From mboxrd@z Thu Jan 1 00:00:00 1970
Return-Path: <hemant.agrawal@nxp.com>
Received: from EUR04-VI1-obe.outbound.protection.outlook.com
(mail-eopbgr80052.outbound.protection.outlook.com [40.107.8.52])
by dpdk.org (Postfix) with ESMTP id 137BA4CAB
for <dev@dpdk.org>; Tue, 26 Mar 2019 13:18:12 +0100 (CET)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=nxp.com; s=selector1;
h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck;
bh=SKlP5dflNKjzmJPAWJ/guIpnSdMPewtpGeObvUam2s0=;
b=u8Wtm96vG6E8cQcly8YpjnR9YxApxdhvz1ofVP76po3q/vpwaqiQCPcGNTTj7BejljrINve5MBrd09J7IGlWdK9rZXPI39cu4Iyrlsok6XlxcBX8f9MUpzDIa5uCADoSSdIlORLWvxL6J7y/uGkYR78X6/LNFYDzeFK4gX17Glc=
Received: from VI1PR0401MB2541.eurprd04.prod.outlook.com (10.168.65.19) by
VI1PR0401MB2575.eurprd04.prod.outlook.com (10.168.65.140) with Microsoft SMTP
Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id
15.20.1730.16; Tue, 26 Mar 2019 12:18:05 +0000
Received: from VI1PR0401MB2541.eurprd04.prod.outlook.com
([fe80::18e3:39b6:c61d:3f18]) by VI1PR0401MB2541.eurprd04.prod.outlook.com
([fe80::18e3:39b6:c61d:3f18%12]) with mapi id 15.20.1730.019; Tue, 26 Mar
2019 12:18:05 +0000
From: Hemant Agrawal <hemant.agrawal@nxp.com>
To: "dev@dpdk.org" <dev@dpdk.org>
CC: Shreyansh Jain <shreyansh.jain@nxp.com>, Yi Liu <yi.liu@nxp.com>
Thread-Topic: [PATCH 3/5] raw/dpaa2_qdma: add burst mode support
Thread-Index: AQHU4833WQmLmfRiSkW+QYJ2NM5WoQ==
Date: Tue, 26 Mar 2019 12:18:05 +0000
Message-ID: <20190326121610.28024-3-hemant.agrawal@nxp.com>
References: <20190326121610.28024-1-hemant.agrawal@nxp.com>
In-Reply-To: <20190326121610.28024-1-hemant.agrawal@nxp.com>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach:
X-MS-TNEF-Correlator:
x-originating-ip: [92.120.1.72]
x-mailer: git-send-email 2.17.1
x-clientproxiedby: BM1PR0101CA0069.INDPRD01.PROD.OUTLOOK.COM
(2603:1096:b00:19::31) To VI1PR0401MB2541.eurprd04.prod.outlook.com
(2603:10a6:800:56::19)
authentication-results: spf=none (sender IP is )
smtp.mailfrom=hemant.agrawal@nxp.com;
x-ms-exchange-messagesentrepresentingtype: 1
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 0f621fc5-6111-4022-0702-08d6b1e5198e
x-ms-office365-filtering-ht: Tenant
x-microsoft-antispam: BCL:0; PCL:0;
RULEID:(2390118)(7020095)(4652040)(8989299)(5600127)(711020)(4605104)(4618075)(4534185)(4627221)(201703031133081)(201702281549075)(8990200)(2017052603328)(7153060)(7193020);
SRVR:VI1PR0401MB2575;
x-ms-traffictypediagnostic: VI1PR0401MB2575:
x-microsoft-antispam-prvs: <VI1PR0401MB2575B022BE7DEA0D9F9A7D3B895F0@VI1PR0401MB2575.eurprd04.prod.outlook.com>
x-forefront-prvs: 09888BC01D
x-forefront-antispam-report: SFV:NSPM;
SFS:(10009020)(396003)(366004)(136003)(346002)(39860400002)(376002)(199004)(189003)(186003)(36756003)(44832011)(53936002)(4326008)(76176011)(71190400001)(11346002)(71200400001)(476003)(6436002)(2616005)(5640700003)(486006)(2501003)(446003)(6486002)(14444005)(25786009)(14454004)(99286004)(256004)(30864003)(2351001)(2906002)(105586002)(86362001)(5660300002)(106356001)(81166006)(7736002)(1730700003)(1076003)(3846002)(97736004)(81156014)(52116002)(54906003)(66066001)(6116002)(68736007)(6506007)(386003)(478600001)(26005)(316002)(6916009)(6512007)(50226002)(8936002)(8676002)(102836004)(305945005);
DIR:OUT; SFP:1101; SCL:1; SRVR:VI1PR0401MB2575;
H:VI1PR0401MB2541.eurprd04.prod.outlook.com; FPR:; SPF:None; LANG:en;
PTR:InfoNoRecords; A:1; MX:1;
received-spf: None (protection.outlook.com: nxp.com does not designate
permitted sender hosts)
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam-message-info: mx1yikZQw2E3pqlQbPbhuRiSwD6KDfzP8MYsgKR7HfBovq3s27J/gr6SzitYjuDnkHnJS/ispHPeAv3z06zeakzC0C8cbDRytqyw3bXCsCTxZjQVqPKmsyJUUHwTUl0XiMflcLOdPMPJpkidwYF8R61GTffb0JN/AWFBz8vkKfpXq6dC2pRzqzkADCaqzK1hqtmyZpYGtMEkOMuwO1UdDZt/Q6iflCY2w5rroeBpjvkLjQk3TrQrQCensqfBMGuIrs//It2PUlTHigyMjjGg0oUSyQtgFUys/mnPHYqEn0/qivDcGLKnBUJXPAq7EI04RyRR1GCcedwxytF9TPciUny8Y769h34i5IZxIC7ZLCEkE5uDuLnxPFbUJgURlH5XEmPqVgVNrZvGVx8HeflpA8l4pnOKlf5CUnHE0jRwmU4=
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: nxp.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 0f621fc5-6111-4022-0702-08d6b1e5198e
