Difficulties shrinking read-only rootfs image

Hello,

My team and I have been running into issues with reaching better disk utilization for our read-only root file system image:

root@beaglebone:~# df -h .
Filesystem      Size  Used Avail Use% Mounted on
/dev/root       604M  340M  219M  61% /
We are building an ext4 image using OpenEmbedded's oe_mkext234fs(), which makes use of the mke2fs command to generate the image from the rootfs directory. We then run a resize2fs to try for a disk utilization equal to the sum of our utilization and 5% of that utilization for root inodes, as we do not need additional space for our read-only image. However, our utilization is quite small.

I tried passing a few options to optimize for our use case:

EXTRA_IMAGECMD_ext4 = " \
   -b 1024 \
   -i 65536 \
   -m 0 \
   -O ^has_journal,^resize,^quota,^dir_index,sparse_super2 \
   -E packed_meta_blocks=1,num_backup_sb=0,lazy_itable_init=1,lazy_journal_init=1 \
   -L ${IMAGE_NAME} \
This did not improve our utilization much. I also tried adding a pretend to oe_mkext234fs(), effectively preparing a dynamically sized image before generating the filesystem from the rootfs directory. While this worked well for our more modular builds closer to 225M in size, we still require a safety buffer of about 20% for our full-sized image. Otherwise, we get:

| Copying files into the device: __populate_fs: Could not allocate block in ext2 filesystem while writing file "hll"
| mkfs.ext4: Could not allocate block in ext2 filesystem while populating file system I am wondering if anyone has experienced similar difficulties, or if anyone might have any more ideas. We are planning on moving to a more appropriate file system such as SquashFS in the future, but we currently rely on ext4 for compatibility reasons.