|
/*
* Copyright (c) 1998, 2022, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
#include "precompiled.hpp"
#include "logging/log.hpp"
#include "memory/resourceArea.hpp"
#include "runtime/interfaceSupport.inline.hpp"
#include "runtime/javaThread.inline.hpp"
#include "runtime/mutex.hpp"
#include "runtime/os.inline.hpp"
#include "runtime/osThread.hpp"
#include "runtime/safepointMechanism.inline.hpp"
#include "runtime/threadCrashProtection.hpp"
#include "utilities/events.hpp"
#include "utilities/macros.hpp"
class InFlightMutexRelease {
private:
Mutex* _in_flight_mutex;
public:
InFlightMutexRelease(Mutex* in_flight_mutex) : _in_flight_mutex(in_flight_mutex) {
assert(in_flight_mutex != NULL, "must be");
}
void operator()(JavaThread* current) {
_in_flight_mutex->release_for_safepoint();
_in_flight_mutex = NULL;
}
bool not_released() { return _in_flight_mutex != NULL; }
};
#ifdef ASSERT
void Mutex::check_block_state(Thread* thread) {
if (!_allow_vm_block && thread->is_VM_thread()) {
// JavaThreads are checked to make sure that they do not hold _allow_vm_block locks during operations
// that could safepoint. Make sure the vm thread never uses locks with _allow_vm_block == false.
fatal("VM thread could block on lock that may be held by a JavaThread during safepoint: %s", name());
}
assert(!ThreadCrashProtection::is_crash_protected(thread),
"locking not allowed when crash protection is set");
}
void Mutex::check_safepoint_state(Thread* thread) {
check_block_state(thread);
// If the lock acquisition checks for safepoint, verify that the lock was created with rank that
// has safepoint checks. Technically this doesn't affect NonJavaThreads since they won't actually
// check for safepoint, but let's make the rule unconditional unless there's a good reason not to.
assert(_rank > nosafepoint,
"This lock should not be taken with a safepoint check: %s", name());
if (thread->is_active_Java_thread()) {
// Also check NoSafepointVerifier, and thread state is _thread_in_vm
JavaThread::cast(thread)->check_for_valid_safepoint_state();
}
}
void Mutex::check_no_safepoint_state(Thread* thread) {
check_block_state(thread);
assert(!thread->is_active_Java_thread() || _rank <= nosafepoint,
"This lock should always have a safepoint check for Java threads: %s",
name());
}
#endif // ASSERT
void Mutex::lock_contended(Thread* self) {
DEBUG_ONLY(int retry_cnt = 0;)
bool is_active_Java_thread = self->is_active_Java_thread();
do {
#ifdef ASSERT
if (retry_cnt++ > 3) {
log_trace(vmmutex)("JavaThread " INTPTR_FORMAT " on %d attempt trying to acquire vmmutex %s", p2i(self), retry_cnt, _name);
}
#endif // ASSERT
// Is it a JavaThread participating in the safepoint protocol.
if (is_active_Java_thread) {
InFlightMutexRelease ifmr(this);
assert(rank() > Mutex::nosafepoint, "Potential deadlock with nosafepoint or lesser rank mutex");
{
ThreadBlockInVMPreprocess<InFlightMutexRelease> tbivmdc(JavaThread::cast(self), ifmr);
_lock.lock();
}
if (ifmr.not_released()) {
// Not unlocked by ~ThreadBlockInVMPreprocess
break;
}
} else {
_lock.lock();
break;
}
} while (!_lock.try_lock());
}
void Mutex::lock(Thread* self) {
assert(owner() != self, "invariant");
check_safepoint_state(self);
check_rank(self);
if (!_lock.try_lock()) {
// The lock is contended, use contended slow-path function to lock
lock_contended(self);
}
assert_owner(NULL);
set_owner(self);
}
void Mutex::lock() {
lock(Thread::current());
}
// Lock without safepoint check - a degenerate variant of lock() for use by
// JavaThreads when it is known to be safe to not check for a safepoint when
// acquiring this lock. If the thread blocks acquiring the lock it is not
// safepoint-safe and so will prevent a safepoint from being reached. If used
// in the wrong way this can lead to a deadlock with the safepoint code.
