X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/9c92907fc7a130fd6abc60b98791daa320742ba2..0f83a8218acedbbea411208a0a9bf89c4aca9455:/src/simix/smx_global.cpp diff --git a/src/simix/smx_global.cpp b/src/simix/smx_global.cpp index 3c94c9dd75..dfcf0bac9f 100644 --- a/src/simix/smx_global.cpp +++ b/src/simix/smx_global.cpp @@ -1,5 +1,4 @@ -/* Copyright (c) 2007-2015. The SimGrid Team. - * All rights reserved. */ +/* Copyright (c) 2007-2017. The SimGrid Team. All rights reserved. */ /* This program is free software; you can redistribute it and/or modify it * under the terms of the license (GNU LGPL) which comes with this package. */ @@ -13,7 +12,8 @@ #include -#include +#include "simgrid/s4u/engine.hpp" +#include "simgrid/s4u/host.hpp" #include "src/surf/surf_interface.hpp" #include "src/surf/storage_interface.hpp" @@ -33,11 +33,9 @@ #if HAVE_MC #include "src/mc/mc_private.h" -#include "src/mc/mc_protocol.h" -#include "src/mc/Client.hpp" - +#include "src/mc/remote/Client.hpp" +#include "src/mc/remote/mc_protocol.h" #include -#include "src/mc/mc_protocol.h" #endif #include "src/mc/mc_record.h" @@ -57,9 +55,8 @@ typedef struct s_smx_timer { double date = 0.0; simgrid::xbt::Task callback; - s_smx_timer() {} - s_smx_timer(double date, simgrid::xbt::Task callback) - : date(date), callback(std::move(callback)) {} + s_smx_timer()=default; + s_smx_timer(double date, simgrid::xbt::Task callback) : date(date), callback(std::move(callback)) {} } s_smx_timer_t; void (*SMPI_switch_data_segment)(int) = nullptr; @@ -81,30 +78,27 @@ static void inthandler(int ignored) static void segvhandler(int signum, siginfo_t *siginfo, void *context) { if (siginfo->si_signo == SIGSEGV && siginfo->si_code == SEGV_ACCERR) { - fprintf(stderr, - "Access violation detected.\n" - "This can result from a programming error in your code or, although less likely,\n" - "from a bug in SimGrid itself. This can also be the sign of a bug in the OS or\n" - "in third-party libraries. Failing hardware can sometimes generate such errors\n" - "too.\n" - "Finally, if nothing of the above applies, this can result from a stack overflow.\n" - "Try to increase stack size with --cfg=contexts/stack_size (current size is %d KiB).\n", + fprintf(stderr, "Access violation detected.\n" + "This probably comes from a programming error in your code, or from a stack\n" + "overflow. If you are certain of your code, try increasing the stack size\n" + " --cfg=contexts/stack-size=XXX (current size is %d KiB).\n" + "\n" + "If it does not help, this may have one of the following causes:\n" + "a bug in SimGrid, a bug in the OS or a bug in a third-party libraries.\n" + "Failing hardware can sometimes generate such errors too.\n" + "\n" + "If you think you've found a bug in SimGrid, please report it along with a\n" + "Minimal Working Example (MWE) reproducing your problem and a full backtrace\n" + "of the fault captured with gdb or valgrind.\n", smx_context_stack_size / 1024); - if (XBT_LOG_ISENABLED(simix_kernel, xbt_log_priority_debug)) { - fprintf(stderr, - "siginfo = {si_signo = %d, si_errno = %d, si_code = %d, si_addr = %p}\n", - siginfo->si_signo, siginfo->si_errno, siginfo->si_code, siginfo->si_addr); - } } else if (siginfo->si_signo == SIGSEGV) { fprintf(stderr, "Segmentation fault.\n"); #if HAVE_SMPI if (smpi_enabled() && !smpi_privatize_global_variables) { #if HAVE_PRIVATIZATION - fprintf(stderr, - "Try to enable SMPI variable privatization with --cfg=smpi/privatize-global-variables:yes.\n"); + fprintf(stderr, "Try to enable SMPI variable privatization with --cfg=smpi/privatize-global-variables:yes.\n"); #else - fprintf(stderr, - "Sadly, your system does not support --cfg=smpi/privatize-global-variables:yes (yet).\n"); + fprintf(stderr, "Sadly, your system does not support --cfg=smpi/privatize-global-variables:yes (yet).