Re: [RFC][PATCH] acpiphp: configure _PRT

[Kristen Accardi <kristen.c.accardi@intel.com>]

On Thu, 2006-03-02 at 22:58 +0900, MUNEDA Takahiro wrote:
> Hi,
> 
> Current acpiphp does not free acpi_device structs when the
> PCI devices are removed. When the PCI device is added,
> acpi_bus_add() fails because acpi_device struct has already
> exists. So, _PRT method does not evaluate.
> 
> This patch fixes this issue.
> 
> This patch is against 2.6.16-rc5-mm1 plus following patches.
> 
>  o Kristen's latest patch set
>  o Kenji's acpiphp: Scan slots under the nested P2P bridge
> 
> I tested this patch on my tiger4 box except for dock eject
> case, because I don't have any pc which has _DCK method.
> 

I tested this on a couple laptops that have _DCK methods - It seems to
work fine, however, see comment below:
<snip>

> Index: linux-2.6.16-rc5-mm1/drivers/pci/hotplug/acpiphp_dock.c
> ===================================================================
> --- linux-2.6.16-rc5-mm1.orig/drivers/pci/hotplug/acpiphp_dock.c
> +++ linux-2.6.16-rc5-mm1/drivers/pci/hotplug/acpiphp_dock.c
> @@ -290,6 +290,7 @@ handle_hotplug_event_dock(acpi_handle ha
>  			dbg("EJECT request\n");
>  			if (!dock_in_progress() && dock_present()) {
>  				hotplug_pci(type);
> +				acpiphp_bus_trim(handle);

This is not necessary.  hotplug_pci() will pass the eject event down to
the acpiphp handle_hotplug_event_func() routine, which will call
disable_device() and remove the bus that way as part of your new patch.

>  				undock();
>  				eject_dock();
>  				if (dock_present())