From patchwork Wed Jul 22 22:10:44 2020 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Alexander Lobakin X-Patchwork-Id: 11679255 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 85EE51510 for ; Wed, 22 Jul 2020 22:13:02 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 68BB322CB1 for ; Wed, 22 Jul 2020 22:13:02 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=fail reason="key not found in DNS" (0-bit key) header.d=marvell.com header.i=@marvell.com header.b="fxBs/9lz" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S2387506AbgGVWNB (ORCPT ); Wed, 22 Jul 2020 18:13:01 -0400 Received: from mx0b-0016f401.pphosted.com ([67.231.156.173]:45016 "EHLO mx0b-0016f401.pphosted.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1733156AbgGVWM7 (ORCPT ); Wed, 22 Jul 2020 18:12:59 -0400 Received: from pps.filterd (m0045851.ppops.net [127.0.0.1]) by mx0b-0016f401.pphosted.com (8.16.0.42/8.16.0.42) with SMTP id 06MM6wNI019882; Wed, 22 Jul 2020 15:12:43 -0700 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=marvell.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-transfer-encoding : content-type; s=pfpt0818; bh=zS0D2xqWH6fEMmnPectJQZgM/k93cgYZ441B5zCe7LU=; b=fxBs/9lzqdnwaRSipj40590evKJZfQNoilZv69t+5tW+Qzn+RgRDgAsarhqSF/5ICHvD 27gSouEq1kb/U/YU1zSdvolliqEcyBba7xmeYDCiVfTVyOkb3WeUsZd7BSFc/JhsFt6Y sQhV1fg8eV+Wpf7MxMjmixFeylBRCgzjywKT1hbWDCU7OT+dK3gdlr+46zv7hg9EYw4s mYSRbMqbMvpC9bgbiCSFfaZMp3pcMpw82Mlz/cd0qvaL8s60YtY9/SXhCpPmD7i9lgOu 87whYmNS2mKX9ilLkoIlndxZx+s0Qbkomh7GKZdfqa/p+Iiwgw118jU9WL5H9AOFYNRC Cw== Received: from sc-exch02.marvell.com ([199.233.58.182]) by mx0b-0016f401.pphosted.com with ESMTP id 32c0kkt0pt-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Wed, 22 Jul 2020 15:12:42 -0700 Received: from DC5-EXCH01.marvell.com (10.69.176.38) by SC-EXCH02.marvell.com (10.93.176.82) with Microsoft SMTP Server (TLS) id 15.0.1497.2; Wed, 22 Jul 2020 15:12:41 -0700 Received: from maili.marvell.com (10.69.176.80) by DC5-EXCH01.marvell.com (10.69.176.38) with Microsoft SMTP Server id 15.0.1497.2 via Frontend Transport; Wed, 22 Jul 2020 15:12:41 -0700 Received: from NN-LT0049.marvell.com (NN-LT0049.marvell.com [10.193.54.6]) by maili.marvell.com (Postfix) with ESMTP id AB4833F703F; Wed, 22 Jul 2020 15:12:34 -0700 (PDT) From: Alexander Lobakin To: "David S. Miller" , Jakub Kicinski CC: Alexander Lobakin , Igor Russkikh , Michal Kalderon , "Ariel Elior" , Denis Bolotin , "Doug Ledford" , Jason Gunthorpe , "Alexei Starovoitov" , Daniel Borkmann , "Jesper Dangaard Brouer" , John Fastabend , Martin KaFai Lau , Song Liu , "Yonghong Song" , Andrii Nakryiko , KP Singh , , , , , Subject: [PATCH v2 net-next 14/15] qede: refactor XDP Tx processing Date: Thu, 23 Jul 2020 01:10:44 +0300 Message-ID: <20200722221045.5436-15-alobakin@marvell.