[PATCH 2/2] vfio/pci: Introduce OpenCAPI devices support.
Alexey Kardashevskiy
aik at ozlabs.ru
Mon Nov 11 15:17:59 AEDT 2019
On 25/10/2019 00:28, christophe lombard wrote:
> This patch adds new IOCTL commands for VFIO PCI driver to support
> configuration and management for OpenCAPI devices, which have been passed
> through from host to QEMU VFIO.
So far we managed to keep all IBM POWER specific inside the IOMMU
subdriver, why is this case different? You really have to outline what
OpenCAPI does and why simple passing through does not work and how the
guest will trigger these ioctls() in QEMU, are you adding hypercalls?
What driver do you expect to work with this in the guest?
arch/powerpc/platforms/powernv/ocxl.c is a powernv-tied driver which
calls into OPAL so it won't work on pseries.
> OpenCAPI (Open Coherent Accelerator Processor Interface) is an interface
> between processors and accelerators.
>
> The main IOCTL command is:
> VFIO_DEVICE_OCXL_OP Handles devices, which supports the OpenCAPI
> interface, using the ocxl pnv_* interface.
>
> The following commands are supported, based on the hcalls defined
> in ocxl/pseries.c that implements the guest-specific callbacks.
> VFIO_DEVICE_OCXL_CONFIG_ADAPTER Used to configure OpenCAPI adapter
> characteristics.
>
> VFIO_DEVICE_OCXL_CONFIG_SPA Used to configure the schedule process
> area (SPA) table for an OpenCAPI device.
>
> VFIO_DEVICE_OCXL_GET_FAULT_STATE Used to retrieve fault information
> from an OpenCAPI device.
>
> VFIO_DEVICE_OCXL_HANDLE_FAULT Used to respond to an OpenCAPI fault.
>
> The platform data is declared in the vfio_pci_ocxl_link which is common
> for each devices sharing the same domain, same bus and same slot.
So this can be IOMMU attributes, no?
> The lpid value, requested to configure the process element in the
> Scheduled Process Area, is not available in the QEMU environment.
> This implies getting it from the host through the iommu group.
lpid does not change after the device is assigned so you can just do
what drivers/vfio/pci/vfio_pci_nvlink2.c does (set it once per IOMMU
group when KVM is attached to a group) and skip the vfio_dev_get_kvm()
part. Thanks,
> Signed-off-by: Christophe Lombard <clombard at linux.vnet.ibm.com>
> ---
> drivers/vfio/pci/Kconfig | 7 +
> drivers/vfio/pci/Makefile | 1 +
> drivers/vfio/pci/vfio_pci.c | 19 ++
> drivers/vfio/pci/vfio_pci_ocxl.c | 287 +++++++++++++++++++++++++++++++
> drivers/vfio/vfio.c | 25 +++
> include/linux/vfio.h | 13 ++
> include/uapi/linux/vfio.h | 22 +++
> 7 files changed, 374 insertions(+)
> create mode 100644 drivers/vfio/pci/vfio_pci_ocxl.c
>
> diff --git a/drivers/vfio/pci/Kconfig b/drivers/vfio/pci/Kconfig
> index ac3c1dd3edef..fd3716d10ded 100644
> --- a/drivers/vfio/pci/Kconfig
> +++ b/drivers/vfio/pci/Kconfig
> @@ -45,3 +45,10 @@ config VFIO_PCI_NVLINK2
> depends on VFIO_PCI && PPC_POWERNV
> help
> VFIO PCI support for P9 Witherspoon machine with NVIDIA V100 GPUs
> +
> +config VFIO_PCI_OCXL
> + depends on VFIO_PCI
> + def_bool y if OCXL_BASE
> + help
> + VFIO PCI support for devices which handle the Open Coherent
> + Accelerator Processor Interface.
