Hi Anoob,
Thank you.
Please see my answers below.
Regards,
Praveen
-----Original Message-----
From: Anoob Joseph <ano...@marvell.com>
Sent: Thursday, March 12, 2020 4:31 PM
To: Shetty, Praveen <praveen.she...@intel.com>; dev@dpdk.org; Doherty, Declan
<declan.dohe...@intel.com>; Iremonger, Bernard <bernard.iremon...@intel.com>;
Ananyev, Konstantin <konstantin.anan...@intel.com>
Subject: RE: [dpdk-dev] [PATCH v1] examples/ipsec-secgw: support flow director
feature
Hi Praveen,
I do have some review comments on the code. Before that, can you give a brief
overview of what is being targeted? My understanding is that the primary
objective is to use rte_flow (or flow director) to redirect a specific
flow(/SA) to a specific queue. Can you confirm?
>>>> Yes, your understanding is correct, the main objective is to support load
>>>> distribution in ipsec-secgw application.
>>>> flow director and RSS features are used achieve the load distribution.
>>>> flow director is used to redirect the specified inbound ipsec flow to a
>>>> specified queue.
Couple of questions,
1. I would assume the new option of "flow-direction" is optional and is
determined per SA. In that case, can I assume that RSS would be active for the
other flows (or SAs). Let's say, I just want to add a SA for which I would like
to enable "flow-direction" but leave the rest as is. How is that handled?
[Praveen]
>>>> We are using fdir_flag to differentiate the mix of SA's(SA's with and
>>>> without flow-direction).
>>>> fdir_flag will be "set" for the SA which has configured with
>>>> flow-direction option(SA rule syntax is extended to add new options
>>>> <action_type> <portid> <queueid> ).
>>>> flow creation is called only for the SA's with fdir_flag is set.
2. I see that the changes are only applicable for LOOKASIDE_PROTOCOL. The same
feature would be useful for other modes as well, right?
[Praveen]
>>>> We are adding this feature for i40e NIC and the i40e NIC doesn't support
>>>> either encryption or decryption, that's why we used only
>>>> LOOKASIDE_PROTOCOL in this case.
3. I'm not sure "flow-direction" is the right wording for the option. This is
just specifying the "rx-queue" per SA. @Akhil, Konstantin, comments?
>>>> @Declan, @Konstantin , @Bernard, @Akhil Could you please suggest a name
>>>> on which we can all agree upon?
Thanks,
Anoob
> -----Original Message-----
> From: dev <dev-boun...@dpdk.org> On Behalf Of Praveen Shetty
> Sent: Wednesday, March 11, 2020 8:25 PM
> To: dev@dpdk.org; declan.dohe...@intel.com;
> bernard.iremon...@intel.com; konstantin.anan...@intel.com
> Subject: [dpdk-dev] [PATCH v1] examples/ipsec-secgw: support flow
> director feature
>
> Modified Secuirty gateway application to support configuration of flow
> director rule to direct inbound IPsec SA to a specified queue.
