2004-08-25 21:39:08 +04:00
|
|
|
/*
|
2007-03-17 02:11:45 +03:00
|
|
|
* Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
|
2005-11-05 22:57:48 +03:00
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
2007-06-14 08:38:06 +04:00
|
|
|
* Copyright (c) 2004-2007 The University of Tennessee and The University
|
2005-11-05 22:57:48 +03:00
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
2004-11-28 23:09:25 +03:00
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
|
|
* University of Stuttgart. All rights reserved.
|
2005-03-24 15:43:37 +03:00
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
Fix a number of OOB issues:
* Remove the connect() timeout code, as it had some nasty race conditions
when connections were established as the trigger was firing. A better
solution has been found for the cluster where this was needed, so just
removing it was easiest.
* When a fatal error (too many connection failures) occurs, set an error
on messages in the queue even if there isn't an active message. The
first message to any peer will be queued without being active (and
so will all subsequent messages until the connection is established),
and the orteds will hang until that first message completes. So if
an orted can never contact it's peer, it will never exit and just sit
waiting for that message to complete.
* Cover an interesting RST condition in the connect code. A connection
can complete the three-way handshake, the connector can even send
some data, but the server side will drop the connection because it
can't move it from the half-connected to fully-connected state because
of space shortage in the listen backlog queue. This causes a RST to
be received first time that recv() is called, which will be when waiting
for the remote side of the OOB ack. In this case, transition the
connection back into a CLOSED state and try to connect again.
* Add levels of debugging, rather than all or nothing, each building on
the previous level. 0 (default) is hard errors. 1 is connection
error debugging info. 2 is all connection info. 3 is more state
info. 4 includes all message info.
* Add some hopefully useful comments
This commit was SVN r14261.
2007-04-08 02:33:30 +04:00
|
|
|
* Copyright (c) 2006-2007 Los Alamos National Security, LLC.
|
|
|
|
* All rights reserved.
|
2004-11-22 04:38:40 +03:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
2004-08-25 21:39:08 +04:00
|
|
|
* $HEADER$
|
2006-03-11 06:09:24 +03:00
|
|
|
*
|
|
|
|
* In windows, many of the socket functions return an EWOULDBLOCK
|
|
|
|
* instead of \ things like EAGAIN, EINPROGRESS, etc. It has been
|
|
|
|
* verified that this will \ not conflict with other error codes that
|
|
|
|
* are returned by these functions \ under UNIX/Linux environments
|
2004-08-25 21:39:08 +04:00
|
|
|
*/
|
2006-03-11 06:09:24 +03:00
|
|
|
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "orte_config.h"
|
2006-08-15 00:14:44 +04:00
|
|
|
#include "opal/opal_socket_errno.h"
|
2005-03-14 23:57:21 +03:00
|
|
|
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "orte/class/orte_proc_table.h"
|
|
|
|
#include "orte/orte_constants.h"
|
|
|
|
#include "orte/mca/ns/ns.h"
|
|
|
|
#include "orte/mca/oob/tcp/oob_tcp.h"
|
|
|
|
#include "orte/mca/oob/tcp/oob_tcp_msg.h"
|
2004-07-13 02:46:57 +04:00
|
|
|
|
|
|
|
|
|
|
|
static void mca_oob_tcp_msg_construct(mca_oob_tcp_msg_t*);
|
|
|
|
static void mca_oob_tcp_msg_destruct(mca_oob_tcp_msg_t*);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
static void mca_oob_tcp_msg_ident(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer);
|
|
|
|
static bool mca_oob_tcp_msg_recv(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer);
|
2004-09-08 21:02:24 +04:00
|
|
|
static void mca_oob_tcp_msg_data(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer);
|
|
|
|
static void mca_oob_tcp_msg_ping(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer);
|
2004-07-13 02:46:57 +04:00
|
|
|
|
|
|
|
OBJ_CLASS_INSTANCE(
|
|
|
|
mca_oob_tcp_msg_t,
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_item_t,
|
2004-07-13 02:46:57 +04:00
|
|
|
mca_oob_tcp_msg_construct,
|
|
|
|
mca_oob_tcp_msg_destruct);
|
|
|
|
|
|
|
|
|
|
|
|
static void mca_oob_tcp_msg_construct(mca_oob_tcp_msg_t* msg)
|
|
|
|
{
|
2005-07-04 02:45:48 +04:00
|
|
|
OBJ_CONSTRUCT(&msg->msg_lock, opal_mutex_t);
|
|
|
|
OBJ_CONSTRUCT(&msg->msg_condition, opal_condition_t);
|
2004-07-13 02:46:57 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void mca_oob_tcp_msg_destruct(mca_oob_tcp_msg_t* msg)
|
|
|
|
{
|
2004-07-15 17:51:40 +04:00
|
|
|
OBJ_DESTRUCT(&msg->msg_lock);
|
|
|
|
OBJ_DESTRUCT(&msg->msg_condition);
|
2004-07-13 02:46:57 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-07-13 02:46:57 +04:00
|
|
|
* Wait for a msg to complete.
|
|
|
|
* @param msg (IN) Message to wait on.
|
2004-08-03 01:24:00 +04:00
|
|
|
* @param rc (OUT) Return code (number of bytes read on success or error code on failure).
|
2006-02-12 04:33:29 +03:00
|
|
|
* @retval ORTE_SUCCESS or error code on failure.
