The 05/24/2021 10:46, Joel Stanley wrote:

On Mon, 24 May 2021 at 02:35, Steven Lee [off-list ref] wrote:

quoted

The 05/21/2021 09:25, Joel Stanley wrote:

quoted

Hi Steven,

On Thu, 20 May 2021 at 10:16, Steven Lee [off-list ref] wrote:

quoted

AST2600 A2(or newer) EVB has gpio regulators for toggling signal voltage
between 3.3v and 1.8v, the patch adds sdhci node and gpio regulator in the
new dts file and adds commment for describing the reference design.

spelling: comment

Thanks, will correct the typo.

quoted

I need you to justify the separate dts for the A2 EVB.

What would happen if this device tree was loaded on to an A1 or A0?

Since the clock default value(SCU210) of A1 and A0 are different to A2,
the following error would happen if A2 device tree was loaded on A1/A0.

[  133.179825] mmc1: Reset 0x4 never completed.
[  133.184599] mmc1: sdhci: ============ SDHCI REGISTER DUMP ===========
[  133.191786] mmc1: sdhci: Sys addr:  0x00000000 | Version:  0x00000002
[  133.198972] mmc1: sdhci: Blk size:  0x00007008 | Blk cnt:  0x00000001
[  133.206158] mmc1: sdhci: Argument:  0x00000c00 | Trn mode: 0x00000013
[  133.213343] mmc1: sdhci: Present:   0x01f70001 | Host ctl: 0x00000011
[  133.220528] mmc1: sdhci: Power:     0x0000000f | Blk gap:  0x00000000
[  133.227713] mmc1: sdhci: Wake-up:   0x00000000 | Clock:    0x00008007
[  133.234898] mmc1: sdhci: Timeout:   0x0000000b | Int stat: 0x00000000
[  133.242083] mmc1: sdhci: Int enab:  0x00ff0083 | Sig enab: 0x00ff0083
[  133.249268] mmc1: sdhci: ACmd stat: 0x00000000 | Slot int: 0x00000000
[  133.256453] mmc1: sdhci: Caps:      0x07f80080 | Caps_1:   0x00000007
[  133.263638] mmc1: sdhci: Cmd:       0x0000341a | Max curr: 0x001f0f08
[  133.270824] mmc1: sdhci: Resp[0]:   0x00000000 | Resp[1]:  0x01dd7f7f
[  133.278009] mmc1: sdhci: Resp[2]:   0x325b5900 | Resp[3]:  0x00400e00
[  133.285193] mmc1: sdhci: Host ctl2: 0x00000000
[  133.290148] mmc1: sdhci: ADMA Err:  0x00000000 | ADMA Ptr: 0xbe041200
[  133.297332] mmc1: sdhci: ============================================

Besides, A1/A0 EVBs don't have regulator, vmmc and vqmmc should be
removed from sdhci node of A1/A0 dts.

quoted

Would this device tree be used for the A3 (and any future revision?)

Yes, A3 can use the A2 dts.

quoted

An alternative proposal: we modify the ast2600-evb.dts to support the
A2 (which I assume would also support the A3).

If we need a separate board file for the A0 and A1 EVB, we add a new
one that supports these earlier revisions. Or we decide to only
support the latest revision in mainline.

In this patch, I add a new dts to support A2 sdhci, and include the
original dts since the other settings can be loaded on A2.
Do you mean creating a new file(e.g. aspeed-ast2600-evb-a1.dts) for A1,
and modifying the original aspeed-ast2600-evb.dts for supporting A2?

Yes, that would be my suggestion. The aspeed-ast2600-evb-a1.dts could
include the aspeed-ast2600-evb.dts.

Thanks for the prompt reply, I will add aspeed-ast2600-evb-a1.dts for
A1 and A0.

quoted

If we decide to only support the latest version in mainline, users
should mark vmmc and vqmmc as comment and modify clk-phase manually
for supporting A1.

If you believe there will be users of the A1 for some time, then I
think it makes sense to support both A1 and future boards in mainline.

Cheers,

Joel