Skip to content

handle errors gracefuly to prevent SEGV #13194

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

Closed
wants to merge 1 commit into from
Closed

handle errors gracefuly to prevent SEGV #13194

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
14 changes: 11 additions & 3 deletions ompi/mca/coll/ucc/coll_ucc_module.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -150,7 +150,7 @@ static ucc_status_t oob_allgather_test(void *req)
size_t msglen = oob_req->msglen;
int probe_count = 5;
int rank, size, sendto, recvfrom, recvdatafrom,
senddatafrom, completed, probe;
senddatafrom, completed, probe, rc;

size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
Expand All @@ -175,10 +175,16 @@ static ucc_status_t oob_allgather_test(void *req)
senddatafrom = (rank - oob_req->iter + size) % size;
tmprecv = (char*)oob_req->rbuf + (ptrdiff_t)recvdatafrom * (ptrdiff_t)msglen;
tmpsend = (char*)oob_req->rbuf + (ptrdiff_t)senddatafrom * (ptrdiff_t)msglen;
MCA_PML_CALL(isend(tmpsend, msglen, MPI_BYTE, sendto, MCA_COLL_BASE_TAG_UCC,
rc = MCA_PML_CALL(isend(tmpsend, msglen, MPI_BYTE, sendto, MCA_COLL_BASE_TAG_UCC,
MCA_PML_BASE_SEND_STANDARD, comm, &oob_req->reqs[0]));
MCA_PML_CALL(irecv(tmprecv, msglen, MPI_BYTE, recvfrom,
if (OMPI_SUCCESS != rc) {
return rc;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not sure if conversion is safe here, return value of this function is ucc_status_t and might be not compatible with ompi error codes. If rc is positive number then it's actually not an error from ucc perspective since all errors in ucc are negative. Maybe just to make it safe the function returns UCC_ERR_NO_MESSAGE in case of error.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is correct. The macro ends up calling the .pml_isend field in the PML struct, which in the case of UCX points to mca_pml_ucx_isend, a function that returns OMPI errors.

}
rc = MCA_PML_CALL(irecv(tmprecv, msglen, MPI_BYTE, recvfrom,
MCA_COLL_BASE_TAG_UCC, comm, &oob_req->reqs[1]));
if (OMPI_SUCCESS != rc) {
return rc;
}
}
probe = 0;
do {
Expand Down Expand Up @@ -206,6 +212,8 @@ static ucc_status_t oob_allgather(void *sbuf, void *rbuf, size_t msglen,
oob_req->msglen = msglen;
oob_req->oob_coll_ctx = oob_coll_ctx;
oob_req->iter = 0;
oob_req->reqs[0] = NULL;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Technically I think this should be MPI_REQUEST_NULL.

oob_req->reqs[1] = NULL;
*req = oob_req;
return UCC_OK;
}
Expand Down
Loading