* [dpdk-dev] [PATCH v1 0/4] some bugs fixes
@ 2020-07-25 3:48 Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 1/4] net/hinic: modify csum offload process Xiaoyun wang
` (3 more replies)
0 siblings, 4 replies; 5+ messages in thread
From: Xiaoyun wang @ 2020-07-25 3:48 UTC (permalink / raw)
To: dev
Cc: ferruh.yigit, bluca, luoxianjun, luoxingyu, zhouguoyang,
yin.yinshi, david.yangxiaoliang, zhaohui8, zhengjingzhou,
guojian365, Xiaoyun wang
This patch modifies csum offload process, optimize Rx
performance for x86, set vhd type with 2 for no offset
in ovs offload scenario, make timeout not affected by
system time jump.
--
v1:
- modify csum offload process
- optimize Rx performance for x86
- modify vhd type for SDI
- make timeout not affected by system time jump
Xiaoyun wang (4):
net/hinic: modify csum offload process
net/hinic: optimize Rx performance for x86
net/hinic/base: modify vhd type for SDI
net/hinic/base: make timeout not affected by system time jump
drivers/net/hinic/base/hinic_compat.h | 11 +-
drivers/net/hinic/base/hinic_pmd_nicio.c | 2 +-
drivers/net/hinic/base/hinic_pmd_nicio.h | 5 +
drivers/net/hinic/hinic_pmd_rx.h | 4 +
drivers/net/hinic/hinic_pmd_tx.c | 371 +++++++++++++++++--------------
5 files changed, 218 insertions(+), 175 deletions(-)
--
1.8.3.1
^ permalink raw reply [flat|nested] 5+ messages in thread
* [dpdk-dev] [PATCH v1 1/4] net/hinic: modify csum offload process
2020-07-25 3:48 [dpdk-dev] [PATCH v1 0/4] some bugs fixes Xiaoyun wang
@ 2020-07-25 3:48 ` Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 2/4] net/hinic: optimize Rx performance for x86 Xiaoyun wang
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Xiaoyun wang @ 2020-07-25 3:48 UTC (permalink / raw)
To: dev
Cc: ferruh.yigit, bluca, luoxianjun, luoxingyu, zhouguoyang,
yin.yinshi, david.yangxiaoliang, zhaohui8, zhengjingzhou,
guojian365, Xiaoyun wang
Encapsulate different types of packet checksum preprocessing
into functions.
Signed-off-by: Xiaoyun wang <cloud.wangxiaoyun@huawei.com>
---
drivers/net/hinic/hinic_pmd_tx.c | 371 +++++++++++++++++++++------------------
1 file changed, 202 insertions(+), 169 deletions(-)
diff --git a/drivers/net/hinic/hinic_pmd_tx.c b/drivers/net/hinic/hinic_pmd_tx.c
index 4d99967..3ef15b2 100644
--- a/drivers/net/hinic/hinic_pmd_tx.c
+++ b/drivers/net/hinic/hinic_pmd_tx.c
@@ -38,9 +38,6 @@
#define HINIC_TSO_PKT_MAX_SGE 127 /* tso max sge 127 */
#define HINIC_TSO_SEG_NUM_INVALID(num) ((num) > HINIC_TSO_PKT_MAX_SGE)
-#define HINIC_TX_OUTER_CHECKSUM_FLAG_SET 1
-#define HINIC_TX_OUTER_CHECKSUM_FLAG_NO_SET 0
-
/* sizeof(struct hinic_sq_bufdesc) == 16, shift 4 */
#define HINIC_BUF_DESC_SIZE(nr_descs) (SIZE_8BYTES(((u32)nr_descs) << 4))
@@ -671,7 +668,7 @@ static inline void hinic_xmit_mbuf_cleanup(struct hinic_txq *txq)
