dkms: use saner compression defaults

Currently our compression options vary from those used in upstream
kernel. As of 6.8-rc2 upstream uses (ignoring force/rm input file etc):

 - gzip -9
 - xz --check=crc32 --lzma2=dict=1MiB
   Note: some older kernels lacked crc32 and/or used 2MiB
 - zstd -T0 (aka -3)

With the higher dictionary size (xz) and compression size (20+ zstd)
more memory is required for decompression. Which may be a problem if the
kernel itself is responsible for the decompression (instead of kmod),
since it uses a smaller/limited ram amount.

Reduce the numbers of be compatible with upstream.

Closes: #386

Signed-off-by: Emil Velikov <[email protected]>

Author: evelikov-work

dkms.in

@@ -1154,9 +1154,9 @@ actual_build()
         if [[ $module_compressed_suffix = .gz ]]; then
             gzip -9f "$built_module" || compressed_module=""
         elif [[ $module_compressed_suffix = .xz ]]; then
-            xz -f "$built_module" || compressed_module=""
+            xz --check=crc32 --lzma2=dict=1MiB -f "$built_module" || compressed_module=""
         elif [[ $module_compressed_suffix = .zst ]]; then
-            zstd -q -f -T0 -20 --ultra "$built_module" || compressed_module=""
+            zstd -q -f -T0 -19 "$built_module" || compressed_module=""
         fi
         if [[ -n $compressed_module ]]; then
             cp -f "$compressed_module" "$base_dir/module/${dest_module_name[$count]}$module_suffix" >/dev/null