> diff --git a/drivers/vfio/pci/Makefile b/drivers/vfio/pci/Makefile
> index f027f8a0e89c..6d55a5fee4b0 100644
> --- a/drivers/vfio/pci/Makefile
> +++ b/drivers/vfio/pci/Makefile
> @@ -3,5 +3,6 @@
> vfio-pci-y := vfio_pci.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
> vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
> vfio-pci-$(CONFIG_VFIO_PCI_NVLINK2) += vfio_pci_nvlink2.o
> +vfio-pci-$(CONFIG_VFIO_PCI_OCXL) += vfio_pci_ocxl.o
>
> obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
> diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
> index 703948c9fbe1..4f9741bbe790 100644
> --- a/drivers/vfio/pci/vfio_pci.c
> +++ b/drivers/vfio/pci/vfio_pci.c
> @@ -1128,6 +1128,25 @@ static long vfio_pci_ioctl(void *device_data,
>
> return vfio_pci_ioeventfd(vdev, ioeventfd.offset,
> ioeventfd.data, count, ioeventfd.fd);
> + } else if (cmd == VFIO_DEVICE_OCXL_OP) {
> + struct vfio_device_ocxl_op ocxl_op;
> + int ret = 0;
> +
> + minsz = offsetofend(struct vfio_device_ocxl_op, data);
> +
> + if (copy_from_user(&ocxl_op, (void __user *)arg, minsz))
> + return -EFAULT;
> +
> + if (ocxl_op.argsz < minsz)
> + return -EINVAL;
> +
> + ret = vfio_pci_ocxl_ioctl(vdev->pdev, &ocxl_op);
> +
> + if (!ret) {
> + if (copy_to_user((void __user *)arg, &ocxl_op, minsz))
> + ret = -EFAULT;
> + }
> + return ret;
> }
>
> return -ENOTTY;
> diff --git a/drivers/vfio/pci/vfio_pci_ocxl.c b/drivers/vfio/pci/vfio_pci_ocxl.c
> new file mode 100644
> index 000000000000..cb5cd4fb416d
> --- /dev/null
> +++ b/drivers/vfio/pci/vfio_pci_ocxl.c
> @@ -0,0 +1,287 @@
> +// SPDX-License-Identifier: GPL-2.0+
> +// Copyright 2019 IBM Corp.
> +
> +#include <asm/kvm_ppc.h>
> +#include <asm/pnv-ocxl.h>
> +#include <linux/vfio.h>
> +#include <linux/slab.h>
> +#include <linux/pci.h>
> +#include <linux/kvm_host.h>
> +
> +struct vfio_device_ocxl_link {
> + struct list_head list;
> + int domain;
> + int bus;
> + int slot;
> + void *platform_data;
> +};
> +static struct list_head links_list = LIST_HEAD_INIT(links_list);
> +static DEFINE_MUTEX(links_list_lock);
> +
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER 1
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_SETUP 1
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_RELEASE 2
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_GET_ACTAG 3
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_GET_PASID 4
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_SET_TL 5
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_ALLOC_IRQ 6
> +#define VFIO_DEVICE_OCXL_CONFIG_ADAPTER_FREE_IRQ 7
> +
> +#define VFIO_DEVICE_OCXL_CONFIG_SPA 2
> +#define VFIO_DEVICE_OCXL_CONFIG_SPA_SET 1
> +#define VFIO_DEVICE_OCXL_CONFIG_SPA_UPDATE 2
> +#define VFIO_DEVICE_OCXL_CONFIG_SPA_REMOVE 3
> +
> +#define VFIO_DEVICE_OCXL_GET_FAULT_STATE 3
> +#define VFIO_DEVICE_OCXL_HANDLE_FAULT 4
> +
> +static struct vfio_device_ocxl_link *find_link(struct pci_dev *pdev)
> +{
> + struct vfio_device_ocxl_link *link;
> +
> + list_for_each_entry(link, &links_list, list) {
> + /* The functions of a device all share the same link */