X-MS-Exchange-CrossTenant-originalarrivaltime: 26 Mar 2019 12:18:05.7192 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 686ea1d3-bc2b-4c6f-a92c-d99c5c301635
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-Transport-CrossTenantHeadersStamped: VI1PR0401MB2575
Subject: [dpdk-dev] [PATCH 3/5] raw/dpaa2_qdma: add burst mode support
X-BeenThere: dev@dpdk.org
X-Mailman-Version: 2.1.15
Precedence: list
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <https://mails.dpdk.org/options/dev>,
<mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://mails.dpdk.org/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <https://mails.dpdk.org/listinfo/dev>,
<mailto:dev-request@dpdk.org?subject=subscribe>
X-List-Received-Date: Tue, 26 Mar 2019 12:18:12 -0000
This patch adds support the batch processing for the qdma jobs
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Signed-off-by: Yi Liu <yi.liu@nxp.com>
---
drivers/raw/dpaa2_qdma/dpaa2_qdma.c | 309 ++++++++++----------
drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h | 6 +-
2 files changed, 167 insertions(+), 148 deletions(-)
diff --git a/drivers/raw/dpaa2_qdma/dpaa2_qdma.c b/drivers/raw/dpaa2_qdma/d=
paa2_qdma.c
index a3f0f7fdd..679bf66e9 100644
--- a/drivers/raw/dpaa2_qdma/dpaa2_qdma.c
+++ b/drivers/raw/dpaa2_qdma/dpaa2_qdma.c
@@ -1,8 +1,10 @@
/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright 2018 NXP
+ * Copyright 2018-2019 NXP
*/
=20
#include <string.h>
+#include <sys/types.h>
+#include <unistd.h>
=20
#include <rte_eal.h>
#include <rte_fslmc.h>
@@ -13,6 +15,7 @@
#include <rte_malloc.h>
#include <rte_ring.h>
#include <rte_mempool.h>
+#include <rte_prefetch.h>
=20
#include <mc/fsl_dpdmai.h>
#include <portal/dpaa2_hw_pvt.h>
@@ -395,21 +398,31 @@ dpaa2_qdma_populate_fle(struct qbman_fle *fle,
DPAA2_SET_FLE_FIN(fle);
}
=20
-static int
-dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_dev,
- uint16_t txq_id,
- uint16_t vq_id,
- struct rte_qdma_job *job)
+int
+rte_qdma_vq_enqueue_multi(uint16_t vq_id,
+ struct rte_qdma_job **job,
+ uint16_t nb_jobs)
{
+ struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
+ struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
+ struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
struct qdma_io_meta *io_meta;
- struct qbman_fd fd;
+ struct qbman_fd fd_arr[MAX_TX_RING_SLOTS];
struct dpaa2_queue *txq;
struct qbman_fle *fle;
struct qbman_eq_desc eqdesc;
struct qbman_swp *swp;
int ret;
+ uint32_t num_to_send =3D 0;
+ uint16_t num_tx =3D 0;
+ uint16_t num_txed =3D 0;
=20
- DPAA2_QDMA_FUNC_TRACE();
+ /* Return error in case of wrong lcore_id */
+ if (rte_lcore_id() !=3D qdma_vq->lcore_id) {
+ DPAA2_QDMA_ERR("QDMA enqueue for vqid %d on wrong core",
+ vq_id);
+ return -1;
+ }
=20
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
ret =3D dpaa2_affine_qbman_swp();
@@ -420,7 +433,7 @@ dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_dev,
}
swp =3D DPAA2_PER_LCORE_PORTAL;
=20
- txq =3D &(dpdmai_dev->tx_queue[txq_id]);
+ txq =3D &(dpdmai_dev->tx_queue[qdma_pq->queue_id]);
=20
/* Prepare enqueue descriptor */
qbman_eq_desc_clear(&eqdesc);
@@ -428,97 +441,86 @@ dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_de=
v,
qbman_eq_desc_set_no_orp(&eqdesc, 0);
qbman_eq_desc_set_response(&eqdesc, 0, 0);
=20
- /*
- * Get an FLE/SDD from FLE pool.
- * Note: IO metadata is before the FLE and SDD memory.
- */
- ret =3D rte_mempool_get(qdma_dev.fle_pool, (void **)(&io_meta));
- if (ret) {
- DPAA2_QDMA_DP_WARN("Memory alloc failed for FLE");
- return ret;
- }
-
- /* Set the metadata */
- io_meta->cnxt =3D (size_t)job;
- io_meta->id =3D vq_id;
-
- fle =3D (struct qbman_fle *)(io_meta + 1);
-
- /* populate Frame descriptor */
- memset(&fd, 0, sizeof(struct qbman_fd));
- DPAA2_SET_FD_ADDR(&fd, DPAA2_VADDR_TO_IOVA(fle));
- DPAA2_SET_FD_COMPOUND_FMT(&fd);
- DPAA2_SET_FD_FRC(&fd, QDMA_SER_CTX);
+ while (nb_jobs > 0) {
+ uint32_t loop;
+
+ num_to_send =3D (nb_jobs > dpaa2_eqcr_size) ?
+ dpaa2_eqcr_size : nb_jobs;
+
+ for (loop =3D 0; loop < num_to_send; loop++) {
+ /*
+ * Get an FLE/SDD from FLE pool.
+ * Note: IO metadata is before the FLE and SDD memory.
