Restart e2fsck only once in case of multiple inodes in uninit range.
Display correct inode number during BG_INO_UNINIT and INOREF_IN_USED errors.
Signed-off-by: Kalpak Shah <kalpak.shah@sun.com>
Signed-off-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Without a force flag, don't allow resize2fs to even start resizing
below what it thinks the minimum safe value is.
This may stop resizes which could otherwise proceed with a bit
of space still left, but seems like a reasonably safe thing to do.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
The extra padding added to the minimum size calculations:
/*
* We need to reserve a few extra blocks if extents are
* enabled, in case we need to grow the extent tree. The more
* we shrink the file system, the more space we need.
*/
if (fs->super->s_feature_incompat & EXT3_FEATURE_INCOMPAT_EXTENTS)
blks_needed += (fs->super->s_blocks_count - blks_needed)/500;
can go quite wrong if we've already added up more "blks_needed"
than our current size, and the above subtraction wraps. This can
easily happen for a filesystem which is almost completely full.
In this case, just return the current fs size as the minimum and
be done with it.
With this fix we could probably call calculate_minimum_resize_size()
for each resize2fs invocation and refuse to resize smaller than that?
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
http://people.redhat.com/esandeen/livecd-creator-imagefile.bz2
contains an image (for now) which, when resized to 578639, corrupts
the filesystem.
This is a bit crazy, I guess, because the fs currently has only
1 free block, but still, we should be graceful about the failure.
Perhaps it would make sense to check the requested valuea against
the minimum value resize2fs would compute for "-P" and fail (at
least without a force).
But in any case, this exposed 2 bugs when moving that one block
required an extent split, which is what hit the ENOSPC.
For starters, ext2fs_extent_set_bmap() in the "(re/un)mapping last
block in extent" case was replacing the old extent before the
new one was created; when the new extent creation failed, it
left us in an inconsistent state. Simply changing the order of
the two should fix this problem.
Next, ext2fs_extent_insert was calling ext2fs_extent_delete()
on *any* error, including one caused by failure to allocate a new
block to split the node to hold that extent ... the handle was left
unchanged, and we deleted the -original- extent.
As a quick fix for this, just don't do the delete if we fail the split,
though this may need to be smarter. I don't think we have terribly
consistent behavior about where a handle is left on various errors.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Fix a regression in e2fsprogs 1.41.5 which would undo updates to the
block group descriptors after a journal replay, caused by commit
b7c5b403. We now use ext2fs_free() instead of ext2fs_close() to make
sure we the library will never try to write out superblock or block
group descriptors.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
One of our customers hit a temporary IO error during an e2fsck run during
the read from the journal. It seems that the read error resulted in
e2fsck automatically discarding the journals and recreating them on several
filesystems on this node without any prompting from the user:
end_request: I/O error, dev sdg, sector 484832
Buffer I/O error on device sdg, logical block 60604
fsck-sdg[8276]: ls2-OST024c: Superblock has an invalid ext3 journal (inode 8).
fsck-sdg[8276]: CLEARED.
fsck-sdg[8276]: *** ext3 journal has been deleted - filesystem is now ext2
only ***
fsck-sdg[8276]: ls2-OST024c was not cleanly unmounted, check forced.
fsck-sdg[8276]: ls2-OST024c: Journal inode is not in use, but contains data.
CLEARED.
fsck-sdg[8276]: ls2-OST024c: Recreate journal to make the filesystem ext3
again?
fsck-sdg[8276]: FIXED.
fsck-sdg[8276]: Creating journal (32768 blocks): Done.
fsck-sdg[8276]:
fsck-sdg[8276]: *** journal has been re-created - filesystem is now ext3 again
***
fsck-sdg[8276]: ls2-OST024c: 39818/20183248 files (8.2% non-contiguous), 222122257/779902976 blocks
fsck-sdg[8276]: exit code 1 (file system errors corrected)
The following patch moves the e2fsck error handler initialization earlier
in the e2fsck startup code before the journal is processed, so that the
user will be prompted for an action. This is the first IO that is not
part of ext2fs_open() where fs->io is first initialized.
