fstools/libfstools, branch master

mount_root: add kernel parameter to specify the overlay fileystem type

2026-03-17T16:16:16Z

The kernel parameter 'fstools_overlay_fstype' can now be used to specify the preferred overlay fileystem type. The supported values are 'ext4', 'f2fs' and 'auto'. Type 'auto' is the default and represents the usual behavior of selecting the filesystem type based on available space. Signed-off-by: Rodrigo Balerdi

libfstools: make get_var_from_file() reusable

2026-03-17T16:16:16Z

Move get_var_from_file() to common library. Signed-off-by: Rodrigo Balerdi

libfstools: Rename PATH_MAX to MTD_PATH_MAX

2026-02-15T18:20:01Z

The PATH_MAX defined here could conflict with PATH_MAX in include/limits.h. The definition in snapshot.h is not used. Just remove the definition from snapshot.h, because it is not used and rename the other definition. Signed-off-by: Hauke Mehrtens

libfstools: add support for EroFS based images

2025-07-05T13:17:57Z

Add support to libfstools to parse and support EroFS based images to permit correct mounting of OverlayFS with it. (in addition to SquashFS support) While at it rework the code to better handle the 2 different filesystem introducing a similar pattern. Signed-off-by: Gao Xiang Co-Developed-by: Christian Marangi Signed-off-by: Christian Marangi

libfstools: skip JFFS2 padding when BLOCKSIZE was given

2024-07-14T18:40:03Z

When BLOCKSIZE is defined we call padjffs2 with the defined alignment to pad towards. That results in one of more block of all-1s instead of the expected 0xdeadc0de and ultimaltely breaks restoring configuration as fstools won't look for a the GZ signature indicating the config backup in this case. Change that and also look for config backup when all-1s are used for padding. Fixes: ee54c6b ("libfstools: skip JFFS2 padding on block devices") Signed-off-by: Daniel Golle

fstools: remove redundant F2FS_MINSIZE definition

2024-04-04T00:53:37Z

It's defined in common.h, which rootdisk.c also includes. Remove the redundant definition from the latter. Signed-off-by: Rui Salvaterra

libfstools: fit: improve fit_volume_find string handling

2024-01-22T00:45:32Z

While string are hardcoded and it's impossible to overflow it, make the string handling more secure to mute Coverity Scan report by using strncpy and adding a define for the max size of the DEVPATHSTR. Fix Coverity Scan CID 1586643: Security best practices violations (STRING_OVERFLOW). Signed-off-by: Christian Marangi

libfstools: force mkfs when formatting overlay

2024-01-15T20:53:46Z

As the rootfs_overlay volume is dynamically allocated on platforms which allow for that, the start of the volume can end up being at the location of previously stored data. This can happen when writing an SD card image to an SD card previously used for something else. sysupgrade should always wipe preexisting data, but I haven't checked if all platform implementations actually always do so). As mkfs is cautious not to accidentally overwrite existing data, this can result in the system to hang on boot, waiting for confirmation on the serial console: mount_root: overlay filesystem in /dev/fitrw has not been formatted yet /dev/fitrw contains `ISO-8859 text, with very long lines (65536), with no line terminators' data Proceed anyway? (y,N) Use the FORCE option (-F for mkfs.ext4, -f for mkfs.f2fs) to make sure the volume is always formatted irrespective of its previous content. Signed-off-by: Daniel Golle

libfstools: partname: raise priority to 25

2023-12-11T22:28:44Z

Make sure partname has precedence over UBI and MTD. We already got ways to prevent partname for matching partitions if needed, however, we would need them as well for UBI and MTD to prevent use of UBI rootfs_data despite having started from MMC on boards having both (such as the BPi-R3). Reported-by: Rui Salvaterra Signed-off-by: Daniel Golle

libfstools: add uImage.FIT fitblk driver

2023-12-01T22:59:52Z

The upcoming 'fitblk' block driver exposing the filesystem sub-image(s) of a uImage.FIT storage on a block device as /dev/fit%u. In case of the block device being a fixed-sized partition, the remaining space (for rootfs_data) is mapped to /dev/fitrw. Recognize both devices and assign the highest priority to the new driver. Signed-off-by: Daniel Golle