static inline struct hinic_sq_wqe *
hinic_get_sq_wqe(struct hinic_txq *txq, int wqebb_cnt,
- struct hinic_wqe_info *wqe_info)
+ struct hinic_wqe_info *wqe_info)
{
u32 cur_pi, end_pi;
u16 remain_wqebbs;
@@ -758,36 +755,33 @@ static inline void hinic_xmit_mbuf_cleanup(struct hinic_txq *txq)
return __rte_raw_cksum_reduce(sum);
}
-static inline void
-hinic_get_pld_offset(struct rte_mbuf *m, struct hinic_tx_offload_info *off_info,
- int outer_cs_flag)
+static inline void hinic_get_outer_cs_pld_offset(struct rte_mbuf *m,
+ struct hinic_tx_offload_info *off_info)
{
uint64_t ol_flags = m->ol_flags;
- if (outer_cs_flag == 1) {
- if ((ol_flags & PKT_TX_UDP_CKSUM) == PKT_TX_UDP_CKSUM) {
- off_info->payload_offset = m->outer_l2_len +
- m->outer_l3_len + m->l2_len + m->l3_len;
- } else if ((ol_flags & PKT_TX_TCP_CKSUM) ||
- (ol_flags & PKT_TX_TCP_SEG)) {
- off_info->payload_offset = m->outer_l2_len +
- m->outer_l3_len + m->l2_len +
- m->l3_len + m->l4_len;
- }
- } else {
- if ((ol_flags & PKT_TX_UDP_CKSUM) == PKT_TX_UDP_CKSUM) {
- off_info->payload_offset = m->l2_len + m->l3_len;
- } else if ((ol_flags & PKT_TX_TCP_CKSUM) ||
- (ol_flags & PKT_TX_TCP_SEG)) {
- off_info->payload_offset = m->l2_len + m->l3_len +
- m->l4_len;
- }
- }
+ if ((ol_flags & PKT_TX_L4_MASK) == PKT_TX_UDP_CKSUM)
+ off_info->payload_offset = m->outer_l2_len + m->outer_l3_len +
+ m->l2_len + m->l3_len;
+ else if ((ol_flags & PKT_TX_TCP_CKSUM) || (ol_flags & PKT_TX_TCP_SEG))
+ off_info->payload_offset = m->outer_l2_len + m->outer_l3_len +
+ m->l2_len + m->l3_len + m->l4_len;
}
-static inline void
-hinic_analyze_tx_info(struct rte_mbuf *mbuf,
- struct hinic_tx_offload_info *off_info)
+static inline void hinic_get_pld_offset(struct rte_mbuf *m,
+ struct hinic_tx_offload_info *off_info)
+{
+ uint64_t ol_flags = m->ol_flags;
+
+ if ((ol_flags & PKT_TX_L4_MASK) == PKT_TX_UDP_CKSUM)
+ off_info->payload_offset = m->l2_len + m->l3_len;
+ else if ((ol_flags & PKT_TX_TCP_CKSUM) || (ol_flags & PKT_TX_TCP_SEG))
+ off_info->payload_offset = m->l2_len + m->l3_len +
+ m->l4_len;
+}
+
+static inline void hinic_analyze_tx_info(struct rte_mbuf *mbuf,
+ struct hinic_tx_offload_info *off_info)
{
struct rte_ether_hdr *eth_hdr;
struct rte_vlan_hdr *vlan_hdr;
@@ -817,17 +811,164 @@ static inline void hinic_xmit_mbuf_cleanup(struct hinic_txq *txq)
}
}
-static inline int
-hinic_tx_offload_pkt_prepare(struct rte_mbuf *m,
- struct hinic_tx_offload_info *off_info)
+static inline void hinic_analyze_outer_ip_vxlan(struct rte_mbuf *mbuf,
+ struct hinic_tx_offload_info *off_info)
+{
+ struct rte_ether_hdr *eth_hdr;
+ struct rte_vlan_hdr *vlan_hdr;
+ struct rte_ipv4_hdr *ipv4_hdr;
+ struct rte_udp_hdr *udp_hdr;
+ u16 eth_type = 0;
+
+ eth_hdr = rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr *);
