Fwd: Re: Phis in /proc/bus/input/devices same for all devices?

Fwd: Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Hi,

can anybody reply to this and share his/her opinion?

Thanks a lot

Michal

----------  Předaná zpráva  ----------

Subject: Re: Phis in /proc/bus/input/devices same for all devices?
Date: středa 21 únor 2007 23:06
From: Marcel Holtmann <marcel@holtmann.org>
To: CIJOML <cijoml@volny.cz>
Cc: bluez-users@lists.sourceforge.net

Hi Michal,

> I use 2.6.20-mh1 patch and would like to differ devices based on Phis, but
> it is the same for all devices??
>
> I: Bus=0005 Vendor=045e Product=007c Version=0035
> N: Name="Microsoft Five Button Mouse"
> P: Phys=00:0D:88:8E:AB:2F
> S: Sysfs=/class/input/input19
> H: Handlers=mouse1 event6
> B: EV=7
> B: KEY=1f0000 0 0 0 0 0 0 0 0
> B: REL=103
>
> I: Bus=0005 Vendor=045e Product=007b Version=0039
> N: Name="Microsoft Bluetooth keyboard"
> P: Phys=00:0D:88:8E:AB:2F
> S: Sysfs=/class/input/input20
> H: Handlers=kbd event7
> B: EV=12000b
> B: KEY=10f80 44007 ffe01878 800d7ff febeffdf f3cfffff ffffffff fffffffe
> B: ABS=700 0
> B: LED=107
>
> I: Bus=0005 Vendor=046d Product=b3e3 Version=2503
> N: Name="Logitech Bluetooth Mediapad"
> P: Phys=00:0D:88:8E:AB:2F
> S: Sysfs=/class/input/input21
> H: Handlers=kbd event8
> B: EV=12000f
> B: KEY=7fff 2c3027 bf004440 0 0 1 10f80 8807c007 ffe67bfa d9415fff febeffdf
> ffefffff ffffffff fffffffe
> B: REL=40
> B: ABS=301 0
> B: LED=1f
>
> Shouldn't be there it's BDADDR?

please post this question to LKML. I am not sure that is the right
approach. Currently we fill in phys and uniq like this:

        strncpy(hid->phys, batostr(&src), 64);
        strncpy(hid->uniq, batostr(&dst), 64);

Maybe the input or HID subsystem should be changed to also export the
uniq values to udev.

Regards

Marcel

-------------------------------------------------------

Re: Re: Phis in /proc/bus/input/devices same for all devices?

[Dmitry Torokhov]

On 2/21/07, CIJOML <cijoml@volny.cz> wrote:
> Hi,
>
> can anybody reply to this and share his/her opinion?

Input core already exports uniq to udev and also via sysfs.

-- 
Dmitry

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

So where is that damned bug, that I don't see those uniq BDADDRESSes there?

Michal

Dne středa 21 únor 2007 23:27 Dmitry Torokhov napsal(a):
> On 2/21/07, CIJOML <cijoml@volny.cz> wrote:
> > Hi,
> >
> > can anybody reply to this and share his/her opinion?
>
> Input core already exports uniq to udev and also via sysfs.

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Hi Dmitry,

there are correct values, but not in 
/proc/bus/input/devices, where is value non of my devices owns.

Michal

Dne středa 21 únor 2007 23:43 Dmitry Torokhov napsal(a):
> On 2/21/07, CIJOML <cijoml@volny.cz> wrote:
> > So where is that damned bug, that I don't see those uniq BDADDRESSes
> > there?
> >
> > Michal
> >
> > Dne středa 21 únor 2007 23:27 Dmitry Torokhov napsal(a):
> > > On 2/21/07, CIJOML <cijoml@volny.cz> wrote:
> > > > Hi,
> > > >
> > > > can anybody reply to this and share his/her opinion?
> > >
> > > Input core already exports uniq to udev and also via sysfs.
>
> If you cat /sys/class/input/inputX/uniq for your BT devices  - is it empty?
>
> Uniq is not exported via  /proc/bus/input/devices.

Re: Re: Phis in /proc/bus/input/devices same for all devices?

