Skip to content

btl/ofi: not sessions friendly for standard OFI EPs #13019

Closed
@hppritcha

Description

@hppritcha

I'm finding that the sessions_init_twice.c test is failing in the OFI BTL on my HPE slingshot11 system (has nothing to do with that in particular though).

[nid001382:149402] *** Process received signal ***
[nid001382:149402] Signal: Segmentation fault (11)
[nid001382:149402] Signal code: Address not mapped (1)
[nid001382:149402] Failing at address: 0x28
[nid001382:149402] [ 0] /lib64/libpthread.so.0(+0x16910)[0x150a22334910]
[nid001382:149402] [ 1] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xe0882)[0x150a22069882]
[nid001382:149402] [ 2] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xe097c)[0x150a2206997c]
[nid001382:149402] [ 3] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xe09ea)[0x150a220699ea]
[nid001382:149402] [ 4] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xe0bbc)[0x150a22069bbc]
[nid001382:149402] [ 5] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(mca_btl_ofi_alloc+0xa8)[0x150a22069e2b]
[nid001382:149402] [ 6] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(mca_btl_ofi_post_recvs+0x41)[0x150a22067683]
[nid001382:149402] [ 7] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xdc1f0)[0x150a220651f0]
[nid001382:149402] [ 8] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(+0xdaf5a)[0x150a22063f5a]
[nid001382:149402] [ 9] foobar/hpp/ompi/install_roci/lib/libopen-pal.so.0(mca_btl_base_select+0x204)[0x150a220464dd]
[nid001382:149402] [10] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(mca_bml_r2_component_init+0x29)[0x150a224dfb56]
[nid001382:149402] [11] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(mca_bml_base_init+0x101)[0x150a224dc90a]
[nid001382:149402] [12] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(+0x3cd75a)[0x150a2273375a]
[nid001382:149402] [13] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(mca_pml_base_select+0x2cc)[0x150a2272c130]
[nid001382:149402] [14] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(+0xc28f4)[0x150a224288f4]
[nid001382:149402] [15] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(ompi_mpi_instance_init+0x75)[0x150a224292cf]
[nid001382:149402] [16] foobar/hpp/ompi/install_roci/lib/libmpi.so.0(PMPI_Session_init+0x1e5)[0x150a224a9b8e]
[nid001382:149402] [17] ./sessions_init_twice[0x4015fc]
[nid001382:149402] [18] /lib64/libc.so.6(__libc_start_main+0xef)[0x150a2215c24d]
[nid001382:149402] [19] ./sessions_init_twice[0x40117a]
[nid001382:149402] *** End of error message ***

This is for standard OFI EPs, not scalable.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions