1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
|
/*
* Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2011 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) 2006-2013 Los Alamos National Security, LLC.
* All rights reserved.
* Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2011 Oak Ridge National Labs. All rights reserved.
* Copyright (c) 2013-2019 Intel, Inc. All rights reserved.
* Copyright (c) 2015 Mellanox Technologies, Inc. All rights reserved.
* Copyright (c) 2021 Nanook Consulting. All rights reserved.
* Copyright (c) 2021 IBM Corporation. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*
*/
#define _GNU_SOURCE
#include <libgen.h>
#include <pthread.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include <unistd.h>
#include "debugger.h"
#include <pmix_tool.h>
/*
* This module is an example of a PMIx debugger daemon. The debugger daemon
* handles interactions with application processes on a node in behalf of the
* front end debugger process.
*/
static pmix_proc_t myproc;
static char *target_namespace = NULL;
/* This is a callback function for the PMIx_Query
* API. The query will callback with a status indicating
* if the request could be fully satisfied, partially
* satisfied, or completely failed. The info parameter
* contains an array of the returned data, with the
* info->key field being the key that was provided in
* the query call. Thus, you can correlate the returned
* data in the info->value field to the requested key.
*
* Once we have dealt with the returned data, we must
* call the release_fn so that the PMIx library can
* cleanup */
static void cbfunc(pmix_status_t status, pmix_info_t *info, size_t ninfo, void *cbdata,
pmix_release_cbfunc_t release_fn, void *release_cbdata)
{
myquery_data_t *mq = (myquery_data_t *) cbdata;
size_t n;
mq->status = status;
printf("%s called as daemon callback for PMIx_Query\n", __FUNCTION__);
/* Save the returned info - it will be * released in the release_fn */
if (0 < ninfo) {
PMIX_INFO_CREATE(mq->info, ninfo);
mq->ninfo = ninfo;
for (n = 0; n < ninfo; n++) {
printf("Transferring %s\n", info[n].key);
PMIX_INFO_XFER(&mq->info[n], &info[n]);
}
}
/* Let the library release the data */
if (NULL != release_fn) {
release_fn(release_cbdata);
}
/* Release the lock */
DEBUG_WAKEUP_THREAD(&mq->lock);
}
/* This is the event notification function we pass down below
* when registering for general events - i.e.,, the default
handler. We don't technically need to register one, but it
* is usually good practice to catch any events that occur */
static void notification_fn(size_t evhdlr_registration_id, pmix_status_t status,
const pmix_proc_t *source, pmix_info_t info[], size_t ninfo,
pmix_info_t results[], size_t nresults,
pmix_event_notification_cbfunc_fn_t cbfunc, void *cbdata)
{
printf("%s called as daemon default event handler for event=%s\n", __FUNCTION__,
PMIx_Error_string(status));
if (NULL != cbfunc) {
cbfunc(PMIX_SUCCESS, NULL, 0, NULL, NULL, cbdata);
}
}
/* This is an event notification function that we explicitly request
* be called when the PMIX_EVENT_JOB_END notification is issued.
