gh-117657: Fix small issues with instrumentation and TSAN #118064
Conversation
DinoV
force-pushed
the
nogil_settrace_tsan
branch
2 times, most recently
from
78f1b64
to
1288090
Compare
| @@ -588,9 +589,10 @@ de_instrument(PyCodeObject *code, int i, int event) | |||
| return; | |||
| } | |||
| CHECK(_PyOpcode_Deopt[deinstrumented] == deinstrumented); | |||
| *opcode_ptr = deinstrumented; | |||
| FT_ATOMIC_STORE_UINT8_RELAXED(*opcode_ptr, deinstrumented); | |||
There was a problem hiding this comment.
Is relaxed strong enough? Do any consumers need to see the updates that are sequenced-before this store? Same comment applies to the below instances of using relaxed stores to update bytecode.
There was a problem hiding this comment.
In the deinstrumented case if we race and still see the instrumented opcode the data will still be there. In the instrumented case I think we do need some more synchronization around the tooling metadata, if we see the instrumented opcode before seeing the writes to the data that'll be bad.
I had originally had synchronizeds reads/writes around that in d8ea6bb#diff-adaefb7da847260ef7aff6e66007bc83f5b226c5389a23e1c9ea622c3b02c419 but the thought was that the synchronization around _PyFrame_GetCode(frame)->_co_instrumentation_version would be sufficient (which is in the latest version of #116775).
But that doesn't actually seem to be good enough - there's still some rare TSAN failures around the instrumentation data. But my plan is to push the synchronization there instead of on the opcode because I don't think we want to do acquire loads on the byte code which would be required to make a release write here meaningful.
DinoV
force-pushed
the
nogil_settrace_tsan
branch
from
1288090
to
d715d18
Compare
|
Apologies for my cluelessness. Does this have any effect on a non-free-threading build? Or do those macros do nothing in that case? |
|
The macros do nothing in the non-free-threaded build (i.e., they are just plain C load/stores). |
DinoV
force-pushed
the
nogil_settrace_tsan
branch
from
d715d18
to
5493306
Compare
DinoV
force-pushed
the
nogil_settrace_tsan
branch
from
5493306
to
2429aaf
Compare
…on#118064) Small TSAN fixups for instrumentation
- Fix a few places where we were not using atomics to (de)instrument opcodes. - Fix a few places where we weren't using atomics to reset adaptive counters. - Remove some redundant non-atomic resets of adaptive counters that presumably snuck as merge artifacts of python#118064 and python#117144 landing close together.
There's some races around reading and updating bytecodes with instrumentation. This switches to using atomic operations as these races are all benign.