diff mbox series

[RFC,bpf-next,1/3] cgroup: add cgroup_freeze_no_kn() to freeze a cgroup from bpf

Message ID 20240327225334.58474-2-tixxdz@gmail.com (mailing list archive)
State New
Headers show
Series bpf: freeze a task cgroup from bpf | expand

Commit Message

Djalal Harouni March 27, 2024, 10:53 p.m. UTC
This patch adds a new cgroup helper cgroup_freeze_no_kn() to freeze a
cgroup hierarchy that is on a default cgroup v2 without going through
kernfs interface.

For some cases we want to freeze the cgroup of a task based on some
signals, doing so from bpf is better than user space which could be
too late.

The cgroup_freeze_no_kn() will acquire the cgroup_mutex and release it
at the end.

It also checks if the cgroup is on the default hierarchy and it is not
a root cgroup.

Signed-off-by: Djalal Harouni <tixxdz@gmail.com>
---
 include/linux/cgroup.h |  2 ++
 kernel/cgroup/cgroup.c | 69 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 71 insertions(+)
diff mbox series

Patch

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index 34aaf0e87def..5019b32ea933 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -137,6 +137,8 @@  int cgroup_init(void);
 
 int cgroup_parse_float(const char *input, unsigned dec_shift, s64 *v);
 
+int cgroup_freeze_no_kn(struct cgroup *cgrp, int freeze);
+
 /*
  * Iteration helpers and macros.
  */
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index a66c088c851c..0aafcd9e39b5 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -1595,6 +1595,26 @@  static u16 cgroup_calc_subtree_ss_mask(u16 subtree_control, u16 this_ss_mask)
 	return cur_ss_mask;
 }
 
+/**
+ * cgroup_dfl_write_no_kn - check if direct writes to cgroup without going
+ * through kernfs is allowed.
+ * @cgrp: the target cgroup
+ *
+ * This helper ensures that the cgroup is on the default hierarchy and it
+ * is not a root cgroup.
+ *
+ * Return: %0 on success or a negative errno code on failure.
+ */
+static int cgroup_dfl_write_no_kn(struct cgroup *cgrp)
+{
+	lockdep_assert_held(&cgroup_mutex);
+
+	if (!cgroup_on_dfl(cgrp) || !cgroup_parent(cgrp))
+		return -EOPNOTSUPP;
+
+	return 0;
+}
+
 /**
  * cgroup_kn_unlock - unlocking helper for cgroup kernfs methods
  * @kn: the kernfs_node being serviced
@@ -1668,6 +1688,25 @@  struct cgroup *cgroup_kn_lock_live(struct kernfs_node *kn, bool drain_offline)
 	return NULL;
 }
 
+/**
+ * cgroup_lock_live_no_kn - locking helper for direct writes to cgroup without
+ * going through kernfs interface.
+ * @cgrp: the target cgroup
+ *
+ * This helper performs cgroup locking and verifies that the associated cgroup
+ * is alive. Returns the cgroup if alive; otherwise, %NULL.
+ * A successful return should be undone by a matching cgroup_unlock()
+ * invocation.
+ */
+static struct cgroup *cgroup_lock_live_no_kn(struct cgroup *cgrp)
+{
+	cgroup_lock();
+	if (!cgroup_is_dead(cgrp))
+		return cgrp;
+	cgroup_unlock();
+	return NULL;
+}
+
 static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
 {
 	char name[CGROUP_FILE_NAME_MAX];
@@ -3930,6 +3969,36 @@  static int cgroup_freeze_show(struct seq_file *seq, void *v)
 	return 0;
 }
 
+/**
+ * cgroup_freeze_no_kn - Freeze a cgroup that is on the default hierarchy
+ * without going through kernfs interface.
+ *
+ * @cgrp: the target cgroup
+ * @freeze: freeze state, passing value 1 causes the freezing of the cgroup
+ * and all descendant cgroups. Processes under this cgroup hierarchy will
+ * be stopped and will not run until the cgroup is explicitly unfrozen.
+ * Passing value 0 unthaws the cgroup hierarchy.
+ *
+ * Return: %0 on success or a negative errno code on failure.
+ */
+int cgroup_freeze_no_kn(struct cgroup *cgrp, int freeze)
+{
+	int ret = 0;
+
+	if (freeze < 0 || freeze > 1)
+		return -ERANGE;
+
+	if (!cgroup_lock_live_no_kn(cgrp))
+		return -ENOENT;
+
+	ret = cgroup_dfl_write_no_kn(cgrp);
+	if (!ret)
+		cgroup_freeze(cgrp, freeze);
+
+	cgroup_unlock();
+	return ret;
+}
+
 static ssize_t cgroup_freeze_write(struct kernfs_open_file *of,
 				   char *buf, size_t nbytes, loff_t off)
 {