\n"); #endif /* HAVE_PRIVATIZATION */ } #endif /* HAVE_SMPI */ @@ -120,7 +114,8 @@ char sigsegv_stack[SIGSTKSZ]; /* alternate stack for SIGSEGV handler */ */ static void install_segvhandler() { - stack_t stack, old_stack; + stack_t stack; + stack_t old_stack; stack.ss_sp = sigsegv_stack; stack.ss_size = sizeof sigsegv_stack; stack.ss_flags = 0; @@ -135,7 +130,8 @@ static void install_segvhandler() sigaltstack(&old_stack, nullptr); } - struct sigaction action, old_action; + struct sigaction action; + struct sigaction old_action; action.sa_sigaction = &segvhandler; action.sa_flags = SA_ONSTACK | SA_RESETHAND | SA_SIGINFO; sigemptyset(&action.sa_mask); @@ -146,8 +142,7 @@ static void install_segvhandler() } if ((old_action.sa_flags & SA_SIGINFO) || old_action.sa_handler != SIG_DFL) { XBT_DEBUG("A signal handler was already installed for SIGSEGV (%p). Restore it.", - (old_action.sa_flags & SA_SIGINFO) ? - (void*)old_action.sa_sigaction : (void*)old_action.sa_handler); + (old_action.sa_flags & SA_SIGINFO) ? (void*)old_action.sa_sigaction : (void*)old_action.sa_handler); sigaction(SIGSEGV, &old_action, nullptr); } } @@ -160,7 +155,7 @@ double SIMIX_timer_next() return xbt_heap_size(simix_timers) > 0 ? xbt_heap_maxkey(simix_timers) : -1.0; } -static void kill_process(smx_process_t process) +static void kill_process(smx_actor_t process) { SIMIX_process_kill(process, nullptr); } @@ -170,6 +165,8 @@ static std::function maestro_code; namespace simgrid { namespace simix { +simgrid::xbt::signal onDeadlock; + XBT_PUBLIC(void) set_maestro(std::function code) { maestro_code = std::move(code); @@ -180,6 +177,9 @@ XBT_PUBLIC(void) set_maestro(std::function code) void SIMIX_set_maestro(void (*code)(void*), void* data) { +#ifdef _WIN32 + XBT_INFO("WARNING, SIMIX_set_maestro is believed to not work on windows. Please help us investigating this issue if you need that feature"); +#endif maestro_code = std::bind(code, data); } @@ -202,9 +202,8 @@ void SIMIX_global_init(int *argc, char **argv) simix_global = std::unique_ptr(new simgrid::simix::Global()); simgrid::simix::ActorImpl proc; - simix_global->process_to_run = xbt_dynar_new(sizeof(smx_process_t), nullptr); - simix_global->process_that_ran = xbt_dynar_new(sizeof(smx_process_t), nullptr); - simix_global->process_list = xbt_swag_new(xbt_swag_offset(proc, process_hookup)); + simix_global->process_to_run = xbt_dynar_new(sizeof(smx_actor_t), nullptr); + simix_global->process_that_ran = xbt_dynar_new(sizeof(smx_actor_t), nullptr); simix_global->process_to_destroy = xbt_swag_new(xbt_swag_offset(proc, destroy_hookup)); simix_global->maestro_process = nullptr; simix_global->create_process_function = &SIMIX_process_create; @@ -227,14 +226,13 @@ void SIMIX_global_init(int *argc, char **argv) #endif /* register a function to be called by SURF after the environment creation */ sg_platf_init(); - simgrid::surf::on_postparse.connect(SIMIX_post_create_environment); + simgrid::s4u::onPlatformCreated.connect(SIMIX_post_create_environment); simgrid::s4u::Host::onCreation.connect([](simgrid::s4u::Host& host) { - SIMIX_host_create(&host); + host.extension_set(new simgrid::simix::Host()); }); - SIMIX_HOST_LEVEL = simgrid::s4u::Host::extension_create(SIMIX_host_destroy); simgrid::surf::storageCreatedCallbacks.connect([](simgrid::surf::Storage* storage) { - const char* name = storage->getName(); + const char* name = storage->cname(); // TODO, create sg_storage_by_name sg_storage_t s = xbt_lib_get_elm_or_null(storage_lib, name); xbt_assert(s != nullptr, "Storage not found for name %s", name); @@ -265,11 +263,12 @@ int smx_cleaned = 0; */ void SIMIX_clean() { - if (smx_cleaned) return; // to avoid double cleaning by java and C + if (smx_cleaned) + return; // to avoid double cleaning by java and C #if HAVE_SMPI if (SIMIX_process_count()>0){ - if(smpi_process_initialized()){ + if(smpi_process()->initialized()){ xbt_die("Process exited without calling MPI_Finalize - Killing simulation"); }else{ XBT_WARN("Process called exit when leaving - Skipping cleanups"); @@ -280,7 +279,7 @@ void SIMIX_clean() smx_cleaned = 1; XBT_DEBUG("SIMIX_clean called. Simulation's over."); - if (!xbt_dynar_is_empty(simix_global->process_to_run) && SIMIX_get_clock() == 0.0) { + if (!xbt_dynar_is_empty(simix_global->process_to_run) && SIMIX_get_clock() <= 0.0) { XBT_CRITICAL(" "); XBT_CRITICAL("The time is still 0, and you still have processes ready to run."); XBT_CRITICAL("It seems that you forgot to run the simulation that you setup."); @@ -298,12 +297,14 @@ void