KVM & mouse wheel

KVM & mouse wheel

[Marko Macek]

[-- Attachment #1: Type: text/plain, Size: 1927 bytes --]

Hello!

A few months ago I posted about problems with 2.6 kernel, KVM and mouse
wheel.

I was using 2.4 kernel until recently, but with the switch to FC2 with
2.6 kernel this problem became much more annoying.

My mouse is Logitech MX 510.

I figured out a few things.

1. Trying to set the mouse/kvm into a stream mode makes things insane.
Since streaming mode is supposed to be the default, I propose not
doing this at all. I haven't researched this further.

-      psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);

2. synaptics_detect hoses imps and exps detection. Resetting the mouse
after failed detect fixes it. This makes 'imps' and 'exps' protocols
work when used as proto=imps or proto=exps. Wheel works, I haven't tried
the buttons.

3. PS2++ detection correctly detects Logitech MX mouse but doesn't
enable the PS2PP protocol, because of unexpected results in this code:

	param[0] = param[1] = param[2] = 0;
         ps2pp_cmd(psmouse, param, 0x39); /* Magic knock */
         ps2pp_cmd(psmouse, param, 0xDB);

         if ((param[0] & 0x78) == 0x48 &&
             (param[1] & 0xf3) == 0xc2 &&
             (param[2] & 0x03) == ((param[1] >> 2) & 3)) {
                 ps2pp_set_smartscroll(psmouse);
	        protocol = PSMOUSE_PS2PP;
         }

The returned param array in my case is: 08 01 00 or 08 00 00 (hex)
(without KVM: C8 C2 64)

I don't understand what this code is trying to check or why the protocol
is only set conditionally. If I set it unconditionally (swap last 2
lines) the PS2++ protocol now works including detection of all buttons
(I don't really need the buttons, just the wheel).

This is not included in the patch. The alternative solution
is to reset the mouse again and resume probing for imps or exps.


I plan on using the mouse in 'exps' mode since I haven't yet determined
out how the PS2++ mode will cooperate with the other OS.

Please consider the patch.

Regards,
Mark


[-- Attachment #2: linux-2.6.7-kvm-mouse.patch --]
[-- Type: text/x-patch, Size: 1996 bytes --]

--- psmouse-base.c.orig	2004-06-16 07:19:01.000000000 +0200
+++ psmouse-base.c	2004-08-03 17:21:24.948668832 +0200
@@ -418,28 +418,36 @@
 /*
  * Try Synaptics TouchPad
  */
-	if (max_proto > PSMOUSE_PS2 && synaptics_detect(psmouse)) {
-		synaptics_hardware = 1;
-
-		if (set_properties) {
-			psmouse->vendor = "Synaptics";
-			psmouse->name = "TouchPad";
-		}
-
-		if (max_proto > PSMOUSE_IMEX) {
-			if (!set_properties || synaptics_init(psmouse) == 0)
-				return PSMOUSE_SYNAPTICS;
+	if (max_proto > PSMOUSE_PS2) {
+		if (synaptics_detect(psmouse)) {
+			synaptics_hardware = 1;
+
+			if (set_properties) {
+				psmouse->vendor = "Synaptics";
+				psmouse->name = "TouchPad";
+			}
+
+			if (max_proto > PSMOUSE_IMEX) {
+				if (!set_properties || synaptics_init(psmouse) == 0)
+					return PSMOUSE_SYNAPTICS;
 /*
  * Some Synaptics touchpads can emulate extended protocols (like IMPS/2).
  * Unfortunately Logitech/Genius probes confuse some firmware versions so
  * we'll have to skip them.
  */
-			max_proto = PSMOUSE_IMEX;
-		}
+				max_proto = PSMOUSE_IMEX;
+			}
 /*
  * Make sure that touchpad is in relative mode, gestures (taps) are enabled
  */
-		synaptics_reset(psmouse);
+			synaptics_reset(psmouse);
+		} else {
+/* 
+ * Reset mouse if synaptics detect failed (KVM switch problem)
+ */
+			psmouse_reset(psmouse);
+			psmouse_command(psmouse, NULL, PSMOUSE_CMD_RESET_DIS);
+		}
 	}
 
