X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/65b31a3b58207bd824d798c37f1815f561bc0d54..d6fc96fb24bb0006584d44bee70df5f67b37f5fb:/src/simix/smx_host.c diff --git a/src/simix/smx_host.c b/src/simix/smx_host.c index d3be3ca9d7..c3c6546fb4 100644 --- a/src/simix/smx_host.c +++ b/src/simix/smx_host.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2007, 2008, 2009, 2010. The SimGrid Team. +/* Copyright (c) 2007-2012. The SimGrid Team. * All rights reserved. */ /* This program is free software; you can redistribute it and/or modify it @@ -13,7 +13,6 @@ XBT_LOG_NEW_DEFAULT_SUBCATEGORY(simix_host, simix, "Logging specific to SIMIX (hosts)"); - static void SIMIX_execution_finish(smx_action_t action); /** @@ -67,7 +66,7 @@ void SIMIX_host_destroy(void *h) SIMIX_display_process_status(); THROWF(arg_error, 0, "%s", msg); } - + xbt_dynar_free(&host->auto_restart_processes); xbt_swag_free(host->process_list); /* Clean host structure */ @@ -77,27 +76,28 @@ void SIMIX_host_destroy(void *h) return; } -/** - * \brief Returns a dict of all hosts. - * - * \return List of all hosts (as a #xbt_dict_t) - */ -xbt_dict_t SIMIX_host_get_dict(void) -{ - xbt_dict_t host_dict = xbt_dict_new_homogeneous(NULL); - xbt_lib_cursor_t cursor = NULL; - char *name = NULL; - void **host = NULL; - - xbt_lib_foreach(host_lib, cursor, name, host){ - if(host[SIMIX_HOST_LEVEL]) - xbt_dict_set(host_dict,name,host[SIMIX_HOST_LEVEL], NULL); - } - return host_dict; -} - -smx_host_t SIMIX_host_get_by_name(const char *name) +///** +// * \brief Returns a dict of all hosts. +// * +// * \return List of all hosts (as a #xbt_dict_t) +// */ +//xbt_dict_t SIMIX_host_get_dict(void) +//{ +// xbt_dict_t host_dict = xbt_dict_new_homogeneous(NULL); +// xbt_lib_cursor_t cursor = NULL; +// char *name = NULL; +// void **host = NULL; +// +// xbt_lib_foreach(host_lib, cursor, name, host){ +// if(host[SIMIX_HOST_LEVEL]) +// xbt_dict_set(host_dict,name,host[SIMIX_HOST_LEVEL], NULL); +// } +// return host_dict; +//} + +smx_host_t SIMIX_host_get_by_name(u_smx_scalar_t *args) { + const char *name = args[0].cc; xbt_assert(((simix_global != NULL) && (host_lib != NULL)), "Environment not set yet"); @@ -119,41 +119,46 @@ const char* SIMIX_host_self_get_name(void) if (host == NULL || SIMIX_process_self() == simix_global->maestro_process) return ""; - return SIMIX_host_get_name(host); + return SIMIX_host_get_name(SIMIX_pack_args(PTR(host))); } -const char* SIMIX_host_get_name(smx_host_t host) +const char* SIMIX_host_get_name(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters"); return host->name; } -xbt_dict_t SIMIX_host_get_properties(smx_host_t host) +xbt_dict_t SIMIX_host_get_properties(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters (simix host is NULL)"); return surf_workstation_model->extension.workstation.get_properties(host->host); } -double SIMIX_host_get_speed(smx_host_t host) +double SIMIX_host_get_speed(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters (simix host is NULL)"); return surf_workstation_model->extension.workstation. get_speed(host->host, 1.0); } -double SIMIX_host_get_available_speed(smx_host_t host) +double SIMIX_host_get_available_speed(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters (simix host is NULL)"); return surf_workstation_model->extension.workstation. get_available_speed(host->host); } -int SIMIX_host_get_state(smx_host_t host) +int SIMIX_host_get_state(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters (simix host is NULL)"); return