Message-ID: <56F6EFFD.9070007@denx.de>
Date: Sat, 26 Mar 2016 21:24:29 +0100
From: Marek Vasut <marex@denx.de>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Icedove/31.7.0
MIME-Version: 1.0
To: Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>,
        "devicetree@vger.kernel.org" <devicetree@vger.kernel.org>
CC: linux-kernel@vger.kernel.org, Rob Herring <robh+dt@kernel.org>,
        Alexey Brodkin <abrodkin@synopsys.com>
Subject: Re: ARC dw-mshc binding compat string
References: <56F66121.8050507@denx.de> <56F6C639.5000301@mentor.com> <56F6C71B.6070002@denx.de> <56F6CC68.5040408@mentor.com> <56F6D083.1020402@denx.de> <56F6D1E9.3050606@mentor.com> <56F6E860.8070207@denx.de> <56F6ED41.5020908@mentor.com>
In-Reply-To: <56F6ED41.5020908@mentor.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 4058
Lines: 98

On 03/26/2016 09:12 PM, Vladimir Zapolskiy wrote:
> On 26.03.2016 21:52, Marek Vasut wrote:
>> On 03/26/2016 07:16 PM, Vladimir Zapolskiy wrote:
>>> On 26.03.2016 20:10, Marek Vasut wrote:
>>>> On 03/26/2016 06:52 PM, Vladimir Zapolskiy wrote:
>>>>> Hi Marek,
>>>>>
>>>>> On 26.03.2016 19:30, Marek Vasut wrote:
>>>>>> On 03/26/2016 06:26 PM, Vladimir Zapolskiy wrote:
>>>>>>> On 26.03.2016 12:14, Marek Vasut wrote:
>>>>>>>> Hi!
>>>>>>>>
>>>>>>>> I noticed that arch/arc/boot/dts/axs10x_mb.dtsi uses "altr," prefix in
>>>>>>>> the DT compatible string:
>>>>>>>>
>>>>>>>> mmc@0x15000 {
>>>>>>>>         compatible = "altr,socfpga-dw-mshc";
>>>>>>>>         reg = < 0x15000 0x400 >;
>>>>>>>>         num-slots = < 1 >;
>>>>>>>>         fifo-depth = < 16 >;
>>>>>>>>         card-detect-delay = < 200 >;
>>>>>>>>         clocks = <&apbclk>, <&mmcclk>;
>>>>>>>>         clock-names = "biu", "ciu";
>>>>>>>>         interrupts = < 7 >;
>>>>>>>>         bus-width = < 4 >;
>>>>>>>> };
>>>>>>>>
>>>>>>>> I don't think this is OK, since ARC is unrelated to Altera, which is
>>>>>>>> what the "altr," prefix stands for. I think the socfpga-dw-mshc shim
>>>>>>>> should be extended with another compatibility string, something like
>>>>>>>> "snps,arc-dw-mshc" and the axs10x_mb.dtsi should be adjusted
>>>>>>>> accordingly. What do you think ?
>>>>>>>>
>>>>>>>
>>>>>>> There is "snps,dw-mshc" described in
>>>>>>> Documentation/devicetree/bindings/mmc/synopsys-dw-mshc.txt and supported by
>>>>>>> dw_mmc host controller driver.
>>>>>>
>>>>>> Thanks, that's even better.
>>>>>>
>>>>>> btw what do you think of using altr, prefix on non-altera system, that
>>>>>> doesn't seem ok, right ?
>>>>>
>>>>> according to ePAPR the prefix should represent a device (IP block here
>>>>> I believe) manufacturer, so it should be okay to use "altr" prefix on
>>>>> non-Altera system, if Altera provides  another hardware vendor with
>>>>> some own IP block.
>>>>
>>>> In this case, it's Synopsys who provides the SD/MMC/MS core to other
>>>> chip makers (Altera etc).
>>>
>>> Correct.
>>>
>>>>> That said, I would rather prefer to see "snps,dw-mshc" prefix on description
>>>>> of an MMC controller found on SoCFPGA series, "altr,socfpga-dw-mshc" seems
>>>>> to be redundant.
>>>>
>>>> According to drivers/mmc/host/dw_mmc-pltfm.c , the Altera SoCFPGA one
>>>> "altr,socfpga-dw-mshc" and also Imagination Technology Pistacio one
>>>> "img,pistachio-dw-mshc" need specialty bit (SDMMC_CMD_USE_HOLD_REG),
>>>> while the stock one "snps,dw-mshc" does not. I am not sure if the ARC
>>>> one needs it as well, but most likely yes.
>>>>
>>>> I wonder if that bit is needed on some particular version of the DWMMC
>>>> core. In that case, should we have "snps,dw-mshc" and "snps,dw-mshc-vN"
>>>> binding ? Or should we use DT property to discern the need for this bit ?
>>>>
>>>
>>> That's the most common way to take into account peculiarities, add
>>> a property and handle it from the driver.
>>
>> And by "that" you mean which of those two I listed , the
>> "snps,dw-mshc-vN" or adding new DT prop ?
>>
> 
> I meant to add a new property, not a new compatible, but that's just
> my experience.
> 
> Let me say it __might__ happen that a particular change you need is
> specific to a particular version of the DWMMC IP (query Synopsys
> by the way), but more probably it might be e.g. the same IP version with
> a different reduced or extended configuration or a minor fix/improvement
> to the IP block without resulting version number bump.
> 
> For example I don't remember that errata fixes in IP blocks result in
> a new compatible, instead there are quite common optional "quirk"
> properties for broken IPs -- e.g. check bindings/usb/dwc3.txt :)

Right, this very much matches how I see it as well. Thanks for confirming.

Alexey, can you tell us if the requirement for setting
SDMMC_CMD_USE_HOLD_REG came with some new revision of the core or
disappeared with some revision OR if this is some configuration
option of the core during synthesis ?

-- 
Best regards,
Marek Vasut