ia64: IOMMU passthrough mode shouldn't trigger swiotlb init
authorDavid Woodhouse <David.Woodhouse@intel.com>
Thu, 13 Aug 2009 17:18:00 +0000 (18:18 +0100)
committerDavid Woodhouse <David.Woodhouse@intel.com>
Thu, 13 Aug 2009 17:18:00 +0000 (18:18 +0100)
Since commit 19943b0e30b05d42e494ae6fef78156ebc8c637e ('intel-iommu:
Unify hardware and software passthrough support'), hardware passthrough
mode will do the same as software passthrough mode was doing -- it'll
still use the IOMMU normally for devices which can't address all of
memory. This means that we don't need to bother with swiotlb.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
arch/ia64/kernel/pci-swiotlb.c

index 223abb134105a814ca755933a7bc9e105c62eaf5..285aae8431c6278d08c6ff862a18fdf66b2e52bf 100644 (file)
@@ -46,7 +46,7 @@ void __init swiotlb_dma_init(void)
 
 void __init pci_swiotlb_init(void)
 {
-       if (!iommu_detected || iommu_pass_through) {
+       if (!iommu_detected) {
 #ifdef CONFIG_IA64_GENERIC
                swiotlb = 1;
                printk(KERN_INFO "PCI-DMA: Re-initialize machine vector.\n");