Is it too strict for VIPT cache flushing in flush_cache_[mm|range]?

2012/10/16 Ning Jiang [off-list ref]:

I found that the implementation of flush_cache_mm() and
flush_cache_range() for VIPT aliasing case is to clean + invalidate
whole data cache, but for a physically tagged cache, why do we need to
perform such an operation? Is it because we'll have aliasing issue?

I searched through the git history and found it is brought in the
commit "d7b6b358"  "[ARM] Fix ARMv6 VIPT cache >= 32K", before that,
there is indeed no flushing, why we "Fix ARMv6 VIPT cache >= 32K"?

Sorry for asking such stupid question, but it really confuses me ;-)

Thanks,
Ning

Maybe someone in the mail list can give an answer ;-)