+ eth_type = rte_be_to_cpu_16(eth_hdr->ether_type);
+
+ if (eth_type == RTE_ETHER_TYPE_VLAN) {
+ vlan_hdr = (struct rte_vlan_hdr *)(eth_hdr + 1);
+ eth_type = rte_be_to_cpu_16(vlan_hdr->eth_proto);
+ }
+
+ if (eth_type == RTE_ETHER_TYPE_IPV4) {
+ ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv4_hdr *,
+ mbuf->outer_l2_len);
+ off_info->outer_l3_type = IPV4_PKT_WITH_CHKSUM_OFFLOAD;
+ ipv4_hdr->hdr_checksum = 0;
+
+ udp_hdr = (struct rte_udp_hdr *)((char *)ipv4_hdr +
+ mbuf->outer_l3_len);
+ udp_hdr->dgram_cksum = 0;
+ } else if (eth_type == RTE_ETHER_TYPE_IPV6) {
+ off_info->outer_l3_type = IPV6_PKT;
+
+ udp_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_udp_hdr *,
+ (mbuf->outer_l2_len +
+ mbuf->outer_l3_len));
+ udp_hdr->dgram_cksum = 0;
+ }
+}
+
+static inline uint8_t hinic_analyze_l3_type(struct rte_mbuf *mbuf)
+{
+ uint8_t l3_type;
+ uint64_t ol_flags = mbuf->ol_flags;
+
+ if (ol_flags & PKT_TX_IPV4)
+ l3_type = (ol_flags & PKT_TX_IP_CKSUM) ?
+ IPV4_PKT_WITH_CHKSUM_OFFLOAD :
+ IPV4_PKT_NO_CHKSUM_OFFLOAD;
+ else if (ol_flags & PKT_TX_IPV6)
+ l3_type = IPV6_PKT;
+ else
+ l3_type = UNKNOWN_L3TYPE;
+
+ return l3_type;
+}
+
+static inline void hinic_calculate_tcp_checksum(struct rte_mbuf *mbuf,
+ struct hinic_tx_offload_info *off_info,
+ uint64_t inner_l3_offset)
{
struct rte_ipv4_hdr *ipv4_hdr;
struct rte_ipv6_hdr *ipv6_hdr;
struct rte_tcp_hdr *tcp_hdr;
+ uint64_t ol_flags = mbuf->ol_flags;
+
+ if (ol_flags & PKT_TX_IPV4) {
+ ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv4_hdr *,
+ inner_l3_offset);
+
+ if (ol_flags & PKT_TX_IP_CKSUM)
+ ipv4_hdr->hdr_checksum = 0;
+
+ tcp_hdr = (struct rte_tcp_hdr *)((char *)ipv4_hdr +
+ mbuf->l3_len);
+ tcp_hdr->cksum = hinic_ipv4_phdr_cksum(ipv4_hdr, ol_flags);
+ } else {
+ ipv6_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv6_hdr *,
+ inner_l3_offset);
+ tcp_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_tcp_hdr *,
+ (inner_l3_offset +
+ mbuf->l3_len));
+ tcp_hdr->cksum = hinic_ipv6_phdr_cksum(ipv6_hdr, ol_flags);
+ }
+
+ off_info->inner_l4_type = TCP_OFFLOAD_ENABLE;
+ off_info->inner_l4_tcp_udp = 1;
+}
+
+static inline void hinic_calculate_udp_checksum(struct rte_mbuf *mbuf,
+ struct hinic_tx_offload_info *off_info,
+ uint64_t inner_l3_offset)
+{
+ struct rte_ipv4_hdr *ipv4_hdr;
+ struct rte_ipv6_hdr *ipv6_hdr;
struct rte_udp_hdr *udp_hdr;
- struct rte_ether_hdr *eth_hdr;
- struct rte_vlan_hdr *vlan_hdr;
- u16 eth_type = 0;
+ uint64_t ol_flags = mbuf->ol_flags;
+
+ if (ol_flags & PKT_TX_IPV4) {
+ ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv4_hdr *,
+ inner_l3_offset);
+
+ if (ol_flags & PKT_TX_IP_CKSUM)
+ ipv4_hdr->hdr_checksum = 0;
+
+ udp_hdr = (struct rte_udp_hdr *)((char *)ipv4_hdr +
+ mbuf->l3_len);
+ udp_hdr->dgram_cksum = hinic_ipv4_phdr_cksum(ipv4_hdr,