SIMIX_clean() xbt_dynar_free(&simix_global->process_to_run); xbt_dynar_free(&simix_global->process_that_ran); xbt_swag_free(simix_global->process_to_destroy); - xbt_swag_free(simix_global->process_list); - simix_global->process_list = nullptr; + simix_global->process_list.clear(); simix_global->process_to_destroy = nullptr; xbt_os_mutex_destroy(simix_global->mutex); simix_global->mutex = nullptr; +#if HAVE_MC + xbt_dynar_free(&simix_global->actors_vector); +#endif /* Let's free maestro now */ delete simix_global->maestro_process->context; @@ -317,7 +318,6 @@ void SIMIX_clean() surf_exit(); simix_global = nullptr; - return; } @@ -338,7 +338,7 @@ double SIMIX_get_clock() static int process_syscall_color(void *p) { - switch ((*(smx_process_t *)p)->simcall.call) { + switch ((*(smx_actor_t *)p)->simcall.call) { case SIMCALL_NONE: case SIMCALL_PROCESS_KILL: return 2; @@ -352,15 +352,13 @@ static int process_syscall_color(void *p) /** Wake up all processes waiting for a Surf action to finish */ static void SIMIX_wake_processes() { - unsigned int iter; - surf_model_t model; surf_action_t action; - xbt_dynar_foreach(all_existing_models, iter, model) { + for(auto model : *all_existing_models) { XBT_DEBUG("Handling the processes whose action failed (if any)"); while ((action = surf_model_extract_failed_action_set(model))) { XBT_DEBUG(" Handling Action %p",action); - SIMIX_simcall_exit((smx_synchro_t) action->getData()); + SIMIX_simcall_exit((smx_activity_t) action->getData()); } XBT_DEBUG("Handling the processes whose action terminated normally (if any)"); while ((action = surf_model_extract_done_action_set(model))) { @@ -368,7 +366,7 @@ static void SIMIX_wake_processes() if (action->getData() == nullptr) XBT_DEBUG("probably vcpu's action %p, skip", action); else - SIMIX_simcall_exit((smx_synchro_t) action->getData()); + SIMIX_simcall_exit((smx_activity_t) action->getData()); } } } @@ -431,17 +429,14 @@ void SIMIX_run() } double time = 0; - smx_process_t process; do { - XBT_DEBUG("New Schedule Round; size(queue)=%lu", - xbt_dynar_length(simix_global->process_to_run)); + XBT_DEBUG("New Schedule Round; size(queue)=%lu", xbt_dynar_length(simix_global->process_to_run)); SIMIX_execute_tasks(); while (!xbt_dynar_is_empty(simix_global->process_to_run)) { - XBT_DEBUG("New Sub-Schedule Round; size(queue)=%lu", - xbt_dynar_length(simix_global->process_to_run)); + XBT_DEBUG("New Sub-Schedule Round; size(queue)=%lu", xbt_dynar_length(simix_global->process_to_run)); /* Run all processes that are ready to run, possibly in parallel */ SIMIX_process_runall(); @@ -505,6 +500,7 @@ void SIMIX_run() */ unsigned int iter; + smx_actor_t process; xbt_dynar_foreach(simix_global->process_that_ran, iter, process) { if (process->simcall.call != SIMCALL_NONE) { SIMIX_simcall_handle(&process->simcall, 0); @@ -519,7 +515,7 @@ void SIMIX_run() } time = SIMIX_timer_next(); - if (time != -1.0 || xbt_swag_size(simix_global->process_list) != 0) { + if (time > -1.0 || simix_global->process_list.empty() == false) { XBT_DEBUG("Calling surf_solve"); time = surf_solve(time); XBT_DEBUG("Moving time ahead : %g", time); @@ -539,29 +535,40 @@ void SIMIX_run() } while (again); /* Autorestart all process */ - char *hostname = nullptr; - unsigned int iter; - xbt_dynar_foreach(host_that_restart,iter,hostname) { - XBT_INFO("Restart processes on host: %s",hostname); - SIMIX_host_autorestart(sg_host_by_name(hostname)); + for (auto host: host_that_restart) { + XBT_INFO("Restart processes on host %s", host->cname()); + SIMIX_host_autorestart(host); } - xbt_dynar_reset(host_that_restart); + host_that_restart.clear(); /* Clean processes to destroy */ SIMIX_process_empty_trash(); - XBT_DEBUG("### time %f, empty %d", time, xbt_dynar_is_empty(simix_global->process_to_run)); + XBT_DEBUG("### time %f, #processes %zu, #to_run %lu", time, simix_global->process_list.size(), + xbt_dynar_length(simix_global->process_to_run)); + + /* If only daemon processes remain, cancel their actions, mark them to die and reschedule them */ + if (simix_global->process_list.size() == simix_global->daemons.size()) + for (const auto& dmon : simix_global->daemons) { + XBT_DEBUG("Kill %s", dmon->cname()); + SIMIX_process_kill(dmon, simix_global->maestro_process); + } + + if (xbt_dynar_is_empty(simix_global->process_to_run) && + !simix_global->process_list.empty()) + simgrid::simix::onDeadlock(); - } while (time != -1.0 || !xbt_dynar_is_empty(simix_global->process_to_run)); + } while (time > -1.0 || !xbt_dynar_is_empty(simix_global->process_to_run)); - if (xbt_swag_size(simix_global->process_list) != 0) { + if (simix_global->process_list.size() != 0) { - TRACE_end(); + TRACE_end(); XBT_CRITICAL("Oops ! Deadlock or code not perfectly clean."); SIMIX_display_process_status(); xbt_abort(); } + simgrid::s4u::onSimulationEnd(); } /** @@ -569,7 +576,7 @@ void SIMIX_run() * * Set the date to execute the function on the surf. * \param date Date to execute function - * \param function Function to be executed + * \param callback Function to be executed * \param arg Parameters of the function * */ @@ -613,13 +620,12 @@ void SIMIX_function_register_process_create(smx_creation_func_t function) /** * \brief Registers a function to kill a process. * - * This function registers a function to be called when a - * process is killed. The function has to call the SIMIX_process_kill(). + * This function registers a function to be called when a process is killed. The function has to call the + * SIMIX_process_kill(). * * \param function Kill process function */ -void SIMIX_function_register_process_kill(void_pfn_smxprocess_t - function) +void SIMIX_function_register_process_kill(void_pfn_smxprocess_t function) { simix_global->kill_process_function = function; } @@ -627,13 +633,11 @@ void SIMIX_function_register_process_kill(void_pfn_smxprocess_t /** * \brief Registers a function to cleanup a process. * - * This function registers a user function to be called when - * a process ends properly. + * This function registers a user function to be called when a process ends properly. * * \param function cleanup process function */ -void SIMIX_function_register_process_cleanup(void_pfn_smxprocess_t - function) +void SIMIX_function_register_process_cleanup(void_pfn_smxprocess_t function) { simix_global->cleanup_process_function = function; } @@ -641,18 +645,13 @@ void SIMIX_function_register_process_cleanup(void_pfn_smxprocess_t void SIMIX_display_process_status() { - if (simix_global->process_list == nullptr) { - return; - } - - smx_process_t process = nullptr; - int nbprocess = xbt_swag_size(simix_global->process_list); + int nbprocess = simix_global->process_list.size(); XBT_INFO("%d processes are still running, waiting for something.", nbprocess); /* List the process and their state */ - XBT_INFO - ("Legend of the following listing: \"Process (@): \""); - xbt_swag_foreach(process, simix_global->process_list) { + XBT_INFO("Legend of the following listing: \"Process (@): \""); + for (auto kv : simix_global->process_list) { + smx_actor_t process = kv.second; if (process->waiting_synchro) { @@ -685,26 +684,18 @@ void SIMIX_display_process_status() break; */ - XBT_INFO("Process %lu (%s@%s): waiting for %s synchro %p (%s) in state %d to finish", - process->pid, process->name.c_str(), sg_host_get_name(process->host), - synchro_description, process->waiting_synchro, - process->waiting_synchro->name.c_str(), (int)process->waiting_synchro->state); + XBT_INFO("Process %lu (%s@%s): waiting for %s synchro %p (%s) in state %d to finish", process->pid, + process->cname(), process->host->cname(), synchro_description, process->waiting_synchro, + process->waiting_synchro->name.c_str(), (int)process->waiting_synchro->state); } else { - XBT_INFO("Process %lu (%s@%s)", process->pid, process->name.c_str(), sg_host_get_name(process->host)); + XBT_INFO("Process %lu (%s@%s)", process->pid, process->cname(), process->host->cname()); } } } -xbt_dict_t simcall_HANDLER_asr_get_properties(smx_simcall_t simcall, const char *name){ - return SIMIX_asr_get_properties(name); -} -xbt_dict_t SIMIX_asr_get_properties(const char *name) -{ - return (xbt_dict_t) xbt_lib_get_or_null(as_router_lib, name, ROUTING_PROP_ASR_LEVEL); -} - int SIMIX_is_maestro() { - return simix_global==nullptr /*SimDag*/|| SIMIX_process_self() == simix_global->maestro_process; + smx_actor_t self = SIMIX_process_self(); + return simix_global == nullptr /*SimDag*/ || self == nullptr || self == simix_global->maestro_process; }