totalview.c

MPI stands for the Message Passing Interface. Written by the MPI Forum (a large committee comprising

/* -*- C -*-
 *
 * Copyright (c) 2004-2006 The Trustees of Indiana University and Indiana
 *                         University Research and Technology
 *                         Corporation.  All rights reserved.
 * Copyright (c) 2004-2006 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      Los Alamos National Security, LLC.  All rights
 *                         reserved. 
 * Copyright (c) 2007      Cisco, Inc.  All rights reserved.
 * Copyright (c) 2007      Sun Microsystems, Inc.  All rights reserved.
 * $COPYRIGHT$
 * 
 * Additional copyrights may follow
 * 
 * $HEADER$
 */

#include "orte_config.h"

/*
 * Debugger support for orterun
 *
 * We interpret the MPICH debugger interface as follows:
 *
 * a) The launcher
 *      - spawns the other processes,
 *      - fills in the table MPIR_proctable, and sets MPIR_proctable_size
 *      - sets MPIR_debug_state to MPIR_DEBUG_SPAWNED ( = 1)
 *      - calls MPIR_Breakpoint() which the debugger will have a
 *	  breakpoint on.
 *
 *  b) Applications start and then spin until MPIR_debug_gate is set
 *     non-zero by the debugger.
 *
 * This file implements (a).
 */

#include <stdio.h>
#ifdef HAVE_STDLIB_H
#include <stdlib.h>
#endif  /* HAVE_STDLIB_H */
#ifdef HAVE_STRINGS_H
#include <strings.h>
#endif  /* HAVE_STRINGS_H */
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif  /* HAVE_UNISTD_H */
#include <ctype.h>

#include "opal/util/opal_environ.h"
#include "opal/util/output.h"
#include "opal/util/argv.h"
#include "opal/util/show_help.h"
#include "opal/util/path.h"
#include "opal/util/os_path.h"
#include "opal/class/opal_list.h"
#include "opal/mca/base/base.h"
#include "orte/mca/errmgr/errmgr.h"
#include "orte/mca/rmgr/rmgr_types.h"
#include "orte/mca/rmaps/rmaps.h"
#include "orte/runtime/runtime.h"
#include "orterun.h"
#include "orterun.h"
#include "totalview.h"

/* +++ begin MPICH/TotalView interface definitions */

#define MPIR_DEBUG_SPAWNED  1
#define MPIR_DEBUG_ABORTING 2

struct MPIR_PROCDESC {
    char *host_name;        /* something that can be passed to inet_addr */
    char *executable_name;  /* name of binary */
    int pid;                /* process pid */
};

ORTE_DECLSPEC struct MPIR_PROCDESC *MPIR_proctable = NULL;
ORTE_DECLSPEC int MPIR_proctable_size = 0;
ORTE_DECLSPEC int MPIR_being_debugged = 0;
ORTE_DECLSPEC int MPIR_force_to_main = 0;
ORTE_DECLSPEC volatile int MPIR_debug_state = 0;
ORTE_DECLSPEC volatile int MPIR_i_am_starter = 0;
ORTE_DECLSPEC volatile int MPIR_debug_gate = 0;
ORTE_DECLSPEC volatile int MPIR_acquired_pre_main = 0;

/* --- end MPICH/TotalView interface definitions */

/*
 * NOTE: The job description in the registry will likely evolve to use
 * the "jobgrp_t", but this works for now.
 * 
 * An initial skeleton of how to implement this with jobgrp_t is
 * available in SVN as orte/tools/orterun/totalview.c, version 7075.
 */


#define DUMP_INT(X) fprintf(stderr, "  %s = %d\n", # X, X);

static void dump(void)
{
    int i;

    DUMP_INT(MPIR_being_debugged);
    DUMP_INT(MPIR_debug_gate);
    DUMP_INT(MPIR_debug_state);
    DUMP_INT(MPIR_acquired_pre_main);
    DUMP_INT(MPIR_i_am_starter);
    DUMP_INT(MPIR_proctable_size);
    fprintf(stderr, "  MPIR_proctable:\n");
    for (i = 0; i < MPIR_proctable_size; i++) {
        fprintf(stderr,
                "    (i, host, exe, pid) = (%d, %s, %s, %d)\n",
                i,
                MPIR_proctable[i].host_name,
                MPIR_proctable[i].executable_name,
                MPIR_proctable[i].pid);
    }
}

/*
 * Process one line from the orte_base_user_debugger MCA param and
 * look for that debugger in the path.  If we find it, fill in
 * new_argv.
 */
static int process(char *orig_line, char *basename, opal_cmd_line_t *cmd_line,
                   int argc, char **argv, char ***new_argv) 
{
    int i;
    char *line, *full_line = strdup(orig_line);
    char *user_argv, *tmp, *tmp2, **tmp_argv, **executable;
    char cwd[PATH_MAX];
    bool used_num_procs = false;
    bool fail_single_app = false;
    bool fail_needed_executable = false;

    line = full_line;
    if (NULL == line) {
        return ORTE_ERR_OUT_OF_RESOURCE;
    }

