From: Pavan Nikhilesh <pbhagavatula@marvell.com> To: <jerinj@marvell.com>, Sunil Kumar Kori <skori@marvell.com>, "Pavan Nikhilesh" <pbhagavatula@marvell.com> Cc: <dev@dpdk.org> Subject: [PATCH 5/6] examples/l2fwd-event: clean up worker state before exit Date: Wed, 27 Apr 2022 02:44:11 +0530 Message-ID: <20220426211412.6138-5-pbhagavatula@marvell.com> (raw) In-Reply-To: <20220426211412.6138-1-pbhagavatula@marvell.com> Event ports are configured to implicitly release the scheduler contexts currently held in the next call to rte_event_dequeue_burst(). A worker core might still hold a scheduling context during exit, as the next call to rte_event_dequeue_burst() is never made. This might lead to deadlock based on the worker exit timing and when there are very less number of flows. Add clean up function to release any scheduling contexts held by the worker by using RTE_EVENT_OP_RELEASE. Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com> --- examples/l2fwd-event/l2fwd_common.c | 34 +++++++++++++++++++++++++++++ examples/l2fwd-event/l2fwd_common.h | 3 +++ examples/l2fwd-event/l2fwd_event.c | 31 ++++++++++++++++---------- 3 files changed, 56 insertions(+), 12 deletions(-) diff --git a/examples/l2fwd-event/l2fwd_common.c b/examples/l2fwd-event/l2fwd_common.c index cf3d1b8aaf..15bfe790a0 100644 --- a/examples/l2fwd-event/l2fwd_common.c +++ b/examples/l2fwd-event/l2fwd_common.c @@ -114,3 +114,37 @@ l2fwd_event_init_ports(struct l2fwd_resources *rsrc) return nb_ports_available; } + +static void +l2fwd_event_vector_array_free(struct rte_event events[], uint16_t num) +{ + uint16_t i; + + for (i = 0; i < num; i++) { + rte_pktmbuf_free_bulk(events[i].vec->mbufs, + events[i].vec->nb_elem); + rte_mempool_put(rte_mempool_from_obj(events[i].vec), + events[i].vec); + } +} + +void +l2fwd_event_worker_cleanup(uint8_t event_d_id, uint8_t port_id, + struct rte_event events[], uint16_t nb_enq, + uint16_t nb_deq, uint8_t is_vector) +{ + int i; + + if (nb_deq) { + if (is_vector) + l2fwd_event_vector_array_free(events + nb_enq, + nb_deq - nb_enq); + else + for (i = nb_enq; i < nb_deq; i++) + rte_pktmbuf_free(events[i].mbuf); + + for (i = 0; i < nb_deq; i++) + events[i].op = RTE_EVENT_OP_RELEASE; + rte_event_enqueue_burst(event_d_id, port_id, events, nb_deq); + } +} diff --git a/examples/l2fwd-event/l2fwd_common.h b/examples/l2fwd-event/l2fwd_common.h index 396e238c6a..bff3b65abf 100644 --- a/examples/l2fwd-event/l2fwd_common.h +++ b/examples/l2fwd-event/l2fwd_common.h @@ -140,5 +140,8 @@ l2fwd_get_rsrc(void) } int l2fwd_event_init_ports(struct l2fwd_resources *rsrc); +void l2fwd_event_worker_cleanup(uint8_t event_d_id, uint8_t port_id, + struct rte_event events[], uint16_t nb_enq, + uint16_t nb_deq, uint8_t is_vector); #endif /* __L2FWD_COMMON_H__ */ diff --git a/examples/l2fwd-event/l2fwd_event.c b/examples/l2fwd-event/l2fwd_event.c index 6df3cdfeab..63450537fe 100644 --- a/examples/l2fwd-event/l2fwd_event.c +++ b/examples/l2fwd-event/l2fwd_event.c @@ -193,6 +193,7 @@ l2fwd_event_loop_single(struct l2fwd_resources *rsrc, evt_rsrc->evq.nb_queues - 1]; const uint64_t timer_period = rsrc->timer_period; const uint8_t event_d_id = evt_rsrc->event_d_id; + uint8_t enq = 0, deq = 0; struct rte_event ev; if (port_id < 0) @@ -203,26 +204,28 @@ l2fwd_event_loop_single(struct l2fwd_resources *rsrc, while (!rsrc->force_quit) { /* Read packet from eventdev */ - if (!rte_event_dequeue_burst(event_d_id, port_id, &ev, 1, 0)) + deq = rte_event_dequeue_burst(event_d_id, port_id, &ev, 1, 0); + if (!deq) continue; l2fwd_event_fwd(rsrc, &ev, tx_q_id, timer_period, flags); if (flags & L2FWD_EVENT_TX_ENQ) { - while (rte_event_enqueue_burst(event_d_id, port_id, - &ev, 1) && - !rsrc->force_quit) - ; + do { + enq = rte_event_enqueue_burst(event_d_id, + port_id, &ev, 1); + } while (!enq && !rsrc->force_quit); } if (flags & L2FWD_EVENT_TX_DIRECT) { - while (!rte_event_eth_tx_adapter_enqueue(event_d_id, - port_id, - &ev, 1, 0) && - !rsrc->force_quit) - ; + do { + enq = rte_event_eth_tx_adapter_enqueue( + event_d_id, port_id, &ev, 1, 0); + } while (!enq && !rsrc->force_quit); } } + + l2fwd_event_worker_cleanup(event_d_id, port_id, &ev, enq, deq, 0); } static __rte_always_inline void @@ -237,7 +240,7 @@ l2fwd_event_loop_burst(struct l2fwd_resources *rsrc, const uint8_t event_d_id = evt_rsrc->event_d_id; const uint8_t deq_len = evt_rsrc->deq_depth; struct rte_event ev[MAX_PKT_BURST]; - uint16_t nb_rx, nb_tx; + uint16_t nb_rx = 0, nb_tx = 0; uint8_t i; if (port_id < 0) @@ -280,6 +283,8 @@ l2fwd_event_loop_burst(struct l2fwd_resources *rsrc, ev + nb_tx, nb_rx - nb_tx, 0); } } + + l2fwd_event_worker_cleanup(event_d_id, port_id, ev, nb_rx, nb_tx, 0); } static __rte_always_inline void @@ -419,7 +424,7 @@ l2fwd_event_loop_vector(struct l2fwd_resources *rsrc, const uint32_t flags) const uint8_t event_d_id = evt_rsrc->event_d_id; const uint8_t deq_len = evt_rsrc->deq_depth; struct rte_event ev[MAX_PKT_BURST]; - uint16_t nb_rx, nb_tx; + uint16_t nb_rx = 0, nb_tx = 0; uint8_t i; if (port_id < 0) @@ -462,6 +467,8 @@ l2fwd_event_loop_vector(struct l2fwd_resources *rsrc, const uint32_t flags) nb_rx - nb_tx, 0); } } + + l2fwd_event_worker_cleanup(event_d_id, port_id, ev, nb_rx, nb_tx, 1); } static void __rte_noinline -- 2.25.1
next prev parent reply other threads:[~2022-04-26 21:14 UTC|newest] Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-04-26 21:14 [PATCH 1/6] app/eventdev: simplify signal handling and teardown Pavan Nikhilesh 2022-04-26 21:14 ` [PATCH 2/6] app/eventdev: clean up worker state before exit Pavan Nikhilesh 2022-05-13 13:40 ` Jerin Jacob 2022-04-26 21:14 ` [PATCH 3/6] examples/eventdev: " Pavan Nikhilesh 2022-04-26 21:14 ` [PATCH 4/6] examples/l3fwd: " Pavan Nikhilesh 2022-04-26 21:14 ` Pavan Nikhilesh [this message] 2022-04-26 21:14 ` [PATCH 6/6] examples/ipsec-secgw: cleanup " Pavan Nikhilesh 2022-05-13 13:41 ` Jerin Jacob 2022-05-13 11:49 ` [PATCH 1/6] app/eventdev: simplify signal handling and teardown Jerin Jacob 2022-05-13 13:39 ` Jerin Jacob 2022-05-13 16:07 ` [PATCH v2 " pbhagavatula 2022-05-13 16:07 ` [PATCH v2 2/6] app/eventdev: clean up worker state before exit pbhagavatula 2022-05-13 16:07 ` [PATCH v2 3/6] examples/eventdev: " pbhagavatula 2022-05-13 16:07 ` [PATCH v2 4/6] examples/l3fwd: " pbhagavatula 2022-05-13 16:07 ` [PATCH v2 5/6] examples/l2fwd-event: " pbhagavatula 2022-05-13 16:07 ` [PATCH v2 6/6] examples/ipsec-secgw: cleanup " pbhagavatula 2022-05-16 16:46 ` [PATCH v2 1/6] app/eventdev: simplify signal handling and teardown Jerin Jacob
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=20220426211412.6138-5-pbhagavatula@marvell.com \ --to=pbhagavatula@marvell.com \ --cc=dev@dpdk.org \ --cc=jerinj@marvell.com \ --cc=skori@marvell.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
DPDK patches and discussions This inbox may be cloned and mirrored by anyone: git clone --mirror http://inbox.dpdk.org/dev/0 dev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 dev dev/ http://inbox.dpdk.org/dev \ dev@dpdk.org public-inbox-index dev Example config snippet for mirrors. Newsgroup available over NNTP: nntp://inbox.dpdk.org/inbox.dpdk.dev AGPL code for this site: git clone https://public-inbox.org/public-inbox.git