Date:   Fri, 3 Feb 2023 16:59:26 -0800
From:   Dmitry Torokhov <dmitry.torokhov@gmail.com>
To:     Richard Kjerstadius <kjerstadius@gmail.com>
Cc:     linux-input@vger.kernel.org,
        Richard Kjerstadius <richard.kjerstadius@teledyne.com>,
        linus.walleij@linaro.org, alistair@alistair23.me,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] input: cyttsp5: Fix bitmask for touch buttons
Message-ID: <Y92t7uqj/F8kFbi8@google.com>
References: <20230127102903.3317089-1-richard.kjerstadius@teledyne.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230127102903.3317089-1-richard.kjerstadius@teledyne.com>
Precedence: bulk

On Fri, Jan 27, 2023 at 11:29:03AM +0100, Richard Kjerstadius wrote:
> Prior to this patch, the bitmask ends up being 0x3, as opposed to 0x1
> which likely was the intention. The erroneous bit results in the driver
> reporting 2 different button activations in designs with 2 or more
> buttons.
> 
> To detect which button has been pressed, cyttsp5_btn_attention() uses a
> for loop to iterate through the input buffer, while shifting and
> applying a bitmask to determine the state for each button.
> Unfortunately, when the bitmask is 0x3 and there are multiple buttons,
> this procedure falls apart.
> 
> Consider a design with 3 buttons. Pressing the third button will result
> in a call to cyttsp5_btn_attention() with the input buffer containing
> 0x4 (binary 0100). In the first iteration of the for loop cur_btn_state
> will be:
> 
> (0x4 >> 0 * 1) & 0x3 = 0x4 & 0x3 = 0x0
> 
> This is correct. However, in the next iteration this happens:
> 
> (0x4 >> 1 * 1) & 0x3 = 0x2 & 0x3 = 0x2
> 
> Which means that a key event for key 1 is generated, even though it's
> not really active. In the third iteration, the loop detects the button
> that was actually pressed:
> 
> (0x4 >> 2 * 1) & 0x3 = 0x1 & 0x3 = 0x1
> 
> This key event is the only one that should have been detected, but it is
> accompanied by the preceding key. Ensuring the applied mask is 0x1
> solves this problem.
> 
> Signed-off-by: Richard Kjerstadius <richard.kjerstadius@teledyne.com>

Applied, thank you.

-- 
Dmitry