From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 7FA96A00C2; Mon, 26 Sep 2022 08:55:38 +0200 (CEST) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 1C2B4400D7; Mon, 26 Sep 2022 08:55:38 +0200 (CEST) Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by mails.dpdk.org (Postfix) with ESMTP id B4D91400D5 for ; Mon, 26 Sep 2022 08:55:35 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1664175336; x=1695711336; h=from:to:cc:subject:date:message-id:references: in-reply-to:content-transfer-encoding:mime-version; bh=ihcjjbAPvVSYT0EODi2Re9lZ+nbJ9oKEfvz/jHSvMpM=; b=mHhiTtEiYCkA5Qh42kKhiYGdmi9j7KRCp426sm7j/O6B5Cdr1dYpZS85 k1rITGkQjZsvRsHX/W16iCirCLUnJEev2LWceDJI9Jvxwyrsd2bwFDCaZ 0tAptectO2749RNd+8lzThxSb1sQMHIvzOIk5m6LFC/ywNTrdm2TCLAqa hegT9jfxvvd9VDWth5sGPgrvkpdReG793CgJw9WHl+Hb4Gtb+Br28gCss JZcFBhpC1C1KLIUna+7GksuV9l4HLdNz3fzK5i1g+Uay0tTKCs1bGGMXP X8WCYzRWSwoC39iKIJzeUEktnDzRJUx8qS9NxfntC8O55/t7WcvLjl8pN g==; X-IronPort-AV: E=McAfee;i="6500,9779,10481"; a="299678297" X-IronPort-AV: E=Sophos;i="5.93,345,1654585200"; d="scan'208";a="299678297" Received: from orsmga001.jf.intel.com ([10.7.209.18]) by fmsmga104.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 25 Sep 2022 23:55:23 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6500,9779,10481"; a="654153699" X-IronPort-AV: E=Sophos;i="5.93,345,1654585200"; d="scan'208";a="654153699" Received: from fmsmsx601.amr.corp.intel.com ([10.18.126.81]) by orsmga001.jf.intel.com with ESMTP; 25 Sep 2022 23:55:22 -0700 Received: from fmsmsx611.amr.corp.intel.com (10.18.126.91) by fmsmsx601.amr.corp.intel.com (10.18.126.81) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31; Sun, 25 Sep 2022 23:55:22 -0700 Received: from fmsmsx608.amr.corp.intel.com (10.18.126.88) by fmsmsx611.amr.corp.intel.com (10.18.126.91) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31; Sun, 25 Sep 2022 23:55:22 -0700 Received: from fmsedg601.ED.cps.intel.com (10.1.192.135) by fmsmsx608.amr.corp.intel.com (10.18.126.88) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2375.31 via Frontend Transport; Sun, 25 Sep 2022 23:55:22 -0700 Received: from NAM11-DM6-obe.outbound.protection.outlook.com (104.47.57.174) by edgegateway.intel.com (192.55.55.70) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.1.2375.31; Sun, 25 Sep 2022 23:55:21 -0700 ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=XYM/IfDkqHIIf8jZddIliVhHy4xUCrRTamlYlvP7QbV/Tk0fiiJgB5LFE0aNVkPclPh1EHou9Qsvrz7m9+5p8lYSiTNUtzMfoXReFp6PelYIDycGEC3U4gLxFufR0JvmePa+dHyPEt/QAVpXY3J15xeWuGwBkU9rA3qaSU9Cr1je4GsyXl0Xft5nWcD2JqyvZ6kl/BpjMuF0HzmWpOpB/2FZiH7xon1/6S9EsgmKnWMBCT6Vyn8QRGo1tDHTdDXhbATK1VwrUqyJ4Yjnp96t5oFEWaezzyNMEoJuDG28uhDuq/ewMKc+JIKiPCXneYgPapWwRWZnsMMQS6RNi8Q96g== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=LHnnH14BYJEqIsLnHPehiYjxaaWJxxpPEeF0ZAf3xj8=; b=BE55FJbNbrdGdYrroEnpQYxDgGA+wRNAwz/RgOVp3/ubei3f1WDkK2FRPpfFlXTnmuzL8y2yHJsOovhdU7w5PLn1oC9p2aJZYIsIdj/o5wxA+QndvNPf2VYKp0rp83QzNLXlkMBsUm8CwotB8QUunw8wGpwMApGmdUa31XxbKZN2ScZNt1TN/9jRvQ1Mj4sDuqau/MUqKL6jMGKQ1EL4gq8doCrA8lWbNjQtFLkyBgEc9DOVMRNrZG9SvVWDy3WVuotGYwsjJpRFF9mG31AbSdKkuJ+ObEA6eFFhW9rfgZLpRNRg0MTS73OXv3V0DWHa3NRIZ+Z7BYT4tKKu2B/KUw== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=intel.com; dmarc=pass action=none header.from=intel.com; dkim=pass header.d=intel.com; arc=none Received: from SN6PR11MB3504.namprd11.prod.outlook.com (2603:10b6:805:d0::17) by CO1PR11MB5186.namprd11.prod.outlook.com (2603:10b6:303:9b::24) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.5654.25; Mon, 26 Sep 2022 06:55:19 +0000 Received: from SN6PR11MB3504.namprd11.prod.outlook.com ([fe80::408e:79cb:b715:b8ac]) by SN6PR11MB3504.namprd11.prod.outlook.com ([fe80::408e:79cb:b715:b8ac%4]) with mapi id 15.20.5654.025; Mon, 26 Sep 2022 06:55:19 +0000 From: "Xia, Chenbo" To: "Wang, YuanX" , "maxime.coquelin@redhat.com" , "dev@dpdk.org" CC: "Hu, Jiayu" , "He, Xingguang" , "Jiang, Cheng1" , "Ma, WenwuX" Subject: RE: [PATCH v3] net/vhost: support asynchronous data path Thread-Topic: [PATCH v3] net/vhost: support asynchronous data path Thread-Index: AQHYts0YCzM5cA8KAk63uznoAs43U63xb24Q Date: Mon, 26 Sep 2022 06:55:19 +0000 Message-ID: References: <20220814150636.2260317-1-jiayu.hu@intel.com> <20220823163534.947869-1-yuanx.wang@intel.com> In-Reply-To: <20220823163534.947869-1-yuanx.wang@intel.com> Accept-Language: en-US, zh-CN Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: dlp-product: dlpe-windows dlp-reaction: no-action dlp-version: 11.6.500.17 authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=intel.com; x-ms-publictraffictype: Email x-ms-traffictypediagnostic: SN6PR11MB3504:EE_|CO1PR11MB5186:EE_ x-ms-office365-filtering-correlation-id: c858e75a-8162-4468-78e5-08da9f8c1323 x-ms-exchange-senderadcheck: 1 x-ms-exchange-antispam-relay: 0 x-microsoft-antispam: BCL:0; x-microsoft-antispam-message-info: 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 x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:; IPV:NLI; SFV:NSPM; H:SN6PR11MB3504.namprd11.prod.outlook.com; PTR:; CAT:NONE; SFS:(13230022)(366004)(346002)(376002)(136003)(396003)(39860400002)(451199015)(110136005)(54906003)(33656002)(86362001)(38070700005)(5660300002)(82960400001)(38100700002)(122000001)(83380400001)(186003)(6506007)(107886003)(9686003)(26005)(41300700001)(76116006)(478600001)(4326008)(66946007)(71200400001)(66476007)(316002)(66556008)(7696005)(8676002)(66446008)(64756008)(55016003)(8936002)(30864003)(2906002)(52536014)(53546011)(579004)(559001); DIR:OUT; SFP:1102; x-ms-exchange-antispam-messagedata-chunkcount: 1 x-ms-exchange-antispam-messagedata-0: =?us-ascii?Q?h3A3q8wTLpaysrO4zRvsHd1xmGdmEImH1mFXqNsnkwYlJNT1D96h60c60/Hr?= =?us-ascii?Q?XvlsyF/t9vAeKNg0QBd3Xaf5nr4OnlJXKk0EEkxWPOFFnBGBTJ1MyB4jEHyL?= =?us-ascii?Q?9sHQCQTgJepL14CJ6Msv4CcTRuCSSOmFy86yYYnOqskgtwt1NjHkE+FSPnHB?= =?us-ascii?Q?49+BRMeY0O8DF15Wu8+dNEO+SNcrKw/7Z11U6HPOeMomlPSVelp5LJxLdePf?= =?us-ascii?Q?KYlbH63ae57Lu7mBmSuZh6+4sZzTquLITDt7vw+vOIAGjuCXHkDD/2nKFP7u?= =?us-ascii?Q?2Nz39xzguTOosly6M47frvlinCa0ms3XIG2suAErER6N6IWV8En+amJoDt2W?= =?us-ascii?Q?Om+ekEgjHJQsH5T8j0oWDX8tv9SYsdA4/L3omzL4j6cf6/Od/3DfvagPCe73?= =?us-ascii?Q?duxbOGwKA/rm4+hZa+CsWbYmGlcxK8QF+p7+Q8VV++8i1eBsZ47mLFdWuase?= =?us-ascii?Q?auF/iRhjG7LA3VOVvqNdwB7kGjiCUFqR+DvE86ZYAKsBxGk+OjH7rTrpXAHt?= =?us-ascii?Q?GkKtHeeIWAGuTmlWVxmAnSdBEEUHjgnb3Il53Elu1S/fmmv2i/2ztm1Fmi/m?= =?us-ascii?Q?1telZIGdpn2E0NJ4nhdCtyYS3j4a8jRN9CFGxcUA6p2/dPhr1UHX2+QZ+Poa?= =?us-ascii?Q?nWyZlp0Jp1SerLCWyYOu0LI1jnPSfekLaoPP3OCzm/cqNNtnjzZxKIsvRDaD?= =?us-ascii?Q?cyxK2dOZUpv4seCtqwWh4RSNNjF2ll6+IIGnAQv9+kePv/v0BNUs4fdY9/FY?= =?us-ascii?Q?zeCgAHo5qKMfPXyXq/ltDBpWwg/Uyyni5QVeQv/yuUZvCJ6m3sbHx0tZMCCd?= =?us-ascii?Q?j7Xzmpwo0OPxfaLJXRBp1G1mFJZGKpC+wmea5Y76SLP0L/bDC8vtPZMgHMwF?= =?us-ascii?Q?bvqLfOcZihpGFyaUtS6hivJdga3Jm1URRvNs9e8WVrxmb36dkCICzdScJA3f?= =?us-ascii?Q?LDu7z/Hbsl/sRFWWYyhXTwUeG4ZcrxCQzlEeHbYVEu3n3Izc6c+XiAUPoO6u?= =?us-ascii?Q?LXz7q4FfXjypC7wR96g6vF0LY4EunlqhuWltcRdd+ulNgHrrQELzTVFQ2Knq?= =?us-ascii?Q?f3pyyUpTV0gc57HMuvbSkVzDpdqC2tHINm/LIneVo1/8sOVYwt+yeILXuzK9?= =?us-ascii?Q?NtwjdVZrR6YrhPdIxQQLjPkH0vgRjyyj5QvziaslDTsJekn6sAABp5UtM0Kr?= =?us-ascii?Q?52lYH4Th+FTx0C4uBUytGuk3T188kiqgra5mloISDW7xkP/iBOB/EaovF5A/?= =?us-ascii?Q?A0IxtZkgUP1ERTBgDn+EX2+FpfUq75sLKHrLTwAvMGE4IbEnVVPBZxIjOsnw?= =?us-ascii?Q?I/0U/Q3nyjoAW1jFpbIhAEoCvV1GwN3lDAJHewWWvK2+s/Dm+rjQZ4XQG6WG?= =?us-ascii?Q?K4VzrziRo/xFOLUiNK6vzRp1M3EHcncO82+hrmPNzpCsQE4UqQ31thZPIC3h?= =?us-ascii?Q?cXI95kgSG78EnJWrVeZMoZXiGxKwfZEBawqShNyG3pX4fmhDVUS3YTy4izG0?= =?us-ascii?Q?Jz/4V9bhEjSaZ4+ynTuGLyFvzS4L6lzLsqIaKoyyp3aF1IS+ZCSVhDCSexB+?= =?us-ascii?Q?rdoMgvDSzG/NUBaVwovKI93zYhwwOs925XE8HdpL?= Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-AuthSource: SN6PR11MB3504.namprd11.prod.outlook.com X-MS-Exchange-CrossTenant-Network-Message-Id: c858e75a-8162-4468-78e5-08da9f8c1323 X-MS-Exchange-CrossTenant-originalarrivaltime: 26 Sep 2022 06:55:19.1519 (UTC) X-MS-Exchange-CrossTenant-fromentityheader: Hosted X-MS-Exchange-CrossTenant-id: 46c98d88-e344-4ed4-8496-4ed7712e255d X-MS-Exchange-CrossTenant-mailboxtype: HOSTED X-MS-Exchange-CrossTenant-userprincipalname: I1kW1bOT1yjlmsI63f5XxvgXCXmxDv6oxlCC1T20an134ChMAIxi+li5W8suuk16YqZ8ObSZtusXfDemTdamDQ== X-MS-Exchange-Transport-CrossTenantHeadersStamped: CO1PR11MB5186 X-OriginatorOrg: intel.com X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org > -----Original Message----- > From: Wang, YuanX > Sent: Wednesday, August 24, 2022 12:36 AM > To: maxime.coquelin@redhat.com; Xia, Chenbo ; > dev@dpdk.org > Cc: Hu, Jiayu ; He, Xingguang ; > Jiang, Cheng1 ; Wang, YuanX ; > Ma, WenwuX > Subject: [PATCH v3] net/vhost: support asynchronous data path How many % will this patch impact the sync data path perf? It should be minor, right? Maxime, should we plan to remove vhost example now? Maintaining vhost examp= le/PMD that have the same functionalities, does not make sense to me. We should se= nd deprecation notice in 22.11 and also get to know tech-board opinion? >=20 > Vhost asynchronous data-path offloads packet copy from the