|
2004-07-13 02:46:57 +04:00
|
|
|
*/
|
|
|
|
|
2004-08-03 01:24:00 +04:00
|
|
|
int mca_oob_tcp_msg_wait(mca_oob_tcp_msg_t* msg, int* rc)
|
2004-07-13 02:46:57 +04:00
|
|
|
{
|
2005-02-16 20:42:07 +03:00
|
|
|
#if OMPI_ENABLE_PROGRESS_THREADS
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&msg->msg_lock);
|
2004-08-28 05:34:48 +04:00
|
|
|
while(msg->msg_complete == false) {
|
2005-07-04 03:09:55 +04:00
|
|
|
if(opal_event_progress_thread()) {
|
2004-08-31 06:57:39 +04:00
|
|
|
int rc;
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
2006-04-06 16:58:09 +04:00
|
|
|
rc = opal_event_loop(OPAL_EVLOOP_ONCE);
|
|
|
|
assert(rc >= 0);
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&msg->msg_lock);
|
2004-08-28 05:34:48 +04:00
|
|
|
} else {
|
2005-07-04 02:45:48 +04:00
|
|
|
opal_condition_wait(&msg->msg_condition, &msg->msg_lock);
|
2004-08-28 05:34:48 +04:00
|
|
|
}
|
|
|
|
}
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
2004-08-28 05:34:48 +04:00
|
|
|
|
|
|
|
#else
|
2004-08-03 01:24:00 +04:00
|
|
|
/* wait for message to complete */
|
2006-04-07 22:13:35 +04:00
|
|
|
while(msg->msg_complete == false) {
|
|
|
|
/* msg_wait() is used in the "barrier" at the completion of
|
|
|
|
MPI_FINALIZE, during which time BTLs may still need to
|
|
|
|
progress pending outgoing communication, so we need to
|
|
|
|
call opal_progress() here to make sure that communication
|
|
|
|
gets pushed out so others can enter finalize (and send us
|
|
|
|
the message we're here waiting for). However, if we're
|
|
|
|
in a callback from the event library that was triggered
|
|
|
|
from a call to opal_progress(), opal_progress() will
|
|
|
|
think another thread is already progressing the event
|
|
|
|
engine (in the case of mpi threads enabled) and not
|
|
|
|
progress the engine, meaning we'll never receive our
|
|
|
|
message. So we also need to progress the event library
|
|
|
|
expicitly. We use EVLOOP_NONBLOCK so that we can
|
|
|
|
progress both the registered callbacks and the event
|
|
|
|
library, as EVLOOP_ONCE may block for a short period of
|
|
|
|
time. */
|
2006-03-29 02:09:40 +04:00
|
|
|
opal_progress();
|
2006-04-07 22:13:35 +04:00
|
|
|
opal_event_loop(OPAL_EVLOOP_NONBLOCK);
|
2007-03-17 02:11:45 +03:00
|
|
|
OPAL_CR_TEST_CHECKPOINT_READY();
|
2006-04-07 22:13:35 +04:00
|
|
|
}
|
2004-08-28 05:34:48 +04:00
|
|
|
#endif
|
2004-08-03 01:24:00 +04:00
|
|
|
|
|
|
|
/* return status */
|
|
|
|
if(NULL != rc) {
|
|
|
|
*rc = msg->msg_rc;
|
|
|
|
}
|
2006-02-12 04:33:29 +03:00
|
|
|
return ORTE_SUCCESS;
|
2004-07-13 02:46:57 +04:00
|
|
|
}
|
|
|
|
|
2004-09-08 21:02:24 +04:00
|
|
|
/*
|
|
|
|
* Wait up to a timeout for the message to complete.
|
|
|
|
* @param msg (IN) Message to wait on.
|
|
|
|
* @param rc (OUT) Return code (number of bytes read on success or error code on failure).
|
2006-02-12 04:33:29 +03:00
|
|
|
* @retval ORTE_SUCCESS or error code on failure.
|
2004-09-08 21:02:24 +04:00
|
|
|
*/
|
|
|
|
|
|
|
|
int mca_oob_tcp_msg_timedwait(mca_oob_tcp_msg_t* msg, int* rc, struct timespec* abstime)
|
|
|
|
{
|
|
|
|
struct timeval tv;
|
|
|
|
uint32_t secs = abstime->tv_sec;
|
|
|
|
uint32_t usecs = abstime->tv_nsec * 1000;
|
|
|
|
gettimeofday(&tv,NULL);
|
|
|
|
|
2005-02-16 20:42:07 +03:00
|
|
|
#if OMPI_ENABLE_PROGRESS_THREADS
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&msg->msg_lock);
|
2004-09-08 21:02:24 +04:00
|
|
|
while(msg->msg_complete == false &&
|
2004-11-21 20:20:42 +03:00
|
|
|
((uint32_t)tv.tv_sec <= secs ||
|
|
|
|
((uint32_t)tv.tv_sec == secs && (uint32_t)tv.tv_usec < usecs))) {
|
2005-07-04 03:09:55 +04:00
|
|
|
if(opal_event_progress_thread()) {
|
2004-09-08 21:02:24 +04:00
|
|
|
int rc;
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
2006-04-06 18:31:38 +04:00
|
|
|
rc = opal_event_loop(OPAL_EVLOOP_ONCE);
|
|
|
|
assert(rc >= 0);
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&msg->msg_lock);
|
2004-09-08 21:02:24 +04:00
|
|
|
} else {
|
2005-07-04 02:45:48 +04:00
|
|
|
opal_condition_timedwait(&msg->msg_condition, &msg->msg_lock, abstime);
|
2004-09-08 21:02:24 +04:00
|
|
|
}
|
2004-09-14 12:06:00 +04:00
|
|
|
gettimeofday(&tv,NULL);
|
2004-09-08 21:02:24 +04:00
|
|
|
}
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
2004-09-08 21:02:24 +04:00
|
|
|
#else
|
|
|
|
/* wait for message to complete */
|
|
|
|
while(msg->msg_complete == false &&
|
2004-11-20 22:12:43 +03:00
|
|
|
((uint32_t)tv.tv_sec <= secs ||
|
|
|
|
((uint32_t)tv.tv_sec == secs && (uint32_t)tv.tv_usec < usecs))) {
|
2006-04-07 22:13:35 +04:00
|
|
|
/* see comment in tcp_msg_wait, above */
|
2006-03-29 02:09:40 +04:00
|
|
|
opal_progress();
|
2006-04-07 22:13:35 +04:00
|
|
|
opal_event_loop(OPAL_EVLOOP_NONBLOCK);
|
2004-09-14 12:06:00 +04:00
|
|
|
gettimeofday(&tv,NULL);
|
2004-09-08 21:02:24 +04:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/* return status */
|
|
|
|
if(NULL != rc) {
|
|
|
|
*rc = msg->msg_rc;
|
|
|
|
}
|
|
|
|
if(msg->msg_rc < 0)
|
|
|
|
return msg->msg_rc;
|
2006-02-12 04:33:29 +03:00
|
|
|
return (msg->msg_complete ? ORTE_SUCCESS : ORTE_ERR_TIMEOUT);
|
2004-09-08 21:02:24 +04:00
|
|
|
}
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-07-13 02:46:57 +04:00
|
|
|
* Signal that a message has completed.
