| Age | Commit message (Collapse) | Author |
|---|
|
Closing 'out_fd' is missed in handling paths in open_target(), fix it
by adding closing operations before returning.
Fixes: a48340c335dab ("mkfs.ubifs: use libubi to format UBI volume")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The libubi could be opened in get_options(), don't forget to close it
in error handling paths in main(). Also close libubi in error handling
paths in open_ubi(). To implement that, extract libubi_close() into a
new helper function close_ubi().
Besides, invoke crypto_cleanup() in error handling paths in main().
Fixes: a48340c335dab ("mkfs.ubifs: use libubi to format UBI volume")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The size of directory should be the total length of encrypted entry name,
otherwise it could trigger errors while checking filesystem:
dbg_check_filesystem [ubifs]: directory inode 89 size is 352, but
calculated size is 400
Fixes: 4c55918dd747d ("Implement filename encryption")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The type of xattr entry should be regular file, otherwise xattr entry
and xattr inode have inconsistent types.
Fixes: 50044efbd6e713 ("mkfs.ubifs: Add extended attribute support")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Variable 'cipher_name' is defined on stack without initialization, when
an user invokes mkfs with '--key' and without '-C', 'cipher_name' is a
random value, which could cause init_fscrypt_context() failed to find a
valid cipher.
Fix it by initializing 'cipher_name' as NULL when it is declared.
Fixes: cc4c5e295f546 ("mkfs.ubifs: Enable support for building without crypto")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Direct writing is not allowd on an UBI volume, unless the volume is set
with direct_write property. The open_target sets direct_write property,
don't forget to clear direct_write property for UBI volume when closing
target.
Fixes: a48340c335dab ("mkfs.ubifs: use libubi to format UBI volume")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
PKG_CHECK_MODULES confusingly does not check to see if a static version
of the library is available.
Actually with older glibc, it might fail as lpthread needs to be passed
to linker flags (Libs.private in pc files) which can only be obtained if
--static is passed to pkg-config.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
remount_tested_fs() and recover_tested_fs() both have
almost the same code for remounting the target FS
RO then RW. Split this sequence into a new
function called remount_ro_rw().
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Link: https://lore.kernel.org/linux-mtd/5530437c-b564-461f-26af-85bde1a00cb3@huawei.com/T/#mad1d043095b3f64a1f1eddd2d87b1545e329e141
Signed-off-by: Csókás, Bence <csokas.bence@prolan.hu>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
remount_tested_fs() and recover_tested_fs() both have
almost the same code for umount'ing the target FS and
mount'ing it back. Split this sequence into a new
function called umount_and_remount().
Signed-off-by: Csókás, Bence <csokas.bence@prolan.hu>
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Calling ubi_update_start() may fail if exclusive access to the UBI
volume cannot be established by the kernel. This is likely to happen if
an UBI volume is updated directly after creation, because current
versions of (systemd-)udevd run their internal 'blkid' also for UBI
volumes as soon as they appear (trying to read the UBIFS UUID).
There are mainly to options for this case:
1. Don't allow udevd to read UBI volumes at all (means loosing
functionatlity)
2. Wait until udevd has finished accessing a new volume (requires
cooperation with udevd)
3. Simply retry after short time in case of EBUSY
For using option 3, the documentation should state that the concrete
error code can be read via errno (this has always been the case since
libubi has been introduced).
Link: https://groups.google.com/g/swupdate/c/8NVooKjD9oo
Signed-off-by: Christian Eggers <ceggers@arri.de>
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
fectest.c formats a struct timeval with the assumption that time_t and
suseconds_t are aliases for long, but some 32-bit systems now define
them as long long in order to support timestamps beyond 2038.
As this is an elapsed time and not a timestamp, both fields should be
within the range of long, so cast them to long before formatting
rather than moving to long long.
Signed-off-by: Ben Hutchings <ben.hutchings@mind.be>
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Remove the strict check of the with or enable value and use the
generated variable instead of our own.
Also, don't use WITH_GETRANDOM as an implicit WITH_UBIHEALTHD, split
the two autoconf/automake variables and handle them separately.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This commit modifes the handling of the OpenSSL dependency to work the
same as the other optional dependencies, except that the variable
"with_crypto" is used instead of "with_openssl".
