[PATCH v2] vsock/virtio: set vsock frontend ready in virtio_vsock_probe()

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian <redacted>
---
 net/vmw_vsock/virtio_transport.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c992a..dc834b8fd 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -639,6 +639,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
 
 	mutex_unlock(&the_virtio_vsock_mutex);
 
+	virtio_device_ready(vdev);
+
 	return 0;
 
 out:

-- 
2.17.1

在 2021/7/20 下午3:13, Xianting Tian 写道:

Just notice this:

commit 5b40a7daf51812b35cf05d1601a779a7043f8414
Author: Rusty Russell [off-list ref]
Date:   Tue Feb 17 16:12:44 2015 +1030

     virtio: don't set VIRTIO_CONFIG_S_DRIVER_OK twice.

     I noticed this with the console device.  It's not *wrong*, just a bit
     weird.

     Signed-off-by: Rusty Russell [off-list ref]

diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
index b9f70dfc4751..5ce2aa48fc6e 100644
--- a/drivers/virtio/virtio.c
+++ b/drivers/virtio/virtio.c

@@ -236,7 +236,10 @@ static int virtio_dev_probe(struct device *_d)
         if (err)
                 goto err;

-       add_status(dev, VIRTIO_CONFIG_S_DRIVER_OK);
+       /* If probe didn't do it, mark device DRIVER_OK ourselves. */
+       if (!(dev->config->get_status(dev) & VIRTIO_CONFIG_S_DRIVER_OK))
+               virtio_device_ready(dev);
+
         if (drv->scan)
                 drv->scan(dev);

So I think we need to be consistent: switch to use virtio_device_ready() 
for all the drivers, and then we can remove this step and warn if 
(DRIVER_OK) is not set.

Thanks

thanks for the findings.

Let me make furture patches to make consistent.

在 2021/7/20 下午3:54, Jason Wang 写道:

On Tue, Jul 20, 2021 at 03:13:37PM +0800, Xianting Tian wrote:

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian <redacted>
---
 net/vmw_vsock/virtio_transport.c | 2 ++
 1 file changed, 2 insertions(+)

Please include a changelog when you send v2, v3, etc patches.

Why is this patch necessary?

The core virtio_dev_probe() code already calls virtio_device_ready for
us:

  static int virtio_dev_probe(struct device *_d)
  {
      ...
      err = drv->probe(dev);
      if (err)
          goto err;
  
      /* If probe didn't do it, mark device DRIVER_OK ourselves. */
      if (!(dev->config->get_status(dev) & VIRTIO_CONFIG_S_DRIVER_OK))
          virtio_device_ready(dev);

OK,

thanks

在 2021/7/20 下午6:23, Stefan Hajnoczi 写道:

On Tue, Jul 20, 2021 at 03:13:37PM +0800, Xianting Tian wrote:

quoted

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian <redacted>
---
  net/vmw_vsock/virtio_transport.c | 2 ++
  1 file changed, 2 insertions(+)

Please include a changelog when you send v2, v3, etc patches.

quoted

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c992a..dc834b8fd 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -639,6 +639,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
  
  	mutex_unlock(&the_virtio_vsock_mutex);
  
+	virtio_device_ready(vdev);

Why is this patch necessary?

The core virtio_dev_probe() code already calls virtio_device_ready for
us:

   static int virtio_dev_probe(struct device *_d)
   {
       ...
       err = drv->probe(dev);
       if (err)
           goto err;
   
       /* If probe didn't do it, mark device DRIVER_OK ourselves. */
       if (!(dev->config->get_status(dev) & VIRTIO_CONFIG_S_DRIVER_OK))
           virtio_device_ready(dev);

在 2021/7/20 下午6:23, Stefan Hajnoczi 写道:

On Tue, Jul 20, 2021 at 03:13:37PM +0800, Xianting Tian wrote:

quoted

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian<redacted>
---
  net/vmw_vsock/virtio_transport.c | 2 ++
  1 file changed, 2 insertions(+)

Please include a changelog when you send v2, v3, etc patches.

OK, thanks.

quoted

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c992a..dc834b8fd 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -639,6 +639,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
  
  	mutex_unlock(&the_virtio_vsock_mutex);
  
+	virtio_device_ready(vdev);

Why is this patch necessary?

