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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
|
// SPDX-License-Identifier: GPL-3.0-or-later
#include "ebpf.h"
#include "ebpf_mdflush.h"
struct config mdflush_config = { .first_section = NULL,
.last_section = NULL,
.mutex = NETDATA_MUTEX_INITIALIZER,
.index = { .avl_tree = { .root = NULL, .compar = appconfig_section_compare },
.rwlock = AVL_LOCK_INITIALIZER } };
#define MDFLUSH_MAP_COUNT 0
static ebpf_local_maps_t mdflush_maps[] = {
{
.name = "tbl_mdflush",
.internal_input = 1024,
.user_input = 0,
.type = NETDATA_EBPF_MAP_STATIC,
.map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED
},
/* end */
{
.name = NULL,
.internal_input = 0,
.user_input = 0,
.type = NETDATA_EBPF_MAP_CONTROLLER,
.map_fd = ND_EBPF_MAP_FD_NOT_INITIALIZED
}
};
// store for "published" data from the reader thread, which the collector
// thread will write to netdata agent.
static avl_tree_lock mdflush_pub;
// tmp store for mdflush values we get from a per-CPU eBPF map.
static mdflush_ebpf_val_t *mdflush_ebpf_vals = NULL;
static struct netdata_static_thread mdflush_threads = {
.name = "MDFLUSH KERNEL",
.config_section = NULL,
.config_name = NULL,
.env_name = NULL,
.enabled = 1,
.thread = NULL,
.init_routine = NULL,
.start_routine = NULL
};
static enum ebpf_threads_status ebpf_mdflush_exited = NETDATA_THREAD_EBPF_RUNNING;
/**
* MDflush exit
*
* Cancel thread and exit.
*
* @param ptr thread data.
*/
static void mdflush_exit(void *ptr)
{
ebpf_module_t *em = (ebpf_module_t *)ptr;
if (!em->enabled) {
em->enabled = NETDATA_MAIN_THREAD_EXITED;
return;
}
ebpf_mdflush_exited = NETDATA_THREAD_EBPF_STOPPING;
}
/**
* CLeanup
*
* Clean allocated memory.
*
* @param ptr thread data.
*/
static void mdflush_cleanup(void *ptr)
{
ebpf_module_t *em = (ebpf_module_t *)ptr;
if (ebpf_mdflush_exited != NETDATA_THREAD_EBPF_STOPPED)
return;
freez(mdflush_ebpf_vals);
freez(mdflush_threads.thread);
mdflush_threads.enabled = NETDATA_MAIN_THREAD_EXITED;
em->enabled = NETDATA_MAIN_THREAD_EXITED;
}
/**
* Compare mdflush values.
*
* @param a `netdata_mdflush_t *`.
* @param b `netdata_mdflush_t *`.
*
* @return 0 if a==b, 1 if a>b, -1 if a<b.
*/
static int mdflush_val_cmp(void *a, void *b)
{
netdata_mdflush_t *ptr1 = a;
netdata_mdflush_t *ptr2 = b;
if (ptr1->unit > ptr2->unit) {
return 1;
}
else if (ptr1->unit < ptr2->unit) {
return -1;
}
else {
return 0;
}
}
static void mdflush_read_count_map()
{
int mapfd = mdflush_maps[MDFLUSH_MAP_COUNT].map_fd;
mdflush_ebpf_key_t curr_key = (uint32_t)-1;
mdflush_ebpf_key_t key = (uint32_t)-1;
netdata_mdflush_t search_v;
netdata_mdflush_t *v = NULL;
while (bpf_map_get_next_key(mapfd, &curr_key, &key) == 0) {
curr_key = key;
// get val for this key.
int test = bpf_map_lookup_elem(mapfd, &key, mdflush_ebpf_vals);
if (unlikely(test < 0)) {
continue;
}
// is this record saved yet?
//
// if not, make a new one, mark it as unsaved for now, and continue; we
// will insert it at the end after all of its values are correctly set,
// so that we can safely publish it to the collector within a single,
// short locked operation.
//
// otherwise simply continue; we will only update the flush count,
// which can be republished safely without a lock.
//
// NOTE: lock isn't strictly necessary for this initial search, as only
// this thread does writing, but the AVL is using a read-write lock so
// there is no congestion.
bool v_is_new = false;
search_v.unit = key;
v = (netdata_mdflush_t *)avl_search_lock(
&mdflush_pub,
(avl_t *)&search_v
);
if (unlikely(v == NULL)) {
// flush count can only be added reliably at a later time.