It doesn't seem possible to initialize the error handlers for the initial
filesystem open without changing the prototype for ext2fs_open2(). If we
are getting a new ext2fs_open3() prototype for 64-bit it might make sense
to add at least "read_error" as a parameter ("write_error" is not strictly
necessary for the open and could be set afterward).
Signed-off-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Jim Garlick <garlick@llnl.gov>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If ext2fs_extent_open() fails due to a corrupt extent header, and the
user declines to clear the inode, check_blocks_extents() should bail
out; otherwise, it will cause a core dump due a null pointer
dereference.
Addresses-Sourceforge-Bug: #2791794
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The Linux kernel (since 2.6.29, patch 784aae735d9b0bba3f8b9faef4c8b30df3bf0128)
exports the real DM device names in /sys/block/<ptname>/dm/name.
The sysfs based solution is nicer and faster than scan for devno in
/dev/mapper/.
CC: Milan Broz <mbroz@redhat.com>
Signed-off-by: Karel Zak <kzak@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cleanup whitespace in the problem.h and problem.c files. Removes a
bunch of places where tabs follow spaces, whitespace on empty lines, etc.
I didn't reformat the indenting of the entire problem.h error codes,
but there is some room for doing this...
Signed-off-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
We've hit a number of cases where the error codes in problem.h have
been assigned duplicate values compared to problems in our own e2fsck
patches, and this can lead to confusing and difficult to find bugs
in e2fsck (e.g. wrong problem messages, incorrect repair action, etc).
Attached is a test case for the problem.c file to ensure that the
problem table is sorted and does not contain any duplicate values.
Having the problem table sorted allows the correctness checking to be
very simple, and if it ever became important for performance we could
use binary searching of the problem table for the specific problem code.
Signed-off-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
The FIEMAP ioctl is more efficient and doesn't require root
privileges. So if it is available, use it in preference to repeated
FIBMAP calls.
Signed-off-by: Kalpak Shah <kalpak.shah@sun.com>
Signed-off-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If the superblock is clean, and we only need to update
s_kbytes_written, then we only need to update the superblock.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we are checking a mounted filesystem (typically the root
filesystem, mounted read/only) and the NEEDS_RECOVERY flag is not set,
skip all of the checks associated with making sure the journal is
consistent. There is the very slight possibility we could lose if the
NEEDS_RECOVERY flag was somehow cleared even though there was data in
the journal, but this has practically never happend in practice, and
it reduces the number of reads required at boot-time, which is a big
deal when trying to reduce boot times with HDD's.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When opening a filesystem, make ext2fs_open2() much more efficient by
reading the normal block group descriptors all at once, instead of one
block at a time.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
E2fsck needs to check to see if the backup superblock differs from the
primary superblock. Previously it was doing so by calling
ext2fs_open(), which does a lot of unnecessary work, including reading
all of the backup block group descriptors. Avoid this by reading in
the backup superblock directly.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Move check_resize_inode() out of check_super_block(), since we only
need to test the resize_inode for correctness only if the filesystem
requires checking. This change avoids a lot of I/O operations which
slows down a 1 second boot.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
We can get into a situation in blkid where whole disks remain
in the cache, even though partitions are found. For labels
such as sun disklabels which may have the first partition
beginning at sector 0, this is even somewhat likely.
1) create a sun disklabel w/partitions
2) mkfs the first partition (at sector 0)
3) remove the partition table
4) run blkid - this finds the fs on the whole disk, places in cache
5) recreate the partition table
6) run blkid - this finds the partition, places in cache
And now we have both /dev/sda and /dev/sda1 in cache.
There are heuristics in probe_all to avoid putting the whole disk
in cache if it has partitions, but there is nothing to remove the
whole-disk entry in the above case. I think the below patch
suffices, although I haven't quite convinced myself that setting
the lens[which]=0; is the right logic for that bit of state...