    /* Trim off whitespace at the beginning and ending of line */

    for (i = 0; '\0' != line[i] && isspace(line[i]); ++line) {
        continue;
    }
    for (i = strlen(line) - 2; i > 0 && isspace(line[i]); ++i) {
        line[i] = '\0';
    }
    if (strlen(line) <= 0) {
        return ORTE_ERROR;
    }

    /* Get the tail of the command line (i.e., the user executable /
       argv) */

    opal_cmd_line_get_tail(cmd_line, &i, &executable);

    /* Remove --debug, --debugger, and -tv from the user command line
       params */

    if (1 == argc) {
        user_argv = strdup("");
    } else {
        tmp_argv = opal_argv_copy(argv);
        for (i = 0; NULL != tmp_argv[i]; ++i) {
            if (0 == strcmp(tmp_argv[i], "-debug") ||
                0 == strcmp(tmp_argv[i], "--debug")) {
                free(tmp_argv[i]);
                tmp_argv[i] = strdup("");
            } else if (0 == strcmp(tmp_argv[i], "-tv") ||
                0 == strcmp(tmp_argv[i], "--tv")) {
                free(tmp_argv[i]);
                tmp_argv[i] = strdup("");
            } else if (0 == strcmp(tmp_argv[i], "--debugger") ||
                       0 == strcmp(tmp_argv[i], "-debugger")) {
                free(tmp_argv[i]);
                tmp_argv[i] = strdup("");
                if (NULL != tmp_argv[i + 1]) {
                    ++i;
                    free(tmp_argv[i]);
                    tmp_argv[i] = strdup("");
                }
            }
        }
        user_argv = opal_argv_join(tmp_argv + 1, ' ');
        opal_argv_free(tmp_argv);
    }

    /* Replace @@ tokens - line should never realistically be bigger
       than MAX_INT, so just cast to int to remove compiler warning */

    for (i = 0; i < (int) strlen(line); ++i) {
        tmp = NULL;
        if (0 == strncmp(line + i, "@mpirun@", 8)) {
            line[i] = '\0';
            asprintf(&tmp, "%s%s%s", line, argv[0], line + i + 8);
        } else if (0 == strncmp(line + i, "@orterun@", 9)) {
            line[i] = '\0';
            asprintf(&tmp, "%s%s%s", line, argv[0], line + i + 9);
        } else if (0 == strncmp(line + i, "@mpirun_args@", 13)) {
            line[i] = '\0';
            asprintf(&tmp, "%s%s%s", line, user_argv, line + i + 13);
        } else if (0 == strncmp(line + i, "@orterun_args@", 14)) {
            line[i] = '\0';
            asprintf(&tmp, "%s%s%s", line, user_argv, line + i + 14);
        } else if (0 == strncmp(line + i, "@np@", 4)) {
            line[i] = '\0';
            asprintf(&tmp, "%s%d%s", line, orterun_globals.num_procs,
                     line + i + 4);
            used_num_procs = true;
        } else if (0 == strncmp(line + i, "@single_app@", 12)) {
            line[i] = '\0';
            /* This token is only a flag; it is not replaced with any
               alternate text */
            asprintf(&tmp, "%s%s", line, line + i + 12);
            tmp2 = opal_argv_join(executable + 1, ' ');
            if (NULL != strchr(tmp2, ':')) {
                fail_single_app = true;
            }
            free(tmp2);
        } else if (0 == strncmp(line + i, "@executable@", 12)) {
            line[i] = '\0';
            /* If we found the executable, paste it in.  Otherwise,
               this is a possible error. */
            if (NULL != executable) {
                asprintf(&tmp, "%s%s%s", line, executable[0], line + i + 12);
            } else {
                fail_needed_executable = true;
            }
        } else if (0 == strncmp(line + i, "@executable_argv@", 17)) {
            line[i] = '\0';
            /* If we found the tail, paste in the argv.  Otherwise,
               this is a possible error. */
            if (NULL != executable) {
                if (NULL != executable[1]) {
                    /* Put in the argv */
                    tmp2 = opal_argv_join(executable + 1, ' ');
                    asprintf(&tmp, "%s%s%s", line, tmp2, line + i + 17);
                    free(tmp2);
                } else {
                    /* There is no argv; just paste the front and back
                       together, removing the @token@ */
                    asprintf(&tmp, "%s%s", line, line + i + 17);
                }
            } else {
                fail_needed_executable = true;
            }
        }

        if (NULL != tmp) {
            free(full_line);
            full_line = line = tmp;
            --i;
        }
    }

    /* Split up into argv */

    *new_argv = opal_argv_split(line, ' ');
    free(line);