diff options
author | Joerg Roedel <joerg.roedel@amd.com> | 2008-09-04 13:01:02 -0400 |
---|---|---|
committer | Ingo Molnar <mingo@elte.hu> | 2008-09-19 06:59:08 -0400 |
commit | 5507eef835c9c941e69d6d96e4b43af23eeb4ac9 (patch) | |
tree | 3072e37193035e1f2ff6a37e265a8bddadbad640 | |
parent | 1c65577398589bb44ab0980f9b9d30804b48a5db (diff) |
AMD IOMMU: add branch hints to completion wait checks
This patch adds branch hints to the cecks if a completion_wait is
necessary. The completion_waits in the mapping paths are unlikly because
they will only happen on software implementations of AMD IOMMU which
don't exists today or with lazy IO/TLB flushing when the allocator wraps
around the address space. With lazy IO/TLB flushing the completion_wait
in the unmapping path is unlikely too.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
-rw-r--r-- | arch/x86/kernel/amd_iommu.c | 12 |
1 files changed, 6 insertions, 6 deletions
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c index 679f2a8e22ee..d743aa0adccc 100644 --- a/arch/x86/kernel/amd_iommu.c +++ b/arch/x86/kernel/amd_iommu.c | |||
@@ -876,7 +876,7 @@ static dma_addr_t map_single(struct device *dev, phys_addr_t paddr, | |||
876 | if (addr == bad_dma_address) | 876 | if (addr == bad_dma_address) |
877 | goto out; | 877 | goto out; |
878 | 878 | ||
879 | if (iommu->need_sync) | 879 | if (unlikely(iommu->need_sync)) |
880 | iommu_completion_wait(iommu); | 880 | iommu_completion_wait(iommu); |
881 | 881 | ||
882 | out: | 882 | out: |
@@ -905,7 +905,7 @@ static void unmap_single(struct device *dev, dma_addr_t dma_addr, | |||
905 | 905 | ||
906 | __unmap_single(iommu, domain->priv, dma_addr, size, dir); | 906 | __unmap_single(iommu, domain->priv, dma_addr, size, dir); |
907 | 907 | ||
908 | if (iommu->need_sync) | 908 | if (unlikely(iommu->need_sync)) |
909 | iommu_completion_wait(iommu); | 909 | iommu_completion_wait(iommu); |
910 | 910 | ||
911 | spin_unlock_irqrestore(&domain->lock, flags); | 911 | spin_unlock_irqrestore(&domain->lock, flags); |
@@ -968,7 +968,7 @@ static int map_sg(struct device *dev, struct scatterlist *sglist, | |||
968 | goto unmap; | 968 | goto unmap; |
969 | } | 969 | } |
970 | 970 | ||
971 | if (iommu->need_sync) | 971 | if (unlikely(iommu->need_sync)) |
972 | iommu_completion_wait(iommu); | 972 | iommu_completion_wait(iommu); |
973 | 973 | ||
974 | out: | 974 | out: |
@@ -1014,7 +1014,7 @@ static void unmap_sg(struct device *dev, struct scatterlist *sglist, | |||
1014 | s->dma_address = s->dma_length = 0; | 1014 | s->dma_address = s->dma_length = 0; |
1015 | } | 1015 | } |
1016 | 1016 | ||
1017 | if (iommu->need_sync) | 1017 | if (unlikely(iommu->need_sync)) |
1018 | iommu_completion_wait(iommu); | 1018 | iommu_completion_wait(iommu); |
1019 | 1019 | ||
1020 | spin_unlock_irqrestore(&domain->lock, flags); | 1020 | spin_unlock_irqrestore(&domain->lock, flags); |
@@ -1061,7 +1061,7 @@ static void *alloc_coherent(struct device *dev, size_t size, | |||
1061 | goto out; | 1061 | goto out; |
1062 | } | 1062 | } |
1063 | 1063 | ||
1064 | if (iommu->need_sync) | 1064 | if (unlikely(iommu->need_sync)) |
1065 | iommu_completion_wait(iommu); | 1065 | iommu_completion_wait(iommu); |
1066 | 1066 | ||
1067 | out: | 1067 | out: |
@@ -1093,7 +1093,7 @@ static void free_coherent(struct device *dev, size_t size, | |||
1093 | 1093 | ||
1094 | __unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL); | 1094 | __unmap_single(iommu, domain->priv, dma_addr, size, DMA_BIDIRECTIONAL); |
1095 | 1095 | ||
1096 | if (iommu->need_sync) | 1096 | if (unlikely(iommu->need_sync)) |
1097 | iommu_completion_wait(iommu); | 1097 | iommu_completion_wait(iommu); |
1098 | 1098 | ||
1099 | spin_unlock_irqrestore(&domain->lock, flags); | 1099 | spin_unlock_irqrestore(&domain->lock, flags); |