Don't call ext2fs_get_pathname() for every single directory; instead,
only call it if we find a matching directory entry. In addition, if
ext2fs_get_pathname() fails, print the number of the parent directory
in angle parents so the user gets some additional information.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Newer versions of glibc no longer export the getpagesize() prototype when
using recent versions of POSIX (_XOPEN_SOURCE). So building tdb.c gives
use implicit function declaration warnings. Fix the issue by using the
portable sysconf() function which returns the same answer.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
There is a slight desync between the mke2fs(8) man page and the mke2fs
help output when it comes to the -t/-T options. Since the man page is
correct, update the mke2fs usage string to match.
Reported-by: Ben Kohler <bkohler@gmail.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
This feature is especially useful for better understanding how e2fsprogs
tools (mainly e2fsck) treats bitmaps and what bitmap backend can be most
suitable for particular bitmap. Backend itself (if implemented) can
provide statistics of its own as well.
[ Changed to provide basic statistics when enabled with the
E2FSPROGS_BITMAPS_STATS environment variable -- tytso]
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Now that we have multiple backend implementations of the bitmap code,
this commit teaches e2fsck to use either the most appropriate backend
for each use case.
Since we don't know for sure if we will get it all right, the default
choices can be overridden via e2fsck.conf. The various definitions
are shown here, with the current defaults (which may change as we add
more bitmap implementations and as learn what works better).
; EXT2FS_BAMP64_BITARRAY is 1
; EXT2FS_BMAP64_RBTREE is 2
; EXT2FS_BMAP64_AUTODIR is 3
[bitmaps]
inode_used_map = 2 ; pass1
inode_dir_map = 3 ; pass1
inode_reg_map = 2 ; pass1
block_found_map = 2 ; pass1
inode_bad_map = 2 ; pass1
inode_imagic_map = 2 ; pass1
block_dup_map = 2 ; pass1
block_ea_map = 2 ; pass1
inode_link_info = 2 ; pass1
inode_dup_map = 2 ; pass1b
inode_done_map = 3 ; pass3
inode_loop_detect = 3 ; pass3
fs_bitmaps = 2
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This change causes the max resident memory of mke2fs, as reported by
/usr/bin/time, to drop from 9296k to 5328k when formatting a 25
gig volume.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The pass5 checks would fail if the expected and current {inode,block}
bitmaps used different back ends that returned different non-zero
values from the test_*_bitmap() functions. Fix this by changing
"(actual == bitmap)" to "(!actual == !bitmap)".
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This backend type will automatically switch between the bitarray and
the rbtree backend based on the number of directories in the file
system.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
For a long time we had a bitarray backend for storing filesystem
metadata bitmaps, however today this approach might hit its limits with
todays huge data storage devices, because of its memory utilization.
Bitarrays stores bitmaps as ..well, as bitmaps. But this is in most
cases highly unefficient because we need to allocate memory even for the
big parts of bitmaps we will never use, resulting in high memory
utilization especially for huge filesystem, when bitmaps might occupy
gigabytes of space.
This commit adds another backend to store bitmaps. It is based on
rbtrees and it stores just used extents of bitmaps. It means that it can
be more memory efficient in most cases.
I have done some limited benchmarking and it shows that rbtree backend
consumes approx 65% less memory that bitarray on 312GB filesystem aged
with Impression (default config). This number may grow significantly
with the filesystem size, but also it may be a lot lower (even negative)
if the inodes are very fragmented (need more benchmarking).
This commit itself does not enable the use of rbtree backend.
[ Simplified the code by avoiding unneeded memory allocation and
deallocation of del_ext. In addition, fixed a bug discovered by the
tst_bitmaps tests: rb_unamrk_bmap() must return true if the bit was
previously set in bitmap, and zero otherwise -- tytso ]
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This commit adds rbtree library into e2fsprogs so it can be used for
various internal data structures. The rbtree implementation is ripped of
kernel rbtree implementation with small changes needed for it to work
outside kernel.
[ I prefixed the exported symbols and interface with ext2fs_ to keep
avoid pulluting the namespace exported by the libext2fs shared
library. -- tytso ]
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This allows a program to control the bitmap backend implementation
that will get used without needing to change the current library API.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This is only an issue for programs compiled against e2fsprogs 1.41
that manipulate bitmaps directly. Fortunately there are very few
programs which do that, especially those that try to clear a bitmap.
