Re: [dpdk-dev] [PATCH 2/2] ethdev: make rte flow API thread safe

[Ori Kam <orika@nvidia.com>]

Hi Suanming,

Small comments,
PSB,

Best,
Ori

> -----Original Message-----
> From: Suanming Mou <suanmingm@nvidia.com>
> Sent: Sunday, September 27, 2020 11:20 AM
> Cc: dev@dpdk.org
> Subject: [PATCH 2/2] ethdev: make rte flow API thread safe
> 
> Currently, the rte flow functions are not defined as thread safety.

For a given port. (between ports they are thread safe)

> DPDK applications either call the functions in single thread or add
> locks around the functions for the critical section.
> 
> For PMDs support the flow operations thread safe natively, the
> redundant protection in application hurts the performance of the
> rte flow operation functions.
> 
> And the restriction of thread safety not guaranteed for the rte
> flow functions also limits the applications' expectation.
> 
> This feature is going to change the rte flow functions to be thread
> safety. As different PMDs have different flow operations, some may

Safety => safe I think this is true also to other places in this patch.

> support thread safety already and others may not. For PMDs don't
> support flow thread safe operation, a new lock is defined in ethdev
> in order to protects thread unsafe PMDs from rte flow level.
> 
> A new RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE device flag is added to
> determine whether the PMD supports thread safe flow operation or not.
> For PMDs support thread safe flow operations, set the
> RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE flag, rte flow level functions will
> skip the thread safe helper lock for these PMDs. Again the rte flow
> level thread safe lock only works when PMD operation functions are
> not thread safety.
safe
> 
> For the PMDs which don't want the default mutex lock, just set the
> flag in the PMD, and add the perfer type of lock in the PMD. Then
> the default mutex lock is easily replaced by the PMD level lock.
> 
> The change has no effect on the current DPDK applications. No change
> is required for the current DPDK applications. If no lock contention
> with the added rte flow level mutex, the mutex only does the atomic
> increasing in pthread_mutex_lock() and decreasing in
> pthread_mutex_unlock(). No futex() syscall will be involved.
> 
> Signed-off-by: Suanming Mou <suanmingm@nvidia.com>
> ---
>  doc/guides/prog_guide/rte_flow.rst  |  7 ++--
>  drivers/net/mlx5/linux/mlx5_os.c    |  2 +
>  lib/librte_ethdev/rte_ethdev.c      |  2 +
>  lib/librte_ethdev/rte_ethdev.h      |  2 +
>  lib/librte_ethdev/rte_ethdev_core.h |  4 ++
>  lib/librte_ethdev/rte_flow.c        | 84 ++++++++++++++++++++++++++++---------
>  6 files changed, 78 insertions(+), 23 deletions(-)
> 
> diff --git a/doc/guides/prog_guide/rte_flow.rst
> b/doc/guides/prog_guide/rte_flow.rst
> index 119b128..6f7997a 100644
> --- a/doc/guides/prog_guide/rte_flow.rst
> +++ b/doc/guides/prog_guide/rte_flow.rst
> @@ -3046,10 +3046,6 @@ Caveats
>  - API operations are synchronous and blocking (``EAGAIN`` cannot be
>    returned).
> 
> -- There is no provision for re-entrancy/multi-thread safety, although nothing
> -  should prevent different devices from being configured at the same
> -  time. PMDs may protect their control path functions accordingly.
> -
>  - Stopping the data path (TX/RX) should not be necessary when managing flow
>    rules. If this cannot be achieved naturally or with workarounds (such as
>    temporarily replacing the burst function pointers), an appropriate error
> @@ -3101,6 +3097,9 @@ This interface additionally defines the following
> helper function:
>  - ``rte_flow_ops_get()``: get generic flow operations structure from a
>    port.
> 
> +If PMD interfaces do not support re-entrancy/multi-thread safety, rte flow
> +level functions will do it by mutex.
> +

I think you should add which flag the app should test to see if the PMD
support thread safe.

