Hi Kumara,
The v3 patch is not complete and it seems to be a patch based on v2.
In addition, did you test the code for tcp4 and tcp6 after your change?
For others, please see replies inline.
Thanks,
Jiayu
> -----Original Message-----
> From: Kumara Parameshwaran <kumaraparamesh92@gmail.com>
> Sent: Friday, June 2, 2023 2:34 PM
> To: Hu, Jiayu <jiayu.hu@intel.com>
> Cc: dev@dpdk.org; Kumara Parameshwaran
> <kumaraparamesh92@gmail.com>
> Subject: [PATCH v3] gro : ipv6 changes to support GRO for TCP/ipv6
>
> The patch adds GRO support for TCP/ipv6 packets. This does not include the
> support for vxlan, udp ipv6 packets.
>
> Signed-off-by: Kumara Parameshwaran <kumaraparamesh92@gmail.com>
> ---
> v1:
> * Changes to support GRO for TCP/ipv6 packets. This does not
> include
> vxlan changes.
> * The GRO is performed only for ipv6 packets that does not contain
> extension headers.
> * The logic for the TCP coalescing remains the same, in ipv6 header
> the source address, destination address, flow label, version fields
> are expected to be the same.
> * Re-organised the code to reuse certain tcp functions for both ipv4
> and
> ipv6 flows.
> v2:
> * Fix comments in gro_tcp6.h header file.
>
> v3:
> * Adderess review comments to fix code duplication for v4 and v6
>
> lib/gro/gro_tcp.c | 160 ++++++++++++++++++++++++
> lib/gro/gro_tcp.h | 63 ++++++++++
> lib/gro/gro_tcp4.c | 255 ++++++++++++---------------------------
> lib/gro/gro_tcp4.h | 18 +--
> lib/gro/gro_tcp6.c | 243 ++++++++++---------------------------
> lib/gro/gro_tcp6.h | 31 +++--
> lib/gro/gro_vxlan_tcp4.c | 18 +--
> lib/gro/meson.build | 1 +
> 8 files changed, 396 insertions(+), 393 deletions(-) create mode 100644
> lib/gro/gro_tcp.c
>
> diff --git a/lib/gro/gro_tcp.c b/lib/gro/gro_tcp.c new file mode 100644 index
> 0000000000..6a5aaada58
> --- /dev/null
> +++ b/lib/gro/gro_tcp.c
> @@ -0,0 +1,160 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2017 Intel Corporation
> + */
> +#include <rte_malloc.h>
> +#include <rte_mbuf.h>
> +#include <rte_ethdev.h>
> +
> +#include "gro_tcp.h"
> +
> +static inline uint32_t
> +find_an_empty_item(struct gro_tcp_item *items,
> + uint32_t table_size)
> +{
> + uint32_t i;
> +
> + for (i = 0; i < table_size; i++)
> + if (items[i].firstseg == NULL)
> + return i;
> + return INVALID_ARRAY_INDEX;
> +}
> +
> +uint32_t
> +insert_new_tcp_item(struct rte_mbuf *pkt,
> + struct gro_tcp_item *items,
> + uint32_t *item_num,
> + uint32_t table_size,
> + uint64_t start_time,
> + uint32_t prev_idx,
> + uint32_t sent_seq,
> + uint16_t ip_id,
> + uint8_t is_atomic)
> +{
> + uint32_t item_idx;
> +
> + item_idx = find_an_empty_item(items, table_size);
> + if (item_idx == INVALID_ARRAY_INDEX)
> + return INVALID_ARRAY_INDEX;
> +
> + items[item_idx].firstseg = pkt;
> + items[item_idx].lastseg = rte_pktmbuf_lastseg(pkt);
> + items[item_idx].start_time = start_time;
> + items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
> + items[item_idx].sent_seq = sent_seq;
> + items[item_idx].ip_id = ip_id;
> + items[item_idx].nb_merged = 1;
> + items[item_idx].is_atomic = is_atomic;
> + (*item_num) += 1;
> +
> + /* if the previous packet exists, chain them together. */
> + if (prev_idx != INVALID_ARRAY_INDEX) {
> + items[item_idx].next_pkt_idx =
> + items[prev_idx].next_pkt_idx;
> + items[prev_idx].next_pkt_idx = item_idx;
> + }
> +
> + return item_idx;
> +}
> +
> +uint32_t
> +delete_tcp_item(struct gro_tcp_item *items, uint32_t item_idx,
> + uint32_t *item_num,
> + uint32_t prev_item_idx)
> +{
> + uint32_t next_idx = items[item_idx].next_pkt_idx;
> +
> + /* NULL indicates an empty item */
> + items[item_idx].firstseg = NULL;
> + (*item_num) -= 1;
> + if (prev_item_idx != INVALID_ARRAY_INDEX)
> + items[prev_item_idx].next_pkt_idx = next_idx;
> +
> + return next_idx;
> +}
> +
> +int32_t
> +gro_tcp_reassemble(struct rte_mbuf *pkt,
> + void *tbl,
> + void *key,
> + int32_t tcp_dl,
> + struct gro_tcp_flow_ops *ops,
> + struct gro_tcp_item *items,
> + uint32_t *item_num,
> + uint32_t table_size,
> + uint16_t ip_id,
> + uint8_t is_atomic,
> + uint64_t start_time)
In general, TCP4 and TCP6 share struct gro_tcp_item and have private flow structures,
i.e., struct gro_tcp4/6_flow, and I like this abstraction. IMO, the code processing
struct gro_tcp_item should be implemented as common functions shared by
gro_tcp4.c and gro_tcp6.c. The code processing struct gro_tcp4/6_flow is tcp4 and
tcp6 dependent and no need to abstract and share.
In gro_tcp_reassemble(), it uses callback functions defined in struct gro_tcp_flow_ops
to provide the different operations on struct gro_tcp4/6_flow. I don't think it's necessary
for abstraction purpose as gro_tcp4/6_flows_ops implementations are alike and it also
introduces extra cost on function calls.
The gro_tcp_reassemble() has two parts: the common part to process struct gro_tcp_item
and the private part to process struct gro_tcp4/6_flow. I think a better way is to remove
gro_tcp_reassemble() and struct gro_tcp_flow_ops, and implement the common part as
an internal function in gro_tcp.c/gro_tcp.h, and make both gro_tcp4/6_reassemble()
implement own private part and invoke the common function to process struct gro_tcp_item.
With this change, there is no callback cost anymore and tcp4/6 can share the common function.
> +{
> + uint32_t item_idx;
> + uint32_t cur_idx;
> + uint32_t prev_idx;
> + struct rte_tcp_hdr *tcp_hdr;
> + int cmp;
> + uint32_t sent_seq;
> +
> + tcp_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_tcp_hdr *, pkt-
> >l2_len + pkt->l3_len);
> + /*
> + * Don't process the packet which has FIN, SYN, RST, PSH, URG, ECE
> + * or CWR set.
> + */
> + if (tcp_hdr->tcp_flags != RTE_TCP_ACK_FLAG)
> + return -1;
> + sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
> +
> + ops->tcp_flow_key_init(key, tcp_hdr);
> +
> + item_idx = ops->tcp_flow_lookup(tbl, key);
> + if (item_idx == INVALID_ARRAY_INDEX) {
> + item_idx = insert_new_tcp_item(pkt, items, item_num,
> table_size, start_time,
> +
> INVALID_ARRAY_INDEX, sent_seq, ip_id,
> + is_atomic);
> + if (item_idx == INVALID_ARRAY_INDEX)
> + return -1;
> + if (ops->tcp_flow_insert(tbl, key, item_idx) ==
> + INVALID_ARRAY_INDEX) {
> + /*
> + * Fail to insert a new flow, so delete the
> + * stored packet.
> + */
> + delete_tcp_item(items, item_idx, item_num,
> INVALID_ARRAY_INDEX);
> + return -1;
> + }
> + return 0;
> + }
> + /*
> + * Check all packets in the flow and try to find a neighbor for
> + * the input packet.
> + */
> + cur_idx = item_idx;
> + prev_idx = cur_idx;
> + do {
> + cmp = check_seq_option(&items[cur_idx], tcp_hdr,
> + sent_seq, ip_id, pkt->l4_len, tcp_dl, 0,
> + is_atomic);
> + if (cmp) {
> + if (merge_two_tcp_packets(&items[cur_idx],
> + pkt, cmp, sent_seq, ip_id, 0))
> + return 1;
> + /*
> + * Fail to merge the two packets, as the packet
> + * length is greater than the max value. Store
> + * the packet into the flow.