|
|
|
|
* @param msg (IN) Message to wait on.
|
2004-07-15 17:51:40 +04:00
|
|
|
* @param peer (IN) the peer of the message
|
2006-02-12 04:33:29 +03:00
|
|
|
* @retval ORTE_SUCCESS or error code on failure.
|
2004-07-13 02:46:57 +04:00
|
|
|
*/
|
2005-03-14 23:57:21 +03:00
|
|
|
int mca_oob_tcp_msg_complete(mca_oob_tcp_msg_t* msg, orte_process_name_t * peer)
|
2004-07-13 02:46:57 +04:00
|
|
|
{
|
2005-10-25 17:48:08 +04:00
|
|
|
OPAL_THREAD_LOCK(&msg->msg_lock);
|
2004-07-15 17:51:40 +04:00
|
|
|
msg->msg_complete = true;
|
|
|
|
if(NULL != msg->msg_cbfunc) {
|
2005-10-25 17:48:08 +04:00
|
|
|
opal_list_item_t* item;
|
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
|
|
|
|
2007-06-14 08:38:06 +04:00
|
|
|
#if defined(__WINDOWS__)
|
|
|
|
/**
|
2007-06-15 02:35:38 +04:00
|
|
|
* In order to be able to generate TCP events recursively, Windows need
|
2007-06-14 08:38:06 +04:00
|
|
|
* to get out of the callback attached to a specific socket. Therefore,
|
|
|
|
* as our OOB allow to block on a connection from a callback on the same
|
|
|
|
* connection, we have to trigger the completion callbacks outside of the
|
|
|
|
* OOB callbacks. We add them to the completed list here, and the progress
|
|
|
|
* engine will call our progress function later once all socket related
|
2007-06-15 02:35:38 +04:00
|
|
|
* events have been processed.
|
2007-06-14 08:38:06 +04:00
|
|
|
*/
|
2007-07-20 05:34:02 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
opal_list_append(&mca_oob_tcp_component.tcp_msg_completed, (opal_list_item_t*)msg);
|
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
return ORTE_SUCCESS;
|
2007-06-14 08:38:06 +04:00
|
|
|
#else
|
2007-07-20 05:34:02 +04:00
|
|
|
|
|
|
|
/* post to a global list of completed messages */
|
|
|
|
if ((msg->msg_flags & ORTE_RML_FLAG_RECURSIVE_CALLBACK) == 0) {
|
|
|
|
int size;
|
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
opal_list_append(&mca_oob_tcp_component.tcp_msg_completed, (opal_list_item_t*)msg);
|
|
|
|
size = opal_list_get_size(&mca_oob_tcp_component.tcp_msg_completed);
|
2005-10-25 17:48:08 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_lock);
|
2007-07-20 05:34:02 +04:00
|
|
|
if(size > 1) {
|
|
|
|
return ORTE_SUCCESS;
|
|
|
|
}
|
2005-10-25 17:48:08 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* invoke message callback */
|
2004-09-02 03:07:40 +04:00
|
|
|
msg->msg_cbfunc(msg->msg_rc, peer, msg->msg_uiov, msg->msg_ucnt, msg->msg_hdr.msg_tag, msg->msg_cbdata);
|
2005-10-25 17:48:08 +04:00
|
|
|
|
2007-07-20 05:34:02 +04:00
|
|
|
/* dispatch any completed events */
|
|
|
|
if ((msg->msg_flags & ORTE_RML_FLAG_RECURSIVE_CALLBACK) == 0) {
|
2005-10-25 17:48:08 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_lock);
|
2007-07-20 05:34:02 +04:00
|
|
|
opal_list_remove_item(&mca_oob_tcp_component.tcp_msg_completed, (opal_list_item_t*)msg);
|
|
|
|
MCA_OOB_TCP_MSG_RETURN(msg);
|
|
|
|
while(NULL !=
|
|
|
|
(item = opal_list_remove_first(&mca_oob_tcp_component.tcp_msg_completed))) {
|
|
|
|
msg = (mca_oob_tcp_msg_t*)item;
|
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
msg->msg_cbfunc(
|
|
|
|
msg->msg_rc,
|
|
|
|
&msg->msg_peer,
|
|
|
|
msg->msg_uiov,
|
|
|
|
msg->msg_ucnt,
|
|
|
|
msg->msg_hdr.msg_tag,
|
|
|
|
msg->msg_cbdata);
|
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
MCA_OOB_TCP_MSG_RETURN(msg);
|
|
|
|
}
|
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_lock);
|
|
|
|
} else {
|
2005-10-31 19:21:11 +03:00
|
|
|
MCA_OOB_TCP_MSG_RETURN(msg);
|
2005-10-25 17:48:08 +04:00
|
|
|
}
|
2007-06-14 08:38:06 +04:00
|
|
|
#endif /* defined(__WINDOWS__) */
|
2004-07-15 17:51:40 +04:00
|
|
|
} else {
|
2005-07-04 02:45:48 +04:00
|
|
|
opal_condition_broadcast(&msg->msg_condition);
|
2005-10-25 17:48:08 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&msg->msg_lock);
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
2006-02-12 04:33:29 +03:00
|
|
|
return ORTE_SUCCESS;
|
2004-07-13 02:46:57 +04:00
|
|
|
}
|
|
|
|
|
2004-07-15 23:08:54 +04:00
|
|
|
/*
|
2004-07-15 17:51:40 +04:00
|
|
|
* The function that actually sends the data!