 	if (max_proto > PSMOUSE_IMEX && genius_detect(psmouse)) {
@@ -459,6 +467,10 @@
 		int type = ps2pp_init(psmouse, set_properties);
 		if (type > PSMOUSE_PS2)
 			return type;
+		else {
+			psmouse_reset(psmouse);
+			psmouse_command(psmouse, NULL, PSMOUSE_CMD_RESET_DIS);
+		}
 	}
 
 	if (max_proto >= PSMOUSE_IMEX && im_explorer_detect(psmouse)) {
@@ -589,12 +601,6 @@
 		psmouse_set_resolution(psmouse);
 		psmouse_command(psmouse,  NULL, PSMOUSE_CMD_SETSCALE11);
 	}
-
-/*
- * We set the mouse into streaming mode.
- */
-
-	psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);
 }
 
 /*

Re: KVM & mouse wheel

[Dmitry Torokhov]

Hi,

On Tuesday 03 August 2004 10:26 am, Marko Macek wrote:
> Hello!
> 
> A few months ago I posted about problems with 2.6 kernel, KVM and mouse
> wheel.
> 
> I was using 2.4 kernel until recently, but with the switch to FC2 with
> 2.6 kernel this problem became much more annoying.
> 
> My mouse is Logitech MX 510.
> 
> I figured out a few things.
> 
> 1. Trying to set the mouse/kvm into a stream mode makes things insane.
> Since streaming mode is supposed to be the default, I propose not
> doing this at all. I haven't researched this further.
> 
> -      psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);

Could you describe what insane mean? If you take the KVM out of the picture
is the mouse still instane?

> 
> 2. synaptics_detect hoses imps and exps detection. Resetting the mouse
> after failed detect fixes it. This makes 'imps' and 'exps' protocols
> work when used as proto=imps or proto=exps. Wheel works, I haven't tried
> the buttons.
> 

Again, does it work without the KVM?

> 3. PS2++ detection correctly detects Logitech MX mouse but doesn't
> enable the PS2PP protocol, because of unexpected results in this code:
> 
> 	param[0] = param[1] = param[2] = 0;
>          ps2pp_cmd(psmouse, param, 0x39); /* Magic knock */
>          ps2pp_cmd(psmouse, param, 0xDB);
> 
>          if ((param[0] & 0x78) == 0x48 &&
>              (param[1] & 0xf3) == 0xc2 &&
>              (param[2] & 0x03) == ((param[1] >> 2) & 3)) {
>                  ps2pp_set_smartscroll(psmouse);
> 	        protocol = PSMOUSE_PS2PP;
>          }
> 
> The returned param array in my case is: 08 01 00 or 08 00 00 (hex)
> (without KVM: C8 C2 64)
> 
> I don't understand what this code is trying to check or why the protocol
> is only set conditionally. If I set it unconditionally (swap last 2
> lines) the PS2++ protocol now works including detection of all buttons
> (I don't really need the buttons, just the wheel).
> 

Apparently your KVM doctors the data stream from the mouse. The driver
tries to play safe and only switches to PS2++ protocol if mouse responds
properly, otherwise there is a chance that it uses PS2++ with mouse that
does not actually support it. 

> This is not included in the patch. The alternative solution
> is to reset the mouse again and resume probing for imps or exps.
> 

It will be probed for imps/exps if PS2++ fails. Now I suspect that your
particular KVM does not expect any extended probes and gets confused by
them.

-- 
Dmitry

Re: KVM & mouse wheel [was PATCH]

[Marko Macek]

Dmitry Torokhov wrote:

>Hi,
>
>On Tuesday 03 August 2004 10:26 am, Marko Macek wrote:
>  
>
>>Hello!
>>
>>A few months ago I posted about problems with 2.6 kernel, KVM and mouse
>>wheel.
>>
>>I was using 2.4 kernel until recently, but with the switch to FC2 with
>>2.6 kernel this problem became much more annoying.
>>
>>My mouse is Logitech MX 510.
>>
>>I figured out a few things.
>>
>>1. Trying to set the mouse/kvm into a stream mode makes things insane.
>>Since streaming mode is supposed to be the default, I propose not
>>doing this at all. I haven't researched this further.
>>
>>-      psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);
>>    
>>
>
>Could you describe what insane mean? If you take the KVM out of the picture
>is the mouse still instane?
>  
>
Insane means that the mouse is moving really slowly, jumping around and 
buttons
work erratically. I did some more experiments and it seems that if I put 
this command
before the resolution/rate/scaling setting it seems to do no harm (the 
mouse works perfectly).
I'd still prefer to remove it since I see no need for it.

