X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/5c19d3d6403ded95f8675596aa250240b6f7e467..60087ab28ce5450ad32f128cffbb09eb492eb550:/src/simix/smx_process.cpp diff --git a/src/simix/smx_process.cpp b/src/simix/smx_process.cpp index 0bcf89e54e..82b4d162a4 100644 --- a/src/simix/smx_process.cpp +++ b/src/simix/smx_process.cpp @@ -11,16 +11,19 @@ #include "xbt/dict.h" #include "mc/mc.h" #include "src/mc/mc_replay.h" -#include "src/mc/mc_client.h" +#include "src/mc/Client.hpp" #include "src/simix/smx_private.hpp" #include "src/msg/msg_private.h" +#include "src/simix/SynchroSleep.hpp" +#include "src/simix/SynchroRaw.hpp" +#include "src/simix/SynchroIo.hpp" + #ifdef HAVE_SMPI #include "src/smpi/private.h" #endif -XBT_LOG_NEW_DEFAULT_SUBCATEGORY(simix_process, simix, - "Logging specific to SIMIX (process)"); +XBT_LOG_NEW_DEFAULT_SUBCATEGORY(simix_process, simix, "Logging specific to SIMIX (process)"); unsigned long simix_process_maxpid = 0; @@ -35,7 +38,7 @@ smx_process_t SIMIX_process_self(void) { smx_context_t self_context = SIMIX_context_self(); - return self_context ? SIMIX_context_get_process(self_context) : NULL; + return self_context ? self_context->process() : NULL; } /** @@ -66,44 +69,43 @@ void SIMIX_process_cleanup(smx_process_t process) /* cancel non-blocking communications */ smx_synchro_t synchro; while ((synchro = (smx_synchro_t) xbt_fifo_pop(process->comms))) { + simgrid::simix::Comm *comm = static_cast(synchro); /* make sure no one will finish the comm after this process is destroyed, * because src_proc or dst_proc would be an invalid pointer */ - SIMIX_comm_cancel(synchro); + SIMIX_comm_cancel(comm); - if (synchro->comm.src_proc == process) { + if (comm->src_proc == process) { XBT_DEBUG("Found an unfinished send comm %p (detached = %d), state %d, src = %p, dst = %p", - synchro, synchro->comm.detached, (int)synchro->state, synchro->comm.src_proc, synchro->comm.dst_proc); - synchro->comm.src_proc = NULL; + comm, comm->detached, (int)comm->state, comm->src_proc, comm->dst_proc); + comm->src_proc = NULL; /* I'm not supposed to destroy a detached comm from the sender side, */ - if (!synchro->comm.detached) - SIMIX_comm_destroy(synchro); + if (comm->detached) + XBT_DEBUG("Don't destroy it since it's a detached comm and I'm the sender"); else - XBT_DEBUG("Don't destroy it since it's a detached comm"); + SIMIX_comm_destroy(comm); } - else if (synchro->comm.dst_proc == process){ + else if (comm->dst_proc == process){ XBT_DEBUG("Found an unfinished recv comm %p, state %d, src = %p, dst = %p", - synchro, (int)synchro->state, synchro->comm.src_proc, synchro->comm.dst_proc); - synchro->comm.dst_proc = NULL; + comm, (int)comm->state, comm->src_proc, comm->dst_proc); + comm->dst_proc = NULL; - if (synchro->comm.detached && synchro->comm.refcount == 1 - && synchro->comm.src_proc != NULL) { + if (comm->detached && comm->refcount == 1 && comm->src_proc != NULL) { /* the comm will be freed right now, remove it from the sender */ - xbt_fifo_remove(synchro->comm.src_proc->comms, synchro); + xbt_fifo_remove(comm->src_proc->comms, comm); } - SIMIX_comm_destroy(synchro); - } - else { - xbt_die("Communication synchro %p is in my list but I'm not the sender " - "or the receiver", synchro); + SIMIX_comm_destroy(comm); + } else { + xbt_die("Communication synchro %p is in my list but I'm not the sender nor the receiver", synchro); } } XBT_DEBUG("%p should not be run anymore",process); xbt_swag_remove(process, simix_global->process_list); - xbt_swag_remove(process, sg_host_simix(process->host)->process_list); + if (process->host) + xbt_swag_remove(process, sg_host_simix(process->host)->process_list); xbt_swag_insert(process, simix_global->process_to_destroy); process->context->iwannadie = 0; @@ -138,25 +140,45 @@ void SIMIX_process_empty_trash(void) } } -/** - * \brief Creates and runs the maestro process - */ -void SIMIX_create_maestro_process() +namespace simgrid { +namespace simix { + +void create_maestro(std::function code) { smx_process_t maestro = NULL; - /* Create maestro process and intilialize it */ maestro = xbt_new0(s_smx_process_t, 1); maestro->pid = simix_process_maxpid++; maestro->ppid = -1; - maestro->name = (char *) ""; + maestro->name = (char*) ""; + maestro->data = nullptr; maestro->running_ctx = (xbt_running_ctx_t*) xbt_malloc0(sizeof(xbt_running_ctx_t)); XBT_RUNNING_CTX_INITIALIZE(maestro->running_ctx); - maestro->context = SIMIX_context_new(NULL, 0, NULL, NULL, maestro); + + if (!code) { + maestro->context = SIMIX_context_new(NULL, 0, nullptr, NULL, maestro); + } else { + if (!simix_global) + xbt_die("simix is not initialized, please call MSG_init first"); + maestro->context = + simix_global->context_factory->create_maestro(code, maestro); + } + maestro->simcall.issuer = maestro; simix_global->maestro_process = maestro; - return; } + +} +} + +/** + * \brief Creates and runs the maestro process + */ +void SIMIX_maestro_create(void (*code)(void*), void* data) +{ + simgrid::simix::create_maestro(std::bind(code, data)); +} + /** * \brief Stops a process. * @@ -168,7 +190,7 @@ void SIMIX_process_stop(smx_process_t arg) { /* execute the on_exit functions */ SIMIX_process_on_exit_runall(arg); /* Add the process to the list of process to restart, only if the host is down */ - if (arg->auto_restart && arg->host->is_off()) { + if (arg->auto_restart && arg->host->isOff()) { SIMIX_host_add_auto_restart_process(arg->host,arg->name,arg->code, arg->data, sg_host_get_name(arg->host), SIMIX_timer_get_date(arg->kill_timer), @@ -247,7 +269,7 @@ smx_process_t SIMIX_process_create( XBT_DEBUG("Start process %s on host '%s'", name, hostname); - if (host->is_off()) { + if (host->isOff()) { int i; XBT_WARN("Cannot launch process '%s' on failed host '%s'", name, hostname); @@ -273,7 +295,7 @@ smx_process_t SIMIX_process_create( process->ppid = SIMIX_process_get_PID(parent_process); /* SMPI process have their own data segment and each other inherit from their father */ -#ifdef HAVE_SMPI +#if HAVE_SMPI if(smpi_privatize_global_variables){ if( parent_process->pid != 0){ SIMIX_segment_index_set(process, parent_process->segment_index); @@ -328,6 +350,110 @@ smx_process_t SIMIX_process_create( return process; } +smx_process_t SIMIX_process_attach( + const char* name, + void *data, + const char* hostname, + xbt_dict_t properties, + smx_process_t parent_process) +{ + // This is mostly a copy/paste from SIMIX_process_new(), + // it'd be nice to share some code between those two functions. + + sg_host_t host = sg_host_by_name(hostname); + XBT_DEBUG("Attach process %s on host '%s'", name, hostname); + + if (host->isOff()) { + XBT_WARN("Cannot launch process '%s' on failed host '%s'", + name, hostname); + return nullptr; + } + + smx_process_t process = xbt_new0(s_smx_process_t, 1); + /* Process data */ + process->pid = simix_process_maxpid++; + process->name = xbt_strdup(name); + process->host = host; + process->data = data; + process->comms = xbt_fifo_new(); + process->simcall.issuer = process; + process->ppid = -1; + /* Initiliaze data segment to default value */ + SIMIX_segment_index_set(process, -1); + if (parent_process != NULL) { + process->ppid = SIMIX_process_get_PID(parent_process); + /* SMPI process have their own data segment and + each