Message-ID: <Y1w9EKH/CZhNGLJj@slm.duckdns.org>
Date:   Fri, 28 Oct 2022 10:35:28 -1000
From:   Tejun Heo <tj@...nel.org>
To:     Zefan Li <lizefan.x@...edance.com>,
        Johannes Weiner <hannes@...xchg.org>
Cc:     kernel-team@...a.com, linux-kernel@...r.kernel.org
Subject: [PATCH cgroup/for-6.1-fixes] cgroup: Implement DEBUG_CGROUP_REF

It's really difficult to debug when cgroup or css refs leak. Let's add a
debug option to force the refcnt function to not be inlined so that they can
be kprobed for debugging.

Signed-off-by: Tejun Heo <tj@...nel.org>
---
Applying this to for-6.1-fixes. Holler for any objections.

Thanks.

 include/linux/cgroup.h        |   97 +++++-------------------------------------
 include/linux/cgroup_refcnt.h |   91 +++++++++++++++++++++++++++++++++++++++
 kernel/cgroup/cgroup.c        |    5 ++
 lib/Kconfig.debug             |   10 ++++
 4 files changed, 118 insertions(+), 85 deletions(-)

--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -309,72 +309,24 @@ void css_task_iter_end(struct css_task_i
  * Inline functions.
  */
 
+#ifdef CONFIG_DEBUG_CGROUP_REF
+void css_get(struct cgroup_subsys_state *css);
+void css_get_many(struct cgroup_subsys_state *css, unsigned int n);
+bool css_tryget(struct cgroup_subsys_state *css);
+bool css_tryget_online(struct cgroup_subsys_state *css);
+void css_put(struct cgroup_subsys_state *css);
+void css_put_many(struct cgroup_subsys_state *css, unsigned int n);
+#else
+#define CGROUP_REF_FN_ATTRS	static inline
+#include <linux/cgroup_refcnt.h>
+#endif
+
 static inline u64 cgroup_id(const struct cgroup *cgrp)
 {
 	return cgrp->kn->id;
 }
 
 /**
- * css_get - obtain a reference on the specified css
- * @css: target css
- *
- * The caller must already have a reference.
- */
-static inline void css_get(struct cgroup_subsys_state *css)
-{
-	if (!(css->flags & CSS_NO_REF))
-		percpu_ref_get(&css->refcnt);
-}
-
-/**
- * css_get_many - obtain references on the specified css
- * @css: target css
- * @n: number of references to get
- *
- * The caller must already have a reference.
- */
-static inline void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
-{
-	if (!(css->flags & CSS_NO_REF))
-		percpu_ref_get_many(&css->refcnt, n);
-}
-
-/**
- * css_tryget - try to obtain a reference on the specified css
- * @css: target css
- *
- * Obtain a reference on @css unless it already has reached zero and is
- * being released.  This function doesn't care whether @css is on or
- * offline.  The caller naturally needs to ensure that @css is accessible
- * but doesn't have to be holding a reference on it - IOW, RCU protected
- * access is good enough for this function.  Returns %true if a reference
- * count was successfully obtained; %false otherwise.
- */
-static inline bool css_tryget(struct cgroup_subsys_state *css)
-{
-	if (!(css->flags & CSS_NO_REF))
-		return percpu_ref_tryget(&css->refcnt);
-	return true;
-}
-
-/**
- * css_tryget_online - try to obtain a reference on the specified css if online
- * @css: target css
- *
- * Obtain a reference on @css if it's online.  The caller naturally needs
- * to ensure that @css is accessible but doesn't have to be holding a
- * reference on it - IOW, RCU protected access is good enough for this
- * function.  Returns %true if a reference count was successfully obtained;
- * %false otherwise.
- */
-static inline bool css_tryget_online(struct cgroup_subsys_state *css)
-{
-	if (!(css->flags & CSS_NO_REF))
-		return percpu_ref_tryget_live(&css->refcnt);
-	return true;
-}
-
-/**
  * css_is_dying - test whether the specified css is dying
  * @css: target css
  *
@@ -394,31 +346,6 @@ static inline bool css_is_dying(struct c
 	return !(css->flags & CSS_NO_REF) && percpu_ref_is_dying(&css->refcnt);
 }
 