>>2. synaptics_detect hoses imps and exps detection. Resetting the mouse
>>after failed detect fixes it. This makes 'imps' and 'exps' protocols
>>work when used as proto=imps or proto=exps. Wheel works, I haven't tried
>>the buttons.
>>
>>    
>>
>
>Again, does it work without the KVM?
>  
>
Yes, without the KVM the mouse works perfectly.

>  
>
>>3. PS2++ detection correctly detects Logitech MX mouse but doesn't
>>enable the PS2PP protocol, because of unexpected results in this code:
>>
>>	param[0] = param[1] = param[2] = 0;
>>         ps2pp_cmd(psmouse, param, 0x39); /* Magic knock */
>>         ps2pp_cmd(psmouse, param, 0xDB);
>>
>>         if ((param[0] & 0x78) == 0x48 &&
>>             (param[1] & 0xf3) == 0xc2 &&
>>             (param[2] & 0x03) == ((param[1] >> 2) & 3)) {
>>                 ps2pp_set_smartscroll(psmouse);
>>	        protocol = PSMOUSE_PS2PP;
>>         }
>>
>>The returned param array in my case is: 08 01 00 or 08 00 00 (hex)
>>(without KVM: C8 C2 64)
>>
>>I don't understand what this code is trying to check or why the protocol
>>is only set conditionally. If I set it unconditionally (swap last 2
>>lines) the PS2++ protocol now works including detection of all buttons
>>(I don't really need the buttons, just the wheel).
>>
>>    
>>
>
>Apparently your KVM doctors the data stream from the mouse. The driver
>tries to play safe and only switches to PS2++ protocol if mouse responds
>properly, otherwise there is a chance that it uses PS2++ with mouse that
>does not actually support it. 
>
>  
>
Yeah, that's why I decided to fallback to im/exps modes instead.

I have now also successfuly used the extra mouse buttons with 'exps' 
protocol.
So only the 'task' button is not supported with my previous patch (because
it requires PS2++ mode).

>>This is not included in the patch. The alternative solution
>>is to reset the mouse again and resume probing for imps or exps.
>>
>>    
>>
>
>It will be probed for imps/exps if PS2++ fails. Now I suspect that your
>particular KVM does not expect any extended probes and gets confused by
>them.
>  
>
I have also seen comments in X sources about needing to reset the mouse 
after
each unsuccessful probe for best compatibility.

Mark

Re: KVM & mouse wheel

[Marko Macek]

Dmitry Torokhov wrote:

 >Hi,
 >
 >On Tuesday 03 August 2004 10:26 am, Marko Macek wrote:
 >
 >>Hello!
 >>
 >>A few months ago I posted about problems with 2.6 kernel, KVM and mouse
 >>wheel.
 >>
 >>I was using 2.4 kernel until recently, but with the switch to FC2 with
 >>2.6 kernel this problem became much more annoying.
 >>
 >>My mouse is Logitech MX 510.
 >>
 >>I figured out a few things.
 >>
 >>1. Trying to set the mouse/kvm into a stream mode makes things insane.
 >>Since streaming mode is supposed to be the default, I propose not
 >>doing this at all. I haven't researched this further.
 >>
 >>-      psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);
 >
 >
 >Could you describe what insane mean? If you take the KVM out of the 
picture
 >is the mouse still instane?
 >
Insane means that the mouse is moving really slowly, jumping around and 
buttons
work erratically. I did some more experiments and it seems that if I put 
this command
before the resolution/rate/scaling setting it seems to do no harm (the 
mouse works perfectly).
I'd still prefer to remove it since I see no need for it.

