[PATCH v4 0/3] sched, cgroup/cpuset: Keep user set cpus affinity

[PATCH v4 0/3] sched, cgroup/cpuset: Keep user set cpus affinity

[Waiman Long]

v4:
 - Update patch 1 to make sched_setaffinity() the only function to
   update user_cpus_ptr to make the logic simpler and
   easier to understand. restrict_cpus_allowed_ptr() and
   relax_compatible_cpus_allowed_ptr() will just use it if present.

v3:
 - Attach a new patch 2 to introduce a copy_user_cpus_mask() to copy
   out user masks with lock protection & use it in patch 3.

v2:
 - Rework the v1 patch by extending the semantics of user_cpus_ptr to
   store user set cpus affinity and keeping to it as much as possible.

The user_cpus_ptr field is added by commit b90ca8badbd1 ("sched:
Introduce task_struct::user_cpus_ptr to track requested affinity")
which uses it narrowly to allow keeping cpus affinity intact with
asymmetric cpu setup.

This patchset extends user_cpus_ptr to store user set cpus affinity via
sched_setaffinity() API. With that information avaiable, it will enable
cpuset to keep cpus afinity as close to what the user wants as possible
within the cpu list constraint of the current cpuset. Otherwise some
change to the cpuset hierarchy may reset the cpumask of the tasks in
the affected cpusets to the default cpuset value even if those tasks
have cpus affinity explicitly set by the users before.

It also means that once sched_setaffinity() is called, user_cpus_ptr
will remain allocated until the task exits.

Waiman Long (3):
  sched: Use user_cpus_ptr for saving user provided cpumask in
    sched_setaffinity()
  sched: Provide copy_user_cpus_mask() to copy out user mask
  cgroup/cpuset: Keep user set cpus affinity

 include/linux/sched.h  |   1 +
 kernel/cgroup/cpuset.c |  28 +++++++++-
 kernel/sched/core.c    | 119 ++++++++++++++++++++++++-----------------
 kernel/sched/sched.h   |   1 -
 4 files changed, 98 insertions(+), 51 deletions(-)

-- 
2.31.1

[PATCH v4 1/3] sched: Use user_cpus_ptr for saving user provided cpumask in sched_setaffinity()

[Waiman Long]

The user_cpus_ptr field is added by commit b90ca8badbd1 ("sched:
Introduce task_struct::user_cpus_ptr to track requested affinity"). It
is currently used only by arm64 arch due to possible asymmetric CPU
setup. This patch extends its usage to save user provided cpumask when
sched_setaffinity() is called for all arches. With this patch applied,
user_cpus_ptr, once allocated after a call to sched_setaffinity(),
will only be freed when the task exits.

Since user_cpus_ptr is supposed to be used for "requested
affinity", there is actually no point to save current cpu affinity in
restrict_cpus_allowed_ptr() if sched_setaffinity() has never been called.
Modify the logic to set user_cpus_ptr only in sched_setaffinity() and use
it in restrict_cpus_allowed_ptr() and relax_compatible_cpus_allowed_ptr()
if defined but not changing it.

This will be some changes in behavior for arm64 systems with asymmetric
CPUs in some corner cases. For instance, if sched_setaffinity()
has never been called and there is a cpuset change before
relax_compatible_cpus_allowed_ptr() is called, its subsequent call will
follow what the cpuset allows but not what the previous cpu affinity
setting allows.

As a call to sched_setaffinity() will no longer clear user_cpus_ptr
but set it instead, the SCA_USER flag is no longer necessary and can
be removed.

Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/sched/core.c  | 100 ++++++++++++++++++++++---------------------
 kernel/sched/sched.h |   1 -
 2 files changed, 52 insertions(+), 49 deletions(-)

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index ee28253c9ac0..9353641efdca 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2848,7 +2848,6 @@ static int __set_cpus_allowed_ptr_locked(struct task_struct *p,
 	const struct cpumask *cpu_allowed_mask = task_cpu_possible_mask(p);
 	const struct cpumask *cpu_valid_mask = cpu_active_mask;
 	bool kthread = p->flags & PF_KTHREAD;
-	struct cpumask *user_mask = NULL;
 	unsigned int dest_cpu;
 	int ret = 0;
 
@@ -2907,14 +2906,7 @@ static int __set_cpus_allowed_ptr_locked(struct task_struct *p,
 
