[3/3] net/cnxk: update IP header of reassembled packets

Message ID 20250321100343.1081664-3-rbhansali@marvell.com (mailing list archive)
State Changes Requested
Delegated to: Jerin Jacob
Headers
Series [1/3] common/cnxk: update steer rule mbox for cn20k |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/loongarch-compilation success Compilation OK
ci/loongarch-unit-testing success Unit Testing PASS
ci/Intel-compilation success Compilation OK
ci/intel-Testing success Testing PASS
ci/intel-Functional success Functional PASS
ci/iol-marvell-Functional success Functional Testing PASS
ci/iol-broadcom-Performance success Performance Testing PASS
ci/github-robot: build fail github build: failed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-abi-testing success Testing PASS
ci/iol-unit-arm64-testing success Testing PASS
ci/iol-sample-apps-testing success Testing PASS
ci/iol-compile-arm64-testing success Testing PASS
ci/iol-compile-amd64-testing success Testing PASS
ci/iol-mellanox-Functional success Functional Testing PASS
ci/iol-unit-amd64-testing success Testing PASS

Commit Message

Rahul Bhansali March 21, 2025, 10:03 a.m. UTC
Updates IP header data of reassembled packets,
1) IPv4 length, MF bit and checksum correction
2) IPv6 length update and fragment extension header remove
Also, corrects data length of multisegment decrypted
packets.

Signed-off-by: Rahul Bhansali <rbhansali@marvell.com>
---
 drivers/net/cnxk/cn20k_rx.h | 112 ++++++++++++++++++++++++++++++++----
 1 file changed, 102 insertions(+), 10 deletions(-)
  

Comments

Jerin Jacob April 1, 2025, 10:24 a.m. UTC | #1
> -----Original Message-----
> From: Rahul Bhansali <rbhansali@marvell.com>
> Sent: Friday, March 21, 2025 3:34 PM
> To: dev@dpdk.org; Nithin Kumar Dabilpuram <ndabilpuram@marvell.com>;
> Kiran Kumar Kokkilagadda <kirankumark@marvell.com>; Sunil Kumar Kori
> <skori@marvell.com>; Satha Koteswara Rao Kottidi
> <skoteshwar@marvell.com>; Harman Kalra <hkalra@marvell.com>
> Cc: Jerin Jacob <jerinj@marvell.com>; Rahul Bhansali <rbhansali@marvell.com>
> Subject: [PATCH 3/3] net/cnxk: update IP header of reassembled packets
> 
> Updates IP header data of reassembled packets,
> 1) IPv4 length, MF bit and checksum correction
> 2) IPv6 length update and fragment extension header remove Also, corrects
> data length of multisegment decrypted packets.
> 
> Signed-off-by: Rahul Bhansali <rbhansali@marvell.com>


https://mails.dpdk.org/archives/test-report/2025-March/863905.html - fix this issue without performance impact.
  

Patch

diff --git a/drivers/net/cnxk/cn20k_rx.h b/drivers/net/cnxk/cn20k_rx.h
index b54d9df662..97a2b2b3d0 100644
--- a/drivers/net/cnxk/cn20k_rx.h
+++ b/drivers/net/cnxk/cn20k_rx.h
@@ -137,6 +137,71 @@  nix_get_mbuf_from_cqe(void *cq, const uint64_t data_off)
 	return (struct rte_mbuf *)(buff - data_off);
 }
 
