From: Benjamin Tissoires <benjamin.tissoires@redhat.com>
To: Eric GOUYER <folays@gmail.com>
Cc: Illia Ostapyshyn <ostapyshyn@sra.uni-hannover.de>,
David Revoy <davidrevoy@protonmail.com>,
jkosina@suse.cz, jason.gerecke@wacom.com,
jose.exposito89@gmail.com, linux-input@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: Requesting your attention and expertise regarding a Tablet/Kernel issue
Date: Wed, 8 Nov 2023 10:04:30 +0100 [thread overview]
Message-ID: <CAO-hwJK_xp1A=dEOV-2v3KJAf0bRLDWNcrFQeBpgEuxT-qSBnw@mail.gmail.com> (raw)
In-Reply-To: <20231108062306.33f5dcd0@dryade>
Thanks Eric and Illia for the logs and the explanations.
On Wed, Nov 8, 2023 at 6:23 AM Eric GOUYER <folays@gmail.com> wrote:
>
> Sorry, below I re-send my email which bounced due to exceeding 100k
> (hid-recorder traces included), so I re-send it without them, for some
> book-keeping.
>
> I had originally replied-to-all, so mainteners involved should have
> received those traces just before (I hope).
>
> Best Regards,
>
> -----8<-----8<-----8<-----8<
>
> Hello, I have the same tablet than OP (David) :
> - XP-Pen Artist Pro 16 Gen 2
> - on Ubuntu 23.10 linux-image-generic 6.5.0.10.12
>
> I am not (yet ?) encountering the problem described above since I guess
> that my kernel is before the suspected regression.
>
> Here below I included much detail, but you can TL;DR + jump at the five
> hid-recorder traces below.
>
> > On Mon, Nov 6, 2023 at 9:06 PM Illia Ostapyshyn
> > <ostapyshyn@sra.uni-hannover.de> wrote:
> > >
> > > On 11/6/23 17:59, Benjamin Tissoires wrote:
> > >
> > > > If the pen has 2 buttons, and an eraser side, it would be a
> > > > serious design flow for XPPEN to report both as eraser.
> > > >
> > > > Could you please use sudo hid-recorder from hid-tools[1] on any
> > > > kernel version and send us the logs here?
> > > > I'll be able to replay the events locally, and understand why the
> > > > kernel doesn't work properly.
> > > >
> > > > And if there is a design flaw that can be fixed, we might even be
> > > > able to use hid-bpf to change it :)
> > >
> > > My wild guess is that XP-Pen 16 Artist Pro reports :
> > > - (1) an Eraser usage without Invert for the upper button
> > > - (2) and Eraser with Invert for the eraser tip.
>
> I think you will agree with below traces that :
> - (1) : correct (it reports invert=0 eraser=1 tipSwitch=1)
> - (2) : no, for the rubber tip, it reports invert=1 eraser=0 tipSwitch=1
>
> > > A device-specific driver could work with that, but
> > > there seems to be no way to incorporate two different erasers
> > > (thus, allowing userspace to map them to different actions
> > > arbitrarily) in the generic driver currently.
> >
> > That's exactly why I want to see the exact event flow. We can not do
> > "wild guesses" unfortunately (not meaning any offenses).
> > And I am very suspicious about the fact that the stylus reports 2
> > identical erasers. Because in the past David seemed to be able to have
> > 2 distincts behaviors for the 2 "buttons" (physical button and eraser
> > tail).
>
> The Pen, hardware-wise, has the following possibilities :
> - The (main) pressure tip
> - The "bottom" button (nearest of the main pressure tip)
> - The "top" button (farthest of the main tip
> - The "back" pressure tip (it has pressure!) somewhat called rubber
>
> All of those works I think -natively- on my kernel version, without
> external 3rd-party kernel modules.
>
> It works especially on Blender, where I can :
> - click (or paint) with main pressure-tip
> - middle-click (to pan viewport) with "bottom" button + move main tip
> - right-click with "top" button
> - erase with backside pressure-tip
>
> I installed a .deb from the official website [1] which does not contain
> kernel modules, and seems to only contains :
> - a udev rule to, I think, chmod 0666 to input character devices
> - a Qt app to configure some behavior, only userland-side.