* We could catch it in the general event notification function and test
* the status to see if it was "job terminated", but it often is simpler
* to declare a use-specific notification callback point. In this case,
* we are asking to know whenever a job terminates, and we will then
* know we can exit */
static void release_fn(size_t evhdlr_registration_id, pmix_status_t status,
const pmix_proc_t *source, pmix_info_t info[], size_t ninfo,
pmix_info_t results[], size_t nresults,
pmix_event_notification_cbfunc_fn_t cbfunc, void *cbdata)
{
myrel_t *lock;
bool found;
int exit_code;
size_t n;
pmix_proc_t *affected = NULL;
printf("%s called as daemon callback for event=%s\n", __FUNCTION__, PMIx_Error_string(status));
/* Be sure notification is for our application process namespace */
if (0 != strcmp(target_namespace, source->nspace)) {
printf("Ignoring termination notification for '%s'\n", source->nspace);
/* Tell the event handler state machine that we are the last step */
if (NULL != cbfunc) {
cbfunc(PMIX_EVENT_ACTION_COMPLETE, NULL, 0, NULL, NULL, cbdata);
}
return;
}
/* Find our return object */
lock = NULL;
found = false;
for (n = 0; n < ninfo; n++) {
/* Retrieve the lock that needs to be released by this callback. */
if (0 == strncmp(info[n].key, PMIX_EVENT_RETURN_OBJECT, PMIX_MAX_KEYLEN)) {
lock = (myrel_t *) info[n].value.data.ptr;
/* Not every RM will provide an exit code, but check if one was
* given */
} else if (0 == strncmp(info[n].key, PMIX_EXIT_CODE, PMIX_MAX_KEYLEN)) {
exit_code = info[n].value.data.integer;
found = true;
} else if (0 == strncmp(info[n].key, PMIX_EVENT_AFFECTED_PROC, PMIX_MAX_KEYLEN)) {
affected = info[n].value.data.proc;
}
}
/* if the lock object wasn't returned, then that is an error */
if (NULL == lock) {
fprintf(stderr, "LOCK WASN'T RETURNED IN RELEASE CALLBACK\n");
/* let the event handler progress */
if (NULL != cbfunc) {
cbfunc(PMIX_SUCCESS, NULL, 0, NULL, NULL, cbdata);
}
return;
}
printf("DEBUGGER DAEMON NAMESPACE %s NOTIFIED THAT JOB TERMINATED - AFFECTED %s\n",
lock->nspace, (NULL == affected) ? "NULL" : affected->nspace);
/* If the lock object was found then store return status in the lock
* object. */
if (found) {
lock->exit_code = exit_code;
lock->exit_code_given = true;
}
/* Tell the event handler state machine that we are the last step */
if (NULL != cbfunc) {
cbfunc(PMIX_EVENT_ACTION_COMPLETE, NULL, 0, NULL, NULL, cbdata);
}
/* Wake up the thread that is waiting for this callback to complete */
DEBUG_WAKEUP_THREAD(&lock->lock);
}
/* Event handler registration is done asynchronously because it
* may involve the PMIx server registering with the host RM for
* external events. So we provide a callback function that returns
* the status of the request (success or an error), plus a numerical index
* to the registered event. The index is used later on to deregister
* an event handler - if we don't explicitly deregister it, then the
* PMIx server will do so when it see us exit */
static void evhandler_reg_callbk(pmix_status_t status, size_t evhandler_ref, void *cbdata)
{
mylock_t *lock = (mylock_t *) cbdata;
printf("%s called by daemon as registration callback\n", __FUNCTION__);
if (PMIX_SUCCESS != status) {
fprintf(stderr, "Client %s:%d EVENT HANDLER REGISTRATION FAILED WITH STATUS %d, ref=%lu\n",
myproc.nspace, myproc.rank, status, (unsigned long) evhandler_ref);
}
lock->status = status;
DEBUG_WAKEUP_THREAD(lock);
}
int main(int argc, char **argv)
{
pmix_status_t rc;
pmix_value_t *val;
void *dirs;
pmix_proc_t proc;
pmix_info_t *info;
size_t ninfo;
pmix_query_t *query;
pmix_proc_info_t *proctable;
size_t nq;
size_t n;
myquery_data_t myquery_data;
pid_t pid;