>  More will be added over time.
> 
>  Device compatibility
> diff --git a/drivers/net/mlx5/linux/mlx5_os.c
> b/drivers/net/mlx5/linux/mlx5_os.c
> index 2a4beb0..daefb7d 100644
> --- a/drivers/net/mlx5/linux/mlx5_os.c
> +++ b/drivers/net/mlx5/linux/mlx5_os.c
> @@ -1351,6 +1351,8 @@
>  			goto error;
>  		}
>  	}
> +	if (priv->config.dv_flow_en)
> +		eth_dev->data->dev_flags =
> RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
>  	return eth_dev;

I don't think this code is relevant for this patch.
I also think you are missing the | when assigning the value.

>  error:
>  	if (priv) {
> diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
> index dfe5c1b..d6fd17e 100644
> --- a/lib/librte_ethdev/rte_ethdev.c
> +++ b/lib/librte_ethdev/rte_ethdev.c
> @@ -500,6 +500,7 @@ struct rte_eth_dev *
>  	strlcpy(eth_dev->data->name, name, sizeof(eth_dev->data->name));
>  	eth_dev->data->port_id = port_id;
>  	eth_dev->data->mtu = RTE_ETHER_MTU;
> +	pthread_mutex_init(&eth_dev->data->fts_mutex, NULL);
> 
>  unlock:
>  	rte_spinlock_unlock(&rte_eth_dev_shared_data->ownership_lock);
> @@ -562,6 +563,7 @@ struct rte_eth_dev *
>  		rte_free(eth_dev->data->mac_addrs);
>  		rte_free(eth_dev->data->hash_mac_addrs);
>  		rte_free(eth_dev->data->dev_private);
> +		pthread_mutex_destroy(&eth_dev->data->fts_mutex);
>  		memset(eth_dev->data, 0, sizeof(struct rte_eth_dev_data));
>  	}
> 
> diff --git a/lib/librte_ethdev/rte_ethdev.h b/lib/librte_ethdev/rte_ethdev.h
> index 645a186..61bd09f 100644
> --- a/lib/librte_ethdev/rte_ethdev.h
> +++ b/lib/librte_ethdev/rte_ethdev.h
> @@ -1669,6 +1669,8 @@ struct rte_eth_dev_owner {
>  #define RTE_ETH_DEV_REPRESENTOR  0x0010
>  /** Device does not support MAC change after started */
>  #define RTE_ETH_DEV_NOLIVE_MAC_ADDR  0x0020
> +/** Device PMD supports thread safety flow operation */
> +#define RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE  0x0040
> 
>  /**
>   * Iterates over valid ethdev ports owned by a specific owner.
> diff --git a/lib/librte_ethdev/rte_ethdev_core.h
> b/lib/librte_ethdev/rte_ethdev_core.h
> index fd3bf92..89df65a 100644
> --- a/lib/librte_ethdev/rte_ethdev_core.h
> +++ b/lib/librte_ethdev/rte_ethdev_core.h
> @@ -5,6 +5,9 @@
>  #ifndef _RTE_ETHDEV_CORE_H_
>  #define _RTE_ETHDEV_CORE_H_
> 
> +#include <pthread.h>
> +#include <sys/types.h>
> +
>  /**
>   * @file
>   *
> @@ -180,6 +183,7 @@ struct rte_eth_dev_data {
>  			 *   Valid if RTE_ETH_DEV_REPRESENTOR in dev_flags.
>  			 */
> 
> +	pthread_mutex_t fts_mutex; /**< rte flow ops thread safety mutex. */
>  	uint64_t reserved_64s[4]; /**< Reserved for future fields */
>  	void *reserved_ptrs[4];   /**< Reserved for future fields */
>  } __rte_cache_aligned;
> diff --git a/lib/librte_ethdev/rte_flow.c b/lib/librte_ethdev/rte_flow.c
> index f8fdd68..ca80b12 100644
> --- a/lib/librte_ethdev/rte_flow.c
> +++ b/lib/librte_ethdev/rte_flow.c
> @@ -207,6 +207,20 @@ struct rte_flow_desc_data {
>  	return -rte_errno;
>  }
> 
> +static void
> +flow_lock(struct rte_eth_dev *dev)
> +{
> +	if (!(dev->data->dev_flags &
> RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE))
> +		pthread_mutex_lock(&dev->data->fts_mutex);
> +}
Why not inline function?
Also I don't think you should check the PMD support in this function
when this function is called it should lock.

