[v6] app/testpmd: enable GTP header parse and Tx checksum offload
Checks
Commit Message
This patch enables testpmd to forward GTP packet in csum fwd mode.
A GTP header structure (without optional fields and extension header)
is defined in new rte_gtp.h. A parser function in testpmd is added.
GTPU and GTPC packets are both supported, with respective UDP
destination port and GTP message type.
Signed-off-by: Ting Xu <ting.xu@intel.com>
---
v6: complete doxygen documentation.
v5: create rte_gtp.h and move GTP header definition
to it.
v4: move GTP header definition to rte_ether.h
v3: correct coding style issue.
v2: modify commit log
depend on patch: lib/mbuf: add GTP tunnel type flag.
---
app/test-pmd/csumonly.c | 97 +++++++++++++++++++++++++++++++++++----
doc/api/doxy-api-index.md | 1 +
lib/librte_net/Makefile | 2 +-
lib/librte_net/rte_gtp.h | 52 +++++++++++++++++++++
4 files changed, 141 insertions(+), 11 deletions(-)
create mode 100644 lib/librte_net/rte_gtp.h
Comments
On 10/22/2019 5:26 PM, Ting Xu wrote:
> This patch enables testpmd to forward GTP packet in csum fwd mode.
> A GTP header structure (without optional fields and extension header)
> is defined in new rte_gtp.h. A parser function in testpmd is added.
> GTPU and GTPC packets are both supported, with respective UDP
> destination port and GTP message type.
>
> Signed-off-by: Ting Xu <ting.xu@intel.com>
>
> ---
> v6: complete doxygen documentation.
> v5: create rte_gtp.h and move GTP header definition
> to it.
> v4: move GTP header definition to rte_ether.h
> v3: correct coding style issue.
> v2: modify commit log
>
> depend on patch: lib/mbuf: add GTP tunnel type flag.
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
Will wait mbuf patch to be applied to get the patch.
On 10/22/2019 4:45 PM, Ferruh Yigit wrote:
> On 10/22/2019 5:26 PM, Ting Xu wrote:
>> This patch enables testpmd to forward GTP packet in csum fwd mode.
>> A GTP header structure (without optional fields and extension header)
>> is defined in new rte_gtp.h. A parser function in testpmd is added.
>> GTPU and GTPC packets are both supported, with respective UDP
>> destination port and GTP message type.
>>
>> Signed-off-by: Ting Xu <ting.xu@intel.com>
>>
>> ---
>> v6: complete doxygen documentation.
>> v5: create rte_gtp.h and move GTP header definition
>> to it.
>> v4: move GTP header definition to rte_ether.h
>> v3: correct coding style issue.
>> v2: modify commit log
>>
>> depend on patch: lib/mbuf: add GTP tunnel type flag.
>
> Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
>
> Will wait mbuf patch to be applied to get the patch.
>
Applied to dpdk-next-net/master, thanks.
23/10/2019 19:23, Ferruh Yigit:
> On 10/22/2019 4:45 PM, Ferruh Yigit wrote:
> > On 10/22/2019 5:26 PM, Ting Xu wrote:
> >> This patch enables testpmd to forward GTP packet in csum fwd mode.
> >> A GTP header structure (without optional fields and extension header)
> >> is defined in new rte_gtp.h. A parser function in testpmd is added.
> >> GTPU and GTPC packets are both supported, with respective UDP
> >> destination port and GTP message type.
> >>
> >> Signed-off-by: Ting Xu <ting.xu@intel.com>
> >>
> >> ---
> >> v6: complete doxygen documentation.
> >> v5: create rte_gtp.h and move GTP header definition
> >> to it.
> >> v4: move GTP header definition to rte_ether.h
> >> v3: correct coding style issue.
> >> v2: modify commit log
> >>
> >> depend on patch: lib/mbuf: add GTP tunnel type flag.
> >
> > Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
> >
> > Will wait mbuf patch to be applied to get the patch.
> >
>
> Applied to dpdk-next-net/master, thanks.
Sorry, dropped from pull in master.
An update of meson.build is missing and it makes me feel
something else may be missing.
22/10/2019 18:26, Ting Xu:
> This patch enables testpmd to forward GTP packet in csum fwd mode.
