Thomas Gleixner debugged a particularly ugly seqlock related livelock:
do not process the seq-read section if we know it beforehand that the
test at the end of the section will fail ...

Signed-off-by: Ingo Molnar <mingo@elte.hu>