other inherit from their father */ + #if HAVE_SMPI + if(smpi_privatize_global_variables){ + if(parent_process->pid != 0){ + SIMIX_segment_index_set(process, parent_process->segment_index); + } else { + SIMIX_segment_index_set(process, process->pid - 1); + } + } + #endif + } + + /* Process data for auto-restart */ + process->auto_restart = false; + process->code = nullptr; + process->argc = 0; + process->argv = nullptr; + + XBT_VERB("Create context %s", process->name); + if (!simix_global) + xbt_die("simix is not initialized, please call MSG_init first"); + process->context = simix_global->context_factory->attach( + simix_global->cleanup_process_function, process); + + process->running_ctx = (xbt_running_ctx_t*) xbt_malloc0(sizeof(xbt_running_ctx_t)); + XBT_RUNNING_CTX_INITIALIZE(process->running_ctx); + + if(MC_is_active()){ + MC_ignore_heap(process->running_ctx, sizeof(*process->running_ctx)); + } + + /* Add properties */ + process->properties = properties; + + /* Add the process to it's host process list */ + xbt_swag_insert(process, sg_host_simix(host)->process_list); + + /* Now insert it in the global process list and in the process to run list */ + xbt_swag_insert(process, simix_global->process_list); + XBT_DEBUG("Inserting %s(%s) in the to_run list", process->name, sg_host_get_name(host)); + xbt_dynar_push_as(simix_global->process_to_run, smx_process_t, process); + + /* Tracing the process creation */ + TRACE_msg_process_create(process->name, process->pid, process->host); + + auto context = dynamic_cast(process->context); + if (!context) + xbt_die("Not a suitable context"); + + context->attach_start(); + return process; +} + +void SIMIX_process_detach(void) +{ + auto context = dynamic_cast(SIMIX_context_self()); + if (!context) + xbt_die("Not a suitable context"); + + simix_global->cleanup_process_function(context->process()); + + // Let maestro ignore we are still alive: + // xbt_swag_remove(context->process(), simix_global->process_list); + + // TODDO, Remove from proces list: + // xbt_swag_remove(process, sg_host_simix(host)->process_list); + + context->attach_stop(); + // delete context; +} + /** * \brief Executes the processes from simix_global->process_to_run. * @@ -371,39 +497,39 @@ void SIMIX_process_kill(smx_process_t process, smx_process_t issuer) { /* destroy the blocking synchro if any */ if (process->waiting_synchro) { - switch (process->waiting_synchro->type) { + simgrid::simix::Exec *exec = dynamic_cast(process->waiting_synchro); + simgrid::simix::Comm *comm = dynamic_cast(process->waiting_synchro); + simgrid::simix::Sleep *sleep = dynamic_cast(process->waiting_synchro); + simgrid::simix::Raw *raw = dynamic_cast(process->waiting_synchro); + simgrid::simix::Io *io = dynamic_cast(process->waiting_synchro); - case SIMIX_SYNC_EXECUTE: - case SIMIX_SYNC_PARALLEL_EXECUTE: - SIMIX_process_execution_destroy(process->waiting_synchro); - break; + if (exec != nullptr) { + SIMIX_execution_destroy(process->waiting_synchro); - case SIMIX_SYNC_COMMUNICATE: + } else if (comm != nullptr) { xbt_fifo_remove(process->comms, process->waiting_synchro); SIMIX_comm_cancel(process->waiting_synchro); xbt_fifo_remove(process->waiting_synchro->simcalls, &process->simcall); SIMIX_comm_destroy(process->waiting_synchro); - break; - - case SIMIX_SYNC_SLEEP: - SIMIX_process_sleep_destroy(process->waiting_synchro); - break; - case SIMIX_SYNC_JOIN: + } else if (sleep != nullptr) { SIMIX_process_sleep_destroy(process->waiting_synchro); - break; - case SIMIX_SYNC_SYNCHRO: + } else if (raw != nullptr) { SIMIX_synchro_stop_waiting(process, &process->simcall); SIMIX_synchro_destroy(process->waiting_synchro); - break; - case