Skip to content

Fix v3.0.x ob1 issues #3852

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
Jul 17, 2017
Merged

Fix v3.0.x ob1 issues #3852

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
bc1c56b
pml/ob1: do not access fragment after calling btl rget
hjelmn Jul 11, 2017
254c46b
btl/vader: work around ob1 pending fragment bug
hjelmn Jul 11, 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
5 changes: 3 additions & 2 deletions ompi/mca/pml/ob1/pml_ob1_recvreq.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@
* Copyright (c) 2008 UT-Battelle, LLC. All rights reserved.
* Copyright (c) 2011 Sandia National Laboratories. All rights reserved.
* Copyright (c) 2012-2015 NVIDIA Corporation. All rights reserved.
* Copyright (c) 2011-2016 Los Alamos National Security, LLC. All rights
* Copyright (c) 2011-2017 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2012 FUJITSU LIMITED. All rights reserved.
* Copyright (c) 2014-2016 Research Organization for Information Science
Expand Down Expand Up @@ -753,13 +753,14 @@ void mca_pml_ob1_recv_request_progress_rget( mca_pml_ob1_recv_request_t* recvreq
frag->rdma_length = bytes_remaining;
}

prev_sent = frag->rdma_length;

/* NTH: TODO -- handle error conditions gracefully */
rc = mca_pml_ob1_recv_request_get_frag(frag);
if (OMPI_SUCCESS != rc) {
break;
}

prev_sent = frag->rdma_length;
bytes_remaining -= prev_sent;
offset += prev_sent;
}
Expand Down
8 changes: 8 additions & 0 deletions opal/mca/btl/vader/btl_vader_send.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,6 +49,10 @@ int mca_btl_vader_send (struct mca_btl_base_module_t *btl,
return 1;
}

/* in order to work around a long standing ob1 bug (see #3845) we have to always
* make the callback. once this is fixed in ob1 we can restore the code below. */
frag->base.des_flags |= MCA_BTL_DES_SEND_ALWAYS_CALLBACK;

/* header (+ optional inline data) */
frag->hdr->len = total_size;
/* type of message, pt-2-pt, one-sided, etc */
Expand All @@ -69,6 +73,9 @@ int mca_btl_vader_send (struct mca_btl_base_module_t *btl,
return OPAL_SUCCESS;
}

return OPAL_SUCCESS;

#if 0
if ((frag->hdr->flags & MCA_BTL_VADER_FLAG_SINGLE_COPY) ||
!(frag->base.des_flags & MCA_BTL_DES_FLAGS_BTL_OWNERSHIP)) {
frag->base.des_flags |= MCA_BTL_DES_SEND_ALWAYS_CALLBACK;
Expand All @@ -79,4 +86,5 @@ int mca_btl_vader_send (struct mca_btl_base_module_t *btl,
/* data is gone (from the pml's perspective). frag callback/release will
happen later */
return 1;
#endif
}