Re: [PATCH 4/7] x86/pci: use pcie_cap to simplify code

[Bjorn Helgaas]

On Tue, Sep 03, 2013 at 03:35:12PM +0800, Yijing Wang wrote:
> PCI core saves PCIe Cap offset in pcie_cap,
> use pcie_cap to simplify code.
> 
> Signed-off-by: Yijing Wang <wangyijing@xxxxxxxxxx>
> ---
>  arch/x86/pci/fixup.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/x86/pci/fixup.c b/arch/x86/pci/fixup.c
> index f5809fa..ee8330d 100644
> --- a/arch/x86/pci/fixup.c
> +++ b/arch/x86/pci/fixup.c
> @@ -288,7 +288,7 @@ static void pcie_rootport_aspm_quirk(struct pci_dev *pdev)
>  		 */
>  		list_for_each_entry(dev, &pbus->devices, bus_list) {
>  			/* There are 0 to 8 devices attached to this bus */
> -			cap_base = pci_find_capability(dev, PCI_CAP_ID_EXP);
> +			cap_base = dev->pcie_cap;
>  			quirk_aspm_offset[GET_INDEX(pdev->device, dev->devfn)] = cap_base + 0x10;

This should use PCI_EXP_LNKCTL instead of "0x10".

>  		}
>  		pbus->ops = &quirk_pcie_aspm_ops;

This quirk replaces the config accessors with ones that silently ignore
writes to ASPM control bits.  That really warrants at least a dev_info()
note here, and we should be using pci_bus_set_ops().

Even that is a little bit dubious because I don't think this is really
safe -- what happens if this quirk replaces the ops, then somebody
else replaces the ops again?  aer_inject.c at least keeps track of
the old ops and seems to fall back to them, but it seems fragile to
depend on every caller of pci_bus_set_ops() to do the right thing
there.

But this is beyond the scope of your patch, so if you just
add a dev_info() note and use pci_bus_set_ops(), that should be
enough for now.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/