Addresses-Sourceforge-Bugs: #3451486
Reported-by: robi6@users.sourceforge.net
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Optimize how the tdb library so that running with [scratch_files] in
/etc/e2fsck.conf is more efficient. Use a better hash function,
supplied by Rogier Wolff, and supply an estimate of the size of the
hash table to tdb_open instead of using the default (which is way too
small in most cases). Also, disable the tdb locking and fsync calls,
since it's not necessary for our use in this case (which is
essentially as cheap swap space; the tdb files do not contain
persistent data.)
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Turns out the Hurd defines MS_SYNC but doesn't define msync(). Go
figure. So check for both.
Reported by Svante Signell.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
We need to do an accounting of duplicate clusters on a per-cluster
instead of a per-block basis so we know when we've correctly accounted
for all of the multiply claimed blocks in a particular inode.
Thanks to Robin Dong for reporting this bug.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
PATH_MAX is not portable (for example, it doesn't exist on the Hurd).
So replace it with a new define, which defines the maximum length of
the base quota name. As it turns out, this is substantially smaller
than PATH_MAX.
Also move the definitions relating to quotaio.c from mkquota.h to
quotaio.h, as a cleanup.
Addresses-Debian-Bug: #649689
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When I create a non-extent file with the maximum size in ext4,
e2fsck detects the following error:
Pass 1: Checking inodes, blocks, and sizes
Inode 12, i_size is 4402345721856, should be 4402345721856. Fix?
As we know, e2fsck checks the size field of the inode in pass 1.
However, in case of the ext4 with the feature of ^extent and huge_file,
the maximum file size calculated in e2fsck is less than the real one.
The patch fixes this problem.
Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Linux's compat_sys_ioctl() function, which is run when executing a
ioctl using a 32-bit binary on a 64-bit kernel, returns EINVAL when an
inode does not exist. Sigh. See /usr/src/linux/fs/compat_ioctl.c.
This is probably a kernel bug, but work around it for now.
Addresses-Debian-Bug: #644989
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we abort fsck (due to ENOMEM for example) we exit
with only the FSCK_ERROR flag. It seems useful
to do the same sorts of checks as we do on normal
exit, and return whether the filesystem was modified,
whether there are still uncorrected errors, etc, even
in the abort case.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
The htree dump code overflows a char buffer if the directory has a long
filename because the buffer is not large enough to hold the characters that are
not part of the filename. Make the buffer larger and use snprintf instead.
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Define EXT2FS_MAX_NESTED_LINKS as 8, and check the link count to make
sure we don't exceed it in ext2fs_find_block_device() and
follow_link(). This fixes a potential infinite loop in
ext2fs_find_block_device() if there are symbolic loop links in the
device directory.
Signed-off-by: Niu Yawei <niu@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
A 32-bit s390 build was failing on a 64-bit s390x host, when
make check failed e2undo tests, like this:
md5sum before mke2fs 922c8a591c882dbdd1a381d18547cfd5
using mke2fs to test e2undo
Overwriting existing filesystem; this can be undone using the command:
e2undo /tmp/mke2fs-tmp.EM9XjmTA81.e2undo /tmp/tmp.EM9XjmTA81
md5sum after mke2fs cbf32fb6c3db45280ad013f42ac294f1
Replayed transaction of size 32768 at location 0
Replayed transaction of size 32768 at location 0
Replayed transaction of size 32768 at location 0
Replayed transaction of size 32768 at location 0
Replayed transaction of size 0 at location 0
md5sum after e2undo 31b4e14307c5b7ccce5b8d300c2ad5f1
Note the "at location 0" for the block number.
A proper cast in e2undo.c fixes this up.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
The freefrag command provides the functionality of e2freefrag on the
currently open file system in debugfs.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Create a version of debugfs which only supports read-only examination
of the file system metadata (but not the data blocks). The idea is
that this version of debugfs might be suitable to be setuid root, and
executable only by members of a particular group, or setgid disk, and
globally executable, depending on the security/privacy policies in
force at a particular site.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
In quota_compute_usage(), the space usage should be in bytes but
not quota block.