> + if (link->domain == pci_domain_nr(pdev->bus) &&
> + link->bus == pdev->bus->number &&
> + link->slot == PCI_SLOT(pdev->devfn)) {
> + return link;
> + }
> + }
> +
> + /* link doesn't exist yet. Allocate one */
> + link = kzalloc(sizeof(struct vfio_device_ocxl_link), GFP_KERNEL);
> + if (!link)
> + return NULL;
> + link->domain = pci_domain_nr(pdev->bus);
> + link->bus = pdev->bus->number;
> + link->slot = PCI_SLOT(pdev->devfn);
> + list_add(&link->list, &links_list);
> + return link;
> +}
> +
> +static long irq_mapped(struct pci_dev *pdev,
> + int host_irq, int guest_irq, bool set)
> +{
> + struct irq_desc *desc;
> + struct kvm *kvm;
> + int ret = 0, virq;
> +
> + virq = irq_create_mapping(NULL, host_irq);
> + if (!virq) {
> + dev_err(&pdev->dev,
> + "irq_create_mapping failed for translation interrupt\n");
> + return -EINVAL;
> + }
> +
> + desc = irq_to_desc(virq);
> + if (!desc) {
> + dev_err(&pdev->dev,
> + "irq_to_desc failed (host_irq: %d, virq: %d)\n",
> + host_irq, virq);
> + return -EIO;
> + }
> +
> + kvm = vfio_dev_get_kvm(&pdev->dev);
> + if (!kvm)
> + return -ENODEV;
> +
> + mutex_lock(&kvm->lock);
> + if (xics_on_xive()) {
> + if (set)
> + ret = kvmppc_xive_set_mapped(kvm, guest_irq, desc);
> + else
> + ret = kvmppc_xive_clr_mapped(kvm, guest_irq, desc);
> + } else {
> + if (set)
> + kvmppc_xics_set_mapped(kvm, guest_irq, host_irq);
> + else
> + kvmppc_xics_clr_mapped(kvm, guest_irq, host_irq);
> + }
> + mutex_unlock(&kvm->lock);
> + kvm_put_kvm(kvm);
> +
> + return ret;
> +}
> +
> +static long config_adapter(struct pci_dev *pdev,
> + struct vfio_device_ocxl_op *ocxl_op,
> + struct vfio_device_ocxl_link *link)
> +{
> + int PE_mask, host_irq, guest_irq, count, tl_dvsec;
> + u16 base, enabled, supported;
> + u64 cmd;
> + int ret = 0;
> +
> + cmd = ocxl_op->data[2];
> + switch (cmd) {
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER_SETUP:
> + PE_mask = ocxl_op->data[3];
> + ret = pnv_ocxl_platform_setup(pdev, PE_mask,
> + &host_irq,
> + &link->platform_data);
> + if (!ret) {
> + guest_irq = ocxl_op->data[4];
> + ret = irq_mapped(pdev, host_irq, guest_irq, true);
> + if (!ret)
> + ocxl_op->data[0] = host_irq;
> + }
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER_RELEASE:
> + pnv_ocxl_platform_release(link->platform_data);
> +
> + host_irq = ocxl_op->data[3];
> + guest_irq = ocxl_op->data[4];
> + if (host_irq && guest_irq)
> + ret = irq_mapped(pdev, host_irq, guest_irq, false);
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER_GET_ACTAG:
> + ret = pnv_ocxl_get_actag(pdev, &base, &enabled,
> + &supported);
> + if (!ret) {
> + ocxl_op->data[0] = base;
> + ocxl_op->data[1] = enabled;
> + ocxl_op->data[2] = supported;
> + }
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER_GET_PASID:
> + ret = pnv_ocxl_get_pasid_count(pdev, &count);
> + if (!ret)
> + ocxl_op->data[0] = count;
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER_SET_TL:
> + tl_dvsec = ocxl_op->data[3];
> + ret = pnv_ocxl_set_TL(pdev, tl_dvsec);
> + break;
> +
> + default:
> + ret = -EINVAL;
> + }
