Why ext4? [LWN.net]

Why ext4?

Posted Oct 11, 2020 21:27 UTC (Sun) by WolfWings (subscriber, #56790)
In reply to: Why ext4? by eru
Parent article: The ABI status of filesystem formats

SquashFS _requires_ compression, so there'd be a compression -> decompression layer involved. CRAMFS has a ~256MB total size limit. ROMFS is 32-bit so it'd run into obstacles around the 4GB mark (or maybe 2GB, I don't think it's been tested for >2GB total size to my knowledge) but it would be the lightest weight option IMHO.

But this does boil down to they were abusing an invalid filesystem data structure configuration that was not previously checked. That they were declaring that a particular flag marked the filesystem as such a monstrosity doesn't change that they were laying multiple copies of a data structure on top of each other instead of pushing upstream for a proper "read only" flag that did away with those structures entirely is the core invalidity.

to post comments

Why ext4?

Posted Oct 13, 2020 15:30 UTC (Tue) by NYKevin (subscriber, #129325) [Link] (1 responses)

What about EROFS? It doesn't look very complicated, but it does appear to support most things you might reasonably want to do.

https://www.kernel.org/doc/html/latest/filesystems/erofs....

Why ext4?

Posted Oct 18, 2020 16:19 UTC (Sun) by WolfWings (subscriber, #56790) [Link]

Oh, yup, I'd missed that one going through the main filesystems, that one looks to nail it and he "ROMFS but a full Linux filesystem" so it's just about ideal.

Why ext4?

Posted Oct 20, 2020 5:38 UTC (Tue) by eru (subscriber, #2753) [Link]

Squashfs, unlike cramfs, does not have problematic size limitations: "Files up to 2^64 bytes are supported; file systems can be up to 2^64 bytes" (https://tldp.org/HOWTO/SquashFS-HOWTO/whatis.html).