Signed-off-by: Niu Yawei <niu@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
When turn on quota by tune2fs, if the old quota file exist, the quota
usage should be recomputed but the old limits should be preserved.
Signed-off-by: Niu Yawei <niu@whamcloud.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
mke2fs was creating both user and group quota inodes on enabling
the quota feature. This patch adds the extended option 'quotatype'
that can be used to exclusively specify the quota type that the
user wants to initialize.
# Ex: Default behavior without extended option creates both
# user and group quota inodes:
$ mke2fs -t ext4 -O quota /dev/ram1
# To enable only user quotas:
$ mke2fs -t ext4 -O quota -E quotatype=usr /dev/ram1
# To enable only group quotas:
$ mke2fs -t ext4 -O quota -E quotatype=grp /dev/ram1
Signed-off-by: Aditya Kali <adityakali@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
When turning on the quota feature, tune2fs would create empty quota inodes and
set their inode numbers in superblock. This required e2fsck to be ran before
using the quota feature. This patch adds adds call to compute_quota() and make
sure that we write correct quota information in the quota files at tune2fs time
itself. This gets rid of the necessity for running e2fsck after setting the
quota feature. Also, tune2fs now does not use existing old quota files
(aquota.user and aquota.group) even if they exist.
Signed-off-by: Aditya Kali <adityakali@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
This patch cleans up the quota code as suggested in previous reviews. This
includes
* remove BUG_ON()s and 'exit()' calls from library code
* remove calls to malloc/free and instead use ext2fs_get/free_mem functions.
* lib/quota/common.c file in not needed anymore and is removed.
* rename exported functions to start with quota_
(ex: init_quota_context --> quota_init_context)
* better error handling in quota library
Signed-off-by: Aditya Kali <adityakali@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Quota support can be enabled using --enable-quota. There are still
some buglets that we need to fix up before it can be considered 100%
supported, so let's disable it for the 1.42 release.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
After:
# mke2fs -O ^has_journal,^resize_inode,^uninit_bg,extent,meta_bg,flex_bg,bigalloc /dev/sda
# e2freefrag /dev/sda
It will report error message like:
Illegal block number passed to ext2fs_test_block_bitmap #1732133 for block bitmap for /dev/sda
Illegal block number passed to ext2fs_test_block_bitmap #1732134 for block bitmap for /dev/sda
Illegal block number passed to ext2fs_test_block_bitmap #1732135 for block bitmap for /dev/sda
One bit in bitmap of bigalloc-ext4 means a cluster not a block,
therefore ext2fs_fast_test_block_bitmap2 should check cluster.
Signed-off-by: Robin Dong <sanbai@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Several of the mmp tests were failing on an s390 guest because
direct IO cannot be done on sub-sector sizes, and they were
doing 1k IOs to files on a 4k logical/physical device.
A few tests could just be changed to 4k, but others
needed fixed up output as well.
With this I have all tests passing on s390 again.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Without this change, we will write data past the end of the
mmp buf. Valgrind catches this:
==6373== Syscall param write(buf) points to unaddressable byte(s)
==6373== at 0x362260E470: __write_nocancel (in /lib64/libpthread-2.12.2.so)
==6373== by 0x41CF83: raw_write_blk (unix_io.c:255)
==6373== by 0x41D2BC: unix_write_blk64 (unix_io.c:757)
==6373== by 0x41A05D: ext2fs_mmp_write (mmp.c:130)
==6373== by 0x40B0C9: do_set_mmp_value (set_fields.c:806)
==6373== by 0x421B61: really_execute_command (execute_cmd.c:108)
==6373== by 0x421C54: ss_execute_line (execute_cmd.c:234)
==6373== by 0x403743: main (debugfs.c:2339)
==6373== Address 0x63f000 is not stack'd, malloc'd or (recently) free'd
and in my testing it led to silent failures while writing the mmp
block in debugfs:
write(3, "xV4\22PMM\342\325V\274N\0\0\0\0host.name."..., 4096) = -1 EFAULT (Bad address)
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Several small fixes:
* Gracefully fail mmp commands if fs is not open
* Show magic number in dump_mmp command
* Fix header in output for set_mmp_value -l
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Theodore Ts'o