diff mbox series

[v5,4/6] vhost: Add worker related functions to support kthread

Message ID 20241230124445.1850997-5-lulu@redhat.com (mailing list archive)
State Not Applicable
Headers show
Series vhost: Add support of kthread API | expand

Checks

Context Check Description
netdev/tree_selection success Not a local patch

Commit Message

Cindy Lu Dec. 30, 2024, 12:43 p.m. UTC 
Restore the previously removed functions kthread_wakeup and
kthread_stop, and add two new function pointers to wake up and stop
the workers. The function vhost_worker_create will initialize these
pointers based on the value of inherit_owner.

The functions vhost_worker_queue() and vhost_worker_destroy() will
use the function pointer in vhost_worker, which is initialized
according to the inherit_owner value.

Signed-off-by: Cindy Lu <lulu@redhat.com>
---
 drivers/vhost/vhost.c | 84 ++++++++++++++++++++++++++++++++++---------
 drivers/vhost/vhost.h |  3 ++
 2 files changed, 71 insertions(+), 16 deletions(-)

Comments

Jason Wang Jan. 2, 2025, 3:33 a.m. UTC | #1 
On Mon, Dec 30, 2024 at 8:45 PM Cindy Lu <lulu@redhat.com> wrote:
>
> Restore the previously removed functions kthread_wakeup and
> kthread_stop, and add two new function pointers to wake up and stop
> the workers. The function vhost_worker_create will initialize these
> pointers based on the value of inherit_owner.
>
> The functions vhost_worker_queue() and vhost_worker_destroy() will
> use the function pointer in vhost_worker, which is initialized
> according to the inherit_owner value.

I'd suggest using "vhost: introduce worker ops to support multiple
thread models" as the title.

>
> Signed-off-by: Cindy Lu <lulu@redhat.com>
> ---
>  drivers/vhost/vhost.c | 84 ++++++++++++++++++++++++++++++++++---------
>  drivers/vhost/vhost.h |  3 ++
>  2 files changed, 71 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
> index 812dfd218bc2..ff17c42e2d1a 100644
> --- a/drivers/vhost/vhost.c
> +++ b/drivers/vhost/vhost.c
> @@ -243,7 +243,7 @@ static void vhost_worker_queue(struct vhost_worker *worker,
>                  * test_and_set_bit() implies a memory barrier.
>                  */
>                 llist_add(&work->node, &worker->work_list);
> -               vhost_task_wake(worker->vtsk);
> +               worker->worker_wakeup(worker);
>         }
>  }
>
> @@ -698,7 +698,7 @@ static void vhost_worker_destroy(struct vhost_dev *dev,
>
>         WARN_ON(!llist_empty(&worker->work_list));
>         xa_erase(&dev->worker_xa, worker->id);
> -       vhost_task_stop(worker->vtsk);
> +       worker->worker_stop(worker);
>         kfree(worker);
>  }
>
> @@ -721,14 +721,36 @@ static void vhost_workers_free(struct vhost_dev *dev)
>         xa_destroy(&dev->worker_xa);
>  }
>
> +static void vhost_task_wakeup_fn(struct vhost_worker *worker)
> +{
> +       return vhost_task_wake(worker->vtsk);
> +}
> +
> +static void vhost_kthread_wakeup_fn(struct vhost_worker *worker)
> +{
> +       wake_up_process(worker->kthread_task);
> +}
> +
> +static void vhost_task_stop_fn(struct vhost_worker *worker)
> +{
> +       return vhost_task_stop(worker->vtsk);
> +}
> +
> +static void vhost_kthread_stop_fn(struct vhost_worker *worker)
> +{
> +       kthread_stop(worker->kthread_task);
> +}
> +
>  static struct vhost_worker *vhost_worker_create(struct vhost_dev *dev)
>  {
>         struct vhost_worker *worker;
> -       struct vhost_task *vtsk;
> +       struct vhost_task *vtsk = NULL;
> +       struct task_struct *task = NULL;
>         char name[TASK_COMM_LEN];
>         int ret;
>         u32 id;
>
> +       /* Allocate resources for the worker */
>         worker = kzalloc(sizeof(*worker), GFP_KERNEL_ACCOUNT);
>         if (!worker)
>                 return NULL;
> @@ -736,27 +758,57 @@ static struct vhost_worker *vhost_worker_create(struct vhost_dev *dev)
>         worker->dev = dev;
>         snprintf(name, sizeof(name), "vhost-%d", current->pid);
>
> -       vtsk = vhost_task_create(vhost_run_work_list, vhost_worker_killed,
> -                                worker, name);
> -       if (!vtsk)
> -               goto free_worker;
> -
>         mutex_init(&worker->mutex);
>         init_llist_head(&worker->work_list);
>         worker->kcov_handle = kcov_common_handle();
> -       worker->vtsk = vtsk;
> +    /*
> +     * If inherit_owner is true we use vhost_tasks to create
> +     * the worker so all settings/limits like cgroups, NPROC,
> +     * scheduler, etc are inherited from the owner.
> +     * If false,we use kthreads and only attach to the same
> +     * cgroups as the owner for compat with older kernels.
> +     */
> +       if (dev->inherit_owner) {
> +               vtsk = vhost_task_create(vhost_run_work_list,
> +                                        vhost_worker_killed, worker, name);
> +               if (!vtsk)
> +                       goto free_worker;
> +
> +               worker->vtsk = vtsk;
> +               worker->worker_wakeup = vhost_task_wakeup_fn;
> +               worker->worker_stop = vhost_task_stop_fn;
> +
> +               vhost_task_start(vtsk);
> +               ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b,
> +                              GFP_KERNEL);
> +               if (ret < 0)
> +                       goto stop_worker;