|
|
|
|
* @param msg a pointer to the message to send
|
|
|
|
* @param peer the peer we are sending to
|
|
|
|
* @retval true if the entire message has been sent
|
|
|
|
* @retval false if the entire message has not been sent
|
|
|
|
*/
|
2004-07-15 23:08:54 +04:00
|
|
|
bool mca_oob_tcp_msg_send_handler(mca_oob_tcp_msg_t* msg, struct mca_oob_tcp_peer_t * peer)
|
2004-07-15 17:51:40 +04:00
|
|
|
{
|
|
|
|
int rc;
|
|
|
|
while(1) {
|
2004-08-03 01:24:00 +04:00
|
|
|
rc = writev(peer->peer_sd, msg->msg_rwptr, msg->msg_rwnum);
|
|
|
|
if(rc < 0) {
|
2006-08-15 00:14:44 +04:00
|
|
|
if(opal_socket_errno == EINTR)
|
2004-07-15 17:51:40 +04:00
|
|
|
continue;
|
2005-01-20 03:40:22 +03:00
|
|
|
/* In windows, many of the socket functions return an EWOULDBLOCK instead of \
|
|
|
|
things like EAGAIN, EINPROGRESS, etc. It has been verified that this will \
|
|
|
|
not conflict with other error codes that are returned by these functions \
|
|
|
|
under UNIX/Linux environments */
|
2006-08-15 00:14:44 +04:00
|
|
|
else if (opal_socket_errno == EAGAIN || opal_socket_errno == EWOULDBLOCK)
|
2004-07-15 17:51:40 +04:00
|
|
|
return false;
|
|
|
|
else {
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s-%s mca_oob_tcp_msg_send_handler: writev failed: %s (%d)",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name),
|
|
|
|
ORTE_NAME_PRINT(&(peer->peer_name)),
|
2006-12-14 21:20:43 +03:00
|
|
|
strerror(opal_socket_errno),
|
2006-08-15 00:14:44 +04:00
|
|
|
opal_socket_errno);
|
2004-08-03 01:24:00 +04:00
|
|
|
mca_oob_tcp_peer_close(peer);
|
2006-02-12 04:33:29 +03:00
|
|
|
msg->msg_rc = ORTE_ERR_CONNECTION_FAILED;
|
2005-03-14 23:57:21 +03:00
|
|
|
return true;
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
|
|
|
}
|
2004-08-03 01:24:00 +04:00
|
|
|
|
|
|
|
msg->msg_rc += rc;
|
|
|
|
do {/* while there is still more iovecs to write */
|
2004-11-20 22:12:43 +03:00
|
|
|
if(rc < (int)msg->msg_rwptr->iov_len) {
|
2004-07-15 17:51:40 +04:00
|
|
|
msg->msg_rwptr->iov_len -= rc;
|
2005-01-26 03:20:35 +03:00
|
|
|
msg->msg_rwptr->iov_base = (ompi_iov_base_ptr_t)((char *) msg->msg_rwptr->iov_base + rc);
|
2004-07-15 17:51:40 +04:00
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
rc -= msg->msg_rwptr->iov_len;
|
2004-08-03 01:24:00 +04:00
|
|
|
(msg->msg_rwnum)--;
|
2004-07-15 17:51:40 +04:00
|
|
|
(msg->msg_rwptr)++;
|
2004-08-03 01:24:00 +04:00
|
|
|
if(0 == msg->msg_rwnum) {
|
2004-07-15 17:51:40 +04:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
2007-04-12 09:01:29 +04:00
|
|
|
} while(1);
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-08-03 01:24:00 +04:00
|
|
|
* Receives message data.
|
2007-06-13 02:47:14 +04:00
|
|
|
* @param msg the message to be received into
|
2007-04-12 09:01:29 +04:00
|
|
|
* @param peer the peer to receive from
|
2004-08-03 01:24:00 +04:00
|
|
|
* @retval true if the whole message was received
|
|
|
|
* @retval false if the whole message was not received
|
2004-07-15 17:51:40 +04:00
|
|
|
*/
|
2004-07-15 23:08:54 +04:00
|
|
|
bool mca_oob_tcp_msg_recv_handler(mca_oob_tcp_msg_t* msg, struct mca_oob_tcp_peer_t * peer)
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
{
|
|
|
|
/* has entire header been received */
|
|
|
|
if(msg->msg_rwptr == msg->msg_rwiov) {
|
|
|
|
if(mca_oob_tcp_msg_recv(msg, peer) == false)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
/* allocate a buffer for the receive */
|
|
|
|
MCA_OOB_TCP_HDR_NTOH(&msg->msg_hdr);
|
|
|
|
if(msg->msg_hdr.msg_size > 0) {
|
|
|
|
msg->msg_rwbuf = malloc(msg->msg_hdr.msg_size);
|
|
|
|
if(NULL == msg->msg_rwbuf) {
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s-%s mca_oob_tcp_msg_recv_handler: malloc(%d) failed\n",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name),
|
|
|
|
ORTE_NAME_PRINT(&(peer->peer_name)),
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
msg->msg_hdr.msg_size);
|
|
|
|
mca_oob_tcp_peer_close(peer);
|
|
|
|
return false;
|
|
|
|
}
|
2005-01-26 03:20:35 +03:00
|
|
|
msg->msg_rwiov[1].iov_base = (ompi_iov_base_ptr_t)msg->msg_rwbuf;
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
msg->msg_rwiov[1].iov_len = msg->msg_hdr.msg_size;
|
|
|
|
msg->msg_rwnum = 1;
|
2005-03-14 23:57:21 +03:00
|
|
|
} else {
|
|
|
|
msg->msg_rwiov[1].iov_base = NULL;
|
|
|
|
msg->msg_rwiov[1].iov_len = 0;
|
|
|
|
msg->msg_rwnum = 0;
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
}
|
2007-04-25 23:08:07 +04:00
|
|
|
if (mca_oob_tcp_component.tcp_debug >= OOB_TCP_DEBUG_INFO) {
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s-%s mca_oob_tcp_msg_recv_handler: size %lu\n",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name),
|
|
|
|
ORTE_NAME_PRINT(&(peer->peer_name)),
|
2007-04-12 09:01:29 +04:00
|
|
|
(unsigned long)(msg->msg_hdr.msg_size) );
|
|
|
|
}
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* do the right thing based on the message type */
|
|
|
|
switch(msg->msg_hdr.msg_type) {
|
|
|
|
case MCA_OOB_TCP_IDENT:
|
|
|
|
/* done - there is nothing else to receive */
|
|
|
|
return true;
|
2004-09-08 21:02:24 +04:00
|
|
|
case MCA_OOB_TCP_PING:
|
|
|
|
/* done - there is nothing else to receive */
|
|