> + */
> + if (insert_new_tcp_item(pkt, items, item_num,
> table_size, start_time, cur_idx,
> + sent_seq, ip_id, is_atomic) ==
> + INVALID_ARRAY_INDEX)
> + return -1;
> + return 0;
> + }
> + prev_idx = cur_idx;
> + cur_idx = items[cur_idx].next_pkt_idx;
> + } while (cur_idx != INVALID_ARRAY_INDEX);
> +
> + /* Fail to find a neighbor, so store the packet into the flow. */
> + if (insert_new_tcp_item(pkt, items, item_num, table_size, start_time,
> prev_idx, sent_seq,
> + ip_id, is_atomic) == INVALID_ARRAY_INDEX)
> + return -1;
> +
> + return 0;
> +
> +}
> diff --git a/lib/gro/gro_tcp.h b/lib/gro/gro_tcp.h index
> c5d248a022..202f485c18 100644
> --- a/lib/gro/gro_tcp.h
> +++ b/lib/gro/gro_tcp.h
> @@ -1,6 +1,8 @@
> #ifndef _GRO_TCP_H_
> #define _GRO_TCP_H_
>
> +#define INVALID_ARRAY_INDEX 0xffffffffUL
> +
> #include <rte_tcp.h>
>
> /*
> @@ -14,6 +16,31 @@
> #define INVALID_TCP_HDRLEN(len) \
> (((len) < sizeof(struct rte_tcp_hdr)) || ((len) > MAX_TCP_HLEN))
>
> +struct gro_tcp_flow {
> + struct rte_ether_addr eth_saddr;
> + struct rte_ether_addr eth_daddr;
> + uint32_t recv_ack;
> + uint16_t src_port;
> + uint16_t dst_port;
> +};
> +
> +#define ASSIGN_TCP_FLOW_KEY(k1, k2) \
> + rte_ether_addr_copy(&(k1->eth_saddr), &(k2->eth_saddr)); \
> + rte_ether_addr_copy(&(k1->eth_daddr), &(k2->eth_daddr)); \
> + k2->recv_ack = k1->recv_ack; \
> + k2->src_port = k1->src_port; \
> + k2->dst_port = k1->dst_port;
For multiline macro, it's better to use do{...}while(0) to avoid unexpected errors
in the future.
> +
> +typedef uint32_t (*gro_tcp_flow_lookup)(void *table, void *key);
> +typedef uint32_t (*gro_tcp_flow_insert)(void *table, void *key,
> +uint32_t item_idx); typedef void (*gro_tcp_flow_key_init)(void *key,
> +struct rte_tcp_hdr *tcp_hdr);
> +
> +struct gro_tcp_flow_ops {
> + gro_tcp_flow_lookup tcp_flow_lookup;
> + gro_tcp_flow_insert tcp_flow_insert;
> + gro_tcp_flow_key_init tcp_flow_key_init; };
> +
> struct gro_tcp_item {
> /*
> * The first MBUF segment of the packet. If the value @@ -44,6
> +71,36 @@ struct gro_tcp_item {
> uint8_t is_atomic;
> };
>
> +uint32_t
> +insert_new_tcp_item(struct rte_mbuf *pkt,
> + struct gro_tcp_item *items,
> + uint32_t *item_num,
> + uint32_t table_size,
> + uint64_t start_time,
> + uint32_t prev_idx,
> + uint32_t sent_seq,
> + uint16_t ip_id,
> + uint8_t is_atomic);
> +
> +uint32_t
> +delete_tcp_item(struct gro_tcp_item *items,
> + uint32_t item_idx,
> + uint32_t *item_num,
> + uint32_t prev_item_idx);
> +
> +int32_t
> +gro_tcp_reassemble(struct rte_mbuf *pkt,
> + void *tbl,
> + void *key,
> + int32_t tcp_dl,
> + struct gro_tcp_flow_ops *ops,
> + struct gro_tcp_item *items,
> + uint32_t *item_num,
> + uint32_t table_size,
> + uint16_t ip_id,
> + uint8_t is_atomic,
> + uint64_t start_time);
> +
> /*
> * Merge two TCP packets without updating checksums.
> * If cmp is larger than 0, append the new packet to the @@ -152,4 +209,10
> @@ check_seq_option(struct gro_tcp_item *item,
> return 0;
> }
>
> +static inline int
> +is_same_tcp_flow(struct gro_tcp_flow *k1, struct gro_tcp_flow *k2) {
> + return (!memcmp(k1, k2, sizeof(struct gro_tcp_flow))); }
> +
> #endif
On Tue, Jun 6, 2023 at 10:05 AM Hu, Jiayu <jiayu.hu@intel.com> wrote:
> Hi Kumara,
>
> The v3 patch is not complete and it seems to be a patch based on v2.
> In addition, did you test the code for tcp4 and tcp6 after your change?
>
>> Will make sure that next patch contains the entire diff.
>>
> For others, please see replies inline.
>
> Thanks,
> Jiayu
>
> > -----Original Message-----
> > From: Kumara Parameshwaran <kumaraparamesh92@gmail.com>
> > Sent: Friday, June 2, 2023 2:34 PM
> > To: Hu, Jiayu <jiayu.hu@intel.com>
> > Cc: dev@dpdk.org; Kumara Parameshwaran
> > <kumaraparamesh92@gmail.com>
> > Subject: [PATCH v3] gro : ipv6 changes to support GRO for TCP/ipv6
> >
> > The patch adds GRO support for TCP/ipv6 packets. This does not include
> the
> > support for vxlan, udp ipv6 packets.
> >
> > Signed-off-by: Kumara Parameshwaran <kumaraparamesh92@gmail.com>
> > ---
> > v1:
> > * Changes to support GRO for TCP/ipv6 packets. This does not
> > include
> > vxlan changes.
> > * The GRO is performed only for ipv6 packets that does not contain
> > extension headers.
> > * The logic for the TCP coalescing remains the same, in ipv6 header
> > the source address, destination address, flow label, version
> fields
> > are expected to be the same.
> > * Re-organised the code to reuse certain tcp functions for both
> ipv4
> > and
> > ipv6 flows.
> > v2:
> > * Fix comments in gro_tcp6.h header file.
> >
> > v3:
> > * Adderess review comments to fix code duplication for v4 and v6
> >
> > lib/gro/gro_tcp.c | 160 ++++++++++++++++++++++++
> > lib/gro/gro_tcp.h | 63 ++++++++++
> > lib/gro/gro_tcp4.c | 255 ++++++++++++---------------------------
> > lib/gro/gro_tcp4.h | 18 +--
> > lib/gro/gro_tcp6.c | 243 ++++++++++---------------------------
> > lib/gro/gro_tcp6.h | 31 +++--
> > lib/gro/gro_vxlan_tcp4.c | 18 +--
> > lib/gro/meson.build | 1 +
> > 8 files changed, 396 insertions(+), 393 deletions(-) create mode 100644
> > lib/gro/gro_tcp.c
> >
> > diff --git a/lib/gro/gro_tcp.c b/lib/gro/gro_tcp.c new file mode 100644
> index
> > 0000000000..6a5aaada58
> > --- /dev/null
> > +++ b/lib/gro/gro_tcp.c
> > @@ -0,0 +1,160 @@
> > +/* SPDX-License-Identifier: BSD-3-Clause
> > + * Copyright(c) 2017 Intel Corporation
> > + */
> > +#include <rte_malloc.h>
> > +#include <rte_mbuf.h>
> > +#include <rte_ethdev.h>
> > +
> > +#include "gro_tcp.h"
> > +
> > +static inline uint32_t
> > +find_an_empty_item(struct gro_tcp_item *items,
> > + uint32_t table_size)
> > +{
> > + uint32_t i;
> > +
> > + for (i = 0; i < table_size; i++)
> > + if (items[i].firstseg == NULL)
> > + return i;
> > + return INVALID_ARRAY_INDEX;
> > +}
> > +
> > +uint32_t
> > +insert_new_tcp_item(struct rte_mbuf *pkt,
> > + struct gro_tcp_item *items,
> > + uint32_t *item_num,
> > + uint32_t table_size,
> > + uint64_t start_time,
> > + uint32_t prev_idx,
> > + uint32_t sent_seq,
> > + uint16_t ip_id,
> > + uint8_t is_atomic)
> > +{
> > + uint32_t item_idx;
> > +
> > + item_idx = find_an_empty_item(items, table_size);
> > + if (item_idx == INVALID_ARRAY_INDEX)
> > + return INVALID_ARRAY_INDEX;
> > +
> > + items[item_idx].firstseg = pkt;
> > + items[item_idx].lastseg = rte_pktmbuf_lastseg(pkt);
> > + items[item_idx].start_time = start_time;
> > + items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
> > + items[item_idx].sent_seq = sent_seq;
> > + items[item_idx].ip_id = ip_id;
> > + items[item_idx].nb_merged = 1;
> > + items[item_idx].is_atomic = is_atomic;
> > + (*item_num) += 1;
> > +
> > + /* if the previous packet exists, chain them together. */
> > + if (prev_idx != INVALID_ARRAY_INDEX) {
> > + items[item_idx].next_pkt_idx =
> > + items[prev_idx].next_pkt_idx;
> > + items[prev_idx].next_pkt_idx = item_idx;
> > + }
> > +
> > + return item_idx;
> > +}
> > +
> > +uint32_t
> > +delete_tcp_item(struct gro_tcp_item *items, uint32_t item_idx,
> > + uint32_t *item_num,
> > + uint32_t prev_item_idx)
> > +{
> > + uint32_t next_idx = items[item_idx].next_pkt_idx;
> > +
> > + /* NULL indicates an empty item */
> > + items[item_idx].firstseg = NULL;
> > + (*item_num) -= 1;
> > + if (prev_item_idx != INVALID_ARRAY_INDEX)
> > + items[prev_item_idx].next_pkt_idx = next_idx;
> > +
> > + return next_idx;
> > +}
> > +
> > +int32_t
> > +gro_tcp_reassemble(struct rte_mbuf *pkt,
> > + void *tbl,
> > + void *key,
> > + int32_t tcp_dl,
> > + struct gro_tcp_flow_ops *ops,
> > + struct gro_tcp_item *items,
> > + uint32_t *item_num,
> > + uint32_t table_size,
> > + uint16_t ip_id,
> > + uint8_t is_atomic,
> > + uint64_t start_time)
>
> In general, TCP4 and TCP6 share struct gro_tcp_item and have private flow
> structures,
> i.e., struct gro_tcp4/6_flow, and I like this abstraction. IMO, the code
> processing
> struct gro_tcp_item should be implemented as common functions shared by
> gro_tcp4.c and gro_tcp6.c. The code processing struct gro_tcp4/6_flow is
> tcp4 and
> tcp6 dependent and no need to abstract and share.
