From patchwork Thu Dec 7 08:12:45 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Yoshihiro Shimoda X-Patchwork-Id: 13482910 X-Patchwork-Delegate: kuba@kernel.org Received: from relmlie6.idc.renesas.com (relmlor2.renesas.com [210.160.252.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id 1E03C110; Thu, 7 Dec 2023 00:12:56 -0800 (PST) X-IronPort-AV: E=Sophos;i="6.04,256,1695654000"; d="scan'208";a="189512039" Received: from unknown (HELO relmlir5.idc.renesas.com) ([10.200.68.151]) by relmlie6.idc.renesas.com with ESMTP; 07 Dec 2023 17:12:54 +0900 Received: from localhost.localdomain (unknown [10.166.13.99]) by relmlir5.idc.renesas.com (Postfix) with ESMTP id 026EE40108EC; Thu, 7 Dec 2023 17:12:54 +0900 (JST) From: Yoshihiro Shimoda To: s.shtylyov@omp.ru, davem@davemloft.net, edumazet@google.com, kuba@kernel.org, pabeni@redhat.com Cc: netdev@vger.kernel.org, linux-renesas-soc@vger.kernel.org, Yoshihiro Shimoda Subject: [PATCH net-next v4 8/9] net: rswitch: Add jumbo frames handling for TX Date: Thu, 7 Dec 2023 17:12:45 +0900 Message-Id: <20231207081246.1557582-9-yoshihiro.shimoda.uh@renesas.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20231207081246.1557582-1-yoshihiro.shimoda.uh@renesas.com> References: <20231207081246.1557582-1-yoshihiro.shimoda.uh@renesas.com> Precedence: bulk X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 X-Patchwork-Delegate: kuba@kernel.org If the driver would like to transmit a jumbo frame like 2KiB or more, it should be split into multiple queues. In the near future, to support this, add handling specific descriptor types F{START,MID,END}. However, such jumbo frames will not happen yet because the maximum MTU size is still default for now. Signed-off-by: Yoshihiro Shimoda --- drivers/net/ethernet/renesas/rswitch.c | 56 +++++++++++++++++++++----- 1 file changed, 46 insertions(+), 10 deletions(-) diff --git a/drivers/net/ethernet/renesas/rswitch.c b/drivers/net/ethernet/renesas/rswitch.c index 36c70131594d..d43f705f410b 100644 --- a/drivers/net/ethernet/renesas/rswitch.c +++ b/drivers/net/ethernet/renesas/rswitch.c @@ -1632,15 +1632,44 @@ static bool rswitch_ext_desc_set(struct rswitch_device *rdev, return true; } +static u8 rswitch_ext_desc_get_die_dt(unsigned int nr_desc, unsigned int index) +{ + if (nr_desc == 1) + return DT_FSINGLE | DIE; + if (index == 0) + return DT_FSTART; + if (nr_desc - 1 == index) + return DT_FEND | DIE; + return DT_FMID; +} + +static u16 rswitch_ext_desc_get_len(u8 die_dt, unsigned int orig_len) +{ + switch (die_dt & DT_MASK) { + case DT_FSINGLE: + case DT_FEND: + return (orig_len % RSWITCH_DESC_BUF_SIZE) ?: RSWITCH_DESC_BUF_SIZE; + case DT_FSTART: + case DT_FMID: + return RSWITCH_DESC_BUF_SIZE; + default: + return 0; + } +} + static netdev_tx_t rswitch_start_xmit(struct sk_buff *skb, struct net_device *ndev) { struct rswitch_device *rdev = netdev_priv(ndev); struct rswitch_gwca_queue *gq = rdev->tx_queue; + dma_addr_t dma_addr, dma_addr_orig; netdev_tx_t ret = NETDEV_TX_OK; struct rswitch_ext_desc *desc; - dma_addr_t dma_addr; + unsigned int i, nr_desc; + u8 die_dt; + u16 len; - if (rswitch_get_num_cur_queues(gq) >= gq->ring_size - 1) { + nr_desc = (skb->len - 1) / RSWITCH_DESC_BUF_SIZE + 1; + if (rswitch_get_num_cur_queues(gq) >= gq->ring_size - nr_desc) { netif_stop_subqueue(ndev, 0); return NETDEV_TX_BUSY; } @@ -1648,25 +1677,32 @@ static netdev_tx_t rswitch_start_xmit(struct sk_buff *skb, struct net_device *nd if (skb_put_padto(skb, ETH_ZLEN)) return ret; - dma_addr = dma_map_single(ndev->dev.parent, skb->data, skb->len, DMA_TO_DEVICE); - if (dma_mapping_error(ndev->dev.parent, dma_addr)) + dma_addr_orig = dma_map_single(ndev->dev.parent, skb->data, skb->len, DMA_TO_DEVICE); + if (dma_mapping_error(ndev->dev.parent, dma_addr_orig)) goto err_kfree; gq->skbs[gq->cur] = skb; - gq->unmap_addrs[gq->cur] = dma_addr; - desc = &gq->tx_ring[gq->cur]; - if (!rswitch_ext_desc_set(rdev, skb, desc, dma_addr, skb->len, DT_FSINGLE | DIE)) - goto err_unmap; + gq->unmap_addrs[gq->cur] = dma_addr_orig; + + /* DT_FSTART should be set at last. So, this is reverse order. */ + for (i = nr_desc; i-- > 0; ) { + desc = &gq->tx_ring[rswitch_next_queue_index(gq, true, i)]; + die_dt = rswitch_ext_desc_get_die_dt(nr_desc, i); + dma_addr = dma_addr_orig + i * RSWITCH_DESC_BUF_SIZE; + len = rswitch_ext_desc_get_len(die_dt, skb->len); + if (!rswitch_ext_desc_set(rdev, skb, desc, dma_addr, len, die_dt)) + goto err_unmap; + } wmb(); /* gq->cur must be incremented after die_dt was set */ - gq->cur = rswitch_next_queue_index(gq, true, 1); + gq->cur = rswitch_next_queue_index(gq, true, nr_desc); rswitch_modify(rdev->addr, GWTRC(gq->index), 0, BIT(gq->index % 32)); return ret; err_unmap: - dma_unmap_single(ndev->dev.parent, dma_addr, skb->len, DMA_TO_DEVICE); + dma_unmap_single(ndev->dev.parent, dma_addr_orig, skb->len, DMA_TO_DEVICE); err_kfree: dev_kfree_skb_any(skb);