> A GTP header structure (without optional fields and extension header)
> is defined in new rte_gtp.h. A parser function in testpmd is added.
> GTPU and GTPC packets are both supported, with respective UDP
> destination port and GTP message type.
>
> Signed-off-by: Ting Xu <ting.xu@intel.com>
>
> ---
> v6: complete doxygen documentation.
> v5: create rte_gtp.h and move GTP header definition
> to it.
> v4: move GTP header definition to rte_ether.h
> v3: correct coding style issue.
> v2: modify commit log
>
> depend on patch: lib/mbuf: add GTP tunnel type flag.
> ---
> app/test-pmd/csumonly.c | 97 +++++++++++++++++++++++++++++++++++----
> doc/api/doxy-api-index.md | 1 +
> lib/librte_net/Makefile | 2 +-
You missed meson.build in addition of Makefile.
> lib/librte_net/rte_gtp.h | 52 +++++++++++++++++++++
> 4 files changed, 141 insertions(+), 11 deletions(-)
> create mode 100644 lib/librte_net/rte_gtp.h
Does it deserve 2 separate patches? GTP header + testpmd use
[...]
> --- /dev/null
> +++ b/lib/librte_net/rte_gtp.h
[...]
> +
> +#endif /* RTE_GTP_H_ */
> +
Please remove this blank line at EOF. Such things are cleaned up sometimes:
http://git.dpdk.org/dpdk/commit/?id=6f3f0acd95
26/10/2019 00:43, Thomas Monjalon:
> 23/10/2019 19:23, Ferruh Yigit:
> > On 10/22/2019 4:45 PM, Ferruh Yigit wrote:
> > > On 10/22/2019 5:26 PM, Ting Xu wrote:
> > >> This patch enables testpmd to forward GTP packet in csum fwd mode.
> > >> A GTP header structure (without optional fields and extension header)
> > >> is defined in new rte_gtp.h. A parser function in testpmd is added.
> > >> GTPU and GTPC packets are both supported, with respective UDP
> > >> destination port and GTP message type.
> > >>
> > >> Signed-off-by: Ting Xu <ting.xu@intel.com>
> > >>
> > >> ---
> > >> v6: complete doxygen documentation.
> > >> v5: create rte_gtp.h and move GTP header definition
> > >> to it.
> > >> v4: move GTP header definition to rte_ether.h
> > >> v3: correct coding style issue.
> > >> v2: modify commit log
> > >>
> > >> depend on patch: lib/mbuf: add GTP tunnel type flag.
> > >
> > > Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
> > >
> > > Will wait mbuf patch to be applied to get the patch.
> > >
> >
> > Applied to dpdk-next-net/master, thanks.
>
> Sorry, dropped from pull in master.
> An update of meson.build is missing and it makes me feel
> something else may be missing.
Changed my mind: I pull it in the master branch with few fixes.
Hi, Thomas,
Sorry for my mistake about missing meson.build. I have fixed it as well as others based on your comments.
However, I am not quite understand your last reply. Do I still need to send a new patch now or not?
Thanks for your patient help!
Best Regards,
Ting Xu
-----Original Message-----
From: Thomas Monjalon <thomas@monjalon.net>
Sent: Saturday, October 26, 2019 6:49 AM
To: Xu, Ting <ting.xu@intel.com>
Cc: dev@dpdk.org; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Wu, Jingjing <jingjing.wu@intel.com>; Iremonger, Bernard <bernard.iremonger@intel.com>; olivier.matz@6wind.com; Yigit, Ferruh <ferruh.yigit@intel.com>
Subject: Re: [dpdk-dev] [PATCH v6] app/testpmd: enable GTP header parse and Tx checksum offload
22/10/2019 18:26, Ting Xu:
> This patch enables testpmd to forward GTP packet in csum fwd mode.
> A GTP header structure (without optional fields and extension header)
> is defined in new rte_gtp.h. A parser function in testpmd is added.
> GTPU and GTPC packets are both supported, with respective UDP
> destination port and GTP message type.
>
> Signed-off-by: Ting Xu <ting.xu@intel.com>
>
> ---
> v6: complete doxygen documentation.
> v5: create rte_gtp.h and move GTP header definition to it.
> v4: move GTP header definition to rte_ether.h
> v3: correct coding style issue.
