Bug 1794491

Please backport the fix from Fedora BZ 1775790:

For kernel builds, /usr/lib/rpm/kmod.prov is fork+execed by rpmbuild in "Processing files:" step about 8000 times, single-threaded, with cumulative run time of ~2 minutes.

Ideally rpmbuild should be changed to parallelize this, but it's not trivial to implement.

The fix in Fedora, meanwhile, speeds up the script, by avoiding additional fork+execing. The new kmod.prov is:

#!/bin/sh +x
# Kernel build can have many thousands of modules.
# kmod.prov is run for every one of them.
# Try to make this script run as fast as we can.
# For example, use shell string ops instead of external programs
# where possible.

IFS=$'\n'

read -r fname || exit

# Only process files from .../lib/modules/... subtree
[ "${fname#*/lib/modules/*}" != "$fname" ] || exit 0

kmod=${fname##*/}  # like basename, but faster

if [ "$kmod" = "modules.builtin" ]; then
        for j in $(cat -- "$fname"); do
                echo "kmod(${j##*/})"
        done
        exit 0
fi

kmod=${kmod%.gz}
kmod=${kmod%.xz}
if [ "${kmod%.ko}" != "$kmod" ]; then
        echo "kmod($kmod)"
fi