Re: [dpdk-dev] [PATCH v1] lib/ipsec: add support for header construction

["Ananyev, Konstantin" <konstantin.ananyev@intel.com>]

Hi,

> 
> Add support for RFC 4301(5.1.2) to update of
> Type of service field and Traffic class field
> bits inside ipv4/ipv6 packets for outbound cases
> and inbound cases which deals with the update of
> the DSCP/ENC bits inside each of the fields.
> 
> Signed-off-by: Marko Kovacevic <marko.kovacevic@intel.com>
> ---
>  examples/ipsec-secgw/sa.c          |   2 +
>  lib/librte_ipsec/esp_inb.c         |  14 ++++-
>  lib/librte_ipsec/esp_outb.c        |   4 +-
>  lib/librte_ipsec/iph.h             | 119 +++++++++++++++++++++++++++++++++++--
>  lib/librte_ipsec/rte_ipsec_sa.h    |  25 ++++++++
>  lib/librte_ipsec/sa.c              |  17 ++++++
>  lib/librte_ipsec/sa.h              |   2 +
>  lib/librte_net/rte_ip.h            |   8 +++
>  lib/librte_security/rte_security.h |   9 +++
>  9 files changed, 191 insertions(+), 9 deletions(-)

Looks good in general, some generic comments:
- I think it is better to split the patch into few sub-pathces:
  One for rte_security, second for rte_net, third - rte_ipsec, forth - examples/ipsec-secgw
- Would be good to add support for other options too (ttl, etc.)
- Would be good to add new test-case for it into examples/ipsec-secgw/test/

