From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 9C7A8C433F5 for ; Tue, 16 Nov 2021 05:18:45 +0000 (UTC) Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by mail.kernel.org (Postfix) with ESMTP id D6AB361BFA for ; Tue, 16 Nov 2021 05:18:44 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org D6AB361BFA Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=emumba.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=dpdk.org Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 1677540141; Tue, 16 Nov 2021 06:18:44 +0100 (CET) Received: from mail-ed1-f41.google.com (mail-ed1-f41.google.com [209.85.208.41]) by mails.dpdk.org (Postfix) with ESMTP id 5FC6940040 for ; Tue, 16 Nov 2021 06:18:43 +0100 (CET) Received: by mail-ed1-f41.google.com with SMTP id m20so36401431edc.5 for ; Mon, 15 Nov 2021 21:18:43 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=emumba-com.20210112.gappssmtp.com; s=20210112; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=r1pe6ICw83v5906lLvVBhnWswkg0gvJVhanAHzLDIBg=; b=qDu5CEkvj9dgqbqTWtmNqizlIJQkEzKgrZAKualE2JlNRi/ttjfLz/fBv764x84jcp H7sBXTbCA4IKlslQ1DRgX470V9uKBdgnmiWuVrL1kjGunc3xMSYBrGKPdjmV/5fskHDi x7IlEza/dq9w5yA4JT5IjzCi3j+8xcp17R5AXOix0p64f0HqhwSLwIlV2EyVrcUndeFU rVf05ozySPuQN01LPKWwo7OkPmJO9cz4w/TZGzJ5+ZxDgUCfybvqx1Slg5HODjtnvC+W n21eO4UZEis5yKzpi/hqPVFKOAtRnBbIztj6e+sOnCisRSVXS2ToRVf5kux6bWl0aPES Sdkg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=r1pe6ICw83v5906lLvVBhnWswkg0gvJVhanAHzLDIBg=; b=3nywAlC970x/YQMCFhV8gzFEiqnR3fWlw/DltCkuGrWGvMW6CKBP+QPUz5iF3Yyy93 wdoYX4HPDvi7mNRCbakNHeOBlO7kT5lPQt1IToCWsqvoAL0WkZXf0apwIvbTsezrNnQ0 GZznPncf2xvvvtDJUrB3wbKJKww4V0bzD/vP6WRduqwhzKSrOqwIr1yN4a3UUWeCSQ5M KKDEkNv6sZHZaako3P2FQRV0u7fvK7QN9TRZRtpuNvWy45c0n6hDPoMNk7JgvgUmlV16 fq2hX8dQBY02Xu2HUoAeHW5qOJeUve4KPfc316eqmYzPnuy2VvbAZvbQ+Ptf4J2PE5hb H5ZQ== X-Gm-Message-State: AOAM532DaZqDALGNsSwquSkPOjoHaXCZEJUZDooSbl7ue3cnN0zT32kF nbQRFfvyFz04HJy0fuyfBfh5ypyqdf+L2wkY01i6sbyRTjUi X-Google-Smtp-Source: ABdhPJzEsh5lVP2mD9oRq7hbplu77N8Q2GgC4xFZSGvhdMeNKtcuQHs/xsXhVpe5M3MV/IHxzQUJewrg2xlbr5ABqoc= X-Received: by 2002:a05:6402:34d3:: with SMTP id w19mr6423628edc.35.1637039923005; Mon, 15 Nov 2021 21:18:43 -0800 (PST) MIME-Version: 1.0 References: <20211008155111.125786-1-usama.nadeem@emumba.com> <20211014184322.5148-1-usama.nadeem@emumba.com> In-Reply-To: From: Usama Nadeem Date: Tue, 16 Nov 2021 10:18:31 +0500 Message-ID: Subject: Re: [dpdk-dev] [PATCH v4] examples/l3fwd: ipv4 and udp/tcp cksum verification through software To: "Ananyev, Konstantin" Cc: "thomas@monjalon.net" , "dev@dpdk.org" Content-Type: multipart/alternative; boundary="00000000000018201e05d0e10ff2" X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org --00000000000018201e05d0e10ff2 Content-Type: text/plain; charset="UTF-8" Hi Ananyev, Konstantin , Yes, you understand it right. The approach you discussed, we have been doing the same thing in our patch. At first, we check the HW offload capabilities. > + if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_IPV4_CKSUM) > + local_port_conf.rxmode.offloads |= > + DEV_RX_OFFLOAD_IPV4_CKSUM; > + else { > + l3_sft_cksum = true; > + printf("WARNING: IPV4 checksum offload not available.\n"); > + } If cksum HW offload capability is missing, we call SFT cksum function, by setting a bool variable to true. We will look into setting up the SFT callbacks in the next version of patch. On Thu, Nov 4, 2021 at 6:19 PM Ananyev, Konstantin < konstantin.ananyev@intel.com> wrote: > > checks if ipv4 and udptcp cksum offload capability available > > If not available, cksum is verified through software > > If cksum is corrupt, packet is dropped, rest of the packets > > are forwarded back. > > From what I see right now l3fwd: > a) enables HW RX cksum offload > b) simply ignores HW provided cksum status > Which came as a real surprise to me. > Feel free to correct me if I missed something obvious here. > > So, I think first we need to add missing check first, > even