Message-ID: <20230928-spectacle-civic-339c0d71d8d7@spud>
Date:   Thu, 28 Sep 2023 17:56:25 +0100
From:   Conor Dooley <conor@...nel.org>
To:     yang tylor <tylor_yang@...ax.corp-partner.google.com>
Cc:     Conor Dooley <conor.dooley@...rochip.com>,
        dmitry.torokhov@...il.com, robh+dt@...nel.org,
        krzysztof.kozlowski+dt@...aro.org, conor+dt@...nel.org,
        jikos@...nel.org, benjamin.tissoires@...hat.com,
        linux-input@...r.kernel.org, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        poyuan_chang@...ax.corp-partner.google.com, hbarnor@...omium.org,
        "jingyliang@...omium.org" <jingyliang@...omium.org>,
        wuxy23@...ovo.com, luolm1@...ovo.com,
        hung poyu <poyu_hung@...ax.corp-partner.google.com>
Subject: Re: [PATCH V2 1/2] dt-bindings: input: Introduce Himax HID-over-SPI
 device

On Thu, Sep 28, 2023 at 10:12:41AM +0800, yang tylor wrote:
> On Tue, Sep 26, 2023 at 8:53 PM Conor Dooley <conor@...nel.org> wrote:
> > On Tue, Sep 26, 2023 at 05:52:39PM +0800, yang tylor wrote:
> > > On Tue, Sep 26, 2023 at 5:02 PM Conor Dooley <conor@...nel.org> wrote:
> > > > On Mon, Sep 25, 2023 at 06:16:29PM +0800, yang tylor wrote:
> > > > > On Mon, Sep 25, 2023 at 4:41 PM Conor Dooley <conor.dooley@...rochip.com> wrote:
> > > > > > On Mon, Sep 25, 2023 at 09:44:21AM +0800, yang tylor wrote:
> > > > > > > On Fri, Sep 22, 2023 at 11:31 PM Conor Dooley <conor@...nel.org> wrote:
> > > > > > > >
> > > > > > > > On Fri, Sep 22, 2023 at 05:43:54PM +0800, yang tylor wrote:
> > > > > > > > > On Fri, Sep 22, 2023 at 5:22 PM Conor Dooley <conor@...nel.org> wrote:
> > > > > > > > > >
> > > > > > > > > > On Fri, Sep 22, 2023 at 03:56:25PM +0800, yang tylor wrote:
> > > > > > > > > > > On Tue, Sep 19, 2023 at 7:09 PM Conor Dooley <conor@...nel.org> wrote:
> > > > > > > > > > > > On Tue, Sep 19, 2023 at 05:31:29PM +0800, yang tylor wrote:
> > > > > > > > > >
> > > > > > > > > > > > > The behavior of "himax,boot_time_fw_upgrade" seems not stable and
> > > > > > > > > > > > > should be removed. "himax,fw_in_flash", I use the kernel config for
> > > > > > > > > > > > > user to select.
> > > > > > > > > > > >
> > > > > > > > > > > > That seems like a bad idea, we want to be able to build one kernel that
> > > > > > > > > > > > works for all hardware at the same time.
> > > > > > > > > > > >
> > > > > > > > > > > I see, so I should take that back?
> > > > > > > > > > > I'll explain more about it.
> > > > > > > > > >
> > > > > > > > > > Are there particular ICs where the firmware would always be in flash and
> > > > > > > > > > others where it would never be? Or is this a choice made by the board or
> > > > > > > > > > system designer?
> > > > > > > > > >
> > > > > > > > > Most cases it's about the system designer's decision. But some ICs may be forced
> > > > > > > > > to use flash because of its architecture(multiple IC inside, need to
> > > > > > > > > load firmware to
> > > > > > > > > multiple IC's sram by master IC). But if there is no limitation on
> > > > > > > > > this part, most system
> > > > > > > > > designers will prefer flashless.
> > > > > > > >
> > > > > > > > Forgive me if I am not understanding correctly, there are some ICs that
> > > > > > > > will need to load the firmware from flash and there are some where it
> > > > > > > > will be a decision made by the designer of the board. Is the flash part
> > > > > > > > of the IC or is it an external flash chip?
> > > > > > > >
> > > > > > >
> > > > > > > Both are possible, it depends on the IC type. For TDDI, the IC is long
> > > > > > > and thin, placed on panel PCB, flash will be located at the external
> > > > > > > flash chip. For the OLED TP, IC is usually placed at FPC and its flash
> > > > > > > is embedded, thus the IC size is large compared to TDDI. But from the
> > > > > > > driver's perspective either external flash or embedded flash, the IC
> > > > > > > itself will load firmware from flash automatically when reset pin is
> > > > > > > released. Only if firmware is loading from the host storage system,
> > > > > > > the driver needs to operate the IC in detail.
> > > > > >
> > > > > >
> > > > > > Since there are ICs that can use the external flash or have it loaded
> > > > > > from the host, it sounds like you do need a property to differentiate
> > > > > > between those cases.
> > > > > Yep.
> > > > >
> > > > > > Is it sufficient to just set the firmware-name property for these cases?
> > > > > > If the property exists, then you know you need to load firmware & what
> > > > > > its name is. If it doesn't, then the firmware either isn't needed or
> > > > > > will be automatically loaded from the external flash.
> > > >
> > > > > We have a default prefix firmware name(like himax_xxxx.bin) in the driver code.
> > > >
> > > > How do you intend generating the name of the firmware file? I assume the
> > > > same firmware doesn't work on every IC, so you'll need to pick a
> > > > different one depending on the compatible?
> > > >
> > > If considering a firmware library line-up for all the incoming panels
> > > of this driver.
> > > We would use PID as part of the file name. Because all the support panels would
> > > have a unique PID associated. Which will make the firmware name like
> > > himax_xxx_{$PID}.bin. The problem is, we need to know PID before firmware load
> > > at no flash condition. Thus PID information is required in dts when
> > > no-flash-flag
> > > is specified.
> >
> > Firstly, where does the "xxx" come from?
> > And you're making it sound more like having firmware-name is suitable
> > for this use case, given you need to determine the name of the file to
> > use based on something that is hardware specific but is not
> > dynamically detectable.
> Current driver patch uses a prefix name "himax_i2chid" which comes
> from the previous project
>  and seems not suitable for this condition, so I use "xxx" and plan to
> replace it in the next version.
> For finding firmware, I think both solutions are reasonable.
> - provide firmware name directly: implies no-flash and use user
> specified firmware, no PID info.
> - provide no-flash-flag and PID info: loading firmware from organized
> names with PID info.
> I prefer the 2nd solution, but it needs more properties in dts. 1st
> has less properties and more
> intuitive.
> 
> I don't know which one is more acceptable by the community, as you
> know I'm a newbie here.

To be honest, I am not all that sure either! Does the panel id have
value in its own right, or is that only used to determine the firmware
filename?
Also, if it does have value in its own right, rather than a "pid",
should the panel be a child node of this hid device with its own
compatible?

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)

Powered by blists - more mailing lists