From patchwork Thu Nov 21 13:34:36 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Zhu, Lingshan" X-Patchwork-Id: 11256243 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id BD4846C1 for ; Thu, 21 Nov 2019 13:37:13 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 8C7D620715 for ; Thu, 21 Nov 2019 13:37:13 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727040AbfKUNhA (ORCPT ); Thu, 21 Nov 2019 08:37:00 -0500 Received: from mga11.intel.com ([192.55.52.93]:27188 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726947AbfKUNg7 (ORCPT ); Thu, 21 Nov 2019 08:36:59 -0500 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by fmsmga102.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 21 Nov 2019 05:36:59 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.69,226,1571727600"; d="scan'208";a="238178535" Received: from unknown (HELO localhost.localdomain.bj.intel.com) ([10.240.193.79]) by fmsmga002.fm.intel.com with ESMTP; 21 Nov 2019 05:36:55 -0800 From: Zhu Lingshan To: mst@redhat.com, jasowang@redhat.com, alex.williamson@redhat.com Cc: linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, kvm@vger.kernel.org, netdev@vger.kernel.org, dan.daly@intel.com, cunming.liang@intel.com, tiwei.bie@intel.com, jason.zeng@intel.com, zhiyuan.lv@intel.com, Zhu Lingshan , Xiao Wang Subject: [RFC V1 1/2] vhost: IFC VF initialization functions Date: Thu, 21 Nov 2019 21:34:36 +0800 Message-Id: <1574343277-8835-2-git-send-email-lingshan.zhu@intel.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1574343277-8835-1-git-send-email-lingshan.zhu@intel.com> References: <1574343277-8835-1-git-send-email-lingshan.zhu@intel.com> Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org This commit intends to add initialization functions for IFC VF, including: probe / remove / mdev operations and sysfs interfaces. Signed-off-by: Zhu Lingshan Signed-off-by: Xiao Wang Signed-off-by: Tiwei Bie --- drivers/vhost/Kconfig | 12 ++ drivers/vhost/Makefile | 2 + drivers/vhost/ifcvf/Makefile | 2 + drivers/vhost/ifcvf/ifcvf_base.c | 129 ++++++++++++++++ drivers/vhost/ifcvf/ifcvf_base.h | 123 +++++++++++++++ drivers/vhost/ifcvf/ifcvf_main.c | 315 +++++++++++++++++++++++++++++++++++++++ 6 files changed, 583 insertions(+) create mode 100644 drivers/vhost/ifcvf/Makefile create mode 100644 drivers/vhost/ifcvf/ifcvf_base.c create mode 100644 drivers/vhost/ifcvf/ifcvf_base.h create mode 100644 drivers/vhost/ifcvf/ifcvf_main.c diff --git a/drivers/vhost/Kconfig b/drivers/vhost/Kconfig index a4cf67a..7fa2a45 100644 --- a/drivers/vhost/Kconfig +++ b/drivers/vhost/Kconfig @@ -46,6 +46,18 @@ config VHOST_MDEV To compile this driver as a module, choose M here: the module will be called vhost_mdev. +config IFCVF + tristate "Intel IFC VF VDPA driver" + depends on MDEV_VIRTIO + select VHOST + default n + ---help--- + This kernel module can drive Intel IFC VF NIC to offload + vhost dataplane traffic to hardware. + + To compile this driver as a module, choose M here: the module will + be called ifcvf. + config VHOST tristate ---help--- diff --git a/drivers/vhost/Makefile b/drivers/vhost/Makefile index ad9c0f8..4cc484d 100644 --- a/drivers/vhost/Makefile +++ b/drivers/vhost/Makefile @@ -14,3 +14,5 @@ obj-$(CONFIG_VHOST_MDEV) += vhost_mdev.o vhost_mdev-y := mdev.o obj-$(CONFIG_VHOST) += vhost.o + +obj-$(CONFIG_IFCVF) += ifcvf/ diff --git a/drivers/vhost/ifcvf/Makefile b/drivers/vhost/ifcvf/Makefile new file mode 100644 index 0000000..aa12148 --- /dev/null +++ b/drivers/vhost/ifcvf/Makefile @@ -0,0 +1,2 @@ +obj-m += ifcvf.o +ifcvf-$(CONFIG_IFCVF)+= ifcvf_main.o ifcvf_base.o diff --git a/drivers/vhost/ifcvf/ifcvf_base.c b/drivers/vhost/ifcvf/ifcvf_base.c new file mode 100644 index 0000000..ec5985f --- /dev/null +++ b/drivers/vhost/ifcvf/ifcvf_base.c @@ -0,0 +1,129 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2019 Intel Corporation. + */ + +#include "ifcvf_base.h" + +struct ifcvf_adapter *vf_to_adapter(struct ifcvf_hw *hw) +{ + return container_of(hw, struct ifcvf_adapter, vf); +} + +static void *get_cap_addr(struct ifcvf_hw *hw, struct virtio_pci_cap *cap) +{ + struct ifcvf_adapter *ifcvf; + u32 length, offset; + u8 bar; + + length = le32_to_cpu(cap->length); + offset = le32_to_cpu(cap->offset); + bar = cap->bar; + + ifcvf = vf_to_adapter(hw); + if (bar >= IFCVF_PCI_MAX_RESOURCE) { + IFCVF_DBG(ifcvf->dev, + "Invalid bar number %u to get capabilities\n", bar); + return NULL; + } + + if (offset + length > hw->mem_resource[bar].len) { + IFCVF_DBG(ifcvf->dev, + "offset(%u) + len(%u) overflows bar%u to get capabilities\n", + offset, length, bar); + return NULL; + } + + return hw->mem_resource[bar].addr + offset; +} + +int ifcvf_read_config_range(struct pci_dev *dev, + uint32_t *val, int size, int where) +{ + int ret, i; + + for (i = 0; i < size; i += 4) { + ret = pci_read_config_dword(dev, where + i, val + i / 4); + if (ret < 0) + return ret; + } + + return 0; +} + +int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *dev) +{ + struct virtio_pci_cap cap; + u16 notify_off; + int ret; + u8 pos; + u32 i; + + ret = pci_read_config_byte(dev, PCI_CAPABILITY_LIST, &pos); + if (ret < 0) { + IFCVF_ERR(&dev->dev, "Failed to read PCI capability list\n"); + return -EIO; + } + + while (pos) { + ret = ifcvf_read_config_range(dev, (u32 *)&cap, + sizeof(cap), pos); + if (ret < 0) { + IFCVF_ERR(&dev->dev, + "Failed to get PCI capability at %x\n", pos); + break; + } + + if (cap.cap_vndr != PCI_CAP_ID_VNDR) + goto next; + + switch (cap.cfg_type) { + case VIRTIO_PCI_CAP_COMMON_CFG: + hw->common_cfg = get_cap_addr(hw, &cap); + IFCVF_DBG(&dev->dev, "hw->common_cfg = %p\n", + hw->common_cfg); + break; + case VIRTIO_PCI_CAP_NOTIFY_CFG: + pci_read_config_dword(dev, pos + sizeof(cap), + &hw->notify_off_multiplier); + hw->notify_bar = cap.bar; + hw->notify_base = get_cap_addr(hw, &cap); + IFCVF_DBG(&dev->dev, "hw->notify_base = %p\n", + hw->notify_base); + break; + case VIRTIO_PCI_CAP_ISR_CFG: + hw->isr = get_cap_addr(hw, &cap); + IFCVF_DBG(&dev->dev, "hw->isr = %p\n", hw->isr); + break; + case VIRTIO_PCI_CAP_DEVICE_CFG: + hw->net_cfg = get_cap_addr(hw, &cap); + IFCVF_DBG(&dev->dev, "hw->net_cfg = %p\n", hw->net_cfg); + break; + } + +next: + pos = cap.cap_next; + } + + if (hw->common_cfg == NULL || hw->notify_base == NULL || + hw->isr == NULL || hw->net_cfg == NULL) { + IFCVF_ERR(&dev->dev, "Incomplete PCI capabilities\n"); + return -EIO; + } + + for (i = 0; i < IFCVF_MAX_QUEUE_PAIRS * 2; i++) { + iowrite16(i, &hw->common_cfg->queue_select); + notify_off = ioread16(&hw->common_cfg->queue_notify_off); + hw->vring->notify_addr[i] = (u16 __iomem *)((u8 *)hw->notify_base + + notify_off * hw->notify_off_multiplier); + } + + hw->lm_cfg = hw->mem_resource[IFCVF_LM_BAR].addr; + + IFCVF_DBG(&dev->dev, + "PCI capability mapping: common cfg: %p, notify base: %p\n, isr cfg: %p, device cfg: %p, multiplier: %u\n", + hw->common_cfg, hw->notify_base, hw->isr, + hw->net_cfg, hw->notify_off_multiplier); + + return 0; +} diff --git a/drivers/vhost/ifcvf/ifcvf_base.h b/drivers/vhost/ifcvf/ifcvf_base.h new file mode 100644 index 0000000..a1e33ef --- /dev/null +++ b/drivers/vhost/ifcvf/ifcvf_base.h @@ -0,0 +1,123 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Copyright (C) 2019 Intel Corporation. + */ + +#ifndef _IFCVF_H_ +#define _IFCVF_H_ + +#include +#include +#include +#include +#include +#include +#include + +#define IFCVF_VENDOR_ID 0x1AF4 +#define IFCVF_DEVICE_ID 0x1041 +#define IFCVF_SUBSYS_VENDOR_ID 0x8086 +#define IFCVF_SUBSYS_DEVICE_ID 0x001A + +#define IFCVF_MDEV_LIMIT 1 + +#define IFCVF_SUPPORTED_FEATURES \ + ((1ULL << VIRTIO_NET_F_MAC) | \ + (1ULL << VIRTIO_F_ANY_LAYOUT) | \ + (1ULL << VIRTIO_F_VERSION_1) | \ + (1ULL << VIRTIO_F_ORDER_PLATFORM) | \ + (1ULL << VIRTIO_F_IOMMU_PLATFORM) | \ + (1ULL << VIRTIO_NET_F_MRG_RXBUF)) + +/* Only one queue pair for now. */ +#define IFCVF_MAX_QUEUE_PAIRS 1 + +#define IFCVF_QUEUE_ALIGNMENT PAGE_SIZE +#define