>
> I think I observed that before running the Qt app, there was only 3
> xinput devices, and after launching it, there were 4 more (7 total).
>
> Blender would not receive pen input without lanching the Qt app.
> Indeed, the "first" 3 /dev/input/event* associated to the pen are
> chmod'ed too restrictively (0660) for an underprivileged user.
This is wrong (not your fault, of course). There is no way Blender
should directly talk to the input nodes. It should use the events from
the compositor, not randomly getting values from input nodes. sigh...
>
> Anyway, after running the Qt App, 4 *ANOTHER* /dev/input/event* pops,
> correctly chmod'ed (thanks to the udev rules), and Blender works.
I got a quick look, and those events are coming from uinput. So their
Qt App is just a horrible user space driver to fix their tablet. Ouch.
>
> My best guess would be that the first 3 very-native /dev/input/
> devices are somewhat "raw", and that the Qt app feeds those inputs to
> the Qt app's configured pressure curve + remap the button to the input
> as chosen by the user.
The "raw" nodes are giving all of the information, but not correctly.
So they are using a separate userspace driver to "fix" it.
>
> Here is the dmesg before launching the Qt App :
> -----8<-----8<-----8<-----8<
> usb 3-9: new full-speed USB device number 64 using xhci_hcd
> usb 3-9: New USB device found, idVendor=28bd, idProduct=095b,
> bcdDevice= 0.00
> usb 3-9: New USB device strings: Mfr=1, Product=2, SerialNumber=3
> usb 3-9: Product: Artist Pro 16 (Gen2)
> usb 3-9: Manufacturer: UGTABLET
> usb 3-9: SerialNumber: 00000
> input: UGTABLET Artist Pro 16 (Gen2) Mouse as
> /devices/pci0000:00/0000:00:02.1/0000:04:00.0/0000:05:08.0/0000:07:00.0/
> 0000:08:0c.0/0000:69:00.0/usb3/3-9/3-9:1.0/0003:28BD:095B.009E/input/input196
> input: UGTABLET Artist Pro 16 (Gen2) Keyboard as
> /devices/pci0000:00/0000:00:02.1/0000:04:00.0/0000:05:08.0/0000:07:00.0/
> 0000:08:0c.0/0000:69:00.0/usb3/3-9/3-9:1.0/0003:28BD:095B.009E/input/input197
> hid-generic 0003:28BD:095B.009E: input,hidraw4: USB HID v1.00 Mouse
> [UGTABLET Artist Pro 16 (Gen2)] on usb-0000:69:00.0-9/input0
> input: UGTABLET Artist Pro 16 (Gen2) as
> /devices/pci0000:00/0000:00:02.1/0000:04:00.0/0000:05:08.0/0000:07:00.0/
> 0000:08:0c.0/0000:69:00.0/usb3/3-9/3-9:1.1/0003:28BD:095B.009F/input/input198
> hid-generic 0003:28BD:095B.009F: input,hidraw7: USB HID v1.00 Device
> [UGTABLET Artist Pro 16 (Gen2)] on usb-0000:69:00.0-9/input1
> hid-generic 0003:28BD:095B.00A0: hiddev0,hidraw12: USB HID v1.00 Device
> [UGTABLET Artist Pro 16 (Gen2)] on usb-0000:69:00.0-9/input2
> -----8<-----8<-----8<-----8<
>
> This make "sudo xinput" reports 3 new devices :
> UGTABLET Artist Pro 16 (Gen2) Mouse id=x [slave pointer (n)]
> UGTABLET Artist Pro 16 (Gen2) Keyboard id=y [slave keyboard (n+1)]
> UGTABLET Artist Pro 16 (Gen2) id=z [slave keyboard (n+1)]
> (sorry I've masked the id to prevent confusion, I did not run "xinput"
> at the same time/power cycle than "dmesg")
>
> Running the Qt App makes "sudo xinput" report 4 new devices :
> XP-Pen Mouse id=xc [slave pointer (n)]
> XP-Pen Eraser id=xa [slave keyboard (n+1)]
> XP-Pen Pen id=xb [slave keyboard (n+1)] << only this chmod'ed 0666
> XP-Pen Mouse id=xd [slave keyboard (n+1)]
>
> And... it suffices to make Blender works without configuring anything.