void Mutex::lock_without_safepoint_check(Thread * self) {
assert(owner() != self, "invariant");
check_no_safepoint_state(self);
check_rank(self);
_lock.lock();
assert_owner(NULL);
set_owner(self);
}
void Mutex::lock_without_safepoint_check() {
lock_without_safepoint_check(Thread::current());
}
// Returns true if thread succeeds in grabbing the lock, otherwise false.
bool Mutex::try_lock_inner(bool do_rank_checks) {
Thread * const self = Thread::current();
// Checking the owner hides the potential difference in recursive locking behaviour
// on some platforms.
if (owner() == self) {
return false;
}
if (do_rank_checks) {
check_rank(self);
}
// Some safepoint checking locks use try_lock, so cannot check
// safepoint state, but can check blocking state.
check_block_state(self);
if (_lock.try_lock()) {
assert_owner(NULL);
set_owner(self);
return true;
}
return false;
}
bool Mutex::try_lock() {
return try_lock_inner(true /* do_rank_checks */);
}
bool Mutex::try_lock_without_rank_check() {
bool res = try_lock_inner(false /* do_rank_checks */);
DEBUG_ONLY(if (res) _skip_rank_check = true;)
return res;
}
void Mutex::release_for_safepoint() {
assert_owner(NULL);
_lock.unlock();
}
void Mutex::unlock() {
DEBUG_ONLY(assert_owner(Thread::current()));
set_owner(NULL);
_lock.unlock();
}
void Monitor::notify() {
DEBUG_ONLY(assert_owner(Thread::current()));
_lock.notify();
}
void Monitor::notify_all() {
DEBUG_ONLY(assert_owner(Thread::current()));
_lock.notify_all();
}
bool Monitor::wait_without_safepoint_check(int64_t timeout) {
Thread* const self = Thread::current();
// timeout is in milliseconds - with zero meaning never timeout
assert(timeout >= 0, "negative timeout");
assert_owner(self);
check_rank(self);
// conceptually set the owner to NULL in anticipation of
// abdicating the lock in wait
set_owner(NULL);
// Check safepoint state after resetting owner and possible NSV.
check_no_safepoint_state(self);
int wait_status = _lock.wait(timeout);
set_owner(self);
return wait_status != 0; // return true IFF timeout
}
bool Monitor::wait(int64_t timeout) {
JavaThread* const self = JavaThread::current();
// Safepoint checking logically implies an active JavaThread.
assert(self->is_active_Java_thread(), "invariant");
// timeout is in milliseconds - with zero meaning never timeout
assert(timeout >= 0, "negative timeout");
assert_owner(self);
check_rank(self);
// conceptually set the owner to NULL in anticipation of
// abdicating the lock in wait
set_owner(NULL);
// Check safepoint state after resetting owner and possible NSV.
check_safepoint_state(self);
int wait_status;
InFlightMutexRelease ifmr(this);
{
ThreadBlockInVMPreprocess<InFlightMutexRelease> tbivmdc(self, ifmr);
OSThreadWaitState osts(self->osthread(), false /* not Object.wait() */);
wait_status = _lock.wait(timeout);
}
if (ifmr.not_released()) {
// Not unlocked by ~ThreadBlockInVMPreprocess
assert_owner(NULL);
// Conceptually reestablish ownership of the lock.
set_owner(self);
} else {
lock(self);
}
return wait_status != 0; // return true IFF timeout
}
Mutex::~Mutex() {
assert_owner(NULL);
os::free(const_cast<char*>(_name));
}
Mutex::Mutex(Rank rank, const char * name, bool allow_vm_block) : _owner(NULL) {
assert(os::mutex_init_done(), "Too early!");
assert(name != NULL, "Mutex requires a name");
_name = os::strdup(name, mtInternal);
#ifdef ASSERT
_allow_vm_block = allow_vm_block;
_rank = rank;
_skip_rank_check = false;
assert(_rank >= static_cast<Rank>(0) && _rank <= safepoint, "Bad lock rank %s: %s", rank_name(), name);
// The allow_vm_block also includes allowing other non-Java threads to block or
// allowing Java threads to block in native.