>
> In gro_tcp_reassemble(), it uses callback functions defined in struct
> gro_tcp_flow_ops
> to provide the different operations on struct gro_tcp4/6_flow. I don't
> think it's necessary
> for abstraction purpose as gro_tcp4/6_flows_ops implementations are alike
> and it also
> introduces extra cost on function calls.
>
>> Sure, would make the changes accordingly.
>>
>
> The gro_tcp_reassemble() has two parts: the common part to process struct
> gro_tcp_item
> and the private part to process struct gro_tcp4/6_flow. I think a better
> way is to remove
> gro_tcp_reassemble() and struct gro_tcp_flow_ops, and implement the common
> part as
> an internal function in gro_tcp.c/gro_tcp.h, and make both
> gro_tcp4/6_reassemble()
> implement own private part and invoke the common function to process
> struct gro_tcp_item.
> With this change, there is no callback cost anymore and tcp4/6 can share
> the common function.
>
>> Sure, got it.
>>
>> > +{
> > + uint32_t item_idx;
> > + uint32_t cur_idx;
> > + uint32_t prev_idx;
> > + struct rte_tcp_hdr *tcp_hdr;
> > + int cmp;
> > + uint32_t sent_seq;
> > +
> > + tcp_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_tcp_hdr *, pkt-
> > >l2_len + pkt->l3_len);
> > + /*
> > + * Don't process the packet which has FIN, SYN, RST, PSH, URG, ECE
> > + * or CWR set.
> > + */
> > + if (tcp_hdr->tcp_flags != RTE_TCP_ACK_FLAG)
> > + return -1;
> > + sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
> > +
> > + ops->tcp_flow_key_init(key, tcp_hdr);
> > +
> > + item_idx = ops->tcp_flow_lookup(tbl, key);
> > + if (item_idx == INVALID_ARRAY_INDEX) {
> > + item_idx = insert_new_tcp_item(pkt, items, item_num,
> > table_size, start_time,
> > +
> > INVALID_ARRAY_INDEX, sent_seq, ip_id,
> > + is_atomic);
> > + if (item_idx == INVALID_ARRAY_INDEX)
> > + return -1;
> > + if (ops->tcp_flow_insert(tbl, key, item_idx) ==
> > + INVALID_ARRAY_INDEX) {
> > + /*
> > + * Fail to insert a new flow, so delete the
> > + * stored packet.
> > + */
> > + delete_tcp_item(items, item_idx, item_num,
> > INVALID_ARRAY_INDEX);
> > + return -1;
> > + }
> > + return 0;
> > + }
> > + /*
> > + * Check all packets in the flow and try to find a neighbor for
> > + * the input packet.
> > + */
> > + cur_idx = item_idx;
> > + prev_idx = cur_idx;
> > + do {
> > + cmp = check_seq_option(&items[cur_idx], tcp_hdr,
> > + sent_seq, ip_id, pkt->l4_len, tcp_dl, 0,
> > + is_atomic);
> > + if (cmp) {
> > + if (merge_two_tcp_packets(&items[cur_idx],
> > + pkt, cmp, sent_seq, ip_id,
> 0))
> > + return 1;
> > + /*
> > + * Fail to merge the two packets, as the packet
> > + * length is greater than the max value. Store
> > + * the packet into the flow.
> > + */
> > + if (insert_new_tcp_item(pkt, items, item_num,
> > table_size, start_time, cur_idx,
> > + sent_seq, ip_id,
> is_atomic) ==
> > + INVALID_ARRAY_INDEX)
> > + return -1;
> > + return 0;
> > + }
> > + prev_idx = cur_idx;
> > + cur_idx = items[cur_idx].next_pkt_idx;
> > + } while (cur_idx != INVALID_ARRAY_INDEX);
> > +
> > + /* Fail to find a neighbor, so store the packet into the flow. */
> > + if (insert_new_tcp_item(pkt, items, item_num, table_size,
> start_time,
> > prev_idx, sent_seq,
> > + ip_id, is_atomic) == INVALID_ARRAY_INDEX)
> > + return -1;
> > +
> > + return 0;
> > +
> > +}
> > diff --git a/lib/gro/gro_tcp.h b/lib/gro/gro_tcp.h index
> > c5d248a022..202f485c18 100644
> > --- a/lib/gro/gro_tcp.h
> > +++ b/lib/gro/gro_tcp.h
> > @@ -1,6 +1,8 @@
> > #ifndef _GRO_TCP_H_
> > #define _GRO_TCP_H_
> >
> > +#define INVALID_ARRAY_INDEX 0xffffffffUL
> > +
> > #include <rte_tcp.h>
> >
> > /*
> > @@ -14,6 +16,31 @@
> > #define INVALID_TCP_HDRLEN(len) \
> > (((len) < sizeof(struct rte_tcp_hdr)) || ((len) > MAX_TCP_HLEN))
> >
> > +struct gro_tcp_flow {
> > + struct rte_ether_addr eth_saddr;
> > + struct rte_ether_addr eth_daddr;
> > + uint32_t recv_ack;
> > + uint16_t src_port;
> > + uint16_t dst_port;
> > +};
> > +
> > +#define ASSIGN_TCP_FLOW_KEY(k1, k2) \
> > + rte_ether_addr_copy(&(k1->eth_saddr), &(k2->eth_saddr)); \
> > + rte_ether_addr_copy(&(k1->eth_daddr), &(k2->eth_daddr)); \
> > + k2->recv_ack = k1->recv_ack; \
> > + k2->src_port = k1->src_port; \
> > + k2->dst_port = k1->dst_port;
>
> For multiline macro, it's better to use do{...}while(0) to avoid
> unexpected errors
> in the future.
>
> > +
> > +typedef uint32_t (*gro_tcp_flow_lookup)(void *table, void *key);
> > +typedef uint32_t (*gro_tcp_flow_insert)(void *table, void *key,
> > +uint32_t item_idx); typedef void (*gro_tcp_flow_key_init)(void *key,
> > +struct rte_tcp_hdr *tcp_hdr);
> > +
> > +struct gro_tcp_flow_ops {
> > + gro_tcp_flow_lookup tcp_flow_lookup;
> > + gro_tcp_flow_insert tcp_flow_insert;
> > + gro_tcp_flow_key_init tcp_flow_key_init; };
> > +
> > struct gro_tcp_item {
> > /*
> > * The first MBUF segment of the packet. If the value @@ -44,6
> > +71,36 @@ struct gro_tcp_item {
> > uint8_t is_atomic;
> > };
> >
> > +uint32_t
> > +insert_new_tcp_item(struct rte_mbuf *pkt,
> > + struct gro_tcp_item *items,
> > + uint32_t *item_num,
> > + uint32_t table_size,
> > + uint64_t start_time,
> > + uint32_t prev_idx,
> > + uint32_t sent_seq,
> > + uint16_t ip_id,
> > + uint8_t is_atomic);
> > +
> > +uint32_t
> > +delete_tcp_item(struct gro_tcp_item *items,
> > + uint32_t item_idx,
> > + uint32_t *item_num,
> > + uint32_t prev_item_idx);
> > +
> > +int32_t
> > +gro_tcp_reassemble(struct rte_mbuf *pkt,
> > + void *tbl,
> > + void *key,
> > + int32_t tcp_dl,
> > + struct gro_tcp_flow_ops *ops,
> > + struct gro_tcp_item *items,
> > + uint32_t *item_num,
> > + uint32_t table_size,
> > + uint16_t ip_id,
> > + uint8_t is_atomic,
> > + uint64_t start_time);
> > +
> > /*
> > * Merge two TCP packets without updating checksums.