SIMIX_SYNC_IO: + } else if (io != nullptr) { SIMIX_io_destroy(process->waiting_synchro); - break; - } + /* + switch (process->waiting_synchro->type) { + case SIMIX_SYNC_JOIN: + SIMIX_process_sleep_destroy(process->waiting_synchro); + break; + } */ + process->waiting_synchro = NULL; } if(!xbt_dynar_member(simix_global->process_to_run, &(process)) && process != issuer) { @@ -428,35 +554,34 @@ void SIMIX_process_throw(smx_process_t process, xbt_errcat_t cat, int value, con /* cancel the blocking synchro if any */ if (process->waiting_synchro) { - switch (process->waiting_synchro->type) { - - case SIMIX_SYNC_EXECUTE: - case SIMIX_SYNC_PARALLEL_EXECUTE: - SIMIX_process_execution_cancel(process->waiting_synchro); - break; + simgrid::simix::Exec *exec = dynamic_cast(process->waiting_synchro); + if (exec != nullptr) { + SIMIX_execution_cancel(process->waiting_synchro); + } - case SIMIX_SYNC_COMMUNICATE: + simgrid::simix::Comm *comm = dynamic_cast(process->waiting_synchro); + if (comm != nullptr) { xbt_fifo_remove(process->comms, process->waiting_synchro); SIMIX_comm_cancel(process->waiting_synchro); - break; + } - case SIMIX_SYNC_SLEEP: - case SIMIX_SYNC_JOIN: + simgrid::simix::Sleep *sleep = dynamic_cast(process->waiting_synchro); + if (sleep != nullptr) { SIMIX_process_sleep_destroy(process->waiting_synchro); if (!xbt_dynar_member(simix_global->process_to_run, &(process)) && process != SIMIX_process_self()) { XBT_DEBUG("Inserting %s in the to_run list", process->name); xbt_dynar_push_as(simix_global->process_to_run, smx_process_t, process); } - break; + } - case SIMIX_SYNC_SYNCHRO: + simgrid::simix::Raw *raw = dynamic_cast(process->waiting_synchro); + if (raw != nullptr) { SIMIX_synchro_stop_waiting(process, &process->simcall); - break; + } - case SIMIX_SYNC_IO: + simgrid::simix::Io *io = dynamic_cast(process->waiting_synchro); + if (io != nullptr) { SIMIX_io_destroy(process->waiting_synchro); - break; - } } process->waiting_synchro = NULL; @@ -512,15 +637,13 @@ void simcall_HANDLER_process_suspend(smx_simcall_t simcall, smx_process_t proces } else { xbt_fifo_push(sync_suspend->simcalls, simcall); process->waiting_synchro = sync_suspend; - SIMIX_host_execution_suspend(process->waiting_synchro); + SIMIX_execution_suspend(process->waiting_synchro); } /* If we are suspending ourselves, then just do not finish the simcall now */ } smx_synchro_t SIMIX_process_suspend(smx_process_t process, smx_process_t issuer) { - xbt_assert((process != NULL), "Invalid parameters"); - if (process->suspended) { XBT_DEBUG("Process '%s' is already suspended", process->name); return NULL; @@ -528,43 +651,17 @@ smx_synchro_t SIMIX_process_suspend(smx_process_t process, smx_process_t issuer) process->suspended = 1; - /* If we are suspending another process, and it is waiting on a sync, - suspend its synchronization. */ + /* If we are suspending another process that is waiting on a sync, suspend its synchronization. */ if (process != issuer) { - if (process->waiting_synchro) { - - switch (process->waiting_synchro->type) { - - case SIMIX_SYNC_EXECUTE: - case SIMIX_SYNC_PARALLEL_EXECUTE: - SIMIX_host_execution_suspend(process->waiting_synchro); - break; - - case SIMIX_SYNC_COMMUNICATE: - SIMIX_comm_suspend(process->waiting_synchro); - break; + if (process->waiting_synchro) + process->waiting_synchro->suspend(); + /* If the other process is not waiting, its suspension is delayed to when the process is rescheduled. */ - case SIMIX_SYNC_SLEEP: - SIMIX_process_sleep_suspend(process->waiting_synchro); - break; - - case SIMIX_SYNC_SYNCHRO: - /* Suspension is delayed