surf_workstation_model->extension.workstation. @@ -162,33 +167,142 @@ int SIMIX_host_get_state(smx_host_t host) void* SIMIX_host_self_get_data(void) { - return SIMIX_host_get_data(SIMIX_host_self()); + smx_host_t self = SIMIX_host_self(); + return SIMIX_host_get_data(SIMIX_pack_args(PTR(self))); } void SIMIX_host_self_set_data(void *data) { - SIMIX_host_set_data(SIMIX_host_self(), data); + smx_host_t self = SIMIX_host_self(); + SIMIX_host_set_data(SIMIX_pack_args(PTR(self), PTR(data))); } -void* SIMIX_host_get_data(smx_host_t host) +void* SIMIX_host_get_data(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; xbt_assert((host != NULL), "Invalid parameters (simix host is NULL)"); return host->data; } +void _SIMIX_host_free_process_arg(void *); +void _SIMIX_host_free_process_arg(void *data) +{ + smx_process_arg_t arg = *(void**)data; + xbt_free(arg->name); + xbt_free(arg); +} +/** + * \brief Add a process to the list of the processes that the host will restart when it comes back + * This function add a process to the list of the processes that will be restarted when the host comes + * back. It is expected that this function is called when the host is down. + * The processes will only be restarted once, meaning that you will have to register the process + * again to restart the process again. + */ +void SIMIX_host_add_auto_restart_process(smx_host_t host, + const char *name, + xbt_main_func_t code, + void *data, + const char *hostname, + double kill_time, + int argc, char **argv, + xbt_dict_t properties, + int auto_restart) +{ + if (!host->auto_restart_processes) { + host->auto_restart_processes = xbt_dynar_new(sizeof(smx_process_arg_t),_SIMIX_host_free_process_arg); + } + smx_process_arg_t arg = xbt_new(s_smx_process_arg_t,1); + arg->name = xbt_strdup(name); + arg->code = code; + arg->data = data; + arg->hostname = hostname; + arg->kill_time = kill_time; + arg->argc = argc; + + arg->argv = xbt_new(char*,argc + 1); + + int i; + for (i = 0; i < argc; i++) { + arg->argv[i] = xbt_strdup(argv[i]); + } + arg->argv[argc] = NULL; + + arg->properties = properties; + arg->auto_restart = auto_restart; + + if( SIMIX_host_get_state(SIMIX_pack_args(PTR(host))) == SURF_RESOURCE_OFF + && !xbt_dict_get_or_null(watched_hosts_lib,host->name)){ + xbt_dict_set(watched_hosts_lib,host->name,host,NULL); + XBT_DEBUG("Have push host %s to watched_hosts_lib because state == SURF_RESOURCE_OFF",host->name); + } + xbt_dynar_push_as(host->auto_restart_processes,smx_process_arg_t,arg); +} +/** + * \brief Restart the list of processes that have been registered to the host + */ +void SIMIX_host_restart_processes(smx_host_t host) +{ + unsigned int cpt; + smx_process_arg_t arg; + xbt_dynar_foreach(host->auto_restart_processes,cpt,arg) { + + smx_process_t process; + + XBT_DEBUG("Restarting Process %s(%s) right now", arg->argv[0], arg->hostname); + if (simix_global->create_process_function) { + simix_global->create_process_function(&process, + arg->argv[0], + arg->code, + NULL, + arg->hostname, + arg->kill_time, + arg->argc, + arg->argv, + arg->properties, + arg->auto_restart); + } + else { + simcall_process_create(&process, + arg->argv[0], + arg->code, + NULL, + arg->hostname, + arg->kill_time, + arg->argc, + arg->argv, + arg->properties, + arg->auto_restart); + + } + } + xbt_dynar_reset(host->auto_restart_processes); +} + +void SIMIX_host_autorestart(smx_host_t host) +{ + if(simix_global->autorestart) + simix_global->autorestart(host); + else + xbt_die("No function for simix_global->autorestart"); +} -void