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
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
|
package generate
import (
"fmt"
"strings"
)
func generateBPFHandler(tp GeneratedTracepoint) string {
f := tp.Format
isEnter := strings.Split(f.Name, "_")[1] == "enter"
// Use the kernel's actual tracepoint context structs (syscall_trace_enter/exit)
// rather than the BTF-emitted trace_event_raw_sys_enter/exit aliases. On RHEL 9
// kernels (5.14 with the rt-merge backport that added preempt_lazy_count to
// trace_entry) the two diverge: trace_event_raw_sys_* grows by 8 bytes and
// the args/ret offsets shift, but the real context handed to the BPF program
// is still syscall_trace_*. Reading via the wider alias trips the verifier's
// max_ctx_offset check and the attach fails with EACCES. The two structs are
// identical on non-RHEL kernels, so this is a no-op everywhere else.
ctxStruct := "syscall_trace_exit"
if isEnter {
ctxStruct = "syscall_trace_enter"
}
eventStruct := eventStructName(tp.Classification.Kind)
comment := fmt.Sprintf("%s (kind=%s)", eventStruct, tp.Classification.Kind.MetadataName())
if tp.Classification.Kind == KindRet {
comment = fmt.Sprintf("%s (%s) (kind=%s)", eventStruct, ClassifyRet(f.Name), tp.Classification.Kind.MetadataName())
}
eventTypeConst := eventTypeConstant(tp.Classification.Kind, isEnter)
extra := generateExtra(tp, isEnter)
return renderHandler(f.Name, ctxStruct, eventStruct, comment, eventTypeConst, extra, isEnter)
}
func renderHandler(name, ctxStruct, eventStruct, comment, eventTypeConst, extra string, isEnter bool) string {
var b strings.Builder
fmt.Fprintf(&b, "/// %s is a struct %s\n", name, comment)
fmt.Fprintf(&b, "SEC(\"tracepoint/syscalls/%s\")\n", name)
fmt.Fprintf(&b, "int handle_%s(struct %s *ctx) {\n", strings.ToLower(name), ctxStruct)
b.WriteString(" __u32 pid, tid;\n")
b.WriteString(" if (filter(&pid, &tid))\n")
b.WriteString(" return 0;\n")
b.WriteString("\n")
if isEnter {
fmt.Fprintf(&b, " if (!ior_on_syscall_enter(tid, %s))\n", strings.ToUpper(name))
b.WriteString(" return 0;\n")
} else {
fmt.Fprintf(&b, " if (!ior_on_syscall_exit(tid, %s, ctx->ret))\n", strings.ToUpper(name))
b.WriteString(" return 0;\n")
}
b.WriteString("\n")
fmt.Fprintf(&b, " struct %s *ev = bpf_ringbuf_reserve(&event_map, sizeof(struct %s), 0);\n", eventStruct, eventStruct)
b.WriteString(" if (!ev)\n")
b.WriteString(" return 0;\n")
b.WriteString("\n")
fmt.Fprintf(&b, " ev->event_type = %s;\n", eventTypeConst)
fmt.Fprintf(&b, " ev->trace_id = %s;\n", strings.ToUpper(name))
b.WriteString(" ev->pid = pid;\n")
b.WriteString(" ev->tid = tid;\n")
b.WriteString(" ev->time = bpf_ktime_get_boot_ns();\n")
if extra != "" {
b.WriteString(extra)
}
b.WriteString("\n")
b.WriteString(" bpf_ringbuf_submit(ev, 0);\n")
b.WriteString(" return 0;\n")
b.WriteString("}\n")
return b.String()
}
// generateExtra returns the kind-specific C body lines for a tracepoint handler,
// dispatching to a per-kind helper so that each case stays concise.