> +
> +static void
> +flow_unlock(struct rte_eth_dev *dev)
> +{
> +	if (!(dev->data->dev_flags &
> RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE))
> +		pthread_mutex_unlock(&dev->data->fts_mutex);
> +}

Same comments as above.

> +
>  static int
>  flow_err(uint16_t port_id, int ret, struct rte_flow_error *error)
>  {
> @@ -346,12 +360,16 @@ struct rte_flow_desc_data {
>  {
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> +	int ret;
> 
>  	if (unlikely(!ops))
>  		return -rte_errno;
> -	if (likely(!!ops->validate))
> -		return flow_err(port_id, ops->validate(dev, attr, pattern,
> -						       actions, error), error);
> +	if (likely(!!ops->validate)) {
> +		flow_lock(dev);
> +		ret = ops->validate(dev, attr, pattern, actions, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -372,7 +390,9 @@ struct rte_flow *
>  	if (unlikely(!ops))
>  		return NULL;
>  	if (likely(!!ops->create)) {
> +		flow_lock(dev);
>  		flow = ops->create(dev, attr, pattern, actions, error);
> +		flow_unlock(dev);
>  		if (flow == NULL)
>  			flow_err(port_id, -rte_errno, error);
>  		return flow;
> @@ -390,12 +410,16 @@ struct rte_flow *
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (unlikely(!ops))
>  		return -rte_errno;
> -	if (likely(!!ops->destroy))
> -		return flow_err(port_id, ops->destroy(dev, flow, error),
> -				error);
> +	if (likely(!!ops->destroy)) {
> +		flow_lock(dev);
> +		ret = ops->destroy(dev, flow, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -408,11 +432,16 @@ struct rte_flow *
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (unlikely(!ops))
>  		return -rte_errno;
> -	if (likely(!!ops->flush))
> -		return flow_err(port_id, ops->flush(dev, error), error);
> +	if (likely(!!ops->flush)) {
> +		flow_lock(dev);
> +		ret = ops->flush(dev, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -428,12 +457,16 @@ struct rte_flow *
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (!ops)
>  		return -rte_errno;
> -	if (likely(!!ops->query))
> -		return flow_err(port_id, ops->query(dev, flow, action, data,
> -						    error), error);
> +	if (likely(!!ops->query)) {
> +		flow_lock(dev);
> +		ret = ops->query(dev, flow, action, data, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -447,11 +480,16 @@ struct rte_flow *
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (!ops)
>  		return -rte_errno;
> -	if (likely(!!ops->isolate))
> -		return flow_err(port_id, ops->isolate(dev, set, error), error);
> +	if (likely(!!ops->isolate)) {
> +		flow_lock(dev);
> +		ret = ops->isolate(dev, set, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -1224,12 +1262,16 @@ enum rte_flow_conv_item_spec_type {
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (unlikely(!ops))
>  		return -rte_errno;
> -	if (likely(!!ops->dev_dump))
> -		return flow_err(port_id, ops->dev_dump(dev, file, error),
> -				error);
> +	if (likely(!!ops->dev_dump)) {
> +		flow_lock(dev);
> +		ret = ops->dev_dump(dev, file, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOSYS,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOSYS));
> @@ -1241,12 +1283,16 @@ enum rte_flow_conv_item_spec_type {
>  {
>  	struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>  	const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error);
> +	int ret;
> 
>  	if (unlikely(!ops))
>  		return -rte_errno;
> -	if (likely(!!ops->get_aged_flows))
> -		return flow_err(port_id, ops->get_aged_flows(dev, contexts,
> -				nb_contexts, error), error);
> +	if (likely(!!ops->get_aged_flows)) {
> +		flow_lock(dev);
> +		ret = ops->get_aged_flows(dev, contexts, nb_contexts, error);
> +		flow_unlock(dev);
> +		return flow_err(port_id, ret, error);
> +	}
>  	return rte_flow_error_set(error, ENOTSUP,
>  				  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  				  NULL, rte_strerror(ENOTSUP));
> --
> 1.8.3.1