SIMIX_host_set_data(smx_host_t host, void *data) +void SIMIX_host_set_data(u_smx_scalar_t *args) { + smx_host_t host = args[0].p; + void *data = args[1].p; xbt_assert((host != NULL), "Invalid parameters"); xbt_assert((host->data == NULL), "Data already set"); host->data = data; } -smx_action_t SIMIX_host_execute(const char *name, smx_host_t host, - double computation_amount, - double priority) +smx_action_t SIMIX_host_execute(u_smx_scalar_t args[]) { + const char *name = args[0].p; + smx_host_t host = args[1].p; + double computation_amount = args[2].d; + double priority = args[3].d; + /* alloc structures and initialize */ smx_action_t action = xbt_mallocator_get(simix_global->action_mallocator); action->type = SIMIX_ACTION_EXECUTE; @@ -201,10 +315,10 @@ smx_action_t SIMIX_host_execute(const char *name, smx_host_t host, #endif /* set surf's action */ - if (!MC_IS_ENABLED) { + if (!MC_is_active()) { action->execution.surf_exec = surf_workstation_model->extension.workstation.execute(host->host, - computation_amount); + computation_amount); surf_workstation_model->action_data_set(action->execution.surf_exec, action); surf_workstation_model->set_priority(action->execution.surf_exec, priority); } @@ -214,11 +328,15 @@ smx_action_t SIMIX_host_execute(const char *name, smx_host_t host, return action; } -smx_action_t SIMIX_host_parallel_execute( const char *name, - int host_nb, smx_host_t *host_list, - double *computation_amount, double *communication_amount, - double amount, double rate) +smx_action_t SIMIX_host_parallel_execute(u_smx_scalar_t *args) { + const char *name = args[0].cc; + int host_nb = args[1].i; + smx_host_t *host_list = args[2].p; + double *computation_amount = args[3].p; + double *communication_amount = args[4].p; + double amount = args[5].d; + double rate = args[6].d; void **workstation_list = NULL; int i; @@ -239,11 +357,11 @@ smx_action_t SIMIX_host_parallel_execute( const char *name, workstation_list[i] = host_list[i]->host; /* set surf's action */ - if (!MC_IS_ENABLED) { + if (!MC_is_active()) { action->execution.surf_exec = surf_workstation_model->extension.workstation. execute_parallel_task(host_nb, workstation_list, computation_amount, - communication_amount, amount, rate); + communication_amount, rate); surf_workstation_model->action_data_set(action->execution.surf_exec, action); } @@ -252,30 +370,31 @@ smx_action_t SIMIX_host_parallel_execute( const char *name, return action; } -void SIMIX_host_execution_destroy(smx_action_t action) +void SIMIX_host_execution_destroy(u_smx_scalar_t *args) { + smx_action_t action = args[0].p; XBT_DEBUG("Destroy action %p", action); - xbt_free(action->name); - if (action->execution.surf_exec) { surf_workstation_model->action_unref(action->execution.surf_exec); action->execution.surf_exec = NULL; } - + xbt_free(action->name); xbt_mallocator_release(simix_global->action_mallocator, action); } -void SIMIX_host_execution_cancel(smx_action_t action) +void SIMIX_host_execution_cancel(u_smx_scalar_t *args) { + smx_action_t action = args[0].p; XBT_DEBUG("Cancel action %p", action); if (action->execution.surf_exec) surf_workstation_model->action_cancel(action->execution.surf_exec); } -double SIMIX_host_execution_get_remains(smx_action_t action) +double SIMIX_host_execution_get_remains(u_smx_scalar_t *args) { + smx_action_t action = args[0].p; double result = 0.0; if (action->state == SIMIX_RUNNING) @@ -284,20 +403,24 @@ double SIMIX_host_execution_get_remains(smx_action_t action) return result; } -e_smx_state_t