From e6918187568dbd01842d8d1d2c808ce16a894239 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 21 Apr 2024 13:54:28 +0200 Subject: Adding upstream version 18.2.2. Signed-off-by: Daniel Baumann --- src/rgw/rgw_coroutine.cc | 1130 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 1130 insertions(+) create mode 100644 src/rgw/rgw_coroutine.cc (limited to 'src/rgw/rgw_coroutine.cc') diff --git a/src/rgw/rgw_coroutine.cc b/src/rgw/rgw_coroutine.cc new file mode 100644 index 000000000..a9c9c38e3 --- /dev/null +++ b/src/rgw/rgw_coroutine.cc @@ -0,0 +1,1130 @@ +// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*- +// vim: ts=8 sw=2 smarttab ft=cpp + +#include "include/Context.h" +#include "common/ceph_json.h" +#include "rgw_coroutine.h" + +// re-include our assert to clobber the system one; fix dout: +#include "include/ceph_assert.h" + +#include + +#define dout_subsys ceph_subsys_rgw +#define dout_context g_ceph_context + +using namespace std; + +class RGWCompletionManager::WaitContext : public Context { + RGWCompletionManager *manager; + void *opaque; +public: + WaitContext(RGWCompletionManager *_cm, void *_opaque) : manager(_cm), opaque(_opaque) {} + void finish(int r) override { + manager->_wakeup(opaque); + } +}; + +RGWCompletionManager::RGWCompletionManager(CephContext *_cct) : cct(_cct), + timer(cct, lock) +{ + timer.init(); +} + +RGWCompletionManager::~RGWCompletionManager() +{ + std::lock_guard l{lock}; + timer.cancel_all_events(); + timer.shutdown(); +} + +void RGWCompletionManager::complete(RGWAioCompletionNotifier *cn, const rgw_io_id& io_id, void *user_info) +{ + std::lock_guard l{lock}; + _complete(cn, io_id, user_info); +} + +void RGWCompletionManager::register_completion_notifier(RGWAioCompletionNotifier *cn) +{ + std::lock_guard l{lock}; + if (cn) { + cns.insert(cn); + } +} + +void RGWCompletionManager::unregister_completion_notifier(RGWAioCompletionNotifier *cn) +{ + std::lock_guard l{lock}; + if (cn) { + cns.erase(cn); + } +} + +void RGWCompletionManager::_complete(RGWAioCompletionNotifier *cn, const rgw_io_id& io_id, void *user_info) +{ + if (cn) { + cns.erase(cn); + } + + if (complete_reqs_set.find(io_id) != complete_reqs_set.end()) { + /* already have completion for this io_id, don't allow multiple completions for it */ + return; + } + complete_reqs.push_back(io_completion{io_id, user_info}); + cond.notify_all(); +} + +int RGWCompletionManager::get_next(io_completion *io) +{ + std::unique_lock l{lock}; + while (complete_reqs.empty()) { + if (going_down) { + return -ECANCELED; + } + cond.wait(l); + } + *io = complete_reqs.front(); + complete_reqs_set.erase(io->io_id); + complete_reqs.pop_front(); + return 0; +} + +bool RGWCompletionManager::try_get_next(io_completion *io) +{ + std::lock_guard l{lock}; + if (complete_reqs.empty()) { + return false; + } + *io = complete_reqs.front(); + complete_reqs_set.erase(io->io_id); + complete_reqs.pop_front(); + return true; +} + +void RGWCompletionManager::go_down() +{ + std::lock_guard l{lock}; + for (auto cn : cns) { + cn->unregister(); + } + going_down = true; + cond.notify_all(); +} + +void RGWCompletionManager::wait_interval(void *opaque, const utime_t& interval, void *user_info) +{ + std::lock_guard l{lock}; + ceph_assert(waiters.find(opaque) == waiters.end()); + waiters[opaque] = user_info; + timer.add_event_after(interval, new WaitContext(this, opaque)); +} + +void RGWCompletionManager::wakeup(void *opaque) +{ + std::lock_guard l{lock}; + _wakeup(opaque); +} + +void