/* * Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana * University Research and Technology * Corporation. All rights reserved. * Copyright (c) 2004-2008 The University of Tennessee and The University * of Tennessee Research Foundation. All rights * reserved. * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, * University of Stuttgart. All rights reserved. * Copyright (c) 2004-2005 The Regents of the University of California. * All rights reserved. * Copyright (c) 2007-2010 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2007 Evergrid, Inc. All rights reserved. * Copyright (c) 2008-2020 Cisco Systems, Inc. All rights reserved * Copyright (c) 2010 IBM Corporation. All rights reserved. * Copyright (c) 2011-2014 Los Alamos National Security, LLC. All rights * reserved. * Copyright (c) 2013-2020 Intel, Inc. All rights reserved. * Copyright (c) 2017 Rutgers, The State University of New Jersey. * All rights reserved. * Copyright (c) 2017 Research Organization for Information Science * and Technology (RIST). All rights reserved. * * Copyright (c) 2021-2022 Nanook Consulting. All rights reserved. * $COPYRIGHT$ * * Additional copyrights may follow * * $HEADER$ */ /* * There is a complicated sequence of events that occurs when the * parent forks a child process that is intended to launch the target * executable. * * Before the child process exec's the target executable, it might tri * to set the affinity of that new child process according to a * complex series of rules. This binding may fail in a myriad of * different ways. A lot of this code deals with reporting that error * occurately to the end user. This is a complex task in itself * because the child process is not "really" an PRTE process -- all * error reporting must be proxied up to the parent who can use normal * PRTE error reporting mechanisms. * * Here's a high-level description of what is occurring in this file: * * - parent opens a pipe * - parent forks a child * - parent blocks reading on the pipe: the pipe will either close * (indicating that the child successfully exec'ed) or the child will * write some proxied error data up the pipe * * - the child tries to set affinity and do other housekeeping in * preparation of exec'ing the target executable * - if the child fails anywhere along the way, it sends a message up * the pipe to the parent indicating what happened -- including a * rendered error message detailing the problem (i.e., human-readable). * - it is important that the child renders the error message: there * are so many errors that are possible that the child is really the * only entity that has enough information to make an accuate error string * to report back to the user. * - the parent reads this message + rendered string in and uses PRTE * reporting mechanisms to display it to the user * - if the problem was only a warning, the child continues processing * (potentially eventually exec'ing the target executable). * - if the problem was an error, the child exits and the parent * handles the death of the child as appropriate (i.e., this ODLS * simply reports the error -- other things decide what to do). */ #include "prte_config.h" #include "constants.h" #include "types.h" #include #include #ifdef HAVE_UNISTD_H # include #endif #include #ifdef HAVE_SYS_TYPES_H # include #endif #ifdef HAVE_SYS_WAIT_H # include #endif #include #ifdef HAVE_FCNTL_H # include #endif #ifdef HAVE_SYS_TIME_H # include #endif #ifdef HAVE_SYS_PARAM_H # include #endif #ifdef HAVE_NETDB_H # include #endif #include #ifdef HAVE_SYS_STAT_H # include #endif /* HAVE_SYS_STAT_H */ #include #ifdef HAVE_SYS_SELECT_H # include #endif #ifdef HAVE_DIRENT_H # include #endif #include "src/class/pmix_pointer_array.h" #include "src/hwloc/hwloc-internal.h" #include "src/pmix/pmix-internal.h" #include "src/util/pmix_fd.h" #include "src/util/pmix_environ.h" #include "src/util/pmix_show_help.h" #include "src/util/sys_limits.h" #include "src/mca/errmgr/errmgr.h" #include "src/mca/ess/ess.h" #include "src/mca/iof/base/iof_base_setup.h" #include "src/mca/plm/plm.h" #include "src/mca/rtc/rtc.h" #include "src/runtime/prte_globals.h" #include "src/runtime/prte_wait.h" #include "src/util/name_fns.h" #include "src/util/pmix_show_help.h" #include "src/mca/odls/alps/odls_alps.h" #include "src/mca/odls/base/base.h" #include "src/prted/pmix/pmix_server.h" /* * Module functions (function pointers used in a struct) */ static int prte_odls_alps_launch_local_procs(pmix_data_buffer_t *data); static int prte_odls_alps_kill_local_procs(pmix_pointer_array_t *procs); static int prte_odls_alps_signal_local_procs(const pmix_proc_t *proc, int32_t signal); static int prte_odls_alps_restart_proc(prte_proc_t *child); /* * Explicitly declared functions so that we can get the noreturn * attribute registered with the compiler. */ static void send_error_show_help(int fd, int exit_status, const char *file, const char *topic, ...) __prte_attribute_noreturn__; static int do_child(prte_odls_spawn_caddy_t *cd, int write_fd) __prte_attribute_noreturn__; /* * Module */ prte_odls_base_module_t prte_odls_alps_module = {prte_odls_base_default_get_add_procs_data, prte_odls_alps_launch_local_procs, prte_odls_alps_kill_local_procs, prte_odls_alps_signal_local_procs, prte_odls_alps_restart_proc}; static int odls_alps_kill_local(pid_t pid, int signum) { pid_t pgrp; #if HAVE_SETPGID pgrp = getpgid(pid); if (-1 != pgrp) { /* target the lead process of the process * group so we ensure that the signal is * seen by all members of that group. This * ensures that the signal is seen by any * child processes our child may have * started */ pid = pgrp; } #endif if (0 != kill(pid, signum)) { if (ESRCH != errno) { PMIX_OUTPUT_VERBOSE((2, prte_odls_base_framework.framework_output, "%s odls:alps:SENT KILL %d TO PID %d GOT ERRNO %d", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), signum, (int) pid, errno)); return errno; } } PMIX_OUTPUT_VERBOSE((2, prte_odls_base_framework.framework_output, "%s odls:alps:SENT KILL %d TO PID %d SUCCESS", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), signum, (int) pid)); return 0; } int prte_odls_alps_kill_local_procs(pmix_pointer_array_t *procs) { int rc; if (PRTE_SUCCESS != (rc = prte_odls_base_default_kill_local_procs(procs, odls_alps_kill_local))) { PRTE_ERROR_LOG(rc); return rc; } return PRTE_SUCCESS; } static void set_handler_alps(int sig) { struct sigaction act; act.sa_handler = SIG_DFL; act.sa_flags = 0; sigemptyset(&act.sa_mask); sigaction(sig, &act, (struct sigaction *) 0); } /* * Internal function to write a rendered show_help message back up the * pipe to the waiting parent. */ static int write_help_msg(int fd, prte_odls_pipe_err_msg_t *msg, const char *file, const char *topic, va_list ap) { int ret; char *str; if (NULL == file || NULL == topic) { return PRTE_ERR_BAD_PARAM; } str = pmix_show_help_vstring(file, topic, true, ap); msg->file_str_len = (int) strlen(file); if (msg->file_str_len > PRTE_ODLS_MAX_FILE_LEN) { PRTE_ERROR_LOG(PRTE_ERR_BAD_PARAM); return PRTE_ERR_BAD_PARAM; } msg->topic_str_len = (int) strlen(topic); if (msg->topic_str_len > PRTE_ODLS_MAX_TOPIC_LEN) { PRTE_ERROR_LOG(PRTE_ERR_BAD_PARAM); return PRTE_ERR_BAD_PARAM; } msg->msg_str_len = (int) strlen(str); /* Only keep writing if each write() succeeds */ if (PRTE_SUCCESS != (ret = pmix_fd_write(fd, sizeof(*msg), msg))) { goto out; } if (msg->file_str_len > 0 && PRTE_SUCCESS != (ret = pmix_fd_write(fd, msg->file_str_len, file))) { goto out; } if (msg->topic_str_len > 0 && PRTE_SUCCESS != (ret = pmix_fd_write(fd, msg->topic_str_len, topic))) { goto out; } if (msg->msg_str_len > 0 && PRTE_SUCCESS != (ret = pmix_fd_write(fd, msg->msg_str_len, str))) { goto out; } out: free(str); return ret; } /* Called from the child to send an error message up the pipe to the waiting parent. */ static void send_error_show_help(int fd, int exit_status, const char *file, const char *topic, ...) { va_list ap; prte_odls_pipe_err_msg_t msg; msg.fatal = true; msg.exit_status = exit_status; /* Send it */ va_start(ap, topic); write_help_msg(fd, &msg, file, topic, ap); va_end(ap); exit(exit_status); } static int close_open_file_descriptors(int write_fd, prte_iof_base_io_conf_t opts) { int rc, fd; DIR *dir = NULL; struct dirent *files; int app_alps_filedes[2], alps_app_filedes[2]; dir = opendir("/proc/self/fd"); if (NULL == dir) { return PRTE_ERR_FILE_OPEN_FAILURE; } /* grab the fd of the opendir above so we don't close in the * middle of the scan. */ int dir_scan_fd = dirfd(dir); if (dir_scan_fd < 0) { return PRTE_ERR_FILE_OPEN_FAILURE; } /* close all file descriptors w/ exception of stdin/stdout/stderr, the pipe used for the IOF INTERNAL messages, and the pipe up to the parent. Be careful to retain all of the pipe fd's set up by the apshephered. These are needed for obtaining RDMA credentials, synchronizing with aprun, etc. */ rc = alps_app_lli_pipes(app_alps_filedes, alps_app_filedes); if (0 != rc) { closedir(dir); return PRTE_ERR_FILE_OPEN_FAILURE; } while ((files = readdir(dir)) != NULL) { if (!strncmp(files->d_name, ".", 1) || !strncmp(files->d_name, "..", 2)) continue; fd = strtoul(files->d_name, NULL, 10); if (EINVAL == errno || ERANGE == errno) { closedir(dir); return PRTE_ERR_TYPE_MISMATCH; } /* * skip over the pipes we have open to apshepherd or slurmd */ if (fd == XTAPI_FD_IDENTITY) continue; if (fd == XTAPI_FD_RESILIENCY) continue; if ((fd == app_alps_filedes[0]) || (fd == app_alps_filedes[1]) || (fd == alps_app_filedes[0]) || (fd == alps_app_filedes[1])) continue; if (fd >= 3 && fd != write_fd && fd != dir_scan_fd) { close(fd); } } closedir(dir); return PRTE_SUCCESS; } static int do_child(prte_odls_spawn_caddy_t *cd, int write_fd) { int i; sigset_t sigs; /* Setup the pipe to be close-on-exec */ pmix_fd_set_cloexec(write_fd); if (NULL != cd->child) { /* setup stdout/stderr so that any error messages that we may print out will get displayed back at prun. NOTE: Definitely do this AFTER we check contexts so that any error message from those two functions doesn't come out to the user. IF we didn't do it in this order, THEN a user who gives us a bad executable name or working directory would get N error messages, where N=num_procs. This would be very annoying for large jobs, so instead we set things up so that prun always outputs a nice, single message indicating what happened */ if (PRTE_SUCCESS != (i = prte_iof_base_setup_child(&cd->opts, &cd->env))) { PRTE_ERROR_LOG(i); send_error_show_help(write_fd, 1, "help-prte-odls-alps.txt", "iof setup failed", prte_process_info.nodename, cd->app->app); /* Does not return */ } /* now set any child-level controls such as binding */ prte_rtc.set(cd, write_fd); } else if (!PRTE_FLAG_TEST(cd->jdata, PRTE_JOB_FLAG_FORWARD_OUTPUT)) { /* tie stdin/out/err/internal to /dev/null */ int fdnull; for (i = 0; i < 3; i++) { fdnull = open("/dev/null", O_RDONLY, 0); if (fdnull > i && i != write_fd) { dup2(fdnull, i); } close(fdnull); } } if (PRTE_SUCCESS != close_open_file_descriptors(write_fd, cd->opts)) { send_error_show_help(write_fd, 1, "help-prte-odls-alps.txt", "close fds", prte_process_info.nodename, cd->app->app, __FILE__, __LINE__); } if (cd->argv == NULL) { cd->argv = malloc(sizeof(char *) * 2); cd->argv[0] = strdup(cd->app->app); cd->argv[1] = NULL; } /* Set signal handlers back to the default. Do this close to the exev() because the event library may (and likely will) reset them. If we don't do this, the event library may have left some set that, at least on some OS's, don't get reset via fork() or exec(). Hence, the launched process could be unkillable (for example). */ set_handler_alps(SIGTERM); set_handler_alps(SIGINT); set_handler_alps(SIGHUP); set_handler_alps(SIGPIPE); set_handler_alps(SIGCHLD); /* Unblock all signals, for many of the same reasons that we set the default handlers, above. This is noticable on Linux where the event library blocks SIGTERM, but we don't want that blocked by the launched process. */ sigprocmask(0, 0, &sigs); sigprocmask(SIG_UNBLOCK, &sigs, 0); /* take us to the correct wdir */ if (NULL != cd->wdir) { if (0 != chdir(cd->wdir)) { send_error_show_help(write_fd, 1, "help-prun.txt", "prun:wdir-not-found", "orted", cd->wdir, prte_process_info.nodename, (NULL == cd->child) ? 