> > * If cmp is larger than 0, append the new packet to the @@ -152,4
> +209,10
> > @@ check_seq_option(struct gro_tcp_item *item,
> > return 0;
> > }
> >
> > +static inline int
> > +is_same_tcp_flow(struct gro_tcp_flow *k1, struct gro_tcp_flow *k2) {
> > + return (!memcmp(k1, k2, sizeof(struct gro_tcp_flow))); }
> > +
> > #endif
>
>
new file mode 100644
@@ -0,0 +1,160 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Intel Corporation
+ */
+#include <rte_malloc.h>
+#include <rte_mbuf.h>
+#include <rte_ethdev.h>
+
+#include "gro_tcp.h"
+
+static inline uint32_t
+find_an_empty_item(struct gro_tcp_item *items,
+ uint32_t table_size)
+{
+ uint32_t i;
+
+ for (i = 0; i < table_size; i++)
+ if (items[i].firstseg == NULL)
+ return i;
+ return INVALID_ARRAY_INDEX;
+}
+
+uint32_t
+insert_new_tcp_item(struct rte_mbuf *pkt,
+ struct gro_tcp_item *items,
+ uint32_t *item_num,
+ uint32_t table_size,
+ uint64_t start_time,
+ uint32_t prev_idx,
+ uint32_t sent_seq,
+ uint16_t ip_id,
+ uint8_t is_atomic)
+{
+ uint32_t item_idx;
+
+ item_idx = find_an_empty_item(items, table_size);
+ if (item_idx == INVALID_ARRAY_INDEX)
+ return INVALID_ARRAY_INDEX;
+
+ items[item_idx].firstseg = pkt;
+ items[item_idx].lastseg = rte_pktmbuf_lastseg(pkt);
+ items[item_idx].start_time = start_time;
+ items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
+ items[item_idx].sent_seq = sent_seq;
+ items[item_idx].ip_id = ip_id;
+ items[item_idx].nb_merged = 1;
+ items[item_idx].is_atomic = is_atomic;
+ (*item_num) += 1;
+
+ /* if the previous packet exists, chain them together. */
+ if (prev_idx != INVALID_ARRAY_INDEX) {
+ items[item_idx].next_pkt_idx =
+ items[prev_idx].next_pkt_idx;
+ items[prev_idx].next_pkt_idx = item_idx;
+ }
+
+ return item_idx;
+}
+
+uint32_t
+delete_tcp_item(struct gro_tcp_item *items, uint32_t item_idx,
+ uint32_t *item_num,
+ uint32_t prev_item_idx)
+{
+ uint32_t next_idx = items[item_idx].next_pkt_idx;
+
+ /* NULL indicates an empty item */
+ items[item_idx].firstseg = NULL;
+ (*item_num) -= 1;
+ if (prev_item_idx != INVALID_ARRAY_INDEX)
+ items[prev_item_idx].next_pkt_idx = next_idx;
+
+ return next_idx;
+}
+
+int32_t
+gro_tcp_reassemble(struct rte_mbuf *pkt,
+ void *tbl,
+ void *key,
+ int32_t tcp_dl,
+ struct gro_tcp_flow_ops *ops,
+ struct gro_tcp_item *items,
+ uint32_t *item_num,
+ uint32_t table_size,
+ uint16_t ip_id,
+ uint8_t is_atomic,
+ uint64_t start_time)
+{
+ uint32_t item_idx;
+ uint32_t cur_idx;
+ uint32_t prev_idx;
+ struct rte_tcp_hdr *tcp_hdr;
+ int cmp;
+ uint32_t sent_seq;
+
+ tcp_hdr = rte_pktmbuf_mtod_offset(pkt, struct rte_tcp_hdr *, pkt->l2_len + pkt->l3_len);
+ /*
+ * Don't process the packet which has FIN, SYN, RST, PSH, URG, ECE
+ * or CWR set.
+ */
+ if (tcp_hdr->tcp_flags != RTE_TCP_ACK_FLAG)
+ return -1;
+ sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
+
+ ops->tcp_flow_key_init(key, tcp_hdr);
+
+ item_idx = ops->tcp_flow_lookup(tbl, key);
+ if (item_idx == INVALID_ARRAY_INDEX) {
+ item_idx = insert_new_tcp_item(pkt, items, item_num, table_size, start_time,
+ INVALID_ARRAY_INDEX, sent_seq, ip_id,
+ is_atomic);
+ if (item_idx == INVALID_ARRAY_INDEX)
+ return -1;
+ if (ops->tcp_flow_insert(tbl, key, item_idx) ==
+ INVALID_ARRAY_INDEX) {
+ /*
+ * Fail to insert a new flow, so delete the
+ * stored packet.
+ */
+ delete_tcp_item(items, item_idx, item_num, INVALID_ARRAY_INDEX);
+ return -1;
+ }
+ return 0;
+ }
+ /*
+ * Check all packets in the flow and try to find a neighbor for
+ * the input packet.
+ */
+ cur_idx = item_idx;
+ prev_idx = cur_idx;
+ do {
+ cmp = check_seq_option(&items[cur_idx], tcp_hdr,
+ sent_seq, ip_id, pkt->l4_len, tcp_dl, 0,
+ is_atomic);
+ if (cmp) {
+ if (merge_two_tcp_packets(&items[cur_idx],
+ pkt, cmp, sent_seq, ip_id, 0))
+ return 1;
+ /*
+ * Fail to merge the two packets, as the packet
+ * length is greater than the max value. Store
+ * the packet into the flow.
+ */
+ if (insert_new_tcp_item(pkt, items, item_num, table_size, start_time, cur_idx,
+ sent_seq, ip_id, is_atomic) ==
+ INVALID_ARRAY_INDEX)
+ return -1;
+ return 0;
+ }
+ prev_idx = cur_idx;
+ cur_idx = items[cur_idx].next_pkt_idx;
+ } while (cur_idx != INVALID_ARRAY_INDEX);
+
+ /* Fail to find a neighbor, so store the packet into the flow. */
+ if (insert_new_tcp_item(pkt, items, item_num, table_size, start_time, prev_idx, sent_seq,
+ ip_id, is_atomic) == INVALID_ARRAY_INDEX)
+ return -1;
+
+ return 0;
+
+}
@@ -1,6 +1,8 @@
#ifndef _GRO_TCP_H_
#define _GRO_TCP_H_
+#define INVALID_ARRAY_INDEX 0xffffffffUL
+
#include <rte_tcp.h>
/*
@@ -14,6 +16,31 @@
#define INVALID_TCP_HDRLEN(len) \
(((len) < sizeof(struct rte_tcp_hdr)) || ((len) > MAX_TCP_HLEN))
+struct gro_tcp_flow {
+ struct rte_ether_addr eth_saddr;
+ struct rte_ether_addr eth_daddr;
+ uint32_t recv_ack;
+ uint16_t src_port;
+ uint16_t dst_port;
+};
+
+#define ASSIGN_TCP_FLOW_KEY(k1, k2) \
+ rte_ether_addr_copy(&(k1->eth_saddr), &(k2->eth_saddr)); \
+ rte_ether_addr_copy(&(k1->eth_daddr), &(k2->eth_daddr)); \
+ k2->recv_ack = k1->recv_ack; \
+ k2->src_port = k1->src_port; \
+ k2->dst_port = k1->dst_port;
+
+typedef uint32_t (*gro_tcp_flow_lookup)(void *table, void *key);
+typedef uint32_t (*gro_tcp_flow_insert)(void *table, void *key, uint32_t item_idx);
+typedef void (*gro_tcp_flow_key_init)(void *key, struct rte_tcp_hdr *tcp_hdr);
+
+struct gro_tcp_flow_ops {
+ gro_tcp_flow_lookup tcp_flow_lookup;
+ gro_tcp_flow_insert tcp_flow_insert;
+ gro_tcp_flow_key_init tcp_flow_key_init;
+};
+
struct gro_tcp_item {
/*
* The first MBUF segment of the packet. If the value
@@ -44,6 +71,36 @@ struct gro_tcp_item {
uint8_t is_atomic;
};
+uint32_t
+insert_new_tcp_item(struct rte_mbuf *pkt,
+ struct gro_tcp_item *items,
+ uint32_t *item_num,
+ uint32_t table_size,
+ uint64_t start_time,
+ uint32_t prev_idx,
+ uint32_t sent_seq,
+ uint16_t ip_id,
+ uint8_t is_atomic);
+
+uint32_t
+delete_tcp_item(struct gro_tcp_item *items,
+ uint32_t item_idx,
+ uint32_t *item_num,
+ uint32_t prev_item_idx);
+
+int32_t
+gro_tcp_reassemble(struct rte_mbuf *pkt,
+ void *tbl,
+ void *key,
+ int32_t tcp_dl,
+ struct gro_tcp_flow_ops *ops,
+ struct gro_tcp_item *items,
+ uint32_t *item_num,
+ uint32_t table_size,
+ uint16_t ip_id,
+ uint8_t is_atomic,
+ uint64_t start_time);
+
/*
* Merge two TCP packets without updating checksums.