>
> Signed-off-by: Praveen Shetty <praveen.she...@intel.com>
> ---
> examples/ipsec-secgw/ep0.cfg | 11 +++++
> examples/ipsec-secgw/ipsec-secgw.c | 56 ++++++++++++++++++++++++-
> examples/ipsec-secgw/ipsec.c | 67 ++++++++++++++++++++++++++++++
> examples/ipsec-secgw/ipsec.h | 11 +++++
> examples/ipsec-secgw/sa.c | 50 +++++++++++++++++++++-
> 5 files changed, 192 insertions(+), 3 deletions(-)
>
> diff --git a/examples/ipsec-secgw/ep0.cfg
> b/examples/ipsec-secgw/ep0.cfg index dfd4aca7d..c9f80e81b 100644
> --- a/examples/ipsec-secgw/ep0.cfg
> +++ b/examples/ipsec-secgw/ep0.cfg
> @@ -29,6 +29,7 @@ sp ipv4 in esp protect 111 pri 1 dst
> 192.168.186.0/24 sport
> 0:65535 dport 0:6553 sp ipv4 in esp protect 115 pri 1 dst
> 192.168.210.0/24 sport
> 0:65535 dport 0:65535 sp ipv4 in esp protect 116 pri 1 dst
> 192.168.211.0/24 sport 0:65535 dport 0:65535 sp ipv4 in esp protect
> 115 pri 1 dst
> 192.168.210.0/24 sport 0:65535 dport 0:65535
> +sp ipv4 in esp protect 117 pri 1 dst 192.168.212.0/24 sport 0:65535
> +dport 0:65535
> sp ipv4 in esp protect 125 pri 1 dst 192.168.65.0/24 sport 0:65535
> dport 0:65535 sp ipv4 in esp protect 125 pri 1 dst 192.168.65.0/24
> sport 0:65535 dport 0:65535 sp ipv4 in esp protect 126 pri 1 dst
> 192.168.66.0/24 sport 0:65535 dport 0:65535 @@ -61,6 +62,8 @@ sp ipv6
> in esp protect 125 pri 1 dst
> ffff:0000:0000:0000:aaaa:aaaa:0000:0000/96
> sport 0:65535 dport 0:65535
> sp ipv6 in esp protect 126 pri 1 dst
> ffff:0000:0000:0000:bbbb:bbbb:0000:0000/96 \ sport 0:65535 dport
> 0:65535
> +sp ipv6 in esp protect 127 pri 1 dst
> +ffff:0000:0000:0000:cccc:dddd:0000:0000/96 \ sport 0:65535 dport
> +0:65535
>
> #SA rules
> sa out 5 cipher_algo aes-128-cbc cipher_key
> 0:0:0:0:0:0:0:0:0:0:0:0:0:0:0:0 \ @@ -118,6 +121,9 @@ dst 172.16.1.5
>
> sa in 116 cipher_algo null auth_algo null mode ipv4-tunnel src
> 172.16.2.6 dst
> 172.16.1.6
>
> +sa in 117 cipher_algo null auth_algo null mode ipv4-tunnel src
> +172.16.2.7 \ dst 172.16.1.7 flow-direction 0 2 port_id 0 type
> +lookaside-protocol-offload
> +
> sa in 125 cipher_algo aes-128-cbc cipher_key
> c3:c3:c3:c3:c3:c3:c3:c3:c3:c3:c3:\
> c3:c3:c3:c3:c3 auth_algo sha1-hmac auth_key
> c3:c3:c3:c3:c3:c3:c3:c3:c3:c3:c3:\
> c3:c3:c3:c3:c3:c3:c3:c3:c3 mode ipv6-tunnel \ @@ -130,6 +136,11 @@ sa
> in
> 126 cipher_algo aes-128-cbc cipher_key
> 4d:4d:4d:4d:4d:4d:4d:4d:4d:4d:4d:\
> src 2222:2222:2222:2222:2222:2222:2222:6666 \ dst
> 1111:1111:1111:1111:1111:1111:1111:6666
>
> +sa in 127 cipher_algo null auth_algo null mode ipv6-tunnel \ src
> +2222:2222:2222:2222:2222:2222:2222:7777 \ dst
> +1111:1111:1111:1111:1111:1111:1111:7777 \ flow-direction 0 3 port_id