though it might cause some perf drop. > Then make changes to actually check provided by HW status and > when HW doesn't provide such offload do check in SW. > > Another alternative would be to remove request for HW offloads > and document l3fwd that it doesn't check checksums at all, > but I don't think it is a good way. > > > Bugzilla ID:545 > > Signed-off-by: Usama Nadeem > > --- > > examples/l3fwd/l3fwd.h | 6 ++++ > > examples/l3fwd/l3fwd_lpm.c | 72 ++++++++++++++++++++++++++++++++++++-- > > examples/l3fwd/main.c | 33 +++++++++++++++-- > > 3 files changed, 105 insertions(+), 6 deletions(-) > > > > diff --git a/examples/l3fwd/l3fwd.h b/examples/l3fwd/l3fwd.h > > index a808d60247..c2c21a91fb 100644 > > --- a/examples/l3fwd/l3fwd.h > > +++ b/examples/l3fwd/l3fwd.h > > @@ -55,6 +55,8 @@ > > #define L3FWD_HASH_ENTRIES (1024*1024*1) > > #endif > > #define HASH_ENTRY_NUMBER_DEFAULT 4 > > +extern bool l3_sft_cksum; > > +extern bool l4_sft_cksum; > > About the approach itself. > We have similar issue for HW PTYPE recognition - some HW doesn't support > it. > So we check HW capabilities and if required we setup SW RX callbacks to do > determine PTYPE in SW. Note that for EM/LPM we have different callbacks. > I think for cksum checks we can do the same: > check HW capabilities, if they are missing add a new callback that would > calculate/check cksum and set RTE_MBUF_F_RX_*_CKSUM_* flags. > That way it will HW/SW cksum will be transparent for the rest of l3fwd > code. > > About cksums required: for LPM/FIB mode just IPv4 cksum seems enough. > For EM we probably need L4 cksum too, though not sure is it really needed. > Wonder what other people think here? > > > struct mbuf_table { > > uint16_t len; > > @@ -210,6 +212,10 @@ em_main_loop(__rte_unused void *dummy); > > int > > lpm_main_loop(__rte_unused void *dummy); > > > > +int > > +check_software_cksum(struct rte_mbuf **pkts_burst, > > +struct rte_mbuf **pkts_burst_to_send, int nb_rx); > > + > > int > > fib_main_loop(__rte_unused void *dummy); > > > > diff --git a/examples/l3fwd/l3fwd_lpm.c b/examples/l3fwd/l3fwd_lpm.c > > index 232b606b54..ecaf323943 100644 > > --- a/examples/l3fwd/l3fwd_lpm.c > > +++ b/examples/l3fwd/l3fwd_lpm.c > > @@ -26,6 +26,7 @@ > > #include > > #include > > #include > > +#include > > > > #include "l3fwd.h" > > #include "l3fwd_event.h" > > @@ -139,16 +140,65 @@ lpm_get_dst_port_with_ipv4(const struct lcore_conf > *qconf, struct rte_mbuf *pkt, > > #include "l3fwd_lpm.h" > > #endif > > > > + > > +int check_software_cksum(struct rte_mbuf **pkts_burst, > > +struct rte_mbuf **pkts_burst_to_send, int nb_rx) > > +{ > > + int j; > > + int i = 0; > > + struct rte_net_hdr_lens hdr_lens; > > + struct rte_ipv4_hdr *ipv4_hdr; > > + void *l3_hdr; > > + void *l4_hdr; > > + rte_be16_t prev_cksum; > > + int dropped_pkts_udp_tcp = 0; > > + int dropped_pkts_ipv4 = 0; > > + bool dropped; > > + for (j = 0; j < nb_rx; j++) { > > + dropped = false; > > + rte_net_get_ptype(pkts_burst[j], &hdr_lens, > RTE_PTYPE_ALL_MASK); > > + l3_hdr = rte_pktmbuf_mtod_offset(pkts_burst[j], > > + void *, hdr_lens.l2_len); > > + l4_hdr = rte_pktmbuf_mtod_offset(pkts_burst[j], > > + void *, hdr_lens.l2_len + hdr_lens.l3_len); > > + ipv4_hdr = l3_hdr; > > + prev_cksum = ipv4_hdr->hdr_checksum; > > + ipv4_hdr->hdr_checksum = 0; > > + ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr); > > + > > + if (l3_sft_cksum && prev_cksum != ipv4_hdr->hdr_checksum) { > > + rte_pktmbuf_free(pkts_burst[j]); > > + dropped_pkts_ipv4++; > > + dropped = true; > > + } else if (l4_sft_cksum && > > + rte_ipv4_udptcp_cksum_verify > > + (l3_hdr, l4_hdr) != 0) { > > + > > + rte_pktmbuf_free(pkts_burst[j]); > > + dropped_pkts_udp_tcp++; > > + dropped = true; > > + } > > + if (dropped == false) { > > + pkts_burst_to_send[i] = pkts_burst[j]; > > + i++; > > + } > > + > > + } > > + return dropped_pkts_udp_tcp+dropped_pkts_ipv4; > > +} > > + > > /* main processing loop */ > > int > > lpm_main_loop(__rte_unused