Let's simply have a new ops like worker_create to avoid the if/else here.

> +       } else {
> +               task = kthread_create(vhost_run_work_kthread_list, worker,
> +                                     "vhost-%d", current->pid);
> +               if (IS_ERR(task)) {
> +                       ret = PTR_ERR(task);
> +                       goto free_worker;
> +               }
> +               worker->kthread_task = task;
> +               worker->worker_wakeup = vhost_kthread_wakeup_fn;
> +               worker->worker_stop = vhost_kthread_stop_fn;
>
> -       vhost_task_start(vtsk);
> +               wake_up_process(task);
> +               ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b,
> +                              GFP_KERNEL);
> +               if (ret < 0)
> +                       goto stop_worker;
>
> -       ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b, GFP_KERNEL);
> -       if (ret < 0)
> -               goto stop_worker;
> -       worker->id = id;
> +               ret = vhost_attach_task_to_cgroups(worker);
> +               if (ret)
> +                       goto stop_worker;
> +       }
>
> +       worker->id = id;
>         return worker;
> -
>  stop_worker:
> -       vhost_task_stop(vtsk);
> +       worker->worker_stop(worker);
>  free_worker:
>         kfree(worker);
>         return NULL;
> diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
> index c650c4506c70..63b1da08a2b0 100644
> --- a/drivers/vhost/vhost.h
> +++ b/drivers/vhost/vhost.h
> @@ -27,6 +27,7 @@ struct vhost_work {
>  };
>
>  struct vhost_worker {
> +       struct task_struct *kthread_task;
>         struct vhost_task       *vtsk;
>         struct vhost_dev        *dev;
>         /* Used to serialize device wide flushing with worker swapping. */
> @@ -36,6 +37,8 @@ struct vhost_worker {
>         u32                     id;
>         int                     attachment_cnt;
>         bool                    killed;
> +       void (*worker_wakeup)(struct vhost_worker *worker);
> +       void (*worker_stop)(struct vhost_worker *worker);

Let's use a dedicated ops structure for this.

Thanks

>  };
>
>  /* Poll a file (eventfd or socket) */
> --
> 2.45.0
>
diff mbox series

Patch

diff --git a/drivers/vhost/vhost.c b/drivers/vhost/vhost.c
index 812dfd218bc2..ff17c42e2d1a 100644
--- a/drivers/vhost/vhost.c
+++ b/drivers/vhost/vhost.c
@@ -243,7 +243,7 @@  static void vhost_worker_queue(struct vhost_worker *worker,
 		 * test_and_set_bit() implies a memory barrier.
 		 */
 		llist_add(&work->node, &worker->work_list);
-		vhost_task_wake(worker->vtsk);
+		worker->worker_wakeup(worker);
 	}
 }
 
@@ -698,7 +698,7 @@  static void vhost_worker_destroy(struct vhost_dev *dev,
 
 	WARN_ON(!llist_empty(&worker->work_list));
 	xa_erase(&dev->worker_xa, worker->id);
-	vhost_task_stop(worker->vtsk);
+	worker->worker_stop(worker);
 	kfree(worker);
 }
 
@@ -721,14 +721,36 @@  static void vhost_workers_free(struct vhost_dev *dev)
 	xa_destroy(&dev->worker_xa);
 }
 
