mtd-utils.git, branch master

ubi-utils: add wear-leveling threshold parameter to ubiattach

2026-08-03T05:42:07+00:00

Add wl_threshold support end-to-end: expose the kernel's wear-leveling threshold via a new --wl-threshold / -w option in ubiattach, plumb it through libubi, and define it by replacing 4 bytes of padding in struct ubi_attach_req. Valid range is 2-65536; 0 means use the kernel default. Signed-off-by: Ran Hongyun Signed-off-by: David Oberhollenzer

lib: libmtd: fix sign extension of ioctl request in mtd_xlock

2026-08-03T05:42:05+00:00

The req parameter of mtd_xlock() was declared as int. On 64-bit big-endian architectures such as ppc64el, _IOW()-derived ioctl numbers have bit 31 set (because _IOC_WRITE=4 is placed at bit 29). Storing such a value in a signed int and then passing it to ioctl(), whose second argument is unsigned long, causes implicit sign extension: 0x80084d05 becomes 0xffffffff80084d05, which does not match the expected constant and fails the cmocka check_expected() assertion in the unit tests for mtd_lock and mtd_unlock. Fix by declaring req as unsigned long, matching the type used by the ioctl() syscall interface. Fixes: test_mtd_lock / test_mtd_unlock failures on ppc64el Signed-off-by: Bastian Germann Signed-off-by: David Oberhollenzer

mtd-utils: don't include features.h

2026-08-03T05:42:02+00:00

This is an internal header that's already included with others. Signed-off-by: Rosen Penev Signed-off-by: David Oberhollenzer

Release mtd-utils-2.3.1

2026-04-13T07:07:48+00:00

Signed-off-by: David Oberhollenzer

Update CHANGELOG.md

2026-04-13T07:04:38+00:00

Signed-off-by: David Oberhollenzer

mtd-utils: tests: jittertest: reject overlong file names

2026-04-13T06:42:56+00:00

plotJittervsFill copies the -f argument into a 250-byte buffer with strncpy(..., sizeof(LogFile)). A 250-byte file name leaves the buffer unterminated, and the subsequent fopen() reads past the end of LogFile. JitterTest uses the same fixed-size file name pattern for -r, while -c still silently truncates overlong names and -f already rejects them. Validate jittertest file name arguments before copying them so these options all reject overlong input instead of truncating it or reading past the end of fixed-size buffers. Add a shell regression test that exercises the accepted and rejected boundary lengths for plotJittervsFill and JitterTest during make check. Signed-off-by: Aviv Daum Signed-off-by: David Oberhollenzer

misc-utils: docfdisk.c: validate partition size to prevent arithmetic overflow

2026-04-13T06:42:56+00:00

report of the static analyzer: Possible integer overflow: right operand is tainted. An integer overflow may occur due to arithmetic operation (addition) between variable 'block' and value { [1, 4294967295] } of 'nblocks[i]', when 'block' is equal to '1' correct explained: Added bounds check before incrementing block counter to ensure that adding nblocks[i] does not exceed totblocks. This prevents potential integer overflow when user-specified partition sizes are too large, which could lead to incorrect partition table layout and device corruption. The validation ensures safe arithmetic by checking block + nblocks[i] <= totblocks using unsigned comparison. Signed-off-by: Anton Moryakov Signed-off-by: David Oberhollenzer

misc-utils: docfdisk.c: fix potential integer underflow in partition size calculation

2026-04-13T06:42:56+00:00

report of the static analyzer: Possible integer underflow: right operand is tainted. An integer underflow may occur due to arithmetic operation (unsigned subtraction) between variables 'totblocks' and 'block', where 'totblocks' is in range { [0, 4294967295] }, and 'block' is tainted { [0, 4294967295] } correct explained: Added validation check before calculating remaining space for partition. The issue occurred when setting the last partition size to 0, which triggers calculation 'totblocks - block'. Without validation, if block >= totblocks, this would result in integer underflow due to unsigned arithmetic, potentially creating a partition with enormous size and leading to device corruption. Signed-off-by: Anton Moryakov Signed-off-by: David Oberhollenzer

fsck.ubifs: fix platform dependant ino_t and loff_t formatting

2026-04-13T05:32:41+00:00

On architectures such as armv7-a, ino_t and loff_t are unsigned long long rather than unsigned long. In such cases, the printf format specifier "%lu" is not appropriate and causes an incorrect address offset. mtd-utils/ubifs-utils/fsck.ubifs/problem.c:224 log_out(c, "problem: %s, ino %lu, unreachable dentry %s, type %s%s", problem->desc, ifp->file->inum, c->encrypted && !ifp->file->ino.is_xattr ? "" : dent_node->name, ubifs_get_type_name(dent_node->type), key_type(c, &dent_node->key) == UBIFS_XENT_KEY ? "(xattr)" : ""); fsck.ubifs[484] (/dev/ubi0_0,danger mode): problem: Dentry is unreachable, ino 917, unreachable dentry (null), type checksum_typefile Furthermore, running fsck.ubifs with the --debug=4 option will almost certainly cause a SEGV at the following point. mtd-utils/ubifs-utils/fsck.ubifs/check_files.c:103 dbg_fsck("construct file(%lu) for %s node, TNC location %d:%d, in %s", inum, ubifs_get_key_name(key_type(c, key)), sn->lnum, sn->offs, c->dev_name); To ensure functionality regardless of environment, cast ino_t to unsigned long, since it will never be more than 4 bytes. For loff_t, use %lld and cast accordingly. Signed-off-by: Yuta Hayama Signed-off-by: Tomas Alvarez Vanoli Reviewed-by: Zhihao Cheng Signed-off-by: David Oberhollenzer

fsck.ubifs: don't use pointers that reference out-of-scope variables

2026-04-13T05:32:41+00:00

sn is a reference to either an ino_node, dent_node, or data_node. When sn is actually used in calls to dbg_fsck() or insert_or_update_file(), these variables must not be out of scope. Signed-off-by: Yuta Hayama Reviewed-by: Zhihao Cheng Signed-off-by: David Oberhollenzer