On Wed, Jan 05, 2011 at 05:08:20PM -0600, Timur Tabi wrote:
My concern is that I think it's unwise to hard-code the default values of the registers in the source file. Who's to say that a newer version of the chip won't have different power-on defaults?
In practical terms this is vanishingly unlikely;
I do want to support a register cache, but I don't want to hard-code the default values into cs4270.c. Is this supported?
If something you'd like to do doesn't seem to be working or is otherwise not there and you can see a way to implement it sensibly then the most obvious thing would seem to be to implement it.
Please remember that this isn't a frozen commercial OS release from a third party, you can contribute to the core code if you see a need to. None of this is fixed in stone and a lot of it comes from looking at the code and taking opportunities to add appropriate abstractions - look at symmetric_rates, or the creation of soc-cache for example.
As you will doubtless have seen when you looked at the code every other reference to reg_cache_default checks to see if it's set before using it. This does rather suggest that the intention of the code is that it be optional.
So are you saying that there's a bug in this patch? Perhaps that code should look like this:
Perhaps it should, though my immediate question (without looking at the existing code again) is why the existing tests are being removed.