Re: [PATCH 0/10] ACPI / PCI / PCIe: Rework _OSC handling (v3)

[Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>]

(2010/08/06 10:03), Rafael J. Wysocki wrote:
> Hi,
> 
> This is the 3rd iteration of the patchset based on
> https://patchwork.kernel.org/patch/114917/ and the comments I received on that
> patch and the previous two iterations of this patchset.
> 
> The patches below do not make the intermediate step to put an unconditional
> _OSC query into acpi_pci_osc_control_set(), but the final result is quite
> similar to the 2nd iteration with https://patchwork.kernel.org/patch/116976/
> applied on top.
> 
> [1/10] - Introduce acpi_pci_osc_control_query() allowing the caller to get a
>          mask of _OSC control bits the BIOS allows the kernel to control
>          for a given PCI root bridge.  This version repeats _OSC queries until
>          the BIOS doesn't mask any requested bits.

This contains 2 parts:
 1-A) change acpi_status acpi_pci_query_osc() to take support and *control
 1-B) introduce acpi_pci_osc_control_query()

and I'd like to kill 1-B since it is removed soon by patch [10/10].

> 
> [2/10] - Introduce pci_aer_available() allowing the caller to check if the
>          AER service driver should be enabled.
> 
> [3/10] - Introduce kernel command line switch pcie_ports=.

These 2 are prep for following big changes.

> 
> [4/10] - Rework the PCIe port driver to request _OSC control for all serives at
>          once.
> 
> [5/10] - Disable PCIe port services (that might be enabled by the BIOS) during
>          initialization.

These 2 are core part to solve the issue.

> 
> [6/10] - Remove the PCIe port driver modules exit function.

This is a cleanup. There is no reason to keep this in the series.

> 
> [7/10] - Make PCI hotplug code query _OSC before requesting control of _OSC
>          features.

Needless since removed soon by patch [10/10].

> 
> [8/10] - Rework acpi_pci_osc_control_set() so that it doesn't use cached
>         result of a query and remove the fields of struct acpi_pci_root that
>         aren't used any more.

This also contains 2 parts:
 8-A: don't use cached result
 8-B: remove query in acpi_pci_osc_control_set()

I think 8-B is not required since query relives in patch [10/10].

> 
> [9/10] - Reorder checks in acpi_pci_osc_control_set().

A minor cleanup can be out of the series.

> 
> [10/10] - Merge acpi_pci_osc_control_set() with acpi_pci_osc_control_query().

This removes many badness in this series.
I think we should have this improved acpi_pci_osc_control_set() before [4/10].

> 
> Comments welcome.

I recommend you to rearrange the order of patches in this series.

1. Misc cleanups
   [6/10]
   [9/10]
2. Prep1: prepare small parts
   [2/10]
   [3/10]
3. Prep2: improve acpi_pci_osc_control_set()
   [1-A]
   [8-A]
   [10/10]
4. Main
   [4/10]
   [5/10]

I think there is no objections to 4 patches from the first ([6,9,2,3/10]).


Thanks,
H.Seto