* If cmp is larger than 0, append the new packet to the
@@ -152,4 +209,10 @@ check_seq_option(struct gro_tcp_item *item,
return 0;
}
+static inline int
+is_same_tcp_flow(struct gro_tcp_flow *k1, struct gro_tcp_flow *k2)
+{
+ return (!memcmp(k1, k2, sizeof(struct gro_tcp_flow)));
+}
+
#endif
@@ -72,18 +72,6 @@ gro_tcp4_tbl_destroy(void *tbl)
rte_free(tcp_tbl);
}
-static inline uint32_t
-find_an_empty_item(struct gro_tcp4_tbl *tbl)
-{
- uint32_t i;
- uint32_t max_item_num = tbl->max_item_num;
-
- for (i = 0; i < max_item_num; i++)
- if (tbl->items[i].firstseg == NULL)
- return i;
- return INVALID_ARRAY_INDEX;
-}
-
static inline uint32_t
find_an_empty_flow(struct gro_tcp4_tbl *tbl)
{
@@ -97,98 +85,76 @@ find_an_empty_flow(struct gro_tcp4_tbl *tbl)
}
static inline uint32_t
-insert_new_item(struct gro_tcp4_tbl *tbl,
- struct rte_mbuf *pkt,
- uint64_t start_time,
- uint32_t prev_idx,
- uint32_t sent_seq,
- uint16_t ip_id,
- uint8_t is_atomic)
+gro_tcp4_insert_flow(void *tbl,
+ void *key,
+ uint32_t item_idx)
{
- uint32_t item_idx;
+ struct gro_tcp4_tbl *tcp4_tbl = (struct gro_tcp4_tbl *)tbl;
+ struct tcp4_flow_key *src_key = (struct tcp4_flow_key *)key;
+ struct tcp4_flow_key *dst_key;
+ struct gro_tcp_flow *src_flow;
+ struct gro_tcp_flow *dst_flow;
+ uint32_t flow_idx;
- item_idx = find_an_empty_item(tbl);
- if (item_idx == INVALID_ARRAY_INDEX)
+ flow_idx = find_an_empty_flow(tcp4_tbl);
+ if (unlikely(flow_idx == INVALID_ARRAY_INDEX))
return INVALID_ARRAY_INDEX;
- tbl->items[item_idx].firstseg = pkt;
- tbl->items[item_idx].lastseg = rte_pktmbuf_lastseg(pkt);
- tbl->items[item_idx].start_time = start_time;
- tbl->items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
- tbl->items[item_idx].sent_seq = sent_seq;
- tbl->items[item_idx].ip_id = ip_id;
- tbl->items[item_idx].nb_merged = 1;
- tbl->items[item_idx].is_atomic = is_atomic;
- tbl->item_num++;
-
- /* if the previous packet exists, chain them together. */
- if (prev_idx != INVALID_ARRAY_INDEX) {
- tbl->items[item_idx].next_pkt_idx =
- tbl->items[prev_idx].next_pkt_idx;
- tbl->items[prev_idx].next_pkt_idx = item_idx;
- }
+ dst_key = &(tcp4_tbl->flows[flow_idx].key);
+ dst_flow = &dst_key->tcp_flow;
+ src_flow = &src_key->tcp_flow;
- return item_idx;
-}
+ ASSIGN_TCP_FLOW_KEY(src_flow, dst_flow)
-static inline uint32_t
-delete_item(struct gro_tcp4_tbl *tbl, uint32_t item_idx,
- uint32_t prev_item_idx)
-{
- uint32_t next_idx = tbl->items[item_idx].next_pkt_idx;
+ dst_key->ip_src_addr = src_key->ip_src_addr;
+ dst_key->ip_dst_addr = src_key->ip_dst_addr;
- /* NULL indicates an empty item */
- tbl->items[item_idx].firstseg = NULL;
- tbl->item_num--;
- if (prev_item_idx != INVALID_ARRAY_INDEX)
- tbl->items[prev_item_idx].next_pkt_idx = next_idx;
+ tcp4_tbl->flows[flow_idx].start_index = item_idx;
+ tcp4_tbl->flow_num++;
- return next_idx;
+ return flow_idx;
}
-static inline uint32_t
-insert_new_flow(struct gro_tcp4_tbl *tbl,
- struct tcp4_flow_key *src,
- uint32_t item_idx)
-{
- struct tcp4_flow_key *dst;
- uint32_t flow_idx;
-
- flow_idx = find_an_empty_flow(tbl);
- if (unlikely(flow_idx == INVALID_ARRAY_INDEX))
- return INVALID_ARRAY_INDEX;
-
- dst = &(tbl->flows[flow_idx].key);
-
- rte_ether_addr_copy(&(src->eth_saddr), &(dst->eth_saddr));
- rte_ether_addr_copy(&(src->eth_daddr), &(dst->eth_daddr));
- dst->ip_src_addr = src->ip_src_addr;
- dst->ip_dst_addr = src->ip_dst_addr;
- dst->recv_ack = src->recv_ack;
- dst->src_port = src->src_port;
- dst->dst_port = src->dst_port;
- tbl->flows[flow_idx].start_index = item_idx;
- tbl->flow_num++;
+static uint32_t
+gro_tcp4_flow_lookup(void *tbl, void *key)
+{
+ struct gro_tcp4_tbl *tcp4_tbl = (struct gro_tcp4_tbl *)tbl;
+ struct tcp4_flow_key *tcp4_key = (struct tcp4_flow_key *)key;
+ uint32_t max_flow_num = tcp4_tbl->max_flow_num;
+ uint32_t remaining_flow_num = tcp4_tbl->flow_num;
+ uint32_t i;
- return flow_idx;
+ /* Search for a matched flow. */
+ for (i = 0; i < max_flow_num && remaining_flow_num; i++) {
+ if (tcp4_tbl->flows[i].start_index != INVALID_ARRAY_INDEX) {
+ if (is_same_tcp4_flow(tcp4_tbl->flows[i].key, *tcp4_key)) {
+ return i;
+ }
+ remaining_flow_num--;
+ }
+ }
+ return INVALID_ARRAY_INDEX;
}
-/*
- * update the packet length for the flushed packet.
- */
-static inline void
-update_header(struct gro_tcp_item *item)
+static void
+gro_tcp4_flow_key_init(void *key, struct rte_tcp_hdr *tcp_hdr)
{
- struct rte_ipv4_hdr *ipv4_hdr;
- struct rte_mbuf *pkt = item->firstseg;
+ struct tcp4_flow_key *tcp4_key = (struct tcp4_flow_key *)key;
+ struct gro_tcp_flow *tcp_flow = &tcp4_key->tcp_flow;
+
+ tcp_flow->src_port = tcp_hdr->src_port;
+ tcp_flow->dst_port = tcp_hdr->dst_port;
+ tcp_flow->recv_ack = tcp_hdr->recv_ack;
- ipv4_hdr = (struct rte_ipv4_hdr *)(rte_pktmbuf_mtod(pkt, char *) +
- pkt->l2_len);
- ipv4_hdr->total_length = rte_cpu_to_be_16(pkt->pkt_len -
- pkt->l2_len);
}
+struct gro_tcp_flow_ops gro_tcp4_flows_ops = {
+ .tcp_flow_lookup = gro_tcp4_flow_lookup,
+ .tcp_flow_insert = gro_tcp4_insert_flow,
+ .tcp_flow_key_init = gro_tcp4_flow_key_init
+};
+
int32_t
gro_tcp4_reassemble(struct rte_mbuf *pkt,
struct gro_tcp4_tbl *tbl,
@@ -196,17 +162,10 @@ gro_tcp4_reassemble(struct rte_mbuf *pkt,
{
struct rte_ether_hdr *eth_hdr;
struct rte_ipv4_hdr *ipv4_hdr;
- struct rte_tcp_hdr *tcp_hdr;
- uint32_t sent_seq;
int32_t tcp_dl;
uint16_t ip_id, hdr_len, frag_off, ip_tlen;
uint8_t is_atomic;
-
struct tcp4_flow_key key;
- uint32_t cur_idx, prev_idx, item_idx;
- uint32_t i, max_flow_num, remaining_flow_num;
- int cmp;
- uint8_t find;
/*
* Don't process the packet whose TCP header length is greater
@@ -217,21 +176,13 @@ gro_tcp4_reassemble(struct rte_mbuf *pkt,
eth_hdr = rte_pktmbuf_mtod(pkt, struct rte_ether_hdr *);
ipv4_hdr = (struct rte_ipv4_hdr *)((char *)eth_hdr + pkt->l2_len);
- tcp_hdr = (struct rte_tcp_hdr *)((char *)ipv4_hdr + pkt->l3_len);
- hdr_len = pkt->l2_len + pkt->l3_len + pkt->l4_len;
-
- /*
- * Don't process the packet which has FIN, SYN, RST, PSH, URG, ECE
- * or CWR set.