to when the process is rescheduled. */ - break; - - default: - xbt_die("Internal error in SIMIX_process_suspend: unexpected synchronization type %d", - (int)process->waiting_synchro->type); - } - return NULL; - } else { - /* Suspension is delayed to when the process is rescheduled. */ - return NULL; - } + return NULL; } else { /* FIXME: computation size is zero. Is it okay that bound is zero ? */ - return SIMIX_process_execute(process, "suspend", 0.0, 1.0, 0.0, 0); + return SIMIX_execution_start(process, "suspend", 0.0, 1.0, 0.0, 0); } } @@ -589,31 +686,7 @@ void SIMIX_process_resume(smx_process_t process, smx_process_t issuer) if (process != issuer) { if (process->waiting_synchro) { - - switch (process->waiting_synchro->type) { - - case SIMIX_SYNC_EXECUTE: - case SIMIX_SYNC_PARALLEL_EXECUTE: - SIMIX_host_execution_resume(process->waiting_synchro); - break; - - case SIMIX_SYNC_COMMUNICATE: - SIMIX_comm_resume(process->waiting_synchro); - break; - - case SIMIX_SYNC_SLEEP: - SIMIX_process_sleep_resume(process->waiting_synchro); - break; - - case SIMIX_SYNC_SYNCHRO: - /* I cannot resume it now. This is delayed to when the process is rescheduled at - * the end of the synchro. */ - break; - - default: - xbt_die("Internal error in SIMIX_process_resume: unexpected synchronization type %d", - (int)process->waiting_synchro->type); - } + process->waiting_synchro->resume(); } } else XBT_WARN("Strange. Process %p is trying to resume himself.", issuer); @@ -643,9 +716,9 @@ int SIMIX_process_get_PPID(smx_process_t self){ return self->ppid; } -void* SIMIX_process_self_get_data(smx_process_t self) +void* SIMIX_process_self_get_data() { - xbt_assert(self == SIMIX_process_self(), "This is not the current process"); + smx_process_t self = SIMIX_process_self(); if (!self) { return NULL; @@ -653,9 +726,9 @@ void* SIMIX_process_self_get_data(smx_process_t self) return SIMIX_process_get_data(self); } -void SIMIX_process_self_set_data(smx_process_t self, void *data) +void SIMIX_process_self_set_data(void *data) { - xbt_assert(self == SIMIX_process_self(), "This is not the current process"); + smx_process_t self = SIMIX_process_self(); SIMIX_process_set_data(self, data); } @@ -724,12 +797,14 @@ void simcall_HANDLER_process_join(smx_simcall_t simcall, smx_process_t process, simcall->issuer->waiting_synchro = sync; } -static int SIMIX_process_join_finish(smx_process_exit_status_t status, smx_synchro_t sync){ - if (sync->sleep.surf_sleep) { - sync->sleep.surf_sleep->cancel(); +static int SIMIX_process_join_finish(smx_process_exit_status_t status, smx_synchro_t synchro){ + simgrid::simix::Sleep *sleep = static_cast(synchro); + + if (sleep->surf_sleep) { + sleep->surf_sleep->cancel(); smx_simcall_t simcall; - while ((simcall = (smx_simcall_t) xbt_fifo_shift(sync->simcalls))) { + while ((simcall = (smx_simcall_t) xbt_fifo_shift(sleep->simcalls))) { simcall_process_sleep__set__result(simcall, SIMIX_DONE); simcall->issuer->waiting_synchro = NULL; if (simcall->issuer->suspended) { @@ -740,17 +815,16 @@ static int SIMIX_process_join_finish(smx_process_exit_status_t status, smx_synch SIMIX_simcall_answer(simcall); } } - sync->sleep.surf_sleep->unref(); - sync->sleep.surf_sleep = NULL; + sleep->surf_sleep->unref(); + sleep->surf_sleep = NULL; } - xbt_mallocator_release(simix_global->synchro_mallocator, sync); + delete sleep; return 0; } smx_synchro_t SIMIX_process_join(smx_process_t issuer, smx_process_t process, double timeout) { smx_synchro_t res = SIMIX_process_sleep(issuer, timeout); - res->type = SIMIX_SYNC_JOIN; SIMIX_process_on_exit(process, (int_f_pvoid_pvoid_t)SIMIX_process_join_finish, res); return