+ ol_flags);
+ } else {
+ ipv6_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_ipv6_hdr *,
+ inner_l3_offset);
+
+ udp_hdr = rte_pktmbuf_mtod_offset(mbuf, struct rte_udp_hdr *,
+ (inner_l3_offset +
+ mbuf->l3_len));
+ udp_hdr->dgram_cksum = hinic_ipv6_phdr_cksum(ipv6_hdr,
+ ol_flags);
+ }
+
+ off_info->inner_l4_type = UDP_OFFLOAD_ENABLE;
+ off_info->inner_l4_tcp_udp = 1;
+}
+
+static inline void
+hinic_calculate_sctp_checksum(struct hinic_tx_offload_info *off_info)
+{
+ off_info->inner_l4_type = SCTP_OFFLOAD_ENABLE;
+ off_info->inner_l4_tcp_udp = 0;
+ off_info->inner_l4_len = sizeof(struct rte_sctp_hdr);
+}
+
+static inline void hinic_calculate_checksum(struct rte_mbuf *mbuf,
+ struct hinic_tx_offload_info *off_info,
+ uint64_t inner_l3_offset)
+{
+ uint64_t ol_flags = mbuf->ol_flags;
+
+ switch (ol_flags & PKT_TX_L4_MASK) {
+ case PKT_TX_UDP_CKSUM:
+ hinic_calculate_udp_checksum(mbuf, off_info, inner_l3_offset);
+ break;
+
+ case PKT_TX_TCP_CKSUM:
+ hinic_calculate_tcp_checksum(mbuf, off_info, inner_l3_offset);
+ break;
+
+ case PKT_TX_SCTP_CKSUM:
+ hinic_calculate_sctp_checksum(off_info);
+ break;
+
+ default:
+ if (ol_flags & PKT_TX_TCP_SEG)
+ hinic_calculate_tcp_checksum(mbuf, off_info,
+ inner_l3_offset);
+ break;
+ }
+}
+
+static inline int hinic_tx_offload_pkt_prepare(struct rte_mbuf *m,
+ struct hinic_tx_offload_info *off_info)
+{
uint64_t inner_l3_offset;
uint64_t ol_flags = m->ol_flags;
@@ -836,8 +977,8 @@ static inline void hinic_xmit_mbuf_cleanup(struct hinic_txq *txq)
return 0;
/* Support only vxlan offload */
- if ((ol_flags & PKT_TX_TUNNEL_MASK) &&
- !(ol_flags & PKT_TX_TUNNEL_VXLAN))
+ if (unlikely((ol_flags & PKT_TX_TUNNEL_MASK) &&
+ !(ol_flags & PKT_TX_TUNNEL_VXLAN)))
return -ENOTSUP;
#ifdef RTE_LIBRTE_ETHDEV_DEBUG
@@ -846,169 +987,61 @@ static inline void hinic_xmit_mbuf_cleanup(struct hinic_txq *txq)
#endif
if (ol_flags & PKT_TX_TUNNEL_VXLAN) {
+ off_info->tunnel_type = TUNNEL_UDP_NO_CSUM;
+
+ /* inner_l4_tcp_udp csum should be setted to calculate outter
+ * udp checksum when vxlan packets without inner l3 and l4
+ */
+ off_info->inner_l4_tcp_udp = 1;
+
if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) ||
(ol_flags & PKT_TX_OUTER_IPV6) ||
(ol_flags & PKT_TX_TCP_SEG)) {
inner_l3_offset = m->l2_len + m->outer_l2_len +
- m->outer_l3_len;
+ m->outer_l3_len;
off_info->outer_l2_len = m->outer_l2_len;
off_info->outer_l3_len = m->outer_l3_len;
/* just support vxlan tunneling pkt */
off_info->inner_l2_len = m->l2_len - VXLANLEN -
- sizeof(*udp_hdr);
- off_info->inner_l3_len = m->l3_len;
- off_info->inner_l4_len = m->l4_len;
+ sizeof(struct rte_udp_hdr);
off_info->tunnel_length = m->l2_len;
- off_info->tunnel_type = TUNNEL_UDP_NO_CSUM;
- hinic_get_pld_offset(m, off_info,
- HINIC_TX_OUTER_CHECKSUM_FLAG_SET);