- */
- if (tcp_hdr->tcp_flags != RTE_TCP_ACK_FLAG)
- return -1;
/* trim the tail padding bytes */
ip_tlen = rte_be_to_cpu_16(ipv4_hdr->total_length);
if (pkt->pkt_len > (uint32_t)(ip_tlen + pkt->l2_len))
rte_pktmbuf_trim(pkt, pkt->pkt_len - ip_tlen - pkt->l2_len);
+ hdr_len = pkt->l2_len + pkt->l3_len + pkt->l4_len;
/*
* Don't process the packet whose payload length is less than or
* equal to 0.
@@ -240,6 +191,10 @@ gro_tcp4_reassemble(struct rte_mbuf *pkt,
if (tcp_dl <= 0)
return -1;
+ rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.tcp_flow.eth_saddr));
+ rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.tcp_flow.eth_daddr));
+ key.ip_src_addr = ipv4_hdr->src_addr;
+ key.ip_dst_addr = ipv4_hdr->dst_addr;
/*
* Save IPv4 ID for the packet whose DF bit is 0. For the packet
* whose DF bit is 1, IPv4 ID is ignored.
@@ -247,87 +202,25 @@ gro_tcp4_reassemble(struct rte_mbuf *pkt,
frag_off = rte_be_to_cpu_16(ipv4_hdr->fragment_offset);
is_atomic = (frag_off & RTE_IPV4_HDR_DF_FLAG) == RTE_IPV4_HDR_DF_FLAG;
ip_id = is_atomic ? 0 : rte_be_to_cpu_16(ipv4_hdr->packet_id);
- sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
- rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.eth_saddr));
- rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.eth_daddr));
- key.ip_src_addr = ipv4_hdr->src_addr;
- key.ip_dst_addr = ipv4_hdr->dst_addr;
- key.src_port = tcp_hdr->src_port;
- key.dst_port = tcp_hdr->dst_port;
- key.recv_ack = tcp_hdr->recv_ack;
-
- /* Search for a matched flow. */
- max_flow_num = tbl->max_flow_num;
- remaining_flow_num = tbl->flow_num;
- find = 0;
- for (i = 0; i < max_flow_num && remaining_flow_num; i++) {
- if (tbl->flows[i].start_index != INVALID_ARRAY_INDEX) {
- if (is_same_tcp4_flow(tbl->flows[i].key, key)) {
- find = 1;
- break;
- }
- remaining_flow_num--;
- }
- }
-
- /*
- * Fail to find a matched flow. Insert a new flow and store the
- * packet into the flow.
- */
- if (find == 0) {
- item_idx = insert_new_item(tbl, pkt, start_time,
- INVALID_ARRAY_INDEX, sent_seq, ip_id,
- is_atomic);
- if (item_idx == INVALID_ARRAY_INDEX)
- return -1;
- if (insert_new_flow(tbl, &key, item_idx) ==
- INVALID_ARRAY_INDEX) {
- /*
- * Fail to insert a new flow, so delete the
- * stored packet.
- */
- delete_item(tbl, item_idx, INVALID_ARRAY_INDEX);
- return -1;
- }
- return 0;
- }
-
- /*
- * Check all packets in the flow and try to find a neighbor for
- * the input packet.
- */
- cur_idx = tbl->flows[i].start_index;
- prev_idx = cur_idx;
- do {
- cmp = check_seq_option(&(tbl->items[cur_idx]), tcp_hdr,
- sent_seq, ip_id, pkt->l4_len, tcp_dl, 0,
- is_atomic);
- if (cmp) {
- if (merge_two_tcp_packets(&(tbl->items[cur_idx]),
- pkt, cmp, sent_seq, ip_id, 0))
- return 1;
- /*
- * Fail to merge the two packets, as the packet
- * length is greater than the max value. Store
- * the packet into the flow.
- */
- if (insert_new_item(tbl, pkt, start_time, cur_idx,
- sent_seq, ip_id, is_atomic) ==
- INVALID_ARRAY_INDEX)
- return -1;
- return 0;
- }
- prev_idx = cur_idx;
- cur_idx = tbl->items[cur_idx].next_pkt_idx;
- } while (cur_idx != INVALID_ARRAY_INDEX);
+ return gro_tcp_reassemble(pkt, tbl, &key, tcp_dl, &gro_tcp4_flows_ops, tbl->items,
+ &tbl->item_num, tbl->max_item_num,
+ ip_id, is_atomic, start_time);
+}
- /* Fail to find a neighbor, so store the packet into the flow. */
- if (insert_new_item(tbl, pkt, start_time, prev_idx, sent_seq,
- ip_id, is_atomic) == INVALID_ARRAY_INDEX)
- return -1;
+/*
+ * update the packet length for the flushed packet.
+ */
+static inline void
+update_header(struct gro_tcp_item *item)
+{
+ struct rte_ipv4_hdr *ipv4_hdr;
+ struct rte_mbuf *pkt = item->firstseg;
- return 0;
+ ipv4_hdr = (struct rte_ipv4_hdr *)(rte_pktmbuf_mtod(pkt, char *) +
+ pkt->l2_len);
+ ipv4_hdr->total_length = rte_cpu_to_be_16(pkt->pkt_len -
+ pkt->l2_len);
}
uint16_t
@@ -354,7 +247,7 @@ gro_tcp4_tbl_timeout_flush(struct gro_tcp4_tbl *tbl,
* Delete the packet and get the next
* packet in the flow.
