Date:   Fri, 12 Apr 2019 00:17:52 +0800
From:   Kai-Heng Feng <kai.heng.feng@...onical.com>
To:     hotwater438@...anota.com
Cc:     Hans de Goede <hdegoede@...hat.com>,
        Dmitry Torokhov <dtor@...omium.org>,
        Vladislav Dalechyn <vlad.dalechin@...il.com>,
        Benjamin Tissoires <benjamin.tissoires@...hat.com>,
        Jiri Kosina <jikos@...nel.org>, Swboyd <swboyd@...omium.org>,
        Bigeasy <bigeasy@...utronix.de>,
        "open list:HID CORE LAYER" <linux-input@...r.kernel.org>,
        lkml <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] ELAN touchpad i2c_hid bugs fix

Hi,

at 05:18, <hotwater438@...anota.com> <hotwater438@...anota.com> wrote:

> Hi.
>
> 1) Run "cat /proc/interrupts | grep ELAN" , note down the value
> 2) Very quickly/briefly touch the touchpad once
> 3) Run "cat /proc/interrupts | grep ELAN" , note down the value again
> 4) Subtract result from 1. from result from 3, this difference is
> the value we are interested in. E.g. my testing got 254 and 257, so
> a difference of 3.
> I've tested that, main diffs are 30, 24, 16 (the most frequent), 2 (the  
> least frequent).
>
> I was using 4.19.13 kernel, because I use ParrotOS (which happens to be  
> Debian distribution).
> But I've installed experimental 5.0.0 kernel and I can't say right now if  
> suspend problem is resolved (i have to rebuild latest kernel with patch).

Can you try below fix?

This can solve what commit 1475af255e18 ("HID: i2c-hid: Ignore input report  
if there's no data present on Elan touchpanels”) tries to workaround.

diff --git a/drivers/pinctrl/intel/pinctrl-intel.c  
b/drivers/pinctrl/intel/pinctrl-intel.c
index c19a4c45f7bb..30e3664f1ae5 100644
--- a/drivers/pinctrl/intel/pinctrl-intel.c
+++ b/drivers/pinctrl/intel/pinctrl-intel.c
@@ -957,6 +957,10 @@ static void intel_gpio_irq_mask_unmask(struct irq_data  
*d, bool mask)
                 reg = community->regs + community->ie_offset + gpp * 4;
 
                 raw_spin_lock_irqsave(&pctrl->lock, flags);
+
+               if (!mask)
+                       writel(BIT(gpp_offset), community->regs +  
community->is_offset + gpp * 4);
+
                 value = readl(reg);
                 if (mask)
                         value &= ~BIT(gpp_offset);


>
> Regards,
> Vladislav.
>
> Apr 3, 2019, 2:18 PM by hdegoede@...hat.com:
> Hi,
>
> On 31-03-19 11:50, hotwater438@...anota.com wrote:
> Hi. I've done everything you said, here are results:
>
> Vladislav can you check the output of /cat/interrupts on a kernel
> without the patch and while *not* using the touchpad; and check
> if the amount of touchpads-interrupts still keeps increasing in this
> case?
>
> IWI or IRQ work interrupts keep increasing with speed at least 3  
> interrupts/s.
>
> I'm really only interested in the touchpad related IRQs, so e.g. the line
> about "intel-gpio 129 ELAN1200:00", if you're seeing 3 interrupts/s on
> some others that is fine, so I take it the ELAN1200:00 interrupt count
> does not increase on an *unpatched* kernel, unless you use the touchpad?
> Also when I am moving touchpad IR-IO-APIC 14-fasteoi INT345D:00 get's  
> triggered and increased.
>
> That is the GPIO controller interrupt, so that one increasing is normal.
>
> If I understand things correctly then this all means that the IRQ indeed
> is a normal level IRQ and Dmitry is likely correct that there is an
> pinctrl / gpiochip driver problem here.
>
> Can you try the following with an *unpatched* kernel? :
>
> 1) Run "cat /proc/interrupts | grep ELAN" , note down the value
> 2) Very quickly/briefly touch the touchpad once
> 3) Run "cat /proc/interrupts | grep ELAN" , note down the value again
> 4) Subtract result from 1. from result from 3, this difference is
> the value we are interested in. E.g. my testing got 254 and 257, so
> a difference of 3.
>
> The goal here is to get an as low as possible difference. Feel free
> to repeat this a couple of times.
>
> On an Apollo Lake laptop with an I2C hid mt touchpad I can get the
> amount of interrupts triggered for a single touch down to 3,
> given the huge interrupt counts of 130000+ reported in:
> https://bugzilla.redhat.com/show_bug.cgi?id=1543769
>
> I expect you to get a much bigger smallest possible difference
> between 2 "cat /proc/interrupts | grep ELAN" commands, note a
> difference of 0 means your touch did not register.
>
> Assuming you indeed see much more interrupts for a very quick
> touch + release, then we indeed have an interrupt handling problem
> we need to investigate further.
> I don't know if it's important or not, but for some reason these  
> interrupts keep popping only on CPU2 (i have 4cpu processor).
>
> That does not matter.
> 1) Suspending the machine by selecting suspend from a menu in your
> desktop environment, or by briefly pressing the power-button, do
> not close the lid
> 2) As soon as the system starts suspending and while it is suspended, move
> your finger around the touchpad
> 3) Wake the system up with the powerbutton while moving your finger around
> 4) Check if the touchpad still works after this
>
> It works, but as it seems, looses edge. JournalCTL is being flooded with  
> i2c_hid_get_input: incomplete report (16/65535)
>
> That is probably a different issue. If you loose the edge IRQ, then the  
> touchpad
> would stop working without any messages. I believe that the suspend /  
> resume
> issue may be fixed by:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=52cf93e63ee672a92f349edc6ddad86ec8808fd8
>
> Does your kernel have this commit? (please always use the latest kernel  
> while
> testing).
> Also a thing to notice, that after manually removing and modprobing  
> i2c_hid module, it says next in journalctl:
>
> i2c_hid i2c-ELAN1200:00: i2c-ELAN1200:00 supply vdd not found, using  
> dummy regulator
> i2c_hid i2c-ELAN1200:00: i2c-ELAN1200:00 supply vddl not found, using  
> dummy regulator
>
> Those messages can safely be ignored.
>
> Regards,
>
> Hans

Powered by blists - more mailing lists