[PATCH V9 11/11] powerpc/powernv: compound PE for VFs

[Wei Yang]

Hi, Michael

Hope you didn't take this yet. We may change this patch a little.

On Fri, Jul 17, 2015 at 02:02:41PM +0800, Wei Yang wrote:
>When VF BAR size is larger than 64MB, we group VFs in terms of M64 BAR,
>which means those VFs in a group should form a compound PE.
>
>This patch links those VF PEs into compound PE in this case.
>
>[gwshan: code refactoring for a bit]
>Signed-off-by: Wei Yang <weiyang at linux.vnet.ibm.com>
>Acked-by: Gavin Shan <gwshan at linux.vnet.ibm.com>
>---
> arch/powerpc/platforms/powernv/pci-ioda.c |   46 +++++++++++++++++++++++++----
> arch/powerpc/platforms/powernv/pci.c      |   17 +++++++++--
> 2 files changed, 56 insertions(+), 7 deletions(-)
>
>diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
>index 5738d31..d1530cb 100644
>--- a/arch/powerpc/platforms/powernv/pci-ioda.c
>+++ b/arch/powerpc/platforms/powernv/pci-ioda.c
>@@ -1359,9 +1359,20 @@ static void pnv_ioda_release_vf_PE(struct pci_dev *pdev, u16 num_vfs)
> 	}
>
> 	list_for_each_entry_safe(pe, pe_n, &phb->ioda.pe_list, list) {
>+		struct pnv_ioda_pe *s, *sn;
> 		if (pe->parent_dev != pdev)
> 			continue;
>
>+		if ((pe->flags & PNV_IODA_PE_MASTER) &&
>+		    (pe->flags & PNV_IODA_PE_VF)) {
>+			list_for_each_entry_safe(s, sn, &pe->slaves, list) {
>+				pnv_pci_ioda2_release_dma_pe(pdev, s);
>+				list_del(&s->list);
>+				pnv_ioda_deconfigure_pe(phb, s);
>+				pnv_ioda_free_pe(phb, s->pe_number);
>+			}
>+		}
>+
> 		pnv_pci_ioda2_release_dma_pe(pdev, pe);
>
> 		/* Remove from list */
>@@ -1414,7 +1425,7 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs)
> 	struct pci_bus        *bus;
> 	struct pci_controller *hose;
> 	struct pnv_phb        *phb;
>-	struct pnv_ioda_pe    *pe;
>+	struct pnv_ioda_pe    *pe, *master_pe;
> 	int                    pe_num;
> 	u16                    vf_index;
> 	struct pci_dn         *pdn;
>@@ -1456,10 +1467,13 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs)
> 			continue;
> 		}
>
>-		/* Put PE to the list */
>-		mutex_lock(&phb->ioda.pe_list_mutex);
>-		list_add_tail(&pe->list, &phb->ioda.pe_list);
>-		mutex_unlock(&phb->ioda.pe_list_mutex);
>+		/* Put PE to the list, or postpone it for compound PEs */
>+		if ((pdn->m64_per_iov != M64_PER_IOV) ||
>+		    (num_vfs <= M64_PER_IOV)) {
>+			mutex_lock(&phb->ioda.pe_list_mutex);
>+			list_add_tail(&pe->list, &phb->ioda.pe_list);
>+			mutex_unlock(&phb->ioda.pe_list_mutex);
>+		}
>
> 		pnv_pci_ioda2_setup_dma_pe(phb, pe);
> 	}
>@@ -1472,10 +1486,32 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs)
> 		vf_per_group = roundup_pow_of_two(num_vfs) / pdn->m64_per_iov;
>
> 		for (vf_group = 0; vf_group < M64_PER_IOV; vf_group++) {
>+			master_pe = NULL;
>+
> 			for (vf_index = vf_group * vf_per_group;
> 			     vf_index < (vf_group + 1) * vf_per_group &&
> 			     vf_index < num_vfs;
> 			     vf_index++) {
>+
>+				/*
>+				 * Figure out the master PE and put all slave
>+				 * PEs to master PE's list.
>+				 */
>+				pe = &phb->ioda.pe_array[pdn->offset + vf_index];
>+				if (!master_pe) {
>+					pe->flags |= PNV_IODA_PE_MASTER;
>+					INIT_LIST_HEAD(&pe->slaves);
>+					master_pe = pe;
>+					mutex_lock(&phb->ioda.pe_list_mutex);
>+					list_add_tail(&pe->list, &phb->ioda.pe_list);
>+					mutex_unlock(&phb->ioda.pe_list_mutex);
>+				} else {
>+					pe->flags |= PNV_IODA_PE_SLAVE;
>+					pe->master = master_pe;
>+					list_add_tail(&pe->list,
>+						&master_pe->slaves);
>+				}
>+
> 				for (vf_index1 = vf_group * vf_per_group;
> 				     vf_index1 < (vf_group + 1) * vf_per_group &&
> 				     vf_index1 < num_vfs;
>diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c
>index 0e4f42e..f3aead0 100644
>--- a/arch/powerpc/platforms/powernv/pci.c
>+++ b/arch/powerpc/platforms/powernv/pci.c
>@@ -739,7 +739,7 @@ void pnv_pci_dma_dev_setup(struct pci_dev *pdev)
> 	struct pci_controller *hose = pci_bus_to_host(pdev->bus);
> 	struct pnv_phb *phb = hose->private_data;
> #ifdef CONFIG_PCI_IOV
>-	struct pnv_ioda_pe *pe;
>+	struct pnv_ioda_pe *pe, *slave;
> 	struct pci_dn *pdn;
>
> 	/* Fix the VF pdn PE number */
>@@ -751,10 +751,23 @@ void pnv_pci_dma_dev_setup(struct pci_dev *pdev)
> 			    (pdev->devfn & 0xff))) {
> 				pdn->pe_number = pe->pe_number;
> 				pe->pdev = pdev;
>-				break;
>+				goto found;
>+			}
>+
>+			if ((pe->flags & PNV_IODA_PE_MASTER) &&
>+			    (pe->flags & PNV_IODA_PE_VF)) {
>+				list_for_each_entry(slave, &pe->slaves, list) {
>+					if (slave->rid == ((pdev->bus->number << 8)
>+					   | (pdev->devfn & 0xff))) {
>+						pdn->pe_number = slave->pe_number;
>+						slave->pdev = pdev;
>+						goto found;
>+					}
>+				}
> 			}
> 		}
> 	}
>+found:
> #endif /* CONFIG_PCI_IOV */
>
> 	if (phb && phb->dma_dev_setup)
>-- 
>1.7.9.5

-- 
Richard Yang
Help you, Help me