Allow a VIOMMU object to allocate VIRQ events. Each VIOMMU is allowed to have multiple VIRQ events but they must not have a duplicated type.
Signed-off-by: Nicolin Chen nicolinc@nvidia.com --- drivers/iommu/iommufd/event.c | 136 ++++++++++++++++++++++++ drivers/iommu/iommufd/iommufd_private.h | 54 ++++++++++ drivers/iommu/iommufd/main.c | 5 + drivers/iommu/iommufd/viommu.c | 2 + include/uapi/linux/iommufd.h | 32 ++++++ 5 files changed, 229 insertions(+)
diff --git a/drivers/iommu/iommufd/event.c b/drivers/iommu/iommufd/event.c index 8fea142e1ac2..f10827ce9cbd 100644 --- a/drivers/iommu/iommufd/event.c +++ b/drivers/iommu/iommufd/event.c @@ -339,6 +339,67 @@ static const struct iommufd_event_ops iommufd_event_iopf_ops = { .write = &iommufd_event_iopf_fops_write, };
+/* IOMMUFD_OBJ_EVENT_VIRQ Functions */ + +void iommufd_event_virq_destroy(struct iommufd_object *obj) +{ + struct iommufd_event *event = + container_of(obj, struct iommufd_event, obj); + struct iommufd_event_virq *event_virq = to_event_virq(event); + struct iommufd_viommu_irq *virq, *next; + + /* + * The iommufd object's reference count is zero at this point. + * We can be confident that no other threads are currently + * accessing this pointer. Therefore, acquiring the mutex here + * is unnecessary. + */ + list_for_each_entry_safe(virq, next, &event->deliver, node) { + list_del(&virq->node); + kfree(virq); + } + destroy_workqueue(event_virq->irq_wq); + list_del(&event_virq->node); + refcount_dec(&event_virq->viommu->obj.users); +} + +static ssize_t +iommufd_event_virq_fops_read(struct iommufd_event *event, + char __user *buf, size_t count, loff_t *ppos) +{ + size_t done = 0; + int rc = 0; + + if (*ppos) + return -ESPIPE; + + mutex_lock(&event->mutex); + while (!list_empty(&event->deliver) && count > done) { + struct iommufd_viommu_irq *virq = + list_first_entry(&event->deliver, + struct iommufd_viommu_irq, node); + void *virq_data = (void *)virq + sizeof(*virq); + + if (virq->irq_len > count - done) + break; + + if (copy_to_user(buf + done, virq_data, virq->irq_len)) { + rc = -EFAULT; + break; + } + done += virq->irq_len; + list_del(&virq->node); + kfree(virq); + } + mutex_unlock(&event->mutex); + + return done == 0 ? rc : done; +} + +static const struct iommufd_event_ops iommufd_event_virq_ops = { + .read = &iommufd_event_virq_fops_read, +}; + /* Common Event Functions */
static ssize_t iommufd_event_fops_read(struct file *filep, char __user *buf, @@ -475,3 +536,78 @@ int iommufd_event_iopf_alloc(struct iommufd_ucmd *ucmd)
return rc; } + +int iommufd_event_virq_alloc(struct iommufd_ucmd *ucmd) +{ + struct iommu_virq_alloc *cmd = ucmd->cmd; + struct iommufd_event_virq *event_virq; + struct workqueue_struct *irq_wq; + struct iommufd_viommu *viommu; + int fdno; + int rc; + + if (cmd->flags) + return -EOPNOTSUPP; + if (cmd->type == IOMMU_VIRQ_TYPE_NONE) + return -EINVAL; + + viommu = iommufd_get_viommu(ucmd, cmd->viommu_id); + if (IS_ERR(viommu)) + return PTR_ERR(viommu); + down_write(&viommu->virqs_rwsem); + + if (iommufd_viommu_find_event_virq(viommu, cmd->type)) { + rc = -EEXIST; + goto out_unlock_virqs; + } + + event_virq = __iommufd_object_alloc(ucmd->ictx, event_virq, + IOMMUFD_OBJ_EVENT_VIRQ, common.obj); + if (IS_ERR(event_virq)) { + rc = PTR_ERR(event_virq); + goto