+ hinic_analyze_outer_ip_vxlan(m, off_info);
+
+ hinic_get_outer_cs_pld_offset(m, off_info);
} else {
inner_l3_offset = m->l2_len;
hinic_analyze_tx_info(m, off_info);
/* just support vxlan tunneling pkt */
off_info->inner_l2_len = m->l2_len - VXLANLEN -
- sizeof(*udp_hdr) - off_info->outer_l2_len -
- off_info->outer_l3_len;
- off_info->inner_l3_len = m->l3_len;
- off_info->inner_l4_len = m->l4_len;
+ sizeof(struct rte_udp_hdr) -
+ off_info->outer_l2_len -
+ off_info->outer_l3_len;
off_info->tunnel_length = m->l2_len -
- off_info->outer_l2_len - off_info->outer_l3_len;
- off_info->tunnel_type = TUNNEL_UDP_NO_CSUM;
+ off_info->outer_l2_len -
+ off_info->outer_l3_len;
+ off_info->outer_l3_type = IPV4_PKT_NO_CHKSUM_OFFLOAD;
- hinic_get_pld_offset(m, off_info,
- HINIC_TX_OUTER_CHECKSUM_FLAG_NO_SET);
+ hinic_get_pld_offset(m, off_info);
}
} else {
inner_l3_offset = m->l2_len;
off_info->inner_l2_len = m->l2_len;
- off_info->inner_l3_len = m->l3_len;
- off_info->inner_l4_len = m->l4_len;
off_info->tunnel_type = NOT_TUNNEL;
- hinic_get_pld_offset(m, off_info,
- HINIC_TX_OUTER_CHECKSUM_FLAG_NO_SET);
+ hinic_get_pld_offset(m, off_info);
}
/* invalid udp or tcp header */
if (unlikely(off_info->payload_offset > MAX_PLD_OFFSET))
return -EINVAL;
- /* Process outter udp pseudo-header checksum */
- if ((ol_flags & PKT_TX_TUNNEL_VXLAN) && ((ol_flags & PKT_TX_TCP_SEG) ||
- (ol_flags & PKT_TX_OUTER_IP_CKSUM) ||
- (ol_flags & PKT_TX_OUTER_IPV6))) {
-
- /* inner_l4_tcp_udp csum should be setted to calculate outter
- * udp checksum when vxlan packets without inner l3 and l4
- */
- off_info->inner_l4_tcp_udp = 1;
-
- eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
- eth_type = rte_be_to_cpu_16(eth_hdr->ether_type);
-
- if (eth_type == RTE_ETHER_TYPE_VLAN) {
- vlan_hdr = (struct rte_vlan_hdr *)(eth_hdr + 1);
- eth_type = rte_be_to_cpu_16(vlan_hdr->eth_proto);
- }
-
- if (eth_type == RTE_ETHER_TYPE_IPV4) {
- ipv4_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
- m->outer_l2_len);
- off_info->outer_l3_type = IPV4_PKT_WITH_CHKSUM_OFFLOAD;
- ipv4_hdr->hdr_checksum = 0;
-
- udp_hdr = (struct rte_udp_hdr *)((char *)ipv4_hdr +
- m->outer_l3_len);
- udp_hdr->dgram_cksum = 0;
- } else if (eth_type == RTE_ETHER_TYPE_IPV6) {
- off_info->outer_l3_type = IPV6_PKT;
- ipv6_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
- m->outer_l2_len);
-
- udp_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_udp_hdr *,
- (m->outer_l2_len +
- m->outer_l3_len));
- udp_hdr->dgram_cksum = 0;
- }
- } else if (ol_flags & PKT_TX_OUTER_IPV4) {
- off_info->tunnel_type = TUNNEL_UDP_NO_CSUM;
- off_info->inner_l4_tcp_udp = 1;
- off_info->outer_l3_type = IPV4_PKT_NO_CHKSUM_OFFLOAD;
- }
-
- if (ol_flags & PKT_TX_IPV4)
- off_info->inner_l3_type = (ol_flags & PKT_TX_IP_CKSUM) ?