IFCVF_QUEUE_MAX 32768 +#define IFCVF_MSI_CONFIG_OFF 0 +#define IFCVF_MSI_QUEUE_OFF 1 +#define IFCVF_PCI_MAX_RESOURCE 6 + +#define IFCVF_LM_CFG_SIZE 0x40 +#define IFCVF_LM_RING_STATE_OFFSET 0x20 +#define IFCVF_LM_BAR 4 + +#define IFCVF_ERR(dev, fmt, ...) dev_err(dev, fmt, ##__VA_ARGS__) +#define IFCVF_DBG(dev, fmt, ...) dev_dbg(dev, fmt, ##__VA_ARGS__) +#define IFCVF_INFO(dev, fmt, ...) dev_info(dev, fmt, ##__VA_ARGS__) + +#define ifcvf_private_to_vf(adapter) \ + (&((struct ifcvf_adapter *)adapter)->vf) + +#define IFCVF_MAX_INTR (IFCVF_MAX_QUEUE_PAIRS * 2 + 1) + +struct ifcvf_pci_mem_resource { + u64 phys_addr; + u64 len; + /* Virtual address, NULL when not mapped. */ + u8 *addr; +}; + +struct vring_info { + u64 desc; + u64 avail; + u64 used; + u16 size; + u16 last_avail_idx; + u16 last_used_idx; + bool ready; + char msix_name[256]; + u16 __iomem *notify_addr[IFCVF_MAX_QUEUE_PAIRS * 2]; + struct virtio_mdev_callback cb; +}; + +struct ifcvf_hw { + u8 __iomem *isr; + /* Live migration */ + u8 __iomem *lm_cfg; + u16 nr_vring; + /* Notification bar number */ + u8 notify_bar; + /* Notificaiton bar address */ + u16 __iomem *notify_base; + u32 notify_off_multiplier; + u64 req_features; + struct virtio_pci_common_cfg __iomem *common_cfg; + struct virtio_net_config __iomem *net_cfg; + struct vring_info vring[IFCVF_MAX_QUEUE_PAIRS * 2]; + struct ifcvf_pci_mem_resource mem_resource[IFCVF_PCI_MAX_RESOURCE]; +}; + +struct ifcvf_adapter { + struct device *dev; + struct mutex mdev_lock; + int mdev_count; + int vectors; + struct ifcvf_hw vf; +}; + +struct ifcvf_vring_lm_cfg { + u32 idx_addr[2]; + u8 reserved[IFCVF_LM_CFG_SIZE - 8]; +}; + +struct ifcvf_lm_cfg { + u8 reserved[IFCVF_LM_RING_STATE_OFFSET]; + struct ifcvf_vring_lm_cfg vring_lm_cfg[IFCVF_MAX_QUEUE_PAIRS]; +}; + +int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *dev); +int ifcvf_start_hw(struct ifcvf_hw *hw); +void ifcvf_stop_hw(struct ifcvf_hw *hw); +void ifcvf_notify_queue(struct ifcvf_hw *hw, u16 qid); +void ifcvf_read_net_config(struct ifcvf_hw *hw, u64 offset, + void *dst, int length); +void ifcvf_write_net_config(struct ifcvf_hw *hw, u64 offset, + const void *src, int length); +u8 ifcvf_get_status(struct ifcvf_hw *hw); +void ifcvf_set_status(struct ifcvf_hw *hw, u8 status); +void io_write64_twopart(u64 val, u32 *lo, u32 *hi); +void ifcvf_reset(struct ifcvf_hw *hw); +u64 ifcvf_get_features(struct ifcvf_hw *hw); +struct ifcvf_adapter *vf_to_adapter(struct ifcvf_hw *hw); +#endif /* _IFCVF_H_ */ diff --git a/drivers/vhost/ifcvf/ifcvf_main.c b/drivers/vhost/ifcvf/ifcvf_main.c new file mode 100644 index 0000000..4f602a3 --- /dev/null +++ b/drivers/vhost/ifcvf/ifcvf_main.c @@ -0,0 +1,315 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Intel IFC VF NIC driver for vhost dataplane offloading + * + * Copyright (C) 2019 Intel Corporation. + * + * Author: Zhu Lingshan + * + */ + +#include +#include +#include +#include +#include +#include "ifcvf_base.h" + +#define VERSION_STRING "0.1" +#define DRIVER_AUTHOR "Intel Corporation" +#define IFCVF_DRIVER_NAME "ifcvf" + +static irqreturn_t ifcvf_intr_handler(int irq, void *arg) +{ + struct vring_info *vring = arg; + + if (vring->cb.callback) + return vring->cb.callback(vring->cb.private); + + return IRQ_HANDLED; +} + +static const struct mdev_virtio_ops ifc_mdev_ops = { + NULL, +}; + +static int ifcvf_init_msix(struct ifcvf_adapter *adapter) +{ + struct pci_dev *pdev = to_pci_dev(adapter->dev); + struct ifcvf_hw *vf = &adapter->vf; + int vector, i, ret, irq; + + ret = pci_alloc_irq_vectors(pdev, IFCVF_MAX_INTR, + IFCVF_MAX_INTR, PCI_IRQ_MSIX); + if (ret < 0) { + IFCVF_ERR(adapter->dev, "Failed to alloc irq vectors\n"); + return ret; + } + + for (i = 0; i < IFCVF_MAX_QUEUE_PAIRS * 2; i++) { + vector = i + IFCVF_MSI_QUEUE_OFF; + irq = pci_irq_vector(pdev, vector); + ret = request_irq(irq, ifcvf_intr_handler, 0, + pci_name(pdev), &vf->vring[i]); + if (ret) { + IFCVF_ERR(adapter->dev, + "Failed to request irq for vq %d\n", i); + return ret; + } + } + + return 0; +} + +static void ifcvf_destroy_adapter(struct ifcvf_adapter *adapter) +{ + struct ifcvf_hw *vf = ifcvf_private_to_vf(adapter); + struct pci_dev *pdev = to_pci_dev(adapter->dev); + int i, vector, irq; + + for (i = 0; i < IFCVF_MAX_QUEUE_PAIRS * 2; i++) { + vector = i + IFCVF_MSI_QUEUE_OFF; + irq = pci_irq_vector(pdev, vector); + free_irq(irq, &vf->vring[i]); + } +} + +static ssize_t name_show(struct kobject *kobj, struct device *dev, char *buf) +{ + const char *name = "IFC VF virtio/vhost accelerator (virtio ring compatible)"; + + return sprintf(buf, "%s\n", name); +} +MDEV_TYPE_ATTR_RO(name); + +static ssize_t device_api_show(struct kobject *kobj, struct device *dev, + char *buf) +{ + return sprintf(buf, "%s\n", "virtio_mdev"); +} +MDEV_TYPE_ATTR_RO(device_api); + +static ssize_t available_instances_show(struct kobject *kobj, + struct device *dev, char *buf) +{ + struct pci_dev *pdev; + struct ifcvf_adapter *adapter; + + pdev = to_pci_dev(dev); + adapter = pci_get_drvdata(pdev); + + return sprintf(buf, "%d\n", adapter->mdev_count); +} +MDEV_TYPE_ATTR_RO(available_instances); + +static ssize_t type_show(struct kobject *kobj, + struct device *dev, char *buf) +{ + return sprintf(buf, "%s\n", "net"); +} +MDEV_TYPE_ATTR_RO(type); + +static struct attribute *mdev_types_attrs[] = { + &mdev_type_attr_name.attr, + &mdev_type_attr_device_api.attr, + &mdev_type_attr_available_instances.attr, + &mdev_type_attr_type.attr, + NULL, +}; + +static struct attribute_group mdev_type_group_virtio = { + .name = "virtio_mdev", + .attrs = mdev_types_attrs, +}; + +static struct attribute_group mdev_type_group_vhost = { + .name = "vhost_mdev", + .attrs = mdev_types_attrs, +}; + +static struct attribute_group *mdev_type_groups[] = { + &mdev_type_group_virtio, + &mdev_type_group_vhost, + NULL, +}; + +const struct attribute_group *mdev_dev_groups[] = { + NULL, +}; + +static int ifcvf_mdev_create(struct kobject *kobj, struct mdev_device *mdev) +{ + struct device *dev = mdev_parent_dev(mdev); + struct pci_dev *pdev = to_pci_dev(dev); + struct ifcvf_adapter *adapter = pci_get_drvdata(pdev); + int ret = 0; + + mutex_lock(&adapter->mdev_lock); + + if (adapter->mdev_count < IFCVF_MDEV_LIMIT) { + ret = -EINVAL; + goto out; + } + + mdev_virtio_set_ops(mdev, &ifc_mdev_ops); + if (!strcmp(kobj->name, "ifcvf-virtio_mdev")) + mdev_virtio_set_class_id(mdev, MDEV_VIRTIO_CLASS_ID_VIRTIO); + + if (!strcmp(kobj->name, "ifcvf-vhost_mdev")) + mdev_virtio_set_class_id(mdev, MDEV_VIRTIO_CLASS_ID_VHOST); + + mdev_set_drvdata(mdev, adapter); + mdev_set_iommu_device(mdev_dev(mdev), dev); + adapter->mdev_count--; + +out: + mutex_unlock(&adapter->mdev_lock); + return ret; +} + +static int ifcvf_mdev_remove(struct mdev_device *mdev) +{ + struct device *dev = mdev_parent_dev(mdev); + struct pci_dev *pdev = to_pci_dev(dev); + struct ifcvf_adapter *adapter = pci_get_drvdata(pdev); + + mutex_lock(&adapter->mdev_lock); + adapter->mdev_count++; + mutex_unlock(&adapter->mdev_lock); + + return 0; +} + +static const struct mdev_parent_ops ifcvf_mdev_fops = { + .owner = THIS_MODULE, + .supported_type_groups = mdev_type_groups, + .mdev_attr_groups = mdev_dev_groups, + .create = ifcvf_mdev_create, + .remove = ifcvf_mdev_remove, +}; + +static int ifcvf_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + struct device *dev = &pdev->dev; + struct ifcvf_adapter *adapter; + struct ifcvf_hw *vf; + int ret, i; + + adapter = kzalloc(sizeof(struct ifcvf_adapter), GFP_KERNEL); + if (adapter == NULL) { + ret = -ENOMEM; + goto fail; + } + + mutex_init(&adapter->mdev_lock); + adapter->mdev_count = IFCVF_MDEV_LIMIT; + adapter->dev = dev; + pci_set_drvdata(pdev, adapter); + ret = pci_enable_device(pdev); + if (ret) { + IFCVF_ERR(adapter->dev, "Failed to enable device\n"); + goto free_adapter; + } + + ret = pci_request_regions(pdev, IFCVF_DRIVER_NAME); + if (ret) { + IFCVF_ERR(adapter->dev, "Failed to request MMIO region\n"); + goto