Use of PS/2 mouse breaks IRQ link that was set up manually

[Lars Erdmann]

For this problem: this system has 2 onboard PS/2 ports (one for mouse, one for keyboard). It also has 3 onboard OHCI USB controllers. 1.) One of the onboard OHCI USB controllers was "dead". That was because its PCI interrupt pin was not properly hooked up to any IRQ line. The USB controller in question is tied to LINK LNKE. I specified this in config.sys to rectify the situation: PSD=ACPI.PSD /OS="Microsoft Windows" /EIS /LNK:LNKE:5 which fixed the problem.

2.) I now plugged in a PS/2 mouse (instead of the USB mouse I normally use). That "kills" the setup of LNKE so that again the affected OHCI USB controller becomes inoperational.

What I already tested: I have this setting my BIOS: a.) "PS/2 Mouse wakeup from S3". I toggled between ON and OFF but it did not make a difference. b.) I used the original IBM 10.162 drivers, then I tried my updated USB drivers (work in progress). However the behaviour for the described problem is exactly the same.

Find attached the "DSDT" and the "FACP" tables of my system.

[Lars Erdmann]

DSDT table

[Lars Erdmann]

FACP table

[Lars Erdmann]

ACPI.PSD log file

[pasha]

Please download experimental ACPI build from Mensys site:

• Experimental build for you:

acpi319ForDesktop.zip

• Read instructions: ​http://ecomstation.ru/projects/acpitools/?action=testcase

[Lars Erdmann]

ACPI log file with ACPI319ForDesktop.zip but unchanged config.sys

[Lars Erdmann]

ACPI log file with ACPI319ForDesktop.zip with LNKG set

[Lars Erdmann]

ACPI log file with ACPI319ForDesktop.zip with LNKB, LNKE, LNKG set

[Lars Erdmann]

ACPI log file with ACPI319ForDesktop.zip with no LINK manually set

[Lars Erdmann]

I now tried acpi310ForDesktop. Initially, the behaviour was completely the same. However, this is what I observed:
1.) with this command line:
PSD=ACPI.PSD /OS="Microsoft Windows" /EIS /LNK:LNKE:5
LNKE would be initially correctly set to IRQ5 but the IRQ setting would then "wander" to LNKG. See attached acpica2.log.
2.) with this command line:
PSD=ACPI.PSD /OS="Microsoft Windows" /EIS /LNK:LNKG:5
the problematic USB HC would then properly work again. See attached acpica3.log.
On the other hand, if LNKE is not set it will be linked to IRQ9 which is the ACPI IRQ. Consequently I get a bunch of "Interrupt not handled" messages in acpica3.log. Funny enough ALL OHCI HCs will work correctly. I tried other variations, see acpica4.log and acpica5.log.
How can we prevent this from happening ? Changing config.sys or acpi.cfg when I plug in a PS/2 mouse is not an option. I also see no apparent logic of how a user would need to set up the LINKS if they do not properly work right out of the box.

[pasha]

can you attach config.sys?

[Lars Erdmann]

Current config.sys

[Lars Erdmann]

Find config.sys attached. As you can see I tried different variations. Specifying /APIC eliminated the problem (I seem to remember, I would have to retest) but was no option to me because it broke DOS window/DOS fullscreen support.

[pasha]

you need check mouse.sys for SMP ready

[Lars Erdmann]

It's the IBM scroll mouse driver. And this mouse driver is SMP ready. It is built with the "smp" flag set. The effect of this is that DevHelp_SetIRQMask is used to mask/unmask interrupts (and not by directly programming the 8259 interrupt controller).

Or what are you looking for that needs to be "SMP ready" ?

[pasha]

mouse.sys from eCS 2.0 GA. Or try this mouse.sys

[Lars Erdmann]

I am closing this bug as there is no longer a need to manually set up IRQ links.