iommu/vt-d: Restore context entry setup order for aliased devices
Subsystems (2)
Lu Baolu
maintainerPATCH 1/1 iommu/vt-d: Restore context entry setup order for aliased devices
Commit 2031c469f816 ("iommu/vt-d: Add support for static identity domain") changed the context entry setup during domain attachment from a set-and-check policy to a clear-and-reset approach. This inadvertently introduced a regression affecting PCI aliased devices behind PCIe-to-PCI bridges.
Specifically, keyboard and touchpad stopped working on several Apple Macbooks with below messages:
kernel: platform pxa2xx-spi.3: Adding to iommu group 20
kernel: input: Apple SPI Keyboard as /devices/pci0000:00/0000:00:1e.3/pxa2xx-spi.3/spi_master/spi2/spi-APP000D:00/input/input0
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA Read NO_PASID] Request device [00:1e.3] fault addr 0xffffa000 [fault reason 0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA Read NO_PASID] Request device [00:1e.3] fault addr 0xffffa000 [fault reason 0x06] PTE Read access is not set
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA Read NO_PASID] Request device [00:1e.3] fault addr 0xffffa000 [fault reason 0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00
Fix this by restoring the previous context setup order.
--- drivers/iommu/intel/iommu.c | 11 +++++++++++ drivers/iommu/intel/iommu.h | 1 + drivers/iommu/intel/nested.c | 4 ++-- 3 files changed, 14 insertions(+), 2 deletions(-)
drivers/iommu/intel/iommu.c +11 -0
@@ -1809,6 +1809,7 @@ static int dmar_domain_attach_device(struct dmar_domain *domain,return ret; info->domain = domain; + info->domain_attached = true;spin_lock_irqsave (&domain->lock, flags);list_add (&info->link, &domain->devices);spin_unlock_irqrestore (&domain->lock, flags);
@@ -3206,6 +3207,10 @@ void device_block_translation(struct device *dev)struct intel_iommu *iommu = info->iommu;unsigned long flags; +/* Device in DMA blocking state. Noting to do. */ +if (!info->domain_attached) +return ; +if (info->domain)cache_tag_unassign_domain (info->domain, dev,IOMMU_NO_PASID );
@@ -3217,6 +3222,9 @@ void device_block_translation(struct device *dev)domain_context_clear (info); } +/* Device now in DMA blocking state. */ + info->domain_attached = false; +if (!info->domain)return ;
@@ -4302,6 +4310,9 @@ static int identity_domain_attach_dev(struct iommu_domain *domain, struct deviceelse ret =device_setup_pass_through (dev); +if (!ret) + info->domain_attached = true; +return ret; }
drivers/iommu/intel/iommu.h +1 -0
@@ -774,6 +774,7 @@ struct device_domain_info { u8 ats_supported:1; u8 ats_enabled:1; u8 dtlb_extra_inval:1; /* Quirk for devices need extra flush */ + u8 domain_attached:1; /* Device has domain attached */u8 ats_qdep;unsigned int iopf_refcount;struct device *dev;/* it's NULL for PCIe-to-PCI bridge */
drivers/iommu/intel/nested.c +2 -2
@@ -27,8 +27,7 @@ static int intel_nested_attach_dev(struct iommu_domain *domain,unsigned long flags;int ret =0 ; -if (info->domain) -device_block_translation (dev); +device_block_translation (dev);if (iommu->agaw < dmar_domain->s2_domain->agaw) {dev_err_ratelimited (dev,"Adjusted guest address width not compatible\n" );
@@ -66,6 +65,7 @@ static int intel_nested_attach_dev(struct iommu_domain *domain, goto disable_iopf; info->domain = dmar_domain; + info->domain_attached = true;spin_lock_irqsave (&dmar_domain->lock, flags);list_add (&info->link, &dmar_domain->devices);spin_unlock_irqrestore (&dmar_domain->lock, flags);
Signature · 1 line
2.43.0
Quoted 29 lines
From: Lu Baolu <[email protected]>
Sent: Wednesday, May 14, 2025 2:05 PMCommit
2031c469f816("iommu/vt-d: Add support for static identity domain") changed the context entry setup during domain attachment from a set-and-check policy to a clear-and-reset approach. This inadvertently introduced a regression affecting PCI aliased devices behind PCIe-to-PCI bridges.Specifically, keyboard and touchpad stopped working on several Apple Macbooks with below messages:
kernel: platform pxa2xx-spi.3: Adding to iommu group 20
kernel: input: Apple SPI Keyboard as /devices/pci0000:00/0000:00:1e.3/pxa2xx-spi.3/spi_master/spi2/spi-
APP000D:00/input/input0
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00
Quoted 6 lines
Fix this by restoring the previous context setup order.
Fixes 2031c469f816 ("iommu/vt-d: Add support for static identity domain")Signed-off-by Lu Baolu <[email protected]>
Hi Baolu,
Quoted 5 lines · Lu Baolu
Commit
2031c469f816("iommu/vt-d: Add support for static identity domain") changed the context entry setup during domain attachment from a set-and-check policy to a clear-and-reset approach. This inadvertently introduced a regression affecting PCI aliased devices behind PCIe-to-PCI bridges.
I got what the patch does. But just bit confused on the above description. I didn't see the commit 2031c469f816 mentioned any policy thing on the context entry setup. To me, the problem looks to be that the info->domain is no more accurate to be used for checking if any domain is attached after the above commit. Maybe I missed something. feel free correct me.
