On Fri, May 01, 2009 at 12:20:51AM +0400, Anton Vorontsov wrote:
Sometimes errors happen even during copy-paste procedure. Or for example HW designers have made some minor (they thought) change, and didn't bother to bump the revision. Later we observe that the change is the cause of some errata, but we can't deal with it because device-tree isn't specific enough.
Of course this is just my imagination, but theoretically possible?
It's possible -- but not likely enough to warrant using the chip name if a usable block number is available (it's more likely that a future rev of the same chip number will have a different logic block version). PVR or information elsewhere in the device tree could be used in such cases.
Note that it can even happen at board level. Accessing power management controller registers (which are SoC-internal) on early revs of the 8313erdb would lock the board (unless JTAG is plugged in or a resistor modification is made), at no fault of the 8313e chip itself. But we don't put the board name in all of the compatibles...
-Scott