Re: [PATCH v4 6/7] dt-bindings: phy-qcom-usb2: Add support to override tuning values

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On Thu, Mar 29, 2018 at 4:04 AM, Manu Gautam <mgautam@xxxxxxxxxxxxxx> wrote:
> To improve eye diagram for PHYs on different boards of same SOC,
> some parameters may need to be changed. Provide device tree
> properties to override these from board specific device tree
> files. While at it, replace "qcom,qusb2-v2-phy" with compatible
> string for USB2 PHY on sdm845 which was earlier added for
> sdm845 only.
>
> Signed-off-by: Manu Gautam <mgautam@xxxxxxxxxxxxxx>
> ---
>  .../devicetree/bindings/phy/qcom-qusb2-phy.txt        | 19 ++++++++++++++++++-
>  1 file changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/Documentation/devicetree/bindings/phy/qcom-qusb2-phy.txt b/Documentation/devicetree/bindings/phy/qcom-qusb2-phy.txt
> index 42c9742..0ed140a 100644
> --- a/Documentation/devicetree/bindings/phy/qcom-qusb2-phy.txt
> +++ b/Documentation/devicetree/bindings/phy/qcom-qusb2-phy.txt
> @@ -6,7 +6,7 @@ QUSB2 controller supports LS/FS/HS usb connectivity on Qualcomm chipsets.
>  Required properties:
>   - compatible: compatible list, contains
>                "qcom,msm8996-qusb2-phy" for 14nm PHY on msm8996,
> -              "qcom,qusb2-v2-phy" for QUSB2 V2 PHY.
> +              "qcom,sdm845-qusb2-phy" for 10nm PHY on sdm845.
>
>   - reg: offset and length of the PHY register set.
>   - #phy-cells: must be 0.
> @@ -27,6 +27,23 @@ Optional properties:
>                 tuning parameter value for qusb2 phy.
>
>   - qcom,tcsr-syscon: Phandle to TCSR syscon register region.

Just to confirm: the new properties below work just fine on the old
msm8996 PHY too, right?


> + - qcom,imp-res-offset-value: It is a 6 bit value that specifies offset to be
> +               added to PHY refgen RESCODE via IMP_CTRL1 register. It is a PHY
> +               tuning paramter that may vary for different boards of same SOC.
> + - qcom,hstx-trim-value: It is a 4 bit value that specifies tuning for HSTX
> +               output current. 0x0 value corresponding to 24mA which is maximum
> +               current and 0xf corresponds to lowest current which is 15mA.
> + - qcom,preemphasis-level: It is a 2 bit value that specifies pre-emphasis level.
> +               Possible values are:
> +               00: NONE
> +               01: +5%
> +               10: +10%
> +               11: +15%

The user of the device tree will expect to specify this in decimal,
right?  So list the above as 0, 1, 2, 3.  ...not 00, 01, 10, 11.
(though below I suggest that specifying 0, 1, 2, 3 is probably not
quite the right way to describe this property).


> +- qcom,preemphasis-width: It is a 1 bit value that specifies how long the HSTX
> +               pre-emphasis (specified using qcom,preemphasis-level) must be in
> +               effect. Possible values are:
> +               0: Full-bit width
> +               1: Half-bit width

Perhaps just make this a boolean property.  If it exists then you get
the non-default case.  AKA: if the default is full bit width, then
you'd allow a boolean property "qcom,preemphasis-half-width" to
override.  If the default is half bit width then you'd allow
"qcom,preemphasis-full-width" to override.


For all the above optional bits, please indicate what the default is
if they aren't specified.  If you have to specify a different default
for sdm845 vs. msm8996 then so be it (though it would be nice to avoid
it by changing the default for sdm845 unless that's totally crazy).


Overall this looks pretty good to me.  Certainly the descriptions are
very understandable and this will make tuning on a per-board /
per-port basis much easier!  Thanks!


One last overall comment is that ideally we could make the device tree
a little bit more human readable.  Right now we'll have a set of
properties that looks like this (numbers made up):

qcom,imp-res-offset-value = <0x13>;
qcom,hstx-trim-value = <0x7>;
qcom,preemphasis-level = <1>;
qcom,preemphasis-width = <0>;


One option to make that more readable would be to change the units, AKA:

qcom,hstx-trim-ma = <18>;
qcom,preemphasis-percent = <5>;

...then the code would translate from these human-readable values to
the real numbers.


Another option would be to add a #include to the bindings.  I'd defer
to the wisdom of the bindings guys about if this is better or worse
than adding the units, but personally I like it better because:
* You get a compile-time error if you use an unsupported value.
* You don't need to add the code to translate.

So you'd add a ".h" file as part of this bindings patch (see git
history for other examples) and you'd have stuff like:

#define SDM845_QUSB2_TRIM_24MA 0
...
#define SDM845_QUSB2_TRIM_15MA 0xf

#define SDM845_QUSB2_PREEMPHASIS_NONE 0
#define SDM845_QUSB2_PREEMPHASIS_5_PERCENT 1
#define SDM845_QUSB2_PREEMPHASIS_10_PERCENT 2
#define SDM845_QUSB2_PREEMPHASIS_15_PERCENT 3

...then you'd use those #defines in the description of the properties.

--

One last note: from the current code
<https://patchwork.kernel.org/patch/10314925/> all of these new
properties need to be specified as 8-bit numbers, not 32-bit.  AKA:

qcom,imp-res-offset-value = /bits/ 8 <0x13>;
qcom,hstx-trim-value = /bits/ 8 <0x7>;
qcom,preemphasis-level = /bits/ 8 <1>;
qcom,preemphasis-width = /bits/ 8 <0>;


If we decide to keep it that way then it needs to be documented in the
bindings file.  IMHO we should just specify that these are normal
(32-bit) properties even if we expect that they'll always be 8-bits.
That seems to be commonplace elsewhere.



-Doug
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux