op-atime-preserve): Mention METHOD. * src/common.h (atime_preserve): New enum. (atime_preserve_option): Now of the enum type rather than bool. All uses changed. * src/compare.c (diff_file): Read with O_NOATIME if asked for. * src/create.c (dump_file0): Read regular and CTG files with O_NOATIME if asked for. * src/tar.c (usage): Mention new usage. (parse_opt): Parse new usage. --- ChangeLog | 93 +++++++++++++-- NEWS | 4 + THANKS | 1 + configure.ac | 6 +- doc/tar.texi | 295 ++++++++++++++++++++++++++++------------------- doc/value.texi | 10 +- src/common.h | 8 +- src/compare.c | 36 +++--- src/create.c | 280 +++++++++++++++++++++++--------------------- src/misc.c | 28 +++++ src/tar.c | 35 +++++- tests/ignfail.at | 4 +- 12 files changed, 508 insertions(+), 292 deletions(-) diff --git a/ChangeLog b/ChangeLog index 385fa94..d6ddd43 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,61 @@ +2005-11-29 Paul Eggert + + * NEWS: New option --atime-preserve=system, which uses O_NOATIME. + + * configure.ac: Prefer AC_CHECK_HEADERS_ONCE to AC_CHECK_HEADERS. + Check for stropts.h and sys/filio.h too, for _FIOSATIME. + * doc/tar.texi: Change "modification time" to "data modification + time", "change time" to "status change time", and "filesystem" to + "file system", so that we use terminology consistent with POSIX. + Use American spacing rather than French for sentence ends. + "non-dependable" -> "undependable". + (Option Summary, Attributes): Explain better the pitfalls of the + --atime-preserve option, and suggest read-only mounts,loopback + mounts, and noatime mounts for older systems. + * doc/value.texi (op-atime-preserve-system): Renamed from + op-atime-preserver-system to fix a misspelling. + * src/common.h (enum atime_preserve): Use lower case for enum values. + * src/compare.c: Don't include utimens.h; no longer needed. + (diff_file): Use set_file_atime rather than utimens; avoid closing + diff_handle until after this, so that we can set the file time stamp + via the file descriptor rather than via its name. + * src/create.c: Don't include utimens.h; no longer needed. + (dump_regular_finish): Remove. All callers now do its work inline. + (dump_dir): New arg FD. All callers changed. + Use fdsavedir rather than savedir. + (unknown_file_error): Arg is a const pointer now. + (dump_file0): 2nd arg is a const pointer now. + Treat directories more like files, with respect to --atime-preserve. + For example, also warn if a directory changes while we are dumping it. + Prefer file descriptors to file names when retrieving/setting file + attributes; this saves path-resolution time and allows us to avoid + changing mtime/ctime on Solaris when restoring atime as root. + Use O_DIRECTORY when opening directories, to avoid some race conditions. + Do not reset atime if mtime has changed. Report an error if + we cannot reset atime. + +2005-11-29 Ian Turner + + First cut at adding support for --atime-preserve=system. + * doc/tar.texi (Option Summary): First cut at documenting it. + All other uses of --atime-preserve changed to --atime-preserve=replace. + * doc/value.texi (op-atime-preserve-replace, op-atime-preserver-system): + New. + (op-atime-preserve): Mention METHOD. + * src/common.h (atime_preserve): New enum. + (atime_preserve_option): Now of the enum type rather than bool. + All uses changed. + * src/compare.c (diff_file): Read with O_NOATIME if asked for. + * src/create.c (dump_file0): Read regular and CTG files with O_NOATIME + if asked for. + * src/tar.c (usage): Mention new usage. + (parse_opt): Parse new usage. + +2005-11-29 Paul Eggert + + * THANKS: Convert back to UTF-8, sort (using LC_ALL=C on Debian + stable), and consistently use tabs rather than spaces. + 2005-11-27 Sergey Poznyakoff * src/xheader.c: Remove parts of code prematurely introduced @@ -23,7 +81,7 @@ tests/star/multi-fail.at: Close stdin so that if something fails causing tar to ask for the next volume, it won't hang the testsuite. - + * src/buffer.c (flush_write,flush_read): Change data type. (flush_archive): Compute actual buffer fill level before calling low level function. @@ -45,22 +103,22 @@ Correct buffer size calculation (allocated too much space). (xheader_write): Increase global_header_count here ... (xheader_write_global): ... instead of here - + * tests/testsuite.at (AT_TAR_CHECK): Define TEST_TAR_FORMAT * tests/multiv01.at: Update * tests/multiv02.at: Update * tests/multiv03.at: Update * tests/sparsemvp.at: Update * tests/star/multi-fail.at: Update - + * scripts/tarcat: Handle archives in pax format. Improve handling of traditional archives. * doc/tar.texi (Tarcat): New node - + 2005-11-10 Sergey Poznyakoff Fix splitting of sparse files between the volumes. - + * src/buffer.c (try_new_volume): Bugfix. Always check continued_file_name. If it is absent, the volume is out of sync. @@ -86,13 +144,13 @@ on the extended header boundary) are still not solved, others (splitting a sparse file between the volume) require additional testing. Wait for the next commit. - + (volume_label,continued_file_name,continued_file_size) (continued_file_offset): New globals. (save_name,save_totsize,save_sizeleft): Make static (mv_begin,mv_end,mv_total_size,mv_size_left): New functions (open_archive,flush_write,flush_read): Rewritten - + * src/common.h (save_name,save_sizeleft,save_totsize): Remove globals. (volume_label,continued_file_name,continued_file_size): New @@ -122,7 +180,19 @@ of related tests in groups. * doc/tar.texi: Update - + +2005-11-07 Paul Eggert + + * src/xheader.c (struct xhdr_tab.coder): Last arg is void const *, + not void *, to avoid create.c warning from GCC "passing argument 3 + of 'xheader_store' discards qualifiers from pointer target type". + (dummy_coder, atime_coder, gid_coder, gname_coder, linkpath_coder): + (ctime_coder, mtime_coder, path_coder, size_coder, uid_coder): + (uname_coder, sparse_size_coder, sparse_numblocks_coder): + (sparse_offset_coder, sparse_numbytes_coder, dumpdir_coder): + (xheader_store): Likewise. + * src/common.h (xheader_store): Likewise. + 2005-11-07 Sergey Poznyakoff and Paul Eggert @@ -188,6 +258,13 @@ 2005-11-06 Paul Eggert + * NEWS: Minor language and white space fixes. + + * tests/truncate.at: Create files whose time stamps must fall on + 1-second boundaries. This prevents tests from failing on hosts + like Solaris 8 that have nanosecond-resolution file time stamps. + * tests/update.at: Likewise. + * src/xheader.c (strtoimax, strtoumax): Remove decls; now done in system.h. diff --git a/NEWS b/NEWS index 5e21457..b6a39bc 100644 --- a/NEWS +++ b/NEWS @@ -38,6 +38,10 @@ it lists member names as stored in the archive, i.e., with any eventual prefixes removed. The option is useful, for example, while comparing `tar cv' and `tar tv' outputs. +* New option --atime-preserve=system, which uses the O_NOATIME feature +of recent Linux kernels to avoid some problems when preserving file +access times. + * Better support for full-resolution time stamps. Tar cannot restore time stamps to full nanosecond resolution, though, until the kernel guys get their act together and give us a system call to set file time diff --git a/THANKS b/THANKS index b547f6d..9395e1c 100644 --- a/THANKS +++ b/THANKS @@ -201,6 +201,7 @@ Hunyue Yau hunyue.yau@picksys.com Ian Jackson ijackson@gnu.org Ian Lance Taylor ian@cygnus.com Ian T. Zimmerman itz@crl.com +Ian Turner ian@zmanda.com Indra Singhal indra@synoptics.com J. Dean Brock brock@cs.unca.edu J.J. Bailey jjb@jagware.bcc.com diff --git a/configure.ac b/configure.ac index d6849fc..a82623c 100644 --- a/configure.ac +++ b/configure.ac @@ -34,9 +34,9 @@ AC_SYS_LARGEFILE AC_ISC_POSIX AC_C_INLINE -AC_CHECK_HEADERS(fcntl.h linux/fd.h memory.h net/errno.h \ - sgtty.h string.h \ - sys/param.h sys/device.h sys/gentape.h \ +AC_CHECK_HEADERS_ONCE(fcntl.h linux/fd.h memory.h net/errno.h \ + sgtty.h string.h stropts.h \ + sys/param.h sys/device.h