lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20221214025101.1268437-2-ming.lei@redhat.com>
Date:   Wed, 14 Dec 2022 10:50:59 +0800
From:   Ming Lei <ming.lei@...hat.com>
To:     Jens Axboe <axboe@...nel.dk>, Tejun Heo <tj@...nel.org>
Cc:     linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        Zhong Jinghua <zhongjinghua@...wei.com>,
        Yu Kuai <yukuai3@...wei.com>, Dennis Zhou <dennis@...nel.org>,
        Ming Lei <ming.lei@...hat.com>
Subject: [PATCH 1/3] lib/percpu-refcount: support to exit refcount automatically during releasing

We only have two users in which percpu_ref_exit() is called from
->release().

Add flag of PERCPU_REF_AUTO_EXIT for avoiding to call percpu_ref_exit()
from ->release() directly since we need to drain ->release() in
percpu_ref_exit() for fixing use-after-free.

Signed-off-by: Ming Lei <ming.lei@...hat.com>
---
 include/linux/percpu-refcount.h | 21 +++++++++++++++++++--
 lib/percpu-refcount.c           |  9 ++++++---
 2 files changed, 25 insertions(+), 5 deletions(-)

diff --git a/include/linux/percpu-refcount.h b/include/linux/percpu-refcount.h
index d73a1c08c3e3..006c6aae261e 100644
--- a/include/linux/percpu-refcount.h
+++ b/include/linux/percpu-refcount.h
@@ -90,6 +90,11 @@ enum {
 	 * Allow switching from atomic mode to percpu mode.
 	 */
 	PERCPU_REF_ALLOW_REINIT	= 1 << 2,
+
+	/*
+	 * call percpu_ref_exit() when releasing
+	 */
+	PERCPU_REF_AUTO_EXIT	= 1 << 3,
 };
 
 struct percpu_ref_data {
@@ -98,6 +103,7 @@ struct percpu_ref_data {
 	percpu_ref_func_t	*confirm_switch;
 	bool			force_atomic:1;
 	bool			allow_reinit:1;
+	bool			auto_exit:1;
 	struct rcu_head		rcu;
 	struct percpu_ref	*ref;
 };
@@ -331,8 +337,19 @@ static inline void percpu_ref_put_many(struct percpu_ref *ref, unsigned long nr)
 
 	if (__ref_is_percpu(ref, &percpu_count))
 		this_cpu_sub(*percpu_count, nr);
-	else if (unlikely(atomic_long_sub_and_test(nr, &ref->data->count)))
-		ref->data->release(ref);
+	else {
+		struct percpu_ref_data *data = ref->data;
+		struct percpu_ref copy = *ref;
+		bool release = false;
+
+		if (unlikely(atomic_long_sub_and_test(nr, &data->count))) {
+			data->release(ref);
+			release = true;
+		}
+
+		if (release && data->auto_exit)
+			percpu_ref_exit(&copy);
+	}
 
 	rcu_read_unlock();
 }
diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
index 668f6aa6a75d..c0cadf92948f 100644
--- a/lib/percpu-refcount.c
+++ b/lib/percpu-refcount.c
@@ -82,6 +82,7 @@ int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release,
 
 	data->force_atomic = flags & PERCPU_REF_INIT_ATOMIC;
 	data->allow_reinit = flags & PERCPU_REF_ALLOW_REINIT;
+	data->auto_exit = flags & PERCPU_REF_AUTO_EXIT;
 
 	if (flags & (PERCPU_REF_INIT_ATOMIC | PERCPU_REF_INIT_DEAD)) {
 		ref->percpu_count_ptr |= __PERCPU_REF_ATOMIC;
@@ -123,9 +124,11 @@ static void __percpu_ref_exit(struct percpu_ref *ref)
  *
  * This function exits @ref.  The caller is responsible for ensuring that
  * @ref is no longer in active use.  The usual places to invoke this
- * function from are the @ref->release() callback or in init failure path
- * where percpu_ref_init() succeeded but other parts of the initialization
- * of the embedding object failed.
+ * function from are where the refcounter is confirmed as idle or in init
+ * failure path where percpu_ref_init() succeeded but other parts of the
+ * initialization of the embedding object failed. For caller which needs
+ * to call percpu_ref_exit() in ->release, please pass PERCPU_REF_AUTO_EXIT
+ * to percpu_ref_init().
  */
 void percpu_ref_exit(struct percpu_ref *ref)
 {
-- 
2.38.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