[Marcel Holtmann]

Hi Dmitry,

> > can anybody reply to this and share his/her opinion?
> 
> Input core already exports uniq to udev and also via sysfs.

so do you think it would be better to create phys as a string of the
source and destination address like "<src>-<dst>" for Bluetooth. And
then keep the uniq empty like all USB devices without serial number do?

Regards

Marcel

Re: Re: Phis in /proc/bus/input/devices same for all devices?

[Dmitry Torokhov]

On 2/21/07, Marcel Holtmann <marcel@holtmann.org> wrote:
> Hi Dmitry,
>
> > > can anybody reply to this and share his/her opinion?
> >
> > Input core already exports uniq to udev and also via sysfs.
>
> so do you think it would be better to create phys as a string of the
> source and destination address like "<src>-<dst>" for Bluetooth. And
> then keep the uniq empty like all USB devices without serial number do?
>

I'd keep uniq as is and do what you proposed with phys. This way you
can either use uniq to identify your device while moving it from one
receiver to another (if you have several) or do exact match on phys to
get receiver/device pair.

Does this make sense?

-- 
Dmitry

Re: Re: Phis in /proc/bus/input/devices same for all devices?

[Marcel Holtmann]

Hi Dmitry,

> > > > can anybody reply to this and share his/her opinion?
> > >
> > > Input core already exports uniq to udev and also via sysfs.
> >
> > so do you think it would be better to create phys as a string of the
> > source and destination address like "<src>-<dst>" for Bluetooth. And
> > then keep the uniq empty like all USB devices without serial number do?
> >
> 
> I'd keep uniq as is and do what you proposed with phys. This way you
> can either use uniq to identify your device while moving it from one
> receiver to another (if you have several) or do exact match on phys to
> get receiver/device pair.
> 
> Does this make sense?

actually it doesn't help to keep uniq around since the Bluetooth HID
always reconnects to the same source/host address. Making it reconnect
results in a virtual cable unplug. So I think that I am going to modify
the phys to include source and destination address. In the end it is
only a string.

Regards

Marcel

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Marcel how can you differ what is src and what is dst, when device can connect 
first time from hub to keyboard and later keyboard to hub?

Michal

Dne středa 21 únor 2007 23:53 Marcel Holtmann napsal(a):
> Hi Dmitry,
>
> > > > > can anybody reply to this and share his/her opinion?
> > > >
> > > > Input core already exports uniq to udev and also via sysfs.
> > >
> > > so do you think it would be better to create phys as a string of the
> > > source and destination address like "<src>-<dst>" for Bluetooth. And
> > > then keep the uniq empty like all USB devices without serial number do?
> >
> > I'd keep uniq as is and do what you proposed with phys. This way you
> > can either use uniq to identify your device while moving it from one
> > receiver to another (if you have several) or do exact match on phys to
> > get receiver/device pair.
> >
> > Does this make sense?
>
> actually it doesn't help to keep uniq around since the Bluetooth HID
> always reconnects to the same source/host address. Making it reconnect
> results in a virtual cable unplug. So I think that I am going to modify
> the phys to include source and destination address. In the end it is
> only a string.
>
> Regards
>
> Marcel

Re: Phis in /proc/bus/input/devices same for all devices?

[Marcel Holtmann]

Hi Michal,

> how can you differ what is src and what is dst, when device can connect 
> first time from hub to keyboard and later keyboard to hub?

the source is always the host (meaning the adapter) and the destination
is always the device.

Regards

Marcel

Re: Re: Phis in /proc/bus/input/devices same for all devices?

