lkml.org 
[lkml]   [2017]   [Oct]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    From
    Date
    Subject[PATCH 3.16 005/192] sched/topology: Fix overlapping sched_group_mask
    3.16.49-rc1 review patch.  If anyone has any objections, please let me know.

    ------------------

    From: Peter Zijlstra <peterz@infradead.org>

    commit 73bb059f9b8a00c5e1bf2f7ca83138c05d05e600 upstream.

    The point of sched_group_mask is to select those CPUs from
    sched_group_cpus that can actually arrive at this balance domain.

    The current code gets it wrong, as can be readily demonstrated with a
    topology like:

    node 0 1 2 3
    0: 10 20 30 20
    1: 20 10 20 30
    2: 30 20 10 20
    3: 20 30 20 10

    Where (for example) domain 1 on CPU1 ends up with a mask that includes
    CPU0:

    [] CPU1 attaching sched-domain:
    [] domain 0: span 0-2 level NUMA
    [] groups: 1 (mask: 1), 2, 0
    [] domain 1: span 0-3 level NUMA
    [] groups: 0-2 (mask: 0-2) (cpu_capacity: 3072), 0,2-3 (cpu_capacity: 3072)

    This causes sched_balance_cpu() to compute the wrong CPU and
    consequently should_we_balance() will terminate early resulting in
    missed load-balance opportunities.

    The fixed topology looks like:

    [] CPU1 attaching sched-domain:
    [] domain 0: span 0-2 level NUMA
    [] groups: 1 (mask: 1), 2, 0
    [] domain 1: span 0-3 level NUMA
    [] groups: 0-2 (mask: 1) (cpu_capacity: 3072), 0,2-3 (cpu_capacity: 3072)

    (note: this relies on OVERLAP domains to always have children, this is
    true because the regular topology domains are still here -- this is
    before degenerate trimming)

    Debugged-by: Lauro Ramos Venancio <lvenanci@redhat.com>
    Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
    Cc: Linus Torvalds <torvalds@linux-foundation.org>
    Cc: Mike Galbraith <efault@gmx.de>
    Cc: Peter Zijlstra <peterz@infradead.org>
    Cc: Thomas Gleixner <tglx@linutronix.de>
    Cc: linux-kernel@vger.kernel.org
    Fixes: e3589f6c81e4 ("sched: Allow for overlapping sched_domain spans")
    Signed-off-by: Ingo Molnar <mingo@kernel.org>
    [bwh: Backported to 3.16:
    - Use span, not sg_span
    - Adjust filename context]
    Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
    ---
    kernel/sched/core.c | 18 +++++++++++++++++-
    1 file changed, 17 insertions(+), 1 deletion(-)

    --- a/kernel/sched/core.c
    +++ b/kernel/sched/core.c
    @@ -5775,6 +5775,8 @@ enum s_alloc {
    * and our sibling sd spans will be empty. Domains should always include the
    * cpu they're built on, so check that.
    *
    + * Only CPUs that can arrive at this group should be considered to continue
    + * balancing.
    */
    static void build_group_mask(struct sched_domain *sd, struct sched_group *sg)
    {
    @@ -5785,11 +5787,24 @@ static void build_group_mask(struct sche

    for_each_cpu(i, span) {
    sibling = *per_cpu_ptr(sdd->sd, i);
    - if (!cpumask_test_cpu(i, sched_domain_span(sibling)))
    +
    + /*
    + * Can happen in the asymmetric case, where these siblings are
    + * unused. The mask will not be empty because those CPUs that
    + * do have the top domain _should_ span the domain.
    + */
    + if (!sibling->child)
    + continue;
    +
    + /* If we would not end up here, we can't continue from here */
    + if (!cpumask_equal(span, sched_domain_span(sibling->child)))
    continue;

    cpumask_set_cpu(i, sched_group_mask(sg));
    }
    +
    + /* We must not have empty masks here */
    + WARN_ON_ONCE(cpumask_empty(sched_group_mask(sg)));
    }

    /*
    \
     
     \ /
      Last update: 2017-10-09 15:56    [W:5.948 / U:0.032 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site