pmix_status_t code = PMIX_EVENT_JOB_END;
mylock_t mylock;
myrel_t myrel;
uint16_t localrank;
int i;
pmix_data_array_t darray;
int cospawned_namespace = 0;
char hostname[256];
pid = getpid();
gethostname(hostname, sizeof hostname);
/* Initialize this daemon - since we were launched by the RM, our
* connection info * will have been provided at startup. */
if (PMIX_SUCCESS != (rc = PMIx_tool_init(&myproc, NULL, 0))) {
fprintf(stderr, "Debugger daemon: PMIx_tool_init failed: %s\n", PMIx_Error_string(rc));
exit(0);
}
printf("Debugger daemon ns %s on host %s rank %d pid %lu: Running\n", myproc.nspace,
hostname, myproc.rank, (unsigned long) pid);
/* Register our default event handler */
DEBUG_CONSTRUCT_LOCK(&mylock);
PMIX_INFO_CREATE(info, 1);
PMIX_INFO_LOAD(&info[0], PMIX_EVENT_HDLR_NAME, "DEFAULT", PMIX_STRING);
PMIx_Register_event_handler(NULL, 0, info, 1, notification_fn, evhandler_reg_callbk,
(void *) &mylock);
DEBUG_WAIT_THREAD(&mylock);
PMIX_INFO_FREE(info, 1);
if (PMIX_SUCCESS != mylock.status) {
rc = mylock.status;
DEBUG_DESTRUCT_LOCK(&mylock);
goto done;
}
DEBUG_DESTRUCT_LOCK(&mylock);
/*
* Get the namespace of the job we are to debug. If the application and the
* debugger daemons are spawned separately or if the debugger is attaching
* to a running application, the debugger will set the application
* namespace in the PMIX_DEBUG_TARGET attribute, and the daemon retrieves
* it by calling PMIx_Get.
*
* If the application processes and debugger daemons are spawned together
* (cospawn), then the debugger cannot pass the application namespace since
* that is not known until after the PMIx_Spawn call completes. However,
* the applicaton processes and the debugger daemons have the same
* namespace, so this module uses the debugger namespace, which it knows.
*/
PMIX_LOAD_PROCID(&proc, myproc.nspace, PMIX_RANK_WILDCARD);
rc = PMIx_Get(&proc, PMIX_DEBUG_TARGET, NULL, 0, &val);
if (PMIX_ERR_NOT_FOUND == rc) {
/* Save the application namespace for later */
// NOTE: This is a bug. The cospawned namespace should be more distinct.
/*
fprintf(stderr,
"[%s:%d:%lu] Warning: Could not find PMIX_DEBUG_TARGET. Assume Cospawn.\n",
myproc.nspace, myproc.rank);
*/
target_namespace = strdup(myproc.nspace);
cospawned_namespace = 1;
} else if (rc != PMIX_SUCCESS) {
fprintf(stderr, "[%s:%d:%lu] Failed to get job being debugged - error %s\n", myproc.nspace,
myproc.rank, (unsigned long) pid, PMIx_Error_string(rc));
goto done;
} else {
/* Verify that the expected data structures were returned */
if (NULL == val || PMIX_PROC != val->type) {
fprintf(stderr, "[%s:%d:%lu] Failed to get job being debugged - NULL data returned\n",
myproc.nspace, myproc.rank, (unsigned long) pid);
goto done;
}
printf("[%s:%d:%lu] PMIX_DEBUG_JOB is '%s'\n", proc.nspace, proc.rank, (unsigned long) pid,
val->data.proc->nspace);
/* Save the application namespace for later */
target_namespace = strdup(val->data.proc->nspace);
PMIX_VALUE_RELEASE(val);
}
printf("[%s:%d:%lu] Debugging '%s'\n", myproc.nspace, myproc.rank, (unsigned long) pid,
target_namespace);
/* Get my local rank so I can determine which local proc is "mine" to
* debug */
val = NULL;
if (PMIX_SUCCESS != (rc = PMIx_Get(&myproc, PMIX_LOCAL_RANK, NULL, 0, &val))) {
fprintf(stderr, "[%s:%d:%lu] Failed to get my local rank - error %s\n", myproc.nspace,
myproc.rank, (unsigned long) pid, PMIx_Error_string(rc));
goto done;
}
/* Verify the expected data object was returned */
if (NULL == val) {
fprintf(stderr, "[%s:%d:%lu] Failed to get my local rank - NULL data returned\n",
myproc.nspace, myproc.rank, (unsigned long) pid);
goto done;
}
if (PMIX_UINT16 != val->type) {