 	__do_set_cpus_allowed(p, new_mask, flags);
 
-	if (flags & SCA_USER)
-		user_mask = clear_user_cpus_ptr(p);
-
-	ret = affine_move_task(rq, p, rf, dest_cpu, flags);
-
-	kfree(user_mask);
-
-	return ret;
+	return affine_move_task(rq, p, rf, dest_cpu, flags);
 
 out:
 	task_rq_unlock(rq, p, rf);
@@ -2949,8 +2941,10 @@ EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
 
 /*
  * Change a given task's CPU affinity to the intersection of its current
- * affinity mask and @subset_mask, writing the resulting mask to @new_mask
- * and pointing @p->user_cpus_ptr to a copy of the old mask.
+ * affinity mask and @subset_mask, writing the resulting mask to @new_mask.
+ * If user_cpus_ptr is defined, use it as the the basis for restricting CPU
+ * affinity or use cpu_online_mask instead.
+ *
  * If the resulting mask is empty, leave the affinity unchanged and return
  * -EINVAL.
  */
@@ -2958,16 +2952,10 @@ static int restrict_cpus_allowed_ptr(struct task_struct *p,
 				     struct cpumask *new_mask,
 				     const struct cpumask *subset_mask)
 {
-	struct cpumask *user_mask = NULL;
 	struct rq_flags rf;
 	struct rq *rq;
 	int err;
-
-	if (!p->user_cpus_ptr) {
-		user_mask = kmalloc(cpumask_size(), GFP_KERNEL);
-		if (!user_mask)
-			return -ENOMEM;
-	}
+	bool not_empty;
 
 	rq = task_rq_lock(p, &rf);
 
@@ -2981,25 +2969,21 @@ static int restrict_cpus_allowed_ptr(struct task_struct *p,
 		goto err_unlock;
 	}
 
-	if (!cpumask_and(new_mask, &p->cpus_mask, subset_mask)) {
+
+	if (p->user_cpus_ptr)
+		not_empty = cpumask_and(new_mask, p->user_cpus_ptr, subset_mask);
+	else
+		not_empty = cpumask_and(new_mask, cpu_online_mask, subset_mask);
+
+	if (!not_empty) {
 		err = -EINVAL;
 		goto err_unlock;
 	}
 
-	/*
-	 * We're about to butcher the task affinity, so keep track of what
-	 * the user asked for in case we're able to restore it later on.
-	 */
-	if (user_mask) {
-		cpumask_copy(user_mask, p->cpus_ptr);
-		p->user_cpus_ptr = user_mask;
-	}
-
 	return __set_cpus_allowed_ptr_locked(p, new_mask, 0, rq, &rf);
 
 err_unlock:
 	task_rq_unlock(rq, p, &rf);
-	kfree(user_mask);
 	return err;
 }
 
@@ -3049,34 +3033,27 @@ void force_compatible_cpus_allowed_ptr(struct task_struct *p)
 }
 
 static int
-__sched_setaffinity(struct task_struct *p, const struct cpumask *mask);
+__sched_setaffinity(struct task_struct *p, const struct cpumask *mask, bool save_mask);
 
 /*
  * Restore the affinity of a task @p which was previously restricted by a
- * call to force_compatible_cpus_allowed_ptr(). This will clear (and free)
- * @p->user_cpus_ptr.
+ * call to force_compatible_cpus_allowed_ptr().
  *
  * It is the caller's responsibility to serialise this with any calls to
  * force_compatible_cpus_allowed_ptr(@p).
  */
 void relax_compatible_cpus_allowed_ptr(struct task_struct *p)
 {
-	struct cpumask *user_mask = p->user_cpus_ptr;
-	unsigned long flags;
+	const struct cpumask *user_mask = p->user_cpus_ptr;
+
+	if (!user_mask)
+		user_mask = cpu_online_mask;
 
 	/*
-	 * Try to restore the old affinity mask. If this fails, then
-	 * we free the mask explicitly to avoid it being inherited across
-	 * a subsequent fork().
+	 * Try to restore the old affinity mask with __sched_setaffinity().
+	 * Cpuset masking will be done there too.
 	 */
-	if (!user_mask || !__sched_setaffinity(p, user_mask))
-		return;
-
-	raw_spin_lock_irqsave(&p->pi_lock, flags);
-	user_mask = clear_user_cpus_ptr(p);
-	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
-
-	kfree(user_mask);
+	__sched_setaffinity(p, user_mask, false);
 }
 