assert(_rank > nosafepoint || _allow_vm_block,
"Locks that don't check for safepoint should always allow the vm to block: %s", name);
#endif
}
bool Mutex::owned_by_self() const {
return owner() == Thread::current();
}
void Mutex::print_on_error(outputStream* st) const {
st->print("[" PTR_FORMAT, p2i(this));
st->print("] %s", _name);
st->print(" - owner thread: " PTR_FORMAT, p2i(owner()));
}
// ----------------------------------------------------------------------------------
// Non-product code
//
#ifdef ASSERT
static Mutex::Rank _ranks[] = { Mutex::event, Mutex::service, Mutex::stackwatermark, Mutex::tty, Mutex::oopstorage,
Mutex::nosafepoint, Mutex::safepoint };
static const char* _rank_names[] = { "event", "service", "stackwatermark", "tty", "oopstorage",
"nosafepoint", "safepoint" };
static const int _num_ranks = 7;
static const char* rank_name_internal(Mutex::Rank r) {
// Find closest rank and print out the name
stringStream st;
for (int i = 0; i < _num_ranks; i++) {
if (r == _ranks[i]) {
return _rank_names[i];
} else if (r > _ranks[i] && (i < _num_ranks-1 && r < _ranks[i+1])) {
int delta = static_cast<int>(_ranks[i+1]) - static_cast<int>(r);
st.print("%s-%d", _rank_names[i+1], delta);
return st.as_string();
}
}
return "fail";
}
const char* Mutex::rank_name() const {
return rank_name_internal(_rank);
}
void Mutex::assert_no_overlap(Rank orig, Rank adjusted, int adjust) {
int i = 0;
while (_ranks[i] < orig) i++;
// underflow is caught in constructor
if (i != 0 && adjusted > event && adjusted <= _ranks[i-1]) {
ResourceMark rm;
assert(adjusted > _ranks[i-1],
"Rank %s-%d overlaps with %s",
rank_name_internal(orig), adjust, rank_name_internal(adjusted));
}
}
#endif // ASSERT
#ifndef PRODUCT
void Mutex::print_on(outputStream* st) const {
st->print("Mutex: [" PTR_FORMAT "] %s - owner: " PTR_FORMAT,
p2i(this), _name, p2i(owner()));
if (_allow_vm_block) {
st->print("%s", " allow_vm_block");
}
DEBUG_ONLY(st->print(" %s", rank_name()));
st->cr();
}
void Mutex::print() const {
print_on(::tty);
}
#endif // PRODUCT
#ifdef ASSERT
void Mutex::assert_owner(Thread * expected) {
const char* msg = "invalid owner";
if (expected == NULL) {
msg = "should be un-owned";
}
else if (expected == Thread::current()) {
msg = "should be owned by current thread";
}
assert(owner() == expected,
"%s: owner=" INTPTR_FORMAT ", should be=" INTPTR_FORMAT,
msg, p2i(owner()), p2i(expected));
}
Mutex* Mutex::get_least_ranked_lock(Mutex* locks) {
Mutex *res, *tmp;
for (res = tmp = locks; tmp != NULL; tmp = tmp->next()) {
if (tmp->rank() < res->rank()) {
res = tmp;
}
}
return res;
}
Mutex* Mutex::get_least_ranked_lock_besides_this(Mutex* locks) {
Mutex *res, *tmp;
for (res = NULL, tmp = locks; tmp != NULL; tmp = tmp->next()) {
if (tmp != this && (res == NULL || tmp->rank() < res->rank())) {
res = tmp;
}
}
assert(res != this, "invariant");
return res;
}
// Tests for rank violations that might indicate exposure to deadlock.
void Mutex::check_rank(Thread* thread) {
Mutex* locks_owned = thread->owned_locks();
// We expect the locks already acquired to be in increasing rank order,
// modulo locks acquired in try_lock_without_rank_check()
for (Mutex* tmp = locks_owned; tmp != NULL; tmp = tmp->next()) {
if (tmp->next() != NULL) {
assert(tmp->rank() < tmp->next()->rank()
|| tmp->skip_rank_check(), "mutex rank anomaly?");
}
}
if (owned_by_self()) {
// wait() case
Mutex* least = get_least_ranked_lock_besides_this(locks_owned);
// For JavaThreads, we enforce not holding locks of rank nosafepoint or lower while waiting
// because the held lock has a NoSafepointVerifier so waiting on a lower ranked lock will not be
// able to check for safepoints first with a TBIVM.