+ */
+ ret =3D rte_mempool_get(qdma_dev.fle_pool,
+ (void **)(&io_meta));
+ if (ret) {
+ DPAA2_QDMA_DP_WARN("Me alloc failed for FLE");
+ return ret;
+ }
=20
- /* Populate FLE */
- memset(fle, 0, QDMA_FLE_POOL_SIZE);
- dpaa2_qdma_populate_fle(fle, job->src, job->dest, job->len, job->flags);
+ /* Set the metadata */
+ io_meta->cnxt =3D (size_t)job[num_tx];
+ io_meta->id =3D vq_id;
=20
- /* Enqueue the packet to the QBMAN */
- do {
- ret =3D qbman_swp_enqueue_multiple(swp, &eqdesc, &fd, NULL, 1);
- if (ret < 0 && ret !=3D -EBUSY)
- DPAA2_QDMA_ERR("Transmit failure with err: %d", ret);
- } while (ret =3D=3D -EBUSY);
+ fle =3D (struct qbman_fle *)(io_meta + 1);
=20
- DPAA2_QDMA_DP_DEBUG("Successfully transmitted a packet");
+ /* populate Frame descriptor */
+ memset(&fd_arr[loop], 0, sizeof(struct qbman_fd));
+ DPAA2_SET_FD_ADDR(&fd_arr[loop],
+ DPAA2_VADDR_TO_IOVA(fle));
+ DPAA2_SET_FD_COMPOUND_FMT(&fd_arr[loop]);
+ DPAA2_SET_FD_FRC(&fd_arr[loop], QDMA_SER_CTX);
=20
- return ret;
-}
+ /* Populate FLE */
+ memset(fle, 0, QDMA_FLE_POOL_SIZE);
+ dpaa2_qdma_populate_fle(fle, job[num_tx]->src,
+ job[num_tx]->dest,
+ job[num_tx]->len,
+ job[num_tx]->flags);
=20
-int __rte_experimental
-rte_qdma_vq_enqueue_multi(uint16_t vq_id,
- struct rte_qdma_job **job,
- uint16_t nb_jobs)
-{
- int i, ret;
+ num_tx++;
+ }
=20
- DPAA2_QDMA_FUNC_TRACE();
+ /* Enqueue the packet to the QBMAN */
+ uint32_t enqueue_loop =3D 0;
+ while (enqueue_loop < num_to_send) {
+ enqueue_loop +=3D qbman_swp_enqueue_multiple(swp,
+ &eqdesc,
+ &fd_arr[enqueue_loop],
+ NULL,
+ num_to_send - enqueue_loop);
+ }
=20
- for (i =3D 0; i < nb_jobs; i++) {
- ret =3D rte_qdma_vq_enqueue(vq_id, job[i]);
- if (ret < 0)
- break;
+ num_txed +=3D num_to_send;
+ nb_jobs -=3D num_to_send;
}
-
- return i;
+ qdma_vq->num_enqueues +=3D num_txed;
+ return num_txed;
}
=20
-int __rte_experimental
+int
rte_qdma_vq_enqueue(uint16_t vq_id,
struct rte_qdma_job *job)
{
- struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
- struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
- struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
int ret;
=20
- DPAA2_QDMA_FUNC_TRACE();
-
- /* Return error in case of wrong lcore_id */
- if (rte_lcore_id() !=3D qdma_vq->lcore_id) {
- DPAA2_QDMA_ERR("QDMA enqueue for vqid %d on wrong core",
- vq_id);
- return -EINVAL;
- }
-
- ret =3D dpdmai_dev_enqueue(dpdmai_dev, qdma_pq->queue_id, vq_id, job);
+ ret =3D rte_qdma_vq_enqueue_multi(vq_id, &job, 1);
if (ret < 0) {
DPAA2_QDMA_ERR("DPDMAI device enqueue failed: %d", ret);
return ret;
}
=20
- qdma_vq->num_enqueues++;
-
return 1;
}
=20
/* Function to receive a QDMA job for a given device and queue*/
static int
-dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
+dpdmai_dev_dequeue_multijob(struct dpaa2_dpdmai_dev *dpdmai_dev,
uint16_t rxq_id,
uint16_t *vq_id,
- struct rte_qdma_job **job)
+ struct rte_qdma_job **job,
+ uint16_t nb_jobs)
{
struct qdma_io_meta *io_meta;
struct dpaa2_queue *rxq;
@@ -531,8 +533,6 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
uint8_t status;
int ret;
=20
- DPAA2_QDMA_FUNC_TRACE();
-
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
ret =3D dpaa2_affine_qbman_swp();
if (ret) {
@@ -541,7 +541,6 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
}
}
swp =3D DPAA2_PER_LCORE_PORTAL;
-
rxq =3D &(dpdmai_dev->rx_queue[rxq_id]);
dq_storage =3D rxq->q_storage->dq_storage[0];
fqid =3D rxq->fqid;
@@ -551,7 +550,10 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev=
,
qbman_pull_desc_set_fq(&pulldesc, fqid);
qbman_pull_desc_set_storage(&pulldesc, dq_storage,
(uint64_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
- qbman_pull_desc_set_numframes(&pulldesc, 1);
+ if (nb_jobs > dpaa2_dqrr_size)
+ qbman_pull_desc_set_numframes(&pulldesc, dpaa2_dqrr_size);
+ else
+ qbman_pull_desc_set_numframes(&pulldesc, nb_jobs);
=20
while (1) {
if (qbman_swp_pull(swp, &pulldesc)) {
@@ -561,125 +563,138 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_=
dev,
break;
}
=20
- /* Check if previous issued command is completed. */
+ rte_prefetch0((void *)((size_t)(dq_storage + 1)));
+ /* Check if the previous issued command is completed. */
while (!qbman_check_command_complete(dq_storage))
;
- /* Loop until dq_storage is updated with new token by QBMAN */
- while (!qbman_check_new_result(dq_storage))
- ;
=20
- /* Check for valid frame. */
- status =3D qbman_result_DQ_flags(dq_storage);
- if (unlikely((status & QBMAN_DQ_STAT_VALIDFRAME) =3D=3D 0)) {
- DPAA2_QDMA_DP_DEBUG("No frame is delivered");
- return 0;
- }
+ int num_pulled =3D 0;
+ int pending =3D 1;
+ do {
+ /* Loop until the dq_storage is updated with
+ * new token by QBMAN
+ */
+ while (!qbman_check_new_result(dq_storage))
+ ;
=20
- /* Get the FD */
- fd =3D qbman_result_DQ_fd(dq_storage);
+ rte_prefetch0((void *)((size_t)(dq_storage + 2)));
+ /* Check whether Last Pull command is Expired and
+ * setting Condition for Loop termination
+ */
+ if (qbman_result_DQ_is_pull_complete(dq_storage)) {
+ pending =3D 0;
+ /* Check for valid frame. */
+ status =3D qbman_result_DQ_flags(dq_storage);
+ if (unlikely((status &
+ QBMAN_DQ_STAT_VALIDFRAME) =3D=3D 0))
+ continue;
+ }
+ fd =3D qbman_result_DQ_fd(dq_storage);
=20
- /*
- * Fetch metadata from FLE. job and vq_id were set
- * in metadata in the enqueue operation.
- */
- fle =3D (struct qbman_fle *)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd));
- io_meta =3D (struct qdma_io_meta *)(fle) - 1;
- if (vq_id)
- *vq_id =3D io_meta->id;
+ /*
+ * Fetch metadata from FLE. job and vq_id were set
+ * in metadata in the enqueue operation.
