This approach eliminates the originally reported race. It also gets rid of the deadlock reported in gh-96071, so we can remove the workaround added then.
* Mark almost all reachable objects before doing collection phase
* Add stats for objects marked
* Visit new frames before each increment
* Remove lazy dict tracking
* Update docs
* Clearer calculation of work to do.
The primary objective here is to allow some later changes to be cleaner. Mostly this involves renaming things and moving a few things around.
* CrossInterpreterData -> XIData
* crossinterpdatafunc -> xidatafunc
* split out pycore_crossinterp_data_registry.h
* add _PyXIData_lookup_t
This is essentially a cleanup, moving a handful of API declarations to the header files where they fit best, creating new ones when needed.
We do the following:
* add pycore_debug_offsets.h and move _Py_DebugOffsets, etc. there
* inline struct _getargs_runtime_state and struct _gilstate_runtime_state in _PyRuntimeState
* move struct _reftracer_runtime_state to the existing pycore_object_state.h
* add pycore_audit.h and move to it _Py_AuditHookEntry , _PySys_Audit(), and _PySys_ClearAuditHooks
* add audit.h and cpython/audit.h and move the existing audit-related API there
*move the perfmap/trampoline API from cpython/sysmodule.h to cpython/ceval.h, and remove the now-empty cpython/sysmodule.h
Stop the world when invalidating function versions
The tier1 interpreter specializes `CALL` instructions based on the values
of certain function attributes (e.g. `__code__`, `__defaults__`). The tier1
interpreter uses function versions to verify that the attributes of a function
during execution of a specialization match those seen during specialization.
A function's version is initialized in `MAKE_FUNCTION` and is invalidated when
any of the critical function attributes are changed. The tier1 interpreter stores
the function version in the inline cache during specialization. A guard is used by
the specialized instruction to verify that the version of the function on the operand
stack matches the cached version (and therefore has all of the expected attributes).
It is assumed that once the guard passes, all attributes will remain unchanged
while executing the rest of the specialized instruction.
Stopping the world when invalidating function versions ensures that all critical
function attributes will remain unchanged after the function version guard passes
in free-threaded builds. It's important to note that this is only true if the remainder
of the specialized instruction does not enter and exit a stop-the-world point.
We will stop the world the first time any of the following function attributes
are mutated:
- defaults
- vectorcall
- kwdefaults
- closure
- code
This should happen rarely and only happens once per function, so the performance
impact on majority of code should be minimal.
Additionally, refactor the API for manipulating function versions to more clearly
match the stated semantics.
Add more offsets to _Py_DebugOffsets
We add a few more offsets that are required by some out-of-process
tools, such as [Austin](https://github.com/p403n1x87/austin).
This adds `_PyMem_FreeDelayed()` and supporting functions. The
`_PyMem_FreeDelayed()` function frees memory with the same allocator as
`PyMem_Free()`, but after some delay to ensure that concurrent lock-free
readers have finished.
This adds a safe memory reclamation scheme based on FreeBSD's "GUS" and
quiescent state based reclamation (QSBR). The API provides a mechanism
for callers to detect when it is safe to free memory that may be
concurrently accessed by readers.
This marks dead ThreadHandles as non-joinable earlier in
`PyOS_AfterFork_Child()` before we execute any Python code. The handles
are stored in a global linked list in `_PyRuntimeState` because `fork()`
affects the entire process.
For interpreters that share state with the main interpreter, this points
to the same static memory structure. For interpreters with their own
obmalloc state, it is heap allocated. Add free_obmalloc_arenas() which
will free the obmalloc arenas and radix tree structures for interpreters
with their own obmalloc state.
Co-authored-by: Eric Snow <ericsnowcurrently@gmail.com>
The `--disable-gil` builds occasionally need to pause all but one thread. Some
examples include:
* Cyclic garbage collection, where this is often called a "stop the world event"
* Before calling `fork()`, to ensure a consistent state for internal data structures
* During interpreter shutdown, to ensure that daemon threads aren't accessing Python objects
This adds the following functions to implement global and per-interpreter pauses:
* `_PyEval_StopTheWorldAll()` and `_PyEval_StartTheWorldAll()` (for the global runtime)
* `_PyEval_StopTheWorld()` and `_PyEval_StartTheWorld()` (per-interpreter)
(The function names may change.)
These functions are no-ops outside of the `--disable-gil` build.
* gh-112529: Track if debug allocator is used as underlying allocator
The GC implementation for free-threaded builds will need to accurately
detect if the debug allocator is used because it affects the offset of
the Python object from the beginning of the memory allocation. The
current implementation of `_PyMem_DebugEnabled` only considers if the
debug allocator is the outer-most allocator; it doesn't handle the case
of "hooks" like tracemalloc being used on top of the debug allocator.
This change enables more accurate detection of the debug allocator by
tracking when debug hooks are enabled.
* Simplify _PyMem_DebugEnabled
Every PyThreadState instance is now actually a _PyThreadStateImpl.