Plus few nits in the code below.
Konstantin

> 
> diff --git a/examples/ipsec-secgw/sa.c b/examples/ipsec-secgw/sa.c
> index b850e9839..4d85d09df 100644
> --- a/examples/ipsec-secgw/sa.c
> +++ b/examples/ipsec-secgw/sa.c
> @@ -991,6 +991,8 @@ fill_ipsec_sa_prm(struct rte_ipsec_sa_prm *prm, const struct ipsec_sa *ss,
>  	prm->ipsec_xform.mode = (ss->flags == TRANSPORT) ?
>  		RTE_SECURITY_IPSEC_SA_MODE_TRANSPORT :
>  		RTE_SECURITY_IPSEC_SA_MODE_TUNNEL;
> +	prm->ipsec_xform.options.ecn = 1;
> +	prm->ipsec_xform.options.copy_dscp = 1;
> 
>  	if (ss->flags == IP4_TUNNEL) {
>  		prm->ipsec_xform.tunnel.type = RTE_SECURITY_IPSEC_TUNNEL_IPV4;
> diff --git a/lib/librte_ipsec/esp_inb.c b/lib/librte_ipsec/esp_inb.c
> index 4e0e12a85..8a3cb8a15 100644
> --- a/lib/librte_ipsec/esp_inb.c
> +++ b/lib/librte_ipsec/esp_inb.c
> @@ -377,9 +377,10 @@ tun_process(const struct rte_ipsec_sa *sa, struct rte_mbuf *mb[],
>  {
>  	uint32_t adj, i, k, tl;
>  	uint32_t hl[num];
> +	void *inner_h;
> +	const void *outter_h;
>  	struct esp_tail espt[num];
>  	struct rte_mbuf *ml[num];
> -
>  	const uint32_t tlen = sa->icv_len + sizeof(espt[0]);
>  	const uint32_t cofs = sa->ctp.cipher.offset;
> 
> @@ -400,9 +401,16 @@ tun_process(const struct rte_ipsec_sa *sa, struct rte_mbuf *mb[],
>  		if (tun_process_check(mb[i], ml[i], espt[i], adj, tl,
>  					sa->proto) == 0) {
> 
> +			outter_h = rte_pktmbuf_mtod_offset(mb[i], uint8_t *,
> +					mb[i]->l2_len);
> +
>  			/* modify packet's layout */
> -			tun_process_step2(mb[i], ml[i], hl[i], adj,
> -				tl, sqn + k);
> +			inner_h = tun_process_step2(mb[i], ml[i], hl[i], adj,
> +					tl, sqn + k);
> +
> +			if ((sa->type & INB_TUN_HDR_MSK) != 0)
> +				update_inb_tun_l3_hdr(sa, inner_h, outter_h);
> +
>  			/* update mbuf's metadata */
>  			tun_process_step3(mb[i], sa->tx_offload.msk,
>  				sa->tx_offload.val);
> diff --git a/lib/librte_ipsec/esp_outb.c b/lib/librte_ipsec/esp_outb.c
> index c798bc4c4..a71164e0c 100644
> --- a/lib/librte_ipsec/esp_outb.c
> +++ b/lib/librte_ipsec/esp_outb.c
> @@ -152,8 +152,8 @@ outb_tun_pkt_prepare(struct rte_ipsec_sa *sa, rte_be64_t sqc,
>  	rte_memcpy(ph, sa->hdr, sa->hdr_len);
> 
>  	/* update original and new ip header fields */
> -	update_tun_l3hdr(sa, ph + sa->hdr_l3_off, mb->pkt_len, sa->hdr_l3_off,
> -			sqn_low16(sqc));
> +	update_outb_tun_l3hdr(sa, ph + sa->hdr_l3_off, ph + hlen, mb->pkt_len,
> +			sa->hdr_l3_off, sqn_low16(sqc));
> 
>  	/* update spi, seqn and iv */
>  	esph = (struct esp_hdr *)(ph + sa->hdr_len);
> diff --git a/lib/librte_ipsec/iph.h b/lib/librte_ipsec/iph.h
> index 58930cf18..f45db5d4a 100644
> --- a/lib/librte_ipsec/iph.h
> +++ b/lib/librte_ipsec/iph.h
> @@ -11,6 +11,11 @@
>   * used internally by ipsec library.
>   */
> 
> +#define IPV6_DSCP_MASK	(DSCP_MASK << IPV6_HDR_TC_SHIFT)
> +#define IPV6_ECN_MASK	(ECN_MASK << IPV6_HDR_TC_SHIFT)
> +#define IPV6_TOS_MASK	(IPV6_ECN_MASK | IPV6_DSCP_MASK)
> +#define IPV6_ECN_CE	IPV6_ECN_MASK
> +
>  /*
>   * Move preceding (L3) headers down to remove ESP header and IV.
>   */
> @@ -35,6 +40,26 @@ insert_esph(char *np, char *op, uint32_t hlen)
>  		np[i] = op[i];
>  }
> 
> +static inline uint8_t
> +get_ipv6_tos(rte_be32_t vtc_flow)
> +{
> +	uint32_t v;
> +
> +	v = rte_be_to_cpu_32(vtc_flow);
> +	return v >> IPV6_HDR_TC_SHIFT;
> +}
> +
> +static inline rte_be32_t
> +set_ipv6_tos(rte_be32_t vtc_flow, uint32_t tos)
> +{
> +	uint32_t v;
> +
> +	v = rte_cpu_to_be_32(tos << IPV6_HDR_TC_SHIFT);
> +	vtc_flow &= ~rte_cpu_to_be_32(IPV6_TOS_MASK);
> +
> +	return (v | vtc_flow);
> +}
> +
>  /* update original ip header fields for transport case */
>  static inline int
>  update_trs_l3hdr(const struct rte_ipsec_sa *sa, void *p, uint32_t plen,
> @@ -64,20 +89,106 @@ update_trs_l3hdr(const struct rte_ipsec_sa *sa, void *p, uint32_t plen,
> 
>  /* update original and new ip header fields for tunnel case */
>  static inline void
> -update_tun_l3hdr(const struct rte_ipsec_sa *sa, void *p, uint32_t plen,
> -		uint32_t l2len, rte_be16_t pid)
> +update_outb_tun_l3hdr(const struct rte_ipsec_sa *sa, void *outh,
> +		const void *inh, uint32_t plen, uint32_t l2len, rte_be16_t pid)
>  {
>  	struct ipv4_hdr *v4h;
>  	struct ipv6_hdr *v6h;
> +	uint32_t itp, otp;
> +	const struct ipv4_hdr *v4in_h;
> +	const struct ipv6_hdr *v6in_h;
> 
>  	if (sa->type & RTE_IPSEC_SATP_MODE_TUNLV4) {
> -		v4h = p;
> +		v4h = outh;
>  		v4h->packet_id = pid;
>  		v4h->total_length = rte_cpu_to_be_16(plen - l2len);

I think it makes sense to invoke the code below, only when:
((sa->type & INB_TUN_HDR_MSK) != 0)
Same as we doing for onbound.
Also probably worth to put it into a separate inline function.