[Dmitry Torokhov]

Hi Marcel,

On 2/21/07, Marcel Holtmann <marcel@holtmann.org> wrote:
> Hi Dmitry,
>
> > > > > can anybody reply to this and share his/her opinion?
> > > >
> > > > Input core already exports uniq to udev and also via sysfs.
> > >
> > > so do you think it would be better to create phys as a string of the
> > > source and destination address like "<src>-<dst>" for Bluetooth. And
> > > then keep the uniq empty like all USB devices without serial number do?
> > >
> >
> > I'd keep uniq as is and do what you proposed with phys. This way you
> > can either use uniq to identify your device while moving it from one
> > receiver to another (if you have several) or do exact match on phys to
> > get receiver/device pair.
> >
> > Does this make sense?
>
> actually it doesn't help to keep uniq around since the Bluetooth HID
> always reconnects to the same source/host address. Making it reconnect
> results in a virtual cable unplug. So I think that I am going to modify
> the phys to include source and destination address. In the end it is
> only a string.
>

Matching on uniq (and having uniq available) makes sense when you want
to perform device-specific setup and want it to work even if you
change your BT adapter. Imagne yopu have a keyboard that yoou want to
handle in a special way and after you set up your BT card gets fried
and you buy another one. When matchign on uniq your setup will not
change at all.

-- 
Dmitry

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

I need this for differing Option "XkbModel" in Xorg, so I can bind to each
device it's model and has correct xev's scancodes based on model.
So I prefer to have as Phis just only BDADDR, because for me this doesn't make 
sense to have there device to which keyboard connect. Then I should have 1 
section for each receiver, which is stupid.

Section "InputDevice"
        Identifier  "Bluetooth Keyboard"
        Driver      "kbd"
        Option  "Name"  "Bluetooth HID Boot Protocol Device"
        Option  "Phis"        "BDADDR"
        Option      "XkbRules"  "xfree86"
        Option      "XkbModel"  "dinovo"
        Option      "XkbLayout" "us"
EndSection

and so on for every other BT keyboard I own.


Michal

Dne středa 21 únor 2007 23:48 Dmitry Torokhov napsal(a):
> On 2/21/07, Marcel Holtmann <marcel@holtmann.org> wrote:
> > Hi Dmitry,
> >
> > > > can anybody reply to this and share his/her opinion?
> > >
> > > Input core already exports uniq to udev and also via sysfs.
> >
> > so do you think it would be better to create phys as a string of the
> > source and destination address like "<src>-<dst>" for Bluetooth. And
> > then keep the uniq empty like all USB devices without serial number do?
>
> I'd keep uniq as is and do what you proposed with phys. This way you
> can either use uniq to identify your device while moving it from one
> receiver to another (if you have several) or do exact match on phys to
> get receiver/device pair.
>
> Does this make sense?

Re: Phis in /proc/bus/input/devices same for all devices?

[Marcel Holtmann]

Hi Michal,

> I need this for differing Option "XkbModel" in Xorg, so I can bind to each
> device it's model and has correct xev's scancodes based on model.
> So I prefer to have as Phis just only BDADDR, because for me this doesn't make 
> sense to have there device to which keyboard connect. Then I should have 1 
> section for each receiver, which is stupid.

it makes sense to make it analogous to the USB driver, where the phys
includes the USB path. The equivalent for Bluetooth is the source plus
destination.

> Section "InputDevice"
>         Identifier  "Bluetooth Keyboard"
>         Driver      "kbd"
>         Option  "Name"  "Bluetooth HID Boot Protocol Device"
>         Option  "Phis"        "BDADDR"
>         Option      "XkbRules"  "xfree86"
>         Option      "XkbModel"  "dinovo"
>         Option      "XkbLayout" "us"
> EndSection

I don't know how "Phis" suppose to work, but I might think it is better
using vendor and product ID for matching.

Regards

Marcel

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Dne čtvrtek 22 únor 2007 00:00 Marcel Holtmann napsal(a):
> Hi Michal,
>
> > I need this for differing Option "XkbModel" in Xorg, so I can bind to
> > each device it's model and has correct xev's scancodes based on model. So
> > I prefer to have as Phis just only BDADDR, because for me this doesn't
> > make sense to have there device to which keyboard connect. Then I should
> > have 1 section for each receiver, which is stupid.
>
> it makes sense to make it analogous to the USB driver, where the phys
> includes the USB path. The equivalent for Bluetooth is the source plus
> destination.
>
> > Section "InputDevice"
> >         Identifier  "Bluetooth Keyboard"
> >         Driver      "kbd"
> >         Option  "Name"  "Bluetooth HID Boot Protocol Device"
> >         Option  "Phis"        "BDADDR"
> >         Option      "XkbRules"  "xfree86"
> >         Option      "XkbModel"  "dinovo"
> >         Option      "XkbLayout" "us"
> > EndSection
>
> I don't know how "Phis" suppose to work, but I might think it is better
> using vendor and product ID for matching.