0 : cd->child->app_rank); /* Does not return */ } } /* Exec the new executable */ if (10 < pmix_output_get_verbosity(prte_odls_base_framework.framework_output)) { int jout; pmix_output(0, "%s STARTING %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), cd->app->app); for (jout = 0; NULL != cd->argv[jout]; jout++) { pmix_output(0, "%s\tARGV[%d]: %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), jout, cd->argv[jout]); } for (jout = 0; NULL != cd->env[jout]; jout++) { pmix_output(0, "%s\tENVIRON[%d]: %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), jout, cd->env[jout]); } } execve(cd->cmd, cd->argv, cd->env); send_error_show_help(write_fd, 1, "help-prte-odls-alps.txt", "execve error", prte_process_info.nodename, cd->app->app, strerror(errno)); /* Does not return */ } static int do_parent(prte_odls_spawn_caddy_t *cd, int read_fd) { int rc; prte_odls_pipe_err_msg_t msg; char file[PRTE_ODLS_MAX_FILE_LEN + 1], topic[PRTE_ODLS_MAX_TOPIC_LEN + 1], *str = NULL; if (cd->opts.connect_stdin) { close(cd->opts.p_stdin[0]); } close(cd->opts.p_stdout[1]); close(cd->opts.p_stderr[1]); /* Block reading a message from the pipe */ while (1) { rc = pmix_fd_read(read_fd, sizeof(msg), &msg); /* If the pipe closed, then the child successfully launched */ if (PMIX_ERR_TIMEOUT == rc) { rc = prte_pmix_convert_status(rc); break; } /* If Something Bad happened in the read, error out */ if (PMIX_SUCCESS != rc) { PMIX_ERROR_LOG(rc); close(read_fd); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_UNDEF; } rc = prte_pmix_convert_status(rc); return rc; } /* Otherwise, we got a warning or error message from the child */ if (NULL != cd->child) { if (msg.fatal) { PRTE_FLAG_UNSET(cd->child, PRTE_PROC_FLAG_ALIVE); } else { PRTE_FLAG_SET(cd->child, PRTE_PROC_FLAG_ALIVE); } } /* Read in the strings; ensure to terminate them with \0 */ if (msg.file_str_len > 0) { rc = pmix_fd_read(read_fd, msg.file_str_len, file); if (PMIX_SUCCESS != rc) { pmix_show_help("help-prte-odls-alps.txt", "syscall fail", true, prte_process_info.nodename, cd->app, "pmix_fd_read", __FILE__, __LINE__); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_UNDEF; } rc = prte_pmix_convert_status(rc); return rc; } file[msg.file_str_len] = '\0'; } if (msg.topic_str_len > 0) { rc = pmix_fd_read(read_fd, msg.topic_str_len, topic); if (PMIX_SUCCESS != rc) { pmix_show_help("help-prte-odls-alps.txt", "syscall fail", true, prte_process_info.nodename, cd->app, "pmix_fd_read", __FILE__, __LINE__); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_UNDEF; } rc = prte_pmix_convert_status(rc); return rc; } topic[msg.topic_str_len] = '\0'; } if (msg.msg_str_len > 0) { str = calloc(1, msg.msg_str_len + 1); if (NULL == str) { pmix_show_help("help-prte-odls-alps.txt", "syscall fail", true, prte_process_info.nodename, cd->app, "pmix_fd_read", __FILE__, __LINE__); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_UNDEF; } return rc; } rc = pmix_fd_read(read_fd, msg.msg_str_len, str); } /* Print out what we got. We already have a rendered string, so use pmix_show_help_norender(). */ if (msg.msg_str_len > 0) { pmix_show_help_norender(file, topic, str); free(str); str = NULL; } /* If msg.fatal is true, then the child exited with an error. Otherwise, whatever we just printed was a warning, so loop around and see what else is on the pipe (or if the pipe closed, indicating that the child launched successfully). */ if (msg.fatal) { if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_FAILED_TO_START; PRTE_FLAG_UNSET(cd->child, PRTE_PROC_FLAG_ALIVE); } close(read_fd); return PRTE_ERR_FAILED_TO_START; } } /* If we got here, it means that the pipe closed without indication of a fatal error, meaning that the child