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Now that we have plumbing in place for both jffsX-utils and mkfs.ubifs,
add the missing autoconf and automake changes to allow mtd-utils to be
built without a hard dependency on zlib.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This one is a bit trickier than adding WITH_ZLIB ifdefs. Some parts
of the code assume that zlib is always present and have complicated
fallback behavior. Particularly the "favor_lzo" compression method
that uses either zlib or lzo, whichever produces the better result.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The jffsX-utils already have a CONFIG_JFFS2_ZLIB define, but it is
statically defined in a header and not used consistently. This patch
first replaces it with a "WITH_ZLIB" define, provided via automake,
to bring it in line with the already existing "WITH_LZO". Then, the
missing checks and typedefs are added in jffs2reader.c.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
- Simple fixup in autoconf.ac
- Make compilation of jffsX-utils compr_lzo.c optional. Simply
don't compile it if we built without LZO and remove the place
holders. They are not used anyway if we build without LZO.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Don't use super pedantic parsing of the argument and work with the
generated variable instead of assigning it to our own and set it
to "check" if not value is assigned. Then search for a dependency
if the with variable is anything other than "no" and fail if it
was set to "yes".
In addition, the WITHOUT_xxx defines are replaced with WITH_xxx defines.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
After removing libtool, there is no more cross-check for AR. Because
some slightly older versions of autoconf don't have AC_PROG_AR, simply
insert the implementation one-liner.
We already tried bumping the minimum autoconf version once in the past,
but apparently that is an issue for a number of distros.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
mkfs.uibfs can add Selinux xattrs from a labeling file using
libselinux to parse it. The commit that added this feature simply
introduced a separate function, inode_add_selinux_xattr, which is
called instead of inode_add_xattr. If no --selinux argument is
specified for mkfs.ubifs, this is a no-op.
The problem is, that this breaks xattr scanning for any build with
Selinux enabled. The Selinux version is always called and it does
not scan for xattrs on the filesystem, or dispatch to the original.
This commit fixes the xattr scanning behavior. We unconditionally call
both functions (they each have no-op implementations if the feature
is missing) and in the regular xattr scanning code, we skip selinux
attributes, if the --selinux option was given.
Fixes: f1feccec5ad8 ("mkfs.ubifs: Implement selinux labelling support")
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <goliath@infraroot.at>
|
|
The ioctl UBI_IOCATT has been extended with need_resv_pool parameter in
[1].
This parameter is used for deciding whether to reserve PEBs for filling
pool/wl_pool for target ubi device. This parameter will be effective
when fastmap is enabled, which will slow down the frequency of updating
fastmap by filling more free PEBs in pool/wl_pool. See details in [2].
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ac085cfe57df2cc1d7a5c4c5e64b8780c8ad452f
[2] https://bugzilla.kernel.org/show_bug.cgi?id=217787
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The ioctl UBI_IOCATT has been extended with disable_fm parameter after
[1].
This parameter is used for disabling fastmap for target ubi device.
If 'disable_fm' is set, ubi doesn't create new fastmap even the module
param 'fm_autoconvert' is set, and existed old fastmap will be destroyed
after attaching process.
A simple test case in [2].
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=669d204469c46e91d99da24914130f78277a71d3
[2] https://bugzilla.kernel.org/show_bug.cgi?id=216278
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
JFFS2 supports buffer mode for ECC'd NOR Flash that cleanmarker size
is rounded up to mtd->writesize, while the '-j' option in flash_erase
utility uses fixed 12 bytes. That makes flash program ops unaligned to
mtd->writesize and causes program error or disables ECC.
The mkfs.jffs2 utility supports '-c' option that allows users to specify
cleanmarker size. This patch implements '-c' option in the flash_erase
that can be used along with '-j' option.
Signed-off-by: Takahiro Kuwano <Takahiro.Kuwano@infineon.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This option is useful for power fail resilient bootloader updates, for
systems that check only the partition header for validity before
attempting to load the bootloader. For example zynq.
Postponing the write of a specified amount of bytes at the beginning of
a bootloader makes sure a written bootloader slot is not considered
valid until all the data is written.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This no-op was remarked by coverity. According to git-blame, it has
been this way since the initial commit in 2006.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
For the most part, this was straight forward. For the very old
releases, there were lots of changes that are not straight forward
to untangle.
The following two releases posed issues:
- Version 1.4.1 does not have a tag. The release tarball on FTP
is from 2010-10-19 and appears to contain 1.4.0 with commit
b864c387e8f16273aad1453d6457d847e29c3d25 cherry-picked on top
of it (except for one line that was fixed in the release, but
broken in the repo).
