Re: [PATCH v5 1/4] i2c: busses: i2c-st: Add ST I2C controller

From: srinivas kandagatla
Date: Fri Oct 18 2013 - 04:31:39 EST


On 17/10/13 15:49, Lucas Stach wrote:
> Am Donnerstag, den 17.10.2013, 15:30 +0100 schrieb srinivas kandagatla:
> [...]
>> Sorry to ask this but, Where is this requirement coming from?
>> I have not spotted any thing as such in ePAPR specs.
>>
>>
>> All the spec says is.
>> ===
>> The compatible property value consists of one or more strings that
>> define the specific programming model for the device. This list of
>> strings should be used by a client program for device driver selection.
>> The property value consists of a concatenated list of null terminated
>> strings, *from most specific to most general.* They allow a device to
>> express its compatibility with a family of similar devices, potentially
>> allowing a single device driver to match against several devices.
>> The recommended format is âmanufacturer,modelâ, where manufacturer is a
>> string describing the name of the manufacturer (such as an OUI), and
>> model specifies the model number.
>>
>> Example:
>> compatible = âfsl,mpc8641-uartâ, âns16550";
>> In this example, an operating system would first try to locate a device
>> driver that supported fsl,mpc8641-uart. If a driver was not found, it
>> would then try to locate a driver that supported the more general
>> ns16550 device type.
>> ===
>>
>> The more general compatible string in this case is "st,comms-ssc-i2c",
>> rather than the first soc name.
>> How can a first SOC name be more general?
>>
>> As this driver is not very specific to StiH415, it is generic driver for
>> ST comms-ssc-i2c block.
>>
>
> You just can't know if someone in the future decides to subtly change
> the register layout or make some other incompatible change to the
> comms-ssc-i2c block.
>
This is not the case for comms-ssc-i2c block, This IP is kind of reused
from past 10+ years(I think!!). Am not predicting the future here, but I
am making a informed guess from past experience that this IP would not
change in future.

Am still not happy with the idea of using first SoC for the compatible
for following reasons:

1> Generic IPs can be integrated into various vendor SoCs. For example
synopsis IP can be integrated by ST parts and other non-ST parts. What
would be the first SoC name in this case?

2> Looking at example like "arm,pl310-cache", "arm,l220-cache"... or any
other generic ips, why are these IPs not encoding the first SoC name in
there compatible string? I think the answer is generic IP.

3> IMHO, the idea of first SoC might solve the problem you described,
but why would some one know about the first SoC when this was available.
In this case this IP was available may be 10+ years back on an ST40
platform. Having such old SoC names in compatible strings in the device
trees for a modern chip looks bit confusing.

4> ST generic drivers which are in kernel still use st,<IP> name, so I
would like this consistency across all the st drivers (at least the ones
which are going to be used by mach-sti platforms).

5> ePAPR spec clearly says that compatible string should contain "most
specific to most general" names. In this case using more generic name
makes more sense than having a specific name because its generic IP.
Allowing a single device driver to match against several devices.

6> IMHO, the compatible string should be "vendor,<IP-name>-<IP-version>"
rather than first SoC.


Thanks,
srini

> So as a general rule of thumb you take the first SoC where a particular
> IP block appeared as the compatible string, so you can just pick the
> name of the SoC where the incompatible change was made as the next
> string and not need to invent generic and unspecific comms-ssc-i2c-v2
> compatibles.
>
> Regards,
> Lucas
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/