func generateExtra(tp GeneratedTracepoint, isEnter bool) string {
f := tp.Format
switch tp.Classification.Kind {
case KindFd:
return generateExtraFd(f)
case KindDup3:
return " ev->fd = (__s32)ctx->args[0];\n ev->flags = (__s32)ctx->args[2];\n"
case KindOpenByHandleAt:
return " ev->flags = (__s32)ctx->args[2];\n"
case KindSocket:
return generateExtraSocket()
case KindSocketpair:
return generateExtraSocketpair(isEnter)
case KindAccept:
return generateExtraAccept(isEnter)
case KindPipe:
return generateExtraPipe(f, isEnter)
case KindEventfd:
return generateExtraEventfd(f, isEnter)
case KindPidfd:
return generateExtraEventfd(f, isEnter)
case KindEpollCtl:
return generateExtraEpollCtl()
case KindTwoFd:
return generateExtraTwoFd(f.Name)
case KindPoll:
return generateExtraPoll(f.Name)
case KindMem:
return generateExtraMem(f.Name)
case KindSleep:
return generateExtraSleep(f.Name)
case KindKeyctl:
return generateExtraKeyctl(f.Name)
case KindPtrace:
return generateExtraPtrace()
case KindPerfOpen:
return generateExtraPerfOpen()
case KindOpen:
return generateExtraOpen(f)
case KindMqOpen:
return generateExtraMqOpen(f)
case KindExec:
return generateExtraExec(f)
case KindPathname:
return generateExtraPathname(tp, f)
case KindName:
return generateExtraName(f)
case KindFcntl:
return generateExtraFcntl(f)
case KindRet:
return fmt.Sprintf(" ev->ret = ctx->ret;\n ev->ret_type = %s;\n", ClassifyRet(f.Name))
case KindNull:
return ""
}
return ""
}
// generateExtraFd returns the fd-capture lines for fd-family events.
func generateExtraFd(f *Format) string {
if f.Name == "sys_enter_pidfd_getfd" {
return " ev->fd = (__s32)ctx->args[0];\n"
}
fdIdx := f.FieldNumber("fd")
if fdIdx >= 0 {
return fmt.Sprintf(" ev->fd = (__s32)ctx->args[%d];\n", fdIdx)
}
return " ev->fd = (__s32)ctx->args[0];\n"
}
// generateExtraOpen returns the filename/comm/flags capture lines for open-family events.
func generateExtraOpen(f *Format) string {
return generateExtraOpenWithFields(f, "filename", "flags")
}
func generateExtraMqOpen(f *Format) string {
return generateExtraOpenWithFields(f, "u_name", "oflag")
}
func generateExtraExec(f *Format) string {
filenameIdx := f.FieldNumber("filename")
dirfdIdx := f.FieldNumber("dfd")
if dirfdIdx < 0 {
dirfdIdx = f.FieldNumber("fd")
}
if dirfdIdx < 0 {
dirfdIdx = f.FieldNumber("dirfd")
}
flagsIdx := f.FieldNumber("flags")
if filenameIdx < 0 {
filenameIdx = 0
}
var b strings.Builder
b.WriteString(" __builtin_memset(&(ev->filename), 0, sizeof(ev->filename) + sizeof(ev->comm));\n")
fmt.Fprintf(&b, " bpf_probe_read_user_str(ev->filename, sizeof(ev->filename), (void *)ctx->args[%d]);\n", filenameIdx)
b.WriteString(" bpf_get_current_comm(&ev->comm, sizeof(ev->comm));\n")
if dirfdIdx > -1 {
fmt.Fprintf(&b, " ev->dirfd = (__s32)ctx->args[%d];\n", dirfdIdx)
} else if f.Name == "sys_enter_execveat" {
b.WriteString(" ev->dirfd = (__s32)ctx->args[0];\n")
} else {
b.WriteString(" ev->dirfd = -1;\n")
}
if flagsIdx > -1 {
fmt.Fprintf(&b, " ev->flags = (__s32)ctx->args[%d];\n", flagsIdx)
} else {
b.WriteString(" ev->flags = 0;\n")
}
return b.String()
}
func generateExtraOpenWithFields(f *Format, pathnameField, flagsField string) string {
filenameIdx := f.FieldNumber(pathnameField)
flagsIdx := f.FieldNumber(flagsField)
var b strings.Builder
b.WriteString(" __builtin_memset(&(ev->filename), 0, sizeof(ev->filename) + sizeof(ev->comm));\n")
fmt.Fprintf(&b, " bpf_probe_read_user_str(ev->filename, sizeof(ev->filename), (void *)ctx->args[%d]);\n", filenameIdx)
b.WriteString(" bpf_get_current_comm(&ev->comm, sizeof(ev->comm));\n")
if flagsIdx > -1 {
fmt.Fprintf(&b, " ev->flags = ctx->args[%d];\n", flagsIdx)
} else {
b.WriteString(" ev->flags = -1; // Probably OK\n")
}
return b.String()
}
// generateExtraPathname returns the pathname capture lines for path-family events.