+ */
+ fle =3D (struct qbman_fle *)
+ DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd));
+ io_meta =3D (struct qdma_io_meta *)(fle) - 1;
+ if (vq_id)
+ vq_id[num_pulled] =3D io_meta->id;
=20
- *job =3D (struct rte_qdma_job *)(size_t)io_meta->cnxt;
- (*job)->status =3D DPAA2_GET_FD_ERR(fd);
+ job[num_pulled] =3D (struct rte_qdma_job *)(size_t)io_meta->cnxt;
+ job[num_pulled]->status =3D DPAA2_GET_FD_ERR(fd);
=20
- /* Free FLE to the pool */
- rte_mempool_put(qdma_dev.fle_pool, io_meta);
+ /* Free FLE to the pool */
+ rte_mempool_put(qdma_dev.fle_pool, io_meta);
=20
- DPAA2_QDMA_DP_DEBUG("packet received");
+ dq_storage++;
+ num_pulled++;
+ } while (pending && (num_pulled <=3D dpaa2_dqrr_size));
=20
- return 1;
+ return num_pulled;
}
=20
-int __rte_experimental
+int
rte_qdma_vq_dequeue_multi(uint16_t vq_id,
struct rte_qdma_job **job,
uint16_t nb_jobs)
-{
- int i;
-
- DPAA2_QDMA_FUNC_TRACE();
-
- for (i =3D 0; i < nb_jobs; i++) {
- job[i] =3D rte_qdma_vq_dequeue(vq_id);
- if (!job[i])
- break;
- }
-
- return i;
-}
-
-struct rte_qdma_job * __rte_experimental
-rte_qdma_vq_dequeue(uint16_t vq_id)
{
struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
- struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
- struct rte_qdma_job *job =3D NULL;
struct qdma_virt_queue *temp_qdma_vq;
- int dequeue_budget =3D QDMA_DEQUEUE_BUDGET;
- int ring_count, ret, i;
- uint16_t temp_vq_id;
-
- DPAA2_QDMA_FUNC_TRACE();
+ struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
+ int ring_count, ret =3D 0, i;
=20
/* Return error in case of wrong lcore_id */
if (rte_lcore_id() !=3D (unsigned int)(qdma_vq->lcore_id)) {
- DPAA2_QDMA_ERR("QDMA dequeue for vqid %d on wrong core",
+ DPAA2_QDMA_WARN("QDMA dequeue for vqid %d on wrong core",
vq_id);
- return NULL;
+ return -1;
}
=20
/* Only dequeue when there are pending jobs on VQ */
if (qdma_vq->num_enqueues =3D=3D qdma_vq->num_dequeues)
- return NULL;
+ return 0;
+
+ if (qdma_vq->num_enqueues < (qdma_vq->num_dequeues + nb_jobs))
+ nb_jobs =3D (qdma_vq->num_enqueues - qdma_vq->num_dequeues);
=20
if (qdma_vq->exclusive_hw_queue) {
/* In case of exclusive queue directly fetch from HW queue */
- ret =3D dpdmai_dev_dequeue(dpdmai_dev, qdma_pq->queue_id,
- NULL, &job);
+ ret =3D dpdmai_dev_dequeue_multijob(dpdmai_dev, qdma_pq->queue_id,
+ NULL, job, nb_jobs);
if (ret < 0) {
DPAA2_QDMA_ERR(
"Dequeue from DPDMAI device failed: %d", ret);
- return NULL;
+ return ret;
}
+ qdma_vq->num_dequeues +=3D ret;
} else {
+ uint16_t temp_vq_id[RTE_QDMA_BURST_NB_MAX];
/*
* Get the QDMA completed jobs from the software ring.
* In case they are not available on the ring poke the HW
* to fetch completed jobs from corresponding HW queues
*/
ring_count =3D rte_ring_count(qdma_vq->status_ring);
- if (ring_count =3D=3D 0) {
+ if (ring_count < nb_jobs) {
/* TODO - How to have right budget */
- for (i =3D 0; i < dequeue_budget; i++) {
- ret =3D dpdmai_dev_dequeue(dpdmai_dev,
- qdma_pq->queue_id, &temp_vq_id, &job);
- if (ret =3D=3D 0)
- break;
- temp_qdma_vq =3D &qdma_vqs[temp_vq_id];
+ ret =3D dpdmai_dev_dequeue_multijob(dpdmai_dev,
+ qdma_pq->queue_id,
+ temp_vq_id, job, nb_jobs);
+ for (i =3D 0; i < ret; i++) {
+ temp_qdma_vq =3D &qdma_vqs[temp_vq_id[i]];
rte_ring_enqueue(temp_qdma_vq->status_ring,
- (void *)(job));
+ (void *)(job[i]));
ring_count =3D rte_ring_count(
qdma_vq->status_ring);
- if (ring_count)
- break;
}
}
=20
- /* Dequeue job from the software ring to provide to the user */
- rte_ring_dequeue(qdma_vq->status_ring, (void **)&job);
- if (job)
- qdma_vq->num_dequeues++;
+ if (ring_count) {
+ /* Dequeue job from the software ring
+ * to provide to the user
+ */
+ ret =3D rte_ring_dequeue_bulk(qdma_vq->status_ring,
+ (void **)job, ring_count, NULL);
+ if (ret)
+ qdma_vq->num_dequeues +=3D ret;
+ }
}
=20
+ return ret;
+}
+
+struct rte_qdma_job *
+rte_qdma_vq_dequeue(uint16_t vq_id)
+{
+ int ret;
+ struct rte_qdma_job *job =3D NULL;
+
+ ret =3D rte_qdma_vq_dequeue_multi(vq_id, &job, 1);
+ if (ret < 0)
+ DPAA2_QDMA_DP_WARN("DPDMAI device dequeue failed: %d", ret);
+
return job;
}
=20
diff --git a/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h b/drivers/raw/dpaa=
2_qdma/rte_pmd_dpaa2_qdma.h
index c9697b4d7..e1ccc19e8 100644
--- a/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h
+++ b/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h
@@ -12,6 +12,9 @@
*
*/
=20
+/** Maximum qdma burst size */
+#define RTE_QDMA_BURST_NB_MAX 32
+
/** Determines the mode of operation */
enum {
/**
@@ -225,7 +228,8 @@ rte_qdma_vq_enqueue(uint16_t vq_id,
* Number of QDMA jobs requested for dequeue by the user.
*
* @returns
- * Number of jobs actually dequeued.
+ * - >=3D0: Number of jobs successfully received
+ * - <0: Error code.
