MPI over mmaped region kills Kernel?

Intel® MPI Library

Get help with building, analyzing, optimizing, and scaling high-performance computing (HPC) applications.

MPI over mmaped region kills Kernel?

318 Views

Hi,

I use MPI (with Infiniband's RDMA enabled) over mmaped region. The size of mmap region is larger than the physical memory size, so I expect TLB is updated often, which may incur 'undefined' behavior of Infiniband's RDMA. The application kills the kernel (kernel panic), which is absolutely not acceptible (User application code never incur it)

I suspect Infiniband's RDMA capability which bypasses translation buffer of CPU and accesses physical memory directly as one of the reasons for corrupting OS 'somehow'.

1. Can RDMA capability occur such a problem I described?

2. I tried with 'I_MPI_DAPL_TRANSLATION_CACHE' disabled, but the issue is not resolved. (I don't see any message saying 'I_MPI_DAPL_TRANSALCTION_CACHE=0' even with 'I_MPI_DEBUG=100', am I miss-using the env. vars?)

many thanks for all :D

Link Copied

0 Replies

Community support is provided during standard business hours (Monday to Friday 7AM - 5PM PST). Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.