On Nahelem systems, there are 43 Host Bridges exposed to the kernel
for each System Socket. With a large socket count, and the pass
through option for iommu is set, too much time is spent in the
identity_mapping function, hunting though the iommu domains to
check if a specific device is "identity mapped".
Speed up the function by setting an "is_identity_mapped" flag in
the pci_dev struct, if this device is mapped to the static identity
domain.
Signed-off-by: Mike Travis <[email protected]>
---
drivers/pci/intel-iommu.c | 11 +++++------
include/linux/pci.h | 1 +
2 files changed, 6 insertions(+), 6 deletions(-)
--- linux.orig/drivers/pci/intel-iommu.c
+++ linux/drivers/pci/intel-iommu.c
@@ -2102,11 +2102,7 @@ static int identity_mapping(struct pci_d
if (likely(!iommu_identity_mapping))
return 0;
-
- list_for_each_entry(info, &si_domain->devices, link)
- if (info->dev == pdev)
- return 1;
- return 0;
+ return pdev->is_identity_mapped;
}
static int domain_add_dev_info(struct dmar_domain *domain,
@@ -2138,6 +2134,7 @@ static int domain_add_dev_info(struct dm
list_add(&info->global, &device_domain_list);
pdev->dev.archdata.iommu = info;
spin_unlock_irqrestore(&device_domain_lock, flags);
+ pdev->is_identity_mapped = (domain == si_domain);
return 0;
}
@@ -3404,8 +3401,10 @@ static void domain_remove_one_dev_info(s
* update iommu count and coherency
*/
if (iommu == device_to_iommu(info->segment, info->bus,
- info->devfn))
+ info->devfn)) {
found = 1;
+ pdev->is_identity_mapped = 0;
+ }
}
if (found == 0) {
--- linux.orig/include/linux/pci.h
+++ linux/include/linux/pci.h
@@ -314,6 +314,7 @@ struct pci_dev {
unsigned int is_virtfn:1;
unsigned int reset_fn:1;
unsigned int is_hotplug_bridge:1;
+ unsigned int is_identity_mapped:1;
unsigned int __aer_firmware_first_valid:1;
unsigned int __aer_firmware_first:1;
pci_dev_flags_t dev_flags;
--
* Mike Travis ([email protected]) wrote:
> @@ -2138,6 +2134,7 @@ static int domain_add_dev_info(struct dm
> list_add(&info->global, &device_domain_list);
> pdev->dev.archdata.iommu = info;
The si_domain is cached right here...can you not use this instead of
adding an iommu impl. specific bit to pci_dev?
> spin_unlock_irqrestore(&device_domain_lock, flags);
> + pdev->is_identity_mapped = (domain == si_domain);
Chris Wright wrote:
> * Mike Travis ([email protected]) wrote:
>> @@ -2138,6 +2134,7 @@ static int domain_add_dev_info(struct dm
>> list_add(&info->global, &device_domain_list);
>> pdev->dev.archdata.iommu = info;
>
> The si_domain is cached right here...can you not use this instead of
> adding an iommu impl. specific bit to pci_dev?
>
>> spin_unlock_irqrestore(&device_domain_lock, flags);
>> + pdev->is_identity_mapped = (domain == si_domain);
Yes, it looks like I came at the solution backwards... I added
the bit then figured out how to set it. I'll update that patch.
Subject: Intel iommu: Speed up processing of the identity_mapping function
When there are a large count of PCI devices, and the pass
through option for iommu is set, much time is spent in the
identity_mapping function hunting though the iommu domains to
check if a specific device is "identity mapped".
Speed up the function by checking the cached info to see if
it's mapped to the static identity domain.
Signed-off-by: Mike Travis <[email protected]>
---
drivers/pci/intel-iommu.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
--- linux-2.6.32.orig/drivers/pci/intel-iommu.c
+++ linux-2.6.32/drivers/pci/intel-iommu.c
@@ -2124,10 +2124,10 @@ static int identity_mapping(struct pci_d
if (likely(!iommu_identity_mapping))
return 0;
+ info = pdev->dev.archdata.iommu;
+ if (info && info != DUMMY_DEVICE_DOMAIN_INFO)
+ return (info->domain == si_domain);
- list_for_each_entry(info, &si_domain->devices, link)
- if (info->dev == pdev)
- return 1;
return 0;
}