> v2: modify commit log
>
> depend on patch: lib/mbuf: add GTP tunnel type flag.
> ---
> app/test-pmd/csumonly.c | 97 +++++++++++++++++++++++++++++++++++----
> doc/api/doxy-api-index.md | 1 +
> lib/librte_net/Makefile | 2 +-
You missed meson.build in addition of Makefile.
> lib/librte_net/rte_gtp.h | 52 +++++++++++++++++++++
> 4 files changed, 141 insertions(+), 11 deletions(-) create mode
> 100644 lib/librte_net/rte_gtp.h
Does it deserve 2 separate patches? GTP header + testpmd use
[...]
> --- /dev/null
> +++ b/lib/librte_net/rte_gtp.h
[...]
> +
> +#endif /* RTE_GTP_H_ */
> +
Please remove this blank line at EOF. Such things are cleaned up sometimes:
http://git.dpdk.org/dpdk/commit/?id=6f3f0acd95
28/10/2019 02:35, Xu, Ting:
> Hi, Thomas,
>
> Sorry for my mistake about missing meson.build. I have fixed it as well as others based on your comments.
> However, I am not quite understand your last reply. Do I still need to send a new patch now or not?
No, this is already merged: http://git.dpdk.org/dpdk/commit/?id=d8e5e69f3a
> Thanks for your patient help!
>
> Best Regards,
> Ting Xu
>
> -----Original Message-----
> From: Thomas Monjalon <thomas@monjalon.net>
> Sent: Saturday, October 26, 2019 6:49 AM
> To: Xu, Ting <ting.xu@intel.com>
> Cc: dev@dpdk.org; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Wu, Jingjing <jingjing.wu@intel.com>; Iremonger, Bernard <bernard.iremonger@intel.com>; olivier.matz@6wind.com; Yigit, Ferruh <ferruh.yigit@intel.com>
> Subject: Re: [dpdk-dev] [PATCH v6] app/testpmd: enable GTP header parse and Tx checksum offload
>
> 22/10/2019 18:26, Ting Xu:
> > This patch enables testpmd to forward GTP packet in csum fwd mode.
> > A GTP header structure (without optional fields and extension header)
> > is defined in new rte_gtp.h. A parser function in testpmd is added.
> > GTPU and GTPC packets are both supported, with respective UDP
> > destination port and GTP message type.
> >
> > Signed-off-by: Ting Xu <ting.xu@intel.com>
> >
> > ---
> > v6: complete doxygen documentation.
> > v5: create rte_gtp.h and move GTP header definition to it.
> > v4: move GTP header definition to rte_ether.h
> > v3: correct coding style issue.
> > v2: modify commit log
> >
> > depend on patch: lib/mbuf: add GTP tunnel type flag.
> > ---
> > app/test-pmd/csumonly.c | 97 +++++++++++++++++++++++++++++++++++----
> > doc/api/doxy-api-index.md | 1 +
> > lib/librte_net/Makefile | 2 +-
>
> You missed meson.build in addition of Makefile.
>
> > lib/librte_net/rte_gtp.h | 52 +++++++++++++++++++++
> > 4 files changed, 141 insertions(+), 11 deletions(-) create mode
> > 100644 lib/librte_net/rte_gtp.h
>
> Does it deserve 2 separate patches? GTP header + testpmd use
>
> [...]
> > --- /dev/null
> > +++ b/lib/librte_net/rte_gtp.h
> [...]
> > +
> > +#endif /* RTE_GTP_H_ */
> > +
>
> Please remove this blank line at EOF. Such things are cleaned up sometimes:
> http://git.dpdk.org/dpdk/commit/?id=6f3f0acd95
@@ -36,6 +36,7 @@
#include <rte_tcp.h>
#include <rte_udp.h>
#include <rte_sctp.h>
+#include <rte_gtp.h>
#include <rte_prefetch.h>
#include <rte_string_fns.h>
#include <rte_flow.h>
@@ -179,6 +180,67 @@ parse_ethernet(struct rte_ether_hdr *eth_hdr, struct testpmd_offload_info *info)
}
}
+/*
+ * Parse a GTP protocol header.
+ * No optional fields and next extension header type.
