linux-kernel - Re: [PATCH v2 5/5] Input: elants_i2c: Delay longer with reset asserted

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 8 Dec 2022 18:20:59 -0800
From:   Dmitry Torokhov <dmitry.torokhov@...il.com>
To:     Douglas Anderson <dianders@...omium.org>
Cc:     Bjorn Andersson <andersson@...nel.org>,
        linux-arm-msm@...r.kernel.org, mka@...omium.org,
        Yunlong Jia <ecs.beijing2022@...il.com>,
        Konrad Dybcio <konrad.dybcio@...aro.org>,
        linux-input@...r.kernel.org, swboyd@...omium.org,
        Johnny Chuang <johnny.chuang.emc@...il.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 5/5] Input: elants_i2c: Delay longer with reset
 asserted

On Thu, Dec 08, 2022 at 06:06:12PM -0800, Douglas Anderson wrote:
> The elan touchscreen datasheet says that the reset GPIO only needs to
> be asserted for 500us in order to reset the regulator. The problem is
> that some boards need a level shifter between the signals on the GPIO
> controller and the signals on the touchscreen. All of these extra
> components on the line can slow the transition of the signals. On one
> board, we measured the reset line and saw that it took almost 1.8ms to
> go low. Even after we bumped up the "drive strength" of the signal
> from the default 2mA to 8mA we still saw it take 421us for the signal
> to go low.
> 
> In order to account for this let's lengthen the amount of time that we
> keep the reset asserted. Let's bump it up from 500us to 5000us.
> That's still a relatively short amount of time and is much safer.
> 
> It should be noted that this fixes real problems. Case in point:
> 1. The touchscreen power rail may be shared with another device (like
>    an eDP panel). That means that at probe time power might already be
>    on.
> 2. In probe we grab the reset GPIO and assert it (make it low).
> 3. We turn on power (a noop since it was already on).
> 4. We wait 500us.
> 5. We deassert the reset GPIO.
> 
> With the above case and only a 500us delay we saw only a partial reset
> asserted, which is bad. Giving it 5ms is overkill but feels safer in
> case someone else has a different level shifter setup.
> 
> Note that bumping up the delay to 5000 means that some configs yell
> about using udelay(). We'll change to using usleep_range(). We give a
> small range here because:
> - This isn't a delay that happens very often so we don't need to worry
>   about giving a big range to allow for power efficiency.
> - usleep_range() is known to almost always pick the upper bound and
>   delay that long and we really don't want to slow down the power on
>   of the touchscreen that much.
> 
> Signed-off-by: Douglas Anderson <dianders@...omium.org>
> Reviewed-by: Matthias Kaehlcke <mka@...omium.org>
> ---
> 
> Changes in v2:
> - Fix typo in commit message (Matthias)
> - udelay -> usleep_range (Patches Robot, Dmitry)
> 
>  drivers/input/touchscreen/elants_i2c.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/input/touchscreen/elants_i2c.c b/drivers/input/touchscreen/elants_i2c.c
> index 879a4d984c90..192d543e5aa9 100644
> --- a/drivers/input/touchscreen/elants_i2c.c
> +++ b/drivers/input/touchscreen/elants_i2c.c
> @@ -114,7 +114,7 @@
>  /* calibration timeout definition */
>  #define ELAN_CALI_TIMEOUT_MSEC	12000
>  
> -#define ELAN_POWERON_DELAY_USEC	500
> +#define ELAN_POWERON_DELAY_USEC	5000
>  #define ELAN_RESET_DELAY_MSEC	20
>  
>  /* FW boot code version */
> @@ -1352,7 +1352,7 @@ static int elants_i2c_power_on(struct elants_data *ts)
>  	 * We need to wait a bit after powering on controller before
>  	 * we are allowed to release reset GPIO.
>  	 */
> -	udelay(ELAN_POWERON_DELAY_USEC);
> +	usleep_range(ELAN_POWERON_DELAY_USEC, ELAN_POWERON_DELAY_USEC + 100);
>  
>  release_reset_gpio:
>  	gpiod_set_value_cansleep(ts->reset_gpio, 0);

This gives me conflict because this label is gone in my tree, so I
adjusted for context and applied.

Thanks.

-- 
Dmitry