From: Jerin Jacob <jerin.jacob@caviumnetworks.com>
To: Nikhil Rao <nikhil.rao@intel.com>
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH v2] eventdev: add callback for Rx adapter SW transfers
Date: Fri, 6 Jul 2018 10:50:23 +0530 [thread overview]
Message-ID: <20180706052021.GB6922@jerin> (raw)
In-Reply-To: <1530168490-23223-1-git-send-email-nikhil.rao@intel.com>
-----Original Message-----
> Date: Thu, 28 Jun 2018 12:18:10 +0530
> From: Nikhil Rao <nikhil.rao@intel.com>
> To: jerin.jacob@caviumnetworks.com
> CC: nikhil.rao@intel.com, dev@dpdk.org
> Subject: [PATCH v2] eventdev: add callback for Rx adapter SW transfers
> X-Mailer: git-send-email 1.8.3.1
>
> External Email
>
> Add ability for application to register a callback function
> for SW transfers, the callback can decide which packets can
> be enqueued to the event device.
>
> Signed-off-by: Nikhil Rao <nikhil.rao@intel.com>
> Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
> ---
Applied to dpdk-next-eventdev/master. Thanks.
>
> Changelog
> =========
>
> v1->v2:
> * Change function names to rte_event_eth_rx_adapter_ namespace. (Jerin Jacob)
> * Sort function names in map file alphabetically. (Jerin Jacob)
> * Doc updated to mention RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT. (Jerin Jacob)
> * Fix logic that handles callback return value.
>
> lib/librte_eventdev/rte_event_eth_rx_adapter.h | 81 ++++++++++++++++++++++
> lib/librte_eventdev/rte_event_eth_rx_adapter.c | 70 ++++++++++++++++++-
> .../prog_guide/event_ethernet_rx_adapter.rst | 17 +++++
> lib/librte_eventdev/rte_eventdev_version.map | 25 +++----
> 4 files changed, 178 insertions(+), 15 deletions(-)
>
> diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.h b/lib/librte_eventdev/rte_event_eth_rx_adapter.h
> index 97f25e9..332ee21 100644
> --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.h
> +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.h
> @@ -63,6 +63,20 @@
> * rte_event_eth_rx_adapter_service_id_get() function can be used to retrieve
> * the service function ID of the adapter in this case.
> *
> + * For SW based packet transfers, i.e., when the
> + * RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT is not set in the adapter's
> + * capabilities flags for a particular ethernet device, the service function
> + * temporarily enqueues mbufs to an event buffer before batch enqueueing these
> + * to the event device. If the buffer fills up, the service function stops
> + * dequeueing packets from the ethernet device. The application may want to
> + * monitor the buffer fill level and instruct the service function to
> + * selectively buffer packets. The application may also use some other
> + * criteria to decide which packets should enter the event device even when
> + * the event buffer fill level is low. The
> + * rte_event_eth_rx_adapter_cb_register() function allows the
> + * application to register a callback that selects which packets to enqueue
> + * to the event device.
> + *
> * Note:
> * 1) Devices created after an instance of rte_event_eth_rx_adapter_create
> * should be added to a new instance of the rx adapter.
> @@ -206,6 +220,47 @@ struct rte_event_eth_rx_adapter_stats {
> * @warning
> * @b EXPERIMENTAL: this API may change without prior notice
> *
> + * Callback function invoked by the SW adapter before it continues
> + * to process packets. The callback is passed the size of the enqueue
> + * buffer in the SW adapter and the occupancy of the buffer. The
> + * callback can use these values to decide which mbufs should be
> + * enqueued to the event device. If the return value of the callback
> + * is less than nb_mbuf then the SW adapter uses the return value to
> + * enqueue enq_mbuf[] to the event device.
> + *
> + * @param eth_dev_id
> + * Port identifier of the Ethernet device.
> + * @param queue_id
> + * Receive queue index.
> + * @param enqueue_buf_size
> + * Total enqueue buffer size.
> + * @param enqueue_buf_count
> + * mbuf count in enqueue buffer.
> + * @param mbuf
> + * mbuf array.
> + * @param nb_mbuf
> + * mbuf count.
> + * @param cb_arg
> + * Callback argument.
> + * @param[out] enq_mbuf
> + * The adapter enqueues enq_mbuf[] if the return value of the
> + * callback is less than nb_mbuf
> + * @return
> + * Returns the number of mbufs should be enqueued to eventdev
> + */
> +typedef uint16_t (*rte_event_eth_rx_adapter_cb_fn)(uint16_t eth_dev_id,
> + uint16_t queue_id,
> + uint32_t enqueue_buf_size,
> + uint32_t enqueue_buf_count,
> + struct rte_mbuf **mbuf,
> + uint16_t nb_mbuf,
> + void *cb_arg,
> + struct rte_mbuf **enq_buf);
> +
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change without prior notice
> + *
> * Create a new ethernet Rx event adapter with the specified identifier.