|
|
return true;
|
|
|
|
case MCA_OOB_TCP_DATA:
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
/* finish receiving message */
|
|
|
|
return mca_oob_tcp_msg_recv(msg, peer);
|
|
|
|
default:
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* Process the current iovec
|
|
|
|
*/
|
|
|
|
|
|
|
|
static bool mca_oob_tcp_msg_recv(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer)
|
2004-07-15 17:51:40 +04:00
|
|
|
{
|
|
|
|
int rc;
|
2005-03-14 23:57:21 +03:00
|
|
|
while(msg->msg_rwnum) {
|
2004-08-03 01:24:00 +04:00
|
|
|
rc = readv(peer->peer_sd, msg->msg_rwptr, msg->msg_rwnum);
|
2004-08-28 05:15:19 +04:00
|
|
|
if(rc < 0) {
|
2006-08-15 00:14:44 +04:00
|
|
|
if(opal_socket_errno == EINTR)
|
2004-07-15 17:51:40 +04:00
|
|
|
continue;
|
2005-01-20 03:40:22 +03:00
|
|
|
/* In windows, many of the socket functions return an EWOULDBLOCK instead of \
|
|
|
|
things like EAGAIN, EINPROGRESS, etc. It has been verified that this will \
|
|
|
|
not conflict with other error codes that are returned by these functions \
|
|
|
|
under UNIX/Linux environments */
|
2006-08-15 00:14:44 +04:00
|
|
|
else if (opal_socket_errno == EAGAIN || opal_socket_errno == EWOULDBLOCK)
|
2004-07-15 17:51:40 +04:00
|
|
|
return false;
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s-%s mca_oob_tcp_msg_recv: readv failed: %s (%d)",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name),
|
|
|
|
ORTE_NAME_PRINT(&(peer->peer_name)),
|
2007-04-12 09:01:29 +04:00
|
|
|
strerror(opal_socket_errno),
|
|
|
|
opal_socket_errno);
|
|
|
|
mca_oob_tcp_peer_close(peer);
|
2007-07-20 05:34:02 +04:00
|
|
|
if (NULL != mca_oob_tcp.oob_exception_callback) {
|
|
|
|
mca_oob_tcp.oob_exception_callback(&peer->peer_name, ORTE_RML_PEER_DISCONNECTED);
|
|
|
|
}
|
2007-04-12 09:01:29 +04:00
|
|
|
return false;
|
2004-08-28 05:15:19 +04:00
|
|
|
} else if (rc == 0) {
|
Fix a number of OOB issues:
* Remove the connect() timeout code, as it had some nasty race conditions
when connections were established as the trigger was firing. A better
solution has been found for the cluster where this was needed, so just
removing it was easiest.
* When a fatal error (too many connection failures) occurs, set an error
on messages in the queue even if there isn't an active message. The
first message to any peer will be queued without being active (and
so will all subsequent messages until the connection is established),
and the orteds will hang until that first message completes. So if
an orted can never contact it's peer, it will never exit and just sit
waiting for that message to complete.
* Cover an interesting RST condition in the connect code. A connection
can complete the three-way handshake, the connector can even send
some data, but the server side will drop the connection because it
can't move it from the half-connected to fully-connected state because
of space shortage in the listen backlog queue. This causes a RST to
be received first time that recv() is called, which will be when waiting
for the remote side of the OOB ack. In this case, transition the
connection back into a CLOSED state and try to connect again.
* Add levels of debugging, rather than all or nothing, each building on
the previous level. 0 (default) is hard errors. 1 is connection
error debugging info. 2 is all connection info. 3 is more state
info. 4 includes all message info.
* Add some hopefully useful comments
This commit was SVN r14261.
2007-04-08 02:33:30 +04:00
|
|
|
if(mca_oob_tcp_component.tcp_debug >= OOB_TCP_DEBUG_CONNECT_FAIL) {
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s-%s mca_oob_tcp_msg_recv: peer closed connection",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name),
|
|
|
|
ORTE_NAME_PRINT(&(peer->peer_name)));
|
2004-09-08 21:02:24 +04:00
|
|
|
}
|
2004-08-28 05:15:19 +04:00
|
|
|
mca_oob_tcp_peer_close(peer);
|
2007-07-20 05:34:02 +04:00
|
|
|
if (NULL != mca_oob_tcp.oob_exception_callback) {
|
|
|
|
mca_oob_tcp.oob_exception_callback(&peer->peer_name, ORTE_RML_PEER_DISCONNECTED);
|
|
|
|
}
|
2004-08-28 05:15:19 +04:00
|
|
|
return false;
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
2004-08-03 01:24:00 +04:00
|
|
|
|
2004-07-15 17:51:40 +04:00
|
|
|
do {
|
2004-11-20 22:12:43 +03:00
|
|
|
if(rc < (int)msg->msg_rwptr->iov_len) {
|
2004-07-15 17:51:40 +04:00
|
|
|
msg->msg_rwptr->iov_len -= rc;
|
2005-01-26 03:20:35 +03:00
|
|
|
msg->msg_rwptr->iov_base = (ompi_iov_base_ptr_t)((char *) msg->msg_rwptr->iov_base + rc);
|
2004-07-15 17:51:40 +04:00
|
|
|
break;
|
|
|
|
} else {
|
|
|
|
rc -= msg->msg_rwptr->iov_len;
|
2004-08-03 01:24:00 +04:00
|
|
|
(msg->msg_rwnum)--;
|
2004-07-15 17:51:40 +04:00
|
|
|
(msg->msg_rwptr)++;
|
2004-08-03 01:24:00 +04:00
|
|
|
if(0 == msg->msg_rwnum) {
|
2007-04-12 09:01:29 +04:00
|
|
|
assert( 0 == rc );
|
2004-07-15 17:51:40 +04:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
2007-04-12 09:01:29 +04:00
|
|
|
} while(1);
|
2004-08-03 01:24:00 +04:00
|
|
|
}
|
2005-03-14 23:57:21 +03:00
|
|
|
return true;
|
2004-08-03 01:24:00 +04:00
|
|
|
}
|
|
|
|
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
/**
|
|
|
|
* Process a completed message.
