Skip to content

v3.x: If we lose connection to the server after initiating a send/recv in P… #3230

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Mar 31, 2017
Merged

v3.x: If we lose connection to the server after initiating a send/recv in P… #3230

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
55022b4
If we lose connection to the server after initiating a send/recv in P…
Mar 23, 2017
a358c7b
Use the correct callback data - the callback function was expecting a…
Mar 29, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 4 additions & 12 deletions opal/mca/pmix/pmix2x/pmix/src/client/pmix_client.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -492,7 +492,7 @@ PMIX_EXPORT pmix_status_t PMIx_Abort(int flag, const char msg[],
pmix_buffer_t *bfr;
pmix_cmd_t cmd = PMIX_ABORT_CMD;
pmix_status_t rc;
pmix_ptl_sr_t cb;
volatile bool active;

pmix_output_verbose(2, pmix_globals.debug_output,
"pmix:client abort called");
Expand Down Expand Up @@ -541,23 +541,15 @@ PMIX_EXPORT pmix_status_t PMIx_Abort(int flag, const char msg[],
}
}

/* create a callback object as we need to pass it to the
* recv routine so we know which callback to use when
* the return message is recvd */
PMIX_CONSTRUCT(&cb, pmix_ptl_sr_t);
cb.active = true;
cb.cbfunc = wait_cbfunc;

/* send to the server */
active = true;
if (PMIX_SUCCESS != (rc = pmix_ptl.send_recv(&pmix_client_globals.myserver, bfr,
wait_cbfunc, &cb))){
PMIX_DESTRUCT(&cb);
wait_cbfunc, (void*)&active))){
return rc;
}

/* wait for the release */
PMIX_WAIT_FOR_COMPLETION(cb.active);
PMIX_DESTRUCT(&cb);
PMIX_WAIT_FOR_COMPLETION(active);
return PMIX_SUCCESS;
}

Expand Down
20 changes: 20 additions & 0 deletions opal/mca/pmix/pmix2x/pmix/src/mca/ptl/base/ptl_base_sendrecv.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,9 @@ static void lost_connection(pmix_peer_t *peer, pmix_status_t err)
pmix_regevents_info_t *reginfoptr, *regnext;
pmix_peer_events_info_t *pr, *pnext;
pmix_rank_info_t *info, *pinfo;
pmix_ptl_posted_recv_t *rcv;
pmix_buffer_t buf;
pmix_ptl_hdr_t hdr;

/* stop all events */
if (peer->recv_ev_active) {
Expand Down Expand Up @@ -143,6 +146,23 @@ static void lost_connection(pmix_peer_t *peer, pmix_status_t err)
pmix_globals.connected = false;
/* set the public error status */
err = PMIX_ERR_LOST_CONNECTION_TO_SERVER;
/* it is possible that we have sendrecv's in progress where
* we are waiting for a response to arrive. Since we have
* lost connection to the server, that will never happen.
* Thus, to preclude any chance of hanging, cycle thru
* the list of posted recvs and complete any that are
* the return call from a sendrecv - i.e., any that are
* waiting on dynamic tags */
PMIX_CONSTRUCT(&buf, pmix_buffer_t);
hdr.nbytes = 0; // initialize the hdr to something safe
PMIX_LIST_FOREACH(rcv, &pmix_ptl_globals.posted_recvs, pmix_ptl_posted_recv_t) {
if (UINT_MAX != rcv->tag && NULL != rcv->cbfunc) {
/* construct and load the buffer */
hdr.tag = rcv->tag;
rcv->cbfunc(pmix_globals.mypeer, &hdr, &buf, rcv->cbdata);
}
}
PMIX_DESTRUCT(&buf);
}
PMIX_REPORT_EVENT(err, _notify_complete);
}
Expand Down