process launched successfully. */ if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_RUNNING; PRTE_FLAG_SET(cd->child, PRTE_PROC_FLAG_ALIVE); } close(read_fd); return PRTE_SUCCESS; } /** * Fork/exec the specified processes */ static int odls_alps_fork_local_proc(void *cdptr) { prte_odls_spawn_caddy_t *cd = (prte_odls_spawn_caddy_t *) cdptr; int p[2]; pid_t pid; /* A pipe is used to communicate between the parent and child to indicate whether the exec ultimately succeeded or failed. The child sets the pipe to be close-on-exec; the child only ever writes anything to the pipe if there is an error (e.g., executable not found, exec() fails, etc.). The parent does a blocking read on the pipe; if the pipe closed with no data, then the exec() succeeded. If the parent reads something from the pipe, then the child was letting us know why it failed. */ if (pipe(p) < 0) { PMIX_ERROR_LOG(PMIX_ERR_SYS_LIMITS_PIPES); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_FAILED_TO_START; cd->child->exit_code = PMIX_ERR_SYS_LIMITS_PIPES; } return PMIX_ERR_SYS_LIMITS_PIPES; } /* Fork off the child */ pid = fork(); if (NULL != cd->child) { cd->child->pid = pid; } if (pid < 0) { PMIX_ERROR_LOG(PMIX_ERR_SYS_LIMITS_CHILDREN); if (NULL != cd->child) { cd->child->state = PRTE_PROC_STATE_FAILED_TO_START; cd->child->exit_code = PMIX_ERR_SYS_LIMITS_CHILDREN; } return PMIX_ERR_SYS_LIMITS_CHILDREN; } if (pid == 0) { close(p[0]); #if HAVE_SETPGID setpgid(0, 0); #endif do_child(cd, p[1]); /* Does not return */ } close(p[1]); return do_parent(cd, p[0]); } /** * Launch all processes allocated to the current node. */ int prte_odls_alps_launch_local_procs(pmix_data_buffer_t *data) { pmix_nspace_t job; int rc; /* construct the list of children we are to launch */ rc = prte_odls_base_default_construct_child_list(data, &job); if (PRTE_SUCCESS != rc) { PMIX_OUTPUT_VERBOSE((2, prte_odls_base_framework.framework_output, "%s odls:alps:launch:local failed to construct child list on error %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), PRTE_ERROR_NAME(rc))); return rc; } /* get the RDMA credentials and push them into the launch environment */ rc = prte_odls_alps_get_rdma_creds(); if (PRTE_SUCCESS != rc) { PMIX_OUTPUT_VERBOSE((2, prte_odls_base_framework.framework_output, "%s odls:alps:launch:failed to get GNI rdma credentials %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), PRTE_ERROR_NAME(rc))); return rc; } /* launch the local procs */ PRTE_ACTIVATE_LOCAL_LAUNCH(job, odls_alps_fork_local_proc); return PRTE_SUCCESS; } /** * Send a signal to a pid. Note that if we get an error, we set the * return value and let the upper layer print out the message. */ static int send_signal(pid_t pid, int signal) { int rc = PRTE_SUCCESS; PMIX_OUTPUT_VERBOSE((1, prte_odls_base_framework.framework_output, "%s sending signal %d to pid %ld", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), signal, (long) pid)); if (kill(pid, signal) != 0) { switch (errno) { case EINVAL: rc = PRTE_ERR_BAD_PARAM; break; case ESRCH: /* This case can occur when we deliver a signal to a process that is no longer there. This can happen if we deliver a signal while the job is shutting down. This does not indicate a real problem, so just ignore the error. */ break; case EPERM: rc = PRTE_ERR_PERM; break; default: rc = PRTE_ERROR; } } return rc; } static int prte_odls_alps_signal_local_procs(const pmix_proc_t *proc, int32_t signal) { int rc; rc = prte_odls_base_default_signal_local_procs(proc, signal, send_signal); if (PRTE_SUCCESS != rc) { PRTE_ERROR_LOG(rc); return rc; } return PRTE_SUCCESS; } static int prte_odls_alps_restart_proc(prte_proc_t *child) { int rc; /* restart the local proc */ rc = prte_odls_base_default_restart_proc(child, odls_alps_fork_local_proc); if (PRTE_SUCCESS != rc) { PMIX_OUTPUT_VERBOSE((2, prte_odls_base_framework.framework_output, "%s odls:alps:restart_proc failed to launch on error %s", PRTE_NAME_PRINT(PRTE_PROC_MY_NAME), PRTE_ERROR_NAME(rc))); } return rc; }