HP nx8220 psmouse vs ACPI at boot up

HP nx8220 psmouse vs ACPI at boot up

[Thomas Renninger]

Hi,

I first thought it's the "unload psmouse" or "cleanup psmouse ports"
at shutdown problem again.

I think this time EC gets confused at boot time.
Compiling psmouse as module helps again.

Symptom:
AC adapter status is wrong if booted with AC unplugged.
Replugging helps.
This sounds like a minor bug, but ACPI errors at other places
with AE_TIME errors also happen and theoretically all kind of
ACPI breakage could happen...

I can reproduce this with a 2.6.18 and a 2.6.16 kernel.
Latest 2.6.21-rc7 does not show the problem.
I wonder whether this really got fixed or whether this
is just because psmouse gets activated at some earlier or later
time because of other changes in the kernel.

This looks a bit related to the 
"Could the k8temp driver be interfering with ACPI"
discussion. It seems as if mouse ports should not get touched
when ACPI is accessing the EC on HP laptops (and others?) in general?

I can't access the machine anymore, I possibly can have it for
some more tests.
I just want to point out that there still could be trouble.

   Thomas

I thought that attached patch helps (exclude psmouse init and ACPI
interpreter). But it seems that not.
Hmm, so strange, I removed processor, fan and thermal module to not be
loaded via initrd -> works. However adding them again still works, maybe
they are loaded in different order. Really confused now, as this problem
is a BIOS problem and what I see cannot get cleanly workarounded I am
giving up.

---
 drivers/input/mouse/psmouse-base.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)

Index: linux-2.6.18-SL102_BRANCH/drivers/input/mouse/psmouse-base.c
===================================================================
--- linux-2.6.18-SL102_BRANCH.orig/drivers/input/mouse/psmouse-base.c
+++ linux-2.6.18-SL102_BRANCH/drivers/input/mouse/psmouse-base.c
@@ -22,6 +22,8 @@
 #include <linux/libps2.h>
 #include <linux/mutex.h>
 
+#include <acpi/acpi.h>
+
 #include "psmouse.h"
 #include "synaptics.h"
 #include "logips2pp.h"
@@ -1513,6 +1515,14 @@ static int psmouse_get_maxproto(char *bu
 
 static int __init psmouse_init(void)
 {
+	acpi_status status;
+
+	status = acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+	if (status) {
+		printk (KERN_ERR "%s - Could not acquire ACPI interpreter"
+			" mutex\n", __FUNCTION__);
+	}
+
 	kpsmoused_wq = create_singlethread_workqueue("kpsmoused");
 	if (!kpsmoused_wq) {
 		printk(KERN_ERR "psmouse: failed to create kpsmoused workqueue\n");
@@ -1521,6 +1531,12 @@ static int __init psmouse_init(void)
 
 	serio_register_driver(&psmouse_drv);
 
+	status = acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+	if (status) {
+		printk (KERN_ERR "%s - Could not release ACPI interpreter"
+			" mutex\n", __FUNCTION__);
+	}
+
 	return 0;
 }
 

Re: HP nx8220 psmouse vs ACPI at boot up

[Dmitry Torokhov]

Hi Thomas,

On 4/19/07, Thomas Renninger <trenn@suse.de> wrote:
> Hi,
>
> I first thought it's the "unload psmouse" or "cleanup psmouse ports"
> at shutdown problem again.
>
> I think this time EC gets confused at boot time.
> Compiling psmouse as module helps again.
>
> Symptom:
> AC adapter status is wrong if booted with AC unplugged.
> Replugging helps.
> This sounds like a minor bug, but ACPI errors at other places
> with AE_TIME errors also happen and theoretically all kind of
> ACPI breakage could happen...
>
> I can reproduce this with a 2.6.18 and a 2.6.16 kernel.
> Latest 2.6.21-rc7 does not show the problem.
> I wonder whether this really got fixed or whether this
> is just because psmouse gets activated at some earlier or later
> time because of other changes in the kernel.
>

I don't think so, unless you changed your .config.

> This looks a bit related to the
> "Could the k8temp driver be interfering with ACPI"
> discussion. It seems as if mouse ports should not get touched
> when ACPI is accessing the EC on HP laptops (and others?) in general?
>
> I can't access the machine anymore, I possibly can have it for
> some more tests.
> I just want to point out that there still could be trouble.
>
>   Thomas
>
> I thought that attached patch helps (exclude psmouse init and ACPI
> interpreter). But it seems that not.

You want to take the lock in psmouse_connect() and
psmouse_disconnect() and bunch of other places. Because all probing
work is offloaded to kseriod you did not really change anything when
protecting psmouse_init.

-- 
Dmitry

Re: HP nx8220 psmouse vs ACPI at boot up

[Bjorn Helgaas]

On Thursday 19 April 2007 06:44, Thomas Renninger wrote:
> I first thought it's the "unload psmouse" or "cleanup psmouse ports"
> at shutdown problem again.
> 
> I think this time EC gets confused at boot time.
> Compiling psmouse as module helps again.
> 
> Symptom:
> AC adapter status is wrong if booted with AC unplugged.
> Replugging helps.
> This sounds like a minor bug, but ACPI errors at other places
> with AE_TIME errors also happen and theoretically all kind of
> ACPI breakage could happen...

Is there a more complete bug report for this, with the whole dmesg
log?  Is this related to suspend/resume, or does it happen on all
shutdown and restart?  Any chance you could collect the DSDT?

Bjorn

Re: HP nx8220 psmouse vs ACPI at boot up

[Thomas Renninger]

On Thu, 2007-04-19 at 09:18 -0600, Bjorn Helgaas wrote:
> On Thursday 19 April 2007 06:44, Thomas Renninger wrote:
> > I first thought it's the "unload psmouse" or "cleanup psmouse ports"
> > at shutdown problem again.
> > 
> > I think this time EC gets confused at boot time.
> > Compiling psmouse as module helps again.
> > 
> > Symptom:
> > AC adapter status is wrong if booted with AC unplugged.
> > Replugging helps.
> > This sounds like a minor bug, but ACPI errors at other places
> > with AE_TIME errors also happen and theoretically all kind of
> > ACPI breakage could happen...
> 
> Is there a more complete bug report for this, with the whole dmesg
> log?  Is this related to suspend/resume, or does it happen on all
> shutdown and restart?  Any chance you could collect the DSDT?

This is the bug report:
https://bugzilla.novell.com/show_bug.cgi?id=260794

Klaus: Could you also attach acpidump output to the bug, pls.

This is not related to suspend/resume.

> Does it happen on all shutdown and restart
-> Good question, first I was sure to be able to always reproduce this
on 2.6.1[68] kernels with psmouse module compiled in. Latest .21-rc7 was
unaffected.
I then removed thermal, fan, processor module to not get loaded via
initrd (gets loaded later at init 3 or 2). Since then the symptom
vanished. (even after adding the modules again. Maybe I mixed something
up in the end, or the order of modules loaded in initrd changed, but I
doubt it was that...)

However, if someone should see any AE_TIME ACPI errors on an HP, I'd try
to compile psmouse as module or compile it out, chances are high that it
helps..., I wonder whether more people see this, would be nice to get
some more reports.

    Thomas