Image generation

Image generation

[claudio]

I played with disk image generation and it's almost usable again. There
are still a few things to fix, like SIBO stuff and ktcp installation,
but these should be easy.

Here is the description of the changes, please let me know what you think:

The previous system had three rules for each image, in the format (for full3):

$(FULL3_TARGET_FS):
       dd if=/dev/zero of=$(FULL3_TARGET_FS) bs=1024 count=$(FULL3_TARGET_BLKS)


mount_full3fs: $(FULL3_TARGET_FS) $(ELKSCMD_DIR)/rootfs_template/etc/issue
       $(MKFS) $(MKFS_OPTS) $(FULL3_TARGET_FS) $(FULL3_TARGET_BLKS)
       sudo -u root mount $(FULL3_TARGET_FS) $(TARGET_MNT) $(LOOP)
       (cd $(ELKSCMD_DIR)/rootfs_template ; tar cf - --exclude CVS *) | \
               (cd $(TARGET_MNT) ; tar xpvf -)
       (cd $(TARGET_MNT)/dev ; ./MAKEDEV)


full3: mount_full3fs $(ELKS_DIR)/arch/i86/boot/Image $(FD_BSECT) $(KHELPER)
       @for i in $(DIRS); do make -C $$i rfs ; done
       cp $(ELKS_DIR)/arch/i86/boot/Image $(TARGET_MNT)/boot/linux
       cp $(KHELPER) $(TARGET_MNT)/boot/boot
       umount $(FULL3_TARGET_FS)
       dd if=$(FD_BSECT) of=$(FULL3_TARGET_FS) bs=512 count=2 conv=notrunc


I changed that to a generic rule called by all targets with parameters for
image name, image size and type of installation, avoiding code duplication
and errors. I'll prepare a patch and submit as soon as I fix the tcp and
SIBO issues.

As Tommy pointed already, there's no space in the 360k root image for all
the utilities. We can save a lot of inodes in devices (e.g. using 16 hd*
instead of 64), but the binaries still eat too much space (even with a
simplified shell in the 360k image).

The other images (full3, full5, comb, etc) are being correctly generated.

Image generation

[Jody]

> *snip*
> >
> > As Tommy pointed already, there's no space in the 360k root image for
all
> > the utilities. We can save a lot of inodes in devices (e.g. using 16
hd*
> > instead of 64), but the binaries still eat too much space (even with a
> > simplified shell in the 360k image.
>
> How about just hd[ab][1234] instead of 16?  It is highly unlikely that a
> system with a 360K drive would even have a second hard disk anyway.
Eight
> inodes tops for hard disks...maybe four since we probably won't need four
> partitions per disk either...
>
> Jody