func generateExtraPathname(tp GeneratedTracepoint, f *Format) string {
fieldName := tp.Classification.PathnameField
fieldIdx := f.FieldNumber(fieldName)
var b strings.Builder
b.WriteString(" __builtin_memset(&(ev->pathname), 0, sizeof(ev->pathname));\n")
fmt.Fprintf(&b, " bpf_probe_read_user_str(ev->pathname, sizeof(ev->pathname), (void*)ctx->args[%d]);\n", fieldIdx)
return b.String()
}
// generateExtraName returns the oldname/newname capture lines for rename/link-family events.
func generateExtraName(f *Format) string {
oldIdx := f.FieldNumber("oldname")
newIdx := f.FieldNumber("newname")
var b strings.Builder
b.WriteString(" __builtin_memset(&(ev->oldname), 0, sizeof(ev->oldname) + sizeof(ev->newname));\n")
fmt.Fprintf(&b, " bpf_probe_read_user_str(ev->oldname, sizeof(ev->oldname), (void*)ctx->args[%d]);\n", oldIdx)
fmt.Fprintf(&b, " bpf_probe_read_user_str(ev->newname, sizeof(ev->newname), (void*)ctx->args[%d]);\n", newIdx)
return b.String()
}
// generateExtraFcntl returns the fd/cmd/arg capture lines for fcntl events.
func generateExtraFcntl(f *Format) string {
fdIdx := f.FieldNumber("fd")
cmdIdx := f.FieldNumber("cmd")
argIdx := f.FieldNumber("arg")
return fmt.Sprintf(
" ev->fd = ctx->args[%d];\n ev->cmd = ctx->args[%d];\n ev->arg = ctx->args[%d];\n",
fdIdx, cmdIdx, argIdx,
)
}
func generateExtraSocket() string {
return " ev->family = (__s32)ctx->args[0];\n ev->type = (__s32)ctx->args[1];\n ev->protocol = (__s32)ctx->args[2];\n"
}
func generateExtraSocketpair(isEnter bool) string {
if isEnter {
return " struct socketpair_ctx pending;\n pending.usockvec = ctx->args[3];\n pending.family = (__s32)ctx->args[0];\n pending.type = (__s32)ctx->args[1];\n pending.protocol = (__s32)ctx->args[2];\n bpf_map_update_elem(&socketpair_ctx_map, &tid, &pending, BPF_ANY);\n ev->family = pending.family;\n ev->type = pending.type;\n ev->protocol = pending.protocol;\n ev->sv0 = -1;\n ev->sv1 = -1;\n ev->ret = 0;\n"
}
return " __s32 family = -1;\n __s32 type = -1;\n __s32 protocol = -1;\n __s32 sv0 = -1;\n __s32 sv1 = -1;\n struct socketpair_ctx *pending = bpf_map_lookup_elem(&socketpair_ctx_map, &tid);\n if (pending) {\n family = pending->family;\n type = pending->type;\n protocol = pending->protocol;\n if (ctx->ret == 0 && pending->usockvec != 0) {\n int sv[2];\n if (bpf_probe_read_user(&sv, sizeof(sv), (void *)pending->usockvec) == 0) {\n sv0 = (__s32)sv[0];\n sv1 = (__s32)sv[1];\n }\n }\n bpf_map_delete_elem(&socketpair_ctx_map, &tid);\n }\n ev->family = family;\n ev->type = type;\n ev->protocol = protocol;\n ev->sv0 = sv0;\n ev->sv1 = sv1;\n ev->ret = ctx->ret;\n"
}
func generateExtraAccept(isEnter bool) string {
if isEnter {
return " ev->fd = (__s32)ctx->args[0];\n ev->ret = -1;\n"
}
return " ev->fd = -1;\n ev->ret = ctx->ret;\n"
}
func generateExtraPipe(f *Format, isEnter bool) string {
if isEnter {
flagsExpr := "0"
if f.Name == "sys_enter_pipe2" {
flagsExpr = "(__s32)ctx->args[1]"
}
return " struct pipe_ctx pending;\n pending.upipefd = ctx->args[0];\n pending.flags = " + flagsExpr + ";\n bpf_map_update_elem(&pipe_ctx_map, &tid, &pending, BPF_ANY);\n ev->flags = pending.flags;\n ev->fd0 = -1;\n ev->fd1 = -1;\n ev->ret = 0;\n"