> +0 type lookaside-protocol-offload
> +
> #Routing rules
> rt ipv4 dst 172.16.2.5/32 port 0
> rt ipv4 dst 172.16.2.6/32 port 1
> diff --git a/examples/ipsec-secgw/ipsec-secgw.c b/examples/ipsec-
> secgw/ipsec-secgw.c index 4799bc90c..132484422 100644
> --- a/examples/ipsec-secgw/ipsec-secgw.c
> +++ b/examples/ipsec-secgw/ipsec-secgw.c
> @@ -166,7 +166,6 @@ static const struct option lgopts[] = {
> {CMD_LINE_OPT_FRAG_TTL, 1, 0, CMD_LINE_OPT_FRAG_TTL_NUM},
> {NULL, 0, 0, 0}
> };
> -
> /* mask of enabled ports */
> static uint32_t enabled_port_mask;
> static uint64_t enabled_cryptodev_mask = UINT64_MAX; @@ -259,6
> +258,30 @@ static struct rte_eth_conf port_conf = {
> .txmode = {
> .mq_mode = ETH_MQ_TX_NONE,
> },
> + .fdir_conf = {
> + .mode = RTE_FDIR_MODE_NONE,
> + .pballoc = RTE_FDIR_PBALLOC_64K,
> + .status = RTE_FDIR_REPORT_STATUS,
> + .mask = {
> + .vlan_tci_mask = 0xFFEF,
> + .ipv4_mask = {
> + .src_ip = 0xFFFFFFFF,
> + .dst_ip = 0xFFFFFFFF,
> + },
> + .ipv6_mask = {
> + .src_ip = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
> + 0xFFFFFFFF},
> + .dst_ip = {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
> + 0xFFFFFFFF},
> + },
> + .src_port_mask = 0xFFFF,
> + .dst_port_mask = 0xFFFF,
> + .mac_addr_byte_mask = 0xFF,
> + .tunnel_type_mask = 1,
> + .tunnel_id_mask = 0xFFFFFFFF,
> + },
> + .drop_queue = 127,
> + }
> };
>
> static struct socket_ctx socket_ctx[NB_SOCKETS]; @@ -1184,7 +1207,6
> @@
> main_loop(__attribute__((unused)) void *dummy)
>
> if (nb_rx > 0)
> process_pkts(qconf, pkts, nb_rx, portid);
> -
> /* dequeue and process completed crypto-ops */
> if (UNPROTECTED_PORT(portid))
> drain_inbound_crypto_queues(qconf,
> @@ -1196,6 +1218,27 @@ main_loop(__attribute__((unused)) void *dummy)
> }
> }
>
> +int check_flow_params(uint16_t fdir_portid, uint8_t fdir_qid) {
> + uint16_t i;
> + uint16_t portid;
> + uint8_t queueid;
> +
> + for (i = 0; i < nb_lcore_params; ++i) {
> + portid = lcore_params_array[i].port_id;
> + if (portid == fdir_portid) {
> + queueid = lcore_params_array[i].queue_id;
> + if (queueid == fdir_qid)
> + break;
> + }
> +
> + if (i == nb_lcore_params - 1)
> + return -1;
> + }
> +
> + return 1;
> +}
> +
> static int32_t
> check_params(void)
> {
> @@ -2503,6 +2546,15 @@ main(int32_t argc, char **argv)
> continue;
>
> sa_check_offloads(portid, &req_rx_offloads, &req_tx_offloads);
> + /* check if FDIR is configured on the port */
> + if (check_fdir_configured(portid)) {
> + /* Enable FDIR */
> + port_conf.fdir_conf.mode =
> RTE_FDIR_MODE_PERFECT;
> + /* Disable RSS */
> + port_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
> + port_conf.rx_adv_conf.rss_conf.rss_hf = 0;
