[openib-general] [PATCH/RFC] busted request IRQ for PCIe ipath HCAs

[Roland Dreier]

I think commit 51f65ebc (fix HT IRQ setting on HT HCAs) busted ipath
on PCIe HCAs, since ipath_irq is set before pci_enable_msi(), which
means it gets some value unrelated to the actual IRQ that is assigned.
I needed the patch below to make 2.6.19 work with my PCIe HCAs.

Bryan/anyone at Qlogic, does this look right?  It worked for me, so if
this is what was intended, I will queue the patch for 2.6.20 and
submit to stable at kernel.org for 2.6.19.x.

 - R.

diff --git a/drivers/infiniband/hw/ipath/ipath_iba6120.c b/drivers/infiniband/hw/ipath/ipath_iba6120.c
index 6af8968..498b596 100644
--- a/drivers/infiniband/hw/ipath/ipath_iba6120.c
+++ b/drivers/infiniband/hw/ipath/ipath_iba6120.c
@@ -851,8 +851,8 @@ static int ipath_setup_pe_config(struct
 	int pos, ret;
 
 	dd->ipath_msi_lo = 0;	/* used as a flag during reset processing */
-	dd->ipath_irq = pdev->irq;
 	ret = pci_enable_msi(dd->pcidev);
+	dd->ipath_irq = pdev->irq;
 	if (ret)
 		ipath_dev_err(dd, "pci_enable_msi failed: %d, "
 			      "interrupts may not work\n", ret);