|
|
|
|
*/
|
|
|
|
|
|
|
|
void mca_oob_tcp_msg_recv_complete(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer)
|
|
|
|
{
|
|
|
|
switch(msg->msg_hdr.msg_type) {
|
|
|
|
case MCA_OOB_TCP_IDENT:
|
|
|
|
mca_oob_tcp_msg_ident(msg,peer);
|
|
|
|
break;
|
2004-09-08 21:02:24 +04:00
|
|
|
case MCA_OOB_TCP_PING:
|
|
|
|
mca_oob_tcp_msg_ping(msg,peer);
|
|
|
|
break;
|
|
|
|
case MCA_OOB_TCP_DATA:
|
|
|
|
mca_oob_tcp_msg_data(msg,peer);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
break;
|
|
|
|
default:
|
2007-07-20 06:34:29 +04:00
|
|
|
opal_output(0, "%s mca_oob_tcp_msg_recv_complete: invalid message type: %d from peer %s\n",
|
|
|
|
ORTE_NAME_PRINT(orte_process_info.my_name), msg->msg_hdr.msg_type,
|
|
|
|
ORTE_NAME_PRINT(&peer->peer_name));
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
MCA_OOB_TCP_MSG_RETURN(msg);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
Bring over the update to terminate orteds that are generated by a dynamic spawn such as comm_spawn. This introduces the concept of a job "family" - i.e., jobs that have a parent/child relationship. Comm_spawn'ed jobs have a parent (the one that spawned them). We track that relationship throughout the lineage - i.e., if a comm_spawned job in turn calls comm_spawn, then it has a parent (the one that spawned it) and a "root" job (the original job that started things).
Accordingly, there are new APIs to the name service to support the ability to get a job's parent, root, immediate children, and all its descendants. In addition, the terminate_job, terminate_orted, and signal_job APIs for the PLS have been modified to accept attributes that define the extent of their actions. For example, doing a "terminate_job" with an attribute of ORTE_NS_INCLUDE_DESCENDANTS will terminate the given jobid AND all jobs that descended from it.
I have tested this capability on a MacBook under rsh, Odin under SLURM, and LANL's Flash (bproc). It worked successfully on non-MPI jobs (both simple and including a spawn), and MPI jobs (again, both simple and with a spawn).
This commit was SVN r12597.
2006-11-14 22:34:59 +03:00
|
|
|
* Process an ident message. In this case, we insist that the two process names
|
|
|
|
* exactly match - hence, we use the orte_ns.compare_fields function, which
|
|
|
|
* checks each field in a literal manner (i.e., no wildcards).
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
*/
|
|
|
|
|
|
|
|
static void mca_oob_tcp_msg_ident(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer)
|
|
|
|
{
|
2005-03-14 23:57:21 +03:00
|
|
|
orte_process_name_t src = msg->msg_hdr.msg_src;
|
|
|
|
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_lock);
|
Bring over the update to terminate orteds that are generated by a dynamic spawn such as comm_spawn. This introduces the concept of a job "family" - i.e., jobs that have a parent/child relationship. Comm_spawn'ed jobs have a parent (the one that spawned them). We track that relationship throughout the lineage - i.e., if a comm_spawned job in turn calls comm_spawn, then it has a parent (the one that spawned it) and a "root" job (the original job that started things).
Accordingly, there are new APIs to the name service to support the ability to get a job's parent, root, immediate children, and all its descendants. In addition, the terminate_job, terminate_orted, and signal_job APIs for the PLS have been modified to accept attributes that define the extent of their actions. For example, doing a "terminate_job" with an attribute of ORTE_NS_INCLUDE_DESCENDANTS will terminate the given jobid AND all jobs that descended from it.
I have tested this capability on a MacBook under rsh, Odin under SLURM, and LANL's Flash (bproc). It worked successfully on non-MPI jobs (both simple and including a spawn), and MPI jobs (again, both simple and with a spawn).
This commit was SVN r12597.
2006-11-14 22:34:59 +03:00
|
|
|
if (orte_ns.compare_fields(ORTE_NS_CMP_ALL, &peer->peer_name, &src) != ORTE_EQUAL) {
|
2005-07-19 16:25:19 +04:00
|
|
|
orte_hash_table_remove_proc(&mca_oob_tcp_component.tcp_peers, &peer->peer_name);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
peer->peer_name = src;
|
2005-07-19 16:25:19 +04:00
|
|
|
orte_hash_table_set_proc(&mca_oob_tcp_component.tcp_peers, &peer->peer_name, peer);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
}
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_lock);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
}
|
|
|
|
|
2004-09-08 21:02:24 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* Process a ping message.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void mca_oob_tcp_msg_ping(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer)
|
|
|
|
{
|
|
|
|
/* for now - we dont do anything - may want to send back a response at some poing */
|
|
|
|
}
|
|
|
|
|
|
|
|
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
/*
|
|
|
|
* Progress a completed recv:
|
|
|
|
* (1) signal a posted recv as complete
|
|
|
|
* (2) queue an unexpected message in the recv list
|
|
|
|
*/
|
|
|
|
|
2004-09-08 21:02:24 +04:00
|
|
|
static void mca_oob_tcp_msg_data(mca_oob_tcp_msg_t* msg, mca_oob_tcp_peer_t* peer)
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
{
|
|
|
|
/* attempt to match unexpected message to a posted recv */
|
|
|
|
mca_oob_tcp_msg_t* post;
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_match_lock);
|
2004-11-20 22:12:43 +03:00
|
|
|
|
2005-10-21 02:06:11 +04:00
|
|
|
/* match msg against posted receives */
|
2005-03-22 03:31:17 +03:00
|
|
|
post = mca_oob_tcp_msg_match_post(&peer->peer_name, msg->msg_hdr.msg_tag);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
