#include <rte_lcore.h>
#include <rte_atomic.h>
#include <rte_branch_prediction.h>
-#include <rte_ring.h>
#include <rte_memory.h>
#include <rte_mempool.h>
#include <rte_mbuf.h>
uint16_t outer_l3_len;
uint8_t outer_l4_proto;
uint16_t tso_segsz;
+ uint16_t tunnel_tso_segsz;
};
/* simplified GRE header */
ipv4_hdr->hdr_checksum = 0;
ol_flags |= PKT_TX_IPV4;
- if (info->tso_segsz != 0 && info->l4_proto == IPPROTO_TCP) {
+ if (info->l4_proto == IPPROTO_TCP &&
+ ((info->is_tunnel && info->tunnel_tso_segsz != 0) ||
+ (!info->is_tunnel && info->tso_segsz != 0))) {
ol_flags |= PKT_TX_IP_CKSUM;
} else {
if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_IP_CKSUM)
} else if (info->l4_proto == IPPROTO_TCP) {
tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + info->l3_len);
tcp_hdr->cksum = 0;
- if (info->tso_segsz != 0) {
+ if ((info->is_tunnel && info->tunnel_tso_segsz != 0) ||
+ (!info->is_tunnel && info->tso_segsz != 0)) {
ol_flags |= PKT_TX_TCP_SEG;
tcp_hdr->cksum = get_psd_sum(l3_hdr, info->ethertype,
ol_flags);
return ol_flags;
}
-/* Calculate the checksum of outer header (only vxlan is supported,
- * meaning IP + UDP). The caller already checked that it's a vxlan
- * packet */
+/* Calculate the checksum of outer header */
static uint64_t
process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
- uint16_t testpmd_ol_flags)
+ uint16_t testpmd_ol_flags, int tso_enabled)
{
struct ipv4_hdr *ipv4_hdr = outer_l3_hdr;
struct ipv6_hdr *ipv6_hdr = outer_l3_hdr;
if (info->outer_l4_proto != IPPROTO_UDP)
return ol_flags;
- /* outer UDP checksum is always done in software as we have no
- * hardware supporting it today, and no API for it. */
-
udp_hdr = (struct udp_hdr *)((char *)outer_l3_hdr + info->outer_l3_len);
+
+ /* outer UDP checksum is done in software as we have no hardware
+ * supporting it today, and no API for it. In the other side, for
+ * UDP tunneling, like VXLAN or Geneve, outer UDP checksum can be
+ * set to zero.
+ *
+ * If a packet will be TSOed into small packets by NIC, we cannot
+ * set/calculate a non-zero checksum, because it will be a wrong
+ * value after the packet be split into several small packets.
+ */
+ if (tso_enabled)
+ udp_hdr->dgram_cksum = 0;
+
/* do not recalculate udp cksum if it was 0 */
if (udp_hdr->dgram_cksum != 0) {
udp_hdr->dgram_cksum = 0;
return ol_flags;
}
+/*
+ * Helper function.
+ * Performs actual copying.
+ * Returns number of segments in the destination mbuf on success,
+ * or negative error code on failure.
+ */
+static int
+mbuf_copy_split(const struct rte_mbuf *ms, struct rte_mbuf *md[],
+ uint16_t seglen[], uint8_t nb_seg)
+{
+ uint32_t dlen, slen, tlen;
+ uint32_t i, len;
+ const struct rte_mbuf *m;
+ const uint8_t *src;
+ uint8_t *dst;
+
+ dlen = 0;
+ slen = 0;
+ tlen = 0;
+
+ dst = NULL;
+ src = NULL;
+
+ m = ms;
+ i = 0;
+ while (ms != NULL && i != nb_seg) {
+
+ if (slen == 0) {
+ slen = rte_pktmbuf_data_len(ms);
+ src = rte_pktmbuf_mtod(ms, const uint8_t *);
+ }
+
+ if (dlen == 0) {
+ dlen = RTE_MIN(seglen[i], slen);
+ md[i]->data_len = dlen;
+ md[i]->next = (i + 1 == nb_seg) ? NULL : md[i + 1];
+ dst = rte_pktmbuf_mtod(md[i], uint8_t *);
+ }
+
+ len = RTE_MIN(slen, dlen);
+ memcpy(dst, src, len);
+ tlen += len;
+ slen -= len;
+ dlen -= len;
+ src += len;
+ dst += len;
+
+ if (slen == 0)
+ ms = ms->next;
+ if (dlen == 0)
+ i++;
+ }
+
+ if (ms != NULL)
+ return -ENOBUFS;
+ else if (tlen != m->pkt_len)
+ return -EINVAL;
+
+ md[0]->nb_segs = nb_seg;
+ md[0]->pkt_len = tlen;
+ md[0]->vlan_tci = m->vlan_tci;
+ md[0]->vlan_tci_outer = m->vlan_tci_outer;
+ md[0]->ol_flags = m->ol_flags;
+ md[0]->tx_offload = m->tx_offload;
+
+ return nb_seg;
+}
+
+/*
+ * Allocate a new mbuf with up to tx_pkt_nb_segs segments.