- IPV4_PKT_WITH_CHKSUM_OFFLOAD :
- IPV4_PKT_NO_CHKSUM_OFFLOAD;
- else if (ol_flags & PKT_TX_IPV6)
- off_info->inner_l3_type = IPV6_PKT;
+ off_info->inner_l3_len = m->l3_len;
+ off_info->inner_l4_len = m->l4_len;
+ off_info->inner_l3_type = hinic_analyze_l3_type(m);
/* Process the pseudo-header checksum */
- if ((ol_flags & PKT_TX_L4_MASK) == PKT_TX_UDP_CKSUM) {
- if (ol_flags & PKT_TX_IPV4) {
- ipv4_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
- inner_l3_offset);
-
- if (ol_flags & PKT_TX_IP_CKSUM)
- ipv4_hdr->hdr_checksum = 0;
-
- udp_hdr = (struct rte_udp_hdr *)((char *)ipv4_hdr +
- m->l3_len);
- udp_hdr->dgram_cksum =
- hinic_ipv4_phdr_cksum(ipv4_hdr, ol_flags);
- } else {
- ipv6_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
- inner_l3_offset);
-
- udp_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_udp_hdr *,
- (inner_l3_offset + m->l3_len));
- udp_hdr->dgram_cksum =
- hinic_ipv6_phdr_cksum(ipv6_hdr, ol_flags);
- }
-
- off_info->inner_l4_type = UDP_OFFLOAD_ENABLE;
- off_info->inner_l4_tcp_udp = 1;
- } else if (((ol_flags & PKT_TX_L4_MASK) == PKT_TX_TCP_CKSUM) ||
- (ol_flags & PKT_TX_TCP_SEG)) {
- if (ol_flags & PKT_TX_IPV4) {
- ipv4_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
- inner_l3_offset);
-
- if (ol_flags & PKT_TX_IP_CKSUM)
- ipv4_hdr->hdr_checksum = 0;
-
- /* non-TSO tcp */
- tcp_hdr = (struct rte_tcp_hdr *)((char *)ipv4_hdr +
- m->l3_len);
- tcp_hdr->cksum =
- hinic_ipv4_phdr_cksum(ipv4_hdr, ol_flags);
- } else {
- ipv6_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
- inner_l3_offset);
- /* non-TSO tcp */
- tcp_hdr =
- rte_pktmbuf_mtod_offset(m, struct rte_tcp_hdr *,
- (inner_l3_offset + m->l3_len));
- tcp_hdr->cksum =
- hinic_ipv6_phdr_cksum(ipv6_hdr, ol_flags);
- }
-
- off_info->inner_l4_type = TCP_OFFLOAD_ENABLE;
- off_info->inner_l4_tcp_udp = 1;
- } else if ((ol_flags & PKT_TX_L4_MASK) == PKT_TX_SCTP_CKSUM) {
- off_info->inner_l4_type = SCTP_OFFLOAD_ENABLE;
- off_info->inner_l4_tcp_udp = 0;
- off_info->inner_l4_len = sizeof(struct rte_sctp_hdr);
- }
+ hinic_calculate_checksum(m, off_info, inner_l3_offset);
return 0;
}
--
1.8.3.1
^ permalink raw reply [flat|nested] 5+ messages in thread
* [dpdk-dev] [PATCH v1 2/4] net/hinic: optimize Rx performance for x86
2020-07-25 3:48 [dpdk-dev] [PATCH v1 0/4] some bugs fixes Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 1/4] net/hinic: modify csum offload process Xiaoyun wang
@ 2020-07-25 3:48 ` Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 3/4] net/hinic/base: modify vhd type for SDI Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 4/4] net/hinic/base: make timeout not affected by system time jump Xiaoyun wang
3 siblings, 0 replies; 5+ messages in thread
From: Xiaoyun wang @ 2020-07-25 3:48 UTC (permalink / raw)
To: dev
Cc: ferruh.yigit, bluca, luoxianjun, luoxingyu, zhouguoyang,
yin.yinshi, david.yangxiaoliang, zhaohui8, zhengjingzhou,
guojian365, Xiaoyun wang, stable
For x86 platform, the rq cqe without cache aligned, which can
improve performace for some gateway scenarios.
