[dpdk-dev] [RFC 2/2] vhost: support vhost-user request only handled by external backend

[Ilya Maximets]

On 27.02.2019 13:02, Maxime Coquelin wrote:
> External backends may have specific requests to handle, and so
> we don't want the vhost-user lib to handle these requests as
> errors.
> 
> This patch also catch the case where a request is neither handled
> by the external backend nor by the vhost library.
> 
> Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
> ---
>  lib/librte_vhost/vhost_user.c | 28 +++++++++++++++-------------
>  1 file changed, 15 insertions(+), 13 deletions(-)
> 
> diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
> index 36c0c676d..bae5ef1cc 100644
> --- a/lib/librte_vhost/vhost_user.c
> +++ b/lib/librte_vhost/vhost_user.c
> @@ -1924,27 +1924,29 @@ vhost_user_msg_handler(int vid, int fd)
>  	}
>  
>  	ret = read_vhost_message(fd, &msg);
> -	if (ret <= 0 || msg.request.master >= VHOST_USER_MAX) {
> +	if (ret <= 0) {
>  		if (ret < 0)
>  			RTE_LOG(ERR, VHOST_CONFIG,
>  				"vhost read message failed\n");
> -		else if (ret == 0)
> +		else
>  			RTE_LOG(INFO, VHOST_CONFIG,
>  				"vhost peer closed\n");
> -		else
> -			RTE_LOG(ERR, VHOST_CONFIG,
> -				"vhost read incorrect message\n");
>  
>  		return -1;
>  	}
>  
>  	ret = 0;
> -	if (msg.request.master != VHOST_USER_IOTLB_MSG)
> -		RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
> -			vhost_message_str[msg.request.master]);
> -	else
> -		RTE_LOG(DEBUG, VHOST_CONFIG, "read message %s\n",
> -			vhost_message_str[msg.request.master]);
> +	request = msg.request.master;
> +	if (request < VHOST_USER_MAX && vhost_message_str[req]) {
> +		if (request != VHOST_USER_IOTLB_MSG)
> +			RTE_LOG(INFO, VHOST_CONFIG, "read message %s\n",
> +				vhost_message_str[request]);
> +		else if (
> +			RTE_LOG(DEBUG, VHOST_CONFIG, "read message %s\n",
> +				vhost_message_str[request]);

There is no need for the 'if' without the body.

> +	} else {
> +		RTE_LOG(INFO, VHOST_CONFIG, "External request %d\n", request);

External requests could be annoying. Maybe we'll need to move them under DBG ?
I'm not sure.

> +	}
>  
>  	ret = vhost_user_check_and_alloc_queue_pair(dev, &msg);
>  	if (ret < 0) {
> @@ -1960,7 +1962,7 @@ vhost_user_msg_handler(int vid, int fd)
>  	 * inactive, so it is safe. Otherwise taking the access_lock
>  	 * would cause a dead lock.
>  	 */
> -	switch (msg.request.master) {
> +	switch (request) {
>  	case VHOST_USER_SET_FEATURES:
>  	case VHOST_USER_SET_PROTOCOL_FEATURES:
>  	case VHOST_USER_SET_OWNER:
> @@ -1985,6 +1987,7 @@ vhost_user_msg_handler(int vid, int fd)
>  
>  	}
>  
> +	ret = RTE_VHOST_MSG_RESULT_ERR;

This will break the 'vhost_crypto', because it has no 'pre_msg_handler'
and master will skip to 'post_msg_handler', but it will not be called
because current status is ERR.

Maybe it's easier to introduce RTE_VHOST_MSG_RESULT_NOT_HANDLED and convert
it to ERR before the reply ?
This will require changing the pre_msg_handlers to return
RTE_VHOST_MSG_RESULT_NOT_HANDLED if message wasn't recognized.
And we'll possibly be able to drop the 'skip_master' in this case.

>  	if (dev->extern_ops.pre_msg_handle) {
>  		ret = (*dev->extern_ops.pre_msg_handle)(dev->vid,
>  				(void *)&msg, &skip_master);
> @@ -1997,7 +2000,6 @@ vhost_user_msg_handler(int vid, int fd)
>  			goto skip_to_post_handle;
>  	}
>  
> -	request = msg.request.master;
>  	if (request > VHOST_USER_NONE && request < VHOST_USER_MAX) {
>  		if (!vhost_message_handlers[request])
>  			goto skip_to_post_handle;
>