if(NULL != post) {
|
2005-10-21 02:06:11 +04:00
|
|
|
|
2007-07-20 05:34:02 +04:00
|
|
|
if(NULL == post->msg_uiov || 0 == post->msg_ucnt) {
|
|
|
|
opal_output(0, "msg_data returning bad param");
|
|
|
|
post->msg_rc = ORTE_ERR_BAD_PARAM;
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
} else {
|
|
|
|
/* copy msg data into posted recv */
|
2007-07-20 05:34:02 +04:00
|
|
|
if (post->msg_flags & ORTE_RML_ALLOC) msg->msg_flags |= ORTE_RML_ALLOC;
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
post->msg_rc = mca_oob_tcp_msg_copy(msg, post->msg_uiov, post->msg_ucnt);
|
2007-07-20 05:34:02 +04:00
|
|
|
if(post->msg_flags & ORTE_RML_TRUNC) {
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
int i, size = 0;
|
|
|
|
for(i=1; i<msg->msg_rwcnt+1; i++)
|
|
|
|
size += msg->msg_rwiov[i].iov_len;
|
|
|
|
post->msg_rc = size;
|
|
|
|
}
|
|
|
|
}
|
2005-10-21 02:06:11 +04:00
|
|
|
|
2007-07-20 05:34:02 +04:00
|
|
|
if(post->msg_flags & ORTE_RML_PEEK) {
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
/* will need message for actual receive */
|
2006-07-18 01:51:50 +04:00
|
|
|
opal_list_append(&mca_oob_tcp_component.tcp_msg_recv, &msg->super.super);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
} else {
|
|
|
|
MCA_OOB_TCP_MSG_RETURN(msg);
|
|
|
|
}
|
2004-09-30 19:09:29 +04:00
|
|
|
mca_oob_tcp_component.tcp_match_count++;
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_match_lock);
|
2004-09-30 19:09:29 +04:00
|
|
|
|
2007-07-20 05:34:02 +04:00
|
|
|
if(post->msg_flags & ORTE_RML_PERSISTENT) {
|
2005-10-21 02:06:11 +04:00
|
|
|
post->msg_cbfunc(
|
|
|
|
post->msg_rc,
|
|
|
|
&peer->peer_name,
|
|
|
|
post->msg_uiov,
|
|
|
|
post->msg_ucnt,
|
|
|
|
post->msg_hdr.msg_tag,
|
|
|
|
post->msg_cbdata);
|
|
|
|
} else {
|
|
|
|
mca_oob_tcp_msg_complete(post, &peer->peer_name);
|
|
|
|
}
|
2004-09-30 19:09:29 +04:00
|
|
|
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_oob_tcp_component.tcp_match_lock);
|
2004-09-30 19:09:29 +04:00
|
|
|
if(--mca_oob_tcp_component.tcp_match_count == 0)
|
2005-07-04 02:45:48 +04:00
|
|
|
opal_condition_signal(&mca_oob_tcp_component.tcp_match_cond);
|
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_match_lock);
|
2004-09-30 19:09:29 +04:00
|
|
|
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
} else {
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_append(&mca_oob_tcp_component.tcp_msg_recv, (opal_list_item_t*)msg);
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_oob_tcp_component.tcp_match_lock);
|
Not as bad as this all may look. Tim and I made a significant change to the way we handle the startup of the oob, the seed, etc. We have made it backwards-compatible so that mpirun2 and singleton operations remain working. We had to adjust the name server and gpr as well, plus the process_info structure.
This also includes a checkpoint update to openmpi.c and ompid.c. I have re-enabled the ompid compile.
This latter raises an important point. The trunk compiles the programs like ompid just fine under Linux. It also does just fine for OSX under the dynamic libraries. However, we are seeing errors when compiling under OSX for the static case - the linker seems to have trouble resolving some variable names, even though linker diagnostics show the variables as being defined. Thus, a warning to Mac users that you may have to locally turn things off if you are trying to do static compiles. We ask, however, that you don't commit those changes that turn things off for everyone else - instead, let's try to figure out why the static compile is having a problem, and let everyone else continue to work.
Thanks
Ralph
This commit was SVN r2534.
2004-09-08 07:59:06 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-08-03 01:24:00 +04:00
|
|
|
* Called to copy the results of a message into user supplied iovec array.
|
|
|
|
* @param msg (IN) Message send that is in progress.
|
|
|
|
* @param iov (IN) Iovec array of user supplied buffers.
|
|
|
|
* @retval count Number of elements in iovec array.
|
|
|
|
*/
|
|
|
|
|
2004-08-13 02:41:42 +04:00
|
|
|
int mca_oob_tcp_msg_copy(mca_oob_tcp_msg_t* msg, struct iovec* iov, int count)
|
2004-08-03 01:24:00 +04:00
|
|
|
{
|
2007-07-20 05:34:02 +04:00
|
|
|
int i, ret = 0;
|
|
|
|
unsigned char* src_ptr = (unsigned char*) msg->msg_rwbuf;
|
2006-04-27 18:27:02 +04:00
|
|
|
size_t src_len = msg->msg_hdr.msg_size;
|
2007-07-20 05:34:02 +04:00
|
|
|
|
|
|
|
for (i = 0 ; i < count ; i++) {
|
|
|
|
if ((msg->msg_flags & ORTE_RML_ALLOC) && (i == count - 1)) {
|
|
|
|
if (i == 0) {
|
|
|
|
iov[i].iov_base = src_ptr;
|
|
|
|
iov[i].iov_len = src_len;
|
|
|
|
msg->msg_rwbuf = NULL;
|
|
|
|
} else {
|
|
|
|
iov[i].iov_base = malloc(src_len);
|
|
|
|
iov[i].iov_len = src_len;
|
|
|
|
memcpy(iov[i].iov_base, src_ptr, src_len);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (iov[i].iov_len > src_len) {
|
|
|
|
memcpy(iov[i].iov_base, src_ptr, src_len);
|
|
|
|
iov[i].iov_len = src_len;
|
|
|
|
} else {
|
|
|
|
memcpy(iov[i].iov_base, src_ptr, iov[i].iov_len);
|
2004-08-03 01:24:00 +04:00
|
|
|
}
|
|
|
|
}
|
2007-07-20 05:34:02 +04:00
|
|
|
|
|
|
|
ret += iov[i].iov_len;
|
|
|
|
src_len -= iov[i].iov_len;
|
|
|
|
src_ptr += iov[i].iov_len;
|
|
|
|
|
|
|
|
if (0 == src_len) break;
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
2007-07-20 05:34:02 +04:00
|
|
|
|
|
|
|
return ret;
|
2004-07-15 17:51:40 +04:00
|
|
|
}
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-08-03 01:24:00 +04:00
|
|
|
* Match name to a message that has been received asynchronously (unexpected).