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20200722221045.5436-1-alobakin@marvell.com> References: <20200722221045.5436-1-alobakin@marvell.com> MIME-Version: 1.0 X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.235,18.0.687 definitions=2020-07-22_16:2020-07-22,2020-07-22 signatures=0 Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org Current XDP Tx logic is suboptimal and can't be reused for XDP_REDIRECT path. Make qede_xdp_{tx_int,xmit}() more universal and effective in general to allow future expanding. Misc: use unlikely() hints where appropriate and replace "fallthrough" comments with pseudo-keywords. Signed-off-by: Alexander Lobakin Signed-off-by: Igor Russkikh Signed-off-by: Michal Kalderon --- drivers/net/ethernet/qlogic/qede/qede.h | 1 + drivers/net/ethernet/qlogic/qede/qede_fp.c | 89 +++++++++++----------- 2 files changed, 45 insertions(+), 45 deletions(-) diff --git a/drivers/net/ethernet/qlogic/qede/qede.h b/drivers/net/ethernet/qlogic/qede/qede.h index e8ed0bb94ee0..308c66a5f98f 100644 --- a/drivers/net/ethernet/qlogic/qede/qede.h +++ b/drivers/net/ethernet/qlogic/qede/qede.h @@ -455,6 +455,7 @@ struct qede_fastpath { u8 id; u8 xdp_xmit; +#define QEDE_XDP_TX BIT(0) struct napi_struct napi; struct qed_sb_info *sb_info; diff --git a/drivers/net/ethernet/qlogic/qede/qede_fp.c b/drivers/net/ethernet/qlogic/qede/qede_fp.c index 1c4ece0713f8..c80bf6d37b89 100644 --- a/drivers/net/ethernet/qlogic/qede/qede_fp.c +++ b/drivers/net/ethernet/qlogic/qede/qede_fp.c @@ -302,48 +302,37 @@ static inline void qede_update_tx_producer(struct qede_tx_queue *txq) wmb(); } -static int qede_xdp_xmit(struct qede_dev *edev, struct qede_fastpath *fp, - struct sw_rx_data *metadata, u16 padding, u16 length) +static int qede_xdp_xmit(struct qede_tx_queue *txq, dma_addr_t dma, u16 pad, + u16 len, struct page *page) { - struct qede_tx_queue *txq = fp->xdp_tx; - struct eth_tx_1st_bd *first_bd; - u16 idx = txq->sw_tx_prod; + struct eth_tx_1st_bd *bd; + struct sw_tx_xdp *xdp; u16 val; - if (!qed_chain_get_elem_left(&txq->tx_pbl)) { + if (unlikely(qed_chain_get_elem_used(&txq->tx_pbl) >= + txq->num_tx_buffers)) { txq->stopped_cnt++; return -ENOMEM; } - first_bd = (struct eth_tx_1st_bd *)qed_chain_produce(&txq->tx_pbl); + bd = qed_chain_produce(&txq->tx_pbl); + bd->data.nbds = 1; + bd->data.bd_flags.bitfields = BIT(ETH_TX_1ST_BD_FLAGS_START_BD_SHIFT); - memset(first_bd, 0, sizeof(*first_bd)); - first_bd->data.bd_flags.bitfields = - BIT(ETH_TX_1ST_BD_FLAGS_START_BD_SHIFT); - - val = (length & ETH_TX_DATA_1ST_BD_PKT_LEN_MASK) << + val = (len & ETH_TX_DATA_1ST_BD_PKT_LEN_MASK) << ETH_TX_DATA_1ST_BD_PKT_LEN_SHIFT; - first_bd->data.bitfields |= cpu_to_le16(val); - first_bd->data.nbds = 1; + bd->data.bitfields = cpu_to_le16(val); /* We can safely ignore the offset, as it's 0 for XDP */ - BD_SET_UNMAP_ADDR_LEN(first_bd, metadata->mapping + padding, length); + BD_SET_UNMAP_ADDR_LEN(bd, dma + pad, len); - /* Synchronize the buffer back to device, as program [probably] - * has changed it. - */ - dma_sync_single_for_device(&edev->pdev->dev, - metadata->mapping + padding, - length, PCI_DMA_TODEVICE); + xdp = txq->sw_tx_ring.xdp + txq->sw_tx_prod; + xdp->mapping = dma; + xdp->page = page; - txq->sw_tx_ring.xdp[idx].page = metadata->data; - txq->sw_tx_ring.xdp[idx].mapping = metadata->mapping; txq->sw_tx_prod = (txq->sw_tx_prod + 1) % txq->num_tx_buffers; - /* Mark the fastpath for future XDP doorbell */ - fp->xdp_xmit = 1; - return 0; } @@ -362,20 +351,21 @@ int qede_txq_has_work(struct qede_tx_queue *txq) static void qede_xdp_tx_int(struct qede_dev *edev, struct qede_tx_queue *txq) { - u16 hw_bd_cons, idx; + struct sw_tx_xdp *xdp_info, *xdp_arr = txq->sw_tx_ring.xdp; + struct device *dev = &edev->pdev->dev; + u16 hw_bd_cons; hw_bd_cons = le16_to_cpu(*txq->hw_cons_ptr); barrier(); while (hw_bd_cons != qed_chain_get_cons_idx(&txq->tx_pbl)) { - qed_chain_consume(&txq->tx_pbl); - idx = txq->sw_tx_cons; + xdp_info = xdp_arr + txq->sw_tx_cons; - dma_unmap_page(&edev->pdev->dev, - txq->sw_tx_ring.xdp[idx].mapping, - PAGE_SIZE, DMA_BIDIRECTIONAL); - __free_page(txq->sw_tx_ring.xdp[idx].page); + dma_unmap_page(dev, xdp_info->mapping, PAGE_SIZE, + DMA_BIDIRECTIONAL); + __free_page(xdp_info->page); + qed_chain_consume(&txq->tx_pbl); txq->sw_tx_cons = (txq->sw_tx_cons + 1) % txq->num_tx_buffers; txq->xmit_pkts++; } @@ -1064,32 +1054,39 @@ static bool qede_rx_xdp(struct qede_dev *edev, switch (act) { case XDP_TX: /* We need the replacement buffer before transmit. */ - if (qede_alloc_rx_buffer(rxq, true)) { + if (unlikely(qede_alloc_rx_buffer(rxq, true))) { qede_recycle_rx_bd_ring(rxq, 1); + trace_xdp_exception(edev->ndev, prog, act); - return false; + break; } /* Now if there's a transmission problem, we'd still have to * throw current buffer, as replacement was already allocated. */ - if (qede_xdp_xmit(edev, fp, bd, *data_offset, *len)) { - dma_unmap_page(rxq->dev, bd->mapping, - PAGE_SIZE, DMA_BIDIRECTIONAL); + if (unlikely(qede_xdp_xmit(fp->xdp_tx, bd->mapping, + *data_offset, *len, bd->data))) { + dma_unmap_page(rxq->dev, bd->mapping, PAGE_SIZE, + rxq->data_direction); __free_page(bd->data); + trace_xdp_exception(edev->ndev, prog, act); + } else { + dma_sync_single_for_device(rxq->dev, + bd->mapping + *data_offset, + *len, rxq->data_direction); + fp->xdp_xmit |= QEDE_XDP_TX; } /* Regardless, we've consumed an Rx BD */ qede_rx_bd_ring_consume(rxq); - return false; - + break; default: bpf_warn_invalid_xdp_action(act); - /* Fall through */ + fallthrough; case XDP_ABORTED: trace_xdp_exception(edev->ndev, prog, act); - /* Fall through */ + fallthrough; case XDP_DROP: qede_recycle_rx_bd_ring(rxq, cqe->bd_num); } @@ -1353,6 +1350,9 @@ int qede_poll(struct napi_struct *napi, int budget) napi); struct qede_dev *edev = fp->edev; int rx_work_done = 0; + u16 xdp_prod; + + fp->xdp_xmit = 0; if (likely(fp->type & QEDE_FASTPATH_TX)) { int cos; @@ -1380,10 +1380,9 @@ int qede_poll(struct napi_struct *napi, int budget) } } - if (fp->xdp_xmit) { - u16 xdp_prod = qed_chain_get_prod_idx(&fp->xdp_tx->tx_pbl); + if (fp->xdp_xmit & QEDE_XDP_TX) { + xdp_prod = qed_chain_get_prod_idx(&fp->xdp_tx->tx_pbl); - fp->xdp_xmit = 0; fp->xdp_tx->tx_db.data.bd_prod = cpu_to_le16(xdp_prod); qede_update_tx_producer(fp->xdp_tx); }