Fixes: 361a9ccf81d6 ("net/hinic: optimize Rx performance")
Cc: stable@dpdk.org
Signed-off-by: Xiaoyun wang <cloud.wangxiaoyun@huawei.com>
---
drivers/net/hinic/hinic_pmd_rx.h | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/net/hinic/hinic_pmd_rx.h b/drivers/net/hinic/hinic_pmd_rx.h
index 49fa565..8a45f2d 100644
--- a/drivers/net/hinic/hinic_pmd_rx.h
+++ b/drivers/net/hinic/hinic_pmd_rx.h
@@ -35,7 +35,11 @@ struct hinic_rq_cqe {
u32 rss_hash;
u32 rsvd[4];
+#if defined(RTE_ARCH_ARM64)
} __rte_cache_aligned;
+#else
+};
+#endif
struct hinic_rq_cqe_sect {
struct hinic_sge sge;
--
1.8.3.1
^ permalink raw reply [flat|nested] 5+ messages in thread
* [dpdk-dev] [PATCH v1 3/4] net/hinic/base: modify vhd type for SDI
2020-07-25 3:48 [dpdk-dev] [PATCH v1 0/4] some bugs fixes Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 1/4] net/hinic: modify csum offload process Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 2/4] net/hinic: optimize Rx performance for x86 Xiaoyun wang
@ 2020-07-25 3:48 ` Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 4/4] net/hinic/base: make timeout not affected by system time jump Xiaoyun wang
3 siblings, 0 replies; 5+ messages in thread
From: Xiaoyun wang @ 2020-07-25 3:48 UTC (permalink / raw)
To: dev
Cc: ferruh.yigit, bluca, luoxianjun, luoxingyu, zhouguoyang,
yin.yinshi, david.yangxiaoliang, zhaohui8, zhengjingzhou,
guojian365, Xiaoyun wang
For ovs offload scenario, when fw processes the virtio header,
there is no need to offset; and for standard card scenarios,
fw does not care about the vhd_type parameter, so in order to
be compatible with the two scenarios, use 0 offset instead.