disable_device; + } + + pci_set_master(pdev); + ret = ifcvf_init_msix(adapter); + if (ret) { + IFCVF_ERR(adapter->dev, "Failed to initialize MSI-X\n"); + goto free_msix; + } + + vf = &adapter->vf; + for (i = 0; i < IFCVF_PCI_MAX_RESOURCE; i++) { + vf->mem_resource[i].phys_addr = pci_resource_start(pdev, i); + vf->mem_resource[i].len = pci_resource_len(pdev, i); + if (!vf->mem_resource[i].len) { + vf->mem_resource[i].addr = NULL; + continue; + } + + vf->mem_resource[i].addr = pci_iomap_range(pdev, i, 0, + vf->mem_resource[i].len); + if (!vf->mem_resource[i].addr) { + IFCVF_ERR(adapter->dev, "Failed to map IO resource %d\n", + i); + ret = -EINVAL; + goto free_msix; + } + } + + if (ifcvf_init_hw(vf, pdev) < 0) { + ret = -EINVAL; + goto destroy_adapter; + } + + ret = mdev_virtio_register_device(dev, &ifcvf_mdev_fops); + if (ret) { + IFCVF_ERR(adapter->dev, "Failed to register mdev device\n"); + goto destroy_adapter; + } + + return 0; + +destroy_adapter: + ifcvf_destroy_adapter(adapter); +free_msix: + pci_free_irq_vectors(pdev); + pci_release_regions(pdev); +disable_device: + pci_disable_device(pdev); +free_adapter: + kfree(adapter); +fail: + return ret; +} + +static void ifcvf_remove(struct pci_dev *pdev) +{ + struct ifcvf_adapter *adapter = pci_get_drvdata(pdev); + struct device *dev = &pdev->dev; + struct ifcvf_hw *vf; + int i; + + mdev_virtio_unregister_device(dev); + vf = &adapter->vf; + for (i = 0; i < IFCVF_PCI_MAX_RESOURCE; i++) { + if (vf->mem_resource[i].addr) { + pci_iounmap(pdev, vf->mem_resource[i].addr); + vf->mem_resource[i].addr = NULL; + } + } + + ifcvf_destroy_adapter(adapter); + pci_free_irq_vectors(pdev); + pci_release_regions(pdev); + pci_disable_device(pdev); + kfree(adapter); +} + +static struct pci_device_id ifcvf_pci_ids[] = { + { PCI_DEVICE_SUB(IFCVF_VENDOR_ID, + IFCVF_DEVICE_ID, + IFCVF_SUBSYS_VENDOR_ID, + IFCVF_SUBSYS_DEVICE_ID) }, + { 0 }, +}; +MODULE_DEVICE_TABLE(pci, ifcvf_pci_ids); + +static struct pci_driver ifcvf_driver = { + .name = IFCVF_DRIVER_NAME, + .id_table = ifcvf_pci_ids, + .probe = ifcvf_probe, + .remove = ifcvf_remove, +}; + +module_pci_driver(ifcvf_driver); + +MODULE_LICENSE("GPL v2"); +MODULE_VERSION(VERSION_STRING); +MODULE_AUTHOR(DRIVER_AUTHOR); From patchwork Thu Nov 21 13:34:37 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Zhu, Lingshan" X-Patchwork-Id: 11256241 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id AD9C7138C for ; Thu, 21 Nov 2019 13:37:05 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 7B3CC2071F for ; Thu, 21 Nov 2019 13:37:05 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727121AbfKUNhE (ORCPT ); Thu, 21 Nov 2019 08:37:04 -0500 Received: from mga11.intel.com ([192.55.52.93]:27188 "EHLO mga11.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1727097AbfKUNhD (ORCPT ); Thu, 21 Nov 2019 08:37:03 -0500 X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by fmsmga102.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 21 Nov 2019 05:37:02 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.69,226,1571727600"; d="scan'208";a="238178558" Received: from unknown (HELO localhost.localdomain.bj.intel.com) ([10.240.193.79]) by fmsmga002.fm.intel.com with ESMTP; 21 Nov 2019 05:37:00 -0800 From: Zhu Lingshan To: mst@redhat.com, jasowang@redhat.com, alex.williamson@redhat.com Cc: linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, kvm@vger.kernel.org, netdev@vger.kernel.org, dan.daly@intel.com, cunming.liang@intel.com, tiwei.bie@intel.com, jason.zeng@intel.com, zhiyuan.lv@intel.com, Zhu Lingshan , Xiao Wang Subject: [RFC V1 2/2] vhost: Support for virtio_mdev and vhost_mdev Date: Thu, 21 Nov 2019 21:34:37 +0800 Message-Id: <1574343277-8835-3-git-send-email-lingshan.zhu@intel.