NetBSD-Bugs archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

bin/50350: rump/rumpkern/t_sp/stress_{long,short} fail on Core 2 Quad



>Number:         50350
>Category:       bin
>Synopsis:       rump/rumpkern/t_sp/stress_{long,short} fail on Core 2 Quad
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    bin-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Oct 21 08:15:01 +0000 2015
>Originator:     Andreas Gustafsson
>Release:        NetBSD-current > 2013-08-14
>Organization:

>Environment:
System: NetBSD
Architecture: x86_64
Machine: amd64
>Description:

I'm running the ATF tests on a physical amd64 host which I recently
upgraded from an old single-core AMD Athlon64 to a slightly less
old Intel Core 2 Quad Q6600 quad-core machine.

When I did that, the stress_short and stress_long test cases of the
rump/rumpkern/t_sp test, which had reliably succeeded on the
single-core machine, started failing in every test run.

I also tried running the t_sp test on two other PCs, both of the same
model, Compaq DC7900, but with different CPUs, one being a dual-core
E8400 and the other a quad-core Q6600, and it only failed on the
quad-core one.

Bisection shows that the tests pass on the Q6600 for source dates
older 2013-08-14, when the following change was committed:

  src/lib/librumpuser/rumpuser.c 1.54

    Change the default value of rump kernels CPUs to 2.  It used to be
    the number of host cores, but that value is overkill for most uses,
    especially with massively multicore hosts.  Dozens of useless virtual
    CPUs are relatively speaking expensive in terms of bootstrap time and
    memory footprint.  On the other end of the spectrum, defaulting to 2
    might shake out some bugs from the qemu test runs.

It looks like defaulting to 2 managed to shake out some bugs from
bare-metal test runs instead of qemu ones.

Log output from one failing test is here:

  http://www.gson.org/netbsd/bugs/build/amd64-baremetal/2015/2015.10.09.17.21.45/test.html#rump_rumpkern_t_sp_stress_long

>How-To-Repeat:

On a quad-core amd64 machine, run

  cd /usr/tests/rump/rumpkern
  atf-run t_sp | atf-report

>Fix:



Home | Main Index | Thread Index | Old Index