Set compression levels to upstream defaults

[C0rn3j]

Currently, the compression defaults are maxed out to maximize storage gains (except for xz which is set to default?), at the cost of taking minutes instead of seconds to compress.

There is previous conversation about this comparing zstd gains at max vs minimum - #455

@skwerlman made a PR that that made the values configurable - #456

This PR changes this configuration to upstream defaults, as suggested in the above discussion.

The suggestion also considered dropping the parameters altogether to just always adapt upstream defaults - it should be more user friendly for those that are trying to modify the configuration to have a set example.

This is likely to break setups where people are putting the modules on an undersized ESP (notably, when people reuse the 100MiB Windows ESP for /boot), so bumping the filesizes by a couple megabytes could tip the size over the max.

Such people can either fix their ESP size, or more simply configure dkms to use the old values that sacrifice speed for space savings.

[christian-heusel]

Wouldn't it make more sense to omit these to also pick changed upstream defaults in the future?

[C0rn3j]

Wouldn't it make more sense to omit these to also pick changed upstream defaults in the future?

The suggestion also considered dropping the parameters altogether to just always adapt upstream defaults - it should be more user friendly for those that are trying to modify the configuration to have a set example.

[scaronni]

@C0rn3j shall we leave the default configuration commented in the configuration file and just not pass any parameter in dkms.in? This way users still have a reference but the default is not explicitly declared.

[C0rn3j]

Could do that for gzip, but I don't know how to cleanly do it for xz/zstd as they set more than just compression level value, and something like this looks mildly bad:

# compress_gzip_opts="-6"
compress_xz_opts="--check=crc32 --lzma2=dict=1MiB"
# compress_xz_opts="--check=crc32 --lzma2=dict=1MiB -6"
compress_zstd_opts="-q -T0"
# compress_zstd_opts="-q -T0 -3"

EDIT: Or do you mean to leave the compression values only in the framework file?

EDIT2: Did that - won't set the defaults, but the commented out framework file has them as an example.

[scaronni]

EDIT: Or do you mean to leave the compression values only in the framework file?

EDIT2: Did that - won't set the defaults, but the commented out framework file has them as an example.

Yes, exactly. Thank you, I think it's fine.

[skwerlman]

This is probably a good place to add a note about the kernel not supporting zstd's --ultra flag or xz's crc64 csum mode: #386

[scaronni]

Merging for now, but the proper solution would be #319