Ompi: bml/r2: closing already closed BTL in some cases.

Created on 19 Sep 2018  路  1Comment  路  Source: open-mpi/ompi

mca_bml_r2_del_procs() will do a double free if the BTL is presented on both btl_send and btl_rdma list, resulted in segfault in MPI_Finalize. (see both for loop at 620 and 637)

https://github.com/open-mpi/ompi/blob/dccf780546e9cb49066bd76121b01293aa6d4b97/ompi/mca/bml/r2/bml_r2.c#L605-L656

btl/uct and btl/ofi did a workaround on this issue by using hash table on del_proc. However, this will still be a problem for other BTL with both send and rdma capabilities. I just reproduced this on btl/openib. Should we fix this from r2 or the BTL has to implement their own workaround?

I can reproduce this on 4.0.0rc1.

Awaiting developer information master v4.0.x bug question

Most helpful comment

(github pro tip: use permalinks to paste in code from a github repo)

I would say that this should be fixed in r2.

>All comments

(github pro tip: use permalinks to paste in code from a github repo)

I would say that this should be fixed in r2.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

sap9433 picture sap9433  路  3Comments

dirk-schubert-arm picture dirk-schubert-arm  路  4Comments

pozdneev picture pozdneev  路  4Comments

egorgam picture egorgam  路  5Comments

ax3l picture ax3l  路  7Comments