> +
> + if (ret)
> + dev_err(&pdev->dev, "Failed to configure the adapter "
> + "(cmd: %#llx, ret: %d)\n",
> + cmd, ret);
> +
> + return ret;
> +}
> +
> +static long config_spa(struct pci_dev *pdev,
> + struct vfio_device_ocxl_op *ocxl_op,
> + struct vfio_device_ocxl_link *link)
> +{
> + int lpid, pid, tid, pasid;
> + int pe_handle, ret = 0;
> + u32 pidr, tidr, amr;
> + struct kvm *kvm;
> + u64 cmd;
> +
> + cmd = ocxl_op->data[2];
> + switch (cmd) {
> + case VFIO_DEVICE_OCXL_CONFIG_SPA_SET:
> + kvm = vfio_dev_get_kvm(&pdev->dev);
> + if (!kvm)
> + return -ENODEV;
> + lpid = kvm->arch.lpid;
> + kvm_put_kvm(kvm);
> +
> + pasid = ocxl_op->data[3];
> + pidr = ocxl_op->data[4];
> + tidr = ocxl_op->data[5];
> + amr = ocxl_op->data[6];
> +
> + ret = pnv_ocxl_set_pe(link->platform_data, lpid, pasid,
> + pidr, tidr, amr, &pe_handle);
> + if (!ret)
> + ocxl_op->data[0] = pe_handle;
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_SPA_UPDATE:
> + pasid = ocxl_op->data[3];
> + tid = ocxl_op->data[4];
> +
> + pnv_ocxl_update_pe(link->platform_data, pasid, tid);
> + break;
> +
> + case VFIO_DEVICE_OCXL_CONFIG_SPA_REMOVE:
> + pasid = ocxl_op->data[3];
> +
> + ret = pnv_ocxl_remove_pe(link->platform_data, pasid,
> + &pid, &tid, &pe_handle);
> + if (!ret) {
> + ocxl_op->data[0] = pid;
> + ocxl_op->data[1] = tid;
> + ocxl_op->data[2] = pe_handle;
> + }
> + break;
> +
> + default:
> + ret = -EINVAL;
> + }
> +
> + if (ret)
> + dev_err(&pdev->dev, "Failed to configure the SPA "
> + "(cmd: %#llx, ret: %d)\n",
> + cmd, ret);
> +
> + return ret;
> +}
> +
> +static void get_fault_state(struct vfio_device_ocxl_op *ocxl_op,
> + struct vfio_device_ocxl_link *link)
> +{
> + u64 dsisr, dar, pe_handle;
> + int pid;
> +
> + pnv_ocxl_get_fault_state(link->platform_data, &dsisr, &dar,
> + &pe_handle, &pid);
> +
> + ocxl_op->data[0] = dsisr;
> + ocxl_op->data[1] = dar;
> + ocxl_op->data[2] = pe_handle;
> + ocxl_op->data[3] = pid;
> +}
> +
> +static void handle_fault(struct vfio_device_ocxl_op *ocxl_op,
> + struct vfio_device_ocxl_link *link)
> +{
> + u64 tfc;
> +
> + tfc = ocxl_op->data[2];
> + pnv_ocxl_handle_fault(link->platform_data, tfc);
> +}
> +
> +long vfio_pci_ocxl_ioctl(struct pci_dev *pdev,
> + struct vfio_device_ocxl_op *ocxl_op)
> +{
> + struct vfio_device_ocxl_link *link;
> + int ret = 0;
> +
> + /* The functions of a device all share the same link */
> + mutex_lock(&links_list_lock);
> + link = find_link(pdev);
> +
> + switch (ocxl_op->op) {
> + case VFIO_DEVICE_OCXL_CONFIG_ADAPTER:
> + ret = config_adapter(pdev, ocxl_op, link);
> + break;
> + case VFIO_DEVICE_OCXL_CONFIG_SPA:
> + ret = config_spa(pdev, ocxl_op, link);
> + break;
> + case VFIO_DEVICE_OCXL_GET_FAULT_STATE:
> + get_fault_state(ocxl_op, link);
> + break;
> + case VFIO_DEVICE_OCXL_HANDLE_FAULT:
> + handle_fault(ocxl_op, link);
> + break;
> + default:
> + ret = -EINVAL;
> + }
> +
> + mutex_unlock(&links_list_lock);
> + return ret;
> +}
> +EXPORT_SYMBOL_GPL(vfio_pci_ocxl_ioctl);
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index 388597930b64..31d64ecac690 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -18,6 +18,7 @@