+ */
+static void
+parse_gtp(struct rte_udp_hdr *udp_hdr,
+ struct testpmd_offload_info *info)
+{
+ struct rte_ipv4_hdr *ipv4_hdr;
+ struct rte_ipv6_hdr *ipv6_hdr;
+ struct rte_gtp_hdr *gtp_hdr;
+ uint8_t gtp_len = sizeof(*gtp_hdr);
+ uint8_t ip_ver;
+
+ /* Check udp destination port. */
+ if (udp_hdr->dst_port != _htons(RTE_GTPC_UDP_PORT) &&
+ udp_hdr->src_port != _htons(RTE_GTPC_UDP_PORT) &&
+ udp_hdr->dst_port != _htons(RTE_GTPU_UDP_PORT))
+ return;
+
+ info->is_tunnel = 1;
+ info->outer_ethertype = info->ethertype;
+ info->outer_l2_len = info->l2_len;
+ info->outer_l3_len = info->l3_len;
+ info->outer_l4_proto = info->l4_proto;
+ info->l2_len = 0;
+
+ gtp_hdr = (struct rte_gtp_hdr *)((char *)udp_hdr +
+ sizeof(struct rte_udp_hdr));
+
+ /*
+ * Check message type. If message type is 0xff, it is
+ * a GTP data packet. If not, it is a GTP control packet
+ */
+ if (gtp_hdr->msg_type == 0xff) {
+ ip_ver = *(uint8_t *)((char *)udp_hdr +
+ sizeof(struct rte_udp_hdr) +
+ sizeof(struct rte_gtp_hdr));
+ ip_ver = (ip_ver) & 0xf0;
+
+ if (ip_ver == RTE_GTP_TYPE_IPV4) {
+ ipv4_hdr = (struct rte_ipv4_hdr *)((char *)gtp_hdr +
+ gtp_len);
+ info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
+ parse_ipv4(ipv4_hdr, info);
+ } else if (ip_ver == RTE_GTP_TYPE_IPV6) {
+ ipv6_hdr = (struct rte_ipv6_hdr *)((char *)gtp_hdr +
+ gtp_len);
+ info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
+ parse_ipv6(ipv6_hdr, info);
+ }
+ } else {
+ info->ethertype = 0;
+ info->l4_len = 0;
+ info->l3_len = 0;
+ info->l4_proto = 0;
+ }
+
+ info->l2_len += RTE_ETHER_GTP_HLEN;
+}
+
/* Parse a vxlan header */
static void
parse_vxlan(struct rte_udp_hdr *udp_hdr,
@@ -478,15 +540,22 @@ process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
if (info->outer_l4_proto != IPPROTO_UDP)
return ol_flags;
+ udp_hdr = (struct rte_udp_hdr *)
+ ((char *)outer_l3_hdr + info->outer_l3_len);
+
/* Skip SW outer UDP checksum generation if HW supports it */
if (tx_offloads & DEV_TX_OFFLOAD_OUTER_UDP_CKSUM) {
+ if (info->outer_ethertype == _htons(RTE_ETHER_TYPE_IPV4))
+ udp_hdr->dgram_cksum
+ = rte_ipv4_phdr_cksum(ipv4_hdr, ol_flags);
+ else
+ udp_hdr->dgram_cksum
+ = rte_ipv6_phdr_cksum(ipv6_hdr, ol_flags);
+
ol_flags |= PKT_TX_OUTER_UDP_CKSUM;
return ol_flags;
}
- udp_hdr = (struct rte_udp_hdr *)
- ((char *)outer_l3_hdr + info->outer_l3_len);
-
/* outer UDP checksum is done in software. In the other side, for
* UDP tunneling, like VXLAN or Geneve, outer UDP checksum can be
* set to zero.
