diff --git a/arch/x86_64/kernel/pci-dma.c b/arch/x86_64/kernel/pci-dma.c
index 7edd1a40fab3..a45844c7e3a3 100644
--- a/arch/x86_64/kernel/pci-dma.c
+++ b/arch/x86_64/kernel/pci-dma.c
@@ -279,3 +279,32 @@ __init int iommu_setup(char *p)
     }
     return 1;
 }
+__setup("iommu=", iommu_setup);
+
+void __init pci_iommu_alloc(void)
+{
+	/*
+	 * The order of these functions is important for
+	 * fall-back/fail-over reasons
+	 */
+#ifdef CONFIG_IOMMU
+	iommu_hole_init();
+#endif
+
+#ifdef CONFIG_SWIOTLB
+	pci_swiotlb_init();
+#endif
+}
+
+static int __init pci_iommu_init(void)
+{
+#ifdef CONFIG_IOMMU
+	gart_iommu_init();
+#endif
+
+	no_iommu_init();
+	return 0;
+}
+
+/* Must execute after PCI subsystem */
+fs_initcall(pci_iommu_init);
diff --git a/arch/x86_64/kernel/pci-gart.c b/arch/x86_64/kernel/pci-gart.c
index 4f67957d2b42..9a93954bed37 100644
--- a/arch/x86_64/kernel/pci-gart.c
+++ b/arch/x86_64/kernel/pci-gart.c
@@ -571,7 +571,7 @@ static struct dma_mapping_ops gart_dma_ops = {
 	.unmap_sg = gart_unmap_sg,
 };
 
-static int __init pci_iommu_init(void)
+void __init gart_iommu_init(void)
 { 
 	struct agp_kern_info info;
 	unsigned long aper_size;
@@ -581,7 +581,7 @@ static int __init pci_iommu_init(void)
 
 	if (cache_k8_northbridges() < 0 || num_k8_northbridges == 0) {
 		printk(KERN_INFO "PCI-GART: No AMD northbridge found.\n");
-		return -ENODEV;
+		return;
 	}
 
 #ifndef CONFIG_AGP_AMD64
@@ -595,11 +595,11 @@ static int __init pci_iommu_init(void)
 #endif	
 
 	if (swiotlb)
-		return -ENODEV;
+		return;
 
 	/* Did we detect a different HW IOMMU? */
 	if (iommu_detected && !iommu_aperture)
-		return -1;
+		return;
 
 	if (no_iommu ||
 	    (!force_iommu && end_pfn <= MAX_DMA32_PFN) ||
@@ -611,7 +611,7 @@ static int __init pci_iommu_init(void)
 					"but IOMMU not available.\n"
 			       KERN_ERR "WARNING 32bit PCI may malfunction.\n");
 		}
-		return -ENODEV;
+		return;
 	}
 
 	printk(KERN_INFO "PCI-DMA: using GART IOMMU.\n");
@@ -678,11 +678,10 @@ static int __init pci_iommu_init(void)
 
 	flush_gart();
 	dma_ops = &gart_dma_ops;
-	return 0;
 } 
 
 /* Must execute after PCI subsystem */
-fs_initcall(pci_iommu_init);
+fs_initcall(gart_iommu_init);
 
 void gart_parse_options(char *p)
 {
diff --git a/arch/x86_64/kernel/setup.c b/arch/x86_64/kernel/setup.c
index a9de8f02671f..04b2d7b92d17 100644
--- a/arch/x86_64/kernel/setup.c
+++ b/arch/x86_64/kernel/setup.c
@@ -63,7 +63,6 @@
 #include <asm/setup.h>
 #include <asm/mach_apic.h>
 #include <asm/numa.h>
-#include <asm/swiotlb.h>
 #include <asm/sections.h>
 #include <asm/dmi.h>
 
@@ -702,10 +701,6 @@ void __init setup_arch(char **cmdline_p)
 
 	e820_setup_gap();
 
-#ifdef CONFIG_IOMMU
-	iommu_hole_init();
-#endif
-
 #ifdef CONFIG_VT
 #if defined(CONFIG_VGA_CONSOLE)
 	conswitchp = &vga_con;
diff --git a/arch/x86_64/mm/init.c b/arch/x86_64/mm/init.c
index b83645a2e02d..2f5f5b11e9d0 100644
--- a/arch/x86_64/mm/init.c
+++ b/arch/x86_64/mm/init.c
@@ -41,8 +41,6 @@
 #include <asm/proto.h>
 #include <asm/smp.h>
 #include <asm/sections.h>
-#include <asm/dma-mapping.h>
-#include <asm/swiotlb.h>
 
 #ifndef Dprintk
 #define Dprintk(x...)
@@ -587,10 +585,7 @@ void __init mem_init(void)
 {
 	long codesize, reservedpages, datasize, initsize;
 
-#ifdef CONFIG_SWIOTLB
-	pci_swiotlb_init();
-#endif
-	no_iommu_init();
+	pci_iommu_alloc();
 
 	/* How many end-of-memory variables you have, grandma! */
 	max_low_pfn = end_pfn;
diff --git a/include/asm-x86_64/pci.h b/include/asm-x86_64/pci.h
index 3374d34c4acd..4dbc07c54f7a 100644
--- a/include/asm-x86_64/pci.h
+++ b/include/asm-x86_64/pci.h
@@ -39,8 +39,8 @@ int pcibios_set_irq_routing(struct pci_dev *dev, int pin, int irq);
 #include <asm/scatterlist.h>
 #include <linux/string.h>
 #include <asm/page.h>
-#include <linux/dma-mapping.h> /* for have_iommu */
 
+extern void pci_iommu_alloc(void);
 extern int iommu_setup(char *opt);
 
 /* The PCI address space does equal the physical memory
diff --git a/include/asm-x86_64/proto.h b/include/asm-x86_64/proto.h
index 9d3335b4e9b6..038fe1f47e6f 100644
--- a/include/asm-x86_64/proto.h
+++ b/include/asm-x86_64/proto.h
@@ -37,7 +37,6 @@ extern void ia32_sysenter_target(void);
 
 extern void config_acpi_tables(void);
 extern void ia32_syscall(void);
-extern void iommu_hole_init(void);
 
 extern int pmtimer_mark_offset(void);
 extern void pmtimer_resume(void);
@@ -101,13 +100,9 @@ extern int unsynchronized_tsc(void);
 
 extern void select_idle_routine(const struct cpuinfo_x86 *c);
 
-extern void gart_parse_options(char *);
-extern void __init no_iommu_init(void);
-
 extern unsigned long table_start, table_end;
 
 extern int exception_trace;
-extern int force_iommu, no_iommu;
 extern int using_apic_timer;
 extern int disable_apic;
 extern unsigned cpu_khz;
@@ -116,8 +111,13 @@ extern int skip_ioapic_setup;
 extern int acpi_ht;
 extern int acpi_disabled;
 
+extern void no_iommu_init(void);
+extern int force_iommu, no_iommu;
 extern int iommu_detected;
 #ifdef CONFIG_IOMMU
+extern void gart_iommu_init(void);
+extern void gart_parse_options(char *);
+extern void iommu_hole_init(void);
 extern int fallback_aper_order;
 extern int fallback_aper_force;
 extern int iommu_aperture;