netdev - Re: [PATCH net-next 2/2] net: enetc: add support for software TSO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20211007083307.6alnxej5qx5ys62k@skbuf>
Date:   Thu, 7 Oct 2021 08:33:08 +0000
From:   Ioana Ciornei <ioana.ciornei@....com>
To:     Claudiu Manoil <claudiu.manoil@....com>
CC:     "davem@...emloft.net" <davem@...emloft.net>,
        "kuba@...nel.org" <kuba@...nel.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Vladimir Oltean <vladimir.oltean@....com>
Subject: Re: [PATCH net-next 2/2] net: enetc: add support for software TSO

On Thu, Oct 07, 2021 at 07:59:25AM +0000, Claudiu Manoil wrote:
> > -----Original Message-----
> > From: Ioana Ciornei <ioana.ciornei@....com>
> > Sent: Wednesday, October 6, 2021 11:13 PM
> [...]
> > +static int enetc_map_tx_tso_buffs(struct enetc_bdr *tx_ring, struct
> > sk_buff *skb)
> > +{
> > +	int hdr_len, total_len, data_len;
> > +	struct enetc_tx_swbd *tx_swbd;
> > +	union enetc_tx_bd *txbd;
> > +	struct tso_t tso;
> > +	__wsum csum, csum2;
> > +	int count = 0, pos;
> > +	int err, i;
> > +
> > +	/* Check that we have enough BDs for this skb */
> > +	if (enetc_bd_unused(tx_ring) < tso_count_descs(skb)) {
> > +		if (net_ratelimit())
> > +			netdev_err(tx_ring->ndev, "Not enough BDs for TSO!\n");
> > +		return 0;
> > +	}
> > +
> 
> On this path, in case the interface is congested, you will drop the packet in the driver,
> and the stack will think transmission was successful and will continue to deliver skbs
> to the driver. Is this the right thing to do?
> 

Good point. I should have mimicked the non-GSO code path when congestion
occurs and stop the subqueue.

For symmetry I'll also move this check outside of the
enetc_map_tx_tso_buffs() to get the code looking somewhat like this:


	if (skb_is_gso(skb)) {
		if (enetc_bd_unused(tx_ring) < tso_count_descs(skb)) {
			netif_stop_subqueue(ndev, tx_ring->index);
			return NETDEV_TX_BUSY;
		}

		enetc_lock_mdio();
		count = enetc_map_tx_tso_buffs(tx_ring, skb);
		enetc_unlock_mdio();
	} else {
		if (unlikely(skb_shinfo(skb)->nr_frags > ENETC_MAX_SKB_FRAGS))
			if (unlikely(skb_linearize(skb)))
				goto drop_packet_err;

		count = skb_shinfo(skb)->nr_frags + 1; /* fragments + head */
		if (enetc_bd_unused(tx_ring) < ENETC_TXBDS_NEEDED(count)) {
			netif_stop_subqueue(ndev, tx_ring->index);
			return NETDEV_TX_BUSY;
		}

		if (skb->ip_summed == CHECKSUM_PARTIAL) {
			err = skb_csum_hwoffload_help(skb, 0);
			if (err)
				goto drop_packet_err;
		}
		enetc_lock_mdio();
		count = enetc_map_tx_buffs(tx_ring, skb);
		enetc_unlock_mdio();
	}


Ioana