+ * Copy packet contents and offload information into then new segmented mbuf.
+ */
+static struct rte_mbuf *
+pkt_copy_split(const struct rte_mbuf *pkt)
+{
+ int32_t n, rc;
+ uint32_t i, len, nb_seg;
+ struct rte_mempool *mp;
+ uint16_t seglen[RTE_MAX_SEGS_PER_PKT];
+ struct rte_mbuf *p, *md[RTE_MAX_SEGS_PER_PKT];
+
+ mp = current_fwd_lcore()->mbp;
+
+ if (tx_pkt_split == TX_PKT_SPLIT_RND)
+ nb_seg = random() % tx_pkt_nb_segs + 1;
+ else
+ nb_seg = tx_pkt_nb_segs;
+
+ memcpy(seglen, tx_pkt_seg_lengths, nb_seg * sizeof(seglen[0]));
+
+ /* calculate number of segments to use and their length. */
+ len = 0;
+ for (i = 0; i != nb_seg && len < pkt->pkt_len; i++) {
+ len += seglen[i];
+ md[i] = NULL;
+ }
+
+ n = pkt->pkt_len - len;
+
+ /* update size of the last segment to fit rest of the packet */
+ if (n >= 0) {
+ seglen[i - 1] += n;
+ len += n;
+ }
+
+ nb_seg = i;
+ while (i != 0) {
+ p = rte_pktmbuf_alloc(mp);
+ if (p == NULL) {
+ RTE_LOG(ERR, USER1,
+ "failed to allocate %u-th of %u mbuf "
+ "from mempool: %s\n",
+ nb_seg - i, nb_seg, mp->name);
+ break;
+ }
+
+ md[--i] = p;
+ if (rte_pktmbuf_tailroom(md[i]) < seglen[i]) {
+ RTE_LOG(ERR, USER1, "mempool %s, %u-th segment: "
+ "expected seglen: %u, "
+ "actual mbuf tailroom: %u\n",
+ mp->name, i, seglen[i],
+ rte_pktmbuf_tailroom(md[i]));
+ break;
+ }
+ }
+
+ /* all mbufs successfully allocated, do copy */
+ if (i == 0) {
+ rc = mbuf_copy_split(pkt, md, seglen, nb_seg);
+ if (rc < 0)
+ RTE_LOG(ERR, USER1,
+ "mbuf_copy_split for %p(len=%u, nb_seg=%hhu) "
+ "into %u segments failed with error code: %d\n",
+ pkt, pkt->pkt_len, pkt->nb_segs, nb_seg, rc);
+
+ /* figure out how many mbufs to free. */
+ i = RTE_MAX(rc, 0);
+ }
+
+ /* free unused mbufs */
+ for (; i != nb_seg; i++) {
+ rte_pktmbuf_free_seg(md[i]);
+ md[i] = NULL;
+ }
+
+ return md[0];
+}
+
/*
* Receive a burst of packets, and for each packet:
* - parse packet, and try to recognize a supported packet type (1)
{
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
struct rte_port *txp;
- struct rte_mbuf *m;
+ struct rte_mbuf *m, *p;
struct ether_hdr *eth_hdr;
void *l3_hdr = NULL, *outer_l3_hdr = NULL; /* can be IPv4 or IPv6 */
uint16_t nb_rx;
uint16_t i;
uint64_t ol_flags;
uint16_t testpmd_ol_flags;
+ uint32_t retry;
uint32_t rx_bad_ip_csum;
uint32_t rx_bad_l4_csum;
struct testpmd_offload_info info;
testpmd_ol_flags = txp->tx_ol_flags;
memset(&info, 0, sizeof(info));
info.tso_segsz = txp->tso_segsz;
+ info.tunnel_tso_segsz = txp->tunnel_tso_segsz;
for (i = 0; i < nb_rx; i++) {
+ if (likely(i < nb_rx - 1))
+ rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[i + 1],
+ void *));
ol_flags = 0;
info.is_tunnel = 0;
if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_PARSE_TUNNEL) {
if (info.l4_proto == IPPROTO_UDP) {
struct udp_hdr *udp_hdr;
+
udp_hdr = (struct udp_hdr *)((char *)l3_hdr +
info.l3_len);
parse_vxlan(udp_hdr, &info, m->packet_type);
+ if (info.is_tunnel)
+ ol_flags |= PKT_TX_TUNNEL_VXLAN;
} else if (info.l4_proto == IPPROTO_GRE) {
struct simple_gre_hdr *gre_hdr;
+
gre_hdr = (struct simple_gre_hdr *)
((char *)l3_hdr + info.l3_len);
parse_gre(gre_hdr, &info);
+ if (info.is_tunnel)