> #include <linux/fs.h>
> #include <linux/idr.h>
> #include <linux/iommu.h>
> +#include <linux/kvm_host.h>
> #include <linux/list.h>
> #include <linux/miscdevice.h>
> #include <linux/module.h>
> @@ -2051,6 +2052,30 @@ void vfio_group_set_kvm(struct vfio_group *group, struct kvm *kvm)
> }
> EXPORT_SYMBOL_GPL(vfio_group_set_kvm);
>
> +struct kvm *vfio_dev_get_kvm(struct device *dev)
> +{
> + struct iommu_group *iommu_group;
> + struct vfio_group *group;
> + struct kvm *kvm;
> +
> + iommu_group = iommu_group_get(dev);
> + if (!iommu_group)
> + return NULL;
> +
> + group = vfio_group_get_from_iommu(iommu_group);
> + if (!group) {
> + iommu_group_put(iommu_group);
> + return NULL;
> + }
> +
> + kvm_get_kvm(kvm = group->kvm);
> + iommu_group_put(iommu_group);
> + vfio_group_put(group);
> +
> + return kvm;
> +}
> +EXPORT_SYMBOL_GPL(vfio_dev_get_kvm);
> +
> static int vfio_register_group_notifier(struct vfio_group *group,
> unsigned long *events,
> struct notifier_block *nb)
> diff --git a/include/linux/vfio.h b/include/linux/vfio.h
> index e42a711a2800..22ee8d007353 100644
> --- a/include/linux/vfio.h
> +++ b/include/linux/vfio.h
> @@ -129,6 +129,7 @@ extern int vfio_unregister_notifier(struct device *dev,
>
> struct kvm;
> extern void vfio_group_set_kvm(struct vfio_group *group, struct kvm *kvm);
> +extern struct kvm *vfio_dev_get_kvm(struct device *dev);
>
> /*
> * Sub-module helpers
> @@ -195,4 +196,16 @@ extern int vfio_virqfd_enable(void *opaque,
> void *data, struct virqfd **pvirqfd, int fd);
> extern void vfio_virqfd_disable(struct virqfd **pvirqfd);
>
> +/* OpenCAPI */
> +#if IS_ENABLED(CONFIG_OCXL_BASE)
> +extern long vfio_pci_ocxl_ioctl(struct pci_dev *pdev,
> + struct vfio_device_ocxl_op *ocxl_op);
> +#else
> +static inline long vfio_pci_ocxl_ioctl(struct pci_dev *pdev,
> + struct vfio_device_ocxl_op *ocxl_op)
> +{
> + return -ENOTTY;
> +}
> +#endif /* CONFIG_OCXL_BASE */
> +
> #endif /* VFIO_H */
> diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
> index 8f10748dac79..4432593c2e65 100644
> --- a/include/uapi/linux/vfio.h
> +++ b/include/uapi/linux/vfio.h
> @@ -912,6 +912,28 @@ struct vfio_iommu_spapr_tce_remove {
> };
> #define VFIO_IOMMU_SPAPR_TCE_REMOVE _IO(VFIO_TYPE, VFIO_BASE + 20)
>
> +/**
> + * VFIO_DEVICE_OCXL_OP - _IOW(VFIO_TYPE, VFIO_BASE + 22, struct vfio_device_ocxl_op)
> + *
> + * Handles devices, which supports the OpenCAPI interface, using the
> + * ocxl pnv_* interface.
> + */
> +struct vfio_device_ocxl_op {
> + __u32 argsz;
> + __u32 flags;
> + __u32 op;
> + __u64 data[9];
> + /* data to be read and written
> + * data[0] = buid
> + * data[1] = config_addr
> + * data[2] = cmd or data[2] = p1, data[3] = p2, ...
> + * data[3] = p1, params[4] = p2, ...
> + *
> + * data[x] = outx ...
> + */
> +};
> +#define VFIO_DEVICE_OCXL_OP _IO(VFIO_TYPE, VFIO_BASE + 22)
> +
> /* ***************************************************************** */
>
> #endif /* _UAPIVFIO_H */
>
--
Alexey
More information about the Linuxppc-dev
mailing list