From: Ivan Malov <ivan.malov@oktetlabs.ru>
To: dev@dpdk.org
Cc: stable@dpdk.org, Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>,
Andy Moreton <amoreton@xilinx.com>,
Igor Romanov <igor.romanov@oktetlabs.ru>,
Robert Stonehouse <rstonehouse@solarflare.com>,
Mark Spender <mspender@solarflare.com>,
Andrew Lee <alee@solarflare.com>
Subject: [dpdk-stable] [PATCH] net/sfc: add 8000-series EF10 hardware TSO bug workaround
Date: Tue, 2 Feb 2021 18:23:45 +0300 [thread overview]
Message-ID: <20210202152345.2878-1-ivan.malov@oktetlabs.ru> (raw)
Innermost IP length and outer UDP datagram length must be
greater than or equal to the corresponding values derived
from the MSS; otherwise, the checksum offloads will break.
Fixes: c1ce2ba218f8 ("net/sfc: support tunnel TSO on EF10 native Tx datapath")
Fixes: 6bc985e41155 ("net/sfc: support TSO in EF10 Tx datapath")
Fixes: fec33d5bb3eb ("net/sfc: support firmware-assisted TSO")
Cc: stable@dpdk.org
Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru>
Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Reviewed-by: Andy Moreton <amoreton@xilinx.com>
---
drivers/net/sfc/sfc_ef10_tx.c | 19 +++++++++++++++++++
drivers/net/sfc/sfc_tso.c | 7 +++++++
drivers/net/sfc/sfc_tso.h | 30 ++++++++++++++++++++++++++++++
3 files changed, 56 insertions(+)
diff --git a/drivers/net/sfc/sfc_ef10_tx.c b/drivers/net/sfc/sfc_ef10_tx.c
index 87fa40f3e..33d2d637c 100644
--- a/drivers/net/sfc/sfc_ef10_tx.c
+++ b/drivers/net/sfc/sfc_ef10_tx.c
@@ -481,6 +481,25 @@ sfc_ef10_xmit_tso_pkt(struct sfc_ef10_txq * const txq, struct rte_mbuf *m_seg,
needed_desc--;
}
+ /*
+ * 8000-series EF10 hardware requires that innermost IP length
+ * be greater than or equal to the value which each segment is
+ * supposed to have; otherwise, TCP checksum will be incorrect.
+ *
+ * The same concern applies to outer UDP datagram length field.
+ */
+ switch (m_seg->ol_flags & PKT_TX_TUNNEL_MASK) {
+ case PKT_TX_TUNNEL_VXLAN:
+ /* FALLTHROUGH */
+ case PKT_TX_TUNNEL_GENEVE:
+ sfc_tso_outer_udp_fix_len(first_m_seg, hdr_addr);
+ break;
+ default:
+ break;
+ }
+
+ sfc_tso_innermost_ip_fix_len(first_m_seg, hdr_addr, iph_off);
+
/*
* Tx prepare has debug-only checks that offload flags are correctly
* filled in in TSO mbuf. Use zero IPID if there is no IPv4 flag.
diff --git a/drivers/net/sfc/sfc_tso.c b/drivers/net/sfc/sfc_tso.c
index d6f111989..b090ef14d 100644
--- a/drivers/net/sfc/sfc_tso.c
+++ b/drivers/net/sfc/sfc_tso.c
@@ -140,6 +140,13 @@ sfc_efx_tso_do(struct sfc_efx_txq *txq, unsigned int idx,
tsoh = rte_pktmbuf_mtod(m, uint8_t *);
}
+ /*
+ * 8000-series EF10 hardware requires that innermost IP length
+ * be greater than or equal to the value which each segment is
+ * supposed to have; otherwise, TCP checksum will be incorrect.
+ */
+ sfc_tso_innermost_ip_fix_len(m, tsoh, nh_off);
+
/*
* Handle IP header. Tx prepare has debug-only checks that offload flags
* are correctly filled in in TSO mbuf. Use zero IPID if there is no
diff --git a/drivers/net/sfc/sfc_tso.h b/drivers/net/sfc/sfc_tso.h
index 8597c2868..361aa2219 100644
--- a/drivers/net/sfc/sfc_tso.h
+++ b/drivers/net/sfc/sfc_tso.h
@@ -38,6 +38,36 @@ sfc_tso_ip4_get_ipid(const uint8_t *pkt_hdrp, size_t ip_hdr_off)
return rte_be_to_cpu_16(ipid);
}
+static inline void
+sfc_tso_outer_udp_fix_len(const struct rte_mbuf *m, uint8_t *tsoh)
+{
+ rte_be16_t len = rte_cpu_to_be_16(m->l2_len + m->l3_len + m->l4_len +
+ m->tso_segsz);
+
+ rte_memcpy(tsoh + m->outer_l2_len + m->outer_l3_len +
+ offsetof(struct rte_udp_hdr, dgram_len),
+ &len, sizeof(len));
+}
+
+static inline void
+sfc_tso_innermost_ip_fix_len(const struct rte_mbuf *m, uint8_t *tsoh,
+ size_t iph_ofst)
+{
+ size_t ip_payload_len = m->l4_len + m->tso_segsz;
+ size_t field_ofst;
+ rte_be16_t len;
+
+ if (m->ol_flags & PKT_TX_IPV4) {
+ field_ofst = offsetof(struct rte_ipv4_hdr, total_length);
+ len = rte_cpu_to_be_16(m->l3_len + ip_payload_len);
+ } else {
+ field_ofst = offsetof(struct rte_ipv6_hdr, payload_len);
+ len = rte_cpu_to_be_16(ip_payload_len);
+ }
+
+ rte_memcpy(tsoh + iph_ofst + field_ofst, &len, sizeof(len));
+}
+
unsigned int sfc_tso_prepare_header(uint8_t *tsoh, size_t header_len,
struct rte_mbuf **in_seg, size_t *in_off);
--
2.20.1
next reply other threads:[~2021-02-02 15:24 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-02-02 15:23 Ivan Malov [this message]
2021-02-03 10:37 ` Ferruh Yigit
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210202152345.2878-1-ivan.malov@oktetlabs.ru \
--to=ivan.malov@oktetlabs.ru \
--cc=alee@solarflare.com \
--cc=amoreton@xilinx.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=igor.romanov@oktetlabs.ru \
--cc=mspender@solarflare.com \
--cc=rstonehouse@solarflare.com \
--cc=stable@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).