DPDK patches and discussions
 help / color / mirror / Atom feed
From: Ganapati Kundapura <ganapati.kundapura@intel.com>
To: dev@dpdk.org
Subject: [PATCH v1] eventdev/crypto: restore opaque field between dequeue and enqueue
Date: Tue,  4 Jun 2024 11:08:47 -0500	[thread overview]
Message-ID: <20240604160847.1573821-1-ganapati.kundapura@intel.com> (raw)

For session-less crypto operations, event info is contained in
crypto op metadata for each event which is restored in event
from the crypto op metadata response info.

For session based crypto operations, crypto op contains per session
based event info in crypto op metadata. If any PMD passes any
implementaion specific data in ev::impl_opaque on each event,
it's not getting restored.

This patch stores ev::impl_opaque in mbuf's dynamic field before
enqueueing to cryptodev and restores ev::impl_opaque from
mbuf's dynamic field after dequeueing crypto op from cryptodev
for session based crypto operations.

Signed-off-by: Ganapati Kundapura <ganapati.kundapura@intel.com>

diff --git a/lib/eventdev/rte_event_crypto_adapter.c b/lib/eventdev/rte_event_crypto_adapter.c
index db1c7f3..91a30ca 100644
--- a/lib/eventdev/rte_event_crypto_adapter.c
+++ b/lib/eventdev/rte_event_crypto_adapter.c
@@ -138,6 +138,27 @@ static struct event_crypto_adapter **event_crypto_adapter;
 	} \
 } while (0)
 
+#define ECA_DYNFIELD_NAME "eca_ev_opaque_data"
+/** Device-specific metadata field type */
+typedef uint8_t eca_dynfield_t;
+/** Dynamic mbuf field for device-specific metadata */
+int eca_dynfield_offset = -1;
+
+static int
+eca_dynfield_register(void)
+{
+	static const struct rte_mbuf_dynfield eca_dynfield_desc = {
+		.name = ECA_DYNFIELD_NAME,
+		.size = sizeof(eca_dynfield_t),
+		.align = alignof(eca_dynfield_t),
+		.flags = 0,
+	};
+
+	eca_dynfield_offset =
+		rte_mbuf_dynfield_register(&eca_dynfield_desc);
+	return eca_dynfield_offset;
+}
+
 static inline int
 eca_valid_id(uint8_t id)
 {
@@ -491,6 +512,24 @@ eca_enq_to_cryptodev(struct event_crypto_adapter *adapter, struct rte_event *ev,
 		crypto_op = ev[i].event_ptr;
 		if (crypto_op == NULL)
 			continue;
+
+		/** ev::impl_opaque field passed on from eventdev PMD could
+		 *  have different value per event.
+		 *  For session-based crypto operations retain ev::impl_opaque
+		 *  into mbuf dynfield and restore it back after copying event
+		 *  information from session event metadata.
+		 *  For session-less, each crypto operation carries event
+		 *  metadata and retains ev::impl_opaque information to be
+		 *  passed back to eventdev PMD.
+		 */
+		if (crypto_op->sess_type == RTE_CRYPTO_OP_WITH_SESSION) {
+			struct rte_mbuf *mbuf = crypto_op->sym->m_src;
+
+			*RTE_MBUF_DYNFIELD(mbuf,
+					eca_dynfield_offset,
+					eca_dynfield_t *) = ev[i].impl_opaque;
+		}
+
 		m_data = rte_cryptodev_session_event_mdata_get(crypto_op);
 		if (m_data == NULL) {
 			rte_pktmbuf_free(crypto_op->sym->m_src);
@@ -657,6 +696,21 @@ eca_ops_enqueue_burst(struct event_crypto_adapter *adapter,
 
 		rte_memcpy(ev, &m_data->response_info, sizeof(*ev));
 		ev->event_ptr = ops[i];
+
+		/** restore ev::impl_opaque from mbuf dyn field
+		 *  for session based crypto operation.
+		 *  For session-less, each crypto operations carries event
+		 *  metadata and retains ev::impl_opaque information to be
+		 *  passed back to eventdev PMD.
+		 */
+		if (ops[i]->sess_type == RTE_CRYPTO_OP_WITH_SESSION) {
+			struct rte_mbuf *mbuf = ops[i]->sym->m_src;
+
+			ev->impl_opaque = *RTE_MBUF_DYNFIELD(mbuf,
+							eca_dynfield_offset,
+							eca_dynfield_t *);
+		}
+
 		ev->event_type = RTE_EVENT_TYPE_CRYPTODEV;
 		if (adapter->implicit_release_disabled)
 			ev->op = RTE_EVENT_OP_FORWARD;
@@ -895,6 +949,16 @@ eca_init_service(struct event_crypto_adapter *adapter, uint8_t id)
 	}
 
 	adapter->implicit_release_disabled = (uint8_t)impl_rel;
+
+	/* Register for mbuf dyn field to store/restore ev::impl_opaque */
+	eca_dynfield_offset = eca_dynfield_register();
+	if (eca_dynfield_offset  < 0) {
+		RTE_EDEV_LOG_ERR("Failed to register eca mbuf dyn field");
+		eca_circular_buffer_free(&adapter->ebuf);
+		rte_free(adapter);
+		return -EINVAL;
+	}
+
 	adapter->service_inited = 1;
 
 	return ret;
-- 
2.6.4


             reply	other threads:[~2024-06-04 16:08 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-06-04 16:08 Ganapati Kundapura [this message]
2024-06-04 16:18 ` [PATCH v2] " Ganapati Kundapura
2024-06-06  8:27   ` Jerin Jacob
2024-06-06 10:02     ` Jerin Jacob
2024-06-06 11:47       ` Kundapura, Ganapati
2024-06-06 11:46   ` [PATCH v3] " Ganapati Kundapura
2024-06-06 12:17     ` Gujjar, Abhinandan S
2024-06-06 13:37       ` Kundapura, Ganapati
2024-06-06 13:35     ` [PATCH v4] " Ganapati Kundapura
2024-06-06 13:48       ` [PATCH v5] " Ganapati Kundapura
2024-06-07  8:20         ` 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=20240604160847.1573821-1-ganapati.kundapura@intel.com \
    --to=ganapati.kundapura@intel.com \
    --cc=dev@dpdk.org \
    /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).