Skip to content

L
linux
Switch branch/tag
  1. 02 Aug, 2009 9 commits
    • Gregory Haskins's avatar
      sched: Enhance the pre/post scheduling logic · 3f029d3c
      Gregory Haskins authored 
      We currently have an explicit "needs_post" vtable method which
returns a stack variable for whether we should later run
post-schedule.  This leads to an awkward exchange of the
variable as it bubbles back up out of the context switch. Peter
Zijlstra observed that this information could be stored in the
run-queue itself instead of handled on the stack.

Therefore, we revert to the method of having context_switch
return void, and update an internal rq->post_schedule variable
when we require further processing.

In addition, we fix a race condition where we try to access
current->sched_class without holding the rq->lock.  This is
technically racy, as the sched-class could change out from
under us.  Instead, we reference the per-rq post_schedule
variable with the runqueue unlocked, but with preemption
disabled to see if we need to reacquire the rq->lock.

Finally, we clean the code up slightly by removing the #ifdef
CONFIG_SMP conditionals from the schedule() call, and implement
some inline helper functions instead.

This patch passes checkpatch, and rt-migrate.
Signed-off-by: default avatarGregory Haskins <ghaskins@novell.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090729150422.17691.55590.stgit@dev.haskins.net>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      3f029d3c
    • Steven Rostedt's avatar
      sched: Add new prio to cpupri before removing old prio · c3a2ae3d
      Steven Rostedt authored 
      We need to add the new prio to the cpupri accounting before
removing the old prio. This is because removing the old prio
first will open a race window where the cpu will be removed
from pri_active. In this case the cpu will not be visible for
RT push and pulls. This could cause a RT task to not migrate
appropriately, and create a very large latency.

This bug was found with the use of ftrace sched events and
trace_printk.
Signed-off-by: default avatarSteven Rostedt <srostedt@redhat.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090729042526.438281019@goodmis.org>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      c3a2ae3d
    • Steven Rostedt's avatar
      sched: Check for pushing rt tasks after all scheduling · da19ab51
      Steven Rostedt authored 
      The current method for pushing RT tasks after scheduling only
happens after a context switch. But we found cases where a task
is set up on a run queue to be pushed but the push never
happens because the schedule chooses the same task.

This bug was found with the help of Gregory Haskins and the use
of ftrace (trace_printk). It tooks several days for both of us
analyzing the code and the trace output to find this.
Signed-off-by: default avatarSteven Rostedt <srostedt@redhat.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090729042526.205923666@goodmis.org>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      da19ab51
    • Peter Zijlstra's avatar
      sched: Optimize unused cgroup configuration · e7097159
      Peter Zijlstra authored 
      When cgroup group scheduling is built in, skip some code paths
if we don't have any (but the root) cgroups configured.
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      e7097159
    • Peter Zijlstra's avatar
      sched: Fix cgroup smp fairness · a5004278
      Peter Zijlstra authored 
      Commit ec4e0e2f ("fix
inconsistency when redistribute per-cpu tg->cfs_rq shares")
broke cgroup smp fairness.

In order to avoid starvation of newly placed tasks, we never
quite set the share of an empty cpu group-task to 0, but
instead we set it as if there's a single NICE-0 task present.

If however we actually set this in cfs_rq[cpu]->shares, that
means the total shares for that group will be slightly inflated
every time we balance, causing the observed unfairness.

Fix this by setting cfs_rq[cpu]->shares to 0 but actually
setting the effective weight of the related se to the inflated
number.
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1248696557.6987.1615.camel@twins>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      a5004278
    • Ingo Molnar's avatar
      Merge branch 'sched/urgent' into sched/core · 8e9ed8b0
      Ingo Molnar authored 
      Merge reason: avoid upcoming patch conflict.
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      8e9ed8b0
    • Gregory Haskins's avatar
      sched: Fix race in cpupri introduced by cpumask_var changes · 07903af1
      Gregory Haskins authored 
      Background:

Several race conditions in the scheduler have cropped up
recently, which Steven and I have tracked down using ftrace.
The most recent one turns out to be a race in how the scheduler
determines a suitable migration target for RT tasks, introduced
recently with commit:

    commit 68e74568
    Date:   Tue Nov 25 02:35:13 2008 +1030

        sched: convert struct cpupri_vec cpumask_var_t.

The original design of cpupri allowed lockless readers to
quickly determine a best-estimate target.  Races between the
pri_active bitmap and the vec->mask were handled in the
original code because we would detect and return "0" when this
occured.  The design was predicated on the *effective*
atomicity (*) of caching the result of cpus_and() between the
cpus_allowed and the vec->mask.

Commit 68e74568 changed the behavior such that vec->mask is
accessed multiple times.  This introduces a subtle race, the
result of which means we can have a result that returns "1",
but with an empty bitmap.

*) yes, we know cpus_and() is not a locked operator across the
   entire composite array, but it is implicitly atomic on a
   per-word basis which is all the design required to work.

Implementation:

Rather than forgoing the lockless design, or reverting to a
stack-based cpumask_t, we simply check for when the race has
been encountered and continue processing in the event that the
race is hit.  This renders the removal race as if the priority
bit had been atomically cleared as well, and allows the
algorithm to execute correctly.
Signed-off-by: default avatarGregory Haskins <ghaskins@novell.com>
CC: Rusty Russell <rusty@rustcorp.com.au>
CC: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20090730145728.25226.92769.stgit@dev.haskins.net>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      07903af1
    • Peter Zijlstra's avatar
      sched: Fix latencytop and sleep profiling vs group scheduling · e414314c
      Peter Zijlstra authored 
      The latencytop and sleep accounting code assumes that any
scheduler entity represents a task, this is not so.

Cc: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      e414314c
    • Frederic Weisbecker's avatar
      sched: Fix cond_resched_lock() in !CONFIG_PREEMPT · 716a4234
      Frederic Weisbecker authored 
      The might_sleep() test inside cond_resched_lock() assumes the
spinlock is held and then preemption is disabled. This is true
with CONFIG_PREEMPT but the preempt_count() doesn't change
otherwise.

Check by starting from the appropriate preempt offset depending
on the config.
Reported-by: default avatarLi Zefan <lizf@cn.fujitsu.com>
Signed-off-by: default avatarFrederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1248458723-12146-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      716a4234
  3. 01 Aug, 2009 1 commit
  5. 31 Jul, 2009 19 commits
  7. 30 Jul, 2009 11 commits
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7