Hi,
For Btrfs and more recently XFS they explicitly support changing the
label while mounted. Documentation is available saying it can be done
and stracing shows it is using a file system specific ioctl() to make
the change.
However for ext2/3/4 file systems I can't find anything explicitly
documented, and stracing "tune2fs -L" shows that it writes directly to
the block device while the file system is mounted. That seems unsafe as
I presume the kernel would be expecting to have exclusive access to the
block device while the file system is mounted. However it always seems
to work. I am wondering if it is safe and supported to change an
ext2/3/4 file system label while it is mounted?
(I have a old GParted enhancement request for this, hence the question.
https://bugzilla.gnome.org/show_bug.cgi?id=600496)
Thanks,
Mike
# mkfs.btrfs /dev/sdb1
# mount /dev/db1 /mnt/1
# strace btrfs filesystem label /mnt/1 btrfs_label
...
open("/mnt/1", O_RDONLY|O_NOATIME) = 3
ioctl(3, BTRFS_IOC_SET_FSLABEL, "btrfs_label") = 0
# mkfs.xfs /dev/sdb2
# mount /dev/sdb2 /mnt/2
# strace xfs_io -c "label -s xfs_label" /mnt/2
...
open("/mnt/2", O_RDONLY) = 3
...
ioctl(3, BTRFS_IOC_SET_FSLABEL, "xfs_label") = 0
# mkfs.ext2 /dev/sdb3
# mount /dev/sdb3 /mnt/3
# strace tune2fs -L ext2_label /dev/sdb3
...
open("/dev/sdb3", O_RDWR) = 3
...
lseek(3, 134217728, SEEK_SET) = 134217728
write(3, "\0\200\0\0\0\0\2\0\231\31\0\0\217\367\1\0\365\177\0\0\0\0\0\0\2\0\0\0\2\0\0\0"...,
1024) = 1024
lseek(3, 402653184, SEEK_SET) = 402653184
write(3, "\0\200\0\0\0\0\2\0\231\31\0\0\217\367\1\0\365\177\0\0\0\0\0\0\2\0\0\0\2\0\0\0"...,
1024) = 1024
fsync(3) = 0
lseek(3, 1072, SEEK_SET) = 1072
write(3, "\vM", 2) = 2
lseek(3, 1144, SEEK_SET) = 1144
write(3, "ext2_label", 10) = 10
fsync(3) = 0
close(3)