[dpdk-dev] [Suspected-Phishing][PATCH] net/vhost: add an API for get queue status

Maxime Coquelin maxime.coquelin at redhat.com
Tue Jun 25 09:49:59 CEST 2019


And about short term, can't you just call
rte_eth_vhost_get_queue_event() in loop at startup until you get -1 and
build the states based on that?

As states->seen is zero-initialized, you would get all queues that have
been enabled before event handler is registered, right?

Thanks,
Maxime

On 6/25/19 9:36 AM, Maxime Coquelin wrote:
> 
> 
> On 6/25/19 9:00 AM, Noa Ezra wrote:
>>
>>
>>> -----Original Message-----
>>> From: Maxime Coquelin [mailto:maxime.coquelin at redhat.com]
>>> Sent: Monday, June 24, 2019 7:47 PM
>>> To: Noa Ezra <noae at mellanox.com>; tiwei.bie at intel.com;
>>> zhihong.wang at intel.com
>>> Cc: Matan Azrad <matan at mellanox.com>; dev at dpdk.org
>>> Subject: Re: [Suspected-Phishing][PATCH] net/vhost: add an API for get
>>> queue status
>>>
>>>
>>>
>>> On 6/24/19 1:08 PM, Noa Ezra wrote:
>>>> Hi,
>>>> What do you say about this patch?
>>>
>>> I acknowledge we miss a way to get the queue state, but I don't like
>>> introducing new APIs for PMD drivers.
>>>
>>> I looked at ethdev ops on Friday, but it seems no one currently 
>>> available can
>>> do the job.
>>>
>>> I would suggest to create new ethdev ops like:
>>> .get_tx_queue_state(int port_id, int queue_id) .get_rx_queue_state(int
>>> port_id, int queue_id)
>>>
>>> That would return wether the queue is enabled or not.
>>>
>>> What do you think?
>>
>> I understand what you say about not having new APIs in the PMD, but I 
>> see that there are already APIs in the vhost pmd.
> 
> Indeed, and I hope we can get rid off them one day.
> 
>> rte_eth_vhost_get_queue_event allows to get the queue_state 
>> asynchronously, so it sounds reasonable to add an API that allows to 
>> get the queue_state synchronously.
> 
> I think we can get rid off rte_eth_vhost_get_queue_event easily.
> Maybe we can even just drop it if we implement the new ethdev ops.
> 
>> Maybe in the future we can create new ethdev ops like you suggested 
>> and remove both APIs for getting queue_state.
>> Currently I don’t have the resources for such implementation and we 
>> really need the ability to get queue_state.
> 
> It is anyway too late for v19.08, as the proposal deadline is over 3
> weeks now. I will only accept fixes or small patches that have limited
> impact.
> 
> Here, the patch is adding a new API, which is significant.
> And with the new API stability rules that is being put in place, it may
> imply that we'll need to support it for more than 2 years.
> 
> So for v19.11, if you think you don't have the bandwidth, maybe I can
> try to do it. I really prefer that than having an API already deprecated
> when merged in master.
> 
> Thanks,
> Maxime
>> Thanks,
>> Noa.
>>
>>
>>> Thanks,
>>> Maxime
>>>> Thanks,
>>>> Noa.
>>>>
>>>>> -----Original Message-----
>>>>> From: Noa Ezra [mailto:noae at mellanox.com]
>>>>> Sent: Wednesday, June 19, 2019 9:15 AM
>>>>> To: maxime.coquelin at redhat.com
>>>>> Cc: Matan Azrad <matan at mellanox.com>; dev at dpdk.org; Noa Ezra
>>>>> <noae at mellanox.com>
>>>>> Subject: [Suspected-Phishing][PATCH] net/vhost: add an API for get
>>>>> queue status
>>>>>
>>>>> Add an API that returns queue status for requested queue in the port.
>>>>> The queue's status can be changed before the user has signed for the
>>>>> queue state event interrupt. In this case the user can't know the
>>>>> current queue's status. This API returns the current status.
>>>>>
>>>>> Signed-off-by: Noa Ezra <noae at mellanox.com>
>>>>> Reviewed-by: Matan Azrad <matan at mellanox.com>
>>>>> ---
>>>>>    drivers/net/vhost/rte_eth_vhost.c           | 47
>>>>> +++++++++++++++++++++++++++++
>>>>>    drivers/net/vhost/rte_eth_vhost.h           | 18 +++++++++++
>>>>>    drivers/net/vhost/rte_pmd_vhost_version.map |  6 ++++
>>>>>    3 files changed, 71 insertions(+)
>>>>>