+static __rte_always_inline void
+nix_sec_reass_first_frag_update(struct rte_mbuf *head, const rte_iova_t *iova_list, uintptr_t cpth,
+				uint64_t cq_w1, uint64_t cq_w5, uint16_t rlen)
+{
+	uint8_t *m_ipptr, *ipptr;
+	uint16_t tot_len;
+	uint32_t cksum;
+	uint8_t lc_ptr;
+	uint8_t lc_off;
+
+	lc_ptr = (cq_w5 >> 16) & 0xFF;
+	lc_off = lc_ptr - (cq_w5 & 0xFF);
+	ipptr = (uint8_t *)*iova_list + lc_off;
+	m_ipptr = (uint8_t *)cpth + lc_ptr;
+
+	/* Find the L3 header length and update inner pkt based on meta lc type */
+	if (((cq_w1 >> 40) & 0xF) == NPC_LT_LC_IP) {
+		const struct rte_ipv4_hdr *m_hdr = (const struct rte_ipv4_hdr *)m_ipptr;
+		struct rte_ipv4_hdr *ip_hdr = (struct rte_ipv4_hdr *)ipptr;
+
+		ip_hdr->fragment_offset = 0;
+		tot_len = rte_cpu_to_be_16(rlen);
+		ip_hdr->total_length = tot_len;
+		/* Perform incremental checksum based on meta pkt ip hdr */
+		cksum = m_hdr->hdr_checksum;
+		cksum += m_hdr->fragment_offset;
+		cksum += 0xFFFF;
+		cksum += m_hdr->total_length;
+		cksum += (uint16_t)(~tot_len);
+		cksum = (cksum & 0xFFFF) + ((cksum & 0xFFFF0000) >> 16);
+		ip_hdr->hdr_checksum = cksum;
+		return;
+	}
+
+	/* Assuming IPv6 packet update */
+	struct rte_ipv6_hdr *ipv6_hdr = (struct rte_ipv6_hdr *)ipptr;
+	size_t ext_len = sizeof(struct rte_ipv6_hdr);
+	uint8_t *nxt_hdr = (uint8_t *)ipv6_hdr;
+	uint8_t *nxt_proto = &ipv6_hdr->proto;
+	int nh = ipv6_hdr->proto;
+
+	tot_len = 0;
+	while (nh != -EINVAL) {
+		nxt_hdr += ext_len;
+		tot_len += ext_len;
+		if (nh == IPPROTO_FRAGMENT) {
+			*nxt_proto = *nxt_hdr;
+			break;
+		}
+		nh = rte_ipv6_get_next_ext(nxt_hdr, nh, &ext_len);
+		nxt_proto = nxt_hdr;
+	}
+
+	/* Remove the frag header by moving header 8 bytes forward */
+	ipv6_hdr->payload_len = rte_cpu_to_be_16(rlen - 8 - sizeof(struct rte_ipv6_hdr));
+
+	/* tot_len is sum of all IP header's length before fragment header */
+	rte_memcpy(rte_pktmbuf_mtod_offset(head, void *, 8), rte_pktmbuf_mtod(head, void *),
+		   lc_off + tot_len);
+
+	head->data_len -= 8;
+	head->data_off += 8;
+	head->pkt_len -= 8;
+}
+
 static __rte_always_inline uint64_t
 nix_sec_meta_to_mbuf_sc(uint64_t cq_w5, uint64_t cpth, const uint64_t sa_base,
 			struct rte_mbuf *mbuf, uint16_t *len, uint64_t *mbuf_init,
@@ -241,19 +306,22 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 {
 	const struct cpt_parse_hdr_s *hdr = (const struct cpt_parse_hdr_s *)cpth;
 	struct cn20k_inb_priv_data *inb_priv = NULL;
+	uint64_t fsz_w1 = 0, cq_w1, cq_w5, sg;
 	uint32_t offset = hdr->w2.ptr_offset;
 	const struct cpt_frag_info_s *finfo;
 	uint8_t num_frags = 0, nxt_frag = 0;
 	struct rte_mbuf *head, *last_mbuf;
-	uint64_t fsz_w1 = 0, cq_w1, sg;
+	uint16_t rlen = hdr->w3.rlen;
 	const rte_iova_t *iova_list;
 	uint8_t sg_cnt = 1, nb_segs;
 	uint16_t later_skip = 0;
 	bool reas_fail = false;
 	const rte_iova_t *eol;
+	uint16_t data_off = 0;
 	uint8_t ts_rx_off;
 	int dyn_off = 0;
-	uint32_t len;
+	uint16_t sg_len;
+	int64_t len;
 	uintptr_t p;
 
 	cq_w1 = *(const uint64_t *)rx;
@@ -266,10 +334,12 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 		if (!hdr->w4.gthr_size && ((flags & NIX_RX_REAS_F) || !hdr->w4.sctr_size))
 			return;
 