fprintf(stderr, "[%s:%d:%lu] Failed to get my local rank - returned wrong type %s\n",
myproc.nspace, myproc.rank, (unsigned long) pid, PMIx_Data_type_string(val->type));
goto done;
}
/* Save the rank */
localrank = val->data.uint16;
PMIX_VALUE_RELEASE(val);
printf("[%s:%d:%lu] my local rank %d\n", myproc.nspace, myproc.rank, (unsigned long) pid,
(int) localrank);
/* Register an event handler specifically for when the target job
* completes */
DEBUG_CONSTRUCT_LOCK(&myrel.lock);
myrel.nspace = strdup(proc.nspace);
PMIX_LOAD_PROCID(&proc, target_namespace, PMIX_RANK_WILDCARD);
PMIX_INFO_LIST_START(dirs);
PMIX_INFO_LIST_ADD(rc, dirs, PMIX_EVENT_HDLR_NAME, "APP-TERMINATION", PMIX_STRING);
/* Pass the lock we will use to wait for notification of the
* PMIX_EVENT_JOB_END event */
PMIX_INFO_LIST_ADD(rc, dirs, PMIX_EVENT_RETURN_OBJECT, &myrel, PMIX_POINTER);
/* Only call me back when this specific job terminates */
PMIX_INFO_LIST_ADD(rc, dirs, PMIX_EVENT_AFFECTED_PROC, &proc, PMIX_PROC);
PMIX_INFO_LIST_CONVERT(rc, dirs, &darray);
PMIX_INFO_LIST_RELEASE(dirs);
ninfo = darray.size;
info = darray.array;
printf("[%s:%d:%lu] registering for termination of '%s'\n", myproc.nspace, myproc.rank,
(unsigned long) pid, proc.nspace);
/* Create a lock to wait for completion of the event registration
* callback */
DEBUG_CONSTRUCT_LOCK(&mylock);
PMIx_Register_event_handler(&code, 1, info, ninfo, release_fn, evhandler_reg_callbk,
(void *) &mylock);
DEBUG_WAIT_THREAD(&mylock);
PMIX_DATA_ARRAY_DESTRUCT(&darray);
if (PMIX_SUCCESS != mylock.status) {
fprintf(stderr, "Failed to register handler for PMIX_EVENT_JOB_END: %s\n",
PMIx_Error_string(mylock.status));
rc = mylock.status;
DEBUG_DESTRUCT_LOCK(&mylock);
goto done;
}
DEBUG_DESTRUCT_LOCK(&mylock);
/* Get our local proctable - for scalability reasons, we don't want to
* have our "root" debugger process get the proctable for everybody and
* send it out to us. So ask the local PMIx server for the pid's of
* our local target processes
*/
nq = 1;
PMIX_QUERY_CREATE(query, nq);
PMIX_ARGV_APPEND(rc, query[0].keys, PMIX_QUERY_LOCAL_PROC_TABLE);
n = 0;
ninfo = 1;
query[0].nqual = ninfo;
PMIX_INFO_CREATE(query[0].qualifiers, ninfo);
/* Set the namespace to query */
PMIX_INFO_LOAD(&query[0].qualifiers[n], PMIX_NSPACE, target_namespace, PMIX_STRING);
/* Create the lock used to wait for query completion */
DEBUG_CONSTRUCT_LOCK(&myquery_data.lock);
myquery_data.info = NULL;
myquery_data.ninfo = 0;
/* Execute the query */
if (PMIX_SUCCESS != (rc = PMIx_Query_info_nb(query, nq, cbfunc, (void *) &myquery_data))) {
fprintf(stderr, "PMIx_Query_info failed: %d\n", rc);
goto done;
}
/* Wait for the query to complete */
DEBUG_WAIT_THREAD(&myquery_data.lock);
DEBUG_DESTRUCT_LOCK(&myquery_data.lock);
PMIX_QUERY_FREE(query, nq);
if (PMIX_SUCCESS != myquery_data.status) {
rc = myquery_data.status;
fprintf(stderr, "Error querying proc table for '%s': %s\n", target_namespace,
PMIx_Error_string(myquery_data.status));
goto done;
}
/* Display the process table */
printf("[%s:%d:%lu] Local proctable received for nspace '%s' has %d entries\n", myproc.nspace,
myproc.rank, (unsigned long) pid, target_namespace,
(int) myquery_data.info[0].value.data.darray->size);
proctable = myquery_data.info[0].value.data.darray->array;
for (i = 0; i < myquery_data.info[0].value.data.darray->size; i++) {
printf("Proctable[%d], namespace %s rank %d exec %s\n", i, proctable[i].proc.nspace,
proctable[i].proc.rank, basename(proctable[i].executable_name));
}
/* Now that we have the proctable for our local processes, this daemon can
* interact with application processes, such as setting initial breakpoints,
* or other setup for the debugging * session.