> + port_conf.rx_adv_conf.rss_conf.rss_key = NULL;
> + }
> port_init(portid, req_rx_offloads, req_tx_offloads);
> }
>
> diff --git a/examples/ipsec-secgw/ipsec.c
> b/examples/ipsec-secgw/ipsec.c index 6e8120702..363809cfd 100644
> --- a/examples/ipsec-secgw/ipsec.c
> +++ b/examples/ipsec-secgw/ipsec.c
> @@ -415,6 +415,73 @@ create_inline_session(struct socket_ctx *skt_ctx,
> struct ipsec_sa *sa,
> return 0;
> }
>
> +int
> +create_ipsec_esp_flow(struct ipsec_sa *sa) {
> + int ret = 0;
> + struct rte_flow_error err;
> + if (sa->direction == RTE_SECURITY_IPSEC_SA_DIR_EGRESS)
> + return 0; /* No Flow director rules for Egress traffic */
> + if (sa->flags == TRANSPORT) {
> + RTE_LOG(ERR, IPSEC,
> + "No Flow director rule for transport mode:");
> + return -1;
> + }
> + sa->action[0].type = RTE_FLOW_ACTION_TYPE_QUEUE;
> + sa->pattern[0].type = RTE_FLOW_ITEM_TYPE_ETH;
> + sa->action[0].conf =
> + &(struct rte_flow_action_queue){
> + .index = sa->fdir_qid,
> + };
> + sa->attr.egress = 0;
> + sa->attr.ingress = 1;
> + if (IS_IP6(sa->flags)) {
> + sa->pattern[1].mask = &rte_flow_item_ipv6_mask;
> + sa->pattern[1].type = RTE_FLOW_ITEM_TYPE_IPV6;
> + sa->pattern[1].spec = &sa->ipv6_spec;
> + memcpy(sa->ipv6_spec.hdr.dst_addr,
> + sa->dst.ip.ip6.ip6_b, IPV6_ADDR_LEN);
> + memcpy(sa->ipv6_spec.hdr.src_addr,
> + sa->src.ip.ip6.ip6_b, IPV6_ADDR_LEN);
> + sa->pattern[2].type = RTE_FLOW_ITEM_TYPE_ESP;
> + sa->pattern[2].spec = &sa->esp_spec;
> + sa->pattern[2].mask = &rte_flow_item_esp_mask;
> + sa->esp_spec.hdr.spi = rte_cpu_to_be_32(sa->spi);
> + sa->pattern[3].type = RTE_FLOW_ITEM_TYPE_END;
> + } else if (IS_IP4(sa->flags)) {
> + sa->pattern[1].mask = &rte_flow_item_ipv4_mask;
> + sa->pattern[1].type = RTE_FLOW_ITEM_TYPE_IPV4;
> + sa->pattern[1].spec = &sa->ipv4_spec;
> + sa->ipv4_spec.hdr.dst_addr = sa->dst.ip.ip4;
> + sa->ipv4_spec.hdr.src_addr = sa->src.ip.ip4;
> + sa->pattern[2].type = RTE_FLOW_ITEM_TYPE_ESP;
> + sa->pattern[2].spec = &sa->esp_spec;
> + sa->pattern[2].mask = &rte_flow_item_esp_mask;
> + sa->esp_spec.hdr.spi = rte_cpu_to_be_32(sa->spi);
> + sa->pattern[3].type = RTE_FLOW_ITEM_TYPE_END;
> + }
> + sa->action[1].type = RTE_FLOW_ACTION_TYPE_END;
> +
> + ret = rte_flow_validate(sa->fdir_portid, &sa->attr,
> + sa->pattern, sa->action,
> + &err);
> + if (ret < 0) {
> + RTE_LOG(ERR, IPSEC,
> + "Flow Validation failed\n");
> + return ret;
> + }
> + sa->flow = rte_flow_create(sa->fdir_portid,
> + &sa->attr, sa->pattern, sa->action,
> + &err);
> + if (!sa->flow) {
> + RTE_LOG(ERR, IPSEC,
> + "Flow Creation failed\n");
> + return -1;
> + }
> +
> + return 0;
> +}
> +
> /*
> * queue crypto-ops into PMD queue.