|
|
|
|
*
|
|
|
|
* @param name (IN) Name associated with peer or wildcard to match first posted recv.
|
|
|
|
* @return msg Matched message or NULL.
|
|
|
|
*
|
|
|
|
* Note - this routine requires the caller to be holding the module lock.
|
|
|
|
*/
|
|
|
|
|
2005-03-14 23:57:21 +03:00
|
|
|
mca_oob_tcp_msg_t* mca_oob_tcp_msg_match_recv(orte_process_name_t* name, int tag)
|
2004-08-03 01:24:00 +04:00
|
|
|
{
|
|
|
|
mca_oob_tcp_msg_t* msg;
|
2005-07-03 20:22:16 +04:00
|
|
|
for(msg = (mca_oob_tcp_msg_t*) opal_list_get_first(&mca_oob_tcp_component.tcp_msg_recv);
|
|
|
|
msg != (mca_oob_tcp_msg_t*) opal_list_get_end(&mca_oob_tcp_component.tcp_msg_recv);
|
|
|
|
msg = (mca_oob_tcp_msg_t*) opal_list_get_next(msg)) {
|
2004-08-03 01:24:00 +04:00
|
|
|
|
Bring over the update to terminate orteds that are generated by a dynamic spawn such as comm_spawn. This introduces the concept of a job "family" - i.e., jobs that have a parent/child relationship. Comm_spawn'ed jobs have a parent (the one that spawned them). We track that relationship throughout the lineage - i.e., if a comm_spawned job in turn calls comm_spawn, then it has a parent (the one that spawned it) and a "root" job (the original job that started things).
Accordingly, there are new APIs to the name service to support the ability to get a job's parent, root, immediate children, and all its descendants. In addition, the terminate_job, terminate_orted, and signal_job APIs for the PLS have been modified to accept attributes that define the extent of their actions. For example, doing a "terminate_job" with an attribute of ORTE_NS_INCLUDE_DESCENDANTS will terminate the given jobid AND all jobs that descended from it.
I have tested this capability on a MacBook under rsh, Odin under SLURM, and LANL's Flash (bproc). It worked successfully on non-MPI jobs (both simple and including a spawn), and MPI jobs (again, both simple and with a spawn).
This commit was SVN r12597.
2006-11-14 22:34:59 +03:00
|
|
|
if(ORTE_EQUAL == orte_dss.compare(name, &msg->msg_peer, ORTE_NAME)) {
|
2005-03-14 23:57:21 +03:00
|
|
|
if (tag == msg->msg_hdr.msg_tag) {
|
2004-08-03 01:24:00 +04:00
|
|
|
return msg;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2004-08-10 03:07:53 +04:00
|
|
|
/*
|
2004-08-03 01:24:00 +04:00
|
|
|
* Match name to a posted recv request.
|
|
|
|
*
|
|
|
|
* @param name (IN) Name associated with peer or wildcard to match first posted recv.
|
|
|
|
* @return msg Matched message or NULL.
|
|
|
|
*
|
|
|
|
* Note - this routine requires the caller to be holding the module lock.
|
|
|
|
*/
|
|
|
|
|
2005-03-22 03:31:17 +03:00
|
|
|
mca_oob_tcp_msg_t* mca_oob_tcp_msg_match_post(orte_process_name_t* name, int tag)
|
2004-08-03 01:24:00 +04:00
|
|
|
{
|
|
|
|
mca_oob_tcp_msg_t* msg;
|
2005-07-03 20:22:16 +04:00
|
|
|
for(msg = (mca_oob_tcp_msg_t*) opal_list_get_first(&mca_oob_tcp_component.tcp_msg_post);
|
|
|
|
msg != (mca_oob_tcp_msg_t*) opal_list_get_end(&mca_oob_tcp_component.tcp_msg_post);
|
|
|
|
msg = (mca_oob_tcp_msg_t*) opal_list_get_next(msg)) {
|
2004-08-03 01:24:00 +04:00
|
|
|
|
Bring over the update to terminate orteds that are generated by a dynamic spawn such as comm_spawn. This introduces the concept of a job "family" - i.e., jobs that have a parent/child relationship. Comm_spawn'ed jobs have a parent (the one that spawned them). We track that relationship throughout the lineage - i.e., if a comm_spawned job in turn calls comm_spawn, then it has a parent (the one that spawned it) and a "root" job (the original job that started things).
Accordingly, there are new APIs to the name service to support the ability to get a job's parent, root, immediate children, and all its descendants. In addition, the terminate_job, terminate_orted, and signal_job APIs for the PLS have been modified to accept attributes that define the extent of their actions. For example, doing a "terminate_job" with an attribute of ORTE_NS_INCLUDE_DESCENDANTS will terminate the given jobid AND all jobs that descended from it.
I have tested this capability on a MacBook under rsh, Odin under SLURM, and LANL's Flash (bproc). It worked successfully on non-MPI jobs (both simple and including a spawn), and MPI jobs (again, both simple and with a spawn).
This commit was SVN r12597.
2006-11-14 22:34:59 +03:00
|
|
|
if(ORTE_EQUAL == orte_dss.compare(name, &msg->msg_peer, ORTE_NAME)) {
|
2005-03-14 23:57:21 +03:00
|
|
|
if (msg->msg_hdr.msg_tag == tag) {
|
2007-07-20 05:34:02 +04:00
|
|
|
if((msg->msg_flags & ORTE_RML_PERSISTENT) == 0) {
|
2006-07-18 01:51:50 +04:00
|
|
|
opal_list_remove_item(&mca_oob_tcp_component.tcp_msg_post, &msg->super.super);
|
2004-08-03 01:24:00 +04:00
|
|
|
}
|
2005-10-21 02:06:11 +04:00
|
|
|
return msg;
|
2004-08-03 01:24:00 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|