sys/filio.h sys/gentape.h \ sys/inet.h sys/io/trioctl.h \ sys/mtio.h sys/time.h sys/tprintf.h sys/tape.h \ unistd.h locale.h) diff --git a/doc/tar.texi b/doc/tar.texi index d148581..5de2ba3 100644 --- a/doc/tar.texi +++ b/doc/tar.texi @@ -227,7 +227,7 @@ Changing How @command{tar} Writes Files * Keep Newer Files:: * Unlink First:: * Recursive Unlink:: -* Modification Times:: +* Data Modification Times:: * Setting Access Permissions:: * Writing to Standard Output:: * remove files:: @@ -262,7 +262,7 @@ Choosing Files and Names for @command{tar} * Wildcards:: * after:: Operating Only on New Files * recurse:: Descending into Directories -* one:: Crossing Filesystem Boundaries +* one:: Crossing File System Boundaries Reading Names from a File @@ -273,7 +273,7 @@ Excluding Some Files * controlling pattern-patching with exclude:: * problems with exclude:: -Crossing Filesystem Boundaries +Crossing File System Boundaries * directory:: Changing Directory * absolute:: Absolute File Names @@ -427,7 +427,7 @@ The @command{tar} program is used to create and manipulate @command{tar} archives. An @dfn{archive} is a single file which contains the contents of many files, while still identifying the names of the files, their owner(s), and so forth. (In addition, archives record access -permissions, user and group, size in bytes, and last modification time. +permissions, user and group, size in bytes, and data modification time. Some archives also record the file names in each archived directory, as well as other file and directory information.) You can use @command{tar} to @dfn{create} a new archive in a specified directory. @@ -440,14 +440,14 @@ The files inside an archive are called @dfn{members}. Within this manual, we use the term @dfn{file} to refer only to files accessible in the normal ways (by @command{ls}, @command{cat}, and so forth), and the term @dfn{member} to refer only to the members of an archive. Similarly, a -@dfn{file name} is the name of a file, as it resides in the filesystem, +@dfn{file name} is the name of a file, as it resides in the file system, and a @dfn{member name} is the name of an archive member within the archive. @cindex extraction @cindex unpacking The term @dfn{extraction} refers to the process of copying an archive -member (or multiple members) into a file in the filesystem. Extracting +member (or multiple members) into a file in the file system. Extracting all the members of an archive is often called @dfn{extracting the archive}. The term @dfn{unpack} can also be used to refer to the extraction of many or all the members of an archive. Extracting an @@ -511,7 +511,7 @@ projects) together on a disk or a tape. This guards against accidental destruction of the information in those files. @GNUTAR{} has special features that allow it to be used to make incremental and full dumps of all the files in a -filesystem. +file system. @item Transportation You can create an archive on one system, transfer it to another system, @@ -692,7 +692,7 @@ about how Unix-type operating systems work, and you should know how to use some basic utilities. For example, you should know how to create, list, copy, rename, edit, and delete files and directories; how to change between directories; and how to figure out where you are in the -filesystem. You should have some basic understanding of directory +file system. You should have some basic understanding of directory structure and how files are named according to which directory they are in. You should understand concepts such as standard output and standard input, what various definitions of the term ``argument'' mean, and the @@ -1473,7 +1473,7 @@ earlier (say, @file{blues}), you can extract it from the archive without changing the archive's structure. It will be identical to the original file @file{blues} that you deleted. @FIXME{At the time of this writing, atime and ctime are not restored. Since this is a tutorial -for a beginnig user, it should hardly be mentioned here. Maybe in +for a beginning user, it should hardly be mentioned here. Maybe in a footnote? --gray}. First, make sure you are in the @file{practice} directory, and list the @@ -1489,7 +1489,7 @@ $ @kbd{tar --extract --file=collection.tar blues} @noindent If you list the files in the directory again, you will see that the file -@file{blues} has been restored, with its original permissions, creation +@file{blues} has been restored, with its original permissions, data modification times, and owner.@FIXME{This is only accidentally true, but not in general. In most cases, one has to be root for restoring the owner, and use a special option for restoring permissions. Here, it just happens @@ -1735,7 +1735,7 @@ option). @value{xref-absolute-names}, for more information about If you give the name of a directory as either a file name or a member name, then @command{tar} acts recursively on all the files and directories beneath that directory. For example, the name @file{/} identifies all -the files in the filesystem to @command{tar}. +the files in the file system to @command{tar}. The distinction between file names and archive member names is especially important when shell globbing is used, and sometimes a source of confusion @@ -2214,16 +2214,51 @@ An exclude pattern must match an initial subsequence of the name's components. @FIXME-xref{} @item --atime-preserve - -Tells @command{tar} to preserve the access time field in a file's inode when -reading it. Due to limitations in the @code{utimes} system call, the -modification time field is also preserved, which may cause problems if -the file is simultaneously being modified by another program. -This option is incompatible with incremental backups, because -preserving the access time involves updating the last-changed time. -Also, this option does not work on files that you do not own, -unless you're root. -@FIXME-xref{} +@itemx --atime-preserve=replace +@itemx --atime-preserve=system + +Attempt to preserve the access time of files when reading them. This +option currently is effective only on files that you own, unless you +have superuser privileges. + +Though this option should work on recent Linux kernel versions, it is +not reliable on other platforms. To preserve the access time reliably +on those platforms, you can mount the file system read-only, or access +the file system via a read-only loopback mount, or use the +@samp{noatime} mount option available on some systems. However, +mounting typically requires superuser privileges and can be a pain to +manage, so the @option{--atime-preserve} option can be useful despite +its glitches on other platforms. + +@value{op-atime-preserve-replace} remembers the access time of a file +before reading it, and then restores the access time afterwards. This +may cause problems if other programs are reading the file at the same +time, as the times of their accesses will be lost. On most platforms +restoring the access time also requires @command{tar} to restore the +data modification time too, so this option may also cause problems if +other programs are writing the file at the same time. (Tar attempts +to detect this situation, but cannot do so reliably due to race +conditions.) Worse, on most platforms restoring the access time also +updates the status change time, which means that this option is +incompatible with incremental backups. + +@value{op-atime-preserve-system} avoids changing time stamps on files +other than directories, without interfering with time stamp updates +caused by other programs, so it works better with incremental backups. +However, it requires a special @code{O_NOATIME} option from the +underlying operating and file system implementation, and it requires +that searching directories does not update their access times. As of +this writing (November 2005) this works only in a few new Linux +kernels. Worse, there is currently no reliable way to know whether +the features actually work. Sometimes @command{tar} knows for sure +that the features are not working, so it will complain and exit right +away if you try to use @value{op-atime-preserve-system}; but other +times @command{tar} might think that the option is supported when it +is not actually working. + +Currently @option{--atime-preserve} with no operand defaults to +@value{op-atime-preserve-replace}, but this may change in the future +as support for @value{op-atime-preserve-system} improves. @item --backup=@var{backup-type} @@ -2403,7 +2438,7 @@ archive, which normally signals EOF. @xref{Reading}. Used to inform @command{tar} that it is working with an old @acronym{GNU}-format incremental backup archive. It is intended primarily for backwards compatibility only. @FIXME{incremental and -listed-incremental}. +listed-incremental}. @item --index-file=@var{file} @@ -2494,7 +2529,7 @@ in cases when such recognition fails. When creating an archive, @command{tar} will only add files that have changed since @var{date}. If @var{date} begins with @samp{/} or @samp{.