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1574343277-8835-1-git-send-email-lingshan.zhu@intel.com> References: <1574343277-8835-1-git-send-email-lingshan.zhu@intel.com> Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org This commit adds functions to support virtio_mdev and vhost_mdev. Implemented interfaces in struct mdev_virtio_ops and supportive functions including status operations, features operations, hardware enable / disable and etc. Signed-off-by: Zhu Lingshan Signed-off-by: Xiao Wang Signed-off-by: Tiwei Bie --- drivers/vhost/ifcvf/ifcvf_base.c | 194 +++++++++++++++++++++++++++++++ drivers/vhost/ifcvf/ifcvf_main.c | 242 ++++++++++++++++++++++++++++++++++++++- 2 files changed, 435 insertions(+), 1 deletion(-) diff --git a/drivers/vhost/ifcvf/ifcvf_base.c b/drivers/vhost/ifcvf/ifcvf_base.c index ec5985f..d56a069 100644 --- a/drivers/vhost/ifcvf/ifcvf_base.c +++ b/drivers/vhost/ifcvf/ifcvf_base.c @@ -127,3 +127,197 @@ int ifcvf_init_hw(struct ifcvf_hw *hw, struct pci_dev *dev) return 0; } + +u8 ifcvf_get_status(struct ifcvf_hw *hw) +{ + return ioread8(&hw->common_cfg->device_status); +} + +void ifcvf_set_status(struct ifcvf_hw *hw, u8 status) +{ + iowrite8(status, &hw->common_cfg->device_status); +} + +void ifcvf_reset(struct ifcvf_hw *hw) +{ + ifcvf_set_status(hw, 0); + ifcvf_get_status(hw); +} + +static void ifcvf_add_status(struct ifcvf_hw *hw, u8 status) +{ + if (status != 0) + status |= ifcvf_get_status(hw); + + ifcvf_set_status(hw, status); + ifcvf_get_status(hw); +} + +u64 ifcvf_get_features(struct ifcvf_hw *hw) +{ + struct virtio_pci_common_cfg *cfg = hw->common_cfg; + u32 features_lo, features_hi; + + iowrite32(0, &cfg->device_feature_select); + features_lo = ioread32(&cfg->device_feature); + + iowrite32(1, &cfg->device_feature_select); + features_hi = ioread32(&cfg->device_feature); + + return ((u64)features_hi << 32) | features_lo; +} + +void ifcvf_read_net_config(struct ifcvf_hw *hw, u64 offset, + void *dst, int length) +{ + u8 old_gen, new_gen, *p; + int i; + + WARN_ON(offset + length > sizeof(struct virtio_net_config)); + do { + old_gen = ioread8(&hw->common_cfg->config_generation); + p = dst; + for (i = 0; i < length; i++) + *p++ = ioread8((u8 *)hw->net_cfg + offset + i); + + new_gen = ioread8(&hw->common_cfg->config_generation); + } while (old_gen != new_gen); +} + +void ifcvf_write_net_config(struct ifcvf_hw *hw, u64 offset, + const void *src, int length) +{ + const u8 *p; + int i; + + p = src; + WARN_ON(offset + length > sizeof(struct virtio_net_config)); + for (i = 0; i < length; i++) + iowrite8(*p++, (u8 *)hw->net_cfg + offset + i); +} + +static void ifcvf_set_features(struct ifcvf_hw *hw, u64 features) +{ + struct virtio_pci_common_cfg *cfg = hw->common_cfg; + + iowrite32(0, &cfg->guest_feature_select); + iowrite32((u32)features, &cfg->guest_feature); + + iowrite32(1, &cfg->guest_feature_select); + iowrite32(features >> 32, &cfg->guest_feature); +} + +static int ifcvf_config_features(struct ifcvf_hw *hw) +{ + struct ifcvf_adapter *ifcvf; + + ifcvf = vf_to_adapter(hw); + ifcvf_set_features(hw, hw->req_features); + ifcvf_add_status(hw, VIRTIO_CONFIG_S_FEATURES_OK); + + if (!(ifcvf_get_status(hw) & VIRTIO_CONFIG_S_FEATURES_OK)) { + IFCVF_ERR(ifcvf->dev, "Failed to set FEATURES_OK status\n"); + return -EIO; + } + + return 0; +} + +void io_write64_twopart(u64 val, u32 *lo, u32 *hi) +{ + iowrite32(val & ((1ULL << 32) - 1), lo); + iowrite32(val >> 32, hi); +} + +static int ifcvf_hw_enable(struct ifcvf_hw *hw) +{ + struct ifcvf_lm_cfg __iomem *ifcvf_lm; + struct virtio_pci_common_cfg *cfg; + struct ifcvf_adapter *ifcvf; + u32 __iomem *idx_addr; + u32 i, val; + + ifcvf_lm = (struct ifcvf_lm_cfg __iomem *)hw->lm_cfg; + ifcvf = vf_to_adapter(hw); + cfg = hw->common_cfg; + iowrite16(IFCVF_MSI_CONFIG_OFF, &cfg->msix_config); + + if (ioread16(&cfg->msix_config) == VIRTIO_MSI_NO_VECTOR) { + IFCVF_ERR(ifcvf->dev, "No msix vector for device config\n"); + return -EINVAL; + } + + for (i = 0; i < hw->nr_vring; i++) { + if (!hw->vring[i].ready) + break; + + iowrite16(i, &cfg->queue_select); + io_write64_twopart(hw->vring[i].desc, &cfg->queue_desc_lo, + &cfg->queue_desc_hi); + io_write64_twopart(hw->vring[i].avail, &cfg->queue_avail_lo, + &cfg->queue_avail_hi); + io_write64_twopart(hw->vring[i].used, &cfg->queue_used_lo, + &cfg->queue_used_hi); + iowrite16(hw->vring[i].size, &cfg->queue_size); + idx_addr = &ifcvf_lm->vring_lm_cfg[i/IFCVF_MAX_QUEUE_PAIRS * 2].idx_addr[i%IFCVF_MAX_QUEUE_PAIRS * 2]; + val = hw->vring[i].last_avail_idx | + ((u32)hw->vring[i].last_used_idx << 16); + iowrite32(val, idx_addr); + iowrite16(i + IFCVF_MSI_QUEUE_OFF, &cfg->queue_msix_vector); + + if (ioread16(&cfg->queue_msix_vector) == + VIRTIO_MSI_NO_VECTOR) { + IFCVF_ERR(ifcvf->dev, + "No msix vector for queue %u\n", i); + return -EINVAL; + } + + iowrite16(1, &cfg->queue_enable); + } + + return 0; +} + +static void ifcvf_hw_disable(struct ifcvf_hw *hw) +{ + struct virtio_pci_common_cfg *cfg; + u32 i; + + cfg = hw->common_cfg; + iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->msix_config); + + for (i = 0; i < hw->nr_vring; i++) { + iowrite16(i, &cfg->queue_select); + iowrite16(VIRTIO_MSI_NO_VECTOR, &cfg->queue_msix_vector); + } + + ioread16(&cfg->queue_msix_vector); +} + +int ifcvf_start_hw(struct ifcvf_hw *hw) +{ + ifcvf_reset(hw); + ifcvf_add_status(hw, VIRTIO_CONFIG_S_ACKNOWLEDGE); + ifcvf_add_status(hw, VIRTIO_CONFIG_S_DRIVER); + + if (ifcvf_config_features(hw) < 0) + return -EINVAL; + + if (ifcvf_hw_enable(hw) < 0) + return -EINVAL; + + ifcvf_add_status(hw, VIRTIO_CONFIG_S_DRIVER_OK); + + return 0; +} + +void ifcvf_stop_hw(struct ifcvf_hw *hw) +{ + ifcvf_hw_disable(hw); + ifcvf_reset(hw); +} + +void ifcvf_notify_queue(struct ifcvf_hw *hw, u16 qid) +{ + iowrite16(qid, hw->vring->notify_addr[qid]); +} diff --git a/drivers/vhost/ifcvf/ifcvf_main.c b/drivers/vhost/ifcvf/ifcvf_main.c index 4f602a3..bce9609 100644 --- a/drivers/vhost/ifcvf/ifcvf_main.c +++ b/drivers/vhost/ifcvf/ifcvf_main.c @@ -29,8 +29,248 @@ static irqreturn_t ifcvf_intr_handler(int irq, void *arg) return IRQ_HANDLED; } +static int ifcvf_start_datapath(void *private) +{ + struct ifcvf_hw *vf = ifcvf_private_to_vf(private); + struct ifcvf_adapter *ifcvf; + u8 status; + int ret; + + ifcvf = vf_to_adapter(vf); + vf->nr_vring = IFCVF_MAX_QUEUE_PAIRS * 2; + ret = ifcvf_start_hw(vf); + if (ret < 0) { + status = ifcvf_get_status(vf); + status |= VIRTIO_CONFIG_S_FAILED; + ifcvf_set_status(vf, status); + } + + return ret; +} + +static int ifcvf_stop_datapath(void *private) +{ + struct ifcvf_hw *vf = ifcvf_private_to_vf(private); + int i; + + for (i = 0; i < IFCVF_MAX_QUEUE_PAIRS * 2; i++) + vf->vring[i].cb.callback = NULL; + + ifcvf_stop_hw(vf); + + return 0; +} + +static void ifcvf_reset_vring(struct ifcvf_adapter *adapter) +{ + struct ifcvf_hw *vf = ifcvf_private_to_vf(adapter); + int i; + + for (i = 0; i < IFCVF_MAX_QUEUE_PAIRS * 2; i++) { + vf->vring[i].last_used_idx = 0; + vf->vring[i].last_avail_idx = 0; + vf->vring[i].desc = 0; + vf->vring[i].avail = 0; + vf->vring[i].used = 0; + vf->vring[i].ready = 0; + vf->vring->cb.callback = NULL; + vf->vring->cb.private = NULL; + } + + ifcvf_reset(vf); +} + +static struct ifcvf_hw *mdev_to_vf(struct mdev_device *mdev) +{ + struct ifcvf_adapter *adapter = mdev_get_drvdata(mdev); + struct ifcvf_hw *vf = ifcvf_private_to_vf(adapter); + + return vf; +} + +static u64 ifcvf_mdev_get_features(struct mdev_device *mdev) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + u64 features; + + features = ifcvf_get_features(vf) & IFCVF_SUPPORTED_FEATURES; + + return features; +} + +static int ifcvf_mdev_set_features(struct mdev_device *mdev, u64 features) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->req_features = features; + + return 0; +} + +static u8 ifcvf_mdev_get_status(struct mdev_device *mdev) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + return ifcvf_get_status(vf); +} + +static void ifcvf_mdev_set_status(struct mdev_device *mdev, u8 status) +{ + struct ifcvf_adapter *adapter = mdev_get_drvdata(mdev); + struct ifcvf_hw *vf = ifcvf_private_to_vf(adapter); + + if (status == 0) { + ifcvf_stop_datapath(adapter); + ifcvf_reset_vring(adapter); + return; + } + + if (status & VIRTIO_CONFIG_S_DRIVER_OK) { + if (ifcvf_start_datapath(adapter) < 0) + IFCVF_ERR(adapter->dev, "Failed to set mdev status %u\n", + status); + } + + ifcvf_set_status(vf, status); +} + +static u16 ifcvf_mdev_get_vq_num_max(struct mdev_device *mdev) +{ + return IFCVF_QUEUE_MAX; +} + +static u64 ifcvf_mdev_get_vq_state(struct mdev_device *mdev, u16 qid) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + u16 last_avail_idx; + u16 __iomem *idx_addr; + + idx_addr = (u16 __iomem *)(vf->lm_cfg + IFCVF_LM_RING_STATE_OFFSET + + (qid / 2) * IFCVF_LM_CFG_SIZE + (qid % 2) * 4); + last_avail_idx = ioread16(idx_addr); + + return last_avail_idx; +} + +static int ifcvf_mdev_set_vq_state(struct mdev_device *mdev, u16 qid, u64 num) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->vring[qid].last_avail_idx = num; + + return 0; +} + +static void ifcvf_mdev_set_vq_cb(struct mdev_device *mdev, u16 qid, + struct virtio_mdev_callback *cb) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->vring[qid].cb = *cb; +} + +static void ifcvf_mdev_set_vq_ready(struct mdev_device *mdev, + u16 qid, bool ready) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->vring[qid].ready = ready; +} + +static bool ifcvf_mdev_get_vq_ready(struct mdev_device *mdev, u16 qid) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + return vf->vring[qid].ready; +} + +static void ifcvf_mdev_set_vq_num(struct mdev_device *mdev, u16 qid, u32 num) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->vring[qid].size = num; +} + +static int ifcvf_mdev_set_vq_address(struct mdev_device *mdev, u16 qid, + u64 desc_area, u64 driver_area, + u64 device_area) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + vf->vring[qid].desc = desc_area; + vf->vring[qid].avail = driver_area; + vf->vring[qid].used = device_area; + + return 0; +} + +static void ifcvf_mdev_kick_vq(struct mdev_device *mdev, u16 qid) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + ifcvf_notify_queue(vf, qid); +} + +static u32 ifcvf_mdev_get_generation(struct mdev_device *mdev) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + return ioread8(&vf->common_cfg->config_generation); +} + +static u32 ifcvf_mdev_get_device_id(struct mdev_device *mdev) +{ + return VIRTIO_ID_NET; +} + +static u32 ifcvf_mdev_get_vendor_id(struct mdev_device *mdev) +{ + return IFCVF_SUBSYS_VENDOR_ID; +} + +static u16 ifcvf_mdev_get_vq_align(struct mdev_device *mdev) +{ + return IFCVF_QUEUE_ALIGNMENT; +} + +static void ifcvf_mdev_get_config(struct mdev_device *mdev, unsigned int offset, + void *buf, unsigned int len) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + WARN_ON(offset + len > sizeof(struct virtio_net_config)); + ifcvf_read_net_config(vf, offset, buf, len); +} + +static void ifcvf_mdev_set_config(struct mdev_device *mdev, unsigned int offset, + const void *buf, unsigned int len) +{ + struct ifcvf_hw *vf = mdev_to_vf(mdev); + + WARN_ON(offset + len > sizeof(struct virtio_net_config)); + ifcvf_write_net_config(vf, offset, buf, len); +} + static const struct mdev_virtio_ops ifc_mdev_ops = { - NULL, + .get_features = ifcvf_mdev_get_features, + .set_features = ifcvf_mdev_set_features, + .get_status = ifcvf_mdev_get_status, + .set_status = ifcvf_mdev_set_status, + .get_vq_num_max = ifcvf_mdev_get_vq_num_max, + .get_vq_state = ifcvf_mdev_get_vq_state, + .set_vq_state = ifcvf_mdev_set_vq_state, + .set_vq_cb = ifcvf_mdev_set_vq_cb, + .set_vq_ready = ifcvf_mdev_set_vq_ready, + .get_vq_ready = ifcvf_mdev_get_vq_ready, + .set_vq_num = ifcvf_mdev_set_vq_num, + .set_vq_address = ifcvf_mdev_set_vq_address, + .kick_vq = ifcvf_mdev_kick_vq, + .get_generation = ifcvf_mdev_get_generation, + .get_device_id = ifcvf_mdev_get_device_id, + .get_vendor_id = ifcvf_mdev_get_vendor_id, + .get_vq_align = ifcvf_mdev_get_vq_align, + .get_config = ifcvf_mdev_get_config, + .set_config = ifcvf_mdev_set_config, }; static int ifcvf_init_msix(struct ifcvf_adapter *adapter)