+		cq_w5 = *((const uint64_t *)rx + 4);
+		len = rlen + ((cq_w5 >> 16) & 0xFF) - (cq_w5 & 0xFF);
 		num_frags = hdr->w0.num_frags;
 		sg_base = cpth + (offset ? (offset << 3) : 256);
 		finfo = (const struct cpt_frag_info_s *)sg_base;
-		sg_base += num_frags ? (num_frags >> 2 ? 32 : 16) : 0;
+		sg_base += num_frags ? (num_frags > 4 ? 32 : 16) : 0;
 		sg = *(uint64_t *)sg_base;
 		nb_segs = (sg >> 48) & 0x3;
 		iova_list = (rte_iova_t *)(sg_base);
@@ -279,9 +349,8 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 		if ((flags & NIX_RX_REAS_F) && num_frags) {
 			void *inb_sa;
 
-			num_frags = hdr->w0.num_frags;
-			inb_sa = roc_nix_inl_ot_ipsec_inb_sa(sa_base, hdr->w0.cookie);
-			inb_priv = roc_nix_inl_ot_ipsec_inb_sa_sw_rsvd(inb_sa);
+			inb_sa = roc_nix_inl_ow_ipsec_inb_sa(sa_base, hdr->w0.cookie);
+			inb_priv = roc_nix_inl_ow_ipsec_inb_sa_sw_rsvd(inb_sa);
 			dyn_off = inb_priv->reass_dynfield_off;
 			num_frags -= 1;
 
@@ -300,16 +369,24 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 		if (nb_segs == 1)
 			return;
 
+		len = rx->pkt_lenm1 + 1;
+
 		/* Skip SG_S and first IOVA */
 		eol = ((const rte_iova_t *)(rx + 1) + ((rx->desc_sizem1 + 1) << 1));
 		iova_list = ((const rte_iova_t *)(rx + 1)) + 2;
 	}
 
 	/* Update data len as per the segment size */
-	mbuf->data_len = sg & 0xFFFF;
+	sg_len = sg & 0xFFFF;
+	mbuf->data_len = sg_len;
 	mbuf->nb_segs = nb_segs;
 	head = mbuf;
 
+	/* Update IP header */
+	if ((flags & NIX_RX_REAS_F) && num_frags && !reas_fail)
+		nix_sec_reass_first_frag_update(mbuf, iova_list - 1, cpth, cq_w1, cq_w5, rlen);
+
+	len -= sg_len;
 	sg = sg >> 16;
 	nb_segs--;
 
@@ -317,7 +394,7 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 
 	while (nb_segs) {
 		last_mbuf = mbuf;
-		if (flags & NIX_RX_REAS_F) {
+		if ((flags & NIX_RX_REAS_F) && num_frags) {
 			offset = (*iova_list) % (buf_sz & 0xFFFFFFFF);
 			mbuf->next = (struct rte_mbuf *)((*iova_list) - offset + (buf_sz >> 32));
 		} else {
@@ -344,6 +421,7 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 			len = fsz_w1 & 0xFFFF;
 			head->pkt_len = len - ts_rx_off;
 			head->nb_segs = sg_cnt;
+			data_off = rearm & 0xFFFF;
 			sg_cnt = 0;
 			nxt_frag = nxt_frag >> 1;
 			fsz_w1 = fsz_w1 >> 16;
@@ -351,14 +429,28 @@  nix_cqe_xtract_mseg(const union nix_rx_parse_u *rx, struct rte_mbuf *mbuf, uint6
 				fsz_w1 = finfo->w1.u64;
 		}
 
-		mbuf->data_len = (sg & 0xFFFF) - ts_rx_off;
+		if ((flags & NIX_RX_REAS_F) && num_frags && !reas_fail)
+			data_off = *iova_list - (uint64_t)mbuf->buf_addr;
+
+		sg_len = sg & 0xFFFF;
+		if ((flags & NIX_RX_OFFLOAD_SECURITY_F) && !(flags & NIX_RX_REAS_F)) {
+			/* Adjust last mbuf data length with negative offset for
+			 * security pkts if needed.
+			 */
+			len -= sg_len;
+			sg_len = (len > 0) ? sg_len : (sg_len + len);
+			len = (len > 0) ? len : 0;
+		}
+
+		mbuf->data_len = sg_len;
 		sg = sg >> 16;
 		p = (uintptr_t)&mbuf->rearm_data;
-		*(uint64_t *)p = rearm & ~0xFFFF;
+		*(uint64_t *)p = (rearm & ~0xFFFF) | data_off;
 
 		sg_cnt++;
 		nb_segs--;
 		iova_list++;
+		data_off = 0;
 
 		if (!nb_segs && (iova_list + 1 < eol)) {
 			sg = *(const uint64_t *)(iova_list);