From: Bartosz Golaszewski <brgl@bgdev.pl>
To: Jakub Kicinski <kuba@kernel.org>
Cc: Rob Herring <robh+dt@kernel.org>,
"David S . Miller" <davem@davemloft.net>,
Matthias Brugger <matthias.bgg@gmail.com>,
John Crispin <john@phrozen.org>,
Sean Wang <sean.wang@mediatek.com>,
Mark Lee <Mark-MC.Lee@mediatek.com>,
Arnd Bergmann <arnd@arndb.de>,
Fabien Parent <fparent@baylibre.com>,
devicetree <devicetree@vger.kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
netdev <netdev@vger.kernel.org>,
Linux ARM <linux-arm-kernel@lists.infradead.org>,
linux-mediatek@lists.infradead.org,
Bartosz Golaszewski <bgolaszewski@baylibre.com>
Subject: Re: [PATCH 06/11] net: ethernet: mtk-eth-mac: new driver
Date: Wed, 6 May 2020 09:09:52 +0200 [thread overview]
Message-ID: <CAMRc=MfmuKd64YaqrkhGFThDZd0_tRecR5H0QLY0cDJWSM-VgQ@mail.gmail.com> (raw)
In-Reply-To: <20200505110447.2404985c@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Hi Jakub,
thanks for the review.
wt., 5 maj 2020 o 20:04 Jakub Kicinski <kuba@kernel.org> napisał(a):
>
> > +/* Represents the actual structure of descriptors used by the MAC. We can
> > + * reuse the same structure for both TX and RX - the layout is the same, only
> > + * the flags differ slightly.
> > + */
> > +struct mtk_mac_ring_desc {
> > + /* Contains both the status flags as well as packet length. */
> > + u32 status;
> > + u32 data_ptr;
> > + u32 vtag;
> > + u32 reserved;
> > +} __aligned(4) __packed;
>
> It will be aligned to 4, because the members are all 4B. And there is
> no possibility of holes. You can safely remove those attrs.
>
I noticed some other drivers whose descriptors are well aligned define
these attributes anyway so I assumed it's a convention. I'll drop them
in v2.
>
> > + status = desc->status;
> > +
> > + if (!(status & MTK_MAC_DESC_BIT_COWN))
> > + return -1;
> > +
> > + desc_data->len = status & MTK_MAC_DESC_MSK_LEN;
> > + desc_data->flags = status & ~MTK_MAC_DESC_MSK_LEN;
> > + desc_data->dma_addr = desc->data_ptr;
> > + desc_data->skb = ring->skbs[ring->tail];
> > +
> > + desc->data_ptr = 0;
> > + desc->status = MTK_MAC_DESC_BIT_COWN;
> > + if (status & MTK_MAC_DESC_BIT_EOR)
> > + desc->status |= MTK_MAC_DESC_BIT_EOR;
> > +
> > + dma_wmb();
>
> What is this separating?
I'll add comments to barriers in v2.
>
> > +/* All processing for TX and RX happens in the napi poll callback. */
> > +static irqreturn_t mtk_mac_handle_irq(int irq, void *data)
> > +{
> > + struct mtk_mac_priv *priv;
> > + struct net_device *ndev;
> > + unsigned int status;
> > +
> > + ndev = data;
> > + priv = netdev_priv(ndev);
> > +
> > + if (netif_running(ndev)) {
> > + mtk_mac_intr_mask_all(priv);
> > + status = mtk_mac_intr_read_and_clear(priv);
> > +
> > + /* RX Complete */
> > + if (status & MTK_MAC_BIT_INT_STS_FNRC)
> > + napi_schedule(&priv->napi);
> > +
> > + /* TX Complete */
> > + if (status & MTK_MAC_BIT_INT_STS_TNTC)
> > + schedule_work(&priv->tx_work);
> > +
> > + /* One of the counter reached 0x8000000 */
> > + if (status & MTK_MAC_REG_INT_STS_MIB_CNT_TH) {
> > + mtk_mac_update_stats(priv);
> > + mtk_mac_reset_counters(priv);
> > + }
> > +
> > + mtk_mac_intr_unmask_all(priv);
>
> Why do you unmask all IRQs here? The usual way to operate is to leave
> TX and RX IRQs masked until NAPI finishes.