out_unlock_virqs; + } + + irq_wq = alloc_workqueue("viommu_irq/%d", WQ_UNBOUND, 0, + event_virq->common.obj.id); + if (!irq_wq) { + rc = -ENOMEM; + goto out_abort; + } + + rc = iommufd_event_init(&event_virq->common, "[iommufd-viommu-irq]", + ucmd->ictx, &fdno, &iommufd_event_virq_ops); + if (rc) + goto out_irq_wq; + + event_virq->irq_wq = irq_wq; + event_virq->viommu = viommu; + event_virq->type = cmd->type; + cmd->out_virq_id = event_virq->common.obj.id; + cmd->out_virq_fd = fdno; + + rc = iommufd_ucmd_respond(ucmd, sizeof(*cmd)); + if (rc) + goto out_put_fdno; + iommufd_object_finalize(ucmd->ictx, &event_virq->common.obj); + + fd_install(fdno, event_virq->common.filep); + + list_add_tail(&event_virq->node, &viommu->virqs); + refcount_inc(&viommu->obj.users); + + goto out_unlock_virqs; +out_put_fdno: + put_unused_fd(fdno); + fput(event_virq->common.filep); + iommufd_event_deinit(&event_virq->common); +out_irq_wq: + destroy_workqueue(irq_wq); +out_abort: + iommufd_object_abort_and_destroy(ucmd->ictx, &event_virq->common.obj); +out_unlock_virqs: + up_write(&viommu->virqs_rwsem); + iommufd_put_object(ucmd->ictx, &viommu->obj); + + return rc; +} diff --git a/drivers/iommu/iommufd/iommufd_private.h b/drivers/iommu/iommufd/iommufd_private.h index c22d72c981c7..be1f1813672e 100644 --- a/drivers/iommu/iommufd/iommufd_private.h +++ b/drivers/iommu/iommufd/iommufd_private.h @@ -133,6 +133,7 @@ enum iommufd_object_type { IOMMUFD_OBJ_IOAS, IOMMUFD_OBJ_ACCESS, IOMMUFD_OBJ_EVENT_IOPF, + IOMMUFD_OBJ_EVENT_VIRQ, IOMMUFD_OBJ_VIOMMU, #ifdef CONFIG_IOMMUFD_TEST IOMMUFD_OBJ_SELFTEST, @@ -567,6 +568,43 @@ static inline int iommufd_hwpt_replace_device(struct iommufd_device *idev, return iommu_group_replace_domain(idev->igroup->group, hwpt->domain); }
+struct iommufd_event_virq { + struct iommufd_event common; + struct iommufd_viommu *viommu; + struct workqueue_struct *irq_wq; + struct list_head node; + + unsigned int type; +}; + +static inline struct iommufd_event_virq * +to_event_virq(struct iommufd_event *event) +{ + return container_of(event, struct iommufd_event_virq, common); +} + +static inline struct iommufd_event_virq * +iommufd_get_event_virq(struct iommufd_ucmd *ucmd, u32 id) +{ + return container_of(iommufd_get_object(ucmd->ictx, id, + IOMMUFD_OBJ_EVENT_VIRQ), + struct iommufd_event_virq, common.obj); +} + +int iommufd_event_virq_alloc(struct iommufd_ucmd *ucmd); +void iommufd_event_virq_destroy(struct iommufd_object *obj); + +struct iommufd_viommu_irq { + struct iommufd_event_virq *event_virq; + struct list_head node; + ssize_t irq_len; +}; + +static inline int iommufd_event_virq_handler(struct iommufd_viommu_irq *virq) +{ + return iommufd_event_notify(&virq->event_virq->common, &virq->node); +} + struct iommufd_viommu { struct iommufd_object obj; struct iommufd_ctx *ictx; @@ -575,6 +613,8 @@ struct iommufd_viommu { /* The locking order is vdev_ids_rwsem -> igroup::lock */ struct rw_semaphore vdev_ids_rwsem; struct xarray vdev_ids; + struct rw_semaphore virqs_rwsem; + struct list_head virqs;
const struct iommufd_viommu_ops *ops;
@@ -595,6 +635,20 @@ iommufd_get_viommu(struct iommufd_ucmd *ucmd, u32 id) struct iommufd_viommu, obj); }
