| 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
| ||
|
Date: Mon, 14 Oct 2019 15:27:04 +0200
From: Lorenzo Bianconi <lorenzo@...nel.org>
To: Jesper Dangaard Brouer <brouer@...hat.com>
Cc: netdev@...r.kernel.org, lorenzo.bianconi@...hat.com,
davem@...emloft.net, thomas.petazzoni@...tlin.com,
ilias.apalodimas@...aro.org, matteo.croce@...hat.com,
mw@...ihalf.com
Subject: Re: [PATCH v3 net-next 5/8] net: mvneta: add basic XDP support
> On Mon, 14 Oct 2019 12:49:52 +0200
> Lorenzo Bianconi <lorenzo@...nel.org> wrote:
>
> > Add basic XDP support to mvneta driver for devices that rely on software
> > buffer management. Currently supported verdicts are:
> > - XDP_DROP
> > - XDP_PASS
> > - XDP_REDIRECT
> > - XDP_ABORTED
> >
> > - iptables drop:
> > $iptables -t raw -I PREROUTING -p udp --dport 9 -j DROP
> > $nstat -n && sleep 1 && nstat
> > IpInReceives 151169 0.0
> > IpExtInOctets 6953544 0.0
> > IpExtInNoECTPkts 151165 0.0
> >
> > - XDP_DROP via xdp1
> > $./samples/bpf/xdp1 3
> > proto 0: 421419 pkt/s
> > proto 0: 421444 pkt/s
> > proto 0: 421393 pkt/s
> > proto 0: 421440 pkt/s
> > proto 0: 421184 pkt/s
> >
> > Tested-by: Matteo Croce <mcroce@...hat.com>
> > Signed-off-by: Lorenzo Bianconi <lorenzo@...nel.org>
> > ---
> > drivers/net/ethernet/marvell/mvneta.c | 147 ++++++++++++++++++++++++--
> > 1 file changed, 138 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c
> > index 1722dffe265d..b47a44cf9610 100644
> > --- a/drivers/net/ethernet/marvell/mvneta.c
> > +++ b/drivers/net/ethernet/marvell/mvneta.c
> > @@ -38,6 +38,7 @@
> > #include <net/ipv6.h>
> > #include <net/tso.h>
> > #include <net/page_pool.h>
> > +#include <linux/bpf_trace.h>
> >
> > /* Registers */
> > #define MVNETA_RXQ_CONFIG_REG(q) (0x1400 + ((q) << 2))
> > @@ -323,8 +324,10 @@
> > ETH_HLEN + ETH_FCS_LEN, \
> > cache_line_size())
> >
> > +#define MVNETA_SKB_HEADROOM (max(XDP_PACKET_HEADROOM, NET_SKB_PAD) + \
> > + NET_IP_ALIGN)
> > #define MVNETA_SKB_PAD (SKB_DATA_ALIGN(sizeof(struct skb_shared_info) + \
> > - NET_SKB_PAD))
> > + MVNETA_SKB_HEADROOM))
> > #define MVNETA_SKB_SIZE(len) (SKB_DATA_ALIGN(len) + MVNETA_SKB_PAD)
> > #define MVNETA_MAX_RX_BUF_SIZE (PAGE_SIZE - MVNETA_SKB_PAD)
> >
> > @@ -352,6 +355,11 @@ struct mvneta_statistic {
> > #define T_REG_64 64
> > #define T_SW 1
> >
> > +#define MVNETA_XDP_PASS BIT(0)
> > +#define MVNETA_XDP_CONSUMED BIT(1)
>
> I find it confusing that you call it "consumed" (MVNETA_XDP_CONSUMED),
> because if I follow the code these are all drop-cases that are due to
> errors.
>
> Can we call it MVNETA_XDP_DROPPED?
Hi Jesper,
thx for the review. Sure I will rename it in v4.
Regards,
Lorenzo
>
> I also checked, your XDP_TX patch[8/8], that all the return paths for
> MVNETA_XDP_CONSUMED also leads to drop of the xdp_buff.
>
>
> > +#define MVNETA_XDP_TX BIT(2)
> > +#define MVNETA_XDP_REDIR BIT(3)
> > +
> > static const struct mvneta_statistic mvneta_statistics[] = {
> > { 0x3000, T_REG_64, "good_octets_received", },
> > { 0x3010, T_REG_32, "good_frames_received", },
> > @@ -431,6 +439,8 @@ struct mvneta_port {
> > u32 cause_rx_tx;
> > struct napi_struct napi;
> >
> > + struct bpf_prog *xdp_prog;
> > +
> > /* Core clock */
> > struct clk *clk;
> > /* AXI clock */
> > @@ -1950,11 +1960,51 @@ int mvneta_rx_refill_queue(struct mvneta_port *pp, struct mvneta_rx_queue *rxq)
> > return i;
> > }
> >
> > +static int
> > +mvneta_run_xdp(struct mvneta_port *pp, struct mvneta_rx_queue *rxq,
> > + struct bpf_prog *prog, struct xdp_buff *xdp)
> > +{
> > + u32 ret, act = bpf_prog_run_xdp(prog, xdp);
> > +
> > + switch (act) {
> > + case XDP_PASS:
> > + ret = MVNETA_XDP_PASS;
> > + break;
> > + case XDP_REDIRECT: {
> > + int err;
> > +
> > + err = xdp_do_redirect(pp->dev, xdp, prog);
> > + if (err) {
> > + ret = MVNETA_XDP_CONSUMED;
> > + xdp_return_buff(xdp);
>
> Dropped in case of errors. As this is an error case, I don't mind that
> you use the slower xdp_return_buff().