>
I actually did it before as the leftover comment says above the
function. Then I thought this way we mask interrupt for a shorter
period of time. I can go back to the previous approach.
> > + }
> > +
> > + return IRQ_HANDLED;
> > +}
>
> > +static int mtk_mac_enable(struct net_device *ndev)
> > +{
> > + /* Reset all counters */
> > + mtk_mac_reset_counters(priv);
>
> This doesn't reset the counters to zero, right?
>
Yes, it does actually. I'll drop it in v2 - it's not necessary.
>
> > +static void mtk_mac_tx_work(struct work_struct *work)
> > +{
> > + struct mtk_mac_priv *priv;
> > + struct mtk_mac_ring *ring;
> > + struct net_device *ndev;
> > + bool wake = false;
> > + int ret;
> > +
> > + priv = container_of(work, struct mtk_mac_priv, tx_work);
> > + ndev = mtk_mac_get_netdev(priv);
> > + ring = &priv->tx_ring;
> > +
> > + for (;;) {
> > + mtk_mac_lock(priv);
> > +
> > + if (!mtk_mac_ring_descs_available(ring)) {
> > + mtk_mac_unlock(priv);
> > + break;
> > + }
> > +
> > + ret = mtk_mac_tx_complete(priv);
> > + mtk_mac_unlock(priv);
> > + if (ret)
> > + break;
> > +
> > + wake = true;
> > + }
> > +
> > + if (wake)
> > + netif_wake_queue(ndev);
>
> This looks racy, if the TX path runs in parallel the queue may have
> already been filled up at the point you wake it up.
>
> > +}
>
> Why do you clean the TX ring from a work rather than from the NAPI
> context?
>
So this was unclear to me, that's why I went with a workqueue. The
budget argument in napi poll is for RX. Should I put some cap on the
number of TX descriptors processed in napi context?
>
> > +static int mtk_mac_receive_packet(struct mtk_mac_priv *priv)
> > +{
> > + struct net_device *ndev = mtk_mac_get_netdev(priv);
> > + struct mtk_mac_ring *ring = &priv->rx_ring;
> > + struct device *dev = mtk_mac_get_dev(priv);
> > + struct mtk_mac_ring_desc_data desc_data;
> > + struct sk_buff *new_skb;
> > + int ret;
> > +
> > + mtk_mac_lock(priv);
> > + ret = mtk_mac_ring_pop_tail(ring, &desc_data);
> > + mtk_mac_unlock(priv);
> > + if (ret)
> > + return -1;
> > +
> > + mtk_mac_dma_unmap_rx(priv, &desc_data);
> > +
> > + if ((desc_data.flags & MTK_MAC_DESC_BIT_RX_CRCE) ||
> > + (desc_data.flags & MTK_MAC_DESC_BIT_RX_OSIZE)) {
> > + /* Error packet -> drop and reuse skb. */
> > + new_skb = desc_data.skb;
> > + goto map_skb;
> > + }
> > +
> > + new_skb = mtk_mac_alloc_skb(ndev);
> > + if (!new_skb) {
> > + netdev_err(ndev, "out of memory for skb\n");
>
> No need for printing, kernel will complain loudly about oom.
>
> > + ndev->stats.rx_dropped++;
> > + new_skb = desc_data.skb;
> > + goto map_skb;
> > + }
> > +
> > + skb_put(desc_data.skb, desc_data.len);
> > + desc_data.skb->ip_summed = CHECKSUM_NONE;
> > + desc_data.skb->protocol = eth_type_trans(desc_data.skb, ndev);
> > + desc_data.skb->dev = ndev;
> > + netif_receive_skb(desc_data.skb);
> > +
> > +map_skb:
> > + desc_data.dma_addr = mtk_mac_dma_map_rx(priv, new_skb);
> > + if (dma_mapping_error(dev, desc_data.dma_addr)) {
> > + dev_kfree_skb(new_skb);
> > + netdev_err(ndev, "DMA mapping error of RX descriptor\n");
> > + return -ENOMEM;
>
> In this case nothing will ever replenish the RX ring right? If we hit
> this condition 128 times the ring will be empty?