Sorry, I didn't notice the check in virtio_dev_probe(),

As Jason comment,  I alsoe think we need to be consistent: switch to use 
virtio_device_ready() for all the drivers. What's opinion about this?

The core virtio_dev_probe() code already calls virtio_device_ready for
us:

   static int virtio_dev_probe(struct device *_d)
   {
       ...
       err = drv->probe(dev);
       if (err)
           goto err;
   
       /* If probe didn't do it, mark device DRIVER_OK ourselves. */
       if (!(dev->config->get_status(dev) & VIRTIO_CONFIG_S_DRIVER_OK))
           virtio_device_ready(dev);

On Tue, Jul 20, 2021 at 07:05:39PM +0800, Xianting Tian wrote:

在 2021/7/20 下午6:23, Stefan Hajnoczi 写道:

quoted

On Tue, Jul 20, 2021 at 03:13:37PM +0800, Xianting Tian wrote:

quoted

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian<redacted>
---
  net/vmw_vsock/virtio_transport.c | 2 ++
  1 file changed, 2 insertions(+)

Please include a changelog when you send v2, v3, etc patches.

OK, thanks.

quoted

quoted

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c992a..dc834b8fd 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -639,6 +639,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
  	mutex_unlock(&the_virtio_vsock_mutex);
+	virtio_device_ready(vdev);

Why is this patch necessary?

Sorry, I didn't notice the check in virtio_dev_probe(),

As Jason comment,  I alsoe think we need to be consistent: switch to use
virtio_device_ready() for all the drivers. What's opinion about this?

According to the documentation the virtio_device_read() API is optional:

  /**
   * virtio_device_ready - enable vq use in probe function
   * @vdev: the device
   *
   * Driver must call this to use vqs in the probe function.
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   *
   * Note: vqs are enabled automatically after probe returns.
     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   */

Many drivers do not use vqs during the ->probe() function. They don't
need to call virtio_device_ready(). That's why the virtio_vsock driver
doesn't call it.

But if a ->probe() function needs to send virtqueue buffers, e.g. to
query the device or activate some device feature, then the driver will
need to call it explicitly.

The documentation is clear and this design is less error-prone than
relying on all drivers to call it manually. I suggest leaving things
unchanged.

Stefan

Got it.

thanks for the comments,


在 2021/7/20 下午9:12, Stefan Hajnoczi 写道:

On Tue, Jul 20, 2021 at 07:05:39PM +0800, Xianting Tian wrote:

quoted

在 2021/7/20 下午6:23, Stefan Hajnoczi 写道:

quoted

On Tue, Jul 20, 2021 at 03:13:37PM +0800, Xianting Tian wrote:

quoted

Add the missed virtio_device_ready() to set vsock frontend ready.

Signed-off-by: Xianting Tian<redacted>
---
   net/vmw_vsock/virtio_transport.c | 2 ++
   1 file changed, 2 insertions(+)

Please include a changelog when you send v2, v3, etc patches.

OK, thanks.

quoted

quoted

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e0c2c992a..dc834b8fd 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -639,6 +639,8 @@ static int virtio_vsock_probe(struct virtio_device *vdev)
   	mutex_unlock(&the_virtio_vsock_mutex);
+	virtio_device_ready(vdev);

Why is this patch necessary?

Sorry, I didn't notice the check in virtio_dev_probe(),

As Jason comment,  I alsoe think we need to be consistent: switch to use
virtio_device_ready() for all the drivers. What's opinion about this?

According to the documentation the virtio_device_read() API is optional:

   /**
    * virtio_device_ready - enable vq use in probe function
    * @vdev: the device
    *
    * Driver must call this to use vqs in the probe function.
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    *
    * Note: vqs are enabled automatically after probe returns.
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    */

Many drivers do not use vqs during the ->probe() function. They don't
need to call virtio_device_ready(). That's why the virtio_vsock driver
doesn't call it.

But if a ->probe() function needs to send virtqueue buffers, e.g. to
query the device or activate some device feature, then the driver will
need to call it explicitly.

The documentation is clear and this design is less error-prone than
relying on all drivers to call it manually. I suggest leaving things
unchanged.

Stefan