>
> Anyway, besides this user experience / userland Qt app, here are some
> five hid-recorder traces ;
Many thanks for attaching them and for digging into them.
> I re-used the terminology appearing in the traces :
> - "barrel" : the "bottom" button nearest of the main tip
> - "erase" : the farther "top" button, just above "barrel"
> - "back" : the "rubber" secondary pressure tip at the back of the pen
>
> (1) hidraw7_inRange_contact_move_lift_outOfRange
> I "contact" the main tip, move it, and lift it out or range.
>
> (2) hidraw7_inRange_contact_barrelPress\
> _move_barrelRelease_lift_outOfRange
> I "contact" the main tip, press "barrel" (which is the button nearest
> of the tip), move the pen, release the button, move it out of range.
>
> (3) hidraw7_inRange_contact_erasePress\
> _move_eraseRelease_lift_outOfRange
> The same than above, except that I instead use the "erase" button,
> i.e. the button just above ("vertically speaking") of the "main" button.
>
> (4) hidraw7_inRange_contact_BOTHbarrelAnderasePress_move_\
> BOTHbarrelAnderaseRelease_lift_outOfRange
> That's just plain stupid, and uncomfortable to do, but it's just to
> show an example of the capability of the hardware ;
> In this one, I'm doing the same than above, except I stupidly press BOTH
> buttons.
> It means that I "contact" the main tip, press both barrel+erase, move
> the pen, release both button, lift it out of range.
>
> (5) hidraw7_inRange_contact_backPress_move_backRelease_lift_outOfRange
> Back to non-stupid tests, here I am using the "rubber" (which has
> pressure !), i.e. I use the back of the pen.
> I contact it, press the rubber to have some pressure, move it, release
> the pressure, move the pen ouf of contact and out of range.
>
> Reading those traces, I think I observed some (to me) very logical
> behavior ;
>
> I agree that the Pen is possibly not acting in respect of the
> specification, by the fact that... the spec does not differentiate the
> "eraser" 2nd button, versus the "backside" rubber pressure tip.
>
> The traces shows that :
> - In range [0 - 1] is correctly reported, both for main tip + rubber tip
> - Tip Pressure [0 - n] is correctly reported, for both pressure tips
Yep, that part is correct and follows the specification
>
> - Tip Switch [0 - 1] is correctly set to ONE for all traces, when the
> in-range tip goes to contact. (either main tip or rubber tip)
This is half wrong (spec-wise):
- tip switch should only be used when the actual tip is in contact.
- when using the rubber tip, we should get an eraser event :(
>
> - Barrel Switch [0 - 1] is correctly set to ONE only when pressing
> button down, and of course only when using the main tip
This is correct (as in it follows the spec)
>
> - Invert [0 - 1] is set to ONE when (and only) using the back rubber
> tip, and the ERASE button (2nd button) does NOT set it, so "INVERT"
In your case, the invert usage is correct.
In Illia's case, the invert usage is forwarded by the eraser usage, so
this is just plain wrong there.
>
> - Eraser [0 - 1] is set to ONE ONLY when pressing the 2nd top button,
> and of :
> - ONLY when using the main itp
> - and if you were to instead use the "back" of the pen (rubber tip),
> then you will have invert=1 + eraser=0
This is not correct (again, according to the specification). XP-Pen
should report SecondaryBarrelSwitch (0x5a in the HID usage table[0])
And again, in Illia's case, this button should be reported as an
invert, and eraser when the tip touches the tablet.
>
> The main extract of those is, I think, that :
> - eraser "top" button is completely independent of the back "rubber"
Yes
> - eraser "top" button NEVER concerns itself to set invert=1
yes
> - backside "rubber" pressure tip will set invert=1, but leave eraser=0
yes, but that's not correct, it should report eraser=1 when you touch
the rubber on the surface.
>
> So, the behavior probably breaks the specs, but sincerely I'm happy to
> have the "eraser" button independent of the "rubber eraser", which
> makes the stylus a somewhat 4-buttons stylus (tip, button1, button2,
> rubber), and I would like to keep this.
Yes, and I'd like to keep it that way, even if 6.6 and 6.5.8
apparently broke it.