void *dummy) > > { > > struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; > > + struct rte_mbuf *pkts_burst_to_send[MAX_PKT_BURST]; > > unsigned lcore_id; > > uint64_t prev_tsc, diff_tsc, cur_tsc; > > int i, nb_rx; > > uint16_t portid; > > uint8_t queueid; > > + int dropped; > > struct lcore_conf *qconf; > > const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / > > US_PER_S * BURST_TX_DRAIN_US; > > @@ -209,19 +259,35 @@ lpm_main_loop(__rte_unused void *dummy) > > if (nb_rx == 0) > > continue; > > > > + if (l3_sft_cksum || l4_sft_cksum) { > > + dropped = check_software_cksum(pkts_burst, > > + pkts_burst_to_send, nb_rx); > > + > > + nb_rx = nb_rx-dropped; > > + } > > + > > + > > #if defined RTE_ARCH_X86 || defined __ARM_NEON \ > > || defined RTE_ARCH_PPC_64 > > + if (l3_sft_cksum == false && l4_sft_cksum == false) > > l3fwd_lpm_send_packets(nb_rx, pkts_burst, > > portid, qconf); > > + else > > + l3fwd_lpm_send_packets(nb_rx, pkts_burst_to_send, > > + portid, qconf); > > + > > #else > > - l3fwd_lpm_no_opt_send_packets(nb_rx, pkts_burst, > > + if (l3_sft_cksum == false && l4_sft_cksum == false) > > + l3fwd_lpm_no_opt_send_packets(nb_rx, > pkts_burst, > > portid, qconf); > > + else > > + l3fwd_lpm_no_opt_send_packets(nb_rx, > > + pkts_burst_to_send, portid, qconf); > > + > > #endif /* X86 */ > > } > > - > > cur_tsc = rte_rdtsc(); > > } > > - > > return 0; > > } > > > > diff --git a/examples/l3fwd/main.c b/examples/l3fwd/main.c > > index 00ac267af1..a54aca070d 100644 > > --- a/examples/l3fwd/main.c > > +++ b/examples/l3fwd/main.c > > @@ -61,6 +61,9 @@ static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; > > /**< Ports set in promiscuous mode off by default. */ > > static int promiscuous_on; > > > > +bool l3_sft_cksum; > > +bool l4_sft_cksum; > > + > > /* Select Longest-Prefix, Exact match or Forwarding Information Base. */ > > enum L3FWD_LOOKUP_MODE { > > L3FWD_LOOKUP_DEFAULT, > > @@ -123,7 +126,6 @@ static struct rte_eth_conf port_conf = { > > .mq_mode = ETH_MQ_RX_RSS, > > .max_rx_pkt_len = RTE_ETHER_MAX_LEN, > > .split_hdr_size = 0, > > - .offloads = DEV_RX_OFFLOAD_CHECKSUM, > > }, > > .rx_adv_conf = { > > .rss_conf = { > > @@ -981,6 +983,7 @@ prepare_ptype_parser(uint16_t portid, uint16_t > queueid) > > return 0; > > } > > > > + > > static void > > l3fwd_poll_resource_setup(void) > > { > > @@ -993,7 +996,8 @@ l3fwd_poll_resource_setup(void) > > unsigned int nb_ports; > > unsigned int lcore_id; > > int ret; > > - > > + l3_sft_cksum = false; > > + l4_sft_cksum = false; > > if (check_lcore_params() < 0) > > rte_exit(EXIT_FAILURE, "check_lcore_params failed\n"); > > > > @@ -1034,11 +1038,34 @@ l3fwd_poll_resource_setup(void) > > rte_exit(EXIT_FAILURE, > > "Error during getting device (port %u) > info: %s\n", > > portid, strerror(-ret)); > > - > > if (dev_info.tx_offload_capa & > DEV_TX_OFFLOAD_MBUF_FAST_FREE) > > local_port_conf.txmode.offloads |= > > DEV_TX_OFFLOAD_MBUF_FAST_FREE; > > > > + if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_IPV4_CKSUM) > > + local_port_conf.rxmode.offloads |= > > + DEV_RX_OFFLOAD_IPV4_CKSUM; > > + else { > > + l3_sft_cksum = true; > > + printf("WARNING: IPV4 checksum offload not > available.\n"); > > + } > > + > > + if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_UDP_CKSUM) > > + local_port_conf.rxmode.offloads |= > > + DEV_RX_OFFLOAD_UDP_CKSUM; > > + else { > > + l4_sft_cksum = true; > > + printf("WARNING: UDP checksum offload not > available.\n"); > > + } > > + > > + if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_TCP_CKSUM) > > + local_port_conf.rxmode.offloads |= > > + DEV_RX_OFFLOAD_TCP_CKSUM; > > + else { > > + l4_sft_cksum = true; > > + printf("WARNING: TCP checksum offload not > available.\n"); > > + } > > + > > local_port_conf.rx_adv_conf.rss_conf.rss_hf &= > > dev_info.flow_type_rss_offloads; > > > > -- > > 2.25.1 > > -- -Usama --00000000000018201e05d0e10ff2 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