It is safe to cast from `PyThreadState*` to `_PyThreadStateImpl*` and back.
The _PyThreadStateImpl will contain fields that we do not want to expose
in the public C API.
This moves several general internal APIs out of _xxsubinterpretersmodule.c and into the new Python/crossinterp.c (and the corresponding internal headers).
Specifically:
* _Py_excinfo, etc.: the initial implementation for non-object exception snapshots (in pycore_pyerrors.h and Python/errors.c)
* _PyXI_exception_info, etc.: helpers for passing an exception beween interpreters (wraps _Py_excinfo)
* _PyXI_namespace, etc.: helpers for copying a dict of attrs between interpreters
* _PyXI_Enter(), _PyXI_Exit(): functions that abstract out the transitions between one interpreter and a second that will do some work temporarily
Again, these were all abstracted out of _xxsubinterpretersmodule.c as generalizations. I plan on proposing these as public API at some point.
Fixes#109894
* set `interp.static_objects.last_resort_memory_error.args` to empty tuple to avoid crash on `PyErr_Display()` call
* allow `_PyExc_InitGlobalObjects()` to be called on subinterpreter init
---------
Co-authored-by: blurb-it[bot] <43283697+blurb-it[bot]@users.noreply.github.com>
In a few places we switch to another interpreter without knowing if it has a thread state associated with the current thread. For the main interpreter there wasn't much of a problem, but for subinterpreters we were *mostly* okay re-using the tstate created with the interpreter (located via PyInterpreterState_ThreadHead()). There was a good chance that tstate wasn't actually in use by another thread.
However, there are no guarantees of that. Furthermore, re-using an already used tstate is currently fragile. To address this, now we create a new thread state in each of those places and use it.
One consequence of this change is that PyInterpreterState_ThreadHead() may not return NULL (though that won't happen for the main interpreter).
Python built with "configure --with-trace-refs" (tracing references)
is now ABI compatible with Python release build and debug build.
Moreover, it now also supports the Limited API.
Change Py_TRACE_REFS build:
* Remove _PyObject_EXTRA_INIT macro.
* The PyObject structure no longer has two extra members (_ob_prev
and _ob_next).
* Use a hash table (_Py_hashtable_t) to trace references (all
objects): PyInterpreterState.object_state.refchain.
* Py_TRACE_REFS build is now ABI compatible with release build and
debug build.
* Limited C API extensions can now be built with Py_TRACE_REFS:
xxlimited, xxlimited_35, _testclinic_limited.
* No longer rename PyModule_Create2() and PyModule_FromDefAndSpec2()
functions to PyModule_Create2TraceRefs() and
PyModule_FromDefAndSpec2TraceRefs().
* _Py_PrintReferenceAddresses() is now called before
finalize_interp_delete() which deletes the refchain hash table.
* test_tracemalloc find_trace() now also filters by size to ignore
the memory allocated by _PyRefchain_Trace().
Test changes for Py_TRACE_REFS:
* Add test.support.Py_TRACE_REFS constant.
* Add test_sys.test_getobjects() to test sys.getobjects() function.
* test_exceptions skips test_recursion_normalizing_with_no_memory()
and test_memory_error_in_PyErr_PrintEx() if Python is built with
Py_TRACE_REFS.
* test_repl skips test_no_memory().
* test_capi skisp test_set_nomemory().
* Add missing includes.
* Remove unused includes.
* Update old include/symbol names to newer names.
* Mention at least one included symbol.
* Sort includes.
* Update Tools/cases_generator/generate_cases.py used to generated
pycore_opcode_metadata.h.
* Update Parser/asdl_c.py used to generate pycore_ast.h.
* Cleanup also includes in _testcapimodule.c and _testinternalcapi.c.
The linked list of objects was a global variable, which broke isolation between interpreters, causing crashes. To solve this, we've moved the linked list to each interpreter.
This fixes a crasher due to a race condition, triggered infrequently when two isolated (own GIL) subinterpreters simultaneously initialize their sys or builtins modules. The crash happened due the combination of the "detached" thread state we were using and the "last holder" logic we use for the GIL. It turns out it's tricky to use the same thread state for different threads. Who could have guessed?
We solve the problem by eliminating the one object we were still sharing between interpreters. We replace it with a low-level hashtable, using the "raw" allocator to avoid tying it to the main interpreter.
We also remove the accommodations for "detached" thread states, which were a dubious idea to start with.
We tried this before with a dict and for all interned strings. That ran into problems due to interpreter isolation. However, exclusively using a per-interpreter cache caused some inconsistency that can eliminate the benefit of interning. Here we circle back to using a global cache, but only for statically allocated strings. We also use a more-basic _Py_hashtable_t for that global cache instead of a dict.
Ideally we would only have the global cache, but the optional isolation of each interpreter's allocator means that a non-static string object must not outlive its interpreter. Thus we would have to store a copy of each such interned string in the global cache, tied to the main interpreter.
