Re: [PATCH v7 3/3] net: ax88796c: ASIX AX88796C SPI Ethernet Adapter Driver

From: Lukasz Stelmach <l.stelmach@samsung.com>
To: Jakub Kicinski <kuba@kernel.org>
Cc: "Andrew Lunn" <andrew@lunn.ch>,
	jim.cromie@gmail.com, "Heiner Kallweit" <hkallweit1@gmail.com>,
	"David S. Miller" <davem@davemloft.net>,
	"Rob Herring" <robh+dt@kernel.org>,
	"Kukjin Kim" <kgene@kernel.org>,
	"Krzysztof Kozlowski" <krzk@kernel.org>,
	"Russell King" <linux@armlinux.org.uk>,
	netdev@vger.kernel.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-samsung-soc@vger.kernel.org,
	"Bartłomiej Żolnierkiewicz" <b.zolnierkie@samsung.com>,
	"Marek Szyprowski" <m.szyprowski@samsung.com>
Subject: Re: [PATCH v7 3/3] net: ax88796c: ASIX AX88796C SPI Ethernet Adapter Driver
Date: Wed, 02 Dec 2020 11:46:28 +0100	[thread overview]
Message-ID: <dleftj8sageb97.fsf%l.stelmach@samsung.com> (raw)
In-Reply-To: <20201125132621.628ac98b@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com> (Jakub Kicinski's message of "Wed, 25 Nov 2020 13:26:21 -0800")

[-- Attachment #1: Type: text/plain, Size: 2598 bytes --]

It was <2020-11-25 śro 13:26>, when Jakub Kicinski wrote:
> On Tue, 24 Nov 2020 13:03:30 +0100 Łukasz Stelmach wrote:
>> +static int
>> +ax88796c_start_xmit(struct sk_buff *skb, struct net_device *ndev)
>> +{
>> +	struct ax88796c_device *ax_local = to_ax88796c_device(ndev);
>> +
>> +	skb_queue_tail(&ax_local->tx_wait_q, skb);
>> +	if (skb_queue_len(&ax_local->tx_wait_q) > TX_QUEUE_HIGH_WATER) {
>> +		netif_err(ax_local, tx_queued, ndev,
>> +			  "Too many TX packets in queue %d\n",
>> +			  skb_queue_len(&ax_local->tx_wait_q));
>
> This will probably happen under heavy traffic. No need to print errors,
> it's normal to back pressure.
>

Removed.

>> +		netif_stop_queue(ndev);
>> +	}
>> +
>> +	set_bit(EVENT_TX, &ax_local->flags);
>> +	schedule_work(&ax_local->ax_work);
>> +
>> +	return NETDEV_TX_OK;
>> +}
>> +
>> +static void
>> +ax88796c_skb_return(struct ax88796c_device *ax_local, struct sk_buff *skb,
>> +		    struct rx_header *rxhdr)
>> +{
>> +	struct net_device *ndev = ax_local->ndev;
>> +	int status;
>> +
>> +	do {
>> +		if (!(ndev->features & NETIF_F_RXCSUM))
>> +			break;
>> +
>> +		/* checksum error bit is set */
>> +		if ((rxhdr->flags & RX_HDR3_L3_ERR) ||
>> +		    (rxhdr->flags & RX_HDR3_L4_ERR))
>> +			break;
>> +
>> +		/* Other types may be indicated by more than one bit. */
>> +		if ((rxhdr->flags & RX_HDR3_L4_TYPE_TCP) ||
>> +		    (rxhdr->flags & RX_HDR3_L4_TYPE_UDP))
>> +			skb->ip_summed = CHECKSUM_UNNECESSARY;
>> +	} while (0);
>> +
>> +	ax_local->stats.rx_packets++;
>> +	ax_local->stats.rx_bytes += skb->len;
>> +	skb->dev = ndev;
>> +
>> +	skb->truesize = skb->len + sizeof(struct sk_buff);
>> +	skb->protocol = eth_type_trans(skb, ax_local->ndev);
>> +
>> +	netif_info(ax_local, rx_status, ndev, "< rx, len %zu, type 0x%x\n",
>> +		   skb->len + sizeof(struct ethhdr), skb->protocol);
>> +
>> +	status = netif_rx(skb);
>
> If I'm reading things right this is in process context, so netif_rx_ni()
>

Is it? The stack looks as follows

    ax88796c_skb_return()
    ax88796c_rx_fixup()
    ax88796c_receive()
    ax88796c_process_isr()
    ax88796c_work()

and ax88796c_work() is a scheduled in the system_wq.

>> +	if (status != NET_RX_SUCCESS)
>> +		netif_info(ax_local, rx_err, ndev,
>> +			   "netif_rx status %d\n", status);
>
> Again, it's inadvisable to put per packet prints without any rate
> limiting in the data path.

Even if limmited by the msglvl flag, which is off by default?

-- 
Łukasz Stelmach
Samsung R&D Institute Poland
Samsung Electronics

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]