+static inline struct iommufd_event_virq * +iommufd_viommu_find_event_virq(struct iommufd_viommu *viommu, u32 type) +{ + struct iommufd_event_virq *event_virq, *next; + + lockdep_assert_held(&viommu->virqs_rwsem); + + list_for_each_entry_safe(event_virq, next, &viommu->virqs, node) { + if (event_virq->type == type) + return event_virq; + } + return NULL; +} + int iommufd_viommu_alloc_ioctl(struct iommufd_ucmd *ucmd); void iommufd_viommu_destroy(struct iommufd_object *obj); int iommufd_viommu_set_vdev_id(struct iommufd_ucmd *ucmd); diff --git a/drivers/iommu/iommufd/main.c b/drivers/iommu/iommufd/main.c index 015f492afab1..22381ba031b5 100644 --- a/drivers/iommu/iommufd/main.c +++ b/drivers/iommu/iommufd/main.c @@ -361,6 +361,8 @@ static const struct iommufd_ioctl_op iommufd_ioctl_ops[] = { IOCTL_OP(IOMMU_DESTROY, iommufd_destroy, struct iommu_destroy, id), IOCTL_OP(IOMMU_FAULT_QUEUE_ALLOC, iommufd_event_iopf_alloc, struct iommu_fault_alloc, out_fault_fd), + IOCTL_OP(IOMMU_VIRQ_ALLOC, iommufd_event_virq_alloc, + struct iommu_virq_alloc, out_virq_fd), IOCTL_OP(IOMMU_GET_HW_INFO, iommufd_get_hw_info, struct iommu_hw_info, __reserved), IOCTL_OP(IOMMU_HWPT_ALLOC, iommufd_hwpt_alloc, struct iommu_hwpt_alloc, @@ -528,6 +530,9 @@ static const struct iommufd_object_ops iommufd_object_ops[] = { [IOMMUFD_OBJ_EVENT_IOPF] = { .destroy = iommufd_event_iopf_destroy, }, + [IOMMUFD_OBJ_EVENT_VIRQ] = { + .destroy = iommufd_event_virq_destroy, + }, [IOMMUFD_OBJ_VIOMMU] = { .destroy = iommufd_viommu_destroy, }, diff --git a/drivers/iommu/iommufd/viommu.c b/drivers/iommu/iommufd/viommu.c index a4ba8bff4a26..9adc9c62ada9 100644 --- a/drivers/iommu/iommufd/viommu.c +++ b/drivers/iommu/iommufd/viommu.c @@ -67,6 +67,8 @@ int iommufd_viommu_alloc_ioctl(struct iommufd_ucmd *ucmd)
xa_init(&viommu->vdev_ids); init_rwsem(&viommu->vdev_ids_rwsem); + INIT_LIST_HEAD(&viommu->virqs); + init_rwsem(&viommu->virqs_rwsem);
refcount_inc(&viommu->hwpt->common.obj.users);
diff --git a/include/uapi/linux/iommufd.h b/include/uapi/linux/iommufd.h index 0d973486b604..f9ec07efed8d 100644 --- a/include/uapi/linux/iommufd.h +++ b/include/uapi/linux/iommufd.h @@ -54,6 +54,7 @@ enum { IOMMUFD_CMD_VIOMMU_ALLOC = 0x8f, IOMMUFD_CMD_VIOMMU_SET_VDEV_ID = 0x90, IOMMUFD_CMD_VIOMMU_UNSET_VDEV_ID = 0x91, + IOMMUFD_CMD_VIRQ_ALLOC = 0x92, };
/** @@ -951,4 +952,35 @@ struct iommu_viommu_unset_vdev_id { __aligned_u64 vdev_id; }; #define IOMMU_VIOMMU_UNSET_VDEV_ID _IO(IOMMUFD_TYPE, IOMMUFD_CMD_VIOMMU_UNSET_VDEV_ID) + +/** + * enum iommu_virq_type - Virtual IRQ Type + * @IOMMU_VIRQ_TYPE_NONE: INVALID type + */ +enum iommu_virq_type { + IOMMU_VIRQ_TYPE_NONE = 0, +}; + +/** + * struct iommu_virq_alloc - ioctl(IOMMU_VIRQ_ALLOC) + * @size: sizeof(struct iommu_virq_alloc) + * @flags: Must be 0 + * @viommu: viommu ID to associate the virtual IRQ with + * @type: Type of the virtual IRQ. Must be defined in enum iommu_virq_type + * @out_virq_id: The ID of the new VIRQ + * @out_fault_fd: The fd of the new VIRQ + * + * Explicitly allocate a virtual IRQ handler for a VIOMMU. A VIOMMU can have + * multiple FDs for different @type, but is confined to have only one FD per + * @type. + */ +struct iommu_virq_alloc { + __u32 size; + __u32 flags; + __u32 viommu_id; + __u32 type; + __u32 out_virq_id; + __u32 out_virq_fd; +}; +#define IOMMU_VIRQ_ALLOC _IO(IOMMUFD_TYPE, IOMMUFD_CMD_VIRQ_ALLOC) #endif