* If the application was launched by the debugger, then all application
* tasks should be suspended in PMIx_Init, usually within the application's
* MPI_Init call.
* Once initial setup is complete, the daemon sends a release event to the
* application processes and those processes resume execution.
*/
(void) strncpy(proc.nspace, target_namespace, PMIX_MAX_NSLEN);
proc.rank = PMIX_RANK_WILDCARD;
// Since we are using the 'wildcard' only one daemon should send
// the release message.
// If we are 'cospawned' then the daemons are not ranked separately
// from the application (this is a bug) so just have everyone
// send the release.
if (0 == myproc.rank || 1 == cospawned_namespace) {
PMIX_INFO_LIST_START(dirs);
/* Send release notification to application namespace */
PMIX_INFO_LIST_ADD(rc, dirs, PMIX_EVENT_CUSTOM_RANGE, &proc, PMIX_PROC);
/* Don't send notification to default event handlers */
PMIX_INFO_LIST_ADD(rc, dirs, PMIX_EVENT_NON_DEFAULT, NULL, PMIX_BOOL);
PMIX_INFO_LIST_CONVERT(rc, dirs, &darray);
PMIX_INFO_LIST_RELEASE(dirs);
info = darray.array;
ninfo = darray.size;
// Todo: Move this to the main tool
// https://github.com/openpmix/prrte/pull/857#discussion_r600849033
sleep(1);
printf("[%s:%u:%lu] Sending release\n", myproc.nspace, myproc.rank, (unsigned long) pid);
rc = PMIx_Notify_event(PMIX_DEBUGGER_RELEASE, NULL, PMIX_RANGE_CUSTOM, info, ninfo, NULL,
NULL);
PMIX_DATA_ARRAY_DESTRUCT(&darray);
if (PMIX_SUCCESS != rc) {
fprintf(stderr, "[%s:%u:%lu] Sending release failed with error %s(%d)\n", myproc.nspace,
myproc.rank, (unsigned long) pid, PMIx_Error_string(rc), rc);
goto done;
}
}
/* At this point the application processes should be running under debugger
* control. The daemons can interact further with application processes as
* needed, or just wait for the application * termination.
* This example just waits for application termination.
* Note that if the application processes and daemon processes are spawned
* by the same PMIx_Spawn call, then no PMIX_EVENT_JOB_END
* notifications are sent since the daemons are part of the same namespace
* and are still running.
*/
if (0 == cospawned_namespace) {
printf("Waiting for application namespace %s to terminate\n", proc.nspace);
DEBUG_WAIT_THREAD(&myrel.lock);
printf("Application namespace %s terminated\n", proc.nspace);
}
done:
if (NULL != target_namespace) {
free(target_namespace);
}
/* Call PMIx_tool_finalize to shut down the PMIx runtime */
printf("Debugger daemon ns %s rank %d pid %lu: Finalizing\n", myproc.nspace, myproc.rank,
(unsigned long) pid);
rc = PMIx_tool_finalize();
fclose(stdout);
fclose(stderr);
sleep(1);
return (rc);
}
|