RGWCompletionManager::_wakeup(void *opaque) +{ + map::iterator iter = waiters.find(opaque); + if (iter != waiters.end()) { + void *user_id = iter->second; + waiters.erase(iter); + _complete(NULL, rgw_io_id{0, -1} /* no IO id */, user_id); + } +} + +RGWCoroutine::~RGWCoroutine() { + for (auto stack : spawned.entries) { + stack->put(); + } +} + +void RGWCoroutine::init_new_io(RGWIOProvider *io_provider) +{ + ceph_assert(stack); // if there's no stack, io_provider won't be uninitialized + stack->init_new_io(io_provider); +} + +void RGWCoroutine::set_io_blocked(bool flag) { + if (stack) { + stack->set_io_blocked(flag); + } +} + +void RGWCoroutine::set_sleeping(bool flag) { + if (stack) { + stack->set_sleeping(flag); + } +} + +int RGWCoroutine::io_block(int ret, int64_t io_id) { + return io_block(ret, rgw_io_id{io_id, -1}); +} + +int RGWCoroutine::io_block(int ret, const rgw_io_id& io_id) { + if (!stack) { + return 0; + } + if (stack->consume_io_finish(io_id)) { + return 0; + } + set_io_blocked(true); + stack->set_io_blocked_id(io_id); + return ret; +} + +void RGWCoroutine::io_complete(const rgw_io_id& io_id) { + if (stack) { + stack->io_complete(io_id); + } +} + +void RGWCoroutine::StatusItem::dump(Formatter *f) const { + ::encode_json("timestamp", timestamp, f); + ::encode_json("status", status, f); +} + +stringstream& RGWCoroutine::Status::set_status() +{ + std::unique_lock l{lock}; + string s = status.str(); + status.str(string()); + if (!timestamp.is_zero()) { + history.push_back(StatusItem(timestamp, s)); + } + if (history.size() > (size_t)max_history) { + history.pop_front(); + } + timestamp = ceph_clock_now(); + + return status; +} + +RGWCoroutinesManager::~RGWCoroutinesManager() { + stop(); + completion_mgr->put(); + if (cr_registry) { + cr_registry->remove(this); + } +} + +int64_t RGWCoroutinesManager::get_next_io_id() +{ + return (int64_t)++max_io_id; +} + +uint64_t RGWCoroutinesManager::get_next_stack_id() { + return (uint64_t)++max_stack_id; +} + +RGWCoroutinesStack::RGWCoroutinesStack(CephContext *_cct, RGWCoroutinesManager *_ops_mgr, RGWCoroutine *start) : cct(_cct), ops_mgr(_ops_mgr), + done_flag(false), error_flag(false), blocked_flag(false), + sleep_flag(false), interval_wait_flag(false), is_scheduled(false), is_waiting_for_child(false), + retcode(0), run_count(0), + env(NULL), parent(NULL) +{ + id = ops_mgr->get_next_stack_id(); + if (start) { + ops.push_back(start); + } + pos = ops.begin(); +} + +RGWCoroutinesStack::~RGWCoroutinesStack() +{ + for (auto op : ops) { + op->put(); + } + + for (auto stack : spawned.entries) { + stack->put(); + } +} + +int RGWCoroutinesStack::operate(const DoutPrefixProvider *dpp, RGWCoroutinesEnv *_env) +{ + env = _env; + RGWCoroutine *op = *pos; + op->stack = this; + ldpp_dout(dpp, 20) << *op << ": operate()" << dendl; + int r = op->operate_wrapper(dpp); + if (r < 0) { + ldpp_dout(dpp, 20) << *op << ": operate() returned r=" << r << dendl; + } + + error_flag = op->is_error(); + + if (op->is_done()) { + int op_retcode = r; + r = unwind(op_retcode); + op->put(); + done_flag = (pos == ops.end()); + blocked_flag &= !done_flag; + if (done_flag) { + retcode = op_retcode; + } + return r; + } + + /* should r ever be negative at this point? */ + ceph_assert(r >= 0); + + return 0; +} + +string RGWCoroutinesStack::error_str() +{ + if (pos != ops.end()) { + return (*pos)->error_str(); + } + return string(); +} + +void RGWCoroutinesStack::call(RGWCoroutine *next_op) { + if (!next_op) { + return; + } + ops.push_back(next_op); + if (pos != ops.end()) { + ++pos; + } else { + pos = ops.begin(); + } +} + +void RGWCoroutinesStack::schedule() +{ + env->manager->schedule(env, this); +} + +void RGWCoroutinesStack::_schedule() +{ + env->manager->_schedule(env, this); +} + +RGWCoroutinesStack *RGWCoroutinesStack::spawn(RGWCoroutine *source_op, RGWCoroutine *op, bool wait) +{ + if (!op) { + return NULL; + } + + rgw_spawned_stacks *s = (source_op ? &source_op->spawned : &spawned); + + RGWCoroutinesStack *stack = env->manager->allocate_stack(); + s->add_pending(stack); + stack->parent = this; + + stack->get(); /* we'll need to collect the stack */ + stack->call(op); + + env->manager->schedule(env, stack); + + if (wait) { + set_blocked_by(stack); + } + + return stack; +} + +RGWCoroutinesStack *RGWCoroutinesStack::spawn(RGWCoroutine *op, bool wait) +{ + return spawn(NULL, op, wait); +} + +int RGWCoroutinesStack::wait(const utime_t& interval) +{ + RGWCompletionManager *completion_mgr = env->manager->get_completion_mgr(); + completion_mgr->wait_interval((void *)this, interval, (void *)this); + set_io_blocked(true); + set_interval_wait(true); + return 0; +} + +void RGWCoroutinesStack::wakeup() +{ + RGWCompletionManager *completion_mgr = env->manager->get_completion_mgr(); + completion_mgr->wakeup((void *)this); +} + +void RGWCoroutinesStack::io_complete(const rgw_io_id& io_id) +{ + RGWCompletionManager *completion_mgr = env->manager->get_completion_mgr(); + completion_mgr->complete(nullptr, io_id, (void *)this); +} + +int RGWCoroutinesStack::unwind(int retcode) +{ + rgw_spawned_stacks *src_spawned = &(*pos)->spawned; + + if (pos == ops.begin()) { + ldout(cct, 15) << "stack " << (void *)this << " end" << dendl; + spawned.inherit(src_spawned); + ops.clear(); + pos = ops.end(); + return retcode; + } + + --pos; + ops.pop_back(); + RGWCoroutine *op = *pos; + op->set_retcode(retcode); + op->spawned.inherit(src_spawned); + return 0; +} + +void RGWCoroutinesStack::cancel() +{ + while (!ops.empty()) { + RGWCoroutine *op = *pos; + unwind(-ECANCELED); + op->put(); + } + put(); +} + +bool RGWCoroutinesStack::collect(RGWCoroutine *op, int *ret, RGWCoroutinesStack *skip_stack, uint64_t *stack_id) /* returns true if needs to be called again */ +{ + bool need_retry = false; + rgw_spawned_stacks *s = (op ? &op->spawned : &spawned); + *ret = 0; + vector new_list; + + for (vector::iterator iter = s->entries.begin(); iter != s->entries.end(); ++iter) { + RGWCoroutinesStack *stack = *iter; + if (stack == skip_stack || !stack->is_done()) { + new_list.push_back(stack); + if (!stack->is_done()) { + ldout(cct, 20) << "collect(): s=" << (void *)this << " stack=" << (void *)stack << " is still running" << dendl; + } else if (stack == skip_stack) { + ldout(cct, 20) << "collect(): s=" << (void *)this << " stack=" << (void *)stack << " explicitly skipping stack" << dendl; + } + continue; + } + if (stack_id) { + *stack_id = stack->get_id(); + } + int r = stack->get_ret_status(); + stack->put(); + if (r < 0) { + *ret = r; + ldout(cct, 20) << "collect(): s=" << (void *)this << " stack=" << (void *)stack << " encountered error (r=" << r << "), skipping next stacks" << dendl; + new_list.insert(new_list.end(), ++iter, s->entries.end()); + need_retry = (iter != s->entries.end()); + break; + } + + ldout(cct, 20) << "collect(): s=" << (void *)this << " stack=" << (void *)stack << " is complete" << dendl; + } + + s->entries.swap(new_list); + return need_retry; +} + +bool RGWCoroutinesStack::collect_next(RGWCoroutine *op, int *ret, RGWCoroutinesStack **collected_stack) /* returns true if found a stack to collect */ +{ + rgw_spawned_stacks *s = (op ? &op->spawned : &spawned); + *ret = 0; + + if (collected_stack) { + *collected_stack = NULL; + } + + for (vector::iterator iter = s->entries.begin(); iter != s->entries.end(); ++iter) { + RGWCoroutinesStack *stack = *iter; + if (!stack->is_done()) { + continue; + } + int r = stack->get_ret_status(); + if (r < 0) { + *ret = r; + } + + if (collected_stack) { + *collected_stack = stack; + } + stack->put(); + + s->entries.erase(iter); + return true; + } + + return false; +} + +bool RGWCoroutinesStack::collect(int *ret, RGWCoroutinesStack *skip_stack, uint64_t *stack_id) /* returns true if needs to be called again */ +{ + return collect(NULL, ret, skip_stack, stack_id); +} + +static void _aio_completion_notifier_cb(librados::completion_t cb, void *arg) +{ + (static_cast(arg))->cb(); +} + +RGWAioCompletionNotifier::RGWAioCompletionNotifier(RGWCompletionManager *_mgr, const rgw_io_id& _io_id, void *_user_data) : completion_mgr(_mgr), + io_id(_io_id), + user_data(_user_data), registered(true) { + c = librados::Rados::aio_create_completion(this, _aio_completion_notifier_cb); +} + +RGWAioCompletionNotifier *RGWCoroutinesStack::create_completion_notifier() +{ + return ops_mgr->create_completion_notifier(this); +} + +RGWCompletionManager *RGWCoroutinesStack::get_completion_mgr() +{ + return ops_mgr->get_completion_mgr(); +} + +bool RGWCoroutinesStack::unblock_stack(RGWCoroutinesStack **s) +{ + if (blocking_stacks.empty()) { + return false; + } + + set::iterator iter = blocking_stacks.begin(); + *s = *iter; + blocking_stacks.erase(iter); + (*s)->blocked_by_stack.erase(this); + + return true; +} + +void RGWCoroutinesManager::report_error(RGWCoroutinesStack *op) +{ + if (!op) { + return; + } + string err = op->error_str(); + if (err.empty()) { + return; + } + lderr(cct) << "ERROR: failed operation: " << op->error_str() << dendl; +} + +void RGWCoroutinesStack::dump(Formatter *f) const { + stringstream ss; + ss << (void *)this; + ::encode_json("stack", ss.str(), f); + ::encode_json("run_count", run_count, f); + f->open_array_section("ops"); + for (auto& i : ops) { + encode_json("op", *i, f); + } + f->close_section(); +} + +void RGWCoroutinesStack::init_new_io(RGWIOProvider *io_provider) +{ + io_provider->set_io_user_info((void *)this); + io_provider->assign_io(env->manager->get_io_id_provider()); +} + +bool RGWCoroutinesStack::try_io_unblock(const rgw_io_id& io_id) +{ + if (!can_io_unblock(io_id)) { + auto p = io_finish_ids.emplace(io_id.id, io_id); + auto& iter = p.first; + bool inserted = p.second; + if (!inserted) { /* could not insert, entry already existed, add channel to completion mask */ + iter->second.channels |= io_id.channels; + } + return false; + } + + return true; +} + +bool RGWCoroutinesStack::consume_io_finish(const rgw_io_id& io_id) +{ + auto iter = io_finish_ids.find(io_id.id); + if (iter == io_finish_ids.end()) { + return false; + } + int finish_mask = iter->second.channels; + bool found = (finish_mask & io_id.channels) != 0; + + finish_mask &= ~(finish_mask & io_id.channels); + + if (finish_mask == 0) { + io_finish_ids.erase(iter); + } + return found; +} + + +void RGWCoroutinesManager::handle_unblocked_stack(set& context_stacks, list& scheduled_stacks, + RGWCompletionManager::io_completion& io, int *blocked_count, int *interval_wait_count) +{ + ceph_assert(ceph_mutex_is_wlocked(lock)); + RGWCoroutinesStack *stack = static_cast(io.user_info); + if (context_stacks.find(stack) == context_stacks.end()) { + return; + } + if (!stack->try_io_unblock(io.io_id)) { + return; + } + if (stack->is_io_blocked()) { + --(*blocked_count); + stack->set_io_blocked(false); + if (stack->is_interval_waiting()) { + --(*interval_wait_count); + } + } + stack->set_interval_wait(false); + if (!stack->is_done()) { + if (!stack->is_scheduled) { + scheduled_stacks.push_back(stack); + stack->set_is_scheduled(true); + } + } else { + context_stacks.erase(stack); + stack->put(); + } +} + +void RGWCoroutinesManager::schedule(RGWCoroutinesEnv *env, RGWCoroutinesStack *stack) +{ + std::unique_lock wl{lock}; + _schedule(env, stack); +} + +void RGWCoroutinesManager::_schedule(RGWCoroutinesEnv *env, RGWCoroutinesStack *stack) +{ + ceph_assert(ceph_mutex_is_wlocked(lock)); + if (!stack->is_scheduled) { + env->scheduled_stacks->push_back(stack); + stack->set_is_scheduled(true); + } + set& context_stacks = run_contexts[env->run_context]; + context_stacks.insert(stack); +} + +void RGWCoroutinesManager::set_sleeping(RGWCoroutine *cr, bool flag) +{ + cr->set_sleeping(flag); +} + +void RGWCoroutinesManager::io_complete(RGWCoroutine *cr, const rgw_io_id& io_id) +{ + cr->io_complete(io_id); +} + +int RGWCoroutinesManager::run(const DoutPrefixProvider *dpp, list& stacks) +{ + int ret = 0; + int blocked_count = 0; + int interval_wait_count = 0; + bool canceled = false; // set on going_down + RGWCoroutinesEnv env; + bool op_not_blocked; + + uint64_t run_context = ++run_context_count; + + lock.lock(); + set& context_stacks = run_contexts[run_context]; + list scheduled_stacks; + for (auto& st : stacks) { + context_stacks.insert(st); + scheduled_stacks.push_back(st); + st->set_is_scheduled(true); + } + env.run_context = run_context; + env.manager = this; + env.scheduled_stacks = &scheduled_stacks; + + for (list::iterator iter = scheduled_stacks.begin(); iter != scheduled_stacks.end() && !going_down;) { + RGWCompletionManager::io_completion io; + RGWCoroutinesStack *stack = *iter; + ++iter; + scheduled_stacks.pop_front(); + + if (context_stacks.find(stack) == context_stacks.end()) { + /* stack was probably schedule more than once due to IO, but was since complete */ + goto next; + } + env.stack = stack; + + lock.unlock(); + + ret = stack->operate(dpp, &env); + + lock.lock(); + + stack->set_is_scheduled(false); + if (ret < 0) { + ldpp_dout(dpp, 20) << "stack->operate() returned ret=" << ret << dendl; + } + + if (stack->is_error()) { + report_error(stack); + } + + op_not_blocked = false; + + if (stack->is_io_blocked()) { + ldout(cct, 20) << __func__ << ":" << " stack=" << (void *)stack << " is io blocked" << dendl; + if (stack->is_interval_waiting()) { + interval_wait_count++; + } + blocked_count++; + } else if (stack->is_blocked()) { + /* do nothing, we'll re-add the stack when the blocking stack is done, + * or when we're awaken + */ + ldout(cct, 20) << __func__ << ":" << " stack=" << (void *)stack << " is_blocked_by_stack()=" << stack->is_blocked_by_stack() + << " is_sleeping=" << stack->is_sleeping() << " waiting_for_child()=" << stack->waiting_for_child() << dendl; + } else if (stack->is_done()) { + ldout(cct, 20) << __func__ << ":" << " stack=" << (void *)stack << " is done" << dendl; + RGWCoroutinesStack *s; + while (stack->unblock_stack(&s)) { + if (!s->is_blocked_by_stack() && !s->is_done()) { + if (s->is_io_blocked()) { + if (stack->is_interval_waiting()) { + interval_wait_count++; + } + blocked_count++; + } else { + s->_schedule(); + } + } + } + if (stack->parent && stack->parent->waiting_for_child()) { + stack->parent->set_wait_for_child(false); + stack->parent->_schedule(); + } + context_stacks.erase(stack); + stack->put(); + stack = NULL; + } else { + op_not_blocked = true; + stack->run_count++; + stack->_schedule(); + } + + if (!op_not_blocked && stack) { + stack->run_count = 0; + } + + while (completion_mgr->try_get_next(&io)) { + handle_unblocked_stack(context_stacks, scheduled_stacks, io, &blocked_count, &interval_wait_count); + } + + /* + * only account blocked operations that are not in interval_wait, these are stacks that + * were put on a wait without any real IO operations. While we mark these as io_blocked, + * these aren't really waiting for IOs + */ + while (blocked_count - interval_wait_count >= ops_window) { + lock.unlock(); + ret = completion_mgr->get_next(&io); + lock.lock(); + if (ret < 0) { + ldout(cct, 5) << "completion_mgr.get_next() returned ret=" << ret << dendl; + } + handle_unblocked_stack(context_stacks, scheduled_stacks, io, &blocked_count, &interval_wait_count); + } + +next: + while (scheduled_stacks.empty() && blocked_count > 0) { + lock.unlock(); + ret = completion_mgr->get_next(&io); + lock.lock(); + if (ret < 0) { + ldout(cct, 5) << "completion_mgr.get_next() returned ret=" << ret << dendl; + } + if (going_down) { + ldout(cct, 5) << __func__ << "(): was stopped, exiting" << dendl; + ret = -ECANCELED; + canceled = true; + break; + } + handle_unblocked_stack(context_stacks, scheduled_stacks, io, &blocked_count, &interval_wait_count); + iter = scheduled_stacks.begin(); + } + if (canceled) { + break; + } + + if (iter == scheduled_stacks.end()) { + iter = scheduled_stacks.begin(); + } + } + + if (!context_stacks.empty() && !going_down) { + JSONFormatter formatter(true); + formatter.open_array_section("context_stacks"); + for (auto& s : context_stacks) { + ::encode_json("entry", *s, &formatter); + } + formatter.close_section(); + lderr(cct) << __func__ << "(): ERROR: deadlock detected, dumping remaining coroutines:\n"; + formatter.flush(*_dout); + *_dout << dendl; + ceph_assert(context_stacks.empty() || going_down); // assert on deadlock + } + + for (auto stack : context_stacks) { + ldout(cct, 20) << "clearing stack on run() exit: stack=" << (void *)stack << " nref=" << stack->get_nref() << dendl; + stack->cancel(); + } + run_contexts.erase(run_context); + lock.unlock(); + + return ret; +} + +int RGWCoroutinesManager::run(const DoutPrefixProvider *dpp, RGWCoroutine *op) +{ + if (!op) { + return 0; + } + list stacks; + RGWCoroutinesStack *stack = allocate_stack(); + op->get(); + stack->call(op); + + stacks.push_back(stack); + + int r = run(dpp, stacks); + if (r < 0) { + ldpp_dout(dpp, 20) << "run(stacks) returned r=" << r << dendl; + } else { + r = op->get_ret_status(); + } + op->put(); + + return r; +} + +RGWAioCompletionNotifier *RGWCoroutinesManager::create_completion_notifier(RGWCoroutinesStack *stack) +{ + rgw_io_id io_id{get_next_io_id(), -1}; + RGWAioCompletionNotifier *cn = new RGWAioCompletionNotifier(completion_mgr, io_id, (void *)stack); + completion_mgr->register_completion_notifier(cn); + return cn; +} + +void RGWCoroutinesManager::dump(Formatter *f) const { + std::shared_lock rl{lock}; + + f->open_array_section("run_contexts"); + for (auto& i : run_contexts) { + f->open_object_section("context"); + ::encode_json("id", i.first, f); + f->open_array_section("entries"); + for (auto& s : i.second) { + ::encode_json("entry", *s, f); + } + f->close_section(); + f->close_section(); + } + f->close_section(); +} + +RGWCoroutinesStack *RGWCoroutinesManager::allocate_stack() { + return new RGWCoroutinesStack(cct, this); +} + +string RGWCoroutinesManager::get_id() +{ + if (!id.empty()) { + return id; + } + stringstream ss; + ss << (void *)this; + return ss.str(); +} + +void RGWCoroutinesManagerRegistry::add(RGWCoroutinesManager *mgr) +{ + std::unique_lock wl{lock}; + if (managers.find(mgr) == managers.end()) { + managers.insert(mgr); + get(); + } +} + +void RGWCoroutinesManagerRegistry::remove(RGWCoroutinesManager *mgr) +{ + std::unique_lock wl{lock}; + if (managers.find(mgr) != managers.end()) { + managers.erase(mgr); + put(); + } +} + +RGWCoroutinesManagerRegistry::~RGWCoroutinesManagerRegistry() +{ + AdminSocket *admin_socket = cct->get_admin_socket(); + if (!admin_command.empty()) { + admin_socket->unregister_commands(this); + } +} + +int RGWCoroutinesManagerRegistry::hook_to_admin_command(const string& command) +{ + AdminSocket *admin_socket = cct->get_admin_socket(); + if (!admin_command.empty()) { + admin_socket->unregister_commands(this); + } + admin_command = command; + int r = admin_socket->register_command(admin_command, this, + "dump current coroutines stack state"); + if (r < 0) { + lderr(cct) << "ERROR: fail to register admin socket command (r=" << r << ")" << dendl; + return r; + } + return 0; +} + +int RGWCoroutinesManagerRegistry::call(std::string_view command, + const cmdmap_t& cmdmap, + const bufferlist&, + Formatter *f, + std::ostream& ss, + bufferlist& out) { + std::shared_lock rl{lock}; + ::encode_json("cr_managers", *this, f); + return 0; +} + +void RGWCoroutinesManagerRegistry::dump(Formatter *f) const { + f->open_array_section("coroutine_managers"); + for (auto m : managers) { + ::encode_json("entry", *m, f); + } + f->close_section(); +} + +void RGWCoroutine::call(RGWCoroutine *op) +{ + if (op) { + stack->call(op); + } else { + // the call()er expects this to set a retcode + retcode = 0; + } +} + +RGWCoroutinesStack *RGWCoroutine::spawn(RGWCoroutine *op, bool wait) +{ + return stack->spawn(this, op, wait); +} + +bool RGWCoroutine::collect(int *ret, RGWCoroutinesStack *skip_stack, uint64_t *stack_id) /* returns true if needs to be called again */ +{ + return stack->collect(this, ret, skip_stack, stack_id); +} + +bool RGWCoroutine::collect_next(int *ret, RGWCoroutinesStack **collected_stack) /* returns true if found a stack to collect */ +{ + return stack->collect_next(this, ret, collected_stack); +} + +int RGWCoroutine::wait(const utime_t& interval) +{ + return stack->wait(interval); +} + +void RGWCoroutine::wait_for_child() +{ + /* should only wait for child if there is a child that is not done yet, and no complete children */ + if (spawned.entries.empty()) { + return; + } + for (vector::iterator iter = spawned.entries.begin(); iter != spawned.entries.end(); ++iter) { + if ((*iter)->is_done()) { + return; + } + } + stack->set_wait_for_child(true); +} + +string RGWCoroutine::to_str() const +{ + return typeid(*this).name(); +} + +ostream& operator<<(ostream& out, const RGWCoroutine& cr) +{ + out << "cr:s=" << (void *)cr.get_stack() << ":op=" << (void *)&cr << ":" << typeid(cr).name(); + return out; +} + +bool RGWCoroutine::drain_children(int num_cr_left, + RGWCoroutinesStack *skip_stack, + std::optional > cb) +{ + bool done = false; + ceph_assert(num_cr_left >= 0); + if (num_cr_left == 0 && skip_stack) { + num_cr_left = 1; + } + reenter(&drain_status.cr) { + while (num_spawned() > (size_t)num_cr_left) { + yield wait_for_child(); + int ret; + uint64_t stack_id; + bool again = false; + do { + again = collect(&ret, skip_stack, &stack_id); + if (ret < 0) { + ldout(cct, 10) << "collect() returned ret=" << ret << dendl; + /* we should have reported this error */ + log_error() << "ERROR: collect() returned error (ret=" << ret << ")"; + } + if (cb) { + (*cb)(stack_id, ret); + } + } while (again); + } + done = true; + } + return done; +} + +bool RGWCoroutine::drain_children(int num_cr_left, + std::optional > cb) +{ + bool done = false; + ceph_assert(num_cr_left >= 0); + + reenter(&drain_status.cr) { + while (num_spawned() > (size_t)num_cr_left) { + yield wait_for_child(); + int ret; + uint64_t stack_id; + bool again = false; + do { + again = collect(&ret, nullptr, &stack_id); + if (ret < 0) { + ldout(cct, 10) << "collect() returned ret=" << ret << dendl; + /* we should have reported this error */ + log_error() << "ERROR: collect() returned error (ret=" << ret << ")"; + } + if (cb && !drain_status.should_exit) { + int r = (*cb)(stack_id, ret); + if (r < 0) { + drain_status.ret = r; + drain_status.should_exit = true; + num_cr_left = 0; /* need to drain all */ + } + } + } while (again); + } + done = true; + } + return done; +} + +void RGWCoroutine::wakeup() +{ + if (stack) { + stack->wakeup(); + } +} + +RGWCoroutinesEnv *RGWCoroutine::get_env() const +{ + return stack->get_env(); +} + +void RGWCoroutine::dump(Formatter *f) const { + if (!description.str().empty()) { + encode_json("description", description.str(), f); + } + encode_json("type", to_str(), f); + if (!spawned.entries.empty()) { + f->open_array_section("spawned"); + for (auto& i : spawned.entries) { + char buf[32]; + snprintf(buf, sizeof(buf), "%p", (void *)i); + encode_json("stack", string(buf), f); + } + f->close_section(); + } + if (!status.history.empty()) { + encode_json("history", status.history, f); + } + + if (!status.status.str().empty()) { + f->open_object_section("status"); + encode_json("status", status.status.str(), f); + encode_json("timestamp", status.timestamp, f); + f->close_section(); + } +} + +RGWSimpleCoroutine::~RGWSimpleCoroutine() +{ + if (!called_cleanup) { + request_cleanup(); + } +} + +void RGWSimpleCoroutine::call_cleanup() +{ + called_cleanup = true; + request_cleanup(); +} + +int RGWSimpleCoroutine::operate(const DoutPrefixProvider *dpp) +{ + int ret = 0; + reenter(this) { + yield return state_init(); + yield return state_send_request(dpp); + yield return state_request_complete(); + yield return state_all_complete(); + drain_all(); + call_cleanup(); + return set_state(RGWCoroutine_Done, ret); + } + return 0; +} + +int RGWSimpleCoroutine::state_init() +{ + int ret = init(); + if (ret < 0) { + call_cleanup(); + return set_state(RGWCoroutine_Error, ret); + } + return 0; +} + +int RGWSimpleCoroutine::state_send_request(const DoutPrefixProvider *dpp) +{ + int ret = send_request(dpp); + if (ret < 0) { + call_cleanup(); + return set_state(RGWCoroutine_Error, ret); + } + return io_block(0); +} + +int RGWSimpleCoroutine::state_request_complete() +{ + int ret = request_complete(); + if (ret < 0) { + call_cleanup(); + return set_state(RGWCoroutine_Error, ret); + } + return 0; +} + +int RGWSimpleCoroutine::state_all_complete() +{ + int ret = finish(); + if (ret < 0) { + call_cleanup(); + return set_state(RGWCoroutine_Error, ret); + } + return 0; +} + + -- cgit v1.2.3