-/**
- * css_put - put a css reference
- * @css: target css
- *
- * Put a reference obtained via css_get() and css_tryget_online().
- */
-static inline void css_put(struct cgroup_subsys_state *css)
-{
-	if (!(css->flags & CSS_NO_REF))
-		percpu_ref_put(&css->refcnt);
-}
-
-/**
- * css_put_many - put css references
- * @css: target css
- * @n: number of references to put
- *
- * Put references obtained via css_get() and css_tryget_online().
- */
-static inline void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
-{
-	if (!(css->flags & CSS_NO_REF))
-		percpu_ref_put_many(&css->refcnt, n);
-}
-
 static inline void cgroup_get(struct cgroup *cgrp)
 {
 	css_get(&cgrp->self);
--- /dev/null
+++ b/include/linux/cgroup_refcnt.h
@@ -0,0 +1,91 @@
+/**
+ * css_get - obtain a reference on the specified css
+ * @css: target css
+ *
+ * The caller must already have a reference.
+ */
+CGROUP_REF_FN_ATTRS
+void css_get(struct cgroup_subsys_state *css)
+{
+	if (!(css->flags & CSS_NO_REF))
+		percpu_ref_get(&css->refcnt);
+}
+
+/**
+ * css_get_many - obtain references on the specified css
+ * @css: target css
+ * @n: number of references to get
+ *
+ * The caller must already have a reference.
+ */
+CGROUP_REF_FN_ATTRS
+void css_get_many(struct cgroup_subsys_state *css, unsigned int n)
+{
+	if (!(css->flags & CSS_NO_REF))
+		percpu_ref_get_many(&css->refcnt, n);
+}
+
+/**
+ * css_tryget - try to obtain a reference on the specified css
+ * @css: target css
+ *
+ * Obtain a reference on @css unless it already has reached zero and is
+ * being released.  This function doesn't care whether @css is on or
+ * offline.  The caller naturally needs to ensure that @css is accessible
+ * but doesn't have to be holding a reference on it - IOW, RCU protected
+ * access is good enough for this function.  Returns %true if a reference
+ * count was successfully obtained; %false otherwise.
+ */
+CGROUP_REF_FN_ATTRS
+bool css_tryget(struct cgroup_subsys_state *css)
+{
+	if (!(css->flags & CSS_NO_REF))
+		return percpu_ref_tryget(&css->refcnt);
+	return true;
+}
+
+/**
+ * css_tryget_online - try to obtain a reference on the specified css if online
+ * @css: target css
+ *
+ * Obtain a reference on @css if it's online.  The caller naturally needs
+ * to ensure that @css is accessible but doesn't have to be holding a
+ * reference on it - IOW, RCU protected access is good enough for this
+ * function.  Returns %true if a reference count was successfully obtained;
+ * %false otherwise.
+ */
+CGROUP_REF_FN_ATTRS
+bool css_tryget_online(struct cgroup_subsys_state *css)
+{
+	if (!(css->flags & CSS_NO_REF))
+		return percpu_ref_tryget_live(&css->refcnt);
+	return true;
+}
+
+/**
+ * css_put - put a css reference
+ * @css: target css
+ *
+ * Put a reference obtained via css_get() and css_tryget_online().
+ */
+CGROUP_REF_FN_ATTRS
+void css_put(struct cgroup_subsys_state *css)
+{
+	if (!(css->flags & CSS_NO_REF))
+		percpu_ref_put(&css->refcnt);
+}
+
+/**
+ * css_put_many - put css references
+ * @css: target css
+ * @n: number of references to put
+ *
+ * Put references obtained via css_get() and css_tryget_online().
+ */
+CGROUP_REF_FN_ATTRS
+void css_put_many(struct cgroup_subsys_state *css, unsigned int n)
+{
+	if (!(css->flags & CSS_NO_REF))
+		percpu_ref_put_many(&css->refcnt, n);
+}
+
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -248,6 +248,11 @@ static int cgroup_addrm_files(struct cgr
 			      struct cgroup *cgrp, struct cftype cfts[],
 			      bool is_add);
 
+#ifdef CONFIG_DEBUG_CGROUP_REF
+#define CGROUP_REF_FN_ATTRS	noinline
+#include <linux/cgroup_refcnt.h>
+#endif
+
 /**
  * cgroup_ssid_enabled - cgroup subsys enabled test by subsys ID
  * @ssid: subsys ID of interest
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1715,6 +1715,16 @@ config LATENCYTOP
 	  Enable this option if you want to use the LatencyTOP tool
 	  to find out which userspace is blocking on what kernel operations.
 
+config DEBUG_CGROUP_REF
+	bool "Disable inlining of cgroup css reference count functions"
+	depends on DEBUG_KERNEL
+	depends on CGROUPS
+	depends on KPROBES
+	default n
+	help
+	  Force cgroup css reference count functions to not be inlined so
+	  that they can be kprobed for debugging.
+
 source "kernel/trace/Kconfig"
 
 config PROVIDE_OHCI1394_DMA_INIT

Powered by blists - more mailing lists