mirror of
https://github.com/xdp-project/bpf-examples.git
synced 2024-05-06 15:54:53 +00:00
1010e53065
Make tc pin the ts_start map when loading the TC-BPF program, and rewrite XDP loader to reuse map pinned by tc. Also add comment with TODO list in pping.c. Testing pping by adding a delay through a netem qdisc in the test environment shows that the reported RTT will approach 100ms for any delay lower than 100ms, but the correct RTT for any delay over 100ms. Root cause is unknown, but Pollere's original pping implementation (as well as a bpftrace based pping implementation) shows the same issue. This issue has not been observed when running on real interfaces without a netem qdisc. Signed-off-by: Simon Sundberg <simon.sundberg@kau.se>
78 lines
2.7 KiB
C
78 lines
2.7 KiB
C
#include <linux/bpf.h>
|
|
#include <bpf/bpf_helpers.h>
|
|
#include <xdp/parsing_helpers.h>
|
|
|
|
#include <linux/in.h>
|
|
#include <linux/if_ether.h>
|
|
#include <linux/ip.h>
|
|
#include <linux/tcp.h>
|
|
|
|
#include <string.h>
|
|
|
|
#include "pping.h"
|
|
#include "pping_helpers.h"
|
|
|
|
char _license[] SEC("license") = "GPL";
|
|
|
|
struct bpf_map_def SEC("maps") ts_start = {
|
|
.type = BPF_MAP_TYPE_HASH,
|
|
.key_size = sizeof(struct ts_key),
|
|
.value_size = sizeof(struct ts_timestamp),
|
|
.max_entries = 16384,
|
|
};
|
|
|
|
struct bpf_map_def SEC("maps") rtt_events = {
|
|
.type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
|
|
.key_size = sizeof(__u32), // CPU ID
|
|
.value_size = sizeof(__u32), // perf file descriptor?
|
|
};
|
|
|
|
// XDP program for parsing TSECR-val from ingress traffic and check for match in map
|
|
SEC("pping_ingress")
|
|
int xdp_prog_ingress(struct xdp_md *ctx)
|
|
{
|
|
void *data = (void *)(long)ctx->data;
|
|
void *data_end = (void *)(long)ctx->data_end;
|
|
int proto = -1;
|
|
struct hdr_cursor nh = {.pos = data };
|
|
struct ethhdr *eth;
|
|
struct iphdr *iph;
|
|
struct tcphdr *tcph;
|
|
|
|
//bpf_printk("Received packet of length %d\n", (int)(data_end - data));
|
|
proto = parse_ethhdr(&nh, data_end, ð);
|
|
if (bpf_ntohs(proto) != ETH_P_IP)
|
|
goto end; // Not IPv4 packet (or failed to parse ethernet header)
|
|
proto = parse_iphdr(&nh, data_end, &iph);
|
|
if (proto != IPPROTO_TCP)
|
|
goto end; // Not a TCP packet (or failed to parse ethernet header)
|
|
proto = parse_tcphdr(&nh, data_end, &tcph);
|
|
if (proto < 0)
|
|
goto end; // Failed parsing TCP-header
|
|
|
|
//bpf_printk("TCP-packet with %d byte header and %lu bytes of data\n", proto, data_end - nh.pos);
|
|
|
|
__u32 tsval, tsecr;
|
|
if (parse_tcp_ts(tcph, data_end, &tsval, &tsecr) < 0) // No TCP timestamp
|
|
goto end;
|
|
|
|
// We have a TCP-timestamp - now we can check if it's in the map
|
|
//bpf_printk("TCP-packet with timestap. TSval: %u, TSecr: %u\n", bpf_ntohl(tsval), bpf_ntohl(tsecr));
|
|
struct ts_key key;
|
|
fill_ipv4_flow(&(key.flow), iph->daddr, iph->saddr, tcph->dest, tcph->source); // Fill in reverse order of egress (dest <--> source)
|
|
key.tsval = tsecr;
|
|
struct ts_timestamp *ts = bpf_map_lookup_elem(&ts_start, &key);
|
|
if (ts && ts->used == 0) { // Only calculate RTT for first packet with matching TSecr
|
|
// As used is not set atomically with the lookup, could potentially have multiple "first" packets (on different CPUs), but all those should then also have very similar RTT, so don't consider it a significant issue
|
|
ts->used = 1;
|
|
|
|
struct rtt_event event = {0};
|
|
memcpy(&(event.flow), &(key.flow), sizeof(struct ipv4_flow));
|
|
event.rtt = bpf_ktime_get_ns() - ts->timestamp;
|
|
bpf_perf_event_output(ctx, &rtt_events, BPF_F_CURRENT_CPU, &event, sizeof(event));
|
|
//bpf_printk("Pushed rtt event with RTT: %llu\n", event.rtt);
|
|
}
|
|
end:
|
|
return XDP_PASS;
|
|
}
|