lkml.org 
[lkml]   [2018]   [Oct]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH v5 4/5] sched/core: Prevent race condition between cpuset and __sched_setscheduler()
On 03/10/18 15:42, Steven Rostedt wrote:
> On Mon, 3 Sep 2018 16:28:00 +0200
> Juri Lelli <juri.lelli@redhat.com> wrote:
>
>
> > diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
> > index 5b43f482fa0f..8dc26005bb1e 100644
> > --- a/kernel/cgroup/cpuset.c
> > +++ b/kernel/cgroup/cpuset.c
> > @@ -2410,6 +2410,24 @@ void __init cpuset_init_smp(void)
> > BUG_ON(!cpuset_migrate_mm_wq);
> > }
> >
> > +/**
> > + * cpuset_read_only_lock - Grab the callback_lock from another subsysytem
> > + *
> > + * Description: Gives the holder read-only access to cpusets.
> > + */
> > +void cpuset_read_only_lock(void)
> > +{
> > + raw_spin_lock(&callback_lock);
>
> This was confusing to figure out why grabbing a spinlock gives read
> only access. So I read the long comment above the definition of
> callback_lock. A couple of notes.
>
> 1) The above description needs to go into more detail as to why
> grabbing a spinlock is "read only".
>
> 2) The comment above the callback_lock needs to incorporate this, as
> reading that comment alone will not give anyone an idea that this
> exists.

Right, does the updated version below look any better?

Thanks for reviewing!

Best,

- Juri

--->8---
From d704536ba80a01116007024ec055efcc4a9ee558 Mon Sep 17 00:00:00 2001
From: Mathieu Poirier <mathieu.poirier@linaro.org>
Date: Thu, 23 Aug 2018 14:52:13 +0200
Subject: [PATCH v5 4/5] sched/core: Prevent race condition between cpuset and
__sched_setscheduler()

No synchronisation mechanism exists between the cpuset subsystem and calls
to function __sched_setscheduler(). As such, it is possible that new root
domains are created on the cpuset side while a deadline acceptance test
is carried out in __sched_setscheduler(), leading to a potential oversell
of CPU bandwidth.

Grab callback_lock from core scheduler, so to prevent situations such as
the one described above from happening.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>

---

v4->v5: grab callback_lock instead of cpuset_mutex, as callback_lock is
enough to get read-only access to cpusets [1] and it can be easily
converted to be a raw_spinlock (done in previous - new - patch).

[1] https://elixir.bootlin.com/linux/latest/source/kernel/cgroup/cpuset.c#L275
---
include/linux/cpuset.h | 6 ++++++
kernel/cgroup/cpuset.c | 25 ++++++++++++++++++++++++-
kernel/sched/core.c | 10 ++++++++++
3 files changed, 40 insertions(+), 1 deletion(-)

diff --git a/include/linux/cpuset.h b/include/linux/cpuset.h
index 934633a05d20..8e5a8dd0622b 100644
--- a/include/linux/cpuset.h
+++ b/include/linux/cpuset.h
@@ -55,6 +55,8 @@ extern void cpuset_init_smp(void);
extern void cpuset_force_rebuild(void);
extern void cpuset_update_active_cpus(void);
extern void cpuset_wait_for_hotplug(void);
+extern void cpuset_read_only_lock(void);
+extern void cpuset_read_only_unlock(void);
extern void cpuset_cpus_allowed(struct task_struct *p, struct cpumask *mask);
extern void cpuset_cpus_allowed_fallback(struct task_struct *p);
extern nodemask_t cpuset_mems_allowed(struct task_struct *p);
@@ -176,6 +178,10 @@ static inline void cpuset_update_active_cpus(void)

static inline void cpuset_wait_for_hotplug(void) { }

+static inline void cpuset_read_only_lock(void) { }
+
+static inline void cpuset_read_only_unlock(void) { }
+
static inline void cpuset_cpus_allowed(struct task_struct *p,
struct cpumask *mask)
{
diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 5b43f482fa0f..bff72b920624 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -273,7 +273,8 @@ static struct cpuset top_cpuset = {
* __alloc_pages().
*
* If a task is only holding callback_lock, then it has read-only
- * access to cpusets.
+ * access to cpusets. Mind that callback_lock might be grabbed from other
+ * subsystems as well (via cpuset_read_only_lock()).
*
* Now, the task_struct fields mems_allowed and mempolicy may be changed
* by other task, we use alloc_lock in the task_struct fields to protect
@@ -2410,6 +2411,28 @@ void __init cpuset_init_smp(void)
BUG_ON(!cpuset_migrate_mm_wq);
}

+/**
+ * cpuset_read_only_lock - Grab the callback_lock from cpuset subsystem.
+ *
+ * Description: As described in full details the comment above cpuset_mutex
+ * and callback_lock definitions, holding callback_lock gives the holder
+ * read-only access to cpusets. Even though it might look counter-intuitive
+ * (as callback_lock is a spinlock), in fact a task must hold both
+ * callback_lock _and_ cpuset_mutex to modify cpusets (write access).
+ */
+void cpuset_read_only_lock(void)
+{
+ raw_spin_lock(&callback_lock);
+}
+
+/**
+ * cpuset_read_only_unlock - Release the callback_lock from cpuset subsystem.
+ */
+void cpuset_read_only_unlock(void)
+{
+ raw_spin_unlock(&callback_lock);
+}
+
/**
* cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
* @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 22f5622cba69..ac11ee599968 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4228,6 +4228,13 @@ static int __sched_setscheduler(struct task_struct *p,
rq = task_rq_lock(p, &rf);
update_rq_clock(rq);

+ /*
+ * Make sure we don't race with the cpuset subsystem where root
+ * domains can be rebuilt or modified while operations like DL
+ * admission checks are carried out.
+ */
+ cpuset_read_only_lock();
+
/*
* Changing the policy of the stop threads its a very bad idea:
*/
@@ -4289,6 +4296,7 @@ static int __sched_setscheduler(struct task_struct *p,
/* Re-check policy now with rq lock held: */
if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
policy = oldpolicy = -1;
+ cpuset_read_only_unlock();
task_rq_unlock(rq, p, &rf);
goto recheck;
}
@@ -4346,6 +4354,7 @@ static int __sched_setscheduler(struct task_struct *p,

/* Avoid rq from going away on us: */
preempt_disable();
+ cpuset_read_only_unlock();
task_rq_unlock(rq, p, &rf);

if (pi)
@@ -4358,6 +4367,7 @@ static int __sched_setscheduler(struct task_struct *p,
return 0;

unlock:
+ cpuset_read_only_unlock();
task_rq_unlock(rq, p, &rf);
return retval;
}
--
2.17.1
\
 
 \ /
  Last update: 2018-10-04 11:05    [W:0.116 / U:0.736 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site