Signed-off-by: Xiaoyun wang <cloud.wangxiaoyun@huawei.com>
---
drivers/net/hinic/base/hinic_pmd_nicio.c | 2 +-
drivers/net/hinic/base/hinic_pmd_nicio.h | 5 +++++
2 files changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/net/hinic/base/hinic_pmd_nicio.c b/drivers/net/hinic/base/hinic_pmd_nicio.c
index 2914e99..576fe59 100644
--- a/drivers/net/hinic/base/hinic_pmd_nicio.c
+++ b/drivers/net/hinic/base/hinic_pmd_nicio.c
@@ -578,7 +578,7 @@ int hinic_init_qp_ctxts(struct hinic_hwdev *hwdev)
rx_buf_sz = nic_io->rq_buf_size;
/* update rx buf size to function table */
- err = hinic_set_rx_vhd_mode(hwdev, 0, rx_buf_sz);
+ err = hinic_set_rx_vhd_mode(hwdev, HINIC_VHD_TYPE_0B, rx_buf_sz);
if (err) {
PMD_DRV_LOG(ERR, "Set rx vhd mode failed, rc: %d", err);
return err;
diff --git a/drivers/net/hinic/base/hinic_pmd_nicio.h b/drivers/net/hinic/base/hinic_pmd_nicio.h
index 9a487d0..600c073 100644
--- a/drivers/net/hinic/base/hinic_pmd_nicio.h
+++ b/drivers/net/hinic/base/hinic_pmd_nicio.h
@@ -8,6 +8,11 @@
#define RX_BUF_LEN_16K 16384
#define RX_BUF_LEN_1_5K 1536
+/* vhd type */
+#define HINIC_VHD_TYPE_0B 2
+#define HINIC_VHD_TYPE_10B 1
+#define HINIC_VHD_TYPE_12B 0
+
#define HINIC_Q_CTXT_MAX 42
/* performance: ci addr RTE_CACHE_SIZE(64B) alignment */
--
1.8.3.1
^ permalink raw reply [flat|nested] 5+ messages in thread
* [dpdk-dev] [PATCH v1 4/4] net/hinic/base: make timeout not affected by system time jump
2020-07-25 3:48 [dpdk-dev] [PATCH v1 0/4] some bugs fixes Xiaoyun wang
` (2 preceding siblings ...)
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 3/4] net/hinic/base: modify vhd type for SDI Xiaoyun wang
@ 2020-07-25 3:48 ` Xiaoyun wang
3 siblings, 0 replies; 5+ messages in thread
From: Xiaoyun wang @ 2020-07-25 3:48 UTC (permalink / raw)
To: dev
Cc: ferruh.yigit, bluca, luoxianjun, luoxingyu, zhouguoyang,
yin.yinshi, david.yangxiaoliang, zhaohui8, zhengjingzhou,
guojian365, Xiaoyun wang, stable
Replace gettimeofday() with clock_gettime(CLOCK_MONOTONIC_RAW, &now),
the reason is same with this patch "make alarm not affected by
system time jump".
Fixes: 81d53291a466 ("net/hinic/base: add various headers")
Cc: stable@dpdk.org
Signed-off-by: Xiaoyun wang <cloud.wangxiaoyun@huawei.com>
---
drivers/net/hinic/base/hinic_compat.h | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/net/hinic/base/hinic_compat.h b/drivers/net/hinic/base/hinic_compat.h
index 2d21b7b..7036b03 100644
--- a/drivers/net/hinic/base/hinic_compat.h
+++ b/drivers/net/hinic/base/hinic_compat.h
@@ -166,16 +166,17 @@ static inline u32 readl(const volatile void *addr)
#define spin_lock(spinlock_prt) rte_spinlock_lock(spinlock_prt)
#define spin_unlock(spinlock_prt) rte_spinlock_unlock(spinlock_prt)
-static inline unsigned long get_timeofday_ms(void)
+static inline unsigned long clock_gettime_ms(void)
{
- struct timeval tv;
+ struct timespec tv;
- (void)gettimeofday(&tv, NULL);
+ (void)clock_gettime(CLOCK_MONOTONIC, &tv);
- return (unsigned long)tv.tv_sec * 1000 + tv.tv_usec / 1000;
+ return (unsigned long)tv.tv_sec * 1000 +
+ (unsigned long)tv.tv_nsec / 1000000;
}
-#define jiffies get_timeofday_ms()
+#define jiffies clock_gettime_ms()
#define msecs_to_jiffies(ms) (ms)
#define time_before(now, end) ((now) < (end))
--
1.8.3.1
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2020-07-25 3:49 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-07-25 3:48 [dpdk-dev] [PATCH v1 0/4] some bugs fixes Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 1/4] net/hinic: modify csum offload process Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 2/4] net/hinic: optimize Rx performance for x86 Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 3/4] net/hinic/base: modify vhd type for SDI Xiaoyun wang
2020-07-25 3:48 ` [dpdk-dev] [PATCH v1 4/4] net/hinic/base: make timeout not affected by system time jump Xiaoyun wang
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).