So, to me:
- 276e14e6c3993317257e1787e93b7166fbc30905 is wrong: this is a
firmware bug (reporting invert through eraser) and should not be
tackled at the generic level, thus it should be reverted
- both of these tablets are forwarding the useful information, but not
correctly, which confuses the kernel
- I should now be able to write regression tests
- I should be able to provide HID-BPF fixes for those tablets so that
we can keep them working with or without
276e14e6c3993317257e1787e93b7166fbc30905
reverted (hopefully)
- problem is I still don't have the mechanics to integrate the HID-BPF
fixes directly in the kernel tree, so maybe I'll have to write a
driver for XP-Pen while these internals are set (it shouldn't
interfere with the HID-BPF out of the tree).
Cheers,
Benjamin
>
> Best Regards,
>
> [1] https://www.xp-pen.fr/download-1027.html ; the .deb Qt app
>
> --
> Eric GOUYER
>
[0] https://usb.org/document-library/hid-usage-tables-14
next prev parent reply other threads:[~2023-11-08 9:10 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <kRKTNDYigUSblpNgSuZ2H4dX03Of1yD4j_L9GgbyKXcDqZ67yh5HOQfcd7_83U3jZuQzxpKT3L6FXcRkkZIGdl_-PQF14oIB0QmRSfvpc2k=@protonmail.com>
2023-11-01 19:37 ` Requesting your attention and expertise regarding a Tablet/Kernel issue Jiri Kosina
2023-11-02 0:44 ` Bagas Sanjaya
2023-11-02 6:31 ` Linux regression tracking (Thorsten Leemhuis)
2023-11-02 7:51 ` Bagas Sanjaya
2023-11-02 7:44 ` Linux regression tracking #adding (Thorsten Leemhuis)
2023-12-01 8:24 ` Linux regression tracking #update (Thorsten Leemhuis)
2023-11-03 20:05 ` Illia Ostapyshyn
2023-11-04 0:46 ` Bagas Sanjaya
2023-11-06 13:17 ` David Revoy
2023-11-06 16:59 ` Benjamin Tissoires
2023-11-06 20:06 ` Illia Ostapyshyn
2023-11-07 7:59 ` Benjamin Tissoires
2023-11-07 13:40 ` Illia Ostapyshyn
2023-11-08 5:23 ` Eric GOUYER
2023-11-08 9:04 ` Benjamin Tissoires [this message]
2023-11-08 9:23 ` José Expósito
2023-11-08 9:34 ` Benjamin Tissoires
2023-11-08 18:21 ` Benjamin Tissoires
2023-11-09 0:32 ` David Revoy
2023-11-09 11:56 ` Benjamin Tissoires
2023-11-09 16:13 ` Benjamin Tissoires
2023-11-09 22:04 ` David Revoy
2023-11-10 10:19 ` Benjamin Tissoires
2023-11-11 8:52 ` Benjamin Tissoires
2023-11-13 22:08 ` David Revoy
2023-11-14 14:35 ` Benjamin Tissoires
2023-11-15 15:14 ` Benjamin Tissoires
2023-11-23 22:12 ` David Revoy
2023-11-24 17:18 ` Benjamin Tissoires
2023-11-29 15:32 ` Benjamin Tissoires
2023-11-30 22:25 ` David Revoy
2023-12-01 15:41 ` Benjamin Tissoires
2023-11-08 19:40 ` Nils Fuhler
2023-11-08 20:34 ` Benjamin Tissoires
2023-11-08 22:31 ` ostapyshyn
2023-11-09 11:46 ` Benjamin Tissoires
2023-11-08 23:18 ` David Revoy
2023-11-09 11:47 ` Benjamin Tissoires
2023-11-11 17:15 ` Thorsten Leemhuis
2023-11-08 22:28 ` David Revoy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAO-hwJK_xp1A=dEOV-2v3KJAf0bRLDWNcrFQeBpgEuxT-qSBnw@mail.gmail.com' \
--to=benjamin.tissoires@redhat.com \
--cc=davidrevoy@protonmail.com \
--cc=folays@gmail.com \
--cc=jason.gerecke@wacom.com \
--cc=jkosina@suse.cz \
--cc=jose.exposito89@gmail.com \
--cc=linux-input@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=ostapyshyn@sra.uni-hannover.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).