1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
|
// SPDX-License-Identifier: GPL-2.0
/* GPLv2, Copyright(c) 2017 Jesper Dangaard Brouer, Red Hat, Inc. */
#include "xdp_sample.bpf.h"
#include <bpf/bpf_tracing.h>
#include <bpf/bpf_core_read.h>
#include <bpf/bpf_helpers.h>
array_map rx_cnt SEC(".maps");
array_map redir_err_cnt SEC(".maps");
array_map cpumap_enqueue_cnt SEC(".maps");
array_map cpumap_kthread_cnt SEC(".maps");
array_map exception_cnt SEC(".maps");
const volatile int nr_cpus = 0;
/* These can be set before loading so that redundant comparisons can be DCE'd by
* the verifier, and only actual matches are tried after loading tp_btf program.
* This allows sample to filter tracepoint stats based on net_device.
*/
const volatile int from_match[32] = {};
const volatile int to_match[32] = {};
int cpumap_map_id = 0;
/* Find if b is part of set a, but if a is empty set then evaluate to true */
#define IN_SET(a, b) \
({ \
bool __res = !(a)[0]; \
for (int i = 0; i < ARRAY_SIZE(a) && (a)[i]; i++) { \
__res = (a)[i] == (b); \
if (__res) \
break; \
} \
__res; \
})
static __always_inline __u32 xdp_get_err_key(int err)
{
switch (err) {
case 0:
return 0;
case -EINVAL:
return 2;
case -ENETDOWN:
return 3;
case -EMSGSIZE:
return 4;
case -EOPNOTSUPP:
return 5;
case -ENOSPC:
return 6;
default:
return 1;
}
}
static __always_inline int xdp_redirect_collect_stat(int from, int err)
{
u32 cpu = bpf_get_smp_processor_id();
u32 key = XDP_REDIRECT_ERROR;
struct datarec *rec;
u32 idx;
if (!IN_SET(from_match, from))
return 0;
key = xdp_get_err_key(err);
idx = key * nr_cpus + cpu;
rec = bpf_map_lookup_elem(&redir_err_cnt, &idx);
if (!rec)
return 0;
if (key)
NO_TEAR_INC(rec->dropped);
else
NO_TEAR_INC(rec->processed);
return 0; /* Indicate event was filtered (no further processing)*/
/*
* Returning 1 here would allow e.g. a perf-record tracepoint
* to see and record these events, but it doesn't work well
* in-practice as stopping perf-record also unload this
* bpf_prog. Plus, there is additional overhead of doing so.
*/
}
SEC("tp_btf/xdp_redirect_err")
int BPF_PROG(tp_xdp_redirect_err, const struct net_device *dev,
const struct bpf_prog *xdp, const void *tgt, int err,
const struct bpf_map *map, u32 index)
{
return xdp_redirect_collect_stat(dev->ifindex, err);
}
SEC("tp_btf/xdp_redirect_map_err")
int BPF_PROG(tp_xdp_redirect_map_err, const struct net_device *dev,
const struct bpf_prog *xdp, const void *tgt, int err,
const struct bpf_map *map, u32 index)
{
return xdp_redirect_collect_stat(dev->ifindex, err);
}
SEC("tp_btf/xdp_redirect")
int BPF_PROG(tp_xdp_redirect, const struct net_device *dev,
const struct bpf_prog *xdp, const void *tgt, int err,
const struct bpf_map *map, u32 index)
{
return xdp_redirect_collect_stat(dev->ifindex, err);
}
SEC("tp_btf/xdp_redirect_map")
int BPF_PROG(tp_xdp_redirect_map, const struct net_device *dev,
const struct bpf_prog *xdp, const void *tgt, int err,
const struct bpf_map *map, u32 index)
{
return xdp_redirect_collect_stat(dev->ifindex, err);
}
SEC("tp_btf/xdp_cpumap_enqueue")
int BPF_PROG(tp_xdp_cpumap_enqueue, int map_id, unsigned int processed,
unsigned int drops, int to_cpu)
{
u32 cpu = bpf_get_smp_processor_id();
struct datarec *rec;
u32 idx;
if (cpumap_map_id && cpumap_map_id != map_id)
return 0;
idx = to_cpu * nr_cpus + cpu;
rec = bpf_map_lookup_elem(&cpumap_enqueue_cnt, &idx);
if (!rec)
return 0;
NO_TEAR_ADD(rec->processed, processed);
NO_TEAR_ADD(rec->dropped, drops);
/* Record bulk events, then userspace can calc average bulk size */
if (processed > 0)
NO_TEAR_INC(rec->issue);
/* Inception: It's possible to detect overload situations, via
* this tracepoint. This can be used for creating a feedback
* loop to XDP, which can take appropriate actions to mitigate
* this overload situation.
*/
return 0;
}
SEC("tp_btf/xdp_cpumap_kthread")
int BPF_PROG(tp_xdp_cpumap_kthread, int map_id, unsigned int processed,
unsigned int drops, int sched, struct xdp_cpumap_stats *xdp_stats)
{
struct datarec *rec;
u32 cpu;
if (cpumap_map_id && cpumap_map_id != map_id)
return 0;
cpu = bpf_get_smp_processor_id();
rec = bpf_map_lookup_elem(&cpumap_kthread_cnt, &cpu);
if (!rec)
return 0;
NO_TEAR_ADD(rec->processed, processed);
NO_TEAR_ADD(rec->dropped, drops);
NO_TEAR_ADD(rec->xdp_pass, xdp_stats->pass);
NO_TEAR_ADD(rec->xdp_drop, xdp_stats->drop);
NO_TEAR_ADD(rec->xdp_redirect, xdp_stats->redirect);
/* Count times kthread yielded CPU via schedule call */
if (sched)
NO_TEAR_INC(rec->issue);
return 0;
}
SEC("tp_btf/xdp_exception")
int BPF_PROG(tp_xdp_exception, const struct net_device *dev,
const struct bpf_prog *xdp, u32 act)
{
u32 cpu = bpf_get_smp_processor_id();
struct datarec *rec;
u32 key = act, idx;
if (!IN_SET(from_match, dev->ifindex))
return 0;
if (!IN_SET(to_match, dev->ifindex))
return 0;
if (key > XDP_REDIRECT)
key = XDP_REDIRECT + 1;
idx = key * nr_cpus + cpu;
rec = bpf_map_lookup_elem(&exception_cnt, &idx);
if (!rec)
return 0;
NO_TEAR_INC(rec->dropped);
return 0;
}
|