}
return " __s32 flags = 0;\n __s32 fd0 = -1;\n __s32 fd1 = -1;\n struct pipe_ctx *pending = bpf_map_lookup_elem(&pipe_ctx_map, &tid);\n if (pending) {\n flags = pending->flags;\n if (ctx->ret == 0 && pending->upipefd != 0) {\n int pipefd[2];\n if (bpf_probe_read_user(&pipefd, sizeof(pipefd), (void *)pending->upipefd) == 0) {\n fd0 = (__s32)pipefd[0];\n fd1 = (__s32)pipefd[1];\n }\n }\n bpf_map_delete_elem(&pipe_ctx_map, &tid);\n }\n ev->flags = flags;\n ev->fd0 = fd0;\n ev->fd1 = fd1;\n ev->ret = ctx->ret;\n"
}
func generateExtraEventfd(f *Format, isEnter bool) string {
if isEnter {
flagsExpr := "0"
switch f.Name {
case "sys_enter_eventfd2":
flagsExpr = "(__s32)ctx->args[1]"
case "sys_enter_memfd_create":
flagsExpr = "(__s32)ctx->args[1]"
case "sys_enter_memfd_secret":
flagsExpr = "(__s32)ctx->args[0]"
case "sys_enter_userfaultfd":
flagsExpr = "(__s32)ctx->args[0]"
case "sys_enter_signalfd4":
flagsExpr = "(__s32)ctx->args[3]"
case "sys_enter_timerfd_create":
flagsExpr = "(__s32)ctx->args[1]"
case "sys_enter_pidfd_open":
flagsExpr = "(__s32)ctx->args[0]"
case "sys_enter_fsmount":
flagsExpr = "(__s32)ctx->args[1]"
case "sys_enter_fsopen":
flagsExpr = "(__s32)ctx->args[1]"
}
return " __s32 flags = " + flagsExpr + ";\n bpf_map_update_elem(&eventfd_flags_map, &tid, &flags, BPF_ANY);\n ev->flags = flags;\n ev->ret = -1;\n"
}
return " __s32 flags = 0;\n __s32 *pending = bpf_map_lookup_elem(&eventfd_flags_map, &tid);\n if (pending) {\n flags = *pending;\n bpf_map_delete_elem(&eventfd_flags_map, &tid);\n }\n ev->flags = flags;\n ev->ret = ctx->ret;\n"
}
func generateExtraEpollCtl() string {
return " ev->epfd = (__s32)ctx->args[0];\n ev->op = (__s32)ctx->args[1];\n ev->fd = (__s32)ctx->args[2];\n ev->events = 0;\n if (ctx->args[3] != 0) {\n __u32 user_events = 0;\n if (bpf_probe_read_user(&user_events, sizeof(user_events), (void *)ctx->args[3]) == 0) {\n ev->events = user_events;\n }\n }\n"
}
func generateExtraTwoFd(name string) string {
switch name {
case "sys_enter_move_mount":
return " ev->fd_a = (__s32)ctx->args[0];\n ev->fd_b = (__s32)ctx->args[2];\n ev->extra = (__u64)ctx->args[4];\n"
case "sys_enter_kcmp":
return " ev->fd_a = (__s32)ctx->args[3];\n ev->fd_b = (__s32)ctx->args[4];\n ev->extra = (__u64)ctx->args[2];\n"
default:
return " ev->fd_a = (__s32)ctx->args[0];\n ev->fd_b = (__s32)ctx->args[1];\n ev->extra = (__u64)ctx->args[2];\n"
}
}
func generateExtraPoll(name string) string {
switch name {
case "sys_enter_poll":
return " ev->nfds = (__s32)ctx->args[1];\n ev->timeout_ns = -1;\n __s32 timeout_ms = (__s32)ctx->args[2];\n if (timeout_ms >= 0) {\n ev->timeout_ns = ((__s64)timeout_ms) * 1000000LL;\n }\n"
case "sys_enter_ppoll":
return " ev->nfds = (__s32)ctx->args[1];\n ev->timeout_ns = -1;\n if (ctx->args[2] != 0) {\n struct __ior_timespec {\n __s64 tv_sec;\n __s64 tv_nsec;\n } ts = {};\n if (bpf_probe_read_user(&ts, sizeof(ts), (void *)ctx->args[2]) == 0) {\n ev->timeout_ns = ts.tv_sec * 1000000000LL + ts.tv_nsec;\n }\n }\n"
case "sys_enter_select":