res; } @@ -773,20 +847,15 @@ smx_synchro_t SIMIX_process_sleep(smx_process_t process, double duration) sg_host_t host = process->host; /* check if the host is active */ - if (host->is_off()) { - THROWF(host_error, 0, "Host %s failed, you cannot call this function", - sg_host_get_name(host)); - } + if (host->isOff()) + THROWF(host_error, 0, "Host %s failed, you cannot call this function", sg_host_get_name(host)); - smx_synchro_t synchro = (smx_synchro_t) xbt_mallocator_get(simix_global->synchro_mallocator); - synchro->type = SIMIX_SYNC_SLEEP; + simgrid::simix::Sleep *synchro = new simgrid::simix::Sleep(); synchro->name = NULL; - synchro->category = NULL; - - synchro->sleep.host = host; - synchro->sleep.surf_sleep = surf_host_sleep(host, duration); - synchro->sleep.surf_sleep->setData(synchro); + synchro->host = host; + synchro->surf_sleep = surf_host_sleep(host, duration); + synchro->surf_sleep->setData(synchro); XBT_DEBUG("Create sleep synchronization %p", synchro); return synchro; @@ -796,18 +865,18 @@ void SIMIX_post_process_sleep(smx_synchro_t synchro) { smx_simcall_t simcall; e_smx_state_t state; - xbt_assert(synchro->type == SIMIX_SYNC_SLEEP || synchro->type == SIMIX_SYNC_JOIN); + simgrid::simix::Sleep *sleep = static_cast(synchro); while ((simcall = (smx_simcall_t) xbt_fifo_shift(synchro->simcalls))) { - switch (synchro->sleep.surf_sleep->getState()){ - case SURF_ACTION_FAILED: + switch (sleep->surf_sleep->getState()){ + case simgrid::surf::Action::State::failed: simcall->issuer->context->iwannadie = 1; //SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed"); state = SIMIX_SRC_HOST_FAILURE; break; - case SURF_ACTION_DONE: + case simgrid::surf::Action::State::done: state = SIMIX_DONE; break; @@ -815,7 +884,7 @@ void SIMIX_post_process_sleep(smx_synchro_t synchro) THROW_IMPOSSIBLE; break; } - if (simcall->issuer->host->is_off()) { + if (simcall->issuer->host->isOff()) { simcall->issuer->context->iwannadie = 1; } simcall_process_sleep__set__result(simcall, state); @@ -835,27 +904,12 @@ void SIMIX_post_process_sleep(smx_synchro_t synchro) void SIMIX_process_sleep_destroy(smx_synchro_t synchro) { XBT_DEBUG("Destroy synchro %p", synchro); - xbt_assert(synchro->type == SIMIX_SYNC_SLEEP || synchro->type == SIMIX_SYNC_JOIN); + simgrid::simix::Sleep *sleep = static_cast(synchro); - if (synchro->sleep.surf_sleep) { - synchro->sleep.surf_sleep->unref(); - synchro->sleep.surf_sleep = NULL; + if (sleep->surf_sleep) { + sleep->surf_sleep->unref(); + sleep->surf_sleep = NULL; } - if (synchro->type == SIMIX_SYNC_SLEEP) - xbt_mallocator_release(simix_global->synchro_mallocator, synchro); -} - -void SIMIX_process_sleep_suspend(smx_synchro_t synchro) -{ - xbt_assert(synchro->type == SIMIX_SYNC_SLEEP); - synchro->sleep.surf_sleep->suspend(); -} - -void SIMIX_process_sleep_resume(smx_synchro_t synchro) -{ - XBT_DEBUG("Synchro state is %d on process_sleep_resume.", synchro->state); - xbt_assert(synchro->type == SIMIX_SYNC_SLEEP); - synchro->sleep.surf_sleep->resume(); } /** @@ -896,7 +950,7 @@ void SIMIX_process_yield(smx_process_t self) if (self->doexception) { XBT_DEBUG("Wait, maestro left me an exception"); self->doexception = 0; - SMX_THROW(); + RETHROW; } if(SMPI_switch_data_segment && self->segment_index != -1){ @@ -907,7 +961,11 @@ void SIMIX_process_yield(smx_process_t self) /* callback: context fetching */ xbt_running_ctx_t *SIMIX_process_get_running_context(void) { - return SIMIX_process_self()->running_ctx; + smx_process_t process = SIMIX_process_self(); + if (process) + return process->running_ctx; + else + return nullptr; } /* callback: termination */