mirror/xdp-project-bpf-examples
1
0
Fork 0
mirror of https://github.com/xdp-project/bpf-examples.git synced 2024-05-06 15:54:53 +00:00
Code Issues Projects Releases Wiki Activity

pping: Format headers and fix TODO.md

Browse Source 
Format the header files in the Linux kernel style (missed in previous
commit). Also fix a formating error in TODO.md that cause empty
checkboxes to not display correctly.

Signed-off-by: Simon Sundberg <[email protected]>
This commit is contained in:
Simon Sundberg
2021-02-04 19:43:39 +01:00
parent d3a5de62c4
commit b1ce4ee69b
3 changed files with 75 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pping/TODO.md
+8 -8
View File
@@ -4,14 +4,14 @@
- [x] Clean up commits and add signed-off-by tags
- [x] Add SPDX-license-identifier tags
- [x] Format C-code in kernel style
- [] Use existing funcionality to reuse maps by using BTF-defined maps
- [ ] Use existing funcionality to reuse maps by using BTF-defined maps
## Future
- [] Use libxdp to load XDP program
- [] Cleanup: Unload TC-BPF at program shutdown, and unpin and delete map - In userspace part
- [] Add IPv6 support - In TC-BPF, XDP and userspace part
- [] Check for existance of reverse flow before adding to hash-map (to avoid adding timestamps for flows that we can't see the reverse traffic for) - In TC-BPF part
- [ ] Use libxdp to load XDP program
- [ ] Cleanup: Unload TC-BPF at program shutdown, and unpin and delete map - In userspace part
- [ ] Add IPv6 support - In TC-BPF, XDP and userspace part
- [ ] Check for existance of reverse flow before adding to hash-map (to avoid adding timestamps for flows that we can't see the reverse traffic for) - In TC-BPF part
- This could miss the first few packets, would not be ideal for short flows
- [] Keep track of minimum RTT for each flow (done by Pollere's pping, and helps identify buffer bloat) - In XDP part
- [] Add configurable rate-limit for how often each flow can add entries to the map (prevent high-rate flows from quickly filling up the map) - In TCP-BPF part
- [] Improve map cleaning: Use a dynamic time to live for hash map entries based on flow's RTT, instead of static 10s limit - In TC-BPF, XDP and userspace
- [ ] Keep track of minimum RTT for each flow (done by Pollere's pping, and helps identify buffer bloat) - In XDP part
- [ ] Add configurable rate-limit for how often each flow can add entries to the map (prevent high-rate flows from quickly filling up the map) - In TCP-BPF part
- [ ] Improve map cleaning: Use a dynamic time to live for hash map entries based on flow's RTT, instead of static 10s limit - In TC-BPF, XDP and userspace
pping/pping.h
+16 -21
View File
@@ -3,33 +3,28 @@
#define TIMESTAMP_MAP_H
#include <linux/types.h>
struct ipv4_flow
{
__u32 saddr;
__u32 daddr;
__u16 sport;
__u16 dport;
struct ipv4_flow {
__u32 saddr;
__u32 daddr;
__u16 sport;
__u16 dport;
};
struct ts_key
{
struct ipv4_flow flow;
__u32 tsval;
struct ts_key {
struct ipv4_flow flow;
__u32 tsval;
};
struct ts_timestamp
{
__u64 timestamp;
//__u64 ttl; // Delete entry after ttl, allows more dynamic map cleaning where entries for flows with short RTTs can be removed earlier
__u8 used;
// __u8 pad[7]; // Need to pad it due to compiler optimization, see "Remove struct padding with aligning members by using #pragma pack." at https://docs.cilium.io/en/v1.9/bpf/
struct ts_timestamp {
__u64 timestamp;
//__u64 ttl; // Delete entry after ttl, allows more dynamic map cleaning where entries for flows with short RTTs can be removed earlier
