[PATCH v3 0/4] cgroup/cpuset: Fix CLONE_INTO_CGROUP problem & other issues

[PATCH v3 0/4] cgroup/cpuset: Fix CLONE_INTO_CGROUP problem & other issues

[Waiman Long]

 v3:
  - Update patches 2 & 3 to put task_cs() call under rcu_read_lock().

 v2:
  - Drop v1 patch 3
  - Add a new patch to fix an issue in cpuset_cancel_attach() and
    another patch to add cpuset_can_fork() and cpuset_cacnel_fork()
    methods.

The first patch in this series fixes a problem in
cpuset_cancel_attach(). Patches 2 and 3 fixes the CLONE_INTO_CGROUP
problem in cpuset. The last one is a minor fix.

Waiman Long (4):
  cgroup/cpuset: Wake up cpuset_attach_wq tasks in
    cpuset_cancel_attach()
  cgroup/cpuset: Make cpuset_fork() handle CLONE_INTO_CGROUP properly
  cgroup/cpuset: Add cpuset_can_fork() and cpuset_cancel_fork() methods
  cgroup/cpuset: Make cpuset_attach_task() skip subpartitions CPUs for
    top_cpuset

 kernel/cgroup/cpuset.c | 161 +++++++++++++++++++++++++++++++++--------
 1 file changed, 129 insertions(+), 32 deletions(-)

-- 
2.31.1

[PATCH v3 1/4] cgroup/cpuset: Wake up cpuset_attach_wq tasks in cpuset_cancel_attach()

[Waiman Long]

After a successful cpuset_can_attach() call which increments the
attach_in_progress flag, either cpuset_cancel_attach() or cpuset_attach()
will be called later. In cpuset_attach(), tasks in cpuset_attach_wq,
if present, will be woken up at the end. That is not the case in
cpuset_cancel_attach(). So missed wakeup is possible if the attach
operation is somehow cancelled. Fix that by doing the wakeup in
cpuset_cancel_attach() as well.

Fixes: e44193d39e8d ("cpuset: let hotplug propagation work wait for task attaching")
Signed-off-by: Waiman Long <longman-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
Reviewed-by: Michal Koutn√Ω <mkoutny-IBi9RG/b67k@public.gmane.org>
---
 kernel/cgroup/cpuset.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index bc4dcfd7bee5..066689a7dcc3 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2507,11 +2507,15 @@ static int cpuset_can_attach(struct cgroup_taskset *tset)
 static void cpuset_cancel_attach(struct cgroup_taskset *tset)
 {
 	struct cgroup_subsys_state *css;
+	struct cpuset *cs;
 
 	cgroup_taskset_first(tset, &css);
+	cs = css_cs(css);
 
 	percpu_down_write(&cpuset_rwsem);
-	css_cs(css)->attach_in_progress--;
+	cs->attach_in_progress--;
+	if (!cs->attach_in_progress)
+		wake_up(&cpuset_attach_wq);
 	percpu_up_write(&cpuset_rwsem);
 }
 
-- 
2.31.1

[PATCH v3 2/4] cgroup/cpuset: Make cpuset_fork() handle CLONE_INTO_CGROUP properly

[Waiman Long]

By default, the clone(2) syscall spawn a child process into the same
cgroup as its parent. With the use of the CLONE_INTO_CGROUP flag
introduced by commit ef2c41cf38a7 ("clone3: allow spawning processes
into cgroups"), the child will be spawned into a different cgroup which
is somewhat similar to writing the child's tid into "cgroup.threads".

The current cpuset_fork() method does not properly handle the
CLONE_INTO_CGROUP case where the cpuset of the child may be different
from that of its parent.  Update the cpuset_fork() method to treat the
CLONE_INTO_CGROUP case similar to cpuset_attach().

Since the newly cloned task has not been running yet, its actual
memory usage isn't known. So it is not necessary to make change to mm
in cpuset_fork().

Fixes: ef2c41cf38a7 ("clone3: allow spawning processes into cgroups")
Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c | 62 ++++++++++++++++++++++++++++--------------
 1 file changed, 42 insertions(+), 20 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index 066689a7dcc3..e954d5abb784 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2520,16 +2520,33 @@ static void cpuset_cancel_attach(struct cgroup_taskset *tset)
 }
 