*/
int
rte_qdma_vq_dequeue_multi(uint16_t vq_id,
--=20
2.17.1
From mboxrd@z Thu Jan 1 00:00:00 1970
Return-Path: <dev-bounces@dpdk.org>
Received: from dpdk.org (dpdk.org [92.243.14.124])
by dpdk.space (Postfix) with ESMTP id 8D004A05D3
for <public@inbox.dpdk.org>; Tue, 26 Mar 2019 13:18:26 +0100 (CET)
Received: from [92.243.14.124] (localhost [127.0.0.1])
by dpdk.org (Postfix) with ESMTP id D8FA14C9C;
Tue, 26 Mar 2019 13:18:13 +0100 (CET)
Received: from EUR04-VI1-obe.outbound.protection.outlook.com
(mail-eopbgr80052.outbound.protection.outlook.com [40.107.8.52])
by dpdk.org (Postfix) with ESMTP id 137BA4CAB
for <dev@dpdk.org>; Tue, 26 Mar 2019 13:18:12 +0100 (CET)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=nxp.com; s=selector1;
h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck;
bh=SKlP5dflNKjzmJPAWJ/guIpnSdMPewtpGeObvUam2s0=;
b=u8Wtm96vG6E8cQcly8YpjnR9YxApxdhvz1ofVP76po3q/vpwaqiQCPcGNTTj7BejljrINve5MBrd09J7IGlWdK9rZXPI39cu4Iyrlsok6XlxcBX8f9MUpzDIa5uCADoSSdIlORLWvxL6J7y/uGkYR78X6/LNFYDzeFK4gX17Glc=
Received: from VI1PR0401MB2541.eurprd04.prod.outlook.com (10.168.65.19) by
VI1PR0401MB2575.eurprd04.prod.outlook.com (10.168.65.140) with Microsoft SMTP
Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id
15.20.1730.16; Tue, 26 Mar 2019 12:18:05 +0000
Received: from VI1PR0401MB2541.eurprd04.prod.outlook.com
([fe80::18e3:39b6:c61d:3f18]) by VI1PR0401MB2541.eurprd04.prod.outlook.com
([fe80::18e3:39b6:c61d:3f18%12]) with mapi id 15.20.1730.019; Tue, 26 Mar
2019 12:18:05 +0000
From: Hemant Agrawal <hemant.agrawal@nxp.com>
To: "dev@dpdk.org" <dev@dpdk.org>
CC: Shreyansh Jain <shreyansh.jain@nxp.com>, Yi Liu <yi.liu@nxp.com>
Thread-Topic: [PATCH 3/5] raw/dpaa2_qdma: add burst mode support
Thread-Index: AQHU4833WQmLmfRiSkW+QYJ2NM5WoQ==
Date: Tue, 26 Mar 2019 12:18:05 +0000
Message-ID: <20190326121610.28024-3-hemant.agrawal@nxp.com>
References: <20190326121610.28024-1-hemant.agrawal@nxp.com>
In-Reply-To: <20190326121610.28024-1-hemant.agrawal@nxp.com>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach:
X-MS-TNEF-Correlator:
x-originating-ip: [92.120.1.72]
x-mailer: git-send-email 2.17.1
x-clientproxiedby: BM1PR0101CA0069.INDPRD01.PROD.OUTLOOK.COM
(2603:1096:b00:19::31) To VI1PR0401MB2541.eurprd04.prod.outlook.com
(2603:10a6:800:56::19)
authentication-results: spf=none (sender IP is )
smtp.mailfrom=hemant.agrawal@nxp.com;
x-ms-exchange-messagesentrepresentingtype: 1
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 0f621fc5-6111-4022-0702-08d6b1e5198e
x-ms-office365-filtering-ht: Tenant
x-microsoft-antispam: BCL:0; PCL:0;
RULEID:(2390118)(7020095)(4652040)(8989299)(5600127)(711020)(4605104)(4618075)(4534185)(4627221)(201703031133081)(201702281549075)(8990200)(2017052603328)(7153060)(7193020);
SRVR:VI1PR0401MB2575;
x-ms-traffictypediagnostic: VI1PR0401MB2575:
x-microsoft-antispam-prvs: <VI1PR0401MB2575B022BE7DEA0D9F9A7D3B895F0@VI1PR0401MB2575.eurprd04.prod.outlook.com>
x-forefront-prvs: 09888BC01D
x-forefront-antispam-report: SFV:NSPM;
SFS:(10009020)(396003)(366004)(136003)(346002)(39860400002)(376002)(199004)(189003)(186003)(36756003)(44832011)(53936002)(4326008)(76176011)(71190400001)(11346002)(71200400001)(476003)(6436002)(2616005)(5640700003)(486006)(2501003)(446003)(6486002)(14444005)(25786009)(14454004)(99286004)(256004)(30864003)(2351001)(2906002)(105586002)(86362001)(5660300002)(106356001)(81166006)(7736002)(1730700003)(1076003)(3846002)(97736004)(81156014)(52116002)(54906003)(66066001)(6116002)(68736007)(6506007)(386003)(478600001)(26005)(316002)(6916009)(6512007)(50226002)(8936002)(8676002)(102836004)(305945005);
DIR:OUT; SFP:1101; SCL:1; SRVR:VI1PR0401MB2575;
H:VI1PR0401MB2541.eurprd04.prod.outlook.com; FPR:; SPF:None; LANG:en;
PTR:InfoNoRecords; A:1; MX:1;
received-spf: None (protection.outlook.com: nxp.com does not designate
permitted sender hosts)
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam-message-info: mx1yikZQw2E3pqlQbPbhuRiSwD6KDfzP8MYsgKR7HfBovq3s27J/gr6SzitYjuDnkHnJS/ispHPeAv3z06zeakzC0C8cbDRytqyw3bXCsCTxZjQVqPKmsyJUUHwTUl0XiMflcLOdPMPJpkidwYF8R61GTffb0JN/AWFBz8vkKfpXq6dC2pRzqzkADCaqzK1hqtmyZpYGtMEkOMuwO1UdDZt/Q6iflCY2w5rroeBpjvkLjQk3TrQrQCensqfBMGuIrs//It2PUlTHigyMjjGg0oUSyQtgFUys/mnPHYqEn0/qivDcGLKnBUJXPAq7EI04RyRR1GCcedwxytF9TPciUny8Y769h34i5IZxIC7ZLCEkE5uDuLnxPFbUJgURlH5XEmPqVgVNrZvGVx8HeflpA8l4pnOKlf5CUnHE0jRwmU4=