CPU > to the DMA engine. As a result, large packet copy can be accelerated > by the DMA engine, and vhost can free CPU cycles for higher level > functions. >=20 > In this patch, we enable asynchronous data-path for vhostpmd. > Asynchronous data path is enabled per tx/rx queue, and users need > to specify the DMA device used by the tx/rx queue. Each tx/rx queue > only supports to use one DMA device, but one DMA device can be shared > among multiple tx/rx queues of different vhostpmd ports. Vhostpmd -> vhost PMD >=20 > Two PMD parameters are added: > - dmas: specify the used DMA device for a tx/rx queue. > (Default: no queues enable asynchronous data path) > - dma-ring-size: DMA ring size. > (Default: 4096). >=20 > Here is an example: > --vdev > 'eth_vhost0,iface=3D./s0,dmas=3D[txq0@0000:00.01.0;rxq0@0000:00.01.1],dma= - > ring-size=3D4096' >=20 > Signed-off-by: Jiayu Hu > Signed-off-by: Yuan Wang > Signed-off-by: Wenwu Ma > --- > v3: > - add the API to version.map >=20 > v2: > - add missing file > - hide async_tx_poll_completed > - change default DMA ring size to 4096 > --- > drivers/net/vhost/meson.build | 1 + > drivers/net/vhost/rte_eth_vhost.c | 494 ++++++++++++++++++++++++++++-- > drivers/net/vhost/rte_eth_vhost.h | 15 + > drivers/net/vhost/version.map | 7 + > drivers/net/vhost/vhost_testpmd.c | 65 ++++ > 5 files changed, 549 insertions(+), 33 deletions(-) > create mode 100644 drivers/net/vhost/vhost_testpmd.c >=20 > diff --git a/drivers/net/vhost/meson.build b/drivers/net/vhost/meson.buil= d > index f481a3a4b8..22a0ab3a58 100644 > --- a/drivers/net/vhost/meson.build > +++ b/drivers/net/vhost/meson.build > @@ -9,4 +9,5 @@ endif >=20 > deps +=3D 'vhost' > sources =3D files('rte_eth_vhost.c') > +testpmd_sources =3D files('vhost_testpmd.c') > headers =3D files('rte_eth_vhost.h') > diff --git a/drivers/net/vhost/rte_eth_vhost.c > b/drivers/net/vhost/rte_eth_vhost.c > index 7e512d94bf..aa069c6b68 100644 > --- a/drivers/net/vhost/rte_eth_vhost.c > +++ b/drivers/net/vhost/rte_eth_vhost.c > @@ -17,6 +17,8 @@ > #include > #include > #include > +#include > +#include >=20 > #include "rte_eth_vhost.h" >=20 > @@ -36,8 +38,13 @@ enum {VIRTIO_RXQ, VIRTIO_TXQ, VIRTIO_QNUM}; > #define ETH_VHOST_LINEAR_BUF "linear-buffer" > #define ETH_VHOST_EXT_BUF "ext-buffer" > #define ETH_VHOST_LEGACY_OL_FLAGS "legacy-ol-flags" > +#define ETH_VHOST_DMA_ARG "dmas" > +#define ETH_VHOST_DMA_RING_SIZE "dma-ring-size" > #define VHOST_MAX_PKT_BURST 32 >=20 > +#define INVALID_DMA_ID -1 > +#define DEFAULT_DMA_RING_SIZE 4096 > + > static const char *valid_arguments[] =3D { > ETH_VHOST_IFACE_ARG, > ETH_VHOST_QUEUES_ARG, > @@ -48,6 +55,8 @@ static const char *valid_arguments[] =3D { > ETH_VHOST_LINEAR_BUF, > ETH_VHOST_EXT_BUF, > ETH_VHOST_LEGACY_OL_FLAGS, > + ETH_VHOST_DMA_ARG, > + ETH_VHOST_DMA_RING_SIZE, > NULL > }; >=20 > @@ -79,8 +88,39 @@ struct vhost_queue { > struct vhost_stats stats; > int intr_enable; > rte_spinlock_t intr_lock; > + > + /* Flag of enabling async data path */ > + bool async_register; > + /* DMA device ID */ > + int16_t dma_id; > + /** > + * For a Rx queue, "txq" points to its peer Tx queue. > + * For a Tx queue, "txq" is never used. > + */ > + struct vhost_queue *txq; > + /* Array to keep DMA completed packets */ > + struct rte_mbuf *cmpl_pkts[VHOST_MAX_PKT_BURST]; > }; >=20 > +struct dma_input_info { > + int16_t dmas[RTE_MAX_QUEUES_PER_PORT * 2]; > + uint16_t dma_ring_size; > +}; > + > +static int16_t configured_dmas[RTE_DMADEV_DEFAULT_MAX]; > +static int dma_count; > + > +/** > + * By default, its Rx path to call rte_vhost_poll_enqueue_completed() fo= r > enqueue operations. > + * However, Rx function is never been called in testpmd "txonly" mode, > thus causing virtio > + * cannot receive DMA completed packets. To make txonly mode work > correctly, we provide a > + * command in testpmd to call rte_vhost_poll_enqueue_completed() in Tx > path. > + * > + * When set async_tx_poll_completed to true, Tx path calls > rte_vhost_poll_enqueue_completed(); > + * otherwise, Rx path calls it. > + */ > +bool async_tx_poll_completed; > + > struct pmd_internal { > rte_atomic32_t dev_attached; > char *iface_name; > @@ -93,6 +133,10 @@ struct pmd_internal { > bool vlan_strip; > bool rx_sw_csum; > bool tx_sw_csum; > + struct { > + int16_t dma_id; > + bool async_register; > + } queue_dmas[RTE_MAX_QUEUES_PER_PORT * 2]; > }; >=20 > struct internal_list { > @@ -123,6 +167,17 @@ struct rte_vhost_vring_state { >=20 > static struct rte_vhost_vring_state *vring_states[RTE_MAX_ETHPORTS]; >=20 > +static bool > +dma_is_configured(int16_t dma_id) > +{ > + int i; > + > + for (i =3D 0; i < dma_count; i++) > + if (configured_dmas[i] =3D=3D dma_id) > + return true; > + return false; > +} > + > static int > vhost_dev_xstats_reset(struct rte_eth_dev *dev) > { > @@ -395,6 +450,17 @@ vhost_dev_rx_sw_csum(struct rte_mbuf *mbuf) > mbuf->ol_flags |=3D RTE_MBUF_F_RX_L4_CKSUM_GOOD; > } >=20 > +static inline void > +vhost_tx_free_completed(uint16_t vid, uint16_t virtqueue_id, int16_t > dma_id, > + struct rte_mbuf **pkts, uint16_t count) > +{ > + uint16_t i, ret; > + > + ret =3D rte_vhost_poll_enqueue_completed(vid, virtqueue_id, pkts, > count, dma_id, 0); > + for (i =3D 0; likely(i < ret); i++) > + rte_pktmbuf_free(pkts[i]); > +} > + > static uint16_t > eth_vhost_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) > { > @@ -403,7 +469,7 @@ eth_vhost_rx(void *q, struct rte_mbuf **bufs, uint16_= t > nb_bufs) > uint16_t nb_receive =3D nb_bufs; >=20 > if (unlikely(rte_atomic32_read(&r->allow_queuing) =3D=3D 0)) > - return 0; > + goto tx_poll; >=20 > rte_atomic32_set(&r->while_queuing, 1); >=20 > @@ -411,19 +477,36 @@ eth_vhost_rx(void *q, struct rte_mbuf **bufs, > uint16_t nb_bufs) > goto out; >=20 > /* Dequeue packets from guest TX queue */ > - while (nb_receive) { > - uint16_t nb_pkts; > - uint16_t num =3D (uint16_t)RTE_MIN(nb_receive, > - VHOST_MAX_PKT_BURST); > - > - nb_pkts =3D rte_vhost_dequeue_burst(r->vid, r->virtqueue_id, > - r->mb_pool, &bufs[nb_rx], > - num); > - > - nb_rx +=3D nb_pkts; > - nb_receive -=3D nb_pkts; > - if (nb_pkts < num) > - break; > + if (!r->async_register) { > + while (nb_receive) { > + uint16_t nb_pkts; > + uint16_t num =3D (uint16_t)RTE_MIN(nb_receive, > + VHOST_MAX_PKT_BURST); > + > + nb_pkts =3D rte_vhost_dequeue_burst(r->vid, r- > >virtqueue_id, > + r->mb_pool, &bufs[nb_rx], > + num); > + > + nb_rx +=3D nb_pkts; > + nb_receive -=3D nb_pkts; > + if (nb_pkts < num) > + break; > + } > + } else { > + while (nb_receive) { > + uint16_t nb_pkts; > + uint16_t num =3D (uint16_t)RTE_MIN(nb_receive, > VHOST_MAX_PKT_BURST); Some variables like nb_pkts and num, which will anyway be used. We don't need to define them in every loop. > + int nr_inflight; > + > + nb_pkts =3D rte_vhost_async_try_dequeue_burst(r->vid, r- > >virtqueue_id, > + r->mb_pool, &bufs[nb_rx], num, > &nr_inflight, > + r->dma_id, 0); > + > + nb_rx +=3D nb_pkts; > + nb_receive -=3D nb_pkts; > + if (nb_pkts < num) > + break; > + } > } >=20 > r->stats.pkts +=3D nb_rx; > @@ -444,6 +527,17 @@ eth_vhost_rx(void *q, struct rte_mbuf **bufs, > uint16_t nb_bufs) > out: > rte_atomic32_set(&r->while_queuing, 0); >=20 > +tx_poll: > + /** > + * Poll and free completed packets for the virtqueue of Tx queue. > + * Note that we access Tx queue's virtqueue, which is protected > + * by vring lock. > + */ > + if (!async_tx_poll_completed && r->txq->async_register) { Logically we should check async_register first. > + vhost_tx_free_completed(r->vid, r->txq->virtqueue_id, r->txq- > >dma_id, > + r->cmpl_pkts, VHOST_MAX_PKT_BURST); > + } > + > return nb_rx; > } >=20 > @@ -485,31 +579,53 @@ eth_vhost_tx(void *q, struct rte_mbuf **bufs, > uint16_t nb_bufs) > } >=20 > /* Enqueue packets to guest RX queue */ > - while (nb_send) { > - uint16_t nb_pkts; > - uint16_t num =3D (uint16_t)RTE_MIN(nb_send, > - VHOST_MAX_PKT_BURST); > + if (!r->async_register) { > + while (nb_send) { > + uint16_t nb_pkts; > + uint16_t num =3D (uint16_t)RTE_MIN(nb_send, > VHOST_MAX_PKT_BURST); > + > + nb_pkts =3D rte_vhost_enqueue_burst(r->vid, r- > >virtqueue_id, > + &bufs[nb_tx], num); > + > + nb_tx +=3D nb_pkts; > + nb_send -=3D nb_pkts; > + if (nb_pkts < num) > + break; > + } >=20 > - nb_pkts =3D rte_vhost_enqueue_burst(r->vid, r->virtqueue_id, > - &bufs[nb_tx], num); > + for (i =3D 0; likely(i < nb_tx); i++) { > + nb_bytes +=3D bufs[i]->pkt_len; > + rte_pktmbuf_free(bufs[i]); > + } >=20 > - nb_tx +=3D nb_pkts; > - nb_send -=3D nb_pkts; > - if (nb_pkts < num) > - break; > - } > + } else { > + while (nb_send) { > + uint16_t nb_pkts; > + uint16_t num =3D (uint16_t)RTE_MIN(nb_send, > VHOST_MAX_PKT_BURST); >=20 > - for (i =3D 0; likely(i < nb_tx); i++) > - nb_bytes +=3D bufs[i]->pkt_len; > + nb_pkts =3D rte_vhost_submit_enqueue_burst(r->vid, r- > >virtqueue_id, > + &bufs[nb_tx], num, r->dma_id, 0); >=20 > - nb_missed =3D nb_bufs - nb_tx; > + nb_tx +=3D nb_pkts; > + nb_send -=3D nb_pkts; > + if (nb_pkts < num) > + break; > + } >=20 > + for (i =3D 0; likely(i < nb_tx); i++) > + nb_bytes +=3D bufs[i]->pkt_len; > + > + if (unlikely(async_tx_poll_completed)) { > + vhost_tx_free_completed(r->vid, r->virtqueue_id, r- > >dma_id, r->cmpl_pkts, > + VHOST_MAX_PKT_BURST); > + } > + } > + > + nb_missed =3D nb_bufs - nb_tx; > r->stats.pkts +=3D nb_tx; > r->stats.bytes +=3D nb_bytes; > r->stats.missed_pkts +=3D nb_missed; >=20 > - for (i =3D 0; likely(i < nb_tx); i++) > - rte_pktmbuf_free(bufs[i]); > out: > rte_atomic32_set(&r->while_queuing, 0); >=20 > @@ -797,6 +913,8 @@ queue_setup(struct rte_eth_dev *eth_dev, struct > pmd_internal *internal) > vq->vid =3D internal->vid; > vq->internal =3D internal; > vq->port =3D eth_dev->data->port_id; > + if (i < eth_dev->data->nb_tx_queues) > + vq->txq =3D eth_dev->data->tx_queues[i]; > } > for (i =3D 0; i < eth_dev->data->nb_tx_queues; i++) { > vq =3D eth_dev->data->tx_queues[i]; > @@ -982,6 +1100,9 @@ vring_state_changed(int vid, uint16_t vring, int > enable) > struct rte_vhost_vring_state *state; > struct rte_eth_dev *eth_dev; > struct internal_list *list; > + struct vhost_queue *queue; > + struct pmd_internal *internal; > + int qid; > char ifname[PATH_MAX]; >=20 > rte_vhost_get_ifname(vid, ifname, sizeof(ifname)); > @@ -1010,6 +1131,65 @@ vring_state_changed(int vid, uint16_t vring, int > enable) >=20 > update_queuing_status(eth_dev, false); >=20 > + qid =3D vring / VIRTIO_QNUM; > + if (vring % VIRTIO_QNUM =3D=3D VIRTIO_RXQ) > + queue =3D eth_dev->data->tx_queues[qid]; > + else > + queue =3D eth_dev->data->rx_queues[qid]; > + > + if (!queue) > + goto skip; > + > + internal =3D eth_dev->data->dev_private; > + > + /* Register async data path for the queue assigned valid DMA device > */ > + if (internal->queue_dmas[queue->virtqueue_id].dma_id =3D=3D > INVALID_DMA_ID) > + goto skip; > + > + if (enable && !queue->async_register) { > + if (rte_vhost_async_channel_register_thread_unsafe(vid, vring)) > { > + VHOST_LOG(ERR, "Failed to register async for vid-%u > vring-%u!\n", vid, > + vring); > + return -1; > + } > + > + queue->async_register =3D true; > + internal->queue_dmas[vring].async_register =3D true; > + > + VHOST_LOG(INFO, "Succeed to register async for vid-%u vring- > %u\n", vid, vring); > + } > + > + if (!enable && queue->async_register) { > + struct rte_mbuf *pkts[VHOST_MAX_PKT_BURST]; > + uint16_t ret, i, nr_done =3D 0; > + uint16_t dma_id =3D queue->dma_id; > + > + while (rte_vhost_async_get_inflight_thread_unsafe(vid, vring) > > 0) { > + ret =3D rte_vhost_clear_queue_thread_unsafe(vid, vring, > pkts, > + VHOST_MAX_PKT_BURST, dma_id, 0); > + > + for (i =3D 0; i < ret ; i++) > + rte_pktmbuf_free(pkts[i]); > + > + nr_done +=3D ret; > + } > + > + VHOST_LOG(INFO, "Completed %u in-flight pkts for vid-%u vring- > %u\n", nr_done, vid, > + vring); > + > + if (rte_vhost_async_channel_unregister_thread_unsafe(vid, > vring)) { > + VHOST_LOG(ERR, "Failed to unregister async for vid-%u > vring-%u\n", vid, > + vring); > + return -1; > + } > + > + queue->async_register =3D false; > + internal->queue_dmas[vring].async_register =3D false; > + > + VHOST_LOG(INFO, "Succeed to unregister async for vid-%u vring- > %u\n", vid, vring); > + } > + > +skip: > VHOST_LOG(INFO, "vring%u is %s\n", > vring, enable ? "enabled" : "disabled"); >=20 > @@ -1158,6 +1338,12 @@ rte_eth_vhost_get_vid_from_port_id(uint16_t port_i= d) > return vid; > } >=20 > +void > +rte_eth_vhost_async_tx_poll_completed(bool enable) > +{ > + async_tx_poll_completed =3D enable; > +} > + > static int > eth_dev_configure(struct rte_eth_dev *dev) > { > @@ -1214,11 +1400,37 @@ eth_dev_stop(struct rte_eth_dev *dev) > return 0; > } >=20 > +static inline int > +async_clear_virtqueue(uint16_t vid, uint16_t virtqueue_id, int16_t dma_i= d) > +{ > + struct rte_mbuf *pkts[VHOST_MAX_PKT_BURST]; > + uint16_t i, ret, nr_done =3D 0; > + > + while (rte_vhost_async_get_inflight(vid, virtqueue_id) > 0) { > + ret =3D rte_vhost_clear_queue(vid, virtqueue_id, pkts, > VHOST_MAX_PKT_BURST, dma_id, > + 0); > + for (i =3D 0; i < ret ; i++) > + rte_pktmbuf_free(pkts[i]); > + > + nr_done +=3D ret; > + } > + VHOST_LOG(INFO, "Completed %u pkts for vid-%u vring-%u\n", nr_done, > vid, virtqueue_id); > + > + if (rte_vhost_async_channel_unregister(vid, virtqueue_id)) { > + VHOST_LOG(ERR, "Failed to unregister async for vid-%u vring- > %u\n", vid, > + virtqueue_id); > + return -1; > + } > + > + return nr_done; > +} If you don't need to check the return value, make it return void. > + > static int > eth_dev_close(struct rte_eth_dev *dev) > { > struct pmd_internal *internal; > struct internal_list *list; > + struct vhost_queue *queue; > unsigned int i, ret; >=20 > if (rte_eal_process_type() !