Re: Graphire4 Pad Button freeze with 0.10.11.

> On Wed, Jun 1, 2011 at 3:53 PM, Jason Gerecke <killertofu@...> wrote:
>> On Tue, May 31, 2011 at 12:43 PM, Chris Bagwell <chris@...> wrote:
>>> On Tue, May 31, 2011 at 1:28 PM, Jason Gerecke <killertofu@...> wrote:
>>>> On Sat, May 28, 2011 at 5:53 PM, Chris Bagwell <chris@...> wrote:
>>>>> On Sat, May 28, 2011 at 3:54 PM, Jason Gerecke <killertofu@...> wrote:
>>>>>> On Fri, May 27, 2011 at 9:11 PM, Chris Bagwell <chris@...> wrote:
>>>>>>> Hi Favux,
>>>>>>>
>>>>>>> For first bug report, it does sound same issue as I looked into a
>>>>>>> while back.  The kernel driver is mistakenly telling userland that the
>>>>>>> PAD device has gone out of proximity when you use pen.  From that
>>>>>>> point, buttons don't work.
>>>>>>>
>>>>>>> Something has changed in xf86-input-wacom that we must l trust kernel
>>>>>>> information more then we used to.  My personal preference is to fix
>>>>>>> the kernel to give us good info instead of reverting back to old
>>>>>>> xf86-input-wacom behavior.
>>>>>>>
>>>>>>> Maybe when Ping gets back next month we can have some progress.  Or
>>>>>>> hopefully we can find a user that doesn't mind hacking kernel drivers.
>>>>>>>  I can give tons of guidance on patch to implement (its like 3 lines
>>>>>>> for simple solution) but ultimately I can't push forward much more
>>>>>>> since I don't have the device.
>>>>>>>
>>>>>>> For second bug report (X server crash), I've no idea.  It sounds like
>>>>>>> its own issue.
>>>>>>>
>>>>>>> Chris
>>>>>>>
>>>>>>
>>>>>> Chris, I just happen to have a Graphire4 at home. If you have an idea
>>>>>> of where the bug may lie, I'd be happy to try patching up the kernel
>>>>>> code :)
>>>>>>
>>>>>> Jason
>>>>>>
>>>>>
>>>>> Cool.  I've attached the patch I think that is needed.  This is
>>>>> against Linus' git that was rebased at some random time in last month
>>>>> or two.
>>>>>
>>>>> Originally, I was going to give you a patch against input-wacom but it
>>>>> looks like that is  different compared to upstream for graphire.
>>>>> Specifically, it is correctly setting MSC_SERIAL when switching over
>>>>> to pen.
>>>>>
>>>>> That means input-wacom may be working with graphire4.  input-wacom is
>>>>> not ever sending BTN_TOOL_FINGER for some reason so if its not working
>>>>> then thats probably only change needed (add it back in to match
>>>>> kernel).
>>>>>
>>>>> Chris
>>>>>
>>>>
>>>> While I can confirm the bug under Ubuntu 11.04 64-bit
>>>> (2.6.38-8-generic), I can't under my 32-bit Arch Linux development
>>>> box. This was tested with the pre-installed "2.6.38-ARCH" kernel, as
>>>> well as the tarballs from kernel.org for 2.6.38.7 or 2.8.39 (with the
>>>> config from /proc/config used as the source). xf86-input-wacom was
>>>> from Git obviously :)
>>>>
>>>
>>> Strange.   Must be distro's backporting patches.
>>>
>>>> There is a similar bug under these kernels (multiple presses to the
>>>> same pad button don't work while pen is in prox -- you either have to
>>>> press another button before the first works again, or take the pen out
>>>> of prox to press buttons repeatedly).
>>>
>>> Yeah, that sounds about right for this bug.  When switching tools (PEN
>>> vs. PAD), we are not correctly changing serial # and that side affect
>>> can be seen a few different ways.
>>>
>>>>
>>>> I'll try setting up a build environment under my virtual Ubuntu box to
>>>> see what the official kernel sources do there.
>>>
>>> Since Graphire is generally stable, I look at git to see whats been
>>> changing.  Here is immediately interesting one.  Looks like a
>>> non-Tablet PC change snuck in this commit.
>>>
>>> commit ab687b18aa77aeda5472d9ea054bf92c45c49c0c
>>> Author: Ping Cheng <pingc@...>
>>> Date:   Mon Apr 5 23:07:41 2010 -0700
>>>
>>>    Input: wacom - streamline 2-finger touch support
>>>
>>>    Clean up 2-finger touch support. This still needs to be converted to
>>>    proper multi-touch protocol.
>>>
>>>    Signed-off-by: Ping Cheng <pingc@...>
>>>    Signed-off-by: Dmitry Torokhov <dtor@...>
>>>
>>> diff --git a/drivers/input/tablet/wacom_wac.c b/drivers/input/tablet/wacom_wac.c
>>> index 950a81d..847fd01 100644
>>> --- a/drivers/input/tablet/wacom_wac.c
>>> +++ b/drivers/input/tablet/wacom_wac.c
>>> @@ -236,13 +236,12 @@ static int wacom_graphire_irq(struct wacom_wac *wacom)
>>>                        rw = ((data[7] & 0x18) >> 3) - ((data[7] & 0x20) >> 3);
>>>                        input_report_rel(input, REL_WHEEL, rw);
>>>                        input_report_key(input, BTN_TOOL_FINGER, 0xf0);
>>> -                       input_report_abs(input, ABS_MISC, wacom->id[1]);
>>>                        if (!prox)
>>>                                wacom->id[1] = 0;
>>>                        input_report_abs(input, ABS_MISC, wacom->id[1]);
>>>                        input_event(input, EV_MSC, MSC_SERIAL, 0xf0);
>>> +                       retval = 1;
>>>                }
>>> -               retval = 1;
>>>                break;
>>>
>>>
>>> There more in that patch but just wanted to show the part about
>>> ABS_MISC being removed.
>>>
>>> Here is my original email ton linuxwacom-devel to describe gist of
>>> problem.  It contains reference to why removing that above ABS_MISC
>>> may have side affect of fixing graphire issue.
>>>
>>> http://sourceforge.net/mailarchive/message.php?msg_id=27202503
>>>
>>> I prefer my patch I included in this thread alaready for its
>>> correctness but I think above happens to make things squeak by.
>>>
>>> Chris
>>>
>>
>> Looks like your patch works when applied to the ubuntu-natty git tree.
>> Buttons continue to work flawlessly after pen enters and exits
>> proximity. I haven't yet tested to see if your patch fixes the related
>> bug I was seeing on my development box, but I'll get around to that
>> soon(-ish).
>>
>> Jason
>>
>> ---
>> Day xee-nee-svsh duu-'ushtlh-ts'it;
>> nuu-wee-ya' duu-xan' 'vm-nvshtlh-ts'it.
>> Huu-chan xuu naa~-gha.
>>
>
> Patch looks good on my Arch box with the kernel.org 2.6.38.7 release
> as well. Fixes the related bug I was experiencing.
>
> I assume Peter knows how to get this upstream?
>