Skip to contentSkip to conversation

Workaround i8042 chips with broken MUX mode

PATCH Dmitry Torokhov 4 participants 11 messages Mar 16, 2004 – Mar 28, 2004 1 message not in the archive

lore
Share & tools
lore
#0 lore raw

Review

Quoted 2 lines · Vojtech Pavlik
(quoted diff) +2 -0
+	
+	/* Workaround for broken chips which seem to

support MUX, but in reality don't. */

Review

Quoted 3 lines
(quoted diff) +3 -0
+	/* They all report version 12.10 */
+	if (mux_version == 0xCA)
+		return -1;

Hi,

I think it should be 0xAC (0xA4 with 4th bit flipped) as the version reported is 10.12:

i8042.c: Detected active multiplexing controller, rev 10.12.

Quoted 1 line

From little debug info that I've been sent ThinkPad's

controllers seem to be flipping 4th bit sometimes, I can't quite pinpoint the exact sequence.

Dmitry

P.S. Sorry for breaking the threading...

#1 lore raw
Quoted 13 lines · Dmitry Torokhov

Review

Quoted 2 lines · Vojtech Pavlik
(quoted diff) +2 -0
+	
+	/* Workaround for broken chips which seem to

support MUX, but in reality don't. */

Review

Quoted 3 lines
(quoted diff) +3 -0
+	/* They all report version 12.10 */
+	if (mux_version == 0xCA)
+		return -1;

Hi,

I think it should be 0xAC (0xA4 with 4th bit flipped) as the version reported is 10.12:

Quoted 6 lines

i8042.c: Detected active multiplexing controller, rev 10.12.

From little debug info that I've been sent ThinkPad's controllers seem to be flipping 4th bit sometimes, I can't quite pinpoint the exact sequence.

Could this be the bit that indicates whether the report is coming from an internal or external device?

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw
Quoted 23 lines · Dmitry Torokhov
Quoted 2 lines

Could this be the bit that indicates whether the report is coming from an internal or external device?

That is what I suspect, but I can not prove it and in fact I am quite surprised that KBC would mangle data requested via CMD_AUX_LOOP. So far I have seen the following sequences (* denotes 4th bit flipped):

2.6.1:

IN:    	FO   
OUT:    F8* 

2.6.2-rc1-bk1:

IN:     5A  F0  56  A4 
OUT:    5A  F0  56  AC*

2.6.3 + my debug

IN:     5A  A4  5B  A4  53  F0  56
OUT:    5A  A4  5B  AC* AC  F0  5E*

2.6.3 + my other debug
IN: 5A A4 A4 A4 A4 F0 56 A4 A4 A4 A4
OUT: 5A A4 A4 A4 A4 F0 56 A4 A4 A4 A4

Quoted 6 lines

... and I am waiting on sequence:
5A 00 00 00 00 F0 56 A4 00 00 00 00

As you can see it does not always flip that bit, but if it will mangle the last sequence I think we can claim that we see "hidden multiplexor" and abort active multiplexor test.

So far on every machine I've got a report from it was caused by BIOS emulation of PS/2 mouse using an USB mouse (even when USB mouse wasn't present). Compiling the USB modules into the kernel fixes the problem.

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR
#4 lore raw

Hi.

Quoted 3 lines

So far on every machine I've got a report from it was caused by BIOS emulation of PS/2 mouse using an USB mouse (even when USB mouse wasn't present). Compiling the USB modules into the kernel fixes the problem.

Could this have anything to do with the fact that my x86-64 kernel nukes on startup if USB keyboard/mouse emul is enabled in the BIOS?

This is on an ASUS K8T800 and an MSI K8T800 board.

If you don't know what I'm talking about I'll give more info of course.

// Stefan

#5 lore raw
Quoted 8 lines · Stefan Smietanowski

Hi.

Quoted 3 lines

So far on every machine I've got a report from it was caused by BIOS emulation of PS/2 mouse using an USB mouse (even when USB mouse wasn't present). Compiling the USB modules into the kernel fixes the problem.

Could this have anything to do with the fact that my x86-64 kernel nukes on startup if USB keyboard/mouse emul is enabled in the BIOS?

Yes, and no. USB keyboard/mouse emulation is simply broken in most BIOSes. It's even more broken on AMD64 BIOSes, because usually noone tests its interaction with 64-bit longmode. And thus when it is invoked in 64-bit longmode, it crashes the machine.

On x86 it just causes problems with PS/2 mouse and keyboard.

So it's caused by the same thing, but these are two different problems.

Quoted 3 lines

This is on an ASUS K8T800 and an MSI K8T800 board.

If you don't know what I'm talking about I'll give more info of course.

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR
#6 lore raw
In reply to Dmitry Torokhov
Quoted 13 lines · Dmitry Torokhov

Review

Quoted 2 lines · Vojtech Pavlik
(quoted diff) +2 -0
+	
+	/* Workaround for broken chips which seem to

support MUX, but in reality don't. */

Review

Quoted 3 lines
(quoted diff) +3 -0
+	/* They all report version 12.10 */
+	if (mux_version == 0xCA)
+		return -1;

Hi,

I think it should be 0xAC (0xA4 with 4th bit flipped) as the version reported is 10.12:

Quoted 6 lines

i8042.c: Detected active multiplexing controller, rev 10.12.