>
Indeed. What should I do if this fails though?
I'll address all other issues in v2.
Bart
next prev parent reply other threads:[~2020-05-06 7:10 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-05 14:02 [PATCH 00/11] mediatek: add support for MediaTek Ethernet MAC Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 01/11] dt-bindings: add a binding document for MediaTek PERICFG controller Bartosz Golaszewski
2020-05-13 2:38 ` Rob Herring
2020-05-13 8:09 ` Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 02/11] dt-bindings: new: add yaml bindings for MediaTek Ethernet MAC Bartosz Golaszewski
2020-05-13 2:41 ` Rob Herring
2020-05-05 14:02 ` [PATCH 03/11] net: ethernet: mediatek: rename Kconfig prompt Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 04/11] net: ethernet: mediatek: remove unnecessary spaces from Makefile Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 05/11] net: core: provide devm_register_netdev() Bartosz Golaszewski
2020-05-05 17:31 ` Jakub Kicinski
2020-05-06 6:39 ` Bartosz Golaszewski
2020-05-06 17:12 ` Jakub Kicinski
2020-05-07 9:25 ` Bartosz Golaszewski
2020-05-07 16:53 ` Jakub Kicinski
2020-05-07 17:03 ` Bartosz Golaszewski
2020-05-07 22:56 ` Jakub Kicinski
2020-05-08 5:54 ` Heiner Kallweit
2020-05-08 18:39 ` Bartosz Golaszewski
2020-05-05 19:25 ` Edwin Peer
2020-05-06 6:46 ` Bartosz Golaszewski
2020-05-06 18:20 ` Edwin Peer
2020-05-05 14:02 ` [PATCH 06/11] net: ethernet: mtk-eth-mac: new driver Bartosz Golaszewski
2020-05-05 17:47 ` Andrew Lunn
2020-05-06 7:02 ` Bartosz Golaszewski
2020-05-05 18:04 ` Jakub Kicinski
2020-05-06 7:09 ` Bartosz Golaszewski [this message]
2020-05-06 17:19 ` Jakub Kicinski
2020-05-06 19:16 ` Leon Romanovsky
2020-05-06 19:23 ` Jakub Kicinski
2020-05-07 5:55 ` Leon Romanovsky
2020-05-07 22:50 ` Jakub Kicinski
2020-05-06 19:24 ` Joe Perches
[not found] ` <1588844771.5921.27.camel@mtksdccf07>
2020-05-07 10:50 ` Bartosz Golaszewski
2020-05-07 13:16 ` Andrew Lunn
2020-05-07 17:01 ` Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 07/11] ARM64: dts: mediatek: add pericfg syscon to mt8516.dtsi Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 08/11] ARM64: dts: mediatek: add the ethernet node " Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 09/11] ARM64: dts: mediatek: add an alias for ethernet0 for pumpkin boards Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 10/11] ARM64: dts: mediatek: add ethernet pins " Bartosz Golaszewski
2020-05-05 14:02 ` [PATCH 11/11] ARM64: dts: mediatek: enable ethernet on " Bartosz Golaszewski
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='CAMRc=MfmuKd64YaqrkhGFThDZd0_tRecR5H0QLY0cDJWSM-VgQ@mail.gmail.com' \
--to=brgl@bgdev.pl \
--cc=Mark-MC.Lee@mediatek.com \
--cc=arnd@arndb.de \
--cc=bgolaszewski@baylibre.com \
--cc=davem@davemloft.net \
--cc=devicetree@vger.kernel.org \
--cc=fparent@baylibre.com \
--cc=john@phrozen.org \
--cc=kuba@kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mediatek@lists.infradead.org \
--cc=matthias.bgg@gmail.com \
--cc=netdev@vger.kernel.org \
--cc=robh+dt@kernel.org \
--cc=sean.wang@mediatek.com \
/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).