Hi Ananyev, Konstantin,=C2=A0

Yes, you understand it right.=

The approach yo= u discussed, we have been doing the same thing in our patch. At first, we c= heck the HW offload capabilities.=C2=A0

> + if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_IPV4_CKSUM)

> + local_port_conf.rxmode.offloads |=3D

> += DEV_RX_OFFLOAD_IPV= 4_CKSUM;

> + else {

> + l3_s= ft_cksum =3D true;

> = + printf("WARNING: IPV4 checksum offloa= d not available.\n");

> + <= span style=3D"font-size:8.5pt;font-family:"Courier New";color:rgb= (0,127,0);font-variant-numeric:normal;font-variant-east-asian:normal;vertic= al-align:baseline;white-space:pre-wrap">}

If cksum HW offload capability is missing, we cal= l SFT cksum function, by setting a bool variable to true. We will look into= setting up the SFT callbacks in the next version of patch.



On Thu, Nov 4, 2021 at 6:19= PM Ananyev, Konstantin <konstantin.ananyev@intel.com> wrote:
> checks if ipv4 and udptcp cksum offload c= apability available
> If not available, cksum is verified through software
> If cksum is corrupt, packet is dropped, rest of the packets
> are forwarded back.

>From what I see right now l3fwd:
=C2=A0 =C2=A0a) enables HW RX cksum offload
=C2=A0 =C2=A0b) simply ignores HW provided cksum status
Which came as a real surprise to me.
Feel free to correct me if I missed something obvious here.

So, I think first we need to add missing check first,
even though it might cause some perf drop.
Then make changes to actually check provided by HW status and
when HW doesn't provide such offload do check in SW.

Another alternative would be to remove request for HW offloads
and document l3fwd that it doesn't check checksums at all,
but I don't think it is a good way.