- Version 1.4.6 also does not have a tag. The release tarball
is from 2011-08-19. This _presumably_ corresponds with commit
b602ab59b9782003d1179488d8c1580bc8563738 that bumps the version
number.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The CHANGELOG tries to summarize the change history of the project
and currently contains changes from 2.0.0 onwards that could easily
be extracted from the release mails.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
On error, lseek returns a negative value. We need to check for that
and abort, instead of passing -1 to malloc and later on to read_nocheck.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
In nandflipbits, nandtest and ubiscan, uint64_t integers are printed
to stdout using "%llu" as a format specifier, but on platforms like
x86_64, uint64_t is actually typedef'd as `unsigned long` only.
For compatibillity across platforms, simply use the C99 printfs
macros instead.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The libubigen.h header relies on external declarations in ubi-media.h.
While not technically needed, it generates warnings in mtdinfo.c if
those are not visible. All other places where libubigen is used include
the header first. This is primarily to silence compiler warnings related
to the missing declrations.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
We do not build any shared libraries, there is no reason to use libtool
or disable static. We do build static libraries, for which we need ranlib.
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Running nandtest on devices with sizes of 4Gb or more does not work, as
the size returned in mtd_info_user is reported as 0 for 4Gb devices for
example.
The patch uses sysfs to figure out the size of the nand device (as
recommended in a comment in include/mtd/mtd-abi.h) and changes sizes
and offsets to 64 bit values.
[From: Christian Wendt <cw@brainaid.de>]
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This introduces a new feature to the MTD command line utilities that
allows MTD devices to be referenced by name instead of device node. For
example this looks like:
> # Display info for the MTD device with name "data"
> mtdinfo mtd:data
> # Copy file to MTD device with name "data"
> flashcp /my/file mtd:data
This follows the syntax supported by the kernel which allows MTD
device's to be mounted by name[1].
Add the function mtd_find_dev_node() that accepts an MTD "identifier"
and returns the MTD's device node. The function accepts a string
starting with "mtd:" which it treats as the MTD's name. It then attempts
to search for the MTD, and if found maps it back to the /dev/mtdX device
node. If the string does not start with "mtd:", then assume it's the old
style and refers directly to a MTD device node.
The function is then hooked into existing tools like flashcp, mtdinfo,
flash_unlock, etc. To load in the new MTD parsing code in a consistent
way across programs.
[1] http://www.linux-mtd.infradead.org/faq/jffs2.html#L_mtdblock
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
This is a convenience function for end users. In some situations it's
easier to reference MTD device's by their name then by MTD number, as
the name may be more reliable if device partitioning is dynamic or for
porting between systems.
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Below failure happens when mkfs.ubifs --selinux=FILE ...
"Error: bad file context FILE 1"
"No such file or directory (error 2)"
It is fixed by this change.
Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Add correct casting for written to prevent overflow that size_t is
only 32 bits on a 32-bit platform.
Signed-off-by: cchoux <chou.cosmo@gmail.com>
Reviewed-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The Read While Write (RWW) feature allows to perform reads from the
flash array into cache while a program (from cache) or an erase
operation happens, provided that the two areas are located on different
banks.
The main benefit is the possible reduced latency when requesting to read
a page while a much longer operation is ongoing, like a write or an
erase.
We can try to compare the positive impact of such a feature by enhancing
the flash_speed test tool with the following test:
- Measure the time taken by an eraseblock write in parallel with an
eraseblock read.
- Measure when the read operation ends.
- Compare the two to get the latency saved with the RWW feature.
To be sure the mtd_write actually starts (and acquires the necessary
locks) before the mtd_read does, we use SCHED_FIFO at rather high
(arbitrary) priorities, respectively 42 and 41.
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
In order to be able to have interleaved measures, let's not use the
start and finish global variables from the time helpers directly,
provide parameters for these variables so that we can provide either the
global entries, or more specific ones when relevant.
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
If the partition size passed in to the BLKPG_ADD_PARTITION ioctl is 0,
Linux will make the partition fill to the end of the device[1]. Document
this feature in the help.
[1] https://elixir.bootlin.com/linux/v6.0.6/source/drivers/mtd/mtdpart.c#L254
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The --partition mode is not verifying that data is being written
successfully.
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
|
The DIFF_BLOCKS code requires that src and dest buffers be large enough
to hold one MTD erasesize. This is because each loop operates on one
eraseblock so that it can erase and write one whole sector. But the src
and dest buffers are fixed at BUFSIZE, so on platforms where the MTD
erasesize are larger then BUFSIZE it will overflow the buffers.
Instead allocate the buffers dynamically so that they can be sized to
fit the erasesize.
Signed-off-by: Brandon Maier <brandon.maier@collins.com>
Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at>
|
