* [dpdk-dev] [PATCH 2/2] net/mlx5: fix L4 packet type support
2017-07-26 19:29 [dpdk-dev] [PATCH 0/2] net/mlx5: fix packet type support Yongseok Koh
2017-07-26 19:29 ` [dpdk-dev] [PATCH 1/2] net/mlx5: fix missing packet type calculation Yongseok Koh
@ 2017-07-26 19:29 ` Yongseok Koh
2017-07-30 9:08 ` Sagi Grimberg
2017-07-31 16:29 ` [dpdk-dev] [PATCH 0/2] net/mlx5: fix " Ferruh Yigit
2 siblings, 1 reply; 6+ messages in thread
From: Yongseok Koh @ 2017-07-26 19:29 UTC (permalink / raw)
To: adrien.mazarguil, nelio.laranjeiro; +Cc: dev, Yongseok Koh
TCP/UDP/NONFRAG/FRAG flags aren't counted for both outer and inner header
even though device supports it.
Fixes: 0603df73a077 ("net/mlx5: fix Rx packet validation and type")
Fixes: 6cb559d67b83 ("net/mlx5: add vectorized Rx/Tx burst for x86")
Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
---
drivers/net/mlx5/mlx5.c | 2 +
drivers/net/mlx5/mlx5_ethdev.c | 10 +-
drivers/net/mlx5/mlx5_rxtx.c | 172 ++++++++++++++++++++++++++---------
drivers/net/mlx5/mlx5_rxtx.h | 3 +-
drivers/net/mlx5/mlx5_rxtx_vec_sse.c | 17 ++--
5 files changed, 151 insertions(+), 53 deletions(-)
diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c
index 49d4dba2a..1f438476a 100644
--- a/drivers/net/mlx5/mlx5.c
+++ b/drivers/net/mlx5/mlx5.c
@@ -887,6 +887,8 @@ RTE_INIT(rte_mlx5_pmd_init);
static void
rte_mlx5_pmd_init(void)
{
+ /* Build the static table for ptype conversion. */
+ mlx5_set_ptype_table();
/*
* RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use
* huge pages. Calling ibv_fork_init() during init allows
diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index b70b7b9a9..d1ad61c20 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -715,12 +715,20 @@ mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev)
{
static const uint32_t ptypes[] = {
/* refers to rxq_cq_to_pkt_type() */
+ RTE_PTYPE_L2_ETHER,
RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
+ RTE_PTYPE_L4_NONFRAG,
+ RTE_PTYPE_L4_FRAG,
+ RTE_PTYPE_L4_TCP,
+ RTE_PTYPE_L4_UDP,
RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
+ RTE_PTYPE_INNER_L4_NONFRAG,
+ RTE_PTYPE_INNER_L4_FRAG,
+ RTE_PTYPE_INNER_L4_TCP,
+ RTE_PTYPE_INNER_L4_UDP,
RTE_PTYPE_UNKNOWN
-
};
if (dev->rx_pkt_burst == mlx5_rx_burst ||
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index 5a24e15ea..2572a1652 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -79,31 +79,123 @@ mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe,
static __rte_always_inline uint32_t
rxq_cq_to_ol_flags(struct rxq *rxq, volatile struct mlx5_cqe *cqe);
-/*
- * The index to the array should have:
- * bit[1:0] = l3_hdr_type, bit[2] = tunneled, bit[3] = outer_l3_type
- */
-const uint32_t mlx5_ptype_table[] = {
- RTE_PTYPE_UNKNOWN,
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN, /* b0001 */
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN, /* b0010 */
- RTE_PTYPE_UNKNOWN, RTE_PTYPE_UNKNOWN,
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
- RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN, /* b0101 */
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
- RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN, /* b0110 */
- RTE_PTYPE_UNKNOWN, RTE_PTYPE_UNKNOWN,
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN, /* b1001 */
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN, /* b1010 */
- RTE_PTYPE_UNKNOWN, RTE_PTYPE_UNKNOWN,
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
- RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN, /* b1101 */
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
- RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN, /* b1110 */
- RTE_PTYPE_ALL_MASK /* b1111 */
+uint32_t mlx5_ptype_table[] __rte_cache_aligned = {
+ [0xff] = RTE_PTYPE_ALL_MASK, /* Last entry for errored packet. */
};
/**
+ * Build a table to translate Rx completion flags to packet type.