> Bugzilla ID:545
> Signed-off-by: Usama Nadeem <usama.nadeem@emumba.com>
> ---
>=C2=A0 examples/l3fwd/l3fwd.h=C2=A0 =C2=A0 =C2=A0|=C2=A0 6 ++++
>=C2=A0 examples/l3fwd/l3fwd_lpm.c | 72 ++++++++++++++++++++++++++++++++= ++++--
>=C2=A0 examples/l3fwd/main.c=C2=A0 =C2=A0 =C2=A0 | 33 +++++++++++++++--=
>=C2=A0 3 files changed, 105 insertions(+), 6 deletions(-)
>
> diff --git a/examples/l3fwd/l3fwd.h b/examples/l3fwd/l3fwd.h
> index a808d60247..c2c21a91fb 100644
> --- a/examples/l3fwd/l3fwd.h
> +++ b/examples/l3fwd/l3fwd.h
> @@ -55,6 +55,8 @@
>=C2=A0 #define L3FWD_HASH_ENTRIES=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0(1024*1024*1)
>=C2=A0 #endif
>=C2=A0 #define HASH_ENTRY_NUMBER_DEFAULT=C2=A0 =C2=A0 4
> +extern bool l3_sft_cksum;
> +extern bool l4_sft_cksum;

About the approach itself.
We have similar issue for HW PTYPE recognition - some HW doesn't suppor= t it.
So we check HW capabilities and if required we setup SW RX callbacks to do<= br> determine PTYPE in SW. Note that for EM/LPM we have different callbacks. I think for cksum checks we can do the same:
check HW capabilities, if they are missing add a new callback that would calculate/check cksum and set=C2=A0 RTE_MBUF_F_RX_*_CKSUM_* flags.
That way it will HW/SW cksum will be transparent for the rest of l3fwd code= .=C2=A0

About cksums required: for LPM/FIB mode just IPv4 cksum seems enough.
For EM we probably need L4 cksum too, though not sure is it really needed.<= br> Wonder what other people think here?

