Comment 40 for bug 585092

Glad the Maverick kernel helps in your case. Though this is not the same problems as the initially reported one which was a long wait time on unmount. AFAIK the writeback code in 2.6.32 had another issue where an active writer might starve other IO under certain conditions. But this should probably get tracked in a separate bug. And I guess this bug should get renamed to address the issue more specifically.