[PATCH 4/5] ARM: dts: sun6i: Add sun6i-a31s.dtsi

Hi,

On Wed, Nov 26, 2014 at 5:05 PM, Hans de Goede [off-list ref] wrote:

Hi,

On 11/26/2014 12:45 AM, Chen-Yu Tsai wrote:

quoted

On Wed, Nov 26, 2014 at 2:18 AM, Maxime Ripard

<snip>

quoted

quoted

quoted

+             ahb1_gates: clk at 01c20060 {
+                     #clock-cells = <1>;
+                     compatible = "allwinner,sun6i-a31-ahb1-gates-clk";
+                     reg = <0x01c20060 0x8>;
+                     clocks = <&ahb1>;
+                     clock-output-names = "ahb1_mipidsi", "ahb1_ss",
+                                     "ahb1_dma", "ahb1_mmc0",
"ahb1_mmc1",
+                                     "ahb1_mmc2", "ahb1_mmc3",
"ahb1_nand1",
+                                     "ahb1_nand0", "ahb1_sdram",
+                                     "ahb1_gmac", "ahb1_ts",
"ahb1_hstimer",
+                                     "ahb1_spi0", "ahb1_spi1",
"ahb1_spi2",
+                                     "ahb1_spi3", "ahb1_otg",
"ahb1_ehci0",
+                                     "ahb1_ehci1", "ahb1_ohci0",
+                                     "ahb1_ohci1", "ahb1_ohci2",
"ahb1_ve",
+                                     "ahb1_lcd0", "ahb1_lcd1",
"ahb1_csi",
+                                     "ahb1_hdmi", "ahb1_de0",
"ahb1_de1",
+                                     "ahb1_fe0", "ahb1_fe1", "ahb1_mp",
+                                     "ahb1_gpu", "ahb1_deu0",
"ahb1_deu1",
+                                     "ahb1_drc0", "ahb1_drc1";
+             };

Are the gates really identical?

As in not even stripped down?

According to the user manuals, the AHB1 gates on the A31s don't have
GMAC (or EMAC in the manual) and MIPI DSI listed. Though it seems
a lot of A31s hardware out there actually use GMAC.

Hmm, interesting I was focussing on the ohci2 gate when comparing them as
the ohci2 is not listed in the pinmux.

The GMAC clock gate definitely actually is there, as I've an A31s based
board
with a working GMAC.

Also Allwinner has confirmed to me that the A31 and A31s are the same die,
so
there is no reason to differentiate between the 2 at this level. Either we
do
no differentiation at all, or only at the pinmux level to reflect that
certain
pins are simply just not there.

My vote goes to differentiating at the pinmux level.

I agree. Since it is confirmed the dies are the same, the hardware
description should be the same as well, except for the external
differences (pinmux). Only downside is a slightly bloated DTB
containing nodes we will never use on the a31s.

IIRC, Allwinner's kernel treats them as the same as well.


ChenYu