=C2=A0>=C2=A0 struct mbuf_table {
>=C2=A0 =C2=A0 =C2=A0 =C2=A0uint16_t len;
> @@ -210,6 +212,10 @@ em_main_loop(__rte_unused void *dummy);
>=C2=A0 int
>=C2=A0 lpm_main_loop(__rte_unused void *dummy);
>
> +int
> +check_software_cksum(struct rte_mbuf **pkts_burst,
> +struct rte_mbuf **pkts_burst_to_send, int nb_rx);
> +
>=C2=A0 int
>=C2=A0 fib_main_loop(__rte_unused void *dummy);
>
> diff --git a/examples/l3fwd/l3fwd_lpm.c b/examples/l3fwd/l3fwd_lpm.c > index 232b606b54..ecaf323943 100644
> --- a/examples/l3fwd/l3fwd_lpm.c
> +++ b/examples/l3fwd/l3fwd_lpm.c
> @@ -26,6 +26,7 @@
>=C2=A0 #include <rte_udp.h>
>=C2=A0 #include <rte_lpm.h>
>=C2=A0 #include <rte_lpm6.h>
> +#include <rte_net.h>
>
>=C2=A0 #include "l3fwd.h"
>=C2=A0 #include "l3fwd_event.h"
> @@ -139,16 +140,65 @@ lpm_get_dst_port_with_ipv4(const struct lcore_co= nf *qconf, struct rte_mbuf *pkt,
>=C2=A0 #include "l3fwd_lpm.h"
>=C2=A0 #endif
>
> +
> +int check_software_cksum(struct rte_mbuf **pkts_burst,
> +struct rte_mbuf **pkts_burst_to_send, int nb_rx)
> +{
> +=C2=A0 =C2=A0 =C2=A0int j;
> +=C2=A0 =C2=A0 =C2=A0int i =3D 0;
> +=C2=A0 =C2=A0 =C2=A0struct rte_net_hdr_lens hdr_lens;
> +=C2=A0 =C2=A0 =C2=A0struct rte_ipv4_hdr *ipv4_hdr;
> +=C2=A0 =C2=A0 =C2=A0void *l3_hdr;
> +=C2=A0 =C2=A0 =C2=A0void *l4_hdr;
> +=C2=A0 =C2=A0 =C2=A0rte_be16_t prev_cksum;
> +=C2=A0 =C2=A0 =C2=A0int dropped_pkts_udp_tcp =3D 0;
> +=C2=A0 =C2=A0 =C2=A0int dropped_pkts_ipv4 =3D 0;
> +=C2=A0 =C2=A0 =C2=A0bool dropped;
> +=C2=A0 =C2=A0 =C2=A0for (j =3D 0; j < nb_rx; j++) {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0dropped =3D false; > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0rte_net_get_ptype(pkt= s_burst[j], &hdr_lens, RTE_PTYPE_ALL_MASK);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0l3_hdr =3D rte_pktmbu= f_mtod_offset(pkts_burst[j],
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0void *, hdr_lens.l2_l= en);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0l4_hdr =3D rte_pktmbu= f_mtod_offset(pkts_burst[j],
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0void *, hdr_lens.l2_l= en + hdr_lens.l3_len);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0ipv4_hdr =3D l3_hdr;<= br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0prev_cksum =3D ipv4_h= dr->hdr_checksum;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0ipv4_hdr->hdr_chec= ksum =3D 0;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0ipv4_hdr->hdr_chec= ksum =3D rte_ipv4_cksum(ipv4_hdr);
> +
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (l3_sft_cksum &= ;& prev_cksum !=3D ipv4_hdr->hdr_checksum) {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0rte_pktmbuf_free(pkts_burst[j]);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0dropped_pkts_ipv4++;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0dropped =3D true;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0} else if (l4_sft_cks= um &&
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0rte_ipv4_udptcp_cksum_verify
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0(l3_hdr, l4_hdr) !=3D 0) {
> +
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0rte_pktmbuf_free(pkts_burst[j]);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0dropped_pkts_udp_tcp++;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0dropped =3D true;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (dropped =3D=3D fa= lse) {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0pkts_burst_to_send[i] =3D pkts_burst[j];
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0i++;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}
> +
> +=C2=A0 =C2=A0 =C2=A0}
> +=C2=A0 =C2=A0 =C2=A0return dropped_pkts_udp_tcp+dropped_pkts_ipv4; > +}
> +
>=C2=A0 /* main processing loop */
>=C2=A0 int
>=C2=A0 lpm_main_loop(__rte_unused void *dummy)
>=C2=A0 {
>=C2=A0 =C2=A0 =C2=A0 =C2=A0struct rte_mbuf *pkts_burst[MAX_PKT_BURST];<= br> > +=C2=A0 =C2=A0 =C2=A0struct rte_mbuf *pkts_burst_to_send[MAX_PKT_BURST= ];
>=C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned lcore_id;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0uint64_t prev_tsc, diff_tsc, cur_tsc;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0int i, nb_rx;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0uint16_t portid;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0uint8_t queueid;
> +=C2=A0 =C2=A0 =C2=A0int dropped;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0struct lcore_conf *qconf;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0const uint64_t drain_tsc =3D (rte_get_tsc_hz= () + US_PER_S - 1) /
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0US_PER_S * BURST= _TX_DRAIN_US;
> @@ -209,19 +259,35 @@ lpm_main_loop(__rte_unused void *dummy)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0if (nb_rx =3D=3D 0)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0continue;
>
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0if (l3_sft_cksum || l4_sft_cksum) {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0dropped =3D check_software_cksum(pkts_bu= rst,
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pkts_burst_to_send,=C2=A0 =C2=A0 =C2=A0n= b_rx);
> +
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0nb_rx =3D nb_rx-dropped;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0}
> +
> +
>=C2=A0 #if defined RTE_ARCH_X86 || defined __ARM_NEON \
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 || defined RTE_ARCH_PPC_64
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (l3_sft_cksum =3D= =3D false && l4_sft_cksum =3D=3D false)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0l3fwd_lpm_send_packets(nb_rx, pkts_burst,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0portid, qconf);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0else