From little debug info that I've been sent ThinkPad's controllers seem to be flipping 4th bit sometimes, I can't quite pinpoint the exact sequence.

Thanks, fixed.

Signature · 2 lines
Vojtech Pavlik
SuSE Labs, SuSE CR
#7 lore raw
In reply to Dmitry Torokhov
Quoted 4 lines · Dmitry Torokhov

Review

Quoted 1 line · Vojtech Pavlik
(quoted diff) +1 -0
+	/* Workaround for broken chips which seem to

support MUX, but in reality don't. */

Why call them "broken"? Better to delete that word.

Quoted 6 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+	/* They all report version 12.10 */
+	if (mux_version == 0xCA)
+		return -1;

I think it should be 0xAC (0xA4 with 4th bit flipped) as the version reported is 10.12.

Yes. I have seen one such report. Have there been more?

The Synaptics multiplexing proposal uses 0xf0, 0x56, 0xa4 to activate and 0xf0, 0x56, 0xa5 to deactivate. In both cases the replies must be 0xf0, 0x56, version.

Thus, I suppose one might get a more robust detection by checking that both the activation and deactivation sequences yield the same version.

Andries

#8 lore raw
Quoted 8 lines · Andries Brouwer
Quoted 4 lines · Dmitry Torokhov

Review

Quoted 1 line · Vojtech Pavlik
(quoted diff) +1 -0
+	/* Workaround for broken chips which seem to

support MUX, but in reality don't. */

Why call them "broken"? Better to delete that word.

Quoted 9 lines
Quoted 6 lines

Review

Quoted 3 lines
(quoted diff) +3 -0
+	/* They all report version 12.10 */
+	if (mux_version == 0xCA)
+		return -1;

I think it should be 0xAC (0xA4 with 4th bit flipped) as the version reported is 10.12.

Yes. I have seen one such report. Have there been more?

As it turned out its not the chip but USB legacy emulation that gets in the way of synaptics query. Actually alot of problems were linked to broken legacy emulation implementations, Vojtech mentioned that PCI quirk to turn legacy emulation off may be appropriate.

Quoted 8 lines

The Synaptics multiplexing proposal uses 0xf0, 0x56, 0xa4 to activate and 0xf0, 0x56, 0xa5 to deactivate. In both cases the replies must be 0xf0, 0x56, version.

Thus, I suppose one might get a more robust detection by checking that both the activation and deactivation sequences yield the same version.

Unfortunately in this particular case it looks like something flips 4th bit on some (but not all, like every 3rd) bytes, so it may very well respond with 0xAC to both queries.

Signature · 1 line
Dmitry
#9 lore raw
Quoted 4 lines · Dmitry Torokhov
Quoted 4 lines

The Synaptics multiplexing proposal uses 0xf0, 0x56, 0xa4 to activate and 0xf0, 0x56, 0xa5 to deactivate. In both cases the replies must be 0xf0, 0x56, version.

Quoted 8 lines
Quoted 4 lines

Thus, I suppose one might get a more robust detection by checking that both the activation and deactivation sequences yield the same version.

Unfortunately in this particular case it looks like something flips 4th bit on some (but not all, like every 3rd) bytes, so it may very well respond with 0xAC to both queries.

If one flips that bit on 0xa5 the result is 0xad, not 0xac.

That the bit is set is not strange.
The standard PS/2 protocol requires bit 3 in the first word of every 3-byte packet to be 1.

Andries

#10 lore raw
Quoted 11 lines · Andries Brouwer
Quoted 9 lines · Dmitry Torokhov
Quoted 8 lines

The Synaptics multiplexing proposal uses 0xf0, 0x56, 0xa4 to activate and 0xf0, 0x56, 0xa5 to deactivate. In both cases the replies must be 0xf0, 0x56, version.

Thus, I suppose one might get a more robust detection by checking that both the activation and deactivation sequences yield the same version.

Quoted 5 lines
Quoted 3 lines

Unfortunately in this particular case it looks like something flips 4th bit on some (but not all, like every 3rd) bytes, so it may very well respond with 0xAC to both queries.

If one flips that bit on 0xa5 the result is 0xad, not 0xac.

Yes, you are right.

Quoted 4 lines

That the bit is set is not strange.
The standard PS/2 protocol requires bit 3 in the first word of every 3-byte packet to be 1.

If it did it in every 3rd byte and not in response to loopback requiest but to real data stream coming from a device I would agree...

Signature · 1 line
Dmitry
#11 lore raw
In reply to Dmitry Torokhov

Hi.

Quoted 2 lines

linked to broken legacy emulation implementations, Vojtech mentioned that PCI quirk to turn legacy emulation off may be appropriate.

Is there any code available that can do that as it is now?

This would be useful especially right now under x86-64.

As it is right now I either have the keyboard plugged into both ps/2 and usb, plug it in only into usb (can't operate grub) or plug it only into usb (enable emulation, can't run linux ... ) and if the kernel could have a way of disabling the emulation that would be good at least in my case.

// Stefan