diff options
| -rw-r--r-- | hurd/translator/ext2fs.mdwn | 16 | ||||
| -rw-r--r-- | hurd/translator/ext2fs/filetype.mdwn | 33 | ||||
| -rw-r--r-- | open_issues/git-core-2.mdwn | 7 | ||||
| -rw-r--r-- | open_issues/git_duplicated_content.mdwn (renamed from open_issues/ext2fs_file_corruption.mdwn) | 86 |
4 files changed, 58 insertions, 84 deletions
diff --git a/hurd/translator/ext2fs.mdwn b/hurd/translator/ext2fs.mdwn index 305576b8..fff2e74b 100644 --- a/hurd/translator/ext2fs.mdwn +++ b/hurd/translator/ext2fs.mdwn @@ -1,4 +1,4 @@ -[[!meta copyright="Copyright © 2007, 2008, 2010 Free Software Foundation, +[[!meta copyright="Copyright © 2007, 2008, 2010, 2011 Free Software Foundation, Inc."]] [[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable @@ -9,14 +9,20 @@ Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled [[GNU Free Documentation License|/fdl]]."]]"""]] -# Large Stores +# Implementation + + * [[filetype]] option + + +## Large Stores The `ext2fs` translator from the upstream Hurd code base can only handle file systems with sizes of less than roughly 2 GiB. [[!tag open_issue_hurd]] -## Ognyan's Work + +### Ognyan's Work * Ognyan Kulev, [[*Supporting Large ext2 File Systems in the Hurd*|ogi-fosdem2005.mgp]], 2005, at FOSDEM @@ -34,4 +40,6 @@ small backend stores, like floppy devices. # Documentation -<http://www.nongnu.org/ext2-doc/> + * <http://e2fsprogs.sourceforge.net/ext2.html> + + * <http://www.nongnu.org/ext2-doc/> diff --git a/hurd/translator/ext2fs/filetype.mdwn b/hurd/translator/ext2fs/filetype.mdwn new file mode 100644 index 00000000..5d85bac9 --- /dev/null +++ b/hurd/translator/ext2fs/filetype.mdwn @@ -0,0 +1,33 @@ +[[!meta copyright="Copyright © 2011 Free Software Foundation, Inc."]] + +[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable +id="license" text="Permission is granted to copy, distribute and/or modify this +document under the terms of the GNU Free Documentation License, Version 1.2 or +any later version published by the Free Software Foundation; with no Invariant +Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license +is included in the section entitled [[GNU Free Documentation +License|/fdl]]."]]"""]] + +The *ext2fs* translator doesn't support the ext2 format's *filetype* option. + +According to *mke2fs(8)*: + +> **filetype**: Store file type information in directory entries. + +By setting directory listings' informational `d_type` field (`readdir`, etc.), +this may avoid the need for subsequent `stat` calls. + +Not all file systems can support this option. + +In `[hurd]/ext2fs/dir.c` the `EXT2_FEATURE_INCOMPAT_FILETYPE` is generally +masked out (is not even considered) when adding a node to a directory in +`diskfs_direnter_hard` and when reading in `diskfs_get_directs`. The Hurd's +ext2fs unconditionally sets this field to 0 (`EXT2_FT_UNKNOWN`). + + +# `e2fsck` + +Running `e2fsck` on a file system with the *filetype* option, will correct the +*filetype* for a lot of files (all `EXT2_FT_UNKNOWN`?) to either 1 (regular +file, `EXT2_FT_REG_FILE`), or 2 (directory, `EXT2_FT_DIR`), and likely others. +The Hurd's ext2fs will again ignore these fields, of course. diff --git a/open_issues/git-core-2.mdwn b/open_issues/git-core-2.mdwn index a1fdbac1..acd82442 100644 --- a/open_issues/git-core-2.mdwn +++ b/open_issues/git-core-2.mdwn @@ -133,3 +133,10 @@ starting from an empty working directory (after an external `git push`): $ git status # On branch master nothing to commit (working directory clean) + +--- + +2011-06-13 + +This seems to be fixed after a rebuild of the Debian git 1.7.5.3-1 with a +recent toolchain. diff --git a/open_issues/ext2fs_file_corruption.mdwn b/open_issues/git_duplicated_content.mdwn index 2af35bb5..3b4bc44b 100644 --- a/open_issues/ext2fs_file_corruption.mdwn +++ b/open_issues/git_duplicated_content.mdwn @@ -8,44 +8,7 @@ Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled [[GNU Free Documentation License|/fdl]]."]]"""]] -[[!tag open_issue_hurd]] - -[[!toc]] - - -# Log - -Create two 5 GiB partitions. - - # # Would default to 256 byte inodes otherwise. - # mke2fs -I 128 -o linux /dev/hd1s1 - [...] - # mke2fs -o hurd /dev/hd1s2 - [...] - -The only difference (apart from UUID, time stamps, hash seeds) in ext2 file -system meta data, according to `dumpe2fs -h`: - - [...] - -Filesystem features: ext_attr resize_inode dir_index filetype sparse_super large_file - +Filesystem features: ext_attr resize_inode dir_index sparse_super large_file - Filesystem flags: signed_directory_hash - Default mount options: (none) - Filesystem state: clean - Errors behavior: Continue - -Filesystem OS type: Linux - +Filesystem OS type: Hurd - [...] - -According to *mke2fs(8)*: - -> *filetype* -> Store file type information in directory entries. - - # settrans -c /media/hd1s1 /hurd/ext2fs /dev/hd1s1 - # settrans -c /media/hd1s2 /hurd/ext2fs /dev/hd1s2 - # mkdir /media/hd1s{1,2}/tmp - # chmod 1777 /media/hd1s{1,2}/tmp +[[!tag open_issue_porting]] $ git-new-workdir ~/tmp/binutils/git /media/hd1s1/tmp/master master error: unable to create file gas/testsuite/gas/arm/attr-mfpu-vfpv3-d16.d (Interrupted system call) @@ -140,8 +103,6 @@ Consistent; thus very likely corrupt on-disk. After a few tries, the pattern generally is that for the files where there are differences, once the file regularely ends, its content appears once more. That is, the files' content appears once (regularely), and then the same again. -From three retries, I've seen this only ever happen on the `-o -linux`/`filetype` option file system. Some more copying: @@ -155,45 +116,10 @@ Some more copying: No further difference. +--- -# Conclusion? - -The ext2 `filetype` option, as used for Linux-type file systems by default, -might cause some confusion inside [[hurd/translator/ext2fs]]? - -[[!tag open_issue_documentation]] The strange thing is that in -`[hurd]/ext2fs/dir.c` the `EXT2_FEATURE_INCOMPAT_FILETYPE` is generally masked -out (is not even considered) when adding a node to a directory in -`diskfs_direnter_hard` and when reading in `diskfs_get_directs`, so unless -setting this file system flag has some other effects on the meta data (which it -doesn't seem to have according to the Linux kernel source code), it shouldn't -make any difference. Also, this information would only be used for setting -directory listing's informational `d_type` field (`readdir`, etc.). Hrm. -Perhaps only a red herring? - - -## `e2fsck` - -Running `e2fsck` on a file system with the `filetypes` option, there are -messages that the `filetype` for a lot of files is set (to either 1 (regular -file, `EXT2_FT_REG_FILE`), or 2 (directory, `EXT2_FT_DIR`)). The Hurd's ext2fs -should unconditionally set this field to 0 (`EXT2_FT_UNKNOWN`). - - -# Next - -Start over with the `filetype` option reversed: - - # mke2fs -I 128 -o linux -O ^filetype /dev/hd1s1 - # mke2fs -o hurd -O filetype /dev/hd1s2 - -In fact, for `-o hurd`, `-O filetype` doesn't have any effect (according to -`dumpe2fs -h`), so the meta data now only differs in *Filesystem OS type*. - -Invoke `git-new-workdir` once more (for each). + $ git-new-workdir git master master + $ diff -x .git -ur tar_master/ master/ > master.diff -Now there is corruption in files on the `-o hurd -O filetype` file system. But -this time not caused by `git-new-workdir`, but instead when copying a binutils -checkout to a new directory. Huh. Does `-O filetype` in fact have any effect -with `-o hurd`? The `e2fsprogs` source code should tell, but is not available -right now. + $ rm -rf ar_master* && (cd git/ && git archive master) | (mkdir ar_master && cd ar_master/ && tar -x) && diff -x .git -ru tar_master/ ar_master/ > ar_master.diff; ls -l ar_master.diff + $ (cd git/ && git archive master) | md5sum |