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
|
import os
import signal
import socket
import subprocess
import select
import ptrace.debugger
from ptrace.debugger import (
ProcessSignal,
ProcessExit,
NewProcessEvent,
)
# ----------------------------------------------------------------------
# Configuration
# ----------------------------------------------------------------------
# If scheduler does not provide input within this time (seconds),
# continue running the last chosen thread.
SCHED_TIMEOUT = 0
# ----------------------------------------------------------------------
# Scheduler (non-blocking)
# ----------------------------------------------------------------------
def schedule_next_nonblocking(sock, processes, current_proc):
"""
processes: dict[tid] -> PtraceProcess
current_proc: PtraceProcess or None
"""
timeout = SCHED_TIMEOUT if SCHED_TIMEOUT > 0 else 0
r, _, _ = select.select([sock], [], [], timeout)
if not r:
return current_proc # no input → continue with current
data = sock.recv(64)
if not data:
return current_proc
try:
tid = int(data.strip())
except ValueError:
print(f"Scheduler: invalid data {data!r}")
return current_proc
proc = processes.get(tid)
if proc is not None:
print(f"Scheduler picked TID {tid}")
return proc
print(f"Scheduler sent inactive TID {tid}, ignoring")
return current_proc
# ----------------------------------------------------------------------
# Main tracing logic
# ----------------------------------------------------------------------
def trace(pid, sched_socket_path):
debugger = ptrace.debugger.PtraceDebugger()
debugger.traceClone()
debugger.traceFork()
debugger.traceExec()
print(f"Attach process {pid}")
proc0 = debugger.addProcess(pid, False)
# ------------------------------------------------------------------
# Create scheduler socket
# ------------------------------------------------------------------
if os.path.exists(sched_socket_path):
os.unlink(sched_socket_path)
srv = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
srv.bind(sched_socket_path)
srv.listen(1)
print(f"Waiting for scheduler connection on {sched_socket_path}")
conn, _ = srv.accept()
print("Scheduler connected")
# ------------------------------------------------------------------
# Prime the very first thread
# ------------------------------------------------------------------
current_proc = proc0
# ignore-first-clone state
first_clone_ignored = False
ignored_tid = None
# Arm first process: run until its first event/syscall
current_proc.syscall()
# ------------------------------------------------------------------
# Global event loop
# ------------------------------------------------------------------
while debugger.list:
try:
event = debugger.waitSyscall()
# --------------------------------------------------------------
# New process / thread via clone/fork/vfork
# --------------------------------------------------------------
except NewProcessEvent as ev:
child = ev.process
parent = child.parent
child_tid = child.pid
if not first_clone_ignored:
# FIRST clone is ignored
first_clone_ignored = True
ignored_tid = child_tid
print(f"First clone: created TID {child_tid} — IGNORING it")
# Detach ignored child so it runs untraced
try:
child.detach()
except Exception:
pass
# Remove from debugger
debugger.deleteProcess(child)
# Resume parent so clone() completes
try:
parent.syscall()
except Exception as e:
print(f"Error resuming parent {parent.pid} after ignored clone: {e}")
else:
# LATER clones are traced
print(f"New traced thread {child_tid} (parent {parent.pid})")
# Both child and parent should be armed again
try:
child.syscall()
except Exception as e:
print(f"Error arming child {child_tid}: {e}")
try:
debugger.deleteProcess(child)
except Exception:
pass
try:
parent.syscall()
except Exception as e:
print(f"Error arming parent {parent.pid}: {e}")
try:
debugger.deleteProcess(parent)
except Exception:
pass
continue
# --------------------------------------------------------------
# Signal delivered to a traced task
# --------------------------------------------------------------
except ProcessSignal as ev:
proc = ev.process
try:
proc.syscall(ev.signum)
except Exception as e:
print(f"Error arming TID {proc.pid} after signal {ev.signum}: {e}")
try:
debugger.deleteProcess(proc)
except Exception:
pass
continue
# --------------------------------------------------------------
# A traced task exited
# --------------------------------------------------------------
except ProcessExit as ev:
dead_proc = ev.process
tid = dead_proc.pid
print(f"TID {tid} exited (exitcode={ev.exitcode})")
try:
dead_proc.detach()
except Exception:
pass
try:
debugger.deleteProcess(dead_proc)
except Exception:
pass
if not debugger.list:
break
# Choose a new current_proc and arm it
current_proc = debugger.list[0]
try:
current_proc.syscall()
except Exception as e:
print(f"Error arming new current TID {current_proc.pid}: {e}")
try:
debugger.deleteProcess(current_proc)
except Exception:
pass
continue
# --------------------------------------------------------------
# NORMAL SYSCALL STOP
# --------------------------------------------------------------
proc = event.process
tid = proc.pid
if tid == ignored_tid:
# Should not happen; just log and continue
print(f"WARNING: ignored TID {tid} hit a syscall-stop")
else:
try:
ip = proc.getInstrPointer()
print(f"TID {tid} syscall-stop at {hex(ip)}")
except Exception as e:
print(f"Error reading IP for TID {tid}: {e}")
# Build a fresh pid->process map from the debugger
processes = {p.pid: p for p in debugger.list}
# Scheduler decides what to run next
current_proc = schedule_next_nonblocking(conn, processes, proc)
if current_proc is None or current_proc not in debugger.list:
# Fallback: pick any alive one
if not debugger.list:
break
current_proc = debugger.list[0]
# Resume chosen thread
try:
current_proc.syscall()
except Exception as e:
print(f"Error arming TID {current_proc.pid}: {e}")
try:
debugger.deleteProcess(current_proc)
except Exception:
pass
# We don't immediately re-arm here; next loop iteration will
# pick another process (if any) and arm it.
conn.close()
srv.close()
debugger.quit()
# ----------------------------------------------------------------------
# Entry point
# ----------------------------------------------------------------------
def quoted(s: str) -> str:
return f'"{s}"'
if __name__ == "__main__":
env = os.environ.copy()
qemu = [
"qemu-x86_64",
"-g",
"12348",
"/nix/store/dmpq06y392i752zwhcna07kb2x5l58l5-memcached-static-x86_64-unknown-linux-musl-1.6.37/bin/memcached",
"-p",
"11211",
"-t",
"4",
"-vv"
]
sched_path = "/tmp/memcached_scheduler.sock"
proc = subprocess.Popen(qemu, env=env)
try:
trace(proc.pid, sched_path)
except Exception as e:
print(f"Got exception: {e}")
proc.kill()
exit(2)
exit(0)
|