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
@@ -8079,10 +8056,11 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
 #endif
 
 static int
-__sched_setaffinity(struct task_struct *p, const struct cpumask *mask)
+__sched_setaffinity(struct task_struct *p, const struct cpumask *mask, bool save_mask)
 {
 	int retval;
 	cpumask_var_t cpus_allowed, new_mask;
+	struct cpumask *user_mask = NULL;
 
 	if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL))
 		return -ENOMEM;
@@ -8098,8 +8076,33 @@ __sched_setaffinity(struct task_struct *p, const struct cpumask *mask)
 	retval = dl_task_check_affinity(p, new_mask);
 	if (retval)
 		goto out_free_new_mask;
+
+	/*
+	 * Save the user requested mask into user_cpus_ptr if save_mask set.
+	 * pi_lock is used for protecting user_cpus_ptr.
+	 */
+	if (save_mask && !p->user_cpus_ptr) {
+		user_mask = kmalloc(cpumask_size(), GFP_KERNEL);
+
+		if (!user_mask) {
+			retval = -ENOMEM;
+			goto out_free_new_mask;
+		}
+	}
+	if (save_mask) {
+		unsigned long flags;
+
+		raw_spin_lock_irqsave(&p->pi_lock, flags);
+		if (!p->user_cpus_ptr) {
+			p->user_cpus_ptr = user_mask;
+			user_mask = NULL;
+		}
+
+		cpumask_copy(p->user_cpus_ptr, mask);
+		raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+	}
 again:
-	retval = __set_cpus_allowed_ptr(p, new_mask, SCA_CHECK | SCA_USER);
+	retval = __set_cpus_allowed_ptr(p, new_mask, SCA_CHECK);
 	if (retval)
 		goto out_free_new_mask;
 
@@ -8113,6 +8116,7 @@ __sched_setaffinity(struct task_struct *p, const struct cpumask *mask)
 		goto again;
 	}
 
+	kfree(user_mask);
 out_free_new_mask:
 	free_cpumask_var(new_mask);
 out_free_cpus_allowed:
@@ -8156,7 +8160,7 @@ long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
 	if (retval)
 		goto out_put_task;
 
-	retval = __sched_setaffinity(p, in_mask);
+	retval = __sched_setaffinity(p, in_mask, true);
 out_put_task:
 	put_task_struct(p);
 	return retval;
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index e26688d387ae..15eefcd65faa 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2283,7 +2283,6 @@ extern struct task_struct *pick_next_task_idle(struct rq *rq);
 #define SCA_CHECK		0x01
 #define SCA_MIGRATE_DISABLE	0x02
 #define SCA_MIGRATE_ENABLE	0x04
-#define SCA_USER		0x08
 
 #ifdef CONFIG_SMP
 
-- 
2.31.1

[PATCH v4 2/3] sched: Provide copy_user_cpus_mask() to copy out user mask

[Waiman Long]

Since accessing the content of the user_cpus_ptr requires lock protection
to ensure its validity, provide a helper function copy_user_cpus_mask()
to facilitate its reading.

Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
---
 include/linux/sched.h |  1 +
 kernel/sched/core.c   | 19 +++++++++++++++++++
 2 files changed, 20 insertions(+)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index e7b2f8a5c711..f2b0340c094e 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1830,6 +1830,7 @@ extern int task_can_attach(struct task_struct *p, const struct cpumask *cs_effec
 extern void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask);
 extern int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask);
 extern int dup_user_cpus_ptr(struct task_struct *dst, struct task_struct *src, int node);
+extern struct cpumask *copy_user_cpus_mask(struct task_struct *p, struct cpumask *user_mask);
 extern void release_user_cpus_ptr(struct task_struct *p);
 extern int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask);
 extern void force_compatible_cpus_allowed_ptr(struct task_struct *p);
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 9353641efdca..ee0fa1635163 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2618,6 +2618,25 @@ void release_user_cpus_ptr(struct task_struct *p)
 	kfree(clear_user_cpus_ptr(p));
 }
 
+/*
+ * Return the copied mask pointer or NULL if user mask not available.
+ * Acquiring pi_lock for read access protection.
+ */
+struct cpumask *copy_user_cpus_mask(struct task_struct *p,
+				    struct cpumask *user_mask)
+{
+	struct cpumask *mask = NULL;
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&p->pi_lock, flags);
+	if (p->user_cpus_ptr) {
+		cpumask_copy(user_mask, p->user_cpus_ptr);
+		mask = user_mask;
+	}
+	raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+	return mask;
+}
+
 /*
  * This function is wildly self concurrent; here be dragons.
  *
-- 
2.31.1

[PATCH v4 3/3] cgroup/cpuset: Keep user set cpus affinity

[Waiman Long]

It was found that any change to the current cpuset hierarchy may reset
the cpumask of the tasks in the affected cpusets to the default cpuset
value even if those tasks have cpus affinity explicitly set by the users
before. That is especially easy to trigger under a cgroup v2 environment
where writing "+cpuset" to the root cgroup's cgroup.subtree_control
file will reset the cpus affinity of all the processes in the system.

