netdev - Re: [PATCH net-next v2] net: stmmac: Enable TSO on VLANs

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: <CANn89i+11L5=tKsa7V7Aeyxaj6nYGRwy35PAbCRYJ73G+b25sg@mail.gmail.com>
Date: Fri, 14 Jun 2024 08:29:48 +0200
From: Eric Dumazet <edumazet@...gle.com>
To: Furong Xu <0x1207@...il.com>
Cc: "Russell King (Oracle)" <linux@...linux.org.uk>, "David S. Miller" <davem@...emloft.net>, 
	Alexandre Torgue <alexandre.torgue@...s.st.com>, Jose Abreu <joabreu@...opsys.com>, 
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>, 
	Maxime Coquelin <mcoquelin.stm32@...il.com>, Joao Pinto <jpinto@...opsys.com>, 
	Corinna Vinschen <vinschen@...hat.com>, netdev@...r.kernel.org, 
	linux-stm32@...md-mailman.stormreply.com, 
	linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org, 
	xfr@...look.com, rock.xu@....com
Subject: Re: [PATCH net-next v2] net: stmmac: Enable TSO on VLANs

On Fri, Jun 14, 2024 at 8:04 AM Furong Xu <0x1207@...il.com> wrote:
>
> The TSO engine works well when the frames are not VLAN Tagged.
> But it will produce broken segments when frames are VLAN Tagged.
>
> The first segment is all good, while the second segment to the
> last segment are broken, they lack of required VLAN tag.
>
> An example here:
> ========
> // 1st segment of a VLAN Tagged TSO frame, nothing wrong.
> MacSrc > MacDst, ethertype 802.1Q (0x8100), length 1518: vlan 100, p 1, ethertype IPv4 (0x0800), HostA:42643 > HostB:5201: Flags [.], seq 1:1449
>
> // 2nd to last segments of a VLAN Tagged TSO frame, VLAN tag is missing.
> MacSrc > MacDst, ethertype IPv4 (0x0800), length 1514: HostA:42643 > HostB:5201: Flags [.], seq 1449:2897
> MacSrc > MacDst, ethertype IPv4 (0x0800), length 1514: HostA:42643 > HostB:5201: Flags [.], seq 2897:4345
> MacSrc > MacDst, ethertype IPv4 (0x0800), length 1514: HostA:42643 > HostB:5201: Flags [.], seq 4345:5793
> MacSrc > MacDst, ethertype IPv4 (0x0800), length 1514: HostA:42643 > HostB:5201: Flags [P.], seq 5793:7241
>
> // normal VLAN Tagged non-TSO frame, nothing wrong.
> MacSrc > MacDst, ethertype 802.1Q (0x8100), length 1022: vlan 100, p 1, ethertype IPv4 (0x0800), HostA:42643 > HostB:5201: Flags [P.], seq 7241:8193
> MacSrc > MacDst, ethertype 802.1Q (0x8100), length 70: vlan 100, p 1, ethertype IPv4 (0x0800), HostA:42643 > HostB:5201: Flags [F.], seq 8193
> ========
>
> When transmitting VLAN Tagged TSO frames, never insert VLAN tag by HW,
> always insert VLAN tag to SKB payload, then TSO works well on VLANs for
> all MAC cores.
>
> Tested on DWMAC CORE 5.10a, DWMAC CORE 5.20a and DWXGMAC CORE 3.20a
>
> Signed-off-by: Furong Xu <0x1207@...il.com>
> ---
>   Changes in v2:
>     - Use __vlan_hwaccel_push_inside() to insert vlan tag to the payload.
> ---
>  .../net/ethernet/stmicro/stmmac/stmmac_main.c | 27 ++++++++++---------
>  1 file changed, 14 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> index bbedf2a8c60f..e8cbfada63ca 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> @@ -4233,18 +4233,27 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
>  {
>         struct dma_desc *desc, *first, *mss_desc = NULL;
>         struct stmmac_priv *priv = netdev_priv(dev);
> -       int nfrags = skb_shinfo(skb)->nr_frags;
> -       u32 queue = skb_get_queue_mapping(skb);
>         unsigned int first_entry, tx_packets;
>         struct stmmac_txq_stats *txq_stats;
> -       int tmp_pay_len = 0, first_tx;
> +       int tmp_pay_len = 0, first_tx, nfrags;
>         struct stmmac_tx_queue *tx_q;
> -       bool has_vlan, set_ic;
> +       bool set_ic;
>         u8 proto_hdr_len, hdr;
> -       u32 pay_len, mss;
> +       u32 pay_len, mss, queue;
>         dma_addr_t des;
>         int i;
>
> +       /* Always insert VLAN tag to SKB payload for TSO frames.
> +        *
> +        * Never insert VLAN tag by HW, since segments splited by
> +        * TSO engine will be un-tagged by mistake.
> +        */
> +       if (skb_vlan_tag_present(skb))
> +               skb = __vlan_hwaccel_push_inside(skb);

skb can be NULL after this point.

> +
> +       nfrags = skb_shinfo(skb)->nr_frags;
> +       queue = skb_get_queue_mapping(skb);
> +
>         tx_q = &priv->dma_conf.tx_queue[queue];
>         txq_stats = &priv->xstats.txq_stats[queue];
>         first_tx = tx_q->cur_tx;
> @@ -4297,9 +4306,6 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
>                         skb->data_len);
>         }
>
> -       /* Check if VLAN can be inserted by HW */
> -       has_vlan = stmmac_vlan_insert(priv, skb, tx_q);
> -
>         first_entry = tx_q->cur_tx;
>         WARN_ON(tx_q->tx_skbuff[first_entry]);
>
> @@ -4309,9 +4315,6 @@ static netdev_tx_t stmmac_tso_xmit(struct sk_buff *skb, struct net_device *dev)
>                 desc = &tx_q->dma_tx[first_entry];
>         first = desc;
>
> -       if (has_vlan)
> -               stmmac_set_desc_vlan(priv, first, STMMAC_VLAN_INSERT);
> -
>         /* first descriptor: fill Headers on Buf1 */
>         des = dma_map_single(priv->device, skb->data, skb_headlen(skb),
>                              DMA_TO_DEVICE);
> @@ -7678,8 +7681,6 @@ int stmmac_dvr_probe(struct device *device,
>                 ndev->features |= NETIF_F_RXHASH;
>
>         ndev->vlan_features |= ndev->features;
> -       /* TSO doesn't work on VLANs yet */
> -       ndev->vlan_features &= ~NETIF_F_TSO;
>
>         /* MTU range: 46 - hw-specific max */
>         ndev->min_mtu = ETH_ZLEN - ETH_HLEN;
> --
> 2.34.1
>