 >>2. synaptics_detect hoses imps and exps detection. Resetting the mouse
 >>after failed detect fixes it. This makes 'imps' and 'exps' protocols
 >>work when used as proto=imps or proto=exps. Wheel works, I haven't tried
 >>the buttons.
 >>
 >
 >Again, does it work without the KVM?
 >
Yes, without the KVM the mouse works perfectly.

 >
 >>3. PS2++ detection correctly detects Logitech MX mouse but doesn't
 >>enable the PS2PP protocol, because of unexpected results in this code:
 >>
 >>    param[0] = param[1] = param[2] = 0;
 >>         ps2pp_cmd(psmouse, param, 0x39); /* Magic knock */
 >>         ps2pp_cmd(psmouse, param, 0xDB);
 >>
 >>         if ((param[0] & 0x78) == 0x48 &&
 >>             (param[1] & 0xf3) == 0xc2 &&
 >>             (param[2] & 0x03) == ((param[1] >> 2) & 3)) {
 >>                 ps2pp_set_smartscroll(psmouse);
 >>            protocol = PSMOUSE_PS2PP;
 >>         }
 >>
 >>The returned param array in my case is: 08 01 00 or 08 00 00 (hex)
 >>(without KVM: C8 C2 64)
 >>
 >>I don't understand what this code is trying to check or why the protocol
 >>is only set conditionally. If I set it unconditionally (swap last 2
 >>lines) the PS2++ protocol now works including detection of all buttons
 >>(I don't really need the buttons, just the wheel).
 >>
 >
 >Apparently your KVM doctors the data stream from the mouse. The driver
 >tries to play safe and only switches to PS2++ protocol if mouse responds
 >properly, otherwise there is a chance that it uses PS2++ with mouse that
 >does not actually support it.
 >
Yeah, that's why I decided to fallback to im/exps modes instead.

I have now also successfuly used the extra mouse buttons with 'exps' 
protocol.
So only the 'task' button is not supported with my previous patch (because
it requires PS2++ mode).

 >>This is not included in the patch. The alternative solution
 >>is to reset the mouse again and resume probing for imps or exps.
 >>
 >
 >It will be probed for imps/exps if PS2++ fails. Now I suspect that your
 >particular KVM does not expect any extended probes and gets confused by
 >them.
 >
I have also seen comments in X sources about needing to reset the mouse 
after
each unsuccessful probe for best compatibility.

Mark

Re: KVM & mouse wheel

[Jesper Juhl]

On Tue, 3 Aug 2004, Marko Macek wrote:

> Hello!
> A few months ago I posted about problems with 2.6 kernel, KVM and mouse
> wheel.
> I was using 2.4 kernel until recently, but with the switch to FC2 with
> 2.6 kernel this problem became much more annoying.

I also had problems with my KVM switch and mouse when I initially moved to 
2.6, but adding this kernel boot parameter fixed it, meybe it will help 
you as well :  psmouse.proto=imps

My mouse was jumping all over the screen, clicking buttons at random, but 
since I added 
append="psmouse.proto=imps"
to my lilo.conf everything has been working perfectly with every 2.6 
kernel I've tried.

My mouse is a "Logitec MouseMan Wheel" (USB mouse, but connected to the 
KVM with a USB->PS/2 adapter), my KVM switch is an old 8port thing that I 
unfortunately don't know the name/brand of since I got it used and the 
label with that info on it is gone.

 Hope that helps you :)


--
Jesper Juhl <juhl-lkml@dif.dk>

Re: KVM & mouse wheel

[David Ford]

[-- Attachment #1: Type: text/plain, Size: 3154 bytes --]

I have a customer using a KVM device, the mouse works fine -until- the 
KVM is switched elsewhere and back.  After that, the mouse sends 
garbage.  X+Y movement data is completely hosed causing the mouse to 
jump around everywhere.  If I remove the module and reload it, the mouse 
partially returns.  X+Y movement is normal again however the buttons are 
now messed up.  Sometimes the buttons seem constantly pressed or 
constantly released.  The only way to recover normal mousing is to reboot.

This didn't happen back on 2.6.0-test8 and I don't think it happened on 
2.6.7.

