Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 13 Aug 2018 17:23:58 +0200
From:   =?UTF-8?B?TXlsw6huZQ==?= Josserand <mylene.josserand@bootlin.com>
To:     Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc:     robh+dt@kernel.org, mark.rutland@arm.com,
        mylene.josserand@free-electrons.com,
        thomas.petazzoni@free-electrons.com,
        maxime.ripard@free-electrons.com, linux-kernel@vger.kernel.org,
        devicetree@vger.kernel.org
Subject: Re: [PATCH v6 0/2] Input: Add Cypress Gen5 Touchscreen driver
Message-ID: <20180813172358.71912d55@dell-desktop.home>
In-Reply-To: <20180724174053.GA61165@dtor-ws>
References: <20180703094309.18514-1-mylene.josserand@bootlin.com>
 <20180704162158.euiw2v3hyv3hzofn@penguin>
 <20180724150046.7d70786c@dell-desktop.home>
 <20180724174053.GA61165@dtor-ws>
Organization: Bootlin
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Dmitry,

On Tue, 24 Jul 2018 10:40:53 -0700
Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:

> Hi Mylène,
> 
> On Tue, Jul 24, 2018 at 03:00:46PM +0200, Mylène Josserand wrote:
> > Hello Dmitry,
> > 
> > On Wed, 4 Jul 2018 16:21:58 +0000
> > Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:
> >   
> > > Hi Mylène,
> > > 
> > > On Tue, Jul 03, 2018 at 11:43:07AM +0200, Mylène Josserand wrote:  
> > > > Hello,
> > > > 
> > > > Here is a V6 series to add the driver of the touchscreen Cypress,
> > > > TrueTouch Generation 5.
> > > > Based on v4.18-rc3.
> > > > 
> > > > This patch series has already been posted in several iterations:
> > > >     - v1: Sent on 2017/05/29
> > > >     - v2: Sent on 2017/08/18
> > > >     - v3: Sent on 2017/09/27
> > > >     - v4: Sent on 2017/12/01
> > > >     - v5: Sent on 2017/12/20
> > > > 
> > > > I did not have any comments the last 4 versions.
> > > > And no reviews on my v5 during 6 months. Could I have any updates
> > > > or feedback on my series to know why it is not merged (to be able to
> > > > correct what is wrong)?    
> > > 
> > > Sorry, I must have missed the v5, sorry about that.
> > > 
> > > I probably asked this question before, but just to make sure - I see
> > > references to HID in the patch - the device is really not HID
> > > compatible? Is there any hope it could be made work with i2c-hid +
> > > hid-multitouch?
> > > 
> > > Thanks.
> > >   
> > 
> > I have checked and, for what I have seen, all the HID descriptor stuff
> > is HID compliant. We could definitely use i2c-hid and hid-multitouch
> > (there is the "hid-cypress" driver that exists also).
> > 
> > The only problem is that this touchscreen has two modes: a bootloader
> > mode and an application mode (which is the one where we can send
> > HID commands). After a power-on-reset, it is always in "bootloader"
> > mode so we need to send some commands (called "bootloader commands") to
> > switch to application mode.  
> 
> Is this a documented or observed behavior? In my practice devices (I am
> talking in general, not about Cypress) that have proper configuration
> loaded and that were brought up with appropriate power up sequence and
> timings automatically switch to application mode. They only end up in
> bootloader mode when proper power up sequence is not respected and they
> are unhappy.

I have checked and indeed, if everything is correctly performed, the
bootloader has a timeout to switch to application mode.
The datasheet says that this timeout can be configured and the "0" value
means that the bootloader will never automatically switch to application
unless a bootloader command is sent.

In our case, you were right, after a timeout, the touchscreen is
correctly switching to Application mode. Great news!

> 
> > These commands are not HID-compliant as the
> > datasheet indicates:
> > 
> > "Bootloader commands are not HID-over-I2C compliant."  
> 
> Any chance you could share the datasheet?

Sorry, it is not possible, the datasheet is under NDA :(

> 
> > 
> > I think that if the touchscreen would start directly in "application"
> > mode, we could directly use i2c-hid and hid-cypress drivers.
> > Unfortunately, this is not the case.
> > 
> > In bootloader mode, the ProductID is 0xc101 and in application mode, it
> > is 0xc001 (already available in hid-ids.h:
> > USB_DEVICE_ID_CYPRESS_TRUETOUCH but not handled)
> > 
> > What would be the better approach here?
> > Should I add a new product ID to detect the bootloader mode in
> > hid-cypress driver and send non-HID commands to switch to
> > "application" mode in this driver?
> > Anyway, I guess that I will drop this cyttsp5 driver and update the
> > existing one, right?  
> 
> So it still accessible through HID, even when in bootloader mode? OK,
> then I guess there are 2 options:
> 
> - if device is documented to always start in bootloader mode, you could
>   have a small stub driver that switches it into application mode in its
>   probe() code. The "bootloader" device will disappear and
>   "application" device will appear, and standard driver (hid-multitouch)
>   will bind to it.

Okay, I see. In our case, we do not have the timeout to 0 as after a
moment, the application mode is automatically switched.

> 
> - if device supposed to come up in application mode unless configuration
>   is damaged: I'd recommend doing what we do on Chrome OS and have some
>   userspace software that runs on boot (or whenever device is
>   discovered) and check if it has the latest firmware/configuration, and
>   repair device if needed.

I see.

I tried to make the i2d-hid & hid-cypress working. This is not
successful for the moment because I can not retrieve the correct
bcdVersion. While debugging, I have noticed that the HID descriptors
don't seem to be exactly the same:

under i2c-hid.c:

struct i2c_hid_desc {
        __le16 wHIDDescLength;
        __le16 bcdVersion;
        __le16 wReportDescLength;
        __le16 wReportDescRegister;
        __le16 wInputRegister;
        __le16 wMaxInputLength;
        __le16 wOutputRegister;
        __le16 wMaxOutputLength;
        __le16 wCommandRegister;
        __le16 wDataRegister;
        __le16 wVendorID;
        __le16 wProductID;
        __le16 wVersionID;
        __le32 reserved;
} __packed;

whereas in my driver, I have:

struct cyttsp5_hid_desc {
        __le16 hid_desc_len;
        u8 packet_id;       <-- Different
        u8 reserved_byte;   <-- Different
        __le16 bcd_version;
        __le16 report_desc_len;
        __le16 report_desc_register;
        __le16 input_register;
        __le16 max_input_len;
        __le16 output_register;
        __le16 max_output_len;
        __le16 command_register;
        __le16 data_register;
        __le16 vendor_id;
        __le16 product_id;
        __le16 version_id;
        u8 reserved[4];
} __packed;

Have you already seen devices that are "HID compatible" with a different
HID descriptor's content like this?

Thank you in advance,

Best regards,

-- 
Mylène Josserand, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com