+ *
+ * @note: fix mlx5_dev_supported_ptypes_get() if any change here.
+ */
+void
+mlx5_set_ptype_table(void)
+{
+ unsigned int i;
+ uint32_t (*p)[RTE_DIM(mlx5_ptype_table)] = &mlx5_ptype_table;
+
+ for (i = 0; i < RTE_DIM(mlx5_ptype_table); ++i)
+ (*p)[i] = RTE_PTYPE_UNKNOWN;
+ /*
+ * The index to the array should have:
+ * bit[1:0] = l3_hdr_type
+ * bit[4:2] = l4_hdr_type
+ * bit[5] = ip_frag
+ * bit[6] = tunneled
+ * bit[7] = outer_l3_type
+ */
+ /* L3 */
+ (*p)[0x01] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_NONFRAG;
+ (*p)[0x02] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_NONFRAG;
+ /* Fragmented */
+ (*p)[0x21] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_FRAG;
+ (*p)[0x22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_FRAG;
+ /* TCP */
+ (*p)[0x05] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0x06] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ /* UDP */
+ (*p)[0x09] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ (*p)[0x0a] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ /* Repeat with outer_l3_type being set. Just in case. */
+ (*p)[0x81] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_NONFRAG;
+ (*p)[0x82] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_NONFRAG;
+ (*p)[0xa1] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_FRAG;
+ (*p)[0xa2] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_FRAG;
+ (*p)[0x85] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0x86] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0x89] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ (*p)[0x8a] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ /* Tunneled - L3 */
+ (*p)[0x41] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_NONFRAG;
+ (*p)[0x42] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_NONFRAG;
+ (*p)[0xc1] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_NONFRAG;
+ (*p)[0xc2] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_NONFRAG;
+ /* Tunneled - Fragmented */
+ (*p)[0x61] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_FRAG;
+ (*p)[0x62] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_FRAG;
+ (*p)[0xe1] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_FRAG;
+ (*p)[0xe2] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L4_FRAG;
+ /* Tunneled - TCP */
+ (*p)[0x45] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0x46] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0xc5] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ (*p)[0xc6] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_TCP;
+ /* Tunneled - UDP */
+ (*p)[0x49] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ (*p)[0x4a] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ (*p)[0xc9] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+ (*p)[0xca] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
+ RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN |
+ RTE_PTYPE_L4_UDP;
+}
+
+/**
* Return the size of tailroom of WQ.
*
* @param txq
@@ -1501,30 +1593,20 @@ mlx5_tx_burst_empw(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)
static inline uint32_t
rxq_cq_to_pkt_type(volatile struct mlx5_cqe *cqe)
{
- uint32_t pkt_type;
- uint16_t flags = ntohs(cqe->hdr_type_etc);
+ uint8_t idx;
+ uint8_t pinfo = cqe->pkt_info;
+ uint16_t ptype = cqe->hdr_type_etc;
- if (cqe->pkt_info & MLX5_CQE_RX_TUNNEL_PACKET) {
- pkt_type =
- TRANSPOSE(flags,
- MLX5_CQE_RX_IPV4_PACKET,
- RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN) |
- TRANSPOSE(flags,
- MLX5_CQE_RX_IPV6_PACKET,
- RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN);
- pkt_type |= ((cqe->pkt_info & MLX5_CQE_RX_OUTER_PACKET) ?