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-OriginatorOrg: nxp.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 0f621fc5-6111-4022-0702-08d6b1e5198e
X-MS-Exchange-CrossTenant-originalarrivaltime: 26 Mar 2019 12:18:05.7192 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 686ea1d3-bc2b-4c6f-a92c-d99c5c301635
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-Transport-CrossTenantHeadersStamped: VI1PR0401MB2575
Subject: [dpdk-dev] [PATCH 3/5] raw/dpaa2_qdma: add burst mode support
X-BeenThere: dev@dpdk.org
X-Mailman-Version: 2.1.15
Precedence: list
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <https://mails.dpdk.org/options/dev>,
<mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://mails.dpdk.org/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <https://mails.dpdk.org/listinfo/dev>,
<mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>
Message-ID: <20190326121805.IZ7-bEfRIK_PTY-Y0Nd56l7ysWDTUwfBSU98wb0xgeY@z>
This patch adds support the batch processing for the qdma jobs
Signed-off-by: Hemant Agrawal <hemant.agrawal@nxp.com>
Signed-off-by: Yi Liu <yi.liu@nxp.com>
---
drivers/raw/dpaa2_qdma/dpaa2_qdma.c | 309 ++++++++++----------
drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h | 6 +-
2 files changed, 167 insertions(+), 148 deletions(-)
diff --git a/drivers/raw/dpaa2_qdma/dpaa2_qdma.c b/drivers/raw/dpaa2_qdma/d=
paa2_qdma.c
index a3f0f7fdd..679bf66e9 100644
--- a/drivers/raw/dpaa2_qdma/dpaa2_qdma.c
+++ b/drivers/raw/dpaa2_qdma/dpaa2_qdma.c
@@ -1,8 +1,10 @@
/* SPDX-License-Identifier: BSD-3-Clause
- * Copyright 2018 NXP
+ * Copyright 2018-2019 NXP
*/
=20
#include <string.h>
+#include <sys/types.h>
+#include <unistd.h>
=20
#include <rte_eal.h>
#include <rte_fslmc.h>
@@ -13,6 +15,7 @@
#include <rte_malloc.h>
#include <rte_ring.h>
#include <rte_mempool.h>
+#include <rte_prefetch.h>
=20
#include <mc/fsl_dpdmai.h>
#include <portal/dpaa2_hw_pvt.h>
@@ -395,21 +398,31 @@ dpaa2_qdma_populate_fle(struct qbman_fle *fle,
DPAA2_SET_FLE_FIN(fle);
}
=20
-static int
-dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_dev,
- uint16_t txq_id,
- uint16_t vq_id,
- struct rte_qdma_job *job)
+int
+rte_qdma_vq_enqueue_multi(uint16_t vq_id,
+ struct rte_qdma_job **job,
+ uint16_t nb_jobs)
{
+ struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
+ struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
+ struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
struct qdma_io_meta *io_meta;
- struct qbman_fd fd;
+ struct qbman_fd fd_arr[MAX_TX_RING_SLOTS];
struct dpaa2_queue *txq;
struct qbman_fle *fle;
struct qbman_eq_desc eqdesc;
struct qbman_swp *swp;
int ret;
+ uint32_t num_to_send =3D 0;
+ uint16_t num_tx =3D 0;
+ uint16_t num_txed =3D 0;
=20
- DPAA2_QDMA_FUNC_TRACE();
+ /* Return error in case of wrong lcore_id */
+ if (rte_lcore_id() !=3D qdma_vq->lcore_id) {
+ DPAA2_QDMA_ERR("QDMA enqueue for vqid %d on wrong core",
+ vq_id);
+ return -1;
+ }
=20
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
ret =3D dpaa2_affine_qbman_swp();
@@ -420,7 +433,7 @@ dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_dev,
}
swp =3D DPAA2_PER_LCORE_PORTAL;
=20
- txq =3D &(dpdmai_dev->tx_queue[txq_id]);
+ txq =3D &(dpdmai_dev->tx_queue[qdma_pq->queue_id]);
=20
/* Prepare enqueue descriptor */
qbman_eq_desc_clear(&eqdesc);
@@ -428,97 +441,86 @@ dpdmai_dev_enqueue(struct dpaa2_dpdmai_dev *dpdmai_de=
v,
qbman_eq_desc_set_no_orp(&eqdesc, 0);
qbman_eq_desc_set_response(&eqdesc, 0, 0);
=20
- /*
- * Get an FLE/SDD from FLE pool.
- * Note: IO metadata is before the FLE and SDD memory.
- */
- ret =3D rte_mempool_get(qdma_dev.fle_pool, (void **)(&io_meta));
- if (ret) {
- DPAA2_QDMA_DP_WARN("Memory alloc failed for FLE");
- return ret;
- }
-
- /* Set the metadata */
- io_meta->cnxt =3D (size_t)job;
- io_meta->id =3D vq_id;
-
- fle =3D (struct qbman_fle *)(io_meta + 1);
-
- /* populate Frame descriptor */
- memset(&fd, 0, sizeof(struct qbman_fd));
- DPAA2_SET_FD_ADDR(&fd, DPAA2_VADDR_TO_IOVA(fle));
- DPAA2_SET_FD_COMPOUND_FMT(&fd);
- DPAA2_SET_FD_FRC(&fd, QDMA_SER_CTX);
+ while (nb_jobs > 0) {
+ uint32_t loop;
+
+ num_to_send =3D (nb_jobs > dpaa2_eqcr_size) ?
+ dpaa2_eqcr_size : nb_jobs;
+
+ for (loop =3D 0; loop < num_to_send; loop++) {
+ /*
+ * Get an FLE/SDD from FLE pool.
+ * Note: IO metadata is before the FLE and SDD memory.