// For all threads, we enforce not holding the tty lock or below, since this could block progress also.
// Also "this" should be the monitor with lowest rank owned by this thread.
if (least != NULL && ((least->rank() <= Mutex::nosafepoint && thread->is_Java_thread()) ||
least->rank() <= Mutex::tty ||
least->rank() <= this->rank())) {
ResourceMark rm(thread);
assert(false, "Attempting to wait on monitor %s/%s while holding lock %s/%s -- "
"possible deadlock. %s", name(), rank_name(), least->name(), least->rank_name(),
least->rank() <= this->rank() ?
"Should wait on the least ranked monitor from all owned locks." :
thread->is_Java_thread() ?
"Should not block(wait) while holding a lock of rank nosafepoint or below." :
"Should not block(wait) while holding a lock of rank tty or below.");
}
} else {
// lock()/lock_without_safepoint_check()/try_lock() case
Mutex* least = get_least_ranked_lock(locks_owned);
// Deadlock prevention rules require us to acquire Mutexes only in
// a global total order. For example, if m1 is the lowest ranked mutex
// that the thread holds and m2 is the mutex the thread is trying
// to acquire, then deadlock prevention rules require that the rank
// of m2 be less than the rank of m1. This prevents circular waits.
if (least != NULL && least->rank() <= this->rank()) {
ResourceMark rm(thread);
if (least->rank() > Mutex::tty) {
// Printing owned locks acquires tty lock. If the least rank was below or equal
// tty, then deadlock detection code would circle back here, until we run
// out of stack and crash hard. Print locks only when it is safe.
thread->print_owned_locks();
}
assert(false, "Attempting to acquire lock %s/%s out of order with lock %s/%s -- "
"possible deadlock", this->name(), this->rank_name(), least->name(), least->rank_name());
}
}
}
// Called immediately after lock acquisition or release as a diagnostic
// to track the lock-set of the thread.
// Rather like an EventListener for _owner (:>).
void Mutex::set_owner_implementation(Thread *new_owner) {
// This function is solely responsible for maintaining
// and checking the invariant that threads and locks
// are in a 1/N relation, with some some locks unowned.
// It uses the Mutex::_owner, Mutex::_next, and
// Thread::_owned_locks fields, and no other function
// changes those fields.
// It is illegal to set the mutex from one non-NULL
// owner to another--it must be owned by NULL as an
// intermediate state.
if (new_owner != NULL) {
// the thread is acquiring this lock
assert(new_owner == Thread::current(), "Should I be doing this?");
assert(owner() == NULL, "setting the owner thread of an already owned mutex");
raw_set_owner(new_owner); // set the owner
// link "this" into the owned locks list
this->_next = new_owner->_owned_locks;
new_owner->_owned_locks = this;
// NSV implied with locking allow_vm_block flag.
// The tty_lock is special because it is released for the safepoint by
// the safepoint mechanism.
if (new_owner->is_Java_thread() && _allow_vm_block && this != tty_lock) {
JavaThread::cast(new_owner)->inc_no_safepoint_count();
}
} else {
// the thread is releasing this lock
Thread* old_owner = owner();
_last_owner = old_owner;
_skip_rank_check = false;
assert(old_owner != NULL, "removing the owner thread of an unowned mutex");
assert(old_owner == Thread::current(), "removing the owner thread of an unowned mutex");
raw_set_owner(NULL); // set the owner
Mutex* locks = old_owner->owned_locks();
// remove "this" from the owned locks list
Mutex* prev = NULL;
bool found = false;
for (; locks != NULL; prev = locks, locks = locks->next()) {
if (locks == this) {
found = true;
break;
}
}
assert(found, "Removing a lock not owned");
if (prev == NULL) {
old_owner->_owned_locks = _next;
} else {
prev->_next = _next;
}
_next = NULL;
// ~NSV implied with locking allow_vm_block flag.
if (old_owner->is_Java_thread() && _allow_vm_block && this != tty_lock) {
JavaThread::cast(old_owner)->dec_no_safepoint_count();
}
}
}
#endif // ASSERT
¤ Dauer der Verarbeitung: 0.25 Sekunden
(vorverarbeitet)
¤
|
Haftungshinweis
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung ist noch experimentell.
|