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN :
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN);
- } else {
- pkt_type =
- TRANSPOSE(flags,
- MLX5_CQE_L3_HDR_TYPE_IPV6,
- RTE_PTYPE_L3_IPV6_EXT_UNKNOWN) |
- TRANSPOSE(flags,
- MLX5_CQE_L3_HDR_TYPE_IPV4,
- RTE_PTYPE_L3_IPV4_EXT_UNKNOWN);
- }
- return pkt_type;
+ /*
+ * The index to the array should have:
+ * bit[1:0] = l3_hdr_type
+ * bit[4:2] = l4_hdr_type
+ * bit[5] = ip_frag
+ * bit[6] = tunneled
+ * bit[7] = outer_l3_type
+ */
+ idx = ((pinfo & 0x3) << 6) | ((ptype & 0xfc00) >> 10);
+ return mlx5_ptype_table[idx];
}
/**
diff --git a/drivers/net/mlx5/mlx5_rxtx.h b/drivers/net/mlx5/mlx5_rxtx.h
index 534aaeb46..09882e173 100644
--- a/drivers/net/mlx5/mlx5_rxtx.h
+++ b/drivers/net/mlx5/mlx5_rxtx.h
@@ -334,8 +334,9 @@ uint16_t mlx5_tx_burst_secondary_setup(void *, struct rte_mbuf **, uint16_t);
/* mlx5_rxtx.c */
-extern const uint32_t mlx5_ptype_table[];
+extern uint32_t mlx5_ptype_table[];
+void mlx5_set_ptype_table(void);
uint16_t mlx5_tx_burst(void *, struct rte_mbuf **, uint16_t);
uint16_t mlx5_tx_burst_mpw(void *, struct rte_mbuf **, uint16_t);
uint16_t mlx5_tx_burst_mpw_inline(void *, struct rte_mbuf **, uint16_t);
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.c b/drivers/net/mlx5/mlx5_rxtx_vec_sse.c
index 95e96baa2..d43ab51d0 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.c
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.c
@@ -756,7 +756,7 @@ rxq_cq_to_ptype_oflags_v(struct rxq *rxq, __m128i cqes[4], __m128i op_err,
__m128i cv_flags;
const __m128i zero = _mm_setzero_si128();
const __m128i ptype_mask =
- _mm_set_epi32(0xd06, 0xd06, 0xd06, 0xd06);
+ _mm_set_epi32(0xfd06, 0xfd06, 0xfd06, 0xfd06);
const __m128i ptype_ol_mask =
_mm_set_epi32(0x106, 0x106, 0x106, 0x106);
const __m128i pinfo_mask =
@@ -817,18 +817,23 @@ rxq_cq_to_ptype_oflags_v(struct rxq *rxq, __m128i cqes[4], __m128i op_err,
}
/*
* Merge the two fields to generate the following:
- * bit[1] = l3_ok, bit[2] = l4_ok
- * bit[8] = cv, bit[11:10] = l3_hdr_type
- * bit[12] = tunneled, bit[13] = outer_l3_type
+ * bit[1] = l3_ok
+ * bit[2] = l4_ok
+ * bit[8] = cv
+ * bit[11:10] = l3_hdr_type
+ * bit[14:12] = l4_hdr_type
+ * bit[15] = ip_frag
+ * bit[16] = tunneled
+ * bit[17] = outer_l3_type
*/
ptype = _mm_and_si128(ptype, ptype_mask);
pinfo = _mm_and_si128(pinfo, pinfo_mask);
- pinfo = _mm_slli_epi32(pinfo, 12);
+ pinfo = _mm_slli_epi32(pinfo, 16);
ptype = _mm_or_si128(ptype, pinfo);
ptype = _mm_srli_epi32(ptype, 10);
ptype = _mm_packs_epi32(ptype, zero);
/* Errored packets will have RTE_PTYPE_ALL_MASK. */
- op_err = _mm_srli_epi16(op_err, 12);
+ op_err = _mm_srli_epi16(op_err, 8);
ptype = _mm_or_si128(ptype, op_err);
pkts[0]->packet_type = mlx5_ptype_table[_mm_extract_epi8(ptype, 0)];
pkts[1]->packet_type = mlx5_ptype_table[_mm_extract_epi8(ptype, 2)];
--
2.11.0
^ permalink raw reply [flat|nested] 6+ messages in thread