// when they're added, only then will we AVL insert.
v = callocz(1, sizeof(netdata_mdflush_t));
v->unit = key;
sprintf(v->disk_name, "md%u", key);
v->dim_exists = false;
v_is_new = true;
}
// we must add up count value for this record across all CPUs.
uint64_t total_cnt = 0;
int i;
int end = (running_on_kernel < NETDATA_KERNEL_V4_15) ? 1 : ebpf_nprocs;
for (i = 0; i < end; i++) {
total_cnt += mdflush_ebpf_vals[i];
}
// can now safely publish count for existing records.
v->cnt = total_cnt;
// can now safely publish new record.
if (v_is_new) {
avl_t *check = avl_insert_lock(&mdflush_pub, (avl_t *)v);
if (check != (avl_t *)v) {
error("Internal error, cannot insert the AVL tree.");
}
}
}
}
/**
* Read eBPF maps for mdflush.
*/
static void *mdflush_reader(void *ptr)
{
netdata_thread_cleanup_push(mdflush_cleanup, ptr);
heartbeat_t hb;
heartbeat_init(&hb);
ebpf_module_t *em = (ebpf_module_t *)ptr;
usec_t step = NETDATA_MDFLUSH_SLEEP_MS * em->update_every;
while (ebpf_mdflush_exited == NETDATA_THREAD_EBPF_RUNNING) {
usec_t dt = heartbeat_next(&hb, step);
UNUSED(dt);
if (ebpf_mdflush_exited == NETDATA_THREAD_EBPF_STOPPING)
break;
mdflush_read_count_map();
}
ebpf_mdflush_exited = NETDATA_THREAD_EBPF_STOPPED;
netdata_thread_cleanup_pop(1);
return NULL;
}
static void mdflush_create_charts(int update_every)
{
ebpf_create_chart(
"mdstat",
"mdstat_flush",
"MD flushes",
"flushes",
"flush (eBPF)",
"md.flush",
NETDATA_EBPF_CHART_TYPE_STACKED,
NETDATA_CHART_PRIO_MDSTAT_FLUSH,
NULL, NULL, 0, update_every,
NETDATA_EBPF_MODULE_NAME_MDFLUSH
);
fflush(stdout);
}
// callback for avl tree traversal on `mdflush_pub`.
static int mdflush_write_dims(void *entry, void *data)
{
UNUSED(data);
netdata_mdflush_t *v = entry;
// records get dynamically added in, so add the dim if we haven't yet.
if (!v->dim_exists) {
ebpf_write_global_dimension(
v->disk_name, v->disk_name,
ebpf_algorithms[NETDATA_EBPF_INCREMENTAL_IDX]
);
v->dim_exists = true;
}
write_chart_dimension(v->disk_name, v->cnt);
return 1;
}
/**
* Main loop for this collector.
*/
static void mdflush_collector(ebpf_module_t *em)
{
mdflush_ebpf_vals = callocz(ebpf_nprocs, sizeof(mdflush_ebpf_val_t));
avl_init_lock(&mdflush_pub, mdflush_val_cmp);
// create reader thread.
mdflush_threads.thread = mallocz(sizeof(netdata_thread_t));
mdflush_threads.start_routine = mdflush_reader;
netdata_thread_create(
mdflush_threads.thread,
mdflush_threads.name,
NETDATA_THREAD_OPTION_DEFAULT,
mdflush_reader,
em
);
// create chart and static dims.
pthread_mutex_lock(&lock);
mdflush_create_charts(em->update_every);
ebpf_update_stats(&plugin_statistics, em);
pthread_mutex_unlock(&lock);
// loop and read from published data until ebpf plugin is closed.
heartbeat_t hb;
heartbeat_init(&hb);
usec_t step = em->update_every * USEC_PER_SEC;
while (!ebpf_exit_plugin) {
(void)heartbeat_next(&hb, step);
if (ebpf_exit_plugin)
break;
// write dims now for all hitherto discovered devices.
write_begin_chart("mdstat", "mdstat_flush");
avl_traverse_lock(&mdflush_pub, mdflush_write_dims, NULL);
write_end_chart();
pthread_mutex_unlock(&lock);
}
}
/**
* mdflush thread.
*
* @param ptr a `ebpf_module_t *`.
* @return always NULL.
*/
void *ebpf_mdflush_thread(void *ptr)
{
|