RE: [patch 2/2] acpi: add ability to derive irq when doing a surpriseremoval of an adapter

[Li, Shaohua]

Hi,
>
>If an adapter is surprise removed, the interrupt pin must be guessed,
as
>any attempts to read it would obviously be invalid.  cycle through all
>possible interrupt pin values until we can either lookup or derive the
>right irq to disable.
>
>Signed-off-by: Kristen Carlson Accardi <kristen.c.accardi@xxxxxxxxx>
>
>diff -uprN -X linux-2.6.14-rc2/Documentation/dontdiff linux-2.6.14-
>rc2/drivers/acpi/pci_irq.c linux-2.6.14-rc2-kca1/drivers/acpi/pci_irq.c
>--- linux-2.6.14-rc2/drivers/acpi/pci_irq.c	2005-09-27
>09:01:28.000000000 -0700
>+++ linux-2.6.14-rc2-kca1/drivers/acpi/pci_irq.c	2005-09-28
>10:40:57.000000000 -0700
>@@ -491,6 +491,79 @@ void __attribute__ ((weak)) acpi_unregis
> {
> }
>
>+
>+
>+/*
>+ * This function will be called only in the case of
>+ * a "surprise" hot plug removal.  For surprise removals,
>+ * the card has either already be yanked out of the slot, or
>+ * the slot's been powered off, so we have to brute force
>+ * our way through all the possible interrupt pins to derive
>+ * the GSI, then we double check with the value stored in the
>+ * pci_dev structure to make sure we have the GSI that belongs
>+ * to this IRQ.
>+ */
>+void acpi_pci_irq_disable_nodev(struct pci_dev *dev)
>+{
>+	int gsi = 0;
>+	u8  pin = 0;
>+	int edge_level = ACPI_LEVEL_SENSITIVE;
>+	int active_high_low = ACPI_ACTIVE_LOW;
>+	int irq;
>+
>+	/*
>+	 * since our device is not present, we
>+	 * can't just read the interrupt pin
>+	 * and use the value to derive the irq.
>+	 * in this case, we are going to check
>+	 * each returned irq value to make
>+	 * sure it matches our already assigned
>+	 * irq before we use it.
>+	 */
>+	for (pin = 0; pin < 4; pin++) {
>+		/*
>+	 	 * First we check the PCI IRQ routing table (PRT) for an
IRQ.
>+	 	 */
>+		gsi = acpi_pci_irq_lookup(dev->bus,
PCI_SLOT(dev->devfn), pin,
>+				  &edge_level, &active_high_low, NULL,
>+				  acpi_pci_free_irq);
acpi_pci_free_irq has side effect. In the link device case, it
deferences a count. The blind guess will mass the reference count. Could
you introduce something like 'acpi_pci_find_irq'?

Thanks,
Shaohua 
-
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/