> *
> * @param id
> @@ -426,6 +481,32 @@ int rte_event_eth_rx_adapter_stats_get(uint8_t id,
> */
> int rte_event_eth_rx_adapter_service_id_get(uint8_t id, uint32_t *service_id);
>
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change without prior notice
> + *
> + * Register callback to process Rx packets, this is supported for
> + * SW based packet transfers.
> + * @see rte_event_eth_rx_cb_fn
> + *
> + * @param id
> + * Adapter identifier.
> + * @param eth_dev_id
> + * Port identifier of Ethernet device.
> + * @param cb_fn
> + * Callback function.
> + * @param cb_arg
> + * Callback arg.
> + * @return
> + * - 0: Success
> + * - <0: Error code on failure.
> + */
> +int __rte_experimental
> +rte_event_eth_rx_adapter_cb_register(uint8_t id,
> + uint16_t eth_dev_id,
> + rte_event_eth_rx_adapter_cb_fn cb_fn,
> + void *cb_arg);
> +
> #ifdef __cplusplus
> }
> #endif
> diff --git a/lib/librte_eventdev/rte_event_eth_rx_adapter.c b/lib/librte_eventdev/rte_event_eth_rx_adapter.c
> index 8b86b18..8821e33 100644
> --- a/lib/librte_eventdev/rte_event_eth_rx_adapter.c
> +++ b/lib/librte_eventdev/rte_event_eth_rx_adapter.c
> @@ -147,6 +147,10 @@ struct rte_event_eth_rx_adapter {
> struct eth_device_info {
> struct rte_eth_dev *dev;
> struct eth_rx_queue_info *rx_queue;
> + /* Rx callback */
> + rte_event_eth_rx_adapter_cb_fn cb_fn;
> + /* Rx callback argument */
> + void *cb_arg;
> /* Set if ethdev->eventdev packet transfer uses a
> * hardware mechanism
> */
> @@ -759,11 +763,12 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
> uint16_t num)
> {
> uint32_t i;
> - struct eth_device_info *eth_device_info =
> + struct eth_device_info *dev_info =
> &rx_adapter->eth_devices[eth_dev_id];
> struct eth_rx_queue_info *eth_rx_queue_info =
> - ð_device_info->rx_queue[rx_queue_id];
> -
> + &dev_info->rx_queue[rx_queue_id];
> + struct rte_eth_event_enqueue_buffer *buf =
> + &rx_adapter->event_enqueue_buffer;
> int32_t qid = eth_rx_queue_info->event_queue_id;
> uint8_t sched_type = eth_rx_queue_info->sched_type;
> uint8_t priority = eth_rx_queue_info->priority;
> @@ -774,6 +779,8 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
> uint32_t rss;
> int do_rss;
> uint64_t ts;
> + struct rte_mbuf *cb_mbufs[BATCH_SIZE];
> + uint16_t nb_cb;
>
> /* 0xffff ffff if PKT_RX_RSS_HASH is set, otherwise 0 */
> rss_mask = ~(((m->ol_flags & PKT_RX_RSS_HASH) != 0) - 1);
> @@ -789,6 +796,19 @@ static uint16_t rxa_gcd_u16(uint16_t a, uint16_t b)
> }
> }
>
> +
> + nb_cb = dev_info->cb_fn ? dev_info->cb_fn(eth_dev_id, rx_queue_id,
> + ETH_EVENT_BUFFER_SIZE,
> + buf->count, mbufs,
> + num,
> + dev_info->cb_arg,
> + cb_mbufs) :
> + num;
> + if (nb_cb < num) {
> + mbufs = cb_mbufs;
> + num = nb_cb;
> + }
> +
> for (i = 0; i < num; i++) {
> m = mbufs[i];
> struct rte_event *ev = &events[i];
> @@ -2365,3 +2385,47 @@ static int rxa_sw_add(struct rte_event_eth_rx_adapter *rx_adapter,
>
> return rx_adapter->service_inited ? 0 : -ESRCH;
> }
> +
> +int rte_event_eth_rx_adapter_cb_register(uint8_t id,
> + uint16_t eth_dev_id,
> + rte_event_eth_rx_adapter_cb_fn cb_fn,
> + void *cb_arg)
> +{
> + struct rte_event_eth_rx_adapter *rx_adapter;
> + struct eth_device_info *dev_info;
> + uint32_t cap;
> + int ret;
> +
> + RTE_EVENT_ETH_RX_ADAPTER_ID_VALID_OR_ERR_RET(id, -EINVAL);
> + RTE_ETH_VALID_PORTID_OR_ERR_RET(eth_dev_id, -EINVAL);
> +
> + rx_adapter = rxa_id_to_adapter(id);
> + if (rx_adapter == NULL)
> + return -EINVAL;
> +
> + dev_info = &rx_adapter->eth_devices[eth_dev_id];
> + if (dev_info->rx_queue == NULL)
> + return -EINVAL;
> +
> + ret = rte_event_eth_rx_adapter_caps_get(rx_adapter->eventdev_id,
> + eth_dev_id,
> + &cap);
> + if (ret) {
> + RTE_EDEV_LOG_ERR("Failed to get adapter caps edev %" PRIu8
> + "eth port %" PRIu16, id, eth_dev_id);
> + return ret;
> + }
> +
