[PATCH] Do not write virq back to PCI config space (revised patch).

[PATCH] Do not write virq back to PCI config space (revised patch).

[Michal Ostrowski]

(Revised version, includes changes for 32 and 64 bit platforms.)

- Drivers will not rely on the PCI config space value, as they've
  already been conditioned to rely on the irq field in "struct pci_dev".

- The virq value may not be < 256 as it has been remapped.

- The PCI config space should reflect the hardware configuration, which
  is not being changed.  We are only creating a virtual irq mapping that
  exists in the kernel only. One would never expect the PCI hardware to
  generate the "virq" interrupt.

Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com>
---
 arch/powerpc/kernel/pci_32.c |    1 -
 arch/powerpc/kernel/pci_64.c |    1 -
 2 files changed, 0 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/kernel/pci_32.c b/arch/powerpc/kernel/pci_32.c
index 0d9ff72..0df9126 100644
--- a/arch/powerpc/kernel/pci_32.c
+++ b/arch/powerpc/kernel/pci_32.c
@@ -1459,7 +1459,6 @@ int pci_read_irq_line(struct pci_dev *pci_dev)
 		return -1;
 	}
 	pci_dev->irq = virq;
-	pci_write_config_byte(pci_dev, PCI_INTERRUPT_LINE, virq);
 
 	return 0;
 }
diff --git a/arch/powerpc/kernel/pci_64.c b/arch/powerpc/kernel/pci_64.c
index 9bae8a5..81a73fb 100644
--- a/arch/powerpc/kernel/pci_64.c
+++ b/arch/powerpc/kernel/pci_64.c
@@ -1305,7 +1305,6 @@ int pci_read_irq_line(struct pci_dev *pci_dev)
 	DBG(" -> mapped to linux irq %d\n", virq);
 
 	pci_dev->irq = virq;
-	pci_write_config_byte(pci_dev, PCI_INTERRUPT_LINE, virq);
 
 	return 0;
 }
-- 
1.4.4.2

Re: [PATCH] Do not write virq back to PCI config space (revised patch).

[Benjamin Herrenschmidt]

On Wed, 2006-12-27 at 22:14 -0600, Michal Ostrowski wrote:
> (Revised version, includes changes for 32 and 64 bit platforms.)
> 
> - Drivers will not rely on the PCI config space value, as they've
>   already been conditioned to rely on the irq field in "struct pci_dev".
> 
> - The virq value may not be < 256 as it has been remapped.
> 
> - The PCI config space should reflect the hardware configuration, which
>   is not being changed.  We are only creating a virtual irq mapping that
>   exists in the kernel only. One would never expect the PCI hardware to
>   generate the "virq" interrupt.
> 
> Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com>

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

> ---
>  arch/powerpc/kernel/pci_32.c |    1 -
>  arch/powerpc/kernel/pci_64.c |    1 -
>  2 files changed, 0 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/powerpc/kernel/pci_32.c b/arch/powerpc/kernel/pci_32.c
> index 0d9ff72..0df9126 100644
> --- a/arch/powerpc/kernel/pci_32.c
> +++ b/arch/powerpc/kernel/pci_32.c
> @@ -1459,7 +1459,6 @@ int pci_read_irq_line(struct pci_dev *pci_dev)
>  		return -1;
>  	}
>  	pci_dev->irq = virq;
> -	pci_write_config_byte(pci_dev, PCI_INTERRUPT_LINE, virq);
>  
>  	return 0;
>  }
> diff --git a/arch/powerpc/kernel/pci_64.c b/arch/powerpc/kernel/pci_64.c
> index 9bae8a5..81a73fb 100644
> --- a/arch/powerpc/kernel/pci_64.c
> +++ b/arch/powerpc/kernel/pci_64.c
> @@ -1305,7 +1305,6 @@ int pci_read_irq_line(struct pci_dev *pci_dev)
>  	DBG(" -> mapped to linux irq %d\n", virq);
>  
>  	pci_dev->irq = virq;
> -	pci_write_config_byte(pci_dev, PCI_INTERRUPT_LINE, virq);
>  
>  	return 0;
>  }