[Log in to get rid of this advertisement]

(Yes, I know, ext2 and ext3 are superseded by ext4, but it seems this no one asked this before.)

On file systems like FAT32 and NTFS and UDF, the maximum volume size scales proportionally with block size or cluster size. This means twice the cluster size means twice the size limit. However, ext2 and ext3 are different:

(from https://www.kernel.org/doc/Documenta...stems/ext2.txt )

Does anyone know why does the filesystem size limit with 2K blocks become four times larger than 1K blocks, but only doubles with 4K and 8K blocks?

Also, why is the size limit so low with 1K blocks? (16G) Is it arbitrarily done for performance reasons (like FAT32 does with its 2 MiB directory entry limit) or is there a technical reason?

And how come it doesn't increase with 4K to 8K blocks while becoming 16 times larger from 1K to 2K?

I can't answer your question, but I can suggest any value those filesystems have before then will expire on 19 January 2038, so creating any more of them would seem to have dubious value other than as educational exercise.

I have to think the size limits were colored by system RAM and bus sizes and speeds when EXT2 was written over three decades ago, with RAM measured in MB rather than GB, and 64MB was a lot rather than 128GB being a lot.

1 members found this post helpful.

you definitely didn't read the directions before formatting the partition

there ARE correct defaults and you ARE encourage to change them depending on what kind of server you are configuring

From the internet:

File size limits are tied to inode pointer architecture, not just block size.

Filesystem size limits are tied to total block count and block size.

The scaling is non-linear due to these architectural constraints.

I thought ext3 was fine, but I wonder if I missed something with ext2/3, so I did a search:

Yes, looks like ext2 will be no more: linuxiac.com

And seems ext3 also: lwn.net.

ext3 is a bit of a surprise to me, but from the articles I can see why they will be deprecated. Reading between the lines, to make ext3 work after 2038, you would need to re-format your file system. So if you have to do that, might as well go to ext4 or another fs.

I left ext2/3 many years ago, but I have a couple ext2 diskettes hanging around. Time to see what is on them, if possible

As long as EXT2 and EXT3 are supported by the EXT4 drivers we will be able to create and use those file systems. There will be issues, and booting or hosting any system critical files there will become an un-recommended and unsupported configuration because the date/time information on disk will be "odd". (Not exactly meaningless, but not very useful.)

The only distributions that I am aware of that still make use of EXT2 by default do so for specific purposes and using a kernel in the 3.X or 4.X version family. Functional EXT3 is more likely to be missed, but EXT4 can be tuned to serve the usage case to replace EXT3 and perform well.

In summary: 1. they are not really going to totally go away as long as EXT4 does not, but will become a lot less useful. 2. I doubt anyone will really notice after 2032. We have larger issues to address.

I would "as a matter of course" upgrade these disks to the latest version of the latest file system. Also, if they are "spinning metal," be aware of the age of the device itself.