|
From: marcus h. <ma...@tu...> - 2005-12-01 17:04:25
|
On Wed, Nov 30, 2005 at 11:05:26PM +0100, Pavel Machek wrote: > You should really go to 2.6... button drivers should be possible there > without hacking kernel. Well, my driver is a loadable module, the only kernel hacking I'm doing is trying to figure out why it isn't working... It works just fine on this hardware under 2.4.20. But the ACPI system underwent big changes between 2.4.20 -> 2.4.28. I have now become convinced that ACPI is not seeing the button presses, so the AML code never runs to "write" into the address space, so my driver never sees the address space activated. That all makes reasonable sense and is supported by the fact that IRQ 9 (which is the only ACPI handled interrupt) has zero hits recorded, regardless of repeatedly pressing buttons. Also, with full debugging turned on, I see no output as a result of button presses. So, the problem seems to be something in the setup. Many of the ACPI functions *are* working.. I can adjust screen brightness, I can see the battery status, I can see the AC status change when I plug in or remove the charger. One other thing that appears to be non-functional is the power switch: <4>evxfevnt-0196 [26] acpi_enable_event : Could not enable power_button even t <4> evxface-0107 [25] acpi_install_fixed_eve: Could not enable fixed event. <4>acpi_button-0437 [24] acpi_button_add : Error installing notify handler My driver also attempts to call acpi_install_fixed_event_handler(ACPI_EVENT_POWER_BUTTON,buttons_acpi_power,0) which fails with code 0x19 (AE_NO_HARDWARE_RESPONSE). I don't know if this is clearly related to the button issue, but it seems likely at least. Is there any additional setup required with the newer ACPI code that might wake things up? I am not running acpid, if that might be significant.. I do want to take things to the 2.6 kernel eventually, but there are still enough ties to the 2.4 kernel elsewhere that I can't do it just yet... Thanks in advance! Marcus Hall |