+ ol_flags |= PKT_TX_TUNNEL_GRE;
} else if (info.l4_proto == IPPROTO_IPIP) {
void *encap_ip_hdr;
+
encap_ip_hdr = (char *)l3_hdr + info.l3_len;
parse_encap_ip(encap_ip_hdr, &info);
+ if (info.is_tunnel)
+ ol_flags |= PKT_TX_TUNNEL_IPIP;
}
}
* processed in hardware. */
if (info.is_tunnel == 1) {
ol_flags |= process_outer_cksums(outer_l3_hdr, &info,
- testpmd_ol_flags);
+ testpmd_ol_flags,
+ !!(ol_flags & PKT_TX_TCP_SEG));
}
/* step 4: fill the mbuf meta data (flags and header lengths) */
if (info.is_tunnel == 1) {
- if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM) {
+ if (info.tunnel_tso_segsz ||
+ testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM) {
m->outer_l2_len = info.outer_l2_len;
m->outer_l3_len = info.outer_l3_len;
m->l2_len = info.l2_len;
m->l3_len = info.l3_len;
m->l4_len = info.l4_len;
+ m->tso_segsz = info.tunnel_tso_segsz;
}
else {
/* if there is a outer UDP cksum
m->l2_len = info.l2_len;
m->l3_len = info.l3_len;
m->l4_len = info.l4_len;
+ m->tso_segsz = info.tso_segsz;
}
- m->tso_segsz = info.tso_segsz;
m->ol_flags = ol_flags;
+ /* Do split & copy for the packet. */
+ if (tx_pkt_split != TX_PKT_SPLIT_OFF) {
+ p = pkt_copy_split(m);
+ if (p != NULL) {
+ rte_pktmbuf_free(m);
+ m = p;
+ pkts_burst[i] = m;
+ }
+ }
+
/* if verbose mode is enabled, dump debug info */
if (verbose_level > 0) {
struct {
{ PKT_TX_OUTER_IPV4, PKT_TX_OUTER_IPV4 },
{ PKT_TX_OUTER_IPV6, PKT_TX_OUTER_IPV6 },
{ PKT_TX_TCP_SEG, PKT_TX_TCP_SEG },
+ { PKT_TX_TUNNEL_VXLAN, PKT_TX_TUNNEL_MASK },
+ { PKT_TX_TUNNEL_GRE, PKT_TX_TUNNEL_MASK },
+ { PKT_TX_TUNNEL_IPIP, PKT_TX_TUNNEL_MASK },
+ { PKT_TX_TUNNEL_GENEVE, PKT_TX_TUNNEL_MASK },
};
unsigned j;
const char *name;
printf("-----------------\n");
+ printf("mbuf=%p, pkt_len=%u, nb_segs=%hhu:\n",
+ m, m->pkt_len, m->nb_segs);
/* dump rx parsed packet info */
printf("rx: l2_len=%d ethertype=%x l3_len=%d "
"l4_proto=%d l4_len=%d\n",
printf("tx: m->l2_len=%d m->l3_len=%d "
"m->l4_len=%d\n",
m->l2_len, m->l3_len, m->l4_len);
- if ((info.is_tunnel == 1) &&
- (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM))
- printf("tx: m->outer_l2_len=%d m->outer_l3_len=%d\n",
- m->outer_l2_len, m->outer_l3_len);
- if (info.tso_segsz != 0)
+ if (info.is_tunnel == 1) {
+ if (testpmd_ol_flags &
+ TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM)
+ printf("tx: m->outer_l2_len=%d "
+ "m->outer_l3_len=%d\n",
+ m->outer_l2_len,
+ m->outer_l3_len);
+ if (info.tunnel_tso_segsz != 0)
+ printf("tx: m->tso_segsz=%d\n",
+ m->tso_segsz);
+ } else if (info.tso_segsz != 0)
printf("tx: m->tso_segsz=%d\n", m->tso_segsz);
printf("tx: flags=");
for (j = 0; j < sizeof(tx_flags)/sizeof(*tx_flags); j++) {
}
}
nb_tx = rte_eth_tx_burst(fs->tx_port, fs->tx_queue, pkts_burst, nb_rx);
+ /*
+ * Retry if necessary
+ */
+ if (unlikely(nb_tx < nb_rx) && fs->retry_enabled) {
+ retry = 0;
+ while (nb_tx < nb_rx && retry++ < burst_tx_retry_num) {
+ rte_delay_us(burst_tx_delay_time);
+ nb_tx += rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
+ &pkts_burst[nb_tx], nb_rx - nb_tx);
+ }
+ }
fs->tx_packets += nb_tx;
fs->rx_bad_ip_csum += rx_bad_ip_csum;
fs->rx_bad_l4_csum += rx_bad_l4_csum;
git.droids-corp.org / dpdk.git / blobdiff