The risk of a race with this state is relatively low, but we play it safe anyway. We do avoid using the lock in performance-sensitive cases where the risk of a race is very, very low.
This is strictly about moving the "obmalloc" runtime state from
`_PyRuntimeState` to `PyInterpreterState`. Doing so improves isolation
between interpreters, specifically most of the memory (incl. objects)
allocated for each interpreter's use. This is important for a
per-interpreter GIL, but such isolation is valuable even without it.
FWIW, a per-interpreter obmalloc is the proverbial
canary-in-the-coalmine when it comes to the isolation of objects between
interpreters. Any object that leaks (unintentionally) to another
interpreter is highly likely to cause a crash (on debug builds at
least). That's a useful thing to know, relative to interpreter
isolation.
Core static types will continue to use the global value. All other types
will use the per-interpreter value. They all share the same range, where
the global types use values < 2^16 and each interpreter uses values
higher than that.
This is the implementation of PEP683
Motivation:
The PR introduces the ability to immortalize instances in CPython which bypasses reference counting. Tagging objects as immortal allows up to skip certain operations when we know that the object will be around for the entire execution of the runtime.
Note that this by itself will bring a performance regression to the runtime due to the extra reference count checks. However, this brings the ability of having truly immutable objects that are useful in other contexts such as immutable data sharing between sub-interpreters.
Sharing mutable (or non-immortal) objects between interpreters is generally not safe. We can work around that but not easily.
There are two restrictions that are critical for objects that break interpreter isolation.
The first is that the object's state be guarded by a global lock. For now the GIL meets this requirement, but a granular global lock is needed once we have a per-interpreter GIL.
The second restriction is that the object (and, for a container, its items) be deallocated/resized only when the interpreter in which it was allocated is the current one. This is because every interpreter has (or will have, see gh-101660) its own object allocator. Deallocating an object with a different allocator can cause crashes.
The dict for the cache of module defs is completely internal, which simplifies what we have to do to meet those requirements. To do so, we do the following:
* add a mechanism for re-using a temporary thread state tied to the main interpreter in an arbitrary thread
* add _PyRuntime.imports.extensions.main_tstate`
* add _PyThreadState_InitDetached() and _PyThreadState_ClearDetached() (pystate.c)
* add _PyThreadState_BindDetached() and _PyThreadState_UnbindDetached() (pystate.c)
* make sure the cache dict (_PyRuntime.imports.extensions.dict) and its items are all owned by the main interpreter)
* add a placeholder using for a granular global lock
Note that the cache is only used for legacy extension modules and not for multi-phase init modules.
https://github.com/python/cpython/issues/100227
This reverts commit 87be8d9.
This approach to keeping the interned strings safe is turning out to be too complex for my taste (due to obmalloc isolation). For now I'm going with the simpler solution, making the dict per-interpreter. We can revisit that later if we want a sharing solution.
This is effectively two changes. The first (the bulk of the change) is where we add _Py_AddToGlobalDict() (and _PyRuntime.cached_objects.main_tstate, etc.). The second (much smaller) change is where we update PyUnicode_InternInPlace() to use _Py_AddToGlobalDict() instead of calling PyDict_SetDefault() directly.
Basically, _Py_AddToGlobalDict() is a wrapper around PyDict_SetDefault() that should be used whenever we need to add a value to a runtime-global dict object (in the few cases where we are leaving the container global rather than moving it to PyInterpreterState, e.g. the interned strings dict). _Py_AddToGlobalDict() does all the necessary work to make sure the target global dict is shared safely between isolated interpreters. This is especially important as we move the obmalloc state to each interpreter (gh-101660), as well as, potentially, the GIL (PEP 684).
https://github.com/python/cpython/issues/100227
* Eliminate all remaining uses of Py_SIZE and Py_SET_SIZE on PyLongObject, adding asserts.
* Change layout of size/sign bits in longobject to support future addition of immortal ints and tagged medium ints.
* Add functions to hide some internals of long object, and for setting sign and digit count.
* Replace uses of IS_MEDIUM_VALUE macro with _PyLong_IsCompact().
Specific changes:
* move the import lock to PyInterpreterState
* move the "find_and_load" diagnostic state to PyInterpreterState
Note that the import lock exists to keep multiple imports of the same module in the same interpreter (but in different threads) from stomping on each other. Independently, we use a distinct global lock to protect globally shared import state, especially related to loaded extension modules. For now we can rely on the GIL as that lock but with a per-interpreter GIL we'll need a new global lock.
The remaining state in _PyRuntimeState.imports will (probably) continue being global.
https://github.com/python/cpython/issues/100227
Eric Snow
Hugo van Kemenade
Mark Shannon
mpage
Matt Wozniski
Pablo Galindo Salgado
Gabriele N. Tornetta
Sam Gross
Neil Schemenauer
Irit Katriel
Radislav Chugunov
Victor Stinner
Eddie Elizondo