=3D RTE_PROC_PRIMARY) > @@ -1232,6 +1444,27 @@ eth_dev_close(struct rte_eth_dev *dev) >=20 > list =3D find_internal_resource(internal->iface_name); > if (list) { > + /* Make sure all in-flight packets are completed before > destroy virtio */ Virtio -> vhost > + if (dev->data->rx_queues) { > + for (i =3D 0; i < dev->data->nb_rx_queues; i++) { > + queue =3D dev->data->rx_queues[i]; > + if (queue->async_register) { > + async_clear_virtqueue(queue->vid, queue- > >virtqueue_id, > + queue->dma_id); > + } > + } > + } > + > + if (dev->data->tx_queues) { > + for (i =3D 0; i < dev->data->nb_tx_queues; i++) { > + queue =3D dev->data->tx_queues[i]; > + if (queue->async_register) { > + async_clear_virtqueue(queue->vid, queue- > >virtqueue_id, > + queue->dma_id); > + } > + } > + } > + > rte_vhost_driver_unregister(internal->iface_name); > pthread_mutex_lock(&internal_list_lock); > TAILQ_REMOVE(&internal_list, list, next); > @@ -1266,6 +1499,7 @@ eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_= t > rx_queue_id, > struct rte_mempool *mb_pool) > { > struct vhost_queue *vq; > + struct pmd_internal *internal =3D dev->data->dev_private; >=20 > vq =3D rte_zmalloc_socket(NULL, sizeof(struct vhost_queue), > RTE_CACHE_LINE_SIZE, socket_id); > @@ -1276,6 +1510,8 @@ eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_= t > rx_queue_id, >=20 > vq->mb_pool =3D mb_pool; > vq->virtqueue_id =3D rx_queue_id * VIRTIO_QNUM + VIRTIO_TXQ; > + vq->async_register =3D internal->queue_dmas[vq- > >virtqueue_id].async_register; > + vq->dma_id =3D internal->queue_dmas[vq->virtqueue_id].dma_id; > rte_spinlock_init(&vq->intr_lock); > dev->data->rx_queues[rx_queue_id] =3D vq; >=20 > @@ -1289,6 +1525,7 @@ eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_= t > tx_queue_id, > const struct rte_eth_txconf *tx_conf __rte_unused) > { > struct vhost_queue *vq; > + struct pmd_internal *internal =3D dev->data->dev_private; >=20 > vq =3D rte_zmalloc_socket(NULL, sizeof(struct vhost_queue), > RTE_CACHE_LINE_SIZE, socket_id); > @@ -1298,6 +1535,8 @@ eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_= t > tx_queue_id, > } >=20 > vq->virtqueue_id =3D tx_queue_id * VIRTIO_QNUM + VIRTIO_RXQ; > + vq->async_register =3D internal->queue_dmas[vq- > >virtqueue_id].async_register; > + vq->dma_id =3D internal->queue_dmas[vq->virtqueue_id].dma_id; > rte_spinlock_init(&vq->intr_lock); > dev->data->tx_queues[tx_queue_id] =3D vq; >=20 > @@ -1508,13 +1747,14 @@ static const struct eth_dev_ops ops =3D { > static int > eth_dev_vhost_create(struct rte_vdev_device *dev, char *iface_name, > int16_t queues, const unsigned int numa_node, uint64_t flags, > - uint64_t disable_flags) > + uint64_t disable_flags, struct dma_input_info *dma_input) > { > const char *name =3D rte_vdev_device_name(dev); > struct rte_eth_dev_data *data; > struct pmd_internal *internal =3D NULL; > struct rte_eth_dev *eth_dev =3D NULL; > struct rte_ether_addr *eth_addr =3D NULL; > + int i; >=20 > VHOST_LOG(INFO, "Creating VHOST-USER backend on numa socket %u\n", > numa_node); > @@ -1563,6 +1803,12 @@ eth_dev_vhost_create(struct rte_vdev_device *dev, > char *iface_name, > eth_dev->rx_pkt_burst =3D eth_vhost_rx; > eth_dev->tx_pkt_burst =3D eth_vhost_tx; >=20 > + for (i =3D 0; i < RTE_MAX_QUEUES_PER_PORT * 2; i++) { > + /* Invalid DMA ID indicates the queue does not want to enable > async data path */ > + internal->queue_dmas[i].dma_id =3D dma_input->dmas[i]; > + internal->queue_dmas[i].async_register =3D false; > + } > + > rte_eth_dev_probing_finish(eth_dev); > return 0; >=20 > @@ -1602,6 +1848,153 @@ open_int(const char *key __rte_unused, const char > *value, void *extra_args) > return 0; > } >=20 > +static int > +init_dma(int16_t dma_id, uint16_t ring_size) > +{ > + struct rte_dma_info info; > + struct rte_dma_conf dev_config =3D { .nb_vchans =3D 1 }; > + struct rte_dma_vchan_conf qconf =3D { > + .direction =3D RTE_DMA_DIR_MEM_TO_MEM, > + }; > + int ret =3D 0; > + > + if (dma_is_configured(dma_id)) > + goto out; > + > + if (rte_dma_info_get(dma_id, &info) !=3D 0) { > + VHOST_LOG(ERR, "dma %u get info failed\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + if (info.max_vchans < 1) { > + VHOST_LOG(ERR, "No channels available on dma %d\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + if (rte_dma_configure(dma_id, &dev_config) !