[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200203200942.GA130652@google.com>
Date: Mon, 3 Feb 2020 14:09:42 -0600
From: Bjorn Helgaas <helgaas@...nel.org>
To: "Jason A. Donenfeld" <Jason@...c4.com>
Cc: linux-kernel@...r.kernel.org, x86@...nel.org, hch@....de
Subject: Re: [PATCH] x86/PCI: ensure to_pci_sysdata usage is guarded by
CONFIG_PCI
On Mon, Feb 03, 2020 at 07:19:06PM +0100, Jason A. Donenfeld wrote:
> Recently, the helper to_pci_sysdata was added inside of the CONFIG_PCI
> guard, but it is used from inside of a CONFIG_PCI_MSI_IRQ_DOMAIN guard,
> which does not require CONFIG_PCI.
AFAICT, CONFIG_PCI_MSI_IRQ_DOMAIN *does* require CONFIG_PCI.
Do you mean to_pci_sysdata() is used inside a *CONFIG_NUMA* guard by
__pcibus_to_node()?
The patch below makes __pcibus_to_node() available only when
CONFIG_PCI, when previously it was always available. Maybe that's
fine, I dunno.
Another possibility would be to move the to_pci_sysdata() definition
outside of #ifdef CONFIG_PCI, just as the struct pci_sysdata is. Then
we wouldn't have to change the availability of __pcibus_to_node().
> This breaks builds on !CONFIG_PCI
> machines. This commit fixes the ifdef to require CONFIG_PCI.
>
> Signed-off-by: Jason A. Donenfeld <Jason@...c4.com>
> Fixes: aad6aa0cd674 ("x86/PCI: Add to_pci_sysdata() helper")
> Cc: Christoph Hellwig <hch@....de>
> ---
> arch/x86/include/asm/pci.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/include/asm/pci.h b/arch/x86/include/asm/pci.h
> index 40ac1330adb2..d8772b75236d 100644
> --- a/arch/x86/include/asm/pci.h
> +++ b/arch/x86/include/asm/pci.h
> @@ -119,7 +119,7 @@ void native_restore_msi_irqs(struct pci_dev *dev);
> /* generic pci stuff */
> #include <asm-generic/pci.h>
>
> -#ifdef CONFIG_NUMA
> +#if defined(CONFIG_NUMA) && defined(CONFIG_PCI)
> /* Returns the node based on pci bus */
> static inline int __pcibus_to_node(const struct pci_bus *bus)
> {
> --
> 2.25.0
>
Powered by blists - more mailing lists