| Message ID | 1398047119-6861-2-git-send-email-weiyang@linux.vnet.ibm.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 3f28c5af3964c11e61e9a58df77cae5ebdb8209e |
| Headers | show |
On Mon, 2014-04-21 at 10:25 +0800, Wei Yang wrote: > Here is a draft call flow: > > pci_device_add > pcibios_add_device > pci_dma_dev_setup > pnv_pci_dma_dev_setup > pnv_pci_ioda_dma_dev_setup > set_iommu_table_base_and_group <--- here > device_add > > When set_iommu_table_base_and_group() is invoked int > pnv_pci_ioda_dma_dev_setup(), the dev->kobj->sd is not initialized. The > dev->kobj->sd is initialized in device_add(). > > After applying this patch, the error > iommu_tce: 0003:05:00.0 has not been added, ret=-14 > is cleared. > > This patch revert the change for pnv_pci_ioda_dma_dev_setup() from commit > d905c5df(PPC: POWERNV: move iommu_add_device earlier). But in that case, is the group still set ? Alexey, is that correct ? Cheers, Ben. > Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com> > --- > arch/powerpc/platforms/powernv/pci-ioda.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c > index fc4edda..b5d8f73 100644 > --- a/arch/powerpc/platforms/powernv/pci-ioda.c > +++ b/arch/powerpc/platforms/powernv/pci-ioda.c > @@ -1425,7 +1425,7 @@ static void pnv_pci_ioda_dma_dev_setup(struct pnv_phb *phb, struct pci_dev *pdev > > pe = &phb->ioda.pe_array[pdn->pe_number]; > WARN_ON(get_dma_ops(&pdev->dev) != &dma_iommu_ops); > - set_iommu_table_base_and_group(&pdev->dev, &pe->tce32_table); > + set_iommu_table_base(&pdev->dev, &pe->tce32_table); > } > > static int pnv_pci_ioda_dma_set_mask(struct pnv_phb *phb,
On Mon, Apr 21, 2014 at 01:37:32PM +1000, Benjamin Herrenschmidt wrote: >On Mon, 2014-04-21 at 10:25 +0800, Wei Yang wrote: >> Here is a draft call flow: >> >> pci_device_add >> pcibios_add_device >> pci_dma_dev_setup >> pnv_pci_dma_dev_setup >> pnv_pci_ioda_dma_dev_setup >> set_iommu_table_base_and_group <--- here >> device_add >> >> When set_iommu_table_base_and_group() is invoked int >> pnv_pci_ioda_dma_dev_setup(), the dev->kobj->sd is not initialized. The >> dev->kobj->sd is initialized in device_add(). >> >> After applying this patch, the error >> iommu_tce: 0003:05:00.0 has not been added, ret=-14 >> is cleared. >> >> This patch revert the change for pnv_pci_ioda_dma_dev_setup() from commit >> d905c5df(PPC: POWERNV: move iommu_add_device earlier). > >But in that case, is the group still set ? I think so. iommu_add_device() will be invoked when device_add() notify the pci bus. > >Alexey, is that correct ? > >Cheers, >Ben. > >> Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com> >> --- >> arch/powerpc/platforms/powernv/pci-ioda.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c >> index fc4edda..b5d8f73 100644 >> --- a/arch/powerpc/platforms/powernv/pci-ioda.c >> +++ b/arch/powerpc/platforms/powernv/pci-ioda.c >> @@ -1425,7 +1425,7 @@ static void pnv_pci_ioda_dma_dev_setup(struct pnv_phb *phb, struct pci_dev *pdev >> >> pe = &phb->ioda.pe_array[pdn->pe_number]; >> WARN_ON(get_dma_ops(&pdev->dev) != &dma_iommu_ops); >> - set_iommu_table_base_and_group(&pdev->dev, &pe->tce32_table); >> + set_iommu_table_base(&pdev->dev, &pe->tce32_table); >> } >> >> static int pnv_pci_ioda_dma_set_mask(struct pnv_phb *phb, >
On Mon, Apr 21, 2014 at 01:22:13PM +0800, Wei Yang wrote: >On Mon, Apr 21, 2014 at 01:37:32PM +1000, Benjamin Herrenschmidt wrote: >>On Mon, 2014-04-21 at 10:25 +0800, Wei Yang wrote: >>> Here is a draft call flow: >>> >>> pci_device_add >>> pcibios_add_device >>> pci_dma_dev_setup >>> pnv_pci_dma_dev_setup >>> pnv_pci_ioda_dma_dev_setup >>> set_iommu_table_base_and_group <--- here >>> device_add >>> >>> When set_iommu_table_base_and_group() is invoked int >>> pnv_pci_ioda_dma_dev_setup(), the dev->kobj->sd is not initialized. The >>> dev->kobj->sd is initialized in device_add(). >>> >>> After applying this patch, the error >>> iommu_tce: 0003:05:00.0 has not been added, ret=-14 >>> is cleared. >>> >>> This patch revert the change for pnv_pci_ioda_dma_dev_setup() from commit >>> d905c5df(PPC: POWERNV: move iommu_add_device earlier). >> >>But in that case, is the group still set ? > >I think so. > >iommu_add_device() will be invoked when device_add() notify the pci bus. > Did a quick test, after applying this patch, the logic would be the same for VF and PF hotplug, the iommu_group is created with corresponding device.
diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c index fc4edda..b5d8f73 100644 --- a/arch/powerpc/platforms/powernv/pci-ioda.c +++ b/arch/powerpc/platforms/powernv/pci-ioda.c @@ -1425,7 +1425,7 @@ static void pnv_pci_ioda_dma_dev_setup(struct pnv_phb *phb, struct pci_dev *pdev pe = &phb->ioda.pe_array[pdn->pe_number]; WARN_ON(get_dma_ops(&pdev->dev) != &dma_iommu_ops); - set_iommu_table_base_and_group(&pdev->dev, &pe->tce32_table); + set_iommu_table_base(&pdev->dev, &pe->tce32_table); } static int pnv_pci_ioda_dma_set_mask(struct pnv_phb *phb,
Here is a draft call flow: pci_device_add pcibios_add_device pci_dma_dev_setup pnv_pci_dma_dev_setup pnv_pci_ioda_dma_dev_setup set_iommu_table_base_and_group <--- here device_add When set_iommu_table_base_and_group() is invoked int pnv_pci_ioda_dma_dev_setup(), the dev->kobj->sd is not initialized. The dev->kobj->sd is initialized in device_add(). After applying this patch, the error iommu_tce: 0003:05:00.0 has not been added, ret=-14 is cleared. This patch revert the change for pnv_pci_ioda_dma_dev_setup() from commit d905c5df(PPC: POWERNV: move iommu_add_device earlier). Signed-off-by: Wei Yang <weiyang@linux.vnet.ibm.com> --- arch/powerpc/platforms/powernv/pci-ioda.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)