Archiso and EROFS compression [Solved]

Mr Green · 2021-03-20 16:08:50

Testing out the new archiso in gitlab to create arch media iso using the erofs compression/ Am getting an error while building.

Iso build fails as it tries to create erofs image with the following error:

mkfs.erofs invalid option -- 'c'

Looking at code I cannot see anything that stands out that would cause this...

Not sure if this is the right place to post as I cannot add an issue over on arch gitlab

Last edited by Mr Green (2021-03-21 09:31:20)

nl6720 · 2021-03-20 16:51:25

What is airootfs_image_tool_options set to in profiledef.sh?

Mr Green · 2021-03-21 07:55:46

I have:

'-comp' 'xz' '-Xbcj' 'x86' '-b' '1M' '-Xdict-size' '1M'

ie stock settings, looking at it now can see the 'c' ;-)

nl6720 · 2021-03-21 08:06:26

Those are mksquashfs specific options. They will not work with mkfs.erofs.

Run

$ mkfs.erofs --help

to see the supported options.

E.g., for currently best compression:

airootfs_image_tool_options=('-zlz4hc,12')

Mr Green · 2021-03-21 09:31:07

Thank you, not sure I would have figured that myself ;-) marking topic as solved....

Mr Green · 2021-03-21 09:48:43

Interesting erofs image size is near twice the size of stock squashfs.

745M squashfs
1.2G EROFS

Not tried ext4+squashfs yet

nl6720 · 2021-03-21 12:43:39

mkarchiso will use mksquashfs -comp xz if no (or empty) airootfs_image_tool_options is set.

Until EROFS implements LZMA support (still WIP), squashfs will produce a smaller image.

As for ext4+squashfs, it will be larger than just squashfs.

hsiangkao · 2021-03-22 06:38:18

Hi,
EROFS wasn't originally designed for maximum compression ratio, but minimum read amplification for better runtime/random performance.
So it used 4kb fixed-sized output compression (you could evaluate with 4kb block size squashfs, although these are not the same due to different compression method.)

Actually, big pcluster feature (rather than just 4kb) will be pending for 5.13 cycle,
https://lore.kernel.org/r/2021022718502 … -66-Pro-G1
and LZMA support is also pending after big pcluster is upstreamed.
https://lore.kernel.org/r/2021022718502 … ro-G1/T/#u

After these two feature, I think compression ratio issue compared with squashfs would be resolved
if you don't care more about random performance.

Thanks,
Gao Xiang

nl6720 · 2021-03-22 07:32:11

hsiangkao wrote:

https://lore.kernel.org/r/2021022718502 … ro-G1/T/#u

I'm getting a 404 for this link.

hsiangkao · 2021-03-22 09:56:18

hi,

...not quite sure what links behaved like this:
https://lore.kernel.org/r/2021022718502 … 66-Pro-G1/
and
https://lore.kernel.org/r/2021011717460 … o@aol.com/

Do all seem ok now?

Thanks,
Gao Xiang

nl6720 · 2021-03-22 12:30:49

These links work. Thanks!

Arch Linux

#1 2021-03-20 16:08:50

Archiso and EROFS compression [Solved]

#2 2021-03-20 16:51:25

Re: Archiso and EROFS compression [Solved]

#3 2021-03-21 07:55:46

Re: Archiso and EROFS compression [Solved]

#4 2021-03-21 08:06:26

Re: Archiso and EROFS compression [Solved]

#5 2021-03-21 09:31:07

Re: Archiso and EROFS compression [Solved]

#6 2021-03-21 09:48:43

Re: Archiso and EROFS compression [Solved]

#7 2021-03-21 12:43:39

Re: Archiso and EROFS compression [Solved]

#8 2021-03-22 06:38:18

Re: Archiso and EROFS compression [Solved]

#9 2021-03-22 07:32:11

Re: Archiso and EROFS compression [Solved]

#10 2021-03-22 09:56:18

Re: Archiso and EROFS compression [Solved]

#11 2021-03-22 12:30:49

Re: Archiso and EROFS compression [Solved]

Board footer