Thanks to visit codestin.com
Credit goes to github.com

Skip to content

gh-137838: Fix JIT trace buffer overrun by pre-reserving exit stub space #138173

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
corona10 wants to merge 8 commits into from
Closed

gh-137838: Fix JIT trace buffer overrun by pre-reserving exit stub space #138173

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
0776a59
gh-137838: Fix JIT trace buffer overrun by pre-reserving exit stub space
corona10 Aug 27, 2025
eff78b4
Add NEWS.d
corona10 Aug 27, 2025
dbc5d9a
Revert "gh-137838: Fix JIT trace buffer overrun by pre-reserving exit…
corona10 Aug 27, 2025
76eec2d
update
corona10 Aug 27, 2025
2a77ad0
Update comment
corona10 Aug 27, 2025
18a8161
lint
corona10 Aug 27, 2025
1d54d87
fix
corona10 Aug 27, 2025
a4e666d
revert to 20%....
corona10 Aug 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions Lib/test/test_sys_settrace.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
import textwrap
import subprocess
import warnings

try:
import _testinternalcapi
except ImportError:
Expand Down Expand Up @@ -360,6 +361,8 @@ class TraceTestCase(unittest.TestCase):
# Disable gc collection when tracing, otherwise the
# deallocators may be traced as well.
def setUp(self):
if os.environ.get('PYTHON_UOPS_OPTIMIZE') == '0':
self.skipTest("Line tracing behavior differs when JIT optimizer is disabled")
self.using_gc = gc.isenabled()
gc.disable()
self.addCleanup(sys.settrace, sys.gettrace())
Expand Down
6 changes: 6 additions & 0 deletions Lib/test/test_trace.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -142,6 +142,8 @@ def test_traced_func_linear(self):

self.assertEqual(self.tracer.results().counts, expected)

@unittest.skipIf(os.environ.get('PYTHON_UOPS_OPTIMIZE') == '0',
"Line counts differ when JIT optimizer is disabled")
def test_traced_func_loop(self):
self.tracer.runfunc(traced_func_loop, 2, 3)

Expand All @@ -166,6 +168,8 @@ def test_traced_func_importing(self):

self.assertEqual(self.tracer.results().counts, expected)

@unittest.skipIf(os.environ.get('PYTHON_UOPS_OPTIMIZE') == '0',
"Line counts differ when JIT optimizer is disabled")
def test_trace_func_generator(self):
self.tracer.runfunc(traced_func_calling_generator)

Expand Down Expand Up @@ -236,6 +240,8 @@ def setUp(self):
self.my_py_filename = fix_ext_py(__file__)
self.addCleanup(sys.settrace, sys.gettrace())

@unittest.skipIf(os.environ.get('PYTHON_UOPS_OPTIMIZE') == '0',
"Line counts differ when JIT optimizer is disabled")
def test_exec_counts(self):
self.tracer = Trace(count=1, trace=0, countfuncs=0, countcallers=0)
code = r'''traced_func_loop(2, 5)'''
Expand Down
2 changes: 2 additions & 0 deletions Misc/NEWS.d/next/Core_and_Builtins/2025-08-27-14-58-26.gh-issue-137838.lK6T0j.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
Fix JIT trace buffer overrun by pre-reserving exit stub space. Patch By
Donghee Na.
24 changes: 10 additions & 14 deletions Python/optimizer.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -540,6 +540,7 @@ add_to_trace(
assert(func == NULL || func->func_code == (PyObject *)code); \
instr = trace_stack[trace_stack_depth].instr;


/* Returns the length of the trace on success,
* 0 if it failed to produce a worthwhile trace,
* and -1 on an error.
Expand All @@ -560,8 +561,10 @@ translate_bytecode_to_trace(
_Py_BloomFilter_Add(dependencies, initial_code);
_Py_CODEUNIT *initial_instr = instr;
int trace_length = 0;
// Leave space for possible trailing _EXIT_TRACE
int max_length = buffer_size-2;
// Leave space for possible trailing _EXIT_TRACE and estimated exit stubs
// Reserve 20% of buffer space for exit stubs (empirically sufficient)
Copy link
Copy Markdown
Member

@Fidget-Spinner Fidget-Spinner Aug 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we have a tighter and more logical bound than this?

Copy link
Copy Markdown
Member Author

@corona10 corona10 Aug 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe 15%? or 10%?

Copy link
Copy Markdown
Member Author

@corona10 corona10 Aug 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

10% is too small. Let me test from 15%.

Copy link
Copy Markdown
Member

@Fidget-Spinner Fidget-Spinner Aug 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No I mean something based on math. Like a loose bound would be:

Every instruction needs a _DEOPT or _EXIT_TRACE

_DEOPT might have _ERROR_POP_N before it

Therefore from n instructions we must reserve 2n instructions.

This bound is overly loose, but it would be good to work it out

Copy link
Copy Markdown
Member Author

@corona10 corona10 Aug 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm, let me think about this differently. I’ll share the next commit with an improvement that tracks each instruction based on a mathematical assumption

int max_exit_stubs = (buffer_size * 20) / 100; // 20% for exit stubs
int max_length = buffer_size - 2 - max_exit_stubs;
struct {
PyFunctionObject *func;
PyCodeObject *code;
Expand Down Expand Up @@ -647,16 +650,7 @@ translate_bytecode_to_trace(
assert(!OPCODE_HAS_DEOPT(opcode));
}

if (OPCODE_HAS_EXIT(opcode)) {
// Make space for side exit and final _EXIT_TRACE:
RESERVE_RAW(2, "_EXIT_TRACE");
max_length--;
}
if (OPCODE_HAS_ERROR(opcode)) {
// Make space for error stub and final _EXIT_TRACE:
RESERVE_RAW(2, "_ERROR_POP_N");
max_length--;
}
// Note: Exit stub space is pre-reserved in max_length calculation above
switch (opcode) {
case POP_JUMP_IF_NONE:
case POP_JUMP_IF_NOT_NONE:
Expand Down Expand Up @@ -731,9 +725,11 @@ translate_bytecode_to_trace(
{
const struct opcode_macro_expansion *expansion = &_PyOpcode_macro_expansion[opcode];
if (expansion->nuops > 0) {
// Reserve space for nuops (+ _SET_IP + _EXIT_TRACE)
// Reserve space for nuops
int nuops = expansion->nuops;
RESERVE(nuops + 1); /* One extra for exit */

// Reserve space for nuops (exit stub space already pre-reserved)
RESERVE(nuops);
int16_t last_op = expansion->uops[nuops-1].uop;
if (last_op == _RETURN_VALUE || last_op == _RETURN_GENERATOR || last_op == _YIELD_VALUE) {
// Check for trace stack underflow now:
Expand Down
Loading