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0l3fwd_lpm_send_packets(nb_rx, pkts_burst_to_send,
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0portid, qconf);
> +
>=C2=A0 #else
> -=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0l3fwd_lpm_no_opt_send_packets(nb_rx, pkts_burst,
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0if (l3_sft_cksum =3D=3D false && l4_sft_cksum =3D=3D false)<= br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0l3fwd_lpm_no_opt_send_packets(nb_rx, pkt= s_burst,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0portid, qconf);
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0else
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0l3fwd_lpm_no_opt_send_packets(nb_rx,
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pkts_burst_to_send, portid, qconf);
> +
>=C2=A0 #endif /* X86 */
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}
> -
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cur_tsc =3D rte_= rdtsc();
>=C2=A0 =C2=A0 =C2=A0 =C2=A0}
> -
>=C2=A0 =C2=A0 =C2=A0 =C2=A0return 0;
>=C2=A0 }
>
> diff --git a/examples/l3fwd/main.c b/examples/l3fwd/main.c
> index 00ac267af1..a54aca070d 100644
> --- a/examples/l3fwd/main.c
> +++ b/examples/l3fwd/main.c
> @@ -61,6 +61,9 @@ static uint16_t nb_txd =3D RTE_TEST_TX_DESC_DEFAULT;=
>=C2=A0 /**< Ports set in promiscuous mode off by default. */
>=C2=A0 static int promiscuous_on;
>
> +bool l3_sft_cksum;
> +bool l4_sft_cksum;
> +
>=C2=A0 /* Select Longest-Prefix, Exact match or Forwarding Information = Base. */
>=C2=A0 enum L3FWD_LOOKUP_MODE {
>=C2=A0 =C2=A0 =C2=A0 =C2=A0L3FWD_LOOKUP_DEFAULT,
> @@ -123,7 +126,6 @@ static struct rte_eth_conf port_conf =3D {
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.mq_mode =3D ETH= _MQ_RX_RSS,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.max_rx_pkt_len = =3D RTE_ETHER_MAX_LEN,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.split_hdr_size = =3D 0,
> -=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.offloads =3D DEV_RX_= OFFLOAD_CHECKSUM,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0},
>=C2=A0 =C2=A0 =C2=A0 =C2=A0.rx_adv_conf =3D {
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.rss_conf =3D {<= br> > @@ -981,6 +983,7 @@ prepare_ptype_parser(uint16_t portid, uint16_t que= ueid)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0return 0;
>=C2=A0 }
>
> +
>=C2=A0 static void
>=C2=A0 l3fwd_poll_resource_setup(void)
>=C2=A0 {
> @@ -993,7 +996,8 @@ l3fwd_poll_resource_setup(void)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned int nb_ports;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0unsigned int lcore_id;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0int ret;
> -
> +=C2=A0 =C2=A0 =C2=A0l3_sft_cksum =3D false;
> +=C2=A0 =C2=A0 =C2=A0l4_sft_cksum =3D false;
>=C2=A0 =C2=A0 =C2=A0 =C2=A0if (check_lcore_params() < 0)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0rte_exit(EXIT_FA= ILURE, "check_lcore_params failed\n");
>
> @@ -1034,11 +1038,34 @@ l3fwd_poll_resource_setup(void)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0rte_exit(EXIT_FAILURE,
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0"Error during getting device = (port %u) info: %s\n",
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0portid, strerror(-ret));
> -
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (dev_info.tx_= offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0local_port_conf.txmode.offloads |=3D
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0DEV_TX_OFFLOAD_MBUF_FAST_FREE;
>
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (dev_info.rx_offlo= ad_capa & DEV_RX_OFFLOAD_IPV4_CKSUM)
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0local_port_conf.rxmode.offloads |=3D
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0DEV_RX_OFFLOAD_IPV4_CKSUM;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0else {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0l3_sft_cksum =3D true;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0printf("WARNING: IPV4 checksum offload not available.\n");=
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0}
> +
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (dev_info.rx_offlo= ad_capa & DEV_RX_OFFLOAD_UDP_CKSUM)
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0local_port_conf.rxmode.offloads |=3D
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0DEV_RX_OFFLOAD_UDP_CKSUM;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0else {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0l4_sft_cksum =3D true;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0printf("WARNING: UDP checksum offload not available.\n");<= br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}
> +
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (dev_info.rx_offlo= ad_capa & DEV_RX_OFFLOAD_TCP_CKSUM)
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0local_port_conf.rxmode.offloads |=3D
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0DEV_RX_OFFLOAD_TCP_CKSUM;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0else {
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0l4_sft_cksum =3D true;
> +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0printf("WARNING: TCP checksum offload not available.\n");<= br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}
> +
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0local_port_conf.= rx_adv_conf.rss_conf.rss_hf &=3D
>=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0dev_info.flow_type_rss_offloads;
>
> --
> 2.25.1



--
-Usama
--00000000000018201e05d0e10ff2--