+ */
+ ret =3D rte_mempool_get(qdma_dev.fle_pool,
+ (void **)(&io_meta));
+ if (ret) {
+ DPAA2_QDMA_DP_WARN("Me alloc failed for FLE");
+ return ret;
+ }
=20
- /* Populate FLE */
- memset(fle, 0, QDMA_FLE_POOL_SIZE);
- dpaa2_qdma_populate_fle(fle, job->src, job->dest, job->len, job->flags);
+ /* Set the metadata */
+ io_meta->cnxt =3D (size_t)job[num_tx];
+ io_meta->id =3D vq_id;
=20
- /* Enqueue the packet to the QBMAN */
- do {
- ret =3D qbman_swp_enqueue_multiple(swp, &eqdesc, &fd, NULL, 1);
- if (ret < 0 && ret !=3D -EBUSY)
- DPAA2_QDMA_ERR("Transmit failure with err: %d", ret);
- } while (ret =3D=3D -EBUSY);
+ fle =3D (struct qbman_fle *)(io_meta + 1);
=20
- DPAA2_QDMA_DP_DEBUG("Successfully transmitted a packet");
+ /* populate Frame descriptor */
+ memset(&fd_arr[loop], 0, sizeof(struct qbman_fd));
+ DPAA2_SET_FD_ADDR(&fd_arr[loop],
+ DPAA2_VADDR_TO_IOVA(fle));
+ DPAA2_SET_FD_COMPOUND_FMT(&fd_arr[loop]);
+ DPAA2_SET_FD_FRC(&fd_arr[loop], QDMA_SER_CTX);
=20
- return ret;
-}
+ /* Populate FLE */
+ memset(fle, 0, QDMA_FLE_POOL_SIZE);
+ dpaa2_qdma_populate_fle(fle, job[num_tx]->src,
+ job[num_tx]->dest,
+ job[num_tx]->len,
+ job[num_tx]->flags);
=20
-int __rte_experimental
-rte_qdma_vq_enqueue_multi(uint16_t vq_id,
- struct rte_qdma_job **job,
- uint16_t nb_jobs)
-{
- int i, ret;
+ num_tx++;
+ }
=20
- DPAA2_QDMA_FUNC_TRACE();
+ /* Enqueue the packet to the QBMAN */
+ uint32_t enqueue_loop =3D 0;
+ while (enqueue_loop < num_to_send) {
+ enqueue_loop +=3D qbman_swp_enqueue_multiple(swp,
+ &eqdesc,
+ &fd_arr[enqueue_loop],
+ NULL,
+ num_to_send - enqueue_loop);
+ }
=20
- for (i =3D 0; i < nb_jobs; i++) {
- ret =3D rte_qdma_vq_enqueue(vq_id, job[i]);
- if (ret < 0)
- break;
+ num_txed +=3D num_to_send;
+ nb_jobs -=3D num_to_send;
}
-
- return i;
+ qdma_vq->num_enqueues +=3D num_txed;
+ return num_txed;
}
=20
-int __rte_experimental
+int
rte_qdma_vq_enqueue(uint16_t vq_id,
struct rte_qdma_job *job)
{
- struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
- struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
- struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
int ret;
=20
- DPAA2_QDMA_FUNC_TRACE();
-
- /* Return error in case of wrong lcore_id */
- if (rte_lcore_id() !=3D qdma_vq->lcore_id) {
- DPAA2_QDMA_ERR("QDMA enqueue for vqid %d on wrong core",
- vq_id);
- return -EINVAL;
- }
-
- ret =3D dpdmai_dev_enqueue(dpdmai_dev, qdma_pq->queue_id, vq_id, job);
+ ret =3D rte_qdma_vq_enqueue_multi(vq_id, &job, 1);
if (ret < 0) {
DPAA2_QDMA_ERR("DPDMAI device enqueue failed: %d", ret);
return ret;
}
=20
- qdma_vq->num_enqueues++;
-
return 1;
}
=20
/* Function to receive a QDMA job for a given device and queue*/
static int
-dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
+dpdmai_dev_dequeue_multijob(struct dpaa2_dpdmai_dev *dpdmai_dev,
uint16_t rxq_id,
uint16_t *vq_id,
- struct rte_qdma_job **job)
+ struct rte_qdma_job **job,
+ uint16_t nb_jobs)
{
struct qdma_io_meta *io_meta;
struct dpaa2_queue *rxq;
@@ -531,8 +533,6 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
uint8_t status;
int ret;
=20
- DPAA2_QDMA_FUNC_TRACE();
-
if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
ret =3D dpaa2_affine_qbman_swp();
if (ret) {
@@ -541,7 +541,6 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev,
}
}
swp =3D DPAA2_PER_LCORE_PORTAL;
-
rxq =3D &(dpdmai_dev->rx_queue[rxq_id]);
dq_storage =3D rxq->q_storage->dq_storage[0];
fqid =3D rxq->fqid;
@@ -551,7 +550,10 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_dev=
,
qbman_pull_desc_set_fq(&pulldesc, fqid);
qbman_pull_desc_set_storage(&pulldesc, dq_storage,
(uint64_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
- qbman_pull_desc_set_numframes(&pulldesc, 1);
+ if (nb_jobs > dpaa2_dqrr_size)
+ qbman_pull_desc_set_numframes(&pulldesc, dpaa2_dqrr_size);
+ else
+ qbman_pull_desc_set_numframes(&pulldesc, nb_jobs);
=20
while (1) {
if (qbman_swp_pull(swp, &pulldesc)) {
@@ -561,125 +563,138 @@ dpdmai_dev_dequeue(struct dpaa2_dpdmai_dev *dpdmai_=
dev,
break;
}
=20
- /* Check if previous issued command is completed. */
+ rte_prefetch0((void *)((size_t)(dq_storage + 1)));
+ /* Check if the previous issued command is completed. */
while (!qbman_check_command_complete(dq_storage))
;
- /* Loop until dq_storage is updated with new token by QBMAN */
- while (!qbman_check_new_result(dq_storage))
- ;
=20
- /* Check for valid frame. */
- status =3D qbman_result_DQ_flags(dq_storage);
- if (unlikely((status & QBMAN_DQ_STAT_VALIDFRAME) =3D=3D 0)) {
- DPAA2_QDMA_DP_DEBUG("No frame is delivered");
- return 0;
- }
+ int num_pulled =3D 0;
+ int pending =3D 1;
+ do {
+ /* Loop until the dq_storage is updated with
+ * new token by QBMAN
+ */
+ while (!qbman_check_new_result(dq_storage))
+ ;
=20
- /* Get the FD */
- fd =3D qbman_result_DQ_fd(dq_storage);
+ rte_prefetch0((void *)((size_t)(dq_storage + 2)));
+ /* Check whether Last Pull command is Expired and
+ * setting Condition for Loop termination
+ */
+ if (qbman_result_DQ_is_pull_complete(dq_storage)) {
+ pending =3D 0;
+ /* Check for valid frame. */
+ status =3D qbman_result_DQ_flags(dq_storage);
+ if (unlikely((status &
+ QBMAN_DQ_STAT_VALIDFRAME) =3D=3D 0))
+ continue;
+ }
+ fd =3D qbman_result_DQ_fd(dq_storage);
=20
- /*
- * Fetch metadata from FLE. job and vq_id were set
- * in metadata in the enqueue operation.
- */
- fle =3D (struct qbman_fle *)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd));
- io_meta =3D (struct qdma_io_meta *)(fle) - 1;
- if (vq_id)
- *vq_id =3D io_meta->id;
+ /*
+ * Fetch metadata from FLE. job and vq_id were set
+ * in metadata in the enqueue operation.
