The s_desc_size in the superblock specifies the group descriptor
size in bytes, but in various places the EXT4_FEATURE_INCOMPAT_64BIT
flag implies that the descriptor size is EXT2_MIN_DESC_SIZE_64BIT
(64 bytes) instead of checking the actual size. In other places,
the s_desc_size field is used without checking for INCOMPAT_64BIT.
In the case of ext2fs_group_desc() the s_desc_size was being ignored,
and assumed to be sizeof(struct ext4_group_desc), which would result
in garbage for any but the first group descriptor. Similarly, in
ext2fs_group_desc_csum() and print_csum() they assumed that the
maximum group descriptor size was sizeof(struct ext4_group_desc).
Fix these functions to use the actual superblock s_desc_size if
INCOMPAT_64BIT.
Conversely, in ext2fs_swap_group_desc2() s_desc_size was used
without checking for INCOMPAT_64BIT being set.
The e2fsprogs behaviour is different than that of the kernel,
which always checks INCOMPAT_64BIT, and only uses s_desc_size to
determine the offset of group descriptors and what range of bytes
to checksum.
Allow specifying the s_desc_size field at mke2fs time with the
"-E desc_size=NNN" option. Allow a power-of-two s_desc_size
value up to s_blocksize if INCOMPAT_64BIT is specified. This
is not expected to be used by regular users at this time, so it
is not currently documented in the mke2fs usage or man page.
Add m_desc_size_128, f_desc_size_128, and f_desc_bad test cases to
verify mke2fs and e2fsck handling of larger group descriptor sizes.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Add a LOG2_CHECK mode for check_super_value() so that it is easy
to verify values that are supposed to be power-of-two values
(s_desc_size and s_inode_size so far). In ext2fs_check_desc()
also check for a power-of-two s_desc_size.
Print out s_desc_size in debugfs "stats" and dumpe2fs output, if
it is non-zero.
It turns out that the s_desc_size validation in check_super_block()
is not currently used by e2fsck, because the group descriptors are
verified earlier by ext2fs_check_desc(), and even without an
explicit check of s_desc_size the group descriptors fail to align
correctly on disk. It makes sense to keep the check_super_block()
regardless, in case the code changes at some point in the future.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Since commit 5ad07acad if $TMP cannot hold large test filesystems
for resize testing the resize_test creates temporary test files
in the local working directory. Since it overrides TMPFILE locally
the calling program does not delete the generated file correctly.
Delete the large $TMPFILE within resize_test if it passes, but leave
it for debugging if the test failed.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Don't check for lost+found in read only mode.
[Note: this patch was originally made against 1.41.14 version of
e2fsprogs found as part of the AOSP (Android Open Source Program)
tree. My Signed-off-by relies on the fact that the original patch
author would have had to have filed a contribution agreement with Open
Handset Alliance before this commit before this commit was allowed
into the AOSP tree. -- tytso]
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Mostly by adding static and removing excess extern qualifiers. Also
convert a few remaining non-ANSI function declarations to ANSI.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
I've added documentation for all available filesystem features
settable by mke2fs.
* Features now include the kernel constant attached to the function.
* Some pre-existing feature explanations have been expanded based on
ext4.wiki.kernel.org.
[ Edited by tytso for stylistic and correctness. ]
Addresses-Debian-Bug: #586218
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
When /etc/mtab is a symlink of /proc/mounts, mke2fs without -FF option
can create a filesystem on the image file that is mounted.
According to mke2fs man page, we should specify -FF option in this case.
This patch protects filesystem from unintended mke2fs caused by human error.
How to reproduce:
# mke2fs -t ext4 -Fq fs.img
# mount -o loop fs.img /mnt/mp1
# mke2fs -t ext4 -Fq fs.img && echo "mke2fs success"
mke2fs success
Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
According to mke2fs man, we can create a filesystem on the mounted
device when -FF option is specified.
However, currently we have to specify -F option third to force mke2fs.
This patch fixes the problem.
Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When we're expanding a directory, check to see if we're doing an
implied cluster allocation; if so, we don't need to allocate a new
block, and we certainly don't need to update the summary counts.