David

Dmitry Torokhov wrote:

>Hi,
>
>On Tuesday 03 August 2004 10:26 am, Marko Macek wrote:
>  
>
>>Hello!
>>
>>A few months ago I posted about problems with 2.6 kernel, KVM and mouse
>>wheel.
>>
>>I was using 2.4 kernel until recently, but with the switch to FC2 with
>>2.6 kernel this problem became much more annoying.
>>
>>My mouse is Logitech MX 510.
>>
>>I figured out a few things.
>>
>>1. Trying to set the mouse/kvm into a stream mode makes things insane.
>>Since streaming mode is supposed to be the default, I propose not
>>doing this at all. I haven't researched this further.
>>
>>-      psmouse_command(psmouse, param, PSMOUSE_CMD_SETSTREAM);
>>    
>>
>
>Could you describe what insane mean? If you take the KVM out of the picture
>is the mouse still instane?
>
>  
>
>>2. synaptics_detect hoses imps and exps detection. Resetting the mouse
>>after failed detect fixes it. This makes 'imps' and 'exps' protocols
>>work when used as proto=imps or proto=exps. Wheel works, I haven't tried
>>the buttons.
>>
>>    
>>
>
>Again, does it work without the KVM?
>
>  
>
>>3. PS2++ detection correctly detects Logitech MX mouse but doesn't
>>enable the PS2PP protocol, because of unexpected results in this code:
>>
>>	param[0] = param[1] = param[2] = 0;
>>         ps2pp_cmd(psmouse, param, 0x39); /* Magic knock */
>>         ps2pp_cmd(psmouse, param, 0xDB);
>>
>>         if ((param[0] & 0x78) == 0x48 &&
>>             (param[1] & 0xf3) == 0xc2 &&
>>             (param[2] & 0x03) == ((param[1] >> 2) & 3)) {
>>                 ps2pp_set_smartscroll(psmouse);
>>	        protocol = PSMOUSE_PS2PP;
>>         }
>>
>>The returned param array in my case is: 08 01 00 or 08 00 00 (hex)
>>(without KVM: C8 C2 64)
>>
>>I don't understand what this code is trying to check or why the protocol
>>is only set conditionally. If I set it unconditionally (swap last 2
>>lines) the PS2++ protocol now works including detection of all buttons
>>(I don't really need the buttons, just the wheel).
>>
>>    
>>
>
>Apparently your KVM doctors the data stream from the mouse. The driver
>tries to play safe and only switches to PS2++ protocol if mouse responds
>properly, otherwise there is a chance that it uses PS2++ with mouse that
>does not actually support it. 
>
>  
>
>>This is not included in the patch. The alternative solution
>>is to reset the mouse again and resume probing for imps or exps.
>>
>>    
>>
>
>It will be probed for imps/exps if PS2++ fails. Now I suspect that your
>particular KVM does not expect any extended probes and gets confused by
>them.
>
>  
>

[-- Attachment #2: david+challenge-response.vcf --]
[-- Type: text/x-vcard, Size: 194 bytes --]

begin:vcard
fn:David Ford
n:Ford;David
email;internet:david@blue-labs.org
title:Industrial Geek
tel;home:Ask please
tel;cell:(203) 650-3611
x-mozilla-html:TRUE
version:2.1
end:vcard

Re: KVM & mouse wheel

[Marko Macek]

Jesper Juhl wrote:

> <>I also had problems with my KVM switch and mouse when I initially 
> moved to
> 2.6, but adding this kernel boot parameter fixed it, meybe it will help
> you as well : psmouse.proto=imps

This doesn't help. Only the patch I sent helps me. The problem is that the
even with psmouse.proto=imps or exps, the driver still probes for 
synaptics which I
consider a bug.

>My mouse is a "Logitec MouseMan Wheel" (USB mouse, but connected to the 
>
I have this one too (besides MX), same problem. The problem is really the
KVM switch and the 2.6 mouse driver.

