Re: Problem with keyboard generating events persists in 2.6.7

[<ai...@ba...>]

Quoting Kjetil Kjernsmo <kj...@kj...>:

> Hi!
>=20
> I'm trying to set aside some time to grok the patch you sent me...
>=20
> For one thing, I have now produced a set of logs and output from variou=
s=20
> parts of the proc fs with the current running kernel, without your=20
> patch. They are now at
> http://dev.kjernsmo.net/tmp/ruby-only-logs1/
> This stuff was produced just after the keyboard started generating=20
> events.=20

Unfortunately that is hardware bug. So nobody can locate the bug
remotely. You must do it by yourself.

> To quote the patch:
>=20
> On 9. july 2004, 10:32, Aivils wrote:
>=20
> > IMHO You should do gross hack by yourself
> >
> > --- hid-core.c  2004-06-17 10:37:47.000000000 +0300
> > +++ hid-core_.c 2004-07-09 11:12:43.045096352 +0300
> > @@ -1491,7 +1491,7 @@ static struct hid_blacklist {
> >         { USB_VENDOR_ID_ATEN, USB_DEVICE_ID_ATEN_2PORTKVM,
> > HID_QUIRK_NOGET }, { USB_VENDOR_ID_ATEN, USB_DEVICE_ID_ATEN_4PORTKVM,
> > HID_QUIRK_NOGET }, { USB_VENDOR_ID_ATEN,
> > USB_DEVICE_ID_ATEN_4PORTKVMC, HID_QUIRK_NOGET }, -       {
> > USB_VENDOR_ID_TANGTOP, USB_DEVICE_ID_TANGTOP_USBPS2, HID_QUIRK_NOGET
> > }, +       { USB_VENDOR_ID_TANGTOP, USB_DEVICE_ID_TANGTOP_USBPS2,
> > HID_QUIRK_NOGET | HID_QUIRK_STOP_TANGTOP},
> >
> >         { USB_VENDOR_ID_A4TECH, USB_DEVICE_ID_A4TECH_WCP32PU,
> > HID_QUIRK_2WHEEL_MOUSE_HACK_BACK }, { USB_VENDOR_ID_CYPRESS,
> > USB_DEVICE_ID_CYPRESS_MOUSE, HID_QUIRK_2WHEEL_MOUSE_HACK_EXTRA },
> >
> >
> >
> >
> >
> >
> > --- hid-input.c 2004-05-10 05:32:28.000000000 +0300
> > +++ hid-input_.c        2004-07-09 11:23:21.278070192 +0300
> > @@ -432,6 +432,13 @@ void hidinput_hid_event(struct hid_devic
> >
> >         input_regs(input, regs);
> >
> > +       /* debug wrong TANGTOP event */
> > +       printk("HID_EVENT: type %d code %d \n", usage->type,
> > usage->code); +       /* stop wrong event */
>=20
>=20
> So, this is going to print some kind of code that I can use?
>=20
> > +       if ((hid->quirks & HID_QUIRK_STOP_TANGTOP) &&
> > +            usage->type =3D=3D UNDESIRED_TYPE && usage->code =3D=3D
> > UNDESIRED_CODE) +               return;
> > +
>=20
> And this will "stop" the TANGTOP if I specify the type and code?=20
>=20
> But what will it mean to "stop"? Will it mean that the keyboard and=20
> mouse freezes rather than spewing out events? Or will it reset itself,=20
> or....? =20

TANGTOP device driver is marked by "HID_QUIRK_STOP_TANGTOP".
If device produce spontaneus keypress events , then my patch is not=20
usable. But TANGTOP may produce recongnizable "wrong" events,
which You stop.

> >         if (((hid->quirks & HID_QUIRK_2WHEEL_MOUSE_HACK_EXTRA) &&
> > (usage->code =3D=3D BTN_EXTRA))
> >
> >                 || ((hid->quirks & HID_QUIRK_2WHEEL_MOUSE_HACK_BACK)
> >                 || && (usage->code =3D=3D BTN_BACK))) {
> >
> >                 if (value)
> >
> >
> >
> > --- hid.h       2004-05-10 05:31:57.000000000 +0300
> > +++ hid_.h      2004-07-09 11:11:41.563442984 +0300
> > @@ -211,6 +211,7 @@ struct hid_item {
> >  #define HID_QUIRK_2WHEEL_MOUSE_HACK_BACK       0x080
> >  #define HID_QUIRK_2WHEEL_MOUSE_HACK_EXTRA      0x100
> >  #define HID_QUIRK_2WHEEL_MOUSE_HACK_ON         0x200
> > +#define HID_QUIRK_STOP_TANGTOP                 0x400
> >
> >  /*
> >   * This is the global environment of the parser. This information is
> >
> >
> >
>=20
>=20
>=20
> > This set of paches means You locate wrong event and leater stop it
> > with # define UNDESIRED_TYPE
> > # define UNDESIRED_CODE
>=20
> Right.... :-)=20
>=20
> So, I should apply this patch, recompile the kernel. Then reproduce the=
=20
> spewing of events, locate the error message beginning with HID_EVENT:
> (in /var/log/messages?), and find the type and code, set the=20
> UNDESIRED_TYPE and UNDESIRED_CODE with them, recompile once more.
>=20
> Have I got that right?=20

Right!

> What would happen then, when the error is encountered?
>=20

You should parry death-blow :)

Aivils

---------------------------------------------------------------------
Radical ringtones, java games, mobile phone logos, backgrounds http://sms=
.BANDA.LV !