[Devel] [PATCH RHEL COMMIT] ve/kthreadd: emulate kthreadd by kworker thread, attached to CT
Konstantin Khorenko
khorenko at virtuozzo.com
Tue Sep 28 14:42:33 MSK 2021
The commit is pushed to "branch-rh9-5.14.vz9.1.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after ark-5.14
------>
commit affacbc14c10da357be5130278d05caffea4f8c7
Author: Stanislav Kinsburskiy <skinsbursky at virtuozzo.com>
Date: Tue Sep 28 14:42:33 2021 +0300
ve/kthreadd: emulate kthreadd by kworker thread, attached to CT
This is the first version of kthreadd in a container (there will some
improvements on top).
With this patch kthreadd won't be visible in a container, but its children
will be.
Signed-off-by: Stanislav Kinsburskiy <skinsbursky at virtuozzo.com>
(cherry-picked from vz8 commit 1cc3a639c3fb ("ve/kthreadd: emulate
kthreadd by kworker thread, attached to CT"))
Signed-off-by: Nikita Yushchenko <nikita.yushchenko at virtuozzo.com>
---
include/linux/ve.h | 2 ++
kernel/ve/ve.c | 94 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 96 insertions(+)
diff --git a/include/linux/ve.h b/include/linux/ve.h
index 18d52c6c63cf..6f75bd084e2e 100644
--- a/include/linux/ve.h
+++ b/include/linux/ve.h
@@ -59,6 +59,8 @@ struct ve_struct {
u64 _uevent_seqnum;
int _randomize_va_space;
+
+ struct kthread_worker *kthreadd_worker;
};
extern int nr_ve;
diff --git a/kernel/ve/ve.c b/kernel/ve/ve.c
index 75299258d677..da86dae1a5f4 100644
--- a/kernel/ve/ve.c
+++ b/kernel/ve/ve.c
@@ -20,6 +20,9 @@
#include <linux/mutex.h>
#include <linux/kmapset.h>
#include <linux/mm.h>
+#include <linux/kthread.h>
+#include <linux/nsproxy.h>
+#include <linux/fs_struct.h>
#include <uapi/linux/vzcalluser.h>
#include <net/rtnetlink.h>
@@ -262,6 +265,87 @@ static void ve_drop_context(struct ve_struct *ve)
ve->init_cred = NULL;
}
+static void ve_stop_kthreadd(struct ve_struct *ve)
+{
+ kthread_destroy_worker(ve->kthreadd_worker);
+ ve->kthreadd_worker = NULL;
+}
+
+struct kthread_attach_work {
+ struct kthread_work work;
+ struct completion done;
+ struct task_struct *target;
+ int result;
+};
+
+static void kthread_attach_fn(struct kthread_work *w)
+{
+ struct kthread_attach_work *work = container_of(w,
+ struct kthread_attach_work, work);
+ struct task_struct *target = work->target;
+ struct cred *cred;
+ int err;
+
+ get_nsproxy(target->nsproxy);
+ switch_task_namespaces(current, target->nsproxy);
+
+ err = unshare_fs_struct();
+ if (err)
+ goto out;
+ set_fs_root(current->fs, &target->fs->root);
+ set_fs_pwd(current->fs, &target->fs->root);
+
+ err = -ENOMEM;
+ cred = prepare_kernel_cred(target);
+ if (!cred)
+ goto out;
+ err = commit_creds(cred);
+ if (err)
+ goto out;
+
+ err = cgroup_attach_task_all(target, current);
+ if (err)
+ goto out;
+out:
+ work->result = err;
+ complete(&work->done);
+}
+
+static struct kthread_worker *ve_create_kworker(struct ve_struct *ve)
+{
+ struct kthread_worker *w;
+ struct kthread_attach_work attach = {
+ KTHREAD_WORK_INIT(attach.work, kthread_attach_fn),
+ COMPLETION_INITIALIZER_ONSTACK(attach.done),
+ .target = current,
+ };
+
+ w = kthread_create_worker(0, "worker/%s", ve_name(ve));
+ if (IS_ERR(w))
+ return w;
+
+ kthread_queue_work(w, &attach.work);
+ wait_for_completion(&attach.done);
+ if (attach.result) {
+ kthread_destroy_worker(w);
+ return ERR_PTR(attach.result);
+ }
+
+ return w;
+}
+
+static int ve_start_kthreadd(struct ve_struct *ve)
+{
+ struct kthread_worker *w;
+
+ w = ve_create_kworker(ve);
+ if (IS_ERR(w))
+ return PTR_ERR(w);
+
+ ve->kthreadd_worker = w;
+ return 0;
+}
+
/* under ve->op_sem write-lock */
static int ve_start_container(struct ve_struct *ve)
{
@@ -307,6 +391,10 @@ static int ve_start_container(struct ve_struct *ve)
if (err)
goto err_list;
+ err = ve_start_kthreadd(ve);
+ if (err)
+ goto err_kthreadd;
+
err = ve_hook_iterate_init(VE_SS_CHAIN, ve);
if (err < 0)
goto err_iterate;
@@ -322,6 +410,8 @@ static int ve_start_container(struct ve_struct *ve)
return 0;
err_iterate:
+ ve_stop_kthreadd(ve);
+err_kthreadd:
ve_list_del(ve);
err_list:
ve_drop_context(ve);
@@ -352,6 +442,10 @@ void ve_stop_ns(struct pid_namespace *pid_ns)
* anymore, setup it again if needed.
*/
ve->is_pseudosuper = 0;
+ /*
+ * Stop kthreads, or zap_pid_ns_processes() will wait them forever.
+ */
+ ve_stop_kthreadd(ve);
unlock:
up_write(&ve->op_sem);
}
More information about the Devel
mailing list