@@ -679,6 +748,7 @@ pkt_copy_split(const struct rte_mbuf *pkt)
* UDP|TCP|SCTP
* Ether / (vlan) / outer IP|IP6 / outer UDP / VXLAN-GPE / IP|IP6 /
* UDP|TCP|SCTP
+ * Ether / (vlan) / outer IP / outer UDP / GTP / IP|IP6 / UDP|TCP|SCTP
* Ether / (vlan) / outer IP|IP6 / GRE / Ether / IP|IP6 / UDP|TCP|SCTP
* Ether / (vlan) / outer IP|IP6 / GRE / IP|IP6 / UDP|TCP|SCTP
* Ether / (vlan) / outer IP|IP6 / IP|IP6 / UDP|TCP|SCTP
@@ -787,16 +857,22 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
udp_hdr = (struct rte_udp_hdr *)
((char *)l3_hdr + info.l3_len);
+ parse_gtp(udp_hdr, &info);
+ if (info.is_tunnel) {
+ tx_ol_flags |= PKT_TX_TUNNEL_GTP;
+ goto tunnel_update;
+ }
parse_vxlan_gpe(udp_hdr, &info);
if (info.is_tunnel) {
- tx_ol_flags |= PKT_TX_TUNNEL_VXLAN_GPE;
- } else {
- parse_vxlan(udp_hdr, &info,
- m->packet_type);
- if (info.is_tunnel)
- tx_ol_flags |=
- PKT_TX_TUNNEL_VXLAN;
+ tx_ol_flags |=
+ PKT_TX_TUNNEL_VXLAN_GPE;
+ goto tunnel_update;
}
+ parse_vxlan(udp_hdr, &info,
+ m->packet_type);
+ if (info.is_tunnel)
+ tx_ol_flags |=
+ PKT_TX_TUNNEL_VXLAN;
} else if (info.l4_proto == IPPROTO_GRE) {
struct simple_gre_hdr *gre_hdr;
@@ -815,6 +891,7 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
}
}
+tunnel_update:
/* update l3_hdr and outer_l3_hdr if a tunnel was parsed */
if (info.is_tunnel) {
outer_l3_hdr = l3_hdr;
@@ -97,6 +97,7 @@ The public API headers are grouped by topics:
[SCTP] (@ref rte_sctp.h),
[TCP] (@ref rte_tcp.h),
[UDP] (@ref rte_udp.h),
+ [GTP] (@ref rte_gtp.h),
[GRO] (@ref rte_gro.h),
[GSO] (@ref rte_gso.h),
[frag/reass] (@ref rte_ip_frag.h),
@@ -20,7 +20,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_NET) += rte_arp.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include := rte_ip.h rte_tcp.h rte_udp.h rte_esp.h
SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include += rte_sctp.h rte_icmp.h rte_arp.h
-SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include += rte_ether.h rte_gre.h rte_net.h
+SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include += rte_ether.h rte_gre.h rte_gtp.h rte_net.h
SYMLINK-$(CONFIG_RTE_LIBRTE_NET)-include += rte_net_crc.h rte_mpls.h
include $(RTE_SDK)/mk/rte.lib.mk
new file mode 100644
@@ -0,0 +1,52 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 1982, 1986, 1990, 1993
+ * The Regents of the University of California.
+ * Copyright(c) 2010-2014 Intel Corporation.
+ * All rights reserved.
+ */
+
+#ifndef _RTE_GTP_H_
+#define _RTE_GTP_H_
+
+/**
+ * @file
+ *
+ * GTP-related defines
+ */
+
+#include <stdint.h>
+#include <rte_byteorder.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * Simplified GTP protocol header.
+ * Contains 8-bit header info, 8-bit message type,
+ * 16-bit payload length after mandatory header, 32-bit TEID.
+ * No optional fields and next extension header.
+ */
+struct rte_gtp_hdr {
+ uint8_t gtp_hdr_info; /**< GTP header info */
+ uint8_t msg_type; /**< GTP message type */
+ uint16_t plen; /**< Total payload length */
+ uint32_t teid; /**< Tunnel endpoint ID */
+} __attribute__((__packed__));
+
+/** GTP header length */
+#define RTE_ETHER_GTP_HLEN \
+ (sizeof(struct rte_udp_hdr) + sizeof(struct rte_gtp_hdr))
+/* GTP next protocal type */
+#define RTE_GTP_TYPE_IPV4 0x40 /**< GTP next protocal type IPv4 */
+#define RTE_GTP_TYPE_IPV6 0x60 /**< GTP next protocal type IPv6 */
+/* GTP destination port number */
+#define RTE_GTPC_UDP_PORT 2123 /**< GTP-C UDP destination port */
+#define RTE_GTPU_UDP_PORT 2152 /**< GTP-U UDP destination port */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* RTE_GTP_H_ */
+