Mark

Re: KVM & mouse wheel

[Dmitry Torokhov]

On Tuesday 03 August 2004 11:29 pm, Marko Macek wrote:
> Jesper Juhl wrote:
> 
> > <>I also had problems with my KVM switch and mouse when I initially 
> > moved to
> > 2.6, but adding this kernel boot parameter fixed it, meybe it will help
> > you as well : psmouse.proto=imps
> 
> This doesn't help. Only the patch I sent helps me. The problem is that the
> even with psmouse.proto=imps or exps, the driver still probes for 
> synaptics which I
> consider a bug.
> 

No it is not - Synaptics with a track-point on a passthrough port will have
track-point disabled if it is not reset after probing for imps/exps.

-- 
Dmitry

Re: KVM & mouse wheel

[Vojtech Pavlik]

On Wed, Aug 04, 2004 at 12:25:19AM -0500, Dmitry Torokhov wrote:

> On Tuesday 03 August 2004 11:29 pm, Marko Macek wrote:
> > Jesper Juhl wrote:
> > 
> > > <>I also had problems with my KVM switch and mouse when I initially 
> > > moved to
> > > 2.6, but adding this kernel boot parameter fixed it, meybe it will help
> > > you as well : psmouse.proto=imps
> > 
> > This doesn't help. Only the patch I sent helps me. The problem is that the
> > even with psmouse.proto=imps or exps, the driver still probes for 
> > synaptics which I
> > consider a bug.
> > 
> 
> No it is not - Synaptics with a track-point on a passthrough port will have
> track-point disabled if it is not reset after probing for imps/exps.
 
Hmm, does the imps/exps probe succeed in this case?

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR

Re: KVM & mouse wheel

[Dmitry Torokhov]

On Wednesday 04 August 2004 02:18 am, Vojtech Pavlik wrote:
> On Wed, Aug 04, 2004 at 12:25:19AM -0500, Dmitry Torokhov wrote:
> 
> > On Tuesday 03 August 2004 11:29 pm, Marko Macek wrote:
> > > Jesper Juhl wrote:
> > > 
> > > > <>I also had problems with my KVM switch and mouse when I initially 
> > > > moved to
> > > > 2.6, but adding this kernel boot parameter fixed it, meybe it will help
> > > > you as well : psmouse.proto=imps
> > > 
> > > This doesn't help. Only the patch I sent helps me. The problem is that the
> > > even with psmouse.proto=imps or exps, the driver still probes for 
> > > synaptics which I
> > > consider a bug.
> > > 
> > 
> > No it is not - Synaptics with a track-point on a passthrough port will have
> > track-point disabled if it is not reset after probing for imps/exps.
>  
> Hmm, does the imps/exps probe succeed in this case?
> 

No, it does not, at least not mine. It either does bare PS/2 or native, but
there are other Synaptics touchpads that can also do imps.

-- 
Dmitry

Re: KVM & mouse wheel

[Vojtech Pavlik]

On Wed, Aug 04, 2004 at 07:38:55AM -0500, Dmitry Torokhov wrote:
> On Wednesday 04 August 2004 02:18 am, Vojtech Pavlik wrote:
> > On Wed, Aug 04, 2004 at 12:25:19AM -0500, Dmitry Torokhov wrote:
> > 
> > > On Tuesday 03 August 2004 11:29 pm, Marko Macek wrote:
> > > > Jesper Juhl wrote:
> > > > 
> > > > > <>I also had problems with my KVM switch and mouse when I initially 
> > > > > moved to
> > > > > 2.6, but adding this kernel boot parameter fixed it, meybe it will help
> > > > > you as well : psmouse.proto=imps
> > > > 
> > > > This doesn't help. Only the patch I sent helps me. The problem is that the
> > > > even with psmouse.proto=imps or exps, the driver still probes for 
> > > > synaptics which I
> > > > consider a bug.
> > > > 
> > > 
> > > No it is not - Synaptics with a track-point on a passthrough port will have
> > > track-point disabled if it is not reset after probing for imps/exps.
> >  
> > Hmm, does the imps/exps probe succeed in this case?
> 
> No, it does not, at least not mine. It either does bare PS/2 or native, but
> there are other Synaptics touchpads that can also do imps.
 
Ok, so how about issuing a reset when the imps probe fails? That'd take
care of all the cases, and I suppose a Synaptics pad that can do imps
will not be confused by it.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR