[bug] [BDX]Ubuntu 16.04.3 with kernel "4.10.0-32-generic_4.10.0-32.36~16.04.1_amd64" hang if set "Cores Enabled" to 2 for both CPU

[quanxian]

Description:
Customer report a Ubuntu kernel issue related to Intel BDX Core numbers setting
Ubuntu 16.04.3 with kernel "4.10.0-32-generic_4.10.0-32.36~16.04.1_amd64" hang if set "Cores Enabled" to 2 for both CPU.

We have reproduced this issue on 17.10 release.

With latest kernel, this issue disappears.

now we bisect this issue is fixed by commit 947134d9b00f (“x86/smpboot: Do not use smp_num_siblings in __max_logical_packages calculation”) which was merged in v4.15-rc4.

But after applying this commit to Ubuntu 16.04.3 kernel v4.13 code base or Linux upstream mainline kernel v4.13 code base, this issue still exists.

Target Release: 16.04 / 17.10

[Alice Liu (dxliu)]

Correct the fix commit info. But applying this patch to v4.13, the issue still exists. It may have more dependencies.

commit 947134d9b00f342415af7eddd42a5fce7262a1b9
Author: Prarit Bhargava <email address hidden>
Date: Mon Dec 4 11:45:21 2017 -0500

    x86/smpboot: Do not use smp_num_siblings in __max_logical_packages calculation

    Documentation/x86/topology.txt defines smp_num_siblings as "The number of
    threads in a core". Since commit bbb65d2d365e ("x86: use cpuid vector 0xb
    when available for detecting cpu topology") smp_num_siblings is the
    maximum number of threads in a core. If Simultaneous MultiThreading
    (SMT) is disabled on a system, smp_num_siblings is 2 and not 1 as
    expected.

    Use topology_max_smt_threads(), which contains the active numer of threads,
    in the __max_logical_packages calculation.

    On a single socket, single core, single thread system __max_smt_threads has
    not been updated when the __max_logical_packages calculation happens, so its
    zero which makes the package estimate fail. Initialize it to one, which is
    the minimum number of threads on a core.

    [ tglx: Folded the __max_smt_threads fix in ]

    Fixes: b4c0a7326f5d ("x86/smpboot: Fix __max_logical_packages estimate")
    Reported-by: Jakub Kicinski <email address hidden>
    Signed-off-by: Prarit Bhargava <<email address hidden>
    Signed-off-by: Thomas Gleixner <email address hidden>
    Tested-by: Jakub Kicinski <email address hidden>

[Thadeu Lima de Souza Cascardo (cascardo)]

As bionic already ships a v4.15 kernel, this should be fixed in bionic.

Can you verify?

Cascardo.

[Joseph Salisbury (jsalisbury)]

I built an Artful test kernel with a back port of commit 947134d9b00. The test kernel can be downloaded from:
http://kernel.ubuntu.com/~jsalisbury/lp1747134

Can you test this kernel and see if it resolves this bug?

Note, to test this kernel, you need to install both the linux-image and linux-image-extra .deb packages.

I'll also work on getting a back port and identify prereq commits for Xenial.

Thanks in advance!

[Joseph Salisbury (jsalisbury)]

A Xenial test kernel is now also available. Xenial required commit 70b8301f6 as a prerequisite. The xenial test kernel can be downloaded from:

http://kernel.ubuntu.com/~jsalisbury/lp1747134/xenial

[Andy Whitcroft (apw)]

This bug was nominated against a series that is no longer supported, ie artful. The bug task representing the artful nomination is being closed as Won't Fix.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.