+static void vhost_task_wakeup_fn(struct vhost_worker *worker)
+{
+	return vhost_task_wake(worker->vtsk);
+}
+
+static void vhost_kthread_wakeup_fn(struct vhost_worker *worker)
+{
+	wake_up_process(worker->kthread_task);
+}
+
+static void vhost_task_stop_fn(struct vhost_worker *worker)
+{
+	return vhost_task_stop(worker->vtsk);
+}
+
+static void vhost_kthread_stop_fn(struct vhost_worker *worker)
+{
+	kthread_stop(worker->kthread_task);
+}
+
 static struct vhost_worker *vhost_worker_create(struct vhost_dev *dev)
 {
 	struct vhost_worker *worker;
-	struct vhost_task *vtsk;
+	struct vhost_task *vtsk = NULL;
+	struct task_struct *task = NULL;
 	char name[TASK_COMM_LEN];
 	int ret;
 	u32 id;
 
+	/* Allocate resources for the worker */
 	worker = kzalloc(sizeof(*worker), GFP_KERNEL_ACCOUNT);
 	if (!worker)
 		return NULL;
@@ -736,27 +758,57 @@  static struct vhost_worker *vhost_worker_create(struct vhost_dev *dev)
 	worker->dev = dev;
 	snprintf(name, sizeof(name), "vhost-%d", current->pid);
 
-	vtsk = vhost_task_create(vhost_run_work_list, vhost_worker_killed,
-				 worker, name);
-	if (!vtsk)
-		goto free_worker;
-
 	mutex_init(&worker->mutex);
 	init_llist_head(&worker->work_list);
 	worker->kcov_handle = kcov_common_handle();
-	worker->vtsk = vtsk;
+    /*
+     * If inherit_owner is true we use vhost_tasks to create
+     * the worker so all settings/limits like cgroups, NPROC,
+     * scheduler, etc are inherited from the owner.
+     * If false,we use kthreads and only attach to the same
+     * cgroups as the owner for compat with older kernels.
+     */
+	if (dev->inherit_owner) {
+		vtsk = vhost_task_create(vhost_run_work_list,
+					 vhost_worker_killed, worker, name);
+		if (!vtsk)
+			goto free_worker;
+
+		worker->vtsk = vtsk;
+		worker->worker_wakeup = vhost_task_wakeup_fn;
+		worker->worker_stop = vhost_task_stop_fn;
+
+		vhost_task_start(vtsk);
+		ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b,
+			       GFP_KERNEL);
+		if (ret < 0)
+			goto stop_worker;
+	} else {
+		task = kthread_create(vhost_run_work_kthread_list, worker,
+				      "vhost-%d", current->pid);
+		if (IS_ERR(task)) {
+			ret = PTR_ERR(task);
+			goto free_worker;
+		}
+		worker->kthread_task = task;
+		worker->worker_wakeup = vhost_kthread_wakeup_fn;
+		worker->worker_stop = vhost_kthread_stop_fn;
 
-	vhost_task_start(vtsk);
+		wake_up_process(task);
+		ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b,
+			       GFP_KERNEL);
+		if (ret < 0)
+			goto stop_worker;
 
-	ret = xa_alloc(&dev->worker_xa, &id, worker, xa_limit_32b, GFP_KERNEL);
-	if (ret < 0)
-		goto stop_worker;
-	worker->id = id;
+		ret = vhost_attach_task_to_cgroups(worker);
+		if (ret)
+			goto stop_worker;
+	}
 
+	worker->id = id;
 	return worker;
-
 stop_worker:
-	vhost_task_stop(vtsk);
+	worker->worker_stop(worker);
 free_worker:
 	kfree(worker);
 	return NULL;
diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
index c650c4506c70..63b1da08a2b0 100644
--- a/drivers/vhost/vhost.h
+++ b/drivers/vhost/vhost.h
@@ -27,6 +27,7 @@  struct vhost_work {
 };
 
 struct vhost_worker {
+	struct task_struct *kthread_task;
 	struct vhost_task	*vtsk;
 	struct vhost_dev	*dev;
 	/* Used to serialize device wide flushing with worker swapping. */
@@ -36,6 +37,8 @@  struct vhost_worker {
 	u32			id;
 	int			attachment_cnt;
 	bool			killed;
+	void (*worker_wakeup)(struct vhost_worker *worker);
+	void (*worker_stop)(struct vhost_worker *worker);
 };
 
 /* Poll a file (eventfd or socket) */