Skip to content

v3.1.x: pml/ob1: Fix data corruption of MPI_BSEND #5436

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 1 commit into from
Aug 7, 2018
Merged

v3.1.x: pml/ob1: Fix data corruption of MPI_BSEND #5436

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
43 changes: 23 additions & 20 deletions ompi/mca/pml/ob1/pml_ob1_sendreq.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@
* Copyright (c) 2015 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2016 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* Copyright (c) 2018 FUJITSU LIMITED. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -417,29 +418,31 @@ int mca_pml_ob1_send_request_start_buffered(
des->des_cbfunc = mca_pml_ob1_rndv_completion;
des->des_cbdata = sendreq;

/* buffer the remainder of the message */
rc = mca_pml_base_bsend_request_alloc((ompi_request_t*)sendreq);
if( OPAL_UNLIKELY(OMPI_SUCCESS != rc) ) {
mca_bml_base_free(bml_btl, des);
return rc;
}
/* buffer the remainder of the message if it is not buffered yet */
if( OPAL_LIKELY(sendreq->req_send.req_addr == sendreq->req_send.req_base.req_addr) ) {
rc = mca_pml_base_bsend_request_alloc((ompi_request_t*)sendreq);
if( OPAL_UNLIKELY(OMPI_SUCCESS != rc) ) {
mca_bml_base_free(bml_btl, des);
return rc;
}

iov.iov_base = (IOVBASE_TYPE*)(((unsigned char*)sendreq->req_send.req_addr) + max_data);
iov.iov_len = max_data = sendreq->req_send.req_bytes_packed - max_data;
iov.iov_base = (IOVBASE_TYPE*)(((unsigned char*)sendreq->req_send.req_addr) + max_data);
iov.iov_len = max_data = sendreq->req_send.req_bytes_packed - max_data;

if((rc = opal_convertor_pack( &sendreq->req_send.req_base.req_convertor,
&iov,
&iov_count,
&max_data)) < 0) {
mca_bml_base_free(bml_btl, des);
return rc;
}
if((rc = opal_convertor_pack( &sendreq->req_send.req_base.req_convertor,
&iov,
&iov_count,
&max_data)) < 0) {
mca_bml_base_free(bml_btl, des);
return rc;
}

/* re-init convertor for packed data */
opal_convertor_prepare_for_send( &sendreq->req_send.req_base.req_convertor,
&(ompi_mpi_byte.dt.super),
sendreq->req_send.req_bytes_packed,
sendreq->req_send.req_addr );
/* re-init convertor for packed data */
opal_convertor_prepare_for_send( &sendreq->req_send.req_base.req_convertor,
&(ompi_mpi_byte.dt.super),
sendreq->req_send.req_bytes_packed,
sendreq->req_send.req_addr );
}

/* wait for ack and completion */
sendreq->req_state = 2;
Expand Down
11 changes: 11 additions & 0 deletions ompi/mca/pml/ob1/pml_ob1_sendreq.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@
* Copyright (c) 2011-2012 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2011-2016 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2018 FUJITSU LIMITED. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
Expand Down Expand Up @@ -468,6 +469,16 @@ mca_pml_ob1_send_request_start_seq (mca_pml_ob1_send_request_t* sendreq, mca_bml
if( OPAL_LIKELY(OMPI_ERR_OUT_OF_RESOURCE != rc) )
return rc;
}
if(MCA_PML_BASE_SEND_BUFFERED == sendreq->req_send.req_send_mode &&
sendreq->req_send.req_addr == sendreq->req_send.req_base.req_addr) {
/* in the buffered mode, the send buffer must be saved to
* the attached buffer before returning it to the user */
int rc;
rc = mca_pml_base_bsend_request_start((ompi_request_t*)sendreq);
if(OMPI_SUCCESS != rc){
return rc;
}
}
add_request_to_send_pending(sendreq, MCA_PML_OB1_SEND_PENDING_START, true);

return OMPI_SUCCESS;
Expand Down