That is problematic in a nohz_full environment where the tasks running
in the nohz_full CPUs usually have their cpus affinity explicitly set
and will behave incorrectly if cpus affinity changes.

Fix this problem by looking at user_cpus_ptr which will be set if
cpus affinity have been explicitly set before and use it to restrcit
the given cpumask unless there is no overlap. In that case, it will
fallback to the given one.

With that change in place, it was verified that tasks that have its
cpus affinity explicitly set will not be affected by changes made to
the v2 cgroup.subtree_control files.

Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c | 28 ++++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 58aadfda9b8b..cabfac540fd8 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -704,6 +704,30 @@ static int validate_change(struct cpuset *cur, struct cpuset *trial)
 	return ret;
 }
 
+/*
+ * Preserve user provided cpumask (if set) as much as possible unless there
+ * is no overlap with the given mask.
+ */
+static int cpuset_set_cpus_allowed_ptr(struct task_struct *p,
+				       const struct cpumask *mask)
+{
+	if (p->user_cpus_ptr) {
+		cpumask_var_t new_mask;
+
+		if (alloc_cpumask_var(&new_mask, GFP_KERNEL) &&
+		    copy_user_cpus_mask(p, new_mask) &&
+		    cpumask_and(new_mask, new_mask, mask)) {
+			int ret = set_cpus_allowed_ptr(p, new_mask);
+
+			free_cpumask_var(new_mask);
+			return ret;
+		}
+		free_cpumask_var(new_mask);
+	}
+
+	return set_cpus_allowed_ptr(p, mask);
+}
+
 #ifdef CONFIG_SMP
 /*
  * Helper routine for generate_sched_domains().
@@ -1130,7 +1154,7 @@ static void update_tasks_cpumask(struct cpuset *cs)
 
 	css_task_iter_start(&cs->css, 0, &it);
 	while ((task = css_task_iter_next(&it)))
-		set_cpus_allowed_ptr(task, cs->effective_cpus);
+		cpuset_set_cpus_allowed_ptr(task, cs->effective_cpus);
 	css_task_iter_end(&it);
 }
 
@@ -2303,7 +2327,7 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 		 * can_attach beforehand should guarantee that this doesn't
 		 * fail.  TODO: have a better way to handle failure here
 		 */
-		WARN_ON_ONCE(set_cpus_allowed_ptr(task, cpus_attach));
+		WARN_ON_ONCE(cpuset_set_cpus_allowed_ptr(task, cpus_attach));
 
 		cpuset_change_task_nodemask(task, &cpuset_attach_nodemask_to);
 		cpuset_update_task_spread_flag(cs, task);
-- 
2.31.1

Re: [PATCH v4 3/3] cgroup/cpuset: Keep user set cpus affinity

[Tejun Heo]

Hello,

So, overall I think this is the right direction.

> +static int cpuset_set_cpus_allowed_ptr(struct task_struct *p,
> +				       const struct cpumask *mask)
> +{
> +	if (p->user_cpus_ptr) {
> +		cpumask_var_t new_mask;
> +
> +		if (alloc_cpumask_var(&new_mask, GFP_KERNEL) &&
> +		    copy_user_cpus_mask(p, new_mask) &&
> +		    cpumask_and(new_mask, new_mask, mask)) {
> +			int ret = set_cpus_allowed_ptr(p, new_mask);
> +
> +			free_cpumask_var(new_mask);
> +			return ret;
> +		}
> +		free_cpumask_var(new_mask);
> +	}
> +
> +	return set_cpus_allowed_ptr(p, mask);
> +}

But this seems racy to me. Let's say attach and setaffinity race. The
expectation should be that we'd end up with the same eventual mask no matter
what the operation order may be. The above code wouldn't do that, right?
There's nothing synchronizing the two and if setaffinity takes place between
the user_cpus_ptr test and set_cpus_allowed_ptr(), it'd get ignored.