__u8 used;
// __u8 pad[7]; // Need to pad it due to compiler optimization, see "Remove struct padding with aligning members by using #pragma pack." at https://docs.cilium.io/en/v1.9/bpf/
};
struct rtt_event
{
struct ipv4_flow flow;
__u64 rtt;
struct rtt_event {
struct ipv4_flow flow;
__u64 rtt;
};
#endif
pping/pping_helpers.h
+51 -44
View File
@@ -5,55 +5,62 @@
#include "pping.h"
#define MAX_TCP_OPTIONS 10
static __always_inline int fill_ipv4_flow(struct ipv4_flow *flow, __u32 saddr, __u32 daddr, __u16 sport, __u16 dport)
static __always_inline int fill_ipv4_flow(struct ipv4_flow *flow, __u32 saddr,
__u32 daddr, __u16 sport, __u16 dport)
{
flow->saddr = saddr;
flow->daddr = daddr;
flow->sport = sport;
flow->dport = dport;
return 0;
flow->saddr = saddr;
flow->daddr = daddr;
flow->sport = sport;
flow->dport = dport;
return 0;
}
// Parses the TSval and TSecr values from the TCP options field - returns 0 if sucessful and -1 on failure
// If sucessful the TSval and TSecr values will be stored at tsval and tsecr (in network byte order!)
static __always_inline int parse_tcp_ts(struct tcphdr *tcph, void *data_end, __u32 *tsval, __u32 *tsecr)
/*
* Parses the TSval and TSecr values from the TCP options field. If sucessful
* the TSval and TSecr values will be stored at tsval and tsecr (in network
* byte order).
* Returns 0 if sucessful and -1 on failure
*/
static __always_inline int parse_tcp_ts(struct tcphdr *tcph, void *data_end,
__u32 *tsval, __u32 *tsecr)
{
if (tcph + 1 > data_end) // To hopefully please verifier
return -1;
int len = tcph->doff << 2;
if (len <= sizeof(struct tcphdr)) // No TCP options
return -1;
void *pos = (void *)(tcph + 1);
void *opt_end = ((void *)tcph + len);
__u8 i, opt, opt_size;
#pragma unroll
for (i = 0; i < MAX_TCP_OPTIONS; i++) {
if (pos+1 > opt_end || pos+1 > data_end)
return -1;
opt = *(__u8 *)pos; // Save value to variable so I don't have to perform any more data_end checks on the option kind
if (opt == 0) // Reached end of TCP options
return -1;
if (opt == 1) {// TCP NOP option - advance one byte
pos++;
continue;
}
// Option > 1, should have option size
if (pos+2 > opt_end || pos+2 > data_end)
return -1;
opt_size = *(__u8 *)(pos+1); // Save value to variable so I don't have to perform any more data_end checks on option size
if (tcph + 1 > data_end)
return -1;
int len = tcph->doff << 2;
if (len <= sizeof(struct tcphdr)) // No TCP options
return -1;
void *pos = (void *)(tcph + 1);
void *opt_end = ((void *)tcph + len);
__u8 i, opt, opt_size;
#pragma unroll
for (i = 0; i < MAX_TCP_OPTIONS; i++) {
if (pos + 1 > opt_end || pos + 1 > data_end)
return -1;
opt = *(__u8 *)pos; // Save value to avoid future data_end comparisons
if (opt == 0) // Reached end of TCP options
return -1;
if (opt == 1) { // TCP NOP option - advance one byte
pos++;
continue;
}
// Option > 1, should have option size
if (pos + 2 > opt_end || pos + 2 > data_end)
return -1;
opt_size = *(__u8 *)(pos + 1); // Save value to avoid future data_end comparisons
if (opt == 8 && opt_size == 10) { // Option-kind is TCP timestap (yey!)
if (pos + opt_size > opt_end || pos + opt_size > data_end)
// Option-kind is TCP timestap (yey!)
if (opt == 8 && opt_size == 10) {
if (pos + opt_size > opt_end ||
pos + opt_size > data_end)
return -1;
*tsval = *(__u32 *)(pos + 2);
*tsecr = *(__u32 *)(pos + 6);
return 0;
}
// Some other TCP option - advance option-length bytes
pos += opt_size;
}
return -1;
*tsval = *(__u32 *)(pos + 2);
*tsecr = *(__u32 *)(pos + 6);
return 0;
}
// Some other TCP option - advance option-length bytes
pos += opt_size;
}
return -1;
}
#endif