This is purely a guess, but I suspect that rather than a character-by-character numerical position change (like that provided by 'seek'), the OP is looking for a "line-by-line cache" - similar to something I did recently.

I'm going to assume that the file is too big to stuff into a scalar and process with something like /($start.*?$match.*?$end)/ - i.e., that it does indeed need to be read one line at a time.

#!/usr/bin/perl
use common::sense;

my $pat1 = "TOP_PATTERN";
my $pat2 = "SEARCH_PATTERN";

my ($flag, $cache);
while (<DATA>){
    if ($flag){
        last if /$pat1/;
        print and next;
    }

    if (/$pat1/){
        $cache = "" and next if $cache;
        $cache .= $_ and next;
    }
    if ($cache){
        $cache .= $_;
        $flag = 1 and print $cache if /$pat2/;
    }
}

__END__
TOP_PATTERN
1
2
3
TOP_PATTERN
4
5
6
SEARCH_PATTERN
7
TOP_PATTERN
8
9
[download]

Output:

TOP_PATTERN
4
5
6
SEARCH_PATTERN
7
[download]

-- 
I hate storms, but calms undermine my spirits.
 -- Bernard Moitessier, "The Long Way"