X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/f05224ef57c30ee1e71be443a160a236f7fe121a..932e745735378c442e2348a3a428d97b69398f86:/src/mc/SafetyChecker.cpp diff --git a/src/mc/SafetyChecker.cpp b/src/mc/SafetyChecker.cpp index 0a05e4a8e7..e7c62e8a3f 100644 --- a/src/mc/SafetyChecker.cpp +++ b/src/mc/SafetyChecker.cpp @@ -7,7 +7,9 @@ #include #include -#include +#include +#include +#include #include #include @@ -51,7 +53,7 @@ static int snapshot_compare(simgrid::mc::State* state1, simgrid::mc::State* stat return snapshot_compare(num1, s1, num2, s2); } -bool SafetyChecker::checkNonDeterminism(simgrid::mc::State* current_state) +bool SafetyChecker::checkNonTermination(simgrid::mc::State* current_state) { for (auto i = stack_.rbegin(); i != stack_.rend(); ++i) if (snapshot_compare(i->get(), current_state) == 0){ @@ -81,12 +83,9 @@ std::vector SafetyChecker::getTextualTrace() // override for (auto const& state : stack_) { int value; smx_simcall_t req = MC_state_get_executed_request(state.get(), &value); - if (req) { - char* req_str = simgrid::mc::request_to_string( - req, value, simgrid::mc::RequestType::executed); - trace.push_back(req_str); - xbt_free(req_str); - } + if (req) + trace.push_back(simgrid::mc::request_to_string( + req, value, simgrid::mc::RequestType::executed)); } return trace; } @@ -96,8 +95,6 @@ int SafetyChecker::run() this->init(); int value; - smx_simcall_t req = nullptr; - std::unique_ptr visited_state = nullptr; while (!stack_.empty()) { @@ -106,176 +103,181 @@ int SafetyChecker::run() XBT_DEBUG("**************************************************"); XBT_DEBUG( - "Exploration depth=%zi (state=%p, num %d)(%u interleave, user_max_depth %d)", + "Exploration depth=%zi (state=%p, num %d)(%zu interleave)", stack_.size(), state, state->num, - MC_state_interleave_size(state), user_max_depth_reached); + state->interleaveSize()); - /* Update statistics */ mc_stats->visited_states++; - /* If there are processes to interleave and the maximum depth has not been reached - then perform one step of the exploration algorithm */ - if (stack_.size() <= (std::size_t) _sg_mc_max_depth && !user_max_depth_reached - && (req = MC_state_get_request(state, &value)) && visited_state == nullptr) { - - char* req_str = simgrid::mc::request_to_string(req, value, simgrid::mc::RequestType::simix); - XBT_DEBUG("Execute: %s", req_str); - xbt_free(req_str); - - if (dot_output != nullptr) - req_str = simgrid::mc::request_get_dot_output(req, value); - - MC_state_set_executed_request(state, req, value); - mc_stats->executed_transitions++; + // The interleave set is empty or the maximum depth is reached, + // let's back-track. + smx_simcall_t req = nullptr; + if (stack_.size() > (std::size_t) _sg_mc_max_depth + || (req = MC_state_get_request(state, &value)) == nullptr + || visitedState_ != nullptr) { + int res = this->backtrack(); + if (res) + return res; + else + continue; + } - // TODO, bundle both operations in a single message - // MC_execute_transition(req, value) + // If there are processes to interleave and the maximum depth has not been + // reached then perform one step of the exploration algorithm. + XBT_DEBUG("Execute: %s", + simgrid::mc::request_to_string( + req, value, simgrid::mc::RequestType::simix).c_str()); - /* Answer the request */ - simgrid::mc::handle_simcall(req, value); - mc_model_checker->wait_for_requests(); + std::string req_str; + if (dot_output != nullptr) + req_str = simgrid::mc::request_get_dot_output(req, value); - /* Create the new expanded state */ - std::unique_ptr next_state = - std::unique_ptr(MC_state_new()); + mc_stats->executed_transitions++; - if (_sg_mc_termination && this->checkNonDeterminism(next_state.get())){ - MC_show_non_termination(); - return SIMGRID_MC_EXIT_NON_TERMINATION; - } + // TODO, bundle both operations in a single message + // MC_execute_transition(req, value) - if (_sg_mc_visited == 0 - || (visited_state = visitedStates_.addVisitedState(next_state.get(), true)) == nullptr) { + /* Answer the request */ + simgrid::mc::handle_simcall(req, value); + mc_model_checker->wait_for_requests(); - /* Get an enabled process and insert it in the interleave set of the next state */ - for (auto& p : mc_model_checker->process().simix_processes()) - if (simgrid::mc::process_is_enabled(&p.copy)) { - MC_state_interleave_process(next_state.get(), &p.copy); - if (reductionMode_ != simgrid::mc::ReductionMode::none) - break; - } + /* Create the new expanded state */ + std::unique_ptr next_state = + std::unique_ptr(MC_state_new()); - if (dot_output != nullptr) - std::fprintf(dot_output, "\"%d\" -> \"%d\" [%s];\n", state->num, next_state->num, req_str); + if (_sg_mc_termination && this->checkNonTermination(next_state.get())) { + MC_show_non_termination(); + return SIMGRID_MC_EXIT_NON_TERMINATION; + } - } else if (dot_output != nullptr) - std::fprintf(dot_output, "\"%d\" -> \"%d\" [%s];\n", state->num, visited_state->other_num == -1 ? visited_state->num : visited_state->other_num, req_str); + if (_sg_mc_visited == 0 + || (visitedState_ = visitedStates_.addVisitedState(next_state.get(), true)) == nullptr) { - stack_.push_back(std::move(next_state)); + /* Get an enabled process and insert it in the interleave set of the next state */ + for (auto& p : mc_model_checker->process().simix_processes()) + if (simgrid::mc::process_is_enabled(&p.copy)) { + MC_state_interleave_process(next_state.get(), &p.copy); + if (reductionMode_ != simgrid::mc::ReductionMode::none) + break; + } if (dot_output != nullptr) - xbt_free(req_str); + std::fprintf(dot_output, "\"%d\" -> \"%d\" [%s];\n", + state->num, next_state->num, req_str.c_str()); - /* Let's loop again */ + } else if (dot_output != nullptr) + std::fprintf(dot_output, "\"%d\" -> \"%d\" [%s];\n", + state->num, + visitedState_->other_num == -1 ? visitedState_->num : visitedState_->other_num, req_str.c_str()); - /* The interleave set is empty or the maximum depth is reached, let's back-track */ - } else { + stack_.push_back(std::move(next_state)); + } - if (stack_.size() > (std::size_t) _sg_mc_max_depth - || user_max_depth_reached - || visited_state != nullptr) { + XBT_INFO("No property violation found."); + MC_print_statistics(mc_stats); + initial_global_state = nullptr; + return SIMGRID_MC_EXIT_SUCCESS; +} - if (user_max_depth_reached && visited_state == nullptr) - XBT_DEBUG("User max depth reached !"); - else if (visited_state == nullptr) - XBT_WARN("/!\\ Max depth reached ! /!\\ "); - else - XBT_DEBUG("State already visited (equal to state %d), exploration stopped on this path.", visited_state->other_num == -1 ? visited_state->num : visited_state->other_num); +int SafetyChecker::backtrack() +{ + if (stack_.size() > (std::size_t) _sg_mc_max_depth + || visitedState_ != nullptr) { + if (visitedState_ == nullptr) + XBT_WARN("/!\\ Max depth reached ! /!\\ "); + else + XBT_DEBUG("State already visited (equal to state %d)," + " exploration stopped on this path.", + visitedState_->other_num == -1 ? visitedState_->num : visitedState_->other_num); + } else + XBT_DEBUG("There are no more processes to interleave. (depth %zi)", + stack_.size() + 1); + + stack_.pop_back(); + + visitedState_ = nullptr; + + /* Check for deadlocks */ + if (mc_model_checker->checkDeadlock()) { + MC_show_deadlock(); + return SIMGRID_MC_EXIT_DEADLOCK; + } - } else - XBT_DEBUG("There are no more processes to interleave. (depth %zi)", - stack_.size() + 1); + /* Traverse the stack backwards until a state with a non empty interleave + set is found, deleting all the states that have it empty in the way. + For each deleted state, check if the request that has generated it + (from it's predecesor state), depends on any other previous request + executed before it. If it does then add it to the interleave set of the + state that executed that previous request. */ - /* Trash the current state, no longer needed */ - XBT_DEBUG("Delete state %d at depth %zi", state->num, stack_.size()); - stack_.pop_back(); + while (!stack_.empty()) { + std::unique_ptr state = std::move(stack_.back()); + stack_.pop_back(); + if (reductionMode_ == simgrid::mc::ReductionMode::dpor) { + smx_simcall_t req = MC_state_get_internal_request(state.get()); + if (req->call == SIMCALL_MUTEX_LOCK || req->call == SIMCALL_MUTEX_TRYLOCK) + xbt_die("Mutex is currently not supported with DPOR, " + "use --cfg=model-check/reduction:none"); + const smx_process_t issuer = MC_smx_simcall_get_issuer(req); + for (auto i = stack_.rbegin(); i != stack_.rend(); ++i) { + simgrid::mc::State* prev_state = i->get(); + if (reductionMode_ != simgrid::mc::ReductionMode::none + && simgrid::mc::request_depend(req, MC_state_get_internal_request(prev_state))) { + if (XBT_LOG_ISENABLED(mc_safety, xbt_log_priority_debug)) { + XBT_DEBUG("Dependent Transitions:"); + int value; + smx_simcall_t prev_req = MC_state_get_executed_request(prev_state, &value); + XBT_DEBUG("%s (state=%d)", + simgrid::mc::request_to_string( + prev_req, value, simgrid::mc::RequestType::internal).c_str(), + prev_state->num); + prev_req = MC_state_get_executed_request(state.get(), &value); + XBT_DEBUG("%s (state=%d)", + simgrid::mc::request_to_string( + prev_req, value, simgrid::mc::RequestType::executed).c_str(), + state->num); + } - visited_state = nullptr; + if (!prev_state->processStates[issuer->pid].done()) + MC_state_interleave_process(prev_state, issuer); + else + XBT_DEBUG("Process %p is in done set", req->issuer); - /* Check for deadlocks */ - if (mc_model_checker->checkDeadlock()) { - MC_show_deadlock(); - return SIMGRID_MC_EXIT_DEADLOCK; - } + break; - /* Traverse the stack backwards until a state with a non empty interleave - set is found, deleting all the states that have it empty in the way. - For each deleted state, check if the request that has generated it - (from it's predecesor state), depends on any other previous request - executed before it. If it does then add it to the interleave set of the - state that executed that previous request. */ - - while (!stack_.empty()) { - std::unique_ptr state = std::move(stack_.back()); - stack_.pop_back(); - if (reductionMode_ == simgrid::mc::ReductionMode::dpor) { - req = MC_state_get_internal_request(state.get()); - if (req->call == SIMCALL_MUTEX_LOCK || req->call == SIMCALL_MUTEX_TRYLOCK) - xbt_die("Mutex is currently not supported with DPOR, " - "use --cfg=model-check/reduction:none"); - const smx_process_t issuer = MC_smx_simcall_get_issuer(req); - for (auto i = stack_.rbegin(); i != stack_.rend(); ++i) { - simgrid::mc::State* prev_state = i->get(); - if (reductionMode_ != simgrid::mc::ReductionMode::none - && simgrid::mc::request_depend(req, MC_state_get_internal_request(prev_state))) { - if (XBT_LOG_ISENABLED(mc_safety, xbt_log_priority_debug)) { - XBT_DEBUG("Dependent Transitions:"); - smx_simcall_t prev_req = MC_state_get_executed_request(prev_state, &value); - char* req_str = simgrid::mc::request_to_string(prev_req, value, simgrid::mc::RequestType::internal); - XBT_DEBUG("%s (state=%d)", req_str, prev_state->num); - xbt_free(req_str); - prev_req = MC_state_get_executed_request(state.get(), &value); - req_str = simgrid::mc::request_to_string(prev_req, value, simgrid::mc::RequestType::executed); - XBT_DEBUG("%s (state=%d)", req_str, state->num); - xbt_free(req_str); - } - - if (!MC_state_process_is_done(prev_state, issuer)) - MC_state_interleave_process(prev_state, issuer); - else - XBT_DEBUG("Process %p is in done set", req->issuer); - - break; - - } else if (req->issuer == MC_state_get_internal_request(prev_state)->issuer) { - - XBT_DEBUG("Simcall %d and %d with same issuer", req->call, MC_state_get_internal_request(prev_state)->call); - break; - - } else { - - const smx_process_t previous_issuer = MC_smx_simcall_get_issuer(MC_state_get_internal_request(prev_state)); - XBT_DEBUG("Simcall %d, process %lu (state %d) and simcall %d, process %lu (state %d) are independant", - req->call, issuer->pid, state->num, - MC_state_get_internal_request(prev_state)->call, - previous_issuer->pid, - prev_state->num); - - } - } - } + } else if (req->issuer == MC_state_get_internal_request(prev_state)->issuer) { - if (MC_state_interleave_size(state.get()) - && stack_.size() < (std::size_t) _sg_mc_max_depth) { - /* We found a back-tracking point, let's loop */ - XBT_DEBUG("Back-tracking to state %d at depth %zi", - state->num, stack_.size() + 1); - stack_.push_back(std::move(state)); - simgrid::mc::replay(stack_); - XBT_DEBUG("Back-tracking to state %d at depth %zi done", - stack_.back()->num, stack_.size()); + XBT_DEBUG("Simcall %d and %d with same issuer", req->call, MC_state_get_internal_request(prev_state)->call); break; + } else { - XBT_DEBUG("Delete state %d at depth %zi", - state->num, stack_.size() + 1); + + const smx_process_t previous_issuer = MC_smx_simcall_get_issuer(MC_state_get_internal_request(prev_state)); + XBT_DEBUG("Simcall %d, process %lu (state %d) and simcall %d, process %lu (state %d) are independant", + req->call, issuer->pid, state->num, + MC_state_get_internal_request(prev_state)->call, + previous_issuer->pid, + prev_state->num); + } } } - } - XBT_INFO("No property violation found."); - MC_print_statistics(mc_stats); - initial_global_state = nullptr; + if (state->interleaveSize() + && stack_.size() < (std::size_t) _sg_mc_max_depth) { + /* We found a back-tracking point, let's loop */ + XBT_DEBUG("Back-tracking to state %d at depth %zi", + state->num, stack_.size() + 1); + stack_.push_back(std::move(state)); + simgrid::mc::replay(stack_); + XBT_DEBUG("Back-tracking to state %d at depth %zi done", + stack_.back()->num, stack_.size()); + break; + } else { + XBT_DEBUG("Delete state %d at depth %zi", + state->num, stack_.size() + 1); + } + } return SIMGRID_MC_EXIT_SUCCESS; }