}, it -is taken to be the name of a file whose last-modified time specifies +is taken to be the name of a file whose data modification time specifies the date. @FIXME-xref{} @item --newer-mtime=@var{date} @@ -2894,8 +2929,8 @@ Displays the total number of bytes written after creating an archive. @item --touch @itemx -m -Sets the modification time of extracted files to the extraction time, -rather than the modification time stored in the archive. +Sets the data modification time of extracted files to the extraction time, +rather than the data modification time stored in the archive. @xref{Writing}. @item --uncompress @@ -3538,7 +3573,7 @@ complex. @command{tar} @emph{allows} you to have infinite number of files with the same name. Some operations treat these same-named members no differently than any other set of archive members: for example, if you view an archive with @value{op-list}, you will see all of those members -listed, with their modification times, owners, etc. +listed, with their data modification times, owners, etc. Other operations don't deal with these members as perfectly as you might prefer; if you were to use @value{op-extract} to extract the archive, @@ -4018,7 +4053,7 @@ Do not exit with nonzero on unreadable files or directories. there's a better way of organizing them.} The previous chapter showed how to use @value{op-extract} to extract -an archive into the filesystem. Various options cause @command{tar} to +an archive into the file system. Various options cause @command{tar} to extract more information than just file contents, such as the owner, the permissions, the modification date, and so forth. This section presents options to be used with @option{--extract} when certain special @@ -4113,6 +4148,7 @@ encountered while reading an archive. Use in conjunction with @subsection Changing How @command{tar} Writes Files @cindex Overwriting old files, prevention @cindex Protecting old files +@cindex Data modification times of extracted files @cindex Modification times of extracted files @cindex Permissions of extracted files @cindex Modes of extracted files @@ -4129,7 +4165,7 @@ encountered while reading an archive. Use in conjunction with * Keep Newer Files:: * Unlink First:: * Recursive Unlink:: -* Modification Times:: +* Data Modification Times:: * Setting Access Permissions:: * Writing to Standard Output:: * remove files:: @@ -4161,7 +4197,7 @@ existing files and to follow existing symbolic links when extracting. Some people argue that @GNUTAR{} should not hesitate to overwrite files with other files when extracting. When extracting a @command{tar} archive, they expect to see a faithful copy of the -state of the filesystem when the archive was created. It is debatable +state of the file system when the archive was created. It is debatable that this would always be a proper behavior. For example, suppose one has an archive in which @file{usr/local} is a link to @file{usr/local2}. Since then, maybe the site removed the link and @@ -4260,22 +4296,22 @@ If you specify the @value{op-recursive-unlink} option, as far as current permissions will allow it. This could include removal of the contents of a full directory hierarchy. -@node Modification Times -@unnumberedsubsubsec Setting Modification Times +@node Data Modification Times +@unnumberedsubsubsec Setting Data Modification Times -Normally, @command{tar} sets the modification times of extracted files to -the modification times recorded for the files in the archive, but +Normally, @command{tar} sets the data modification times of extracted +files to the corresponding times recorded for the files in the archive, but limits the permissions of extracted files by the current @code{umask} setting. -To set the modification times of extracted files to the time when +To set the data modification times of extracted files to the time when the files were extracted, use the @value{op-touch} option in conjunction with @value{op-extract}. @table @option @item --touch @itemx -m -Sets the modification time of extracted archive members to the time +Sets the data modification time of extracted archive members to the time they were extracted, not the time recorded for them in the archive. Use in conjunction with @value{op-extract}. @end table @@ -4666,7 +4702,7 @@ called @dfn{dumps}. @cindex corrupted archives Full dumps should only be made when no other people or programs -are modifying files in the filesystem. If files are modified while +are modifying files in the file system. If files are modified while @command{tar} is making the backup, they may not be stored properly in the archive, in which case you won't be able to restore them if you have to. (Files not being modified are written with no trouble, and do @@ -4676,18 +4712,18 @@ You will want to use the @value{op-label} option to give the archive a volume label, so you can tell what this archive is even if the label falls off the tape, or anything like that. -Unless the filesystem you are dumping is guaranteed to fit on +Unless the file system you are dumping is guaranteed to fit on one volume, you will need to use the @value{op-multi-volume} option. Make sure you have enough tapes on hand to complete the backup. -If you want to dump each filesystem separately you will need to use +If you want to dump each file system separately you will need to use the @value{op-one-file-system} option to prevent @command{tar} from crossing -filesystem boundaries when storing (sub)directories. +file system boundaries when storing (sub)directories. The @value{op-incremental} (@FIXME-pxref{}) option is not needed, -since this is a complete copy of everything in the filesystem, and a +since this is a complete copy of everything in the file system, and a full restore from this backup would only be done onto a completely -empty disk. +empty disk. Unless you are in a hurry, and trust the @command{tar} program (and your tapes), it is a good idea to use the @value{op-verify} option, to make @@ -4700,16 +4736,16 @@ capable of being verified, unfortunately. @section Using @command{tar} to Perform Incremental Dumps @dfn{Incremental backup} is a special form of @GNUTAR{} archive that -stores additional metadata so that exact state of the filesystem -can be restored when extracting the archive. +stores additional metadata so that exact state of the file system +can be restored when extracting the archive. @GNUTAR{} currently offers two options for handling incremental -backups: @value{op-listed-incremental} and @value{op-incremental}. +backups: @value{op-listed-incremental} and @value{op-incremental}. The option @option{--listed-incremental} instructs tar to operate on an incremental archive with additional metadata stored in a standalone -file, called @dfn{snapshot file}. The purpose of this file is to help -determine what files have been changed, added or deleted since the +file, called a @dfn{snapshot file}. The purpose of this file is to help +determine which files have been changed, added or deleted since the last backup, so that the next incremental backup will contain only modified files. The name of the snapshot file is given as an argument to the option: @@ -4732,13 +4768,13 @@ $ @kbd{tar --create \ @end smallexample This will create in @file{archive.1.tar} an incremental backup of -@file{/usr} filesystem, storing additional metadata in the file +the @file{/usr} file system, storing additional metadata in the file @file{/var/log/usr.snar}. If this file does not exist, it will be -created. The created archive will then be called @dfn{level 0 backup} -(see the next section for more info on backup levels). +created. The created archive will then be a @dfn{level 0 backup}; +please see the next section for more on backup levels. -Otherwise, if the file @file{/var/log/usr.snar} exists, it is used to -determine the modified files. In this case only these files will be +Otherwise, if the file @file{/var/log/usr.snar} exists, it +determines which files are modified. In this case only these files will be stored in the archive. Suppose, for example, that after running the above command, you delete file @file{/usr/doc/old} and create directory @file{/usr/local/db} with the following contents: @@ -4749,7 +4785,7 @@ $ @kbd{ls /usr/local/db} /usr/local/db/index @end smallexample -Some time later you create another incremental backup. You will +Some time later you create another incremental backup. You will then see: @smallexample @@ -4765,7 +4801,7 @@ usr/local/db/index @noindent The created archive @file{archive.2.tar} will contain only these -three members. This archive is called @dfn{level 1 backup}. Notice, +three members. This archive is called a @dfn{level 1 backup}. Notice that @file{/var/log/usr.snar} will be updated with the new data, so if you plan to create more @samp{level 1} backups, it is necessary to create a working copy of the snapshot file before running @@ -4781,17 +4817,17 @@ $ @kbd{tar --create \ Incremental dumps depend crucially on time stamps, so the results are unreliable if you modify a file's time stamps during dumping (e.g., -with the @option{--atime-preserve} option), or if you set the clock +with the @option{--atime-preserve=replace} option), or if you set the clock backwards. Metadata stored in snapshot files include device numbers, which, obviously is supposed to be a non-volatile value. However, it turns -out that NFS devices have non-dependable values when an automounter +out that NFS devices have undependable values when an automounter gets in the picture. This can lead to a great deal of spurious redumping in incremental dumps, so it is somewhat useless to compare two NFS devices numbers over time. The solution implemented currently is to considers all NFS devices as being equal when it comes to -comparing directories; this is fairly gross, but there does not seem +comparing directories; this is fairly gross, but there does not seem to be a better way to go. Note that incremental archives use @command{tar} extensions and may @@ -4799,9 +4835,9 @@ not be readable by non-@acronym{GNU} versions of the @command{tar} program. To extract from the incremental dumps, use @option{--listed-incremental} together with @option{--extract} -option (@pxref{extracting files}). In this case, @command{tar} does +option (@pxref{extracting files}). In this case, @command{tar} does not need to access snapshot file, since all the data necessary for -extraction are stored in the archive itself. So, when extracting, you +extraction are stored in the archive itself. So, when extracting, you can give whatever argument to @option{--listed-incremental}, the usual practice is to use @option{--listed-incremental=/dev/null}. Alternatively, you can use @option{--incremental}, which needs no @@ -4817,10 +4853,10 @@ system that did not exist in their directories when the archive was created. If you have created several levels of incremental files, then in order to restore the exact contents the file system had when the last level was created, you will need to restore from all backups -in turn. Continuing our example, to restore the state of @file{/usr} +in turn. Continuing our example, to restore the state of @file{/usr} file system, one would do@footnote{Notice, that since both archives were created withouth @option{-P} option (@pxref{absolute}), these -commands should be run from the root filesystem.}: +commands should be run from the root file system.}: @smallexample $ @kbd{tar --extract \ @@ -4832,7 +4868,7 @@ $ @kbd{tar --extract \ @end smallexample To list the contents of an incremental archive, use @option{--list} -(@pxref{list}), as usual. To obtain more information about the +(@pxref{list}), as usual. To obtain more information about the archive, use @option{--listed-incremental} or @option{--incremental} combined with two @option{--verbose} options@footnote{Two @option{--verbose} options were selected to avoid breaking usual @@ -4865,7 +4901,7 @@ if the file is present in the archive, @samp{N} if the file is not included in the archive, or a @samp{D} if the file is a directory (and is included in the archive).@FIXME-xref{dumpdir format}. Each such line is terminated by a newline character. The last line is followed -by an additional newline to indicate the end of the data. +by an additional newline to indicate the end of the data. @anchor{incremental-op}The option @option{--incremental} (@option{-G}) gives the same behavior as @option{--listed-incremental} when used @@ -4919,7 +4955,7 @@ designed to be used together. While it is possible to restore files by hand from an archive which was created using a backup script, and to create an archive by hand which could then be extracted using the restore script, it is easier to use the scripts. @value{xref-incremental}, before -making such an attempt. +making such an attempt. @node Backup Parameters @section Setting Parameters for Backups and Restoration @@ -5006,7 +5042,7 @@ in a separate file. This file is usually named @defvr {Backup variable} DIRLIST -A path to the file containing the list of the filesystems to backup +A path to the file containing the list of the file systems to backup or restore. By default it is @file{/etc/backup/dirs}. @end defvr @@ -5050,7 +5086,7 @@ of @GNUTAR{}. @defvr {Backup variable} VOLNO_FILE Name of temporary file to hold volume numbers. This needs to be accessible -by all the machines which have filesystems to be dumped. +by all the machines which have file systems to be dumped. @end defvr @defvr {Backup variable} XLIST @@ -5067,7 +5103,7 @@ This variable affects only @code{backup}. @defvr {Backup variable} SLEEP_TIME -Time to sleep between dumps of any two successive filesystems +Time to sleep between dumps of any two successive file systems This variable affects only @code{backup}. @end defvr @@ -5174,10 +5210,10 @@ Current backup or restore level. Name or IP address of the host machine being dumped or restored. @item fs -Full path name to the filesystem being dumped or restored. +Full path name to the file system being dumped or restored. @item fsname -Filesystem name with directory separators replaced with colons. This +File system name with directory separators replaced with colons. This is useful, e.g., for creating unique files. @end table @end deffn @@ -5185,19 +5221,19 @@ is useful, e.g., for creating unique files. Following variables keep the names of user hook functions @defvr {Backup variable} DUMP_BEGIN -Dump begin function. It is executed before dumping the filesystem. +Dump begin function. It is executed before dumping the file system. @end defvr @defvr {Backup variable} DUMP_END -Executed after dumping the filesystem. +Executed after dumping the file system. @end