> + if (cap & RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT) {
> + RTE_EDEV_LOG_ERR("Rx callback not supported for eth port %"
> + PRIu16, eth_dev_id);
> + return -EINVAL;
> + }
> +
> + rte_spinlock_lock(&rx_adapter->rx_lock);
> + dev_info->cb_fn = cb_fn;
> + dev_info->cb_arg = cb_arg;
> + rte_spinlock_unlock(&rx_adapter->rx_lock);
> +
> + return 0;
> +}
> diff --git a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> index 2f055ec..91ecc1b 100644
> --- a/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> +++ b/doc/guides/prog_guide/event_ethernet_rx_adapter.rst
> @@ -168,3 +168,20 @@ received on a polled Rx queue. The interrupt thread is affinitized to the same
> CPUs as the lcores of the Rx adapter service function, if the Rx adapter
> service function has not been mapped to any lcores, the interrupt thread
> is mapped to the master lcore.
> +
> +Rx Callback for SW Rx Adapter
> +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> +
> +For SW based packet transfers, i.e., when the
> +``RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT`` is not set in the adapter's
> +capabilities flags for a particular ethernet device, the service function
> +temporarily enqueues mbufs to an event buffer before batch enqueueing these
> +to the event device. If the buffer fills up, the service function stops
> +dequeueing packets from the ethernet device. The application may want to
> +monitor the buffer fill level and instruct the service function to selectively
> +enqueue packets to the event device. The application may also use some other
> +criteria to decide which packets should enter the event device even when
> +the event buffer fill level is low. The
> +``rte_event_eth_rx_adapter_cb_register()`` function allow the application
> +to register a callback that selects which packets to enqueue to the event
> +device.
> diff --git a/lib/librte_eventdev/rte_eventdev_version.map b/lib/librte_eventdev/rte_eventdev_version.map
> index c3f18d6..12835e9 100644
> --- a/lib/librte_eventdev/rte_eventdev_version.map
> +++ b/lib/librte_eventdev/rte_eventdev_version.map
> @@ -83,6 +83,19 @@ DPDK_18.05 {
> EXPERIMENTAL {
> global:
>
> + rte_event_crypto_adapter_caps_get;
> + rte_event_crypto_adapter_create;
> + rte_event_crypto_adapter_create_ext;
> + rte_event_crypto_adapter_event_port_get;
> + rte_event_crypto_adapter_free;
> + rte_event_crypto_adapter_queue_pair_add;
> + rte_event_crypto_adapter_queue_pair_del;
> + rte_event_crypto_adapter_service_id_get;
> + rte_event_crypto_adapter_start;
> + rte_event_crypto_adapter_stats_get;
> + rte_event_crypto_adapter_stats_reset;
> + rte_event_crypto_adapter_stop;
> + rte_event_eth_rx_adapter_cb_register;
> rte_event_timer_adapter_caps_get;
> rte_event_timer_adapter_create;
> rte_event_timer_adapter_create_ext;
> @@ -97,16 +110,4 @@ EXPERIMENTAL {
> rte_event_timer_arm_burst;
> rte_event_timer_arm_tmo_tick_burst;
> rte_event_timer_cancel_burst;
> - rte_event_crypto_adapter_caps_get;
> - rte_event_crypto_adapter_create;
> - rte_event_crypto_adapter_create_ext;
> - rte_event_crypto_adapter_event_port_get;
> - rte_event_crypto_adapter_free;
> - rte_event_crypto_adapter_queue_pair_add;
> - rte_event_crypto_adapter_queue_pair_del;
> - rte_event_crypto_adapter_service_id_get;
> - rte_event_crypto_adapter_start;
> - rte_event_crypto_adapter_stats_get;
> - rte_event_crypto_adapter_stats_reset;
> - rte_event_crypto_adapter_stop;
> };
> --
> 1.8.3.1
>
prev parent reply other threads:[~2018-07-06 5:20 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-06-28 6:48 Nikhil Rao
2018-07-06 5:20 ` Jerin Jacob [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180706052021.GB6922@jerin \
--to=jerin.jacob@caviumnetworks.com \
--cc=dev@dpdk.org \
--cc=nikhil.rao@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).