Merging upstream version 4.19.282.

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

3 files changed, 66 insertions, 16 deletions

diff --git a/kernel/cgroup/cgroup-v1.c b/kernel/cgroup/cgroup-v1.c
index 616449762..c0ebb7080 100644
--- a/kernel/cgroup/cgroup-v1.c
+++ b/kernel/cgroup/cgroup-v1.c
@@ -13,6 +13,7 @@
 #include <linux/delayacct.h>
 #include <linux/pid_namespace.h>
 #include <linux/cgroupstats.h>
+#include <linux/cpu.h>
 
 #include <trace/events/cgroup.h>
 
@@ -55,6 +56,7 @@ int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
 	int retval = 0;
 
 	mutex_lock(&cgroup_mutex);
+	get_online_cpus();
 	percpu_down_write(&cgroup_threadgroup_rwsem);
 	for_each_root(root) {
 		struct cgroup *from_cgrp;
@@ -71,6 +73,7 @@ int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
 			break;
 	}
 	percpu_up_write(&cgroup_threadgroup_rwsem);
+	put_online_cpus();
 	mutex_unlock(&cgroup_mutex);
 
 	return retval;
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index a892a99eb..a8185cdb8 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -55,6 +55,7 @@
 #include <linux/nsproxy.h>
 #include <linux/file.h>
 #include <linux/sched/cputime.h>
+#include <linux/cpu.h>
 #include <net/sock.h>
 
 #define CREATE_TRACE_POINTS
@@ -2210,6 +2211,45 @@ int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
 EXPORT_SYMBOL_GPL(task_cgroup_path);
 
 /**
+ * cgroup_attach_lock - Lock for ->attach()
+ * @lock_threadgroup: whether to down_write cgroup_threadgroup_rwsem
+ *
+ * cgroup migration sometimes needs to stabilize threadgroups against forks and
+ * exits by write-locking cgroup_threadgroup_rwsem. However, some ->attach()
+ * implementations (e.g. cpuset), also need to disable CPU hotplug.
+ * Unfortunately, letting ->attach() operations acquire cpus_read_lock() can
+ * lead to deadlocks.
+ *
+ * Bringing up a CPU may involve creating and destroying tasks which requires
+ * read-locking threadgroup_rwsem, so threadgroup_rwsem nests inside
+ * cpus_read_lock(). If we call an ->attach() which acquires the cpus lock while
+ * write-locking threadgroup_rwsem, the locking order is reversed and we end up
+ * waiting for an on-going CPU hotplug operation which in turn is waiting for
+ * the threadgroup_rwsem to be released to create new tasks. For more details:
+ *
+ *   http://lkml.kernel.org/r/20220711174629.uehfmqegcwn2lqzu@wubuntu
+ *
+ * Resolve the situation by always acquiring cpus_read_lock() before optionally
+ * write-locking cgroup_threadgroup_rwsem. This allows ->attach() to assume that
+ * CPU hotplug is disabled on entry.
+ */
+static void cgroup_attach_lock(void)
+{
+	get_online_cpus();
+	percpu_down_write(&cgroup_threadgroup_rwsem);
+}
+
+/**
+ * cgroup_attach_unlock - Undo cgroup_attach_lock()
+ * @lock_threadgroup: whether to up_write cgroup_threadgroup_rwsem
+ */
+static void cgroup_attach_unlock(void)
+{
+	percpu_up_write(&cgroup_threadgroup_rwsem);
+	put_online_cpus();
+}
+
+/**
  * cgroup_migrate_add_task - add a migration target task to a migration context
  * @task: target task
  * @mgctx: target migration context
@@ -2694,7 +2734,7 @@ struct task_struct *cgroup_procs_write_start(char *buf, bool threadgroup)
 	if (kstrtoint(strstrip(buf), 0, &pid) || pid < 0)
 		return ERR_PTR(-EINVAL);
 
-	percpu_down_write(&cgroup_threadgroup_rwsem);
+	cgroup_attach_lock();
 
 	rcu_read_lock();
 	if (pid) {
@@ -2725,7 +2765,7 @@ struct task_struct *cgroup_procs_write_start(char *buf, bool threadgroup)
 	goto out_unlock_rcu;
 
 out_unlock_threadgroup:
-	percpu_up_write(&cgroup_threadgroup_rwsem);
+	cgroup_attach_unlock();
 out_unlock_rcu:
 	rcu_read_unlock();
 	return tsk;
@@ -2740,7 +2780,7 @@ void cgroup_procs_write_finish(struct task_struct *task)
 	/* release reference from cgroup_procs_write_start() */
 	put_task_struct(task);
 
-	percpu_up_write(&cgroup_threadgroup_rwsem);
+	cgroup_attach_unlock();
 	for_each_subsys(ss, ssid)
 		if (ss->post_attach)
 			ss->post_attach();
@@ -2799,7 +2839,7 @@ static int cgroup_update_dfl_csses(struct cgroup *cgrp)
 
 	lockdep_assert_held(&cgroup_mutex);
 
-	percpu_down_write(&cgroup_threadgroup_rwsem);
+	cgroup_attach_lock();
 
 	/* look up all csses currently attached to @cgrp's subtree */
 	spin_lock_irq(&css_set_lock);
@@ -2830,7 +2870,7 @@ static int cgroup_update_dfl_csses(struct cgroup *cgrp)
 	ret = cgroup_migrate_execute(&mgctx);
 out_finish:
 	cgroup_migrate_finish(&mgctx);
-	percpu_up_write(&cgroup_threadgroup_rwsem);
+	cgroup_attach_unlock();
 	return ret;
 }
 
diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
index dcd5755b1..3067d3e5a 100644
--- a/kernel/cgroup/cpuset.c
+++ b/kernel/cgroup/cpuset.c
@@ -830,8 +830,8 @@ static void rebuild_sched_domains_locked(void)
 	cpumask_var_t *doms;
 	int ndoms;
 