> +
> +		if (sa->proto == IPPROTO_IPIP) {

For consistency with the check above, seems a bit better:
if ((sa->type & RTE_IPSEC_SATP_IPV_MASK) == RTE_IPSEC_SATP_IPV4)


> +			/* ipv4 inner header */
> +			v4in_h = inh;
> +
> +			otp = v4h->type_of_service & ~sa->tos_mask;
> +			itp = v4in_h->type_of_service & sa->tos_mask;
> +			v4h->type_of_service = (otp | itp);
> +		} else {
> +			/* ipv6 inner header */
> +			v6in_h = inh;
> +
> +			otp = v4h->type_of_service & ~sa->tos_mask;
> +			itp = get_ipv6_tos(v6in_h->vtc_flow) & sa->tos_mask;
> +			v4h->type_of_service = (otp | itp);
> +		}
>  	} else {
> -		v6h = p;
> +		v6h = outh;
>  		v6h->payload_len = rte_cpu_to_be_16(plen - l2len -
>  				sizeof(*v6h));
> +
> +		if (sa->proto == IPPROTO_IPIP) {

Same comment as above here.

> +			/* ipv4 inner header */
> +			v4in_h = inh;
> +
> +			otp = get_ipv6_tos(v6h->vtc_flow) & ~sa->tos_mask;
> +			itp = v4in_h->type_of_service & sa->tos_mask;
> +			v6h->vtc_flow = set_ipv6_tos(v6h->vtc_flow, otp | itp);
> +		} else {
> +			/* ipv6 inner header */
> +			v6in_h = inh;
> +
> +			otp = get_ipv6_tos(v6h->vtc_flow) & ~sa->tos_mask;
> +			itp = get_ipv6_tos(v6in_h->vtc_flow) & sa->tos_mask;
> +			v6h->vtc_flow = set_ipv6_tos(v6h->vtc_flow, otp | itp);
> +		}
> +	}
> +}
> +
> +static inline void
> +update_inb_tun_l3_hdr(const struct rte_ipsec_sa *sa, void *ip_inner,
> +		const void *ip_outter)
> +{
> +	struct ipv4_hdr *inner_v4h;
> +	const struct ipv4_hdr *outter_v4h;
> +	struct ipv6_hdr *inner_v6h;
> +	const struct ipv6_hdr *outter_v6h;
> +	uint8_t ecn_v4out, ecn_v4in;
> +	uint32_t ecn_v6out, ecn_v6in;
> +
> +	inner_v4h = ip_inner;
> +	outter_v4h = ip_outter;
> +
> +	inner_v6h = ip_inner;
> +	outter_v6h = ip_outter;
> +
> +	/* <update ecn bits in inner IP header> */
> +	if (sa->type & RTE_IPSEC_SATP_MODE_TUNLV4) {
> +
> +		ecn_v4out = outter_v4h->type_of_service & ECN_MASK;
> +
> +		if ((sa->type & RTE_IPSEC_SATP_IPV_MASK) == RTE_IPSEC_SATP_IPV4) {
> +			ecn_v4in = inner_v4h->type_of_service & ECN_MASK;
> +			if (ecn_v4out == ECN_CE && ecn_v4in != 0)
> +				inner_v4h->type_of_service |= ECN_CE;
> +		} else {
> +			ecn_v6in = inner_v6h->vtc_flow &
> +					rte_cpu_to_be_32(IPV6_ECN_MASK);
> +			if (ecn_v4out == ECN_CE && ecn_v6in != 0)
> +				inner_v6h->vtc_flow |=
> +						rte_cpu_to_be_32(IPV6_ECN_CE);
> +		}
> +	} else {
> +		ecn_v6out = outter_v6h->vtc_flow &
> +				rte_cpu_to_be_32(IPV6_ECN_MASK);
> +
> +		if ((sa->type & RTE_IPSEC_SATP_IPV_MASK) == RTE_IPSEC_SATP_IPV6) {
> +			ecn_v6in = inner_v6h->vtc_flow &
> +					rte_cpu_to_be_32(IPV6_ECN_MASK);