Reported-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If pass5 finds bitmap errors in a range of clusters, don't print each
cluster number individually when we could print only the start and end
cluster number. e2fsck already does this for the non-bigalloc case.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When the rehash process is running on a bigalloc filesystem, it
compresses all the directory entries and hash structures into the
beginning of the directory file and then uses block_iterate3() to free
the blocks off the end of the file. It seems to call
ext2fs_block_alloc_stats2() for every block in a cluster, which is
unfortunate because this function allocates and frees entire clusters
(and updates the summary counts accordingly). In this case e2fsck
writes out incorrect summary counts.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When we're appending a block to a directory file or the journal file,
and the new block is part of a cluster that has already been allocated
to the file (implied cluster allocation), don't update the bitmap or
the summary counts because that was performed when the cluster was
allocated.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When bigalloc is enabled, using ext2fs_block_alloc_stats2() to free
any block in a cluster has the effect of freeing the entire cluster.
This is problematic if a caller instructs us to punch, say, blocks
12-15 of a 16-block cluster, because blocks 0-11 now point to a "free"
cluster.
The naive way to solve this problem is to see if any of the other
blocks in this logical cluster map to a physical cluster. If so, then
we know that the cluster is still in use and it mustn't be freed.
Otherwise, we are punching the last mapped block in this cluster, so
we can free the cluster.
The implementation given only does the rigorous checks for the partial
clusters at the beginning and end of the punching range.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The LIST_HEAD() macro conflicts with the <sys/queue.h> declaration
of the same name. Delete the unused LIST_HEAD() macro from the
libext2fs and libblkid headers to avoid compiler warnings. It can
be replaced by INIT_LIST_HEAD() or LIST_HEAD_INIT() if needed.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Include ext2fsP.h in fileio.c for ext2fs_file_block_offset_too_big()
declaration. Fix up the declaration to mark it extern in the header.
Include <strings.h> header for strcasecmp() in tune2fs.c if available,
as described in the strcasecmp(3) man page, instead of doing this
indirectly by declaring _BSD_SOURCE and getting it from <string.h>.
If CONFIG_QUOTA is undefined, parse_quota_opts() is unused in
tune2fs.c so #ifdef it out.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Clean up the formatting of the debugfs(8) man page:
- add command aliases to the main descriptions
- use bold for sub-commands and italics for their arguments
- remove trailing spaces
- add "dirty" sub-command
- clarify "block_dump" and "zap_block" sub-commands
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Compiling with LLVM generates a large number of warnings due
to the use of _() for wrapping strings for i18n:
warning: format string is not a string literal
(potentially insecure) [-Wformat-security]
./nls-enable.h:4:14: note: expanded from macro '_'
#define _(a) (gettext (a))
^~~~~~~~~~~~
These warnings are fixed by using "%s" as the format string,
and then _() is used as the string argument.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Don't print a verbose configure error in parse-types.h if
<asm/types.h> missing and __[SU]*_TYPEDEF is unset. This is
always the case for non-Linux builds.
The printf formatting strings all use "%llu" for printing 64-bit
values and this it produces a large number of warnings if __u64
is defined as "unsigned long". If __U64_TYPEDEF isn't set use
"unsigned long long" for __u64 in ext2-types.h and blkid-types.h
by default instead of using "unsigned long".
Fix a few places where "%d" or "%u" or "%Lu" were used to print a
64-bit value, by converting them to use "%lld" or "%llu" instead.
Fix a few places where "%lu" was used to print .tv_usec, by casting
the variable to "(long)" since .tv_usec is "int" on some systems.
Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The old uninit_bg checksums depend on the UUID, so prohibit changes to
the UUID if a checksumming filesystem is mounted, because this
introduces a nasty race where the kernel and tune2fs are both trying
to rewrite group descriptors at the same time, with different ideas
about what the UUID is.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Use the new ext2fs_punch() call to truncate the quota file. This also
eliminates the need to fix it to work with bigalloc.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
In the ^extent case, passing ~0ULL as the 'end' parameter to
ext2fs_punch() causes the (end - start + 1) calculation to overflow to
zero. Since the old-style mapped block files cannot have more than
2^32 blocks, just clamp it to ~0U.