+	lockdep_assert_cpus_held();
 	lockdep_assert_held(&cpuset_mutex);
-	get_online_cpus();
 
 	/*
 	 * We have raced with CPU hotplug. Don't do anything to avoid
@@ -839,15 +839,13 @@ static void rebuild_sched_domains_locked(void)
 	 * Anyways, hotplug work item will rebuild sched domains.
 	 */
 	if (!cpumask_equal(top_cpuset.effective_cpus, cpu_active_mask))
-		goto out;
+		return;
 
 	/* Generate domain masks and attrs */
 	ndoms = generate_sched_domains(&doms, &attr);
 
 	/* Have scheduler rebuild the domains */
 	partition_sched_domains(ndoms, doms, attr);
-out:
-	put_online_cpus();
 }
 #else /* !CONFIG_SMP */
 static void rebuild_sched_domains_locked(void)
@@ -857,9 +855,11 @@ static void rebuild_sched_domains_locked(void)
 
 void rebuild_sched_domains(void)
 {
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 	rebuild_sched_domains_locked();
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 }
 
 /**
@@ -1504,7 +1504,9 @@ static void cpuset_cancel_attach(struct cgroup_taskset *tset)
 	cs = css_cs(css);
 
 	mutex_lock(&cpuset_mutex);
-	css_cs(css)->attach_in_progress--;
+	cs->attach_in_progress--;
+	if (!cs->attach_in_progress)
+		wake_up(&cpuset_attach_wq);
 	mutex_unlock(&cpuset_mutex);
 }
 
@@ -1528,13 +1530,9 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 	cgroup_taskset_first(tset, &css);
 	cs = css_cs(css);
 
+	lockdep_assert_cpus_held();     /* see cgroup_attach_lock() */
 	mutex_lock(&cpuset_mutex);
 
-	/*
-	 * It should hold cpus lock because a cpu offline event can
-	 * cause set_cpus_allowed_ptr() failed.
-	 */
-	get_online_cpus();
 	/* prepare for attach */
 	if (cs == &top_cpuset)
 		cpumask_copy(cpus_attach, cpu_possible_mask);
@@ -1553,7 +1551,6 @@ static void cpuset_attach(struct cgroup_taskset *tset)
 		cpuset_change_task_nodemask(task, &cpuset_attach_nodemask_to);
 		cpuset_update_task_spread_flag(cs, task);
 	}
-       put_online_cpus();
 
 	/*
 	 * Change mm for all threadgroup leaders. This is expensive and may
@@ -1617,6 +1614,7 @@ static int cpuset_write_u64(struct cgroup_subsys_state *css, struct cftype *cft,
 	cpuset_filetype_t type = cft->private;
 	int retval = 0;
 
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 	if (!is_cpuset_online(cs)) {
 		retval = -ENODEV;
@@ -1654,6 +1652,7 @@ static int cpuset_write_u64(struct cgroup_subsys_state *css, struct cftype *cft,
 	}
 out_unlock:
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 	return retval;
 }
 
@@ -1664,6 +1663,7 @@ static int cpuset_write_s64(struct cgroup_subsys_state *css, struct cftype *cft,
 	cpuset_filetype_t type = cft->private;
 	int retval = -ENODEV;
 
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 	if (!is_cpuset_online(cs))
 		goto out_unlock;
@@ -1678,6 +1678,7 @@ static int cpuset_write_s64(struct cgroup_subsys_state *css, struct cftype *cft,
 	}
 out_unlock:
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 	return retval;
 }
 
@@ -1716,6 +1717,7 @@ static ssize_t cpuset_write_resmask(struct kernfs_open_file *of,
 	kernfs_break_active_protection(of->kn);
 	flush_work(&cpuset_hotplug_work);
 
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 	if (!is_cpuset_online(cs))
 		goto out_unlock;
@@ -1741,6 +1743,7 @@ static ssize_t cpuset_write_resmask(struct kernfs_open_file *of,
 	free_trial_cpuset(trialcs);
 out_unlock:
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 	kernfs_unbreak_active_protection(of->kn);
 	css_put(&cs->css);
 	flush_workqueue(cpuset_migrate_mm_wq);
@@ -1985,6 +1988,7 @@ static int cpuset_css_online(struct cgroup_subsys_state *css)
 	if (!parent)
 		return 0;
 
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 
 	set_bit(CS_ONLINE, &cs->flags);
@@ -2035,6 +2039,7 @@ static int cpuset_css_online(struct cgroup_subsys_state *css)
 	spin_unlock_irq(&callback_lock);
 out_unlock:
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 	return 0;
 }
 
@@ -2048,6 +2053,7 @@ static void cpuset_css_offline(struct cgroup_subsys_state *css)
 {
 	struct cpuset *cs = css_cs(css);
 
+	get_online_cpus();
 	mutex_lock(&cpuset_mutex);
 
 	if (is_sched_load_balance(cs))
@@ -2057,6 +2063,7 @@ static void cpuset_css_offline(struct cgroup_subsys_state *css)
 	clear_bit(CS_ONLINE, &cs->flags);
 
 	mutex_unlock(&cpuset_mutex);
+	put_online_cpus();
 }
 
 static void cpuset_css_free(struct cgroup_subsys_state *css)