[PATCH v2] IA64 Compilation Error Fix for Intel IOMMU Identity Mapping Support

[Yu, Fenghua]

>-----Original Message-----
>From: David Woodhouse [mailto:dwmw2 at infradead.org]
>Sent: Thursday, June 25, 2009 2:52 PM
>To: FUJITA Tomonori
>Cc: Yu, Fenghua; chrisw at redhat.com; torvalds at linux-foundation.org;
>akpm at linux-foundation.org; Luck, Tony; linux-kernel at vger.kernel.org;
>iommu at lists.linux-foundation.org; linux-ia64 at vger.kernel.org
>Subject: Re: [PATCH v2] IA64 Compilation Error Fix for Intel IOMMU Identity
>Mapping Support
>
>On Thu, 2009-06-25 at 08:11 +0100, David Woodhouse wrote:
>> It raises the question: Why are we using firmware-specific interfaces to
>> list the available memory -- can't we get that from somewhere _generic_?
>>
>> The less we tie our code to these crappy BIOS, EFI and ACPI interfaces,
>> the better off we'll be.
>
>Does this work everywhere... ?
>
>(Fenghua, why are we doing the whole setup per pci dev anyway -- why not
>set up the page tables once and point all devices at the same page
>tables?)
>

That's what the ia64 fix patch does...the first pci device set up the page table. Then all following pci devices use that page table by setting up the context mapping.

+		if (first_pdev) {
+			/*
+			 * first pdev sets up the whole page table for
+			 * si_domain.
+			 */
+			first_pdev = 0;
+			ret = iommu_setup_identity_map(pdev);
+			if (ret)
+				return ret;
+		} else {
+			/*
+			 * following pdev's just use the page table.
+			 */
+			printk(KERN_INFO "IOMMU: identity mapping for device %s\n",
+			       pci_name(pdev));
+			ret = domain_context_mapping(si_domain, pdev,
+						     CONTEXT_TT_MULTI_LEVEL);
+			if (ret)
Thanks.

-Fenghua