On Mon, Oct 15, 2018 at 10:56 AM Rafał Miłecki <rafal@xxxxxxxxxx> wrote: > > Perhaps 'cru' in the compatible if that's what the h/w is called? > > > > Also, if this is a sub-block, then it should be a child of the block > > which should be defined here. > > It took me some time to do some extra research on the whole CRU thing. > > Valuable resources: > [1] bcm5301x_dmu.c (from the SDK) > [2] bcm5301x_pcie.c (from the SDK) > [3] https://patchwork.kernel.org/patch/7888651/ > [4] https://patchwork.kernel.org/patch/5051471/ > > First of all CRU seems to be a sub-block of the DMU (which stands for > "Device Management Unit" according to the [1]). DMU seems to be > independent block at 0x1800c000 with a size of 0x1000. > > It isn't actually clear what the CRU stands for. In Broadcom's case: > 1. According to the [3] it's "Clock and Reset Unit" > 2. According to the [4] it's a "central resource unit" > Other vendors seem to use CRU name for "Clock and Reset Unit". > > In any case, you're right Rob, it's a sub-block, a set of random > registers that control SoC. > > So I think that: > 1) We should have a node for DMU and CRU > 2) We should not include "cru" in bindings as pinmuxing seems to be part > of SoC that just happens to be controlled using CRU registers OK do you want me to revert the patches or do you want to fix this stuff on top of the patches that are already in the tree? I'm fine either way. Yours, Linus Walleij