=3D 0) { > + VHOST_LOG(ERR, "dma %u configure failed\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + rte_dma_info_get(dma_id, &info); > + if (info.nb_vchans !=3D 1) { > + VHOST_LOG(ERR, "dma %u has no queues\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + qconf.nb_desc =3D RTE_MIN(ring_size, info.max_desc); > + if (rte_dma_vchan_setup(dma_id, 0, &qconf) !=3D 0) { > + VHOST_LOG(ERR, "dma %u queue setup failed\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + if (rte_dma_start(dma_id) !=3D 0) { > + VHOST_LOG(ERR, "dma %u start failed\n", dma_id); > + ret =3D -1; > + goto out; > + } > + > + configured_dmas[dma_count++] =3D dma_id; > + > +out: > + return ret; > +} > + > +static int > +open_dma(const char *key __rte_unused, const char *value, void > *extra_args) > +{ > + struct dma_input_info *dma_input =3D extra_args; > + char *input =3D strndup(value, strlen(value) + 1); > + char *addrs =3D input; > + char *ptrs[2]; > + char *start, *end, *substr; > + uint16_t qid, virtqueue_id; > + int16_t dma_id; > + int i, ret =3D 0; > + > + while (isblank(*addrs)) > + addrs++; > + if (*addrs =3D=3D '\0') { > + VHOST_LOG(ERR, "No input DMA addresses\n"); > + ret =3D -1; > + goto out; > + } > + > + /* process DMA devices within bracket. */ > + addrs++; > + substr =3D strtok(addrs, ";]"); > + if (!substr) { > + VHOST_LOG(ERR, "No input DMA addresse\n"); > + ret =3D -1; > + goto out; > + } > + > + do { > + rte_strsplit(substr, strlen(substr), ptrs, 2, '@'); > + > + char *txq, *rxq; > + bool is_txq; > + > + txq =3D strstr(ptrs[0], "txq"); > + rxq =3D strstr(ptrs[0], "rxq"); > + if (txq =3D=3D NULL && rxq =3D=3D NULL) { > + VHOST_LOG(ERR, "Illegal queue\n"); > + ret =3D -1; > + goto out; > + } else if (txq) { > + is_txq =3D true; > + start =3D txq; > + } else { > + is_txq =3D false; > + start =3D rxq; > + } > + > + start +=3D 3; > + qid =3D strtol(start, &end, 0); > + if (end =3D=3D start) { > + VHOST_LOG(ERR, "No input queue ID\n"); > + ret =3D -1; > + goto out; > + } > + > + virtqueue_id =3D is_txq ? qid * 2 + VIRTIO_RXQ : qid * 2 + > VIRTIO_TXQ; > + > + dma_id =3D rte_dma_get_dev_id_by_name(ptrs[1]); > + if (dma_id < 0) { > + VHOST_LOG(ERR, "Fail to find DMA device %s.\n", ptrs[1]); > + ret =3D -1; > + goto out; > + } > + > + ret =3D init_dma(dma_id, dma_input->dma_ring_size); > + if (ret !=3D 0) { > + VHOST_LOG(ERR, "Fail to initialize DMA %u\n", dma_id); > + ret =3D -1; > + break; > + } > + > + dma_input->dmas[virtqueue_id] =3D dma_id; > + > + substr =3D strtok(NULL, ";]"); > + } while (substr); > + > + for (i =3D 0; i < dma_count; i++) { > + if (rte_vhost_async_dma_configure(configured_dmas[i], 0) < 0) > { > + VHOST_LOG(ERR, "Fail to configure DMA %u to vhost\n", > configured_dmas[i]); > + ret =3D -1; > + } > + } > + > +out: > + free(input); > + return ret; > +} > + > static int > rte_pmd_vhost_probe(struct rte_vdev_device *dev) > { > @@ -1620,6 +2013,10 @@ rte_pmd_vhost_probe(struct rte_vdev_device *dev) > int legacy_ol_flags =3D 0; > struct rte_eth_dev *eth_dev; > const char *name =3D rte_vdev_device_name(dev); > + struct dma_input_info dma_input; > + > + memset(dma_input.dmas, INVALID_DMA_ID, sizeof(dma_input.dmas)); > + dma_input.dma_ring_size =3D DEFAULT_DMA_RING_SIZE; >=20 > VHOST_LOG(INFO, "Initializing pmd_vhost for %s\n", name); >=20 > @@ -1735,6 +2132,35 @@ rte_pmd_vhost_probe(struct rte_vdev_device *dev) > goto out_free; > } >=20 > + if (rte_kvargs_count(kvlist, ETH_VHOST_DMA_RING_SIZE) =3D=3D 1) { > + ret =3D rte_kvargs_process(kvlist, ETH_VHOST_DMA_RING_SIZE, > + &open_int, &dma_input.dma_ring_size); > + if (ret < 0) > + goto out_free; > + > + if (!rte_is_power_of_2(dma_input.dma_ring_size)) { > + dma_input.dma_ring_size =3D > rte_align32pow2(dma_input.dma_ring_size); > + VHOST_LOG(INFO, "Convert dma_ring_size to the power of > two %u\n", > + dma_input.dma_ring_size); > + } > + } > + > + if (rte_kvargs_count(kvlist, ETH_VHOST_DMA_ARG) =3D=3D 1) { > + ret =3D rte_kvargs_process(kvlist, ETH_VHOST_DMA_ARG, > + &open_dma, &dma_input); > + if (ret < 0) { > + VHOST_LOG(ERR, "Failed to parse %s\n", > ETH_VHOST_DMA_ARG); > + goto out_free; > + } > + > + flags |=3D RTE_VHOST_USER_ASYNC_COPY; > + /** > + * Don't support live migration when enable > + * DMA acceleration. > + */ > + disable_flags |=3D (1ULL << VHOST_F_LOG_ALL); > + } I think we should at least give a warning if some user only specify the=20 dma ring size but does not specify 'dmas' Similarly, only configuring dmas case, we can give a info to user it's usin= g default value. > + > if (legacy_ol_flags =3D=3D 0) > flags |=3D RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS; >=20 > @@ -1742,7 +2168,7 @@ rte_pmd_vhost_probe(struct rte_vdev_device *dev) > dev->device.numa_node =3D rte_socket_id(); >=20 > ret =3D eth_dev_vhost_create(dev, iface_name, queues, > - dev->device.numa_node, flags, disable_flags); > + dev->device.numa_node, flags, disable_flags, &dma_input); > if (ret =3D=3D -1) > VHOST_LOG(ERR, "Failed to create %s\n", name); >=20 > @@ -1786,4 +2212,6 @@ RTE_PMD_REGISTER_PARAM_STRING(net_vhost, > "postcopy-support=3D<0|1> " > "tso=3D<0|1> " > "linear-buffer=3D<0|1> " > - "ext-buffer=3D<0|1>"); > + "ext-buffer=3D<0|1> " > + "dma-ring-size=3D" Missing a space before " IIUC, only the last one does not need that space. > + "dmas=3D[txq0@dma_addr;rxq0@dma_addr] "); > diff --git a/drivers/net/vhost/rte_eth_vhost.h > b/drivers/net/vhost/rte_eth_vhost.h > index 0e68b9f668..146c98803d 100644 > --- a/drivers/net/vhost/rte_eth_vhost.h > +++ b/drivers/net/vhost/rte_eth_vhost.h > @@ -52,6 +52,21 @@ int rte_eth_vhost_get_queue_event(uint16_t port_id, > */ > int rte_eth_vhost_get_vid_from_port_id(uint16_t port_id); >=20 > +/** > + * @warning > + * @b EXPERIMENTAL: this API may change, or be removed, without prior > notice > + * > + * By default, rte_vhost_poll_enqueue_completed() is called in Rx path. > + * This function enables Tx path, rather than Rx path, to poll completed > + * packets for vhost async enqueue operations. Note that virtio may neve= r > + * receive DMA completed packets if there are no more Tx operations. > + * > + * @param enable > + * True indicates Tx path to call rte_vhost_poll_enqueue_completed(). > + */ > +__rte_experimental > +void rte_eth_vhost_async_tx_poll_completed(bool enable); > + > #ifdef __cplusplus > } > #endif > diff --git a/drivers/net/vhost/version.map b/drivers/net/vhost/version.ma= p > index e42c89f1eb..0a40441227 100644 > --- a/drivers/net/vhost/version.map > +++ b/drivers/net/vhost/version.map > @@ -6,3 +6,10 @@ DPDK_23 { >=20 > local: *; > }; > + > +EXPERIMENTAL { > + global: > + > + # added in 22.11 > + rte_eth_vhost_async_tx_poll_completed; > +}; > diff --git a/drivers/net/vhost/vhost_testpmd.c > b/drivers/net/vhost/vhost_testpmd.c > new file mode 100644 > index 0000000000..b8227d1086 > --- /dev/null > +++ b/drivers/net/vhost/vhost_testpmd.c > @@ -0,0 +1,65 @@ > +/* SPDX-License-Identifier: BSD-3-Clause > + * Copyright(c) 2022 Intel Corporation. > + */ > +#include > +#include > +#include > + > +#include "testpmd.h" > + > +struct cmd_tx_poll_result { > + cmdline_fixed_string_t async_vhost; > + cmdline_fixed_string_t tx; > + cmdline_fixed_string_t poll; > + cmdline_fixed_string_t completed; > + cmdline_fixed_string_t what; > +}; > + > +static cmdline_parse_token_string_t cmd_tx_async_vhost =3D > + TOKEN_STRING_INITIALIZER(struct cmd_tx_poll_result, async_vhost, > "async_vhost"); > +static cmdline_parse_token_string_t cmd_tx_tx =3D > + TOKEN_STRING_INITIALIZER(struct cmd_tx_poll_result, tx, "tx"); > +static cmdline_parse_token_string_t cmd_tx_poll =3D > + TOKEN_STRING_INITIALIZER(struct cmd_tx_poll_result, poll, "poll"); > +static cmdline_parse_token_string_t cmd_tx_completed =3D > + TOKEN_STRING_INITIALIZER(struct cmd_tx_poll_result, completed, > "completed"); > +static cmdline_parse_token_string_t cmd_tx_what =3D > + TOKEN_STRING_INITIALIZER(struct cmd_tx_poll_result, what, "on#off"); > + > +static void > +cmd_tx_poll_parsed(void *parsed_result, __rte_unused struct cmdline *cl, > __rte_unused void *data) > +{ > + struct cmd_tx_poll_result *res =3D parsed_result; > + > + if (!strcmp(res->what, "on")) > + rte_eth_vhost_async_tx_poll_completed(true); > + else if (!strcmp(res->what, "off")) > + rte_eth_vhost_async_tx_poll_completed(false); We should print log when the user input is not on/off. Thanks, Chenbo > +} > + > +static cmdline_parse_inst_t async_vhost_cmd_tx_poll =3D { > + .f =3D cmd_tx_poll_parsed, > + .data =3D NULL, > + .help_str =3D "async-vhost tx poll completed on|off", > + .tokens =3D { > + (void *)&cmd_tx_async_vhost, > + (void *)&cmd_tx_tx, > + (void *)&cmd_tx_poll, > + (void *)&cmd_tx_completed, > + (void *)&cmd_tx_what, > + NULL, > + }, > +}; > + > +static struct testpmd_driver_commands async_vhost_cmds =3D { > + .commands =3D { > + { > + &async_vhost_cmd_tx_poll, > + "async_vhost tx poll completed (on|off)\n" > + " Poll and free DMA completed packets in Tx path.\n", > + }, > + { NULL, NULL }, > + }, > +}; > + > +TESTPMD_ADD_DRIVER_COMMANDS(async_vhost_cmds) > -- > 2.25.1