+ */
+ fle =3D (struct qbman_fle *)
+ DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd));
+ io_meta =3D (struct qdma_io_meta *)(fle) - 1;
+ if (vq_id)
+ vq_id[num_pulled] =3D io_meta->id;
=20
- *job =3D (struct rte_qdma_job *)(size_t)io_meta->cnxt;
- (*job)->status =3D DPAA2_GET_FD_ERR(fd);
+ job[num_pulled] =3D (struct rte_qdma_job *)(size_t)io_meta->cnxt;
+ job[num_pulled]->status =3D DPAA2_GET_FD_ERR(fd);
=20
- /* Free FLE to the pool */
- rte_mempool_put(qdma_dev.fle_pool, io_meta);
+ /* Free FLE to the pool */
+ rte_mempool_put(qdma_dev.fle_pool, io_meta);
=20
- DPAA2_QDMA_DP_DEBUG("packet received");
+ dq_storage++;
+ num_pulled++;
+ } while (pending && (num_pulled <=3D dpaa2_dqrr_size));
=20
- return 1;
+ return num_pulled;
}
=20
-int __rte_experimental
+int
rte_qdma_vq_dequeue_multi(uint16_t vq_id,
struct rte_qdma_job **job,
uint16_t nb_jobs)
-{
- int i;
-
- DPAA2_QDMA_FUNC_TRACE();
-
- for (i =3D 0; i < nb_jobs; i++) {
- job[i] =3D rte_qdma_vq_dequeue(vq_id);
- if (!job[i])
- break;
- }
-
- return i;
-}
-
-struct rte_qdma_job * __rte_experimental
-rte_qdma_vq_dequeue(uint16_t vq_id)
{
struct qdma_virt_queue *qdma_vq =3D &qdma_vqs[vq_id];
struct qdma_hw_queue *qdma_pq =3D qdma_vq->hw_queue;
- struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
- struct rte_qdma_job *job =3D NULL;
struct qdma_virt_queue *temp_qdma_vq;
- int dequeue_budget =3D QDMA_DEQUEUE_BUDGET;
- int ring_count, ret, i;
- uint16_t temp_vq_id;
-
- DPAA2_QDMA_FUNC_TRACE();
+ struct dpaa2_dpdmai_dev *dpdmai_dev =3D qdma_pq->dpdmai_dev;
+ int ring_count, ret =3D 0, i;
=20
/* Return error in case of wrong lcore_id */
if (rte_lcore_id() !=3D (unsigned int)(qdma_vq->lcore_id)) {
- DPAA2_QDMA_ERR("QDMA dequeue for vqid %d on wrong core",
+ DPAA2_QDMA_WARN("QDMA dequeue for vqid %d on wrong core",
vq_id);
- return NULL;
+ return -1;
}
=20
/* Only dequeue when there are pending jobs on VQ */
if (qdma_vq->num_enqueues =3D=3D qdma_vq->num_dequeues)
- return NULL;
+ return 0;
+
+ if (qdma_vq->num_enqueues < (qdma_vq->num_dequeues + nb_jobs))
+ nb_jobs =3D (qdma_vq->num_enqueues - qdma_vq->num_dequeues);
=20
if (qdma_vq->exclusive_hw_queue) {
/* In case of exclusive queue directly fetch from HW queue */
- ret =3D dpdmai_dev_dequeue(dpdmai_dev, qdma_pq->queue_id,
- NULL, &job);
+ ret =3D dpdmai_dev_dequeue_multijob(dpdmai_dev, qdma_pq->queue_id,
+ NULL, job, nb_jobs);
if (ret < 0) {
DPAA2_QDMA_ERR(
"Dequeue from DPDMAI device failed: %d", ret);
- return NULL;
+ return ret;
}
+ qdma_vq->num_dequeues +=3D ret;
} else {
+ uint16_t temp_vq_id[RTE_QDMA_BURST_NB_MAX];
/*
* Get the QDMA completed jobs from the software ring.
* In case they are not available on the ring poke the HW
* to fetch completed jobs from corresponding HW queues
*/
ring_count =3D rte_ring_count(qdma_vq->status_ring);
- if (ring_count =3D=3D 0) {
+ if (ring_count < nb_jobs) {
/* TODO - How to have right budget */
- for (i =3D 0; i < dequeue_budget; i++) {
- ret =3D dpdmai_dev_dequeue(dpdmai_dev,
- qdma_pq->queue_id, &temp_vq_id, &job);
- if (ret =3D=3D 0)
- break;
- temp_qdma_vq =3D &qdma_vqs[temp_vq_id];
+ ret =3D dpdmai_dev_dequeue_multijob(dpdmai_dev,
+ qdma_pq->queue_id,
+ temp_vq_id, job, nb_jobs);
+ for (i =3D 0; i < ret; i++) {
+ temp_qdma_vq =3D &qdma_vqs[temp_vq_id[i]];
rte_ring_enqueue(temp_qdma_vq->status_ring,
- (void *)(job));
+ (void *)(job[i]));
ring_count =3D rte_ring_count(
qdma_vq->status_ring);
- if (ring_count)
- break;
}
}
=20
- /* Dequeue job from the software ring to provide to the user */
- rte_ring_dequeue(qdma_vq->status_ring, (void **)&job);
- if (job)
- qdma_vq->num_dequeues++;
+ if (ring_count) {
+ /* Dequeue job from the software ring
+ * to provide to the user
+ */
+ ret =3D rte_ring_dequeue_bulk(qdma_vq->status_ring,
+ (void **)job, ring_count, NULL);
+ if (ret)
+ qdma_vq->num_dequeues +=3D ret;
+ }
}
=20
+ return ret;
+}
+
+struct rte_qdma_job *
+rte_qdma_vq_dequeue(uint16_t vq_id)
+{
+ int ret;
+ struct rte_qdma_job *job =3D NULL;
+
+ ret =3D rte_qdma_vq_dequeue_multi(vq_id, &job, 1);
+ if (ret < 0)
+ DPAA2_QDMA_DP_WARN("DPDMAI device dequeue failed: %d", ret);
+
return job;
}
=20
diff --git a/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h b/drivers/raw/dpaa=
2_qdma/rte_pmd_dpaa2_qdma.h
index c9697b4d7..e1ccc19e8 100644
--- a/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h
+++ b/drivers/raw/dpaa2_qdma/rte_pmd_dpaa2_qdma.h
@@ -12,6 +12,9 @@
*
*/
=20
+/** Maximum qdma burst size */
+#define RTE_QDMA_BURST_NB_MAX 32
+
/** Determines the mode of operation */
enum {
/**
@@ -225,7 +228,8 @@ rte_qdma_vq_enqueue(uint16_t vq_id,
* Number of QDMA jobs requested for dequeue by the user.
*
* @returns
- * Number of jobs actually dequeued.
+ * - >=3D0: Number of jobs successfully received
+ * - <0: Error code.
*/
int
rte_qdma_vq_dequeue_multi(uint16_t vq_id,
--=20
2.17.1