defvr @defvr {Backup variable} RESTORE_BEGIN -Executed before restoring the filesystem. +Executed before restoring the file system. @end defvr @defvr {Backup variable} RESTORE_END -Executed after restoring the filesystem. +Executed after restoring the file system. @end defvr @node backup-specs example @@ -5353,10 +5389,10 @@ Display program version and exit. To restore files that were archived using a scripted backup, use the @code{restore} script. Its usage is quite straightforward. In the simplest form, invoke @code{restore --all}, it will -then restore all the filesystems and files specified in +then restore all the file systems and files specified in @file{backup-specs} (@pxref{General-Purpose Variables,BACKUP_DIRS}). -You may select the filesystems (and/or files) to restore by +You may select the file systems (and/or files) to restore by giving @code{restore} list of @dfn{patterns} in its command line. For example, running @@ -5365,7 +5401,7 @@ restore 'albert:*' @end smallexample @noindent -will restore all filesystems on the machine @samp{albert}. A more +will restore all file systems on the machine @samp{albert}. A more complicated example: @smallexample @@ -5373,8 +5409,8 @@ restore 'albert:*' '*:/var' @end smallexample @noindent -This command will restore all filesystems on the machine @samp{albert} -as well as @file{/var} filesystem on all machines. +This command will restore all file systems on the machine @samp{albert} +as well as @file{/var} file system on all machines. By default @code{restore} will start restoring files from the lowest available dump level (usually zero) and will continue through @@ -5392,7 +5428,7 @@ The full list of options accepted by @code{restore} follows: @table @option @item -a @itemx --all -Restore all filesystems and files specified in @file{backup-specs} +Restore all file systems and files specified in @file{backup-specs} @item -l @var{level} @itemx --level=@var{level} @@ -5431,7 +5467,7 @@ system if they were not in the file system when the archive was made. @end quotation @value{xref-incremental}, for an explanation of how the script makes -that determination. +that determination. @node Choosing @chapter Choosing Files and Names for @command{tar} @@ -5453,7 +5489,7 @@ are in specified directories. * Wildcards:: * after:: Operating Only on New Files * recurse:: Descending into Directories -* one:: Crossing Filesystem Boundaries +* one:: Crossing File System Boundaries @end menu @node file @@ -5959,7 +5995,7 @@ patterns listed in a file. @samp{*} or @samp{?} for example, are replaced and expanded into all existing files matching the given pattern. However, @command{tar} often uses wildcard patterns for matching (or globbing) archive members instead -of actual files in the filesystem. Wildcard patterns are also used for +of actual files in the file system. Wildcard patterns are also used for verifying volume labels of @command{tar} archives. This section has the purpose of explaining wildcard syntax for @command{tar}. @@ -6009,21 +6045,22 @@ string: excluding a directory also excludes all the files beneath it. @node after @section Operating Only on New Files @cindex Excluding file by age +@cindex Data Modification time, excluding files by @cindex Modification time, excluding files by @cindex Age, excluding files by @UNREVISED The @value{op-after-date} option causes @command{tar} to only work on files -whose modification or inode-changed times are newer than the @var{date} +whose data modification or status change times are newer than the @var{date} given. If @var{date} starts with @samp{/} or @samp{.}, it is taken to -be a file name; the last-modified time of that file is used as the date. +be a file name; the data modification time of that file is used as the date. If you use this option when creating or appending to an archive, the archive will only include new files. If you use @option{--after-date} when extracting an archive, @command{tar} will only extract files newer than the @var{date} you specify. If you only want @command{tar} to make the date comparison based on -modification of the actual contents of the file (rather than inode +modification of the file's data (rather than status changes), then use the @value{op-newer-mtime} option. You may use these options with any operation. Note that these options @@ -6037,27 +6074,29 @@ deciding whether or not to archive the files. @itemx -N @var{date} Only store files newer than @var{date}. -Acts on files only if their modification or inode-changed times are +Acts on files only if their data modification or status change times are later than @var{date}. Use in conjunction with any operation. If @var{date} starts with @samp{/} or @samp{.}, it is taken to be a file -name; the last-modified time of that file is used as the date. +name; the data modification time of that file is used as the date. @item --newer-mtime=@var{date} -Acts like @value{op-after-date}, but only looks at modification times. +Acts like @value{op-after-date}, but only looks at data modification times. @end table -These options limit @command{tar} to only operating on files which have -been modified after the date specified. A file is considered to have -changed if the contents have been modified, or if the owner, +These options limit @command{tar} to operate only on files which have +been modified after the date specified. A file's status is considered to have +changed if its contents have been modified, or if its owner, permissions, and so forth, have been changed. (For more information on how to specify a date, see @ref{Date input formats}; remember that the entire date argument must be quoted if it contains any spaces.) -Gurus would say that @value{op-after-date} tests both the @code{mtime} -(time the contents of the file were last modified) and @code{ctime} -(time the file's status was last changed: owner, permissions, etc) -fields, while @value{op-newer-mtime} tests only @code{mtime} field. +Gurus would say that @value{op-after-date} tests both the data +modification time (@code{mtime}, the time the contents of the file +were last modified) and the status change time (@code{ctime}, the time +the file's status was last changed: owner, permissions, etc.@:) +fields, while @value{op-newer-mtime} tests only the @code{mtime} +field. To be precise, @value{op-after-date} checks @emph{both} @code{mtime} and @code{ctime} and processes the file if either one is more recent than @@ -6153,7 +6192,7 @@ contents of @file{grape/concord}, but no entries under @file{grape} other than @file{grape/concord}. @node one -@section Crossing Filesystem Boundaries +@section Crossing File System Boundaries @cindex File system boundaries, not crossing @UNREVISED @@ -6174,7 +6213,7 @@ archiving. Use in conjunction with any write operation. The @option{--one-file-system} option causes @command{tar} to modify its normal behavior in archiving the contents of directories. If a file in -a directory is not on the same filesystem as the directory itself, then +a directory is not on the same file system as the directory itself, then @command{tar} will not archive that file. If the file is a directory itself, @command{tar} will not archive anything beneath it; in other words, @command{tar} will not cross mount points. @@ -6566,7 +6605,7 @@ than System V's. Normally, when @command{tar} archives a symbolic link, it writes a block to the archive naming the target of the link. In that way, the -@command{tar} archive is a faithful record of the filesystem contents. +@command{tar} archive is a faithful record of the file system contents. @value{op-dereference} is used with @value{op-create}, and causes @command{tar} to archive the files symbolic links point to, instead of the links themselves. When this option is used, when @command{tar} @@ -6930,7 +6969,7 @@ treatment of sparse files may be done automatically with any special @acronym{GNU} options. For now, it is an option needing to be specified on the command line with the creation or updating of an archive. -Files in the filesystem occasionally have ``holes.'' A hole in a file +Files in the file system occasionally have ``holes.'' A hole in a file is a section of the file's contents which was never written. The contents of a hole read as all zeros. On many operating systems, actual disk storage is not allocated for holes, but they are counted @@ -7027,29 +7066,44 @@ get it right. @section Handling File Attributes @UNREVISED -When @command{tar} reads files, this causes them to have the access -times updated. To have @command{tar} attempt to set the access times -back to what they were before they were read, use the -@value{op-atime-preserve} option. +When @command{tar} reads files, it updates their access times. To +avoid this, use the @value{op-atime-preserve} option, which can either +reset the access time retroactively or avoid changing it in the first +place. Handling of file attributes @table @option @item --atime-preserve -Preserve access times on files that are read. -This doesn't work for files that -you don't own, unless you're root, and it doesn't interact with -incremental dumps nicely (@pxref{Backups}), and it can set access or -modification times incorrectly if other programs access the file while -@command{tar} is running; but it is good enough for some purposes. +@itemx --atime-preserve=replace +@itemx --atime-preserve=system +Preserve the access times of files that are read. This works only for +files that you own, unless you have superuser privileges. + +@value{op-atime-preserve-replace} works on most systems, but it also +restores the data modification time and updates the status change +time. Hence it doesn't interact with incremental dumps nicely +(@pxref{Backups}), and it can set access or data modification times +incorrectly if other programs access the file while @command{tar} is +running. + +@value{op-atime-preserve-system} avoids changing the access time in +the first place, if the operating system supports this. +Unfortunately, this may or may not work on any given operating system +or file system. If @command{tar} knows for sure it won't work, it +complains right away. + +Currently @option{--atime-preserve} with no operand defaults to +@value{op-atime-preserve-replace}, but this is intended to change to +@value{op-atime-preserve-system} when the latter is better-supported. @item -m @itemx --touch -Do not extract file modified time. +Do not extract data modification time. -When this option is used, @command{tar} leaves the modification times -of the files it extracts as the time when the files were extracted, -instead of setting it to the time recorded in the archive. +When this option is used, @command{tar} leaves the data modification times +of the files it extracts as the times when the files were extracted, +instead of setting it to the times recorded in the archive. This option is meaningless with @value{op-list}. @@ -7090,7 +7144,7 @@ This is useful in certain circumstances, when restoring a backup from an emergency floppy with different passwd/group files for example. It is otherwise impossible to extract files with the right ownerships if the password file in use during the extraction does not match the -one belonging to the filesystem(s) being extracted. This occurs, +one belonging to the file system(s) being extracted. This occurs, for example, if you are restoring your files after a major crash and had booted from an emergency floppy with no password file or put your disk into another machine to do the restore. @@ -7244,9 +7298,10 @@ The @code{size} field is the size of the file in bytes; linked files are archived with this field specified as zero. @FIXME-xref{Modifiers, in particular the @value{op-incremental} option.} -The @code{mtime} field is the modification time of the file at the time -it was archived. It is the ASCII representation of the octal value of -the last time the file was modified, represented as an integer number of +The @code{mtime} field is the data modification time of the file at +the time it was archived. It is the ASCII representation of the octal +value of the last time the file's contents were modified, represented +as an integer number of seconds since January 1, 1970, 00:00 Coordinated Universal Time. The @code{chksum} field is the ASCII representation of the octal value @@ -7262,8 +7317,8 @@ type, the file will be extracted as if it were a regular file. As this action occurs, @command{tar} issues a warning to the standard error. The @code{atime} and @code{ctime} fields are used in making incremental -backups; they store, respectively, the particular file's access time -and last inode-change time. +backups; they store, respectively, the particular file's access and +status change times. The @code{offset} is used by the @value{op-multi-volume} option, when making a multi-volume archive. The offset is number of bytes into @@ -8426,7 +8481,7 @@ failed. Often you might want to write a large archive, one larger than will fit on the actual tape you are using. In such a case, you can run multiple @command{tar} commands, but this can be inconvenient, particularly if you -are using options like @value{op-exclude} or dumping entire filesystems. +are using options like @value{op-exclude} or dumping entire file systems. Therefore, @command{tar} supports multiple tapes automatically. Use @value{op-multi-volume} on the command line, and then @command{tar} will, @@ -8670,7 +8725,7 @@ the first volume file and to decide how to process the rest of the files. However, it makes no attempt to verify whether the files are given in order or even if they are valid @command{tar} archives. It uses @command{dd} and does not filter its standard error, so you -will usually see lots of spurious messages. +will usually see lots of spurious messages. @FIXME{The script is not installed. Should we install it?} diff --git a/doc/value.texi b/doc/value.texi index 41f67c4..869f872 100644 --- a/doc/value.texi +++ b/doc/value.texi @@ -22,7 +22,9 @@ @set xref-append @xref{add} @set pxref-append @pxref{add} -@set op-atime-preserve @option{--atime-preserve} +@set op-atime-preserve @option{--atime-preserve[=METHOD]} +@set op-atime-preserve-replace @option{--atime-preserve=replace} +@set op-atime-preserve-system @option{--atime-preserve=system} @set ref-atime-preserve @ref{Attributes} @set xref-atime-preserve @xref{Attributes} @set pxref-atime-preserve @pxref{Attributes} @@ -174,7 +176,7 @@ @set xref-keep-old-files @xref{Keep Old Files} @set pxref-keep-old-files @pxref{Keep Old Files} -@set op-keep-newer-files @option{--keep-old-files} +@set op-keep-newer-files @option{--keep-old-files} @set ref-keep-newer-files @ref{Keep Newer Files} @set xref-keep-newer-files @xref{Keep Newer Files} @set pxref-keep-newer-files @pxref{Keep Newer Files} @@ -232,7 +234,7 @@ @set xref-numeric-owner @xref{Attributes} @set pxref-numeric-owner @pxref{Attributes} -@set op-occurrence @option{--occurrence} +@set op-occurrence @option{--occurrence} @set ref-occurrence @ref{--occurrence} @set xref-occurrence @xref{--occurrence} @set pxref-occurrence @pxref{--occurrence} @@ -334,7 +336,7 @@ @set xref-starting-file @xref{Scarce} @set pxref-starting-file @pxref{Scarce} -@set op-strip-components @option{--strip-components} +@set op-strip-components @option{--strip-components} @set ref-strip-components @ref{--strip-components} @set xref-strip-components @xref{--strip-components} @set pxref-strip-components @pxref{--strip-components} diff --git a/src/common.h b/src/common.h index 58d2798..77c353b 100644 --- a/src/common.h +++ b/src/common.h @@ -114,7 +114,13 @@ GLOBAL bool utc_option; than newer_mtime_option. */ GLOBAL int after_date_option; -GLOBAL bool atime_preserve_option; +enum atime_preserve +{ + no_atime_preserve, + replace_atime_preserve, + system_atime_preserve +}; +GLOBAL enum atime_preserve atime_preserve_option; GLOBAL bool backup_option; diff --git a/src/compare.c b/src/compare.c index cdbf945..5445ac0 100644 --- a/src/compare.c +++ b/src/compare.c @@ -25,10 +25,8 @@ # include #endif -#include -#include - #include "common.h" +#include #include #include @@ -205,9 +203,10 @@ diff_dir (void) static void diff_file (void) { + char const *file_name = current_stat_info.file_name; struct stat stat_data; - if (!get_stat_data (current_stat_info.file_name, &stat_data)) + if (!get_stat_data (file_name, &stat_data)) skip_member (); else if (!S_ISREG (stat_data.st_mode)) { @@ -228,19 +227,24 @@ diff_file (void) if (tar_timespec_cmp (get_stat_mtime (&stat_data), current_stat_info.mtime)) report_difference (¤t_stat_info, _("Mod time differs")); - if (current_header->header.typeflag != GNUTYPE_SPARSE && - stat_data.st_size != current_stat_info.stat.st_size) + if (current_header->header.typeflag != GNUTYPE_SPARSE + && stat_data.st_size != current_stat_info.stat.st_size) { report_difference (¤t_stat_info, _("Size differs")); skip_member (); } else { - diff_handle = open (current_stat_info.file_name, O_RDONLY | O_BINARY); + int atime_flag = + (atime_preserve_option == system_atime_preserve + ? O_NOATIME + : 0); + + diff_handle = open (file_name, O_RDONLY | O_BINARY | atime_flag); if (diff_handle < 0) { - open_error (current_stat_info.file_name); + open_error (file_name); skip_member (); report_difference (¤t_stat_info, NULL); } @@ -253,18 +257,18 @@ diff_file (void) else read_and_process (¤t_stat_info, process_rawdata); - status = close (diff_handle); - if (status != 0) - close_error (current_stat_info.file_name); - - if (atime_preserve_option) + if (atime_preserve_option == replace_atime_preserve) { struct timespec ts[2]; ts[0] = get_stat_atime (&stat_data); ts[1] = get_stat_mtime (&stat_data); - if (utimens (current_stat_info.file_name, ts) != 0) - utime_error (current_stat_info.file_name); + if (set_file_atime (diff_handle, file_name, ts) != 0) + utime_error (file_name); } + + status = close (diff_handle); + if (status != 0) + close_error (file_name); } } } @@ -357,7 +361,7 @@ diff_dumpdir (void) } else dev = stat.st_dev; - + dumpdir_buffer = get_directory_contents (current_stat_info.file_name, dev); if (dumpdir_buffer) diff --git a/src/create.c b/src/create.c index ff69876..087cacd 100644 --- a/src/create.c +++ b/src/create.c @@ -22,7 +22,6 @@ #include #include -#include #include "common.h" #include @@ -452,7 +451,7 @@ write_gnu_long_link (struct tar_stat_info *st, const char *p, char type) finish_header (st, header, -1); header = find_next_block (); - + bufsize = available_space_after (header); while (bufsize < size) @@ -582,7 +581,7 @@ write_extended (bool global, struct tar_stat_info *st, union block *old_header) union block *header, hp; char *p; int type; - + if (extended_header.buffer || extended_header.stk == NULL) return old_header; @@ -950,35 +949,6 @@ dump_regular_file (int fd, struct tar_stat_info *st) return dump_status_ok; } -static void -dump_regular_finish (int fd, struct tar_stat_info *st, - struct timespec original_ctime) -{ - if (fd >= 0) - { - struct stat final_stat; - if (fstat (fd, &final_stat) != 0) - { - stat_diag (st->orig_file_name); - } - else if (timespec_cmp (get_stat_ctime (&final_stat), original_ctime) - != 0) - { - WARN ((0, 0, _("%s: file changed as we read it"), - quotearg_colon (st->orig_file_name))); - } - if (close (fd) != 0) - { - close_diag (st->orig_file_name); - } - } - if (remove_files_option) - { - if (unlink (st->orig_file_name) == -1) - unlink_error (st->orig_file_name); - } -} - /* Look in directory DIRNAME for a cache directory tag file with the magic name "CACHEDIR.TAG" and a standard header, as described at: @@ -1059,9 +1029,9 @@ dump_dir0 (char *directory, size_t bufsize; ssize_t count; const char *buffer, *p_buffer; - + block_ordinal = current_block_ordinal (); - buffer = gnu_list_name->dir_contents; + buffer = gnu_list_name->dir_contents; if (buffer) totsize = dumpdir_size (buffer); else @@ -1070,7 +1040,7 @@ dump_dir0 (char *directory, finish_header (st, blk, block_ordinal); p_buffer = buffer; size_left = totsize; - + mv_begin (st); mv_total_size (totsize); while (size_left > 0) @@ -1161,11 +1131,9 @@ ensure_slash (char **pstr) } static bool -dump_dir (struct tar_stat_info *st, int top_level, dev_t parent_device) +dump_dir (int fd, struct tar_stat_info *st, int top_level, dev_t parent_device) { - char *directory; - - directory = savedir (st->orig_file_name); + char *directory = fdsavedir (fd); if (!directory) { savedir_diag (st->orig_file_name); @@ -1273,7 +1241,7 @@ compare_links (void const *entry1, void const *entry2) } static void -unknown_file_error (char *p) +unknown_file_error (char const *p) { WARN ((0, 0, _("%s: Unknown file type; file ignored"), quotearg_colon (p))); @@ -1289,8 +1257,8 @@ unknown_file_error (char *p) again if we've done it once already. */ static Hash_table *link_table; -/* Try to dump stat as a hard link to another file in the archive. If - succeeded returns true */ +/* Try to dump stat as a hard link to another file in the archive. + Return true if successful. */ static bool dump_hard_link (struct tar_stat_info *st) { @@ -1392,7 +1360,7 @@ check_links (void) exit_status to failure, a clear diagnostic has been issued. */ static void -dump_file0 (struct tar_stat_info *st, char *p, +dump_file0 (struct tar_stat_info *st, char const *p, int top_level, dev_t parent_device) { union block *header; @@ -1400,6 +1368,7 @@ dump_file0 (struct tar_stat_info *st, char *p, struct timespec original_ctime; struct timespec restore_times[2]; off_t block_ordinal = -1; + bool is_dir; if (interactive_option && !confirm ("add", p)) return; @@ -1458,43 +1427,48 @@ dump_file0 (struct tar_stat_info *st, char *p, if (is_avoided_name (p)) return; - if (S_ISDIR (st->stat.st_mode)) - { - dump_dir (st, top_level, parent_device); - if (atime_preserve_option) - utimens (p, restore_times); - return; - } - else - { - /* Check for multiple links. */ - if (dump_hard_link (st)) - return; - /* This is not a link to a previously dumped file, so dump it. */ + is_dir = S_ISDIR (st->stat.st_mode) != 0; - if (S_ISREG (st->stat.st_mode) - || S_ISCTG (st->stat.st_mode)) - { - int fd; - enum dump_status status; + if (!is_dir && dump_hard_link (st)) + return; - if (file_dumpable_p (st)) + if (is_dir || S_ISREG (st->stat.st_mode) || S_ISCTG (st->stat.st_mode)) + { + bool ok; + int fd = -1; + struct stat final_stat; + + if (is_dir || file_dumpable_p (st)) + { + fd = open (p, + (O_RDONLY | O_BINARY + | (is_dir ? O_DIRECTORY | O_NONBLOCK : 0) + | (atime_preserve_option == system_atime_preserve + ? O_NOATIME + : 0))); + if (fd < 0) { - fd = open (st->orig_file_name, - O_RDONLY | O_BINARY); - if (fd < 0) - { - if (!top_level && errno == ENOENT) - WARN ((0, 0, _("%s: File removed before we read it"), - quotearg_colon (st->orig_file_name))); - else - open_diag (st->orig_file_name); - return; - } + if (!top_level && errno == ENOENT) + WARN ((0, 0, _("%s: File removed before we read it"), + quotearg_colon (p))); + else + open_diag (p); + return; } - else + } + + if (is_dir) + { + ok = dump_dir (fd, st, top_level, parent_device); + + /* dump_dir consumes FD if successful. */ + if (ok) fd = -1; + } + else + { + enum dump_status status; if (fd != -1 && sparse_option && sparse_file_p (st)) { @@ -1508,88 +1482,130 @@ dump_file0 (struct tar_stat_info *st, char *p, switch (status) { case dump_status_ok: - mv_end (); - dump_regular_finish (fd, st, original_ctime); - break; - case dump_status_short: mv_end (); - close (fd); break; case dump_status_fail: - close (fd); - return; + break; case dump_status_not_implemented: abort (); } - if (atime_preserve_option) - utimens (st->orig_file_name, restore_times); - file_count_links (st); - return; + ok = status == dump_status_ok; } -#ifdef