OpenBLAS related segfaults with Julia 1.7.0 on Linux

[carstenbauer]

I could produce the following segfault with fresh Julia 1.7.0 installs on two different HPC clusters (redhat enterprise) and a regular desktop machine (ubuntu). Note that while on one machine it would already segfault for N=5 on a different machine I had to set N>20. So try varying this if you can't reproduce.

(Julia started with multiple threads, e.g. julia -t 8)

julia> n = 1000;

julia> N = 20;

julia> Threads.@threads for i in 1:N
           A = randn(Float64, n, n); inv(A);
       end

signal (11): Segmentation fault
in expression starting at REPL[2]:1
dgetrf_parallel at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
dgetrf_parallel at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
dgetrf_parallel at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
dgetrf_parallel at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
dgetrf_parallel at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
dgetrf_64_ at /cm/shared/apps/pc2/EB-SW/software/Julia/1.7.0-linux-x86_64/bin/../lib/julia/libopenblas64_.so (unknown line)
getrf! at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lapack.jl:575
#lu!#146 at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lu.jl:81 [inlined]
lu!##kw at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lu.jl:81 [inlined]
#lu#153 at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lu.jl:279 [inlined]
lu at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lu.jl:278 [inlined]
lu at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/lu.jl:278 [inlined]
inv at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.7/LinearAlgebra/src/dense.jl:876
macro expansion at ./REPL[2]:2 [inlined]
JuliaLang/julia#40#threadsfor_fun at ./threadingconstructs.jl:85
JuliaLang/julia#40#threadsfor_fun at ./threadingconstructs.jl:52
unknown function (ip: 0x1554f0112d5f)
_jl_invoke at /buildworker/worker/package_linux64/build/src/gf.c:2247 [inlined]
jl_apply_generic at /buildworker/worker/package_linux64/build/src/gf.c:2429
jl_apply at /buildworker/worker/package_linux64/build/src/julia.h:1788 [inlined]
start_task at /buildworker/worker/package_linux64/build/src/task.c:877
Allocations: 4321396 (Pool: 4319432; Big: 1964); GC: 5
Segmentation fault (core dumped)

The segfault disappears when one sets BLAS.set_num_threads(1).

Likely related to https://github.com/JuliaLang/julia/issues/43301 and #886 (caused by the same issue?). However, not really a duplicate because these issues are all about macOS and also Float64 seems to be fine there.

(Let me note that this is the result of me trying to create a MWE. Originally I encountered this segfault and StackOverflowErrors (as in the issues linked above) as part of CI testing a private package with 1.6.4 and 1.7.0. See https://discourse.julialang.org/t/inv-causes-stack-overflow-on-julia-1-7-0-and-mac-os/72411/10). When switching back to 1.6.3 or 1.7.0-rc1 the issues went away.)

[gbaraldi]

Likely fixed by JuliaLang/julia#43300 since those are all similar issues. I get a StackOverFlow error but the new version will probably fix those issues.

[ViralBShah]

Closing as a dup of those other issues discussed above. Can reopen if those don't fix this one.