> +			if (ecn_v6out == IPV6_ECN_CE && ecn_v6in != 0)
> +				inner_v6h->vtc_flow |=
> +						rte_cpu_to_be_32(IPV6_ECN_CE);
> +		} else {
> +			ecn_v4in = inner_v4h->type_of_service & ECN_MASK;
> +			if (ecn_v6out == ECN_CE && ecn_v4in != 0)
> +				inner_v4h->type_of_service |= ECN_CE;
> +		}
>  	}
>  }
> 
> diff --git a/lib/librte_ipsec/rte_ipsec_sa.h b/lib/librte_ipsec/rte_ipsec_sa.h
> index fd9b3ed60..8f179ee9d 100644
> --- a/lib/librte_ipsec/rte_ipsec_sa.h
> +++ b/lib/librte_ipsec/rte_ipsec_sa.h
> @@ -95,6 +95,11 @@ enum {
>  	RTE_SATP_LOG2_MODE,
>  	RTE_SATP_LOG2_SQN = RTE_SATP_LOG2_MODE + 2,
>  	RTE_SATP_LOG2_ESN,
> +	RTE_SATP_LOG2_ECN,
> +	RTE_SATP_LOG2_DSCP,
> +	RTE_SATP_LOG2_TTL,
> +	RTE_SATP_LOG2_DF,
> +	RTE_SATP_LOG2_FLABEL,
>  	RTE_SATP_LOG2_NUM
>  };
> 
> @@ -123,6 +128,26 @@ enum {
>  #define RTE_IPSEC_SATP_ESN_DISABLE	(0ULL << RTE_SATP_LOG2_ESN)
>  #define RTE_IPSEC_SATP_ESN_ENABLE	(1ULL << RTE_SATP_LOG2_ESN)
> 
> +#define RTE_IPSEC_SATP_ECN_MASK		(1ULL << RTE_SATP_LOG2_ECN)
> +#define RTE_IPSEC_SATP_ECN_DISABLE	(0ULL << RTE_SATP_LOG2_ECN)
> +#define RTE_IPSEC_SATP_ECN_ENABLE	(1ULL << RTE_SATP_LOG2_ECN)
> +
> +#define RTE_IPSEC_SATP_DSCP_MASK	(1ULL << RTE_SATP_LOG2_DSCP)
> +#define RTE_IPSEC_SATP_DSCP_DISABLE	(0ULL << RTE_SATP_LOG2_DSCP)
> +#define RTE_IPSEC_SATP_DSCP_ENABLE	(1ULL << RTE_SATP_LOG2_DSCP)
> +
> +#define RTE_IPSEC_SATP_TTL_MASK		(1ULL << RTE_SATP_LOG2_TTL)
> +#define RTE_IPSEC_SATP_TTL_DISABLE	(0ULL << RTE_SATP_LOG2_TTL)
> +#define RTE_IPSEC_SATP_TTL_ENABLE	(1ULL << RTE_SATP_LOG2_TTL)
> +
> +#define RTE_IPSEC_SATP_DF_MASK		(1ULL << RTE_SATP_LOG2_DF)
> +#define RTE_IPSEC_SATP_DF_DISABLE	(0ULL << RTE_SATP_LOG2_DF)
> +#define RTE_IPSEC_SATP_DF_ENABLE	(1ULL << RTE_SATP_LOG2_DF)
> +
> +#define RTE_IPSEC_SATP_FLABEL_MASK	(1ULL << RTE_SATP_LOG2_FLABEL)
> +#define RTE_IPSEC_SATP_FLABEL_DISABLE	(0ULL << RTE_SATP_LOG2_FLABEL)
> +#define RTE_IPSEC_SATP_FLABEL_ENABLE	(1ULL << RTE_SATP_LOG2_FLABEL)
> +
>  /**
>   * get type of given SA
>   * @return
> diff --git a/lib/librte_ipsec/sa.c b/lib/librte_ipsec/sa.c
> index 846e317fe..d48acd117 100644
> --- a/lib/librte_ipsec/sa.c
> +++ b/lib/librte_ipsec/sa.c
> @@ -220,6 +220,17 @@ fill_sa_type(const struct rte_ipsec_sa_prm *prm, uint64_t *type)
>  	else
>  		tp |= RTE_IPSEC_SATP_SQN_RAW;
> 
> +	/* check for ECN flag */
> +	if (prm->ipsec_xform.options.ecn == 0)
> +		tp |= RTE_IPSEC_SATP_ECN_DISABLE;
> +	else
> +		tp |= RTE_IPSEC_SATP_ECN_ENABLE;
> +	/* check for DSCP flag */
> +	if (prm->ipsec_xform.options.copy_dscp == 0)
> +		tp |= RTE_IPSEC_SATP_DSCP_DISABLE;
> +	else
> +		tp |= RTE_IPSEC_SATP_DSCP_ENABLE;