SIMIX_host_execution_get_state(smx_action_t action) +e_smx_state_t SIMIX_host_execution_get_state(u_smx_scalar_t *args) { + smx_action_t action = args[0].p; return action->state; } -void SIMIX_host_execution_set_priority(smx_action_t action, double priority) +void SIMIX_host_execution_set_priority(u_smx_scalar_t *args) { + smx_action_t action = args[0].p; + double priority = args[1].d; if(action->execution.surf_exec) surf_workstation_model->set_priority(action->execution.surf_exec, priority); } -void SIMIX_pre_host_execution_wait(smx_simcall_t simcall) +void SIMIX_pre_host_execution_wait(u_smx_scalar_t *args) { - smx_action_t action = simcall->host_execution_wait.execution; + smx_simcall_t simcall = args[0].p; + smx_action_t action = args[1].p; XBT_DEBUG("Wait for execution of action %p, state %d", action, (int)action->state); @@ -306,7 +429,7 @@ void SIMIX_pre_host_execution_wait(smx_simcall_t simcall) simcall->issuer->waiting_action = action; /* set surf's action */ - if (MC_IS_ENABLED) { + if (MC_is_active()) { action->state = SIMIX_DONE; SIMIX_execution_finish(action); return; @@ -340,12 +463,13 @@ void SIMIX_execution_finish(smx_action_t action) case SIMIX_DONE: /* do nothing, action done */ - XBT_DEBUG("SIMIX_execution_finished: execution successful"); + XBT_DEBUG("SIMIX_execution_finished: execution successful"); break; case SIMIX_FAILED: XBT_DEBUG("SIMIX_execution_finished: host '%s' failed", simcall->issuer->smx_host->name); - SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed"); + simcall->issuer->context->iwannadie = 1; + //SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed"); break; case SIMIX_CANCELED: @@ -357,25 +481,36 @@ void SIMIX_execution_finish(smx_action_t action) xbt_die("Internal error in SIMIX_execution_finish: unexpected action state %d", (int)action->state); } - simcall->issuer->waiting_action = NULL; + /* check if the host is down */ + if (surf_workstation_model->extension. + workstation.get_state(simcall->issuer->smx_host->host) != SURF_RESOURCE_ON) { + simcall->issuer->context->iwannadie = 1; + } + + simcall->issuer->waiting_action = NULL; simcall->host_execution_wait.result = action->state; SIMIX_simcall_answer(simcall); } /* We no longer need it */ - SIMIX_host_execution_destroy(action); + SIMIX_host_execution_destroy(SIMIX_pack_args(PTR(action))); } void SIMIX_post_host_execute(smx_action_t action) { - /* FIXME: check if the host running the action failed or not*/ - /*if(surf_workstation_model->extension.workstation.get_state(action->host->host))*/ - - /* If the host running the action didn't fail, then the action was canceled */ - if (surf_workstation_model->action_state_get(action->execution.surf_exec) == SURF_ACTION_FAILED) - action->state = SIMIX_CANCELED; - else - action->state = SIMIX_DONE; + if (action->type == SIMIX_ACTION_EXECUTE && /* FIMXE: handle resource failure + * for parallel tasks too */ + surf_workstation_model->extension.workstation.get_state(action->execution.host->host) == SURF_RESOURCE_OFF) { + /* If the host running the action failed, notice it so that the asking + * process can be killed if it runs on that host itself */ + action->state = SIMIX_FAILED; + } else if (surf_workstation_model->action_state_get(action->execution.surf_exec) == SURF_ACTION_FAILED) { + /* If the host running the action didn't fail, then the action was + * canceled */ + action->state = SIMIX_CANCELED; + } else { + action->state = SIMIX_DONE; + } if (action->execution.surf_exec) { surf_workstation_model->action_unref(action->execution.surf_exec);