| Message ID | 20230320180528.281755-1-jean-philippe@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | [RESEND] ACPI: VIOT: Initialize the correct IOMMU fwspec | expand |
Hi Jean, On 3/20/23 19:05, Jean-Philippe Brucker wrote: > When setting up DMA for a PCI device, we need to initialize its > iommu_fwspec with all possible alias RIDs (such as PCI bridges). To do > this we use pci_for_each_dma_alias() which calls > viot_pci_dev_iommu_init(). This function incorrectly initializes the > fwspec of the bridge instead of the device being configured. Fix it by > passing the original device as context to pci_for_each_dma_alias(). > > Fixes: 3cf485540e7b ("ACPI: Add driver for the VIOT table") > Reported-by: Eric Auger <eric.auger@redhat.com> > Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> Reviewed-by: Eric Auger <eric.auger@redhat.com> Tested-by: Eric Auger <eric.auger@redhat.com> With this patch the pcie-to-pci bridge is assigned an iommu group. The iommu group topology is not yet correct (ie. end points downstream to the pcie-to-pci bridge are put in a separate iommu group) however this is not related to that patch and not supposed to be fixed here (https://lore.kernel.org/all/4fead092-1058-198a-b430-3dee0fffcd51@arm.com/) Thanks Eric > --- > This fixes issue (1) reported here: > https://lore.kernel.org/all/Y8qzOKm6kvhGWG1T@myrica/ > > resend: Added linux-pci list > --- > drivers/acpi/viot.c | 5 +++-- > 1 file changed, 3 insertions(+), 2 deletions(-) > > diff --git a/drivers/acpi/viot.c b/drivers/acpi/viot.c > index ed752cbbe636..c8025921c129 100644 > --- a/drivers/acpi/viot.c > +++ b/drivers/acpi/viot.c > @@ -328,6 +328,7 @@ static int viot_pci_dev_iommu_init(struct pci_dev *pdev, u16 dev_id, void *data) > { > u32 epid; > struct viot_endpoint *ep; > + struct device *aliased_dev = data; > u32 domain_nr = pci_domain_nr(pdev->bus); > > list_for_each_entry(ep, &viot_pci_ranges, list) { > @@ -338,7 +339,7 @@ static int viot_pci_dev_iommu_init(struct pci_dev *pdev, u16 dev_id, void *data) > epid = ((domain_nr - ep->segment_start) << 16) + > dev_id - ep->bdf_start + ep->endpoint_id; > > - return viot_dev_iommu_init(&pdev->dev, ep->viommu, > + return viot_dev_iommu_init(aliased_dev, ep->viommu, > epid); > } > } > @@ -372,7 +373,7 @@ int viot_iommu_configure(struct device *dev) > { > if (dev_is_pci(dev)) > return pci_for_each_dma_alias(to_pci_dev(dev), > - viot_pci_dev_iommu_init, NULL); > + viot_pci_dev_iommu_init, dev); > else if (dev_is_platform(dev)) > return viot_mmio_dev_iommu_init(to_platform_device(dev)); > return -ENODEV;
On Wed, Mar 22, 2023 at 11:04 AM Eric Auger <eric.auger@redhat.com> wrote: > > Hi Jean, > > On 3/20/23 19:05, Jean-Philippe Brucker wrote: > > When setting up DMA for a PCI device, we need to initialize its > > iommu_fwspec with all possible alias RIDs (such as PCI bridges). To do > > this we use pci_for_each_dma_alias() which calls > > viot_pci_dev_iommu_init(). This function incorrectly initializes the > > fwspec of the bridge instead of the device being configured. Fix it by > > passing the original device as context to pci_for_each_dma_alias(). > > > > Fixes: 3cf485540e7b ("ACPI: Add driver for the VIOT table") > > Reported-by: Eric Auger <eric.auger@redhat.com> > > Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> > Reviewed-by: Eric Auger <eric.auger@redhat.com> > > Tested-by: Eric Auger <eric.auger@redhat.com> Applied as 6.4 material, thanks!
diff --git a/drivers/acpi/viot.c b/drivers/acpi/viot.c index ed752cbbe636..c8025921c129 100644 --- a/drivers/acpi/viot.c +++ b/drivers/acpi/viot.c @@ -328,6 +328,7 @@ static int viot_pci_dev_iommu_init(struct pci_dev *pdev, u16 dev_id, void *data) { u32 epid; struct viot_endpoint *ep; + struct device *aliased_dev = data; u32 domain_nr = pci_domain_nr(pdev->bus); list_for_each_entry(ep, &viot_pci_ranges, list) { @@ -338,7 +339,7 @@ static int viot_pci_dev_iommu_init(struct pci_dev *pdev, u16 dev_id, void *data) epid = ((domain_nr - ep->segment_start) << 16) + dev_id - ep->bdf_start + ep->endpoint_id; - return viot_dev_iommu_init(&pdev->dev, ep->viommu, + return viot_dev_iommu_init(aliased_dev, ep->viommu, epid); } } @@ -372,7 +373,7 @@ int viot_iommu_configure(struct device *dev) { if (dev_is_pci(dev)) return pci_for_each_dma_alias(to_pci_dev(dev), - viot_pci_dev_iommu_init, NULL); + viot_pci_dev_iommu_init, dev); else if (dev_is_platform(dev)) return viot_mmio_dev_iommu_init(to_platform_device(dev)); return -ENODEV;
When setting up DMA for a PCI device, we need to initialize its iommu_fwspec with all possible alias RIDs (such as PCI bridges). To do this we use pci_for_each_dma_alias() which calls viot_pci_dev_iommu_init(). This function incorrectly initializes the fwspec of the bridge instead of the device being configured. Fix it by passing the original device as context to pci_for_each_dma_alias(). Fixes: 3cf485540e7b ("ACPI: Add driver for the VIOT table") Reported-by: Eric Auger <eric.auger@redhat.com> Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org> --- This fixes issue (1) reported here: https://lore.kernel.org/all/Y8qzOKm6kvhGWG1T@myrica/ resend: Added linux-pci list --- drivers/acpi/viot.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)