>>>>> diff --git a/drivers/net/vhost/rte_eth_vhost.c
>>>>> b/drivers/net/vhost/rte_eth_vhost.c
>>>>> index 9a54020..cad1e5c 100644
>>>>> --- a/drivers/net/vhost/rte_eth_vhost.c
>>>>> +++ b/drivers/net/vhost/rte_eth_vhost.c
>>>>> @@ -855,6 +855,7 @@ struct vhost_xstats_name_off {
>>>>>        /* won't be NULL */
>>>>>        state = vring_states[eth_dev->data->port_id];
>>>>>        rte_spinlock_lock(&state->lock);
>>>>> +
>>>>>        state->cur[vring] = enable;
>>>>>        state->max_vring = RTE_MAX(vring, state->max_vring);
>>>>>        rte_spinlock_unlock(&state->lock);
>>>>> @@ -874,6 +875,52 @@ struct vhost_xstats_name_off {  };
>>>>>
>>>>>    int
>>>>> +rte_eth_vhost_get_queue_status(uint16_t port_id, bool rx, uint16_t
>>>>> queue_id,
>>>>> +        bool *queue_status)
>>>>> +{
>>>>> +    struct rte_vhost_vring_state *state;
>>>>> +    struct internal_list *list;
>>>>> +    struct rte_eth_dev *eth_dev;
>>>>> +    int found = 0;
>>>>> +    uint16_t nb_q = 0;
>>>>> +
>>>>> +    if (port_id >= RTE_MAX_ETHPORTS) {
>>>>> +        VHOST_LOG(ERR, "Invalid port id\n");
>>>>> +        return -1;
>>>>> +    }
>>>>> +    TAILQ_FOREACH(list, &internal_list, next) {
>>>>> +        eth_dev = list->eth_dev;
>>>>> +        if (eth_dev->data->port_id == port_id) {
>>>>> +            nb_q = rx ? eth_dev->data->nb_rx_queues :
>>>>> +                    eth_dev->data->nb_tx_queues;
>>>>> +            found = 1;
>>>>> +            break;
>>>>> +        }
>>>>> +    }
>>>>> +    if (!found) {
>>>>> +        VHOST_LOG(ERR, "No device found for port id %u\n",
>>>>> port_id);
>>>>> +        return -1;
>>>>> +    }
>>>>> +    if (queue_id >= nb_q) {
>>>>> +        VHOST_LOG(ERR, "Invalid queue id\n");
>>>>> +        return -1;
>>>>> +    }
>>>>> +
>>>>> +    state = vring_states[port_id];
>>>>> +    if (!state) {
>>>>> +        VHOST_LOG(ERR, "Unused port\n");
>>>>> +        return -1;
>>>>> +    }
>>>>> +
>>>>> +    rte_spinlock_lock(&state->lock);
>>>>> +    *queue_status = rx ? state->cur[queue_id * 2 + 1] :
>>>>> +            state->cur[queue_id * 2];
>>>>> +    rte_spinlock_unlock(&state->lock);
>>>>> +
>>>>> +    return 0;
>>>>> +}
>>>>> +
>>>>> +int
>>>>>    rte_eth_vhost_get_queue_event(uint16_t port_id,
>>>>>            struct rte_eth_vhost_queue_event *event)  { diff --git
>>>>> a/drivers/net/vhost/rte_eth_vhost.h
>>>>> b/drivers/net/vhost/rte_eth_vhost.h
>>>>> index 0e68b9f..1e65c69 100644
>>>>> --- a/drivers/net/vhost/rte_eth_vhost.h
>>>>> +++ b/drivers/net/vhost/rte_eth_vhost.h
>>>>> @@ -44,6 +44,24 @@ int rte_eth_vhost_get_queue_event(uint16_t
>>> port_id,
>>>>>            struct rte_eth_vhost_queue_event *event);
>>>>>
>>>>>    /**
>>>>> + * Get queue status for specific queue in the port.
>>>>> + *
>>>>> + * @param[in] port_id
>>>>> + *  Port id.
>>>>> + * @param[in] rx
>>>>> + *  True is rx, False if tx
>>>>> + * @paran[in] queue_id
>>>>> + *  Queue_id
>>>>> + * @param[out] queue_status
>>>>> + *  Pointer to a boolean, True is enable, False if disable.
>>>>> + * @return
>>>>> + *  - On success, zero, queue_status is updated.
>>>>> + *  - On failure, a negative value, queue_status is not updated.
>>>>> + */
>>>>> +int rte_eth_vhost_get_queue_status(uint16_t port_id, bool rx,
>>>>> +uint16_t
>>>>> queue_id,
>>>>> +        bool *queue_status);
>>>>> +
>>>>> +/**
>>>>>     * Get the 'vid' value associated with the specified port.
>>>>>     *
>>>>>     * @return
>>>>> diff --git a/drivers/net/vhost/rte_pmd_vhost_version.map
>>>>> b/drivers/net/vhost/rte_pmd_vhost_version.map
>>>>> index 695db85..1eabfd2 100644
>>>>> --- a/drivers/net/vhost/rte_pmd_vhost_version.map
>>>>> +++ b/drivers/net/vhost/rte_pmd_vhost_version.map
>>>>> @@ -11,3 +11,9 @@ DPDK_16.11 {
>>>>>
>>>>>        rte_eth_vhost_get_vid_from_port_id;
>>>>>    };
>>>>> +
>>>>> +DPDK_19.08 {
>>>>> +    global:
>>>>> +
>>>>> +    rte_eth_vhost_get_queue_status;
>>>>> +};
>>>>> -- 
>>>>> 1.8.3.1
>>>>


More information about the dev mailing list