Reviewed-by: Xiongfeng Wang <wangxiongfeng2@huawei.com> On 2021/6/18 16:00, LeoLiuoc wrote:

When using ACPI device to create identity mapping, if the physical node of the ACPI device is a pci device, domain_1 will be created for these two devices. But if the pci device and other pci devices belong to another domain_2, this will happen conflict and destroy another domain. Such as PCI devices under the PCIE-to-PCI bridge. Therefore, when the physical node of the ACPI device is a PCI device, this patch uses the PCI device to create the domain. In this way, there is only one domain.

Signed-off-by: LeoLiu-oc <LeoLiu-oc@zhaoxin.com> ---  v1->v2: fix some logic mistake

 drivers/iommu/intel-iommu.c | 21 +++++++++++++++++++++  1 file changed, 21 insertions(+)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c index e111e5d6d269..453ec0eb3a4f 100644 --- a/drivers/iommu/intel-iommu.c +++ b/drivers/iommu/intel-iommu.c @@ -2736,12 +2736,33 @@ static int domain_prepare_identity_map(struct device *dev,      return iommu_domain_identity_map(domain, start, end);  }

+static struct device *acpi_dev_find_pci_dev(struct device *dev) +{ +    struct acpi_device_physical_node *pn; +    struct acpi_device *adev; + +    if (dev->bus == &acpi_bus_type) { +        adev = to_acpi_device(dev); +        mutex_lock(&adev->physical_node_lock); +        list_for_each_entry(pn, &adev->physical_node_list, node) { +            if (dev_is_pci(pn->dev)) { +                mutex_unlock(&adev->physical_node_lock); +                return pn->dev; +            } +        } +        mutex_unlock(&adev->physical_node_lock); +    } + +    return dev; +} +  static int iommu_prepare_identity_map(struct device *dev,                        unsigned long long start,                        unsigned long long end)  {      struct dmar_domain *domain;      int ret; +    dev = acpi_dev_find_pci_dev(dev);

     domain = get_domain_for_dev(dev, DEFAULT_DOMAIN_ADDRESS_WIDTH);      if (!domain)