return " ev->nfds = (__s32)ctx->args[0];\n ev->timeout_ns = -1;\n if (ctx->args[4] != 0) {\n struct __ior_timeval {\n __s64 tv_sec;\n __s64 tv_usec;\n } tv = {};\n if (bpf_probe_read_user(&tv, sizeof(tv), (void *)ctx->args[4]) == 0) {\n ev->timeout_ns = tv.tv_sec * 1000000000LL + tv.tv_usec * 1000LL;\n }\n }\n"
case "sys_enter_pselect6":
return " ev->nfds = (__s32)ctx->args[0];\n ev->timeout_ns = -1;\n if (ctx->args[4] != 0) {\n struct __ior_timespec {\n __s64 tv_sec;\n __s64 tv_nsec;\n } ts = {};\n if (bpf_probe_read_user(&ts, sizeof(ts), (void *)ctx->args[4]) == 0) {\n ev->timeout_ns = ts.tv_sec * 1000000000LL + ts.tv_nsec;\n }\n }\n"
default:
return " ev->nfds = -1;\n ev->timeout_ns = -1;\n"
}
}
func generateExtraMem(name string) string {
switch name {
case "sys_enter_munmap":
return " ev->addr = (__u64)ctx->args[0];\n ev->length = (__u64)ctx->args[1];\n ev->length2 = 0;\n ev->flags = 0;\n"
case "sys_enter_mremap":
return " ev->addr = (__u64)ctx->args[0];\n ev->length = (__u64)ctx->args[1];\n ev->length2 = (__u64)ctx->args[2];\n ev->flags = (__u64)ctx->args[3];\n"
default:
return " ev->addr = 0;\n ev->length = 0;\n ev->length2 = 0;\n ev->flags = 0;\n"
}
}
func generateExtraSleep(name string) string {
ptrExpr := "0"
switch name {
case "sys_enter_nanosleep":
ptrExpr = "ctx->args[0]"
case "sys_enter_clock_nanosleep":
ptrExpr = "ctx->args[2]"
}
return " ev->requested_ns = -1;\n if (" + ptrExpr + " != 0) {\n struct __ior_timespec {\n __s64 tv_sec;\n __s64 tv_nsec;\n } ts = {};\n if (bpf_probe_read_user(&ts, sizeof(ts), (void *)" + ptrExpr + ") == 0) {\n ev->requested_ns = ts.tv_sec * 1000000000LL + ts.tv_nsec;\n }\n }\n"
}
func generateExtraKeyctl(name string) string {
switch name {
case "sys_enter_keyctl":
return " ev->option = (__s32)ctx->args[0];\n ev->key_serial = (__s32)ctx->args[1];\n ev->value = (__u64)ctx->args[2];\n"
case "sys_enter_add_key":
return " ev->option = -1;\n ev->key_serial = (__s32)ctx->args[4];\n ev->value = (__u64)ctx->args[3];\n"
case "sys_enter_request_key":
return " ev->option = -2;\n ev->key_serial = (__s32)ctx->args[3];\n ev->value = 0;\n"
default:
return " ev->option = 0;\n ev->key_serial = 0;\n ev->value = 0;\n"
}
}
func generateExtraPtrace() string {
return " ev->request = (__s64)ctx->args[0];\n ev->target_pid = (__s32)ctx->args[1];\n ev->data = (__u64)ctx->args[3];\n"
}
func generateExtraPerfOpen() string {
return " ev->attr_type = 0;\n ev->attr_size = 0;\n ev->config = 0;\n if (ctx->args[0] != 0) {\n struct __ior_perf_event_attr {\n __u32 type;\n __u32 size;\n __u64 config;\n } attr = {};\n if (bpf_probe_read_user(&attr, sizeof(attr), (void *)ctx->args[0]) == 0) {\n ev->attr_type = attr.type;\n ev->attr_size = attr.size;\n ev->config = attr.config;\n }\n }\n ev->target_pid = (__s32)ctx->args[1];\n ev->cpu = (__s32)ctx->args[2];\n ev->group_fd = (__s32)ctx->args[3];\n ev->flags = (__u32)ctx->args[4];\n"
}
// eventStructName returns the C struct name for a TracepointKind. The mapping
// is driven by kindRegistry so adding a new kind only requires a registry entry.
func eventStructName(kind TracepointKind) string {
return lookupKind(kind).structName
}
func eventTypeConstant(kind TracepointKind, isEnter bool) string {
prefix := "EXIT_"
if isEnter {
prefix = "ENTER_"
}
return prefix + strings.ToUpper(eventStructName(kind))
}
|
