7.2. File system development in barebox

The barebox file system support is based heavily on Linux.

Access to all file systems goes through the virtual file system layer, which provides uniform access to all mounted file systems under the same root.

As power-fail safe writing of file system couldn’t be guaranteed, most file systems supported by barebox are read-only. Safe writing is possible, however, via the Barebox State Framework.

For an up-to-date listing of writable filesystems, refer to the CONFIG_FS_WRITABLE Kconfig symbol.

7.2.1. Testing File systems

Nearly all file system operations have commands that directly exercise them:

Command	Operations
cat - concatenate file(s) to stdout	open, close, read
echo - echo args to console	create, write
sync - flush cached writes	flush
erase - erase flash memory	erase
protect - enable flash write protection	protect
md - memory display	lseek, memmap
rm - remove files	unlink
mkdir - make directories	mkdir
rmdir - remove empty directory(s)	rmdir
ln - create symlink (make a new name for a file)	symlink
readlink - read value of a symbolic link or canonical file name	readlink
ls - list a file or directory tree - list contents of directories in a tree-like format.	opendir, readdir, closedir
stat - display file status	stat
truncate - truncate files to size	truncate

This leaves two specialized operations that can’t be easily tested via the shell:

• discard_range: Advise that a range need not be preserved

• ioctl: Issue device-specific output and input control commands

7.2.2. Unused metadata

barebox currently ignores ownership and permission information inside file systems as well as special nodes like FIFOs or sockets. When porting file systems, these parts can be omitted.

7.2.3. Background execution

Outside command context (i.e. in pollers and secondary barebox threads), virtual file system access is only permitted with ramfs.