*/
- j = delete_item(tbl, j, INVALID_ARRAY_INDEX);
+ j = delete_tcp_item(tbl->items, j, &tbl->item_num, INVALID_ARRAY_INDEX);
tbl->flows[i].start_index = j;
if (j == INVALID_ARRAY_INDEX)
tbl->flow_num--;
@@ -5,19 +5,15 @@
#ifndef _GRO_TCP4_H_
#define _GRO_TCP4_H_
-#define INVALID_ARRAY_INDEX 0xffffffffUL
+#include <gro_tcp.h>
+
#define GRO_TCP4_TBL_MAX_ITEM_NUM (1024UL * 1024UL)
/* Header fields representing a TCP/IPv4 flow */
struct tcp4_flow_key {
- struct rte_ether_addr eth_saddr;
- struct rte_ether_addr eth_daddr;
+ struct gro_tcp_flow tcp_flow;
uint32_t ip_src_addr;
uint32_t ip_dst_addr;
-
- uint32_t recv_ack;
- uint16_t src_port;
- uint16_t dst_port;
};
struct gro_tcp4_flow {
@@ -143,13 +139,9 @@ uint32_t gro_tcp4_tbl_pkt_count(void *tbl);
static inline int
is_same_tcp4_flow(struct tcp4_flow_key k1, struct tcp4_flow_key k2)
{
- return (rte_is_same_ether_addr(&k1.eth_saddr, &k2.eth_saddr) &&
- rte_is_same_ether_addr(&k1.eth_daddr, &k2.eth_daddr) &&
- (k1.ip_src_addr == k2.ip_src_addr) &&
+ return ((k1.ip_src_addr == k2.ip_src_addr) &&
(k1.ip_dst_addr == k2.ip_dst_addr) &&
- (k1.recv_ack == k2.recv_ack) &&
- (k1.src_port == k2.src_port) &&
- (k1.dst_port == k2.dst_port));
+ is_same_tcp_flow(&k1.tcp_flow, &k2.tcp_flow));
}
#endif
@@ -72,18 +72,6 @@ gro_tcp6_tbl_destroy(void *tbl)
rte_free(tcp_tbl);
}
-static inline uint32_t
-find_an_empty_item(struct gro_tcp6_tbl *tbl)
-{
- uint32_t i;
- uint32_t max_item_num = tbl->max_item_num;
-
- for (i = 0; i < max_item_num; i++)
- if (tbl->items[i].firstseg == NULL)
- return i;
- return INVALID_ARRAY_INDEX;
-}
-
static inline uint32_t
find_an_empty_flow(struct gro_tcp6_tbl *tbl)
{
@@ -97,79 +85,32 @@ find_an_empty_flow(struct gro_tcp6_tbl *tbl)
}
static inline uint32_t
-insert_new_item(struct gro_tcp6_tbl *tbl,
- struct rte_mbuf *pkt,
- uint64_t start_time,
- uint32_t prev_idx,
- uint32_t sent_seq,
- uint8_t is_atomic)
-{
- uint32_t item_idx;
-
- item_idx = find_an_empty_item(tbl);
- if (item_idx == INVALID_ARRAY_INDEX)
- return INVALID_ARRAY_INDEX;
-
- tbl->items[item_idx].firstseg = pkt;
- tbl->items[item_idx].lastseg = rte_pktmbuf_lastseg(pkt);
- tbl->items[item_idx].start_time = start_time;
- tbl->items[item_idx].next_pkt_idx = INVALID_ARRAY_INDEX;
- tbl->items[item_idx].sent_seq = sent_seq;
- tbl->items[item_idx].nb_merged = 1;
- tbl->items[item_idx].is_atomic = is_atomic;
- tbl->item_num++;
-
- /* if the previous packet exists, chain them together. */
- if (prev_idx != INVALID_ARRAY_INDEX) {
- tbl->items[item_idx].next_pkt_idx =
- tbl->items[prev_idx].next_pkt_idx;
- tbl->items[prev_idx].next_pkt_idx = item_idx;
- }
-
- return item_idx;
-}
-
-static inline uint32_t
-delete_item(struct gro_tcp6_tbl *tbl, uint32_t item_idx,
- uint32_t prev_item_idx)
-{
- uint32_t next_idx = tbl->items[item_idx].next_pkt_idx;
-
- /* NULL indicates an empty item */
- tbl->items[item_idx].firstseg = NULL;
- tbl->item_num--;
- if (prev_item_idx != INVALID_ARRAY_INDEX)
- tbl->items[prev_item_idx].next_pkt_idx = next_idx;
-
- return next_idx;
-}
-
-static inline uint32_t
-insert_new_flow(struct gro_tcp6_tbl *tbl,
- struct tcp6_flow_key *src,
- rte_be32_t vtc_flow,
+gro_tcp6_insert_flow(void *tbl,
+ void *key,
uint32_t item_idx)
{
- struct tcp6_flow_key *dst;
+ struct gro_tcp6_tbl *tcp6_tbl = (struct gro_tcp6_tbl *)tbl;
+ struct tcp6_flow_key *dst_key;
+ struct tcp6_flow_key *src_key = (struct tcp6_flow_key *)key;
uint32_t flow_idx;
+ struct gro_tcp_flow *src_flow;
+ struct gro_tcp_flow *dst_flow;
- flow_idx = find_an_empty_flow(tbl);
+ flow_idx = find_an_empty_flow(tcp6_tbl);
if (unlikely(flow_idx == INVALID_ARRAY_INDEX))
return INVALID_ARRAY_INDEX;
- dst = &(tbl->flows[flow_idx].key);
+ dst_key = &(tcp6_tbl->flows[flow_idx].key);
+ dst_flow = &dst_key->tcp_flow;
+ src_flow = &src_key->tcp_flow;
- rte_ether_addr_copy(&(src->eth_saddr), &(dst->eth_saddr));
- rte_ether_addr_copy(&(src->eth_daddr), &(dst->eth_daddr));
- memcpy(&dst->src_addr[0], &src->src_addr[0], sizeof(dst->src_addr));
- memcpy(&dst->dst_addr[0], &src->dst_addr[0], sizeof(dst->dst_addr));
- dst->recv_ack = src->recv_ack;
- dst->src_port = src->src_port;
- dst->dst_port = src->dst_port;
+ ASSIGN_TCP_FLOW_KEY(src_flow, dst_flow)
+ memcpy(&dst_key->src_addr[0], &src_key->src_addr[0], sizeof(dst_key->src_addr));
+ memcpy(&dst_key->dst_addr[0], &src_key->dst_addr[0], sizeof(dst_key->dst_addr));
+ dst_key->vtc_flow = src_key->vtc_flow;
- tbl->flows[flow_idx].start_index = item_idx;
- tbl->flow_num++;
- tbl->flows->vtc_flow = vtc_flow;
+ tcp6_tbl->flows[flow_idx].start_index = item_idx;
+ tcp6_tbl->flow_num++;
return flow_idx;
}
@@ -189,6 +130,47 @@ update_header(struct gro_tcp_item *item)
pkt->l2_len - pkt->l3_len);
}
+static uint32_t
+gro_tcp6_flow_lookup(void *tbl, void *key)
+{
+ struct gro_tcp6_tbl *tcp6_tbl = (struct gro_tcp6_tbl *)tbl;
+ struct tcp6_flow_key *tcp6_key = (struct tcp6_flow_key *)key;
+ uint32_t max_flow_num = tcp6_tbl->max_flow_num;
+ uint32_t remaining_flow_num = tcp6_tbl->flow_num;
+ uint32_t i;
+
+ /* Search for a matched flow. */
+ max_flow_num = tcp6_tbl->max_flow_num;
+ remaining_flow_num = tcp6_tbl->flow_num;
+ for (i = 0; i < max_flow_num && remaining_flow_num; i++) {
+ if (tcp6_tbl->flows[i].start_index != INVALID_ARRAY_INDEX) {
+ if (is_same_tcp6_flow(&tcp6_tbl->flows[i].key, tcp6_key)) {
+ return i;
+ }
+ remaining_flow_num--;
+ }
+ }
+
+ return INVALID_ARRAY_INDEX;
+}
+
+static void
+gro_tcp6_flow_key_init(void *key, struct rte_tcp_hdr *tcp_hdr)
+{
+ struct tcp6_flow_key *tcp6_key = (struct tcp6_flow_key *)key;
+ struct gro_tcp_flow *tcp_flow = &tcp6_key->tcp_flow;
+
+ tcp_flow->src_port = tcp_hdr->src_port;
+ tcp_flow->dst_port = tcp_hdr->dst_port;
+ tcp_flow->recv_ack = tcp_hdr->recv_ack;
+}
+
+struct gro_tcp_flow_ops gro_tcp6_flows_ops = {
+ .tcp_flow_lookup = gro_tcp6_flow_lookup,
+ .tcp_flow_insert = gro_tcp6_insert_flow,
+ .tcp_flow_key_init = gro_tcp6_flow_key_init
+};
+
int32_t
gro_tcp6_reassemble(struct rte_mbuf *pkt,
struct gro_tcp6_tbl *tbl,
@@ -196,16 +178,9 @@ gro_tcp6_reassemble(struct rte_mbuf *pkt,
{
struct rte_ether_hdr *eth_hdr;
struct rte_ipv6_hdr *ipv6_hdr;
- struct rte_tcp_hdr *tcp_hdr;
- uint32_t sent_seq;
int32_t tcp_dl;
uint16_t ip_tlen;
struct tcp6_flow_key key;
- uint32_t cur_idx, prev_idx, item_idx;
- uint32_t i, max_flow_num, remaining_flow_num;
- int cmp;
- uint8_t find;
- rte_be32_t vtc_flow_diff;
/*
* Don't process the packet whose TCP header length is greater
@@ -216,14 +191,6 @@ gro_tcp6_reassemble(struct rte_mbuf *pkt,
eth_hdr = rte_pktmbuf_mtod(pkt, struct rte_ether_hdr *);
ipv6_hdr = (struct rte_ipv6_hdr *)((char *)eth_hdr + pkt->l2_len);
- tcp_hdr = (struct rte_tcp_hdr *)((char *)ipv6_hdr + pkt->l3_len);
-
- /*
- * Don't process the packet which has FIN, SYN, RST, PSH, URG, ECE
- * or CWR set.