 /*
- * Protected by cpuset_rwsem.  cpus_attach is used only by cpuset_attach()
+ * Protected by cpuset_rwsem. cpus_attach is used only by cpuset_attach_task()
  * but we can't allocate it dynamically there.  Define it global and
  * allocate from cpuset_init().
  */
 static cpumask_var_t cpus_attach;
+static nodemask_t cpuset_attach_nodemask_to;
+
+static void cpuset_attach_task(struct cpuset *cs, struct task_struct *task)
+{
+	percpu_rwsem_assert_held(&cpuset_rwsem);
+
+	if (cs != &top_cpuset)
+		guarantee_online_cpus(task, cpus_attach);
+	else
+		cpumask_copy(cpus_attach, task_cpu_possible_mask(task));
+	/*
+	 * 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));
+
+	cpuset_change_task_nodemask(task, &cpuset_attach_nodemask_to);
+	cpuset_update_task_spread_flags(cs, task);
+}
 
 static void cpuset_attach(struct cgroup_taskset *tset)
 {
-	/* static buf protected by cpuset_rwsem */
-	static nodemask_t cpuset_attach_nodemask_to;
 	struct task_struct *task;
 	struct task_struct *leader;
 	struct cgroup_subsys_state *css;
@@ -2560,20 +2577,8 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 
 	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
 
-	cgroup_taskset_for_each(task, css, tset) {
-		if (cs != &top_cpuset)
-			guarantee_online_cpus(task, cpus_attach);
-		else
-			cpumask_copy(cpus_attach, task_cpu_possible_mask(task));
-		/*
-		 * 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));
-
-		cpuset_change_task_nodemask(task, &cpuset_attach_nodemask_to);
-		cpuset_update_task_spread_flags(cs, task);
-	}
+	cgroup_taskset_for_each(task, css, tset)
+		cpuset_attach_task(cs, task);
 
 	/*
 	 * Change mm for all threadgroup leaders. This is expensive and may
@@ -3271,11 +3276,28 @@ static void cpuset_bind(struct cgroup_subsys_state *root_css)
  */
 static void cpuset_fork(struct task_struct *task)
 {
-	if (task_css_is_root(task, cpuset_cgrp_id))
+	struct cpuset *cs;
+	bool same_cs;
+
+	rcu_read_lock();
+	cs = task_cs(task);
+	same_cs = (cs == task_cs(current));
+	rcu_read_unlock();
+
+	if (same_cs) {
+		if (cs == &top_cpuset)
+			return;
+
+		set_cpus_allowed_ptr(task, current->cpus_ptr);
+		task->mems_allowed = current->mems_allowed;
 		return;
+	}
 
-	set_cpus_allowed_ptr(task, current->cpus_ptr);
-	task->mems_allowed = current->mems_allowed;
+	/* CLONE_INTO_CGROUP */
+	percpu_down_write(&cpuset_rwsem);
+	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+	cpuset_attach_task(cs, task);
+	percpu_up_write(&cpuset_rwsem);
 }
 
 struct cgroup_subsys cpuset_cgrp_subsys = {
-- 
2.31.1

[PATCH v3 3/4] cgroup/cpuset: Add cpuset_can_fork() and cpuset_cancel_fork() methods

[Waiman Long]

In the case of CLONE_INTO_CGROUP, not all cpusets are ready to accept
new tasks. It is too late to check that in cpuset_fork(). So we need
to add the cpuset_can_fork() and cpuset_cancel_fork() methods to
pre-check it before we can allow attachment to a different cpuset.

We also need to set the attach_in_progress flag to alert other code
that a new task is going to be added to the cpuset.

Fixes: ef2c41cf38a7 ("clone3: allow spawning processes into cgroups")
Signed-off-by: Waiman Long <longman@redhat.com>
---
 kernel/cgroup/cpuset.c | 92 +++++++++++++++++++++++++++++++++++++-----
 1 file changed, 81 insertions(+), 11 deletions(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index e954d5abb784..a331e11e3d20 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2458,6 +2458,20 @@ static int fmeter_getrate(struct fmeter *fmp)
 
 static struct cpuset *cpuset_attach_old_cs;
 
+/*
+ * Check to see if a cpuset can accept a new task
+ * For v1, cpus_allowed and mems_allowed can't be empty.
+ * For v2, effective_cpus can't be empty.
+ * Note that in v1, effective_cpus = cpus_allowed.
+ */
+static int cpuset_can_attach_check(struct cpuset *cs)
+{
+	if (cpumask_empty(cs->effective_cpus) ||
+	   (!is_in_v2_mode() && nodes_empty(cs->mems_allowed)))
+		return -ENOSPC;
+	return 0;
+}
+
 /* Called by cgroups to determine if a cpuset is usable; cpuset_rwsem held */
 static int cpuset_can_attach(struct cgroup_taskset *tset)
 {
@@ -2472,16 +2486,9 @@ static int cpuset_can_attach(struct cgroup_taskset *tset)
 
 	percpu_down_write(&cpuset_rwsem);
 
