G'day Genmai,

Welcome to the Monastery.

Your test data is poor:

• There's no way to differentiate between all the identical keys.
• There's no way to differentiate between all the identical values.
• I question whether it is representative of your real data.

I also note that, after you previewed posts and saw that they didn't match your real code, you went ahead and posted them anyway. You'll find that monks are not particularly inclined to spend much time responding when the poster can't be bothered to make much effort. Please aim to do better in future: we don't like looking at crap and you're the one who suffers.

Here's the guts of what you (probably) need:

#!/usr/bin/env perl

use strict;
use warnings;

use Data::Dump;

my (%data, $topkey);

while (<DATA>) {
    chomp;
    $data{$topkey}{$1} = $2 and next if /^(.*?)=(.*?)$/;
    $topkey = $_; 
}

dd \%data;

__DATA__
Paragraph1
key1=valueA
Paragraph2
key2=valueB
key3=valueC
Paragraph3
[download]

Output:

{
  Paragraph1 => { key1 => "valueA" },
  Paragraph2 => { key2 => "valueB", key3 => "valueC" },
}
[download]

Update: Fixed typo: s/can be bothered/can't be bothered/. Many thanks, hippo, for catching this.