>
>
> > + } else {
> > + ret = MVNETA_XDP_REDIR;
> > + }
> > + break;
> > + }
> > + default:
> > + bpf_warn_invalid_xdp_action(act);
> > + /* fall through */
> > + case XDP_ABORTED:
> > + trace_xdp_exception(pp->dev, prog, act);
> > + /* fall through */
> > + case XDP_DROP:
> > + page_pool_recycle_direct(rxq->page_pool,
> > + virt_to_head_page(xdp->data));
> > + ret = MVNETA_XDP_CONSUMED;
>
> Also drop case.
>
> > + break;
> > + }
> > +
> > + return ret;
> > +}
> > +
> > static int
> > mvneta_swbm_rx_frame(struct mvneta_port *pp,
> > struct mvneta_rx_desc *rx_desc,
> > struct mvneta_rx_queue *rxq,
> > - struct page *page)
> > + struct xdp_buff *xdp,
> > + struct bpf_prog *xdp_prog,
> > + struct page *page, u32 *xdp_ret)
> > {
> > unsigned char *data = page_address(page);
> > int data_len = -MVNETA_MH_SIZE, len;
> > @@ -1974,7 +2024,26 @@ mvneta_swbm_rx_frame(struct mvneta_port *pp,
> > rx_desc->buf_phys_addr,
> > len, dma_dir);
> >
> > - rxq->skb = build_skb(data, PAGE_SIZE);
> > + xdp->data_hard_start = data;
> > + xdp->data = data + MVNETA_SKB_HEADROOM + MVNETA_MH_SIZE;
> > + xdp->data_end = xdp->data + data_len;
> > + xdp_set_data_meta_invalid(xdp);
> > +
> > + if (xdp_prog) {
> > + u32 ret;
> > +
> > + ret = mvneta_run_xdp(pp, rxq, xdp_prog, xdp);
> > + if (ret != MVNETA_XDP_PASS) {
> > + mvneta_update_stats(pp, 1,
> > + xdp->data_end - xdp->data,
>
> Good, you took into account that data_len cannot be used, as BPF/XDP program could have changed data pointer, thus affecting the length.
>
> > + false);
> > + rx_desc->buf_phys_addr = 0;
> > + *xdp_ret |= ret;
> > + return ret;
> > + }
> > + }
> > +
> > + rxq->skb = build_skb(xdp->data_hard_start, PAGE_SIZE);
> > if (unlikely(!rxq->skb)) {
> > netdev_err(dev,
> > "Can't allocate skb on queue %d\n",
> > @@ -1985,8 +2054,9 @@ mvneta_swbm_rx_frame(struct mvneta_port *pp,
> > }
> > page_pool_release_page(rxq->page_pool, page);
> >
> > - skb_reserve(rxq->skb, MVNETA_MH_SIZE + NET_SKB_PAD);
> > - skb_put(rxq->skb, data_len);
> > + skb_reserve(rxq->skb,
> > + xdp->data - xdp->data_hard_start);
> > + skb_put(rxq->skb, xdp->data_end - xdp->data);
> > mvneta_rx_csum(pp, rx_desc->status, rxq->skb);
> >
> > rxq->left_size = rx_desc->data_size - len;
> > @@ -2020,7 +2090,7 @@ mvneta_swbm_add_rx_fragment(struct mvneta_port *pp,
> > /* refill descriptor with new buffer later */
> > skb_add_rx_frag(rxq->skb,
> > skb_shinfo(rxq->skb)->nr_frags,
> > - page, NET_SKB_PAD, data_len,
> > + page, MVNETA_SKB_HEADROOM, data_len,
> > PAGE_SIZE);
> > }
> > page_pool_release_page(rxq->page_pool, page);
> > @@ -2035,10 +2105,17 @@ static int mvneta_rx_swbm(struct napi_struct *napi,
> > {
> > int rcvd_pkts = 0, rcvd_bytes = 0;
> > int rx_pending, refill, done = 0;
> > + struct bpf_prog *xdp_prog;
> > + struct xdp_buff xdp_buf;
> > + u32 xdp_ret = 0;
> >
> > /* Get number of received packets */
> > rx_pending = mvneta_rxq_busy_desc_num_get(pp, rxq);
> >
> > + rcu_read_lock();
> > + xdp_prog = READ_ONCE(pp->xdp_prog);
> > + xdp_buf.rxq = &rxq->xdp_rxq;
>
> Ok, thanks for following my review comments from last.
>
> > /* Fairness NAPI loop */
> > while (done < budget && done < rx_pending) {
> > struct mvneta_rx_desc *rx_desc = mvneta_rxq_next_desc_get(rxq);
> > @@ -2066,7 +2143,8 @@ static int mvneta_rx_swbm(struct napi_struct *napi,
> > continue;
> > }
> >
> > - err = mvneta_swbm_rx_frame(pp, rx_desc, rxq, page);
> > + err = mvneta_swbm_rx_frame(pp, rx_desc, rxq, &xdp_buf,
> > + xdp_prog, page, &xdp_ret);
> > if (err)
> > continue;
> > } else {
> > @@ -2101,6 +2179,10 @@ static int mvneta_rx_swbm(struct napi_struct *napi,
> > /* clean uncomplete skb pointer in queue */
> > rxq->skb = NULL;
> > }
> > + rcu_read_unlock();
> > +
> > + if (xdp_ret & MVNETA_XDP_REDIR)
> > + xdp_do_flush_map();
> >
> > mvneta_update_stats(pp, rcvd_pkts, rcvd_bytes, false);
> >
> [...]
>
>
>
> --
> Best regards,
> Jesper Dangaard Brouer
> MSc.CS, Principal Kernel Engineer at Red Hat
> LinkedIn: http://www.linkedin.com/in/brouer
Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)
Powered by blists - more mailing lists