HAVE_READLINK - else if (S_ISLNK (st->stat.st_mode)) + + if (ok) { - char *buffer; - int size; - size_t linklen = st->stat.st_size; - if (linklen != st->stat.st_size || linklen + 1 == 0) - xalloc_die (); - buffer = (char *) alloca (linklen + 1); - size = readlink (p, buffer, linklen + 1); - if (size < 0) + /* If possible, reopen a directory if we are preserving + atimes, so that we can set just the atime on systems with + _FIOSATIME. */ + if (fd < 0 && is_dir + && atime_preserve_option == replace_atime_preserve) + fd = open (p, O_RDONLY | O_BINARY | O_DIRECTORY | O_NONBLOCK); + + if ((fd < 0 + ? deref_stat (dereference_option, p, &final_stat) + : fstat (fd, &final_stat)) + != 0) { - readlink_diag (p); - return; + stat_diag (p); + ok = false; } - buffer[size] = '\0'; - assign_string (&st->link_name, buffer); - if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT) < size) - write_long_link (st); + } - block_ordinal = current_block_ordinal (); - st->stat.st_size = 0; /* force 0 size on symlink */ - header = start_header (st); - if (!header) - return; - tar_copy_str (header->header.linkname, buffer, NAME_FIELD_SIZE); - header->header.typeflag = SYMTYPE; - finish_header (st, header, block_ordinal); - /* nothing more to do to it */ - - if (remove_files_option) + if (ok) + { + if (timespec_cmp (get_stat_ctime (&final_stat), original_ctime) != 0) + WARN ((0, 0, _("%s: file changed as we read it"), + quotearg_colon (p))); + else if (atime_preserve_option == replace_atime_preserve + && set_file_atime (fd, p, restore_times) != 0) + utime_error (p); + } + + if (0 <= fd && close (fd) != 0) + { + close_diag (p); + ok = false; + } + + if (ok && remove_files_option) + { + if (is_dir) + { + if (rmdir (p) != 0 && errno != ENOTEMPTY) + rmdir_error (p); + } + else { - if (unlink (p) == -1) + if (unlink (p) != 0) unlink_error (p); } - file_count_links (st); - return; } -#endif - else if (S_ISCHR (st->stat.st_mode)) - type = CHRTYPE; - else if (S_ISBLK (st->stat.st_mode)) - type = BLKTYPE; - else if (S_ISFIFO (st->stat.st_mode)) - type = FIFOTYPE; - else if (S_ISSOCK (st->stat.st_mode)) - { - WARN ((0, 0, _("%s: socket ignored"), quotearg_colon (p))); - return; - } - else if (S_ISDOOR (st->stat.st_mode)) + + return; + } +#ifdef HAVE_READLINK + else if (S_ISLNK (st->stat.st_mode)) + { + char *buffer; + int size; + size_t linklen = st->stat.st_size; + if (linklen != st->stat.st_size || linklen + 1 == 0) + xalloc_die (); + buffer = (char *) alloca (linklen + 1); + size = readlink (p, buffer, linklen + 1); + if (size < 0) { - WARN ((0, 0, _("%s: door ignored"), quotearg_colon (p))); + readlink_diag (p); return; } - else + buffer[size] = '\0'; + assign_string (&st->link_name, buffer); + if (NAME_FIELD_SIZE - (archive_format == OLDGNU_FORMAT) < size) + write_long_link (st); + + block_ordinal = current_block_ordinal (); + st->stat.st_size = 0; /* force 0 size on symlink */ + header = start_header (st); + if (!header) + return; + tar_copy_str (header->header.linkname, buffer, NAME_FIELD_SIZE); + header->header.typeflag = SYMTYPE; + finish_header (st, header, block_ordinal); + /* nothing more to do to it */ + + if (remove_files_option) { - unknown_file_error (p); - return; + if (unlink (p) == -1) + unlink_error (p); } + file_count_links (st); + return; + } +#endif + else if (S_ISCHR (st->stat.st_mode)) + type = CHRTYPE; + else if (S_ISBLK (st->stat.st_mode)) + type = BLKTYPE; + else if (S_ISFIFO (st->stat.st_mode)) + type = FIFOTYPE; + else if (S_ISSOCK (st->stat.st_mode)) + { + WARN ((0, 0, _("%s: socket ignored"), quotearg_colon (p))); + return; + } + else if (S_ISDOOR (st->stat.st_mode)) + { + WARN ((0, 0, _("%s: door ignored"), quotearg_colon (p))); + return; + } + else + { + unknown_file_error (p); + return; } if (archive_format == V7_FORMAT) diff --git a/src/misc.c b/src/misc.c index 7322dd2..78844ff 100644 --- a/src/misc.c +++ b/src/misc.c @@ -23,6 +23,14 @@ #include #include #include +#include + +#if HAVE_STROPTS_H +# include +#endif +#if HAVE_SYS_FILIO_H +# include +#endif /* Handling strings. */ @@ -490,6 +498,26 @@ deref_stat (bool deref, char const *name, struct stat *buf) return deref ? stat (name, buf) : lstat (name, buf); } +/* Set FD's (i.e., FILE's) access time to TIMESPEC[0]. If that's not + possible to do by itself, set its access and data modification + times to TIMESPEC[0] and TIMESPEC[1], respectively. */ +int +set_file_atime (int fd, char const *file, struct timespec const timespec[2]) +{ +#ifdef _FIOSATIME + if (0 <= fd) + { + struct timeval timeval; + timeval.tv_sec = timespec[0].tv_sec; + timeval.tv_usec = timespec[0].tv_nsec / 1000; + if (ioctl (fd, _FIOSATIME, &timeval) == 0) + return 0; + } +#endif + + return futimens (fd, file, timespec); +} + /* A description of a working directory. */ struct wd { diff --git a/src/tar.c b/src/tar.c index 8fd0548..b0cddaa 100644 --- a/src/tar.c +++ b/src/tar.c @@ -36,6 +36,7 @@ #define GLOBAL #include "common.h" +#include #include #include #include @@ -113,7 +114,7 @@ confirm (const char *message_action, const char *message_name) status = rpmatch (response) > 0; free (response); } - + if (confirm_file_EOF) { fputc ('\n', stdlis); @@ -267,7 +268,7 @@ The version control may be set with --backup or VERSION_CONTROL, values are:\n\n [For Solaris tar compatibility =/= Is it important at all?] e exit immediately with a nonzero exit status if unexpected errors occur E use extended headers (--format=posix) - + [q alias for --occurrence=1 =/= this would better be used for quiet?] [I same as T =/= will harm star compatibility] @@ -357,8 +358,11 @@ static struct argp_option options[] = { N_("force NAME as group for added files"), 51 }, {"mode", MODE_OPTION, N_("CHANGES"), 0, N_("force (symbolic) mode CHANGES for added files"), 51 }, - {"atime-preserve", ATIME_PRESERVE_OPTION, 0, 0, - N_("don't change access times on dumped files"), 51 }, + {"atime-preserve", ATIME_PRESERVE_OPTION, + N_("METHOD"), OPTION_ARG_OPTIONAL, + N_("preserve access times on dumped files, either by restoring the times" + " after reading (METHOD='replace'; default) or by not setting the times" + " in the first place (METHOD='system')"), 51 }, {"touch", 'm', 0, 0, N_("don't extract file modified time"), 51 }, {"same-owner", SAME_OWNER_OPTION, 0, 0, @@ -578,6 +582,17 @@ static struct argp_option options[] = { {0, 0, 0, 0, 0, 0} }; +static char const *const atime_preserve_args[] = +{ + "replace", "system", NULL +}; +static enum atime_preserve const atime_preserve_types[] = +{ + replace_atime_preserve, system_atime_preserve +}; +ARGMATCH_VERIFY (atime_preserve_args, atime_preserve_types); + + struct tar_args { char const *textual_date_option; int exclude_options; @@ -1117,7 +1132,15 @@ parse_opt (int key, char *arg, struct argp_state *state) break; case ATIME_PRESERVE_OPTION: - atime_preserve_option = true; + atime_preserve_option = + (arg + ? XARGMATCH ("--atime-preserve", arg, + atime_preserve_args, atime_preserve_types) + : replace_atime_preserve); + if (! O_NOATIME && atime_preserve_option == system_atime_preserve) + FATAL_ERROR ((0, 0, + _("--atime-preserve='system' is not supported" + " on this platform\n"))); break; case CHECKPOINT_OPTION: @@ -1872,7 +1895,7 @@ main (int argc, char **argv) /* Make sure we have first three descriptors available */ stdopen (); - + /* Pre-allocate a few structures. */ allocated_archive_names = 10; diff --git a/tests/ignfail.at b/tests/ignfail.at index 156b3ff..6a57316 100644 --- a/tests/ignfail.at +++ b/tests/ignfail.at @@ -70,10 +70,10 @@ tar: Error exit delayed from previous errors ----- tar: file: Warning: Cannot open: Permission denied ----- -tar: directory: Cannot savedir: Permission denied +tar: directory: Cannot open: Permission denied tar: Error exit delayed from previous errors ----- -tar: directory: Warning: Cannot savedir: Permission denied +tar: directory: Warning: Cannot open: Permission denied ]) AT_CLEANUP -- 2.45.2