This fixes a regression in t_quota_2off with the patch "libext2fs: use
ext2fs_punch() to truncate quota file" applied.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When we set the file size, find the block containing EOF, and zero
everything in that block past EOF so that we can't return stale data
if we ever use fallocate or truncate to lengthen the file.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When we're rehashing directories, it's possible that an extent block
(or a map block) could be (silently) allocated by the underlying
libext2fs when expanding the directory. This silent allocation is not
captured in block_found_map, which is disastrous if later the rehash
process expands another directory and uses that same block from
before without realizing that it's now in use.
Therefore, if we notice that the free block count has dropped by more
than what e2fsck allocated itself during the expansion, we iterate the
directory's blocks a second time to ensure that these silent
allocations are marked in the found blocks bitmap.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we're asked to punch a file with no data blocks mapped to it and a
non-zero length, we don't need to do any work in ext2fs_punch_extent()
and can return success. Unfortunately, the extent_get() function
returns "no current node" because it (correctly) failed to find any
extents, which is bubbled up to callers. Since no extents being found
is not an error in this corner case, fix up ext2fs_punch_extent() to
return 0 to callers.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When deleting an entire extent, we cannot always slip to the previous
leaf extent because there might not /be/ a previous extent.
Attempting to correct for that error by asking for the 'current' leaf
extent also doesn't work, because the failed attempt to change to the
previous extent leaves us with no current extent.
Fix this problem by recording the lblk of the next extent before
deleting the current extent and _goto()ing to the next extent after
the deletion.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we're using ext2fs_file_write() to write to a hole in a file,
ensure that we can actually allocate the block before updating i_size.
In other words, don't update i_size and don't return success if we hit
an error while allocating space.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fix up a few places where we ignore return values.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Fix memory allocation calculations and check for NULL pointer returns.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Check the return values from ext2fs_get_block_bitmap_range2(); if an
error happened, print that and don't print garbage bitmap.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
sysconf(_SC_PAGESIZE) will probably never return an error, but just in
case it does, we shouldn't pass what looks like a huge number to
sync_file_range() and posix_fadvise().
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If someone umounts the filesystem between statfs64() and the getmntent()
iteration, we can exit the loop having never set mnt_type, and strcmp
can crash. Fix the potential NULL dereference.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
ext2fs_free_mem() takes a pointer to a pointer, similar to
ext2fs_get_mem(). Improve the documentation, and fix debugfs.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The caller of dump_file provides a fd to write to, so the caller
should also dispose of the fd. Also, the fd never gets closed if
preserve=1.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
When reading or writing file blocks, use the IO manager routines that
can handle 64bit block numbers.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
debugfs should use strtoull wrappers for reading block numbers from
the command line. "unsigned long" isn't wide enough to handle block
numbers on 32bit platforms.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
If we have to create a big symlink (i.e. one that doesn't fit into
i_block[]), we are not 64bit block safe and the namei code does not
handle extents at all. Fix both.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
We should really use the ext2fs memory allocator functions in
copy_file(), and we really should return a value if there's allocation
problems.
Also fix up a minor bogosity in an error message.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Cc: Robert Yang <liezhi.yang@windriver.com>
Cc: Darren Hart <dvhart@linux.intel.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Forbid clients from trying to map logical block numbers that are
larger than the lblk->pblk data structures are capable of handling.
While we're at it, don't let clients set the file size to a number
that's beyond what can be mapped.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
The help text for debugfs' init_filesys command is incorrect; the
second parameter is the size of the filesystem in blocks, not the size
of an individual filesystem block. There is in fact no way to set
that parameter.
Reported-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
mke2fs has a series of checks to ensure that we don't create a
filesystem too big for its blocksize -- if auto-64bit is on, then it
turns on 64bit; otherwise it complains. Unfortunately, it performs
these checks before looking in mke2fs.conf for a blocksize, which
means that the checks are incorrect if the user specifies a non-4096
blocksize in the config file and says nothing on the command line.
The bug also has the effect of mandating a 4k block size on any block
device larger than 4T in that situation. Therefore, read the block
size from the config file before performing the 64bit checks.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
For each site where we test for a large file (> 2GB) and set the
LARGE_FILE feature, use a helper function to make the size test
consistent with the test that's in e2fsck. This fixes the fsck
complaints when we try to create a 2GB journal (not so hard with 64k
block size) and fixes the incorrect test in fileio.c.
Reviewed-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Andreas Dilger