> +
>  	*type = tp;
>  	return 0;
>  }
> @@ -308,6 +319,12 @@ esp_sa_init(struct rte_ipsec_sa *sa, const struct rte_ipsec_sa_prm *prm,
>  	static const uint64_t msk = RTE_IPSEC_SATP_DIR_MASK |
>  				RTE_IPSEC_SATP_MODE_MASK;
> 
> +	if (prm->ipsec_xform.options.ecn)
> +		sa->tos_mask |= ECN_MASK;
> +
> +	if (prm->ipsec_xform.options.copy_dscp)
> +		sa->tos_mask |= DSCP_MASK;
> +
>  	if (cxf->aead != NULL) {
>  		switch (cxf->aead->algo) {
>  		case RTE_CRYPTO_AEAD_AES_GCM:
> diff --git a/lib/librte_ipsec/sa.h b/lib/librte_ipsec/sa.h
> index ffb5fb4f8..41e0b78c9 100644
> --- a/lib/librte_ipsec/sa.h
> +++ b/lib/librte_ipsec/sa.h
> @@ -10,6 +10,7 @@
>  #define IPSEC_MAX_HDR_SIZE	64
>  #define IPSEC_MAX_IV_SIZE	16
>  #define IPSEC_MAX_IV_QWORD	(IPSEC_MAX_IV_SIZE / sizeof(uint64_t))
> +#define INB_TUN_HDR_MSK (RTE_IPSEC_SATP_ECN_MASK | RTE_IPSEC_SATP_DSCP_MASK)
> 
>  /* padding alignment for different algorithms */
>  enum {
> @@ -103,6 +104,7 @@ struct rte_ipsec_sa {
>  	uint8_t iv_ofs; /* offset for algo-specific IV inside crypto op */
>  	uint8_t iv_len;
>  	uint8_t pad_align;
> +	uint8_t tos_mask;
> 
>  	/* template for tunnel header */
>  	uint8_t hdr[IPSEC_MAX_HDR_SIZE];
> diff --git a/lib/librte_net/rte_ip.h b/lib/librte_net/rte_ip.h
> index f9b909090..6592637f7 100644
> --- a/lib/librte_net/rte_ip.h
> +++ b/lib/librte_net/rte_ip.h
> @@ -47,6 +47,14 @@ struct ipv4_hdr {
>  					   (((c) & 0xff) << 8)  | \
>  					   ((d) & 0xff))
> 
> +
> +/** RFC 3168 */
> +#define ECN_MASK	(0x03)
> +#define ECN_CE		ECN_MASK
> +
> +/** Packet Option Masks */
> +#define DSCP_MASK		(0xFC)


Might be worth to add some prefix: IP_ECN_...
Or even RTE_IP_ECN_...

> +
>  /** Maximal IPv4 packet length (including a header) */
>  #define IPV4_MAX_PKT_LEN        65535
> 
> diff --git a/lib/librte_security/rte_security.h b/lib/librte_security/rte_security.h
> index 76f54e0e0..577eff766 100644
> --- a/lib/librte_security/rte_security.h
> +++ b/lib/librte_security/rte_security.h
> @@ -163,6 +163,15 @@ struct rte_security_ipsec_sa_options {
>  	 * * 0: Inner packet is not modified.
>  	 */
>  	uint32_t dec_ttl : 1;
> +
> +	/**< Explicit Congestion Notification (ECN)
> +	 *
> +	 * * ECT(1) (ECN-Capable Transport(1))
> +	 * * ECT(0) (ECN-Capable Transport(0))
> +	 * * ECT(CE)(CE (Congestion Experienced))

I think, that comment (possible ECN values) better move into rte_ip.h.
And here explain briefly what would be behavior for ipsec implementation
for 0/1 values.

> +	 */
> +
> +	uint32_t ecn : 1;
>  };
> 
>  /** IPSec security association direction */
> --
> 2.13.6