[Devel] [PATCH RHEL8 COMMIT] ve/cgroup: Private per-cgroup-root data container

[Konstantin Khorenko]

The commit is pushed to "branch-rh8-4.18.0-240.1.1.vz8.5.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh8-4.18.0-240.1.1.vz8.5.5
------>
commit 4082809201fbb52062c3cd13e44cbae5f5ff4dc2
Author: Valeriy Vdovin <valeriy.vdovin at virtuozzo.com>
Date:   Wed Mar 3 20:21:13 2021 +0300

    ve/cgroup: Private per-cgroup-root data container
    
    As long as each ve is internally attached to a particular
    css_set via it's init_task, it's good to have container with parameters,
    which are common to each cgroup subsystem hierarchy, rooting from it's
    virtual root.
    
    Cherry-picked from vz7 commit
    4a98f07102fd ("ve/cgroup: private per-cgroup-root data container")
    
    Signed-off-by: Valeriy Vdovin <valeriy.vdovin at virtuozzo.com>
    Reviewed-by: Kirill Tkhai <ktkhai at virtuozzo.com>
    
    =====================
    Patchset description:
    
    ve/cgroup: Port release_agent virtualization from vz7
    
    This patchset ports virtualization of cgroup release_agent
    virtualization from vz7.
    
    Major challanges of porting are differences between vz7 and vz8 cgroup
    implementations:
    - transition of cgroups to kernfs
    - slightly changed locking scheme, which relies on css_set_lock in
      places, previously relied on cgroup_mutex.
    
    There is a small number of patches that have been ported without
    modifications, but most of the patches had suffered a lot of
    modification due to the factors described above.
    
    v1:
      - original patchset
    v2:
      - removed port of CGRP_REMOVED due to the use of CSS_ONLINE in VZ8 for
        same reason
      - changed ve_set(get)_release_agent_path signature for more optimal
      - added ve->is_running check before calling userspace executable
    v3:
      - use goto after check for ve->is_running in last patch
---
 include/linux/ve.h |  7 +++++
 kernel/ve/ve.c     | 75 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 82 insertions(+)

diff --git a/include/linux/ve.h b/include/linux/ve.h
index 2ab39b607708..44369dddeb24 100644
--- a/include/linux/ve.h
+++ b/include/linux/ve.h
@@ -114,6 +114,13 @@ struct ve_struct {
 
 	struct workqueue_struct	*wq;
 	struct work_struct	release_agent_work;
+
+	/*
+	 * List of data, private for each root cgroup in
+	 * ve's css_set.
+	 */
+	struct list_head	per_cgroot_list;
+	spinlock_t		per_cgroot_list_lock;
 };
 
 struct ve_devmnt {
diff --git a/kernel/ve/ve.c b/kernel/ve/ve.c
index 934a5ff1c9bb..a108cb63bc9f 100644
--- a/kernel/ve/ve.c
+++ b/kernel/ve/ve.c
@@ -30,6 +30,14 @@
 
 #include "../cgroup/cgroup-internal.h" /* For cgroup_task_count() */
 
+struct per_cgroot_data {
+	struct list_head list;
+	/*
+	 * data is related to this cgroup
+	 */
+	struct cgroup *cgroot;
+};
+
 extern struct kmapset_set sysfs_ve_perms_set;
 
 static struct kmem_cache *ve_cachep;
@@ -67,6 +75,9 @@ struct ve_struct ve0 = {
 	.release_list		= LIST_HEAD_INIT(ve0.release_list),
 	.release_agent_work	= __WORK_INITIALIZER(ve0.release_agent_work,
 					cgroup1_release_agent),
+	.per_cgroot_list	= LIST_HEAD_INIT(ve0.per_cgroot_list),
+	.per_cgroot_list_lock	= __SPIN_LOCK_UNLOCKED(
+					ve0.per_cgroot_list_lock),
 };
 EXPORT_SYMBOL(ve0);
 
@@ -199,6 +210,53 @@ int nr_threads_ve(struct ve_struct *ve)
 }
 EXPORT_SYMBOL(nr_threads_ve);
 
+static struct per_cgroot_data *per_cgroot_data_find_locked(
+	struct list_head *per_cgroot_list, struct cgroup *cgroot)
+{
+	struct per_cgroot_data *data;
+
+	list_for_each_entry(data, per_cgroot_list, list) {
+		if (data->cgroot == cgroot)
+			return data;
+	}
+	return NULL;
+}
+
+static inline struct per_cgroot_data *per_cgroot_get_or_create(
+	struct ve_struct *ve, struct cgroup *cgroot)
+{
+	struct per_cgroot_data *data, *other_data;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ve->per_cgroot_list_lock, flags);
+	data = per_cgroot_data_find_locked(&ve->per_cgroot_list,
+		cgroot);
+	spin_unlock_irqrestore(&ve->per_cgroot_list_lock, flags);
+
+	if (data)
+		return data;
+
+	data = kzalloc(sizeof(struct per_cgroot_data), GFP_KERNEL);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	spin_lock_irqsave(&ve->per_cgroot_list_lock, flags);
+	other_data = per_cgroot_data_find_locked(&ve->per_cgroot_list,
+		cgroot);
+
+	if (other_data) {
+		spin_unlock_irqrestore(&ve->per_cgroot_list_lock, flags);
+		kfree(data);
+		return other_data;
+	}
+
+	data->cgroot = cgroot;
+	list_add(&data->list, &ve->per_cgroot_list);
+
+	spin_unlock_irqrestore(&ve->per_cgroot_list_lock, flags);
+	return data;
+}
+
 struct cgroup_subsys_state *ve_get_init_css(struct ve_struct *ve, int subsys_id)
 {
 	struct cgroup_subsys_state *css;
@@ -533,6 +591,19 @@ static int ve_start_container(struct ve_struct *ve)
 	return err;
 }
 
+static void ve_per_cgroot_free(struct ve_struct *ve)
+{
+	struct per_cgroot_data *data, *saved;
+	unsigned long flags;
+
+	spin_lock_irqsave(&ve->per_cgroot_list_lock, flags);
+	list_for_each_entry_safe(data, saved, &ve->per_cgroot_list, list) {
+		list_del_init(&data->list);
+		kfree(data);
+	}
+	spin_unlock_irqrestore(&ve->per_cgroot_list_lock, flags);
+}
+
 void ve_stop_ns(struct pid_namespace *pid_ns)
 {
 	struct ve_struct *ve = current->task_ve;
@@ -589,6 +660,8 @@ void ve_exit_ns(struct pid_namespace *pid_ns)
 
 	ve_workqueue_stop(ve);
 
+	ve_per_cgroot_free(ve);
+
 	/*
 	 * At this point all userspace tasks in container are dead.
 	 */
@@ -699,6 +772,7 @@ static struct cgroup_subsys_state *ve_create(struct cgroup_subsys_state *parent_
 
 	INIT_WORK(&ve->release_agent_work, cgroup1_release_agent);
 	spin_lock_init(&ve->release_list_lock);
+	spin_lock_init(&ve->per_cgroot_list_lock);
 
 	ve->_randomize_va_space = ve0._randomize_va_space;
 
@@ -721,6 +795,7 @@ static struct cgroup_subsys_state *ve_create(struct cgroup_subsys_state *parent_
 #endif
 	INIT_LIST_HEAD(&ve->devmnt_list);
 	INIT_LIST_HEAD(&ve->release_list);
+	INIT_LIST_HEAD(&ve->per_cgroot_list);
 	mutex_init(&ve->devmnt_mutex);
 
 #ifdef CONFIG_AIO