Xorg just reads lines from devices and match devices against it. So Phis makes 
configuration easier ;), because only one line is enough to match device.

>
> Regards
>
> Marcel

Michal

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Dne čt 22. února 2007 Dmitry Torokhov napsal(a):
> On 2/21/07, CIJOML <cijoml@volny.cz> wrote:
> > Dne čtvrtek 22 únor 2007 00:00 Marcel Holtmann napsal(a):
> > > Hi Michal,
> > >
> > > > I need this for differing Option "XkbModel" in Xorg, so I can bind to
> > > > each device it's model and has correct xev's scancodes based on
> > > > model. So I prefer to have as Phis just only BDADDR, because for me
> > > > this doesn't make sense to have there device to which keyboard
> > > > connect. Then I should have 1 section for each receiver, which is
> > > > stupid.
> > >
> > > it makes sense to make it analogous to the USB driver, where the phys
> > > includes the USB path. The equivalent for Bluetooth is the source plus
> > > destination.
> > >
> > > > Section "InputDevice"
> > > >         Identifier  "Bluetooth Keyboard"
> > > >         Driver      "kbd"
> > > >         Option  "Name"  "Bluetooth HID Boot Protocol Device"
> > > >         Option  "Phis"        "BDADDR"
> > > >         Option      "XkbRules"  "xfree86"
> > > >         Option      "XkbModel"  "dinovo"
> > > >         Option      "XkbLayout" "us"
> > > > EndSection
> > >
> > > I don't know how "Phis" suppose to work, but I might think it is better
> > > using vendor and product ID for matching.
> >
> > Xorg just reads lines from devices and match devices against it. So Phis
> > makes configuration easier ;), because only one line is enough to match
> > device.
>
> They can get vendor and id from /proc/input/devices as well. Plus I
> will export uniq in /proc/bus/input/devices and then yo can ask X guys
> to allow matching on uniq as well.

Hi Dmitri,

I watched at 2.6.23 and this is still not done yet. Are there any patches 
available to use???

Thanks a lot for reply

Michal

Re: Phis in /proc/bus/input/devices same for all devices?

[Dmitry Torokhov]

Hi Michal,

On Monday 29 October 2007, CIJOML wrote:
> Dne čt 22. února 2007 Dmitry Torokhov napsal(a):
>
> > Plus I
> > will export uniq in /proc/bus/input/devices and then yo can ask X guys
> > to allow matching on uniq as well.
> 
> Hi Dmitri,
> 
> I watched at 2.6.23 and this is still not done yet. Are there any patches 
> available to use???
> 

Input core does export uniq in /proc/bus/input/devices so now it is
up to individual drivers to populate this field in input_dev
structure.

-- 
Dmitry

Re: Phis in /proc/bus/input/devices same for all devices?

[CIJOML]

Hi Marcel,

are there any patches available to support uniq? I really want to have easy 
setup of my 2 same diNovo desktops keyboarad sets each with different keymap 
etc...

Dimitri thank you a lot for support so I can push Xguys now ;)

Michal

Dne út 30. října 2007 Dmitry Torokhov napsal(a):
> Hi Michal,
>
> On Monday 29 October 2007, CIJOML wrote:
> > Dne čt 22. února 2007 Dmitry Torokhov napsal(a):
> > > Plus I
> > > will export uniq in /proc/bus/input/devices and then yo can ask X guys
> > > to allow matching on uniq as well.
> >
> > Hi Dmitri,
> >
> > I watched at 2.6.23 and this is still not done yet. Are there any patches
> > available to use???
>
> Input core does export uniq in /proc/bus/input/devices so now it is
> up to individual drivers to populate this field in input_dev
> structure.