> */
> diff --git a/examples/ipsec-secgw/ipsec.h
> b/examples/ipsec-secgw/ipsec.h index 4f2fd6184..00147895a 100644
> --- a/examples/ipsec-secgw/ipsec.h
> +++ b/examples/ipsec-secgw/ipsec.h
> @@ -46,6 +46,8 @@
>
> #define IP6_VERSION (6)
>
> +#define IPV6_ADDR_LEN 16
> +
> struct rte_crypto_xform;
> struct ipsec_xform;
> struct rte_mbuf;
> @@ -138,6 +140,9 @@ struct ipsec_sa {
> };
> enum rte_security_ipsec_sa_direction direction;
> uint16_t portid;
> + uint16_t fdir_portid;
> + uint8_t fdir_qid;
> + uint8_t fdir_flag;
>
> #define MAX_RTE_FLOW_PATTERN (4)
> #define MAX_RTE_FLOW_ACTIONS (3)
> @@ -383,5 +388,11 @@ create_lookaside_session(struct ipsec_ctx
> *ipsec_ctx, struct ipsec_sa *sa, int create_inline_session(struct
> socket_ctx *skt_ctx, struct ipsec_sa *sa,
> struct rte_ipsec_session *ips);
> +int
> +check_flow_params(uint16_t fdir_portid, uint8_t fdir_qid);
> +
> +int
> +create_ipsec_esp_flow(struct ipsec_sa *sa);
>
> +int check_fdir_configured(uint16_t portid);
> #endif /* __IPSEC_H__ */
> diff --git a/examples/ipsec-secgw/sa.c b/examples/ipsec-secgw/sa.c
> index 4822d6bda..9955dfcbe 100644
> --- a/examples/ipsec-secgw/sa.c
> +++ b/examples/ipsec-secgw/sa.c
> @@ -20,6 +20,9 @@
> #include <rte_random.h>
> #include <rte_ethdev.h>
> #include <rte_malloc.h>
> +#include <rte_common.h>
> +#include <rte_string_fns.h>
> +#include <rte_ethdev_driver.h>
>
> #include "ipsec.h"
> #include "esp.h"
> @@ -271,6 +274,7 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
> uint32_t type_p = 0;
> uint32_t portid_p = 0;
> uint32_t fallback_p = 0;
> + int16_t status_p = 0;
>
> if (strcmp(tokens[0], "in") == 0) {
> ri = &nb_sa_in;
> @@ -681,6 +685,25 @@ parse_sa_tokens(char **tokens, uint32_t n_tokens,
> fallback_p = 1;
> continue;
> }
> + if (strcmp(tokens[ti], "flow-direction") == 0) {
> + rule->fdir_flag = 1;
> + INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
> + if (status->status < 0)
> + return;
> + rule->fdir_portid = atoi(tokens[ti]);
> + INCREMENT_TOKEN_INDEX(ti, n_tokens, status);
> + if (status->status < 0)
> + return;
> + rule->fdir_qid = atoi(tokens[ti]);
> + /* validating portid and queueid */
> + status_p = check_flow_params(rule->fdir_portid,
> + rule->fdir_qid);
> + if (status_p < 0) {
> + printf("port id %u / queue id %u is not
> valid\n",
> + rule->fdir_portid, rule->fdir_qid);
> + }
> + continue;
> + }
>
> /* unrecognizeable input */
> APP_CHECK(0, status, "unrecognized input \"%s\"", @@ -823,6
> +846,9 @@ print_one_sa_rule(const struct ipsec_sa *sa, int inbound)
> break;
> }
> }
> + if (sa->fdir_flag == 1)
> + printf("flow-direction %d %d", sa->fdir_portid, sa->fdir_qid);
> +
> printf("\n");
> }
>
> @@ -1153,7 +1179,15 @@ sa_add_rules(struct sa_ctx *sa_ctx, const
> struct ipsec_sa entries[],
> return -EINVAL;
> }
> }
> -
> + if (sa->fdir_flag &&
> + ips->type ==
> + RTE_SECURITY_ACTION_TYPE_LOOKASIDE_PROTOCOL
> &&
> + inbound) {
> + rc = create_ipsec_esp_flow(sa);
> + if (rc != 0)
> + RTE_LOG(ERR, IPSEC_ESP,
> + "create_ipsec_esp flow failed\n");
> + }
> print_one_sa_rule(sa, inbound);
> }
>
> @@ -1256,6 +1290,20 @@ fill_ipsec_session(struct rte_ipsec_session
> *ss, struct rte_ipsec_sa *sa)
> return rc;
> }
>
> +int
> +check_fdir_configured(uint16_t portid) {
> + struct ipsec_sa *sa = NULL;
> + uint32_t idx_sa = 0;
> +
> + for (idx_sa = 0; idx_sa < nb_sa_in; idx_sa++) {
> + sa = &sa_in[idx_sa];
> + if (sa->fdir_portid == portid)
> + return sa->fdir_flag;
> + }
> + return 0;
> +}
> +
> /*
> * Initialise related rte_ipsec_sa object.
> */
> --
> 2.17.1
