Re: RFC: Input: edt-ft5x06 - FT6336G Touch Panel

From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
To: Oliver Graute <oliver.graute@gmail.com>
Cc: Marco Felsch <m.felsch@pengutronix.de>,
	linux-input@vger.kernel.org, dmitry.torokhov@gmail.com,
	rydberg@bitmath.org, robh@kernel.org,
	mylene.josserand@bootlin.com, p.zabel@pengutronix.de
Subject: Re: RFC: Input: edt-ft5x06 - FT6336G Touch Panel
Date: Tue, 17 Mar 2020 19:40:54 +0200	[thread overview]
Message-ID: <20200317174054.GE1922688@smile.fi.intel.com> (raw)
In-Reply-To: <20200317165705.GA16514@ripley>

On Tue, Mar 17, 2020 at 05:57:05PM +0100, Oliver Graute wrote:
> On 16/03/20, Marco Felsch wrote:
> > On 20-03-16 18:30, Andy Shevchenko wrote:
> > > On Mon, Mar 16, 2020 at 03:57:36PM +0100, Oliver Graute wrote:
> > > > On 02/03/20, Oliver Graute wrote:
> > > > > Hello,
> > > > > 
> > > > > I have on of these FT6336G Touch Panels here and I try to get it work
> > > > > with the ft6236 driver. It comes up with model identification "0x11". So
> > > > > I added that identification to the switch case in
> > > > > edt_ft5x06_ts_identify(). But it crashes directly after the probe with a
> > > > > corrupted stack. No clue why. Some ideas?
> > > > 
> > > > I still observe sometimes crashes with ft6236G. When it occurs it is
> > > > always a stack corruption in the probe function on bootup.
> > > > 
> > > > No clue why stack is corrupted sometimes. Is this related to the buggy
> > > > revision field?
> > > > 
> > > > Some further suggestions for debugging this?
> > > 
> > > Are you sure you have proper I²C bus settings on electrical level
> > > (pull-up strength, line capacitance, etc.)?
> > > 
> > > > [    3.806780] edt_ft5x06 1-0038: Model "EP0110M09", Rev. "�", 0x0 sensors
> > > > [    3.808468] at24 2-0050: 2-0050 supply vcc not found, using dummy regulator
> > > > [    3.814864] DEBUG: edt_ft5x06_ts_probe 1201
> > > > [    3.826215] DEBUG: edt_ft5x06_ts_probe 1211
> > > > [    3.830512] DEBUG: edt_ft5x06_ts_probe 1214
> > > > [    3.834904] input: EP0110M09 as /devices/platform/bus@5a000000/5a820000.i2c/i2c-1/1-0038/input/input0
> > > > [    3.844164] at24 2-0050: 4096 byte 24c32 EEPROM, writable, 1 bytes/write
> > > > [    3.844264] DEBUG: edt_ft5x06_ts_probe 1222
> > > > [    3.850919] i2c i2c-2: LPI2C adapter registered
> > > > [    3.855187] edt_ft5x06 1-0038: EDT FT5x06 initialized: IRQ 55, WAKE pin -1, Reset pin -1.
> > > 
> > > > [    3.867878] Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: edt_ft5x06_ts_probe+0x5c4/0xb00
> > > 
> > > Maybe you can look at the assembly to see how compiler puts variables on stack. Also enabling KASAN might help.
> > 
> > This would be the easiest way to debug such failures and since ARMv8
> > support it you just need to enable it =)
> 
> ok, I enabled KASAN and I got the following trace:

Nice. I think it has nothing to do with touch screen.

By the way...

> [    0.000000] Linux version 5.6.0-rc1-next-20200214-00057-g9e78f1560095-dirty (alarm@imx8qm) (gcc version 9.2.0 (GCC)) #100 SMP PREEMPT Tue Mar 17 15:54:17 UTC 2020

May you try the latest and greatest release candidate, i.e. v5.6-rc6 w/o any patches applied?
Then, enable only touchscreen and try again.

Otherwise it's probably good to inform ARM + kasan maintainers about the issue.

-- 
With Best Regards,
Andy Shevchenko