lkml.org 
[lkml]   [2019]   [Dec]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    From
    Date
    SubjectRe: [PATCH v3 4/7] dt-bindings: gpio: Add gpio-repeater bindings
    Hi Rob,

    On Thu, Dec 5, 2019 at 10:06 PM Rob Herring <robh@kernel.org> wrote:
    > On Wed, Nov 27, 2019 at 09:42:50AM +0100, Geert Uytterhoeven wrote:
    > > Add Device Tree bindings for a GPIO repeater, with optional translation
    > > of physical signal properties. This is useful for describing explicitly
    > > the presence of e.g. an inverter on a GPIO line, and was inspired by the
    > > non-YAML gpio-inverter bindings by Harish Jenny K N
    > > <harish_kandiga@mentor.com>[1].
    > >
    > > Note that this is different from a GPIO Nexus Node[2], which cannot do
    > > physical signal property translation.
    >
    > It can't? Why not? The point of the passthru mask is to not do
    > translation of flags, but without it you are always doing translation of
    > cells.

    Thanks for pushing me deeper into nexuses!
    You're right, you can map from one type to another.
    However, you cannot handle the "double inversion" of an ACTIVE_LOW
    signal with a physical inverter added:

    nexus: led-nexus {
    #gpio-cells = <2>;
    gpio-map = <0 0 &gpio2 19 GPIO_ACTIVE_LOW>, // inverted
    <1 0 &gpio2 20 GPIO_ACTIVE_HIGH>, // noninverted
    <2 0 &gpio2 21 GPIO_ACTIVE_LOW>; // inverted
    gpio-map-mask = <3 0>;
    // default gpio-map-pass-thru = <0 0>;
    };

    leds {
    compatible = "gpio-leds";
    led6-inverted {
    gpios = <&nexus 0 GPIO_ACTIVE_HIGH>;
    };
    led7-noninverted {
    gpios = <&nexus 1 GPIO_ACTIVE_HIGH>;
    };
    led8-double-inverted { // FAILS: still inverted
    gpios = <&nexus 2 GPIO_ACTIVE_LOW>;
    };
    };

    It "works" if the last entry in gpio-map is changed to GPIO_ACTIVE_HIGH.
    Still, the consumer would see the final translated polarity, and not the
    actual one it needs to program the consumer for.

    > > While an inverter can be described implicitly by exchanging the
    > > GPIO_ACTIVE_HIGH and GPIO_ACTIVE_LOW flags, this has its limitations.
    > > Each GPIO line has only a single GPIO_ACTIVE_* flag, but applies to both
    > > th provider and consumer sides:
    > > 1. The GPIO provider (controller) looks at the flags to know the
    > > polarity, so it can translate between logical (active/not active)
    > > and physical (high/low) signal levels.
    > > 2. While the signal polarity is usually fixed on the GPIO consumer
    > > side (e.g. an LED is tied to either the supply voltage or GND),
    > > it may be configurable on some devices, and both sides need to
    > > agree. Hence the GPIO_ACTIVE_* flag as seen by the consumer must
    > > match the actual polarity.
    > > There exists a similar issue with interrupt flags, where both the
    > > interrupt controller and the device generating the interrupt need
    > > to agree, which breaks in the presence of a physical inverter not
    > > described in DT (see e.g. [3]).
    >
    > Adding an inverted flag as I've suggested would also solve this issue.

    As per your suggestion in "Re: [PATCH V4 2/2] gpio: inverter: document
    the inverter bindings"?
    https://lore.kernel.org/linux-devicetree/CAL_JsqLp___2O-naU+2PPQy0QmJX6+aN3hByz-OB9+qFvWgN9Q@mail.gmail.com/

    Oh, now I understand. I was misguided by Harish' interpretation
    https://lore.kernel.org/linux-devicetree/dde73334-a26d-b53f-6b97-4101c1cdc185@mentor.com/
    which assumed an "inverted" property, e.g.

    inverted = /bits/ 8 <0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0>;

    But you actually meant a new GPIO_INVERTED flag, to be ORed into the 2nd
    cell of a GPIO specifier? I.e. add to include/dt-bindings/gpio/gpio.h"

    /* Bit 6 expresses the presence of a physical inverter */
    #define GPIO_INVERTED 64

    We need to be very careful in defining to which side the GPIO_ACTIVE_*
    applies to (consumer?), and which side the GPIO_INVERTED flag (provider?).
    Still, this doesn't help if e.g. a FET is used instead of a push-pull
    inverter, as the former needs translation of other flags (which the
    nexus can do, the caveats above still applies, though).

    Same for adding IRQ_TYPE_INVERTED.

    Related issue: how to handle physical inverters on SPI chip select lines,
    if the SPI slave can be configured for both polarities?

    Gr{oetje,eeting}s,

    Geert


    --
    Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

    In personal conversations with technical people, I call myself a hacker. But
    when I'm talking to journalists I just say "programmer" or something like that.
    -- Linus Torvalds

    \
     
     \ /
      Last update: 2019-12-06 10:18    [W:3.331 / U:0.736 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site