Skip to contentSkip to conversation

2.6.9-rc2-mm1: i8042.c: Can't read CTR while initializing i8042

DISCUSSION Hans-Frieder Vogt 2 participants 2 messages Sep 19, 2004 – Sep 20, 2004

lore
Share & tools
lore
#0 lore raw
Quoted 1 line · Felipe Alfaro Solana

My keyboard has suddenly stopped working with 2.6.9-rc2-mm1-VP-S1 and

2.6.9-rc2-mm1. This is part of the output of dmesg:

Quoted 6 lines

i8042: ACPI [P2KI] at I/O 0x0, 0x0, irq 1
i8042: ACPI [P2MI] at irq 12
i8042.c: Can't read CTR while initializing i8042.

This does happen on 2.6.9-rc2-mm1-VP-S1 and 2.6.9-rc2-mm1 on a NEC Chrom@

laptop, with a 440BX motherboard, Pentium III Mobile and integrated PS/2 keyboard and mouse. It doesn't happen in 2.6.8.1, not does it happen on my Pentium 4 machine, however.

Quoted 2 lines

Any ideas?

I had the same problem on my AMD64 system (MSI K8T Neo board). The reason, why the ioports are not recognised on this board (i.e., why they show up as 0x0, 0x0) is, that the ACPI defines these ports as FixedIO () and not as IO () as expected by the current linux code. FixedIO is perfectly fine and in line with the ACPI standard. So the DSDT table is NOT bad but linux is currently simply not flexible enough to cope with all possible (and allowable) combinations.

With the small attached patch the i8042 IO-ports were recognised on my board (only tested with keyboard).

Voytech, could you include this patch into your next patch set?

--- linux-2.6.9-rc1-mm5.orig/drivers/input/serio/i8042-x86ia64io.h 2004-09-13 15:39:39.061522663 +0200
+++ linux-2.6.9-rc1-mm5/drivers/input/serio/i8042-x86ia64io.h 2004-09-20 01:10:32.124593201 +0200
@@ -105,6 +105,7 @@ static acpi_status i8042_acpi_parse_reso
{
struct i8042_acpi_resources *i8042_res = data;
struct acpi_resource_io *io;
+ struct acpi_resource_fixed_io *fixed_io;
struct acpi_resource_irq *irq;
struct acpi_resource_ext_irq *ext_irq;

@@ -119,6 +120,16 @@ static acpi_status i8042_acpi_parse_reso

    }
    break;

+ case ACPI_RSTYPE_FIXED_IO:
+ fixed_io = &res->data.fixed_io;
+ if (fixed_io->range_length) {
+ if (!i8042_res->port1)
+ i8042_res->port1 = fixed_io->base_address;
+ else
+ i8042_res->port2 = fixed_io->base_address;
+ }
+ break;
+
case ACPI_RSTYPE_IRQ:

    irq = &res->data.irq;
    if (irq->number_of_interrupts > 0)
Signature · 3 lines
--
Hans-Frieder Vogt                 e-mail: hfvogt <at> arcor .dot. de
                                          hfvogt <at> gmx .dot. net
#1 lore raw
Quoted 51 lines · Hans-Frieder Vogt
Quoted 1 line · Felipe Alfaro Solana

My keyboard has suddenly stopped working with 2.6.9-rc2-mm1-VP-S1 and

2.6.9-rc2-mm1. This is part of the output of dmesg:

Quoted 6 lines

i8042: ACPI [P2KI] at I/O 0x0, 0x0, irq 1
i8042: ACPI [P2MI] at irq 12
i8042.c: Can't read CTR while initializing i8042.

This does happen on 2.6.9-rc2-mm1-VP-S1 and 2.6.9-rc2-mm1 on a NEC Chrom@

laptop, with a 440BX motherboard, Pentium III Mobile and integrated PS/2 keyboard and mouse. It doesn't happen in 2.6.8.1, not does it happen on my Pentium 4 machine, however.

Quoted 2 lines

Any ideas?

I had the same problem on my AMD64 system (MSI K8T Neo board). The reason, why the ioports are not recognised on this board (i.e., why they show up as 0x0, 0x0) is, that the ACPI defines these ports as FixedIO () and not as IO () as expected by the current linux code. FixedIO is perfectly fine and in line with the ACPI standard. So the DSDT table is NOT bad but linux is currently simply not flexible enough to cope with all possible (and allowable) combinations.

With the small attached patch the i8042 IO-ports were recognised on my board (only tested with keyboard).

Voytech, could you include this patch into your next patch set?

--- linux-2.6.9-rc1-mm5.orig/drivers/input/serio/i8042-x86ia64io.h 2004-09-13
15:39:39.061522663 +0200
+++ linux-2.6.9-rc1-mm5/drivers/input/serio/i8042-x86ia64io.h 2004-09-20
01:10:32.124593201 +0200

(quoted diff) +4 -0
@@ -105,6 +105,7 @@ static acpi_status i8042_acpi_parse_reso
 {
  struct i8042_acpi_resources *i8042_res = data;
  struct acpi_resource_io *io;
+ struct acpi_resource_fixed_io *fixed_io;
  struct acpi_resource_irq *irq;
  struct acpi_resource_ext_irq *ext_irq;
 
@@ -119,6 +120,16 @@ static acpi_status i8042_acpi_parse_reso
    }
    break;
 
+  case ACPI_RSTYPE_FIXED_IO:
+   fixed_io = &res->data.fixed_io;
+   if (fixed_io->range_length) {

Review static acpi_status i8042_acpi_parse_reso · line 135

Quoted 10 lines
(quoted diff) +7 -0
+    if (!i8042_res->port1)
+     i8042_res->port1 = fixed_io->base_address;
+    else
+     i8042_res->port2 = fixed_io->base_address;
+   }
+   break;
+
   case ACPI_RSTYPE_IRQ:
    irq = &res->data.irq;
    if (irq->number_of_interrupts > 0)

Bingo!
This patch solves my problems with i8042.
Please, include it upstream.
Thanks!