[PATCH] x86/mrst/pci: avoid enabling intx for msi capable devices

[PATCH] x86/mrst/pci: avoid enabling intx for msi capable devices

[Jacob Pan]

On Moorestown platform, true PCI devices with MSI capabilities do not support
INTx mode. IRQ line# for those devices are zeros in the PCI shim, an attempt
to enable INTx on these MSI capable devices will cause conflict in the system
such as IRQ0 for the system timers.
If the device driver probes/enables MSI before pci_enable_device(), the conflict
is not shown since INTx will be disabled. But if the driver tries to enable INTx
before MSI, the conflict will cause system timer IRQ0 to break.

This patch will ensure the ordering of INTx and MSI setup by the driver is
not relavent to INTx conflict. We will always skip INTx setup for MSI capable
devices on Moorestown.

Signed-off-by: Jacob Pan <jacob.jun.pan@linux.intel.com>
---
 arch/x86/pci/mrst.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/arch/x86/pci/mrst.c b/arch/x86/pci/mrst.c
index d5c7aef..ba3f08b 100644
--- a/arch/x86/pci/mrst.c
+++ b/arch/x86/pci/mrst.c
@@ -203,6 +203,9 @@ static int mrst_pci_irq_enable(struct pci_dev *dev)
 	u8 pin;
 	struct io_apic_irq_attr irq_attr;
 
+	if (pci_find_capability(dev, PCI_CAP_ID_MSI))
+		return 0;
+
 	pci_read_config_byte(dev, PCI_INTERRUPT_PIN, &pin);
 
 	/* MRST only have IOAPIC, the PCI irq lines are 1:1 mapped to
-- 
1.6.3.3

Re: [PATCH] x86/mrst/pci: avoid enabling intx for msi capable devices

[H. Peter Anvin]

Jesse: are you taking this or should I?

	-hpa

On 06/21/2010 11:30 AM, Jacob Pan wrote:
> On Moorestown platform, true PCI devices with MSI capabilities do not support
> INTx mode. IRQ line# for those devices are zeros in the PCI shim, an attempt
> to enable INTx on these MSI capable devices will cause conflict in the system
> such as IRQ0 for the system timers.
> If the device driver probes/enables MSI before pci_enable_device(), the conflict
> is not shown since INTx will be disabled. But if the driver tries to enable INTx
> before MSI, the conflict will cause system timer IRQ0 to break.
> 
> This patch will ensure the ordering of INTx and MSI setup by the driver is
> not relavent to INTx conflict. We will always skip INTx setup for MSI capable
> devices on Moorestown.

Re: [PATCH] x86/mrst/pci: avoid enabling intx for msi capable devices

[Jesse Barnes]

On Mon, 21 Jun 2010 12:01:30 -0700
"H. Peter Anvin" <hpa@zytor.com> wrote:

> Jesse: are you taking this or should I?
> 
> 	-hpa
> 
> On 06/21/2010 11:30 AM, Jacob Pan wrote:
> > On Moorestown platform, true PCI devices with MSI capabilities do not support
> > INTx mode. IRQ line# for those devices are zeros in the PCI shim, an attempt
> > to enable INTx on these MSI capable devices will cause conflict in the system
> > such as IRQ0 for the system timers.
> > If the device driver probes/enables MSI before pci_enable_device(), the conflict
> > is not shown since INTx will be disabled. But if the driver tries to enable INTx
> > before MSI, the conflict will cause system timer IRQ0 to break.
> > 
> > This patch will ensure the ordering of INTx and MSI setup by the driver is
> > not relavent to INTx conflict. We will always skip INTx setup for MSI capable
> > devices on Moorestown.

You go ahead.  You can add my:

Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>

Thanks,
-- 
Jesse Barnes, Intel Open Source Technology Center