RE: TG3 fix for slow switches (Was: TG3 driver failure on HP 16-way)

Would this condition be possible if something, such as ASF, were
continually polling the PHY in parallel with the driver?

Without proper handshake with ASF, I think it may be possible if ASF is in a very tight loop polling the PHY. The tg3_readphy() poll loop is not very tight so it is possible to continually hit the busy condition if ASF is polling PHY registers. If this is the case, even if tg3_readphy() eventually gets the data, the data will most likely be from a different PHY register (that ASF is trying to read).
 
But I don't think this is an ASF related problem because if it were, the patch would not have fixed it.
 
Michael