This gotta be more integrated. There is what the user requested and there
are restrictions from CPU hotplug state and cpuset. All three should be
synchronized so that there is one synchronzied way to obtain and apply the
current effective mask.

Thanks.

-- 
tejun

Re: [PATCH v4 3/3] cgroup/cpuset: Keep user set cpus affinity

[Waiman Long]

On 8/16/22 13:20, Tejun Heo wrote:
> Hello,
>
> So, overall I think this is the right direction.
>
>> +static int cpuset_set_cpus_allowed_ptr(struct task_struct *p,
>> +				       const struct cpumask *mask)
>> +{
>> +	if (p->user_cpus_ptr) {
>> +		cpumask_var_t new_mask;
>> +
>> +		if (alloc_cpumask_var(&new_mask, GFP_KERNEL) &&
>> +		    copy_user_cpus_mask(p, new_mask) &&
>> +		    cpumask_and(new_mask, new_mask, mask)) {
>> +			int ret = set_cpus_allowed_ptr(p, new_mask);
>> +
>> +			free_cpumask_var(new_mask);
>> +			return ret;
>> +		}
>> +		free_cpumask_var(new_mask);
>> +	}
>> +
>> +	return set_cpus_allowed_ptr(p, mask);
>> +}
> But this seems racy to me. Let's say attach and setaffinity race. The
> expectation should be that we'd end up with the same eventual mask no matter
> what the operation order may be. The above code wouldn't do that, right?
> There's nothing synchronizing the two and if setaffinity takes place between
> the user_cpus_ptr test and set_cpus_allowed_ptr(), it'd get ignored.

Yes, a race like this is possible. To completely eliminate the race may 
require taking task_rq_lock() and then calling 
__set_cpus_allowed_ptr_locked() which is internal to kernel/sched/core.c.

Alternatively, we can check user_cpus_ptr again after the scond 
set_cpus_allowed_ptr() and retry it with the other path if set. That 
will probably address your concern. Please let me know if you are OK 
with that.

Cheers,
Longman

>
> This gotta be more integrated. There is what the user requested and there
> are restrictions from CPU hotplug state and cpuset. All three should be
> synchronized so that there is one synchronzied way to obtain and apply the
> current effective mask.
>
> Thanks.
>

Re: [PATCH v4 3/3] cgroup/cpuset: Keep user set cpus affinity

[Tejun Heo]

Hello,

On Tue, Aug 16, 2022 at 01:38:17PM -0400, Waiman Long wrote:
> Yes, a race like this is possible. To completely eliminate the race may
> require taking task_rq_lock() and then calling
> __set_cpus_allowed_ptr_locked() which is internal to kernel/sched/core.c.
> 
> Alternatively, we can check user_cpus_ptr again after the scond
> set_cpus_allowed_ptr() and retry it with the other path if set. That will
> probably address your concern. Please let me know if you are OK with that.

I think this would look better if structured the other way around - make the
scheduler side call out to cpuset to query the current restrictions and
apply it atomically.

Thanks.

-- 
tejun

Re: [PATCH v4 3/3] cgroup/cpuset: Keep user set cpus affinity

[Waiman Long]

On 8/16/22 13:52, Tejun Heo wrote:
> Hello,
>
> On Tue, Aug 16, 2022 at 01:38:17PM -0400, Waiman Long wrote:
>> Yes, a race like this is possible. To completely eliminate the race may
>> require taking task_rq_lock() and then calling
>> __set_cpus_allowed_ptr_locked() which is internal to kernel/sched/core.c.
>>
>> Alternatively, we can check user_cpus_ptr again after the scond
>> set_cpus_allowed_ptr() and retry it with the other path if set. That will
>> probably address your concern. Please let me know if you are OK with that.
> I think this would look better if structured the other way around - make the
> scheduler side call out to cpuset to query the current restrictions and
> apply it atomically.

The sched_setaffinity() function does call cpuset_cpus_allowed() to 
apply the cpuset constraint. However, making set_cpus_allowed_ptr() call 
cpuset function is a major change. It will disturb the current locking 
sequences and may cause circular locking dependency problem. We 
certainly need more time to figure out if that is feasible.

In the mean time, I would prefer to do a retry if user_cpus_ptr status 
changes for now. We can then do a follow up patch to make this 
structural change if there is a consensus of doing so.

Cheers,
Longman