Quoted 90 lines
Specifically, keyboard and touchpad stopped working on several Apple Macbooks with below messages:
kernel: platform pxa2xx-spi.3: Adding to iommu group 20
kernel: input: Apple SPI Keyboard as /devices/pci0000:00/0000:00:1e.3/pxa2xx-spi.3/spi_master/spi2/spi-APP000D:00/input/input0
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00
kernel: DMAR: DRHD: handling fault status reg 3
kernel: DMAR: [DMA ReadNO_PASID] Request device [00:1e.3] fault addr0xffffa000[fault reason0x06] PTE Read access is not set
kernel: DMAR: DRHD: handling fault status reg 3
kernel: applespi spi-APP000D:00: Error writing to device: 01 0e 00 00Fix this by restoring the previous context setup order.
Fixes 2031c469f816 ("iommu/vt-d: Add support for static identity domain")Signed-off-by Lu Baolu <[email protected]>--- drivers/iommu/intel/iommu.c | 11 +++++++++++ drivers/iommu/intel/iommu.h | 1 + drivers/iommu/intel/nested.c | 4 ++-- 3 files changed, 14 insertions(+), 2 deletions(-)drivers/iommu/intel/iommu.c +11
-0@@ -1809,6 +1809,7 @@ static int dmar_domain_attach_device(struct dmar_domain *domain,return ret; info->domain = domain; + info->domain_attached = true;spin_lock_irqsave (&domain->lock, flags);list_add (&info->link, &domain->devices);spin_unlock_irqrestore (&domain->lock, flags);@@ -3206,6 +3207,10 @@ void device_block_translation(struct device *dev)struct intel_iommu *iommu = info->iommu;unsigned long flags; +/* Device in DMA blocking state. Noting to do. */ +if (!info->domain_attached) +return ; +if (info->domain)cache_tag_unassign_domain (info->domain, dev,IOMMU_NO_PASID );@@ -3217,6 +3222,9 @@ void device_block_translation(struct device *dev)domain_context_clear (info); } +/* Device now in DMA blocking state. */ + info->domain_attached = false; +if (!info->domain)return ;@@ -4302,6 +4310,9 @@ static int identity_domain_attach_dev(struct iommu_domain *domain, struct deviceelse ret =device_setup_pass_through (dev); +if (!ret) + info->domain_attached = true; +return ret; }drivers/iommu/intel/iommu.h +1
-0@@ -774,6 +774,7 @@ struct device_domain_info { u8 ats_supported:1; u8 ats_enabled:1; u8 dtlb_extra_inval:1; /* Quirk for devices need extra flush */ + u8 domain_attached:1; /* Device has domain attached */u8 ats_qdep;unsigned int iopf_refcount;struct device *dev;/* it's NULL for PCIe-to-PCI bridge */ drivers/iommu/intel/nested.c +0
-0
drivers/iommu/intel/nested.c static int intel_nested_attach_dev(struct iommu_domain *domain, · line 30
Quoted 6 lines
drivers/iommu/intel/nested.c +1
-2@@ -27,8 +27,7 @@ static int intel_nested_attach_dev(struct iommu_domain *domain,unsigned long flags;int ret =0 ; -if (info->domain) -device_block_translation (dev); +device_block_translation (dev);
good catch. :)
drivers/iommu/intel/nested.c static int intel_nested_attach_dev(struct iommu_domain *domain, · line 71
Quoted 10 lines
drivers/iommu/intel/nested.c +1
-0if (iommu->agaw < dmar_domain->s2_domain->agaw) {dev_err_ratelimited (dev,"Adjusted guest address width not compatible\n" );@@ -66,6 +65,7 @@ static int intel_nested_attach_dev(struct iommu_domain *domain, goto disable_iopf; info->domain = dmar_domain; + info->domain_attached = true;spin_lock_irqsave (&dmar_domain->lock, flags);list_add (&info->link, &dmar_domain->devices);spin_unlock_irqrestore (&dmar_domain->lock, flags);
LGTM.
Regards,
Yi Liu
Baolu Lu
maintainerQuoted 8 lines · Yi Liu
Quoted 5 lines · Lu Baolu
Commit
2031c469f816("iommu/vt-d: Add support for static identity domain") changed the context entry setup during domain attachment from a set-and-check policy to a clear-and-reset approach. This inadvertently introduced a regression affecting PCI aliased devices behind PCIe-to-PCI bridges.
Quoted 5 lines
I got what the patch does. But just bit confused on the above description. I didn't see the commit
2031c469f816mentioned any policy thing on the context entry setup. To me, the problem looks to be that theinfo->domainis no more accurate to be used for checking if any domain is attached after the above commit. Maybe I missed something. feel free correct me.
The problem was introduced by below change:
drivers/iommu/intel/iommu.c +1 -3
@@ -3691,11 +3691,9 @@ int prepare_domain_attach_device(struct iommu_domain *domain,static int intel_iommu_attach_device (struct iommu_domain *domain,struct device *dev) { -struct device_domain_info *info =dev_iommu_priv_get (dev);int ret; -if (info->domain) -device_block_translation (dev); +device_block_translation (dev);
And after the introduction of static identity domain, "info->domain == NULL" doesn't mean no domain attaching to device anymore. So this patch uses a specific flag bit to indicate this.
Thanks,
baolu