On Tue, Apr 25, 2017 at 1:56 AM, Hauke Mehrtens <hauke@xxxxxxxxxx> wrote: > > > On 04/20/2017 04:48 PM, Rob Herring wrote: >> On Mon, Apr 17, 2017 at 09:29:35PM +0200, Hauke Mehrtens wrote: >>> From: Martin Blumenstingl <martin.blumenstingl@xxxxxxxxxxxxxx> >>> >>> This allows using the xbar driver on ARX300 based SoCs which require the >>> same xbar setup as the xRX200 chipsets because the xbar driver >>> initialization is not guarded by an xRX200 specific >>> of_machine_is_compatible condition anymore. Additionally the new driver >>> takes a syscon phandle to configure the XBAR endianness bits in RCU >>> (before this was done in arch/mips/lantiq/xway/reset.c and also >>> guarded by an xRX200 specific if-statement). >>> >>> Signed-off-by: Hauke Mehrtens <hauke@xxxxxxxxxx> >>> --- >>> .../devicetree/bindings/mips/lantiq/xbar.txt | 22 +++++ >>> MAINTAINERS | 1 + >>> arch/mips/lantiq/xway/reset.c | 4 - >>> arch/mips/lantiq/xway/sysctrl.c | 41 --------- >>> drivers/soc/Makefile | 1 + >>> drivers/soc/lantiq/Makefile | 1 + >>> drivers/soc/lantiq/xbar.c | 100 +++++++++++++++++++++ >>> 7 files changed, 125 insertions(+), 45 deletions(-) >>> create mode 100644 Documentation/devicetree/bindings/mips/lantiq/xbar.txt >>> create mode 100644 drivers/soc/lantiq/Makefile >>> create mode 100644 drivers/soc/lantiq/xbar.c >>> >>> diff --git a/Documentation/devicetree/bindings/mips/lantiq/xbar.txt b/Documentation/devicetree/bindings/mips/lantiq/xbar.txt >>> new file mode 100644 >>> index 000000000000..86e53ff3b0d5 >>> --- /dev/null >>> +++ b/Documentation/devicetree/bindings/mips/lantiq/xbar.txt >>> @@ -0,0 +1,22 @@ >>> +Lantiq XWAY SoC XBAR binding >>> +============================ >>> + >>> + >>> +------------------------------------------------------------------------------- >>> +Required properties: >>> +- compatible : Should be "lantiq,xbar-xway" >> >> This compatible is already in use so it is fine, but you should also >> have per SoC compatible strings. > > I will add a new SoC specific one. > What does per SoC device tree mean? Does it mean for the same silicon, > for the same silicon revision, for the same fusing of a silicon or for > the same marketing name? Depends how specific you need to some extent. For fusing, packaging, metal fixes, speed grading, etc. probably use the same compatible. Different design and dies from the start, then they should have different compatibles. > I would like to make it per silicon or per silicon revision for the IP > cores which I know are different. Being "the same IP" doesn't really matter. The errata can be different and often there is no visibility into what h/w designers may have changed. On some IP you can rely on revision/feature registers though forgetting to rev revision registers is not uncommon. You need to have sufficient information that you can work-around a problem in the future without requiring a new dtb. Rob
