fix(cli): report stage-3 placement misses instead of retrying silently
ci / macos-latest / py3.12 (push) Canceled after 0s
ci / macos-latest / py3.9 (push) Canceled after 0s
ci / ubuntu-latest / py3.12 (push) Canceled after 0s
ci / ubuntu-latest / py3.9 (push) Canceled after 0s
ci / windows-latest / py3.12 (push) Canceled after 0s
ci / windows-latest / py3.9 (push) Canceled after 0s
ci / macos-latest / py3.12 (push) Canceled after 0s
ci / macos-latest / py3.9 (push) Canceled after 0s
ci / ubuntu-latest / py3.12 (push) Canceled after 0s
ci / ubuntu-latest / py3.9 (push) Canceled after 0s
ci / windows-latest / py3.12 (push) Canceled after 0s
ci / windows-latest / py3.9 (push) Canceled after 0s
This commit is contained in:
@@ -234,6 +234,14 @@ retries**, with a counter file in `/out` incrementing 1→5 — so the worker su
|
||||
cycles; what does not exist is a shared session. Anything that must survive between commands has
|
||||
to be written to disk (`/out`) rather than held in memory.
|
||||
|
||||
Stage 3 can miss intermittently: A2's cheap OOM leaves the stale pointer at a slightly different
|
||||
spot, B2 then completes normally (`200`, worker alive) instead of firing the destructor, and no
|
||||
output appears. `poc exploit` prints
|
||||
`no command output; B2 status=... - stage-3 placement missed` and retries the whole chain; keep
|
||||
`--retries` at 3 or higher. Three consecutive cold starts (fresh container each) all succeeded on
|
||||
attempt 1, so the miss is occasional rather than systematic.
|
||||
|
||||
|
||||
For commands longer than the 200-byte carrier limit, write a script into `out/` and run
|
||||
`python -m poc exec --cmd 'sh</out/c'`.
|
||||
|
||||
|
||||
@@ -210,6 +210,8 @@ def exploit_once(args, prof):
|
||||
print(proof)
|
||||
return 0, proof
|
||||
time.sleep(0.2)
|
||||
print(" no command output; B2 status=%r - stage-3 placement missed"
|
||||
% wire.status(r4))
|
||||
return 1, ""
|
||||
finally:
|
||||
wire.release(held)
|
||||
|
||||
Reference in New Issue
Block a user