Ahh, the power of regular expressions:

use strict;
use warnings;

my $data;
{
    local $/;
    $data = <DATA>;
}

my %drives = ();
while ($data =~ /^(\w+)\s*.*?Serial Number\.*(\w*)\s/smg) {
    $drives{$1} = $2;
}

while ( my ($key, $value) = each %drives) {
    print "$key  $value\n";
}

__DATA__
hdisk0   P1/Z1-Aa     16 Bit LVD SCSI Disk Drive (18200 MB)

        Manufacturer................IBM
        Machine Type and Model......ST318305LC
        FRU Number..................09P4437
        Serial Number...............0009E05B
        Part Number.................09P4436

hdisk1  P1/Z1-Ab     16 Bit LVD SCSI Disk Drive (18200 MB)

        Manufacturer................IBM
        Machine Type and Model......ST318305LC
        FRU Number..................09P4437
        Serial Number...............0004A0D2
        Part Number.................09P4436
[download]

For information on these marvels of modern science, read perlretut. What I've done in the above is:

1. ^(\w+) - Find a line that starts with alphanumeric characters
2. \s*.*? - Advance past the smallest amount of arbitrary text until:
3. Serial Number - I encounter the phrase 'Serial Number'
4. \.* - Advance past an arbitrary number of '.' characters
5. (\w*)\s - Find a series of alphanumerics followed by whitespace

The parenthesis capture the results into the variables $1 and $2, which are stored in the hash. The modifiers s and m control how new-lines are treated for matching, and the g modifier makes the while loop repeat so long as a match exists. Also note I localized the default record separator $/, so the entire data section is read in as a single string.