> I think this code is backwards. In xhci-pci.c, when we detect the xHCI
> chipsets with broken MSI, we should clear the HCD_MSI_FIRST flag, and
> allow the USB core to set up the legacy PCI interrupt. Then this
> function should just return when it detects the XHCI_BROKEN_MSI quirk
> (before it sets hcd->irq to -1). That way any changes to the legacy PCI
> IRQ setup normal flow in the USB core will also effect these xHCI hosts
> that really need it.
Sorry for response later, I backed to my hometown from Jan 16 yo Jan 28.
and have a bad Internet connection. :(
how about the following:
----------
>From b3c01db6f652ca788d0dff498e038445365915bb Mon Sep 17 00:00:00 2001
From: Alex Shi <alex.shi@xxxxxxxxx>
Date: Thu, 19 Jan 2012 22:24:02 +0800
Subject: [PATCH 2/2] USB: try MSI before legacy irq on pci xhci HCD
This patch try to make sure that the USB core allows the xHCI driver to
enable MSI/MSIX before the legacy PCI IRQ. If the MSI/MSIX broken for
the HCD. It will still fall back to legacy PCI IRQ. This change can
avoid a unnecessary request/free legacy PCI IRQ when MSI is workable.
Signed-off-by: Alex Shi <alex.shi@xxxxxxxxx>
---
drivers/usb/core/hcd.c | 5 ++---
drivers/usb/host/xhci-pci.c | 11 +++++------
2 files changed, 7 insertions(+), 9 deletions(-)
diff --git a/drivers/usb/core/hcd.c b/drivers/usb/core/hcd.c
index 52566c6..b3a7920 100644
--- a/drivers/usb/core/hcd.c
+++ b/drivers/usb/core/hcd.c
@@ -2447,9 +2447,8 @@ int usb_add_hcd(struct usb_hcd *hcd,
&& device_can_wakeup(&hcd->self.root_hub->dev))
dev_dbg(hcd->self.controller, "supports USB remote wakeup\n");
- /* enable irqs just before we start the controller. But Intel USB3
- * hcd can't do this here on some platform, they will do it in
- * following driver->start();
+ /* enable irqs just before we start the controller, except MSI
+ * first try HCD. That will do it in following driver->start();
*/
if (usb_hcd_is_primary_hcd(hcd) &&
!(hcd->driver->flags & HCD_MSI_FIRST)) {
diff --git a/drivers/usb/host/xhci-pci.c b/drivers/usb/host/xhci-pci.c
index d453cb8..5185ab7 100644
--- a/drivers/usb/host/xhci-pci.c
+++ b/drivers/usb/host/xhci-pci.c
@@ -55,6 +55,8 @@ static int xhci_pci_reinit(struct xhci_hcd *xhci, struct pci_dev *pdev)
static void xhci_pci_quirks(struct device *dev, struct xhci_hcd *xhci)
{
struct pci_dev *pdev = to_pci_dev(dev);
+ struct hc_driver *hcp = (struct hc_driver *) xhci_to_hcd(xhci)
+ ->driver;
/* Look for vendor-specific quirks */
if (pdev->vendor == PCI_VENDOR_ID_FRESCO_LOGIC &&
@@ -69,6 +71,7 @@ static void xhci_pci_quirks(struct device *dev, struct xhci_hcd *xhci)
* capabilities.
*/
xhci->quirks |= XHCI_BROKEN_MSI;
+ hcp->flags &= ~HCD_MSI_FIRST;
xhci_dbg(xhci, "QUIRK: Fresco Logic revision %u "
"has broken MSI implementation\n",
pdev->revision);
@@ -137,11 +140,6 @@ static int xhci_pci_probe(struct pci_dev *dev, const struct pci_device_id *id)
driver = (struct hc_driver *)id->driver_data;
- /* stop line IRQ checking in xhci_hcd_pci_probe. */
- if (dev->vendor == PCI_VENDOR_ID_INTEL &&
- dev->device == PCI_DEVICE_ID_INTEL_PANTHERPOINT_XHCI)
- driver->flags |= HCD_MSI_FIRST;
-
/* Register the USB 2.0 roothub.
* FIXME: USB core must know to register the USB 2.0 roothub first.
* This is sort of silly, because we could just set the HCD driver flags
@@ -249,7 +247,8 @@ static const struct hc_driver xhci_pci_hc_driver = {
* generic hardware linkage
*/
.irq = xhci_irq,
- .flags = HCD_MEMORY | HCD_USB3 | HCD_SHARED,
+ .flags = HCD_MEMORY | HCD_USB3 | HCD_SHARED |
+ HCD_MSI_FIRST,
/*
* basic lifecycle operations
--
1.6.3.3
> > - if (!ret)
> > - /* hcd->irq is -1, we have MSI */
> > - return 0;
> > + if (!ret)
> > + /* hcd->irq is -1, we have MSI */
> > + return 0;
> > + }
> >
> > if (!pdev->irq) {
> > xhci_err(xhci, "No msi-x/msi found and no IRQ in BIOS\n");
>
> The first patch looks fine, although I would have to apply it to see the
> changes. BTW, have you tested the code on the Intel Panther Point xHCI
> host to make sure it works when CONFIG_PCI_MSI is turned off? And by
> "works" I mean that xHCI probe fails when MSI is turned off for Panther
> Point systems.
>
I am sorry for can not try this. Because I need the PCIe NIC to remote
access on the machine from home, if turn off the MSI, It will lose the
only connection for this machine.
Are ther anyone like to do this try?
or after I am back to office.
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
