linux: File systems, build ext4 into the kernel.

[thomas]

Having ext4 support as a module hampers debugging of Mariux kernels.
Think of a Linux VM that tries to boot from an ext4 filesystem :)

Also ext4 my be an other candidate for the root-fs on Mariux.
(and while on it, putting in ext2/3 will not hurt, or?)

[donald]

Agreed to EXT4. But EXT2 and EXT3 should be N or M. Why waste memory for filesystems which we don't boot from and don't want to boot from in the future?

[thomas]

As modules ext2/3 support needed, think of old USB-sticks. So N is IMHO not an option.
With M we waste harddisk space (modules each have a full bloat of debug info), and have a way split situation ext4 in kernel but the others are not.

[donald]

Hmm. Does a compiled-in module have less debug-info? Just courious, compiled-in modules surly need less disk space than dynamic modules, for example because of less elf-headers and filesystem metadata.

But does it matter? How much disk space has a kernel with static ext2 vs dynamic ext?. Is it worth losing the option to, for example, disable ext2 by rm ext2.ko vs reboot if an exploitable security problem is discovered in it?

[donald]

When filesystems are mounted with -t auto are the candidate filesystems (if dynamic) loaded into the kernel or is the filesystem identified by some external.means or how does it work?

Because it has, of course, a lot of unneeded overhead to have a filesystem loaded and initialized which is not used by 99,9% of the boots.