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The functions which test to see if a device is mounted can get
confused by entries in /etc/mtab for virtual filesystems:
rootfs / rootfs rw 0 0
none /sys sysfs rw,nosuid,nodev,noexec,relatime 0 0
If the device name doesn't start with a slash, ignore the /etc/mtab
entry, so that relative pathnames passed into functions such as
ext2fs_check_mount_point() or ext2fs_check_if_mounted() don't return
false positives.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fixed a potential bug where by partial returns from the write(2)
system call could some bytes to be lost when writing to the log file.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Resize2fs will attempt to truncate an image file of a filesystem down
to size for the convenience of the system administrator. If the
truncate operation fails, print a warning message. This also avoids a
gcc warning message.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fixed a potential bug where by partial returns from the write(2)
system call could lost characters to be sent to external progress bar
display program.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fix gcc warnings caused by wrong types passed com_err() in
ext2fs_inode_alloc_stats2 and in ext2fs_block_alloc_stats().
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fixed a potential bug where by partial returns from the write system
call could the fallback pager to drop characters.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This is primarily to silence a gcc warning, but it's better to reflect
the error from system() up to the caller. In this case we don't
actually use it for anything, but that's OK.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Define the prototypes for et_list_lock() and et_list_unlock() in
com_err.h. This promotes better error checking and avoids warnings
when compiling the library and programs that call these functions.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Define the prototypes for e2p_jrnl_feature2string() and
e2p_jrnl_string2feature() in e2p.h. This promotes better error
checking and avoids warnings when compiling the library and programs
that call these functions.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Add an option to switch between the private (in-tree) libblkid and
public (in-system installed) library. The private version is still
enabled by default.
If --disable-libblkid is specified the findfs(8) program, which is a
variant of tune2fs, is also not built or installed.
Signed-off-by: Karel Zak <kzak@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
This fixes the Lintian warning:
The copyright file refers to the versionless symlink in
/usr/share/common-licenses for the full text of the GPL, LGPL, or GFDL
license, but the package does not appear to allow distribution under
later versions of the license. This symlink will change with each
release of a new version of the license and may therefore point to a
different version than the package is released under. debian/copyright
should instead refers to the specific version of the license that the
package references.
For example, if the package says something like "you can redistribute it
and/or modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; version 2 dated June, 1991,"
the debian/copyright file should refer to
/usr/share/common-licenses/GPL-2, not /GPL.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Due to a fencepost bug, when skipping a block group whose block bitmap
was uninitialized (and hence could not contain any blocks eligible for
relaocation), the block immediately following the block group wasn't
checked as well. If it was in use and required relocation, it
wouldn't get properly relocated, with the result that an inode using
such a block would end up, post resize, with a pointer to a block now
outside the bounds of the filesystem.
This commit fixes this fencepost error.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we need to shrink the inode table, we need to make sure the inodes
contained in the part of the inode table we are vacating don't get
reused as part of the filesystem shrink operation. This wasn't a
problem with ext3 filesystems, since the inode table was located in
the block group that was going away, so that location was not eligible
for reallocation.
However with ext4 filesystems with flex_bg enabled, it's possible for
a portion of the inode table in the last flex_bg group to be
deallocated, but in a part of the filesystem which could be used as
data blocks. So we must mark those blocks as reserved to prevent
their reuse, and adjust the minimum filesystem size calculation to
assure that we don't shrink a filesystem too small for the resize
operation to succeed.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When allocating a new set of block group metadata as part of growing
the filesystem, the resize2fs code assumes that the bitmap and inode
table blocks are in their own block group; an assumption which is
changed by the flex_bg feature. This commit works around the problem
by temporarily turning off flex_bg while allocating the new block
group metadata, to avoid potentially overwriting previously allocated
data blocks.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Andreas Dilger