- */
- if (tcp_hdr->tcp_flags != RTE_TCP_ACK_FLAG)
- return -1;
ip_tlen = rte_be_to_cpu_16(ipv6_hdr->payload_len);
/*
@@ -242,95 +209,15 @@ gro_tcp6_reassemble(struct rte_mbuf *pkt,
if (tcp_dl <= 0)
return -1;
- sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
-
- rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.eth_saddr));
- rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.eth_daddr));
+ rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.tcp_flow.eth_saddr));
+ rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.tcp_flow.eth_daddr));
memcpy(&key.src_addr[0], &ipv6_hdr->src_addr, sizeof(key.src_addr));
memcpy(&key.dst_addr[0], &ipv6_hdr->dst_addr, sizeof(key.dst_addr));
- key.src_port = tcp_hdr->src_port;
- key.dst_port = tcp_hdr->dst_port;
- key.recv_ack = tcp_hdr->recv_ack;
+ key.vtc_flow = ipv6_hdr->vtc_flow;
- /* Search for a matched flow. */
- max_flow_num = tbl->max_flow_num;
- remaining_flow_num = tbl->flow_num;
- find = 0;
- for (i = 0; i < max_flow_num && remaining_flow_num; i++) {
- if (tbl->flows[i].start_index != INVALID_ARRAY_INDEX) {
- if (is_same_tcp6_flow(tbl->flows[i].key, key)) {
- /*
- * IP version (4) Traffic Class (8) Flow Label (20)
- * All fields except Traffic class should be same
- */
- vtc_flow_diff = (ipv6_hdr->vtc_flow ^ tbl->flows->vtc_flow);
- if (vtc_flow_diff & htonl(0xF00FFFFF))
- continue;
- find = 1;
- break;
- }
- remaining_flow_num--;
- }
- }
-
- /*
- * Fail to find a matched flow. Insert a new flow and store the
- * packet into the flow.
- */
- if (find == 0) {
- item_idx = insert_new_item(tbl, pkt, start_time,
- INVALID_ARRAY_INDEX, sent_seq, true);
- if (item_idx == INVALID_ARRAY_INDEX)
- return -1;
- if (insert_new_flow(tbl, &key, ipv6_hdr->vtc_flow, item_idx) ==
- INVALID_ARRAY_INDEX) {
- /*
- * Fail to insert a new flow, so delete the
- * stored packet.
- */
- delete_item(tbl, item_idx, INVALID_ARRAY_INDEX);
- return -1;
- }
- return 0;
- }
-
- /*
- * Check all packets in the flow and try to find a neighbor for
- * the input packet.
- */
- cur_idx = tbl->flows[i].start_index;
- prev_idx = cur_idx;
- do {
- cmp = check_seq_option(&(tbl->items[cur_idx]), tcp_hdr,
- sent_seq, 0, pkt->l4_len, tcp_dl, 0,
- true);
- if (cmp) {
- if (merge_two_tcp_packets(&(tbl->items[cur_idx]),
- pkt, cmp, sent_seq, 0, 0)) {
- return 1;
- }
-
- /*
- * Fail to merge the two packets, as the packet
- * length is greater than the max value. Store
- * the packet into the flow.
- */
- if (insert_new_item(tbl, pkt, start_time, cur_idx,
- sent_seq, true) ==
- INVALID_ARRAY_INDEX)
- return -1;
- return 0;
- }
- prev_idx = cur_idx;
- cur_idx = tbl->items[cur_idx].next_pkt_idx;
- } while (cur_idx != INVALID_ARRAY_INDEX);
-
- /* Fail to find a neighbor, so store the packet into the flow. */
- if (insert_new_item(tbl, pkt, start_time, prev_idx, sent_seq,
- true) == INVALID_ARRAY_INDEX)
- return -1;
-
- return 0;
+ return gro_tcp_reassemble(pkt, tbl, &key, tcp_dl, &gro_tcp6_flows_ops, tbl->items,
+ &tbl->item_num, tbl->max_item_num,
+ 0, true, start_time);
}
uint16_t
@@ -357,7 +244,7 @@ gro_tcp6_tbl_timeout_flush(struct gro_tcp6_tbl *tbl,
* Delete the packet and get the next
* packet in the flow.
*/
- j = delete_item(tbl, j, INVALID_ARRAY_INDEX);
+ j = delete_tcp_item(tbl->items, j, &tbl->item_num, INVALID_ARRAY_INDEX);
tbl->flows[i].start_index = j;
if (j == INVALID_ARRAY_INDEX)
tbl->flow_num--;
@@ -5,24 +5,21 @@
#ifndef _GRO_TCP6_H_
#define _GRO_TCP6_H_
+#include <gro_tcp.h>
+
#define INVALID_ARRAY_INDEX 0xffffffffUL
#define GRO_TCP6_TBL_MAX_ITEM_NUM (1024UL * 1024UL)
/* Header fields representing a TCP/IPv6 flow */
struct tcp6_flow_key {
- struct rte_ether_addr eth_saddr;
- struct rte_ether_addr eth_daddr;
+ struct gro_tcp_flow tcp_flow;
uint8_t src_addr[16];
uint8_t dst_addr[16];
-
- uint32_t recv_ack;
- uint16_t src_port;
- uint16_t dst_port;
+ rte_be32_t vtc_flow;
};
struct gro_tcp6_flow {
struct tcp6_flow_key key;
- rte_be32_t vtc_flow;
/*
* The index of the first packet in the flow.
* INVALID_ARRAY_INDEX indicates an empty flow.
@@ -142,9 +139,25 @@ uint32_t gro_tcp6_tbl_pkt_count(void *tbl);
* Check if two TCP/IPv6 packets belong to the same flow.
*/
static inline int
-is_same_tcp6_flow(struct tcp6_flow_key k1, struct tcp6_flow_key k2)
+is_same_tcp6_flow(struct tcp6_flow_key *k1, struct tcp6_flow_key *k2)
{
- return (!memcmp(&k1, &k2, sizeof(struct tcp6_flow_key)));
+ rte_be32_t vtc_flow_diff;
+
+ if (memcmp(&k1->src_addr, &k2->src_addr, 16)) {
+ return 0;
+ }
+ if (memcmp(&k1->dst_addr, &k2->dst_addr, 16)) {
+ return 0;
+ }
+ /*
+ * IP version (4) Traffic Class (8) Flow Label (20)
+ * All fields except Traffic class should be same
+ */
+ vtc_flow_diff = (k1->vtc_flow ^ k2->vtc_flow);
+ if (vtc_flow_diff & htonl(0xF00FFFFF))
+ return 0;
+
+ return is_same_tcp_flow(&k1->tcp_flow, &k2->tcp_flow);
}
#endif
@@ -164,15 +164,9 @@ insert_new_flow(struct gro_vxlan_tcp4_tbl *tbl,
dst = &(tbl->flows[flow_idx].key);
- rte_ether_addr_copy(&(src->inner_key.eth_saddr),
- &(dst->inner_key.eth_saddr));
- rte_ether_addr_copy(&(src->inner_key.eth_daddr),
- &(dst->inner_key.eth_daddr));
+ ASSIGN_TCP_FLOW_KEY((&(src->inner_key.tcp_flow)), (&(dst->inner_key.tcp_flow)));
dst->inner_key.ip_src_addr = src->inner_key.ip_src_addr;
dst->inner_key.ip_dst_addr = src->inner_key.ip_dst_addr;
- dst->inner_key.recv_ack = src->inner_key.recv_ack;
- dst->inner_key.src_port = src->inner_key.src_port;
- dst->inner_key.dst_port = src->inner_key.dst_port;
dst->vxlan_hdr.vx_flags = src->vxlan_hdr.vx_flags;
dst->vxlan_hdr.vx_vni = src->vxlan_hdr.vx_vni;
@@ -358,13 +352,13 @@ gro_vxlan_tcp4_reassemble(struct rte_mbuf *pkt,
sent_seq = rte_be_to_cpu_32(tcp_hdr->sent_seq);
- rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.inner_key.eth_saddr));
- rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.inner_key.eth_daddr));
+ rte_ether_addr_copy(&(eth_hdr->src_addr), &(key.inner_key.tcp_flow.eth_saddr));
+ rte_ether_addr_copy(&(eth_hdr->dst_addr), &(key.inner_key.tcp_flow.eth_daddr));
key.inner_key.ip_src_addr = ipv4_hdr->src_addr;
key.inner_key.ip_dst_addr = ipv4_hdr->dst_addr;
- key.inner_key.recv_ack = tcp_hdr->recv_ack;
- key.inner_key.src_port = tcp_hdr->src_port;
- key.inner_key.dst_port = tcp_hdr->dst_port;
+ key.inner_key.tcp_flow.recv_ack = tcp_hdr->recv_ack;
+ key.inner_key.tcp_flow.src_port = tcp_hdr->src_port;
+ key.inner_key.tcp_flow.dst_port = tcp_hdr->dst_port;
key.vxlan_hdr.vx_flags = vxlan_hdr->vx_flags;
key.vxlan_hdr.vx_vni = vxlan_hdr->vx_vni;
@@ -3,6 +3,7 @@
sources = files(
'rte_gro.c',
+ 'gro_tcp.c',
'gro_tcp4.c',
'gro_tcp6.c',
'gro_udp4.c',