-	/* allow moving tasks into an empty cpuset if on default hierarchy */
-	ret = -ENOSPC;
-	if (!is_in_v2_mode() &&
-	    (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed)))
-		goto out_unlock;
-
-	/*
-	 * Task cannot be moved to a cpuset with empty effective cpus.
-	 */
-	if (cpumask_empty(cs->effective_cpus))
+	/* Check to see if task is allowed in the cpuset */
+	ret = cpuset_can_attach_check(cs);
+	if (ret)
 		goto out_unlock;
 
 	cgroup_taskset_for_each(task, css, tset) {
@@ -2498,7 +2505,6 @@ static int cpuset_can_attach(struct cgroup_taskset *tset)
 	 * changes which zero cpus/mems_allowed.
 	 */
 	cs->attach_in_progress++;
-	ret = 0;
 out_unlock:
 	percpu_up_write(&cpuset_rwsem);
 	return ret;
@@ -3269,6 +3275,63 @@ static void cpuset_bind(struct cgroup_subsys_state *root_css)
 	percpu_up_write(&cpuset_rwsem);
 }
 
+/*
+ * In case the child is cloned into a cpuset different from its parent,
+ * additional checks are done to see if the move is allowed.
+ */
+static int cpuset_can_fork(struct task_struct *task, struct css_set *cset)
+{
+	struct cpuset *cs = css_cs(cset->subsys[cpuset_cgrp_id]);
+	bool same_cs;
+	int ret;
+
+	rcu_read_lock();
+	same_cs = (cs == task_cs(current));
+	rcu_read_unlock();
+
+	if (same_cs)
+		return 0;
+
+	lockdep_assert_held(&cgroup_mutex);
+	percpu_down_write(&cpuset_rwsem);
+
+	/* Check to see if task is allowed in the cpuset */
+	ret = cpuset_can_attach_check(cs);
+	if (ret)
+		goto out_unlock;
+
+	ret = task_can_attach(task, cs->effective_cpus);
+	if (ret)
+		goto out_unlock;
+
+	ret = security_task_setscheduler(task);
+	if (ret)
+		goto out_unlock;
+
+	/*
+	 * Mark attach is in progress.  This makes validate_change() fail
+	 * changes which zero cpus/mems_allowed.
+	 */
+	cs->attach_in_progress++;
+out_unlock:
+	percpu_up_write(&cpuset_rwsem);
+	return ret;
+}
+
+static void cpuset_cancel_fork(struct task_struct *task, struct css_set *cset)
+{
+	struct cpuset *cs = css_cs(cset->subsys[cpuset_cgrp_id]);
+
+	if (cs == task_cs(current))
+		return;
+
+	percpu_down_write(&cpuset_rwsem);
+	cs->attach_in_progress--;
+	if (!cs->attach_in_progress)
+		wake_up(&cpuset_attach_wq);
+	percpu_up_write(&cpuset_rwsem);
+}
+
 /*
  * Make sure the new task conform to the current state of its parent,
  * which could have been changed by cpuset just after it inherits the
@@ -3297,6 +3360,11 @@ static void cpuset_fork(struct task_struct *task)
 	percpu_down_write(&cpuset_rwsem);
 	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
 	cpuset_attach_task(cs, task);
+
+	cs->attach_in_progress--;
+	if (!cs->attach_in_progress)
+		wake_up(&cpuset_attach_wq);
+
 	percpu_up_write(&cpuset_rwsem);
 }
 
@@ -3310,6 +3378,8 @@ struct cgroup_subsys cpuset_cgrp_subsys = {
 	.attach		= cpuset_attach,
 	.post_attach	= cpuset_post_attach,
 	.bind		= cpuset_bind,
+	.can_fork	= cpuset_can_fork,
+	.cancel_fork	= cpuset_cancel_fork,
 	.fork		= cpuset_fork,
 	.legacy_cftypes	= legacy_files,
 	.dfl_cftypes	= dfl_files,
-- 
2.31.1

[PATCH v3 4/4] cgroup/cpuset: Make cpuset_attach_task() skip subpartitions CPUs for top_cpuset

[Waiman Long]

It is found that attaching a task to the top_cpuset does not currently
ignore CPUs allocated to subpartitions in cpuset_attach_task(). So the
code is changed to fix that.

Signed-off-by: Waiman Long <longman-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
Reviewed-by: Michal Koutn√Ω <mkoutny-IBi9RG/b67k@public.gmane.org>
---
 kernel/cgroup/cpuset.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index a331e11e3d20..8fd90f7bdeb0 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -2540,7 +2540,8 @@ static void cpuset_attach_task(struct cpuset *cs, struct task_struct *task)
 	if (cs != &top_cpuset)
 		guarantee_online_cpus(task, cpus_attach);
 	else
-		cpumask_copy(cpus_attach, task_cpu_possible_mask(task));
+		cpumask_andnot(cpus_attach, task_cpu_possible_mask(task),
+			       cs->subparts_cpus);
 	/*
 	 * can_attach beforehand should guarantee that this doesn't
 	 * fail.  TODO: have a better way to handle failure here
-- 
2.31.1